Re: [dpdk-dev] [PATCH] net/virtio: revert forcing IOVA as VA mode for virtio-user
Hello Maxime,
On Wed, Sep 29, 2021 at 10:18 PM Maxime Coquelin
wrote:
>
> This patch removes the simplification in Virtio descriptors
> handling, where their buffer addresses are IOVAs for Virtio
> PCI devices, and VA-only for Virtio-user devices, which
> added a requirement on Virtio-user that it only supported
> IOVA as VA.
>
> This change introduced a regression for applications using
> Virtio-user and other physical PMDs that require IOVA as PA
> because they don't use an IOMMU.
>
> This patch reverts to the old behaviour, but needed to be
> reworked because of the refactoring that happened in v21.02.
>
> Fixes: 17043a2909bb ("net/virtio: force IOVA as VA mode for virtio-user")
> Cc: sta...@dpdk.org
>
> Reported-by: Olivier Matz
> Signed-off-by: Maxime Coquelin
This patch does not apply on next-virtio, but you are best placed to
figure this out :-).
Quick look, only nits otherwise patch lgtm.
> ---
> drivers/net/virtio/virtio.h | 1 +
> drivers/net/virtio/virtio_ethdev.c | 25 +
> drivers/net/virtio/virtio_rxtx.c | 28
> drivers/net/virtio/virtio_rxtx_packed.h | 2 +-
> drivers/net/virtio/virtio_rxtx_packed_avx.h | 8 +++---
> drivers/net/virtio/virtio_rxtx_packed_neon.h | 8 +++---
> drivers/net/virtio/virtio_rxtx_simple.h | 3 ++-
> drivers/net/virtio/virtio_user_ethdev.c | 7 -
> drivers/net/virtio/virtqueue.h | 22 ++-
> 9 files changed, 76 insertions(+), 28 deletions(-)
>
> diff --git a/drivers/net/virtio/virtio.h b/drivers/net/virtio/virtio.h
> index b4f21dc0c7..7118e5d24c 100644
> --- a/drivers/net/virtio/virtio.h
> +++ b/drivers/net/virtio/virtio.h
> @@ -221,6 +221,7 @@ struct virtio_hw {
> uint8_t *rss_key;
> uint64_t req_guest_features;
> struct virtnet_ctl *cvq;
> + bool use_va;
> };
>
> struct virtio_ops {
> diff --git a/drivers/net/virtio/virtio_ethdev.c
> b/drivers/net/virtio/virtio_ethdev.c
> index b4bd1f07c1..8055be88a2 100644
> --- a/drivers/net/virtio/virtio_ethdev.c
> +++ b/drivers/net/virtio/virtio_ethdev.c
> @@ -567,12 +567,16 @@ virtio_init_queue(struct rte_eth_dev *dev, uint16_t
> queue_idx)
>
> memset(mz->addr, 0, mz->len);
>
> - vq->vq_ring_mem = mz->iova;
> + if (hw->use_va)
> + vq->vq_ring_mem = (uintptr_t)mz->addr;
> + else
> + vq->vq_ring_mem = mz->iova;
> +
> vq->vq_ring_virt_mem = mz->addr;
> PMD_INIT_LOG(DEBUG, "vq->vq_ring_mem: 0x%" PRIx64,
> -(uint64_t)mz->iova);
> +(uint64_t)vq->vq_ring_mem);
vq_ring_mem is a rte_iova_t which is a uint64_t.
Cast is unneeded.
> PMD_INIT_LOG(DEBUG, "vq->vq_ring_virt_mem: 0x%" PRIx64,
> -(uint64_t)(uintptr_t)mz->addr);
> +(uint64_t)(uintptr_t)vq->vq_ring_virt_mem);
Why not display with %p and drop casts?
>
> virtio_init_vring(vq);
>
> @@ -622,17 +626,28 @@ virtio_init_queue(struct rte_eth_dev *dev, uint16_t
> queue_idx)
> txvq->port_id = dev->data->port_id;
> txvq->mz = mz;
> txvq->virtio_net_hdr_mz = hdr_mz;
> - txvq->virtio_net_hdr_mem = hdr_mz->iova;
> + if (hw->use_va)
> + txvq->virtio_net_hdr_mem = (uintptr_t)hdr_mz->addr;
> + else
> + txvq->virtio_net_hdr_mem = hdr_mz->iova;
> } else if (queue_type == VTNET_CQ) {
> cvq = &vq->cq;
> cvq->mz = mz;
> cvq->virtio_net_hdr_mz = hdr_mz;
> - cvq->virtio_net_hdr_mem = hdr_mz->iova;
> + if (hw->use_va)
> + cvq->virtio_net_hdr_mem = (uintptr_t)hdr_mz->addr;
> + else
> + cvq->virtio_net_hdr_mem = hdr_mz->iova;
> memset(cvq->virtio_net_hdr_mz->addr, 0, rte_mem_page_size());
>
> hw->cvq = cvq;
> }
>
> + if (hw->use_va)
> + vq->mbuf_addr_offset = offsetof(struct rte_mbuf, buf_addr);
> + else
> + vq->mbuf_addr_offset = offsetof(struct rte_mbuf, buf_iova);
> +
> if (queue_type == VTNET_TQ) {
> struct virtio_tx_region *txr;
> unsigned int i;
> diff --git a/drivers/net/virtio/virtio_rxtx.c
> b/drivers/net/virtio/virtio_rxtx.c
> index b9d7c8d18f..0f3c286438 100644
> --- a/drivers/net/virtio/virtio_rxtx.c
> +++ b/drivers/net/virtio/virtio_rxtx.c
> @@ -271,10 +271,13 @@ virtqueue_enqueue_refill_inorder(struct virtqueue *vq,
> dxp->cookie = (void *)cookies[i];
> dxp->ndescs = 1;
>
> - start_dp[idx].addr = cookies[i]->buf_iova +
> - RTE_PKTMBUF_HEADROOM - hw->vtnet_hdr_size;
> - start_dp[idx].len = cookies[i]->buf_len -
> - RTE_PKTMBUF_HEADROOM + hw->vtnet_hdr_size;
> +
Re: [dpdk-dev] DTS WG material review
Hello Honnappa, On Thu, Sep 30, 2021 at 8:21 AM Honnappa Nagarahalli wrote: > > Hello, > Please find the slides and the excel sheet containing the > work items to fix DTS. We will review the slides in this meeting. Please > provide your review comments before the meeting if possible. > > The update to Techboard will happen on 6th October. If you attached those files to the mail, the mailing list dropped them. Else, can you share a link to them? Thanks. -- David Marchand
Re: [dpdk-dev] [PATCH] net/virtio: revert forcing IOVA as VA mode for virtio-user
Hi David,
On 9/30/21 09:26, David Marchand wrote:
Hello Maxime,
On Wed, Sep 29, 2021 at 10:18 PM Maxime Coquelin
wrote:
This patch removes the simplification in Virtio descriptors
handling, where their buffer addresses are IOVAs for Virtio
PCI devices, and VA-only for Virtio-user devices, which
added a requirement on Virtio-user that it only supported
IOVA as VA.
This change introduced a regression for applications using
Virtio-user and other physical PMDs that require IOVA as PA
because they don't use an IOMMU.
This patch reverts to the old behaviour, but needed to be
reworked because of the refactoring that happened in v21.02.
Fixes: 17043a2909bb ("net/virtio: force IOVA as VA mode for virtio-user")
Cc: sta...@dpdk.org
Reported-by: Olivier Matz
Signed-off-by: Maxime Coquelin
This patch does not apply on next-virtio, but you are best placed to
figure this out :-).
:) I can confirm, I missed my RSS series in between.
Quick look, only nits otherwise patch lgtm.
---
drivers/net/virtio/virtio.h | 1 +
drivers/net/virtio/virtio_ethdev.c | 25 +
drivers/net/virtio/virtio_rxtx.c | 28
drivers/net/virtio/virtio_rxtx_packed.h | 2 +-
drivers/net/virtio/virtio_rxtx_packed_avx.h | 8 +++---
drivers/net/virtio/virtio_rxtx_packed_neon.h | 8 +++---
drivers/net/virtio/virtio_rxtx_simple.h | 3 ++-
drivers/net/virtio/virtio_user_ethdev.c | 7 -
drivers/net/virtio/virtqueue.h | 22 ++-
9 files changed, 76 insertions(+), 28 deletions(-)
diff --git a/drivers/net/virtio/virtio.h b/drivers/net/virtio/virtio.h
index b4f21dc0c7..7118e5d24c 100644
--- a/drivers/net/virtio/virtio.h
+++ b/drivers/net/virtio/virtio.h
@@ -221,6 +221,7 @@ struct virtio_hw {
uint8_t *rss_key;
uint64_t req_guest_features;
struct virtnet_ctl *cvq;
+ bool use_va;
};
struct virtio_ops {
diff --git a/drivers/net/virtio/virtio_ethdev.c
b/drivers/net/virtio/virtio_ethdev.c
index b4bd1f07c1..8055be88a2 100644
--- a/drivers/net/virtio/virtio_ethdev.c
+++ b/drivers/net/virtio/virtio_ethdev.c
@@ -567,12 +567,16 @@ virtio_init_queue(struct rte_eth_dev *dev, uint16_t
queue_idx)
memset(mz->addr, 0, mz->len);
- vq->vq_ring_mem = mz->iova;
+ if (hw->use_va)
+ vq->vq_ring_mem = (uintptr_t)mz->addr;
+ else
+ vq->vq_ring_mem = mz->iova;
+
vq->vq_ring_virt_mem = mz->addr;
PMD_INIT_LOG(DEBUG, "vq->vq_ring_mem: 0x%" PRIx64,
-(uint64_t)mz->iova);
+(uint64_t)vq->vq_ring_mem);
vq_ring_mem is a rte_iova_t which is a uint64_t.
Cast is unneeded.
PMD_INIT_LOG(DEBUG, "vq->vq_ring_virt_mem: 0x%" PRIx64,
-(uint64_t)(uintptr_t)mz->addr);
+(uint64_t)(uintptr_t)vq->vq_ring_virt_mem);
Why not display with %p and drop casts?
Agree, I'll rework these undeed casts.
virtio_init_vring(vq);
@@ -622,17 +626,28 @@ virtio_init_queue(struct rte_eth_dev *dev, uint16_t
queue_idx)
txvq->port_id = dev->data->port_id;
txvq->mz = mz;
txvq->virtio_net_hdr_mz = hdr_mz;
- txvq->virtio_net_hdr_mem = hdr_mz->iova;
+ if (hw->use_va)
+ txvq->virtio_net_hdr_mem = (uintptr_t)hdr_mz->addr;
+ else
+ txvq->virtio_net_hdr_mem = hdr_mz->iova;
} else if (queue_type == VTNET_CQ) {
cvq = &vq->cq;
cvq->mz = mz;
cvq->virtio_net_hdr_mz = hdr_mz;
- cvq->virtio_net_hdr_mem = hdr_mz->iova;
+ if (hw->use_va)
+ cvq->virtio_net_hdr_mem = (uintptr_t)hdr_mz->addr;
+ else
+ cvq->virtio_net_hdr_mem = hdr_mz->iova;
memset(cvq->virtio_net_hdr_mz->addr, 0, rte_mem_page_size());
hw->cvq = cvq;
}
+ if (hw->use_va)
+ vq->mbuf_addr_offset = offsetof(struct rte_mbuf, buf_addr);
+ else
+ vq->mbuf_addr_offset = offsetof(struct rte_mbuf, buf_iova);
+
if (queue_type == VTNET_TQ) {
struct virtio_tx_region *txr;
unsigned int i;
diff --git a/drivers/net/virtio/virtio_rxtx.c b/drivers/net/virtio/virtio_rxtx.c
index b9d7c8d18f..0f3c286438 100644
--- a/drivers/net/virtio/virtio_rxtx.c
+++ b/drivers/net/virtio/virtio_rxtx.c
@@ -271,10 +271,13 @@ virtqueue_enqueue_refill_inorder(struct virtqueue *vq,
dxp->cookie = (void *)cookies[i];
dxp->ndescs = 1;
- start_dp[idx].addr = cookies[i]->buf_iova +
- RTE_PKTMBUF_HEADROOM - hw->vtnet_hdr_size;
- start_dp[idx].len = cookies[i]->buf_len -
- RTE_PKTMBUF_HEADROOM + hw->vtnet_hdr_size;
+ start_dp[i
Re: [dpdk-dev] [PATCH] lib/cmdline: release cl when cmdline exit
2021-09-30 06:53 (UTC+), Peng, ZhihongX: > > -Original Message- > > From: Dmitry Kozlyuk > > Sent: Monday, September 6, 2021 3:34 PM > > To: Peng, ZhihongX > > Cc: olivier.m...@6wind.com; dev@dpdk.org; sta...@dpdk.org > > Subject: Re: [dpdk-dev] [PATCH] lib/cmdline: release cl when cmdline exit > > > > 2021-09-06 05:51 (UTC+), Peng, ZhihongX: > > > > -Original Message- > > > > From: Dmitry Kozlyuk > > > > Sent: Wednesday, September 1, 2021 1:52 AM > > > > To: Peng, ZhihongX > > > > Cc: olivier.m...@6wind.com; dev@dpdk.org; sta...@dpdk.org > > > > Subject: Re: [dpdk-dev] [PATCH] lib/cmdline: release cl when cmdline > > > > exit > > > > > > > > 2021-08-31 10:28 (UTC+0800), zhihongx.p...@intel.com: > > > > > From: Zhihong Peng > > > > > > > > > > Malloc cl in the cmdline_stdin_new function, so release in the > > > > > cmdline_stdin_exit function is logical, so that cl will not be > > > > > released alone. > > > > > > > > > > Fixes: af75078fece3 (first public release) > > > > > Cc: sta...@dpdk.org > > > > > > > > > > Signed-off-by: Zhihong Peng > > > > > --- > > > > > lib/cmdline/cmdline_socket.c | 1 + > > > > > 1 file changed, 1 insertion(+) > > > > > > > > > > diff --git a/lib/cmdline/cmdline_socket.c > > > > > b/lib/cmdline/cmdline_socket.c index 998e8ade25..ebd5343754 100644 > > > > > --- a/lib/cmdline/cmdline_socket.c > > > > > +++ b/lib/cmdline/cmdline_socket.c > > > > > @@ -53,4 +53,5 @@ cmdline_stdin_exit(struct cmdline *cl) > > > > > return; > > > > > > > > > > terminal_restore(cl); > > > > > + cmdline_free(cl); > > > > > } > > > > > > > > Now cmdline_free() may not be called after cmdline_stdin_exit(). > > > > User code that does so needs to be changed to avoid double-free. > > > > This behavior change must be documented in the release notes. > > > > I'm not sure it should be backported because of the above. > > > Using the asan tool, I found that many dpdk apps did not call > > > cmdline_free, > > only one app called. > > > > I mean external programs that use DPDK, not DPDK bundled apps only. > > If some of them use a stable DPDK branch and the change is backported, a > > double-free will be introduced by upgrading DPDK to a minor version. > > Users of current DPDK version that call cmdline_free() after > > cmdline_stdin_exit() will have to upgrade their code, release notes are the > > place to inform them about this need. > > The patch itself is good and now it is the right time for it. > > Can you give me an ack, I have submitted v2: > http://patches.dpdk.org/project/dpdk/patch/20210917021502.502560-1-zhihongx.p...@intel.com/ Hi Zhihong, v2 doesn't address my concerns above. Do you have any objections?
Re: [dpdk-dev] [RFC V1] examples/l3fwd-power: fix memory leak for rte_pci_device
30/09/2021 08:28, Huisong Li: > Hi. Thomas > > I've summed up our previous discussion. > > Can you look at the final proposal again? > > Do you think we should deal with the problem better? I don't understand what is the final proposal. > 在 2021/9/27 9:44, Huisong Li 写道: > > > > 在 2021/9/27 3:16, Thomas Monjalon 写道: > >> 26/09/2021 14:20, Huisong Li: > >>> 在 2021/9/18 16:46, Thomas Monjalon 写道: > 18/09/2021 05:24, Huisong Li: > > 在 2021/9/17 20:50, Thomas Monjalon 写道: > >> 17/09/2021 04:13, Huisong Li: > >>> How should PMD free it? What should we do? Any good suggestions? > >> Check that there is no other port sharing the same PCI device, > >> then call the PMD callback for rte_pci_remove_t. > > For primary and secondary processes, their rte_pci_device is > > independent. > Yes it requires to free on both primary and secondary. > > > Is this for a scenario where there are multiple representor ports > > under > > the same PCI address in the same processe? > A PCI device can have multiple physical or representor ports. > >>> Got it. > >>> Would it be more appropriate to do this in rte_eal_cleanup() if it > >>> cann't be done in the API above? > >> rte_eal_cleanup is a last cleanup for what was not done earlier. > >> We could do that but first we should properly free devices when > >> closed. > >> > > Totally, it is appropriate that rte_eal_cleanup is responsible for > > releasing devices under the pci bus. > Yes, but if a device is closed while the rest of the app keep running, > we should not wait to free it. > >>> From this point of view, it seems to make sense. However, > >>> according to > >>> the OVS-DPDK > >>> > >>> usage, it calls dev_close() first, and then check whether all ports > >>> under the PCI address are > >>> > >>> closed to free rte_pci_device by calling rte_dev_remove(). > >>> > >>> > >>> If we do not want the user to be aware of this, and we want > >>> rte_pci_device to be freed > >>> > >>> in a timely manner. Can we add a code logic calculating the number of > >>> ports under a PCI address > >>> > >>> and calling rte_dev_remove() to rte_eth_dev_close() to free > >>> rte_pci_device and delete it from rte_pci_bus? > >>> > >>> If we do, we may need to make some extra work, otherwise some > >>> applications, such as OVS-DPDK, will > >>> > >>> fail due to a second call to rte_dev_remove(). > >> I don't understand the proposal. > >> Please could explain again the code path? > > > > 1. This RFC patch intended to free rte_pci_device in DPDK app by calling > > > > rte_dev_remove() after calling dev_close(). > > > > 2. For the above-mentioned usage in OVS-DPDK, please see function > > > > netdev_dpdk_destruct() in lib/netdev-dpdk.c. > > > > 3. Later, you suggest that the release of rte_pci_device should be done > > > > in the dev_close() API, not in the rte_eal_init() which is not real-time. > > > > To sum up, the above proposal comes out. > > > >> It may deserve a separate mail thread. > >> > >> > >> . > > . >
Re: [dpdk-dev] [PATCH 1/3] bus/vmbus: fix leak on device scan
On Wed, Sep 29, 2021 at 10:57 PM Long Li wrote:
>
> > Subject: [PATCH 1/3] bus/vmbus: fix leak on device scan
> >
> > Caught running ASAN.
> >
> > The device name is leaked on scan.
> > rte_device name field being a const, use the private vmbus struct to store
> > the
> > device name and point at it.
> >
> > Fixes: 831dba47bd36 ("bus/vmbus: add Hyper-V virtual bus support")
> > Cc: sta...@dpdk.org
> >
> > Signed-off-by: David Marchand
> > ---
> > drivers/bus/vmbus/linux/vmbus_bus.c | 5 -
> > drivers/bus/vmbus/rte_bus_vmbus.h | 1 +
> > 2 files changed, 5 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/bus/vmbus/linux/vmbus_bus.c
> > b/drivers/bus/vmbus/linux/vmbus_bus.c
> > index 3c924eee14..d8eb07d398 100644
> > --- a/drivers/bus/vmbus/linux/vmbus_bus.c
> > +++ b/drivers/bus/vmbus/linux/vmbus_bus.c
> > @@ -242,7 +242,7 @@ vmbus_scan_one(const char *name)
> > return -1;
> >
> > dev->device.bus = &rte_vmbus_bus.bus;
> > - dev->device.name = strdup(name);
> > + dev->device.name = dev->name = strdup(name);
>
>
> I suggest not to add a "name" in struct rte_vmbus_device. How about defining
> a local variable in this function, like:
> dev->device.name = dev_name = strdup(name);
What is your concern for storing the name?
rte_device name only points at some location where the name is stored.
In general this storage is in the bus object or (in some buses) the
devarg that resulted in the rte_device object creation.
If we won't store the name in the bus object, then we lose the ability
to release the name later.
This is probably fine as long as we never release rte_vmbus_device
objects which is the case atm.
But I don't understand why vmbus should be an exception when comparing
to other buses.
--
David Marchand
[dpdk-dev] [PATCH] doc: remove ethdev deprecation note for flag name
Proposed name change of offload flag PKT_RX_EIP_CKSUM_BAD
to PKT_RX_OUTER_IP_CKSUM_BAD has already been done in the
code as per the deprecation note.
Signed-off-by: Aman Singh
---
doc/guides/rel_notes/deprecation.rst | 5 -
lib/mbuf/rte_mbuf_core.h | 7 ---
2 files changed, 12 deletions(-)
diff --git a/doc/guides/rel_notes/deprecation.rst
b/doc/guides/rel_notes/deprecation.rst
index 05fc2fdee7..549e9416c4 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -159,11 +159,6 @@ Deprecation Notices
will be limited to maximum 256 queues.
Also compile time flag ``RTE_ETHDEV_QUEUE_STAT_CNTRS`` will be removed.
-* ethdev: The offload flag ``PKT_RX_EIP_CKSUM_BAD`` will be removed and
- replaced by the new flag ``PKT_RX_OUTER_IP_CKSUM_BAD``. The new name is more
- consistent with existing outer header checksum status flag naming, which
- should help in reducing confusion about its usage.
-
* i40e: As there are both i40evf and iavf pmd, the functions of them are
duplicated. And now more and more advanced features are developed on iavf.
To keep consistent with kernel driver's name
diff --git a/lib/mbuf/rte_mbuf_core.h b/lib/mbuf/rte_mbuf_core.h
index 9d8e3ddc86..d6f1679944 100644
--- a/lib/mbuf/rte_mbuf_core.h
+++ b/lib/mbuf/rte_mbuf_core.h
@@ -79,13 +79,6 @@ extern "C" {
*/
#define PKT_RX_OUTER_IP_CKSUM_BAD (1ULL << 5)
-/**
- * Deprecated.
- * This flag has been renamed, use PKT_RX_OUTER_IP_CKSUM_BAD instead.
- */
-#define PKT_RX_EIP_CKSUM_BAD \
- RTE_DEPRECATED(PKT_RX_EIP_CKSUM_BAD) PKT_RX_OUTER_IP_CKSUM_BAD
-
/**
* A vlan has been stripped by the hardware and its tci is saved in
* mbuf->vlan_tci. This can only happen if vlan stripping is enabled
--
2.17.1
[dpdk-dev] [PATCH v2] net/virtio: revert forcing IOVA as VA mode for virtio-user
This patch removes the simplification in Virtio descriptors
handling, where their buffer addresses are IOVAs for Virtio
PCI devices, and VA-only for Virtio-user devices, which
added a requirement on Virtio-user that it only supported
IOVA as VA.
This change introduced a regression for applications using
Virtio-user and other physical PMDs that require IOVA as PA
because they don't use an IOMMU.
This patch reverts to the old behaviour, but needed to be
reworked because of the refactoring that happened in v21.02.
Fixes: 17043a2909bb ("net/virtio: force IOVA as VA mode for virtio-user")
Cc: sta...@dpdk.org
Reported-by: Olivier Matz
Signed-off-by: Maxime Coquelin
---
Changes in v2:
==
- Fix cosmetics issues reported by David
---
drivers/net/virtio/virtio.h | 1 +
drivers/net/virtio/virtio_ethdev.c | 27 -
drivers/net/virtio/virtio_rxtx.c | 31 +---
drivers/net/virtio/virtio_rxtx_packed.h | 2 +-
drivers/net/virtio/virtio_rxtx_packed_avx.h | 8 ++---
drivers/net/virtio/virtio_rxtx_packed_neon.h | 8 ++---
drivers/net/virtio/virtio_rxtx_simple.h | 2 +-
drivers/net/virtio/virtio_user_ethdev.c | 7 -
drivers/net/virtio/virtqueue.h | 22 +-
9 files changed, 72 insertions(+), 36 deletions(-)
diff --git a/drivers/net/virtio/virtio.h b/drivers/net/virtio/virtio.h
index 525e2dad4c..e78b2e429e 100644
--- a/drivers/net/virtio/virtio.h
+++ b/drivers/net/virtio/virtio.h
@@ -192,6 +192,7 @@ struct virtio_hw {
uint16_t max_queue_pairs;
uint64_t req_guest_features;
struct virtnet_ctl *cvq;
+ bool use_va;
};
struct virtio_ops {
diff --git a/drivers/net/virtio/virtio_ethdev.c
b/drivers/net/virtio/virtio_ethdev.c
index b08109c61c..b60eeb24ab 100644
--- a/drivers/net/virtio/virtio_ethdev.c
+++ b/drivers/net/virtio/virtio_ethdev.c
@@ -515,12 +515,14 @@ virtio_init_queue(struct rte_eth_dev *dev, uint16_t
queue_idx)
memset(mz->addr, 0, mz->len);
- vq->vq_ring_mem = mz->iova;
+ if (hw->use_va)
+ vq->vq_ring_mem = (uintptr_t)mz->addr;
+ else
+ vq->vq_ring_mem = mz->iova;
+
vq->vq_ring_virt_mem = mz->addr;
- PMD_INIT_LOG(DEBUG, "vq->vq_ring_mem: 0x%" PRIx64,
-(uint64_t)mz->iova);
- PMD_INIT_LOG(DEBUG, "vq->vq_ring_virt_mem: 0x%" PRIx64,
-(uint64_t)(uintptr_t)mz->addr);
+ PMD_INIT_LOG(DEBUG, "vq->vq_ring_mem: 0x%" PRIx64, vq->vq_ring_mem);
+ PMD_INIT_LOG(DEBUG, "vq->vq_ring_virt_mem: %p", vq->vq_ring_virt_mem);
virtio_init_vring(vq);
@@ -570,17 +572,28 @@ virtio_init_queue(struct rte_eth_dev *dev, uint16_t
queue_idx)
txvq->port_id = dev->data->port_id;
txvq->mz = mz;
txvq->virtio_net_hdr_mz = hdr_mz;
- txvq->virtio_net_hdr_mem = hdr_mz->iova;
+ if (hw->use_va)
+ txvq->virtio_net_hdr_mem = (uintptr_t)hdr_mz->addr;
+ else
+ txvq->virtio_net_hdr_mem = hdr_mz->iova;
} else if (queue_type == VTNET_CQ) {
cvq = &vq->cq;
cvq->mz = mz;
cvq->virtio_net_hdr_mz = hdr_mz;
- cvq->virtio_net_hdr_mem = hdr_mz->iova;
+ if (hw->use_va)
+ cvq->virtio_net_hdr_mem = (uintptr_t)hdr_mz->addr;
+ else
+ cvq->virtio_net_hdr_mem = hdr_mz->iova;
memset(cvq->virtio_net_hdr_mz->addr, 0, rte_mem_page_size());
hw->cvq = cvq;
}
+ if (hw->use_va)
+ vq->mbuf_addr_offset = offsetof(struct rte_mbuf, buf_addr);
+ else
+ vq->mbuf_addr_offset = offsetof(struct rte_mbuf, buf_iova);
+
if (queue_type == VTNET_TQ) {
struct virtio_tx_region *txr;
unsigned int i;
diff --git a/drivers/net/virtio/virtio_rxtx.c b/drivers/net/virtio/virtio_rxtx.c
index b9d7c8d18f..e8e6ed20a5 100644
--- a/drivers/net/virtio/virtio_rxtx.c
+++ b/drivers/net/virtio/virtio_rxtx.c
@@ -271,7 +271,7 @@ virtqueue_enqueue_refill_inorder(struct virtqueue *vq,
dxp->cookie = (void *)cookies[i];
dxp->ndescs = 1;
- start_dp[idx].addr = cookies[i]->buf_iova +
+ start_dp[idx].addr = VIRTIO_MBUF_ADDR(cookies[i], vq) +
RTE_PKTMBUF_HEADROOM - hw->vtnet_hdr_size;
start_dp[idx].len = cookies[i]->buf_len -
RTE_PKTMBUF_HEADROOM + hw->vtnet_hdr_size;
@@ -310,10 +310,10 @@ virtqueue_enqueue_recv_refill(struct virtqueue *vq,
struct rte_mbuf **cookie,
dxp->cookie = (void *)cookie[i];
dxp->ndescs = 1;
- start_dp[idx].addr = cookie[i]->buf_iova +
+ start_dp[idx].addr = VIRTIO_MBUF_ADDR(cookie[i], vq) +
RTE_P
Re: [dpdk-dev] [PATCH v4] net/af_packet: reinsert the stripped vlan tag
On 9/29/2021 3:08 PM, Tudor Cornea wrote: > The af_packet pmd driver binds to a raw socket and allows > sending and receiving of packets through the kernel. > > Since commit [1], the kernel strips the vlan tags early in > __netif_receive_skb_core(), so we receive untagged packets while > running with the af_packet pmd. > > Luckily for us, the skb vlan-related fields are still populated from the > stripped vlan tags, so we end up having all the information > that we need in the mbuf. > > Having the pmd driver support DEV_RX_OFFLOAD_VLAN_STRIP allows the > application to control the desired vlan stripping behavior, > until we have a way to describe offloads that can't be disabled by > pmd drivers. > > This patch will cause a change in the default way that the af_packet > pmd treats received vlan-tagged frames. While previously, the > application was required to check the PKT_RX_VLAN_STRIPPED flag, after > this patch, the pmd will re-insert the vlan tag transparently to the > user, unless the DEV_RX_OFFLOAD_VLAN_STRIP is enabled in > rxmode.offloads. > > I've attempted a preliminary benchmark to understand if the change could > cause a sizable performance hit. > > Setup: > Two virtual machines running on top of an ESXi hypervisor > > Tx: DPDK app (running on top of vmxnet3 PMD) > Rx: af_packet (running on top of a kernel vmxnet3 interface) > Packet size :68 (packet contains a vlan tag) > > Rates: > Tx - 1.419 Mpps > Rx (without vlan insertion) - 1227636 pps > Rx (with vlan insertion)- 1220081 pps > > At a first glance, we don't seem to have a large degradation in terms > of packet rate. > > [1] > https://github.com/torvalds/linux/commit/bcc6d47903612c3861201cc3a866fb604f26b8b2 > > Signed-off-by: Tudor Cornea > > --- > v4: > * Updated the af_packet documentation > v3: > * Updated release note and documentation > * Updated commit with performance measurements > v2: > * Added DEV_RX_OFFLOAD_VLAN_STRIP to rxmode->offloads > --- > doc/guides/nics/af_packet.rst | 5 + > doc/guides/rel_notes/release_21_11.rst| 4 > drivers/net/af_packet/rte_eth_af_packet.c | 12 > 3 files changed, 21 insertions(+) > > diff --git a/doc/guides/nics/af_packet.rst b/doc/guides/nics/af_packet.rst > index efd6f1c..c87310b 100644 > --- a/doc/guides/nics/af_packet.rst > +++ b/doc/guides/nics/af_packet.rst > @@ -65,3 +65,8 @@ framecnt=512): > .. code-block:: console > > > --vdev=eth_af_packet0,iface=tap0,blocksz=4096,framesz=2048,framecnt=512,qpairs=1,qdisc_bypass=0 > + > +Features and Limitations of the af_packet PMD > +- > + > +Af_packet PMD now works with VLAN's on Linux Thanks Tudor. I see this is suggested by Stephen, but I think 'now' is confusing in the driver guide. Also driver was already working with VLAN, the change is VLAN is not force stripped anymore. I think following part from your previous version is better, what do you think something like following? " The PMD will re-insert the VLAN tag transparently to the packet if the kernel strips it, as long as the ``DEV_RX_OFFLOAD_VLAN_STRIP`` is not enabled by application. " > diff --git a/doc/guides/rel_notes/release_21_11.rst > b/doc/guides/rel_notes/release_21_11.rst > index ad7c1af..095fd5b 100644 > --- a/doc/guides/rel_notes/release_21_11.rst > +++ b/doc/guides/rel_notes/release_21_11.rst > @@ -66,6 +66,10 @@ New Features > >* Added rte_flow support for dual VLAN insert and strip actions. > > +* **Updated af_packet ethdev driver.** > + > + * Added DEV_RX_OFFLOAD_VLAN_STRIP capability. > + I think change in the default behavior is more important in the release notes, again what do you think to have your following update here: " Default VLAN strip behavior changed. If previously, the vlan tag was stripped, if the application now requires the same behavior, it will need to configure ``DEV_RX_OFFLOAD_VLAN_STRIP``. "
Re: [dpdk-dev] [PATCH v6 1/2] Enable ASan for memory detector on DPDK
Hello,
I see v6 is superseded in pw, I have been cleaning my queue... maybe my fault.
On Thu, Sep 30, 2021 at 7:37 AM wrote:
>
> From: Zhihong Peng
>
> AddressSanitizer (ASan) is a google memory error detect
> standard tool. It could help to detect use-after-free and
> {heap,stack,global}-buffer overflow bugs in C/C++ programs,
> print detailed error information when error happens, large
> improve debug efficiency.
>
> `AddressSanitizer
> ` (ASan)
> is a widely-used debugging tool to detect memory access errors.
> It helps detect issues like use-after-free, various kinds of buffer
> overruns in C/C++ programs, and other similar errors, as well as
> printing out detailed debug information whenever an error is detected.
This patch mixes how to use ASan and instrumenting the DPDK mem allocator.
I would split this patch in two.
The first patch can add the documentation on enabling/using ASan and
describe the known issues on enabling it.
I'd find it better (from a user pov) if we hide all those details
about b_lundef and installation of libasan on Centos.
Something like (only quickly tested):
diff --git a/config/meson.build b/config/meson.build
index 4cdf589e20..7d8b71da79 100644
--- a/config/meson.build
+++ b/config/meson.build
@@ -411,6 +411,33 @@ if get_option('b_lto')
endif
endif
+if get_option('b_sanitize') == 'address'
+asan_dep = cc.find_library('asan', required: true)
+if (not cc.links('int main(int argc, char *argv[]) { return 0; }',
+ dependencies: asan_dep))
+error('broken dependency, "libasan"')
+endif
+add_project_link_arguments('-lasan', language: 'c')
+dpdk_extra_ldflags += '-lasan'
+endif
+
if get_option('default_library') == 'both'
error( '''
Unsupported value "both" for "default_library" option.
Bruce, do you see an issue with this approach?
Then a second patch adds the rte_malloc instrumentation, with a check
at configuration time.
endif
add_project_link_arguments('-lasan', language: 'c')
dpdk_extra_ldflags += '-lasan'
+if arch_subdir == 'x86'
+asan_check_code = '''
+#ifdef __SANITIZE_ADDRESS__
+#define RTE_MALLOC_ASAN
+#elif defined(__has_feature)
+# if __has_feature(address_sanitizer)
+#define RTE_MALLOC_ASAN
+# endif
+#endif
+
+#ifndef RTE_MALLOC_ASAN
+#error ASan not available.
+#endif
+'''
+if cc.compiles(asan_check_code)
+dpdk_conf.set10('RTE_MALLOC_ASAN', true)
+endif
+endif
endif
if get_option('default_library') == 'both'
Few more comments:
>
> DPDK ASan functionality is currently only supported Linux x86_64.
> Support other platforms, need to define ASAN_SHADOW_OFFSET value
> according to google ASan document.
>
> Here is an example of heap-buffer-overflow bug:
> ..
> char *p = rte_zmalloc(NULL, 7, 0);
> p[7] = 'a';
> ..
>
> Here is an example of use-after-free bug:
> ..
> char *p = rte_zmalloc(NULL, 7, 0);
> rte_free(p);
> *p = 'a';
> ..
>
> If you want to use this feature,
> you need to add below compilation options when compiling code:
> -Dbuildtype=debug -Db_lundef=false -Db_sanitize=address
> "-Dbuildtype=debug": This is a non-essential option. When this option
> is added, if a memory error occurs, ASan can clearly show where the
> code is wrong.
> "-Db_lundef=false": When use clang to compile DPDK, this option must
> be added.
>
> Signed-off-by: Xueqin Lin
> Signed-off-by: Zhihong Peng
> ---
> devtools/words-case.txt | 1 +
> doc/guides/prog_guide/ASan.rst | 108 +
> doc/guides/prog_guide/index.rst | 1 +
> examples/helloworld/main.c | 5 +
> lib/eal/common/malloc_elem.c| 26 +++-
> lib/eal/common/malloc_elem.h| 204 +++-
> lib/eal/common/malloc_heap.c| 12 ++
> lib/eal/common/rte_malloc.c | 9 +-
> 8 files changed, 361 insertions(+), 5 deletions(-)
> create mode 100644 doc/guides/prog_guide/ASan.rst
>
> diff --git a/devtools/words-case.txt b/devtools/words-case.txt
> index 0bbad48626..3655596d47 100644
> --- a/devtools/words-case.txt
> +++ b/devtools/words-case.txt
> @@ -86,3 +86,4 @@ VXLAN
> Windows
> XDP
> XOR
> +ASan
Alphabetical order please.
> diff --git a/doc/guides/prog_guide/ASan.rst b/doc/guides/prog_guide/ASan.rst
Filenames are lowercase in the doc.
> new file mode 100644
> index 00..7145a3b1a1
> --- /dev/null
> +++ b/doc/guides/prog_guide/ASan.rst
> @@ -0,0 +1,108 @@
> +.. Copyright (c) <2021>, Intel Corporation
> + All rights reserved.
> +
> +Memory error detect standard tool - AddressSanitizer(ASan)
> +==
> +
> +AddressSanitizer (ASan) is a google memory error detect
> +standard tool. It could help to detect use-after-free and
> +{heap,stack,global}-buffer overflow bugs in C/C++ programs,
> +print detailed error infor
Re: [dpdk-dev] [PATCH v2] net/virtio: revert forcing IOVA as VA mode for virtio-user
On Thu, Sep 30, 2021 at 10:13 AM Maxime Coquelin
wrote:
>
> This patch removes the simplification in Virtio descriptors
> handling, where their buffer addresses are IOVAs for Virtio
> PCI devices, and VA-only for Virtio-user devices, which
> added a requirement on Virtio-user that it only supported
> IOVA as VA.
>
> This change introduced a regression for applications using
> Virtio-user and other physical PMDs that require IOVA as PA
> because they don't use an IOMMU.
>
> This patch reverts to the old behaviour, but needed to be
> reworked because of the refactoring that happened in v21.02.
>
> Fixes: 17043a2909bb ("net/virtio: force IOVA as VA mode for virtio-user")
> Cc: sta...@dpdk.org
>
> Reported-by: Olivier Matz
> Signed-off-by: Maxime Coquelin
I think you can keep:
Tested-by: Olivier Matz
Reviewed-by: David Marchand
Thanks.
--
David Marchand
Re: [dpdk-dev] [PATCH v2 0/5] kvargs: promote or remove experimental api
On Wed, Sep 29, 2021 at 11:40 PM Olivier Matz wrote: > > This patchset promotes 2 functions rte_kvargs_parse_delim() and > rte_kvargs_get() as stable. > > It also replaces rte_kvargs_strcmp() by a new one > rte_kvargs_get_with_value(), which is easier to use. > > v2 > * remove rte_kvargs_strcmp from version.map > > Olivier Matz (5): > kvargs: promote delimited parsing as stable > kvargs: promote get from key as stable > kvargs: new function to get from key and value > kvargs: remove experimental function to compare string > kvargs: fix comments style Thanks, for the series, Reviewed-by: David Marchand -- David Marchand
Re: [dpdk-dev] [PATCH] net/virtio: revert forcing IOVA as VA mode for virtio-user
On 9/29/21 23:15, Olivier Matz wrote:
Hi Maxime,
On Wed, Sep 29, 2021 at 10:17:39PM +0200, Maxime Coquelin wrote:
This patch removes the simplification in Virtio descriptors
handling, where their buffer addresses are IOVAs for Virtio
PCI devices, and VA-only for Virtio-user devices, which
added a requirement on Virtio-user that it only supported
IOVA as VA.
This change introduced a regression for applications using
Virtio-user and other physical PMDs that require IOVA as PA
because they don't use an IOMMU.
This patch reverts to the old behaviour, but needed to be
reworked because of the refactoring that happened in v21.02.
Fixes: 17043a2909bb ("net/virtio: force IOVA as VA mode for virtio-user")
Cc: sta...@dpdk.org
Reported-by: Olivier Matz
Signed-off-by: Maxime Coquelin
Tested-by: Olivier Matz
Many thanks for your quick solution on this!
You're welcome, thanks for reporting.
I just notice your reply to v1, so I missed to report your Tested-by on
v2 (which only has cosmetic changes). Feel free to add it.
Maxime
[dpdk-dev] [PATCH v4 1/5] eventdev/rx_adapter: add event buffer size configurability
Currently event buffer is static array with a default size defined
internally.
To configure event buffer size from application,
``rte_event_eth_rx_adapter_create_with_params`` api is added which
takes ``struct rte_event_eth_rx_adapter_params`` to configure event
buffer size in addition other params . The event buffer size is
rounded up for better buffer utilization and performance . In case
of NULL params argument, default event buffer size is used.
Signed-off-by: Naga Harish K S V
Signed-off-by: Ganapati Kundapura
---
v4:
* rebased with latest dpdk-next-eventdev branch
v3:
* updated documentation and code comments as per review comments.
* updated new create api test case name with suitable one.
v2:
* Updated header file and rx adapter documentation as per review comments.
* new api name is modified as rte_event_eth_rx_adapter_create_with_params
as per review comments.
* rxa_params pointer argument Value NULL is allowed to represent the
default values
v1:
* Initial implementation with documentation and unit tests.
---
---
.../prog_guide/event_ethernet_rx_adapter.rst | 7 ++
lib/eventdev/rte_event_eth_rx_adapter.c | 98 +--
lib/eventdev/rte_event_eth_rx_adapter.h | 41 +++-
lib/eventdev/version.map | 2 +
4 files changed, 140 insertions(+), 8 deletions(-)
diff --git a/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
b/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
index ce23d8a474..8526aecf57 100644
--- a/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
+++ b/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
@@ -62,6 +62,13 @@ service function and needs to create an event port for it.
The callback is
expected to fill the ``struct rte_event_eth_rx_adapter_conf structure``
passed to it.
+If the application desires to control the event buffer size, it can use the
+``rte_event_eth_rx_adapter_create_with_params()`` api. The event buffer size is
+specified using ``struct rte_event_eth_rx_adapter_params::event_buf_size``.
+The function is passed the event device to be associated with the adapter
+and port configuration for the adapter to setup an event port if the
+adapter needs to use a service function.
+
Adding Rx Queues to the Adapter Instance
diff --git a/lib/eventdev/rte_event_eth_rx_adapter.c
b/lib/eventdev/rte_event_eth_rx_adapter.c
index 10491ca07b..606db241b8 100644
--- a/lib/eventdev/rte_event_eth_rx_adapter.c
+++ b/lib/eventdev/rte_event_eth_rx_adapter.c
@@ -85,7 +85,9 @@ struct rte_eth_event_enqueue_buffer {
/* Count of events in this buffer */
uint16_t count;
/* Array of events in this buffer */
- struct rte_event events[ETH_EVENT_BUFFER_SIZE];
+ struct rte_event *events;
+ /* size of event buffer */
+ uint16_t events_size;
/* Event enqueue happens from head */
uint16_t head;
/* New packets from rte_eth_rx_burst is enqued from tail */
@@ -946,7 +948,7 @@ rxa_buffer_mbufs(struct rte_event_eth_rx_adapter
*rx_adapter,
dropped = 0;
nb_cb = dev_info->cb_fn(eth_dev_id, rx_queue_id,
buf->last |
- (RTE_DIM(buf->events) & ~buf->last_mask),
+ (buf->events_size & ~buf->last_mask),
buf->count >= BATCH_SIZE ?
buf->count - BATCH_SIZE : 0,
&buf->events[buf->tail],
@@ -972,7 +974,7 @@ rxa_pkt_buf_available(struct rte_eth_event_enqueue_buffer
*buf)
uint32_t nb_req = buf->tail + BATCH_SIZE;
if (!buf->last) {
- if (nb_req <= RTE_DIM(buf->events))
+ if (nb_req <= buf->events_size)
return true;
if (buf->head >= BATCH_SIZE) {
@@ -2206,12 +2208,15 @@ rxa_ctrl(uint8_t id, int start)
return 0;
}
-int
-rte_event_eth_rx_adapter_create_ext(uint8_t id, uint8_t dev_id,
- rte_event_eth_rx_adapter_conf_cb conf_cb,
- void *conf_arg)
+static int
+rxa_create(uint8_t id, uint8_t dev_id,
+ struct rte_event_eth_rx_adapter_params *rxa_params,
+ rte_event_eth_rx_adapter_conf_cb conf_cb,
+ void *conf_arg)
{
struct rte_event_eth_rx_adapter *rx_adapter;
+ struct rte_eth_event_enqueue_buffer *buf;
+ struct rte_event *events;
int ret;
int socket_id;
uint16_t i;
@@ -2226,6 +2231,7 @@ rte_event_eth_rx_adapter_create_ext(uint8_t id, uint8_t
dev_id,
RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL);
RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+
if (conf_cb == NULL)
return -EINVAL;
@@ -2273,11 +2279,30 @@ rte_event_eth_rx_adapter_create_ext(uint8_t id, uint8_t
dev_id,
rte_
[dpdk-dev] [PATCH v4 2/5] test/event: add unit test for Rx adapter
this patch adds unit test for rte_event_eth_rx_adapter_create_with_params
api and validate all possible input combinations.
Signed-off-by: Naga Harish K S V
---
app/test/test_event_eth_rx_adapter.c | 53 +---
1 file changed, 49 insertions(+), 4 deletions(-)
diff --git a/app/test/test_event_eth_rx_adapter.c
b/app/test/test_event_eth_rx_adapter.c
index 13664a3a3b..7c2cf0dd70 100644
--- a/app/test/test_event_eth_rx_adapter.c
+++ b/app/test/test_event_eth_rx_adapter.c
@@ -428,6 +428,50 @@ adapter_create_free(void)
return TEST_SUCCESS;
}
+static int
+adapter_create_free_with_params(void)
+{
+ int err;
+
+ struct rte_event_port_conf rx_p_conf = {
+ .dequeue_depth = 8,
+ .enqueue_depth = 8,
+ .new_event_threshold = 1200,
+ };
+
+ struct rte_event_eth_rx_adapter_params rxa_params = {
+ .event_buf_size = 1024
+ };
+
+ err = rte_event_eth_rx_adapter_create_with_params(TEST_INST_ID,
+ TEST_DEV_ID, NULL, NULL);
+ TEST_ASSERT(err == -EINVAL, "Expected -EINVAL got %d", err);
+
+ err = rte_event_eth_rx_adapter_create_with_params(TEST_INST_ID,
+ TEST_DEV_ID, &rx_p_conf, &rxa_params);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
+ err = rte_event_eth_rx_adapter_create_with_params(TEST_INST_ID,
+ TEST_DEV_ID, &rx_p_conf, &rxa_params);
+ TEST_ASSERT(err == -EEXIST, "Expected -EEXIST %d got %d", -EEXIST, err);
+
+ rxa_params.event_buf_size = 0;
+ err = rte_event_eth_rx_adapter_create_with_params(TEST_INST_ID,
+ TEST_DEV_ID, &rx_p_conf, &rxa_params);
+ TEST_ASSERT(err == -EINVAL, "Expected -EINVAL got %d", err);
+
+ err = rte_event_eth_rx_adapter_free(TEST_INST_ID);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
+ err = rte_event_eth_rx_adapter_free(TEST_INST_ID);
+ TEST_ASSERT(err == -EINVAL, "Expected -EINVAL %d got %d", -EINVAL, err);
+
+ err = rte_event_eth_rx_adapter_free(1);
+ TEST_ASSERT(err == -EINVAL, "Expected -EINVAL %d got %d", -EINVAL, err);
+
+ return TEST_SUCCESS;
+}
+
static int
adapter_queue_add_del(void)
{
@@ -435,7 +479,7 @@ adapter_queue_add_del(void)
struct rte_event ev;
uint32_t cap;
- struct rte_event_eth_rx_adapter_queue_conf queue_config;
+ struct rte_event_eth_rx_adapter_queue_conf queue_config = {0};
err = rte_event_eth_rx_adapter_caps_get(TEST_DEV_ID, TEST_ETHDEV_ID,
&cap);
@@ -523,7 +567,7 @@ adapter_multi_eth_add_del(void)
uint16_t port_index, port_index_base, drv_id = 0;
char driver_name[50];
- struct rte_event_eth_rx_adapter_queue_conf queue_config;
+ struct rte_event_eth_rx_adapter_queue_conf queue_config = {0};
ev.queue_id = 0;
ev.sched_type = RTE_SCHED_TYPE_ATOMIC;
@@ -594,7 +638,7 @@ adapter_intr_queue_add_del(void)
struct rte_event ev;
uint32_t cap;
uint16_t eth_port;
- struct rte_event_eth_rx_adapter_queue_conf queue_config;
+ struct rte_event_eth_rx_adapter_queue_conf queue_config = {0};
if (!default_params.rx_intr_port_inited)
return 0;
@@ -687,7 +731,7 @@ adapter_start_stop(void)
ev.sched_type = RTE_SCHED_TYPE_ATOMIC;
ev.priority = 0;
- struct rte_event_eth_rx_adapter_queue_conf queue_config;
+ struct rte_event_eth_rx_adapter_queue_conf queue_config = {0};
queue_config.rx_queue_flags = 0;
if (default_params.caps &
@@ -774,6 +818,7 @@ static struct unit_test_suite event_eth_rx_tests = {
.teardown = testsuite_teardown,
.unit_test_cases = {
TEST_CASE_ST(NULL, NULL, adapter_create_free),
+ TEST_CASE_ST(NULL, NULL, adapter_create_free_with_params),
TEST_CASE_ST(adapter_create, adapter_free,
adapter_queue_add_del),
TEST_CASE_ST(adapter_create, adapter_free,
--
2.25.1
[dpdk-dev] [PATCH v4 3/5] eventdev/rx_adapter: introduce per queue event buffer
To configure per queue event buffer size, application sets
``rte_event_eth_rx_adapter_params::use_queue_event_buf`` flag
as true and is passed to ``rte_event_eth_rx_adapter_create_with_params``
api.
The per queue event buffer size is populated in
``rte_event_eth_rx_adapter_queue_conf::event_buf_size`` and passed
to ``rte_event_eth_rx_adapter_queue_add`` api.
Signed-off-by: Naga Harish K S V
---
.../prog_guide/event_ethernet_rx_adapter.rst | 19 ---
lib/eventdev/rte_event_eth_rx_adapter.h | 4
2 files changed, 16 insertions(+), 7 deletions(-)
diff --git a/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
b/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
index 8526aecf57..8b58130fc5 100644
--- a/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
+++ b/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
@@ -62,12 +62,14 @@ service function and needs to create an event port for it.
The callback is
expected to fill the ``struct rte_event_eth_rx_adapter_conf structure``
passed to it.
-If the application desires to control the event buffer size, it can use the
-``rte_event_eth_rx_adapter_create_with_params()`` api. The event buffer size is
-specified using ``struct rte_event_eth_rx_adapter_params::event_buf_size``.
-The function is passed the event device to be associated with the adapter
-and port configuration for the adapter to setup an event port if the
-adapter needs to use a service function.
+If the application desires to control the event buffer size at adapter level,
+it can use the ``rte_event_eth_rx_adapter_create_with_params()`` api. The event
+buffer size is specified using ``struct rte_event_eth_rx_adapter_params::
+event_buf_size``. To configure the event buffer size at queue level, the
boolean
+flag ``struct rte_event_eth_rx_adapter_params::use_queue_event_buf`` need to be
+set to true. The function is passed the event device to be associated with
+the adapter and port configuration for the adapter to setup an event port
+if the adapter needs to use a service function.
Adding Rx Queues to the Adapter Instance
@@ -79,7 +81,9 @@ parameter. Event information for packets from this Rx queue
is encoded in the
``ev`` field of ``struct rte_event_eth_rx_adapter_queue_conf``. The
servicing_weight member of the struct rte_event_eth_rx_adapter_queue_conf
is the relative polling frequency of the Rx queue and is applicable when the
-adapter uses a service core function.
+adapter uses a service core function. The applications can configure queue
+event buffer size in ``struct
rte_event_eth_rx_adapter_queue_conf::event_buf_size``
+parameter.
.. code-block:: c
@@ -90,6 +94,7 @@ adapter uses a service core function.
queue_config.rx_queue_flags = 0;
queue_config.ev = ev;
queue_config.servicing_weight = 1;
+queue_config.event_buf_size = 1024;
err = rte_event_eth_rx_adapter_queue_add(id,
eth_dev_id,
diff --git a/lib/eventdev/rte_event_eth_rx_adapter.h
b/lib/eventdev/rte_event_eth_rx_adapter.h
index 846ca569e9..70ca427d66 100644
--- a/lib/eventdev/rte_event_eth_rx_adapter.h
+++ b/lib/eventdev/rte_event_eth_rx_adapter.h
@@ -200,6 +200,8 @@ struct rte_event_eth_rx_adapter_queue_conf {
* Valid when RTE_EVENT_ETH_RX_ADAPTER_QUEUE_EVENT_VECTOR flag is set in
* @see rte_event_eth_rx_adapter_queue_conf::rx_queue_flags.
*/
+ uint16_t event_buf_size;
+ /**< event buffer size for this queue */
};
/**
@@ -267,6 +269,8 @@ struct rte_event_eth_rx_adapter_params {
* This value is rounded up for better buffer utilization
* and performance.
*/
+ bool use_queue_event_buf;
+ /**< flag to indicate that event buffer is separate for each queue */
};
/**
--
2.25.1
[dpdk-dev] [PATCH v4 4/5] eventdev/rx_adapter: implement per queue event buffer
this patch implement the per queue event buffer with
required validations.
Signed-off-by: Naga Harish K S V
---
lib/eventdev/rte_event_eth_rx_adapter.c | 187 +---
1 file changed, 138 insertions(+), 49 deletions(-)
diff --git a/lib/eventdev/rte_event_eth_rx_adapter.c
b/lib/eventdev/rte_event_eth_rx_adapter.c
index 606db241b8..b61af0e75e 100644
--- a/lib/eventdev/rte_event_eth_rx_adapter.c
+++ b/lib/eventdev/rte_event_eth_rx_adapter.c
@@ -102,10 +102,12 @@ struct rte_event_eth_rx_adapter {
uint8_t rss_key_be[RSS_KEY_SIZE];
/* Event device identifier */
uint8_t eventdev_id;
- /* Per ethernet device structure */
- struct eth_device_info *eth_devices;
/* Event port identifier */
uint8_t event_port_id;
+ /* Flag indicating per rxq event buffer */
+ bool use_queue_event_buf;
+ /* Per ethernet device structure */
+ struct eth_device_info *eth_devices;
/* Lock to serialize config updates with service function */
rte_spinlock_t rx_lock;
/* Max mbufs processed in any service function invocation */
@@ -241,6 +243,7 @@ struct eth_rx_queue_info {
uint32_t flow_id_mask; /* Set to ~0 if app provides flow id else 0 */
uint64_t event;
struct eth_rx_vector_data vector_data;
+ struct rte_eth_event_enqueue_buffer *event_buf;
};
static struct rte_event_eth_rx_adapter **event_eth_rx_adapter;
@@ -767,10 +770,9 @@ rxa_enq_block_end_ts(struct rte_event_eth_rx_adapter
*rx_adapter,
/* Enqueue buffered events to event device */
static inline uint16_t
-rxa_flush_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter)
+rxa_flush_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter,
+ struct rte_eth_event_enqueue_buffer *buf)
{
- struct rte_eth_event_enqueue_buffer *buf =
- &rx_adapter->event_enqueue_buffer;
struct rte_event_eth_rx_adapter_stats *stats = &rx_adapter->stats;
uint16_t count = buf->last ? buf->last - buf->head : buf->count;
@@ -888,15 +890,14 @@ rxa_buffer_mbufs(struct rte_event_eth_rx_adapter
*rx_adapter,
uint16_t eth_dev_id,
uint16_t rx_queue_id,
struct rte_mbuf **mbufs,
- uint16_t num)
+ uint16_t num,
+ struct rte_eth_event_enqueue_buffer *buf)
{
uint32_t i;
struct eth_device_info *dev_info =
&rx_adapter->eth_devices[eth_dev_id];
struct eth_rx_queue_info *eth_rx_queue_info =
&dev_info->rx_queue[rx_queue_id];
- struct rte_eth_event_enqueue_buffer *buf =
- &rx_adapter->event_enqueue_buffer;
uint16_t new_tail = buf->tail;
uint64_t event = eth_rx_queue_info->event;
uint32_t flow_id_mask = eth_rx_queue_info->flow_id_mask;
@@ -995,11 +996,10 @@ rxa_eth_rx(struct rte_event_eth_rx_adapter *rx_adapter,
uint16_t queue_id,
uint32_t rx_count,
uint32_t max_rx,
- int *rxq_empty)
+ int *rxq_empty,
+ struct rte_eth_event_enqueue_buffer *buf)
{
struct rte_mbuf *mbufs[BATCH_SIZE];
- struct rte_eth_event_enqueue_buffer *buf =
- &rx_adapter->event_enqueue_buffer;
struct rte_event_eth_rx_adapter_stats *stats =
&rx_adapter->stats;
uint16_t n;
@@ -1012,7 +1012,7 @@ rxa_eth_rx(struct rte_event_eth_rx_adapter *rx_adapter,
*/
while (rxa_pkt_buf_available(buf)) {
if (buf->count >= BATCH_SIZE)
- rxa_flush_event_buffer(rx_adapter);
+ rxa_flush_event_buffer(rx_adapter, buf);
stats->rx_poll_count++;
n = rte_eth_rx_burst(port_id, queue_id, mbufs, BATCH_SIZE);
@@ -1021,14 +1021,14 @@ rxa_eth_rx(struct rte_event_eth_rx_adapter *rx_adapter,
*rxq_empty = 1;
break;
}
- rxa_buffer_mbufs(rx_adapter, port_id, queue_id, mbufs, n);
+ rxa_buffer_mbufs(rx_adapter, port_id, queue_id, mbufs, n, buf);
nb_rx += n;
if (rx_count + nb_rx > max_rx)
break;
}
if (buf->count > 0)
- rxa_flush_event_buffer(rx_adapter);
+ rxa_flush_event_buffer(rx_adapter, buf);
return nb_rx;
}
@@ -1169,7 +1169,7 @@ rxa_intr_ring_dequeue(struct rte_event_eth_rx_adapter
*rx_adapter)
ring_lock = &rx_adapter->intr_ring_lock;
if (buf->count >= BATCH_SIZE)
- rxa_flush_event_buffer(rx_adapter);
+ rxa_flush_event_buffer(rx_adapter, buf);
while (rxa_pkt_buf_available(buf)) {
struct eth_device_info *dev_info;
@@ -1221,7 +1221,7 @@ rxa_intr_ring_dequeue(struct rte_event_eth_rx_adapter
[dpdk-dev] [PATCH v4 5/5] test/event: add unit test for Rx adapter
this patch adds unit tests for checking per rx queue event buffer
feature using rte_event_eth_rx_adapter_queue_add api.
fix segfault in adapter_queue_conf unit test
Signed-off-by: Naga Harish K S V
---
app/test/test_event_eth_rx_adapter.c | 129 ++-
1 file changed, 126 insertions(+), 3 deletions(-)
diff --git a/app/test/test_event_eth_rx_adapter.c
b/app/test/test_event_eth_rx_adapter.c
index 7c2cf0dd70..5e69971b54 100644
--- a/app/test/test_event_eth_rx_adapter.c
+++ b/app/test/test_event_eth_rx_adapter.c
@@ -387,6 +387,90 @@ adapter_create(void)
return err;
}
+static int
+adapter_create_with_params(void)
+{
+ int err;
+ struct rte_event_dev_info dev_info;
+ struct rte_event_port_conf rx_p_conf;
+ struct rte_event_eth_rx_adapter_params rxa_params;
+
+ memset(&rx_p_conf, 0, sizeof(rx_p_conf));
+
+ err = rte_event_dev_info_get(TEST_DEV_ID, &dev_info);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
+ rx_p_conf.new_event_threshold = dev_info.max_num_events;
+ rx_p_conf.dequeue_depth = dev_info.max_event_port_dequeue_depth;
+ rx_p_conf.enqueue_depth = dev_info.max_event_port_enqueue_depth;
+
+ rxa_params.use_queue_event_buf = false;
+ rxa_params.event_buf_size = 0;
+
+ err = rte_event_eth_rx_adapter_create_with_params(TEST_INST_ID,
+ TEST_DEV_ID, &rx_p_conf, &rxa_params);
+ TEST_ASSERT(err == -EINVAL, "Expected -EINVAL got %d", err);
+
+ rxa_params.use_queue_event_buf = true;
+
+ err = rte_event_eth_rx_adapter_create_with_params(TEST_INST_ID,
+ TEST_DEV_ID, &rx_p_conf, &rxa_params);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
+ err = rte_event_eth_rx_adapter_create_with_params(TEST_INST_ID,
+ TEST_DEV_ID, &rx_p_conf, &rxa_params);
+ TEST_ASSERT(err == -EEXIST, "Expected -EEXIST got %d", err);
+
+ return TEST_SUCCESS;
+}
+
+static int
+adapter_queue_event_buf_test(void)
+{
+ int err;
+ struct rte_event ev;
+ uint32_t cap;
+
+ struct rte_event_eth_rx_adapter_queue_conf queue_config = {0};
+
+ err = rte_event_eth_rx_adapter_caps_get(TEST_DEV_ID, TEST_ETHDEV_ID,
+&cap);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
+ ev.queue_id = 0;
+ ev.sched_type = RTE_SCHED_TYPE_ATOMIC;
+ ev.priority = 0;
+
+ queue_config.rx_queue_flags = 0;
+ if (cap & RTE_EVENT_ETH_RX_ADAPTER_CAP_OVERRIDE_FLOW_ID) {
+ ev.flow_id = 1;
+ queue_config.rx_queue_flags =
+ RTE_EVENT_ETH_RX_ADAPTER_QUEUE_FLOW_ID_VALID;
+ }
+ queue_config.ev = ev;
+ queue_config.servicing_weight = 1;
+ queue_config.event_buf_size = 0;
+
+ err = rte_event_eth_rx_adapter_queue_add(TEST_INST_ID,
+ TEST_ETHDEV_ID, 0,
+ &queue_config);
+ TEST_ASSERT(err == -EINVAL, "Expected -EINVAL got %d", err);
+
+ queue_config.event_buf_size = 1024;
+
+ err = rte_event_eth_rx_adapter_queue_add(TEST_INST_ID,
+ TEST_ETHDEV_ID, 0,
+ &queue_config);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
+ err = rte_event_eth_rx_adapter_queue_del(TEST_INST_ID,
+ TEST_ETHDEV_ID,
+ 0);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
+ return TEST_SUCCESS;
+}
+
static void
adapter_free(void)
{
@@ -795,20 +879,57 @@ static int
adapter_queue_conf(void)
{
int err;
- struct rte_event_eth_rx_adapter_queue_conf queue_conf;
+ struct rte_event ev;
+ uint32_t cap;
+ struct rte_event_eth_rx_adapter_queue_conf queue_config = {0};
+
+ err = rte_event_eth_rx_adapter_caps_get(TEST_DEV_ID, TEST_ETHDEV_ID,
+&cap);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
+ ev.queue_id = 0;
+ ev.sched_type = RTE_SCHED_TYPE_ATOMIC;
+ ev.priority = 0;
+
+ queue_config.rx_queue_flags = 0;
+ if (cap & RTE_EVENT_ETH_RX_ADAPTER_CAP_OVERRIDE_FLOW_ID) {
+ ev.flow_id = 1;
+ queue_config.rx_queue_flags =
+ RTE_EVENT_ETH_RX_ADAPTER_QUEUE_FLOW_ID_VALID;
+ }
+ queue_config.ev = ev;
+ queue_config.servicing_weight = 1;
+ queue_config.event_buf_size = 1024;
+
+ err = rte_event_eth_rx_adapter_queue_add(TEST_INST_ID,
+ TEST_ETHDEV_ID, 0,
+ &queue_config);
+ TEST_ASSERT(err == -EINVAL, "Expected -EINVAL got %d", err);
+
+ queue_config.event_buf_size = 0;
+
+ err = rte_event_eth_rx_adapter_queue_add(TEST_INST_ID
Re: [dpdk-dev] [dpdk-stable] [PATCH v6 2/2] lib/pipeline: Fix gcc compilation error using ASan
On Thu, Sep 30, 2021 at 7:37 AM wrote:
>
> From: Zhihong Peng
Commit titles don't start with lib/.
>
> After adding ASan, the gcc compilation check will be stricter.
> "Control reaches end of non-void function" error occurs here.
Fwiw, I could not pinpoint the right version where this warning appears.
I can't see it with gcc v4.8.5 (rhel7), but I get it with gcc 11.2.1 (fc34).
Do you know which versions are affected? Just asking for info.
>
> Fixes: f38913b7fb8e (pipeline: add meter array to SWX)
Should be formatted as:
Fixes: f38913b7fb8e ("pipeline: add meter array to SWX")
Please use a git alias as suggested in the contribution guide.
> Cc: sta...@dpdk.org
>
> Signed-off-by: Xueqin Lin
> Signed-off-by: Zhihong Peng
> ---
> lib/pipeline/rte_swx_pipeline.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/lib/pipeline/rte_swx_pipeline.c b/lib/pipeline/rte_swx_pipeline.c
> index 1cd09a4b44..0acd6c6752 100644
> --- a/lib/pipeline/rte_swx_pipeline.c
> +++ b/lib/pipeline/rte_swx_pipeline.c
> @@ -4642,7 +4642,7 @@ instr_meter_translate(struct rte_swx_pipeline *p,
> return 0;
> }
>
> - CHECK(0, EINVAL);
> + return -EINVAL;
> }
>
> static inline void
> @@ -5937,7 +5937,7 @@ instr_translate(struct rte_swx_pipeline *p,
> instr,
> data);
>
> - CHECK(0, EINVAL);
> + return -EINVAL;
> }
>
> static struct instruction_data *
There are two other functions (instr_table_translate, and
instr_extern_translate) which have the same pattern in this file.
Odd that the compiler is not reporting them.
--
David Marchand
[dpdk-dev] [PATCH] net/softnic: fix memory leak of meter policy
From: Dapeng Yu
After the meter policies are created, they are not freed on device
close.
This patch fixes it.
Fixes: 5f0d54f372f0 ("ethdev: add pre-defined meter policy API")
Cc: sta...@dpdk.org
Signed-off-by: Dapeng Yu
---
drivers/net/softnic/rte_eth_softnic_meter.c | 12
1 file changed, 12 insertions(+)
diff --git a/drivers/net/softnic/rte_eth_softnic_meter.c
b/drivers/net/softnic/rte_eth_softnic_meter.c
index 5831892a39..6b02f43e31 100644
--- a/drivers/net/softnic/rte_eth_softnic_meter.c
+++ b/drivers/net/softnic/rte_eth_softnic_meter.c
@@ -52,6 +52,18 @@ softnic_mtr_free(struct pmd_internals *p)
TAILQ_REMOVE(&p->mtr.meter_profiles, mp, node);
free(mp);
}
+
+ /* Remove meter policies */
+ for ( ; ; ) {
+ struct softnic_mtr_meter_policy *mp;
+
+ mp = TAILQ_FIRST(&p->mtr.meter_policies);
+ if (mp == NULL)
+ break;
+
+ TAILQ_REMOVE(&p->mtr.meter_policies, mp, node);
+ free(mp);
+ }
}
struct softnic_mtr_meter_profile *
--
2.27.0
Re: [dpdk-dev] [PATCH v3 1/5] eventdev/rx_adapter: add event buffer size configurability
> -Original Message-
> From: Jerin Jacob
> Sent: Wednesday, September 29, 2021 10:46 AM
> To: Naga Harish K, S V
> Cc: Jerin Jacob ; Jayatheerthan, Jay
> ; dpdk-dev ; Kundapura,
> Ganapati
> Subject: Re: [dpdk-dev] [PATCH v3 1/5] eventdev/rx_adapter: add event buffer
> size configurability
>
> On Wed, Sep 22, 2021 at 8:44 PM Naga Harish K S V
> wrote:
> >
> > Currently event buffer is static array with a default size defined
> > internally.
> >
> > To configure event buffer size from application,
> > ``rte_event_eth_rx_adapter_create_with_params`` api is added which
> > takes ``struct rte_event_eth_rx_adapter_params`` to configure event
> > buffer size in addition other params . The event buffer size is
> > rounded up for better buffer utilization and performance . In case
> > of NULL params argument, default event buffer size is used.
> >
> > Signed-off-by: Naga Harish K S V
> > Signed-off-by: Ganapati Kundapura
>
>
> Changes look good to me.
>
> @Jayatheerthan, Jay Could review and Ack it?
@Jerin, sure. Will be able to get to it next week.
>
>
> >
> > ---
> > v3:
> > * updated documentation and code comments as per review comments.
> > * updated new create api test case name with suitable one.
> >
> > v2:
> > * Updated header file and rx adapter documentation as per review comments.
> > * new api name is modified as rte_event_eth_rx_adapter_create_with_params
> > as per review comments.
> > * rxa_params pointer argument Value NULL is allowed to represent the
> > default values
> >
> > v1:
> > * Initial implementation with documentation and unit tests.
> > ---
> > .../prog_guide/event_ethernet_rx_adapter.rst | 7 ++
> > lib/eventdev/rte_event_eth_rx_adapter.c | 98 +--
> > lib/eventdev/rte_event_eth_rx_adapter.h | 41 +++-
> > lib/eventdev/version.map | 2 +
> > 4 files changed, 140 insertions(+), 8 deletions(-)
> >
> > diff --git a/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
> > b/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
> > index 0780b6f711..dd753613bd 100644
> > --- a/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
> > +++ b/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
> > @@ -62,6 +62,13 @@ service function and needs to create an event port for
> > it. The callback is
> > expected to fill the ``struct rte_event_eth_rx_adapter_conf structure``
> > passed to it.
> >
> > +If the application desires to control the event buffer size, it can use the
> > +``rte_event_eth_rx_adapter_create_with_params()`` api. The event buffer
> > size is
> > +specified using ``struct rte_event_eth_rx_adapter_params::event_buf_size``.
> > +The function is passed the event device to be associated with the adapter
> > +and port configuration for the adapter to setup an event port if the
> > +adapter needs to use a service function.
> > +
> > Adding Rx Queues to the Adapter Instance
> >
> >
> > diff --git a/lib/eventdev/rte_event_eth_rx_adapter.c
> > b/lib/eventdev/rte_event_eth_rx_adapter.c
> > index f2dc69503d..7dec9a8734 100644
> > --- a/lib/eventdev/rte_event_eth_rx_adapter.c
> > +++ b/lib/eventdev/rte_event_eth_rx_adapter.c
> > @@ -82,7 +82,9 @@ struct rte_eth_event_enqueue_buffer {
> > /* Count of events in this buffer */
> > uint16_t count;
> > /* Array of events in this buffer */
> > - struct rte_event events[ETH_EVENT_BUFFER_SIZE];
> > + struct rte_event *events;
> > + /* size of event buffer */
> > + uint16_t events_size;
> > /* Event enqueue happens from head */
> > uint16_t head;
> > /* New packets from rte_eth_rx_burst is enqued from tail */
> > @@ -919,7 +921,7 @@ rxa_buffer_mbufs(struct rte_event_eth_rx_adapter
> > *rx_adapter,
> > dropped = 0;
> > nb_cb = dev_info->cb_fn(eth_dev_id, rx_queue_id,
> >buf->last |
> > - (RTE_DIM(buf->events) &
> > ~buf->last_mask),
> > + (buf->events_size & ~buf->last_mask),
> >buf->count >= BATCH_SIZE ?
> > buf->count - BATCH_SIZE : 0,
> >&buf->events[buf->tail],
> > @@ -945,7 +947,7 @@ rxa_pkt_buf_available(struct
> > rte_eth_event_enqueue_buffer *buf)
> > uint32_t nb_req = buf->tail + BATCH_SIZE;
> >
> > if (!buf->last) {
> > - if (nb_req <= RTE_DIM(buf->events))
> > + if (nb_req <= buf->events_size)
> > return true;
> >
> > if (buf->head >= BATCH_SIZE) {
> > @@ -2164,12 +2166,15 @@ rxa_ctrl(uint8_t id, int start)
> > return 0;
> > }
> >
> > -int
> > -rte_event_eth_rx_adapter_create_ext(uint8_t id, uint8_t dev_id,
> > - rte_event_eth_rx_adapter_conf_cb con
Re: [dpdk-dev] [PATCH v2 0/7] Removal of PCI bus ABIs
Hello Chenbo, On Wed, Sep 29, 2021 at 9:38 AM Xia, Chenbo wrote: > > Gentle ping for comments.. Sorry, I'll try to look at it. > > @David, could you help me understand what is the compile error in Fedora 31? > DPDK_compile_spdk failure is expected as the header name for SPDK is changed, > I am not sure if it's the same error... The error log is odd (no compilation "backtrace"). You'll need to test spdk manually I guess. -- David Marchand
Re: [dpdk-dev] [v2] telemetry: fix json output buffer size
Hi Gowrishankar,
>-Original Message-
>From: Gowrishankar Muthukrishnan
>Sent: Thursday 23 September 2021 06:53
>To: Power, Ciara ; dev@dpdk.org
>Cc: Richardson, Bruce
>Subject: RE: [v2] telemetry: fix json output buffer size
>
>Hi Ciara,
>> I am not sure about why we would want this to allow for
>> "RTE_TEL_MAX_SINGLE_STRING_LEN - 6".
>> The RTE_TEL_MAX_SINGLE_STRING_LEN is used to represent the max size
>of a
>> singular string value e.g. the response to client being {"" : "value
>> here up to max size in length>" }
>>
>> I wonder could we use the "len" parameter in some way here, that would
>> be the available space to be filled of the "buf" being passed in,
>> allowing the function to copy in the maximum amount to fill the buffer.
>
>Got it. Yeah, "len" is actual available space. I ll send next version based on
>this.
>
>Also, I propose if we can have platform defined upper limits (esp
>MAX_CMD_LEN, MAX_SINGLE_STRING_LEN etc) so that, we need not revisit
>lib/telemetry for platform needs (and I don't think one size fits all platform,
>may be excess too).
>Thoughts ?
I am not sure why it is needed to have limits defined per platform - can you
explain further about why it is necessary?
Thanks,
Ciara
>
>Thanks,
>Gowrishankar
>
>>
>> Thanks,
>> Ciara
Re: [dpdk-dev] [v2] telemetry: fix json output buffer size
> >Also, I propose if we can have platform defined upper limits (esp > >MAX_CMD_LEN, MAX_SINGLE_STRING_LEN etc) so that, we need not revisit > >lib/telemetry for platform needs (and I don't think one size fits all > >platform, may be excess too). > >Thoughts ? > > I am not sure why it is needed to have limits defined per platform - can you > explain further about why it is necessary? > Mainly, for the endpoint in driver. In case, if the endpoint data is bigger than MAX_SINGLE_STRING_LEN at the worst case, endpoint will not work correctly. Thanks, Gowrishankar
Re: [dpdk-dev] [PATCH v6 04/10] mbuf: add IPsec ESP tunnel type
On 9/23/2021 1:59 PM, Ananyev, Konstantin wrote:
Add tunnel type for IPsec ESP tunnels
Signed-off-by: Declan Doherty
Signed-off-by: Radu Nicolau
Signed-off-by: Abhijit Sinha
Signed-off-by: Daniel Martin Buckley
Acked-by: Fan Zhang
Acked-by: Akhil Goyal
Acked-by: Olivier Matz
---
lib/mbuf/rte_mbuf_core.h | 1 +
1 file changed, 1 insertion(+)
diff --git a/lib/mbuf/rte_mbuf_core.h b/lib/mbuf/rte_mbuf_core.h
index bb38d7f581..a4d95deee6 100644
--- a/lib/mbuf/rte_mbuf_core.h
+++ b/lib/mbuf/rte_mbuf_core.h
@@ -253,6 +253,7 @@ extern "C" {
#define PKT_TX_TUNNEL_MPLSINUDP (0x5ULL << 45)
#define PKT_TX_TUNNEL_VXLAN_GPE (0x6ULL << 45)
#define PKT_TX_TUNNEL_GTP (0x7ULL << 45)
+#define PKT_TX_TUNNEL_ESP (0x8ULL << 45)
As I can see, that's not ptype, that's TX flag.
Could you clarify what exactly what flag would mean for PMD TX:
- what is expected from the user who sets this flag
- what is expected from PMD that claims to support it.
BTW, would we need new DEV_TX_OFFLOAD_* for it?
There is documentation above for the other tunnel types, they are
supposed to be used for TSO purposes. I will update the commit message
to clarify this.
/**
* Generic IP encapsulated tunnel type, used for TSO and checksum offload.
* It can be used for tunnels which are not standards or listed above.
--
2.25.1
Re: [dpdk-dev] Questions about vm2vm vhost-user/virtio-net test
Hi, all, I got the answer: testpmd fwd mode should be set "io", then ping OK. IO fwd mode, it will not change packet MAC address info, and ping OK MAC fwd mode, it will change packet MAC address info, like: "rte_ether_addr_copy(&peer_eth_addrs[fs->peer_addr], ð_hdr->d_addr); rte_ether_addr_copy(&ports[fs->tx_port].eth_addr, ð_hdr->s_addr)" Then, ping failed. So, everyone, I got one question: how could this happen? router change packet src MAC and dst MAC, but ping OK, testpmd vhost, treated as switch, also changes packert src MAC and dst MAC, but ping failed ? 在 2021/9/29 20:09, Min Hu (Connor) 写道: Hi, Coquelin, Chenbo, all, I want to seek help about vm2vm vhost-user/virtio-net test from you. When I set up vm2vm vhost-user/virtio-net test, I cannot ping vm2 in vm1. That is, ping failed between vm1 and vm2. Detailed description are as flows: host configuration: Linux localhost 5.14.0-rc4+ #1 SMP PREEMPT Sun Sep 26 16:52:13 CST 2021 aarch64 aarch64 aarch64 GNU/Linux 1, set up vhost on host cd /home/humin/virtio_test rm -rf ./vhost-net* /usr/app/testpmd -l 2-4 -n 4 --no-pci --file-prefix=vhost --vdev 'net_vhost0,iface=vhost-net0,queues=1' --vdev 'net_vhost1,iface=vhost-net1,queues=1' -- -i --nb-cores=2 --txd=1024 --rxd=1024 testpmd> set fwd mac testpmd> start 2, create vm1 and vm2 taskset -c 13 qemu-system-aarch64 \ -name us-vhost-vm1 \ -kernel /home/humin/virtio_test/Image \ -initrd /home/humin/virtio_test/rootfs.cpio.gz \ -machine virt,gic-version=3 -nographic \ -cpu host -enable-kvm -m 4096 \ -object memory-backend-file,id=mem,size=4096M,mem-path=/dev/hugepages,share=on \ -numa node,memdev=mem \ -mem-prealloc \ -smp cores=4,sockets=1 \ -monitor unix:/tmp/vm2_monitor.sock,server,nowait \ -net user,hostfwd=tcp:127.0.0.1:6004-:22 \ -chardev socket,id=char0,path=/home/humin/virtio_test/vhost-net0 \ -netdev type=vhost-user,id=mynet1,chardev=char0,vhostforce \ -device virtio-net-pci,mac=52:54:00:00:00:01,netdev=mynet1,mrg_rxbuf=on,csum=on,guest_csum=on,guest_ecn=on \ -vnc :15 taskset -c 15 qemu-system-aarch64 \ -name us-vhost-vm2 \ -kernel /home/humin/virtio_test/Image \ -initrd /home/humin/virtio_test/rootfs.cpio.gz \ -machine virt,gic-version=3 -nographic \ -cpu host -enable-kvm -m 4096 \ -object memory-backend-file,id=mem,size=4096M,mem-path=/dev/hugepages,share=on \ -numa node,memdev=mem \ -mem-prealloc \ -smp cores=4,sockets=1 \ -monitor unix:/tmp/vm2_monitor.sock,server,nowait \ -net user,hostfwd=tcp:127.0.0.1:6005-:22 \ -chardev socket,id=char1,path=/home/humin/virtio_test/vhost-net1 \ -netdev type=vhost-user,id=mynet2,chardev=char1,vhostforce \ -device virtio-net-pci,mac=52:54:00:00:00:02,netdev=mynet2,mrg_rxbuf=on,csum=on,guest_csum=on,guest_ecn=on \ -vnc :16 3, do test in vm1: ifconfig eth0 up ifconfig eth0 1.1.1.2 arp -s 1.1.1.8 52:54:00:00:00:02 in vm2: ifconfig eth0 up ifconfig eth0 1.1.1.8 arp -s 1.1.1.2 52:54:00:00:00:01 Then in vm1: ping 1.1.1.8 the result: no icmp reply, ping failed. same for vm2. 4, Try to debug using tcpdump to capture packets in vm2, we can get only the ICMP request packet which is from vm1, but no ICMP reply. It looks like ping task does not send ICMP reply, but it is just guess for it. BTW, I also use iperf test: UDP packet test: vm1: iperf -s -i 1 vm2: iperf -c 1.1.1.2 -u -i 1 -t 60 Then vm1 can get the packet. Same result if vm1 is used as client. TCP packet test: vm1: iperf -s -i 1 vm2: iperf -c 1.1.1.2 -i 1 -t 60 Then vm1 can NOT get the packet. Same result if vm1 is used as client. 5, reference I refre to the website https://doc.dpdk.org/dts/test_plans/vm2vm_virtio_net_perf_test_plan.html In 199.7. Test Case 5: scp test success. But my scp test also failed, any one could help me? thanks very much. .
[dpdk-dev] [PATCH v3 01/27] common/cnxk: update policer MBOX APIs and HW definitions
From: Sunil Kumar Kori
To support ingress policer on CN10K, MBOX interfaces and HW
definitions are synced.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/hw/nix.h | 13 ++---
drivers/common/cnxk/roc_mbox.h | 34 +-
2 files changed, 43 insertions(+), 4 deletions(-)
diff --git a/drivers/common/cnxk/hw/nix.h b/drivers/common/cnxk/hw/nix.h
index 6b86002ead..53cdfbb142 100644
--- a/drivers/common/cnxk/hw/nix.h
+++ b/drivers/common/cnxk/hw/nix.h
@@ -692,9 +692,16 @@
#define NIX_RX_BAND_PROF_ACTIONRESULT_DROP (0x1ull) /* [CN10K, .) */
#define NIX_RX_BAND_PROF_ACTIONRESULT_RED (0x2ull) /* [CN10K, .) */
-#define NIX_RX_BAND_PROF_LAYER_LEAF (0x0ull) /* [CN10K, .) */
-#define NIX_RX_BAND_PROF_LAYER_MIDDLE (0x1ull) /* [CN10K, .) */
-#define NIX_RX_BAND_PROF_LAYER_TOP(0x2ull) /* [CN10K, .) */
+#define NIX_RX_BAND_PROF_LAYER_LEAF(0x0ull) /* [CN10K, .) */
+#define NIX_RX_BAND_PROF_LAYER_INVALID (0x1ull) /* [CN10K, .) */
+#define NIX_RX_BAND_PROF_LAYER_MIDDLE (0x2ull) /* [CN10K, .) */
+#define NIX_RX_BAND_PROF_LAYER_TOP (0x3ull) /* [CN10K, .) */
+#define NIX_RX_BAND_PROF_LAYER_MAX (0x4ull) /* [CN10K, .) */
+
+#define NIX_RX_BAND_PROF_PC_MODE_VLAN (0x0ull) /* [CN10K, .) */
+#define NIX_RX_BAND_PROF_PC_MODE_DSCP (0x1ull) /* [CN10K, .) */
+#define NIX_RX_BAND_PROF_PC_MODE_GEN (0x2ull) /* [CN10K, .) */
+#define NIX_RX_BAND_PROF_PC_MODE_RSVD (0x3ull) /* [CN10K, .) */
#define NIX_RX_COLORRESULT_GREEN (0x0ull) /* [CN10K, .) */
#define NIX_RX_COLORRESULT_YELLOW (0x1ull) /* [CN10K, .) */
diff --git a/drivers/common/cnxk/roc_mbox.h b/drivers/common/cnxk/roc_mbox.h
index b5da931b81..c8b97e9aee 100644
--- a/drivers/common/cnxk/roc_mbox.h
+++ b/drivers/common/cnxk/roc_mbox.h
@@ -234,7 +234,11 @@ struct mbox_msghdr {
nix_inline_ipsec_lf_cfg, msg_rsp)\
M(NIX_CN10K_AQ_ENQ, 0x801b, nix_cn10k_aq_enq, nix_cn10k_aq_enq_req,\
nix_cn10k_aq_enq_rsp)\
- M(NIX_GET_HW_INFO, 0x801c, nix_get_hw_info, msg_req, nix_hw_info)
+ M(NIX_GET_HW_INFO, 0x801c, nix_get_hw_info, msg_req, nix_hw_info) \
+ M(NIX_BANDPROF_ALLOC, 0x801d, nix_bandprof_alloc, \
+ nix_bandprof_alloc_req, nix_bandprof_alloc_rsp) \
+ M(NIX_BANDPROF_FREE, 0x801e, nix_bandprof_free, nix_bandprof_free_req, \
+ msg_rsp)
/* Messages initiated by AF (range 0xC00 - 0xDFF) */
#define MBOX_UP_CGX_MESSAGES
\
@@ -771,6 +775,10 @@ struct nix_cn10k_aq_enq_req {
__io struct nix_rsse_s rss;
/* Valid when op == WRITE/INIT and ctype == NIX_AQ_CTYPE_MCE */
__io struct nix_rx_mce_s mce;
+ /* Valid when op == WRITE/INIT and
+* ctype == NIX_AQ_CTYPE_BAND_PROF
+*/
+ __io struct nix_band_prof_s prof;
};
/* Mask data when op == WRITE (1=write, 0=don't write) */
union {
@@ -784,6 +792,8 @@ struct nix_cn10k_aq_enq_req {
__io struct nix_rsse_s rss_mask;
/* Valid when op == WRITE and ctype == NIX_AQ_CTYPE_MCE */
__io struct nix_rx_mce_s mce_mask;
+ /* Valid when op == WRITE and ctype == NIX_AQ_CTYPE_BAND_PROF */
+ __io struct nix_band_prof_s prof_mask;
};
};
@@ -795,6 +805,7 @@ struct nix_cn10k_aq_enq_rsp {
struct nix_cq_ctx_s cq;
struct nix_rsse_s rss;
struct nix_rx_mce_s mce;
+ struct nix_band_prof_s prof;
};
};
@@ -1129,6 +1140,27 @@ struct nix_hw_info {
uint16_t __io rsvd[15];
};
+struct nix_bandprof_alloc_req {
+ struct mbox_msghdr hdr;
+ /* Count of profiles needed per layer */
+ uint16_t __io prof_count[NIX_RX_BAND_PROF_LAYER_MAX];
+};
+
+struct nix_bandprof_alloc_rsp {
+ struct mbox_msghdr hdr;
+ uint16_t __io prof_count[NIX_RX_BAND_PROF_LAYER_MAX];
+
+#define BANDPROF_PER_PFFUNC 64
+ uint16_t __io prof_idx[NIX_RX_BAND_PROF_LAYER_MAX][BANDPROF_PER_PFFUNC];
+};
+
+struct nix_bandprof_free_req {
+ struct mbox_msghdr hdr;
+ uint8_t __io free_all;
+ uint16_t __io prof_count[NIX_RX_BAND_PROF_LAYER_MAX];
+ uint16_t __io prof_idx[NIX_RX_BAND_PROF_LAYER_MAX][BANDPROF_PER_PFFUNC];
+};
+
/* SSO mailbox error codes
* Range 501 - 600.
*/
--
2.25.1
[dpdk-dev] [PATCH v3 02/27] common/cnxk: support RoC API to get level to index
From: Sunil Kumar Kori
CN10K platform supports policer up to 3 level of hierarchy.
Implement RoC API to get corresponding index for given level.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/meson.build| 1 +
drivers/common/cnxk/roc_nix.h | 11 +++
drivers/common/cnxk/roc_nix_bpf.c | 22 ++
drivers/common/cnxk/roc_nix_priv.h | 1 +
drivers/common/cnxk/roc_utils.c| 3 +++
drivers/common/cnxk/version.map| 1 +
6 files changed, 39 insertions(+)
create mode 100644 drivers/common/cnxk/roc_nix_bpf.c
diff --git a/drivers/common/cnxk/meson.build b/drivers/common/cnxk/meson.build
index 8a551d15d6..62901e66e7 100644
--- a/drivers/common/cnxk/meson.build
+++ b/drivers/common/cnxk/meson.build
@@ -25,6 +25,7 @@ sources = files(
'roc_mbox.c',
'roc_model.c',
'roc_nix.c',
+'roc_nix_bpf.c',
'roc_nix_debug.c',
'roc_nix_fc.c',
'roc_nix_irq.c',
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index b0e6fabe31..1488c24f59 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -6,6 +6,8 @@
#define _ROC_NIX_H_
/* Constants */
+#define ROC_NIX_BPF_LEVEL_IDX_INVALID 0xFF
+
enum roc_nix_rss_reta_sz {
ROC_NIX_RSS_RETA_SZ_64 = 64,
ROC_NIX_RSS_RETA_SZ_128 = 128,
@@ -29,6 +31,12 @@ enum roc_nix_vlan_type {
ROC_NIX_VLAN_TYPE_OUTER = 0x02,
};
+enum roc_nix_bpf_level_flag {
+ ROC_NIX_BPF_LEVEL_F_LEAF = BIT(0),
+ ROC_NIX_BPF_LEVEL_F_MID = BIT(1),
+ ROC_NIX_BPF_LEVEL_F_TOP = BIT(2),
+};
+
struct roc_nix_vlan_config {
uint32_t type;
union {
@@ -468,6 +476,9 @@ int __roc_api roc_nix_tm_rsrc_count(struct roc_nix *roc_nix,
int __roc_api roc_nix_tm_node_name_get(struct roc_nix *roc_nix,
uint32_t node_id, char *buf,
size_t buflen);
+/* Ingress Policer API */
+uint8_t __roc_api
+roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag lvl_flag);
/* MAC */
int __roc_api roc_nix_mac_rxtx_start_stop(struct roc_nix *roc_nix, bool start);
diff --git a/drivers/common/cnxk/roc_nix_bpf.c
b/drivers/common/cnxk/roc_nix_bpf.c
new file mode 100644
index 00..b588cc16e4
--- /dev/null
+++ b/drivers/common/cnxk/roc_nix_bpf.c
@@ -0,0 +1,22 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2021 Marvell.
+ */
+
+#include "roc_api.h"
+#include "roc_priv.h"
+
+uint8_t
+roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag level_f)
+{
+ uint8_t idx;
+
+ if (level_f & ROC_NIX_BPF_LEVEL_F_LEAF)
+ idx = 0;
+ else if (level_f & ROC_NIX_BPF_LEVEL_F_MID)
+ idx = 1;
+ else if (level_f & ROC_NIX_BPF_LEVEL_F_TOP)
+ idx = 2;
+ else
+ idx = ROC_NIX_BPF_LEVEL_IDX_INVALID;
+ return idx;
+}
diff --git a/drivers/common/cnxk/roc_nix_priv.h
b/drivers/common/cnxk/roc_nix_priv.h
index 9dc0c88a6f..94040fc744 100644
--- a/drivers/common/cnxk/roc_nix_priv.h
+++ b/drivers/common/cnxk/roc_nix_priv.h
@@ -170,6 +170,7 @@ enum nix_err_status {
NIX_ERR_INVALID_RANGE,
NIX_ERR_INTERNAL,
NIX_ERR_OP_NOTSUP,
+ NIX_ERR_HW_NOTSUP,
NIX_ERR_QUEUE_INVALID_RANGE,
NIX_ERR_AQ_READ_FAILED,
NIX_ERR_AQ_WRITE_FAILED,
diff --git a/drivers/common/cnxk/roc_utils.c b/drivers/common/cnxk/roc_utils.c
index 9cb8708a74..77dc279bcb 100644
--- a/drivers/common/cnxk/roc_utils.c
+++ b/drivers/common/cnxk/roc_utils.c
@@ -34,6 +34,9 @@ roc_error_msg_get(int errorcode)
case NIX_ERR_OP_NOTSUP:
err_msg = "Operation not supported";
break;
+ case NIX_ERR_HW_NOTSUP:
+ err_msg = "Hardware does not support";
+ break;
case NIX_ERR_QUEUE_INVALID_RANGE:
err_msg = "Invalid Queue range";
break;
diff --git a/drivers/common/cnxk/version.map b/drivers/common/cnxk/version.map
index 5df2e56ce6..c19f74fe91 100644
--- a/drivers/common/cnxk/version.map
+++ b/drivers/common/cnxk/version.map
@@ -77,6 +77,7 @@ INTERNAL {
roc_model;
roc_se_auth_key_set;
roc_se_ciph_key_set;
+ roc_nix_bpf_level_to_idx;
roc_nix_cq_dump;
roc_nix_cq_fini;
roc_nix_cq_init;
--
2.25.1
[dpdk-dev] [PATCH v3 03/27] common/cnxk: support RoC API to get profile count
From: Sunil Kumar Kori
Implement interface to get available profile count for given
nixlf.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/roc_nix.h | 5
drivers/common/cnxk/roc_nix_bpf.c | 46 +++
drivers/common/cnxk/version.map | 1 +
3 files changed, 52 insertions(+)
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index 1488c24f59..3d3e169977 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -7,6 +7,7 @@
/* Constants */
#define ROC_NIX_BPF_LEVEL_IDX_INVALID 0xFF
+#define ROC_NIX_BPF_LEVEL_MAX3
enum roc_nix_rss_reta_sz {
ROC_NIX_RSS_RETA_SZ_64 = 64,
@@ -477,6 +478,10 @@ int __roc_api roc_nix_tm_node_name_get(struct roc_nix
*roc_nix,
uint32_t node_id, char *buf,
size_t buflen);
/* Ingress Policer API */
+int __roc_api
+roc_nix_bpf_count_get(struct roc_nix *roc_nix, uint8_t lvl_mask,
+ uint16_t count[ROC_NIX_BPF_LEVEL_MAX] /* Out */);
+
uint8_t __roc_api
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag lvl_flag);
diff --git a/drivers/common/cnxk/roc_nix_bpf.c
b/drivers/common/cnxk/roc_nix_bpf.c
index b588cc16e4..af9dffa90c 100644
--- a/drivers/common/cnxk/roc_nix_bpf.c
+++ b/drivers/common/cnxk/roc_nix_bpf.c
@@ -5,6 +5,14 @@
#include "roc_api.h"
#include "roc_priv.h"
+#define NIX_MAX_BPF_COUNT_LEAF_LAYER 64
+#define NIX_MAX_BPF_COUNT_MID_LAYER 8
+#define NIX_MAX_BPF_COUNT_TOP_LAYER 1
+
+#define NIX_BPF_LEVEL_F_MASK
\
+ (ROC_NIX_BPF_LEVEL_F_LEAF | ROC_NIX_BPF_LEVEL_F_MID | \
+ROC_NIX_BPF_LEVEL_F_TOP)
+
uint8_t
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag level_f)
{
@@ -20,3 +28,41 @@ roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag level_f)
idx = ROC_NIX_BPF_LEVEL_IDX_INVALID;
return idx;
}
+
+int
+roc_nix_bpf_count_get(struct roc_nix *roc_nix, uint8_t lvl_mask,
+ uint16_t count[ROC_NIX_BPF_LEVEL_MAX])
+{
+ uint8_t mask = lvl_mask & NIX_BPF_LEVEL_F_MASK;
+ uint8_t leaf_idx, mid_idx, top_idx;
+
+ PLT_SET_USED(roc_nix);
+
+ if (roc_model_is_cn9k())
+ return NIX_ERR_HW_NOTSUP;
+
+ if (!mask)
+ return NIX_ERR_PARAM;
+
+ /* Currently No MBOX interface is available to get number
+* of bandwidth profiles. So numbers per level are hard coded,
+* considering 3 RPM blocks and each block has 4 LMAC's.
+* So total 12 physical interfaces are in system. Each interface
+* supports following bandwidth profiles.
+*/
+
+ leaf_idx = roc_nix_bpf_level_to_idx(mask & ROC_NIX_BPF_LEVEL_F_LEAF);
+ mid_idx = roc_nix_bpf_level_to_idx(mask & ROC_NIX_BPF_LEVEL_F_MID);
+ top_idx = roc_nix_bpf_level_to_idx(mask & ROC_NIX_BPF_LEVEL_F_TOP);
+
+ if (leaf_idx != ROC_NIX_BPF_LEVEL_IDX_INVALID)
+ count[leaf_idx] = NIX_MAX_BPF_COUNT_LEAF_LAYER;
+
+ if (mid_idx != ROC_NIX_BPF_LEVEL_IDX_INVALID)
+ count[mid_idx] = NIX_MAX_BPF_COUNT_MID_LAYER;
+
+ if (top_idx != ROC_NIX_BPF_LEVEL_IDX_INVALID)
+ count[top_idx] = NIX_MAX_BPF_COUNT_TOP_LAYER;
+
+ return 0;
+}
diff --git a/drivers/common/cnxk/version.map b/drivers/common/cnxk/version.map
index c19f74fe91..790a32e2e0 100644
--- a/drivers/common/cnxk/version.map
+++ b/drivers/common/cnxk/version.map
@@ -77,6 +77,7 @@ INTERNAL {
roc_model;
roc_se_auth_key_set;
roc_se_ciph_key_set;
+ roc_nix_bpf_count_get;
roc_nix_bpf_level_to_idx;
roc_nix_cq_dump;
roc_nix_cq_fini;
--
2.25.1
[dpdk-dev] [PATCH v3 04/27] common/cnxk: support RoC API to alloc bandwidth profiles
From: Sunil Kumar Kori
Implement RoC API to allocate HW resources i.e. bandwidth
profiles for policer processing on CN10K platform.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/roc_nix.h | 11
drivers/common/cnxk/roc_nix_bpf.c | 104 ++
drivers/common/cnxk/version.map | 1 +
3 files changed, 116 insertions(+)
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index 3d3e169977..b55333a01c 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -6,6 +6,7 @@
#define _ROC_NIX_H_
/* Constants */
+#define ROC_NIX_BPF_PER_PFFUNC 64
#define ROC_NIX_BPF_LEVEL_IDX_INVALID 0xFF
#define ROC_NIX_BPF_LEVEL_MAX3
@@ -38,6 +39,12 @@ enum roc_nix_bpf_level_flag {
ROC_NIX_BPF_LEVEL_F_TOP = BIT(2),
};
+struct roc_nix_bpf_objs {
+ uint16_t level;
+ uint16_t count;
+ uint16_t ids[ROC_NIX_BPF_PER_PFFUNC];
+};
+
struct roc_nix_vlan_config {
uint32_t type;
union {
@@ -482,6 +489,10 @@ int __roc_api
roc_nix_bpf_count_get(struct roc_nix *roc_nix, uint8_t lvl_mask,
uint16_t count[ROC_NIX_BPF_LEVEL_MAX] /* Out */);
+int __roc_api roc_nix_bpf_alloc(struct roc_nix *roc_nix, uint8_t lvl_mask,
+ uint16_t per_lvl_cnt[ROC_NIX_BPF_LEVEL_MAX],
+ struct roc_nix_bpf_objs *profs /* Out */);
+
uint8_t __roc_api
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag lvl_flag);
diff --git a/drivers/common/cnxk/roc_nix_bpf.c
b/drivers/common/cnxk/roc_nix_bpf.c
index af9dffa90c..06394bda07 100644
--- a/drivers/common/cnxk/roc_nix_bpf.c
+++ b/drivers/common/cnxk/roc_nix_bpf.c
@@ -13,6 +13,19 @@
(ROC_NIX_BPF_LEVEL_F_LEAF | ROC_NIX_BPF_LEVEL_F_MID | \
ROC_NIX_BPF_LEVEL_F_TOP)
+static uint8_t sw_to_hw_lvl_map[] = {NIX_RX_BAND_PROF_LAYER_LEAF,
+NIX_RX_BAND_PROF_LAYER_MIDDLE,
+NIX_RX_BAND_PROF_LAYER_TOP};
+
+static inline struct mbox *
+get_mbox(struct roc_nix *roc_nix)
+{
+ struct nix *nix = roc_nix_to_nix_priv(roc_nix);
+ struct dev *dev = &nix->dev;
+
+ return dev->mbox;
+}
+
uint8_t
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag level_f)
{
@@ -66,3 +79,94 @@ roc_nix_bpf_count_get(struct roc_nix *roc_nix, uint8_t
lvl_mask,
return 0;
}
+
+int
+roc_nix_bpf_alloc(struct roc_nix *roc_nix, uint8_t lvl_mask,
+ uint16_t per_lvl_cnt[ROC_NIX_BPF_LEVEL_MAX],
+ struct roc_nix_bpf_objs *profs)
+{
+ uint8_t mask = lvl_mask & NIX_BPF_LEVEL_F_MASK;
+ struct mbox *mbox = get_mbox(roc_nix);
+ struct nix_bandprof_alloc_req *req;
+ struct nix_bandprof_alloc_rsp *rsp;
+ uint8_t leaf_idx, mid_idx, top_idx;
+ int rc = -ENOSPC, i;
+
+ if (roc_model_is_cn9k())
+ return NIX_ERR_HW_NOTSUP;
+
+ if (!mask)
+ return NIX_ERR_PARAM;
+
+ leaf_idx = roc_nix_bpf_level_to_idx(mask & ROC_NIX_BPF_LEVEL_F_LEAF);
+ mid_idx = roc_nix_bpf_level_to_idx(mask & ROC_NIX_BPF_LEVEL_F_MID);
+ top_idx = roc_nix_bpf_level_to_idx(mask & ROC_NIX_BPF_LEVEL_F_TOP);
+
+ if ((leaf_idx != ROC_NIX_BPF_LEVEL_IDX_INVALID) &&
+ (per_lvl_cnt[leaf_idx] > NIX_MAX_BPF_COUNT_LEAF_LAYER))
+ return NIX_ERR_INVALID_RANGE;
+
+ if ((mid_idx != ROC_NIX_BPF_LEVEL_IDX_INVALID) &&
+ (per_lvl_cnt[mid_idx] > NIX_MAX_BPF_COUNT_MID_LAYER))
+ return NIX_ERR_INVALID_RANGE;
+
+ if ((top_idx != ROC_NIX_BPF_LEVEL_IDX_INVALID) &&
+ (per_lvl_cnt[top_idx] > NIX_MAX_BPF_COUNT_TOP_LAYER))
+ return NIX_ERR_INVALID_RANGE;
+
+ req = mbox_alloc_msg_nix_bandprof_alloc(mbox);
+ if (req == NULL)
+ goto exit;
+
+ if (leaf_idx != ROC_NIX_BPF_LEVEL_IDX_INVALID) {
+ req->prof_count[sw_to_hw_lvl_map[leaf_idx]] =
+ per_lvl_cnt[leaf_idx];
+ }
+
+ if (mid_idx != ROC_NIX_BPF_LEVEL_IDX_INVALID) {
+ req->prof_count[sw_to_hw_lvl_map[mid_idx]] =
+ per_lvl_cnt[mid_idx];
+ }
+
+ if (top_idx != ROC_NIX_BPF_LEVEL_IDX_INVALID) {
+ req->prof_count[sw_to_hw_lvl_map[top_idx]] =
+ per_lvl_cnt[top_idx];
+ }
+
+ rc = mbox_process_msg(mbox, (void *)&rsp);
+ if (rc)
+ goto exit;
+
+ if (leaf_idx != ROC_NIX_BPF_LEVEL_IDX_INVALID) {
+ profs[leaf_idx].level = leaf_idx;
+ profs[leaf_idx].count =
+ rsp->prof_count[sw_to_hw_lvl_map[leaf_idx]];
+ for (i = 0; i < profs[leaf_
[dpdk-dev] [PATCH v3 05/27] common/cnxk: support RoC API to free bandwidth profiles
From: Sunil Kumar Kori
Implement RoC interface to free HW bandwidth profiles on
CN10K platform.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/roc_nix.h | 6 +
drivers/common/cnxk/roc_nix_bpf.c | 40 +++
drivers/common/cnxk/version.map | 2 ++
3 files changed, 48 insertions(+)
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index b55333a01c..bf451ecdbc 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -493,6 +493,12 @@ int __roc_api roc_nix_bpf_alloc(struct roc_nix *roc_nix,
uint8_t lvl_mask,
uint16_t per_lvl_cnt[ROC_NIX_BPF_LEVEL_MAX],
struct roc_nix_bpf_objs *profs /* Out */);
+int __roc_api roc_nix_bpf_free(struct roc_nix *roc_nix,
+ struct roc_nix_bpf_objs *profs,
+ uint8_t num_prof);
+
+int __roc_api roc_nix_bpf_free_all(struct roc_nix *roc_nix);
+
uint8_t __roc_api
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag lvl_flag);
diff --git a/drivers/common/cnxk/roc_nix_bpf.c
b/drivers/common/cnxk/roc_nix_bpf.c
index 06394bda07..41d31bc6cd 100644
--- a/drivers/common/cnxk/roc_nix_bpf.c
+++ b/drivers/common/cnxk/roc_nix_bpf.c
@@ -170,3 +170,43 @@ roc_nix_bpf_alloc(struct roc_nix *roc_nix, uint8_t
lvl_mask,
exit:
return rc;
}
+
+int
+roc_nix_bpf_free(struct roc_nix *roc_nix, struct roc_nix_bpf_objs *profs,
+uint8_t num_prof)
+{
+ struct mbox *mbox = get_mbox(roc_nix);
+ struct nix_bandprof_free_req *req;
+ uint8_t level;
+ int i, j;
+
+ if (num_prof >= NIX_RX_BAND_PROF_LAYER_MAX)
+ return NIX_ERR_INVALID_RANGE;
+
+ req = mbox_alloc_msg_nix_bandprof_free(mbox);
+ if (req == NULL)
+ return -ENOSPC;
+
+ for (i = 0; i < num_prof; i++) {
+ level = sw_to_hw_lvl_map[profs[i].level];
+ req->prof_count[level] = profs[i].count;
+ for (j = 0; j < profs[i].count; j++)
+ req->prof_idx[level][j] = profs[i].ids[j];
+ }
+
+ return mbox_process(mbox);
+}
+
+int
+roc_nix_bpf_free_all(struct roc_nix *roc_nix)
+{
+ struct mbox *mbox = get_mbox(roc_nix);
+ struct nix_bandprof_free_req *req;
+
+ req = mbox_alloc_msg_nix_bandprof_free(mbox);
+ if (req == NULL)
+ return -ENOSPC;
+
+ req->free_all = true;
+ return mbox_process(mbox);
+}
diff --git a/drivers/common/cnxk/version.map b/drivers/common/cnxk/version.map
index 81e76919b5..c45d524d65 100644
--- a/drivers/common/cnxk/version.map
+++ b/drivers/common/cnxk/version.map
@@ -79,6 +79,8 @@ INTERNAL {
roc_se_ciph_key_set;
roc_nix_bpf_alloc;
roc_nix_bpf_count_get;
+ roc_nix_bpf_free;
+ roc_nix_bpf_free_all;
roc_nix_bpf_level_to_idx;
roc_nix_cq_dump;
roc_nix_cq_fini;
--
2.25.1
[dpdk-dev] [PATCH v3 06/27] common/cnxk: support RoC API to configure bandwidth profile
From: Sunil Kumar Kori
Implement RoC API to configure HW bandwidth profile for
CN10K platform.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/hw/nix.h | 49 +++
drivers/common/cnxk/roc_nix.h | 60
drivers/common/cnxk/roc_nix_bpf.c | 233 ++
drivers/common/cnxk/version.map | 1 +
4 files changed, 343 insertions(+)
diff --git a/drivers/common/cnxk/hw/nix.h b/drivers/common/cnxk/hw/nix.h
index 53cdfbb142..2c1487a105 100644
--- a/drivers/common/cnxk/hw/nix.h
+++ b/drivers/common/cnxk/hw/nix.h
@@ -2112,6 +2112,55 @@ struct nix_lso_format {
#define NIX_RPM_MAX_HW_FRS 16380UL
#define NIX_MIN_HW_FRS 60UL
+/** NIX policer rate limits */
+#define NIX_BPF_MAX_RATE_DIV_EXP 12
+#define NIX_BPF_MAX_RATE_EXPONENT 0xf
+#define NIX_BPF_MAX_RATE_MANTISSA 0xff
+
+#define NIX_BPF_RATE_CONST 200ULL
+
+/* NIX rate calculation in Bits/Sec
+ * PIR_ADD = ((256 + NIX_*_PIR[RATE_MANTISSA])
+ * << NIX_*_PIR[RATE_EXPONENT]) / 256
+ * PIR = (2E6 * PIR_ADD / (1 << NIX_*_PIR[RATE_DIVIDER_EXPONENT]))
+ *
+ * CIR_ADD = ((256 + NIX_*_CIR[RATE_MANTISSA])
+ * << NIX_*_CIR[RATE_EXPONENT]) / 256
+ * CIR = (2E6 * CIR_ADD / (CCLK_TICKS << NIX_*_CIR[RATE_DIVIDER_EXPONENT]))
+ */
+#define NIX_BPF_RATE(exponent, mantissa, div_exp)
\
+ ((NIX_BPF_RATE_CONST * ((256 + (mantissa)) << (exponent))) / \
+(((1ull << (div_exp)) * 256)))
+
+/* Meter rate limits in Bits/Sec */
+#define NIX_BPF_RATE_MIN NIX_BPF_RATE(0, 0, NIX_BPF_MAX_RATE_DIV_EXP)
+#define NIX_BPF_RATE_MAX
\
+ NIX_BPF_RATE(NIX_BPF_MAX_RATE_EXPONENT, NIX_BPF_MAX_RATE_MANTISSA, 0)
+
+#define NIX_BPF_DEFAULT_ADJUST_MANTISSA 511
+#define NIX_BPF_DEFAULT_ADJUST_EXPONENT 0
+
+/** NIX burst limits */
+#define NIX_BPF_MAX_BURST_EXPONENT 0xf
+#define NIX_BPF_MAX_BURST_MANTISSA 0xff
+
+/* NIX burst calculation
+ * PIR_BURST = ((256 + NIX_*_PIR[BURST_MANTISSA])
+ * << (NIX_*_PIR[BURST_EXPONENT] + 1))
+ * / 256
+ *
+ * CIR_BURST = ((256 + NIX_*_CIR[BURST_MANTISSA])
+ * << (NIX_*_CIR[BURST_EXPONENT] + 1))
+ * / 256
+ */
+#define NIX_BPF_BURST(exponent, mantissa)
\
+ (((256 + (mantissa)) << ((exponent) + 1)) / 256)
+
+/** Meter burst limits */
+#define NIX_BPF_BURST_MIN NIX_BPF_BURST(0, 0)
+#define NIX_BPF_BURST_MAX
\
+ NIX_BPF_BURST(NIX_BPF_MAX_BURST_EXPONENT, NIX_BPF_MAX_BURST_MANTISSA)
+
/* NIX rate limits */
#define NIX_TM_MAX_RATE_DIV_EXP 12
#define NIX_TM_MAX_RATE_EXPONENT 0xf
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index bf451ecdbc..42af66ccde 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -39,6 +39,62 @@ enum roc_nix_bpf_level_flag {
ROC_NIX_BPF_LEVEL_F_TOP = BIT(2),
};
+enum roc_nix_bpf_color {
+ ROC_NIX_BPF_COLOR_GREEN,
+ ROC_NIX_BPF_COLOR_YELLOW,
+ ROC_NIX_BPF_COLOR_RED,
+ ROC_NIX_BPF_COLOR_MAX
+};
+
+enum roc_nix_bpf_algo {
+ ROC_NIX_BPF_ALGO_NONE,
+ ROC_NIX_BPF_ALGO_2698,
+ ROC_NIX_BPF_ALGO_4115,
+ ROC_NIX_BPF_ALGO_2697
+};
+
+enum roc_nix_bpf_lmode { ROC_NIX_BPF_LMODE_BYTE, ROC_NIX_BPF_LMODE_PACKET };
+
+enum roc_nix_bpf_action {
+ ROC_NIX_BPF_ACTION_PASS,
+ ROC_NIX_BPF_ACTION_DROP,
+ ROC_NIX_BPF_ACTION_RED
+};
+
+struct roc_nix_bpf_cfg {
+ enum roc_nix_bpf_algo alg;
+ enum roc_nix_bpf_lmode lmode;
+ union {
+ /* Valid when *alg* is set to ROC_NIX_BPF_ALGO_2697. */
+ struct {
+ uint64_t cir;
+ uint64_t cbs;
+ uint64_t ebs;
+ } algo2697;
+
+ /* Valid when *alg* is set to ROC_NIX_BPF_ALGO_2698. */
+ struct {
+ uint64_t cir;
+ uint64_t pir;
+ uint64_t cbs;
+ uint64_t pbs;
+ } algo2698;
+
+ /* Valid when *alg* is set to ROC_NIX_BPF_ALGO_4115. */
+ struct {
+ uint64_t cir;
+ uint64_t eir;
+ uint64_t cbs;
+ uint64_t ebs;
+ } algo4115;
+ };
+
+ enum roc_nix_bpf_action action[ROC_NIX_BPF_COLOR_MAX];
+
+ /* Reserved for future config*/
+ uint32_t rsvd[3];
+};
+
struct roc_nix_bpf_objs {
uint16_t level;
uint16_t count;
@@ -499,6 +555,10 @@ int __roc_api roc_nix_bpf_free(struct roc_nix *roc_nix
[dpdk-dev] [PATCH v3 07/27] common/cnxk: support RoC API to toggle profile state
From: Sunil Kumar Kori
Implement RoC API to enable or disable HW bandwidth profiles
on CN10K platform.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/roc_nix.h | 4
drivers/common/cnxk/roc_nix_bpf.c | 39 +++
drivers/common/cnxk/version.map | 1 +
3 files changed, 44 insertions(+)
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index 42af66ccde..2576e938be 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -214,6 +214,7 @@ struct roc_nix_stats_queue {
struct roc_nix_rq {
/* Input parameters */
uint16_t qid;
+ uint16_t bpf_id;
uint64_t aura_handle;
bool ipsech_ena;
uint16_t first_skip;
@@ -559,6 +560,9 @@ int __roc_api roc_nix_bpf_config(struct roc_nix *roc_nix,
uint16_t id,
enum roc_nix_bpf_level_flag lvl_flag,
struct roc_nix_bpf_cfg *cfg);
+int __roc_api roc_nix_bpf_ena_dis(struct roc_nix *roc_nix, uint16_t id,
+ struct roc_nix_rq *rq, bool enable);
+
uint8_t __roc_api
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag lvl_flag);
diff --git a/drivers/common/cnxk/roc_nix_bpf.c
b/drivers/common/cnxk/roc_nix_bpf.c
index 37631c134c..557b6279e6 100644
--- a/drivers/common/cnxk/roc_nix_bpf.c
+++ b/drivers/common/cnxk/roc_nix_bpf.c
@@ -443,3 +443,42 @@ roc_nix_bpf_config(struct roc_nix *roc_nix, uint16_t id,
return mbox_process(mbox);
}
+
+int
+roc_nix_bpf_ena_dis(struct roc_nix *roc_nix, uint16_t id, struct roc_nix_rq
*rq,
+ bool enable)
+{
+ struct nix *nix = roc_nix_to_nix_priv(roc_nix);
+ struct mbox *mbox = get_mbox(roc_nix);
+ struct nix_cn10k_aq_enq_req *aq;
+ int rc;
+
+ if (roc_model_is_cn9k())
+ return NIX_ERR_HW_NOTSUP;
+
+ if (rq->qid >= nix->nb_rx_queues)
+ return NIX_ERR_QUEUE_INVALID_RANGE;
+
+ aq = mbox_alloc_msg_nix_cn10k_aq_enq(mbox);
+ if (aq == NULL)
+ return -ENOSPC;
+ aq->qidx = rq->qid;
+ aq->ctype = NIX_AQ_CTYPE_RQ;
+ aq->op = NIX_AQ_INSTOP_WRITE;
+
+ aq->rq.policer_ena = enable;
+ aq->rq_mask.policer_ena = ~(aq->rq_mask.policer_ena);
+ if (enable) {
+ aq->rq.band_prof_id = id;
+ aq->rq_mask.band_prof_id = ~(aq->rq_mask.band_prof_id);
+ }
+
+ rc = mbox_process(mbox);
+ if (rc)
+ goto exit;
+
+ rq->bpf_id = id;
+
+exit:
+ return rc;
+}
diff --git a/drivers/common/cnxk/version.map b/drivers/common/cnxk/version.map
index 6a009eaf35..4c5adb8212 100644
--- a/drivers/common/cnxk/version.map
+++ b/drivers/common/cnxk/version.map
@@ -80,6 +80,7 @@ INTERNAL {
roc_nix_bpf_alloc;
roc_nix_bpf_config;
roc_nix_bpf_count_get;
+ roc_nix_bpf_ena_dis;
roc_nix_bpf_free;
roc_nix_bpf_free_all;
roc_nix_bpf_level_to_idx;
--
2.25.1
[dpdk-dev] [PATCH v3 08/27] common/cnxk: support RoC API to dump bandwidth profile
From: Sunil Kumar Kori
Implement RoC API to dump bandwidth profile on CN10K
platform.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/roc_nix.h | 3 ++
drivers/common/cnxk/roc_nix_bpf.c | 86 ++
drivers/common/cnxk/roc_platform.h | 1 +
drivers/common/cnxk/version.map| 1 +
4 files changed, 91 insertions(+)
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index 2576e938be..15df1e593f 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -563,6 +563,9 @@ int __roc_api roc_nix_bpf_config(struct roc_nix *roc_nix,
uint16_t id,
int __roc_api roc_nix_bpf_ena_dis(struct roc_nix *roc_nix, uint16_t id,
struct roc_nix_rq *rq, bool enable);
+int __roc_api roc_nix_bpf_dump(struct roc_nix *roc_nix, uint16_t id,
+ enum roc_nix_bpf_level_flag lvl_flag);
+
uint8_t __roc_api
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag lvl_flag);
diff --git a/drivers/common/cnxk/roc_nix_bpf.c
b/drivers/common/cnxk/roc_nix_bpf.c
index 557b6279e6..970c960186 100644
--- a/drivers/common/cnxk/roc_nix_bpf.c
+++ b/drivers/common/cnxk/roc_nix_bpf.c
@@ -125,6 +125,60 @@ meter_burst_to_nix(uint64_t value, uint64_t *exponent_p,
uint64_t *mantissa_p)
return NIX_BPF_BURST(exponent, mantissa);
}
+static inline void
+nix_lf_bpf_dump(__io struct nix_band_prof_s *bpf)
+{
+ plt_dump("W0: cir_mantissa \t\t\t%d\nW0: pebs_mantissa \t\t\t0x%03x",
+bpf->cir_mantissa, bpf->pebs_mantissa);
+ plt_dump("W0: peir_matissa \t\t\t\t%d\nW0: cbs_exponent \t\t\t%d",
+bpf->peir_mantissa, bpf->cbs_exponent);
+ plt_dump("W0: cir_exponent \t\t\t%d\nW0: pebs_exponent \t\t\t%d",
+bpf->cir_exponent, bpf->pebs_exponent);
+ plt_dump("W0: peir_exponent \t\t\t%d\n", bpf->peir_exponent);
+ plt_dump("W0: tnl_ena \t\t\t%d\n", bpf->tnl_ena);
+ plt_dump("W0: icolor \t\t\t%d\n", bpf->icolor);
+ plt_dump("W0: pc_mode \t\t\t%d\n", bpf->pc_mode);
+ plt_dump("W1: hl_en \t\t%d\nW1: band_prof_id \t\t%d", bpf->hl_en,
+bpf->band_prof_id);
+ plt_dump("W1: meter_algo \t\t%d\nW1: rc_action \t\t%d", bpf->meter_algo,
+bpf->rc_action);
+ plt_dump("W1: yc_action \t\t\t%d\nW1: gc_action \t\t\t%d",
+bpf->yc_action, bpf->gc_action);
+ plt_dump("W1: adjust_mantissa\t\t\t%d\nW1: adjust_exponent \t\t\t%d",
+bpf->adjust_mantissa, bpf->adjust_exponent);
+ plt_dump("W1: rdiv \t\t\t%d\n", bpf->rdiv);
+ plt_dump("W1: l_select \t\t%d\nW2: lmode \t\t%d", bpf->l_sellect,
+bpf->lmode);
+ plt_dump("W1: cbs_mantissa \t\t\t%d\n", bpf->cbs_mantissa);
+ plt_dump("W2: tsa \t\t\t0x%" PRIx64 "\n", (uint64_t)bpf->ts);
+ plt_dump("W3: c_accum \t\t%d\nW3: pe_accum \t\t%d", bpf->c_accum,
+bpf->pe_accum);
+ plt_dump("W4: green_pkt_pass \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->green_pkt_pass);
+ plt_dump("W5: yellow_pkt_pass \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->yellow_pkt_pass);
+ plt_dump("W6: red_pkt_pass \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->red_pkt_pass);
+ plt_dump("W7: green_octs_pass \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->green_octs_pass);
+ plt_dump("W8: yellow_octs_pass \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->yellow_octs_pass);
+ plt_dump("W9: red_octs_pass \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->red_octs_pass);
+ plt_dump("W10: green_pkt_drop \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->green_pkt_drop);
+ plt_dump("W11: yellow_pkt_drop \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->yellow_pkt_drop);
+ plt_dump("W12: red_pkt_drop \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->red_pkt_drop);
+ plt_dump("W13: green_octs_drop \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->green_octs_drop);
+ plt_dump("W14: yellow_octs_drop \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->yellow_octs_drop);
+ plt_dump("W15: red_octs_drop \t\t\t0x%" PRIx64 "",
+(uint64_t)bpf->red_octs_drop);
+}
+
uint8_t
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag level_f)
{
@@ -482,3 +536,35 @@ roc_nix_bpf_ena_dis(struct roc_nix *roc_nix, uint16_t id,
struct roc_nix_rq *rq,
exit:
return rc;
}
+
+int
+roc_nix_bpf_dump(struct roc_nix *roc_nix, uint16_t id,
+enum roc_nix_bpf_level_flag lvl_flag)
+{
+ struct mbox *mbox = get_mbox(roc_nix);
+ struct nix_cn10k_aq_enq_rsp *rsp;
+ struct nix_cn10k_aq_enq_req *aq;
+ uint8_t level_idx;
+
[dpdk-dev] [PATCH v3 09/27] common/cnxk: support RoC API to setup precolor table
From: Sunil Kumar Kori
For initial coloring of input packet, CN10K platform maintains
precolor table for VLAN, DSCP and Generic. Implement RoC
interface to setup pre color table.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/roc_nix.h | 20
drivers/common/cnxk/roc_nix_bpf.c | 193 ++
drivers/common/cnxk/version.map | 1 +
3 files changed, 214 insertions(+)
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index 15df1e593f..6d17d46388 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -39,6 +39,15 @@ enum roc_nix_bpf_level_flag {
ROC_NIX_BPF_LEVEL_F_TOP = BIT(2),
};
+enum roc_nix_bpf_pc_mode {
+ ROC_NIX_BPF_PC_MODE_VLAN_INNER,
+ ROC_NIX_BPF_PC_MODE_VLAN_OUTER,
+ ROC_NIX_BPF_PC_MODE_DSCP_INNER,
+ ROC_NIX_BPF_PC_MODE_DSCP_OUTER,
+ ROC_NIX_BPF_PC_MODE_GEN_INNER,
+ ROC_NIX_BPF_PC_MODE_GEN_OUTER
+};
+
enum roc_nix_bpf_color {
ROC_NIX_BPF_COLOR_GREEN,
ROC_NIX_BPF_COLOR_YELLOW,
@@ -101,6 +110,13 @@ struct roc_nix_bpf_objs {
uint16_t ids[ROC_NIX_BPF_PER_PFFUNC];
};
+struct roc_nix_bpf_precolor {
+#define ROC_NIX_BPF_PRE_COLOR_MAX 64
+ uint8_t count;
+ enum roc_nix_bpf_pc_mode mode;
+ enum roc_nix_bpf_color color[ROC_NIX_BPF_PRE_COLOR_MAX];
+};
+
struct roc_nix_vlan_config {
uint32_t type;
union {
@@ -566,6 +582,10 @@ int __roc_api roc_nix_bpf_ena_dis(struct roc_nix *roc_nix,
uint16_t id,
int __roc_api roc_nix_bpf_dump(struct roc_nix *roc_nix, uint16_t id,
enum roc_nix_bpf_level_flag lvl_flag);
+int __roc_api roc_nix_bpf_pre_color_tbl_setup(
+ struct roc_nix *roc_nix, uint16_t id,
+ enum roc_nix_bpf_level_flag lvl_flag, struct roc_nix_bpf_precolor *tbl);
+
uint8_t __roc_api
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag lvl_flag);
diff --git a/drivers/common/cnxk/roc_nix_bpf.c
b/drivers/common/cnxk/roc_nix_bpf.c
index 970c960186..b5a1c484a9 100644
--- a/drivers/common/cnxk/roc_nix_bpf.c
+++ b/drivers/common/cnxk/roc_nix_bpf.c
@@ -9,6 +9,10 @@
#define NIX_MAX_BPF_COUNT_MID_LAYER 8
#define NIX_MAX_BPF_COUNT_TOP_LAYER 1
+#define NIX_BPF_PRECOLOR_GEN_TABLE_SIZE 16
+#define NIX_BPF_PRECOLOR_VLAN_TABLE_SIZE 16
+#define NIX_BPF_PRECOLOR_DSCP_TABLE_SIZE 64
+
#define NIX_BPF_LEVEL_F_MASK
\
(ROC_NIX_BPF_LEVEL_F_LEAF | ROC_NIX_BPF_LEVEL_F_MID | \
ROC_NIX_BPF_LEVEL_F_TOP)
@@ -179,6 +183,107 @@ nix_lf_bpf_dump(__io struct nix_band_prof_s *bpf)
(uint64_t)bpf->red_octs_drop);
}
+static inline void
+nix_precolor_conv_table_write(struct roc_nix *roc_nix, uint64_t val,
+ uint32_t off)
+{
+ struct nix *nix = roc_nix_to_nix_priv(roc_nix);
+ int64_t *addr;
+
+ addr = PLT_PTR_ADD(nix->base, off);
+ /* FIXME: Currently writing to this register throwing kernel dump.
+* plt_write64(val, addr);
+*/
+ PLT_SET_USED(val);
+ PLT_SET_USED(addr);
+}
+
+static uint8_t
+nix_precolor_vlan_table_update(struct roc_nix *roc_nix,
+ struct roc_nix_bpf_precolor *tbl)
+{
+ uint64_t val = 0, i;
+ uint8_t tn_ena;
+ uint32_t off;
+
+ for (i = 0; i < tbl->count; i++)
+ val |= (((uint64_t)tbl->color[i]) << (2 * i));
+
+ if (tbl->mode == ROC_NIX_BPF_PC_MODE_VLAN_INNER) {
+ off = NIX_LF_RX_VLAN1_COLOR_CONV;
+ tn_ena = true;
+ } else {
+ off = NIX_LF_RX_VLAN0_COLOR_CONV;
+ tn_ena = false;
+ }
+
+ nix_precolor_conv_table_write(roc_nix, val, off);
+ return tn_ena;
+}
+
+static uint8_t
+nix_precolor_inner_dscp_table_update(struct roc_nix *roc_nix,
+struct roc_nix_bpf_precolor *tbl)
+{
+ uint64_t val_lo = 0, val_hi = 0, i, j;
+
+ for (i = 0, j = 0; i < (tbl->count / 2); i++, j++)
+ val_lo |= (((uint64_t)tbl->color[i]) << (2 * j));
+
+ for (j = 0; i < tbl->count; i++, j++)
+ val_hi |= (((uint64_t)tbl->color[i]) << (2 * j));
+
+ nix_precolor_conv_table_write(roc_nix, val_lo,
+ NIX_LF_RX_IIP_COLOR_CONV_LO);
+ nix_precolor_conv_table_write(roc_nix, val_hi,
+ NIX_LF_RX_IIP_COLOR_CONV_HI);
+
+ return true;
+}
+
+static uint8_t
+nix_precolor_outer_dscp_table_update(struct roc_nix *roc_nix,
+struct roc_nix_bpf_precolor *tbl)
+{
+ uint64_t val_lo = 0, val_hi = 0, i, j;
+
+ for (i = 0, j = 0; i < (tbl->count /
[dpdk-dev] [PATCH v3 10/27] common/cnxk: support RoC API to connect bandwidth profiles
From: Sunil Kumar Kori
To maintain chain of bandwidth profiles, they needs to be
connected. Implement RoC API to connect two bandwidth profiles
at different levels.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/roc_nix.h | 6 ++
drivers/common/cnxk/roc_nix_bpf.c | 36 +++
drivers/common/cnxk/version.map | 1 +
3 files changed, 43 insertions(+)
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index 6d17d46388..af76068030 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -7,6 +7,7 @@
/* Constants */
#define ROC_NIX_BPF_PER_PFFUNC 64
+#define ROC_NIX_BPF_ID_INVALID 0x
#define ROC_NIX_BPF_LEVEL_IDX_INVALID 0xFF
#define ROC_NIX_BPF_LEVEL_MAX3
@@ -586,6 +587,11 @@ int __roc_api roc_nix_bpf_pre_color_tbl_setup(
struct roc_nix *roc_nix, uint16_t id,
enum roc_nix_bpf_level_flag lvl_flag, struct roc_nix_bpf_precolor *tbl);
+/* Use ROC_NIX_BPF_ID_INVALID as dst_id to disconnect */
+int __roc_api roc_nix_bpf_connect(struct roc_nix *roc_nix,
+ enum roc_nix_bpf_level_flag lvl_flag,
+ uint16_t src_id, uint16_t dst_id);
+
uint8_t __roc_api
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag lvl_flag);
diff --git a/drivers/common/cnxk/roc_nix_bpf.c
b/drivers/common/cnxk/roc_nix_bpf.c
index b5a1c484a9..f5822136d1 100644
--- a/drivers/common/cnxk/roc_nix_bpf.c
+++ b/drivers/common/cnxk/roc_nix_bpf.c
@@ -761,3 +761,39 @@ roc_nix_bpf_pre_color_tbl_setup(struct roc_nix *roc_nix,
uint16_t id,
exit:
return rc;
}
+
+int
+roc_nix_bpf_connect(struct roc_nix *roc_nix,
+ enum roc_nix_bpf_level_flag lvl_flag, uint16_t src_id,
+ uint16_t dst_id)
+{
+ struct mbox *mbox = get_mbox(roc_nix);
+ struct nix_cn10k_aq_enq_req *aq;
+ uint8_t level_idx;
+
+ if (roc_model_is_cn9k())
+ return NIX_ERR_HW_NOTSUP;
+
+ level_idx = roc_nix_bpf_level_to_idx(lvl_flag);
+ if (level_idx == ROC_NIX_BPF_LEVEL_IDX_INVALID)
+ return NIX_ERR_PARAM;
+
+ aq = mbox_alloc_msg_nix_cn10k_aq_enq(mbox);
+ if (aq == NULL)
+ return -ENOSPC;
+ aq->qidx = (sw_to_hw_lvl_map[level_idx] << 14) | src_id;
+ aq->ctype = NIX_AQ_CTYPE_BAND_PROF;
+ aq->op = NIX_AQ_INSTOP_WRITE;
+
+ if (dst_id == ROC_NIX_BPF_ID_INVALID) {
+ aq->prof.hl_en = false;
+ aq->prof_mask.hl_en = ~(aq->prof_mask.hl_en);
+ } else {
+ aq->prof.hl_en = true;
+ aq->prof.band_prof_id = dst_id;
+ aq->prof_mask.hl_en = ~(aq->prof_mask.hl_en);
+ aq->prof_mask.band_prof_id = ~(aq->prof_mask.band_prof_id);
+ }
+
+ return mbox_process(mbox);
+}
diff --git a/drivers/common/cnxk/version.map b/drivers/common/cnxk/version.map
index a08fbe6013..c04a8ca9da 100644
--- a/drivers/common/cnxk/version.map
+++ b/drivers/common/cnxk/version.map
@@ -79,6 +79,7 @@ INTERNAL {
roc_se_ciph_key_set;
roc_nix_bpf_alloc;
roc_nix_bpf_config;
+ roc_nix_bpf_connect;
roc_nix_bpf_count_get;
roc_nix_bpf_dump;
roc_nix_bpf_ena_dis;
--
2.25.1
[dpdk-dev] [PATCH v3 11/27] common/cnxk: support RoC API to get stats to index
From: Sunil Kumar Kori
CN10K platform supports different stats for HW bandwidth profiles.
Implement RoC API to get index for given stats type.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/roc_nix.h | 18
drivers/common/cnxk/roc_nix_bpf.c | 34 +++
drivers/common/cnxk/version.map | 1 +
3 files changed, 53 insertions(+)
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index af76068030..7bff69d39f 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -10,6 +10,7 @@
#define ROC_NIX_BPF_ID_INVALID 0x
#define ROC_NIX_BPF_LEVEL_IDX_INVALID 0xFF
#define ROC_NIX_BPF_LEVEL_MAX3
+#define ROC_NIX_BPF_STATS_MAX12
enum roc_nix_rss_reta_sz {
ROC_NIX_RSS_RETA_SZ_64 = 64,
@@ -71,6 +72,21 @@ enum roc_nix_bpf_action {
ROC_NIX_BPF_ACTION_RED
};
+enum roc_nix_bpf_stats {
+ ROC_NIX_BPF_GREEN_PKT_F_PASS = BIT_ULL(0),
+ ROC_NIX_BPF_GREEN_OCTS_F_PASS = BIT_ULL(1),
+ ROC_NIX_BPF_GREEN_PKT_F_DROP = BIT_ULL(2),
+ ROC_NIX_BPF_GREEN_OCTS_F_DROP = BIT_ULL(3),
+ ROC_NIX_BPF_YELLOW_PKT_F_PASS = BIT_ULL(4),
+ ROC_NIX_BPF_YELLOW_OCTS_F_PASS = BIT_ULL(5),
+ ROC_NIX_BPF_YELLOW_PKT_F_DROP = BIT_ULL(6),
+ ROC_NIX_BPF_YELLOW_OCTS_F_DROP = BIT_ULL(7),
+ ROC_NIX_BPF_RED_PKT_F_PASS = BIT_ULL(8),
+ ROC_NIX_BPF_RED_OCTS_F_PASS = BIT_ULL(9),
+ ROC_NIX_BPF_RED_PKT_F_DROP = BIT_ULL(10),
+ ROC_NIX_BPF_RED_OCTS_F_DROP = BIT_ULL(11),
+};
+
struct roc_nix_bpf_cfg {
enum roc_nix_bpf_algo alg;
enum roc_nix_bpf_lmode lmode;
@@ -595,6 +611,8 @@ int __roc_api roc_nix_bpf_connect(struct roc_nix *roc_nix,
uint8_t __roc_api
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag lvl_flag);
+uint8_t __roc_api roc_nix_bpf_stats_to_idx(enum roc_nix_bpf_stats lvl_flag);
+
/* MAC */
int __roc_api roc_nix_mac_rxtx_start_stop(struct roc_nix *roc_nix, bool start);
int __roc_api roc_nix_mac_link_event_start_stop(struct roc_nix *roc_nix,
diff --git a/drivers/common/cnxk/roc_nix_bpf.c
b/drivers/common/cnxk/roc_nix_bpf.c
index f5822136d1..8f45650d18 100644
--- a/drivers/common/cnxk/roc_nix_bpf.c
+++ b/drivers/common/cnxk/roc_nix_bpf.c
@@ -300,6 +300,40 @@ roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag
level_f)
return idx;
}
+uint8_t
+roc_nix_bpf_stats_to_idx(enum roc_nix_bpf_stats level_f)
+{
+ uint8_t idx;
+
+ if (level_f & ROC_NIX_BPF_GREEN_PKT_F_PASS)
+ idx = 0;
+ else if (level_f & ROC_NIX_BPF_GREEN_OCTS_F_PASS)
+ idx = 1;
+ else if (level_f & ROC_NIX_BPF_GREEN_PKT_F_DROP)
+ idx = 2;
+ else if (level_f & ROC_NIX_BPF_GREEN_OCTS_F_DROP)
+ idx = 3;
+ else if (level_f & ROC_NIX_BPF_YELLOW_PKT_F_PASS)
+ idx = 4;
+ else if (level_f & ROC_NIX_BPF_YELLOW_OCTS_F_PASS)
+ idx = 5;
+ else if (level_f & ROC_NIX_BPF_YELLOW_PKT_F_DROP)
+ idx = 6;
+ else if (level_f & ROC_NIX_BPF_YELLOW_OCTS_F_DROP)
+ idx = 7;
+ else if (level_f & ROC_NIX_BPF_RED_PKT_F_PASS)
+ idx = 8;
+ else if (level_f & ROC_NIX_BPF_RED_OCTS_F_PASS)
+ idx = 9;
+ else if (level_f & ROC_NIX_BPF_RED_PKT_F_DROP)
+ idx = 10;
+ else if (level_f & ROC_NIX_BPF_RED_OCTS_F_DROP)
+ idx = 11;
+ else
+ idx = ROC_NIX_BPF_STATS_MAX;
+ return idx;
+}
+
int
roc_nix_bpf_count_get(struct roc_nix *roc_nix, uint8_t lvl_mask,
uint16_t count[ROC_NIX_BPF_LEVEL_MAX])
diff --git a/drivers/common/cnxk/version.map b/drivers/common/cnxk/version.map
index c04a8ca9da..5d4bdd57a9 100644
--- a/drivers/common/cnxk/version.map
+++ b/drivers/common/cnxk/version.map
@@ -87,6 +87,7 @@ INTERNAL {
roc_nix_bpf_free_all;
roc_nix_bpf_level_to_idx;
roc_nix_bpf_pre_color_tbl_setup;
+ roc_nix_bpf_stats_to_idx;
roc_nix_cq_dump;
roc_nix_cq_fini;
roc_nix_cq_init;
--
2.25.1
[dpdk-dev] [PATCH v3 12/27] common/cnxk: support RoC API to read profile statistics
From: Sunil Kumar Kori
CN10K platform provides statistics per bandwidth profile and
per nixlf. Implement RoC API to read stats for given bandwidth
profile.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/roc_nix.h | 9 ++
drivers/common/cnxk/roc_nix_bpf.c | 197 ++
drivers/common/cnxk/version.map | 2 +
3 files changed, 208 insertions(+)
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index 7bff69d39f..0081ccc9ee 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -608,6 +608,15 @@ int __roc_api roc_nix_bpf_connect(struct roc_nix *roc_nix,
enum roc_nix_bpf_level_flag lvl_flag,
uint16_t src_id, uint16_t dst_id);
+int __roc_api
+roc_nix_bpf_stats_read(struct roc_nix *roc_nix, uint16_t id, uint64_t mask,
+ enum roc_nix_bpf_level_flag lvl_flag,
+ uint64_t stats[ROC_NIX_BPF_STATS_MAX] /* Out */);
+
+int __roc_api
+roc_nix_bpf_lf_stats_read(struct roc_nix *roc_nix, uint64_t mask,
+ uint64_t stats[ROC_NIX_BPF_STATS_MAX] /* Out */);
+
uint8_t __roc_api
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag lvl_flag);
diff --git a/drivers/common/cnxk/roc_nix_bpf.c
b/drivers/common/cnxk/roc_nix_bpf.c
index 8f45650d18..2169164d9c 100644
--- a/drivers/common/cnxk/roc_nix_bpf.c
+++ b/drivers/common/cnxk/roc_nix_bpf.c
@@ -17,6 +17,9 @@
(ROC_NIX_BPF_LEVEL_F_LEAF | ROC_NIX_BPF_LEVEL_F_MID | \
ROC_NIX_BPF_LEVEL_F_TOP)
+#define NIX_RD_STATS(val) plt_read64(nix->base + NIX_LF_RX_STATX(val))
+#define NIX_RST_STATS(val) plt_write64(0, nix->base + NIX_LF_RX_STATX(val))
+
static uint8_t sw_to_hw_lvl_map[] = {NIX_RX_BAND_PROF_LAYER_LEAF,
NIX_RX_BAND_PROF_LAYER_MIDDLE,
NIX_RX_BAND_PROF_LAYER_TOP};
@@ -831,3 +834,197 @@ roc_nix_bpf_connect(struct roc_nix *roc_nix,
return mbox_process(mbox);
}
+
+int
+roc_nix_bpf_stats_read(struct roc_nix *roc_nix, uint16_t id, uint64_t mask,
+ enum roc_nix_bpf_level_flag lvl_flag,
+ uint64_t stats[ROC_NIX_BPF_STATS_MAX])
+{
+ uint8_t yellow_pkt_pass, yellow_octs_pass, yellow_pkt_drop;
+ uint8_t green_octs_drop, yellow_octs_drop, red_octs_drop;
+ uint8_t green_pkt_pass, green_octs_pass, green_pkt_drop;
+ uint8_t red_pkt_pass, red_octs_pass, red_pkt_drop;
+ struct mbox *mbox = get_mbox(roc_nix);
+ struct nix_cn10k_aq_enq_rsp *rsp;
+ struct nix_cn10k_aq_enq_req *aq;
+ uint8_t level_idx;
+ int rc;
+
+ if (roc_model_is_cn9k())
+ return NIX_ERR_HW_NOTSUP;
+
+ level_idx = roc_nix_bpf_level_to_idx(lvl_flag);
+ if (level_idx == ROC_NIX_BPF_LEVEL_IDX_INVALID)
+ return NIX_ERR_PARAM;
+
+ aq = mbox_alloc_msg_nix_cn10k_aq_enq(mbox);
+ if (aq == NULL)
+ return -ENOSPC;
+ aq->qidx = (sw_to_hw_lvl_map[level_idx] << 14 | id);
+ aq->ctype = NIX_AQ_CTYPE_BAND_PROF;
+ aq->op = NIX_AQ_INSTOP_READ;
+ rc = mbox_process_msg(mbox, (void *)&rsp);
+ if (rc)
+ return rc;
+
+ green_pkt_pass =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_GREEN_PKT_F_PASS);
+ green_octs_pass =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_GREEN_OCTS_F_PASS);
+ green_pkt_drop =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_GREEN_PKT_F_DROP);
+ green_octs_drop =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_GREEN_OCTS_F_DROP);
+ yellow_pkt_pass =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_YELLOW_PKT_F_PASS);
+ yellow_octs_pass =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_YELLOW_OCTS_F_PASS);
+ yellow_pkt_drop =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_YELLOW_PKT_F_DROP);
+ yellow_octs_drop =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_YELLOW_OCTS_F_DROP);
+ red_pkt_pass =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_RED_PKT_F_PASS);
+ red_octs_pass =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_RED_OCTS_F_PASS);
+ red_pkt_drop =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_RED_PKT_F_DROP);
+ red_octs_drop =
+ roc_nix_bpf_stats_to_idx(mask & ROC_NIX_BPF_RED_OCTS_F_DROP);
+
+ if (green_pkt_pass != ROC_NIX_BPF_STATS_MAX)
+ stats[green_pkt_pass] = rsp->prof.green_pkt_pass;
+
+ if (green_octs_pass != ROC_NIX_BPF_STATS_MAX)
+ stats[green_octs_pass] = rsp->prof.
[dpdk-dev] [PATCH v3 13/27] common/cnxk: support RoC API to reset profile stats
From: Sunil Kumar Kori
Implement RoC API to reset stats per bandwidth profile
or per nixlf.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/roc_nix.h | 7 ++
drivers/common/cnxk/roc_nix_bpf.c | 113 ++
drivers/common/cnxk/version.map | 2 +
3 files changed, 122 insertions(+)
diff --git a/drivers/common/cnxk/roc_nix.h b/drivers/common/cnxk/roc_nix.h
index 0081ccc9ee..b6715ef2ed 100644
--- a/drivers/common/cnxk/roc_nix.h
+++ b/drivers/common/cnxk/roc_nix.h
@@ -613,10 +613,17 @@ roc_nix_bpf_stats_read(struct roc_nix *roc_nix, uint16_t
id, uint64_t mask,
enum roc_nix_bpf_level_flag lvl_flag,
uint64_t stats[ROC_NIX_BPF_STATS_MAX] /* Out */);
+int __roc_api roc_nix_bpf_stats_reset(struct roc_nix *roc_nix, uint16_t id,
+ uint64_t mask,
+ enum roc_nix_bpf_level_flag lvl_flag);
+
int __roc_api
roc_nix_bpf_lf_stats_read(struct roc_nix *roc_nix, uint64_t mask,
uint64_t stats[ROC_NIX_BPF_STATS_MAX] /* Out */);
+int __roc_api roc_nix_bpf_lf_stats_reset(struct roc_nix *roc_nix,
+uint64_t mask);
+
uint8_t __roc_api
roc_nix_bpf_level_to_idx(enum roc_nix_bpf_level_flag lvl_flag);
diff --git a/drivers/common/cnxk/roc_nix_bpf.c
b/drivers/common/cnxk/roc_nix_bpf.c
index 2169164d9c..bba906a59b 100644
--- a/drivers/common/cnxk/roc_nix_bpf.c
+++ b/drivers/common/cnxk/roc_nix_bpf.c
@@ -931,6 +931,86 @@ roc_nix_bpf_stats_read(struct roc_nix *roc_nix, uint16_t
id, uint64_t mask,
return 0;
}
+int
+roc_nix_bpf_stats_reset(struct roc_nix *roc_nix, uint16_t id, uint64_t mask,
+ enum roc_nix_bpf_level_flag lvl_flag)
+{
+ struct mbox *mbox = get_mbox(roc_nix);
+ struct nix_cn10k_aq_enq_req *aq;
+ uint8_t level_idx;
+
+ if (roc_model_is_cn9k())
+ return NIX_ERR_HW_NOTSUP;
+
+ level_idx = roc_nix_bpf_level_to_idx(lvl_flag);
+ if (level_idx == ROC_NIX_BPF_LEVEL_IDX_INVALID)
+ return NIX_ERR_PARAM;
+
+ aq = mbox_alloc_msg_nix_cn10k_aq_enq(mbox);
+ if (aq == NULL)
+ return -ENOSPC;
+ aq->qidx = (sw_to_hw_lvl_map[level_idx] << 14 | id);
+ aq->ctype = NIX_AQ_CTYPE_BAND_PROF;
+ aq->op = NIX_AQ_INSTOP_WRITE;
+
+ if (mask & ROC_NIX_BPF_GREEN_PKT_F_PASS) {
+ aq->prof.green_pkt_pass = 0;
+ aq->prof_mask.green_pkt_pass = ~(aq->prof_mask.green_pkt_pass);
+ }
+ if (mask & ROC_NIX_BPF_GREEN_OCTS_F_PASS) {
+ aq->prof.green_octs_pass = 0;
+ aq->prof_mask.green_octs_pass =
+ ~(aq->prof_mask.green_octs_pass);
+ }
+ if (mask & ROC_NIX_BPF_GREEN_PKT_F_DROP) {
+ aq->prof.green_pkt_drop = 0;
+ aq->prof_mask.green_pkt_drop = ~(aq->prof_mask.green_pkt_drop);
+ }
+ if (mask & ROC_NIX_BPF_GREEN_OCTS_F_DROP) {
+ aq->prof.green_octs_drop = 0;
+ aq->prof_mask.green_octs_drop =
+ ~(aq->prof_mask.green_octs_drop);
+ }
+ if (mask & ROC_NIX_BPF_YELLOW_PKT_F_PASS) {
+ aq->prof.yellow_pkt_pass = 0;
+ aq->prof_mask.yellow_pkt_pass =
+ ~(aq->prof_mask.yellow_pkt_pass);
+ }
+ if (mask & ROC_NIX_BPF_YELLOW_OCTS_F_PASS) {
+ aq->prof.yellow_octs_pass = 0;
+ aq->prof_mask.yellow_octs_pass =
+ ~(aq->prof_mask.yellow_octs_pass);
+ }
+ if (mask & ROC_NIX_BPF_YELLOW_PKT_F_DROP) {
+ aq->prof.yellow_pkt_drop = 0;
+ aq->prof_mask.yellow_pkt_drop =
+ ~(aq->prof_mask.yellow_pkt_drop);
+ }
+ if (mask & ROC_NIX_BPF_YELLOW_OCTS_F_DROP) {
+ aq->prof.yellow_octs_drop = 0;
+ aq->prof_mask.yellow_octs_drop =
+ ~(aq->prof_mask.yellow_octs_drop);
+ }
+ if (mask & ROC_NIX_BPF_RED_PKT_F_PASS) {
+ aq->prof.red_pkt_pass = 0;
+ aq->prof_mask.red_pkt_pass = ~(aq->prof_mask.red_pkt_pass);
+ }
+ if (mask & ROC_NIX_BPF_RED_OCTS_F_PASS) {
+ aq->prof.red_octs_pass = 0;
+ aq->prof_mask.red_octs_pass = ~(aq->prof_mask.red_octs_pass);
+ }
+ if (mask & ROC_NIX_BPF_RED_PKT_F_DROP) {
+ aq->prof.red_pkt_drop = 0;
+ aq->prof_mask.red_pkt_drop = ~(aq->prof_mask.red_pkt_drop);
+ }
+ if (mask & ROC_NIX_BPF_RED_OCTS_F_DROP) {
+ aq->prof.red_octs_drop = 0;
+ aq->prof_mask.red_octs_drop = ~(aq->prof_mask.red_octs_dro
[dpdk-dev] [PATCH v3 14/27] common/cnxk: support meter in action list
From: Sunil Kumar Kori
Meter action is added in supported action list.
Signed-off-by: Sunil Kumar Kori
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/common/cnxk/roc_npc.c | 3 +++
drivers/common/cnxk/roc_npc.h | 7 ++-
2 files changed, 9 insertions(+), 1 deletion(-)
diff --git a/drivers/common/cnxk/roc_npc.c b/drivers/common/cnxk/roc_npc.c
index 52a54b3990..8f88e254b3 100644
--- a/drivers/common/cnxk/roc_npc.c
+++ b/drivers/common/cnxk/roc_npc.c
@@ -447,6 +447,9 @@ npc_parse_actions(struct npc *npc, const struct
roc_npc_attr *attr,
case ROC_NPC_ACTION_TYPE_VLAN_PCP_INSERT:
req_act |= ROC_NPC_ACTION_TYPE_VLAN_PCP_INSERT;
break;
+ case ROC_NPC_ACTION_TYPE_METER:
+ req_act |= ROC_NPC_ACTION_TYPE_METER;
+ break;
default:
errcode = NPC_ERR_ACTION_NOTSUP;
goto err_exit;
diff --git a/drivers/common/cnxk/roc_npc.h b/drivers/common/cnxk/roc_npc.h
index 4d6f8f8cd9..86365df754 100644
--- a/drivers/common/cnxk/roc_npc.h
+++ b/drivers/common/cnxk/roc_npc.h
@@ -58,7 +58,7 @@ struct roc_npc_flow_item_raw {
const uint8_t *pattern; /**< Byte string to look for. */
};
-#define ROC_NPC_MAX_ACTION_COUNT 12
+#define ROC_NPC_MAX_ACTION_COUNT 17
enum roc_npc_action_type {
ROC_NPC_ACTION_TYPE_END = (1 << 0),
@@ -77,6 +77,7 @@ enum roc_npc_action_type {
ROC_NPC_ACTION_TYPE_VLAN_INSERT = (1 << 13),
ROC_NPC_ACTION_TYPE_VLAN_ETHTYPE_INSERT = (1 << 14),
ROC_NPC_ACTION_TYPE_VLAN_PCP_INSERT = (1 << 15),
+ ROC_NPC_ACTION_TYPE_METER = (1 << 16),
};
struct roc_npc_action {
@@ -110,6 +111,10 @@ struct roc_npc_action_of_set_vlan_pcp {
uint8_t vlan_pcp; /**< VLAN priority. */
};
+struct roc_npc_action_meter {
+ uint32_t mtr_id; /**< Meter id to be applied. > */
+};
+
struct roc_npc_attr {
uint32_t priority; /**< Rule priority level within group. */
uint32_t ingress : 1; /**< Rule applies to ingress traffic. */
--
2.25.1
[dpdk-dev] [PATCH v3 15/27] net/cnxk: support meter ops get API
From: Sunil Kumar Kori
To enable support for ingress meter, supported operations
are exposed for CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev.c | 1 +
drivers/net/cnxk/cn10k_ethdev.h | 2 ++
drivers/net/cnxk/cn10k_ethdev_mtr.c | 18 ++
drivers/net/cnxk/meson.build| 1 +
4 files changed, 22 insertions(+)
create mode 100644 drivers/net/cnxk/cn10k_ethdev_mtr.c
diff --git a/drivers/net/cnxk/cn10k_ethdev.c b/drivers/net/cnxk/cn10k_ethdev.c
index 7caec6cf14..8c1f6a4408 100644
--- a/drivers/net/cnxk/cn10k_ethdev.c
+++ b/drivers/net/cnxk/cn10k_ethdev.c
@@ -457,6 +457,7 @@ nix_eth_dev_ops_override(void)
cnxk_eth_dev_ops.dev_ptypes_set = cn10k_nix_ptypes_set;
cnxk_eth_dev_ops.timesync_enable = cn10k_nix_timesync_enable;
cnxk_eth_dev_ops.timesync_disable = cn10k_nix_timesync_disable;
+ cnxk_eth_dev_ops.mtr_ops_get = cn10k_nix_mtr_ops_get;
}
static void
diff --git a/drivers/net/cnxk/cn10k_ethdev.h b/drivers/net/cnxk/cn10k_ethdev.h
index 8b6e0f2b3f..117aa2a62d 100644
--- a/drivers/net/cnxk/cn10k_ethdev.h
+++ b/drivers/net/cnxk/cn10k_ethdev.h
@@ -38,4 +38,6 @@ struct cn10k_eth_rxq {
void cn10k_eth_set_rx_function(struct rte_eth_dev *eth_dev);
void cn10k_eth_set_tx_function(struct rte_eth_dev *eth_dev);
+/* MTR */
+int cn10k_nix_mtr_ops_get(struct rte_eth_dev *dev, void *ops);
#endif /* __CN10K_ETHDEV_H__ */
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
new file mode 100644
index 00..9b46032858
--- /dev/null
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -0,0 +1,18 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2021 Marvell.
+ */
+
+#include "cn10k_ethdev.h"
+#include
+
+const struct rte_mtr_ops nix_mtr_ops = {
+};
+
+int
+cn10k_nix_mtr_ops_get(struct rte_eth_dev *dev, void *ops)
+{
+ RTE_SET_USED(dev);
+
+ *(const void **)ops = &nix_mtr_ops;
+ return 0;
+}
diff --git a/drivers/net/cnxk/meson.build b/drivers/net/cnxk/meson.build
index d4cdd1744a..91afc1de4c 100644
--- a/drivers/net/cnxk/meson.build
+++ b/drivers/net/cnxk/meson.build
@@ -35,6 +35,7 @@ sources += files(
# CN10K
sources += files(
'cn10k_ethdev.c',
+'cn10k_ethdev_mtr.c',
'cn10k_rte_flow.c',
'cn10k_rx.c',
'cn10k_rx_mseg.c',
--
2.25.1
[dpdk-dev] [PATCH v3 16/27] net/cnxk: support ops to get meter capabilities
From: Sunil Kumar Kori
Implement ethdev operation to get meter capabilities for
CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev_mtr.c | 48 +
1 file changed, 48 insertions(+)
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index 9b46032858..a1b42831a5 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -5,7 +5,55 @@
#include "cn10k_ethdev.h"
#include
+#define NIX_MTR_COUNT_MAX 73 /* 64(leaf) + 8(mid) + 1(top) */
+#define NIX_MTR_COUNT_PER_FLOW 3 /* 1(leaf) + 1(mid) + 1(top) */
+
+static struct rte_mtr_capabilities mtr_capa = {
+ .n_max = NIX_MTR_COUNT_MAX,
+ .n_shared_max = NIX_MTR_COUNT_PER_FLOW,
+ /* .identical = , */
+ .shared_identical = true,
+ /* .shared_n_flows_per_mtr_max = ,*/
+ .chaining_n_mtrs_per_flow_max = NIX_MTR_COUNT_PER_FLOW,
+ .chaining_use_prev_mtr_color_supported = true,
+ .chaining_use_prev_mtr_color_enforced = true,
+ .meter_srtcm_rfc2697_n_max = NIX_MTR_COUNT_MAX,
+ .meter_trtcm_rfc2698_n_max = NIX_MTR_COUNT_MAX,
+ .meter_trtcm_rfc4115_n_max = NIX_MTR_COUNT_MAX,
+ .meter_rate_max = NIX_BPF_RATE_MAX / 8, /* Bytes per second */
+ .meter_policy_n_max = NIX_MTR_COUNT_MAX,
+ .color_aware_srtcm_rfc2697_supported = true,
+ .color_aware_trtcm_rfc2698_supported = true,
+ .color_aware_trtcm_rfc4115_supported = true,
+ .srtcm_rfc2697_byte_mode_supported = true,
+ .srtcm_rfc2697_packet_mode_supported = true,
+ .trtcm_rfc2698_byte_mode_supported = true,
+ .trtcm_rfc2698_packet_mode_supported = true,
+ .trtcm_rfc4115_byte_mode_supported = true,
+ .trtcm_rfc4115_packet_mode_supported = true,
+ .stats_mask = RTE_MTR_STATS_N_PKTS_GREEN | RTE_MTR_STATS_N_PKTS_YELLOW |
+ RTE_MTR_STATS_N_PKTS_RED | RTE_MTR_STATS_N_PKTS_DROPPED |
+ RTE_MTR_STATS_N_BYTES_GREEN |
+ RTE_MTR_STATS_N_BYTES_YELLOW | RTE_MTR_STATS_N_BYTES_RED |
+ RTE_MTR_STATS_N_BYTES_DROPPED};
+
+static int
+cn10k_nix_mtr_capabilities_get(struct rte_eth_dev *dev,
+ struct rte_mtr_capabilities *capa,
+ struct rte_mtr_error *error)
+{
+ RTE_SET_USED(dev);
+
+ if (!capa)
+ return -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
+ "NULL input parameter");
+ *capa = mtr_capa;
+ return 0;
+}
+
const struct rte_mtr_ops nix_mtr_ops = {
+ .capabilities_get = cn10k_nix_mtr_capabilities_get,
};
int
--
2.25.1
[dpdk-dev] [PATCH v3 17/27] net/cnxk: support ops to create meter profile
From: Sunil Kumar Kori
Implement API to add meter profile for CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev_mtr.c | 138
drivers/net/cnxk/cnxk_ethdev.c | 14 +++
drivers/net/cnxk/cnxk_ethdev.h | 13 +++
3 files changed, 165 insertions(+)
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index a1b42831a5..69efe395b5 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -37,6 +37,106 @@ static struct rte_mtr_capabilities mtr_capa = {
RTE_MTR_STATS_N_BYTES_YELLOW | RTE_MTR_STATS_N_BYTES_RED |
RTE_MTR_STATS_N_BYTES_DROPPED};
+static struct cnxk_mtr_profile_node *
+nix_mtr_profile_find(struct cnxk_eth_dev *dev, uint32_t profile_id)
+{
+ struct cnxk_mtr_profiles *fmps = &dev->mtr_profiles;
+ struct cnxk_mtr_profile_node *fmp;
+
+ TAILQ_FOREACH(fmp, fmps, next)
+ if (profile_id == fmp->id)
+ return fmp;
+
+ return NULL;
+}
+
+static int
+nix_mtr_profile_validate(struct cnxk_eth_dev *dev, uint32_t profile_id,
+struct rte_mtr_meter_profile *profile,
+struct rte_mtr_error *error)
+{
+ int rc = 0;
+
+ PLT_SET_USED(dev);
+
+ if (profile == NULL)
+ return -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE,
+ NULL, "Meter profile is null.");
+
+ if (profile_id == UINT32_MAX)
+ return -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+ NULL, "Meter profile id not valid.");
+
+ switch (profile->alg) {
+ case RTE_MTR_SRTCM_RFC2697:
+ if (profile->srtcm_rfc2697.cir > mtr_capa.meter_rate_max)
+ rc = -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE, NULL,
+ "CIR exceeds max meter rate");
+
+ if (profile->srtcm_rfc2697.cbs > NIX_BPF_BURST_MAX)
+ rc = -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE, NULL,
+ "CBS exceeds max meter burst size");
+
+ if (profile->srtcm_rfc2697.ebs > NIX_BPF_BURST_MAX)
+ rc = -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE, NULL,
+ "EBS exceeds max meter burst size");
+ break;
+
+ case RTE_MTR_TRTCM_RFC2698:
+ if (profile->trtcm_rfc2698.cir > mtr_capa.meter_rate_max)
+ rc = -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE, NULL,
+ "CIR exceeds max meter rate");
+
+ if (profile->trtcm_rfc2698.pir > mtr_capa.meter_rate_max)
+ rc = -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE, NULL,
+ "PIR exceeds max meter rate");
+
+ if (profile->trtcm_rfc2698.cbs > NIX_BPF_BURST_MAX)
+ rc = -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE, NULL,
+ "CBS exceeds max meter burst size");
+
+ if (profile->trtcm_rfc2698.pbs > NIX_BPF_BURST_MAX)
+ rc = -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE, NULL,
+ "PBS exceeds max meter burst size");
+ break;
+
+ case RTE_MTR_TRTCM_RFC4115:
+ if ((profile->trtcm_rfc4115.cir + profile->trtcm_rfc4115.eir) >
+ mtr_capa.meter_rate_max)
+ rc = -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE, NULL,
+ "PIR + EIR exceeds max rate");
+
+ if (profile->trtcm_rfc4115.cbs > NIX_BPF_BURST_MAX)
+ rc = -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE, NULL,
+ "CBS exceeds max meter burst size");
+
+ if (profile->trtcm_rfc4115.ebs > NIX_BPF_BURST_MAX)
+ rc = -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_T
[dpdk-dev] [PATCH v3 18/27] net/cnxk: support ops to delete meter profile
From: Sunil Kumar Kori
Implement API to delete meter profile for CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev_mtr.c | 30 +
1 file changed, 30 insertions(+)
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index 69efe395b5..ea4d898cd7 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -189,9 +189,39 @@ cn10k_nix_mtr_profile_add(struct rte_eth_dev *eth_dev,
uint32_t profile_id,
return 0;
}
+static int
+cn10k_nix_mtr_profile_delete(struct rte_eth_dev *eth_dev, uint32_t profile_id,
+struct rte_mtr_error *error)
+{
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ struct cnxk_mtr_profile_node *fmp;
+
+ if (profile_id == UINT32_MAX)
+ return -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+ NULL, "Meter profile id not valid.");
+
+ fmp = nix_mtr_profile_find(dev, profile_id);
+ if (fmp == NULL)
+ return -rte_mtr_error_set(error, ENOENT,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+ &profile_id,
+ "Meter profile is invalid.");
+
+ if (fmp->ref_cnt)
+ return -rte_mtr_error_set(error, EBUSY,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+ NULL, "Meter profile is in use.");
+
+ TAILQ_REMOVE(&dev->mtr_profiles, fmp, next);
+ plt_free(fmp);
+ return 0;
+}
+
const struct rte_mtr_ops nix_mtr_ops = {
.capabilities_get = cn10k_nix_mtr_capabilities_get,
.meter_profile_add = cn10k_nix_mtr_profile_add,
+ .meter_profile_delete = cn10k_nix_mtr_profile_delete,
};
int
--
2.25.1
[dpdk-dev] [PATCH v3 19/27] net/cnxk: support ops to validate meter policy
From: Sunil Kumar Kori
Implement API to validate meter policy for CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev_mtr.c | 49 +
1 file changed, 49 insertions(+)
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index ea4d898cd7..4cf4ebd6fd 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -218,10 +218,59 @@ cn10k_nix_mtr_profile_delete(struct rte_eth_dev *eth_dev,
uint32_t profile_id,
return 0;
}
+static int
+cn10k_nix_mtr_policy_validate(struct rte_eth_dev *dev,
+ struct rte_mtr_meter_policy_params *policy,
+ struct rte_mtr_error *error)
+{
+ static const char *const action_color[] = {"Green", "Yellow", "Red"};
+ bool supported[RTE_COLORS] = {false, false, false};
+ const struct rte_flow_action *action;
+ char message[1024];
+ uint32_t i;
+
+ RTE_SET_USED(dev);
+
+ if (!policy)
+ return 0; /* Nothing to be validated */
+
+ for (i = 0; i < RTE_COLORS; i++) {
+ if (policy->actions[i]) {
+ for (action = policy->actions[i];
+action->type != RTE_FLOW_ACTION_TYPE_END;
+action++) {
+ if (action->type == RTE_FLOW_ACTION_TYPE_METER)
+ supported[i] = true;
+
+ if (action->type == RTE_FLOW_ACTION_TYPE_DROP)
+ supported[i] = true;
+
+ if (!supported[i]) {
+ sprintf(message,
+ "%s action is not valid",
+ action_color[i]);
+ return -rte_mtr_error_set(error,
+ ENOTSUP,
+ RTE_MTR_ERROR_TYPE_METER_POLICY, NULL,
+ message);
+ }
+ }
+ } else {
+ sprintf(message, "%s action is null", action_color[i]);
+ return -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_METER_POLICY, NULL,
+ message);
+ }
+ }
+
+ return 0;
+}
+
const struct rte_mtr_ops nix_mtr_ops = {
.capabilities_get = cn10k_nix_mtr_capabilities_get,
.meter_profile_add = cn10k_nix_mtr_profile_add,
.meter_profile_delete = cn10k_nix_mtr_profile_delete,
+ .meter_policy_validate = cn10k_nix_mtr_policy_validate,
};
int
--
2.25.1
[dpdk-dev] [PATCH v3 20/27] net/cnxk: support ops to create meter policy
From: Sunil Kumar Kori
Implement API to add meter policy for CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev_mtr.c | 84 +
drivers/net/cnxk/cnxk_ethdev.c | 1 +
drivers/net/cnxk/cnxk_ethdev.h | 31 +++
3 files changed, 116 insertions(+)
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index 4cf4ebd6fd..bb191666c4 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -50,6 +50,18 @@ nix_mtr_profile_find(struct cnxk_eth_dev *dev, uint32_t
profile_id)
return NULL;
}
+static struct cnxk_mtr_policy_node *
+nix_mtr_policy_find(struct cnxk_eth_dev *dev, uint32_t meter_policy_id)
+{
+ struct cnxk_mtr_policy *fmps = &dev->mtr_policy;
+ struct cnxk_mtr_policy_node *fmp;
+
+ TAILQ_FOREACH(fmp, fmps, next)
+ if (meter_policy_id == fmp->id)
+ return fmp;
+ return NULL;
+}
+
static int
nix_mtr_profile_validate(struct cnxk_eth_dev *dev, uint32_t profile_id,
struct rte_mtr_meter_profile *profile,
@@ -266,11 +278,83 @@ cn10k_nix_mtr_policy_validate(struct rte_eth_dev *dev,
return 0;
}
+static void
+cn10k_fill_policy_actions(struct cnxk_mtr_policy_node *fmp,
+ struct rte_mtr_meter_policy_params *policy)
+
+{
+ const struct rte_flow_action_meter *mtr;
+ const struct rte_flow_action *action;
+ int i;
+
+ for (i = 0; i < RTE_COLORS; i++) {
+ if (policy->actions[i]) {
+ for (action = policy->actions[i];
+action->type != RTE_FLOW_ACTION_TYPE_END;
+action++) {
+ if (action->type ==
+ RTE_FLOW_ACTION_TYPE_METER) {
+ fmp->actions[i].action_fate =
+ action->type;
+ mtr = (const struct
+ rte_flow_action_meter *)
+ action->conf;
+ fmp->actions[i].mtr_id = mtr->mtr_id;
+ }
+
+ if (action->type == RTE_FLOW_ACTION_TYPE_DROP) {
+ fmp->actions[i].action_fate =
+ action->type;
+ }
+ }
+ }
+ }
+}
+
+static int
+cn10k_nix_mtr_policy_add(struct rte_eth_dev *eth_dev, uint32_t policy_id,
+struct rte_mtr_meter_policy_params *policy,
+struct rte_mtr_error *error)
+{
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ struct cnxk_mtr_policy *fmps = &dev->mtr_policy;
+ struct cnxk_mtr_policy_node *fmp;
+ int rc;
+
+ fmp = nix_mtr_policy_find(dev, policy_id);
+ if (fmp) {
+ return -rte_mtr_error_set(error, EEXIST,
+ RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+ NULL, "Policy already exist");
+ }
+
+ fmp = plt_zmalloc(sizeof(struct cnxk_mtr_policy_node), ROC_ALIGN);
+ if (fmp == NULL) {
+ return -rte_mtr_error_set(error, ENOMEM,
+ RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
+ "Memory allocation failure");
+ } else {
+ rc = cn10k_nix_mtr_policy_validate(eth_dev, policy, error);
+ if (rc)
+ goto exit;
+ }
+
+ fmp->id = policy_id;
+ cn10k_fill_policy_actions(fmp, policy);
+ TAILQ_INSERT_TAIL(fmps, fmp, next);
+ return 0;
+
+exit:
+ plt_free(fmp);
+ return rc;
+}
+
const struct rte_mtr_ops nix_mtr_ops = {
.capabilities_get = cn10k_nix_mtr_capabilities_get,
.meter_profile_add = cn10k_nix_mtr_profile_add,
.meter_profile_delete = cn10k_nix_mtr_profile_delete,
.meter_policy_validate = cn10k_nix_mtr_policy_validate,
+ .meter_policy_add = cn10k_nix_mtr_policy_add,
};
int
diff --git a/drivers/net/cnxk/cnxk_ethdev.c b/drivers/net/cnxk/cnxk_ethdev.c
index e952aa5ec5..9e75060513 100644
--- a/drivers/net/cnxk/cnxk_ethdev.c
+++ b/drivers/net/cnxk/cnxk_ethdev.c
@@ -508,6 +508,7 @@ static int
nix_ingress_policer_setup(struct cnxk_eth_dev *dev)
{
TAILQ_INIT(&dev->mtr_profiles);
+ TAILQ_INIT(&dev->mtr_policy);
return 0;
}
diff --g
Re: [dpdk-dev] [PATCH v2 1/3] security: add SA config option for inner pkt csum
Hi Anoob, > > > > External Email > > > > -- > > Hi Anoob, > > > > > Hi Konstanin, > > > > > > Please see inline. > > > > > > Thanks, > > > Anoob > > > > > > > -Original Message- > > > > From: Ananyev, Konstantin > > > > Sent: Wednesday, September 29, 2021 4:26 PM > > > > To: Archana Muniganti ; Akhil Goyal > > > > ; Nicolau, Radu ; Zhang, > > > > Roy Fan ; hemant.agra...@nxp.com > > > > Cc: Anoob Joseph ; Tejasree Kondoj > > > > ; Ankur Dwivedi ; Jerin > > > > Jacob Kollanukkaran ; dev@dpdk.org > > > > Subject: [EXT] RE: [PATCH v2 1/3] security: add SA config option for > > > > inner pkt csum > > > > > > > > External Email > > > > > > > > > > > > -- > > > > > Add inner packet IPv4 hdr and L4 checksum enable options in conf. > > > > > These will be used in case of protocol offload. > > > > > Per SA, application could specify whether the > > > > > checksum(compute/verify) can be offloaded to security device. > > > > > > > > > > Signed-off-by: Archana Muniganti > > > > > --- > > > > > doc/guides/cryptodevs/features/default.ini | 1 + > > > > > doc/guides/rel_notes/deprecation.rst | 4 ++-- > > > > > doc/guides/rel_notes/release_21_11.rst | 4 > > > > > lib/cryptodev/rte_cryptodev.h | 2 ++ > > > > > lib/security/rte_security.h| 18 ++ > > > > > 5 files changed, 27 insertions(+), 2 deletions(-) > > > > > > > > > > diff --git a/doc/guides/cryptodevs/features/default.ini > > > > > b/doc/guides/cryptodevs/features/default.ini > > > > > index c24814de98..96d95ddc81 100644 > > > > > --- a/doc/guides/cryptodevs/features/default.ini > > > > > +++ b/doc/guides/cryptodevs/features/default.ini > > > > > @@ -33,6 +33,7 @@ Non-Byte aligned data = Sym raw data path API > > > > > = Cipher multiple data units = > > > > > Cipher wrapped key = > > > > > +Inner checksum = > > > > > > > > > > ; > > > > > ; Supported crypto algorithms of a default crypto driver. > > > > > diff --git a/doc/guides/rel_notes/deprecation.rst > > > > > b/doc/guides/rel_notes/deprecation.rst > > > > > index 05fc2fdee7..8308e00ed4 100644 > > > > > --- a/doc/guides/rel_notes/deprecation.rst > > > > > +++ b/doc/guides/rel_notes/deprecation.rst > > > > > @@ -232,8 +232,8 @@ Deprecation Notices > > > > >IPsec payload MSS (Maximum Segment Size), and ESN (Extended > > > > > Sequence > > > > Number). > > > > > > > > > > * security: The IPsec SA config options ``struct > > > > > rte_security_ipsec_sa_options`` > > > > > - will be updated with new fields to support new features like > > > > > IPsec inner > > > > > - checksum, TSO in case of protocol offload. > > > > > + will be updated with new fields to support new features like > > > > > + TSO in case of protocol offload. > > > > > > > > > > * ipsec: The structure ``rte_ipsec_sa_prm`` will be extended with a > > > > > new > > field > > > > >``hdr_l3_len`` to configure tunnel L3 header length. > > > > > diff --git a/doc/guides/rel_notes/release_21_11.rst > > > > > b/doc/guides/rel_notes/release_21_11.rst > > > > > index 8da851..93d1b36889 100644 > > > > > --- a/doc/guides/rel_notes/release_21_11.rst > > > > > +++ b/doc/guides/rel_notes/release_21_11.rst > > > > > @@ -194,6 +194,10 @@ ABI Changes > > > > >``rte_security_ipsec_xform`` to allow applications to configure SA > > > > > soft > > > > >and hard expiry limits. Limits can be either in number of packets > > > > > or bytes. > > > > > > > > > > +* security: The new options ``ip_csum_enable`` and > > > > > +``l4_csum_enable`` were added > > > > > + in structure ``rte_security_ipsec_sa_options`` to indicate > > > > > +whether inner > > > > > + packet IPv4 header checksum and L4 checksum need to be > > > > > +offloaded to > > > > > + security device. > > > > > > > > > > Known Issues > > > > > > > > > > diff --git a/lib/cryptodev/rte_cryptodev.h > > > > > b/lib/cryptodev/rte_cryptodev.h index bb01f0f195..d9271a6c45 > > > > > 100644 > > > > > --- a/lib/cryptodev/rte_cryptodev.h > > > > > +++ b/lib/cryptodev/rte_cryptodev.h > > > > > @@ -479,6 +479,8 @@ rte_cryptodev_asym_get_xform_enum(enum > > > > > rte_crypto_asym_xform_type *xform_enum, /**< Support operations > > > > > on > > > > multiple data-units message */ > > > > > #define RTE_CRYPTODEV_FF_CIPHER_WRAPPED_KEY (1ULL > > << 26) > > > > > /**< Support wrapped key in cipher xform */ > > > > > +#define RTE_CRYPTODEV_FF_SECURITY_INNER_CSUM (1ULL > > > > << 27) > > > > > +/**< Support inner checksum computation/verification */ > > > > > > > > > > /** > > > > > * Get the name of a crypto device feature flag diff --git > > > > > a/lib/security/rte_security.h b/lib/security/rte_security.h index > > > > > ab1a6e1f65..945f45ad76 100644 > > > > > --- a/lib/security/rte_security.h > > > > > +++ b/lib/security/rte_securit
[dpdk-dev] [PATCH v3 21/27] net/cnxk: support ops to delete meter policy
From: Sunil Kumar Kori
Implement API to delete meter policy for CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev_mtr.c | 26 ++
1 file changed, 26 insertions(+)
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index bb191666c4..ff865b8fa6 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -349,12 +349,38 @@ cn10k_nix_mtr_policy_add(struct rte_eth_dev *eth_dev,
uint32_t policy_id,
return rc;
}
+static int
+cn10k_nix_mtr_policy_delete(struct rte_eth_dev *eth_dev, uint32_t policy_id,
+ struct rte_mtr_error *error)
+{
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ struct cnxk_mtr_policy_node *fmp;
+
+ fmp = nix_mtr_policy_find(dev, policy_id);
+ if (fmp == NULL) {
+ return -rte_mtr_error_set(error, ENOENT,
+ RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+ NULL, "No policy found");
+ }
+
+ if (fmp->ref_cnt)
+ return -rte_mtr_error_set(error, EBUSY,
+ RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+ NULL, "Meter policy is in use.");
+
+ TAILQ_REMOVE(&dev->mtr_policy, fmp, next);
+ plt_free(fmp);
+
+ return 0;
+}
+
const struct rte_mtr_ops nix_mtr_ops = {
.capabilities_get = cn10k_nix_mtr_capabilities_get,
.meter_profile_add = cn10k_nix_mtr_profile_add,
.meter_profile_delete = cn10k_nix_mtr_profile_delete,
.meter_policy_validate = cn10k_nix_mtr_policy_validate,
.meter_policy_add = cn10k_nix_mtr_policy_add,
+ .meter_policy_delete = cn10k_nix_mtr_policy_delete,
};
int
--
2.25.1
[dpdk-dev] [PATCH v3 22/27] net/cnxk: support ops to create meter
From: Sunil Kumar Kori
Implement API to create meter instance for CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev_mtr.c | 97 +
drivers/net/cnxk/cnxk_ethdev.c | 1 +
drivers/net/cnxk/cnxk_ethdev.h | 25
3 files changed, 123 insertions(+)
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index ff865b8fa6..8916dc329f 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -37,6 +37,18 @@ static struct rte_mtr_capabilities mtr_capa = {
RTE_MTR_STATS_N_BYTES_YELLOW | RTE_MTR_STATS_N_BYTES_RED |
RTE_MTR_STATS_N_BYTES_DROPPED};
+static struct cnxk_meter_node *
+nix_mtr_find(struct cnxk_eth_dev *dev, uint32_t meter_id)
+{
+ struct cnxk_mtr *fms = &dev->mtr;
+ struct cnxk_meter_node *fm;
+
+ TAILQ_FOREACH(fm, fms, next)
+ if (meter_id == fm->id)
+ return fm;
+ return NULL;
+}
+
static struct cnxk_mtr_profile_node *
nix_mtr_profile_find(struct cnxk_eth_dev *dev, uint32_t profile_id)
{
@@ -374,6 +386,90 @@ cn10k_nix_mtr_policy_delete(struct rte_eth_dev *eth_dev,
uint32_t policy_id,
return 0;
}
+static int
+cn10k_nix_mtr_create(struct rte_eth_dev *eth_dev, uint32_t mtr_id,
+struct rte_mtr_params *params, int shared,
+struct rte_mtr_error *error)
+{
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ struct cnxk_mtr_profile_node *profile;
+ struct cnxk_mtr_policy_node *policy;
+ struct cnxk_mtr *fm = &dev->mtr;
+ struct cnxk_meter_node *mtr;
+ int i;
+
+ RTE_SET_USED(shared);
+
+ if (params == NULL)
+ return -rte_mtr_error_set(error, ENOENT,
+ RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
+ "Meter params are invalid.");
+
+ profile = nix_mtr_profile_find(dev, params->meter_profile_id);
+ if (profile == NULL)
+ return -rte_mtr_error_set(error, ENOENT,
+ RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+ ¶ms->meter_profile_id,
+ "Meter profile is invalid.");
+
+ policy = nix_mtr_policy_find(dev, params->meter_policy_id);
+ if (policy == NULL)
+ return -rte_mtr_error_set(error, ENOENT,
+ RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+ ¶ms->meter_policy_id,
+ "Meter policy is invalid.");
+
+ mtr = nix_mtr_find(dev, mtr_id);
+ if (mtr == NULL) {
+ mtr = plt_zmalloc(sizeof(struct cnxk_meter_node), ROC_ALIGN);
+ if (mtr == NULL) {
+ return -rte_mtr_error_set(error, ENOMEM,
+ RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
+ "Meter memory alloc failed.");
+ } else {
+ mtr->id = mtr_id;
+ mtr->profile = profile;
+ mtr->policy = policy;
+ mtr->params = *params;
+ mtr->bpf_id = ROC_NIX_BPF_ID_INVALID;
+ mtr->prev_cnt = 0;
+ for (i = 0; i < MAX_PRV_MTR_NODES; i++)
+ mtr->prev_id[i] = ROC_NIX_BPF_ID_INVALID;
+ mtr->next_id = ROC_NIX_BPF_ID_INVALID;
+ mtr->is_prev = false;
+ mtr->is_next = false;
+ mtr->level = ROC_NIX_BPF_LEVEL_IDX_INVALID;
+
+ if (params->dscp_table) {
+ mtr->params.dscp_table =
+ plt_zmalloc(ROC_NIX_BPF_PRE_COLOR_MAX,
+ ROC_ALIGN);
+ if (mtr->params.dscp_table == NULL) {
+ plt_free(mtr);
+ return -rte_mtr_error_set(error, ENOMEM,
+ RTE_MTR_ERROR_TYPE_UNSPECIFIED,
+ NULL, "Memory alloc failed.");
+ }
+
+ for (i = 0; i < ROC_NIX_BPF_PRE_COLOR_MAX;
+i++) {
+ mtr->params.dscp_table[i] =
+ params->dscp_table[i];
+
[dpdk-dev] [PATCH v3 23/27] net/cnxk: support ops to delete meter
From: Sunil Kumar Kori
Implement API to delete meter instance for CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev_mtr.c | 82 +
1 file changed, 82 insertions(+)
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index 8916dc329f..6d48680c19 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -8,6 +8,10 @@
#define NIX_MTR_COUNT_MAX 73 /* 64(leaf) + 8(mid) + 1(top) */
#define NIX_MTR_COUNT_PER_FLOW 3 /* 1(leaf) + 1(mid) + 1(top) */
+static const enum roc_nix_bpf_level_flag lvl_map[] = {ROC_NIX_BPF_LEVEL_F_LEAF,
+ ROC_NIX_BPF_LEVEL_F_MID,
+ ROC_NIX_BPF_LEVEL_F_TOP};
+
static struct rte_mtr_capabilities mtr_capa = {
.n_max = NIX_MTR_COUNT_MAX,
.n_shared_max = NIX_MTR_COUNT_PER_FLOW,
@@ -470,6 +474,83 @@ cn10k_nix_mtr_create(struct rte_eth_dev *eth_dev, uint32_t
mtr_id,
return 0;
}
+static int
+cn10k_nix_mtr_destroy(struct rte_eth_dev *eth_dev, uint32_t mtr_id,
+ struct rte_mtr_error *error)
+{
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ struct roc_nix_bpf_objs profs = {0};
+ struct cnxk_mtr *fm = &dev->mtr;
+ struct roc_nix *nix = &dev->nix;
+ struct cnxk_meter_node *mtr;
+ int rc = 0;
+
+ mtr = nix_mtr_find(dev, mtr_id);
+ if (mtr == NULL) {
+ return -rte_mtr_error_set(error, ENOENT,
+ RTE_MTR_ERROR_TYPE_MTR_ID, &mtr_id,
+ "Meter id is invalid.");
+ }
+
+ if (mtr->ref_cnt) {
+ return -rte_mtr_error_set(error, EADDRINUSE,
+ RTE_MTR_ERROR_TYPE_MTR_ID, &mtr_id,
+ "Meter id in use.");
+ }
+
+ switch (lvl_map[mtr->level]) {
+ case ROC_NIX_BPF_LEVEL_F_LEAF:
+ if (mtr->is_next) {
+ rc = roc_nix_bpf_connect(nix, ROC_NIX_BPF_LEVEL_F_LEAF,
+mtr_id,
+ROC_NIX_BPF_ID_INVALID);
+ }
+ break;
+ case ROC_NIX_BPF_LEVEL_F_MID:
+ while (mtr->prev_cnt) {
+ rc = roc_nix_bpf_connect(nix, ROC_NIX_BPF_LEVEL_F_LEAF,
+mtr->prev_id[mtr->prev_cnt],
+ROC_NIX_BPF_ID_INVALID);
+ mtr->prev_cnt--;
+ }
+ if (mtr->is_next) {
+ rc = roc_nix_bpf_connect(nix, ROC_NIX_BPF_LEVEL_F_MID,
+mtr_id,
+ROC_NIX_BPF_ID_INVALID);
+ }
+ break;
+ case ROC_NIX_BPF_LEVEL_F_TOP:
+ while (mtr->prev_cnt) {
+ rc = roc_nix_bpf_connect(nix, ROC_NIX_BPF_LEVEL_F_MID,
+mtr->prev_id[mtr->prev_cnt],
+ROC_NIX_BPF_ID_INVALID);
+ mtr->prev_cnt--;
+ }
+ break;
+ default:
+ return -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_MTR_ID, NULL,
+ "Invalid meter level");
+ }
+
+ if (rc)
+ goto exit;
+
+ profs.level = mtr->level;
+ profs.count = 1;
+ profs.ids[0] = mtr->bpf_id;
+ rc = roc_nix_bpf_free(nix, &profs, 1);
+ if (rc)
+ goto exit;
+
+ TAILQ_REMOVE(fm, mtr, next);
+ plt_free(mtr->params.dscp_table);
+ plt_free(mtr);
+
+exit:
+ return rc;
+}
+
const struct rte_mtr_ops nix_mtr_ops = {
.capabilities_get = cn10k_nix_mtr_capabilities_get,
.meter_profile_add = cn10k_nix_mtr_profile_add,
@@ -478,6 +559,7 @@ const struct rte_mtr_ops nix_mtr_ops = {
.meter_policy_add = cn10k_nix_mtr_policy_add,
.meter_policy_delete = cn10k_nix_mtr_policy_delete,
.create = cn10k_nix_mtr_create,
+ .destroy = cn10k_nix_mtr_destroy,
};
int
--
2.25.1
[dpdk-dev] [PATCH v3 24/27] net/cnxk: support ops to enable/disable meter
From: Sunil Kumar Kori
Implement API to enable or disable meter instance for
CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev_mtr.c | 60 +
1 file changed, 60 insertions(+)
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index 6d48680c19..88153563c2 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -551,6 +551,64 @@ cn10k_nix_mtr_destroy(struct rte_eth_dev *eth_dev,
uint32_t mtr_id,
return rc;
}
+static int
+cn10k_nix_mtr_enable(struct rte_eth_dev *eth_dev, uint32_t mtr_id,
+struct rte_mtr_error *error)
+{
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ struct roc_nix *nix = &dev->nix;
+ struct cnxk_meter_node *mtr;
+ struct roc_nix_rq *rq;
+ uint32_t i;
+ int rc = 0;
+
+ mtr = nix_mtr_find(dev, mtr_id);
+ if (mtr == NULL) {
+ return -rte_mtr_error_set(error, ENOENT,
+ RTE_MTR_ERROR_TYPE_MTR_ID, NULL,
+ "Meter id is invalid.");
+ }
+
+ if (mtr->level != 0)
+ return 0;
+
+ for (i = 0; i < mtr->rq_num; i++) {
+ rq = &dev->rqs[mtr->rq_id[i]];
+ rc |= roc_nix_bpf_ena_dis(nix, mtr->bpf_id, rq, true);
+ }
+
+ return rc;
+}
+
+static int
+cn10k_nix_mtr_disable(struct rte_eth_dev *eth_dev, uint32_t mtr_id,
+ struct rte_mtr_error *error)
+{
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ struct roc_nix *nix = &dev->nix;
+ struct cnxk_meter_node *mtr;
+ struct roc_nix_rq *rq;
+ uint32_t i;
+ int rc = 0;
+
+ mtr = nix_mtr_find(dev, mtr_id);
+ if (mtr == NULL) {
+ return -rte_mtr_error_set(error, ENOENT,
+ RTE_MTR_ERROR_TYPE_MTR_ID, NULL,
+ "Meter id is invalid.");
+ }
+
+ if (mtr->level != 0)
+ return 0;
+
+ for (i = 0; i < mtr->rq_num; i++) {
+ rq = &dev->rqs[mtr->rq_id[i]];
+ rc |= roc_nix_bpf_ena_dis(nix, mtr->bpf_id, rq, false);
+ }
+
+ return rc;
+}
+
const struct rte_mtr_ops nix_mtr_ops = {
.capabilities_get = cn10k_nix_mtr_capabilities_get,
.meter_profile_add = cn10k_nix_mtr_profile_add,
@@ -560,6 +618,8 @@ const struct rte_mtr_ops nix_mtr_ops = {
.meter_policy_delete = cn10k_nix_mtr_policy_delete,
.create = cn10k_nix_mtr_create,
.destroy = cn10k_nix_mtr_destroy,
+ .meter_enable = cn10k_nix_mtr_enable,
+ .meter_disable = cn10k_nix_mtr_disable,
};
int
--
2.25.1
[dpdk-dev] [PATCH v3 25/27] net/cnxk: support ops to update precolor DSCP table
From: Sunil Kumar Kori
Implement API to update DSCP table for pre-coloring for
incoming packet per nixlf for CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev_mtr.c | 43 +
1 file changed, 43 insertions(+)
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index 88153563c2..55485601fb 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -609,6 +609,48 @@ cn10k_nix_mtr_disable(struct rte_eth_dev *eth_dev,
uint32_t mtr_id,
return rc;
}
+static int
+cn10k_nix_mtr_dscp_table_update(struct rte_eth_dev *eth_dev, uint32_t mtr_id,
+ enum rte_color *dscp_table,
+ struct rte_mtr_error *error)
+{
+ enum roc_nix_bpf_color nix_dscp_tbl[ROC_NIX_BPF_PRE_COLOR_MAX];
+ enum roc_nix_bpf_color color_map[] = {ROC_NIX_BPF_COLOR_GREEN,
+ ROC_NIX_BPF_COLOR_YELLOW,
+ ROC_NIX_BPF_COLOR_RED};
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ uint8_t lvl_flag = ROC_NIX_BPF_LEVEL_F_LEAF;
+ struct roc_nix_bpf_precolor table;
+ struct roc_nix *nix = &dev->nix;
+ int rc, i;
+
+ if (!dscp_table) {
+ for (i = 0; i < ROC_NIX_BPF_PRE_COLOR_MAX; i++)
+ nix_dscp_tbl[i] = ROC_NIX_BPF_COLOR_GREEN;
+ } else {
+ for (i = 0; i < ROC_NIX_BPF_PRE_COLOR_MAX; i++)
+ nix_dscp_tbl[i] = color_map[dscp_table[i]];
+ }
+
+ table.count = ROC_NIX_BPF_PRE_COLOR_MAX;
+ table.mode = ROC_NIX_BPF_PC_MODE_DSCP_OUTER;
+ for (i = 0; i < ROC_NIX_BPF_PRE_COLOR_MAX; i++)
+ table.color[i] = nix_dscp_tbl[i];
+
+ rc = roc_nix_bpf_pre_color_tbl_setup(nix, mtr_id, lvl_flag, &table);
+ if (rc) {
+ rte_mtr_error_set(error, rc, RTE_MTR_ERROR_TYPE_UNSPECIFIED,
+ NULL, NULL);
+ goto exit;
+ }
+
+ for (i = 0; i < ROC_NIX_BPF_PRE_COLOR_MAX; i++)
+ dev->precolor_tbl[i] = nix_dscp_tbl[i];
+
+exit:
+ return rc;
+}
+
const struct rte_mtr_ops nix_mtr_ops = {
.capabilities_get = cn10k_nix_mtr_capabilities_get,
.meter_profile_add = cn10k_nix_mtr_profile_add,
@@ -620,6 +662,7 @@ const struct rte_mtr_ops nix_mtr_ops = {
.destroy = cn10k_nix_mtr_destroy,
.meter_enable = cn10k_nix_mtr_enable,
.meter_disable = cn10k_nix_mtr_disable,
+ .meter_dscp_table_update = cn10k_nix_mtr_dscp_table_update,
};
int
--
2.25.1
[dpdk-dev] [PATCH v3 26/27] net/cnxk: support ops to read/update meter stats
From: Sunil Kumar Kori
Implement API to read and update stats corresponding to
given meter instance for CN10K platform.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
drivers/net/cnxk/cn10k_ethdev_mtr.c | 141
1 file changed, 141 insertions(+)
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index 55485601fb..62f48c534f 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -8,6 +8,21 @@
#define NIX_MTR_COUNT_MAX 73 /* 64(leaf) + 8(mid) + 1(top) */
#define NIX_MTR_COUNT_PER_FLOW 3 /* 1(leaf) + 1(mid) + 1(top) */
+#define NIX_BPF_STATS_MASK_ALL
\
+ { \
+ ROC_NIX_BPF_GREEN_PKT_F_PASS | ROC_NIX_BPF_GREEN_OCTS_F_PASS | \
+ ROC_NIX_BPF_GREEN_PKT_F_DROP | \
+ ROC_NIX_BPF_GREEN_OCTS_F_DROP |\
+ ROC_NIX_BPF_YELLOW_PKT_F_PASS |\
+ ROC_NIX_BPF_YELLOW_OCTS_F_PASS | \
+ ROC_NIX_BPF_YELLOW_PKT_F_DROP |\
+ ROC_NIX_BPF_YELLOW_OCTS_F_DROP | \
+ ROC_NIX_BPF_RED_PKT_F_PASS | \
+ ROC_NIX_BPF_RED_OCTS_F_PASS | \
+ ROC_NIX_BPF_RED_PKT_F_DROP | \
+ ROC_NIX_BPF_RED_OCTS_F_DROP\
+ }
+
static const enum roc_nix_bpf_level_flag lvl_map[] = {ROC_NIX_BPF_LEVEL_F_LEAF,
ROC_NIX_BPF_LEVEL_F_MID,
ROC_NIX_BPF_LEVEL_F_TOP};
@@ -651,6 +666,130 @@ cn10k_nix_mtr_dscp_table_update(struct rte_eth_dev
*eth_dev, uint32_t mtr_id,
return rc;
}
+static int
+cn10k_nix_mtr_stats_update(struct rte_eth_dev *eth_dev, uint32_t mtr_id,
+ uint64_t stats_mask, struct rte_mtr_error *error)
+{
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ struct cnxk_meter_node *mtr;
+
+ if (!stats_mask)
+ return -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
+ "no bit is set to stats mask");
+
+ mtr = nix_mtr_find(dev, mtr_id);
+ if (mtr == NULL) {
+ return -rte_mtr_error_set(error, ENOENT,
+ RTE_MTR_ERROR_TYPE_MTR_ID, NULL,
+ "Meter object not found");
+ }
+
+ mtr->params.stats_mask = stats_mask;
+ return 0;
+}
+
+static int
+cn10k_nix_mtr_stats_read(struct rte_eth_dev *eth_dev, uint32_t mtr_id,
+struct rte_mtr_stats *stats, uint64_t *stats_mask,
+int clear, struct rte_mtr_error *error)
+{
+ uint8_t yellow_pkt_pass, yellow_octs_pass, yellow_pkt_drop;
+ uint8_t green_octs_drop, yellow_octs_drop, red_octs_drop;
+ uint8_t green_pkt_pass, green_octs_pass, green_pkt_drop;
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ uint8_t red_pkt_pass, red_octs_pass, red_pkt_drop;
+ uint64_t bpf_stats[ROC_NIX_BPF_STATS_MAX] = {0};
+ uint8_t lvl_flag = ROC_NIX_BPF_LEVEL_F_LEAF;
+ uint64_t mask = NIX_BPF_STATS_MASK_ALL;
+ struct roc_nix *nix = &dev->nix;
+ struct cnxk_meter_node *mtr;
+ int rc;
+
+ if (!stats)
+ return -rte_mtr_error_set(error, EINVAL,
+ RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
+ "stats pointer is NULL");
+
+ mtr = nix_mtr_find(dev, mtr_id);
+ if (mtr == NULL) {
+ return -rte_mtr_error_set(error, ENOENT,
+ RTE_MTR_ERROR_TYPE_MTR_ID, NULL,
+ "Meter object not found");
+ }
+
+ rc = roc_nix_bpf_stats_read(nix, mtr->bpf_id, mask, lvl_flag,
+ bpf_stats);
+ if (rc) {
+ rte_mtr_error_set(error, rc, RTE_MTR_ERROR_TYPE_UNSPECIFIED,
+ NULL, NULL);
+ goto exit;
+ }
+
+ green_pkt_pass = roc_nix_bpf_stats_to_idx(ROC_NIX_BPF_GREEN_PKT_F_PASS);
+ green_octs_pass =
+ roc_nix_bpf_stats_to_idx(ROC_NIX_BPF_GREEN_OCTS_F_PASS);
+ green_pkt_drop =
[dpdk-dev] [PATCH v3 27/27] net/cnxk: support meter action to flow create
From: Sunil Kumar Kori
Meters are configured per flow using rte_flow_create API.
Implement support for meter action applied on the flow.
Signed-off-by: Sunil Kumar Kori
Signed-off-by: Rakesh Kudurumalla
---
v3:
- Rebase support on latest DPDK
- Handled multilevel chaining for tree hierarchy
- Fix naming convention
v2:
- Rebase support on latest DPDK
- Handled multilevel chaining for linear hierarchy
- Review comments incorporated
doc/guides/nics/features/cnxk.ini| 1 +
doc/guides/nics/features/cnxk_vf.ini | 1 +
drivers/net/cnxk/cn10k_ethdev_mtr.c | 514 +++
drivers/net/cnxk/cn10k_rte_flow.c| 171 -
drivers/net/cnxk/cnxk_ethdev.h | 18 +
drivers/net/cnxk/cnxk_rte_flow.c | 4 +
6 files changed, 708 insertions(+), 1 deletion(-)
diff --git a/doc/guides/nics/features/cnxk.ini
b/doc/guides/nics/features/cnxk.ini
index 5d456257bd..7bbce7dafc 100644
--- a/doc/guides/nics/features/cnxk.ini
+++ b/doc/guides/nics/features/cnxk.ini
@@ -78,6 +78,7 @@ count= Y
drop = Y
flag = Y
mark = Y
+meter= Y
of_pop_vlan = Y
of_push_vlan = Y
of_set_vlan_pcp = Y
diff --git a/doc/guides/nics/features/cnxk_vf.ini
b/doc/guides/nics/features/cnxk_vf.ini
index 7b4299f0be..89802a27f9 100644
--- a/doc/guides/nics/features/cnxk_vf.ini
+++ b/doc/guides/nics/features/cnxk_vf.ini
@@ -70,6 +70,7 @@ count= Y
drop = Y
flag = Y
mark = Y
+meter= Y
of_pop_vlan = Y
of_push_vlan = Y
of_set_vlan_pcp = Y
diff --git a/drivers/net/cnxk/cn10k_ethdev_mtr.c
b/drivers/net/cnxk/cn10k_ethdev_mtr.c
index 62f48c534f..0b98489dee 100644
--- a/drivers/net/cnxk/cn10k_ethdev_mtr.c
+++ b/drivers/net/cnxk/cn10k_ethdev_mtr.c
@@ -814,3 +814,517 @@ cn10k_nix_mtr_ops_get(struct rte_eth_dev *dev, void *ops)
*(const void **)ops = &nix_mtr_ops;
return 0;
}
+
+int
+nix_mtr_validate(struct rte_eth_dev *eth_dev, uint32_t id)
+{
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ struct cnxk_mtr_profile_node *profile;
+ struct cnxk_mtr_policy_node *policy;
+ struct cnxk_meter_node *mtr;
+
+ mtr = nix_mtr_find(dev, id);
+ if (mtr == NULL)
+ return -EINVAL;
+
+ profile = nix_mtr_profile_find(dev, mtr->params.meter_profile_id);
+ if (profile == NULL)
+ return -EINVAL;
+
+ policy = nix_mtr_policy_find(dev, mtr->params.meter_policy_id);
+ if (policy == NULL)
+ return -EINVAL;
+
+ return 0;
+}
+
+int
+nix_mtr_policy_act_get(struct rte_eth_dev *eth_dev, uint32_t id,
+ struct cnxk_mtr_policy_node **policy_act)
+{
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ struct cnxk_mtr_policy_node *policy;
+ struct cnxk_meter_node *mtr;
+
+ mtr = nix_mtr_find(dev, id);
+ if (mtr == NULL)
+ return -EINVAL;
+
+ policy = nix_mtr_policy_find(dev, mtr->params.meter_policy_id);
+ if (policy == NULL)
+ return -EINVAL;
+
+ *policy_act = policy;
+
+ return 0;
+}
+
+int
+nix_mtr_rq_update(struct rte_eth_dev *eth_dev, uint32_t id, uint32_t queue_num,
+ const uint16_t *queue)
+{
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ struct cnxk_meter_node *mtr;
+ uint32_t i;
+
+ mtr = nix_mtr_find(dev, id);
+ if (mtr == NULL)
+ return -EINVAL;
+
+ mtr->rq_id = plt_zmalloc(queue_num * sizeof(uint32_t), ROC_ALIGN);
+ if (mtr->rq_id == NULL)
+ return -ENOMEM;
+
+ mtr->rq_num = queue_num;
+ for (i = 0; i < queue_num; i++)
+ mtr->rq_id[i] = queue[i];
+
+ return 0;
+}
+
+int
+nix_mtr_chain_reset(struct rte_eth_dev *eth_dev, uint32_t cur_id)
+{
+ struct cnxk_meter_node *mtr[ROC_NIX_BPF_LEVEL_MAX] = {0};
+ struct cnxk_eth_dev *dev = cnxk_eth_pmd_priv(eth_dev);
+ uint32_t mtr_id = cur_id;
+ int i = 0, j = 0;
+
+ for (i = 0; i < ROC_NIX_BPF_LEVEL_MAX; i++) {
+ mtr[i] = nix_mtr_find(dev, mtr_id);
+ if (mtr[i])
+ mtr_id = mtr[i]->next_id;
+ }
+ for (i = 0; i < ROC_NIX_BPF_LEVEL_MAX; i++) {
+ if (mtr[i]) {
+ for (j = 0; j < MAX_PRV_MTR_NODES; j++)
+ mtr[i]->prev_id[i] = ROC_NIX_BPF_ID_INVALID;
+ mtr[i]->level = ROC_NIX_BPF_LEVEL_IDX_INVALID;
+ mtr[i]->next_id = ROC_NIX_BPF_ID_INVALID;
+ mtr[i]->is_prev = false;
+ mtr[i]->is_next = false;
+ mtr[i]->prev_cnt = 0;
+ }
+ }
+ return 0;
+}
+
+int
+nix_mtr_chain_update(struct rte_eth_dev *eth_dev, uint32_t cur_id,
+uint32_t p
Re: [dpdk-dev] [dpdk-stable] [PATCH] net/iavf: fix multi-process shared data
> -Original Message-
> From: Yigit, Ferruh
> Sent: Thursday, September 30, 2021 12:28 AM
> To: Yu, DapengX ; Richardson, Bruce
> ; Ananyev, Konstantin
> ; Wu, Jingjing ;
> Xing, Beilei
> Cc: dev@dpdk.org; sta...@dpdk.org
> Subject: Re: [dpdk-stable] [PATCH] net/iavf: fix multi-process shared data
>
> On 9/28/2021 4:37 AM, dapengx...@intel.com wrote:
> > From: Dapeng Yu
> >
> > When the iavf_adapter instance is not initialized completedly in the
> > primary process, the secondary process accesses its "rte_eth_dev"
> > member, it causes secondary process crash.
> >
> > This patch replaces adapter->eth_dev with rte_eth_devices[port_id] in
> > the data paths where rte_eth_dev instance is accessed.
> >
> > Fixes: f978c1c9b3b5 ("net/iavf: add RSS hash parsing in AVX path")
> > Fixes: 9c9aa0040344 ("net/iavf: add offload path for Rx AVX512 flex
> > descriptor")
> > Fixes: 63660ea3ee0b ("net/iavf: add RSS hash parsing in SSE path")
> > Cc: sta...@dpdk.org
> >
> > Signed-off-by: Dapeng Yu
> > ---
> > drivers/net/iavf/iavf_rxtx_vec_avx2.c | 5 +++--
> > drivers/net/iavf/iavf_rxtx_vec_avx512.c | 5 +++--
> > drivers/net/iavf/iavf_rxtx_vec_sse.c| 3 ++-
> > 3 files changed, 8 insertions(+), 5 deletions(-)
> >
> > diff --git a/drivers/net/iavf/iavf_rxtx_vec_avx2.c
> > b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
> > index 475070e036..59b086ade5 100644
> > --- a/drivers/net/iavf/iavf_rxtx_vec_avx2.c
> > +++ b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
> > @@ -525,6 +525,7 @@ _iavf_recv_raw_pkts_vec_avx2_flex_rxd(struct
> > iavf_rx_queue *rxq, #define IAVF_DESCS_PER_LOOP_AVX 8
> >
> > const uint32_t *type_table = rxq->vsi->adapter->ptype_tbl;
> > + struct rte_eth_dev *dev = &rte_eth_devices[rxq->port_id];
> >
>
> It is not good idea to access global variable directly from the driver.
In "lib/ethdev/rte_ethdev.h", the global variable rte_eth_devices is used.
So I think use it in a PMD should be also acceptable since it is just read.
rte_eth_rx_burst(uint16_t port_id, uint16_t queue_id,
struct rte_mbuf **rx_pkts, const uint16_t nb_pkts)
{
struct rte_eth_dev *dev = &rte_eth_devices[port_id];
>
> The problem definition is correct, eth_dev is unique per process, so it can't
> be saved to a shared struct.
>
> But here I assume real intention is to be able to access PMD specific data
> from queue struct, for this what about storing 'rte_eth_dev_data' in the
> 'iavf_rx_queue', this should sove the problem without accessing the global
> variable.
The intention is to read the offload properties of device configuration, so it
not
queue specific or PMD specific. It is already in public data structure.
If it is stored in 'iavf_rx_queue' again, the data will be duplicate.
Re: [dpdk-dev] [PATCH v2] net/virtio: fix virtio-user init when using existing tap
On 9/29/21 15:19, Ferruh Yigit wrote: On 9/28/2021 9:51 AM, David Marchand wrote: When attaching to an existing mono queue tap, the virtio-user was not reporting that the virtio device was not properly initialised which prevented from starting the port later. $ ip tuntap add test mode tap $ dpdk-testpmd --vdev \ net_virtio_user0,iface=test,path=/dev/vhost-net,queues=2 -- -i ... virtio_user_dev_init_mac(): (/dev/vhost-net) No valid MAC in devargs or device, use random vhost_kernel_open_tap(): TUNSETIFF failed: Invalid argument vhost_kernel_enable_queue_pair(): fail to open tap for vhost kernel virtio_user_start_device(): (/dev/vhost-net) Failed to start device ... Configuring Port 0 (socket 0) vhost_kernel_open_tap(): TUNSETIFF failed: Invalid argument vhost_kernel_enable_queue_pair(): fail to open tap for vhost kernel virtio_set_multiple_queues(): Multiqueue configured but send command failed, this is too late now... Fail to start port 0: Invalid argument Please stop the ports first Done The virtio-user with vhost-kernel backend was going through a lot of complications to initialise tap fds only when using them. For each qp enabled for the first time, a tapfd was created via TUNSETIFF with unneeded additional steps (see below) and then mapped to the right qp in the vhost-net backend. Unneeded steps (as long as it has been done once for the port): - tap features were queried while this is a constant on a running system, - the device name in DPDK was updated, - the mac address of the tap was set, On subsequent qps state change, the vhost-net backend fd mapping was updated and the associated queue/tapfd were disabled/enabled via TUNSETQUEUE. Now, this patch simplifies the whole logic by keeping all tapfds opened and in enabled state (from the tap point of view) at all time. Unused ioctl defines are removed. Tap features are validated earlier to fail initialisation asap. Tap name discovery and mac address configuration are moved when configuring qp 0. To support attaching to mono queue tap, the virtio-user driver now tries to attach in multi queue first, then fallbacks to mono queue. Finally (but this is more for consistency), VIRTIO_NET_F_MQ feature is exposed only if the underlying tap supports multi queue. Signed-off-by: David Marchand Do we want to backport this patch? If so can you please provide a fixes tag? No, we don't want to backport it. Thanks, Maxime
[dpdk-dev] FW: [dpdk-dev v1] test/crypto: maxlen calculation update
CCing to sta...@dpdk.org
> -Original Message-
> From: Ji, Kai
> Sent: Wednesday, September 29, 2021 4:38 PM
> To: dev@dpdk.org
> Cc: Zhang, Roy Fan ; Ji, Kai
> Subject: [dpdk-dev v1] test/crypto: maxlen calculation update
>
> Update the calculation of the max length needed when converting mbuf to
> data vec in partial digest test case. This update make sure the enough
> vec buffers are allocated for the appended digest in sgl op for QAT raw
> datapath api.
>
> Fixes: 4868f6591c6f ("test/crypto: add cases for raw datapath API")
> Cc: roy.fan.zh...@intel.com
>
> Signed-off-by: Kai Ji
> ---
> app/test/test_cryptodev.c | 35 +++
> 1 file changed, 31 insertions(+), 4 deletions(-)
>
> diff --git a/app/test/test_cryptodev.c b/app/test/test_cryptodev.c
> index 16d770a17f..ea46911648 100644
> --- a/app/test/test_cryptodev.c
> +++ b/app/test/test_cryptodev.c
> @@ -167,6 +167,10 @@ post_process_raw_dp_op(void *user_data,
> uint32_t index __rte_unused,
> RTE_CRYPTO_OP_STATUS_ERROR;
> }
>
> +static struct crypto_testsuite_params testsuite_params = { NULL };
> +struct crypto_testsuite_params *p_testsuite_params = &testsuite_params;
> +static struct crypto_unittest_params unittest_params;
> +
> void
> process_sym_raw_dp_op(uint8_t dev_id, uint16_t qp_id,
> struct rte_crypto_op *op, uint8_t is_cipher, uint8_t is_auth,
> @@ -181,6 +185,7 @@ process_sym_raw_dp_op(uint8_t dev_id, uint16_t
> qp_id,
> struct rte_crypto_sgl sgl;
> uint32_t max_len;
> union rte_cryptodev_session_ctx sess;
> + uint64_t auth_end_iova;
> uint32_t count = 0;
> struct rte_crypto_raw_dp_ctx *ctx;
> uint32_t cipher_offset = 0, cipher_len = 0, auth_offset = 0,
> @@ -190,6 +195,9 @@ process_sym_raw_dp_op(uint8_t dev_id, uint16_t
> qp_id,
> int ctx_service_size;
> int32_t status = 0;
> int enqueue_status, dequeue_status;
> + struct crypto_unittest_params *ut_params = &unittest_params;
> + /* oop is not supported in raw hw dp api*/
> + int is_sgl = sop->m_src->nb_segs > 1;
>
> ctx_service_size = rte_cryptodev_get_raw_dp_ctx_size(dev_id);
> if (ctx_service_size < 0) {
> @@ -255,6 +263,29 @@ process_sym_raw_dp_op(uint8_t dev_id, uint16_t
> qp_id,
> digest.va = (void *)sop->auth.digest.data;
> digest.iova = sop->auth.digest.phys_addr;
>
> + if (is_sgl) {
> + uint32_t remaining_off = auth_offset + auth_len;
> + struct rte_mbuf *sgl_buf = sop->m_src;
> +
> + while (remaining_off >=
> rte_pktmbuf_data_len(sgl_buf)
> + && sgl_buf->next != NULL) {
> + remaining_off -=
> rte_pktmbuf_data_len(sgl_buf);
> + sgl_buf = sgl_buf->next;
> + }
> +
> + auth_end_iova =
> (uint64_t)rte_pktmbuf_iova_offset(
> + sgl_buf, remaining_off);
> + } else {
> + /* oop is not supported in raw hw dp api */
> + auth_end_iova = rte_pktmbuf_iova(op->sym-
> >m_src) +
> + auth_offset +
> auth_len;
> + }
> + /* Then check if digest-encrypted conditions are met */
> + if ((auth_offset + auth_len < cipher_offset + cipher_len) &&
> + (digest.iova == auth_end_iova) && is_sgl)
> + max_len = RTE_MAX(max_len, auth_offset +
> auth_len +
> + ut_params->auth_xform.auth.digest_length);
> +
> } else if (is_cipher) {
> cipher_offset = sop->cipher.data.offset;
> cipher_len = sop->cipher.data.length;
> @@ -477,10 +508,6 @@ process_crypto_request(uint8_t dev_id, struct
> rte_crypto_op *op)
> return op;
> }
>
> -static struct crypto_testsuite_params testsuite_params = { NULL };
> -struct crypto_testsuite_params *p_testsuite_params = &testsuite_params;
> -static struct crypto_unittest_params unittest_params;
> -
> static int
> testsuite_setup(void)
> {
> --
> 2.17.1
[dpdk-dev] [PATCH V5 1/5] common/mlx5: update new MMO HCA capabilities
New MMO HCA capabilities were added and others were renamed.
Align hca capabilities with new prm.
Add support in devx interface for changes in HCA capabilities.
Signed-off-by: Raja Zidane
Acked-by: Matan Azrad
---
drivers/common/mlx5/mlx5_devx_cmds.c | 15 ---
drivers/common/mlx5/mlx5_devx_cmds.h | 11 ---
drivers/common/mlx5/mlx5_prm.h| 20 ++--
drivers/compress/mlx5/mlx5_compress.c | 4 ++--
4 files changed, 36 insertions(+), 14 deletions(-)
diff --git a/drivers/common/mlx5/mlx5_devx_cmds.c
b/drivers/common/mlx5/mlx5_devx_cmds.c
index ac554cca05..00c78b1288 100644
--- a/drivers/common/mlx5/mlx5_devx_cmds.c
+++ b/drivers/common/mlx5/mlx5_devx_cmds.c
@@ -858,9 +858,18 @@ mlx5_devx_cmd_query_hca_attr(void *ctx,
attr->log_max_srq_sz = MLX5_GET(cmd_hca_cap, hcattr, log_max_srq_sz);
attr->reg_c_preserve =
MLX5_GET(cmd_hca_cap, hcattr, reg_c_preserve);
- attr->mmo_dma_en = MLX5_GET(cmd_hca_cap, hcattr, dma_mmo);
- attr->mmo_compress_en = MLX5_GET(cmd_hca_cap, hcattr, compress);
- attr->mmo_decompress_en = MLX5_GET(cmd_hca_cap, hcattr, decompress);
+ attr->mmo_regex_qp_en = MLX5_GET(cmd_hca_cap, hcattr, regexp_mmo_qp);
+ attr->mmo_regex_sq_en = MLX5_GET(cmd_hca_cap, hcattr, regexp_mmo_sq);
+ attr->mmo_dma_sq_en = MLX5_GET(cmd_hca_cap, hcattr, dma_mmo_sq);
+ attr->mmo_compress_sq_en = MLX5_GET(cmd_hca_cap, hcattr,
+ compress_mmo_sq);
+ attr->mmo_decompress_sq_en = MLX5_GET(cmd_hca_cap, hcattr,
+ decompress_mmo_sq);
+ attr->mmo_dma_qp_en = MLX5_GET(cmd_hca_cap, hcattr, dma_mmo_qp);
+ attr->mmo_compress_qp_en = MLX5_GET(cmd_hca_cap, hcattr,
+ compress_mmo_qp);
+ attr->mmo_decompress_qp_en = MLX5_GET(cmd_hca_cap, hcattr,
+ decompress_mmo_qp);
attr->compress_min_block_size = MLX5_GET(cmd_hca_cap, hcattr,
compress_min_block_size);
attr->log_max_mmo_dma = MLX5_GET(cmd_hca_cap, hcattr, log_dma_mmo_size);
diff --git a/drivers/common/mlx5/mlx5_devx_cmds.h
b/drivers/common/mlx5/mlx5_devx_cmds.h
index c071629904..b21df0fd9b 100644
--- a/drivers/common/mlx5/mlx5_devx_cmds.h
+++ b/drivers/common/mlx5/mlx5_devx_cmds.h
@@ -173,9 +173,14 @@ struct mlx5_hca_attr {
uint32_t log_max_srq;
uint32_t log_max_srq_sz;
uint32_t rss_ind_tbl_cap;
- uint32_t mmo_dma_en:1;
- uint32_t mmo_compress_en:1;
- uint32_t mmo_decompress_en:1;
+ uint32_t mmo_dma_sq_en:1;
+ uint32_t mmo_compress_sq_en:1;
+ uint32_t mmo_decompress_sq_en:1;
+ uint32_t mmo_dma_qp_en:1;
+ uint32_t mmo_compress_qp_en:1;
+ uint32_t mmo_decompress_qp_en:1;
+ uint32_t mmo_regex_qp_en:1;
+ uint32_t mmo_regex_sq_en:1;
uint32_t compress_min_block_size:4;
uint32_t log_max_mmo_dma:5;
uint32_t log_max_mmo_compress:5;
diff --git a/drivers/common/mlx5/mlx5_prm.h b/drivers/common/mlx5/mlx5_prm.h
index d361bcf90e..ec5f871c61 100644
--- a/drivers/common/mlx5/mlx5_prm.h
+++ b/drivers/common/mlx5/mlx5_prm.h
@@ -1386,10 +1386,10 @@ struct mlx5_ifc_cmd_hca_cap_bits {
u8 rtr2rts_qp_counters_set_id[0x1];
u8 rts2rts_udp_sport[0x1];
u8 rts2rts_lag_tx_port_affinity[0x1];
- u8 dma_mmo[0x1];
+ u8 dma_mmo_sq[0x1];
u8 compress_min_block_size[0x4];
- u8 compress[0x1];
- u8 decompress[0x1];
+ u8 compress_mmo_sq[0x1];
+ u8 decompress_mmo_sq[0x1];
u8 log_max_ra_res_qp[0x6];
u8 end_pad[0x1];
u8 cc_query_allowed[0x1];
@@ -1519,7 +1519,9 @@ struct mlx5_ifc_cmd_hca_cap_bits {
u8 num_lag_ports[0x4];
u8 reserved_at_280[0x10];
u8 max_wqe_sz_sq[0x10];
- u8 reserved_at_2a0[0x10];
+ u8 reserved_at_2a0[0xc];
+ u8 regexp_mmo_sq[0x1];
+ u8 reserved_at_2b0[0x3];
u8 max_wqe_sz_rq[0x10];
u8 max_flow_counter_31_16[0x10];
u8 max_wqe_sz_sq_dc[0x10];
@@ -1632,7 +1634,12 @@ struct mlx5_ifc_cmd_hca_cap_bits {
u8 num_vhca_ports[0x8];
u8 reserved_at_618[0x6];
u8 sw_owner_id[0x1];
- u8 reserved_at_61f[0x1e1];
+ u8 reserved_at_61f[0x109];
+ u8 dma_mmo_qp[0x1];
+ u8 regexp_mmo_qp[0x1];
+ u8 compress_mmo_qp[0x1];
+ u8 decompress_mmo_qp[0x1];
+ u8 reserved_at_624[0xd4];
};
struct mlx5_ifc_qos_cap_bits {
@@ -3244,7 +3251,8 @@ struct mlx5_ifc_create_qp_in_bits {
u8 uid[0x10];
u8 reserved_at_20[0x10];
u8 op_mod[0x10];
- u8 reserved_at_40[0x40];
+ u8 qpc_ext[0x1];
+ u8 reserved_at_41[0x3f];
u8 opt_param_mask[0x20];
u8 reserved_at_a0[0x20];
struct mlx5_ifc_qpc_bits qpc;
diff --git a/drivers/compress/mlx5/mlx5_compress.c
b/drivers/compress/mlx5/mlx5_compress.c
index c5e0a83a8c..1e03030510 100644
--- a/drivers/compress/mlx5/mlx5_compr
[dpdk-dev] [PATCH V5 2/5] common/mlx5: add MMO configuration for the DevX QP
A new configuration MMO was added to QP Context.
If set, MMO WQEs are supported on this QP.
For DMA MMO, supported only when dma_mmo_qp==1.
For REGEXP MMO, supported only when regexp_mmo_qp==1.
For COMPRESS MMO, supported only when compress_mmo_qp==1.
For DECOMPRESS MMO, supported only when decompress_mmo_qp==1.
Add support to DevX interface to set MMO bit.
Signed-off-by: Raja Zidane
Acked-by: Matan Azrad
---
drivers/common/mlx5/mlx5_devx_cmds.c | 7 +++
drivers/common/mlx5/mlx5_devx_cmds.h | 1 +
drivers/common/mlx5/mlx5_prm.h | 28 +++-
3 files changed, 35 insertions(+), 1 deletion(-)
diff --git a/drivers/common/mlx5/mlx5_devx_cmds.c
b/drivers/common/mlx5/mlx5_devx_cmds.c
index 00c78b1288..eefb869b7d 100644
--- a/drivers/common/mlx5/mlx5_devx_cmds.c
+++ b/drivers/common/mlx5/mlx5_devx_cmds.c
@@ -2032,6 +2032,13 @@ mlx5_devx_cmd_create_qp(void *ctx,
MLX5_SET(qpc, qpc, ts_format, attr->ts_format);
MLX5_SET(qpc, qpc, user_index, attr->user_index);
if (attr->uar_index) {
+ if (attr->mmo) {
+ void *qpc_ext_and_pas_list = MLX5_ADDR_OF(create_qp_in,
+ in, qpc_extension_and_pas_list);
+ void *qpc_ext = MLX5_ADDR_OF(qpc_extension_and_pas_list,
+ qpc_ext_and_pas_list, qpc_data_extension);
+ MLX5_SET(qpc_extension, qpc_ext, mmo, 1);
+ }
MLX5_SET(qpc, qpc, pm_state, MLX5_QP_PM_MIGRATED);
MLX5_SET(qpc, qpc, uar_page, attr->uar_index);
if (attr->log_page_size > MLX5_ADAPTER_PAGE_SHIFT)
diff --git a/drivers/common/mlx5/mlx5_devx_cmds.h
b/drivers/common/mlx5/mlx5_devx_cmds.h
index b21df0fd9b..e149f8b4f5 100644
--- a/drivers/common/mlx5/mlx5_devx_cmds.h
+++ b/drivers/common/mlx5/mlx5_devx_cmds.h
@@ -403,6 +403,7 @@ struct mlx5_devx_qp_attr {
uint32_t wq_umem_id;
uint64_t wq_umem_offset;
uint32_t user_index:24;
+ uint32_t mmo:1;
};
struct mlx5_devx_virtio_q_couners_attr {
diff --git a/drivers/common/mlx5/mlx5_prm.h b/drivers/common/mlx5/mlx5_prm.h
index ec5f871c61..54e62aa153 100644
--- a/drivers/common/mlx5/mlx5_prm.h
+++ b/drivers/common/mlx5/mlx5_prm.h
@@ -3243,6 +3243,28 @@ struct mlx5_ifc_create_qp_out_bits {
u8 reserved_at_60[0x20];
};
+struct mlx5_ifc_qpc_extension_bits {
+ u8 reserved_at_0[0x2];
+ u8 mmo[0x1];
+ u8 reserved_at_3[0x5fd];
+};
+
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-Wpedantic"
+#endif
+struct mlx5_ifc_qpc_pas_list_bits {
+ u8 pas[0][0x40];
+};
+
+#ifdef PEDANTIC
+#pragma GCC diagnostic ignored "-Wpedantic"
+#endif
+struct mlx5_ifc_qpc_extension_and_pas_list_bits {
+ struct mlx5_ifc_qpc_extension_bits qpc_data_extension;
+ u8 pas[0][0x40];
+};
+
+
#ifdef PEDANTIC
#pragma GCC diagnostic ignored "-Wpedantic"
#endif
@@ -3260,7 +3282,11 @@ struct mlx5_ifc_create_qp_in_bits {
u8 wq_umem_id[0x20];
u8 wq_umem_valid[0x1];
u8 reserved_at_861[0x1f];
- u8 pas[0][0x40];
+ union {
+ struct mlx5_ifc_qpc_pas_list_bits qpc_pas_list;
+ struct mlx5_ifc_qpc_extension_and_pas_list_bits
+ qpc_extension_and_pas_list;
+ };
};
#ifdef PEDANTIC
#pragma GCC diagnostic error "-Wpedantic"
--
2.17.1
[dpdk-dev] [PATCH V5 4/5] regex/mlx5: refactor HW queue objects
The mlx5 PMD for regex class uses an MMO WQE operated by the GGA engine
in BF devices.
Currently, all the MMO WQEs are managed by the SQ object.
Starting from BF3, the queue of the MMO WQEs should be connected to the
GGA engine using a new configuration, MMO, that will be supported only
in the QP object.
The FW introduced new capabilities to define whether the MMO
configuration should be configured for the GGA queue.
Replace all the GGA queue objects to QP, set MMO configuration according
to the new FW capabilities.
Signed-off-by: Raja Zidane
Acked-by: Matan Azrad
---
drivers/regex/mlx5/mlx5_regex.c | 7 +-
drivers/regex/mlx5/mlx5_regex.h | 16 ++-
drivers/regex/mlx5/mlx5_regex_control.c | 65 +
drivers/regex/mlx5/mlx5_regex_fastpath.c | 170 ---
4 files changed, 133 insertions(+), 125 deletions(-)
diff --git a/drivers/regex/mlx5/mlx5_regex.c b/drivers/regex/mlx5/mlx5_regex.c
index 8866a4d0c6..5aa988be6d 100644
--- a/drivers/regex/mlx5/mlx5_regex.c
+++ b/drivers/regex/mlx5/mlx5_regex.c
@@ -146,7 +146,8 @@ mlx5_regex_dev_probe(struct rte_device *rte_dev)
DRV_LOG(ERR, "Unable to read HCA capabilities.");
rte_errno = ENOTSUP;
goto dev_error;
- } else if (!attr.regex || attr.regexp_num_of_engines == 0) {
+ } else if (((!attr.regex) && (!attr.mmo_regex_sq_en) &&
+ (!attr.mmo_regex_qp_en)) || attr.regexp_num_of_engines == 0) {
DRV_LOG(ERR, "Not enough capabilities to support RegEx, maybe "
"old FW/OFED version?");
rte_errno = ENOTSUP;
@@ -164,7 +165,9 @@ mlx5_regex_dev_probe(struct rte_device *rte_dev)
rte_errno = ENOMEM;
goto dev_error;
}
- priv->sq_ts_format = attr.sq_ts_format;
+ priv->mmo_regex_qp_cap = attr.mmo_regex_qp_en;
+ priv->mmo_regex_sq_cap = attr.mmo_regex_sq_en;
+ priv->qp_ts_format = attr.qp_ts_format;
priv->ctx = ctx;
priv->nb_engines = 2; /* attr.regexp_num_of_engines */
ret = mlx5_devx_regex_register_read(priv->ctx, 0,
diff --git a/drivers/regex/mlx5/mlx5_regex.h b/drivers/regex/mlx5/mlx5_regex.h
index 514f3408f9..2242d250a3 100644
--- a/drivers/regex/mlx5/mlx5_regex.h
+++ b/drivers/regex/mlx5/mlx5_regex.h
@@ -17,12 +17,12 @@
#include "mlx5_rxp.h"
#include "mlx5_regex_utils.h"
-struct mlx5_regex_sq {
+struct mlx5_regex_hw_qp {
uint16_t log_nb_desc; /* Log 2 number of desc for this object. */
- struct mlx5_devx_sq sq_obj; /* The SQ DevX object. */
+ struct mlx5_devx_qp qp_obj; /* The QP DevX object. */
size_t pi, db_pi;
size_t ci;
- uint32_t sqn;
+ uint32_t qpn;
};
struct mlx5_regex_cq {
@@ -34,10 +34,10 @@ struct mlx5_regex_cq {
struct mlx5_regex_qp {
uint32_t flags; /* QP user flags. */
uint32_t nb_desc; /* Total number of desc for this qp. */
- struct mlx5_regex_sq *sqs; /* Pointer to sq array. */
- uint16_t nb_obj; /* Number of sq objects. */
+ struct mlx5_regex_hw_qp *qps; /* Pointer to qp array. */
+ uint16_t nb_obj; /* Number of qp objects. */
struct mlx5_regex_cq cq; /* CQ struct. */
- uint32_t free_sqs;
+ uint32_t free_qps;
struct mlx5_regex_job *jobs;
struct ibv_mr *metadata;
struct ibv_mr *outputs;
@@ -73,8 +73,10 @@ struct mlx5_regex_priv {
/**< Called by memory event callback. */
struct mlx5_mr_share_cache mr_scache; /* Global shared MR cache. */
uint8_t is_bf2; /* The device is BF2 device. */
- uint8_t sq_ts_format; /* Whether SQ supports timestamp formats. */
+ uint8_t qp_ts_format; /* Whether SQ supports timestamp formats. */
uint8_t has_umr; /* The device supports UMR. */
+ uint32_t mmo_regex_qp_cap:1;
+ uint32_t mmo_regex_sq_cap:1;
};
#ifdef HAVE_IBV_FLOW_DV_SUPPORT
diff --git a/drivers/regex/mlx5/mlx5_regex_control.c
b/drivers/regex/mlx5/mlx5_regex_control.c
index 8ce2dabb55..572ecc6d86 100644
--- a/drivers/regex/mlx5/mlx5_regex_control.c
+++ b/drivers/regex/mlx5/mlx5_regex_control.c
@@ -106,12 +106,12 @@ regex_ctrl_create_cq(struct mlx5_regex_priv *priv, struct
mlx5_regex_cq *cq)
* 0 on success, a negative errno value otherwise and rte_errno is set.
*/
static int
-regex_ctrl_destroy_sq(struct mlx5_regex_qp *qp, uint16_t q_ind)
+regex_ctrl_destroy_hw_qp(struct mlx5_regex_qp *qp, uint16_t q_ind)
{
- struct mlx5_regex_sq *sq = &qp->sqs[q_ind];
+ struct mlx5_regex_hw_qp *qp_obj = &qp->qps[q_ind];
- mlx5_devx_sq_destroy(&sq->sq_obj);
- memset(sq, 0, sizeof(*sq));
+ mlx5_devx_qp_destroy(&qp_obj->qp_obj);
+ memset(qp, 0, sizeof(*qp));
return 0;
}
@@ -131,45 +131,44 @@ regex_ctrl_destroy_sq(struct mlx5_regex_qp *qp, uint16_t
q_ind)
* 0 on success, a negative errno value otherwise and rte_errno is set.
*/
static int
-regex_ctrl_create_sq(struct mlx5_re
[dpdk-dev] [PATCH V5 0/5] mlx5: replaced hardware queue object
The mlx5 PMDs for compress and regex classes use an MMO WQE operated by the GGA engine in BF devices. Currently, all the MMO WQEs are managed by the SQ object. Starting from BF3, the queue of the MMO WQEs should be connected to the GGA engine using a new configuration, mmo, that will be supported only in the QP object. The FW introduced new capabilities to define whether the mmo configuration should be configured for the GGA queue. Replace all the GGA queue objects to QP, set mmo configuration according to the new FW capabilities. V2: fix checkpatch errors. V3: rebase. V4: compilation error in commit 2/5. V5: rebase. Raja Zidane (5): common/mlx5: update new MMO HCA capabilities common/mlx5: add MMO configuration for the DevX QP compress/mlx5: refactor queue HW object regex/mlx5: refactor HW queue objects compress/mlx5: allow partial transformations support drivers/common/mlx5/mlx5_devx_cmds.c | 22 ++- drivers/common/mlx5/mlx5_devx_cmds.h | 12 +- drivers/common/mlx5/mlx5_prm.h | 48 ++- drivers/compress/mlx5/mlx5_compress.c| 128 +++-- drivers/regex/mlx5/mlx5_regex.c | 7 +- drivers/regex/mlx5/mlx5_regex.h | 16 ++- drivers/regex/mlx5/mlx5_regex_control.c | 65 + drivers/regex/mlx5/mlx5_regex_fastpath.c | 170 --- 8 files changed, 287 insertions(+), 181 deletions(-) -- 2.17.1
[dpdk-dev] [PATCH V5 5/5] compress/mlx5: allow partial transformations support
Currently compress, decompress and dma are allowed only when all 3
capabilities are on.
A case where the user wants decompress offload, if decompress capability
is on but one of compress, dma is off, is not allowed.
Split compress/decompress/dma support check to allow partial
transformations.
Signed-off-by: Raja Zidane
Acked-by: Matan Azrad
---
drivers/compress/mlx5/mlx5_compress.c | 61 ---
1 file changed, 46 insertions(+), 15 deletions(-)
diff --git a/drivers/compress/mlx5/mlx5_compress.c
b/drivers/compress/mlx5/mlx5_compress.c
index 5c5aa87a18..e94e8fb0c6 100644
--- a/drivers/compress/mlx5/mlx5_compress.c
+++ b/drivers/compress/mlx5/mlx5_compress.c
@@ -291,17 +291,44 @@ mlx5_compress_xform_create(struct rte_compressdev *dev,
struct mlx5_compress_xform *xfrm;
uint32_t size;
- if (xform->type == RTE_COMP_COMPRESS && xform->compress.level ==
- RTE_COMP_LEVEL_NONE) {
- DRV_LOG(ERR, "Non-compressed block is not supported.");
- return -ENOTSUP;
- }
- if ((xform->type == RTE_COMP_COMPRESS && xform->compress.hash_algo !=
-RTE_COMP_HASH_ALGO_NONE) || (xform->type == RTE_COMP_DECOMPRESS &&
- xform->decompress.hash_algo != RTE_COMP_HASH_ALGO_NONE)) {
- DRV_LOG(ERR, "SHA is not supported.");
+ switch (xform->type) {
+ case RTE_COMP_COMPRESS:
+ if (xform->compress.algo == RTE_COMP_ALGO_NULL &&
+ !priv->mmo_dma_qp && !priv->mmo_dma_sq) {
+ DRV_LOG(ERR, "Not enough capabilities to support DMA
operation, maybe old FW/OFED version?");
+ return -ENOTSUP;
+ } else if (!priv->mmo_comp_qp && !priv->mmo_comp_sq) {
+ DRV_LOG(ERR, "Not enough capabilities to support
compress operation, maybe old FW/OFED version?");
+ return -ENOTSUP;
+ }
+ if (xform->compress.level == RTE_COMP_LEVEL_NONE) {
+ DRV_LOG(ERR, "Non-compressed block is not supported.");
+ return -ENOTSUP;
+ }
+ if (xform->compress.hash_algo != RTE_COMP_HASH_ALGO_NONE) {
+ DRV_LOG(ERR, "SHA is not supported.");
+ return -ENOTSUP;
+ }
+ break;
+ case RTE_COMP_DECOMPRESS:
+ if (xform->decompress.algo == RTE_COMP_ALGO_NULL &&
+ !priv->mmo_dma_qp && !priv->mmo_dma_sq) {
+ DRV_LOG(ERR, "Not enough capabilities to support DMA
operation, maybe old FW/OFED version?");
+ return -ENOTSUP;
+ } else if (!priv->mmo_decomp_qp && !priv->mmo_decomp_sq) {
+ DRV_LOG(ERR, "Not enough capabilities to support
decompress operation, maybe old FW/OFED version?");
+ return -ENOTSUP;
+ }
+ if (xform->compress.hash_algo != RTE_COMP_HASH_ALGO_NONE) {
+ DRV_LOG(ERR, "SHA is not supported.");
+ return -ENOTSUP;
+ }
+ break;
+ default:
+ DRV_LOG(ERR, "Xform type should be compress/decompress");
return -ENOTSUP;
}
+
xfrm = rte_zmalloc_socket(__func__, sizeof(*xfrm), 0,
priv->dev_config.socket_id);
if (xfrm == NULL)
@@ -816,12 +843,16 @@ mlx5_compress_dev_probe(struct rte_device *dev)
rte_errno = ENODEV;
return -rte_errno;
}
- if (mlx5_devx_cmd_query_hca_attr(ctx, &att) != 0 ||
- ((att.mmo_compress_sq_en == 0 || att.mmo_decompress_sq_en == 0 ||
- att.mmo_dma_sq_en == 0) && (att.mmo_compress_qp_en == 0 ||
- att.mmo_decompress_qp_en == 0 || att.mmo_dma_qp_en == 0))) {
- DRV_LOG(ERR, "Not enough capabilities to support compress "
- "operations, maybe old FW/OFED version?");
+ if (mlx5_devx_cmd_query_hca_attr(ctx, &att) != 0) {
+ DRV_LOG(ERR, "Failed to query device capabilities");
+ claim_zero(mlx5_glue->close_device(ctx));
+ rte_errno = ENOTSUP;
+ return -ENOTSUP;
+ }
+ if (!att.mmo_decompress_qp_en && !att.mmo_decompress_sq_en
+ && !att.mmo_compress_qp_en && !att.mmo_compress_sq_en
+ && !att.mmo_dma_qp_en && !att.mmo_dma_sq_en) {
+ DRV_LOG(ERR, "Not enough capabilities to support compress
operations, maybe old FW/OFED version?");
claim_zero(mlx5_glue->close_device(ctx));
rte_errno = ENOTSUP;
return -ENOTSUP;
--
2.17.1
[dpdk-dev] [PATCH V5 3/5] compress/mlx5: refactor queue HW object
The mlx5 PMD for compress class uses an MMO WQE operated by the GGA
engine in BF devices.
Currently, all the MMO WQEs are managed by the SQ object.
Starting from BF3, the queue of the MMO WQEs should be connected to the
GGA engine using a new configuration, MMO, that will be supported only
in the QP object.
The FW introduced new capabilities to define whether the MMO
configuration should be configured for the GGA queue.
Replace all the GGA queue objects to QP, set MMO configuration according
to the new FW capabilities.
Signed-off-by: Raja Zidane
Acked-by: Matan Azrad
---
drivers/compress/mlx5/mlx5_compress.c | 73 +++
1 file changed, 42 insertions(+), 31 deletions(-)
diff --git a/drivers/compress/mlx5/mlx5_compress.c
b/drivers/compress/mlx5/mlx5_compress.c
index 1e03030510..5c5aa87a18 100644
--- a/drivers/compress/mlx5/mlx5_compress.c
+++ b/drivers/compress/mlx5/mlx5_compress.c
@@ -40,7 +40,7 @@ struct mlx5_compress_priv {
void *uar;
uint32_t pdn; /* Protection Domain number. */
uint8_t min_block_size;
- uint8_t sq_ts_format; /* Whether SQ supports timestamp formats. */
+ uint8_t qp_ts_format; /* Whether SQ supports timestamp formats. */
/* Minimum huffman block size supported by the device. */
struct ibv_pd *pd;
struct rte_compressdev_config dev_config;
@@ -48,6 +48,13 @@ struct mlx5_compress_priv {
rte_spinlock_t xform_sl;
struct mlx5_mr_share_cache mr_scache; /* Global shared MR cache. */
volatile uint64_t *uar_addr;
+ /* HCA caps*/
+ uint32_t mmo_decomp_sq:1;
+ uint32_t mmo_decomp_qp:1;
+ uint32_t mmo_comp_sq:1;
+ uint32_t mmo_comp_qp:1;
+ uint32_t mmo_dma_sq:1;
+ uint32_t mmo_dma_qp:1;
#ifndef RTE_ARCH_64
rte_spinlock_t uar32_sl;
#endif /* RTE_ARCH_64 */
@@ -61,7 +68,7 @@ struct mlx5_compress_qp {
struct mlx5_mr_ctrl mr_ctrl;
int socket_id;
struct mlx5_devx_cq cq;
- struct mlx5_devx_sq sq;
+ struct mlx5_devx_qp qp;
struct mlx5_pmd_mr opaque_mr;
struct rte_comp_op **ops;
struct mlx5_compress_priv *priv;
@@ -134,8 +141,8 @@ mlx5_compress_qp_release(struct rte_compressdev *dev,
uint16_t qp_id)
{
struct mlx5_compress_qp *qp = dev->data->queue_pairs[qp_id];
- if (qp->sq.sq != NULL)
- mlx5_devx_sq_destroy(&qp->sq);
+ if (qp->qp.qp != NULL)
+ mlx5_devx_qp_destroy(&qp->qp);
if (qp->cq.cq != NULL)
mlx5_devx_cq_destroy(&qp->cq);
if (qp->opaque_mr.obj != NULL) {
@@ -152,12 +159,12 @@ mlx5_compress_qp_release(struct rte_compressdev *dev,
uint16_t qp_id)
}
static void
-mlx5_compress_init_sq(struct mlx5_compress_qp *qp)
+mlx5_compress_init_qp(struct mlx5_compress_qp *qp)
{
volatile struct mlx5_gga_wqe *restrict wqe =
- (volatile struct mlx5_gga_wqe *)qp->sq.wqes;
+ (volatile struct mlx5_gga_wqe *)qp->qp.wqes;
volatile struct mlx5_gga_compress_opaque *opaq = qp->opaque_mr.addr;
- const uint32_t sq_ds = rte_cpu_to_be_32((qp->sq.sq->id << 8) | 4u);
+ const uint32_t sq_ds = rte_cpu_to_be_32((qp->qp.qp->id << 8) | 4u);
const uint32_t flags = RTE_BE32(MLX5_COMP_ALWAYS <<
MLX5_COMP_MODE_OFFSET);
const uint32_t opaq_lkey = rte_cpu_to_be_32(qp->opaque_mr.lkey);
@@ -182,15 +189,10 @@ mlx5_compress_qp_setup(struct rte_compressdev *dev,
uint16_t qp_id,
struct mlx5_devx_cq_attr cq_attr = {
.uar_page_id = mlx5_os_get_devx_uar_page_id(priv->uar),
};
- struct mlx5_devx_create_sq_attr sq_attr = {
+ struct mlx5_devx_qp_attr qp_attr = {
+ .pd = priv->pdn,
+ .uar_index = mlx5_os_get_devx_uar_page_id(priv->uar),
.user_index = qp_id,
- .wq_attr = (struct mlx5_devx_wq_attr){
- .pd = priv->pdn,
- .uar_page = mlx5_os_get_devx_uar_page_id(priv->uar),
- },
- };
- struct mlx5_devx_modify_sq_attr modify_attr = {
- .state = MLX5_SQC_STATE_RDY,
};
uint32_t log_ops_n = rte_log2_u32(max_inflight_ops);
uint32_t alloc_size = sizeof(*qp);
@@ -242,24 +244,26 @@ mlx5_compress_qp_setup(struct rte_compressdev *dev,
uint16_t qp_id,
DRV_LOG(ERR, "Failed to create CQ.");
goto err;
}
- sq_attr.cqn = qp->cq.cq->id;
- sq_attr.ts_format = mlx5_ts_format_conv(priv->sq_ts_format);
- ret = mlx5_devx_sq_create(priv->ctx, &qp->sq, log_ops_n, &sq_attr,
+ qp_attr.cqn = qp->cq.cq->id;
+ qp_attr.ts_format = mlx5_ts_format_conv(priv->qp_ts_format);
+ qp_attr.rq_size = 0;
+ qp_attr.sq_size = RTE_BIT32(log_ops_n);
+ qp_attr.mmo = priv->mmo_decomp_qp && priv->mmo_comp_qp
+ && priv->mmo_dma_qp;
+
Re: [dpdk-dev] [PATCH v3 01/10] drivers/crypto: introduce IPsec-mb framework
On 29/09/2021 17:30, Ciara Power wrote: > From: Fan Zhang > > This patch introduces the new framework to share common code between > the SW crypto PMDs that depend on the intel-ipsec-mb library. > This change helps to reduce future effort on the code maintenance and > feature updates. > > The PMDs that will be added to this framework in subsequent patches are: > - AESNI MB > - AESNI GCM > - KASUMI > - SNOW3G > - ZUC > > The use of these PMDs will not change, they will still be supported for > x86, and will use the same EAL args as before. > > The minimum required version for the intel-ipsec-mb library is now v1.0. > > Signed-off-by: Fan Zhang > Signed-off-by: Ciara Power > > --- > v3: > - Updated intel-ipsec-mb macros. > - Added use of auto init function for IMB_MGR. > - Added detail to commit log. > v2: > - Added qp NULL check in get stats function. > - Added maintainers file entry. > - Replaced strlcpy with rte_strlcpy. > --- Acked-by: Ray Kinsella (for the series).
Re: [dpdk-dev] [PATCH 02/10] common/sfc_efx/base: add API to set RECIRC ID in outer rules
On 29/09/2021 21:57, Ivan Malov wrote: > When an outer rule is hit, it can pass recirculation ID down > to action rule lookup, and action rules can match on this ID > instead of matching on the outer rule allocation handle. > By default, recirculation ID is assumed to be zero. > > Add an API to set recirculation ID in outer rules. > > Signed-off-by: Ivan Malov > Reviewed-by: Andrew Rybchenko > --- > drivers/common/sfc_efx/base/efx.h | 9 + > drivers/common/sfc_efx/base/efx_impl.h | 1 + > drivers/common/sfc_efx/base/efx_mae.c | 24 > drivers/common/sfc_efx/version.map | 1 + > 4 files changed, 35 insertions(+) > Acked-by: Ray Kinsella
Re: [dpdk-dev] [PATCH v2 0/5] kvargs: promote or remove experimental api
On 30/09/2021 09:25, David Marchand wrote: > On Wed, Sep 29, 2021 at 11:40 PM Olivier Matz wrote: >> >> This patchset promotes 2 functions rte_kvargs_parse_delim() and >> rte_kvargs_get() as stable. >> >> It also replaces rte_kvargs_strcmp() by a new one >> rte_kvargs_get_with_value(), which is easier to use. >> >> v2 >> * remove rte_kvargs_strcmp from version.map >> >> Olivier Matz (5): >> kvargs: promote delimited parsing as stable >> kvargs: promote get from key as stable >> kvargs: new function to get from key and value >> kvargs: remove experimental function to compare string >> kvargs: fix comments style > > Thanks, for the series, > Reviewed-by: David Marchand > Acked-by: Ray Kinsella
Re: [dpdk-dev] [PATCH v1] ethdev: introduce shared Rx queue
> On Wed, 2021-09-29 at 10:20 +, Ananyev, Konstantin wrote: > > > > > > > > > > > > > > > > > > In current DPDK framework, each RX > > > > > > > > > > > > > > > > > > queue > > > > > > > > > > > > > > > > > > is > > > > > > > > > > > > > > > > > > pre-loaded with mbufs > > > > > > > > > > > > > > > > > > for incoming packets. When number of > > > > > > > > > > > > > > > > > > representors scale out in a > > > > > > > > > > > > > > > > > > switch domain, the memory consumption > > > > > > > > > > > > > > > > > > became > > > > > > > > > > > > > > > > > > significant. Most > > > > > > > > > > > > > > > > > > important, polling all ports leads to > > > > > > > > > > > > > > > > > > high > > > > > > > > > > > > > > > > > > cache miss, high > > > > > > > > > > > > > > > > > > latency and low throughput. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > This patch introduces shared RX > > > > > > > > > > > > > > > > > > queue. > > > > > > > > > > > > > > > > > > Ports > > > > > > > > > > > > > > > > > > with same > > > > > > > > > > > > > > > > > > configuration in a switch domain > > > > > > > > > > > > > > > > > > could > > > > > > > > > > > > > > > > > > share > > > > > > > > > > > > > > > > > > RX queue set by specifying sharing > > > > > > > > > > > > > > > > > > group. > > > > > > > > > > > > > > > > > > Polling any queue using same shared > > > > > > > > > > > > > > > > > > RX > > > > > > > > > > > > > > > > > > queue > > > > > > > > > > > > > > > > > > receives packets from > > > > > > > > > > > > > > > > > > all member ports. Source port is > > > > > > > > > > > > > > > > > > identified > > > > > > > > > > > > > > > > > > by mbuf->port. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Port queue number in a shared group > > > > > > > > > > > > > > > > > > should be > > > > > > > > > > > > > > > > > > identical. Queue > > > > > > > > > > > > > > > > > > index is > > > > > > > > > > > > > > > > > > 1:1 mapped in shared group. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Share RX queue is supposed to be > > > > > > > > > > > > > > > > > > polled > > > > > > > > > > > > > > > > > > on > > > > > > > > > > > > > > > > > > same thread. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Multiple groups is supported by group > > > > > > > > > > > > > > > > > > ID. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Is this offload specific to the > > > > > > > > > > > > > > > > > representor? If > > > > > > > > > > > > > > > > > so can this name be changed > > > > > > > > > > > > > > > > > specifically to > > > > > > > > > > > > > > > > > representor? > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Yes, PF and representor in switch domain > > > > > > > > > > > > > > > > could > > > > > > > > > > > > > > > > take advantage. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > If it is for a generic case, how the > > > > > > > > > > > > > > > > > flow > > > > > > > > > > > > > > > > > ordering will be maintained? > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Not quite sure that I understood your > > > > > > > > > > > > > > > > question. > > > > > > > > > > > > > > > > The control path of is > > > > > > > > > > > > > > > > almost same as before, PF and representor > > > > > > > > > > > > > > > > port > > > > > > > > > > > > > > > > still needed, rte flows not impacted. > > > > > > > > > > > > > > > > Queues still needed for each member port, > > > > > > > > > > > > > > > > descriptors(mbuf) will be > > > > > > > > > > > > > > > > supplied from shared Rx queue in my PMD > > > > > > > > > > > > > > > > implementation. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > My question was if create a generic > > > > > > > > > > > > > > > RTE_ETH_RX_OFFLOAD_SHARED_RXQ offload, > > > > > > > > > > > > > > > multiple > > > > > > > > > > > > > > > ethdev receive queues land into > > > > > > > the same > > > > > > > > > > > > > > > receive queue, In that case, how the flow > > > > > > > > > > > > > > > order > > > > > > > > > > > > > > > is > > > > > > > > > > > > > > > maintained for respective receive queues. > > > > > > > > > > > > > > > > > > > > > > > > > > > > I guess the question is testpmd forward > > > > > > > > > > > > > > stream? > > > > > > > > > > > > > > The > > > > > > > > > > > > > > forwarding logic has to be changed slightly > > > > > > > > > > > > > > in > > > > > > > > > > > > > > case > > > > > > > > > > > > > > of shared rxq. > > > > > > > > > > > > > > basically for each packet in rx_burst result, > > > > > > > > > > > > > > lookup > > > > > > > > > > > > > > source stream according to mbuf->port, > > > > > > > > > > > > > > forwarding > > > > > > > > > > > > > > to > > > > > > > > > > > > > > target fs. > > > > > > > > > > > > > > Packets from same source port could be > > > > > > > > > > > > > > grouped as > > > > > > > > > > > > > >
Re: [dpdk-dev] [PATCH v3 01/27] common/cnxk: update policer MBOX APIs and HW definitions
On 30/09/2021 10:08, sk...@marvell.com wrote: > From: Sunil Kumar Kori > > To support ingress policer on CN10K, MBOX interfaces and HW > definitions are synced. > > Signed-off-by: Sunil Kumar Kori > --- > v3: > - Rebase support on latest DPDK > - Handled multilevel chaining for tree hierarchy > - Fix naming convention > v2: > - Rebase support on latest DPDK > - Handled multilevel chaining for linear hierarchy > - Review comments incorporated > > drivers/common/cnxk/hw/nix.h | 13 ++--- > drivers/common/cnxk/roc_mbox.h | 34 +- > 2 files changed, 43 insertions(+), 4 deletions(-) > Acked-by: Ray Kinsella (for the series).
Re: [dpdk-dev] [PATCH 2/2] security: build on Windows
> Subject: Re: [PATCH 2/2] security: build on Windows
>
> External email: Use caution opening links or attachments
>
>
> On 9/15/2021 9:26 AM, Tal Shnaiderman wrote:
> > Build the security library on Windows.
> >
> > Remove unneeded export from version file.
> >
> > Signed-off-by: Tal Shnaiderman
> > ---
> > Depends-on: patch-98796 ("cryptodev: build on Windows")
> > ---
> > lib/meson.build | 1 +
> > lib/security/version.map | 1 -
> > 2 files changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/lib/meson.build b/lib/meson.build index
> > 5f3b1cf365..e5f0094a82 100644
> > --- a/lib/meson.build
> > +++ b/lib/meson.build
> > @@ -86,6 +86,7 @@ if is_windows
> > 'gso',
> > 'latencystats',
> > 'pdump',
> > +'security',
> > ] # only supported libraries for windows
> > endif
> >
> > diff --git a/lib/security/version.map b/lib/security/version.map index
> > c44c7f5f60..6067051552 100644
> > --- a/lib/security/version.map
> > +++ b/lib/security/version.map
> > @@ -1,7 +1,6 @@
> > DPDK_22 {
> > global:
> >
> > - rte_security_attach_session;
> > rte_security_capabilities_get;
> > rte_security_capability_get;
> > rte_security_session_create;
>
> Getting linking errors using clang for 'rte_security_get_userdata' and
> 'rte_security_set_pkt_metadata' as below:
>
> FAILED: lib/rte_security-22.dll
> "clang" -Wl,/MACHINE:X64 -Wl,/OUT:lib/rte_security-22.dll
> lib/librte_security.a.p/security_rte_security.c.obj "-Wl,/nologo"
> "-Wl,/release" "-Wl,/nologo" "-Wl,/OPT:REF" "-Wl,/DLL"
> "-Wl,/IMPLIB:lib\rte_security.lib" "lib\rte_eal.lib"
> "lib\rte_kvargs.lib" "lib\rte_mempool.lib" "lib\rte_ring.lib"
> "lib\rte_cryptodev.lib" "lib\rte_mbuf.lib" "lib\rte_rcu.lib"
> "lib\rte_net.lib"
> "-Wl,/def:C:\dpdk\ixgbe\dpdk\build\lib\rte_security_exports.def"
> "-ldbghelp" "-lsetupapi" "-lws2_32" "-lmincore" "-lkernel32" "-luser32"
> "-lgdi32" "-lwinspool" "-lshell32" "-lole32" "-loleaut32" "-luuid"
> "-lcomdlg32" "-ladvapi32"
> rte_security_exports.def : error LNK2001: unresolved external symbol
> rte_security_get_userdata rte_security_exports.def : error LNK2001:
> unresolved external symbol rte_security_set_pkt_metadata
> lib\rte_security.lib : fatal error LNK1120: 2 unresolved externals
> clang: error: linker command failed with exit code 1120 (use -v to see
> invocation)
>
> Should it be removed from version file?
Yes, those 2 functions were changed in the patch below after my v1 was sent and
no longer need export.
https://git.dpdk.org/dpdk/commit/?id=d08dcd28c3b245468a9859b7b9a288247dfc95f1
I'll remove them from the version file in v2 as suggested, thanks.
Re: [dpdk-dev] [RFC V2] ethdev: fix issue that dev close in PMD calls twice
在 2021/9/28 15:19, Singh, Aman Deep 写道: On 9/22/2021 9:01 AM, Huisong Li wrote: 在 2021/9/20 22:07, Ferruh Yigit 写道: On 8/25/2021 10:53 AM, Huisong Li wrote: 在 2021/8/24 22:42, Ferruh Yigit 写道: On 8/19/2021 4:45 AM, Huisong Li wrote: 在 2021/8/18 19:24, Ferruh Yigit 写道: On 8/13/2021 9:16 AM, Huisong Li wrote: 在 2021/8/13 14:12, Thomas Monjalon 写道: 13/08/2021 04:11, Huisong Li: Hi, all This patch can enhance the security of device uninstallation to eliminate dependency on user usage methods. Can you check this patch? 在 2021/8/3 10:30, Huisong Li 写道: Ethernet devices in DPDK can be released by rte_eth_dev_close() and rte_dev_remove(). These APIs both call xxx_dev_close() in PMD layer to uninstall hardware. However, the two APIs do not have explicit invocation restrictions. In other words, at the ethdev layer, it is possible to call rte_eth_dev_close() before calling rte_dev_remove() or rte_eal_hotplug_remove(). In such a bad scenario, It is not a bad scenario. If there is no more port for the device after calling close, the device should be removed automatically. Keep in mind "close" is for one port, "remove" is for the entire device which can have more than one port. I know. dev_close() is for removing an eth device. And rte_dev_remove() can be used for removing the rte device and all its eth devices belonging to the rte device. In rte_dev_remove(), "remove" is executed in primary or one of secondary, all eth devices having same pci address will be closed and removed. the primary process may be fine, but it may cause that xxx_dev_close() in the PMD layer will be called twice in the secondary process. So this patch fixes it. If a port is closed in primary, it should be the same in secondary. + /* + * The eth_dev->data->name doesn't be cleared by the secondary process, + * so above "eth_dev" isn't NULL after rte_eth_dev_close() called. This assumption is not clear. All should be closed together. However, dev_close() does not have the feature similar to rte_dev_remove(). Namely, it is not guaranteed that all eth devices are closed together in ethdev layer. It depends on app or user. If the app does not close together, the operation of repeatedly uninstalling an eth device in the secondary process will be triggered when dev_close() is first called by one secondary process, and then rte_dev_remove() is called. So I think it should be avoided. First of all, I am not sure about calling 'rte_eth_dev_close()' or 'rte_dev_remove()' from the secondary process. There are explicit checks in various locations to prevent clearing resources completely from secondary process. There's no denying that. Generally, hardware resources of eth device and shared data of the primary and secondary process are cleared by primary, which are controled by ethdev layer or PMD layer. But there may be some private data or resources of each process (primary or secondary ), such as mp action registered by rte_mp_action_register() or others. For these resources, the secondary process still needs to clear. Namely, both primary and secondary processes need to prevent repeated offloading of resources. Calling 'rte_eth_dev_close()' or 'rte_dev_remove()' by secondary is technically can be done but application needs to be extra cautious and should take extra measures and synchronization to make it work. Regular use-case is secondary processes do the packet processing and all control commands run by primary. You are right. We have a consensus that 'rte_eth_dev_close()' or 'rte_dev_remove()' can be called by primary and secondary processes. But DPDK framework cannot assume user behavior.😁 We need to make it more secure and reliable for both primary and secondary processes. In primary, if you call 'rte_eth_dev_close()' it will clear all ethdev resources and further 'rte_dev_remove()' call will detect missing ethdev resources and won't try to clear them again. In secondary, if you call 'rte_eth_dev_close()', it WON'T clear all resources and further 'rte_dev_remove()' call (either from primary or secondary) will try to clean ethdev resources again. You are trying to prevent this retry in remove happening for secondary process. Right. However, if secondary process in PMD layer has its own private resources to be cleared, it still need to do it by calling 'rte_eth_dev_close()' or 'rte_dev_remove()'. In secondary it won't free ethdev resources anyway if you let it continue, but I guess here you are trying to prevent the PMD dev_close() called again. Why? Is it just for optimization or does it cause unexpected behavior in the PMD? Overall, to free resources you need to do the 'rte_eth_dev_close()' or 'rte_dev_remove()' in the primary anyway. So instead of this workaround, I would suggest making PMD dev_close() safe to be called multiple times (if this is the problem.) In conclusion, primary and secondary processes in PMD layer
Re: [dpdk-dev] [dpdk-stable] [PATCH] net/iavf: fix multi-process shared data
On 9/30/2021 10:11 AM, Yu, DapengX wrote:
>
>
>> -Original Message-
>> From: Yigit, Ferruh
>> Sent: Thursday, September 30, 2021 12:28 AM
>> To: Yu, DapengX ; Richardson, Bruce
>> ; Ananyev, Konstantin
>> ; Wu, Jingjing ;
>> Xing, Beilei
>> Cc: dev@dpdk.org; sta...@dpdk.org
>> Subject: Re: [dpdk-stable] [PATCH] net/iavf: fix multi-process shared data
>>
>> On 9/28/2021 4:37 AM, dapengx...@intel.com wrote:
>>> From: Dapeng Yu
>>>
>>> When the iavf_adapter instance is not initialized completedly in the
>>> primary process, the secondary process accesses its "rte_eth_dev"
>>> member, it causes secondary process crash.
>>>
>>> This patch replaces adapter->eth_dev with rte_eth_devices[port_id] in
>>> the data paths where rte_eth_dev instance is accessed.
>>>
>>> Fixes: f978c1c9b3b5 ("net/iavf: add RSS hash parsing in AVX path")
>>> Fixes: 9c9aa0040344 ("net/iavf: add offload path for Rx AVX512 flex
>>> descriptor")
>>> Fixes: 63660ea3ee0b ("net/iavf: add RSS hash parsing in SSE path")
>>> Cc: sta...@dpdk.org
>>>
>>> Signed-off-by: Dapeng Yu
>>> ---
>>> drivers/net/iavf/iavf_rxtx_vec_avx2.c | 5 +++--
>>> drivers/net/iavf/iavf_rxtx_vec_avx512.c | 5 +++--
>>> drivers/net/iavf/iavf_rxtx_vec_sse.c| 3 ++-
>>> 3 files changed, 8 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/drivers/net/iavf/iavf_rxtx_vec_avx2.c
>>> b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
>>> index 475070e036..59b086ade5 100644
>>> --- a/drivers/net/iavf/iavf_rxtx_vec_avx2.c
>>> +++ b/drivers/net/iavf/iavf_rxtx_vec_avx2.c
>>> @@ -525,6 +525,7 @@ _iavf_recv_raw_pkts_vec_avx2_flex_rxd(struct
>>> iavf_rx_queue *rxq, #define IAVF_DESCS_PER_LOOP_AVX 8
>>>
>>> const uint32_t *type_table = rxq->vsi->adapter->ptype_tbl;
>>> + struct rte_eth_dev *dev = &rte_eth_devices[rxq->port_id];
>>>
>>
>> It is not good idea to access global variable directly from the driver.
> In "lib/ethdev/rte_ethdev.h", the global variable rte_eth_devices is used.
> So I think use it in a PMD should be also acceptable since it is just read.
It is expected for ehtdev APIs to access the array. Application knows only
port_id, ethdev layer converts this port_id to device struct by accessing the
global array, and drivers should be able to operate only with its device.
> rte_eth_rx_burst(uint16_t port_id, uint16_t queue_id,
>struct rte_mbuf **rx_pkts, const uint16_t nb_pkts)
> {
> struct rte_eth_dev *dev = &rte_eth_devices[port_id];
>
>>
>> The problem definition is correct, eth_dev is unique per process, so it can't
>> be saved to a shared struct.
>>
>> But here I assume real intention is to be able to access PMD specific data
>> from queue struct, for this what about storing 'rte_eth_dev_data' in the
>> 'iavf_rx_queue', this should sove the problem without accessing the global
>> variable.
>
> The intention is to read the offload properties of device configuration, so
> it not
> queue specific or PMD specific. It is already in public data structure.
> If it is stored in 'iavf_rx_queue' again, the data will be duplicate.
>
I can see the intention. This is more design concern, you can access to that
data structure doesn't mean you should.
You will just store the pointer of the 'data', is it duplication?
Re: [dpdk-dev] [dpdk-stable] [PATCH v2] net/virtio: revert forcing IOVA as VA mode for virtio-user
On 9/30/2021 9:24 AM, David Marchand wrote:
> On Thu, Sep 30, 2021 at 10:13 AM Maxime Coquelin
> wrote:
>>
>> This patch removes the simplification in Virtio descriptors
>> handling, where their buffer addresses are IOVAs for Virtio
>> PCI devices, and VA-only for Virtio-user devices, which
>> added a requirement on Virtio-user that it only supported
>> IOVA as VA.
>>
>> This change introduced a regression for applications using
>> Virtio-user and other physical PMDs that require IOVA as PA
>> because they don't use an IOMMU.
>>
>> This patch reverts to the old behaviour, but needed to be
>> reworked because of the refactoring that happened in v21.02.
>>
>> Fixes: 17043a2909bb ("net/virtio: force IOVA as VA mode for virtio-user")
>> Cc: sta...@dpdk.org
>>
>> Reported-by: Olivier Matz
>> Signed-off-by: Maxime Coquelin
>
> I think you can keep:
> Tested-by: Olivier Matz
>
> Reviewed-by: David Marchand
>
Applied to dpdk-next-net/main, thanks.
Re: [dpdk-dev] [RFC V2] ethdev: fix issue that dev close in PMD calls twice
On 9/30/2021 11:54 AM, Huisong Li wrote: > > 在 2021/9/28 15:19, Singh, Aman Deep 写道: >> >> On 9/22/2021 9:01 AM, Huisong Li wrote: >>> >>> 在 2021/9/20 22:07, Ferruh Yigit 写道: On 8/25/2021 10:53 AM, Huisong Li wrote: > 在 2021/8/24 22:42, Ferruh Yigit 写道: >> On 8/19/2021 4:45 AM, Huisong Li wrote: >>> 在 2021/8/18 19:24, Ferruh Yigit 写道: On 8/13/2021 9:16 AM, Huisong Li wrote: > 在 2021/8/13 14:12, Thomas Monjalon 写道: >> 13/08/2021 04:11, Huisong Li: >>> Hi, all >>> >>> This patch can enhance the security of device uninstallation to >>> eliminate dependency on user usage methods. >>> >>> Can you check this patch? >>> >>> >>> 在 2021/8/3 10:30, Huisong Li 写道: Ethernet devices in DPDK can be released by rte_eth_dev_close() and rte_dev_remove(). These APIs both call xxx_dev_close() in PMD layer to uninstall hardware. However, the two APIs do not have explicit invocation restrictions. In other words, at the ethdev layer, it is possible to call rte_eth_dev_close() before calling rte_dev_remove() or rte_eal_hotplug_remove(). In such a bad scenario, >> It is not a bad scenario. >> If there is no more port for the device after calling close, >> the device should be removed automatically. >> Keep in mind "close" is for one port, "remove" is for the entire >> device >> which can have more than one port. > I know. > > dev_close() is for removing an eth device. And rte_dev_remove() can be > used > > for removing the rte device and all its eth devices belonging to the > rte > device. > > In rte_dev_remove(), "remove" is executed in primary or one of > secondary, > > all eth devices having same pci address will be closed and removed. > the primary process may be fine, but it may cause that xxx_dev_close() in the PMD layer will be called twice in the secondary process. So this patch fixes it. >> If a port is closed in primary, it should be the same in secondary. >> >> + /* + * The eth_dev->data->name doesn't be cleared by the secondary process, + * so above "eth_dev" isn't NULL after rte_eth_dev_close() called. >> This assumption is not clear. All should be closed together. > However, dev_close() does not have the feature similar to > rte_dev_remove(). > > Namely, it is not guaranteed that all eth devices are closed together > in > ethdev > layer. It depends on app or user. > > If the app does not close together, the operation of repeatedly > uninstalling an > eth device in the secondary process > > will be triggered when dev_close() is first called by one secondary > process, and > then rte_dev_remove() is called. > > So I think it should be avoided. First of all, I am not sure about calling 'rte_eth_dev_close()' or 'rte_dev_remove()' from the secondary process. There are explicit checks in various locations to prevent clearing resources completely from secondary process. >>> There's no denying that. >>> >>> Generally, hardware resources of eth device and shared data of the >>> primary and >>> secondary process >>> >>> are cleared by primary, which are controled by ethdev layer or PMD >>> layer. >>> >>> But there may be some private data or resources of each process >>> (primary or >>> secondary ), such as mp action >>> >>> registered by rte_mp_action_register() or others. For these resources, >>> the >>> secondary process still needs to clear. >>> >>> Namely, both primary and secondary processes need to prevent repeated >>> offloading >>> of resources. >>> Calling 'rte_eth_dev_close()' or 'rte_dev_remove()' by secondary is technically can be done but application needs to be extra cautious and should take extra measures and synchronization to make it work. Regular use-case is secondary processes do the packet processing and all control commands run by primary. >>> You are right. We have a consensus that 'rte_eth_dev_close()' or >>> 'rte_dev_remove()' >>> >>> can be called by primary and secondary processes. >>> >>> But DPDK framework cannot assume user behavior.😁 >>> >>> We need to make it more secure and reliable for both primary and >>> secondary >>> processes. >>> In primary, if you call '
Re: [dpdk-dev] [PATCH v5] ethdev: fix representor port ID search by name
On 9/29/21 2:13 PM, Singh, Aman Deep wrote: > > On 9/13/2021 4:56 PM, Andrew Rybchenko wrote: >> From: Viacheslav Galaktionov >> >> Getting a list of representors from a representor does not make sense. >> Instead, a parent device should be used. >> >> To this end, extend the rte_eth_dev_data structure to include the port ID >> of the backing device for representors. >> >> Signed-off-by: Viacheslav Galaktionov >> >> Signed-off-by: Andrew Rybchenko >> Acked-by: Haiyue Wang >> Acked-by: Beilei Xing >> --- >> The new field is added into the hole in rte_eth_dev_data structure. >> The patch does not change ABI, but extra care is required since ABI >> check is disabled for the structure because of the libabigail bug [1]. >> It should not be a problem anyway since 21.11 is a ABI breaking release. >> >> Potentially it is bad for out-of-tree drivers which implement >> representors but do not fill in a new parert_port_id field in >> rte_eth_dev_data structure. Get ID by name will not work. > Did we change name of new field from parert_port_id to backer_port_id. Yes, see v5 changelog below. It is done to address review notes from Ferruh on v4. >> >> mlx5 changes should be reviwed by maintainers very carefully, since >> we are not sure if we patch it correctly. >> >> [1] https://sourceware.org/bugzilla/show_bug.cgi?id=28060 >> >> v5: >> - try to improve name: backer_port_id instead of parent_port_id >> - init new field to RTE_MAX_ETHPORTS on allocation to avoid >> zero port usage by default >> >> v4: >> - apply mlx5 review notes: remove fallback from generic ethdev >> code and add fallback to mlx5 code to handle legacy usecase >> >> v3: >> - fix mlx5 build breakage >> >> v2: >> - fix mlx5 review notes >> - try device port ID first before parent in order to address >> backward compatibility issue [snip]
Re: [dpdk-dev] [dpdk-stable] [PATCH v5 2/2] ethdev: fix docs of drivers callbacks getting xstats by IDs
On 9/29/2021 12:54 PM, Andrew Rybchenko wrote:
> On 9/29/21 11:44 AM, Ferruh Yigit wrote:
>> On 9/28/2021 5:53 PM, Andrew Rybchenko wrote:
>>> On 9/28/21 7:50 PM, Ferruh Yigit wrote:
On 9/28/2021 1:05 PM, Andrew Rybchenko wrote:
> From: Ivan Ilchenko
>
> Update xstats by IDs callbacks documentation in accordance with
> ethdev usage of these callbacks. Document valid combinations of
> input arguments to make driver implementation simpler.
>
> Fixes: 79c913a42f0 ("ethdev: retrieve xstats by ID")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Ivan Ilchenko
> Signed-off-by: Andrew Rybchenko
> Reviewed-by: Andy Moreton
> ---
> lib/ethdev/ethdev_driver.h | 42 --
> 1 file changed, 40 insertions(+), 2 deletions(-)
>
> diff --git a/lib/ethdev/ethdev_driver.h b/lib/ethdev/ethdev_driver.h
> index 40e474aa7e..c89eefcc42 100644
> --- a/lib/ethdev/ethdev_driver.h
> +++ b/lib/ethdev/ethdev_driver.h
> @@ -187,11 +187,28 @@ typedef int (*eth_xstats_get_t)(struct rte_eth_dev
> *dev,
> struct rte_eth_xstat *stats, unsigned int n);
> /**< @internal Get extended stats of an Ethernet device. */
>
> +/**
> + * @internal
> + * Get extended stats of an Ethernet device.
> + *
> + * @param dev
> + * ethdev handle of port.
> + * @param ids
> + * IDs array to retrieve specific statistics. Must not be NULL.
> + * @param values
> + * A pointer to a table to be filled with device statistics values.
> + * Must not be NULL.
> + * @param n
> + * Element count in @p ids and @p values.
> + *
> + * @return
> + * - A number of filled in stats.
> + * - A negative value on error.
> + */
> typedef int (*eth_xstats_get_by_id_t)(struct rte_eth_dev *dev,
> const uint64_t *ids,
> uint64_t *values,
> unsigned int n);
> -/**< @internal Get extended stats of an Ethernet device. */
>
> /**
> * @internal
> @@ -218,10 +235,31 @@ typedef int (*eth_xstats_get_names_t)(struct
> rte_eth_dev *dev,
> struct rte_eth_xstat_name *xstats_names, unsigned int size);
> /**< @internal Get names of extended stats of an Ethernet device. */
>
> +/**
> + * @internal
> + * Get names of extended stats of an Ethernet device.
> + * For name count, set @p xstats_names and @p ids to NULL.
Why limiting this behavior to 'xstats_get_names_by_id'?
Internally 'xstats_get_names_by_id' is used to get the count, but I think
this
is not intentionally selected, just one of the xstats_*_by_id dev_ops used.
I think it is confusing to have this support for one of the '_by_id'
dev_ops but
not for other. Why not require both to support returning 'count'?
>>>
>>> Simply because it is dead code. There is no point to require
>>> from driver to have dead code.
>>>
>>
>> Let me step back a little, both ethdev APIs can be used to return xstats
>> count
>> by providing 'values/names' & 'ids' pointers as NULL and 'size' as 0:
>> 'rte_eth_xstats_get_names_by_id()'
>> 'rte_eth_xstats_get_by_id()'
>>
>> But internally both APIs use 'xstats_get_names_by_id' dev_ops to get the
>> count,
>> as said above I believe this selection is done unintentionally.
>>
>>
>> I am for below two options:
>>
>> a) Internally use 'xstats_get_names' || 'xstats_get' dev_ops to get the
>> xstats
>> count, and doesn't support getting xstats count for both '_by_id' dev_ops,
>> this
>> simplifies driver code. As far as I remember I suggested this before, still I
>> prefer this one.
>>
>> b) If we will support getting xstats count from '_by_id' dev_ops, I think
>> both
>> should support it, to not make it more complex to figure out which one
>> support
>> what. As sample both 'xstats_get_names' and 'xstats_get' supports getting
>> xstats
>> count, not just one.
>>
>
> In (b) do you suggest to change ethdev to use xstats_get_by_id
> driver op if users asks for a number of xstats using
> rte_eth_xstats_get_by_id()? It will complicate ethdev and
> will complicate drivers. Just for the symmetry?
>
Not just for symmetry, but simplify the logic. Both dev_ops are very similar and
they are having different behavior with same parameters is confusing.
If you want to simplify the drivers why not go with option (a)? Option (a) also
doesn't change external API, and has only minor change in the ethdev layer.
> The patch does not change external API, does not change etcdev
> bahaviour. It just clarify requirements on driver ops to
> allow to check less in all drivers and support less cases
> in all drivers.
>
It is not clarifying the requirements of dev_ops, but changing it. Previously
there was nothing saying only one of the '_by_id' dev_ops should support
returning element count.
Re: [dpdk-dev] [PATCH] doc: remove ethdev deprecation note for flag name
On 9/30/2021 8:47 AM, Aman Singh wrote: > Proposed name change of offload flag PKT_RX_EIP_CKSUM_BAD > to PKT_RX_OUTER_IP_CKSUM_BAD has already been done in the > code as per the deprecation note. > > Signed-off-by: Aman Singh > --- > doc/guides/rel_notes/deprecation.rst | 5 - > lib/mbuf/rte_mbuf_core.h | 7 --- Acked-by: Ferruh Yigit But I would do the patch as mbuf patch, instead of doc patch. If you send a new version with this change please feel free to keep the ack.
[dpdk-dev] [PATCH v1] eventdev/rx-adapter: segfault in queue conf get
rte_event_eth_rx_adapter_queue_conf_get() segfaults if called
without queue added to the Rx adapter.
Added check to no queues in Rx adapter and error out on being
called with no queue in Rx adapter.
Added test case to call queue conf get without queues in
Rx adapter.
Signed-off-by: Ganapati Kundapura
diff --git a/app/test/test_event_eth_rx_adapter.c
b/app/test/test_event_eth_rx_adapter.c
index 13664a3..76ce57b 100644
--- a/app/test/test_event_eth_rx_adapter.c
+++ b/app/test/test_event_eth_rx_adapter.c
@@ -751,20 +751,48 @@ static int
adapter_queue_conf(void)
{
int err;
- struct rte_event_eth_rx_adapter_queue_conf queue_conf;
+ struct rte_event_eth_rx_adapter_queue_conf queue_conf = {0};
- err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID, TEST_DEV_ID,
+ /* Case 1: queue conf get without any queues in Rx adapter */
+ err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID,
+ TEST_ETHDEV_ID,
+ 0, &queue_conf);
+ TEST_ASSERT(err == -EINVAL, "Expected -EINVAL got %d", err);
+
+ /* Add queue to Rx adapter */
+ queue_conf.ev.queue_id = 0;
+ queue_conf.ev.sched_type = RTE_SCHED_TYPE_ATOMIC;
+ queue_conf.ev.priority = RTE_EVENT_DEV_PRIORITY_NORMAL;
+
+ err = rte_event_eth_rx_adapter_queue_add(TEST_INST_ID,
+TEST_ETHDEV_ID,
+0, &queue_conf);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
+ /* Case 2: queue conf get with queue added to Rx adapter */
+ err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID,
+ TEST_ETHDEV_ID,
0, &queue_conf);
TEST_ASSERT(err == 0, "Expected 0 got %d", err);
- err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID, TEST_DEV_ID,
+ /* Case 3: queue conf get with invalue rx queue id */
+ err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID,
+ TEST_ETHDEV_ID,
-1, &queue_conf);
TEST_ASSERT(err == -EINVAL, "Expected -EINVAL got %d", err);
- err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID, TEST_DEV_ID,
+ /* Case 4: queue conf get with NULL queue conf struct */
+ err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID,
+ TEST_ETHDEV_ID,
0, NULL);
TEST_ASSERT(err == -EINVAL, "Expected -EINVAL got %d", err);
+ /* Delete queue from the Rx adapter */
+ err = rte_event_eth_rx_adapter_queue_del(TEST_INST_ID,
+TEST_ETHDEV_ID,
+0);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
return TEST_SUCCESS;
}
diff --git a/lib/eventdev/rte_event_eth_rx_adapter.c
b/lib/eventdev/rte_event_eth_rx_adapter.c
index 10491ca..2a84490 100644
--- a/lib/eventdev/rte_event_eth_rx_adapter.c
+++ b/lib/eventdev/rte_event_eth_rx_adapter.c
@@ -2844,12 +2844,13 @@ rte_event_eth_rx_adapter_queue_conf_get(uint8_t id,
return -EINVAL;
dev_info = &rx_adapter->eth_devices[eth_dev_id];
- queue_info = &dev_info->rx_queue[rx_queue_id];
- if (!queue_info->queue_enabled) {
+ if (dev_info->rx_queue == NULL ||
+ !dev_info->rx_queue[rx_queue_id].queue_enabled) {
RTE_EDEV_LOG_ERR("Rx queue %u not added", rx_queue_id);
return -EINVAL;
}
+ queue_info = &dev_info->rx_queue[rx_queue_id];
qi_ev = (struct rte_event *)&queue_info->event;
memset(queue_conf, 0, sizeof(*queue_conf));
--
2.6.4
Re: [dpdk-dev] [PATCH v5] ethdev: fix representor port ID search by name
On 9/30/2021 5:33 PM, Andrew Rybchenko wrote: On 9/29/21 2:13 PM, Singh, Aman Deep wrote: On 9/13/2021 4:56 PM, Andrew Rybchenko wrote: From: Viacheslav Galaktionov Getting a list of representors from a representor does not make sense. Instead, a parent device should be used. To this end, extend the rte_eth_dev_data structure to include the port ID of the backing device for representors. Signed-off-by: Viacheslav Galaktionov Signed-off-by: Andrew Rybchenko Acked-by: Haiyue Wang Acked-by: Beilei Xing --- The new field is added into the hole in rte_eth_dev_data structure. The patch does not change ABI, but extra care is required since ABI check is disabled for the structure because of the libabigail bug [1]. It should not be a problem anyway since 21.11 is a ABI breaking release. Potentially it is bad for out-of-tree drivers which implement representors but do not fill in a new parert_port_id field in rte_eth_dev_data structure. Get ID by name will not work. Did we change name of new field from parert_port_id to backer_port_id. Yes, see v5 changelog below. It is done to address review notes from Ferruh on v4. Maybe I did not put it clearly, my bad. Just wanted, in above lines also the usage of "parert_port_id" should be changed. mlx5 changes should be reviwed by maintainers very carefully, since we are not sure if we patch it correctly. [1] https://sourceware.org/bugzilla/show_bug.cgi?id=28060 v5: - try to improve name: backer_port_id instead of parent_port_id - init new field to RTE_MAX_ETHPORTS on allocation to avoid zero port usage by default v4: - apply mlx5 review notes: remove fallback from generic ethdev code and add fallback to mlx5 code to handle legacy usecase v3: - fix mlx5 build breakage v2: - fix mlx5 review notes - try device port ID first before parent in order to address backward compatibility issue [snip]
[dpdk-dev] [PATCH v4 0/3] add SA config option for inner pkt csum
Add inner packet IPv4 hdr and L4 checksum enable options in conf. These will be used in case of protocol offload. Per SA, application could specify whether the checksum(compute/verify) can be offloaded to security device. Changes in v4: - Rebased to ToT - Added documentation for per packet checksum(comment from Konstantin) Changes in v3: - Removed code unrelated to this series. Changes in v2: - Fixed release notes - Added feature flag in default.ini and cn10k.ini - Fixed test patch subject Archana Muniganti (3): security: add SA config option for inner pkt csum crypto/cnxk: add inner checksum test/crypto: add inner checksum cases app/test/test_cryptodev.c | 34 +++ app/test/test_cryptodev_security_ipsec.c | 195 ++ app/test/test_cryptodev_security_ipsec.h | 2 + ...st_cryptodev_security_ipsec_test_vectors.h | 6 + doc/guides/cryptodevs/features/cn10k.ini | 1 + doc/guides/cryptodevs/features/default.ini| 1 + doc/guides/rel_notes/deprecation.rst | 4 +- doc/guides/rel_notes/release_21_11.rst| 6 + drivers/crypto/cnxk/cn10k_cryptodev_ops.c | 65 -- drivers/crypto/cnxk/cn10k_ipsec.c | 49 - drivers/crypto/cnxk/cn10k_ipsec.h | 1 + drivers/crypto/cnxk/cn10k_ipsec_la_ops.h | 9 +- drivers/crypto/cnxk/cnxk_cryptodev.c | 3 + .../crypto/cnxk/cnxk_cryptodev_capabilities.c | 2 + lib/cryptodev/rte_cryptodev.h | 2 + lib/security/rte_security.h | 31 +++ 16 files changed, 391 insertions(+), 20 deletions(-) -- 2.22.0
[dpdk-dev] [PATCH v4 1/3] security: add SA config option for inner pkt csum
Add inner packet IPv4 hdr and L4 checksum enable options
in conf. These will be used in case of protocol offload.
Per SA, application could specify whether the
checksum(compute/verify) can be offloaded to security device.
Signed-off-by: Archana Muniganti
---
doc/guides/cryptodevs/features/default.ini | 1 +
doc/guides/rel_notes/deprecation.rst | 4 +--
doc/guides/rel_notes/release_21_11.rst | 4 +++
lib/cryptodev/rte_cryptodev.h | 2 ++
lib/security/rte_security.h| 31 ++
5 files changed, 40 insertions(+), 2 deletions(-)
diff --git a/doc/guides/cryptodevs/features/default.ini
b/doc/guides/cryptodevs/features/default.ini
index c24814de98..96d95ddc81 100644
--- a/doc/guides/cryptodevs/features/default.ini
+++ b/doc/guides/cryptodevs/features/default.ini
@@ -33,6 +33,7 @@ Non-Byte aligned data =
Sym raw data path API =
Cipher multiple data units =
Cipher wrapped key =
+Inner checksum =
;
; Supported crypto algorithms of a default crypto driver.
diff --git a/doc/guides/rel_notes/deprecation.rst
b/doc/guides/rel_notes/deprecation.rst
index 05fc2fdee7..8308e00ed4 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -232,8 +232,8 @@ Deprecation Notices
IPsec payload MSS (Maximum Segment Size), and ESN (Extended Sequence Number).
* security: The IPsec SA config options ``struct
rte_security_ipsec_sa_options``
- will be updated with new fields to support new features like IPsec inner
- checksum, TSO in case of protocol offload.
+ will be updated with new fields to support new features like TSO in case of
+ protocol offload.
* ipsec: The structure ``rte_ipsec_sa_prm`` will be extended with a new field
``hdr_l3_len`` to configure tunnel L3 header length.
diff --git a/doc/guides/rel_notes/release_21_11.rst
b/doc/guides/rel_notes/release_21_11.rst
index 3ade7fe5ac..5480f05a99 100644
--- a/doc/guides/rel_notes/release_21_11.rst
+++ b/doc/guides/rel_notes/release_21_11.rst
@@ -196,6 +196,10 @@ ABI Changes
``rte_security_ipsec_xform`` to allow applications to configure SA soft
and hard expiry limits. Limits can be either in number of packets or bytes.
+* security: The new options ``ip_csum_enable`` and ``l4_csum_enable`` were
added
+ in structure ``rte_security_ipsec_sa_options`` to indicate whether inner
+ packet IPv4 header checksum and L4 checksum need to be offloaded to
+ security device.
Known Issues
diff --git a/lib/cryptodev/rte_cryptodev.h b/lib/cryptodev/rte_cryptodev.h
index bb01f0f195..d9271a6c45 100644
--- a/lib/cryptodev/rte_cryptodev.h
+++ b/lib/cryptodev/rte_cryptodev.h
@@ -479,6 +479,8 @@ rte_cryptodev_asym_get_xform_enum(enum
rte_crypto_asym_xform_type *xform_enum,
/**< Support operations on multiple data-units message */
#define RTE_CRYPTODEV_FF_CIPHER_WRAPPED_KEY(1ULL << 26)
/**< Support wrapped key in cipher xform */
+#define RTE_CRYPTODEV_FF_SECURITY_INNER_CSUM (1ULL << 27)
+/**< Support inner checksum computation/verification */
/**
* Get the name of a crypto device feature flag
diff --git a/lib/security/rte_security.h b/lib/security/rte_security.h
index ab1a6e1f65..0c5636377e 100644
--- a/lib/security/rte_security.h
+++ b/lib/security/rte_security.h
@@ -230,6 +230,37 @@ struct rte_security_ipsec_sa_options {
* * 0: Do not match UDP ports
*/
uint32_t udp_ports_verify : 1;
+
+ /** Compute/verify inner packet IPv4 header checksum in tunnel mode
+*
+* * 1: For outbound, compute inner packet IPv4 header checksum
+* before tunnel encapsulation and for inbound, verify after
+* tunnel decapsulation.
+* * 0: Inner packet IP header checksum is not computed/verified.
+*
+* The checksum verification status would be set in mbuf using
+* PKT_RX_IP_CKSUM_xxx flags.
+*
+* Inner IP checksum computation can also be enabled(per operation)
+* by setting the flag PKT_TX_IP_CKSUM in mbuf.
+*/
+ uint32_t ip_csum_enable : 1;
+
+ /** Compute/verify inner packet L4 checksum in tunnel mode
+*
+* * 1: For outbound, compute inner packet L4 checksum before
+* tunnel encapsulation and for inbound, verify after
+* tunnel decapsulation.
+* * 0: Inner packet L4 checksum is not computed/verified.
+*
+* The checksum verification status would be set in mbuf using
+* PKT_RX_L4_CKSUM_xxx flags.
+*
+* Inner L4 checksum computation can also be enabled(per operation)
+* by setting the flags PKT_TX_TCP_CKSUM or PKT_TX_SCTP_CKSUM or
+* PKT_TX_UDP_CKSUM or PKT_TX_L4_MASK in mbuf.
+*/
+ uint32_t l4_csum_enable : 1;
};
/** IPSec security association direction */
--
2.22.0
[dpdk-dev] [PATCH v4 2/3] crypto/cnxk: add inner checksum
Add inner checksum support for cn10k.
Signed-off-by: Archana Muniganti
---
doc/guides/cryptodevs/features/cn10k.ini | 1 +
doc/guides/rel_notes/release_21_11.rst| 1 +
drivers/crypto/cnxk/cn10k_cryptodev_ops.c | 65 +++
drivers/crypto/cnxk/cn10k_ipsec.c | 49 +-
drivers/crypto/cnxk/cn10k_ipsec.h | 1 +
drivers/crypto/cnxk/cn10k_ipsec_la_ops.h | 9 ++-
drivers/crypto/cnxk/cnxk_cryptodev.c | 3 +
.../crypto/cnxk/cnxk_cryptodev_capabilities.c | 2 +
8 files changed, 113 insertions(+), 18 deletions(-)
diff --git a/doc/guides/cryptodevs/features/cn10k.ini
b/doc/guides/cryptodevs/features/cn10k.ini
index f5552feca3..9d08bd5c04 100644
--- a/doc/guides/cryptodevs/features/cn10k.ini
+++ b/doc/guides/cryptodevs/features/cn10k.ini
@@ -15,6 +15,7 @@ OOP SGL In SGL Out = Y
OOP LB In LB Out = Y
Symmetric sessionless = Y
Digest encrypted = Y
+Inner checksum = Y
;
; Supported crypto algorithms of 'cn10k' crypto driver.
diff --git a/doc/guides/rel_notes/release_21_11.rst
b/doc/guides/rel_notes/release_21_11.rst
index 5480f05a99..576bc01a87 100644
--- a/doc/guides/rel_notes/release_21_11.rst
+++ b/doc/guides/rel_notes/release_21_11.rst
@@ -73,6 +73,7 @@ New Features
* Added UDP encapsulation support in lookaside protocol (IPsec) for CN10K.
* Added support for lookaside protocol (IPsec) offload for CN9K.
* Added support for ZUC algorithm with 256 bit key length for CN10K.
+ * Added inner checksum support in lookaside protocol (IPsec) for CN10K.
* **Added support for event crypto adapter on Marvell CN10K and CN9K.**
diff --git a/drivers/crypto/cnxk/cn10k_cryptodev_ops.c
b/drivers/crypto/cnxk/cn10k_cryptodev_ops.c
index 3caf05aab9..c25c8e67b2 100644
--- a/drivers/crypto/cnxk/cn10k_cryptodev_ops.c
+++ b/drivers/crypto/cnxk/cn10k_cryptodev_ops.c
@@ -50,7 +50,7 @@ cn10k_cpt_sym_temp_sess_create(struct cnxk_cpt_qp *qp, struct
rte_crypto_op *op)
static __rte_always_inline int __rte_hot
cpt_sec_inst_fill(struct rte_crypto_op *op, struct cn10k_sec_session *sess,
- struct cpt_inst_s *inst)
+ struct cpt_inflight_req *infl_req, struct cpt_inst_s *inst)
{
struct rte_crypto_sym_op *sym_op = op->sym;
union roc_ot_ipsec_sa_word2 *w2;
@@ -72,8 +72,10 @@ cpt_sec_inst_fill(struct rte_crypto_op *op, struct
cn10k_sec_session *sess,
if (w2->s.dir == ROC_IE_SA_DIR_OUTBOUND)
ret = process_outb_sa(op, sa, inst);
- else
+ else {
+ infl_req->op_flags |= CPT_OP_FLAGS_IPSEC_DIR_INBOUND;
ret = process_inb_sa(op, sa, inst);
+ }
return ret;
}
@@ -122,7 +124,8 @@ cn10k_cpt_fill_inst(struct cnxk_cpt_qp *qp, struct
rte_crypto_op *ops[],
if (op->sess_type == RTE_CRYPTO_OP_SECURITY_SESSION) {
sec_sess = get_sec_session_private_data(
sym_op->sec_session);
- ret = cpt_sec_inst_fill(op, sec_sess, &inst[0]);
+ ret = cpt_sec_inst_fill(op, sec_sess, infl_req,
+ &inst[0]);
if (unlikely(ret))
return 0;
w7 = sec_sess->sa.inst.w7;
@@ -342,6 +345,49 @@ cn10k_cpt_sec_post_process(struct rte_crypto_op *cop,
m->pkt_len = m_len;
}
+static inline void
+cn10k_cpt_sec_ucc_process(struct rte_crypto_op *cop,
+ struct cpt_inflight_req *infl_req,
+ const uint8_t uc_compcode)
+{
+ struct cn10k_sec_session *sess;
+ struct cn10k_ipsec_sa *sa;
+ struct rte_mbuf *mbuf;
+
+ if (uc_compcode == ROC_IE_OT_UCC_SUCCESS_SA_SOFTEXP_FIRST)
+ cop->aux_flags = RTE_CRYPTO_OP_AUX_FLAGS_IPSEC_SOFT_EXPIRY;
+
+ if (!(infl_req->op_flags & CPT_OP_FLAGS_IPSEC_DIR_INBOUND))
+ return;
+
+ sess = get_sec_session_private_data(cop->sym->sec_session);
+ sa = &sess->sa;
+
+ mbuf = cop->sym->m_src;
+
+ switch (uc_compcode) {
+ case ROC_IE_OT_UCC_SUCCESS:
+ if (sa->ip_csum_enable)
+ mbuf->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
+ break;
+ case ROC_IE_OT_UCC_SUCCESS_PKT_IP_BADCSUM:
+ mbuf->ol_flags |= PKT_RX_IP_CKSUM_BAD;
+ break;
+ case ROC_IE_OT_UCC_SUCCESS_PKT_L4_GOODCSUM:
+ mbuf->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
+ if (sa->ip_csum_enable)
+ mbuf->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
+ break;
+ case ROC_IE_OT_UCC_SUCCESS_PKT_L4_BADCSUM:
+ mbuf->ol_flags |= PKT_RX_L4_CKSUM_BAD;
+ if (sa->ip_csum_enable)
+ mbuf->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
+ break;
+ default:
+ break;
+ }
+}
+
static inline void
cn10k_cpt_deque
[dpdk-dev] [PATCH v4 3/3] test/crypto: add inner checksum cases
This patch adds tests for inner IP and inner L4 checksum
in IPsec mode.
Signed-off-by: Archana Muniganti
---
app/test/test_cryptodev.c | 34 +++
app/test/test_cryptodev_security_ipsec.c | 195 ++
app/test/test_cryptodev_security_ipsec.h | 2 +
...st_cryptodev_security_ipsec_test_vectors.h | 6 +
doc/guides/rel_notes/release_21_11.rst| 1 +
5 files changed, 238 insertions(+)
diff --git a/app/test/test_cryptodev.c b/app/test/test_cryptodev.c
index e6ceeb487f..65b64e1af0 100644
--- a/app/test/test_cryptodev.c
+++ b/app/test/test_cryptodev.c
@@ -18,6 +18,8 @@
#include
#include
#include
+#include
+#include
#ifdef RTE_CRYPTO_SCHEDULER
#include
@@ -9299,6 +9301,30 @@ test_ipsec_proto_udp_ports_verify(const void *data
__rte_unused)
return test_ipsec_proto_all(&flags);
}
+static int
+test_ipsec_proto_inner_ip_csum(const void *data __rte_unused)
+{
+ struct ipsec_test_flags flags;
+
+ memset(&flags, 0, sizeof(flags));
+
+ flags.ip_csum = true;
+
+ return test_ipsec_proto_all(&flags);
+}
+
+static int
+test_ipsec_proto_inner_l4_csum(const void *data __rte_unused)
+{
+ struct ipsec_test_flags flags;
+
+ memset(&flags, 0, sizeof(flags));
+
+ flags.l4_csum = true;
+
+ return test_ipsec_proto_all(&flags);
+}
+
static int
test_PDCP_PROTO_all(void)
{
@@ -14255,6 +14281,14 @@ static struct unit_test_suite ipsec_proto_testsuite =
{
"Tunnel src and dst addr verification",
ut_setup_security, ut_teardown,
test_ipsec_proto_tunnel_src_dst_addr_verify),
+ TEST_CASE_NAMED_ST(
+ "Inner IP checksum",
+ ut_setup_security, ut_teardown,
+ test_ipsec_proto_inner_ip_csum),
+ TEST_CASE_NAMED_ST(
+ "Inner L4 checksum",
+ ut_setup_security, ut_teardown,
+ test_ipsec_proto_inner_l4_csum),
TEST_CASES_END() /**< NULL terminate unit test array */
}
};
diff --git a/app/test/test_cryptodev_security_ipsec.c
b/app/test/test_cryptodev_security_ipsec.c
index 764e77bbff..bcd9746c98 100644
--- a/app/test/test_cryptodev_security_ipsec.c
+++ b/app/test/test_cryptodev_security_ipsec.c
@@ -7,6 +7,7 @@
#include
#include
#include
+#include
#include
#include "test.h"
@@ -103,6 +104,22 @@ test_ipsec_sec_caps_verify(struct rte_security_ipsec_xform
*ipsec_xform,
return -ENOTSUP;
}
+ if (ipsec_xform->options.ip_csum_enable == 1 &&
+ sec_cap->ipsec.options.ip_csum_enable == 0) {
+ if (!silent)
+ RTE_LOG(INFO, USER1,
+ "Inner IP checksum is not supported\n");
+ return -ENOTSUP;
+ }
+
+ if (ipsec_xform->options.l4_csum_enable == 1 &&
+ sec_cap->ipsec.options.l4_csum_enable == 0) {
+ if (!silent)
+ RTE_LOG(INFO, USER1,
+ "Inner L4 checksum is not supported\n");
+ return -ENOTSUP;
+ }
+
return 0;
}
@@ -160,6 +177,56 @@ test_ipsec_td_in_from_out(const struct ipsec_test_data
*td_out,
}
}
+static bool
+is_ipv4(void *ip)
+{
+ struct rte_ipv4_hdr *ipv4 = ip;
+ uint8_t ip_ver;
+
+ ip_ver = (ipv4->version_ihl & 0xf0) >> RTE_IPV4_IHL_MULTIPLIER;
+ if (ip_ver == IPVERSION)
+ return true;
+ else
+ return false;
+}
+
+static void
+test_ipsec_csum_init(void *ip, bool l3, bool l4)
+{
+ struct rte_ipv4_hdr *ipv4;
+ struct rte_tcp_hdr *tcp;
+ struct rte_udp_hdr *udp;
+ uint8_t next_proto;
+ uint8_t size;
+
+ if (is_ipv4(ip)) {
+ ipv4 = ip;
+ size = sizeof(struct rte_ipv4_hdr);
+ next_proto = ipv4->next_proto_id;
+
+ if (l3)
+ ipv4->hdr_checksum = 0;
+ } else {
+ size = sizeof(struct rte_ipv6_hdr);
+ next_proto = ((struct rte_ipv6_hdr *)ip)->proto;
+ }
+
+ if (l4) {
+ switch (next_proto) {
+ case IPPROTO_TCP:
+ tcp = (struct rte_tcp_hdr *)RTE_PTR_ADD(ip, size);
+ tcp->cksum = 0;
+ break;
+ case IPPROTO_UDP:
+ udp = (struct rte_udp_hdr *)RTE_PTR_ADD(ip, size);
+ udp->dgram_cksum = 0;
+ break;
+ default:
+ return;
+ }
+ }
+}
+
void
test_ipsec_td_prepare(const struct crypto_param *param1,
const struct crypto_param *param2,
@@ -194,6 +261,17 @@ test_ipsec_td_prepare(const struct crypto_param *param1,
if (flags->sa_expiry_pkts_soft)
[dpdk-dev] [PATCH v2] eventdev/rx-adapter: segfault in queue conf get
rte_event_eth_rx_adapter_queue_conf_get() segfaults if called
without queue added to the Rx adapter.
Added check to no queues in Rx adapter and error out on being
called with no queue in Rx adapter.
Added test case to call queue conf get without queues in
Rx adapter.
Signed-off-by: Ganapati Kundapura
---
v2:
* Corrected typo in the comment
---
diff --git a/app/test/test_event_eth_rx_adapter.c
b/app/test/test_event_eth_rx_adapter.c
index 13664a3..d0dc552 100644
--- a/app/test/test_event_eth_rx_adapter.c
+++ b/app/test/test_event_eth_rx_adapter.c
@@ -751,20 +751,48 @@ static int
adapter_queue_conf(void)
{
int err;
- struct rte_event_eth_rx_adapter_queue_conf queue_conf;
+ struct rte_event_eth_rx_adapter_queue_conf queue_conf = {0};
- err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID, TEST_DEV_ID,
+ /* Case 1: queue conf get without any queues in Rx adapter */
+ err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID,
+ TEST_ETHDEV_ID,
+ 0, &queue_conf);
+ TEST_ASSERT(err == -EINVAL, "Expected -EINVAL got %d", err);
+
+ /* Add queue to Rx adapter */
+ queue_conf.ev.queue_id = 0;
+ queue_conf.ev.sched_type = RTE_SCHED_TYPE_ATOMIC;
+ queue_conf.ev.priority = RTE_EVENT_DEV_PRIORITY_NORMAL;
+
+ err = rte_event_eth_rx_adapter_queue_add(TEST_INST_ID,
+TEST_ETHDEV_ID,
+0, &queue_conf);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
+ /* Case 2: queue conf get with queue added to Rx adapter */
+ err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID,
+ TEST_ETHDEV_ID,
0, &queue_conf);
TEST_ASSERT(err == 0, "Expected 0 got %d", err);
- err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID, TEST_DEV_ID,
+ /* Case 3: queue conf get with invalid rx queue id */
+ err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID,
+ TEST_ETHDEV_ID,
-1, &queue_conf);
TEST_ASSERT(err == -EINVAL, "Expected -EINVAL got %d", err);
- err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID, TEST_DEV_ID,
+ /* Case 4: queue conf get with NULL queue conf struct */
+ err = rte_event_eth_rx_adapter_queue_conf_get(TEST_INST_ID,
+ TEST_ETHDEV_ID,
0, NULL);
TEST_ASSERT(err == -EINVAL, "Expected -EINVAL got %d", err);
+ /* Delete queue from the Rx adapter */
+ err = rte_event_eth_rx_adapter_queue_del(TEST_INST_ID,
+TEST_ETHDEV_ID,
+0);
+ TEST_ASSERT(err == 0, "Expected 0 got %d", err);
+
return TEST_SUCCESS;
}
diff --git a/lib/eventdev/rte_event_eth_rx_adapter.c
b/lib/eventdev/rte_event_eth_rx_adapter.c
index 10491ca..2a84490 100644
--- a/lib/eventdev/rte_event_eth_rx_adapter.c
+++ b/lib/eventdev/rte_event_eth_rx_adapter.c
@@ -2844,12 +2844,13 @@ rte_event_eth_rx_adapter_queue_conf_get(uint8_t id,
return -EINVAL;
dev_info = &rx_adapter->eth_devices[eth_dev_id];
- queue_info = &dev_info->rx_queue[rx_queue_id];
- if (!queue_info->queue_enabled) {
+ if (dev_info->rx_queue == NULL ||
+ !dev_info->rx_queue[rx_queue_id].queue_enabled) {
RTE_EDEV_LOG_ERR("Rx queue %u not added", rx_queue_id);
return -EINVAL;
}
+ queue_info = &dev_info->rx_queue[rx_queue_id];
qi_ev = (struct rte_event *)&queue_info->event;
memset(queue_conf, 0, sizeof(*queue_conf));
--
2.6.4
Re: [dpdk-dev] [PATCH v2] drivers/net: remove queue xstats auto-fill flag
On 9/29/2021 7:38 AM, Andrew Rybchenko wrote:
> On 9/28/21 8:10 PM, Stephen Hemminger wrote:
>> On Tue, 28 Sep 2021 19:48:54 +0300
>> Andrew Rybchenko wrote:
>>
>>> Some drivers do not provide per-queue statistics. So, there is no point
>>> to have these misleading zeros in xstats.
>>>
>>> Fixes: f30e69b41f94 ("ethdev: add device flag to bypass auto-filled queue
>>> xstats")
>>> Cc: sta...@dpdk.org
>>>
>>> Signed-off-by: Andrew Rybchenko
>> Really?
>> It is useful to have zeros rather than random data there.
>
> I guess there is a misunderstanding here. Auto-filling xstats is
> an addition of per-queue basic statistics to xstats by ethdev
> layer. It makes sense to do it if and only if there is some
> sensible data there.
>
> There is a related deprecation notice saying that per-queue
> stats should be removed from basic stats since per-queue
> stats should be provided by xstats API natively.
>
> Basically RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS means
> that corresponding driver is not ready vs the deprecation notice.
> So, I want to clean it up to see not yet ready drivers only.
>
As you said, 'RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS' flag set by driver means,
driver is not ready on representing queue stats in xstats and ethdev layer is
filling it automatically from basic stats.
First we should wait for drivers to implement it, later clean queue stats from
basic stats and remove the flag.
I am not sure if we can remove the deprecation notice in this release, but agree
to add a deadline for the drivers, which can be 22.11.
[dpdk-dev] [PATCH v6 1/2] Enable ASan for memory detector on DPDK
From: Zhihong Peng
AddressSanitizer (ASan) is a google memory error detect
standard tool. It could help to detect use-after-free and
{heap,stack,global}-buffer overflow bugs in C/C++ programs,
print detailed error information when error happens, large
improve debug efficiency.
`AddressSanitizer
` (ASan)
is a widely-used debugging tool to detect memory access errors.
It helps detect issues like use-after-free, various kinds of buffer
overruns in C/C++ programs, and other similar errors, as well as
printing out detailed debug information whenever an error is detected.
DPDK ASan functionality is currently only supported Linux x86_64.
Support other platforms, need to define ASAN_SHADOW_OFFSET value
according to google ASan document.
Here is an example of heap-buffer-overflow bug:
..
char *p = rte_zmalloc(NULL, 7, 0);
p[7] = 'a';
..
Here is an example of use-after-free bug:
..
char *p = rte_zmalloc(NULL, 7, 0);
rte_free(p);
*p = 'a';
..
If you want to use this feature,
you need to add below compilation options when compiling code:
-Dbuildtype=debug -Db_lundef=false -Db_sanitize=address
"-Dbuildtype=debug": This is a non-essential option. When this option
is added, if a memory error occurs, ASan can clearly show where the
code is wrong.
"-Db_lundef=false": When use clang to compile DPDK, this option must
be added.
Signed-off-by: Xueqin Lin
Signed-off-by: Zhihong Peng
---
devtools/words-case.txt | 1 +
doc/guides/prog_guide/ASan.rst | 112 ++
doc/guides/prog_guide/index.rst | 1 +
lib/eal/common/malloc_elem.c| 26 +++-
lib/eal/common/malloc_elem.h| 204 +++-
lib/eal/common/malloc_heap.c| 12 ++
lib/eal/common/rte_malloc.c | 9 +-
7 files changed, 360 insertions(+), 5 deletions(-)
create mode 100644 doc/guides/prog_guide/ASan.rst
diff --git a/devtools/words-case.txt b/devtools/words-case.txt
index 0bbad48626..3655596d47 100644
--- a/devtools/words-case.txt
+++ b/devtools/words-case.txt
@@ -86,3 +86,4 @@ VXLAN
Windows
XDP
XOR
+ASan
diff --git a/doc/guides/prog_guide/ASan.rst b/doc/guides/prog_guide/ASan.rst
new file mode 100644
index 00..1fd495150b
--- /dev/null
+++ b/doc/guides/prog_guide/ASan.rst
@@ -0,0 +1,112 @@
+.. Copyright (c) <2021>, Intel Corporation
+ All rights reserved.
+
+Memory error detect standard tool - AddressSanitizer(ASan)
+==
+
+AddressSanitizer (ASan) is a google memory error detect
+standard tool. It could help to detect use-after-free and
+{heap,stack,global}-buffer overflow bugs in C/C++ programs,
+print detailed error information when error happens, large
+improve debug efficiency.
+
+By referring to its implementation algorithm
+(https://github.com/google/sanitizers/wiki/AddressSanitizerAlgorithm),
+enabled heap-buffer-overflow and use-after-free functions on DPDK.
+DPDK ASan function currently only supports on Linux x86_64.
+
+AddressSanitizer is a part of LLVM(3.1+)and GCC(4.8+).
+
+DPDK ASan functionality is currently only supported Linux x86_64.
+Support other platforms, need to define ASAN_SHADOW_OFFSET value
+according to google ASan document.
+
+Example heap-buffer-overflow error
+--
+
+Following error was reported when ASan was enabled::
+
+Applied 9 bytes of memory, but accessed the 10th byte of memory,
+so heap-buffer-overflow appeared.
+
+Below code results in this error::
+
+char *p = rte_zmalloc(NULL, 9, 0);
+if (!p) {
+printf("rte_zmalloc error.");
+return -1;
+}
+p[9] = 'a';
+
+The error log::
+
+==49433==ERROR: AddressSanitizer: heap-buffer-overflow on address
0x7f773fafa249 at pc 0x5556b13bdae4 bp 0x7ffeb4965e40 sp 0x7ffeb4965e30 WRITE
of size 1 at 0x7f773fafa249 thread T0
+#0 0x5556b13bdae3 in asan_heap_buffer_overflow
../app/test/test_asan_heap_buffer_overflow.c:25
+#1 0x5556b043e9d4 in cmd_autotest_parsed ../app/test/commands.c:71
+#2 0x5556b1cdd4b0 in cmdline_parse ../lib/cmdline/cmdline_parse.c:290
+#3 0x5556b1cd8987 in cmdline_valid_buffer ../lib/cmdline/cmdline.c:26
+#4 0x5556b1ce477a in rdline_char_in ../lib/cmdline/cmdline_rdline.c:421
+#5 0x5556b1cd923e in cmdline_in ../lib/cmdline/cmdline.c:149
+#6 0x5556b1cd9769 in cmdline_interact ../lib/cmdline/cmdline.c:223
+#7 0x5556b045f53b in main ../app/test/test.c:234
+#8 0x7f7f1eba90b2 in __libc_start_main
(/lib/x86_64-linux-gnu/libc.so.6+0x270b2)
+#9 0x5556b043e70d in _start
(/home/pzh/yyy/x86_64-native-linuxapp-gcc/app/test/dpdk-test+0x7ce70d)
+
+Address 0x7f773fafa249 is a wild pointer.
+SUMMARY: AddressSanitizer: heap-buffer-overflow
../app/test/test_asan_heap_buffer_overflow.c:25 in asan_heap_buffer_overflow
+
+Example use-after-free error
+
+
+Following er
[dpdk-dev] [PATCH v6 2/2] lib/pipeline: Fix compilation error with gcc ASan
From: Zhihong Peng After adding ASan, the gcc compilation check will be stricter. "Control reaches end of non-void function" error occurs here. Fixes: f38913b7fb8e (pipeline: add meter array to SWX) Cc: sta...@dpdk.org Signed-off-by: Xueqin Lin Signed-off-by: Zhihong Peng --- lib/pipeline/rte_swx_pipeline.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/lib/pipeline/rte_swx_pipeline.c b/lib/pipeline/rte_swx_pipeline.c index 1cd09a4b44..0acd6c6752 100644 --- a/lib/pipeline/rte_swx_pipeline.c +++ b/lib/pipeline/rte_swx_pipeline.c @@ -4642,7 +4642,7 @@ instr_meter_translate(struct rte_swx_pipeline *p, return 0; } - CHECK(0, EINVAL); + return -EINVAL; } static inline void @@ -5937,7 +5937,7 @@ instr_translate(struct rte_swx_pipeline *p, instr, data); - CHECK(0, EINVAL); + return -EINVAL; } static struct instruction_data * -- 2.25.1
[dpdk-dev] [PATCH] crypto/cnxk: add max queue pairs limit devargs
Adds max queue pairs limit devargs for crypto cnxk driver. This
can be used to set a limit on the number of maximum queue pairs
supported by the device. The default value is 63.
Signed-off-by: Ankur Dwivedi
Reviewed-by: Anoob Joseph
Reviewed-by: Jerin Jacob Kollanukkaran
---
doc/guides/cryptodevs/cnxk.rst | 15 +
drivers/crypto/cnxk/cn10k_cryptodev.c| 7 +++
drivers/crypto/cnxk/cn9k_cryptodev.c | 7 +++
drivers/crypto/cnxk/cnxk_cryptodev.h | 2 +
drivers/crypto/cnxk/cnxk_cryptodev_devargs.c | 61
drivers/crypto/cnxk/cnxk_cryptodev_ops.c | 5 +-
drivers/crypto/cnxk/meson.build | 1 +
7 files changed, 97 insertions(+), 1 deletion(-)
create mode 100644 drivers/crypto/cnxk/cnxk_cryptodev_devargs.c
diff --git a/doc/guides/cryptodevs/cnxk.rst b/doc/guides/cryptodevs/cnxk.rst
index 752316fd37..85171a50a6 100644
--- a/doc/guides/cryptodevs/cnxk.rst
+++ b/doc/guides/cryptodevs/cnxk.rst
@@ -158,6 +158,21 @@ Bind the CPT VF device to the vfio_pci driver:
./usertools/dpdk-devbind.py -u 0002:20:00.1
./usertools/dpdk-devbind.py -b vfio-pci 0002:20:00.1
+Runtime Config Options
+--
+
+- ``Maximum queue pairs limit`` (default ``63``)
+
+ The number of maximum queue pairs supported by the device, can be limited
+ during runtime by using ``max_qps_limit`` ``devargs`` parameter.
+
+ For example::
+
+ -a 0002:20:00.1,max_qps_limit=4
+
+ With the above configuration, the number of maximum queue pairs supported
+ by the device is limited to 4.
+
Debugging Options
-
diff --git a/drivers/crypto/cnxk/cn10k_cryptodev.c
b/drivers/crypto/cnxk/cn10k_cryptodev.c
index 012eb0c051..869d322d9b 100644
--- a/drivers/crypto/cnxk/cn10k_cryptodev.c
+++ b/drivers/crypto/cnxk/cn10k_cryptodev.c
@@ -68,6 +68,13 @@ cn10k_cpt_pci_probe(struct rte_pci_driver *pci_drv
__rte_unused,
if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
roc_cpt->pci_dev = pci_dev;
+
+ rc = cnxk_cpt_parse_devargs(dev->device->devargs, vf);
+ if (rc) {
+ plt_err("Failed to parse devargs rc=%d", rc);
+ goto pmd_destroy;
+ }
+
rc = roc_cpt_dev_init(roc_cpt);
if (rc) {
plt_err("Failed to initialize roc cpt rc=%d", rc);
diff --git a/drivers/crypto/cnxk/cn9k_cryptodev.c
b/drivers/crypto/cnxk/cn9k_cryptodev.c
index 6b8cb01a12..54df06eec0 100644
--- a/drivers/crypto/cnxk/cn9k_cryptodev.c
+++ b/drivers/crypto/cnxk/cn9k_cryptodev.c
@@ -68,6 +68,13 @@ cn9k_cpt_pci_probe(struct rte_pci_driver *pci_drv
__rte_unused,
if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
roc_cpt->pci_dev = pci_dev;
+
+ rc = cnxk_cpt_parse_devargs(dev->device->devargs, vf);
+ if (rc) {
+ plt_err("Failed to parse devargs rc=%d", rc);
+ goto pmd_destroy;
+ }
+
rc = roc_cpt_dev_init(roc_cpt);
if (rc) {
plt_err("Failed to initialize roc cpt rc=%d", rc);
diff --git a/drivers/crypto/cnxk/cnxk_cryptodev.h
b/drivers/crypto/cnxk/cnxk_cryptodev.h
index 8e051fa0fa..cfb9d291a9 100644
--- a/drivers/crypto/cnxk/cnxk_cryptodev.h
+++ b/drivers/crypto/cnxk/cnxk_cryptodev.h
@@ -25,9 +25,11 @@ struct cnxk_cpt_vf {
struct rte_security_capability sec_caps[CNXK_SEC_MAX_CAPS];
uint64_t cnxk_fpm_iova[CNXK_AE_EC_ID_MAX];
struct roc_ae_ec_group *ec_grp[CNXK_AE_EC_ID_MAX];
+ uint16_t max_qps_limit;
};
uint64_t cnxk_cpt_default_ff_get(void);
int cnxk_cpt_eng_grp_add(struct roc_cpt *roc_cpt);
+int cnxk_cpt_parse_devargs(struct rte_devargs *devargs, struct cnxk_cpt_vf
*vf);
#endif /* _CNXK_CRYPTODEV_H_ */
diff --git a/drivers/crypto/cnxk/cnxk_cryptodev_devargs.c
b/drivers/crypto/cnxk/cnxk_cryptodev_devargs.c
new file mode 100644
index 00..c3e9bdb2d1
--- /dev/null
+++ b/drivers/crypto/cnxk/cnxk_cryptodev_devargs.c
@@ -0,0 +1,61 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2021 Marvell.
+ */
+
+#include
+
+#include "cnxk_cryptodev.h"
+
+#define CNXK_MAX_QPS_LIMIT "max_qps_limit"
+#define CNXK_MAX_QPS_LIMIT_MIN 1
+#define CNXK_MAX_QPS_LIMIT_MAX (ROC_CPT_MAX_LFS - 1)
+
+static int
+parse_max_qps_limit(const char *key, const char *value, void *extra_args)
+{
+ RTE_SET_USED(key);
+ uint32_t val;
+
+ val = atoi(value);
+
+ if (val < CNXK_MAX_QPS_LIMIT_MIN || val > CNXK_MAX_QPS_LIMIT_MAX)
+ return -EINVAL;
+
+ *(uint16_t *)extra_args = val;
+
+ return 0;
+}
+
+int
+cnxk_cpt_parse_devargs(struct rte_devargs *devargs, struct cnxk_cpt_vf *vf)
+{
+ uint16_t max_qps_limit = CNXK_MAX_QPS_LIMIT_MAX;
+ struct rte_kvargs *kvlist;
+ int rc;
+
+ if (devargs == NULL)
+ goto null_devargs;
+
+ kvlist = rte_kvarg
Re: [dpdk-dev] [PATCH] ethdev: remove legacy Rx descriptor done API
On 9/28/2021 4:48 PM, Andrew Rybchenko wrote: > rte_eth_rx_descriptor_status() should be used as a replacement. > > Signed-off-by: Andrew Rybchenko Reviewed-by: Ferruh Yigit
Re: [dpdk-dev] [PATCH v5] mbuf: fix reset on mbuf free
> -Original Message-
> From: Olivier Matz
> Sent: Thursday, September 30, 2021 12:37 AM
> To: dev@dpdk.org
> Cc: ajit.khapa...@broadcom.com; ajitkhapa...@gmail.com; Ali Alnubani
> ; andrew.rybche...@oktetlabs.ru;
> konstantin.anan...@intel.com; m...@smartsharesystems.com;
> sta...@dpdk.org; Slava Ovsiienko
> Subject: [PATCH v5] mbuf: fix reset on mbuf free
>
> m->nb_seg must be reset on mbuf free whatever the value of m->next,
> because it can happen that m->nb_seg is != 1. For instance in this
> case:
>
> m1 = rte_pktmbuf_alloc(mp);
> rte_pktmbuf_append(m1, 500);
> m2 = rte_pktmbuf_alloc(mp);
> rte_pktmbuf_append(m2, 500);
> rte_pktmbuf_chain(m1, m2);
> m0 = rte_pktmbuf_alloc(mp);
> rte_pktmbuf_append(m0, 500);
> rte_pktmbuf_chain(m0, m1);
>
> As rte_pktmbuf_chain() does not reset nb_seg in the initial m1 segment (this
> is not required), after this code the mbuf chain have 3 segments:
> - m0: next=m1, nb_seg=3
> - m1: next=m2, nb_seg=2
> - m2: next=NULL, nb_seg=1
>
> Then split this chain between m1 and m2, it would result in 2 packets:
> - first packet
> - m0: next=m1, nb_seg=2
> - m1: next=NULL, nb_seg=2
> - second packet
> - m2: next=NULL, nb_seg=1
>
> Freeing the first packet will not restore nb_seg=1 in the second segment.
> This is an issue because it is expected that mbufs stored in pool have their
> nb_seg field set to 1.
>
> Fixes: 8f094a9ac5d7 ("mbuf: set mbuf fields while in pool")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Olivier Matz
> Acked-by: Morten Brørup
> Acked-by: Ajit Khaparde
> Acked-by: Konstantin Ananyev
> ---
Tested-by: Ali Alnubani
Re: [dpdk-dev] [dpdk-stable] [PATCH v4] mbuf: fix reset on mbuf free
> -Original Message- > From: Olivier Matz > Sent: Thursday, September 30, 2021 12:39 AM > To: Ali Alnubani > Cc: dev@dpdk.org; David Marchand ; > Alexander Kozyrev ; NBU-Contact-Thomas Monjalon > ; Ferruh Yigit ; Slava > Ovsiienko ; zhaoyan.c...@intel.com; Morten > Brørup ; Andrew Rybchenko > ; Ananyev, Konstantin > ; Ajit Khaparde > ; jer...@marvell.com > Subject: Re: [dpdk-dev] [dpdk-stable] [PATCH v4] mbuf: fix reset on mbuf > free > > Hi Ali, > > > On Wed, Sep 29, 2021 at 08:03:17AM +, Ali Alnubani wrote: > > Hi Olivier, > > > > I wanted to retest the patch on latest main, but it no longer applies, could > you please rebase it? > > I rebased the patch: > https://patchwork.dpdk.org/project/dpdk/patch/20210929213707.17727-1- > olivier.m...@6wind.com/ > Thanks for the update, I tested v5 on main (84b3e4555a1f) and the regression we previously saw no longer reproduces. Will the patch be backported to 20.11? I can still reproduce a 2% drop in single core packet forwarding performance with v4 applied on v20.11.3. Testpmd: dpdk-testpmd -n 4 -a :82:00.0 -a :82:00.1 -c 0x9000 -- -i --nb-cores=1 --txd=256 --rxd=256 --auto-start Traffic generator is TREX with pattern (Ether / IP / UDP). Drop is from ~50.5 mpps to ~49.5 mpps. OS: 18.04.5 LTS Device: ConnectX-5 MLNX_OFED: MLNX_OFED_LINUX-5.4-1.0.3.0 Firmware: 16.31.1014 Regards, Ali
Re: [dpdk-dev] [PATCH] net/memif: allocate socket hash on any NUMA socket
On 9/28/2021 2:51 PM, Junxiao Shi wrote: > Previously, memif socket hash is always allocated on NUMA socket 0. > If the application is entirely running on another NUMA socket and EAL > --socket-limit prevents memory allocation on NUMA socket 0, memif > creation fails with "HASH: memory allocation failed" error. > > This patch allows allocating memif socket hash on any NUMA socket. > > Signed-off-by: Junxiao Shi > --- > drivers/net/memif/memif_socket.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/drivers/net/memif/memif_socket.c > b/drivers/net/memif/memif_socket.c > index f58ff4c0cb..364e818d65 100644 > --- a/drivers/net/memif/memif_socket.c > +++ b/drivers/net/memif/memif_socket.c > @@ -946,6 +946,7 @@ memif_create_socket_hash(void) > params.key_len = MEMIF_SOCKET_UN_SIZE; > params.hash_func = rte_jhash; > params.hash_func_init_val = 0; > + params.socket_id = SOCKET_ID_ANY; > return rte_hash_create(¶ms); > } > > looks good to me. cc'ed the memif maintainer.
Re: [dpdk-dev] [PATCH v5] ethdev: fix representor port ID search by name
On 9/30/21 3:51 PM, Singh, Aman Deep wrote: > > On 9/30/2021 5:33 PM, Andrew Rybchenko wrote: >> On 9/29/21 2:13 PM, Singh, Aman Deep wrote: >>> On 9/13/2021 4:56 PM, Andrew Rybchenko wrote: From: Viacheslav Galaktionov Getting a list of representors from a representor does not make sense. Instead, a parent device should be used. To this end, extend the rte_eth_dev_data structure to include the port ID of the backing device for representors. Signed-off-by: Viacheslav Galaktionov Signed-off-by: Andrew Rybchenko Acked-by: Haiyue Wang Acked-by: Beilei Xing --- The new field is added into the hole in rte_eth_dev_data structure. The patch does not change ABI, but extra care is required since ABI check is disabled for the structure because of the libabigail bug [1]. It should not be a problem anyway since 21.11 is a ABI breaking release. Potentially it is bad for out-of-tree drivers which implement representors but do not fill in a new parert_port_id field in rte_eth_dev_data structure. Get ID by name will not work. >>> Did we change name of new field from parert_port_id to backer_port_id. >> Yes, see v5 changelog below. >> It is done to address review notes from Ferruh on v4. > > Maybe I did not put it clearly, my bad. Just wanted, in above lines also > the usage > of "parert_port_id" should be changed. Thanks, I'll fix it in v6, but I think it does not worse to respin it since it is not a part of description. Just extra information. >> mlx5 changes should be reviwed by maintainers very carefully, since we are not sure if we patch it correctly. [1] https://sourceware.org/bugzilla/show_bug.cgi?id=28060 v5: - try to improve name: backer_port_id instead of parent_port_id - init new field to RTE_MAX_ETHPORTS on allocation to avoid zero port usage by default v4: - apply mlx5 review notes: remove fallback from generic ethdev code and add fallback to mlx5 code to handle legacy usecase v3: - fix mlx5 build breakage v2: - fix mlx5 review notes - try device port ID first before parent in order to address backward compatibility issue >> [snip] >>
Re: [dpdk-dev] [PATCH v2] drivers/net: remove queue xstats auto-fill flag
On 9/30/21 4:00 PM, Ferruh Yigit wrote:
> On 9/29/2021 7:38 AM, Andrew Rybchenko wrote:
>> On 9/28/21 8:10 PM, Stephen Hemminger wrote:
>>> On Tue, 28 Sep 2021 19:48:54 +0300
>>> Andrew Rybchenko wrote:
>>>
Some drivers do not provide per-queue statistics. So, there is no point
to have these misleading zeros in xstats.
Fixes: f30e69b41f94 ("ethdev: add device flag to bypass auto-filled queue
xstats")
Cc: sta...@dpdk.org
Signed-off-by: Andrew Rybchenko
>>> Really?
>>> It is useful to have zeros rather than random data there.
>>
>> I guess there is a misunderstanding here. Auto-filling xstats is
>> an addition of per-queue basic statistics to xstats by ethdev
>> layer. It makes sense to do it if and only if there is some
>> sensible data there.
>>
>> There is a related deprecation notice saying that per-queue
>> stats should be removed from basic stats since per-queue
>> stats should be provided by xstats API natively.
>>
>> Basically RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS means
>> that corresponding driver is not ready vs the deprecation notice.
>> So, I want to clean it up to see not yet ready drivers only.
>>
>
> As you said, 'RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS' flag set by driver means,
> driver is not ready on representing queue stats in xstats and ethdev layer is
> filling it automatically from basic stats.
>
> First we should wait for drivers to implement it, later clean queue stats from
> basic stats and remove the flag.
>
> I am not sure if we can remove the deprecation notice in this release, but
> agree
> to add a deadline for the drivers, which can be 22.11.
>
I'm going to cleanup deprecation. I don't touch it in the
patch. I just want to cleanup list of drivers which
require attention/changes. Drivers covered here do not
provide per-queue stats in basic stats. So, there is no
point to set the flag to show it in xstats.
Re: [dpdk-dev] [PATCH] app/testpmd: support unequal number of RXQ and TXQ
On 9/28/2021 12:08 PM, nipun.gu...@nxp.com wrote:
> From: Jun Yang
>
> The existing forwarding mode configures the total number of
> queues as the minimum of rxq and txq, so eventually the number
> of txq are same as rxq.
> However in some scenarios, specially for flow control the
> number of rxq and txq can be different.
> This patch makes the txq and function of rxq for all such
> scenario instead of keeping 1:1 relationship between the two.
>
Hi Nipun,
I expect more code exists that assumes Rx queue number and Tx queue number is
equal, did you able to test unequal queue numbers with below change?
And can you please describe your usecase more, is the device has uneven Rx/Tx
queues?
> Signed-off-by: Jun Yang
> ---
> app/test-pmd/config.c | 6 ++
> 1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
> index f5765b34f7..7e17f233ba 100644
> --- a/app/test-pmd/config.c
> +++ b/app/test-pmd/config.c
> @@ -3000,8 +3000,6 @@ rss_fwd_config_setup(void)
> int end;
>
> nb_q = nb_rxq;
> - if (nb_q > nb_txq)
> - nb_q = nb_txq;
> cur_fwd_config.nb_fwd_lcores = (lcoreid_t) nb_fwd_lcores;
> cur_fwd_config.nb_fwd_ports = nb_fwd_ports;
> cur_fwd_config.nb_fwd_streams =
> @@ -3038,7 +3036,7 @@ rss_fwd_config_setup(void)
> fs->rx_port = fwd_ports_ids[rxp];
> fs->rx_queue = rxq;
> fs->tx_port = fwd_ports_ids[txp];
> - fs->tx_queue = rxq;
> + fs->tx_queue = (rxq % nb_txq);
Is this assumes number of Rx queue is always more than number of Tx queue?
> fs->peer_addr = fs->tx_port;
> fs->retry_enabled = retry_enabled;
> rxp++;
> @@ -3253,7 +3251,7 @@ fwd_config_setup(void)
> return;
> }
>
> - if ((nb_rxq > 1) && (nb_txq > 1)){
> + if ((nb_rxq > 1) && (nb_txq > 1)) {
> if (dcb_config) {
> for (i = 0; i < nb_fwd_ports; i++) {
> pt_id = fwd_ports_ids[i];
>
[dpdk-dev] [PATCH] net/af_xdp: do not attempt probe for secondary processes
Since the AF_XDP PMD does not work for secondary processes as reported
in Bugzilla 805, check for the process type at the beginning of probe
and return ENOTSUP if the process type is secondary.
Bugzilla ID: 805
Fixes: f1debd77efaf ("net/af_xdp: introduce AF_XDP PMD")
Cc: sta...@dpdk.org
Signed-off-by: Ciara Loftus
Reported-by: Stephen Hemminger
---
drivers/net/af_xdp/rte_eth_af_xdp.c | 15 +--
1 file changed, 5 insertions(+), 10 deletions(-)
diff --git a/drivers/net/af_xdp/rte_eth_af_xdp.c
b/drivers/net/af_xdp/rte_eth_af_xdp.c
index 9bea0a895a..d61cb0aa7c 100644
--- a/drivers/net/af_xdp/rte_eth_af_xdp.c
+++ b/drivers/net/af_xdp/rte_eth_af_xdp.c
@@ -1790,16 +1790,11 @@ rte_pmd_af_xdp_probe(struct rte_vdev_device *dev)
rte_vdev_device_name(dev));
name = rte_vdev_device_name(dev);
- if (rte_eal_process_type() == RTE_PROC_SECONDARY &&
- strlen(rte_vdev_device_args(dev)) == 0) {
- eth_dev = rte_eth_dev_attach_secondary(name);
- if (eth_dev == NULL) {
- AF_XDP_LOG(ERR, "Failed to probe %s\n", name);
- return -EINVAL;
- }
- eth_dev->dev_ops = &ops;
- rte_eth_dev_probing_finish(eth_dev);
- return 0;
+ if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
+ AF_XDP_LOG(ERR, "Failed to probe %s. "
+ "AF_XDP PMD does not support secondary
processes.\n",
+ name);
+ return -ENOTSUP;
}
kvlist = rte_kvargs_parse(rte_vdev_device_args(dev), valid_arguments);
--
2.17.1
Re: [dpdk-dev] [PATCH v2] drivers/net: remove queue xstats auto-fill flag
On 9/30/21 4:45 PM, Andrew Rybchenko wrote:
> On 9/30/21 4:00 PM, Ferruh Yigit wrote:
>> On 9/29/2021 7:38 AM, Andrew Rybchenko wrote:
>>> On 9/28/21 8:10 PM, Stephen Hemminger wrote:
On Tue, 28 Sep 2021 19:48:54 +0300
Andrew Rybchenko wrote:
> Some drivers do not provide per-queue statistics. So, there is no point
> to have these misleading zeros in xstats.
>
> Fixes: f30e69b41f94 ("ethdev: add device flag to bypass auto-filled queue
> xstats")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Andrew Rybchenko
Really?
It is useful to have zeros rather than random data there.
>>>
>>> I guess there is a misunderstanding here. Auto-filling xstats is
>>> an addition of per-queue basic statistics to xstats by ethdev
>>> layer. It makes sense to do it if and only if there is some
>>> sensible data there.
>>>
>>> There is a related deprecation notice saying that per-queue
>>> stats should be removed from basic stats since per-queue
>>> stats should be provided by xstats API natively.
>>>
>>> Basically RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS means
>>> that corresponding driver is not ready vs the deprecation notice.
>>> So, I want to clean it up to see not yet ready drivers only.
>>>
>>
>> As you said, 'RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS' flag set by driver means,
>> driver is not ready on representing queue stats in xstats and ethdev layer is
>> filling it automatically from basic stats.
>>
>> First we should wait for drivers to implement it, later clean queue stats
>> from
>> basic stats and remove the flag.
>>
>> I am not sure if we can remove the deprecation notice in this release, but
>> agree
>> to add a deadline for the drivers, which can be 22.11.
>>
>
> I'm going to cleanup deprecation. I don't touch it in the
Sorry "I'm NOT going to ..."
> patch. I just want to cleanup list of drivers which
> require attention/changes. Drivers covered here do not
> provide per-queue stats in basic stats. So, there is no
> point to set the flag to show it in xstats.
>
Re: [dpdk-dev] [PATCH v6 1/2] Enable ASan for memory detector on DPDK
On 30-Sep-21 1:59 PM, zhihongx.p...@intel.com wrote:
From: Zhihong Peng
AddressSanitizer (ASan) is a google memory error detect
standard tool. It could help to detect use-after-free and
{heap,stack,global}-buffer overflow bugs in C/C++ programs,
print detailed error information when error happens, large
improve debug efficiency.
`AddressSanitizer
` (ASan)
is a widely-used debugging tool to detect memory access errors.
It helps detect issues like use-after-free, various kinds of buffer
overruns in C/C++ programs, and other similar errors, as well as
printing out detailed debug information whenever an error is detected.
DPDK ASan functionality is currently only supported Linux x86_64.
Support other platforms, need to define ASAN_SHADOW_OFFSET value
according to google ASan document.
Here is an example of heap-buffer-overflow bug:
..
char *p = rte_zmalloc(NULL, 7, 0);
p[7] = 'a';
..
Here is an example of use-after-free bug:
..
char *p = rte_zmalloc(NULL, 7, 0);
rte_free(p);
*p = 'a';
..
If you want to use this feature,
you need to add below compilation options when compiling code:
-Dbuildtype=debug -Db_lundef=false -Db_sanitize=address
"-Dbuildtype=debug": This is a non-essential option. When this option
is added, if a memory error occurs, ASan can clearly show where the
code is wrong.
"-Db_lundef=false": When use clang to compile DPDK, this option must
be added.
Signed-off-by: Xueqin Lin
Signed-off-by: Zhihong Peng
---
Acked-by: Anatoly Burakov
--
Thanks,
Anatoly
Re: [dpdk-dev] [PATCH v2 4/5] kvargs: remove experimental function to compare string
On Wed, Sep 29, 2021 at 11:39:42PM +0200, Olivier Matz wrote: > The function was designed to be used as a handler for > rte_kvargs_process() to compare the value string in a kvlist. For > readability, its usages in DPDK have been replaced by > rte_kvargs_get_with_value() in previous commit. > > Remove this function, as it is not used anymore. > > Signed-off-by: Olivier Matz > Reviewed-by: Xueming Li Maybe a quick note could be added in the release note. Something like this: * kvargs: The experimental function ``rte_kvargs_strcmp()`` has been removed. Its usages have been replaced by a new function ``rte_kvargs_get_with_value()``. David, do you want me to send a v3 with this note? Thanks, Olivier
Re: [dpdk-dev] [dpdk-stable] [PATCH v5 2/2] ethdev: fix docs of drivers callbacks getting xstats by IDs
On 9/30/21 3:08 PM, Ferruh Yigit wrote:
> On 9/29/2021 12:54 PM, Andrew Rybchenko wrote:
>> On 9/29/21 11:44 AM, Ferruh Yigit wrote:
>>> On 9/28/2021 5:53 PM, Andrew Rybchenko wrote:
On 9/28/21 7:50 PM, Ferruh Yigit wrote:
> On 9/28/2021 1:05 PM, Andrew Rybchenko wrote:
>> From: Ivan Ilchenko
>>
>> Update xstats by IDs callbacks documentation in accordance with
>> ethdev usage of these callbacks. Document valid combinations of
>> input arguments to make driver implementation simpler.
>>
>> Fixes: 79c913a42f0 ("ethdev: retrieve xstats by ID")
>> Cc: sta...@dpdk.org
>>
>> Signed-off-by: Ivan Ilchenko
>> Signed-off-by: Andrew Rybchenko
>> Reviewed-by: Andy Moreton
>> ---
>> lib/ethdev/ethdev_driver.h | 42 --
>> 1 file changed, 40 insertions(+), 2 deletions(-)
>>
>> diff --git a/lib/ethdev/ethdev_driver.h b/lib/ethdev/ethdev_driver.h
>> index 40e474aa7e..c89eefcc42 100644
>> --- a/lib/ethdev/ethdev_driver.h
>> +++ b/lib/ethdev/ethdev_driver.h
>> @@ -187,11 +187,28 @@ typedef int (*eth_xstats_get_t)(struct rte_eth_dev
>> *dev,
>> struct rte_eth_xstat *stats, unsigned int n);
>> /**< @internal Get extended stats of an Ethernet device. */
>>
>> +/**
>> + * @internal
>> + * Get extended stats of an Ethernet device.
>> + *
>> + * @param dev
>> + * ethdev handle of port.
>> + * @param ids
>> + * IDs array to retrieve specific statistics. Must not be NULL.
>> + * @param values
>> + * A pointer to a table to be filled with device statistics values.
>> + * Must not be NULL.
>> + * @param n
>> + * Element count in @p ids and @p values.
>> + *
>> + * @return
>> + * - A number of filled in stats.
>> + * - A negative value on error.
>> + */
>> typedef int (*eth_xstats_get_by_id_t)(struct rte_eth_dev *dev,
>>const uint64_t *ids,
>>uint64_t *values,
>>unsigned int n);
>> -/**< @internal Get extended stats of an Ethernet device. */
>>
>> /**
>> * @internal
>> @@ -218,10 +235,31 @@ typedef int (*eth_xstats_get_names_t)(struct
>> rte_eth_dev *dev,
>> struct rte_eth_xstat_name *xstats_names, unsigned int size);
>> /**< @internal Get names of extended stats of an Ethernet device. */
>>
>> +/**
>> + * @internal
>> + * Get names of extended stats of an Ethernet device.
>> + * For name count, set @p xstats_names and @p ids to NULL.
>
> Why limiting this behavior to 'xstats_get_names_by_id'?
>
> Internally 'xstats_get_names_by_id' is used to get the count, but I think
> this
> is not intentionally selected, just one of the xstats_*_by_id dev_ops
> used.
>
> I think it is confusing to have this support for one of the '_by_id'
> dev_ops but
> not for other. Why not require both to support returning 'count'?
Simply because it is dead code. There is no point to require
from driver to have dead code.
>>>
>>> Let me step back a little, both ethdev APIs can be used to return xstats
>>> count
>>> by providing 'values/names' & 'ids' pointers as NULL and 'size' as 0:
>>> 'rte_eth_xstats_get_names_by_id()'
>>> 'rte_eth_xstats_get_by_id()'
>>>
>>> But internally both APIs use 'xstats_get_names_by_id' dev_ops to get the
>>> count,
>>> as said above I believe this selection is done unintentionally.
>>>
>>>
>>> I am for below two options:
>>>
>>> a) Internally use 'xstats_get_names' || 'xstats_get' dev_ops to get the
>>> xstats
>>> count, and doesn't support getting xstats count for both '_by_id' dev_ops,
>>> this
>>> simplifies driver code. As far as I remember I suggested this before, still
>>> I
>>> prefer this one.
>>>
>>> b) If we will support getting xstats count from '_by_id' dev_ops, I think
>>> both
>>> should support it, to not make it more complex to figure out which one
>>> support
>>> what. As sample both 'xstats_get_names' and 'xstats_get' supports getting
>>> xstats
>>> count, not just one.
>>>
>>
>> In (b) do you suggest to change ethdev to use xstats_get_by_id
>> driver op if users asks for a number of xstats using
>> rte_eth_xstats_get_by_id()? It will complicate ethdev and
>> will complicate drivers. Just for the symmetry?
>>
>
> Not just for symmetry, but simplify the logic. Both dev_ops are very similar
> and
I'm sorry, but could you point of which logic you'd
like to simply. Less requirements on driver ops
means less code required inside.
> they are having different behavior with same parameters is confusing.
Ah, logic from PMD maintainer point of view. Does it
really worse to require extra code inside because of it?
> If you want to simplify the drivers why not go with option (a)? Option (a)
> als
[dpdk-dev] [PATCH v6 1/4] ethdev: fix docs of functions getting xstats by IDs
From: Ivan Ilchenko
Document valid combinations of input arguments in accordance with
current implementation in ethdev.
Fixes: 79c913a42f0 ("ethdev: retrieve xstats by ID")
Cc: sta...@dpdk.org
Signed-off-by: Ivan Ilchenko
Signed-off-by: Andrew Rybchenko
Reviewed-by: Andy Moreton
---
lib/ethdev/rte_ethdev.h | 35 +++
1 file changed, 19 insertions(+), 16 deletions(-)
diff --git a/lib/ethdev/rte_ethdev.h b/lib/ethdev/rte_ethdev.h
index afdc53b674..2ad2bf2019 100644
--- a/lib/ethdev/rte_ethdev.h
+++ b/lib/ethdev/rte_ethdev.h
@@ -2896,21 +2896,23 @@ int rte_eth_xstats_get(uint16_t port_id, struct
rte_eth_xstat *xstats,
* @param port_id
* The port identifier of the Ethernet device.
* @param xstats_names
- * An rte_eth_xstat_name array of at least *size* elements to
- * be filled. If set to NULL, the function returns the required number
- * of elements.
- * @param ids
- * IDs array given by app to retrieve specific statistics
+ * Array to be filled in with names of requested device statistics.
+ * Must not be NULL if @p ids are specified (not NULL).
* @param size
- * The size of the xstats_names array (number of elements).
+ * Number of elements in @p xstats_names array (if not NULL) and in
+ * @p ids array (if not NULL). Must be 0 if both array pointers are NULL.
+ * @param ids
+ * IDs array given by app to retrieve specific statistics. May be NULL to
+ * retrieve names of all available statistics or, if @p xstats_names is
+ * NULL as well, just the number of available statistics.
* @return
* - A positive value lower or equal to size: success. The return value
* is the number of entries filled in the stats table.
- * - A positive value higher than size: error, the given statistics table
+ * - A positive value higher than size: success. The given statistics table
* is too small. The return value corresponds to the size that should
* be given to succeed. The entries in the table are not valid and
* shall not be used by the caller.
- * - A negative value on error (invalid port id).
+ * - A negative value on error.
*/
int
rte_eth_xstats_get_names_by_id(uint16_t port_id,
@@ -2923,22 +2925,23 @@ rte_eth_xstats_get_names_by_id(uint16_t port_id,
* @param port_id
* The port identifier of the Ethernet device.
* @param ids
- * A pointer to an ids array passed by application. This tells which
- * statistics values function should retrieve. This parameter
- * can be set to NULL if size is 0. In this case function will retrieve
- * all available statistics.
+ * IDs array given by app to retrieve specific statistics. May be NULL to
+ * retrieve all available statistics or, if @p values is NULL as well,
+ * just the number of available statistics.
* @param values
- * A pointer to a table to be filled with device statistics values.
+ * Array to be filled in with requested device statistics.
+ * Must not be NULL if ids are specified (not NULL).
* @param size
- * The size of the ids array (number of elements).
+ * Number of elements in @p values array (if not NULL) and in @p ids
+ * array (if not NULL). Must be 0 if both array pointers are NULL.
* @return
* - A positive value lower or equal to size: success. The return value
* is the number of entries filled in the stats table.
- * - A positive value higher than size: error, the given statistics table
+ * - A positive value higher than size: success: The given statistics table
* is too small. The return value corresponds to the size that should
* be given to succeed. The entries in the table are not valid and
* shall not be used by the caller.
- * - A negative value on error (invalid port id).
+ * - A negative value on error.
*/
int rte_eth_xstats_get_by_id(uint16_t port_id, const uint64_t *ids,
uint64_t *values, unsigned int size);
--
2.30.2
