[dpdk-dev] [PATCH] net/mlx5: fix invalid count in xstats
With the commit af4f09f28294 ("net/mlx5: prefix all functions with mlx5"),
mlx5_xstats_get() is not compliant any longer with the api.
It always returns the caller max entries count while it should return how
many entries it wrote/wanted to write.
Fixes: af4f09f28294 ("net/mlx5: prefix all functions with mlx5")
Signed-off-by: David Marchand
---
drivers/net/mlx5/mlx5_stats.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/mlx5/mlx5_stats.c b/drivers/net/mlx5/mlx5_stats.c
index 8e427e7..875dd10 100644
--- a/drivers/net/mlx5/mlx5_stats.c
+++ b/drivers/net/mlx5/mlx5_stats.c
@@ -325,7 +325,7 @@ mlx5_xstats_get(struct rte_eth_dev *dev, struct
rte_eth_xstat *stats,
stats[i].value = (counters[i] - xstats_ctrl->base[i]);
}
}
- return n;
+ return xstats_n;
}
/**
--
2.7.4
Re: [dpdk-dev] [RFC v2] vhost: new rte_vhost API proposal
On Fri, May 18, 2018 at 03:50:45PM +0200, Maxime Coquelin wrote: > Hi Dariusz, > > On 05/18/2018 03:01 PM, Dariusz Stojaczyk wrote: > > rte_vhost is not vhost-user spec compliant. Some Vhost drivers have > > been already confirmed not to work with rte_vhost. virtio-user-scsi-pci > > in QEMU 2.12 doesn't fully initialize its management queues at SeaBIOS > > stage. This is perfectly fine from the Vhost-user spec perspective, but > > doesn't meet rte_vhost expectations. rte_vhost waits for all queues > > to be fully initialized before it allows the entire device to be > > processed. qFixing rte_vhost directly would require quite a big amount > > of changes, which would completely break backwards compatibility. > > > > This rte_vhost2 library is intended to smooth out the transition. > > It exposes a low-level API for implementing new Vhost-user slaves. > > The existing rte_vhost is about to be refactored to use rte_vhost2 > > library underneath, and demanding backends could now use rte_vhost2 > > directly. > > I like the idea, and the proposed way to smooth the transition. Please be aware of that I just had a quick glimpse of this patch and it's likely I don't have too much time to follow this. However, I also like this idea. And thank you for working on it. --yliu
Re: [dpdk-dev] release date for 17.11.3?
On Thu, May 17, 2018 at 10:11:40PM +0200, Thomas Monjalon wrote: > 17/05/2018 18:16, Montorsi, Francesco: > > Hi all, > > I'd like to build DPDK on Centos 7.5... I found a commit > > (http://dpdk.org/browse/dpdk-stable/commit/?h=17.11&id=3ee847054cc9ab62fa2c9c6dc6ba68899d620e3a) > > that allows that, but it will go into DPDK 17.11.3 right? > > Is there a release date for 17.11.3 already defined? > > Hopefully June 8th: > http://dpdk.org/dev/roadmap#stable Yes, something like that. I probably will make it a little bit earlier than that. --yliu > > +Cc Yuanhan
[dpdk-dev] [PATCH] eventdev: remove ethdev Rx adapter event type
The initial thought was to create a separate event type for each adapter, in order to make consistency with new event adapters, remove RTE_EVENT_TYPE_ETH_RX_ADAPTER event type and use RTE_EVENT_TYPE_ETHDEV event type instead like other event adapters. Cc: Nikhil Rao Signed-off-by: Jerin Jacob --- lib/librte_eventdev/rte_event_eth_rx_adapter.c | 2 +- lib/librte_eventdev/rte_eventdev.h | 2 -- 2 files changed, 1 insertion(+), 3 deletions(-) diff --git a/lib/librte_eventdev/rte_event_eth_rx_adapter.c b/lib/librte_eventdev/rte_event_eth_rx_adapter.c index 6f7050952..8b9887b3d 100644 --- a/lib/librte_eventdev/rte_event_eth_rx_adapter.c +++ b/lib/librte_eventdev/rte_event_eth_rx_adapter.c @@ -468,7 +468,7 @@ fill_event_buffer(struct rte_event_eth_rx_adapter *rx_adapter, ev->op = RTE_EVENT_OP_NEW; ev->sched_type = sched_type; ev->queue_id = qid; - ev->event_type = RTE_EVENT_TYPE_ETH_RX_ADAPTER; + ev->event_type = RTE_EVENT_TYPE_ETHDEV; ev->sub_event_type = 0; ev->priority = priority; ev->mbuf = m; diff --git a/lib/librte_eventdev/rte_eventdev.h b/lib/librte_eventdev/rte_eventdev.h index b6fd6ee7f..cef0a55d8 100644 --- a/lib/librte_eventdev/rte_eventdev.h +++ b/lib/librte_eventdev/rte_eventdev.h @@ -948,8 +948,6 @@ rte_event_dev_close(uint8_t dev_id); /**< The event generated from cpu for pipelining. * Application may use *sub_event_type* to further classify the event */ -#define RTE_EVENT_TYPE_ETH_RX_ADAPTER 0x4 -/**< The event generated from event eth Rx adapter */ #define RTE_EVENT_TYPE_MAX 0x10 /**< Maximum number of event types */ -- 2.17.0
Re: [dpdk-dev] [PATCH] eventdev: remove ethdev Rx adapter event type
-Original Message- > Date: Sun, 20 May 2018 22:18:43 +0530 > From: Jerin Jacob > To: dev@dpdk.org > Cc: Jerin Jacob , Nikhil Rao > > Subject: [dpdk-dev] [PATCH] eventdev: remove ethdev Rx adapter event type > X-Mailer: git-send-email 2.17.0 > > The initial thought was to create a separate event type > for each adapter, in order to make consistency with > new event adapters, remove RTE_EVENT_TYPE_ETH_RX_ADAPTER > event type and use RTE_EVENT_TYPE_ETHDEV event type instead > like other event adapters. > > Cc: Nikhil Rao > Signed-off-by: Jerin Jacob Self - NACK. external PCIe NW cards + HW eventdev case will failing this case, as RTE_EVENT_TYPE_ETHDEV identifier was used for converting to mbuf in eventdev_dequeue case. > --- > lib/librte_eventdev/rte_event_eth_rx_adapter.c | 2 +- > lib/librte_eventdev/rte_eventdev.h | 2 -- > 2 files changed, 1 insertion(+), 3 deletions(-) > > diff --git a/lib/librte_eventdev/rte_event_eth_rx_adapter.c > b/lib/librte_eventdev/rte_event_eth_rx_adapter.c > index 6f7050952..8b9887b3d 100644 > --- a/lib/librte_eventdev/rte_event_eth_rx_adapter.c > +++ b/lib/librte_eventdev/rte_event_eth_rx_adapter.c > @@ -468,7 +468,7 @@ fill_event_buffer(struct rte_event_eth_rx_adapter > *rx_adapter, > ev->op = RTE_EVENT_OP_NEW; > ev->sched_type = sched_type; > ev->queue_id = qid; > - ev->event_type = RTE_EVENT_TYPE_ETH_RX_ADAPTER; > + ev->event_type = RTE_EVENT_TYPE_ETHDEV; > ev->sub_event_type = 0; > ev->priority = priority; > ev->mbuf = m; > diff --git a/lib/librte_eventdev/rte_eventdev.h > b/lib/librte_eventdev/rte_eventdev.h > index b6fd6ee7f..cef0a55d8 100644 > --- a/lib/librte_eventdev/rte_eventdev.h > +++ b/lib/librte_eventdev/rte_eventdev.h > @@ -948,8 +948,6 @@ rte_event_dev_close(uint8_t dev_id); > /**< The event generated from cpu for pipelining. > * Application may use *sub_event_type* to further classify the event > */ > -#define RTE_EVENT_TYPE_ETH_RX_ADAPTER 0x4 > -/**< The event generated from event eth Rx adapter */ > #define RTE_EVENT_TYPE_MAX 0x10 > /**< Maximum number of event types */ > > -- > 2.17.0 >
Re: [dpdk-dev] [PATCH v2] app/test: enhance sanity script
Hello Thomas, -Original Message- From: Thomas Monjalon [mailto:tho...@monjalon.net] Sent: Friday, May 18, 2018 7:18 PM To: Parthasarathy, JananeeX M Cc: dev@dpdk.org; Burakov, Anatoly ; Pattan, Reshma Subject: Re: [dpdk-dev] [PATCH v2] app/test: enhance sanity script 18/05/2018 15:32, Parthasarathy, JananeeX M: > From: Thomas Monjalon [mailto:tho...@monjalon.net] > > 15/05/2018 18:14, Parthasarathy, JananeeX M: > > > From: Thomas Monjalon [mailto:tho...@monjalon.net] > > > > I think we need to keep a fast subset of tests. > > > > Please check in mk/rte.sdktest.mk: > > > > PERFLIST=ring_perf,mempool_perf,memcpy_perf,hash_perf,timer_perf > > test-fast: BLACKLIST=-$(PERFLIST) > > We have added the below test cases to blacklist in mk/rte.sdktest.mk > reciprocal_division,reciprocal_division_perf,lpm_perf,red_all > > Now time taken is around 4mins. > Please let us know is there any minimum time limit for test execution. > Is this time duration fine? 4 minutes, it's a bit long for developer oriented tests. We need to distinguish which tests are enough for quick tests, and which ones can be in another category for full validation purpose. Please can you make a summary of which tests you are adding, and propose a priority classification. Thanks Please find below the test cases newly added to sanity script (autotest_data.py) with priority classification. P1 - Main test cases P2 - Cryptodev test cases which requires dependent libraries (Can be added if the environment is set accordingly) P3 - Perf test cases and others which takes more time (> 10s) (these will be in the blacklist test cases) P4 - Logging/Dump test cases (to be checked whether these are really necessary) The blacklist is updated in makefile "rte.sdktest.mk" with below list of test cases to reduce the total test execution time. Blacklist test cases are ring_perf,mempool_perf,memcpy_perf,hash_perf,timer_perf,reciprocal_division,reciprocal_division_perf,lpm_perf, red_all,barrier,hash_multiwriter,timer_racecond,efd,hash_functions,eventdev_selftest_sw,member_perf,efd_perf,lpm6_perf,red_perf,distributor_perf Currently after the above blacklist update, the total execution time of "make test-fast" is 54s. Eventdev selftest octeontx P1 Event ring autotest P1 Table autotest P1 Flow classify autotest P1 Dump struct sizes P4 Dump mempool P4 Dump malloc stats P4 Dump devargs P4 Dump log types P4 Dump_ring P4 Quit P4 Set rxtx anchor P4 Set rxtx sc P4 Event eth rx adapter autotest P1 Rawdev autotest P1 Kvargs autotest P1 Dump physmem P1 Dump memzone P4 Devargs autotest P4 Link bonding autotest P1 Link bonding mode4 autotest P1 Link bonding rssconf autotest P1 Crc autotest P1 Distributor autotest P1 Reorder autotest P1 Barrier autotest P3 User delay P1 Bitmap test P1 Hash scaling autotest P1 Hash multiwriter autotest P3 Service autotest P1 Timer racecond autotest P3 Member autotest P3 Efd_autotest P3 Thash autotest P1 Hash function autotest P3 Cryptodev sw mrvl autotest P2 Cryptodev dpaa2 sec autotest P2 Cryptodev dpaa sec autotest P2 Cryptodev qat autotest P2 Cryptodev aesni mb autotest P2 Cryptodev openssl autotest P2 Cryptodev scheduler autotest P2 Cryptodev aesni gcm autotest P2 Cryptodev null autotest P2 Cryptodev sw snow3g autotest P2 Cryptodev sw kasumi autotest P2 Cryptodev_sw_zuc_autotest P2 Reciprocal division P3 Red all P3 Mempool performance autotest P3 Memcpy performance autotest P3 Hash performance autotest P3 Timer performance autotest P3 Reciprocal division perf P3 Member perf autotest P3 Efd perf autotest P3 Lpm perf autotest P3 Lpm6 perf autotest P3 Red_perf P3 Distributor perf autotest P3 Ring pmd perf autotest P1 Pmd perf autotest P1 Ring performance autotest P1 Thanks M.P.Jananee -- Intel Research and Development Ireland Limited Registered in Ireland Registered Office: Collinstown Industrial Park, Leixlip, County Kildare Registered Number: 308263 This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
Re: [dpdk-dev] [PATCH v2] app/test: enhance sanity script
20/05/2018 19:41, Parthasarathy, JananeeX M: > From: Thomas Monjalon [mailto:tho...@monjalon.net] > > 18/05/2018 15:32, Parthasarathy, JananeeX M: > > > From: Thomas Monjalon [mailto:tho...@monjalon.net] > > > > 15/05/2018 18:14, Parthasarathy, JananeeX M: > > > > > From: Thomas Monjalon [mailto:tho...@monjalon.net] > > > > > > > > I think we need to keep a fast subset of tests. > > > > > > > > Please check in mk/rte.sdktest.mk: > > > > PERFLIST=ring_perf,mempool_perf,memcpy_perf,hash_perf,ti > > > > mer_perf > > > > > > > > test-fast: BLACKLIST=-$(PERFLIST) > > > > > > We have added the below test cases to blacklist in mk/rte.sdktest.mk > > > reciprocal_division,reciprocal_division_perf,lpm_perf,red_all > > > > > > Now time taken is around 4mins. > > > Please let us know is there any minimum time limit for test execution. > > > Is this time duration fine? > > > > 4 minutes, it's a bit long for developer oriented tests. > > > > We need to distinguish which tests are enough for quick tests, > > and which ones can be in another category for full validation purpose. > > > > Please can you make a summary of which tests you are adding, > > and propose a priority classification. > > > > Thanks > > Please find below the test cases newly added to sanity script > (autotest_data.py) with priority classification. > > P1 - Main test cases We can move some tests from P1. Examples: - bonding can be P2 with other drivers. - PMD perf and ring perf can P3 with other perf tests. > P2 - Cryptodev test cases which requires dependent libraries (Can be added > if the environment is set accordingly) > P3 - Perf test cases and others which takes more time (> 10s) (these will be > in the blacklist test cases) > P4 - Logging/Dump test cases (to be checked whether these are really > necessary) I think this classification is good. We can adjust the categories in the makefile to reflect it. The goal is to have a really quick and core subset (equal P1 category), and other categories available with a different make command (drivers for P2, perf for P3 and debug for P4). > The blacklist is updated in makefile "rte.sdktest.mk" with below list > of test cases to reduce the total test execution time. > > Blacklist test cases are > ring_perf,mempool_perf,memcpy_perf,hash_perf,timer_perf, > reciprocal_division,reciprocal_division_perf,lpm_perf,red_all,barrier, > hash_multiwriter,timer_racecond,efd,hash_functions,eventdev_selftest_sw, > member_perf,efd_perf,lpm6_perf,red_perf,distributor_perf > > Currently after the above blacklist update, the total execution time of "make > test-fast" is 54s. Less than a minute is a good limit! Maybe we can do even better with above suggestions. Thanks for working on it. Feel free to send a new patch when you think it is ready. We can try to have the new autotests when starting 18.08. Then we should communicate to enforce a better coverage of the autotests.
Re: [dpdk-dev] mellanox connectx-2 support
чт, 17 мая 2018 г. в 9:02, Olga Shern : > Vasily, > > ConnectX-2 is very, very old card. > We don't support it. > You can run DPDK on ConnectX-3 Pro, ConnectX-4 /LX and ConnectX-5 NICs. > The best performance and better feature set you will get with ConnectX-5 > > Best Regards, > Thanks for info, web page says about connect-x3 (without pro) If i have Connect-X3 vpi does it works with intel dpdk mellanox driver? -- Vasiliy Tolstov, e-mail: v.tols...@selfip.ru
Re: [dpdk-dev] [PATCH] pci: remove unused function
Hi,
Looks good to me. I noticed that the same is also with
the rte_pci_detach() function, and I will send a patch to
remove it as well.
Regards,
Rami Rosen
-Original Message-
From: dev [mailto:dev-boun...@dpdk.org] On Behalf Of Stephen Hemminger
Sent: Friday, May 18, 2018 02:45
To: dev@dpdk.org
Cc: Stephen Hemminger
Subject: [dpdk-dev] [PATCH] pci: remove unused function
The function rte_pci_probe_one is private to PCI and is not used anywhere in
current code base. Remove dead code.
Signed-off-by: Stephen Hemminger
---
drivers/bus/pci/pci_common.c | 38
drivers/bus/pci/private.h| 15 --
2 files changed, 53 deletions(-)
diff --git a/drivers/bus/pci/pci_common.c b/drivers/bus/pci/pci_common.c index
7215aaec3add..3d21d7373b60 100644
--- a/drivers/bus/pci/pci_common.c
+++ b/drivers/bus/pci/pci_common.c
@@ -254,44 +254,6 @@ pci_probe_all_drivers(struct rte_pci_device *dev)
return 1;
}
-/*
- * Find the pci device specified by pci address, then invoke probe function of
- * the driver of the device.
- */
-int
-rte_pci_probe_one(const struct rte_pci_addr *addr) -{
- struct rte_pci_device *dev = NULL;
-
- int ret = 0;
-
- if (addr == NULL)
- return -1;
-
- /* update current pci device in global list, kernel bindings might have
-* changed since last time we looked at it.
-*/
- if (pci_update_device(addr) < 0)
- goto err_return;
-
- FOREACH_DEVICE_ON_PCIBUS(dev) {
- if (rte_pci_addr_cmp(&dev->addr, addr))
- continue;
-
- ret = pci_probe_all_drivers(dev);
- if (ret)
- goto err_return;
- return 0;
- }
- return -1;
-
-err_return:
- RTE_LOG(WARNING, EAL,
- "Requested device " PCI_PRI_FMT " cannot be used\n",
- addr->domain, addr->bus, addr->devid, addr->function);
- return -1;
-}
-
/*
* Detach device specified by its pci address.
*/
diff --git a/drivers/bus/pci/private.h b/drivers/bus/pci/private.h index
88fa587e74dc..748abb6617fe 100644
--- a/drivers/bus/pci/private.h
+++ b/drivers/bus/pci/private.h
@@ -32,21 +32,6 @@ rte_pci_probe(void);
*/
int rte_pci_scan(void);
-/**
- * Probe the single PCI device.
- *
- * Scan the content of the PCI bus, and find the pci device specified by pci
- * address, then call the probe() function for registered driver that has a
- * matching entry in its id_table for discovered device.
- *
- * @param addr
- * The PCI Bus-Device-Function address to probe.
- * @return
- * - 0 on success.
- * - Negative on error.
- */
-int rte_pci_probe_one(const struct rte_pci_addr *addr);
-
/**
* Close the single PCI device.
*
--
2.17.0
[dpdk-dev] [PATCH] pci: remove unused function
The function rte_pci_detach() is private to PCI and is
not used anywhere in current code base. Remove dead code.
Signed-off-by: Rami Rosen
---
drivers/bus/pci/pci_common.c | 36
drivers/bus/pci/private.h| 14 --
2 files changed, 50 deletions(-)
diff --git a/drivers/bus/pci/pci_common.c b/drivers/bus/pci/pci_common.c
index 7215aae..e6d2c1c 100644
--- a/drivers/bus/pci/pci_common.c
+++ b/drivers/bus/pci/pci_common.c
@@ -292,42 +292,6 @@ static struct rte_devargs *pci_devargs_lookup(struct
rte_pci_device *dev)
return -1;
}
-/*
- * Detach device specified by its pci address.
- */
-int
-rte_pci_detach(const struct rte_pci_addr *addr)
-{
- struct rte_pci_device *dev = NULL;
- int ret = 0;
-
- if (addr == NULL)
- return -1;
-
- FOREACH_DEVICE_ON_PCIBUS(dev) {
- if (rte_pci_addr_cmp(&dev->addr, addr))
- continue;
-
- ret = rte_pci_detach_dev(dev);
- if (ret < 0)
- /* negative value is an error */
- goto err_return;
- if (ret > 0)
- /* positive value means driver doesn't support it */
- continue;
-
- rte_pci_remove_device(dev);
- free(dev);
- return 0;
- }
- return -1;
-
-err_return:
- RTE_LOG(WARNING, EAL, "Requested device " PCI_PRI_FMT
- " cannot be used\n", dev->addr.domain, dev->addr.bus,
- dev->addr.devid, dev->addr.function);
- return -1;
-}
/*
* Scan the content of the PCI bus, and call the probe() function for
diff --git a/drivers/bus/pci/private.h b/drivers/bus/pci/private.h
index 88fa587..63bd876 100644
--- a/drivers/bus/pci/private.h
+++ b/drivers/bus/pci/private.h
@@ -47,20 +47,6 @@
*/
int rte_pci_probe_one(const struct rte_pci_addr *addr);
-/**
- * Close the single PCI device.
- *
- * Scan the content of the PCI bus, and find the pci device specified by pci
- * address, then call the remove() function for registered driver that has a
- * matching entry in its id_table for discovered device.
- *
- * @param addr
- * The PCI Bus-Device-Function address to close.
- * @return
- * - 0 on success.
- * - Negative on error.
- */
-int rte_pci_detach(const struct rte_pci_addr *addr);
/**
* Find the name of a PCI device.
--
1.9.1
Re: [dpdk-dev] [PATCH] net/cxgbevf: fix inter-vm traffic when physical link down
On 5/18/2018 7:59 PM, Rahul Lakkireddy wrote:
> From: Shagun Agrawal
>
> Add force_link_up devargs to always force link as up for VFs.
> This enables VFs on the same NIC to send traffic to each other
> even when physical link is down.
>
> Fixes: 011ebc236ddc ("net/cxgbe: add skeleton VF driver")
>
> Signed-off-by: Shagun Agrawal
> Signed-off-by: Kumar Sanghvi
> Signed-off-by: Rahul Lakkireddy
> ---
> doc/guides/nics/cxgbe.rst| 6 ++
> drivers/net/cxgbe/cxgbe.h| 4
> drivers/net/cxgbe/cxgbe_ethdev.c | 14 ++
> drivers/net/cxgbe/cxgbe_main.c | 17 -
> 4 files changed, 32 insertions(+), 9 deletions(-)
>
> diff --git a/doc/guides/nics/cxgbe.rst b/doc/guides/nics/cxgbe.rst
> index 38d434802..78e391473 100644
> --- a/doc/guides/nics/cxgbe.rst
> +++ b/doc/guides/nics/cxgbe.rst
> @@ -143,6 +143,12 @@ be passed as part of EAL arguments. For example,
>enabled, the outer VLAN tag is preserved in Q-in-Q packets. Otherwise,
>the outer VLAN tag is stripped in Q-in-Q packets.
>
> +- ``force_link_up`` (default **0**)
> +
> + When set to 1, CXGBEVF PMD always forces link as up for all VFs on
> + underlying Chelsio NICs. This enables multiple VFs on the same NIC
> + to send traffic to each other even when the physical link is down.
> +
> .. _driver-compilation:
>
> Driver compilation and testing
> diff --git a/drivers/net/cxgbe/cxgbe.h b/drivers/net/cxgbe/cxgbe.h
> index b1a4e08df..75e0b21a8 100644
> --- a/drivers/net/cxgbe/cxgbe.h
> +++ b/drivers/net/cxgbe/cxgbe.h
> @@ -25,6 +25,9 @@
> ETH_RSS_NONFRAG_IPV6_TCP | \
> ETH_RSS_NONFRAG_IPV6_UDP)
>
> +#define CXGBE_DEVARG_FORCE_LINK_UP "force_link_up"
You may consider to move existing "CXGBE_DEVARG_KEEP_OVLAN" to here, to keep
them together.
Also can you please add "RTE_PMD_REGISTER_PARAM_STRING" to document devargs,
please put it below RTE_PMD_REGISTER_KMOD_DEP.
> +
> +bool force_linkup(struct adapter *adap);
> int cxgbe_probe(struct adapter *adapter);
> int cxgbevf_probe(struct adapter *adapter);
> void cxgbe_get_speed_caps(struct port_info *pi, u32 *speed_caps);
> @@ -45,5 +48,6 @@ int setup_rss(struct port_info *pi);
> void cxgbe_enable_rx_queues(struct port_info *pi);
> void print_port_info(struct adapter *adap);
> void print_adapter_info(struct adapter *adap);
> +int cxgbe_get_devargs(struct rte_devargs *devargs, const char *key);
>
> #endif /* _CXGBE_H_ */
> diff --git a/drivers/net/cxgbe/cxgbe_ethdev.c
> b/drivers/net/cxgbe/cxgbe_ethdev.c
> index 3ee7c44b1..2f1e4f691 100644
> --- a/drivers/net/cxgbe/cxgbe_ethdev.c
> +++ b/drivers/net/cxgbe/cxgbe_ethdev.c
> @@ -204,19 +204,17 @@ int cxgbe_dev_link_update(struct rte_eth_dev *eth_dev,
> struct port_info *pi = (struct port_info *)(eth_dev->data->dev_private);
> struct adapter *adapter = pi->adapter;
> struct sge *s = &adapter->sge;
> - struct rte_eth_link *old_link = ð_dev->data->dev_link;
> + struct rte_eth_link new_link;
> unsigned int work_done, budget = 4;
>
> cxgbe_poll(&s->fw_evtq, NULL, budget, &work_done);
> - if (old_link->link_status == pi->link_cfg.link_ok)
> - return -1; /* link not changed */
>
> - eth_dev->data->dev_link.link_status = pi->link_cfg.link_ok;
> - eth_dev->data->dev_link.link_duplex = ETH_LINK_FULL_DUPLEX;
> - eth_dev->data->dev_link.link_speed = pi->link_cfg.speed;
> + new_link.link_status = force_linkup(adapter) ?
> +ETH_LINK_UP : pi->link_cfg.link_ok;
> + new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
> + new_link.link_speed = pi->link_cfg.speed;
>
> - /* link has changed */
> - return 0;
> + return rte_eth_linkstatus_set(eth_dev, &new_link);
+1 to use rte_eth_linkstatus_set()
Re: [dpdk-dev] [PATCH 1/4] net/qede: fix multicast filtering
On 5/20/2018 1:15 AM, Rasesh Mody wrote:
> From: Harish Patil
>
> This patch is to fix multicast filtering using set_mc_addr_list().
>
> Fixes: 77fac1b54fc9 ("net/qede: fix filtering code")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Harish Patil
> Signed-off-by: Shahed Shaikh
> Signed-off-by: Rasesh Mody
Series applied to dpdk-next-net/master, thanks.
Re: [dpdk-dev] [PATCH v5 00/21] Fixes for GCC8 against lagopus
> Andy Green (21): > lib/librte_ethdev: change eth-dev-ops API to return int > rte_string_fns.h: fix gcc8.1 sign conv warning in lstrcpy > lib/librte_eal: explicit tmp cast > /lib/librte_eal: stage cast from uint64 to long > rte_ring_generic.h: stack declarations before code > rte_ring.h: remove signed type flipflopping > rte_mbuf.h: avoid warnings from inadvertant promotion > rte_mbuf.h: explicit casts for int16 to uint16 > rte_mbuf.h: make sure RTE-MIN compares same types > rte_mbuf.h: explicit cast restricting ptrdiff to uint16 > rte_ether.h: explicit cast avoiding truncation warning > rte_rwlock.h: gcc8 sign conversion warnings > rte_ip.h: cast input to bswap16 to be uint16 > rte_ip.h: cast around promotion to int > rte_ip.h: cast type decided by sizeof to uint32 > rte_ip.h: cast return checksum size to uint16 > rte_ip.h: cast away gcc8 warning on rte_ipv6_phdr_cksum > rte_mbuf.h: explicit cast for size type to uint32 > rte_mbuf.h: explicit casts to uint16 to avoid warnings > rte_ethdev.h: align sign and scope of temp var > rte_byteorder.h: explicit cast for return promotion 16 patches have been applied. The tags Fixes and Cc:stable have been added, so they can be backported. 5 patches are missing: lib/librte_eal: explicit tmp cast rte_rwlock.h: gcc8 sign conversion warnings rte_ip.h: cast type decided by sizeof to uint32 rte_mbuf.h: explicit casts to uint16 to avoid warnings rte_ethdev.h: align sign and scope of temp var Those patches are either not reviewed, or not safe enough at this release stage. Please, feel free to send them again in a v6 to make clear they need more review. I think the mbuf one (for uint16_t) may deserve to be split. Thanks
Re: [dpdk-dev] [PATCH v3 3/3] raw/ifpga_rawdev: fix the clang compiling issue
17/05/2018 03:47, Gavin Hu: > If the subject changes to raw/ifpga, it will get a Wrong headline prefix(can > be ignored?) warning when running git-log-fixes.h Yes, this warning can be ignored.
Re: [dpdk-dev] [PATCH v4 1/3] net/mlx5: fix the clang compiling issue
17/05/2018 04:46, Gavin Hu:
> From: gavin hu
>
> This patch adds a pair of "()" to embrace the argument
> input to the function-like macro invocation.
>
> drivers/net/mlx5/mlx5_rxtx_vec.c:37:
> drivers/net/mlx5/mlx5_rxtx_vec_neon.h:170:24: error: too many arguments
> provided to function-like macro invocation
> (uint16x8_t) { 0, 0, cs_flags,
> rte_cpu_to_be_16(len),
>
> Fixes: 570acdb1da ("net/mlx5: add vectorized Rx/Tx burst for ARM")
> Cc: sta...@dpdk.org
>
> Signed-off-by: Gavin Hu
> Signed-off-by: Sirshak Das
> Reviewed-by: Phil Yang
> Reviewed-by: Honnappa Nagarahalli
> Reviewed-by: Jerin Jacob
> Acked-by: Yongseok Koh
Series applied, thanks
Re: [dpdk-dev] [PATCH] eal: fix directory permissions
18/05/2018 15:55, Adrien Mazarguil:
> Executable bit must be set on directories for normal users to enter them.
>
> This patch addresses the inability to start DPDK applications as non-root
> due to errors such as:
>
> EAL: failed to bind /tmp/dpdk/rte/mp_socket: Permission denied
>
> Fixes: 56236363b481 ("eal: add directory for runtime data")
> Cc: Anatoly Burakov
>
> Signed-off-by: Adrien Mazarguil
Applied, thanks
Re: [dpdk-dev] [PATCH v2] test/eventdev: update the port id to 16 bit
15/05/2018 20:06, Vipin Varghese:
> port id in test case is update to unsigned int 16 bit, to
> prevent compilation failure when RTE_MAX_ETHPORTS port is
> set more than 255.
>
> Signed-off-by: Vipin Varghese
> Acked-by: Nikhil Rao
Fixes: fc8030eb8fe9 ("test/eventdev: add tests for eth Rx adapter APIs")
Cc: sta...@dpdk.org
Applied, thanks
Re: [dpdk-dev] [PATCH] bus/vdev: don't double space log messages
17/05/2018 11:33, Ferruh Yigit:
> On 5/16/2018 10:51 PM, Stephen Hemminger wrote:
> > The VDEV_LOG() macro already adds a newline, don't duplicate.
> >
> > Signed-off-by: Stephen Hemminger
>
> Fixes: d22fcb225c24 ("bus/vdev: change log type")
>
> Reviewed-by: Ferruh Yigit
Applied, thanks
[dpdk-dev] [PATCH] net liquidio fix Unable to update lio_dev->linfo.link var
Signed-off-by: yaochuhong
---
drivers/net/liquidio/lio_ethdev.c | 6 ++
1 file changed, 6 insertions(+)
mode change 100644 => 100755 drivers/net/liquidio/lio_ethdev.c
diff --git a/drivers/net/liquidio/lio_ethdev.c
b/drivers/net/liquidio/lio_ethdev.c
old mode 100644
new mode 100755
index 0e0b5d8..bd47911
--- a/drivers/net/liquidio/lio_ethdev.c
+++ b/drivers/net/liquidio/lio_ethdev.c
@@ -1405,6 +1405,12 @@ struct rte_lio_xstats_name_off {
/* Configure RSS if device configured with multiple RX queues. */
lio_dev_mq_rx_configure(eth_dev);
+/* Before update the link info, set link_status64 to 0,
+ * Otherwise, when eth_dev->data->mtu != mtu, the link state information
cannot be obtained,
+ * and lio_dev->intf_open will be set to 0.
+ */
+lio_dev->linfo.link.link_status64 = 0;
+
/* start polling for lsc */
ret = rte_eal_alarm_set(LIO_LSC_TIMEOUT,
lio_sync_link_state_check,
--
1.8.3.1
[dpdk-dev] [PATCH 2/2] fix Coding Style
---
drivers/net/liquidio/lio_ethdev.c | 5 +
1 file changed, 1 insertion(+), 4 deletions(-)
mode change 100755 => 100644 drivers/net/liquidio/lio_ethdev.c
diff --git a/drivers/net/liquidio/lio_ethdev.c
b/drivers/net/liquidio/lio_ethdev.c
old mode 100755
new mode 100644
index bd47911..0a6eb55
--- a/drivers/net/liquidio/lio_ethdev.c
+++ b/drivers/net/liquidio/lio_ethdev.c
@@ -1405,10 +1405,7 @@ struct rte_lio_xstats_name_off {
/* Configure RSS if device configured with multiple RX queues. */
lio_dev_mq_rx_configure(eth_dev);
-/* Before update the link info, set link_status64 to 0,
- * Otherwise, when eth_dev->data->mtu != mtu, the link state information
cannot be obtained,
- * and lio_dev->intf_open will be set to 0.
- */
+ /* Before update the link info, must set link_status64 to 0. */
lio_dev->linfo.link.link_status64 = 0;
/* start polling for lsc */
--
1.8.3.1
Re: [dpdk-dev] [PATCH v5] net/ixgbe: fix to config VLAN strip on the fly
> -Original Message-
> From: dev [mailto:dev-boun...@dpdk.org] On Behalf Of Wei Dai
> Sent: Saturday, May 19, 2018 6:11 PM
> To: Zhang, Qi Z; Yigit, Ferruh; Lu, Wenzhuo; Ananyev, Konstantin; Wu,
> Yanglong
> Cc: dev@dpdk.org; Dai, Wei; sta...@dpdk.org; Wei Dai
> Subject: [dpdk-dev] [PATCH v5] net/ixgbe: fix to config VLAN strip on the fly
>
> The old ixgbe_vlan_offload_set() is called by
> rte_eth_dev_set_vlan_offload() which is meant to config VLAN
> strip/filter/extend on all queues.
> This old function is also called by rte_eth_dev_start()/ixgbe_dev_start()
> which need support per-queue VALN strip on only parts of queues.
>
> So add new function ixgbe_vlan_offload_config() = old
> ixgbe_vlan_offload_set().
> This new function is called by ixgbe_dev_start() to support VLAN strip per-
> queue configuration.
>
> New ixgbe_vlan_offload_set() = codes to align VLAN strip flags on all queues
> with port level setting + new ixgbe_vlan_offload_configure().
> The 2nd function is called by rte_eth_dev_set_vlan_offload to support
> configure VLAN strip on all queues of whole port.
>
> Fixes: 216f78f4d53f ("net/ixgbe: support VLAN strip per queue offloading in
> PF")
> Fixes: 860a94d3c692 ("net/ixgbe: support VLAN strip per queue offloading in
> VF")
>
> Cc: sta...@dpdk.org
>
> Signed-off-by: Wei Dai
> Signed-off-by: Yanglong Wu
>
> Acked-by: Qi Zhang
Applied to dpdk-next-net-intel, thanks!
/Helin
Re: [dpdk-dev] [PATCH v2] i40e: fix corruption of VF stats via port representor
> -Original Message-
> From: Zhang, Qi Z
> Sent: Saturday, May 19, 2018 10:18 AM
> To: Horton, Remy; dev@dpdk.org
> Cc: Yigit, Ferruh; Xing, Beilei; Zhang, Helin; Wu, Jingjing
> Subject: RE: [PATCH v2] i40e: fix corruption of VF stats via port representor
>
>
>
> > -Original Message-
> > From: Horton, Remy
> > Sent: Friday, May 18, 2018 7:50 PM
> > To: dev@dpdk.org
> > Cc: Yigit, Ferruh ; Zhang, Qi Z
> > ; Xing, Beilei ; Zhang,
> > Helin ; Wu, Jingjing
> > Subject: [PATCH v2] i40e: fix corruption of VF stats via port
> > representor
> >
> > The resetting of stats on a VF involves the setting of an offset that
> > is subtracted from future calls, rather zeroing of counters.
> > However doing a stats reset on the port representor was also adjusting
> > the values forwarded to the VF, which had the effect of corrupting the
> > VF's counters. The fix is for the port representor to maintain its own
> > stats offset, so the port representor and VF maintain independently-
> resettable counters.
> >
> > Fixes: e0cb96204b71 ("net/i40e: add support for representor ports")
> >
> > Signed-off-by: Remy Horton
>
> Title prefix should be "net/i40e", could be fix during apply
>
> Acked-by: Qi Zhang
Applied to dpdk-next-net-intel, thanks!
/Helin
[dpdk-dev] [PATCH 1/3] net liquidio fix Unable to update lio_dev->linfo.link var
Signed-off-by: yaochuhong
---
drivers/net/liquidio/lio_ethdev.c | 6 ++
1 file changed, 6 insertions(+)
mode change 100644 => 100755 drivers/net/liquidio/lio_ethdev.c
diff --git a/drivers/net/liquidio/lio_ethdev.c
b/drivers/net/liquidio/lio_ethdev.c
old mode 100644
new mode 100755
index 0e0b5d8..bd47911
--- a/drivers/net/liquidio/lio_ethdev.c
+++ b/drivers/net/liquidio/lio_ethdev.c
@@ -1405,6 +1405,12 @@ struct rte_lio_xstats_name_off {
/* Configure RSS if device configured with multiple RX queues. */
lio_dev_mq_rx_configure(eth_dev);
+/* Before update the link info, set link_status64 to 0,
+ * Otherwise, when eth_dev->data->mtu != mtu, the link state information
cannot be obtained,
+ * and lio_dev->intf_open will be set to 0.
+ */
+lio_dev->linfo.link.link_status64 = 0;
+
/* start polling for lsc */
ret = rte_eal_alarm_set(LIO_LSC_TIMEOUT,
lio_sync_link_state_check,
--
1.8.3.1
[dpdk-dev] [PATCH 3/3] add signoff
Signed-off-by: yaochuhong
---
drivers/net/liquidio/lio_ethdev.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/liquidio/lio_ethdev.c
b/drivers/net/liquidio/lio_ethdev.c
index 0a6eb55..a6a37c8 100644
--- a/drivers/net/liquidio/lio_ethdev.c
+++ b/drivers/net/liquidio/lio_ethdev.c
@@ -1405,7 +1405,7 @@ struct rte_lio_xstats_name_off {
/* Configure RSS if device configured with multiple RX queues. */
lio_dev_mq_rx_configure(eth_dev);
- /* Before update the link info, must set link_status64 to 0. */
+ /* Before update the link info, must set linfo.link.link_status64 to 0.
*/
lio_dev->linfo.link.link_status64 = 0;
/* start polling for lsc */
--
1.8.3.1
[dpdk-dev] [PATCH 2/3] fix Coding Style
---
drivers/net/liquidio/lio_ethdev.c | 5 +
1 file changed, 1 insertion(+), 4 deletions(-)
mode change 100755 => 100644 drivers/net/liquidio/lio_ethdev.c
diff --git a/drivers/net/liquidio/lio_ethdev.c
b/drivers/net/liquidio/lio_ethdev.c
old mode 100755
new mode 100644
index bd47911..0a6eb55
--- a/drivers/net/liquidio/lio_ethdev.c
+++ b/drivers/net/liquidio/lio_ethdev.c
@@ -1405,10 +1405,7 @@ struct rte_lio_xstats_name_off {
/* Configure RSS if device configured with multiple RX queues. */
lio_dev_mq_rx_configure(eth_dev);
-/* Before update the link info, set link_status64 to 0,
- * Otherwise, when eth_dev->data->mtu != mtu, the link state information
cannot be obtained,
- * and lio_dev->intf_open will be set to 0.
- */
+ /* Before update the link info, must set link_status64 to 0. */
lio_dev->linfo.link.link_status64 = 0;
/* start polling for lsc */
--
1.8.3.1
[dpdk-dev] [PATCH v6 3/8] rte_ip.h: cast type decided by sizeof to uint32
/projects/lagopus/src/dpdk/build/include/rte_ip.h:
In function 'rte_ipv4_udptcp_cksum':
/projects/lagopus/src/dpdk/build/include/rte_byteorder.h:
51:24: warning: conversion from 'long unsigned int' to
'uint32_t' {aka 'unsigned int'} may change value
[-Wconversion]
#define rte_bswap16(x) ((uint16_t)
(__builtin_constant_p(x) ? \
^
/projects/lagopus/src/dpdk/build/include/rte_byteorder.h:
85:29: note: in expansion of macro 'rte_bswap16'
#define rte_be_to_cpu_16(x) rte_bswap16(x)
^~~
/projects/lagopus/src/dpdk/build/include/rte_ip.h:321:11:
note: in expansion of macro 'rte_be_to_cpu_16'
l4_len = rte_be_to_cpu_16(ipv4_hdr->total_length) -
^~~~
Also with this one, it is a cast that always occurred
and is just being done explicitly, with no changes to
the generated code.
The warning stack is misleading, it points to the last
element in the macro that produced the lhs of the subtraction
above. But the only "unsigned long int" in the expression is
the result of the sizeof() on the rhs, it promotes the
subtraction result to unsigned long. So the error actually
relates to the result of the outer subtraction.
The actual error is "you are trying to put an unsigned long
into a uint32_t". We always did so, the fix is just to inform
the compiler it is intentional with an explicit cast.
Fixes: 6006818cfb ("net: new checksum functions")
Signed-off-by: Andy Green
---
lib/librte_net/rte_ip.h |4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/lib/librte_net/rte_ip.h b/lib/librte_net/rte_ip.h
index c924aca7f..72dc2456a 100644
--- a/lib/librte_net/rte_ip.h
+++ b/lib/librte_net/rte_ip.h
@@ -318,8 +318,8 @@ rte_ipv4_udptcp_cksum(const struct ipv4_hdr *ipv4_hdr,
const void *l4_hdr)
uint32_t cksum;
uint32_t l4_len;
- l4_len = rte_be_to_cpu_16(ipv4_hdr->total_length) -
- sizeof(struct ipv4_hdr);
+ l4_len = (uint32_t)(rte_be_to_cpu_16(ipv4_hdr->total_length) -
+ sizeof(struct ipv4_hdr));
cksum = rte_raw_cksum(l4_hdr, l4_len);
cksum += rte_ipv4_phdr_cksum(ipv4_hdr, 0);
[dpdk-dev] [PATCH v6 2/8] rte_rwlock.h: gcc8 sign conversion warnings
In file included from /projects/lagopus/src/dpdk/
build/include/rte_rwlock.h:12,
from ./mgr/lock.h:24,
from ./mgr/sock_io.c:54:
/projects/lagopus/src/dpdk/build/include/generic/
rte_rwlock.h: In function 'rte_rwlock_read_lock':
/projects/lagopus/src/dpdk/build/include/generic/
rte_rwlock.h:74:12: warning: conversion to 'uint32_t'
{aka 'unsigned int'} from 'int32_t' {aka 'int'} may
change the sign of the result [-Wsign-conversion]
x, x + 1);
^
/projects/lagopus/src/dpdk/build/include/generic/
rte_rwlock.h:74:17: warning: conversion to 'uint32_t'
{aka 'unsigned int'} from 'int' may change the sign
of the result [-Wsign-conversion]
x, x + 1);
~~^~~
/projects/lagopus/src/dpdk/build/include/generic/
rte_rwlock.h: In function 'rte_rwlock_write_lock':
/projects/lagopus/src/dpdk/build/include/generic/
rte_rwlock.h:110:15: warning: unsigned conversion
from 'int' to 'uint32_t' {aka 'unsigned int'}
changes value from '-1' to '4294967295'
[-Wsign-conversion]
0, -1);
^~
Again in this case we are making explicit the exact cast
that was always happening implicitly. The patch does not
change the generated code.
The int32_t temp "x" is required to be signed to detect
a < 0 error condition from the lock status. Afterwards,
it has always been implicitly cast to uint32_t when it
is used in the arguments to rte_atomic32_cmpset()...
gcc8.1 objects to the implicit cast now and requires us
to cast it explicitly.
Fixes: af75078fec ("first public release")
Signed-off-by: Andy Green
---
lib/librte_eal/common/include/generic/rte_rwlock.h |4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/lib/librte_eal/common/include/generic/rte_rwlock.h
b/lib/librte_eal/common/include/generic/rte_rwlock.h
index 899e9bc43..5751a0e6d 100644
--- a/lib/librte_eal/common/include/generic/rte_rwlock.h
+++ b/lib/librte_eal/common/include/generic/rte_rwlock.h
@@ -71,7 +71,7 @@ rte_rwlock_read_lock(rte_rwlock_t *rwl)
continue;
}
success = rte_atomic32_cmpset((volatile uint32_t *)&rwl->cnt,
- x, x + 1);
+ (uint32_t)x, (uint32_t)(x + 1));
}
}
@@ -107,7 +107,7 @@ rte_rwlock_write_lock(rte_rwlock_t *rwl)
continue;
}
success = rte_atomic32_cmpset((volatile uint32_t *)&rwl->cnt,
- 0, -1);
+ 0, (uint32_t)-1);
}
}
[dpdk-dev] [PATCH v6 0/8] Fixes for GCC8 against lagopus
The following series fixes build problems in dpdk master headers, found when using it as the dpdk subproject in lagopus. These errors are coming when you try to use the dpdk headers, not when you build dpdk itself. These are the remaining 5 patches not yet applied, with one split into three for clarity. There is no new patch content. Extra explanation is added (along with Fixes: already). I confirmed with master plus these patches, the only remaing compile errors in lagopus are from lagopus itself. --- Andy Green (8): lib/librte_eal: explicit tmp cast rte_rwlock.h: gcc8 sign conversion warnings rte_ip.h: cast type decided by sizeof to uint32 rte_mbuf.h: reduce temp to match uint16 types rte_mbuf.h: avoid implicit demotion in 64-bit arith rte_mbuf.h: choose correct type for temp rte_mbuf.h: add and subtract explicitly to avoid promotion rte_ethdev.h: align sign and scope of temp var .../common/include/arch/x86/rte_memcpy.h |8 +++--- lib/librte_eal/common/include/generic/rte_rwlock.h |4 ++- lib/librte_ethdev/rte_ethdev.h | 25 lib/librte_mbuf/rte_mbuf.h | 14 ++- lib/librte_net/rte_ip.h|4 ++- 5 files changed, 31 insertions(+), 24 deletions(-) -- Signature
[dpdk-dev] [PATCH v6 1/8] lib/librte_eal: explicit tmp cast
/projects/lagopus/src/dpdk/build/include/rte_memcpy.h:
793:2: note: in expansion of macro 'MOVEUNALIGNED_LEFT47'
MOVEUNALIGNED_LEFT47(dst, src, n, srcofs);
^~~~
/projects/lagopus/src/dpdk/build/include/rte_memcpy.h:
649:51: warning: conversion from 'size_t' {aka 'long
unsigned int'} to 'int' may change value [-Wconversion]
case 0x0B: MOVEUNALIGNED_LEFT47_IMM(dst, src,
n, 0x0B); break;\
^
/projects/lagopus/src/dpdk/build/include/rte_memcpy.h:
616:15: note: in definition of macro 'MOVEUNALIGNED_LEFT47_IMM'
tmp = len;
\
^~~
/projects/lagopus/src/dpdk/build/include/rte_memcpy.h:
793:2: note: in expansion of macro 'MOVEUNALIGNED_LEFT47'
MOVEUNALIGNED_LEFT47(dst, src, n, srcofs);
^~~~
/projects/lagopus/src/dpdk/build/include/rte_memcpy.h:
618:13: warning: conversion to 'size_t' {aka 'long
unsigned int'} from 'int' may change the sign of the
result [-Wsign-conversion]
tmp -= len;
\
^~
/projects/lagopus/src/dpdk/build/include/rte_memcpy.h:
649:16: note: in expansion of macro 'MOVEUNALIGNED_LEFT47_IMM'
case 0x0B: MOVEUNALIGNED_LEFT47_IMM(dst, src,
n, 0x0B); break;\
^~~~
/projects/lagopus/src/dpdk/build/include/rte_memcpy.h:
793:2: note: in expansion of macro 'MOVEUNALIGNED_LEFT47'
MOVEUNALIGNED_LEFT47(dst, src, n, srcofs);
^~~~
This provides an explicit cast for a cast that has always been
happening, and still happens exactly the same after the cast.
It doesn't change the generated code.
/projects/lagopus/src/dpdk/build/include/rte_memcpy.h:
618:13: warning: conversion to 'size_t' {aka 'long
unsigned int'} from 'int' may change the sign of the
result [-Wsign-conversion]
tmp -= len;
^~
int tmp;
...
-tmp -= len;
+tmp -= (int)len;
Signed-off-by: Andy Green
Fixes: d35cc1fe6a ("eal/x86: revert select optimized memcpy at run-time")
---
.../common/include/arch/x86/rte_memcpy.h |8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
b/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
index 5ead68ab2..f9ea0ab69 100644
--- a/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
+++ b/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
@@ -597,9 +597,9 @@ __extension__ ({
_mm_storeu_si128((__m128i *)((uint8_t *)dst + 7 * 16),
_mm_alignr_epi8(xmm8, xmm7, offset));\
dst = (uint8_t *)dst + 128;
\
}
\
-tmp = len;
\
+tmp = (int)len;
\
len = ((len - 16 + offset) & 127) + 16 - offset;
\
-tmp -= len;
\
+tmp -= (int)len;
\
src = (const uint8_t *)src + tmp;
\
dst = (uint8_t *)dst + tmp;
\
if (len >= 32 + 16 - offset) {
\
@@ -613,9 +613,9 @@ __extension__ ({
_mm_storeu_si128((__m128i *)((uint8_t *)dst + 1 * 16),
_mm_alignr_epi8(xmm2, xmm1, offset));\
dst = (uint8_t *)dst + 32;
\
}
\
-tmp = len;
\
+tmp = (int)len;
\
len = ((len - 16 + offset) & 31) + 16 - offset;
\
-tmp -= len;
\
+tmp -= (int)len;
\
src = (const uint8_t *)src + tmp;
\
dst = (uint8_t *)dst + tmp;
[dpdk-dev] [PATCH v6 4/8] rte_mbuf.h: reduce temp to match uint16 types
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
In function 'rte_pktmbuf_detach':
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
1583:17: warning: conversion from 'uint32_t' {aka
'unsigned int'} to 'uint16_t' {aka
'short unsigned int'} may change value [-Wconversion]
m->priv_size = priv_size;
^
The temp priv_size is declared as a uint32_t. But it
only deals in uint16_t. m->priv_size is a uint16_t.
Change it to a uint16_t.
Fixes: 355e6735b3 ("mbuf: fix cloning with private mbuf data")
Fixes: 1a60a0daa6 ("mbuf: fix segments number type increase")
Signed-off-by: Andy Green
---
lib/librte_mbuf/rte_mbuf.h |3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
index 28fd4ad52..76e37a2f8 100644
--- a/lib/librte_mbuf/rte_mbuf.h
+++ b/lib/librte_mbuf/rte_mbuf.h
@@ -1571,7 +1571,8 @@ __rte_pktmbuf_free_direct(struct rte_mbuf *m)
static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
{
struct rte_mempool *mp = m->pool;
- uint32_t mbuf_size, buf_len, priv_size;
+ uint32_t mbuf_size, buf_len;
+ uint16_t priv_size;
if (RTE_MBUF_HAS_EXTBUF(m))
__rte_pktmbuf_free_extbuf(m);
[dpdk-dev] [PATCH v6 8/8] rte_ethdev.h: align sign and scope of temp var
/projects/lagopus/src/dpdk/build/include/rte_ethdev.h:
In function 'rte_eth_rx_burst':
/projects/lagopus/src/dpdk/build/include/rte_ethdev.h:
3836:18: warning: conversion to 'int16_t' {aka 'short
int'} from 'uint16_t' {aka 'short unsigned int'} may
change the sign of the result [-Wsign-conversion]
int16_t nb_rx = (*dev->rx_pkt_burst)(dev->data->
rx_queues[queue_id],
^
/projects/lagopus/src/dpdk/build/include/rte_ethdev.h:
3844:50: warning: conversion to 'uint16_t' {aka 'short
unsigned int'} from 'int16_t' {aka 'short int'} may
change the sign of the result [-Wsign-conversion]
nb_rx = cb->fn.rx(port_id, queue_id, rx_pkts, nb_rx,
^
/projects/lagopus/src/dpdk/build/include/rte_ethdev.h:
3844:12: warning: conversion to 'int16_t' {aka 'short
int'} from 'uint16_t' {aka 'short unsigned int'} may
change the sign of the result [-Wsign-conversion]
nb_rx = cb->fn.rx(port_id, queue_id, rx_pkts, nb_rx,
^~
/projects/lagopus/src/dpdk/build/include/rte_ethdev.h:
3851:9: warning: conversion to 'uint16_t' {aka 'short
unsigned int'} from 'int16_t' {aka 'short int'} may
change the sign of the result [-Wsign-conversion]
return nb_rx;
^
The second part of the patch is solved by its own basic
block because it is inside a preprocessor conditional.
Bringing the declaration of the var to the top of the
function would require that also being given its own
preprocessor conditional, or a (void)var to avoid an
unused var warning. The basic block is no worse than
those imho.
Signed-off-by: Andy Green
---
lib/librte_ethdev/rte_ethdev.h | 25 +++--
1 file changed, 15 insertions(+), 10 deletions(-)
diff --git a/lib/librte_ethdev/rte_ethdev.h b/lib/librte_ethdev/rte_ethdev.h
index d52c1bed9..4d059a2a7 100644
--- a/lib/librte_ethdev/rte_ethdev.h
+++ b/lib/librte_ethdev/rte_ethdev.h
@@ -3823,6 +3823,7 @@ 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];
+ uint16_t nb_rx;
#ifdef RTE_LIBRTE_ETHDEV_DEBUG
RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, 0);
@@ -3833,18 +3834,22 @@ rte_eth_rx_burst(uint16_t port_id, uint16_t queue_id,
return 0;
}
#endif
- int16_t nb_rx = (*dev->rx_pkt_burst)(dev->data->rx_queues[queue_id],
- rx_pkts, nb_pkts);
+ nb_rx = (*dev->rx_pkt_burst)(dev->data->rx_queues[queue_id],
+rx_pkts, nb_pkts);
#ifdef RTE_ETHDEV_RXTX_CALLBACKS
- struct rte_eth_rxtx_callback *cb = dev->post_rx_burst_cbs[queue_id];
-
- if (unlikely(cb != NULL)) {
- do {
- nb_rx = cb->fn.rx(port_id, queue_id, rx_pkts, nb_rx,
- nb_pkts, cb->param);
- cb = cb->next;
- } while (cb != NULL);
+ {
+ struct rte_eth_rxtx_callback *cb =
+ dev->post_rx_burst_cbs[queue_id];
+
+ if (unlikely(cb != NULL)) {
+ do {
+ nb_rx = cb->fn.rx(port_id, queue_id,
+ rx_pkts, nb_rx,
+ nb_pkts, cb->param);
+ cb = cb->next;
+ } while (cb != NULL);
+ }
}
#endif
[dpdk-dev] [PATCH v6 5/8] rte_mbuf.h: avoid implicit demotion in 64-bit arith
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
In function 'rte_validate_tx_offload':
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
2112:19: warning: conversion to 'uint64_t'
{aka 'long unsigned int'} from 'int' may change the
sign of the result [-Wsign-conversion]
inner_l3_offset += m->outer_l2_len + m->outer_l3_len;
^~
uint64_t inner_l3_offset...
/* fields for TX offloading of tunnels */
uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
We want to do the arithmetic entirely in uint64_t
space, but there is an implicit demotion to int created by
the +=. Remove the +=.
Fixes: 4fb7e803eb ("ethdev: add Tx preparation")
Signed-off-by: Andy Green
---
lib/librte_mbuf/rte_mbuf.h |3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
index 76e37a2f8..55fba3b14 100644
--- a/lib/librte_mbuf/rte_mbuf.h
+++ b/lib/librte_mbuf/rte_mbuf.h
@@ -2112,7 +2112,8 @@ rte_validate_tx_offload(const struct rte_mbuf *m)
return 0;
if (ol_flags & PKT_TX_OUTER_IP_CKSUM)
- inner_l3_offset += m->outer_l2_len + m->outer_l3_len;
+ inner_l3_offset = inner_l3_offset + m->outer_l2_len +
+ m->outer_l3_len;
/* Headers are fragmented */
if (rte_pktmbuf_data_len(m) < inner_l3_offset + m->l3_len + m->l4_len)
[dpdk-dev] [PATCH v6 7/8] rte_mbuf.h: add and subtract explicitly to avoid promotion
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
In function 'rte_pktmbuf_prepend':
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
1908:17: warning: conversion from 'int' to 'uint16_t'
{aka 'short unsigned int'} may change value [-Wconversion]
m->data_off -= len;
^~~
m->data_off is a uint16_t
uint16_t data_off;
len (a uint16_t) is promoted to an int using -=. Do the
subtraction explicitly and cast the result to uint16_t.
- m->data_off -= len;
+ m->data_off = (uint16_t)(m->data_off - len);
The below += or -= changes are solving the same thing.
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
In function 'rte_pktmbuf_adj':
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
1969:17: warning: conversion from 'int' to 'uint16_t'
{aka 'short unsigned int'} may change value [-Wconversion]
m->data_off += len;
^~~
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
In function 'rte_pktmbuf_chain':
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
2082:19: warning: conversion from 'int' to 'uint16_t'
{aka 'short unsigned int'} may change value [-Wconversion]
head->nb_segs += tail->nb_segs;
^~~~
Also uint16_t
uint16_t nb_segs; /**< Number of segments. */
Fixes: 08b563ffb1 ("mbuf: replace data pointer by an offset")
Signed-off-by: Andy Green
---
lib/librte_mbuf/rte_mbuf.h |6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
index a0423a548..beb104c69 100644
--- a/lib/librte_mbuf/rte_mbuf.h
+++ b/lib/librte_mbuf/rte_mbuf.h
@@ -1908,7 +1908,7 @@ static inline char *rte_pktmbuf_prepend(struct rte_mbuf
*m,
if (unlikely(len > rte_pktmbuf_headroom(m)))
return NULL;
- m->data_off -= len;
+ m->data_off = (uint16_t)(m->data_off - len);
m->data_len = (uint16_t)(m->data_len + len);
m->pkt_len = (m->pkt_len + len);
@@ -1969,7 +1969,7 @@ static inline char *rte_pktmbuf_adj(struct rte_mbuf *m,
uint16_t len)
return NULL;
m->data_len = (uint16_t)(m->data_len - len);
- m->data_off += len;
+ m->data_off = (uint16_t)(m->data_off + len);
m->pkt_len = (m->pkt_len - len);
return (char *)m->buf_addr + m->data_off;
}
@@ -2082,7 +2082,7 @@ static inline int rte_pktmbuf_chain(struct rte_mbuf
*head, struct rte_mbuf *tail
cur_tail->next = tail;
/* accumulate number of segments and total length. */
- head->nb_segs += tail->nb_segs;
+ head->nb_segs = (uint16_t)(head->nb_segs + tail->nb_segs);
head->pkt_len += tail->pkt_len;
/* pkt_len is only set in the head */
[dpdk-dev] [PATCH v6 6/8] rte_mbuf.h: choose correct type for temp
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
In function 'rte_pktmbuf_linearize':
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
1873:32: warning: conversion to 'int' from 'uint32_t'
{aka 'unsigned int'} may change the sign of the
result [-Wsign-conversion]
#define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
^
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
2166:13: note: in expansion of macro 'rte_pktmbuf_pkt_len'
copy_len = rte_pktmbuf_pkt_len(mbuf) -
rte_pktmbuf_data_len(mbuf);
^~~
/projects/lagopus/src/dpdk/build/include/rte_mbuf.h:
2180:51: warning: conversion to 'size_t' {aka
'long unsigned int'} from 'int' may change the
sign of the result [-Wsign-conversion]
rte_memcpy(buffer, rte_pktmbuf_mtod(m, char *), seg_len);
^~~
The temp is consumed as a size_t. So let's make it
a size_t in the first place.
Fixes: 1feda4d8fc ("mbuf: add a function to linearize a packet")
Signed-off-by: Andy Green
---
lib/librte_mbuf/rte_mbuf.h |2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
index 55fba3b14..a0423a548 100644
--- a/lib/librte_mbuf/rte_mbuf.h
+++ b/lib/librte_mbuf/rte_mbuf.h
@@ -2158,7 +2158,7 @@ rte_validate_tx_offload(const struct rte_mbuf *m)
static inline int
rte_pktmbuf_linearize(struct rte_mbuf *mbuf)
{
- int seg_len, copy_len;
+ size_t seg_len, copy_len;
struct rte_mbuf *m;
struct rte_mbuf *m_next;
char *buffer;
Re: [dpdk-dev] [PATCH v5 00/21] Fixes for GCC8 against lagopus
On 05/21/2018 06:18 AM, Thomas Monjalon wrote: Andy Green (21): lib/librte_ethdev: change eth-dev-ops API to return int rte_string_fns.h: fix gcc8.1 sign conv warning in lstrcpy lib/librte_eal: explicit tmp cast /lib/librte_eal: stage cast from uint64 to long rte_ring_generic.h: stack declarations before code rte_ring.h: remove signed type flipflopping rte_mbuf.h: avoid warnings from inadvertant promotion rte_mbuf.h: explicit casts for int16 to uint16 rte_mbuf.h: make sure RTE-MIN compares same types rte_mbuf.h: explicit cast restricting ptrdiff to uint16 rte_ether.h: explicit cast avoiding truncation warning rte_rwlock.h: gcc8 sign conversion warnings rte_ip.h: cast input to bswap16 to be uint16 rte_ip.h: cast around promotion to int rte_ip.h: cast type decided by sizeof to uint32 rte_ip.h: cast return checksum size to uint16 rte_ip.h: cast away gcc8 warning on rte_ipv6_phdr_cksum rte_mbuf.h: explicit cast for size type to uint32 rte_mbuf.h: explicit casts to uint16 to avoid warnings rte_ethdev.h: align sign and scope of temp var rte_byteorder.h: explicit cast for return promotion 16 patches have been applied. The tags Fixes and Cc:stable have been added, so they can be backported. Thanks a lot for the help. 5 patches are missing: lib/librte_eal: explicit tmp cast rte_rwlock.h: gcc8 sign conversion warnings rte_ip.h: cast type decided by sizeof to uint32 rte_mbuf.h: explicit casts to uint16 to avoid warnings rte_ethdev.h: align sign and scope of temp var Those patches are either not reviewed, or not safe enough at this release stage. Well, at least several of them are actually a NOP wrt "safety", since they just do the cast that pre-gcc v8 was doing silently until now. I can see it's not exactly easy to know that at a glance though. I added a rationale for what each patch is doing making it clear where it is effectively a NOP just making explicit what was implicit before. Please, feel free to send them again in a v6 to make clear they need more review. ... well, you may find them easier to parse in the v6 I just pushed. I think the mbuf one (for uint16_t) may deserve to be split. Yes looking at it, it is three patches in one. I split it out. -Andy Thanks
Re: [dpdk-dev] [PATCH] net/i40e: add a specific API to control the LLDP agent
> > Add a new API rte_pmd_i40e_set_lldp_cmd to control LLDP agent for i40e. > > It supports the following i40e debug lldp commands: > > - start/stop of the LLDP agent. > > - get local/remote of the LLDP MIB (Management Information Base). > > > > Signed-off-by: Laurent Hardy > > Signed-off-by: Zijie Pan > > Does it mean that this feature is enabled by default in i40e devices? > Should it be disabled by default? The new API is just an interface to control the LLDP agent. No functions of i40e call it by default. It is only used to debug LLDP of i40e devices. For example we can add a new command for test-pmd to debug lldp of i40e by calling rte_pmd_i40e_set_lldp_cmd().
Re: [dpdk-dev] [PATCH] net/mlx5: fix invalid count in xstats
On Sun, May 20, 2018 at 10:13:58AM +0200, David Marchand wrote:
> With the commit af4f09f28294 ("net/mlx5: prefix all functions with mlx5"),
> mlx5_xstats_get() is not compliant any longer with the api.
> It always returns the caller max entries count while it should return how
> many entries it wrote/wanted to write.
>
> Fixes: af4f09f28294 ("net/mlx5: prefix all functions with mlx5")
>
> Signed-off-by: David Marchand
Acked-by: Nelio Laranjeiro
> ---
> drivers/net/mlx5/mlx5_stats.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/mlx5/mlx5_stats.c b/drivers/net/mlx5/mlx5_stats.c
> index 8e427e7..875dd10 100644
> --- a/drivers/net/mlx5/mlx5_stats.c
> +++ b/drivers/net/mlx5/mlx5_stats.c
> @@ -325,7 +325,7 @@ mlx5_xstats_get(struct rte_eth_dev *dev, struct
> rte_eth_xstat *stats,
> stats[i].value = (counters[i] - xstats_ctrl->base[i]);
> }
> }
> - return n;
> + return xstats_n;
> }
>
> /**
> --
> 2.7.4
Thanks,
--
Nélio Laranjeiro
6WIND
Re: [dpdk-dev] [PATCH v5 01/21] lib/librte_ethdev: change eth-dev-ops API to return int
On Sun, 20 May 2018 02:43:58 + Shreyansh Jain wrote: > > > This doesn't feel correct. A counter, especially the number of > > descriptors in a queue, doesn't have a negative value. So, 1) this is > > an unnatural return and 2) we litter the code with unnecessary > > typecast. > > > > > > In fact, even in the above change, the debug messages continue to > > print unsigned values. So, another typecast would be required there. > > > > > > I don't agree with this change - at least not until some strong gcc 8 > > warning reason is triggering this. Can you please point me to some > > conversation on mailing list which enforces this? > > > > > > > hm no, it's not my idea. > > > > If you don't like it, don't do it, I don't mind either way. I sent a > > patch that just solved the compiler error only already, and was told on > > the list it would be cooler if these things returned an int instead. > > > > There's no point challenging me about the wisdom of it, although it > > seems reasonable to me. I sent a patch, list guy $1 says do X instead, > > I do X and then list guy $2 says EXPLAIN YOURSELF. > > That is what a community is. Consensus has to be built, not expected > automagically. If you touch a line, you are responsible for it (also, because > in future git blame would point *you* out for a change). My comment was a suggestion, not a "you must do it this way". The reason was it was cleaner change for Gcc fix and it allowed for possibility that some driver might not detect an error (for example if device was removed by hot plug).
Re: [dpdk-dev] [PATCH 1/3] net liquidio fix Unable to update lio_dev->linfo.link var
On Mon, 21 May 2018 09:20:53 +0800
yaochuhong wrote:
> Signed-off-by: yaochuhong
> ---
> drivers/net/liquidio/lio_ethdev.c | 6 ++
> 1 file changed, 6 insertions(+)
> mode change 100644 => 100755 drivers/net/liquidio/lio_ethdev.c
>
> diff --git a/drivers/net/liquidio/lio_ethdev.c
> b/drivers/net/liquidio/lio_ethdev.c
> old mode 100644
> new mode 100755
> index 0e0b5d8..bd47911
> --- a/drivers/net/liquidio/lio_ethdev.c
> +++ b/drivers/net/liquidio/lio_ethdev.c
> @@ -1405,6 +1405,12 @@ struct rte_lio_xstats_name_off {
> /* Configure RSS if device configured with multiple RX queues. */
> lio_dev_mq_rx_configure(eth_dev);
>
> +/* Before update the link info, set link_status64 to 0,
> + * Otherwise, when eth_dev->data->mtu != mtu, the link state information
> cannot be obtained,
> + * and lio_dev->intf_open will be set to 0.
> + */
> +lio_dev->linfo.link.link_status64 = 0;
> +
> /* start polling for lsc */
> ret = rte_eal_alarm_set(LIO_LSC_TIMEOUT,
> lio_sync_link_state_check,
Your mail system or editor is converting tabs to spaces.
Please run checkpatch on your submissions.
### [dpdk-dev] [PATCH 1/3] net liquidio fix Unable to update
lio_dev->linfo.link var
ERROR:EXECUTE_PERMISSIONS: do not set execute permissions for source files
#88: FILE: drivers/net/liquidio/lio_ethdev.c
ERROR:TRAILING_WHITESPACE: trailing whitespace
#97: FILE: drivers/net/liquidio/lio_ethdev.c:1409:
+ * Otherwise, when eth_dev->data->mtu != mtu, the link state information
cannot be obtained, $
WARNING:LONG_LINE_COMMENT: line over 80 characters
#97: FILE: drivers/net/liquidio/lio_ethdev.c:1409:
+ * Otherwise, when eth_dev->data->mtu != mtu, the link state information
cannot be obtained,
WARNING:LEADING_SPACE: please, no spaces at the start of a line
#100: FILE: drivers/net/liquidio/lio_ethdev.c:1412:
+lio_dev->linfo.link.link_status64 = 0;$
ERROR:TRAILING_WHITESPACE: trailing whitespace
#101: FILE: drivers/net/liquidio/lio_ethdev.c:1413:
+$
WARNING:LEADING_SPACE: please, no spaces at the start of a line
#101: FILE: drivers/net/liquidio/lio_ethdev.c:1413:
+$
total: 3 errors, 3 warnings, 0 checks, 12 lines checked
