[dpdk-dev] ABI changes in release notes
Hi all, I've just noticed that some ABI changes were announced and done but the corresponding entries in the release notes have not been added. Please check.
[dpdk-dev] [dpdk-announce] release candidate 2.2.0-rc2
A new DPDK release candidate is ready for testing: http://dpdk.org/browse/dpdk/tag/?id=v2.2.0-rc2 Changelog (main changes since 2.2.0-rc1) - enhancements: * ARM v7 * ARM v8 * option -d works with directories * keep alive monitoring * more sched subports * precision time protocol with Intel devices * szedata2 100G networking driver * crypto API (experimental) * Intel AES-NI multi-buffer crypto driver * Intel QuickAssist crypto driver * more in packet framework - fixes for: * compilation * drivers * unit tests The release 2.2.0 is planned for mid-December. It means there are only two weeks remaining to do some code changes. So only fixes should be accepted with very few exceptions. The pending patches about the build system and the examples are still candidates for the RC3. There are still some compilation issues. Please consider them as the highest priority in the coming days. Thank you everyone
[dpdk-dev] [PATCH 1/2] examples/distributor: Fix compile issue
examples/distributor/main.c(338): error #167:
argument of type "struct rte_mbuf *"
is incompatible with parameter of type "const char *"
_mm_prefetch(bufs[0], 0);
The first param of _mm_prefetch should be "const char *" and
need convert "struct rte_mbuf *" to "const void *".
Signed-off-by: Michael Qiu
---
examples/distributor/main.c | 8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/examples/distributor/main.c b/examples/distributor/main.c
index 972bddb..a4d8e34 100644
--- a/examples/distributor/main.c
+++ b/examples/distributor/main.c
@@ -335,13 +335,13 @@ lcore_tx(struct rte_ring *in_r)
/* for traffic we receive, queue it up for transmit */
uint16_t i;
- _mm_prefetch(bufs[0], 0);
- _mm_prefetch(bufs[1], 0);
- _mm_prefetch(bufs[2], 0);
+ _mm_prefetch((const void *)bufs[0], 0);
+ _mm_prefetch((const void *)bufs[1], 0);
+ _mm_prefetch((const void *)bufs[2], 0);
for (i = 0; i < nb_rx; i++) {
struct output_buffer *outbuf;
uint8_t outp;
- _mm_prefetch(bufs[i + 3], 0);
+ _mm_prefetch((const void *)bufs[i + 3], 0);
/*
* workers should update in_port to hold the
* output port value
--
1.9.3
[dpdk-dev] [PATCH 2/2] Fix compile issue in i686 platform
In i686 platform, long is 32bit, so XXX_CYCLECOUNTER_MASK need define as 'ULL' Signed-off-by: Michael Qiu --- drivers/net/e1000/igb_ethdev.c | 2 +- drivers/net/i40e/i40e_ethdev.c | 2 +- drivers/net/ixgbe/ixgbe_ethdev.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/net/e1000/igb_ethdev.c b/drivers/net/e1000/igb_ethdev.c index 31452ae..518b6c9 100644 --- a/drivers/net/e1000/igb_ethdev.c +++ b/drivers/net/e1000/igb_ethdev.c @@ -78,7 +78,7 @@ #define IGB_8_BIT_MASK UINT8_MAX /* Additional timesync values. */ -#define E1000_CYCLECOUNTER_MASK 0x +#define E1000_CYCLECOUNTER_MASK 0xULL #define E1000_ETQF_FILTER_1588 3 #define IGB_82576_TSYNC_SHIFT16 #define E1000_INCPERIOD_82576(1 << E1000_TIMINCA_16NS_SHIFT) diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c index 13ab81a..5cd6e88 100644 --- a/drivers/net/i40e/i40e_ethdev.c +++ b/drivers/net/i40e/i40e_ethdev.c @@ -131,7 +131,7 @@ #define I40E_PTP_1GB_INCVAL 0x20ULL #define I40E_PRTTSYN_TSYNENA 0x8000 #define I40E_PRTTSYN_TSYNTYPE0x0e00 -#define I40E_CYCLECOUNTER_MASK 0x +#define I40E_CYCLECOUNTER_MASK 0xULL #define I40E_MAX_PERCENT100 #define I40E_DEFAULT_DCB_APP_NUM1 diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c index 49f2410..808ac69 100644 --- a/drivers/net/ixgbe/ixgbe_ethdev.c +++ b/drivers/net/ixgbe/ixgbe_ethdev.c @@ -136,7 +136,7 @@ #define IXGBE_INCVAL_SHIFT_82599 7 #define IXGBE_INCPER_SHIFT_82599 24 -#define IXGBE_CYCLECOUNTER_MASK 0x +#define IXGBE_CYCLECOUNTER_MASK 0xULL static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev); static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev); -- 1.9.3
[dpdk-dev] no hugepage with UIO poll-mode driver
Hello,
Thank you all for helping us understand on issues with no hugepage option.
As Konstantin mentioned at the end, I tried using VFIO module instead of
IGB UIO module. I enabled all necessary parameters (IOMMU,
virtualization, vfio-pci, VFIO permission) and ran my code with no
hugepage option.
At first, it seemed to receive packets fine, but after a while, it
stopped receiving packets. I could temporarily remove this issue by not
calling rte_eth_tx_burst(). Also, when I looked at the received packets,
they all contained 0s instead of actual data. Was there anything that I
missed in running with VFIO? I'm curious if no hugepage with no hugepage
option was confirmed to run with VFIO.
Thank you,
Younghwan
2015-11-25 ?? 11:12? Ananyev, Konstantin ?(?) ? ?:
>
>> -Original Message-
>> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Sergio Gonzalez
>> Monroy
>> Sent: Wednesday, November 25, 2015 1:44 PM
>> To: Thomas Monjalon
>> Cc: dev at dpdk.org
>> Subject: Re: [dpdk-dev] no hugepage with UIO poll-mode driver
>>
>> On 25/11/2015 13:22, Thomas Monjalon wrote:
>>> 2015-11-25 12:02, Bruce Richardson:
On Wed, Nov 25, 2015 at 12:03:05PM +0100, Thomas Monjalon wrote:
> 2015-11-25 11:00, Bruce Richardson:
>> On Wed, Nov 25, 2015 at 11:23:57AM +0100, Thomas Monjalon wrote:
>>> 2015-11-25 10:08, Bruce Richardson:
On Wed, Nov 25, 2015 at 03:39:17PM +0900, Younghwan Go wrote:
> Hi Jianfeng,
>
> Thanks for the email. rte mempool was successfully created without any
> error. Now the next problem is that rte_eth_rx_burst() is always
> returning 0
> as if there was no packet to receive... Do you have any suggestion on
> what
> might be causing this issue? In the meantime, I will be digging
> through
> ixgbe driver code to see what's going on.
>
> Thank you,
> Younghwan
>
The problem is that with --no-huge we don't have the physical address
of the memory
to write to the network card. That's what it's marked as for testing
only.
>>> Even with rte_mem_virt2phy() + rte_mem_lock_page() ?
>>>
>> With no-huge, we just set up a single memory segment at startup and set
>> its
>> "physaddr" to be the virtual address.
>>
>> /* hugetlbfs can be disabled */
>> if (internal_config.no_hugetlbfs) {
>> addr = mmap(NULL, internal_config.memory, PROT_READ |
>> PROT_WRITE,
>> MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
>> if (addr == MAP_FAILED) {
>> RTE_LOG(ERR, EAL, "%s: mmap() failed: %s\n",
>> __func__,
>> strerror(errno));
>> return -1;
>> }
>> mcfg->memseg[0].phys_addr =
>> (phys_addr_t)(uintptr_t)addr;
> rte_mem_virt2phy() does not use memseg.phys_addr but /proc/self/pagemap:
>
> /*
>* the pfn (page frame number) are bits 0-54 (see
>* pagemap.txt in linux Documentation)
>*/
> physaddr = ((page & 0x7fULL) * page_size)
> + ((unsigned long)virtaddr % page_size);
>
Yes, you are right. I was not aware that that function was used as part of
the
mempool init, but now I see that "rte_mempool_virt2phy()" does indeed call
that
function if hugepages are disabled, so my bad.
>>> Do you think we could move --no-huge in the main section (not only for
>>> testing)?
>> Hi,
>>
>> I think the main issue is going to be the HW descriptors queues.
>> AFAIK drivers now call rte_eth_dma_zone_reserve, which is basically a
>> wrapper around
>> rte_memzone_reserve, to get a chunk of phys memory, and in the case of
>> --no-huge is
>> not going to be really phys contiguous.
>>
>> Ideally we would move and expand the functionality of dma_zone reserve
>> API to the EAL,
>> so we could detect what page size we have and set the boundary for such
>> page size.
>> dma_zone_reserve does something similar to work on Xen target by
>> reserving memzones
>> on 2MB boundary.
> With xen we have a special kernel driver that allocates physically continuous
> chunks of memory for us.
> So we can guarantee that each such chunk would be at least 2MB long.
> That's enough to allocate HW rings (max HW ring size for let say ixgbe is
> ~64KB).
> Here there is absolutely no guarantee that memory allocated by kernel will be
> memory continuous.
> Of course you can search though all pages that you allocated and most likely
> you'll find a continuous
> chunk big enough for that.
> Another problem - mbufs.
> You need to be sure that each mbuf doesn't cross page boundary
> (in case next page is not adjacent to current one).
> So you'll probably need to use rte_mempool_xmem_cr
[dpdk-dev] [PATCH] app/test: Fix compile issue with icc
app/test/test_cryptodev_perf.c(1837): error #192: unrecognized
character escape sequence
printf("\n%u\t%u\t\%u\t\t%u\t\t%u", dev_num, 0,
"\%u" is the root cause of this issue, just fix it.
Signed-off-by: Michael Qiu
---
app/test/test_cryptodev_perf.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/app/test/test_cryptodev_perf.c b/app/test/test_cryptodev_perf.c
index f0cca8b..1744e13 100644
--- a/app/test/test_cryptodev_perf.c
+++ b/app/test/test_cryptodev_perf.c
@@ -1834,7 +1834,7 @@ test_perf_crypto_qp_vary_burst_size(uint16_t dev_num)
num_received += burst_received;
}
- printf("\n%u\t%u\t\%u\t\t%u\t\t%u", dev_num, 0,
+ printf("\n%u\t%u\t%u\t\t%u\t\t%u", dev_num, 0,
num_sent, num_received, burst_size);
printf("\t\t%"PRIu64, retries);
printf("\t\t\t%"PRIu64, total_cycles/num_received);
--
1.9.3
[dpdk-dev] [PATCH] cryptodev: mark experimental state
On 11/25/2015 07:38 PM, Thomas Monjalon wrote: > The crypto API is in an early state. > It requires more discussions and experiments to declare it stable, > as discussed in http://dpdk.org/ml/archives/dev/2015-November/028634.html > > A documentation section will be required in the guides. > > Signed-off-by: Thomas Monjalon > --- [...] > --- a/config/common_bsdapp > +++ b/config/common_bsdapp > @@ -310,6 +310,7 @@ CONFIG_RTE_PMD_PACKET_PREFETCH=y > > # > # Compile generic crypto device library > +# EXPERIMENTAL: API may change without prior notice > # > CONFIG_RTE_LIBRTE_CRYPTODEV=y > CONFIG_RTE_LIBRTE_CRYPTODEV_DEBUG=n > diff --git a/config/common_linuxapp b/config/common_linuxapp > index eaad8d6..2866986 100644 > --- a/config/common_linuxapp > +++ b/config/common_linuxapp > @@ -319,6 +319,7 @@ CONFIG_RTE_PMD_PACKET_PREFETCH=y > > # > # Compile generic crypto device library > +# EXPERIMENTAL: API may change without prior notice > # > CONFIG_RTE_LIBRTE_CRYPTODEV=y > CONFIG_RTE_LIBRTE_CRYPTODEV_DEBUG=n [...] I think an experimental library which declares itself exempt from the ABI policy should not be compiled by default. That way anybody wanting to try it out will be forced to notice the experimental status. More generally / longer term, perhaps there should be a CONFIG_RTE_EXPERIMENTAL which wraps all experimental features and defaults to off. - Panu -
[dpdk-dev] [PATCH] cryptodev: mark experimental state
On 11/26/2015 09:39 AM, Panu Matilainen wrote: > On 11/25/2015 07:38 PM, Thomas Monjalon wrote: >> The crypto API is in an early state. >> It requires more discussions and experiments to declare it stable, >> as discussed in http://dpdk.org/ml/archives/dev/2015-November/028634.html >> >> A documentation section will be required in the guides. >> >> Signed-off-by: Thomas Monjalon >> --- > [...] >> --- a/config/common_bsdapp >> +++ b/config/common_bsdapp >> @@ -310,6 +310,7 @@ CONFIG_RTE_PMD_PACKET_PREFETCH=y >> >> # >> # Compile generic crypto device library >> +# EXPERIMENTAL: API may change without prior notice >> # >> CONFIG_RTE_LIBRTE_CRYPTODEV=y >> CONFIG_RTE_LIBRTE_CRYPTODEV_DEBUG=n >> diff --git a/config/common_linuxapp b/config/common_linuxapp >> index eaad8d6..2866986 100644 >> --- a/config/common_linuxapp >> +++ b/config/common_linuxapp >> @@ -319,6 +319,7 @@ CONFIG_RTE_PMD_PACKET_PREFETCH=y >> >> # >> # Compile generic crypto device library >> +# EXPERIMENTAL: API may change without prior notice >> # >> CONFIG_RTE_LIBRTE_CRYPTODEV=y >> CONFIG_RTE_LIBRTE_CRYPTODEV_DEBUG=n > [...] > > I think an experimental library which declares itself exempt from the > ABI policy should not be compiled by default. That way anybody wanting > to try it out will be forced to notice the experimental status. > > More generally / longer term, perhaps there should be a > CONFIG_RTE_EXPERIMENTAL which wraps all experimental features and > defaults to off. On a related note, librte_mbuf_offload cannot be built if CONFIG_RTE_LIBRTE_CRYPTODEV is disabled. Which seems to suggest its (at least currently) so tightly couple to cryptodev that perhaps it too should be marked experimental and default to off. - Panu -
[dpdk-dev] [PATCH] mk: fix the combined library problems by replacing it with a linker script
On 11/25/2015 06:08 PM, Stephen Hemminger wrote: > On Wed, 25 Nov 2015 10:38:48 +0200 > Panu Matilainen wrote: > >> On 11/25/2015 12:46 AM, Stephen Hemminger wrote: >>> On Tue, 24 Nov 2015 16:31:17 +0200 >>> Panu Matilainen wrote: >>> The physically linked-together combined library has been an increasing source of problems, as was predicted when library and symbol versioning was introduced. Replace the complex and fragile construction with a simple linker script which achieves the same without all the problems, remove the related kludges from eg mlx drivers. Since creating the linker script is practically zero cost, remove the config option and just create it always. Based on a patch by Sergio Gonzales Monroy, linker script approach initially suggested by Neil Horman. Suggested-by: Sergio Gonzalez Monroy Suggested-by: Neil Horman Signed-off-by: Panu Matilainen >>> >>> But it now means distros have to ship 20 libraries which seems like >>> a step back. >> >> That's how Fedora and RHEL are shipping it already and nobody has so >> much as noticed anything strange, much less complained about it. 20 >> libraries is but a drop in the ocean on a average distro. But more to >> the point, distros will prefer 50 working libraries over one that doesn't. >> >> The combined library as it is simply is no longer a viable option. >> Besides just being broken (witness the strange hacks people are coming >> up with to work around issues in it) its ugly because it basically gives >> the middle finger to all the effort going into version compatibility, >> and its also big. Few projects will use every library in DPDK, but with >> the combined library they're forced to lug the 800 pound gorilla along >> needlessly. >> >> - Panu - >> > > Fixing the combined library took less than an hour for us. > An hour (and many more in times to come) that I bet you could've used on something more interesting than fighting that abomination. - Panu -
[dpdk-dev] [PATCH v9 4/9] nfp: adding stats
Signed-off-by: Alejandro Lucero
Signed-off-by: Rolf Neugebauer
---
drivers/net/nfp/nfp_net.c | 179 +
1 file changed, 179 insertions(+)
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 8451a49..fc02916 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -90,6 +90,9 @@ static int nfp_net_tx_queue_setup(struct rte_eth_dev *dev,
uint16_t queue_idx,
uint16_t nb_desc, unsigned int socket_id,
const struct rte_eth_txconf *tx_conf);
static int nfp_net_start(struct rte_eth_dev *dev);
+static void nfp_net_stats_get(struct rte_eth_dev *dev,
+ struct rte_eth_stats *stats);
+static void nfp_net_stats_reset(struct rte_eth_dev *dev);
static void nfp_net_stop(struct rte_eth_dev *dev);
static uint16_t nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
uint16_t nb_pkts);
@@ -679,6 +682,177 @@ nfp_net_close(struct rte_eth_dev *dev)
*/
}
+static void
+nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
+{
+ int i;
+ struct nfp_net_hw *hw;
+ struct rte_eth_stats nfp_dev_stats;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ /* RTE_ETHDEV_QUEUE_STAT_CNTRS default value is 16 */
+
+ /* reading per RX ring stats */
+ for (i = 0; i < dev->data->nb_rx_queues; i++) {
+ if (i == RTE_ETHDEV_QUEUE_STAT_CNTRS)
+ break;
+
+ nfp_dev_stats.q_ipackets[i] =
+ nn_cfg_readq(hw, NFP_NET_CFG_RXR_STATS(i));
+
+ nfp_dev_stats.q_ipackets[i] -=
+ hw->eth_stats_base.q_ipackets[i];
+
+ nfp_dev_stats.q_ibytes[i] =
+ nn_cfg_readq(hw, NFP_NET_CFG_RXR_STATS(i) + 0x8);
+
+ nfp_dev_stats.q_ibytes[i] -=
+ hw->eth_stats_base.q_ibytes[i];
+ }
+
+ /* reading per TX ring stats */
+ for (i = 0; i < dev->data->nb_tx_queues; i++) {
+ if (i == RTE_ETHDEV_QUEUE_STAT_CNTRS)
+ break;
+
+ nfp_dev_stats.q_opackets[i] =
+ nn_cfg_readq(hw, NFP_NET_CFG_TXR_STATS(i));
+
+ nfp_dev_stats.q_opackets[i] -=
+ hw->eth_stats_base.q_opackets[i];
+
+ nfp_dev_stats.q_obytes[i] =
+ nn_cfg_readq(hw, NFP_NET_CFG_TXR_STATS(i) + 0x8);
+
+ nfp_dev_stats.q_obytes[i] -=
+ hw->eth_stats_base.q_obytes[i];
+ }
+
+ nfp_dev_stats.ipackets =
+ nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_FRAMES);
+
+ nfp_dev_stats.ipackets -= hw->eth_stats_base.ipackets;
+
+ nfp_dev_stats.ibytes =
+ nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_OCTETS);
+
+ nfp_dev_stats.ibytes -= hw->eth_stats_base.ibytes;
+
+ nfp_dev_stats.opackets =
+ nn_cfg_readq(hw, NFP_NET_CFG_STATS_TX_FRAMES);
+
+ nfp_dev_stats.opackets -= hw->eth_stats_base.opackets;
+
+ nfp_dev_stats.obytes =
+ nn_cfg_readq(hw, NFP_NET_CFG_STATS_TX_OCTETS);
+
+ nfp_dev_stats.obytes -= hw->eth_stats_base.obytes;
+
+ nfp_dev_stats.imcasts =
+ nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
+
+ nfp_dev_stats.imcasts -= hw->eth_stats_base.imcasts;
+
+ /* reading general device stats */
+ nfp_dev_stats.ierrors =
+ nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_ERRORS);
+
+ nfp_dev_stats.ierrors -= hw->eth_stats_base.ierrors;
+
+ nfp_dev_stats.oerrors =
+ nn_cfg_readq(hw, NFP_NET_CFG_STATS_TX_ERRORS);
+
+ nfp_dev_stats.oerrors -= hw->eth_stats_base.oerrors;
+
+ /* Multicast frames received */
+ nfp_dev_stats.imcasts =
+ nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
+
+ nfp_dev_stats.imcasts -= hw->eth_stats_base.imcasts;
+
+ /* RX ring mbuf allocation failures */
+ nfp_dev_stats.rx_nombuf = dev->data->rx_mbuf_alloc_failed;
+
+ nfp_dev_stats.imissed =
+ nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_DISCARDS);
+
+ nfp_dev_stats.imissed -= hw->eth_stats_base.imissed;
+
+ if (stats)
+ memcpy(stats, &nfp_dev_stats, sizeof(*stats));
+}
+
+static void
+nfp_net_stats_reset(struct rte_eth_dev *dev)
+{
+ int i;
+ struct nfp_net_hw *hw;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ /*
+* hw->eth_stats_base records the per counter starting point.
+* Lets update it now
+*/
+
+ /* reading per RX ring stats */
+ for (i = 0; i < dev->data->nb_rx_queues; i++) {
+ if (i == RTE_ETHDEV_QUEUE_STAT_CNTRS)
+ break;
+
+ hw->eth_stats_base.q_ipackets[i] =
+ nn_cfg_readq(hw, NFP_NET_CFG_RXR_STAT
[dpdk-dev] [PATCH v9 3/9] nfp: adding rss
Signed-off-by: Alejandro Lucero
Signed-off-by: Rolf Neugebauer
---
drivers/net/nfp/nfp_net.c | 218 +
1 file changed, 218 insertions(+)
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 7e30774..8451a49 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -1501,12 +1501,230 @@ xmit_end:
return i;
}
+/* Update Redirection Table(RETA) of Receive Side Scaling of Ethernet device */
+static int
+nfp_net_reta_update(struct rte_eth_dev *dev,
+ struct rte_eth_rss_reta_entry64 *reta_conf,
+ uint16_t reta_size)
+{
+ uint32_t reta, mask;
+ int i, j;
+ int idx, shift;
+ uint32_t update;
+ struct nfp_net_hw *hw =
+ NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS))
+ return -EINVAL;
+
+ if (reta_size != NFP_NET_CFG_RSS_ITBL_SZ) {
+ RTE_LOG(ERR, PMD, "The size of hash lookup table configured "
+ "(%d) doesn't match the number hardware can supported "
+ "(%d)\n", reta_size, NFP_NET_CFG_RSS_ITBL_SZ);
+ return -EINVAL;
+ }
+
+ /*
+* Update Redirection Table. There are 128 8bit-entries which can be
+* manage as 32 32bit-entries
+*/
+ for (i = 0; i < reta_size; i += 4) {
+ /* Handling 4 RSS entries per loop */
+ idx = i / RTE_RETA_GROUP_SIZE;
+ shift = i % RTE_RETA_GROUP_SIZE;
+ mask = (uint8_t)((reta_conf[idx].mask >> shift) & 0xF);
+
+ if (!mask)
+ continue;
+
+ reta = 0;
+ /* If all 4 entries were set, don't need read RETA register */
+ if (mask != 0xF)
+ reta = nn_cfg_readl(hw, NFP_NET_CFG_RSS_ITBL + i);
+
+ for (j = 0; j < 4; j++) {
+ if (!(mask & (0x1 << j)))
+ continue;
+ if (mask != 0xF)
+ /* Clearing the entry bits */
+ reta &= ~(0xFF << (8 * j));
+ reta |= reta_conf[idx].reta[shift + j] << (8 * j);
+ }
+ nn_cfg_writel(hw, NFP_NET_CFG_RSS_ITBL + shift, reta);
+ }
+
+ update = NFP_NET_CFG_UPDATE_RSS;
+
+ if (nfp_net_reconfig(hw, hw->ctrl, update) < 0)
+ return -EIO;
+
+ return 0;
+}
+
+ /* Query Redirection Table(RETA) of Receive Side Scaling of Ethernet device.
*/
+static int
+nfp_net_reta_query(struct rte_eth_dev *dev,
+ struct rte_eth_rss_reta_entry64 *reta_conf,
+ uint16_t reta_size)
+{
+ uint8_t i, j, mask;
+ int idx, shift;
+ uint32_t reta;
+ struct nfp_net_hw *hw;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS))
+ return -EINVAL;
+
+ if (reta_size != NFP_NET_CFG_RSS_ITBL_SZ) {
+ RTE_LOG(ERR, PMD, "The size of hash lookup table configured "
+ "(%d) doesn't match the number hardware can supported "
+ "(%d)\n", reta_size, NFP_NET_CFG_RSS_ITBL_SZ);
+ return -EINVAL;
+ }
+
+ /*
+* Reading Redirection Table. There are 128 8bit-entries which can be
+* manage as 32 32bit-entries
+*/
+ for (i = 0; i < reta_size; i += 4) {
+ /* Handling 4 RSS entries per loop */
+ idx = i / RTE_RETA_GROUP_SIZE;
+ shift = i % RTE_RETA_GROUP_SIZE;
+ mask = (uint8_t)((reta_conf[idx].mask >> shift) & 0xF);
+
+ if (!mask)
+ continue;
+
+ reta = nn_cfg_readl(hw, NFP_NET_CFG_RSS_ITBL + shift);
+ for (j = 0; j < 4; j++) {
+ if (!(mask & (0x1 << j)))
+ continue;
+ reta_conf->reta[shift + j] =
+ (uint8_t)((reta >> (8 * j)) & 0xF);
+ }
+ }
+ return 0;
+}
+
+static int
+nfp_net_rss_hash_update(struct rte_eth_dev *dev,
+ struct rte_eth_rss_conf *rss_conf)
+{
+ uint32_t update;
+ uint32_t cfg_rss_ctrl = 0;
+ uint8_t key;
+ uint64_t rss_hf;
+ int i;
+ struct nfp_net_hw *hw;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ rss_hf = rss_conf->rss_hf;
+
+ /* Checking if RSS is enabled */
+ if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS)) {
+ if (rss_hf != 0) { /* Enable RSS? */
+ RTE_LOG(ERR, PMD, "RSS unsupported\n");
+ return -EINVAL;
+ }
+ return 0; /* Nothing to do */
+ }
+
+ if (rss_conf->rss_key_len > NFP_NET_CFG_RSS_KEY_
[dpdk-dev] [PATCH v9 0/9] support for netronome nfp-6xxx card
This patchset adds a new PMD for Netronome nfp-6xxx card. Just PCI Virtual Functions supported. Using this PMD requires previous Netronome BSP installation. v9: - Adding flag RTE_PCI_DRV_INTR_LSC - Makefile changes for compilation as a shared library - Adding map file for linker version script info v8: - removing remaining unnecessary flags to PMD Makefile v7: - Adding support for link status changes interrupts. - removing unnecessary flags when compiling the PMD. v6: - Making each patch compilable. v5: - Splitting up patches per functionality. v4: - Getting rid of nfp_uio. Just submitting PMD. - Removing LSC interrupt support v3: - Making all patches independent for applying and building. - changing commits messages following standard v2: - Code style changes based on checkpatch.pl and DPDK style guide. - Documentation changes using the right rst format. - Moving the documentation files to a new patch file. - Adding info to MAINTAINERS and release files. Alejandro Lucero (9): nfp: basic initialization nfp: adding rx/tx functionality nfp: adding rss nfp: adding stats nfp: adding link functionality nfp: adding extra functionality nfp: link status change interrupt support nfp: adding nic guide nfp: updating maintainers MAINTAINERS |4 + config/common_linuxapp |6 + doc/guides/nics/index.rst |1 + doc/guides/nics/nfp.rst | 265 doc/guides/rel_notes/release_2_2.rst|4 + drivers/net/Makefile|1 + drivers/net/nfp/Makefile| 56 + drivers/net/nfp/nfp_net.c | 2499 +++ drivers/net/nfp/nfp_net_ctrl.h | 324 drivers/net/nfp/nfp_net_logs.h | 75 + drivers/net/nfp/nfp_net_pmd.h | 453 ++ drivers/net/nfp/rte_pmd_nfp_version.map |3 + mk/rte.app.mk |1 + 13 files changed, 3692 insertions(+) create mode 100644 doc/guides/nics/nfp.rst create mode 100644 drivers/net/nfp/Makefile create mode 100644 drivers/net/nfp/nfp_net.c create mode 100644 drivers/net/nfp/nfp_net_ctrl.h create mode 100644 drivers/net/nfp/nfp_net_logs.h create mode 100644 drivers/net/nfp/nfp_net_pmd.h create mode 100644 drivers/net/nfp/rte_pmd_nfp_version.map -- 1.7.9.5
[dpdk-dev] [PATCH v9 6/9] nfp: adding extra functionality
Signed-off-by: Alejandro Lucero
Signed-off-by: Rolf Neugebauer
---
drivers/net/nfp/nfp_net.c | 191 +
1 file changed, 191 insertions(+)
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 567ea26..5383f51 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -73,8 +73,13 @@
/* Prototypes */
static void nfp_net_close(struct rte_eth_dev *dev);
static int nfp_net_configure(struct rte_eth_dev *dev);
+static int nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
+static void nfp_net_infos_get(struct rte_eth_dev *dev,
+ struct rte_eth_dev_info *dev_info);
static int nfp_net_init(struct rte_eth_dev *eth_dev);
static int nfp_net_link_update(struct rte_eth_dev *dev, int wait_to_complete);
+static void nfp_net_promisc_enable(struct rte_eth_dev *dev);
+static void nfp_net_promisc_disable(struct rte_eth_dev *dev);
static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
static uint32_t nfp_net_rx_queue_count(struct rte_eth_dev *dev,
uint16_t queue_idx);
@@ -734,6 +739,65 @@ nfp_net_close(struct rte_eth_dev *dev)
*/
}
+static void
+nfp_net_promisc_enable(struct rte_eth_dev *dev)
+{
+ uint32_t new_ctrl, update = 0;
+ struct nfp_net_hw *hw;
+
+ PMD_DRV_LOG(DEBUG, "Promiscuous mode enable\n");
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ if (!(hw->cap & NFP_NET_CFG_CTRL_PROMISC)) {
+ PMD_INIT_LOG(INFO, "Promiscuous mode not supported\n");
+ return;
+ }
+
+ if (hw->ctrl & NFP_NET_CFG_CTRL_PROMISC) {
+ PMD_DRV_LOG(INFO, "Promiscuous mode already enabled\n");
+ return;
+ }
+
+ new_ctrl = hw->ctrl | NFP_NET_CFG_CTRL_PROMISC;
+ update = NFP_NET_CFG_UPDATE_GEN;
+
+ /*
+* DPDK sets promiscuous mode on just after this call assuming
+* it can not fail ...
+*/
+ if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
+ return;
+
+ hw->ctrl = new_ctrl;
+}
+
+static void
+nfp_net_promisc_disable(struct rte_eth_dev *dev)
+{
+ uint32_t new_ctrl, update = 0;
+ struct nfp_net_hw *hw;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ if ((hw->ctrl & NFP_NET_CFG_CTRL_PROMISC) == 0) {
+ PMD_DRV_LOG(INFO, "Promiscuous mode already disabled\n");
+ return;
+ }
+
+ new_ctrl = hw->ctrl & ~NFP_NET_CFG_CTRL_PROMISC;
+ update = NFP_NET_CFG_UPDATE_GEN;
+
+ /*
+* DPDK sets promiscuous mode off just before this call
+* assuming it can not fail ...
+*/
+ if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
+ return;
+
+ hw->ctrl = new_ctrl;
+}
+
/*
* return 0 means link status changed, -1 means not changed
*
@@ -948,6 +1012,65 @@ nfp_net_stats_reset(struct rte_eth_dev *dev)
nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_DISCARDS);
}
+static void
+nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
+{
+ struct nfp_net_hw *hw;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ dev_info->driver_name = dev->driver->pci_drv.name;
+ dev_info->max_rx_queues = (uint16_t)hw->max_rx_queues;
+ dev_info->max_tx_queues = (uint16_t)hw->max_tx_queues;
+ dev_info->min_rx_bufsize = ETHER_MIN_MTU;
+ dev_info->max_rx_pktlen = hw->mtu;
+ /* Next should change when PF support is implemented */
+ dev_info->max_mac_addrs = 1;
+
+ if (hw->cap & NFP_NET_CFG_CTRL_RXVLAN)
+ dev_info->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP;
+
+ if (hw->cap & NFP_NET_CFG_CTRL_RXCSUM)
+ dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_IPV4_CKSUM |
+DEV_RX_OFFLOAD_UDP_CKSUM |
+DEV_RX_OFFLOAD_TCP_CKSUM;
+
+ if (hw->cap & NFP_NET_CFG_CTRL_TXVLAN)
+ dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT;
+
+ if (hw->cap & NFP_NET_CFG_CTRL_TXCSUM)
+ dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_IPV4_CKSUM |
+DEV_RX_OFFLOAD_UDP_CKSUM |
+DEV_RX_OFFLOAD_TCP_CKSUM;
+
+ dev_info->default_rxconf = (struct rte_eth_rxconf) {
+ .rx_thresh = {
+ .pthresh = DEFAULT_RX_PTHRESH,
+ .hthresh = DEFAULT_RX_HTHRESH,
+ .wthresh = DEFAULT_RX_WTHRESH,
+ },
+ .rx_free_thresh = DEFAULT_RX_FREE_THRESH,
+ .rx_drop_en = 0,
+ };
+
+ dev_info->default_txconf = (struct rte_eth_txconf) {
+ .tx_thresh = {
+ .pthresh = DEFAULT_TX_PTHRESH,
+ .hthresh = DEFAULT_TX_HTHRESH,
+
[dpdk-dev] [PATCH v9 5/9] nfp: adding link functionality
Signed-off-by: Alejandro Lucero
Signed-off-by: Rolf Neugebauer
---
drivers/net/nfp/nfp_net.c | 96 +
1 file changed, 96 insertions(+)
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index fc02916..567ea26 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -74,6 +74,7 @@
static void nfp_net_close(struct rte_eth_dev *dev);
static int nfp_net_configure(struct rte_eth_dev *dev);
static int nfp_net_init(struct rte_eth_dev *eth_dev);
+static int nfp_net_link_update(struct rte_eth_dev *dev, int wait_to_complete);
static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
static uint32_t nfp_net_rx_queue_count(struct rte_eth_dev *dev,
uint16_t queue_idx);
@@ -226,6 +227,57 @@ ring_dma_zone_reserve(struct rte_eth_dev *dev, const char
*ring_name,
NFP_MEMZONE_ALIGN);
}
+/*
+ * Atomically reads link status information from global structure rte_eth_dev.
+ *
+ * @param dev
+ * - Pointer to the structure rte_eth_dev to read from.
+ * - Pointer to the buffer to be saved with the link status.
+ *
+ * @return
+ * - On success, zero.
+ * - On failure, negative value.
+ */
+static inline int
+nfp_net_dev_atomic_read_link_status(struct rte_eth_dev *dev,
+ struct rte_eth_link *link)
+{
+ struct rte_eth_link *dst = link;
+ struct rte_eth_link *src = &dev->data->dev_link;
+
+ if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
+ *(uint64_t *)src) == 0)
+ return -1;
+
+ return 0;
+}
+
+/*
+ * Atomically writes the link status information into global
+ * structure rte_eth_dev.
+ *
+ * @param dev
+ * - Pointer to the structure rte_eth_dev to read from.
+ * - Pointer to the buffer to be saved with the link status.
+ *
+ * @return
+ * - On success, zero.
+ * - On failure, negative value.
+ */
+static inline int
+nfp_net_dev_atomic_write_link_status(struct rte_eth_dev *dev,
+struct rte_eth_link *link)
+{
+ struct rte_eth_link *dst = &dev->data->dev_link;
+ struct rte_eth_link *src = link;
+
+ if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
+ *(uint64_t *)src) == 0)
+ return -1;
+
+ return 0;
+}
+
static void
nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq)
{
@@ -682,6 +734,49 @@ nfp_net_close(struct rte_eth_dev *dev)
*/
}
+/*
+ * return 0 means link status changed, -1 means not changed
+ *
+ * Wait to complete is needed as it can take up to 9 seconds to get the Link
+ * status.
+ */
+static int
+nfp_net_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete)
+{
+ struct nfp_net_hw *hw;
+ struct rte_eth_link link, old;
+ uint32_t nn_link_status;
+
+ PMD_DRV_LOG(DEBUG, "Link update\n");
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ memset(&old, 0, sizeof(old));
+ nfp_net_dev_atomic_read_link_status(dev, &old);
+
+ nn_link_status = nn_cfg_readl(hw, NFP_NET_CFG_STS);
+
+ memset(&link, 0, sizeof(struct rte_eth_link));
+
+ if (nn_link_status & NFP_NET_CFG_STS_LINK)
+ link.link_status = 1;
+
+ link.link_duplex = ETH_LINK_FULL_DUPLEX;
+ /* Other cards can limit the tx and rx rate per VF */
+ link.link_speed = ETH_LINK_SPEED_40G;
+
+ if (old.link_status != link.link_status) {
+ nfp_net_dev_atomic_write_link_status(dev, &link);
+ if (link.link_status)
+ PMD_DRV_LOG(INFO, "NIC Link is Up\n");
+ else
+ PMD_DRV_LOG(INFO, "NIC Link is Down\n");
+ return 0;
+ }
+
+ return -1;
+}
+
static void
nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
{
@@ -1895,6 +1990,7 @@ static struct eth_dev_ops nfp_net_eth_dev_ops = {
.dev_start = nfp_net_start,
.dev_stop = nfp_net_stop,
.dev_close = nfp_net_close,
+ .link_update= nfp_net_link_update,
.stats_get = nfp_net_stats_get,
.stats_reset= nfp_net_stats_reset,
.reta_update= nfp_net_reta_update,
--
1.7.9.5
[dpdk-dev] [PATCH v9 7/9] nfp: link status change interrupt support
Signed-off-by: Alejandro Lucero
Signed-off-by: Rolf Neugebauer
---
drivers/net/nfp/nfp_net.c | 123 +
1 file changed, 123 insertions(+)
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 5383f51..3763790 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -73,6 +73,9 @@
/* Prototypes */
static void nfp_net_close(struct rte_eth_dev *dev);
static int nfp_net_configure(struct rte_eth_dev *dev);
+static void nfp_net_dev_interrupt_handler(struct rte_intr_handle *handle,
+ void *param);
+static void nfp_net_dev_interrupt_delayed_handler(void *param);
static int nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
static void nfp_net_infos_get(struct rte_eth_dev *dev,
struct rte_eth_dev_info *dev_info);
@@ -731,6 +734,7 @@ nfp_net_close(struct rte_eth_dev *dev)
nfp_net_stop(dev);
+ rte_intr_disable(&dev->pci_dev->intr_handle);
nn_cfg_writeb(hw, NFP_NET_CFG_LSC, 0xff);
/*
@@ -1115,6 +1119,114 @@ nfp_net_rx_queue_count(struct rte_eth_dev *dev,
uint16_t queue_idx)
return count;
}
+static void
+nfp_net_dev_link_status_print(struct rte_eth_dev *dev)
+{
+ struct rte_eth_link link;
+
+ memset(&link, 0, sizeof(link));
+ nfp_net_dev_atomic_read_link_status(dev, &link);
+ if (link.link_status)
+ RTE_LOG(INFO, PMD, "Port %d: Link Up - speed %u Mbps - %s\n",
+ (int)(dev->data->port_id), (unsigned)link.link_speed,
+ link.link_duplex == ETH_LINK_FULL_DUPLEX
+ ? "full-duplex" : "half-duplex");
+ else
+ RTE_LOG(INFO, PMD, " Port %d: Link Down\n",
+ (int)(dev->data->port_id));
+
+ RTE_LOG(INFO, PMD, "PCI Address: %04d:%02d:%02d:%d\n",
+ dev->pci_dev->addr.domain, dev->pci_dev->addr.bus,
+ dev->pci_dev->addr.devid, dev->pci_dev->addr.function);
+}
+
+/* Interrupt configuration and handling */
+
+/*
+ * nfp_net_irq_unmask - Unmask an interrupt
+ *
+ * If MSI-X auto-masking is enabled clear the mask bit, otherwise
+ * clear the ICR for the entry.
+ */
+static void
+nfp_net_irq_unmask(struct rte_eth_dev *dev)
+{
+ struct nfp_net_hw *hw;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ if (hw->ctrl & NFP_NET_CFG_CTRL_MSIXAUTO) {
+ /* If MSI-X auto-masking is used, clear the entry */
+ rte_wmb();
+ rte_intr_enable(&dev->pci_dev->intr_handle);
+ } else {
+ /* Make sure all updates are written before un-masking */
+ rte_wmb();
+ nn_cfg_writeb(hw, NFP_NET_CFG_ICR(NFP_NET_IRQ_LSC_IDX),
+ NFP_NET_CFG_ICR_UNMASKED);
+ }
+}
+
+static void
+nfp_net_dev_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
+ void *param)
+{
+ int64_t timeout;
+ struct rte_eth_link link;
+ struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
+
+ PMD_DRV_LOG(DEBUG, "We got a LSC interrupt!!!\n");
+
+ /* get the link status */
+ memset(&link, 0, sizeof(link));
+ nfp_net_dev_atomic_read_link_status(dev, &link);
+
+ nfp_net_link_update(dev, 0);
+
+ /* likely to up */
+ if (!link.link_status) {
+ /* handle it 1 sec later, wait it being stable */
+ timeout = NFP_NET_LINK_UP_CHECK_TIMEOUT;
+ /* likely to down */
+ } else {
+ /* handle it 4 sec later, wait it being stable */
+ timeout = NFP_NET_LINK_DOWN_CHECK_TIMEOUT;
+ }
+
+ if (rte_eal_alarm_set(timeout * 1000,
+ nfp_net_dev_interrupt_delayed_handler,
+ (void *)dev) < 0) {
+ RTE_LOG(ERR, PMD, "Error setting alarm");
+ /* Unmasking */
+ nfp_net_irq_unmask(dev);
+ }
+}
+
+/*
+ * Interrupt handler which shall be registered for alarm callback for delayed
+ * handling specific interrupt to wait for the stable nic state. As the NIC
+ * interrupt state is not stable for nfp after link is just down, it needs
+ * to wait 4 seconds to get the stable status.
+ *
+ * @param handle Pointer to interrupt handle.
+ * @param paramThe address of parameter (struct rte_eth_dev *)
+ *
+ * @return void
+ */
+static void
+nfp_net_dev_interrupt_delayed_handler(void *param)
+{
+ struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
+
+ nfp_net_link_update(dev, 0);
+ _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
+
+ nfp_net_dev_link_status_print(dev);
+
+ /* Unmasking */
+ nfp_net_irq_unmask(dev);
+}
+
static int
nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
{
@@ -2315,6 +2427,17 @@ nfp_net_init(struct rte_eth_dev *eth_dev)
[dpdk-dev] [PATCH v9 8/9] nfp: adding nic guide
Signed-off-by: Alejandro Lucero Signed-off-by: Rolf Neugebauer --- doc/guides/nics/index.rst |1 + doc/guides/nics/nfp.rst | 265 + 2 files changed, 266 insertions(+) create mode 100644 doc/guides/nics/nfp.rst diff --git a/doc/guides/nics/index.rst b/doc/guides/nics/index.rst index 0a0b724..7bf2938 100644 --- a/doc/guides/nics/index.rst +++ b/doc/guides/nics/index.rst @@ -46,6 +46,7 @@ Network Interface Controller Drivers intel_vf mlx4 mlx5 +nfp szedata2 virtio vmxnet3 diff --git a/doc/guides/nics/nfp.rst b/doc/guides/nics/nfp.rst new file mode 100644 index 000..55ba64d --- /dev/null +++ b/doc/guides/nics/nfp.rst @@ -0,0 +1,265 @@ +.. BSD LICENSE +Copyright(c) 2015 Netronome Systems, Inc. All rights reserved. +All rights reserved. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: + +* Redistributions of source code must retain the above copyright +notice, this list of conditions and the following disclaimer. +* Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in +the documentation and/or other materials provided with the +distribution. +* Neither the name of Intel Corporation nor the names of its +contributors may be used to endorse or promote products derived +from this software without specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +NFP poll mode driver library + + +Netronome's sixth generation of flow processors pack 216 programmable +cores and over 100 hardware accelerators that uniquely combine packet, +flow, security and content processing in a single device that scales +up to 400 Gbps. + +This document explains how to use DPDK with the Netronome Poll Mode +Driver (PMD) supporting Netronome's Network Flow Processor 6xxx +(NFP-6xxx). + +Currently the driver supports virtual functions (VFs) only. + +Dependencies + + +Before using the Netronome's DPDK PMD some NFP-6xxx configuration, +which is not related to DPDK, is required. The system requires +installation of **Netronome's BSP (Board Support Package)** which includes +Linux drivers, programs and libraries. + +If you have a NFP-6xxx device you should already have the code and +documentation for doing this configuration. Contact +**support at netronome.com** to obtain the latest available firmware. + +The NFP Linux kernel drivers (including the required PF driver for the +NFP) are available on Github at +**https://github.com/Netronome/nfp-drv-kmods** along with build +instructions. + +DPDK runs in userspace and PMDs uses the Linux kernel UIO interface to +allow access to physical devices from userspace. The NFP PMD requires +a separate UIO driver, **nfp_uio**, to perform correct +initialization. This driver is part of Netronome?s BSP and it is +equivalent to Intel's igb_uio driver. + +Building the software +- + +Netronome's PMD code is provided in the **drivers/net/nfp** directory. +Because Netronome?s BSP dependencies the driver is disabled by default +in DPDK build using **common_linuxapp configuration** file. Enabling the +driver or if you use another configuration file and want to have NFP +support, this variable is needed: + +- **CONFIG_RTE_LIBRTE_NFP_PMD=y** + +Once DPDK is built all the DPDK apps and examples include support for +the NFP PMD. + + +System configuration + + +Using the NFP PMD is not different to using other PMDs. Usual steps are: + +#. **Configure hugepages:** All major Linux distributions have the hugepages + functionality enabled by default. By default this allows the system uses for + working with transparent hugepages. But in this case some hugepages need to + be created/reserved for use with the DPDK through the hugetlbfs file system. + First the virtual file system need to be mounted: + + .. code-block:: console + + mount -t hugetlbfs none /mnt/hugetlbfs + + The command uses the common mount point for this file system an
[dpdk-dev] [PATCH v9 2/9] nfp: adding rx/tx functionality
Signed-off-by: Alejandro Lucero
Signed-off-by: Rolf Neugebauer
---
drivers/net/nfp/nfp_net.c | 993 +
1 file changed, 993 insertions(+)
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 18067c0..7e30774 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -74,8 +74,25 @@
static void nfp_net_close(struct rte_eth_dev *dev);
static int nfp_net_configure(struct rte_eth_dev *dev);
static int nfp_net_init(struct rte_eth_dev *eth_dev);
+static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
+static uint32_t nfp_net_rx_queue_count(struct rte_eth_dev *dev,
+ uint16_t queue_idx);
+static uint16_t nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+ uint16_t nb_pkts);
+static void nfp_net_rx_queue_release(void *rxq);
+static int nfp_net_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+ uint16_t nb_desc, unsigned int socket_id,
+ const struct rte_eth_rxconf *rx_conf,
+ struct rte_mempool *mp);
+static int nfp_net_tx_free_bufs(struct nfp_net_txq *txq);
+static void nfp_net_tx_queue_release(void *txq);
+static int nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+ uint16_t nb_desc, unsigned int socket_id,
+ const struct rte_eth_txconf *tx_conf);
static int nfp_net_start(struct rte_eth_dev *dev);
static void nfp_net_stop(struct rte_eth_dev *dev);
+static uint16_t nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+ uint16_t nb_pkts);
/*
* The offset of the queue controller queues in the PCIe Target. These
@@ -186,6 +203,100 @@ nn_cfg_writeq(struct nfp_net_hw *hw, int off, uint64_t
val)
nn_writeq(rte_cpu_to_le_64(val), hw->ctrl_bar + off);
}
+/* Creating memzone for hardware rings. */
+static const struct rte_memzone *
+ring_dma_zone_reserve(struct rte_eth_dev *dev, const char *ring_name,
+ uint16_t queue_id, uint32_t ring_size, int socket_id)
+{
+ char z_name[RTE_MEMZONE_NAMESIZE];
+ const struct rte_memzone *mz;
+
+ snprintf(z_name, sizeof(z_name), "%s_%s_%d_%d",
+dev->driver->pci_drv.name,
+ring_name, dev->data->port_id, queue_id);
+
+ mz = rte_memzone_lookup(z_name);
+ if (mz)
+ return mz;
+
+ return rte_memzone_reserve_aligned(z_name, ring_size, socket_id, 0,
+ NFP_MEMZONE_ALIGN);
+}
+
+static void
+nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq)
+{
+ unsigned i;
+
+ if (rxq->rxbufs == NULL)
+ return;
+
+ for (i = 0; i < rxq->rx_count; i++) {
+ if (rxq->rxbufs[i].mbuf) {
+ rte_pktmbuf_free_seg(rxq->rxbufs[i].mbuf);
+ rxq->rxbufs[i].mbuf = NULL;
+ }
+ }
+}
+
+static void
+nfp_net_rx_queue_release(void *rx_queue)
+{
+ struct nfp_net_rxq *rxq = rx_queue;
+
+ if (rxq) {
+ nfp_net_rx_queue_release_mbufs(rxq);
+ rte_free(rxq->rxbufs);
+ rte_free(rxq);
+ }
+}
+
+static void
+nfp_net_reset_rx_queue(struct nfp_net_rxq *rxq)
+{
+ nfp_net_rx_queue_release_mbufs(rxq);
+ rxq->wr_p = 0;
+ rxq->rd_p = 0;
+ rxq->nb_rx_hold = 0;
+}
+
+static void
+nfp_net_tx_queue_release_mbufs(struct nfp_net_txq *txq)
+{
+ unsigned i;
+
+ if (txq->txbufs == NULL)
+ return;
+
+ for (i = 0; i < txq->tx_count; i++) {
+ if (txq->txbufs[i].mbuf) {
+ rte_pktmbuf_free_seg(txq->txbufs[i].mbuf);
+ txq->txbufs[i].mbuf = NULL;
+ }
+ }
+}
+
+static void
+nfp_net_tx_queue_release(void *tx_queue)
+{
+ struct nfp_net_txq *txq = tx_queue;
+
+ if (txq) {
+ nfp_net_tx_queue_release_mbufs(txq);
+ rte_free(txq->txbufs);
+ rte_free(txq);
+ }
+}
+
+static void
+nfp_net_reset_tx_queue(struct nfp_net_txq *txq)
+{
+ nfp_net_tx_queue_release_mbufs(txq);
+ txq->wr_p = 0;
+ txq->rd_p = 0;
+ txq->tail = 0;
+}
+
static int
__nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update)
{
@@ -423,6 +534,18 @@ nfp_net_disable_queues(struct rte_eth_dev *dev)
hw->ctrl = new_ctrl;
}
+static int
+nfp_net_rx_freelist_setup(struct rte_eth_dev *dev)
+{
+ int i;
+
+ for (i = 0; i < dev->data->nb_rx_queues; i++) {
+ if (nfp_net_rx_fill_freelist(dev->data->rx_queues[i]) < 0)
+ return -1;
+ }
+ return 0;
+}
+
static void
nfp_net_params_setup(struct nfp_net_hw *hw)
{
@@ -451,6 +574,7 @@ nfp_net_start(struct rte_eth_dev *dev)
{
uint32_t new_ctrl, update = 0;
[dpdk-dev] [PATCH v9 9/9] nfp: updating maintainers
Signed-off-by: Alejandro Lucero Signed-off-by: Rolf Neugebauer --- MAINTAINERS |1 + 1 file changed, 1 insertion(+) diff --git a/MAINTAINERS b/MAINTAINERS index df5b962..ece737a 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -317,6 +317,7 @@ F: app/test/test_pmd_ring_perf.c Null PMD M: Tetsuya Mukawa F: drivers/net/null/ +F: doc/guides/nics/nfp.rst Netronome nfp M: Alejandro Lucero -- 1.7.9.5
[dpdk-dev] [PATCH v9 1/9] nfp: basic initialization
Signed-off-by: Alejandro Lucero Signed-off-by: Rolf Neugebauer --- MAINTAINERS |3 + config/common_linuxapp |6 + doc/guides/rel_notes/release_2_2.rst|4 + drivers/net/Makefile|1 + drivers/net/nfp/Makefile| 56 +++ drivers/net/nfp/nfp_net.c | 699 +++ drivers/net/nfp/nfp_net_ctrl.h | 324 ++ drivers/net/nfp/nfp_net_logs.h | 75 drivers/net/nfp/nfp_net_pmd.h | 453 drivers/net/nfp/rte_pmd_nfp_version.map |3 + mk/rte.app.mk |1 + 11 files changed, 1625 insertions(+) create mode 100644 drivers/net/nfp/Makefile create mode 100644 drivers/net/nfp/nfp_net.c create mode 100644 drivers/net/nfp/nfp_net_ctrl.h create mode 100644 drivers/net/nfp/nfp_net_logs.h create mode 100644 drivers/net/nfp/nfp_net_pmd.h create mode 100644 drivers/net/nfp/rte_pmd_nfp_version.map diff --git a/MAINTAINERS b/MAINTAINERS index 840faeb..df5b962 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -318,6 +318,9 @@ Null PMD M: Tetsuya Mukawa F: drivers/net/null/ +Netronome nfp +M: Alejandro Lucero +F: drivers/net/nfp/ Packet processing - diff --git a/config/common_linuxapp b/config/common_linuxapp index 4a68da4..1d77db7 100644 --- a/config/common_linuxapp +++ b/config/common_linuxapp @@ -279,6 +279,12 @@ CONFIG_RTE_LIBRTE_VMXNET3_DEBUG_TX_FREE=n CONFIG_RTE_LIBRTE_VMXNET3_DEBUG_DRIVER=n # +# Compile burst-oriented Netronome NFP PMD driver +# +CONFIG_RTE_LIBRTE_NFP_PMD=n +CONFIG_RTE_LIBRTE_NFP_DEBUG=n + +# # Compile example software rings based PMD # CONFIG_RTE_LIBRTE_PMD_RING=y diff --git a/doc/guides/rel_notes/release_2_2.rst b/doc/guides/rel_notes/release_2_2.rst index 8c77768..8154db7 100644 --- a/doc/guides/rel_notes/release_2_2.rst +++ b/doc/guides/rel_notes/release_2_2.rst @@ -195,6 +195,10 @@ Drivers Fixed issue when releasing null control queue. +* **nfp: adding new PMD for Netronome nfp-6xxx card.** + + Support for using Netronome nfp-6xxx with PCI VFs. + Libraries ~ diff --git a/drivers/net/Makefile b/drivers/net/Makefile index cddcd57..6e4497e 100644 --- a/drivers/net/Makefile +++ b/drivers/net/Makefile @@ -43,6 +43,7 @@ DIRS-$(CONFIG_RTE_LIBRTE_IXGBE_PMD) += ixgbe DIRS-$(CONFIG_RTE_LIBRTE_MLX4_PMD) += mlx4 DIRS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5 DIRS-$(CONFIG_RTE_LIBRTE_MPIPE_PMD) += mpipe +DIRS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp DIRS-$(CONFIG_RTE_LIBRTE_PMD_NULL) += null DIRS-$(CONFIG_RTE_LIBRTE_PMD_PCAP) += pcap DIRS-$(CONFIG_RTE_LIBRTE_PMD_RING) += ring diff --git a/drivers/net/nfp/Makefile b/drivers/net/nfp/Makefile new file mode 100644 index 000..ef7a13d --- /dev/null +++ b/drivers/net/nfp/Makefile @@ -0,0 +1,56 @@ +# BSD LICENSE +# +# Copyright(c) 2010-2014 Intel Corporation. All rights reserved. +# All rights reserved. +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions +# are met: +# +# * Redistributions of source code must retain the above copyright +# notice, this list of conditions and the following disclaimer. +# * Redistributions in binary form must reproduce the above copyright +# notice, this list of conditions and the following disclaimer in +# the documentation and/or other materials provided with the +# distribution. +# * Neither the name of Intel Corporation nor the names of its +# contributors may be used to endorse or promote products derived +# from this software without specific prior written permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +include $(RTE_SDK)/mk/rte.vars.mk + +# +# library name +# +LIB = librte_pmd_nfp.a + +CFLAGS += -O3 +CFLAGS += $(WERROR_FLAGS) + +EXPORT_MAP := rte_pmd_nfp_version.map + +LIBABIVER := 1 + +# +# all source are stored in SRCS-y +# +SRCS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += nfp_net.c + +# this lib depends upon: +DEPDIRS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += lib/librte_eal lib/librte_ether +DEPDIRS-$(CONFIG_RTE_LIBRTE_NFP_PMD) += lib/librte_mempool lib/librte_
[dpdk-dev] [PATCH v2] examples/bond: fix bsd compile error
>-Original Message- >From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Ferruh Yigit >Sent: Wednesday, November 25, 2015 6:41 PM >To: dev at dpdk.org >Subject: [dpdk-dev] [PATCH v2] examples/bond: fix bsd compile error > >Error: >== bond > CC main.o >/.../examples/bond/main.c:431:24: error: use of undeclared identifier 'AF_INET' >if (res->ip.family == AF_INET) > ^ >1 error generated. >/.../mk/internal/rte.compile-pre.mk:126: recipe for target 'main.o' failed > >AF_INET defined in sys/socket.h > >This header included for Linux: >. //include/rte_ip.h >.. /usr/include/netinet/in.h >... /usr/include/sys/socket.h > >But not for FreeBSD: >. //include/rte_ip.h >.. /usr/include/netinet/in.h >... /usr/include/machine/endian.h >... /usr/include/netinet6/in6.h >. //include/rte_tcp.h > >Signed-off-by: Ferruh Yigit >--- > examples/bond/main.c | 1 + > 1 file changed, 1 insertion(+) > >diff --git a/examples/bond/main.c b/examples/bond/main.c >index 4622283..19f4f05 100644 >--- a/examples/bond/main.c >+++ b/examples/bond/main.c >@@ -45,6 +45,7 @@ > #include > #include > #include >+#include > > #include > #include >-- >2.5.0 Acked-by: Daniel Mrzyglod
[dpdk-dev] [PATCH] examples/bond: add header to support freebsd compilation
>-Original Message- >From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com] >Sent: Wednesday, November 25, 2015 7:08 PM > >2015-11-25 19:03, Daniel Mrzyglod: >> definition of 'AF_INET' enum was missing - is available in >> >> Signed-off-by: Daniel Mrzyglod > >It is definitely the right fix as Ferruh submitted the same one >less than one hour ago. >Should we understand it is an ack? > >It seems Ferruh was hesitating about the line where inserting the >include. I would say I prefer your choice :) I acked Ferruh patch. I didn't notice he send patch first. You can combine both patches :> He has much better description :>
[dpdk-dev] [PATCH] cryptodev: mark experimental state
2015-11-26 10:00, Panu Matilainen: > On 11/26/2015 09:39 AM, Panu Matilainen wrote: > > On 11/25/2015 07:38 PM, Thomas Monjalon wrote: > >> --- a/config/common_linuxapp > >> +++ b/config/common_linuxapp > >> @@ -319,6 +319,7 @@ CONFIG_RTE_PMD_PACKET_PREFETCH=y > >> > >> # > >> # Compile generic crypto device library > >> +# EXPERIMENTAL: API may change without prior notice > >> # > >> CONFIG_RTE_LIBRTE_CRYPTODEV=y > >> CONFIG_RTE_LIBRTE_CRYPTODEV_DEBUG=n > > [...] > > > > I think an experimental library which declares itself exempt from the > > ABI policy should not be compiled by default. That way anybody wanting > > to try it out will be forced to notice the experimental status. > > > > More generally / longer term, perhaps there should be a > > CONFIG_RTE_EXPERIMENTAL which wraps all experimental features and > > defaults to off. > > On a related note, librte_mbuf_offload cannot be built if > CONFIG_RTE_LIBRTE_CRYPTODEV is disabled. Which seems to suggest its (at > least currently) so tightly couple to cryptodev that perhaps it too > should be marked experimental and default to off. I think you are right. Declan, what is your opinion?
[dpdk-dev] [PATCH] lib/librte_sched: Fix compile with gcc 4.3.4
gcc 4.3.4 does not include "immintrin.h", and will post below error: lib/librte_sched/rte_sched.c:56:23: error: immintrin.h: No such file or directory To avoid this issue, a gcc version check is need and a flag to indicate vector ablility. Signed-off-by: Michael Qiu --- lib/librte_sched/rte_sched.c | 16 +++- 1 file changed, 15 insertions(+), 1 deletion(-) diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c index d47cfc2..780b314 100644 --- a/lib/librte_sched/rte_sched.c +++ b/lib/librte_sched/rte_sched.c @@ -53,7 +53,21 @@ #endif #ifdef RTE_SCHED_VECTOR + +#if (defined(__ICC) || (__GNUC__ == 4 && __GNUC_MINOR__ < 4)) + +#if defined(__AVX__) #include +#define SCHED_VECTOR_ENABLE +#endif + +#else + +#include +#define SCHED_VECTOR_ENABLE + +#endif + #endif #define RTE_SCHED_TB_RATE_CONFIG_ERR (1e-7) @@ -1667,7 +1681,7 @@ grinder_schedule(struct rte_sched_port *port, uint32_t pos) return 1; } -#ifdef RTE_SCHED_VECTOR +#ifdef SCHED_VECTOR_ENABLE static inline int grinder_pipe_exists(struct rte_sched_port *port, uint32_t base_pipe) -- 1.9.3
[dpdk-dev] remaining deprecation notices
> -Original Message- > From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com] > Sent: Thursday, November 26, 2015 12:44 AM > To: stephen at networkplumber.org; Dumitrescu, Cristian > > Cc: dev at dpdk.org > Subject: remaining deprecation notices > > Stephen, Cristian, > There are 2 remaining deprecation notices: > > * The scheduler statistics structure will change to allow keeping track of > RED actions. > > * librte_pipeline: The prototype for the pipeline input port, output port > and table action handlers will be updated: > the pipeline parameter will be added, the packets mask parameter will be > either removed (for input port action handler) or made input-only. > Due to the delay in applying the packet framework patches to dpdk.org mainline, we decided to postpone the implementation for this deprecation notice for release 2.3 in order to avoid an unpredictable negative effect on performance. We plan to do this work for 2.3 while keeping a close eye on performance impact, so let's keep this deprecation notice for now, thanks! > Shall we remove them?
[dpdk-dev] [PATCH] cryptodev: mark experimental state
> -Original Message- > From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com] > Sent: Thursday, November 26, 2015 10:09 AM > To: Panu Matilainen; Doherty, Declan > Cc: dev at dpdk.org > Subject: Re: [dpdk-dev] [PATCH] cryptodev: mark experimental state > > 2015-11-26 10:00, Panu Matilainen: > > On 11/26/2015 09:39 AM, Panu Matilainen wrote: > > > On 11/25/2015 07:38 PM, Thomas Monjalon wrote: > > >> --- a/config/common_linuxapp > > >> +++ b/config/common_linuxapp > > >> @@ -319,6 +319,7 @@ CONFIG_RTE_PMD_PACKET_PREFETCH=y > > >> > > >> # > > >> # Compile generic crypto device library > > >> +# EXPERIMENTAL: API may change without prior notice > > >> # > > >> CONFIG_RTE_LIBRTE_CRYPTODEV=y > > >> CONFIG_RTE_LIBRTE_CRYPTODEV_DEBUG=n > > > [...] > > > > > > I think an experimental library which declares itself exempt from the > > > ABI policy should not be compiled by default. That way anybody wanting > > > to try it out will be forced to notice the experimental status. > > > > > > More generally / longer term, perhaps there should be a > > > CONFIG_RTE_EXPERIMENTAL which wraps all experimental features and > > > defaults to off. > > > > On a related note, librte_mbuf_offload cannot be built if > > CONFIG_RTE_LIBRTE_CRYPTODEV is disabled. Which seems to suggest its (at > > least currently) so tightly couple to cryptodev that perhaps it too > > should be marked experimental and default to off. > > I think you are right. > Declan, what is your opinion? Hey Thomas, yes librte_mbuf_offload should also be set as experimental, it's probably one of the areas which will most likely change in the future. On the issue of turning off experimental libraries in the build by default, my preference would be not to turn them off unless the library has external dependencies, otherwise the possibility of patches being submitted which could break an experimental library will be much higher. In my opinion the fewer build configurations developers have to test against the better.
[dpdk-dev] [PATCH 1/2] examples/distributor: Fix compile issue
Hi Michael, Thank you very much for working on compilation issues. Please could you try to insert a "Fixes:" line in your logs?
[dpdk-dev] no hugepage with UIO poll-mode driver
On Wed, Nov 25, 2015 at 11:47 PM, Younghwan Go wrote: > Hello, > > Thank you all for helping us understand on issues with no hugepage option. > > As Konstantin mentioned at the end, I tried using VFIO module instead of IGB > UIO module. I enabled all necessary parameters (IOMMU, virtualization, > vfio-pci, VFIO permission) and ran my code with no hugepage option. > > At first, it seemed to receive packets fine, but after a while, it stopped > receiving packets. I could temporarily remove this issue by not calling > rte_eth_tx_burst(). Also, when I looked at the received packets, they all > contained 0s instead of actual data. In my experience, if there's all zeroes it's because the NIC didn't actually write any packet data. In my case, to prove that, I initialized the mbuf data to a well-known pattern, and saw that after packet reception, the mbuf data was still that pattern, not the data I actually sent in the packet. From my analysis, I concluded that the ixgbe NIC was failing to actually write the data to memory because the corresponding PCI transaction was failing. The PCI transaction failed because something was going wrong with how IOMMU was mapping my VM's guest physical address to the host's physical address, and how that interacted with the PCI tree. I don't know if that's what you're seeing. However, it may be that the physical address being given to the NIC for the mbuf is invalid, or cannot be properly mapped to a real physical address if you're using IOMMU. When this happens, can you check the pcie AER registers on the NIC, if they exist? Maybe they'll show that something happened. Thanks, Kyle > Was there anything that I missed in > running with VFIO? I'm curious if no hugepage with no hugepage option was > confirmed to run with VFIO. > > Thank you, > Younghwan > > 2015-11-25 ?? 11:12? Ananyev, Konstantin ?(?) ? ?: >
[dpdk-dev] [PATCH v9 1/9] nfp: basic initialization
On Thu, 26 Nov 2015 09:49:21 +
Alejandro Lucero wrote:
> +static inline void
> +nfp_qcp_ptr_add(__u8 *q, enum nfp_qcp_ptr ptr, uint32_t val)
> +{
> + uint32_t off;
Minor nit. why mix use of Linux specific basic size typedefs (__u8)
with Posix standard values (uint32_t). The DPDK style is to use
the Posix types except in kernel drivers or code that is coming
from unified drivers (ie ixgbe/base)
[dpdk-dev] API feature check _HAS_
When introducing LRO, Vlad has defined the macro RTE_ETHDEV_HAS_LRO_SUPPORT: http://dpdk.org/browse/dpdk/commit/lib/librte_ether/rte_ethdev.h?id=8eecb329 It allows to use the feature without version check (before the release or after a backport). Do you think it is useful? Should we define other macros RTE_[API]_HAS_[FEATURE] for each new feature or API change? It's time to fix it before releasing the 2.2 version.
[dpdk-dev] [PATCH 1/2] examples/distributor: Fix compile issue
2015-11-26 09:35, Michael Qiu: > examples/distributor/main.c(338): error #167: > argument of type "struct rte_mbuf *" > is incompatible with parameter of type "const char *" > _mm_prefetch(bufs[0], 0); > > The first param of _mm_prefetch should be "const char *" and > need convert "struct rte_mbuf *" to "const void *". [...] > - _mm_prefetch(bufs[0], 0); > - _mm_prefetch(bufs[1], 0); > - _mm_prefetch(bufs[2], 0); > + _mm_prefetch((const void *)bufs[0], 0); > + _mm_prefetch((const void *)bufs[1], 0); > + _mm_prefetch((const void *)bufs[2], 0); With clang: error: cast from 'const void *' to 'void *' drops const qualifier
[dpdk-dev] [PATCH 2/2] Fix compile issue in i686 platform
2015-11-26 09:35, Michael Qiu:
> In i686 platform, long is 32bit, so XXX_CYCLECOUNTER_MASK
> need define as 'ULL'
>
> Signed-off-by: Michael Qiu
This patch is correct but the description is not exact:
I have no issue with my i686 compiler.
For future reference, please could you be more precise
about the reproduction environment? Is it related to a specific compiler?
We also need to add these lines:
Fixes: 9c857bf6be87 ("igb: support ieee1588 functions for device time")
Fixes: 1c4445e1f28e ("ixgbe: support ieee1588 functions for device time")
Fixes: f3a4e40eca0c ("i40e: support ieee1588 functions for device time")
[dpdk-dev] [PATCH] app/test: Fix compile issue with icc
2015-11-26 15:21, Michael Qiu:
> app/test/test_cryptodev_perf.c(1837): error #192: unrecognized
> character escape sequence
> printf("\n%u\t%u\t\%u\t\t%u\t\t%u", dev_num, 0,
>
> "\%u" is the root cause of this issue, just fix it.
>
> Signed-off-by: Michael Qiu
Fixes: 202d375c60bc ("app/test: add cryptodev unit and performance tests")
Applied, thanks
[dpdk-dev] [PATCH] lib/librte_sched: Fix compile with gcc 4.3.4
2015-11-26 18:49, Michael Qiu: > gcc 4.3.4 does not include "immintrin.h", and will post below error: > lib/librte_sched/rte_sched.c:56:23: error: > immintrin.h: No such file or directory > > To avoid this issue, a gcc version check is need and a flag to indicate > vector ablility. [...] > +#if (defined(__ICC) || (__GNUC__ == 4 && __GNUC_MINOR__ < 4)) > + > +#if defined(__AVX__) > #include > +#define SCHED_VECTOR_ENABLE > +#endif > + > +#else > + > +#include > +#define SCHED_VECTOR_ENABLE > + > +#endif This kind of complication is managed by EAL. I think we should include rte_vect.h.
[dpdk-dev] Issues with running multicore
On Wed, Nov 25, 2015 at 12:59:01PM -0500, Sugumaran, Varthamanan wrote:
> Hi All,
> Iam trying to create multi core application by mapping separate function for
> each core.
> I had followed Multi process example and I had used pthread.
> The use case is:
>
> 1. In a single process, launch pthread and run rte_eal_init
>
> 2. Once rte_eal_init is done, create 3 pthreads
>
> 3. Launch master core.
>
> 4. In each of 3 pthread, launch slave core.
>
> The issue is, I don't see 4 cores running. Instead I see only single core
> running.
>
> Here is the Pseudo code:
>
> Init_pmd{
>ret = rte_eal_init(args, argv) ==> I had set the number of cores to 4 in
> the params.
Just to clarify here, what did you actually use as the coremask. If you used
-c 4, then that will cause only a single thread to be used running on core 2.
>
> pthread_create(&thread1, NULL, (void*)threadFun1, NULL);
> pthread_create(&thread2, NULL, (void*)threadFun2, NULL);
> pthread_create(&thread3, NULL, (void*)threadFun3, NULL);
These threads are separate from any threads used by DPDK, and DPDK will not use
them for anything, or assign them to any lcores. Unless you configure them
otherwise in your code, they will run on the same physical core as the master
thread.
>
>rte_eal_mp_remote_launch(launch_pmd_single_lcore, NULL, CALL_MASTER);
>
This call will only launch functions on the threads created by DPDK - not the
other threads you created above.
> }
>
> Iam mapping each core to threadFun1, threadFun2 and threadFun3.
>
> threadFun1
> {
>RTE_LCORE_FOREACH_SLAVE(lcore_id)
>{
> if( lcore_id == 1 )
> rte_eal_remote_launch(threadFun1_launch_one_lcore, NULL, lcore_id);
>}
> }
>
> Has anyone seen this issue?
> Please let me know if iam missing something?
I'm not sure your understanding of how DPDK threads work is correct. See my
comments above, and perhaps check how things are done in a number of the DPDK
sample applications to see if what you want can be done without additional
pthread calls in your application.
/Bruce
>
> Thanks
> Vartha
>
[dpdk-dev] Problems with SR-IOV and transparent vlan under DPDK
Hi, I wonder if anyone might have advice in relation to a major problem we're encountering. We have an SR-IOV setup on Debian 8 with 4 VFs. Each VF is configured with transparent vlan 1000. We have 4 VMs via KVM. 2xCentos (C1, C2) and 2xDebian8 (VR1, VR2) with DPDK based dataplane. C1 can ping correctly to C2. However, when C1 pings VR2, it fails because the received packets on VR2 are tagged with the "transparent" vlan 1000. VR2 is not configured for vlans and discards the packets. All VMs are configured identically from the KVM perspective. SR-IOV VFs are attached using: "virsh attach-device " where xmlfile is of the form: "ip link show" shows the various VFs with the associated vlans. $ ip link | grep vlan vf 0 MAC 52:54:00:68:cf:67, vlan 1000, spoof checking off, link-state auto vf 1 MAC 52:54:00:33:22:1a, vlan 1000, spoof checking off, link-state auto vf 2 MAC ba:10:05:10:00:03, vlan 1000, spoof checking off, link-state auto vf 3 MAC ba:10:05:10:00:04, vlan 1000, spoof checking off, link-state auto vf 0 MAC 52:54:00:fe:e0:6e, vlan 1000, spoof checking off, link-state auto vf 1 MAC 52:54:00:9c:f8:96, vlan 1000, spoof checking off, link-state auto vf 2 MAC ba:10:05:10:01:03, vlan 1000, spoof checking off, link-state auto vf 3 MAC ba:10:05:10:01:04, vlan 1000, spoof checking off, link-state auto Wireshark on the Centos VMs show the correct packets being received but show the packets with the incorrect vlan header on the DPDK based VR1/2. So my question is: How should DPDK deal with the transparent vlan handling ? Where should it be notifying the hardware to apply and strip the transparent vlan tags on traffic from and to the VMs ? Any insights gratefuly received. Thanks, Tom
