DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH] drivers/net/i40e/i40e_fdir.c: Improved i40e FDIR programming times
@ 2017-04-26 11:15 Michael Lilja
  2017-04-28  7:54 ` Wu, Jingjing
  0 siblings, 1 reply; 2+ messages in thread
From: Michael Lilja @ 2017-04-26 11:15 UTC (permalink / raw)
  To: elin.zhang, jingjing.wu; +Cc: dev


During my work (https://www.napatech.com/hw-acceleration-via-rte_flow/)on a flowtable application example that use rte_flow I discovered that the rte_flow programming times on a i40e was +11ms. The patch below result in an average programming time of 22usec with a max of 60usec instead of +11ms.

Could the following patch be useful? There might be a good reason for the original code, I'm unable to tell, so I will let it up to the maintainer to decide.


Signed-off-by: Michael Lilja <ml@napatech.com>
---
drivers/net/i40e/i40e_fdir.c | 16 +++++++++++-----
1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/drivers/net/i40e/i40e_fdir.c b/drivers/net/i40e/i40e_fdir.c
index 28cc554f5..a1ec66063 100644
--- a/drivers/net/i40e/i40e_fdir.c
+++ b/drivers/net/i40e/i40e_fdir.c
@@ -1296,23 +1296,29 @@ i40e_fdir_filter_programming(struct i40e_pf *pf,
     rte_wmb();
     I40E_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);
-     for (i = 0; i < I40E_FDIR_WAIT_COUNT; i++) {
-           rte_delay_us(I40E_FDIR_WAIT_INTERVAL_US);
+     for (i = 0; i < (I40E_FDIR_WAIT_COUNT * I40E_FDIR_WAIT_INTERVAL_US); i++) {
           if ((txdp->cmd_type_offset_bsz &
                       rte_cpu_to_le_64(I40E_TXD_QW1_DTYPE_MASK)) ==
                       rte_cpu_to_le_64(I40E_TX_DESC_DTYPE_DESC_DONE))
                 break;
+           rte_delay_us(1);
     }
-     if (i >= I40E_FDIR_WAIT_COUNT) {
+     if (i >= (I40E_FDIR_WAIT_COUNT * I40E_FDIR_WAIT_INTERVAL_US)) {
           PMD_DRV_LOG(ERR, "Failed to program FDIR filter:"
                     " time out to get DD on tx queue.");
           return -ETIMEDOUT;
     }
     /* totally delay 10 ms to check programming status*/
-     rte_delay_us((I40E_FDIR_WAIT_COUNT - i) * I40E_FDIR_WAIT_INTERVAL_US);
+  uint32_t w;
+     for (w = 0; w < (I40E_FDIR_WAIT_COUNT * I40E_FDIR_WAIT_INTERVAL_US); w++) {
+           if (i40e_check_fdir_programming_status(rxq) >= 0) {
+                 break;
+           }
+           rte_delay_us(1);
+     }
     if (i40e_check_fdir_programming_status(rxq) < 0) {
           PMD_DRV_LOG(ERR, "Failed to program FDIR filter:"
-                     " programming status reported.");
+                           " programming status reported.");
           return -ENOSYS;
     }
--
Disclaimer: This email and any files transmitted with it may contain confidential information intended for the addressee(s) only. The information is not to be surrendered or copied to unauthorized persons. If you have received this communication in error, please notify the sender immediately and delete this e-mail from your system.

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [dpdk-dev] [PATCH] drivers/net/i40e/i40e_fdir.c: Improved i40e FDIR programming times
  2017-04-26 11:15 [dpdk-dev] [PATCH] drivers/net/i40e/i40e_fdir.c: Improved i40e FDIR programming times Michael Lilja
@ 2017-04-28  7:54 ` Wu, Jingjing
  0 siblings, 0 replies; 2+ messages in thread
From: Wu, Jingjing @ 2017-04-28  7:54 UTC (permalink / raw)
  To: Michael Lilja, elin.zhang; +Cc: dev

Hi, Lijia

Thanks for the patch? Are you using git send-email.
The mail is html format, not friendly to discuss by mail thread.

About the change, just one comments:

>+  uint32_t w;
Could you move it above?

Thanks

From: Michael Lilja [mailto:ml@napatech.com] 
Sent: Wednesday, April 26, 2017 7:15 PM
To: elin.zhang@intel.com; Wu, Jingjing <jingjing.wu@intel.com>
Cc: dev@dpdk.org
Subject: [PATCH] drivers/net/i40e/i40e_fdir.c: Improved i40e FDIR programming times


During my work (https://www.napatech.com/hw-acceleration-via-rte_flow/)on a flowtable application example that use rte_flow I discovered that the rte_flow programming times on a i40e was +11ms. The patch below result in an average programming time of 22usec with a max of 60usec instead of +11ms.

Could the following patch be useful? There might be a good reason for the original code, I'm unable to tell, so I will let it up to the maintainer to decide.  


Signed-off-by: Michael Lilja <ml@napatech.com>
---
drivers/net/i40e/i40e_fdir.c | 16 +++++++++++-----
1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/drivers/net/i40e/i40e_fdir.c b/drivers/net/i40e/i40e_fdir.c
index 28cc554f5..a1ec66063 100644
--- a/drivers/net/i40e/i40e_fdir.c
+++ b/drivers/net/i40e/i40e_fdir.c
@@ -1296,23 +1296,29 @@ i40e_fdir_filter_programming(struct i40e_pf *pf,
     rte_wmb();
     I40E_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);

-     for (i = 0; i < I40E_FDIR_WAIT_COUNT; i++) {
-           rte_delay_us(I40E_FDIR_WAIT_INTERVAL_US);
+     for (i = 0; i < (I40E_FDIR_WAIT_COUNT * I40E_FDIR_WAIT_INTERVAL_US); i++) {
           if ((txdp->cmd_type_offset_bsz &
                       rte_cpu_to_le_64(I40E_TXD_QW1_DTYPE_MASK)) ==
                       rte_cpu_to_le_64(I40E_TX_DESC_DTYPE_DESC_DONE))
                 break;
+           rte_delay_us(1);
     }
-     if (i >= I40E_FDIR_WAIT_COUNT) {
+     if (i >= (I40E_FDIR_WAIT_COUNT * I40E_FDIR_WAIT_INTERVAL_US)) {
           PMD_DRV_LOG(ERR, "Failed to program FDIR filter:"
                     " time out to get DD on tx queue.");
           return -ETIMEDOUT;
     }
     /* totally delay 10 ms to check programming status*/
-     rte_delay_us((I40E_FDIR_WAIT_COUNT - i) * I40E_FDIR_WAIT_INTERVAL_US);
+  uint32_t w;
+     for (w = 0; w < (I40E_FDIR_WAIT_COUNT * I40E_FDIR_WAIT_INTERVAL_US); w++) {
+           if (i40e_check_fdir_programming_status(rxq) >= 0) {
+                 break;
+           }
+           rte_delay_us(1);
+     }
     if (i40e_check_fdir_programming_status(rxq) < 0) {
           PMD_DRV_LOG(ERR, "Failed to program FDIR filter:"
-                     " programming status reported.");
+                           " programming status reported.");
           return -ENOSYS;
     }

-- 
Disclaimer: This email and any files transmitted with it may contain confidential information intended for the addressee(s) only. The information is not to be surrendered or copied to unauthorized persons. If you have received this communication in error, please notify the sender immediately and delete this e-mail from your system. 

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2017-04-28  7:54 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-04-26 11:15 [dpdk-dev] [PATCH] drivers/net/i40e/i40e_fdir.c: Improved i40e FDIR programming times Michael Lilja
2017-04-28  7:54 ` Wu, Jingjing

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).