[PATCH intel-net] i40e: fix receiving of single packets in xsk zero-copy mode

Subsystems: intel ethernet drivers, networking drivers, the rest

STALE1985d

4 messages, 3 authors, 2021-03-31 · open the first message on its own page

[PATCH intel-net] i40e: fix receiving of single packets in xsk zero-copy mode

From: Magnus Karlsson <hidden>
Date: 2021-03-19 09:45:25

From: Magnus Karlsson <magnus.karlsson@intel.com>

Fix so that single packets are received immediately instead of in
batches of 8. If you sent 1 pss to a system, you received 8 packets
every 8 seconds instead of 1 packet every second. The problem behind
this was that the work_done reporting from the Tx part of the driver
was broken. The work_done reporting in i40e controls not only the
reporting back to the napi logic but also the setting of the interrupt
throttling logic. When Tx or Rx reports that it has more to do,
interrupts are throttled or coalesced and when they both report that
they are done, interrupts are armed right away. If the wrong work_done
value is returned, the logic will start to throttle interrupts in a
situation where it should have just enabled them. This leads to the
undesired batching behavior seen in user-space.

Fix this by returning the correct boolean value from the Tx xsk
zero-copy path. Return true if there is nothing to do or if we got
fewer packets to process than we asked for. Return false if we got as
many packets as the budget since there might be more packets we can
process.

Fixes: 3106c580fb7c ("i40e: Use batched xsk Tx interfaces to increase performance")
Reported-by: Sreedevi Joshi <redacted>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_xsk.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
index fc32c5019b0f..12ca84113587 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
@@ -471,7 +471,7 @@ static bool i40e_xmit_zc(struct i40e_ring *xdp_ring, unsigned int budget)
 
 	nb_pkts = xsk_tx_peek_release_desc_batch(xdp_ring->xsk_pool, descs, budget);
 	if (!nb_pkts)
-		return false;
+		return true;
 
 	if (xdp_ring->next_to_use + nb_pkts >= xdp_ring->count) {
 		nb_processed = xdp_ring->count - xdp_ring->next_to_use;
@@ -488,7 +488,7 @@ static bool i40e_xmit_zc(struct i40e_ring *xdp_ring, unsigned int budget)
 
 	i40e_update_tx_stats(xdp_ring, nb_pkts, total_bytes);
 
-	return true;
+	return nb_pkts < budget;
 }
 
 /**
base-commit: c79a707072fe3fea0e3c92edee6ca85c1e53c29f
-- 
2.29.0

Re: [PATCH intel-net] i40e: fix receiving of single packets in xsk zero-copy mode

From: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Date: 2021-03-19 11:11:25

On Fri, Mar 19, 2021 at 10:44:10AM +0100, Magnus Karlsson wrote:
From: Magnus Karlsson <magnus.karlsson@intel.com>

Fix so that single packets are received immediately instead of in
batches of 8. If you sent 1 pss to a system, you received 8 packets
pps?
every 8 seconds instead of 1 packet every second. The problem behind
this was that the work_done reporting from the Tx part of the driver
was broken. The work_done reporting in i40e controls not only the
reporting back to the napi logic but also the setting of the interrupt
throttling logic. When Tx or Rx reports that it has more to do,
interrupts are throttled or coalesced and when they both report that
they are done, interrupts are armed right away. If the wrong work_done
value is returned, the logic will start to throttle interrupts in a
situation where it should have just enabled them. This leads to the
undesired batching behavior seen in user-space.

Fix this by returning the correct boolean value from the Tx xsk
zero-copy path. Return true if there is nothing to do or if we got
fewer packets to process than we asked for. Return false if we got as
many packets as the budget since there might be more packets we can
process.

Fixes: 3106c580fb7c ("i40e: Use batched xsk Tx interfaces to increase performance")
Reported-by: Sreedevi Joshi <redacted>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
quoted hunk
---
 drivers/net/ethernet/intel/i40e/i40e_xsk.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
index fc32c5019b0f..12ca84113587 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
@@ -471,7 +471,7 @@ static bool i40e_xmit_zc(struct i40e_ring *xdp_ring, unsigned int budget)
 
 	nb_pkts = xsk_tx_peek_release_desc_batch(xdp_ring->xsk_pool, descs, budget);
 	if (!nb_pkts)
-		return false;
+		return true;
 
 	if (xdp_ring->next_to_use + nb_pkts >= xdp_ring->count) {
 		nb_processed = xdp_ring->count - xdp_ring->next_to_use;
@@ -488,7 +488,7 @@ static bool i40e_xmit_zc(struct i40e_ring *xdp_ring, unsigned int budget)
 
 	i40e_update_tx_stats(xdp_ring, nb_pkts, total_bytes);
 
-	return true;
+	return nb_pkts < budget;
 }
 
 /**
base-commit: c79a707072fe3fea0e3c92edee6ca85c1e53c29f
-- 
2.29.0

Re: [PATCH intel-net] i40e: fix receiving of single packets in xsk zero-copy mode

From: Magnus Karlsson <hidden>
Date: 2021-03-19 12:15:53

On Fri, Mar 19, 2021 at 12:10 PM Maciej Fijalkowski
[off-list ref] wrote:
On Fri, Mar 19, 2021 at 10:44:10AM +0100, Magnus Karlsson wrote:
quoted
From: Magnus Karlsson <magnus.karlsson@intel.com>

Fix so that single packets are received immediately instead of in
batches of 8. If you sent 1 pss to a system, you received 8 packets
pps?
Arghh, yes it should be pps, i.e. packets per second. I wonder what pss is?
quoted
every 8 seconds instead of 1 packet every second. The problem behind
this was that the work_done reporting from the Tx part of the driver
was broken. The work_done reporting in i40e controls not only the
reporting back to the napi logic but also the setting of the interrupt
throttling logic. When Tx or Rx reports that it has more to do,
interrupts are throttled or coalesced and when they both report that
they are done, interrupts are armed right away. If the wrong work_done
value is returned, the logic will start to throttle interrupts in a
situation where it should have just enabled them. This leads to the
undesired batching behavior seen in user-space.

Fix this by returning the correct boolean value from the Tx xsk
zero-copy path. Return true if there is nothing to do or if we got
fewer packets to process than we asked for. Return false if we got as
many packets as the budget since there might be more packets we can
process.

Fixes: 3106c580fb7c ("i40e: Use batched xsk Tx interfaces to increase performance")
Reported-by: Sreedevi Joshi <redacted>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
quoted
---
 drivers/net/ethernet/intel/i40e/i40e_xsk.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
index fc32c5019b0f..12ca84113587 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
@@ -471,7 +471,7 @@ static bool i40e_xmit_zc(struct i40e_ring *xdp_ring, unsigned int budget)

      nb_pkts = xsk_tx_peek_release_desc_batch(xdp_ring->xsk_pool, descs, budget);
      if (!nb_pkts)
-             return false;
+             return true;

      if (xdp_ring->next_to_use + nb_pkts >= xdp_ring->count) {
              nb_processed = xdp_ring->count - xdp_ring->next_to_use;
@@ -488,7 +488,7 @@ static bool i40e_xmit_zc(struct i40e_ring *xdp_ring, unsigned int budget)

      i40e_update_tx_stats(xdp_ring, nb_pkts, total_bytes);

-     return true;
+     return nb_pkts < budget;
 }

 /**
base-commit: c79a707072fe3fea0e3c92edee6ca85c1e53c29f
--
2.29.0

RE: [Intel-wired-lan] [PATCH intel-net] i40e: fix receiving of single packets in xsk zero-copy mode

From: Bhandare, KiranX <hidden>
Date: 2021-03-31 12:58:31

-----Original Message-----
From: Intel-wired-lan <redacted> On Behalf Of
Magnus Karlsson
Sent: Friday, March 19, 2021 3:14 PM
To: Karlsson, Magnus <magnus.karlsson@intel.com>; intel-wired-
lan@lists.osuosl.org; Nguyen, Anthony L [off-list ref];
Fijalkowski, Maciej [off-list ref]
Cc: netdev@vger.kernel.org; Joshi, Sreedevi <redacted>
Subject: [Intel-wired-lan] [PATCH intel-net] i40e: fix receiving of single packets
in xsk zero-copy mode

From: Magnus Karlsson <magnus.karlsson@intel.com>

Fix so that single packets are received immediately instead of in batches of 8.
If you sent 1 pss to a system, you received 8 packets every 8 seconds instead
of 1 packet every second. The problem behind this was that the work_done
reporting from the Tx part of the driver was broken. The work_done
reporting in i40e controls not only the reporting back to the napi logic but
also the setting of the interrupt throttling logic. When Tx or Rx reports that it
has more to do, interrupts are throttled or coalesced and when they both
report that they are done, interrupts are armed right away. If the wrong
work_done value is returned, the logic will start to throttle interrupts in a
situation where it should have just enabled them. This leads to the undesired
batching behavior seen in user-space.

Fix this by returning the correct boolean value from the Tx xsk zero-copy
path. Return true if there is nothing to do or if we got fewer packets to
process than we asked for. Return false if we got as many packets as the
budget since there might be more packets we can process.

Fixes: 3106c580fb7c ("i40e: Use batched xsk Tx interfaces to increase
performance")
Reported-by: Sreedevi Joshi <redacted>
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_xsk.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
Tested-by: Kiran Bhandare <redacted>  A Contingent Worker at Intel
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help