Currently, vmxnet3 uses GRO callback only if LRO is disabled. However,
on smartNic based setups where UPT is supported, LRO can be enabled
from guest VM but UPT devicve does not support LRO as of now. In such
cases, there can be performance degradation as GRO is not being done.
This patch fixes this issue by calling GRO API when UPT is enabled. We
use updateRxProd to determine if UPT mode is active or not.
To clarify few things discussed over the thread:
The patch is not neglecting any feature bits nor disabling GRO. It uses
GRO callback when UPT is active as LRO is not available in UPT.
GRO callback cannot be used as default for all cases as it degrades
performance for non-UPT cases or for cases when LRO is already done in
ESXi.
Cc: stable(a)vger.kernel.org
Fixes: 6f91f4ba046e ("vmxnet3: add support for capability registers")
Signed-off-by: Ronak Doshi <doshir(a)vmware.com>
Reviewed-by: Simon Horman <simon.horman(a)corigine.com>
--
v1->v2: split if check on multiple lines
---
drivers/net/vmxnet3/vmxnet3_drv.c | 4 +++-
1 files changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c b/drivers/net/vmxnet3/vmxnet3_drv.c
index 682987040ea8..da488cbb0542 100644
--- a/drivers/net/vmxnet3/vmxnet3_drv.c
+++ b/drivers/net/vmxnet3/vmxnet3_drv.c
@@ -1688,7 +1688,9 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
if (unlikely(rcd->ts))
__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), rcd->tci);
- if (adapter->netdev->features & NETIF_F_LRO)
+ /* Use GRO callback if UPT is enabled */
+ if ((adapter->netdev->features & NETIF_F_LRO) &&
+ !rq->shared->updateRxProd)
netif_receive_skb(skb);
else
napi_gro_receive(&rq->napi, skb);
--
2.11.0
Hi There,
Greeting. Hope this email find you are well.
We are one of the biggest ones in south of China manufacturer specialized in Adapter products since 2001.
Now I'm writing for keeping in touch with you for further business.
As our factory is manufacturer, we can Customizes, OEM & ODM also acceptable.
If you need the catalogue, please feel free to contact me.
Andy
Hi,
This commit that landed in 6.3 was marked to got to stable, but it failed
to apply to both 6.2.y and 6.1.y.
cbd6c1b17d3b ("drm/amd/display: Fix DP MST sinks removal issue")
The code is needed for some timing issues with USB4 CM teardown.
It failed to apply because the code it fixed was moved to a new file in
6.3. The code is otherwise unchanged in the old file, so this has been
manually fixed up.
Please apply for both 6.2.y and 6.1.y.
Thanks!
Cruise Hung (1):
drm/amd/display: Fix DP MST sinks removal issue
drivers/gpu/drm/amd/display/dc/core/dc_link.c | 9 +++++++++
1 file changed, 9 insertions(+)
--
2.34.1
Hi Greg, Sasha,
Recently, 3 patches related to MPTCP have not been backported due to
conflicts:
- 3a236aef280e ("mptcp: refactor passive socket initialization")
- b6985b9b8295 ("mptcp: use the workqueue to destroy unaccepted sockets")
- 0a3f4f1f9c27 ("mptcp: fix UaF in listener shutdown")
Yesterday, Sasha has resolved the conflicts for the first one and he has
already added this one to v6.1.
In fact, this first patch is a requirement for the two others.
I then here resolved the conflicts for the two other patches, documented
that in each patch and ran our tests suite. Everything seems OK.
Do you mind adding these two patches to v6.1 queue as well if you don't
mind?
Signed-off-by: Matthieu Baerts <matthieu.baerts(a)tessares.net>
---
Paolo Abeni (2):
mptcp: use the workqueue to destroy unaccepted sockets
mptcp: fix UaF in listener shutdown
net/mptcp/protocol.c | 46 ++++++++++++++++++---------
net/mptcp/protocol.h | 6 ++--
net/mptcp/subflow.c | 89 +++++++---------------------------------------------
3 files changed, 47 insertions(+), 94 deletions(-)
---
base-commit: 0866b93e23cb1d66eb4b105d305cdb185ca17b7d
change-id: 20230323-upstream-stable-conflicts-6-1-e0baa081983d
Best regards,
--
Matthieu Baerts <matthieu.baerts(a)tessares.net>
The commit 49c47cc21b5b (net: tls: fix possible race condition between
do_tls_getsockopt_conf() and do_tls_setsockopt_conf()) fixes race
condition and use after free. This patch didn't apply cleanly in 5.15
kernel due to the added switch cases in do_tls_getsockopt_conf function.
Hangyu Hua (1):
net: tls: fix possible race condition between do_tls_getsockopt_conf()
and do_tls_setsockopt_conf()
net/tls/tls_main.c | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)
--
2.40.0.348.gf938b09366-goog