vmxnet3: couple of fixes
Subsystems (3)
cover vmxnet3: couple of fixes
This series fixes following issues:
Patch 1:
This patch provides a fix to correctly report encapsulated LRO'ed packet.
Patch 2:
This patch provides a fix to use correct intrConf reference.
Changes in v2:
- declare generic descriptor to be used
Ronak Doshi (2):
vmxnet3: correctly report encapsulated LRO packet
vmxnet3: use correct intrConf reference when using extended queues
drivers/net/vmxnet3/vmxnet3_drv.c | 28 ++++++++++++++++++++++++---- 1 file changed, 24 insertions(+), 4 deletions(-)
Signature · 1 line
2.11.0
PATCH 1/2 vmxnet3: correctly report encapsulated LRO packet
Commit dacce2be3312 ("vmxnet3: add geneve and vxlan tunnel offload support") added support for encapsulation offload. However, the pathc did not report correctly the encapsulated packet which is LRO'ed by the hypervisor.
This patch fixes this issue by using correct callback for the LRO'ed encapsulated packet.
--- drivers/net/vmxnet3/vmxnet3_drv.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-)
drivers/net/vmxnet3/vmxnet3_drv.c +10 -2
@@ -1396,6 +1396,7 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq, };u32 num_pkts =0 ;bool skip_page_frags = false; +bool encap_lro = false;struct Vmxnet3_RxCompDesc *rcd;struct vmxnet3_rx_ctx *ctx = &rq->rx_ctx;u16 segCnt =0 , mss =0 ;
@@ -1556,13 +1557,18 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,if (VMXNET3_VERSION_GE_2 (adapter) && rcd->type ==VMXNET3_CDTYPE_RXCOMP_LRO ) {struct Vmxnet3_RxCompDescExt *rcdlro; +union Vmxnet3_GenericDesc *gdesc; + rcdlro = (struct Vmxnet3_RxCompDescExt *)rcd; + gdesc = (union Vmxnet3_GenericDesc *)rcd; segCnt = rcdlro->segCnt;WARN_ON_ONCE (segCnt ==0 ); mss = rcdlro->mss;if (unlikely (segCnt <=1 )) segCnt =0 ; + encap_lro = (le32_to_cpu (gdesc->dword[0 ]) & + (1UL <<VMXNET3_RCD_HDR_INNER_SHIFT )); }else { segCnt =0 ; }
@@ -1630,7 +1636,7 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,vmxnet3_rx_csum (adapter, skb, (union Vmxnet3_GenericDesc *)rcd); skb->protocol =eth_type_trans (skb, adapter->netdev); -if (!rcd->tcp || +if ((!rcd->tcp && !encap_lro) || !(adapter->netdev->features &NETIF_F_LRO )) goto not_lro;
@@ -1639,7 +1645,7 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq, SKB_GSO_TCPV4 :SKB_GSO_TCPV6 ;skb_shinfo (skb)->gso_size = mss;skb_shinfo (skb)->gso_segs = segCnt; - }else if (segCnt !=0 || skb->len > mtu) { + }else if ((segCnt !=0 || skb->len > mtu) && !encap_lro) {u32 hlen; hlen = vmxnet3_get_hdr_len(adapter, skb,
@@ -1668,6 +1674,8 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,napi_gro_receive (&rq->napi, skb); ctx->skb =NULL ; +if (encap_lro) + encap_lro = false; num_pkts++; }
Signature · 1 line
2.11.0
PATCH 2/2 vmxnet3: use correct intrConf reference when using extended queues
In reply to Ronak Doshi'Commit 39f9895a00f4 ("vmxnet3: add support for 32 Tx/Rx queues")' added support for 32Tx/Rx queues. As a part of this patch, intrConf structure was extended to incorporate increased queues.
This patch fixes the issue where incorrect reference is being used.
--- drivers/net/vmxnet3/vmxnet3_drv.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-)
drivers/net/vmxnet3/vmxnet3_drv.c +14 -2
@@ -75,8 +75,14 @@ vmxnet3_enable_all_intrs(struct vmxnet3_adapter *adapter)for (i =0 ; i < adapter->intr.num_intrs; i++)vmxnet3_enable_intr (adapter, i); - adapter->shared->devRead.intrConf.intrCtrl &= +if (!VMXNET3_VERSION_GE_6 (adapter) || + !adapter->queuesExtEnabled) { + adapter->shared->devRead.intrConf.intrCtrl &=cpu_to_le32 (~VMXNET3_IC_DISABLE_ALL ); + }else { + adapter->shared->devReadExt.intrConfExt.intrCtrl &= +cpu_to_le32 (~VMXNET3_IC_DISABLE_ALL ); + } }
@@ -85,8 +91,14 @@ vmxnet3_disable_all_intrs(struct vmxnet3_adapter *adapter) {int i; - adapter->shared->devRead.intrConf.intrCtrl |= +if (!VMXNET3_VERSION_GE_6 (adapter) || + !adapter->queuesExtEnabled) { + adapter->shared->devRead.intrConf.intrCtrl |=cpu_to_le32 (VMXNET3_IC_DISABLE_ALL ); + }else { + adapter->shared->devReadExt.intrConfExt.intrCtrl |= +cpu_to_le32 (VMXNET3_IC_DISABLE_ALL ); + }for (i =0 ; i < adapter->intr.num_intrs; i++)vmxnet3_disable_intr (adapter, i); }
Signature · 1 line
2.11.0
Re: [PATCH v1 net 2/2] vmxnet3: use correct intrConf reference when using extended queues
Quoted 4 lines · Ronak Doshi
'Commit
39f9895a00f4("vmxnet3: add support for 32 Tx/Rx queues")' added support for 32Tx/Rx queues. As a part of this patch, intrConf structure was extended to incorporate increased queues.
Nit: no need to quote around the commit reference here.
I don't personally think its worth a re-roll to fix that, but good to be aware of in future submission.
The patch itself makes sense.
Quoted 45 lines
This patch fixes the issue where incorrect reference is being used.
Fixes 39f9895a00f4 ("vmxnet3: add support for 32 Tx/Rx queues")Signed-off-by Ronak Doshi <[email protected]>Acked-by Guolin Yang <[email protected]>--- drivers/net/vmxnet3/vmxnet3_drv.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-)drivers/net/vmxnet3/vmxnet3_drv.c +14
-2@@ -75,8 +75,14 @@ vmxnet3_enable_all_intrs(struct vmxnet3_adapter *adapter)for (i =0 ; i < adapter->intr.num_intrs; i++)vmxnet3_enable_intr (adapter, i); - adapter->shared->devRead.intrConf.intrCtrl &= +if (!VMXNET3_VERSION_GE_6 (adapter) || + !adapter->queuesExtEnabled) { + adapter->shared->devRead.intrConf.intrCtrl &=cpu_to_le32 (~VMXNET3_IC_DISABLE_ALL ); + }else { + adapter->shared->devReadExt.intrConfExt.intrCtrl &= +cpu_to_le32 (~VMXNET3_IC_DISABLE_ALL ); + } }@@ -85,8 +91,14 @@ vmxnet3_disable_all_intrs(struct vmxnet3_adapter *adapter) {int i; - adapter->shared->devRead.intrConf.intrCtrl |= +if (!VMXNET3_VERSION_GE_6 (adapter) || + !adapter->queuesExtEnabled) { + adapter->shared->devRead.intrConf.intrCtrl |=cpu_to_le32 (VMXNET3_IC_DISABLE_ALL ); + }else { + adapter->shared->devReadExt.intrConfExt.intrCtrl |= +cpu_to_le32 (VMXNET3_IC_DISABLE_ALL ); + }for (i =0 ; i < adapter->intr.num_intrs; i++)vmxnet3_disable_intr (adapter, i); }
Re: [PATCH v1 net 2/2] vmxnet3: use correct intrConf reference when using extended queues
Quoted 8 lines · Jacob Keller
!! External Email
Quoted 4 lines · Ronak Doshi
'Commit
39f9895a00f4("vmxnet3: add support for 32 Tx/Rx queues")' added support for 32Tx/Rx queues. As a part of this patch, intrConf structure was extended to incorporate increased queues.
Quoted 8 lines
Nit: no need to quote around the commit reference here.
I don't personally think its worth a re-roll to fix that, but good to be aware of in future submission.
The patch itself makes sense.
Reviewed-by Jacob Keller <[email protected]>
Thanks, Jacob, for the review.
Jakub Kicinski
maintainerRe: [PATCH v1 net 1/2] vmxnet3: correctly report encapsulated LRO packet
In reply to Ronak Doshidrivers/net/vmxnet3/vmxnet3_drv.c vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq, · line 1561
Quoted 2 lines · On Mon, 28 Nov 2022 11:32:03 -0800 Ronak Doshi wrote:
drivers/net/vmxnet3/vmxnet3_drv.c +2
-0+union Vmxnet3_GenericDesc *gdesc; +
Trailing white space here.
Please use checkpatch.