[PATCH net v2] vmxnet3: segCnt can be 1 for LRO packets

[Shrikrishna Khare]

The device emulation may send segCnt of 1 for LRO packets.

Signed-off-by: Shrikrishna Khare 
Signed-off-by: Jin Heo 

---
v2: fix subject line
---
 drivers/net/vmxnet3/vmxnet3_drv.c | 2 +-
 drivers/net/vmxnet3/vmxnet3_int.h | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c 
b/drivers/net/vmxnet3/vmxnet3_drv.c
index db8022a..6f399b2 100644
--- a/drivers/net/vmxnet3/vmxnet3_drv.c
+++ b/drivers/net/vmxnet3/vmxnet3_drv.c
@@ -1369,7 +1369,7 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
rcdlro = (struct Vmxnet3_RxCompDescExt *)rcd;
 
segCnt = rcdlro->segCnt;
-   BUG_ON(segCnt <= 1);
+   BUG_ON(segCnt == 0);
mss = rcdlro->mss;
if (unlikely(segCnt <= 1))
segCnt = 0;
diff --git a/drivers/net/vmxnet3/vmxnet3_int.h 
b/drivers/net/vmxnet3/vmxnet3_int.h
index c482539..3d2b64e 100644
--- a/drivers/net/vmxnet3/vmxnet3_int.h
+++ b/drivers/net/vmxnet3/vmxnet3_int.h
@@ -69,10 +69,10 @@
 /*
  * Version numbers
  */
-#define VMXNET3_DRIVER_VERSION_STRING   "1.4.7.0-k"
+#define VMXNET3_DRIVER_VERSION_STRING   "1.4.8.0-k"
 
 /* a 32-bit int, each byte encode a verion number in VMXNET3_DRIVER_VERSION */
-#define VMXNET3_DRIVER_VERSION_NUM  0x01040700
+#define VMXNET3_DRIVER_VERSION_NUM  0x01040800
 
 #if defined(CONFIG_PCI_MSI)
/* RSS only makes sense if MSI-X is supported. */
-- 
2.8.2

Re: [PATCH net v2] vmxnet3: segCnt can be 1 for LRO packets

[David Miller]

From: Shrikrishna Khare 
Date: Tue,  7 Jun 2016 23:13:35 -0700

> The device emulation may send segCnt of 1 for LRO packets.
> 
> Signed-off-by: Shrikrishna Khare 
> Signed-off-by: Jin Heo 

Applied.

Re: [Pv-drivers] [PATCH net v2] vmxnet3: segCnt can be 1 for LRO packets

[Thomas Hellstrom]

On 06/08/2016 08:13 AM, Shrikrishna Khare wrote:
> The device emulation may send segCnt of 1 for LRO packets.
>
> Signed-off-by: Shrikrishna Khare 
> Signed-off-by: Jin Heo 
>
> ---
> v2: fix subject line
> ---
>  drivers/net/vmxnet3/vmxnet3_drv.c | 2 +-
>  drivers/net/vmxnet3/vmxnet3_int.h | 4 ++--
>  2 files changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c 
> b/drivers/net/vmxnet3/vmxnet3_drv.c
> index db8022a..6f399b2 100644
> --- a/drivers/net/vmxnet3/vmxnet3_drv.c
> +++ b/drivers/net/vmxnet3/vmxnet3_drv.c
> @@ -1369,7 +1369,7 @@ vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
>   rcdlro = (struct Vmxnet3_RxCompDescExt *)rcd;
>  
>   segCnt = rcdlro->segCnt;
> - BUG_ON(segCnt <= 1);
> + BUG_ON(segCnt == 0);

What will the vmxnet3 driver do otherwise if segCnt == 0? Please see below.


>   mss = rcdlro->mss;
>   if (unlikely(segCnt <= 1))
>   segCnt = 0;

Based on this code, it looks like it can handle the case without taking
down the kernel completely, so instead of a
"BUG_ON", would it make sense with a WARN_ON_ONCE() or WARN_ON()?

Thanks,
Thomas

Re: [Pv-drivers] [PATCH net v2] vmxnet3: segCnt can be 1 for LRO packets

[Shrikrishna Khare]

On Wed, 8 Jun 2016, Thomas Hellstrom wrote:

> > mss = rcdlro->mss;
> > if (unlikely(segCnt <= 1))
> > segCnt = 0;
> 
> Based on this code, it looks like it can handle the case without taking
> down the kernel completely, so instead of a
> "BUG_ON", would it make sense with a WARN_ON_ONCE() or WARN_ON()?

Thanks for the review Thomas. I fixed the code to use WARN_ON_ONCE. Sent 
out revised patch.