Re: [PATCH v3 0/3] net/txgbe: fix flow create errors

[Stephen Hemminger <stephen@networkplumber.org>]

On Thu, 17 Sep 2026 23:32:39 +0800
Zhang Tengfei <zhtfdev@gmail.com> wrote:

> v3:
> - reject mask-only FDIR before programming the global mask
> - log flex offset mismatch
> - no code change in 1/3 and 2/3
> 
> v2:
> - split L2 tunnel add-failure goto out into its own patch
> - allocate FDIR object before installing the global mask
> - add Fixes tag for the VF FDIR path
> - set -EINVAL on remaining FDIR goto out paths with bad ret
> - use struct assignment instead of rte_memcpy for filter copies
> 
> 1/3 fix L2 tunnel error on flow create
> 2/3 fix leak of filters on flow create
> 3/3 fix FDIR error handling on flow create
> 
> Zhang Tengfei (3):
>   net/txgbe: fix L2 tunnel error on flow create
>   net/txgbe: fix leak of filters on flow create
>   net/txgbe: fix FDIR error handling on flow create
> 
>  drivers/net/txgbe/txgbe_flow.c | 266 +++++++++++++++++----------------
>  1 file changed, 140 insertions(+), 126 deletions(-)
> 

I applied this to next-net.

AI review had some Info level observations, if you want to send
a new version, I will replace the one in next-net.

Review: [PATCH v3 0/3] net/txgbe: flow create fixes
Author: Zhang Tengfei <zhtfdev@gmail.com>

Applies to main, each of the three commits builds with -Dwerror=true.
Fixes tags resolve; both referenced commits are in released versions,
so Cc: stable is right.

Patches 1 and 2 are unchanged from v2 apart from commit message
wording. The v2 warning is addressed: mask-only FDIR rules are now
rejected before the global input mask is programmed, so a failed
create no longer leaves mask_added set. With that check in place the
b_spec block is unconditional, which reads better than the old
nesting. The memcmp and flex offset paths now set -EINVAL, and the
flex mismatch has a log message.


Patch 3/3: net/txgbe: fix FDIR error handling on flow create

Info: the b_spec check sits after the rte_zmalloc(), so it allocates
  and immediately frees on the reject path. Moving it above the
  allocation drops the rte_free() and one level of churn:

	if (!fdir_rule.b_spec) {
		ret = -EINVAL;
		goto out;
	}

	fdir_rule_ptr = rte_zmalloc(...);

Info: the VF path still reaches "out" with a base driver status code.
  txgbevf_fdir_filter_program() propagates TXGBE_ERR_FEATURE_NOT_
  SUPPORTED (-292) and TXGBE_ERR_NOSUPP from txgbevf_set_fdir(), so
  rte_flow_error_set(error, -ret) gets 292, not an errno. The PF
  paths checked out: txgbe_fdir_set_input_mask() and
  txgbe_fdir_filter_program() return -ENOTSUP, -EINVAL, -ENOMEM or
  -ETIMEDOUT. Converting the VF codes is a separate patch.

Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>