e1000_reset_hw_vf() sends E1000_VF_RESET through the VF
mailbox and then waits for the PF response. The posted write can
fail before the reset message is written to VMBMEM or before REQ is
set, for example when the VF cannot obtain the mailbox lock.
Return the write error directly so the reset path does not treat a
failed current request as a later read failure or as unrelated
mailbox state.
Fixes: af75078fece3 ("first public release")
Cc: [email protected]
Signed-off-by: Artem Novikov <[email protected]>
---
drivers/net/intel/e1000/base/e1000_vf.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/net/intel/e1000/base/e1000_vf.c
b/drivers/net/intel/e1000/base/e1000_vf.c
index 488ac3a94b..86ec1ea058 100644
--- a/drivers/net/intel/e1000/base/e1000_vf.c
+++ b/drivers/net/intel/e1000/base/e1000_vf.c
@@ -253,7 +253,9 @@ STATIC s32 e1000_reset_hw_vf(struct e1000_hw *hw)
mbx->timeout = E1000_VF_MBX_INIT_TIMEOUT;
msgbuf[0] = E1000_VF_RESET;
- mbx->ops.write_posted(hw, msgbuf, 1, 0);
+ ret_val = mbx->ops.write_posted(hw, msgbuf, 1, 0);
+ if (ret_val)
+ return ret_val;
msec_delay(10);
--
2.43.0
