[hns3-0114]net: hns3: fix a wrong reset interrupt status mask

[Fred Kimmy]

[Bug Description]
There is timing window between ring_space checking and
netif_stop_subqueue when transmiting a SKB, and the TX BD
cleaning may be executed during the time window, which may
caused TX queue not restarted problem.

[Steps to Reproduce]
1.run IO with high throughput

[Actual Results]
cause tx timeout

[Expected Results]
IO ok

[Reproducibility]
Inevitably

[Additional information]
Hardware: D06
Firmware: NA
Kernel: NA

[Resolution]
This patch fixes it by rechecking the ring_space after
netif_stop_subqueue to make sure TX queue is restarted.

Also, the ring->next_to_clean is updated even when pkts is
zero, because all the TX BD cleaned may be non-SKB, so it
needs to check if TX queue need to be restarted.

[Ike Panhc (ikepanhc)]

Is this patch the fix? It can not match the bug description.

commit 74e78d6bae1904e87469da5ed87e9f6bd1131f46
Author: Huazhong Tan <email address hidden>
Date: Wed Nov 20 10:07:15 2019 +0800

    net: hns3: fix a wrong reset interrupt status mask

    According to hardware user manual, bits5~7 in register
    HCLGE_MISC_VECTOR_INT_STS means reset interrupts status,
    but HCLGE_RESET_INT_M is defined as bits0~2 now. So it
    will make hclge_reset_err_handle() read the wrong reset
    interrupt status.

    This patch fixes this wrong bit mask.

    Fixes: 2336f19d7892 ("net: hns3: check reset interrupt status when reset fails")
    Signed-off-by: Huazhong Tan <email address hidden>
    Signed-off-by: David S. Miller <email address hidden>

[Ike Panhc (ikepanhc)]

If patch 74e78d6bae ("net: hns3: fix a wrong reset interrupt status mask") is the fix, it is not necessary to backport to bionic GA kernel since it's fixes 2336f19d7892 ("net: hns3: check reset interrupt status when reset fails") is not applied to bionic GA kernel.

And patch 74e78d6bae is already in focal GA kernel.

[Fred Kimmy (kongzizaixian)]

this two pathset should merge it. this patch 74e78d6bae ("net: hns3: fix a wrong reset interrupt status mask") also fix other bug such as hns resetting status.

[Ike Panhc (ikepanhc)]

ikepanhc@ubuntu:~/ubuntu-focal$ gitoneline | grep -e 'fix a wrong reset interrupt status mask' -e 'check reset interrupt status when reset fails'
74e78d6bae19 <email address hidden> 2019-11-19 19:09:53 -0800 net: hns3: fix a wrong reset interrupt status mask
2336f19d7892 <email address hidden> 2019-08-29 16:57:44 -0700 net: hns3: check reset interrupt status when reset fails

Patches in comment #2 already in focal kernel, so it will be in bionic HWE kernel when 18.04.5 release

[Ike Panhc (ikepanhc)]

Patch 74e78d6bae19 ("net: hns3: fix a wrong reset interrupt status mask") fixes 2336f19d7892 ("net: hns3: check reset interrupt status when reset fails")

Patch 2336f19d7892 ("net: hns3: check reset interrupt status when reset fails") fixes 72e2fb07997c ("net: hns3: clear reset interrupt status in hclge_irq_handle()")

Patch 72e2fb07997c ("net: hns3: clear reset interrupt status in hclge_irq_handle()") fixes 4ed340ab8f49 ("net: hns3: Add reset process in hclge_main")

But patch 72e2fb07997c fails to clean cherry-pick.

ikepanhc@ubuntu:~/ubuntu-bionic$ git cherry-pick -x -s 72e2fb07997c
Auto-merging drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.h
Auto-merging drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
CONFLICT (content): Merge conflict in drivers/net/ethernet/hisilicon/hns3/hns3vf/hclgevf_main.c
Auto-merging drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
CONFLICT (content): Merge conflict in drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_main.c
Auto-merging drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.h
warning: inexact rename detection was skipped due to too many files.
warning: you may want to set your merge.renamelimit variable to at least 10159 and retry the command.
error: could not apply 72e2fb07997c... net: hns3: clear reset interrupt status in hclge_irq_handle()
hint: after resolving the conflicts, mark the corrected paths
hint: with 'git add <paths>' or 'git rm <paths>'
hint: and commit the result with 'git commit'

Since we can run iperf testing on hns3 and get > 80% of limitation for an hour and multiple times, I don't think this is a issue we need to fix for bionic GA kernel.