Re: scsi: lpfc: Fix hardlockup in lpfc_abort_handler

Re: scsi: lpfc: Fix hardlockup in lpfc_abort_handler

[James Smart]

On 10/10/2019 1:59 AM, Zhangguanghui wrote:
> Hi everyone
> 
> Please refer to the latest patch.
> 
> There is a race deadlock in the function lpfc_abort_handler
> 
> potential deadlocks arising from lock ordering problems.
> 
> It’s the correct order
> 
> spin_unlock(&lpfc_cmd->buf_lock)
> 
> spin_unlock_irqrestore(&phba->hbalock, flags);
> 
> How to solve it ? I think that the patch is reasonable,
> 
> can you help me review and commit this patch, Best regards
> 
> diff --git a/src/lpfc-12.2.0.0/lpfc_scsi.c b/src/lpfc-12.2.0.0/lpfc_scsi.c
> 
> index 3f1375a..19c8505 100644
> 
> --- a/src/lpfc-12.2.0.0/lpfc_scsi.c
> 
> +++ b/src/lpfc-12.2.0.0/lpfc_scsi.c
> 

We confirmed the issue you stated.  We are looking at what you proposed 
and will be adding a patch that will be posted in our next
patch set after we've put it through some regression testing.

-- james