[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <abffa0b0-300d-8ecd-8319-c99621198c86@linux.vnet.ibm.com>
Date: Wed, 20 Apr 2022 10:23:28 -0700
From: David Christensen <drc@...ux.vnet.ibm.com>
To: Manish Chopra <manishc@...vell.com>,
Netdev <netdev@...r.kernel.org>,
Ariel Elior <aelior@...vell.com>,
Sudarsana Reddy Kalluru <skalluru@...vell.com>
Subject: Re: [EXT] EEH Causes bnx2x to Hang in napi_disable
>> ----------------------------------------------------------------------
>> Experiencing an inaccessible system when bnx2x attempts to recover from an
>> injected EEH error. This is a POWER10 system running SLES 15 SP3
>> (5.3.18-150300.59.49-default) with a BCM57810 dual port NIC.
>>
> Thanks for reporting this issue, I tried to reproduce the same issue where in my case it leaded to system crash but it was not exactly the same
> symptom of hang/stuck in napi_disable(). But I believe they all have the same root cause for different issues as you stated about the incorrect
> sequence of NAPI API usage in this particular driver flow (unlike in regular unload flow (ndo_stop()) where it deletes NAPI after it disables the NAPI).
> So driver actually needs to disable NAPI first and delete the NAPI afterwards, I tried below change which fixed the crash which I was always seeing on
> my setup.
>
> Can you please try this below change to see if that helps for the issue you reported or not ?
Glad to report that it fixed our problem as well. Thanks for the help.
Dave
Powered by blists - more mailing lists