[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CACKFLimy_DUPvMzApbgy7j701WCwkzTy_EARd1nmd+XKWC28kw@mail.gmail.com>
Date: Wed, 15 Nov 2023 10:56:57 -0800
From: Michael Chan <michael.chan@...adcom.com>
To: Thinh Tran <thinhtr@...ux.vnet.ibm.com>
Cc: netdev@...r.kernel.org, siva.kallam@...adcom.com, prashant@...adcom.com,
mchan@...adcom.com, pavan.chebbi@...adcom.com, drc@...ux.vnet.ibm.com,
venkata.sai.duggi@....com
Subject: Re: [PATCH v2] net/tg3: fix race condition in tg3_reset_task()
On Wed, Nov 15, 2023 at 10:23 AM Thinh Tran <thinhtr@...ux.vnet.ibm.com> wrote:
>
>
> On 11/14/2023 3:03 PM, Michael Chan wrote:
> >
> > Could you provide more information about the crashes? The
> > dev_watchdog() code already checks for netif_device_present() and
> > netif_running() and netif_carrier_ok() before proceeding to check for
> > TX timeout. Why would adding some additional checks for PCI errors
> > cause problems? Of course the additional checks should only be done
> > on PCI devices only. Thanks.
>
> The checking for PCI errors is not the problem, avoiding calling drivers
> ->ndo_tx_timeout() function, causing some issue.
I see. By skipping TX timeout during PCI errors, bnx2x crashes in
.ndo_start_xmit() after EEH error recovery.
I think it should be fine to fix the original EEH issue in tg3 then.
Please re-post the tg3 patch. Thanks.
Download attachment "smime.p7s" of type "application/pkcs7-signature" (4209 bytes)
Powered by blists - more mailing lists