[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20260108020518.27086-1-ryazanov.s.a@gmail.com>
Date: Thu, 8 Jan 2026 04:05:17 +0200
From: Sergey Ryazanov <ryazanov.s.a@...il.com>
To: Loic Poulain <loic.poulain@....qualcomm.com>,
Johannes Berg <johannes@...solutions.net>
Cc: Andrew Lunn <andrew+netdev@...n.ch>,
Eric Dumazet <edumazet@...gle.com>,
"David S . Miller" <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>,
netdev@...r.kernel.org,
Slark Xiao <slark_xiao@....com>,
Daniele Palmas <dnlplm@...il.com>
Subject: [RFC PATCH 0/1] prevent premature device unregister via
Initially I was unable to hit or reproduce the issue with hwsim since it
unregister the WWAN device ops as a last step effectively holding the
WWAN device when all the regular WWAN ports are already removed. Thanks
to the detiled report of Daniele and the fix proposed by Loic, it became
obvious what a releasing sequence leads to the crash.
With WWAN device ops unregistration done first in hwsim, I was able to
easily reproduce the WWAN device premature unregister, and develop
another fix avoiding a dummy port allocation and relying on a reference
counting. See details in the RFC patch.
Loic, what do you think about this way of the users tracking?
Slark, if you would like to go with the proposed patch, just remove the
patch #7 from the series and insert the proposed patch between between
#1 and #2. Of if you prefer, I can reassemble the whole series and send
it as RFC v5.
CC: Slark Xiao <slark_xiao@....com>
CC: Daniele Palmas <dnlplm@...il.com>
--
2.52.0
Powered by blists - more mailing lists