linux-kernel - netpoll: raspberrypi [4 5] driver locking woes

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <4698029bbe7d7a33edd905a027e3a183ac51ad8a.camel@gmx.de>
Date: Mon, 25 Aug 2025 07:53:42 +0200
From: Mike Galbraith <efault@....de>
To: lkml <linux-kernel@...r.kernel.org>
Cc: Robert Hancock <robert.hancock@...ian.com>, Doug Berger
 <opendmb@...il.com>,  Florian Fainelli <florian.fainelli@...adcom.com>,
 Breno Leitao <leitao@...ian.org>
Subject: netpoll:  raspberrypi [4 5]  driver locking woes

Greetings,

After meeting some netconsole+ieee80211 woes recently, I decided to
build a lockdep enabled 6.17-rc2 kernel for my rpi[45] and give it a go
on their wired nics just to make sure there were no surprises lurking..
but damn the bad luck, both cute/dinky boxen whimpered.

The pi5 met locking woes due to...

   138badbc21a0 net: macb: use NAPI for TX completion path

...injecting spin_lock/unlock_irq() into polling, which inspired
netpoll.c to point a finger directly at the driver.  A quick poke
didn't produce instant gratification, but a wedged in revert did.

On the pi4, lockdep muttered about ring->lock in bcmgenet_xmit() and
bcmgenet_tx_poll() - switching to irqsave/restore spinlock variants
there silenced it.

Note to readers: before turning on lockdep, perhaps consider whacking
lockdep_assert_held() instances in gpiolib.c.. ymmv, but here it would
have saved time and naughty language expenditure.

	-Mike