Received-SPF: softfail (google.com: domain of transitioning linux-wireless-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
Message-ID: <ae6ebb34ba100fa8e17cc7eb187b7cfdf7a20a56.camel@sipsolutions.net>
Subject: Re: [BUG] deadlock in nl80211_vendor_cmd
From:   Johannes Berg <johannes@sipsolutions.net>
To:     William McVicker <willmcvicker@google.com>
Cc:     Jakub Kicinski <kuba@kernel.org>, linux-wireless@vger.kernel.org,
        Marek Szyprowski <m.szyprowski@samsung.com>,
        Kalle Valo <kvalo@codeaurora.org>,
        "David S. Miller" <davem@davemloft.net>, netdev@vger.kernel.org,
        Amitkumar Karwar <amitkarwar@gmail.com>,
        Ganapathi Bhat <ganapathi.bhat@nxp.com>,
        Xinming Hu <huxinming820@gmail.com>, kernel-team@android.com,
        Paolo Abeni <pabeni@redhat.com>
Date:   Fri, 25 Mar 2022 21:21:11 +0100
In-Reply-To: <Yj4FFIXi//ivQC3X@google.com>
References: <0000000000009e9b7105da6d1779@google.com>
         <99eda6d1dad3ff49435b74e539488091642b10a8.camel@sipsolutions.net>
         <5d5cf050-7de0-7bad-2407-276970222635@quicinc.com>
         <YjpGlRvcg72zNo8s@google.com>
         <dc556455-51a2-06e8-8ec5-b807c2901b7e@quicinc.com>
         <Yjzpo3TfZxtKPMAG@google.com>
         <19e12e6b5f04ba9e5b192001fbe31a3fc47d380a.camel@sipsolutions.net>
         <20220325094952.10c46350@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
         <f4f8a27dc07c1adaab470fde302ed841113e6b7f.camel@sipsolutions.net>
         <Yj4FFIXi//ivQC3X@google.com>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.42.4 (3.42.4-1.fc35) 
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

On Fri, 2022-03-25 at 18:08 +0000, William McVicker wrote:
> 
> I'm able to reproduce this issue pretty easily with a Pixel 6 when I add
> support to allow vendor commands to request for the RTNL. 
> 

Hm, wait, which of the two issues?

> For this case, I just
> delay unlocking the RTNL until nl80211_vendor_cmds() at which point I check the
> flags to see if I should unlock before calling doit(). That allows me to run my
> tests again and hit this issue. I imagine that I could hit this issue without
> any changes if I re-work my vendor ops to not need the RTNL.

What are the vendor ops doing though?

If they're actually unregistering a netdev - which I believe you
mentioned earlier - then that's quite clearly going to cause an issue,
if you unlock RTNL while the wiphy mutex is still held.

If not, then I don't see right now how you'd be able to trigger any
issue here at all.

The original issue - that you rtnl_lock() yourself while the wiphy mutex
is held - can't happen anymore with your rework I guess.


johannes