[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190816190234.2aaab5b6@cakuba.netronome.com>
Date: Fri, 16 Aug 2019 19:02:34 -0700
From: Jakub Kicinski <jakub.kicinski@...ronome.com>
To: syzbot <syzbot+6a9ff159672dfbb41c95@...kaller.appspotmail.com>
Cc: ast@...nel.org, aviadye@...lanox.com, borisp@...lanox.com,
bpf@...r.kernel.org, daniel@...earbox.net, davejwatson@...com,
davem@...emloft.net, hdanton@...a.com, john.fastabend@...il.com,
netdev@...r.kernel.org, syzkaller-bugs@...glegroups.com,
herbert@...dor.apana.org.au, linux-crypto@...r.kernel.org
Subject: Re: INFO: task hung in tls_sw_release_resources_tx
On Thu, 15 Aug 2019 11:06:00 -0700, syzbot wrote:
> syzbot has bisected this bug to:
>
> commit 130b392c6cd6b2aed1b7eb32253d4920babb4891
> Author: Dave Watson <davejwatson@...com>
> Date: Wed Jan 30 21:58:31 2019 +0000
>
> net: tls: Add tls 1.3 support
>
> bisection log: https://syzkaller.appspot.com/x/bisect.txt?x=118e8dee600000
> start commit: 6d5afe20 sctp: fix memleak in sctp_send_reset_streams
> git tree: net
> final crash: https://syzkaller.appspot.com/x/report.txt?x=138e8dee600000
> console output: https://syzkaller.appspot.com/x/log.txt?x=158e8dee600000
> kernel config: https://syzkaller.appspot.com/x/.config?x=a4c9e9f08e9e8960
> dashboard link: https://syzkaller.appspot.com/bug?extid=6a9ff159672dfbb41c95
> syz repro: https://syzkaller.appspot.com/x/repro.syz?x=17cb0502600000
> C reproducer: https://syzkaller.appspot.com/x/repro.c?x=14d5dc22600000
>
> Reported-by: syzbot+6a9ff159672dfbb41c95@...kaller.appspotmail.com
> Fixes: 130b392c6cd6 ("net: tls: Add tls 1.3 support")
>
> For information about bisection process see: https://goo.gl/tpsmEJ#bisection
CC Herbert, linux-crypto
This is got to be something in the crypto code :S
The test case opens a ktls socket and back log writes to it.
Then it opens a AF_ALG socket, binds "pcrypt(gcm(aes))" and dies.
The ktls socket upon close waits for async crypto callbacks, but they
never come. If I unset CRYPTO_USER_API_AEAD or change the alg to bind
to "gcm(aes)" the bug does not trigger.
Any suggestions?
Powered by blists - more mailing lists