Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
From:   Dmitry Vyukov <dvyukov@google.com>
Date:   Mon, 1 Feb 2021 11:52:12 +0100
Message-ID: <CACT4Y+YytChe9Tv5etacBL4snEDu+A2fNzF4zKiKtfmH0C0roQ@mail.gmail.com>
Subject: syzbot reporting less duplicates
To:     LKML <linux-kernel@vger.kernel.org>,
        syzkaller <syzkaller@googlegroups.com>
Cc:     Eric Dumazet <edumazet@google.com>,
        Hillf Danton <hdanton@sina.com>,
        Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi,

FYI syzbot should report less duplicates now.
A significant number of reported duplicates was one of common
complaints. syzbot will now merge some slightly differently looking
crashes together, for example, these 2 recently reported bugs will be
merged into a single bug and reported once in future:

KASAN: use-after-free Read in skb_segment
KASAN: slab-out-of-bounds Read in skb_segment

You can see an example here:
https://syzkaller.appspot.com/bug?id=9936b32dd3a4a278f06a2cb07eb13df9e113ca84
which contains crashes for both "BUG: unable to handle kernel paging
request in cdev_del" and "general protection fault in cdev_del".

Hope you will find this useful.

If you are interested in heuristics details, grep for "alt:" here:
https://github.com/google/syzkaller/blob/master/pkg/report/linux.go
Crashes with the same alt title are merged together. In short,
currently there is one cluster for "bad access in function foo" (GPF,
fault, KASAN, KMSAN, corrupted list, etc); another cluster for passive
hangs (task hung, task can't die); and another cluster for active
stalls (rcu stall, cpu hang). These rules are based on analysis of all
previously reported bugs and duplicates among them.

Thanks