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
References: <CGME20210422081531epcas5p23d6c72ebf28a23b2efc150d581319ffa@epcas5p2.samsung.com>
 <1619079317-1131-1-git-send-email-maninder1.s@samsung.com>
In-Reply-To: <1619079317-1131-1-git-send-email-maninder1.s@samsung.com>
From:   Marco Elver <elver@google.com>
Date:   Thu, 22 Apr 2021 16:10:37 +0200
Message-ID: <CANpmjNOT7xVbv4P1n3X24-HH8VMBs7Ny33DFYbzjO6Gqza2mZA@mail.gmail.com>
Subject: Re: [PATCH 1/2] mm/kasan: avoid duplicate KASAN issues from reporting
To:     Maninder Singh <maninder1.s@samsung.com>
Cc:     Andrey Ryabinin <ryabinin.a.a@gmail.com>,
        Alexander Potapenko <glider@google.com>,
        Andrey Konovalov <andreyknvl@gmail.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Dmitry Vyukov <dvyukov@google.com>,
        kasan-dev <kasan-dev@googlegroups.com>,
        Linux Memory Management List <linux-mm@kvack.org>,
        LKML <linux-kernel@vger.kernel.org>, a.sahrawat@samsung.com,
        Vaneet Narang <v.narang@samsung.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, 22 Apr 2021 at 11:17, Maninder Singh <maninder1.s@samsung.com> wrote:
>
> when KASAN multishot is ON and some buggy code hits same code path
> of KASAN issue repetetively, it can flood logs on console.
>
> Check for allocaton, free and backtrace path at time of KASAN error,
> if these are same then it is duplicate error and avoid these prints
> from KASAN.

On a more fundamental level, I think this sort of filtering is the
wrong solution to your problem. One reason why it's good that
multishot is off by default is, because _every_ KASAN report is
critical and can destabilize the system. Therefore, any report after
the first one might be completely bogus, because the system is in a
potentially bad state and its behaviour might be completely random.

The correct solution is to not leave the system running, fix the first
bug found, continue; rinse and repeat. Therefore, this patch adds a
lot of code for little benefit.

The much simpler solution that will likely yield a similar result is
to simply define an upper bound on the number of reports if multishot
is on. Because if I've seen 1000 reports, I already know the system is
completely trashed and whatever else it's reporting might just be
random.

Thanks,
-- Marco