Re: [PATCH 2/2] mm/kasan: proc interface to read KASAN errors at any time

From: Alexander Potapenko <glider@google.com>
To: Dmitry Vyukov <dvyukov@google.com>
Cc: Maninder Singh <maninder1.s@samsung.com>,
	Marco Elver <elver@google.com>,
	Andrey Ryabinin <ryabinin.a.a@gmail.com>,
	Andrey Konovalov <andreyknvl@gmail.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	kasan-dev <kasan-dev@googlegroups.com>,
	Linux-MM <linux-mm@kvack.org>,
	LKML <linux-kernel@vger.kernel.org>,
	AMIT SAHRAWAT <a.sahrawat@samsung.com>,
	Vaneet Narang <v.narang@samsung.com>
Subject: Re: [PATCH 2/2] mm/kasan: proc interface to read KASAN errors at any time
Date: Thu, 22 Apr 2021 12:38:11 +0200	[thread overview]
Message-ID: <CAG_fn=Xdx-hEna_S4u4nRXYXW+HrJCcyPPiGqQoU8Ri1UYZu0g@mail.gmail.com> (raw)
In-Reply-To: <CACT4Y+ZJ95KiFNHeT9k0p6ezDz-apkJVp586UBSdJeHtCYR_Qg@mail.gmail.com>

> Alex, Marco, can the recently added error_report_notify interface be
> used for this? Looks like they are doing roughly the same thing with
> the same intentions.

We've recently attempted to build a universal library capturing every
error report, but then were pointed to tracefs, which was just enough
for our purpose (https://lkml.org/lkml/2021/1/15/609).
Greg also stated that procfs is a bad place for storing reports:
https://lkml.org/lkml/2021/1/15/929.

Maninder, which exactly problem are you trying to solve?
Note that KASAN already triggers a trace_error_report_end tracepoint
on every error report:
https://elixir.bootlin.com/linux/v5.12-rc8/source/mm/kasan/report.c#L90
Would it help if you used that one? It could probably be extended with
more parameters.

Another option if you want verbatim reports is to use the console
tracepoints, as this is done in
https://elixir.bootlin.com/linux/v5.12-rc8/source/mm/kfence/kfence_test.c
Note that there are many caveats with error report collection (see the
links above), but for testing purpose it might be enough.

Alex