Re: [PATCH] mm/slub: fix to return errno if kmalloc() fails

From: Matthew Wilcox <willy@infradead.org>
To: "Vlastimil Babka (SUSE)" <vbabka@kernel.org>
Cc: Muchun Song <muchun.song@linux.dev>, Chao Yu <chao@kernel.org>,
	Linux MM <linux-mm@kvack.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	linux-kernel@vger.kernel.org, jaegeuk@kernel.org,
	Chao Yu <chao.yu@oppo.com>,
	stable@kernel.org,
	syzbot+81684812ea68216e08c5@syzkaller.appspotmail.com,
	David Rientjes <rientjes@google.com>,
	Hyeonggon Yoo <42.hyeyoo@gmail.com>,
	Christoph Lameter <cl@linux.com>
Subject: Re: [PATCH] mm/slub: fix to return errno if kmalloc() fails
Date: Fri, 9 Sep 2022 21:06:11 +0100	[thread overview]
Message-ID: <Yxucs30FNkbWQHVh@casper.infradead.org> (raw)
In-Reply-To: <dbef29b2-ab0a-c3df-638c-381916a0d15a@kernel.org>

On Thu, Sep 08, 2022 at 11:25:08PM +0200, Vlastimil Babka (SUSE) wrote:
> > I tend to agree with you. A mount operation shouldn’t panic the
> > kernel.
> 
> Hmm kmalloc(64) shouldn't normally due that due to the the underlying page
> allocation falling into the "too small to fail" category, wonder if
> syzkaller was doing anything special here?

Here's the repro:

https://syzkaller.appspot.com/x/repro.c?x=17cd7fa3080000

you can see it does:

  fd = open("/proc/thread-self/fail-nth", O_RDWR);
  if (fd == -1)
    exit(1);
  char buf[16];
  sprintf(buf, "%d", nth);
  if (write(fd, buf, strlen(buf)) != (ssize_t)strlen(buf))

so this is the kind of stupid nitpicky bug that we shouldn't be
reporting, let alone fixing, IMO.