Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Sat, 29 Jul 2023 01:50:27 +0100
From:   Matthew Wilcox <willy@infradead.org>
To:     Joan Bruguera =?iso-8859-1?Q?Mic=F3?= <joanbrugueram@gmail.com>
Cc:     Steven Price <steven.price@arm.com>, zhengqi.arch@bytedance.com,
        Roman Gushchin <roman.gushchin@linux.dev>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org,
        Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH] mm: shrinkers: fix race condition on debugfs cleanup
Message-ID: <ZMRiUw8TGdbDsj1t@casper.infradead.org>
References: <c7dbaa5e-f5ae-5dc5-4063-b8099a3058e0@arm.com>
 <20230726000827.724389-1-joanbrugueram@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20230726000827.724389-1-joanbrugueram@gmail.com>
Precedence: bulk

On Wed, Jul 26, 2023 at 12:08:26AM +0000, Joan Bruguera Mic? wrote:
> Hmmm, indeed linux-next crashes with a null pointer dereference when
> calling `ida_free(..., -1)`.
> 
> It appears to me that 69cb69ea5542 ("ida: Remove assertions that an ID was
> allocated") didn't intend to make `ida_free(..., -1)` invalid; after all,
> it was authored & introduced immediately after fc82bbf4dede ("ida: don't
> use BUG_ON() for debugging") whose commit message calls for making it
> legal, with Matthew's support.
> And the referenced Bluetooth HCI code that also calls `ida_free(..., -1)`
> is still there, as far as I can tell ([1]).
> 
> Rather, probably the `((int)id < 0)` was accidentally dropped, or the idea
> was that it would be handled by the `not_found:` label in `ida_free`, but
> for that to work you'd need to change the `!test_bit(bit, bitmap->bitmap)`
> condition to `!bitmap || !test_bit(bit, bitmap->bitmap)` since otherwise
> `bitmap->bitmap` is a null pointer dereference.

I've been in two minds about that patch ever since I wrote it.  I've
dropped it from my tree for now.

But, er, you have pointed out a bug which is that we don't handle
!bitmap properly.  That should be fixed.