[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <BNP2PqwI58CzAxJejDx9BcX2nWxSOo-hoZ1qh_7ZrmvqcNwz49RCBxZscYZCoHTlG7bqz2V12bhc5r88dPKopRgoy_XxO4QypcWoRFe3fMI=@protonmail.com>
Date: Mon, 14 Nov 2022 15:54:30 +0000
From: Barnabás Pőcze <pobrn@...tonmail.com>
To: Thomas Gleixner <tglx@...utronix.de>
Cc: linux-kernel@...r.kernel.org, John Stultz <jstultz@...gle.com>
Subject: Re: [PATCH v1] timerqueue: use rb_entry_safe() in timerqueue_getnext()
Hi
2022. november 14., hétfő 1:17 keltezéssel, Thomas Gleixner írta:
> On Thu, Oct 27 2022 at 21:37, Barnabás Pőcze wrote:
>
> > When `timerqueue_getnext()` is called on an empty timerqueue
> > the returned rb_node pointer will be NULL. Using `rb_entry()`
> > on a potentially NULL pointer will only - coincidentally - work
> > if the offset of the rb_node member is 0. This is currently the
> > case for `timerqueue_node`, but should this ever change,
> > `timerqueue_getnext()` will no longer work correctly on empty
> > timerqueues. To avoid this, switch to using `rb_entry_safe()`.
>
> I agree with the change but not with the argumentation above.
>
> Whatever the current offset of node is does not matter at all,
> really. This is a blantant missing NULL pointer check which works by
> chance.
>
> So there is no weasel wording justfied about "coincidentally" and "might
> not longer work correctly".
>
> Just spell it out that this is a blantant bug and nothing else.
I agree, I was just trying to describe why it has not caused any issues yet.
Would this be better?
When `timerqueue_getnext()` is called on an empty timer queue, it will
use `rb_entry()` on a NULL pointer, which is invalid. Fix that by using
`rb_entry_safe()` which handles NULL pointers.
This has not caused any issues so far because the offset of the `rb_node`
member in `timerqueue_node` is 0, so `rb_entry()` is essentially a no-op.
>
> Back then when that code got introduced rb_entry_safe() did not exist at
> all so it's even more obvious that this is simply a missing NULL pointer
> check, right?
As far as I can tell it did exist and it was actually used
when the offending change was committed (511885d7061e).
>
> This also requires a Fixes: tag which flags the commit which introduces
> this bug.
>
> Thanks,
>
> tglx
>
Regards,
Barnabás Pőcze
Powered by blists - more mailing lists