Re: [PATCH] Btrfs: Fix memory corruption by ulist_add_merge() on 32bit arch

From: Takashi Iwai
Date: Wed Jul 30 2014 - 12:35:48 EST


At Wed, 30 Jul 2014 12:01:31 -0400,
Josef Bacik wrote:
>
> On 07/30/2014 11:52 AM, Takashi Iwai wrote:
> > At Wed, 30 Jul 2014 11:40:14 -0400,
> > Josef Bacik wrote:
> >>
> >> On 07/30/2014 11:05 AM, Takashi Iwai wrote:
> >>> At Wed, 30 Jul 2014 17:01:52 +0200,
> >>> Takashi Iwai wrote:
> >>>>
> >>>> At Wed, 30 Jul 2014 10:29:46 -0400,
> >>>> Josef Bacik wrote:
> >>>>>
> >>>>> On 07/30/2014 05:57 AM, Takashi Iwai wrote:
> >>>>>> At Mon, 28 Jul 2014 16:01:55 +0200,
> >>>>>> Takashi Iwai wrote:
> >>>>>>>
> >>>>>>> At Mon, 28 Jul 2014 15:48:41 +0200,
> >>>>>>> Takashi Iwai wrote:
> >>>>>>>>
> >>>>>>>> At Mon, 28 Jul 2014 09:16:48 -0400,
> >>>>>>>> Josef Bacik wrote:
> >>>>>>>>>
> >>>>>>>>> On 07/28/2014 04:57 AM, Takashi Iwai wrote:
> >>>>>>>>>> We've got bug reports that btrfs crashes when quota is enabled on
> >>>>>>>>>> 32bit kernel, typically with the Oops like below:
> >>>>>>>>>> BUG: unable to handle kernel NULL pointer dereference at 00000004
> >>>>>>>>>> IP: [<f9234590>] find_parent_nodes+0x360/0x1380 [btrfs]
> >>>>>>>>>> *pde = 00000000
> >>>>>>>>>> Oops: 0000 [#1] SMP
> >>>>>>>>>> CPU: 0 PID: 151 Comm: kworker/u8:2 Tainted: G S W 3.15.2-1.gd43d97e-default #1
> >>>>>>>>>> Workqueue: btrfs-qgroup-rescan normal_work_helper [btrfs]
> >>>>>>>>>> task: f1478130 ti: f147c000 task.ti: f147c000
> >>>>>>>>>> EIP: 0060:[<f9234590>] EFLAGS: 00010213 CPU: 0
> >>>>>>>>>> EIP is at find_parent_nodes+0x360/0x1380 [btrfs]
> >>>>>>>>>> EAX: f147dda8 EBX: f147ddb0 ECX: 00000011 EDX: 00000000
> >>>>>>>>>> ESI: 00000000 EDI: f147dda4 EBP: f147ddf8 ESP: f147dd38
> >>>>>>>>>> DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
> >>>>>>>>>> CR0: 8005003b CR2: 00000004 CR3: 00bf3000 CR4: 00000690
> >>>>>>>>>> Stack:
> >>>>>>>>>> 00000000 00000000 f147dda4 00000050 00000001 00000000 00000001 00000050
> >>>>>>>>>> 00000001 00000000 d3059000 00000001 00000022 000000a8 00000000 00000000
> >>>>>>>>>> 00000000 000000a1 00000000 00000000 00000001 00000000 00000000 11800000
> >>>>>>>>>> Call Trace:
> >>>>>>>>>> [<f923564d>] __btrfs_find_all_roots+0x9d/0xf0 [btrfs]
> >>>>>>>>>> [<f9237bb1>] btrfs_qgroup_rescan_worker+0x401/0x760 [btrfs]
> >>>>>>>>>> [<f9206148>] normal_work_helper+0xc8/0x270 [btrfs]
> >>>>>>>>>> [<c025e38b>] process_one_work+0x11b/0x390
> >>>>>>>>>> [<c025eea1>] worker_thread+0x101/0x340
> >>>>>>>>>> [<c026432b>] kthread+0x9b/0xb0
> >>>>>>>>>> [<c0712a71>] ret_from_kernel_thread+0x21/0x30
> >>>>>>>>>> [<c0264290>] kthread_create_on_node+0x110/0x110
> >>>>>>>>>>
> >>>>>>>>>> This indicates a NULL corruption in prefs_delayed list. The further
> >>>>>>>>>> investigation and bisection pointed that the call of ulist_add_merge()
> >>>>>>>>>> results in the corruption.
> >>>>>>>>>>
> >>>>>>>>>> ulist_add_merge() takes u64 as aux and writes a 64bit value into
> >>>>>>>>>> old_aux. The callers of this function in backref.c, however, pass a
> >>>>>>>>>> pointer of a pointer to old_aux. That is, the function overwrites
> >>>>>>>>>> 64bit value on 32bit pointer. This caused a NULL in the adjacent
> >>>>>>>>>> variable, in this case, prefs_delayed.
> >>>>>>>>>>
> >>>>>>>>>> Here is a quick attempt to band-aid over this: a new function,
> >>>>>>>>>> ulist_add_merge_ptr() is introduced to pass/store properly a pointer
> >>>>>>>>>> value instead of u64. There are still ugly void ** cast remaining
> >>>>>>>>>> in the callers because void ** cannot be taken implicitly. But, it's
> >>>>>>>>>> safer than explicit cast to u64, anyway.
> >>>>>>>>>>
> >>>>>>>>>> Bugzilla: https://urldefense.proofpoint.com/v1/url?u=https://bugzilla.novell.com/show_bug.cgi?id%3D887046&k=ZVNjlDMF0FElm4dQtryO4A%3D%3D%0A&r=cKCbChRKsMpTX8ybrSkonQ%3D%3D%0A&m=m3qrbo6ngjqKO%2B7ofuwRfQflb9Cx%2FXrF8TKejkPjxfA%3D%0A&s=199a5b6f0ed181925e9ba2c1060fe20d1c8ad2831dd1d96cc7eddd2a343fa72b
> >>>>>>>>>> Cc: <stable@xxxxxxxxxxxxxxx> [v3.11+]
> >>>>>>>>>> Signed-off-by: Takashi Iwai <tiwai@xxxxxxx>
> >>>>>>>>>> ---
> >>>>>>>>>>
> >>>>>>>>>> Alternatively, we can change the argument of aux and old_aux to a
> >>>>>>>>>> pointer from u64, as backref.c is the only user of ulist_add_merge()
> >>>>>>>>>> function. I'll cook up another patch if it's the preferred way.
> >>>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> Yeah lets just use a pointer and see how that works out. Thanks,
> >>>>>>>>
> >>>>>>>> Oops, I forgot that ulist_add() takes aux as u64 and it calling
> >>>>>>>> ulist_add_merge() internally. So, we can't change the type blindly
> >>>>>>>> there, unfortunately.
> >>>>>>>
> >>>>>>> Looking back at the code, it seems that all aux arguments passed to
> >>>>>>> ulist_add() in qgroup.c are pointers, too. So, indeed, all aux values
> >>>>>>> are pointers, so far, and it'd be even cleaner to replace all these
> >>>>>>> from u64 to void *.
> >>>>>>>
> >>>>>>> But, such a replacement patch will become difficult for backporting to
> >>>>>>> stable kernels (the bug existed since 3.11, at least). So IMO, we
> >>>>>>> should put a smaller fix like my previous one, let it backported to
> >>>>>>> stable kernels, and do more comprehensive replacements to pointer on
> >>>>>>> its top.
> >>>>>>
> >>>>>> Ping. Could you guys take my original patch as is, or do you prefer
> >>>>>> changing in a different way? If so, how?
> >>>>>>
> >>>>>
> >>>>> I don't care how hard it is to backport to stable,
> >>>>
> >>>> You must do care as a maintainer. It's a long-standing and serious
> >>>> bug since 3.11. The kernel hangs up immediately when you enable quota
> >>>> on 32bit kernel. And it's really hard to revert it when the rootfs is
> >>>> btrfs. (The mount follows the immediate hang up after reboot.)
> >>>>
> >>
> >> What I mean is that we want the right fix first, not something that is easier to
> >> pull back to stable and then the right fix later. Do it right first and then
> >> backport it to the stable kernels, it's perfectly acceptable to adjust patches
> >> when sending them to the stable team. "But it's hard" is not a valid excuse for
> >> not doing it right the first time.
> >
> > Well, I guess this underestimates the burden of backports. Currently,
> > there is stable kernel for each kernel release. Anyone has to
> > backport for each version, and you'll be asked. I, as a long-time
> > subsystem maintainer, wouldn't go in that way :)
> >
> > And, speaking of "rightness" -- replacing the callers with a wrapper
> > is also a right fix. It's even a safer fix. That's basically why I
> > posted it as the primary patch.
> >
> > The merit of replacing all callers is that you can eliminate nasty
> > casts by that. This is however rather a cleanup, which is a different
> > bonus from what we need to fix.
> >
> >>>>> since we're using pointers
> >>>>> everywhere just change it to void * and be done with it. Thanks,
> >>>>
> >>>> Fix it quickly, then do cleanup. This is the golden rule for
> >>>> regression :)
> >>>
> >>> Also, another question is whether you guys are OK to change the type
> >>> to a pointer. Through a glance, the ulist code was intended to handle
> >>> any generic data, thus it uses u64, right? Using void pointer breaks
> >>> this concept.
> >>>
> >>
> >> It's fine, ulist today resembles very little from what it was originally. The
> >> current users all shove pointers into there, so we might as well just make it a
> >> pointer. Thanks,
> >
> > OK, good to know.
> >
> > If this post still doesn't convince you, I'll prepare the patch to do
> > all replacements.
> >
> >
>
> I don't care that much, do it however you want. Thanks,

Yes, I do care because I know of this kind of horror very well.

OK, I'm going to send the new patch(es).


Takashi
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/