Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755345AbaG3Qfq (ORCPT ); Wed, 30 Jul 2014 12:35:46 -0400 Received: from cantor2.suse.de ([195.135.220.15]:42184 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754159AbaG3Qfo (ORCPT ); Wed, 30 Jul 2014 12:35:44 -0400 Date: Wed, 30 Jul 2014 18:35:40 +0200 Message-ID: From: Takashi Iwai To: Josef Bacik Cc: Chris Mason , , Subject: Re: [PATCH] Btrfs: Fix memory corruption by ulist_add_merge() on 32bit arch In-Reply-To: <53D916DB.9020901@fb.com> References: <1406537824-2376-1-git-send-email-tiwai@suse.de> <53D64D40.80606@fb.com> <53D9015A.90001@fb.com> <53D911DE.7000806@fb.com> <53D916DB.9020901@fb.com> User-Agent: Wanderlust/2.15.9 (Almost Unreal) SEMI/1.14.6 (Maruoka) FLIM/1.14.9 (=?UTF-8?B?R29qxY0=?=) APEL/10.8 Emacs/24.3 (x86_64-suse-linux-gnu) MULE/6.0 (HANACHIRUSATO) MIME-Version: 1.0 (generated by SEMI 1.14.6 - "Maruoka") Content-Type: text/plain; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org 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: [] 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:[] 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: > >>>>>>>>>> [] __btrfs_find_all_roots+0x9d/0xf0 [btrfs] > >>>>>>>>>> [] btrfs_qgroup_rescan_worker+0x401/0x760 [btrfs] > >>>>>>>>>> [] normal_work_helper+0xc8/0x270 [btrfs] > >>>>>>>>>> [] process_one_work+0x11b/0x390 > >>>>>>>>>> [] worker_thread+0x101/0x340 > >>>>>>>>>> [] kthread+0x9b/0xb0 > >>>>>>>>>> [] ret_from_kernel_thread+0x21/0x30 > >>>>>>>>>> [] 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: [v3.11+] > >>>>>>>>>> Signed-off-by: Takashi Iwai > >>>>>>>>>> --- > >>>>>>>>>> > >>>>>>>>>> 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@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/