Re: fs: NULL deref in atime_needs_update

From: Ian Kent <raven@themaw.net>
To: "Mickaël Salaün" <mic@digikod.net>,
	"Dmitry Vyukov" <dvyukov@google.com>,
	"Alexander Viro" <viro@zeniv.linux.org.uk>,
	"linux-fsdevel@vger.kernel.org" <linux-fsdevel@vger.kernel.org>,
	LKML <linux-kernel@vger.kernel.org>
Cc: syzkaller <syzkaller@googlegroups.com>,
	Kostya Serebryany <kcc@google.com>,
	Alexander Potapenko <glider@google.com>,
	Sasha Levin <sasha.levin@oracle.com>
Subject: Re: fs: NULL deref in atime_needs_update
Date: Wed, 24 Feb 2016 11:12:13 +0800	[thread overview]
Message-ID: <1456283533.2933.20.camel@themaw.net> (raw)
In-Reply-To: <56C3B35E.6020109@digikod.net>

On Wed, 2016-02-17 at 00:40 +0100, Mickaël Salaün wrote:
> Hi,
> 
> Actually I found the same bug (without fuzzing) and I can reproduce it
> in a deterministic way (e.g. by creating a LSM that return 1 for the
> security_file_open hook). At least, from v4.2.8 I can easily trigger
> traces like this :

Reading through this thread I wonder if this is a new problem.

Is there a previous kernel it can't be reproduced on?
Perhaps a bisect will shed some light on what's happening.

> 
> BUG: unable to handle kernel NULL pointer dereference at
> 0000000000000050
> IP: [<ffffffff81170871>] atime_needs_update+0x11/0xc0
> PGD 127b17067 PUD 12ab2e067 PMD 0 
> Oops: 0000 [#45] SMP 
> [...]
> RIP: 0010:[<ffffffff81170871>]  [<ffffffff81170871>]
> atime_needs_update+0x11/0xc0
> RSP: 0018:ffff880127853c18  EFLAGS: 00010246
> RAX: ffff88012ad0c080 RBX: ffff88012ad0c1d8 RCX: ffff88012ad0c080
> RDX: 0000000000000000 RSI: ffff88012ad0c1d8 RDI: ffff880127853d98
> RBP: ffff880127853c28 R08: ffff8800cc0a2540 R09: ffff8800cfbfc320
> R10: ffff8800cc0a2540 R11: 0000000000000001 R12: ffff8800cb5d6300
> R13: 0000000000000000 R14: ffff88012ad0c080 R15: ffff880127853e7c
> FS:  00007f1054aae700(0000) GS:ffff88012fc40000(0000)
> knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 0000000000000050 CR3: 0000000127977000 CR4: 00000000000406e0
> Stack:
> ffff88012ad0c1d8 ffff8800cb5d6300 ffff880127853c60 ffffffff8117094e
> ffff8800c9ade3c0 0000000000000000 00000000a670294f ffff880127853d70
> ffff880127853d98 ffff880127853c98 ffffffff8116071c ffff8800cb4ada80
> Call Trace:
> [<ffffffff8117094e>] ? touch_atime+0x2e/0xd0
> [<ffffffff8116071c>] ? trailing_symlink+0xec/0x280
> [<ffffffff81163a78>] ? path_openat+0x468/0x1240
> [<ffffffff8111856d>] ? pagevec_lru_move_fn+0xed/0x110
> [<ffffffff81117ff0>] ? __activate_page+0x130/0x130
> [<ffffffff8116593c>] ? do_filp_open+0x8c/0x100
> [<ffffffff81164dec>] ? filename_lookup+0xec/0x180
> [<ffffffff8115bc24>] ? do_open_execat+0x74/0x170
> [<ffffffff8115d437>] ? do_execveat_common.isra.42+0x1a7/0x6a0
> [<ffffffff8115db90>] ? SyS_execve+0x30/0x40
> [<ffffffff8156ad65>] ? stub_execve+0x5/0x5
> [<ffffffff8156aadb>] ? entry_SYSCALL_64_fastpath+0x16/0x6a
> Code: 89 c7 e8 63 eb ff ff 48 89 d8 5b c3 0f 1f 40 00 66 2e 0f 1f 84
> 00 00 00 00 00 55 48 89 e5 41 54 53 f6 46 0c 02 75 72 48 8b 56 28 <48>
> 8b 42 50 a9 01 04 00 00 75 63 f6 c4 08 75 65 4c 8b 27 41 8b 
> RIP  [<ffffffff81170871>] atime_needs_update+0x11/0xc0
> RSP <ffff880127853c18>
> CR2: 0000000000000050
> ---[ end trace 97dc4f4bb0214bd8 ]---
> 
> 
> Regards,
>  Mickaël
> 
> 
> On 05/02/2016 22:11, Dmitry Vyukov wrote:
> > Hello,
> > 
> > I've hit the following GPF while running syzkaller fuzzer:
> > 
> > general protection fault: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN
> > Modules linked in:
> > CPU: 1 PID: 5178 Comm: syz-executor Not tainted 4.5.0-rc2+ #65
> > Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs
> > 01/01/2011
> > task: ffff880064768000 ti: ffff8800622c0000 task.ti:
> > ffff8800622c0000
> > RIP: 0010:[<ffffffff8181aa5d>]  [<ffffffff8181aa5d>]
> > atime_needs_update+0x2d/0x460
> > RSP: 0018:ffff8800622c7a30  EFLAGS: 00010203
> > RAX: dffffc0000000000 RBX: 0000000000000000 RCX: dffffc0000000000
> > RDX: 0000000000000001 RSI: 0000000000000000 RDI: 000000000000000c
> > RBP: ffff8800622c7a58 R08: 0000000000000001 R09: 0000000000000000
> > R10: 0000000000000000 R11: 0000000000000001 R12: ffff8800622c7c08
> > R13: ffff8800622c7c08 R14: ffff8800301ca322 R15: ffff8800622c7bb0
> > FS:  00007fd1c9f8b700(0000) GS:ffff88003ed00000(0000)
> > knlGS:0000000000000000
> > CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
> > CR2: 0000000020f31000 CR3: 0000000062274000 CR4: 00000000000006e0
> > Stack:
> >  ffff8800622c7bf4 0000000000000000 ffff8800622c7c08 ffff8800301ca322
> >  ffff8800622c7bb0 ffff8800622c7b38 ffffffff817ecd91 ffff880030bf5200
> >  ffff8800622c7bb8 1ffff1000c458f56 ffff8800622c7c00 ffff8800622c7be0
> > Call Trace:
> >  [<     inline     >] get_link fs/namei.c:1006
> >  [<ffffffff817ecd91>] link_path_walk+0xaf1/0x1030 fs/namei.c:1968
> >  [<ffffffff817ed311>] path_parentat+0x41/0x150 fs/namei.c:2176
> >  [<ffffffff817f4c5c>] filename_parentat+0x17c/0x3c0 fs/namei.c:2198
> >  [<     inline     >] user_path_parent fs/namei.c:2412
> >  [<     inline     >] SYSC_renameat2 fs/namei.c:4411
> >  [<     inline     >] SyS_renameat2 fs/namei.c:4375
> >  [<     inline     >] SYSC_renameat fs/namei.c:4521
> >  [<ffffffff817f9a72>] SyS_renameat+0x192/0x820 fs/namei.c:4518
> >  [<ffffffff8669e0b6>] entry_SYSCALL_64_fastpath+0x16/0x7a
> > arch/x86/entry/entry_64.S:185
> > Code: 89 e5 41 57 41 56 41 55 41 54 49 89 fc 53 48 89 f3 e8 08 25 d5
> > ff 48 8d 7b 0c 48 b8 00 00 00 00 00 fc ff df 48 89 fa 48 c1 ea 03
> > <0f>
> > b6 14 02 48 89 f8 83 e0 07 83 c0 03 38 d0 7c 08 84 d2 0f 85
> > RIP  [<ffffffff8181aa5d>] atime_needs_update+0x2d/0x460
> > fs/inode.c:1611
> >  RSP <ffff8800622c7a30>
> > ---[ end trace 1a4c9bda4680ce46 ]---
> > 
> > On commit df48ab3c2f5ffca88b7803ffbadd074bd5a0a2ef.
> > 
> > Objdump shows that inode is NULL in atime_needs_update.
> > 
> > Unfortunately reproduction of this crash is very hard. The program
> > executes something along the lines of:
> > 
> > mmap(0x20000000, 15945728, PROT_READ|PROT_WRITE,
> > MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0x20000000
> > mkdir("./bus", 0662515705056234013740)  = 0
> > openat(AT_FDCWD, "./bus", O_RDONLY|O_EXCL) = 3
> > symlinkat("../bus", 3, "./bus")         = 0
> > renameat(3, "./bus", 3, "./bus/file0")  = 0
> > mmap(0x20f35000, 4096, PROT_READ|PROT_WRITE,
> > MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0x20f35000
> > mount("./bus", "./bus", 0x20f2aee4,
> > MS_RDONLY|MS_NODEV|MS_RELATIME|MS_NODIRATIME|MS_BIND|MS_MOVE|MS_REC|
> > MS_UNBINDABLE|MS_SLAVE|MS_SHARED|0xc000380,
> > 0x20093f5f) = 0
> > open("./bus/file0", O_RDWR|O_EXCL)      = -1 EISDIR (Is a directory)
> > exit_group(0)                           = ?
> > 
> > But in multiple threads so that some calls can be doubled and/or
> > overlapped. And all this happens on a tmpfs mount.
> > 
> > But I was able to reproduce it 8 or so times, so I am sure that it
> > is real.
> > 
> > For future reference, I was running these programs:
> > https://gist.githubusercontent.com/dvyukov/124c457d308fa724d88a/raw/
> > fec2d86e125a7fd2fa2916791d65d7daead7cbbb/gistfile1.txt
> > Following these instructions:
> > https://github.com/google/syzkaller/wiki/How-to-execute-syzkaller-pr
> > ograms
> > 
> 
> 