Re: [PATCH] fork: return proper negative error code

[Christian Brauner]

On Mon, Jul 01, 2019 at 04:48:08PM +0200, Christian Brauner wrote:
> Make sure to return a proper negative error code from copy_process()
> when anon_inode_getfile() fails with CLONE_PIDFD.
> Otherwise _do_fork() will not detect an error and get_task_pid() will
> operator on a nonsensical pointer:
> 
> R10: 0000000000000000 R11: 0000000000000246 R12: 00000000006dbc2c
> R13: 00007ffc15fbb0ff R14: 00007ff07e47e9c0 R15: 0000000000000000
> kasan: CONFIG_KASAN_INLINE enabled
> kasan: GPF could be caused by NULL-ptr deref or user memory access
> general protection fault: 0000 [#1] PREEMPT SMP KASAN
> CPU: 1 PID: 7990 Comm: syz-executor290 Not tainted 5.2.0-rc6+ #9
> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS
> Google 01/01/2011
> RIP: 0010:__read_once_size include/linux/compiler.h:194 [inline]
> RIP: 0010:get_task_pid+0xe1/0x210 kernel/pid.c:372
> Code: 89 ff e8 62 27 5f 00 49 8b 07 44 89 f1 4c 8d bc c8 90 01 00 00 eb 0c
> e8 0d fe 25 00 49 81 c7 38 05 00 00 4c 89 f8 48 c1 e8 03 <80> 3c 18 00 74
> 08 4c 89 ff e8 31 27 5f 00 4d 8b 37 e8 f9 47 12 00
> RSP: 0018:ffff88808a4a7d78 EFLAGS: 00010203
> RAX: 00000000000000a7 RBX: dffffc0000000000 RCX: ffff888088180600
> RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
> RBP: ffff88808a4a7d90 R08: ffffffff814fb3a8 R09: ffffed1015d66bf8
> R10: ffffed1015d66bf8 R11: 1ffff11015d66bf7 R12: 0000000000041ffc
> R13: 1ffff11011494fbc R14: 0000000000000000 R15: 000000000000053d
> FS:  00007ff07e47e700(0000) GS:ffff8880aeb00000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 00000000004b5100 CR3: 0000000094df2000 CR4: 00000000001406e0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> Call Trace:
>   _do_fork+0x1b9/0x5f0 kernel/fork.c:2360
>   __do_sys_clone kernel/fork.c:2454 [inline]
>   __se_sys_clone kernel/fork.c:2448 [inline]
>   __x64_sys_clone+0xc1/0xd0 kernel/fork.c:2448
>   do_syscall_64+0xfe/0x140 arch/x86/entry/common.c:301
>   entry_SYSCALL_64_after_hwframe+0x49/0xbe
> 
> Link: https://lore.kernel.org/lkml/000000000000e0dc0d058c9e7142@xxxxxxxxxx
> Reported-and-tested-by: syzbot+002e636502bc4b64eb5c@xxxxxxxxxxxxxxxxxxxxxxxxx
> Fixes: 6fd2fe494b17 ("copy_process(): don't use ksys_close() on cleanups")
> Cc: Jann Horn <jannh@xxxxxxxxxx>
> Cc: Al Viro <viro@xxxxxxxxxxxxxxxxxx>
> Signed-off-by: Christian Brauner <christian@xxxxxxxxxx>

syzkaller has verified that this fixes the bug. The patch is sitting in
my fixes branch. I'm sending a PR for this in a little bit to get this
out of the way.

Thanks!
Christian