Re: [syzbot] [keyrings?] [lsm?] WARNING in __mod_timer

Re: [syzbot] [keyrings?] [lsm?] WARNING in __mod_timer

[Hillf Danton]

On Sun, 26 Feb 2023 19:55:37 -0800
> Hello,
> 
> syzbot found the following issue on:
> 
> HEAD commit:    d8ca6dbb8de7 Merge tag 'nfs-for-6.3-1' of git://git.linux-..
> git tree:       upstream
> console output: https://syzkaller.appspot.com/x/log.txt?x=109cec8cc80000
> kernel config:  https://syzkaller.appspot.com/x/.config?x=636897a0dac3d81e
> dashboard link: https://syzkaller.appspot.com/bug?extid=5093ba19745994288b53
> compiler:       Debian clang version 15.0.7, GNU ld (GNU Binutils for Debian) 2.35.2
> 
> Unfortunately, I don't have any reproducer for this issue yet.
> 
> Downloadable assets:
> disk image: https://storage.googleapis.com/syzbot-assets/6eea6a7459e5/disk-d8ca6dbb.raw.xz
> vmlinux: https://storage.googleapis.com/syzbot-assets/6bcbaedb88c0/vmlinux-d8ca6dbb.xz
> kernel image: https://storage.googleapis.com/syzbot-assets/981f6d71a5c5/bzImage-d8ca6dbb.xz
> 
> IMPORTANT: if you fix the issue, please add the following tag to the commit:
> Reported-by: syzbot+5093ba19745994288b53@syzkaller.appspotmail.com
> 
> ------------[ cut here ]------------
> ODEBUG: assert_init not available (active state 0) object: ffffffff8d4fcbc0 object type: timer_list hint: key_gc_timer_func+0x0/0x80 security/keys/gc.c:117
> WARNING: CPU: 1 PID: 10646 at lib/debugobjects.c:512 debug_print_object lib/debugobjects.c:509 [inline]
> WARNING: CPU: 1 PID: 10646 at lib/debugobjects.c:512 debug_object_assert_init+0x1f2/0x240 lib/debugobjects.c:899
> Modules linked in:
> CPU: 1 PID: 10646 Comm: syz-executor.3 Not tainted 6.2.0-syzkaller-06695-gd8ca6dbb8de7 #0
> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/21/2023
> RIP: 0010:debug_print_object lib/debugobjects.c:509 [inline]
> RIP: 0010:debug_object_assert_init+0x1f2/0x240 lib/debugobjects.c:899
> Code: a0 fd 4c 8b 4d 00 48 c7 c7 20 68 38 8b 48 c7 c6 00 65 38 8b 48 c7 c2 e0 69 38 8b 31 c9 4d 89 e8 53 e8 42 6f 10 fd 48 83 c4 08 <0f> 0b ff 05 42 e1 10 0a 48 83 c5 38 48 89 e8 48 c1 e8 03 42 80 3c
> RSP: 0018:ffffc9000b1276f8 EFLAGS: 00010282
> RAX: caefb79dfa053800 RBX: ffffffff83f63700 RCX: 0000000000040000
> RDX: ffffc90014421000 RSI: 0000000000018ba9 RDI: 0000000000018baa
> RBP: ffffffff8aec5400 R08: ffffffff81536d42 R09: fffff52001624e55
> R10: 0000000000000000 R11: dffffc0000000001 R12: dffffc0000000000
> R13: ffffffff8d4fcbc0 R14: 0000000000000004 R15: ffffffff91d3e4e8
> FS:  00007f122cc29700(0000) GS:ffff8880b9900000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 000055555632f708 CR3: 00000000690df000 CR4: 00000000003526e0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> Call Trace:
>  <TASK>
>  debug_timer_assert_init kernel/time/timer.c:792 [inline]
>  debug_assert_init kernel/time/timer.c:837 [inline]
>  __mod_timer+0x10d/0xf40 kernel/time/timer.c:1020
>  key_reject_and_link+0x3f5/0x6e0 security/keys/key.c:610
>  key_negate_and_link include/linux/key-type.h:187 [inline]
>  complete_request_key security/keys/request_key.c:64 [inline]
>  call_sbin_request_key+0xa7b/0xcd0 security/keys/request_key.c:213
>  construct_key security/keys/request_key.c:244 [inline]
>  construct_key_and_link security/keys/request_key.c:503 [inline]
>  request_key_and_link+0x11e3/0x18e0 security/keys/request_key.c:637
>  __do_sys_request_key security/keys/keyctl.c:222 [inline]
>  __se_sys_request_key+0x271/0x3b0 security/keys/keyctl.c:167
>  do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>  do_syscall_64+0x41/0xc0 arch/x86/entry/common.c:80
>  entry_SYSCALL_64_after_hwframe+0x63/0xcd

Given static DEFINE_TIMER(key_gc_timer, key_gc_timer_func), is this report
a false positive one because DEFINE_TIMER is not covered by
CONFIG_DEBUG_OBJECTS_TIMERS?

Hillf

Re: [syzbot] [keyrings?] [lsm?] WARNING in __mod_timer

[Hillf Danton]

On Mon, 27 Feb 2023 21:33:03 +0100 Thomas Gleixner <tglx@linutronix.de>
> On Sun, Feb 26 2023 at 19:55, syzbot wrote:
> > ODEBUG: assert_init not available (active state 0) object: ffffffff8d4fcbc0 object type: timer_list hint: key_gc_timer_func+0x0/0x80 security/keys/gc.c:117
> 
> >  WARNING: CPU: 1 PID: 10646 at lib/debugobjects.c:512 debug_object_assert_init+0x1f2/0x240 lib/debugobjects.c:899
> >  debug_assert_init kernel/time/timer.c:837 [inline]
> >  __mod_timer+0x10d/0xf40 kernel/time/timer.c:1020
> >  key_reject_and_link+0x3f5/0x6e0 security/keys/key.c:610
> >  key_negate_and_link include/linux/key-type.h:187 [inline]
> >  complete_request_key security/keys/request_key.c:64 [inline]
> >  call_sbin_request_key+0xa7b/0xcd0 security/keys/request_key.c:213
> >  construct_key security/keys/request_key.c:244 [inline]
> >  construct_key_and_link security/keys/request_key.c:503 [inline]
> >  request_key_and_link+0x11e3/0x18e0 security/keys/request_key.c:637
> >  __do_sys_request_key security/keys/keyctl.c:222 [inline]
> >  __se_sys_request_key+0x271/0x3b0 security/keys/keyctl.c:167
> 
> This is odd. The timer object is statically allocated via
> DEFINE_TIMER(). That macro sets
> 
>        timer.entry.next = TIMER_ENTRY_STATIC
> 
> which is used to detect statically allocated timer objects via
> timer_is_static_object() and that checks for:
> 
>      timer.entry.pprev == NULL && timer.entry.next == TIMER_ENTRY_STATIC

List operations like hlist_add_head() and __hlist_del() make
timer_is_static_object() return false positive result.

> 
> The only function which touches key_gc_timer is
> 
>     key_reject_and_link()
>       mod_timer()
>         __mod_timer()
>           debug_assert_init()

Commit d02e382cef06 ("timers: Silently ignore timers with a NULL function")
added this assert.

>             debug_timer_assert_init()
>               debug_object_assert_init()
>                 if (!lookup_object()) {
>                    if (!check_for_static_object()) <- Invokes timer_is_static_object()
>                       WARN()


	cpu 0				cpu 2
	---				---
	mod_timer()			mod_timer()
	__mod_timer()			__mod_timer()
	...				...
	debug_object_assert_init()
	raw_spin_lock_irqsave(&db->lock, flags);
	if (!lookup_object()) {
		raw_spin_unlock_irqrestore(&db->lock, flags);

					raw_spin_lock_irqsave(&db->lock, flags);
					if (!lookup_object()) {
						raw_spin_unlock_irqrestore(&db->lock, flags);

		if (check_for_static_object())
			fine;
						if (!check_for_static_object())
							WARN;


Depending on TIMER_ENTRY_STATIC makes check_for_static_object() fragile as a
static timer is always static regardless if it is enqueued.

The fragility is another explanation.

> 
> If this is the first invocation of mod_timer(&key_gc_timer,...) then
> key_gc_timer is corrupted.
> 
> If this is not the first invocation of mod_timer(&key_gc_timer,...) then
> the debugobjects hash is corrupted.
> 
> Either way neither the timer code nor debugobjects have been changed
> since the 6.2 release and certainly are innocent here.
> 
> That smells like a nasty memory corruption issue and the two other
> syzbot reports which arrived in my filtered inbox:
> 
>  https://lore.kernel.org/all/000000000000d7894b05f5924787@google.com
>  https://lore.kernel.org/all/000000000000840dae05f5a7fb53@google.com
> 
> point to memory corruption as well.
> 
> The first one has a C reproducer. Can that be used for bisection?
> 
> Thanks,
> 
>         tglx
>