On 2016/4/6 19:37, Baozeng Ding wrote:

On 2016/4/5 22:18, Takashi Iwai wrote:

...

On Tue, 05 Apr 2016 15:51:30 +0200, Baozeng Ding wrote:

...

Hi all, I've got the following report (use-after-free in kill_fasync) while running syzkaller. Unfortunately no reproducer.The kernel version is 4.5 (on Mar 16 commit 09fd671ccb2475436bd5f597f751ca4a7d177aea).

================================================================== BUG: KASAN: use-after-free in kill_fasync+0x3fb/0x420 at addr ffff880067691d88 Read of size 8 by task swapper/2/0 =============================================================================

BUG kmalloc-2048 (Not tainted): kasan: bad access detected

Disabling lock debugging due to kernel taint INFO: Allocated in 0xbbbbbbbbbbbbbbbb age=18446678412249576073 cpu=2245704320 pid=-1 [< inline >] kmalloc /kernel/include/linux/slab.h:472 [< inline >] kzalloc /kernel/include/linux/slab.h:616 [< none >] snd_pcm_attach_substream+0x3b4/0xb10 /kernel/sound/core/pcm.c:966 [< none >] ___slab_alloc+0x4c7/0x500 /kernel/mm/slub.c:2446 [< none >] __slab_alloc+0x4c/0x90 /kernel/mm/slub.c:2475 [< inline >] slab_alloc_node /kernel/mm/slub.c:2538 [< inline >] slab_alloc /kernel/mm/slub.c:2580 [< none >] kmem_cache_alloc_trace+0x262/0x300 /kernel/mm/slub.c:2597 [< inline >] kmalloc /kernel/include/linux/slab.h:472 [< inline >] kzalloc /kernel/include/linux/slab.h:616 [< none >] snd_pcm_attach_substream+0x3b4/0xb10 /kernel/sound/core/pcm.c:966 [< none >] snd_pcm_open_substream+0x84/0x450 /kernel/sound/core/pcm_native.c:2262 [< inline >] snd_pcm_oss_open_file /kernel/sound/core/oss/pcm_oss.c:2346 [< none >] snd_pcm_oss_open.part.17+0x5a4/0x1100 /kernel/sound/core/oss/pcm_oss.c:2428 [< none >] snd_pcm_oss_open+0x35/0x50 /kernel/sound/core/oss/pcm_oss.c:2392 [< none >] soundcore_open+0x30f/0x640 /kernel/sound/sound_core.c:639 [< none >] chrdev_open+0x22a/0x4c0 /kernel/fs/char_dev.c:388 [< none >] do_dentry_open+0x6a2/0xcb0 /kernel/fs/open.c:736 [< none >] vfs_open+0x17b/0x1f0 /kernel/fs/open.c:853 [< inline >] do_last /kernel/fs/namei.c:3258 [< none >] path_openat+0x4837/0x5830 /kernel/fs/namei.c:3394 [< none >] do_filp_open+0x18e/0x250 /kernel/fs/namei.c:3429 [< none >] do_sys_open+0x201/0x420 /kernel/fs/open.c:1022 [< inline >] SYSC_open /kernel/fs/open.c:1040 [< none >] SyS_open+0x2d/0x40 /kernel/fs/open.c:1035 INFO: Freed in 0x10000b076 age=18446678416544543380 cpu=0 pid=0 [< none >] snd_pcm_detach_substream+0x134/0x280 /kernel/sound/core/pcm.c:1017 [< none >] __slab_free+0x1e8/0x300 /kernel/mm/slub.c:2657 [< inline >] slab_free /kernel/mm/slub.c:2810 [< none >] kfree+0x24e/0x2d0 /kernel/mm/slub.c:3661 [< none >] snd_pcm_detach_substream+0x134/0x280 /kernel/sound/core/pcm.c:1017 [< none >] snd_pcm_release_substream.part.38+0x219/0x2f0 /kernel/sound/core/pcm_native.c:2250 [< none >] snd_pcm_release_substream+0x59/0x70 /kernel/sound/core/pcm_native.c:2251 [< none >] snd_pcm_oss_release_file+0x45/0xb0 /kernel/sound/core/oss/pcm_oss.c:2305 [< none >] snd_pcm_oss_release+0xfa/0x250 /kernel/sound/core/oss/pcm_oss.c:2485 [< none >] __fput+0x236/0x780 /kernel/fs/file_table.c:208 [< none >] ____fput+0x15/0x20 /kernel/fs/file_table.c:244 [< none >] task_work_run+0x16b/0x200 /kernel/kernel/task_work.c:115 [< inline >] exit_task_work /kernel/include/linux/task_work.h:21 [< none >] do_exit+0x87f/0x2c90 /kernel/kernel/exit.c:748 [< none >] do_group_exit+0x108/0x330 /kernel/kernel/exit.c:878 [< inline >] SYSC_exit_group /kernel/kernel/exit.c:889 [< none >] SyS_exit_group+0x1d/0x20 /kernel/kernel/exit.c:887 [< none >] entry_SYSCALL_64_fastpath+0x23/0xc1 /kernel/arch/x86/entry/entry_64.S:207 INFO: Slab 0xffffea00019da400 objects=13 used=8 fp=0xffff880067691be0 flags=0x5fffc0000004080 INFO: Object 0xffff880067691bd8 @offset=7128 fp=0xbbbbbbbbbbbbbbbb Call Trace: <IRQ> [< inline >] __dump_stack /kernel/lib/dump_stack.c:15 <IRQ> [<ffffffff82945051>] dump_stack+0xb3/0x112 /kernel/lib/dump_stack.c:51 [<ffffffff817009ad>] print_trailer+0x10d/0x190 /kernel/mm/slub.c:668 [<ffffffff817074af>] object_err+0x2f/0x40 /kernel/mm/slub.c:675 [< inline >] print_address_description /kernel/mm/kasan/report.c:138 [<ffffffff81709ca5>] kasan_report_error+0x215/0x530 /kernel/mm/kasan/report.c:236 [< inline >] ? spin_lock /kernel/include/linux/spinlock.h:302 [<ffffffff84a1c550>] ? snd_pcm_stream_lock+0x80/0xd0 /kernel/sound/core/pcm_native.c:104 [< inline >] kasan_report /kernel/mm/kasan/report.c:259 [<ffffffff8170a0be>] __asan_report_load8_noabort+0x3e/0x40 /kernel/mm/kasan/report.c:280 [<ffffffff81792a0b>] ? kill_fasync+0x3fb/0x420 /kernel/fs/fcntl.c:729 [<ffffffff81792a0b>] kill_fasync+0x3fb/0x420 /kernel/fs/fcntl.c:729 [<ffffffff84a36e08>] snd_pcm_period_elapsed+0x1c8/0x230 /kernel/sound/core/pcm_lib.c:1890 [<ffffffff84b6e2e0>] ? get_bound_vga.isra.21.part.22+0x140/0x140 /kernel/sound/pci/hda/hda_intel.c:1327 [<ffffffff84b0637d>] stream_update+0xad/0xe0 /kernel/sound/pci/hda/hda_controller.c:923 [<ffffffff84c1583e>] snd_hdac_bus_handle_stream_irq+0x24e/0x350 /kernel/sound/hda/hdac_controller.c:449 [<ffffffff84b062d0>] ? azx_init_chip+0x100/0x100 /kernel/sound/pci/hda/hda_controller.c:890 [<ffffffff84b04d24>] ? azx_interrupt+0x14/0x4d0 /kernel/sound/pci/hda/hda_controller.c:929 [<ffffffff84b04eee>] azx_interrupt+0x1de/0x4d0 /kernel/sound/pci/hda/hda_controller.c:954 [<ffffffff84b04d10>] ? azx_stop_chip+0x20/0x20 /kernel/sound/pci/hda/hda_controller.c:908 [<ffffffff8143d9e3>] handle_irq_event_percpu+0xf3/0x790 /kernel/kernel/irq/handle.c:145 [<ffffffff8143e127>] handle_irq_event+0xa7/0x140 /kernel/kernel/irq/handle.c:192 [<ffffffff81447f81>] handle_edge_irq+0x1e1/0x8d0 /kernel/kernel/irq/chip.c:623 [< inline >] generic_handle_irq_desc /kernel/include/linux/irqdesc.h:146 [<ffffffff811ad739>] handle_irq+0x109/0x2a0 /kernel/arch/x86/kernel/irq_64.c:78 [< inline >] ? rcu_lock_release /kernel/include/linux/rcupdate.h:491 [< inline >] ? rcu_read_unlock /kernel/include/linux/rcupdate.h:926 [< inline >] ? __atomic_notifier_call_chain /kernel/kernel/notifier.c:184 [<ffffffff81369c4f>] ? atomic_notifier_call_chain+0xbf/0x140 /kernel/kernel/notifier.c:193 [<ffffffff81369b90>] ? __atomic_notifier_call_chain+0x150/0x150 /kernel/include/linux/rcupdate.h:922 [<ffffffff811ac36d>] do_IRQ+0x7d/0x1a0 /kernel/arch/x86/kernel/irq.c:240 [<ffffffff85dac40c>] common_interrupt+0x8c/0x8c /kernel/arch/x86/entry/entry_64.S:454 <EOI> [<ffffffff812245c6>] ? native_safe_halt+0x6/0x10 /kernel/./arch/x86/include/asm/irqflags.h:49 [<ffffffff81405ced>] ? trace_hardirqs_on+0xd/0x10 /kernel/kernel/locking/lockdep.c:2635 [< inline >] arch_safe_halt /kernel/./arch/x86/include/asm/paravirt.h:117 [<ffffffff811bf152>] default_idle+0x22/0x2d0 /kernel/arch/x86/kernel/process.c:307 [<ffffffff811c053a>] arch_cpu_idle+0xa/0x10 /kernel/arch/x86/kernel/process.c:298 [<ffffffff813eb618>] default_idle_call+0x48/0x70 /kernel/kernel/sched/idle.c:93 [< inline >] cpuidle_idle_call /kernel/kernel/sched/idle.c:151 [< inline >] cpu_idle_loop /kernel/kernel/sched/idle.c:242 [<ffffffff813ebaa7>] cpu_startup_entry+0x467/0x600 /kernel/kernel/sched/idle.c:291 [<ffffffff81201432>] start_secondary+0x2b2/0x380 /kernel/arch/x86/kernel/smpboot.c:259 [<ffffffff81201180>] ? set_cpu_sibling_map+0x18a0/0x18a0 /kernel/include/linux/topology.h:80 Memory state around the buggy address: ffff880067691c80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff880067691d00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb

...

ffff880067691d80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb

                    ^

ffff880067691e00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff880067691e80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ==================================================================

It seems that in function snd_pcm_detach_substream, when freeing runtime structure, it did not handle its fasync data pointer. Any ideas? Thanks.

Well, fasync doesn't need a cleanup usually. Do you have a proper reproducer code?

Sorry, I cannot reproduce it. I will apply your patch and test it. I will tell you the result whether it is okay a few days later. Thanks.

...

In anyway, try the patch below. I'm not sure whether this would help, but if the path is via kill_fasync(), it'd be good to put in the protected context.

During the last few weeks, I have not seen the bug any more after applying your patch. Thanks.

...

Takashi

---

diff --git a/sound/core/pcm_lib.c b/sound/core/pcm_lib.c index 3a9b66c6e09c..0aca39762ed0 100644 --- a/sound/core/pcm_lib.c +++ b/sound/core/pcm_lib.c @@ -1886,8 +1886,8 @@ void snd_pcm_period_elapsed(struct snd_pcm_substream *substream) snd_timer_interrupt(substream->timer, 1); #endif _end:

• snd_pcm_stream_unlock_irqrestore(substream, flags); kill_fasync(&runtime->fasync, SIGIO, POLL_IN);

• snd_pcm_stream_unlock_irqrestore(substream, flags); } EXPORT_SYMBOL(snd_pcm_period_elapsed);