binder epoll bug (was KASAN: use-after-free Read in __lock_acquire (2))

[Eric Biggers]

[+Cc binder maintainers and list]
[-Cc lockdep maintainers, USB maintainers, and other random people]

On Sat, Dec 02, 2017 at 08:08:01AM -0800, syzbot wrote:
> BUG: KASAN: use-after-free in __lock_acquire+0x465e/0x47f0
> kernel/locking/lockdep.c:3378
> Read of size 8 at addr ffff8801cd8e13f0 by task syzkaller236979/3086
> 
> CPU: 1 PID: 3086 Comm: syzkaller236979 Not tainted 4.15.0-rc1+ #115
> Hardware name: Google Google Compute Engine/Google Compute Engine,
> BIOS Google 01/01/2011
> Call Trace:
>  __dump_stack lib/dump_stack.c:17 [inline]
>  dump_stack+0x194/0x257 lib/dump_stack.c:53
>  print_address_description+0x73/0x250 mm/kasan/report.c:252
>  kasan_report_error mm/kasan/report.c:351 [inline]
>  kasan_report+0x25b/0x340 mm/kasan/report.c:409
>  __asan_report_load8_noabort+0x14/0x20 mm/kasan/report.c:430
>  __lock_acquire+0x465e/0x47f0 kernel/locking/lockdep.c:3378
>  lock_acquire+0x1d5/0x580 kernel/locking/lockdep.c:4004
>  __raw_spin_lock_irqsave include/linux/spinlock_api_smp.h:110 [inline]
>  _raw_spin_lock_irqsave+0x96/0xc0 kernel/locking/spinlock.c:159
>  remove_wait_queue+0x81/0x350 kernel/sched/wait.c:50
>  ep_remove_wait_queue fs/eventpoll.c:595 [inline]
>  ep_unregister_pollwait.isra.7+0x18c/0x590 fs/eventpoll.c:613
>  ep_free+0x13f/0x320 fs/eventpoll.c:830
>  ep_eventpoll_release+0x44/0x60 fs/eventpoll.c:862
>  __fput+0x333/0x7f0 fs/file_table.c:210
>  ____fput+0x15/0x20 fs/file_table.c:244
>  task_work_run+0x199/0x270 kernel/task_work.c:113
>  exit_task_work include/linux/task_work.h:22 [inline]
>  do_exit+0x9bb/0x1ae0 kernel/exit.c:865
>  do_group_exit+0x149/0x400 kernel/exit.c:968
>  SYSC_exit_group kernel/exit.c:979 [inline]
>  SyS_exit_group+0x1d/0x20 kernel/exit.c:977
>  do_syscall_32_irqs_on arch/x86/entry/common.c:327 [inline]
>  do_fast_syscall_32+0x3ee/0xf9d arch/x86/entry/common.c:389
>  entry_SYSENTER_compat+0x51/0x60 arch/x86/entry/entry_64_compat.S:125
> RIP: 0023:0xf7f97c79
> RSP: 002b:00000000ffcb51bc EFLAGS: 00000296 ORIG_RAX: 00000000000000fc
> RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00000000080f0298
> RDX: 0000000000000000 RSI: 00000000080d9b18 RDI: 00000000080f02a0
> RBP: 0000000000000001 R08: 0000000000000000 R09: 0000000000000000
> R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
> R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
> 
> Allocated by task 3086:
>  save_stack+0x43/0xd0 mm/kasan/kasan.c:447
>  set_track mm/kasan/kasan.c:459 [inline]
>  kasan_kmalloc+0xad/0xe0 mm/kasan/kasan.c:551
>  kmem_cache_alloc_trace+0x136/0x750 mm/slab.c:3613
>  kmalloc include/linux/slab.h:499 [inline]
>  kzalloc include/linux/slab.h:688 [inline]
>  binder_get_thread+0x1cf/0x870 drivers/android/binder.c:4184
>  binder_poll+0x8c/0x390 drivers/android/binder.c:4286
>  ep_item_poll.isra.10+0xec/0x320 fs/eventpoll.c:884
>  ep_insert+0x6a3/0x1b10 fs/eventpoll.c:1455
>  SYSC_epoll_ctl fs/eventpoll.c:2106 [inline]
>  SyS_epoll_ctl+0x12e4/0x1ab0 fs/eventpoll.c:1992
>  do_syscall_32_irqs_on arch/x86/entry/common.c:327 [inline]
>  do_fast_syscall_32+0x3ee/0xf9d arch/x86/entry/common.c:389
>  entry_SYSENTER_compat+0x51/0x60 arch/x86/entry/entry_64_compat.S:125
> 
> Freed by task 3086:
>  save_stack+0x43/0xd0 mm/kasan/kasan.c:447
>  set_track mm/kasan/kasan.c:459 [inline]
>  kasan_slab_free+0x71/0xc0 mm/kasan/kasan.c:524
>  __cache_free mm/slab.c:3491 [inline]
>  kfree+0xca/0x250 mm/slab.c:3806
>  binder_free_thread drivers/android/binder.c:4211 [inline]
>  binder_thread_dec_tmpref+0x27f/0x310 drivers/android/binder.c:1808
>  binder_thread_release+0x27d/0x540 drivers/android/binder.c:4275
>  binder_ioctl+0xc05/0x141a drivers/android/binder.c:4492
>  C_SYSC_ioctl fs/compat_ioctl.c:1473 [inline]
>  compat_SyS_ioctl+0x151/0x2a30 fs/compat_ioctl.c:1419
>  do_syscall_32_irqs_on arch/x86/entry/common.c:327 [inline]
>  do_fast_syscall_32+0x3ee/0xf9d arch/x86/entry/common.c:389
>  entry_SYSENTER_compat+0x51/0x60 arch/x86/entry/entry_64_compat.S:125

This is a bug in the binder driver.  binder_poll() tells the poll system to use
the wait queue embedded in the 'struct binder_thread', but then the
BINDER_THREAD_EXIT ioctl will free the 'struct binder_thread' out from under it.

Perhaps either the waitqueue should be associated with the 'struct binder_proc'
rather than the 'struct binder_thread', or binder_poll() should take a reference
to the 'struct binder_thread'.  But I'm not too familiar with binder or epoll,
so I don't know which solution is best.

Here is a program which reproduces the bug, assuming /dev/binder0 exists:

	#include <fcntl.h>
	#include <sys/epoll.h>
	#include <sys/ioctl.h>
	#include <unistd.h>

	#define BINDER_THREAD_EXIT 0x40046208ul

	int main()
	{
		int fd, epfd;
		struct epoll_event event = { .events = EPOLLIN };
		
		fd = open("/dev/binder0", O_RDONLY);
		epfd = epoll_create(1000);
		epoll_ctl(epfd, EPOLL_CTL_ADD, fd, &event);
		ioctl(fd, BINDER_THREAD_EXIT, NULL);
	}