A deadlock [1] occurs in mfill_get_vma() because the locks mmap_lock
and map_changing_lock are not released; the failure to release them
properly stems from the assignment of the vma variable occurring at
an inappropriate stage.

Moving the vma assignment operation within mfill_get_vma() to after
the vma has been got.

[1]
WARNING: possible circular locking dependency detected
syzkaller #0 Not tainted
------------------------------------------------------
syz.0.17/5990 is trying to acquire lock:
ffff88802caef3b8 (&mm->mmap_lock){++++}-{4:4}, at: __might_fault+0xaf/0x130 mm/memory.c:7249

but task is already holding lock:
ffff88807cdbccf0 (&ctx->map_changing_lock){.+.+}-{4:4}, at: mfill_get_vma+0x162/0x660 mm/userfaultfd.c:226

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:
-> #2 (&ctx->map_changing_lock){.+.+}-{4:4}:
       down_read+0x47/0x2e0 kernel/locking/rwsem.c:1568
       mfill_get_vma+0x162/0x660 mm/userfaultfd.c:226
       mfill_atomic mm/userfaultfd.c:900 [inline]
       mfill_atomic_continue+0x189/0x12c0 mm/userfaultfd.c:974
       userfaultfd_continue fs/userfaultfd.c:1806 [inline]
       userfaultfd_ioctl+0x232d/0x4c70 fs/userfaultfd.c:2071
       vfs_ioctl fs/ioctl.c:51 [inline]
       __do_sys_ioctl fs/ioctl.c:597 [inline]
       __se_sys_ioctl+0xfc/0x170 fs/ioctl.c:583
       do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
       do_syscall_64+0x14d/0xf80 arch/x86/entry/syscall_64.c:94
       entry_SYSCALL_64_after_hwframe+0x77/0x7f

-> #1 (vm_lock){++++}-{0:0}:
       __vma_start_exclude_readers+0x28a/0x940 mm/mmap_lock.c:125
       __vma_start_write+0xdc/0x290 mm/mmap_lock.c:148
       vma_start_write include/linux/mmap_lock.h:303 [inline]
       mprotect_fixup+0x5eb/0xa80 mm/mprotect.c:768
       setup_arg_pages+0x565/0xac0 fs/exec.c:670
       load_elf_binary+0xc5e/0x2980 fs/binfmt_elf.c:1029
       search_binary_handler fs/exec.c:1664 [inline]
       exec_binprm fs/exec.c:1696 [inline]
       bprm_execve+0x949/0x1470 fs/exec.c:1748
       kernel_execve+0x844/0x930 fs/exec.c:1892
       try_to_run_init_process+0x13/0x60 init/main.c:1514
       kernel_init+0xad/0x1d0 init/main.c:1642
       ret_from_fork+0x51e/0xb90 arch/x86/kernel/process.c:158
       ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:245

-> #0 (&mm->mmap_lock){++++}-{4:4}:
       check_prev_add kernel/locking/lockdep.c:3165 [inline]
       check_prevs_add kernel/locking/lockdep.c:3284 [inline]
       validate_chain kernel/locking/lockdep.c:3908 [inline]
       __lock_acquire+0x15a5/0x2cf0 kernel/locking/lockdep.c:5237
       lock_acquire+0xf0/0x2e0 kernel/locking/lockdep.c:5868
       __might_fault+0xcb/0x130 mm/memory.c:7249
       userfaultfd_continue fs/userfaultfd.c:1813 [inline]
       userfaultfd_ioctl+0x2372/0x4c70 fs/userfaultfd.c:2071
       vfs_ioctl fs/ioctl.c:51 [inline]
       __do_sys_ioctl fs/ioctl.c:597 [inline]
       __se_sys_ioctl+0xfc/0x170 fs/ioctl.c:583
       do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
       do_syscall_64+0x14d/0xf80 arch/x86/entry/syscall_64.c:94
       entry_SYSCALL_64_after_hwframe+0x77/0x7f

other info that might help us debug this:

Chain exists of:
  &mm->mmap_lock --> vm_lock --> &ctx->map_changing_lock

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  rlock(&ctx->map_changing_lock);
                               lock(vm_lock);
                               lock(&ctx->map_changing_lock);
  rlock(&mm->mmap_lock);

 *** DEADLOCK ***

Fixes: 7d4d4de3ac3e ("userfaultfd: introduce mfill_get_vma() and mfill_put_vma()")
Reported-by: syzbot+c473aa669b5e8a6f48d2@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=c473aa669b5e8a6f48d2
Tested-by: syzbot+c473aa669b5e8a6f48d2@syzkaller.appspotmail.com
Signed-off-by: Edward Adam Davis <eadavis@qq.com>
---
 mm/userfaultfd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
index 9ffc80d0a51b..a3333d5c6454 100644
--- a/mm/userfaultfd.c
+++ b/mm/userfaultfd.c
@@ -218,6 +218,7 @@ static int mfill_get_vma(struct mfill_state *state)
 	if (IS_ERR(dst_vma))
 		return PTR_ERR(dst_vma);
 
+	state->vma = dst_vma;
 	/*
 	 * If memory mappings are changing because of non-cooperative
 	 * operation (e.g. mremap) running in parallel, bail out and
@@ -257,7 +258,6 @@ static int mfill_get_vma(struct mfill_state *state)
 		goto out_unlock;
 
 out:
-	state->vma = dst_vma;
 	return 0;
 
 out_unlock:
-- 
2.43.0