Re: [PATCH] mm: Fix mmap_assert_locked() in follow_pte()

[Sergey Senozhatsky]

On (24/07/11 23:33), David Hildenbrand wrote:
[..]
> > @@ -1815,9 +1815,16 @@ static void unmap_single_vma(struct mmu_gather *tlb,
> >   	if (vma->vm_file)
> >   		uprobe_munmap(vma, start, end);
> > -	if (unlikely(vma->vm_flags & VM_PFNMAP))
> > +	if (unlikely(vma->vm_flags & VM_PFNMAP)) {
> > +		if (!mm_wr_locked)
> > +			mmap_read_lock(vma->vm_mm);
> > +
> >   		untrack_pfn(vma, 0, 0, mm_wr_locked);
> > +		if (!mm_wr_locked)
> > +			mmap_read_unlock(vma->vm_mm);
> > +	}
> > +
> >   	if (start != end) {
> >   		if (unlikely(is_vm_hugetlb_page(vma))) {
>
> I'm not sure if this is the right fix. I like to understand how we end up
> without the mmap lock at least in read mode in that path?

I suspect this is causing a deadlock:

[   10.263161] ============================================
[   10.263165] WARNING: possible recursive locking detected
[   10.263170] 6.10.0-rc7-next-20240712+ #645 Tainted: G                 N
[   10.263177] --------------------------------------------
[   10.263179] (direxec)/166 is trying to acquire lock:
[   10.263184] ffff88810b4f0198 (&mm->mmap_lock){++++}-{3:3}, at: mmap_read_lock+0x12/0x40
[   10.263217]
[   10.263217] but task is already holding lock:
[   10.263219] ffff88810b4f0198 (&mm->mmap_lock){++++}-{3:3}, at: exit_mmap+0x9c/0x830
[   10.263238]
[   10.263238] other info that might help us debug this:
[   10.263241]  Possible unsafe locking scenario:
[   10.263241]
[   10.263243]        CPU0
[   10.263245]        ----
[   10.263247]   lock(&mm->mmap_lock);
[   10.263252]   lock(&mm->mmap_lock);
[   10.263257]
[   10.263257]  *** DEADLOCK ***
[   10.263257]
[   10.263259]  May be due to missing lock nesting notation
[   10.263259]
[   10.263262] 3 locks held by (direxec)/166:
[   10.263267]  #0: ffff88810b4e8548 (&sig->cred_guard_mutex){+.+.}-{3:3}, at: bprm_execve+0x70/0x1110
[   10.263286]  #1: ffff88810b4e85e0 (&sig->exec_update_lock){+.+.}-{3:3}, at: exec_mmap+0x9f/0x510
[   10.263302]  #2: ffff88810b4f0198 (&mm->mmap_lock){++++}-{3:3}, at: exit_mmap+0x9c/0x830
[   10.263318]
[   10.263318] stack backtrace:
[   10.263329] CPU: 6 UID: 0 PID: 166 Comm: (direxec) Tainted: G                 N 6.10.0-rc7-next-20240712+ #645
[   10.263340] Tainted: [N]=TEST
[   10.263349] Call Trace:
[   10.263355]  <TASK>
[   10.263360]  dump_stack_lvl+0xa3/0xeb
[   10.263375]  print_deadlock_bug+0x4d5/0x680
[   10.263387]  __lock_acquire+0x65fb/0x7830
[   10.263408]  ? lock_is_held_type+0xdd/0x150
[   10.263425]  lock_acquire+0x14c/0x3e0
[   10.263433]  ? mmap_read_lock+0x12/0x40
[   10.263445]  ? lock_is_held_type+0xdd/0x150
[   10.263454]  down_read+0x58/0x9a0
[   10.263461]  ? mmap_read_lock+0x12/0x40
[   10.263476]  mmap_read_lock+0x12/0x40
[   10.263485]  unmap_single_vma+0x1bf/0x240
[   10.263497]  unmap_vmas+0x146/0x1c0
[   10.263511]  exit_mmap+0x13d/0x830
[   10.263533]  __mmput+0xc2/0x2c0
[   10.263556]  exec_mmap+0x4cb/0x510
[   10.263580]  begin_new_exec+0xfe6/0x1ba0
[   10.263612]  load_elf_binary+0x797/0x22a0
[   10.263637]  ? load_misc_binary+0x53a/0x930
[   10.263656]  ? lock_release+0x50f/0x830
[   10.263673]  ? bprm_execve+0x6d7/0x1110
[   10.263693]  bprm_execve+0x70d/0x1110
[   10.263730]  do_execveat_common+0x44b/0x600
[   10.263745]  __x64_sys_execve+0x8e/0xa0
[   10.263754]  do_syscall_64+0x71/0x110
[   10.263764]  entry_SYSCALL_64_after_hwframe+0x4b/0x53