]> git.baikalelectronics.ru Git - kernel.git/commit
binder: validate alloc->mm in ->mmap() handler
authorCarlos Llamas <cmllamas@google.com>
Fri, 4 Nov 2022 23:12:35 +0000 (23:12 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 9 Nov 2022 14:41:27 +0000 (15:41 +0100)
commita68acf5d5f33fdd4814a493fec9119bcc52d8a23
treecdf8653469264589c5b4f51350c0e8555ab14f33
parent73c1db414d657b9aaae0b72a573a07954178bb83
binder: validate alloc->mm in ->mmap() handler

Since commit a72a2f42faac ("binder: fix alloc->vma_vm_mm null-ptr
dereference") binder caches a pointer to the current->mm during open().
This fixes a null-ptr dereference reported by syzkaller. Unfortunately,
it also opens the door for a process to update its mm after the open(),
(e.g. via execve) making the cached alloc->mm pointer invalid.

Things get worse when the process continues to mmap() a vma. From this
point forward, binder will attempt to find this vma using an obsolete
alloc->mm reference. Such as in binder_update_page_range(), where the
wrong vma is obtained via vma_lookup(), yet binder proceeds to happily
insert new pages into it.

To avoid this issue fail the ->mmap() callback if we detect a mismatch
between the vma->vm_mm and the original alloc->mm pointer. This prevents
alloc->vm_addr from getting set, so that any subsequent vma_lookup()
calls fail as expected.

Fixes: a72a2f42faac ("binder: fix alloc->vma_vm_mm null-ptr dereference")
Reported-by: Jann Horn <jannh@google.com>
Cc: <stable@vger.kernel.org> # 5.15+
Signed-off-by: Carlos Llamas <cmllamas@google.com>
Acked-by: Todd Kjos <tkjos@google.com>
Link: https://lore.kernel.org/r/20221104231235.348958-1-cmllamas@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/android/binder_alloc.c