Re: mm: fault in __do_fault

[Kirill A. Shutemov]

On Tue, Feb 28, 2017 at 03:04:53PM +0100, Dmitry Vyukov wrote:
> Hello,
> 
> The following program triggers GPF in __do_fault:
> https://gist.githubusercontent.com/dvyukov/27345737fca18d92ef761e7fa08aec9b/raw/d99d02511d0bf9a8d6f6bd9c79d373a26924e974/gistfile1.txt
> 
> general protection fault: 0000 [#1] SMP KASAN
> Modules linked in:
> CPU: 3 PID: 2955 Comm: a.out Not tainted 4.10.0+ #230
> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
> task: ffff88006b4e2480 task.stack: ffff88006b0b0000
> RIP: 0010:__read_once_size include/linux/compiler.h:247 [inline]
> RIP: 0010:compound_head include/linux/page-flags.h:146 [inline]
> RIP: 0010:trylock_page include/linux/pagemap.h:442 [inline]
> RIP: 0010:lock_page include/linux/pagemap.h:452 [inline]
> RIP: 0010:__do_fault+0x247/0x3e0 mm/memory.c:2898
> RSP: 0000:ffff88006b0b7470 EFLAGS: 00010202
> RAX: dffffc0000000000 RBX: 1ffff1000d616e91 RCX: 0000000000000000
> RDX: 0000000000000004 RSI: dffffc0000000000 RDI: 0000000000000020
> RBP: ffff88006b0b7550 R08: 0000000000000001 R09: 0000000000000001
> R10: 0000000000000000 R11: 0000000000000004 R12: ffff88006b0b74e8
> R13: 0000000000000000 R14: ffff88006b0b7528 R15: 0000000000000000
> FS:  00007fef8bb55700(0000) GS:ffff88006d180000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 0000000020012fe0 CR3: 000000006a8dc000 CR4: 00000000001406e0
> Call Trace:
>  do_read_fault mm/memory.c:3266 [inline]
>  do_fault+0x5ee/0x2080 mm/memory.c:3366
>  handle_pte_fault mm/memory.c:3596 [inline]
>  __handle_mm_fault+0x1062/0x2cb0 mm/memory.c:3710
>  handle_mm_fault+0x1e2/0x480 mm/memory.c:3747
>  __do_page_fault+0x4f6/0xb60 arch/x86/mm/fault.c:1396
>  trace_do_page_fault+0x141/0x6c0 arch/x86/mm/fault.c:1489
>  do_async_page_fault+0x72/0xc0 arch/x86/kernel/kvm.c:264
>  async_page_fault+0x28/0x30 arch/x86/entry/entry_64.S:1014
> RIP: 0033:0x43dda0
> RSP: 002b:00007fef8bb54878 EFLAGS: 00010202
> RAX: 00007fef8bb548b0 RBX: 0000000000000000 RCX: 000000000000000e
> RDX: 0000000000000400 RSI: 0000000020012fe0 RDI: 00007fef8bb548b0
> RBP: 00007fef8bb54cd0 R08: 0000000000000400 R09: 00007fef8bb54d10
> R10: 00007fef8bb559d0 R11: 0000000000000202 R12: 0000000000000000
> R13: 0000000000000000 R14: 00007fef8bb559c0 R15: 00007fef8bb55700
> Code: 00 e8 0e 35 b8 ff e8 79 09 a2 02 4c 89 e2 49 8d 7f 20 48 b8 00
> 00 00 00 00 fc ff df 48 c1 ea 03 c6 04 02 00 48 89 fa 48 c1 ea 03 <80>
> 3c 02 00 0f 85 ec 00 00 00 4c 89 e2 48 b8 00 00 00 00 00 fc
> RIP: __read_once_size include/linux/compiler.h:247 [inline] RSP:
> ffff88006b0b7470
> RIP: compound_head include/linux/page-flags.h:146 [inline] RSP: ffff88006b0b7470
> RIP: trylock_page include/linux/pagemap.h:442 [inline] RSP: ffff88006b0b7470
> RIP: lock_page include/linux/pagemap.h:452 [inline] RSP: ffff88006b0b7470
> RIP: __do_fault+0x247/0x3e0 mm/memory.c:2898 RSP: ffff88006b0b7470

__do_fault() expects to see a page if ->fault() returns zero, but
handle_userfault() mess this up, making shmem_fault() return 0 without
setting vmf->page.

I'm not entirely sure it's fully correct, but the patch below fixes the
crash for me.

Andrea, does it look okay for you?

diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c
index 625b7285a37b..56f61f1a1dc1 100644
--- a/fs/userfaultfd.c
+++ b/fs/userfaultfd.c
@@ -489,7 +489,7 @@ int handle_userfault(struct vm_fault *vmf, unsigned long reason)
                         * in such case.
                         */
                        down_read(&mm->mmap_sem);
-                       ret = 0;
+                       ret = VM_FAULT_NOPAGE;
                }
        }

-- 
 Kirill A. Shutemov