Re: [PATCH v2 1/3] pidfd: extend pidfd_get_pid() and de-duplicate pid lookup

On Wed, Oct 16, 2024 at 04:50:56PM +0800, kernel test robot wrote:

Hello,

kernel test robot noticed "BUG:unable_to_handle_page_fault_for_address" on:

Thanks, see below for analysis.

commit: e65dbb5c9051a4da2305787fd558e1d60de2275a ("[PATCH v2 1/3] pidfd: extend pidfd_get_pid() and de-duplicate pid lookup")
url: https://github.com/intel-lab-lkp/linux/commits/Lorenzo-Stoakes/pidfd-extend-pidfd_get_pid-and-de-duplicate-pid-lookup/20241011-191241
base: https://git.kernel.org/cgit/linux/kernel/git/shuah/linux-kselftest.git next
patch link: https://lore.kernel.org/all/8e7edaf2f648fb01a71def749f17f76c0502dee1.1728643714.git.lorenzo.stoakes@oracle.com/ (local)
patch subject: [PATCH v2 1/3] pidfd: extend pidfd_get_pid() and de-duplicate pid lookup

in testcase: trinity
version: trinity-i386-abe9de86-1_20230429
with following parameters:

	runtime: 600s



config: x86_64-randconfig-072-20241015
compiler: gcc-12
test machine: qemu-system-x86_64 -enable-kvm -cpu SandyBridge -smp 2 -m 16G

(please refer to attached dmesg/kmsg for entire log/backtrace)



If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot [off-list ref]
| Closes: https://lore.kernel.org/oe-lkp/202410161634.abca3854-lkp@intel.com (local)


[  416.054386][ T1959] BUG: unable to handle page fault for address: ffffffff8fed9474
[  416.055651][ T1959] #PF: supervisor write access in kernel mode
[  416.056550][ T1959] #PF: error_code(0x0003) - permissions violation
[  416.057502][ T1959] PGD 3e90f5067 P4D 3e90f5067 PUD 3e90f6063 PMD 3e50001a1
[  416.058587][ T1959] Oops: Oops: 0003 [#1] PREEMPT SMP KASAN
[  416.059414][ T1959] CPU: 1 UID: 65534 PID: 1959 Comm: trinity-c3 Not tainted 6.12.0-rc1-00004-ge65dbb5c9051 #1 d7a38916ac9252f968706afc2c77f70fbdabe689
[  416.061328][ T1959] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.2-debian-1.16.2-1 04/01/2014
[ 416.062850][ T1959] RIP: 0010:fput (arch/x86/include/asm/atomic64_64.h:61 include/linux/atomic/atomic-arch-fallback.h:4404 include/linux/atomic/atomic-long.h:1571 include/linux/atomic/atomic-instrumented.h:4540 fs/file_table.c:482)
[ 416.063578][ T1959] Code: ff ff 66 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00 f3 0f 1e fa 55 48 89 e5 41 55 41 54 53 48 89 fb be 08 00 00 00 e8 96 c6 f7 ff <f0> 48 ff 0b 0f 85 dd 00 00 00 65 4c 8b 25 04 ff 0e 70 4c 8d 6b 48
All code
========
   0:	ff                   	(bad)
   1:	ff 66 66             	jmp    *0x66(%rsi)
   4:	2e 0f 1f 84 00 00 00 	cs nopl 0x0(%rax,%rax,1)
   b:	00 00
   d:	0f 1f 00             	nopl   (%rax)
  10:	f3 0f 1e fa          	endbr64
  14:	55                   	push   %rbp
  15:	48 89 e5             	mov    %rsp,%rbp
  18:	41 55                	push   %r13
  1a:	41 54                	push   %r12
  1c:	53                   	push   %rbx
  1d:	48 89 fb             	mov    %rdi,%rbx
  20:	be 08 00 00 00       	mov    $0x8,%esi
  25:	e8 96 c6 f7 ff       	call   0xfffffffffff7c6c0
  2a:*	f0 48 ff 0b          	lock decq (%rbx)		<-- trapping instruction

OK so this looks like the fput() invoking atomic_long_dec_and_test() on an
invalid &file->f_count.

It looks like 0xffffffff8fed9474 in RBX is the file...

And that's because I'm not setting f in
SYSCALL_DEFINE4(pidfd_send_signal, ...) at:

	pidfd_to_pid_proc(pidfd, &f_flags, &f);

On error and yet then jump to

err:
	fdput(f);
	return ret;

Which is trying to fdput() (thus fput()) the f, ugh.

OK I will fix this + respin, thanks for the report!

[snip]