Re: 94d3a25408: kernel_BUG_at_kernel/fork.c

From: Mark Rutland
Date: Mon Mar 12 2018 - 09:46:33 EST


On Sat, Mar 10, 2018 at 05:11:39AM +0800, kernel test robot wrote:
> FYI, we noticed the following commit (built with gcc-7):
>
> commit: 94d3a254089a7cd4f11b7071b4323afd98eea0a6 ("Detect early free of a live mm")
> url: https://github.com/0day-ci/linux/commits/Mark-Rutland/Detect-early-free-of-a-live-mm/20180303-144149

> [ 47.208935] kernel BUG at kernel/fork.c:599!
> [ 47.210365] invalid opcode: 0000 [#1] SMP PTI
> [ 47.211336] Modules linked in:
> [ 47.212145] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 4.16.0-rc3-00001-g94d3a25 #1
> [ 47.213966] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
> [ 47.215869] RIP: 0010:__mmdrop+0x136/0x170
> [ 47.216866] RSP: 0018:ffffffff82803dd8 EFLAGS: 00010293
> [ 47.218160] RAX: ffffffff82818500 RBX: ffff880115770000 RCX: ffffffff810ae876
> [ 47.219758] RDX: 0000000000000000 RSI: 0000000000000001 RDI: ffff880115770000
> [ 47.221306] RBP: ffffffff82803e00 R08: 0000000000000001 R09: 0000000000000000
> [ 47.223268] R10: 0000000000000000 R11: 0000000000000000 R12: ffffffff82818500
> [ 47.224961] R13: ffffffff82a8ce20 R14: ffff88013ff534c0 R15: 00000000000003e7
> [ 47.226716] FS: 0000000000000000(0000) GS:ffff88013b200000(0000) knlGS:0000000000000000
> [ 47.228550] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 47.229884] CR2: 00007fbfc2cc0190 CR3: 0000000002812000 CR4: 00000000000006f0
> [ 47.231580] Call Trace:
> [ 47.232144] idle_task_exit+0x53/0x60

Luckily this is a spurious warning. In idle_task_exit(), we switch to
the init_mm, but leave active_mm stale before calling mmdrop().

In addition to the WARN_ON[_ONCE] changes, I'll drop the following in:

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index e7c535eee0a6..0ef844abc2da 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -5506,6 +5506,7 @@ void idle_task_exit(void)

if (mm != &init_mm) {
switch_mm(mm, &init_mm, current);
+ current->active_mm = &init_mm;
finish_arch_post_lock_switch();
}
mmdrop(mm);

Thanks,
Mark.