Re: [PATCH] sched: fix migration to invalid cpu in __set_cpus_allowed_ptr

From: Valentin Schneider
Date: Thu Sep 12 2019 - 18:10:01 EST


On 12/09/2019 02:55, shikemeng wrote:
> From 089dbf0216628ac6ae98742ab90725ca9c2bf201 Mon Sep 17 00:00:00 2001
> From: <shikemeng@xxxxxxxxxx>
> Date: Tue, 10 Sep 2019 09:44:58 -0400
> Subject: [PATCH] sched: fix migration to invalid cpu in __set_cpus_allowed_ptr
>
> reason: migration to invalid cpu in __set_cpus_allowed_ptr
> archive path: patches/euleros/sched
>

The above should probably be trimmed from the log.

> Oops occur when running qemu on arm64:
> Unable to handle kernel paging request at virtual address ffff000008effe40
> Internal error: Oops: 96000007 [#1] SMP
> Process migration/0 (pid: 12, stack limit = 0x00000000084e3736)
> pstate: 20000085 (nzCv daIf -PAN -UAO)
> pc : __ll_sc___cmpxchg_case_acq_4+0x4/0x20
> lr : move_queued_task.isra.21+0x124/0x298
> ...
> Call trace:
> __ll_sc___cmpxchg_case_acq_4+0x4/0x20
> __migrate_task+0xc8/0xe0
> migration_cpu_stop+0x170/0x180
> cpu_stopper_thread+0xec/0x178
> smpboot_thread_fn+0x1ac/0x1e8
> kthread+0x134/0x138
> ret_from_fork+0x10/0x18
>
> __set_cpus_allowed_ptr will choose an active dest_cpu in affinity mask to migrage the process if process is not
> currently running on any one of the CPUs specified in affinity mask.__set_cpus_allowed_ptr will choose an invalid
> dest_cpu(>= nr_cpu_ids, 1024 in my virtual machine) if CPUS in affinity mask are deactived by cpu_down after
> cpumask_intersects check.

Right, cpumask_any_and() returns >= nr_cpu_ids when there isn't any valid CPU
bit set.

> Cpumask_test_cpu of dest_cpu afterwards is overflow and may passes if corresponding bit
> is coincidentally set.

Ouch. I was going to say the cpu_active_mask check from is_cpu_allowed()
should've stopped the whole thing there, but AFAICT there's no safeguard
against > nr_cpu_ids bit accesses. I see CONFIG_DEBUG_PER_CPU_MAPS should
fire a warning for such accesses, but we don't enable it by default.

Would it make sense to do something like

return test_bit(...) && cpu < nr_cpu_ids;

for cpumask_test_cpu()? We still get the warn with the right config, but we
prevent sneaky mistakes like this one. And it seems it's not the only one
according to:

--
virtual patch
virtual report

@nocheck@
expression E;
identifier any_func =~ "^cpumask_any";
position pos;
@@

E@pos = any_func(...);
... when != E >= nr_cpu_ids
when != E < nr_cpu_ids

@script:python depends on nocheck && report@
p << nocheck.pos;
@@
coccilib.report.print_report(p[0], "Missing cpumask_any_*() return value check!")
---

Some of those seem benign since they are on e.g. cpu_online_mask, some other
are somewhat dubious (e.g. deadline.c::dl_task_can_attach()).

As a consequence, kernel will access a invalid rq address associate with the invalid cpu in
> migration_cpu_stop->__migrate_task->move_queued_task and the Oops occurs. Process as follows may trigger the Oops:
> 1) A process repeatedly bind itself to cpu0 and cpu1 in turn by calling sched_setaffinity
> 2) A shell script repeatedly "echo 0 > /sys/devices/system/cpu/cpu1/online" and "echo 1 > /sys/devices/system/cpu/cpu1/online" in turn
> 3) Oops appears if the invalid cpu is set in memory after tested cpumask.
>
>
> Change-Id: I9c2f95aecd3da568991b7408397215f26c990e40

- This doesn't respect the 75 char per line limit
- Change IDs don't belong here (we're not using Gerrit)
- You're missing a Signed-off-by.

You'll find all the guidelines you need for formatting patches in
Documentation/process/submitting-patches.rst.

> ---
> kernel/sched/core.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 4b63fef..5181ea9 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -1112,7 +1112,8 @@ static int __set_cpus_allowed_ptr(struct task_struct *p,
> if (cpumask_equal(&p->cpus_allowed, new_mask))
> goto out;
>
> - if (!cpumask_intersects(new_mask, cpu_valid_mask)) {
> + dest_cpu = cpumask_any_and(cpu_valid_mask, new_mask);
> + if (dest_cpu >= nr_cpu_ids) {
> ret = -EINVAL;
> goto out;
> }
> @@ -1133,7 +1134,6 @@ static int __set_cpus_allowed_ptr(struct task_struct *p,
> if (cpumask_test_cpu(task_cpu(p), new_mask))
> goto out;
>
> - dest_cpu = cpumask_any_and(cpu_valid_mask, new_mask);> if (task_running(rq, p) || p->state == TASK_WAKING) {
> struct migration_arg arg = { p, dest_cpu };
> /* Need help from migration thread: drop lock and wait. */
> --
> 1.8.5.6
>