On 13/02/2019 14:00, Will Deacon wrote:
> Hi Peter,
>
> On Wed, Feb 13, 2019 at 02:17:20PM +0100, Peter Zijlstra wrote:
>> On Wed, Feb 13, 2019 at 10:50:21AM +0000, Julien Thierry wrote:
>>> On 13/02/2019 10:35, Peter Zijlstra wrote:
>>>> On Tue, Feb 12, 2019 at 09:15:13AM +0000, Julien Thierry wrote:
>>>>
>>>>>>>> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
>>>>>>>> index a674c7db..b1bb7e9 100644
>>>>>>>> --- a/kernel/sched/core.c
>>>>>>>> +++ b/kernel/sched/core.c
>>>>>>>> @@ -3289,6 +3289,14 @@ static inline void schedule_debug(struct
>>>>>>>> task_struct *prev)
>>>>>>>> __schedule_bug(prev);
>>>>>>>> preempt_count_set(PREEMPT_DISABLED);
>>>>>>>> }
>>>>>>>> +
>>>>>>>> + if (IS_ENABLED(CONFIG_DEBUG_UACCESS_SLEEP) &&
>>>>>>>> + unlikely(unsafe_user_region_active())) {
>>>>>>>> + printk(KERN_ERR "BUG: scheduling while user_access
>>>>>>>> enabled: %s/%d/0x%08x\n",
>>>>>>>> + prev->comm, prev->pid, preempt_count());
>>>>>>>> + dump_stack();
>>>>>>>> + }
>>>>>>>> +
>>>>>>>> rcu_sleep_check();
>>>>>>>>
>>>>>>>> profile_hit(SCHED_PROFILING, __builtin_return_address(0));
>>>>
>>>>> I guess I'll drop the might_resched() part of this patch if that sounds
>>>>> alright.
>>>>
>>>> I'm still confused by the schedule_debug() part. How is that not broken?
>>>
>>> Hmmm, I am not exactly sure which part you expect to be broken, I guess
>>> it's because of the nature of the uaccess unsafe accessor usage.
>>>
>>> Basically, the following is a definite no:
>>> if (user_access_begin(ptr, size)) {
>>>
>>> [...]
>>>
>>> //something that calls schedule
>>>
>>> [...]
>>>
>>> user_access_end();
>>> }
>>>
>>>
>>> However the following is fine:
>>>
>>> - user_access_begin(ptr, size)
>>> - taking irq/exception
>>> - get preempted
>>
>> This; how is getting preempted fundamentally different from scheduling
>> ourselves?
>
> The difference is because getting preempted in the sequence above is
> triggered off the back of an interrupt. On arm64, and I think also on x86,
> the user access state (SMAP or PAN) is saved and restored across exceptions
> but not across context switch. Consequently, taking an irq in a
> user_access_{begin,end} section and then scheduling is fine, but calling
> schedule directly within such a section is not.
>
> Julien -- please yell if I've missed some crucial detail, but I think that's
> the gist of what we're trying to describe here.
>
Yes, this summarizes things correctly. Thanks!
I might also stress out that this limitation is already existing for x86
(and is in the arm64 patches picked by Catalin for 5.1), as was
discussed in here:
https://lkml.org/lkml/2018/11/23/430
So this patch is not introducing new semantics, it is only making
existing ones explicit.
If the current state is not good, we need to re-discuss the semantics of
user_access regions.
Thanks,
--
Julien Thierry
