[PATCH v2 03/11] asm-generic/mmiowb: Mark accesses to fix KCSAN warnings

Rohan McLure rmclure at linux.ibm.com
Tue May 23 10:36:02 AEST 2023


On 23 May 2023, at 10:28 am, Rohan McLure <rmclure at linux.ibm.com> wrote:
> 
> On Wed May 10, 2023 at 1:31 PM AEST, Rohan McLure wrote:
>> Prior to this patch, data races are detectable by KCSAN of the following
>> forms:
>> 
>> [1] Asynchronous calls to mmiowb_set_pending() from an interrupt context
>>    or otherwise outside of a critical section
>> [2] Interrupted critical sections, where the interrupt will itself
>>    acquire a lock
>> 
>> In case [1], calling context does not need an mmiowb() call to be
>> issued, otherwise it would do so itself. Such calls to
>> mmiowb_set_pending() are either idempotent or no-ops.
>> 
>> In case [2], irrespective of when the interrupt occurs, the interrupt
>> will acquire and release its locks prior to its return, nesting_count
>> will continue balanced. In the worst case, the interrupted critical
>> section during a mmiowb_spin_unlock() call observes an mmiowb to be
>> pending and afterward is interrupted, leading to an extraneous call to
>> mmiowb(). This data race is clearly innocuous.
>> 
>> Mark all potentially asynchronous memory accesses with READ_ONCE or
>> WRITE_ONCE, including increments and decrements to nesting_count. This
>> has the effect of removing KCSAN warnings at consumer's callsites.
>> 
>> Signed-off-by: Rohan McLure <rmclure at linux.ibm.com>
>> Reported-by: Michael Ellerman <mpe at ellerman.id.au>
>> Reported-by: Gautam Menghani <gautam at linux.ibm.com>
>> Tested-by: Gautam Menghani <gautam at linux.ibm.com>
>> Acked-by: Arnd Bergmann <arnd at arndb.de>
>> ---
>> v2: Remove extraneous READ_ONCE in mmiowb_set_pending for nesting_count
>> ---
>> include/asm-generic/mmiowb.h | 14 +++++++++-----
>> 1 file changed, 9 insertions(+), 5 deletions(-)
>> 
>> diff --git a/include/asm-generic/mmiowb.h b/include/asm-generic/mmiowb.h
>> index 5698fca3bf56..6dea28c8835b 100644
>> --- a/include/asm-generic/mmiowb.h
>> +++ b/include/asm-generic/mmiowb.h
>> @@ -37,25 +37,29 @@ static inline void mmiowb_set_pending(void)
>> struct mmiowb_state *ms = __mmiowb_state();
>> 
>> if (likely(ms->nesting_count))
>> - ms->mmiowb_pending = ms->nesting_count;
>> + WRITE_ONCE(ms->mmiowb_pending, ms->nesting_count);
>> }
>> 
>> static inline void mmiowb_spin_lock(void)
>> {
>> struct mmiowb_state *ms = __mmiowb_state();
>> - ms->nesting_count++;
>> +
>> + /* Increment need not be atomic. Nestedness is balanced over interrupts. */
>> + WRITE_ONCE(ms->nesting_count, READ_ONCE(ms->nesting_count) + 1);
>> }
>> 
>> static inline void mmiowb_spin_unlock(void)
>> {
>> struct mmiowb_state *ms = __mmiowb_state();
>> + u16 pending = READ_ONCE(ms->mmiowb_pending);
>> 
>> - if (unlikely(ms->mmiowb_pending)) {
>> - ms->mmiowb_pending = 0;
>> + WRITE_ONCE(ms->mmiowb_pending, 0);
>> + if (unlikely(pending)) {
>> mmiowb();
>> }
>> 
>> - ms->nesting_count--;
>> + /* Decrement need not be atomic. Nestedness is balanced over interrupts. */
>> + WRITE_ONCE(ms->nesting_count, READ_ONCE(ms->nesting_count) - 1);
> 
> Still think the nesting_counts don't need WRITE_ONCE/READ_ONCE.
> data_race() maybe but I don't know if it's even classed as a data
> race. How does KCSAN handle/annotate preempt_count, for example?

Wow sorry my mail client has some unhelpful keybindings - I don’t know why it
thought I’d want to resend your last item!

Yeah I agree, we don’t need the compiler guarantees of WRITE_ONCE/READ_ONCE, and
yet it’s also not a real data-race. I think I’ll apply data_race() and comment as
I’m still seeing KCSAN warnings here.

Just from inspection, it appears as if __preempt_count_{add,sub} are unmarked and
so likely to generate KCSAN warnings also, but also asm-generic/preempt.h I think
hasn’t been updated to address any such warnings.

> 
> Thanks,
> Nick




More information about the Linuxppc-dev mailing list