Re: [PATCH 1/2] mm: fix missing handler for __GFP_NOWARN

[Qi Zheng <zhengqi.arch@bytedance.com>]

On 2022/5/11 1:12 PM, Qi Zheng wrote:
> 
> 
> On 2022/5/10 7:38 PM, Qi Zheng wrote:
>> We expect no warnings to be issued when we specify __GFP_NOWARN, but
>> currently in paths like alloc_pages() and kmalloc(), there are still
>> some warnings printed, fix it.
> 
> Hi Andrew,
> 
> Maybe we only need to deal with memory allocation failures, such as
> should_fail() (This leads to deadlock in PATCH[2/2]). These 
> WARN_ON_ONCE()s report usage problems and should not be printed. If they

such as WARN_ON_ONCE((gfp_flags & __GFP_NOFAIL) && (order > 1),
gfp_flags).

> are printed, we should fix these usage problems.
> 
> Thanks,
> Qi
> 
> 
>>
>> Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com>
>> ---
>>   include/linux/fault-inject.h |  2 ++
>>   lib/fault-inject.c           |  3 +++
>>   mm/failslab.c                |  3 +++
>>   mm/internal.h                | 11 +++++++++++
>>   mm/page_alloc.c              | 22 ++++++++++++----------
>>   5 files changed, 31 insertions(+), 10 deletions(-)
>>
>> diff --git a/include/linux/fault-inject.h b/include/linux/fault-inject.h
>> index 2d04f6448cde..9f6e25467844 100644
>> --- a/include/linux/fault-inject.h
>> +++ b/include/linux/fault-inject.h
>> @@ -20,6 +20,7 @@ struct fault_attr {
>>       atomic_t space;
>>       unsigned long verbose;
>>       bool task_filter;
>> +    bool no_warn;
>>       unsigned long stacktrace_depth;
>>       unsigned long require_start;
>>       unsigned long require_end;
>> @@ -39,6 +40,7 @@ struct fault_attr {
>>           .ratelimit_state = RATELIMIT_STATE_INIT_DISABLED,    \
>>           .verbose = 2,                        \
>>           .dname = NULL,                        \
>> +        .no_warn = false,                    \
>>       }
>>   #define DECLARE_FAULT_ATTR(name) struct fault_attr name = 
>> FAULT_ATTR_INITIALIZER
>> diff --git a/lib/fault-inject.c b/lib/fault-inject.c
>> index ce12621b4275..423784d9c058 100644
>> --- a/lib/fault-inject.c
>> +++ b/lib/fault-inject.c
>> @@ -41,6 +41,9 @@ EXPORT_SYMBOL_GPL(setup_fault_attr);
>>   static void fail_dump(struct fault_attr *attr)
>>   {
>> +    if (attr->no_warn)
>> +        return;
>> +
>>       if (attr->verbose > 0 && __ratelimit(&attr->ratelimit_state)) {
>>           printk(KERN_NOTICE "FAULT_INJECTION: forcing a failure.\n"
>>                  "name %pd, interval %lu, probability %lu, "
>> diff --git a/mm/failslab.c b/mm/failslab.c
>> index f92fed91ac23..58df9789f1d2 100644
>> --- a/mm/failslab.c
>> +++ b/mm/failslab.c
>> @@ -30,6 +30,9 @@ bool __should_failslab(struct kmem_cache *s, gfp_t 
>> gfpflags)
>>       if (failslab.cache_filter && !(s->flags & SLAB_FAILSLAB))
>>           return false;
>> +    if (gfpflags & __GFP_NOWARN)
>> +        failslab.attr.no_warn = true;
>> +
>>       return should_fail(&failslab.attr, s->object_size);
>>   }
>> diff --git a/mm/internal.h b/mm/internal.h
>> index cf16280ce132..7a268fac6559 100644
>> --- a/mm/internal.h
>> +++ b/mm/internal.h
>> @@ -35,6 +35,17 @@ struct folio_batch;
>>   /* Do not use these with a slab allocator */
>>   #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
>> +#define WARN_ON_ONCE_GFP(cond, gfp)    ({                \
>> +    static bool __section(".data.once") __warned;            \
>> +    int __ret_warn_once = !!(cond);                    \
>> +                                    \
>> +    if (unlikely(!(gfp & __GFP_NOWARN) && __ret_warn_once && 
>> !__warned)) { \
>> +        __warned = true;                    \
>> +        WARN_ON(1);                        \
>> +    }                                \
>> +    unlikely(__ret_warn_once);                    \
>> +})
>> +
>>   void page_writeback_init(void);
>>   static inline void *folio_raw_mapping(struct folio *folio)
>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>> index 0e42038382c1..2bf4ce4d0e2f 100644
>> --- a/mm/page_alloc.c
>> +++ b/mm/page_alloc.c
>> @@ -3722,7 +3722,7 @@ struct page *rmqueue(struct zone *preferred_zone,
>>        * We most definitely don't want callers attempting to
>>        * allocate greater than order-1 page units with __GFP_NOFAIL.
>>        */
>> -    WARN_ON_ONCE((gfp_flags & __GFP_NOFAIL) && (order > 1));
>> +    WARN_ON_ONCE_GFP((gfp_flags & __GFP_NOFAIL) && (order > 1), 
>> gfp_flags);
>>       do {
>>           page = NULL;
>> @@ -3799,6 +3799,9 @@ static bool __should_fail_alloc_page(gfp_t 
>> gfp_mask, unsigned int order)
>>               (gfp_mask & __GFP_DIRECT_RECLAIM))
>>           return false;
>> +    if (gfp_mask & __GFP_NOWARN)
>> +        fail_page_alloc.attr.no_warn = true;
>> +
>>       return should_fail(&fail_page_alloc.attr, 1 << order);
>>   }
>> @@ -4346,7 +4349,8 @@ __alloc_pages_may_oom(gfp_t gfp_mask, unsigned 
>> int order,
>>        */
>>       /* Exhausted what can be done so it's blame time */
>> -    if (out_of_memory(&oc) || WARN_ON_ONCE(gfp_mask & __GFP_NOFAIL)) {
>> +    if (out_of_memory(&oc) ||
>> +        WARN_ON_ONCE_GFP(gfp_mask & __GFP_NOFAIL, gfp_mask)) {
>>           *did_some_progress = 1;
>>           /*
>> @@ -4902,8 +4906,8 @@ __alloc_pages_slowpath(gfp_t gfp_mask, unsigned 
>> int order,
>>        * We also sanity check to catch abuse of atomic reserves being 
>> used by
>>        * callers that are not in atomic context.
>>        */
>> -    if (WARN_ON_ONCE((gfp_mask & (__GFP_ATOMIC|__GFP_DIRECT_RECLAIM)) ==
>> -                (__GFP_ATOMIC|__GFP_DIRECT_RECLAIM)))
>> +    if (WARN_ON_ONCE_GFP((gfp_mask & 
>> (__GFP_ATOMIC|__GFP_DIRECT_RECLAIM)) ==
>> +                (__GFP_ATOMIC|__GFP_DIRECT_RECLAIM), gfp_mask))
>>           gfp_mask &= ~__GFP_ATOMIC;
>>   retry_cpuset:
>> @@ -5117,7 +5121,7 @@ __alloc_pages_slowpath(gfp_t gfp_mask, unsigned 
>> int order,
>>            * All existing users of the __GFP_NOFAIL are blockable, so 
>> warn
>>            * of any new users that actually require GFP_NOWAIT
>>            */
>> -        if (WARN_ON_ONCE(!can_direct_reclaim))
>> +        if (WARN_ON_ONCE_GFP(!can_direct_reclaim, gfp_mask))
>>               goto fail;
>>           /*
>> @@ -5125,7 +5129,7 @@ __alloc_pages_slowpath(gfp_t gfp_mask, unsigned 
>> int order,
>>            * because we cannot reclaim anything and only can loop waiting
>>            * for somebody to do a work for us
>>            */
>> -        WARN_ON_ONCE(current->flags & PF_MEMALLOC);
>> +        WARN_ON_ONCE_GFP(current->flags & PF_MEMALLOC, gfp_mask);
>>           /*
>>            * non failing costly orders are a hard requirement which we
>> @@ -5133,7 +5137,7 @@ __alloc_pages_slowpath(gfp_t gfp_mask, unsigned 
>> int order,
>>            * so that we can identify them and convert them to something
>>            * else.
>>            */
>> -        WARN_ON_ONCE(order > PAGE_ALLOC_COSTLY_ORDER);
>> +        WARN_ON_ONCE_GFP(order > PAGE_ALLOC_COSTLY_ORDER, gfp_mask);
>>           /*
>>            * Help non-failing allocations by giving them access to memory
>> @@ -5379,10 +5383,8 @@ struct page *__alloc_pages(gfp_t gfp, unsigned 
>> int order, int preferred_nid,
>>        * There are several places where we assume that the order value 
>> is sane
>>        * so bail out early if the request is out of bound.
>>        */
>> -    if (unlikely(order >= MAX_ORDER)) {
>> -        WARN_ON_ONCE(!(gfp & __GFP_NOWARN));
>> +    if (WARN_ON_ONCE_GFP(order >= MAX_ORDER, gfp))
>>           return NULL;
>> -    }
>>       gfp &= gfp_allowed_mask;
>>       /*
> 

-- 
Thanks,
Qi