Re: [PATCH v2 01/12] f2fs: cache: implement metadata cache

[Chao Yu <chao@kernel.org>]

On 8/25/26 13:39, Wenjie Qi wrote:
> Hi Chao,
> 
>> next:
>> 	spin_lock_irqsave(&cache->tree_lock, flags);
>> 	nr = radix_tree_gang_lookup(&cache->root, (void **)entries, index,
>> 		min((unsigned long)F2FS_ONSTACK_CACHES, end - index));
>> 	...
>> 	spin_unlock_irqrestore(&cache->tree_lock, flags);
>> 	...
>> 	if (index < end)
>> 		goto next;
> 
> `nr` bounds each gang lookup, while the outer loop continues until `index`
> reaches `end`.  Dropping `tree_lock` between batches is not itself a
> scheduling point.
> 
> Is there a scheduling point or another bound on the number of entries that
> this range invalidation can process in one invocation?

Wenjie,

I think it's good to have cond_resched() before "goto next", will fix in below
cases:

- f2fs_cache_wait_on_all_writeback
- f2fs_drop_cache_range
- f2fs_invalidate_compress_pages

Thanks,