Re: [PATCH] ext4: Fix deadlock on page reclaim

[Andreas Dilger <adilger@dilger.ca>]

[-- Attachment #1: Type: text/plain, Size: 1185 bytes --]

On Jul 25, 2019, at 5:54 AM, Christoph Hellwig <hch@infradead.org> wrote:
> 
> On Thu, Jul 25, 2019 at 06:33:58PM +0900, Damien Le Moal wrote:
>> +	gfp_t gfp_mask;
>> +
>> 	switch (ext4_inode_journal_mode(inode)) {
>> 	case EXT4_INODE_ORDERED_DATA_MODE:
>> 	case EXT4_INODE_WRITEBACK_DATA_MODE:
>> @@ -4019,6 +4019,14 @@ void ext4_set_aops(struct inode *inode)
>> 		inode->i_mapping->a_ops = &ext4_da_aops;
>> 	else
>> 		inode->i_mapping->a_ops = &ext4_aops;
>> +
>> +	/*
>> +	 * Ensure all page cache allocations are done from GFP_NOFS context to
>> +	 * prevent direct reclaim recursion back into the filesystem and blowing
>> +	 * stacks or deadlocking.
>> +	 */
>> +	gfp_mask = mapping_gfp_mask(inode->i_mapping);
>> +	mapping_set_gfp_mask(inode->i_mapping, (gfp_mask & ~(__GFP_FS)));
> 
> This looks like something that could hit every file systems, so
> shouldn't we fix this in common code?

It also has the drawback that it prevents __GFP_FS reclaim when ext4
is *not* at the bottom of the IO stack.

> We could also look into just using memalloc_nofs_save for the page
> cache allocation path instead of the per-mapping gfp_mask.

That makes more sense.

Cheers, Andreas






[-- Attachment #2: Message signed with OpenPGP --]
[-- Type: application/pgp-signature, Size: 873 bytes --]