[Top] [All Lists]

Re: [PATCH 1/5] xfs: remove ->write_super and stop maintaining ->s_dirt

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Subject: Re: [PATCH 1/5] xfs: remove ->write_super and stop maintaining ->s_dirt
From: Eric Sandeen <sandeen@xxxxxxxxxxx>
Date: Sun, 10 May 2009 11:30:04 -0500
Cc: xfs@xxxxxxxxxxx
In-reply-to: <4A06FFF3.2090103@xxxxxxxxxxx>
References: <20090426140305.113371000@xxxxxxxxxxxxxxxxxxxxxx> <20090426140707.533370000@xxxxxxxxxxxxxxxxxxxxxx> <4A06FFF3.2090103@xxxxxxxxxxx>
User-agent: Thunderbird (Macintosh/20090302)
Eric Sandeen wrote:
> Christoph Hellwig wrote:
>> the write_super method is used for
>>  (1) writing back the superblock periodically from pdflush
>>  (2) called just before ->sync_fs for data integerity syncs
>>  (3) just before ->put_super
>> We don't need (1) because we have our own peridoc writeout through xfssyncd,
>> we don't need (2) because xfs_fs_sync_fs performs a proper synchronous
>> superblock writeout after all other data and metadata has been written out,
>> and we don't need (3) because we synchronously write the superblock in
>> ->put_super once the filesystem is fully shut down.
>> Also remove ->s_dirt tracking as it's only used to decide when too call
>> ->write_super.
> Just to double check, what about sync_filesystems():
>                 if (sb->s_root && (wait || sb->s_dirt))
>                         sb->s_op->sync_fs(sb, wait);
> if we lose s_dirt does that mean we are potentially doing one less ->sync_fs 
> here when called with wait = 0, and is that ok?  (/me waves hands about
> sync; sync; sync magic) :)
> -Eric

gah, never mind, I forgot that 0/5 talked about Jan's patches, and here:


takes care of this concern.


<Prev in Thread] Current Thread [Next in Thread>