Re: [PATCH 2/7] writeback: switch to per-bdi threads for flushing data

Re: [PATCH 2/7] writeback: switch to per-bdi threads for flushing data

[Christoph Hellwig]

On Thu, Mar 12, 2009 at 03:33:43PM +0100, Jens Axboe wrote:
> +static void bdi_kupdated(struct backing_dev_info *bdi)
> +{
> +	long nr_to_write;
> +	struct writeback_control wbc = {
> +		.bdi		= bdi,
> +		.sync_mode	= WB_SYNC_NONE,
> +		.nr_to_write	= 0,
> +		.for_kupdate	= 1,
> +		.range_cyclic	= 1,
> +	};
> +
> +	sync_supers();

Not directly related to your patch, but can someone explain WTF
sync_supers is doing here or in the old kupdated?  We're writing back
dirty pages from the VM, and for some reason we try to also write back
superblocks.   This doesn't really make any sense.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 2/7] writeback: switch to per-bdi threads for flushing data

[Chris Mason]

On Mon, 2009-03-16 at 06:22 -0400, Christoph Hellwig wrote:
> On Thu, Mar 12, 2009 at 03:33:43PM +0100, Jens Axboe wrote:
> > +static void bdi_kupdated(struct backing_dev_info *bdi)
> > +{
> > +	long nr_to_write;
> > +	struct writeback_control wbc = {
> > +		.bdi		= bdi,
> > +		.sync_mode	= WB_SYNC_NONE,
> > +		.nr_to_write	= 0,
> > +		.for_kupdate	= 1,
> > +		.range_cyclic	= 1,
> > +	};
> > +
> > +	sync_supers();
> 
> Not directly related to your patch, but can someone explain WTF
> sync_supers is doing here or in the old kupdated?  We're writing back
> dirty pages from the VM, and for some reason we try to also write back
> superblocks.   This doesn't really make any sense.

Some of our poor filesystem cousins don't write the super until kupdate
kicks them (see ext2_write_super).  kupdate has always been the periodic
FS thread of last resort.

-chris




--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 2/7] writeback: switch to per-bdi threads for flushing data

[Christoph Hellwig]

On Mon, Mar 16, 2009 at 09:30:14AM -0400, Chris Mason wrote:
> Some of our poor filesystem cousins don't write the super until kupdate
> kicks them (see ext2_write_super).  kupdate has always been the periodic
> FS thread of last resort.

Yikes, looks like this is indeed the only peridocial sb update for many
simpler filesystems.  We should really have a separate thread for that
instead of hacking it into VM writeback.  Especially with the per-bdi
one where the current setup doesn't make any sense.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>