Patchwork [4/6] block: Optionally snapshot page contents to provide stable pages during write

login
register
mail settings
Submitter Darrick J. Wong
Date Jan. 15, 2013, 5:43 a.m.
Message ID <20130115054305.1563.48197.stgit@blackbox.djwong.org>
Download mbox | patch
Permalink /patch/212001/
State Superseded
Headers show

Comments

Darrick J. Wong - Jan. 15, 2013, 5:43 a.m.
This provides a band-aid to provide stable page writes on jbd without needing
to backport the fixed locking and page writeback bit handling schemes of jbd2.
The band-aid works by using bounce buffers to snapshot page contents instead of
waiting.

For those wondering about the ext3 bandage -- fixing the jbd locking (which was
done as part of ext4dev years ago) is a lot of surgery, and setting
PG_writeback on data pages when we actually hold the page lock dropped ext3
performance by nearly an order of magnitude.  If we're going to migrate iscsi
and raid to use stable page writes, the complaints about high latency will
likely return.  We might as well centralize their page snapshotting thing to
one place.

Tested-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 arch/tile/Kconfig       |    6 ------
 block/blk-core.c        |    8 +++++---
 fs/ext3/super.c         |    1 +
 include/uapi/linux/fs.h |    1 +
 mm/Kconfig              |   13 +++++++++++++
 mm/bounce.c             |   47 +++++++++++++++++++++++++++++++++++++++++++----
 mm/page-writeback.c     |    4 ++++
 7 files changed, 67 insertions(+), 13 deletions(-)



--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Jan Kara - Jan. 16, 2013, 2 a.m.
On Mon 14-01-13 21:43:05, Darrick J. Wong wrote:
> This provides a band-aid to provide stable page writes on jbd without needing
> to backport the fixed locking and page writeback bit handling schemes of jbd2.
> The band-aid works by using bounce buffers to snapshot page contents instead of
> waiting.
> 
> For those wondering about the ext3 bandage -- fixing the jbd locking (which was
> done as part of ext4dev years ago) is a lot of surgery, and setting
> PG_writeback on data pages when we actually hold the page lock dropped ext3
> performance by nearly an order of magnitude.  If we're going to migrate iscsi
> and raid to use stable page writes, the complaints about high latency will
> likely return.  We might as well centralize their page snapshotting thing to
> one place.
> 
> Tested-by: Andy Lutomirski <luto@amacapital.net>
> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> ---
>  arch/tile/Kconfig       |    6 ------
>  block/blk-core.c        |    8 +++++---
>  fs/ext3/super.c         |    1 +
>  include/uapi/linux/fs.h |    1 +
>  mm/Kconfig              |   13 +++++++++++++
>  mm/bounce.c             |   47 +++++++++++++++++++++++++++++++++++++++++++----
>  mm/page-writeback.c     |    4 ++++
>  7 files changed, 67 insertions(+), 13 deletions(-)
> 
> 
> diff --git a/arch/tile/Kconfig b/arch/tile/Kconfig
> index 875d008..c671fda 100644
> --- a/arch/tile/Kconfig
> +++ b/arch/tile/Kconfig
> @@ -410,12 +410,6 @@ config TILE_USB
>  	  Provides USB host adapter support for the built-in EHCI and OHCI
>  	  interfaces on TILE-Gx chips.
>  
> -# USB OHCI needs the bounce pool since tilegx will often have more
> -# than 4GB of memory, but we don't currently use the IOTLB to present
> -# a 32-bit address to OHCI.  So we need to use a bounce pool instead.
> -config NEED_BOUNCE_POOL
> -	def_bool USB_OHCI_HCD
> -
>  source "drivers/pci/hotplug/Kconfig"
>  
>  endmenu
> diff --git a/block/blk-core.c b/block/blk-core.c
> index c973249..277134c 100644
> --- a/block/blk-core.c
> +++ b/block/blk-core.c
> @@ -1474,6 +1474,11 @@ void blk_queue_bio(struct request_queue *q, struct bio *bio)
>  	 */
>  	blk_queue_bounce(q, &bio);
>  
> +	if (bio_integrity_enabled(bio) && bio_integrity_prep(bio)) {
> +		bio_endio(bio, -EIO);
> +		return;
> +	}
> +
>  	if (bio->bi_rw & (REQ_FLUSH | REQ_FUA)) {
>  		spin_lock_irq(q->queue_lock);
>  		where = ELEVATOR_INSERT_FLUSH;
> @@ -1714,9 +1719,6 @@ generic_make_request_checks(struct bio *bio)
>  	 */
>  	blk_partition_remap(bio);
>  
> -	if (bio_integrity_enabled(bio) && bio_integrity_prep(bio))
> -		goto end_io;
> -
  Umm, why did you move this hunk?

>  	if (bio_check_eod(bio, nr_sectors))
>  		goto end_io;

> diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
> index 780d4c6..0144fbb 100644
> --- a/include/uapi/linux/fs.h
> +++ b/include/uapi/linux/fs.h
> @@ -69,6 +69,7 @@ struct inodes_stat_t {
>  #define MS_REMOUNT	32	/* Alter flags of a mounted FS */
>  #define MS_MANDLOCK	64	/* Allow mandatory locks on an FS */
>  #define MS_DIRSYNC	128	/* Directory modifications are synchronous */
> +#define MS_SNAP_STABLE	256	/* Snapshot pages during writeback, if needed */
>  #define MS_NOATIME	1024	/* Do not update access times. */
>  #define MS_NODIRATIME	2048	/* Do not update directory access times */
>  #define MS_BIND		4096
  Please don't mix MS_SNAP_STABLE flag among flags passed by mount(2)
syscall. I think putting it at 1 << 27 might be acceptable. I remember
Al Viro saying something along the lines that kernel internal superblock
flags should be separated from those passed from userspace into a special
superblock entry but that's a different story I guess.

> diff --git a/mm/Kconfig b/mm/Kconfig
> index 278e3ab..7901d83 100644
> --- a/mm/Kconfig
> +++ b/mm/Kconfig
> @@ -258,6 +258,19 @@ config BOUNCE
>  	def_bool y
>  	depends on BLOCK && MMU && (ZONE_DMA || HIGHMEM)
>  
> +# On the 'tile' arch, USB OHCI needs the bounce pool since tilegx will often
> +# have more than 4GB of memory, but we don't currently use the IOTLB to present
> +# a 32-bit address to OHCI.  So we need to use a bounce pool instead.
> +#
> +# We also use the bounce pool to provide stable page writes for jbd.  jbd
> +# initiates buffer writeback without locking the page or setting PG_writeback,
> +# and fixing that behavior (a second time; jbd2 doesn't have this problem) is
> +# a major rework effort.  Instead, use the bounce buffer to snapshot pages
> +# (until jbd goes away).  The only jbd user is ext3.
> +config NEED_BOUNCE_POOL
> +	bool
> +	default y if (TILE && USB_OHCI_HCD) || (BLK_DEV_INTEGRITY && JBD)
> +
>  config NR_QUICK
>  	int
>  	depends on QUICKLIST
> diff --git a/mm/bounce.c b/mm/bounce.c
> index 0420867..a5b30f9 100644
> --- a/mm/bounce.c
> +++ b/mm/bounce.c
> @@ -178,8 +178,44 @@ static void bounce_end_io_read_isa(struct bio *bio, int err)
>  	__bounce_end_io_read(bio, isa_page_pool, err);
>  }
>  
> +#ifdef CONFIG_NEED_BOUNCE_POOL
> +static int must_snapshot_stable_pages(struct bio *bio)
> +{
> +	struct page *page;
> +	struct backing_dev_info *bdi;
> +	struct address_space *mapping;
> +	struct bio_vec *from;
> +	int i;
> +
> +	if (bio_data_dir(bio) != WRITE)
> +		return 0;
> +
> +	/*
> +	 * Based on the first page that has a valid mapping, decide whether or
> +	 * not we have to employ bounce buffering to guarantee stable pages.
> +	 */
> +	bio_for_each_segment(from, bio, i) {
> +		page = from->bv_page;
> +		mapping = page_mapping(page);
> +		if (!mapping)
> +			continue;
> +		bdi = mapping->backing_dev_info;
> +		if (!bdi_cap_stable_pages_required(bdi))
> +			return 0;
> +		return mapping->host->i_sb->s_flags & MS_SNAP_STABLE;
> +	}
> +
> +	return 0;
> +}
  How about using q->backing_dev_info for the
bdi_cap_stable_pages_required() check? It will be a fast path and this check
will be faster..

								Honza
Darrick J. Wong - Jan. 17, 2013, 3:01 a.m.
On Wed, Jan 16, 2013 at 03:00:00AM +0100, Jan Kara wrote:
> On Mon 14-01-13 21:43:05, Darrick J. Wong wrote:
> > This provides a band-aid to provide stable page writes on jbd without needing
> > to backport the fixed locking and page writeback bit handling schemes of jbd2.
> > The band-aid works by using bounce buffers to snapshot page contents instead of
> > waiting.
> > 
> > For those wondering about the ext3 bandage -- fixing the jbd locking (which was
> > done as part of ext4dev years ago) is a lot of surgery, and setting
> > PG_writeback on data pages when we actually hold the page lock dropped ext3
> > performance by nearly an order of magnitude.  If we're going to migrate iscsi
> > and raid to use stable page writes, the complaints about high latency will
> > likely return.  We might as well centralize their page snapshotting thing to
> > one place.
> > 
> > Tested-by: Andy Lutomirski <luto@amacapital.net>
> > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> > ---
> >  arch/tile/Kconfig       |    6 ------
> >  block/blk-core.c        |    8 +++++---
> >  fs/ext3/super.c         |    1 +
> >  include/uapi/linux/fs.h |    1 +
> >  mm/Kconfig              |   13 +++++++++++++
> >  mm/bounce.c             |   47 +++++++++++++++++++++++++++++++++++++++++++----
> >  mm/page-writeback.c     |    4 ++++
> >  7 files changed, 67 insertions(+), 13 deletions(-)
> > 
> > 
> > diff --git a/arch/tile/Kconfig b/arch/tile/Kconfig
> > index 875d008..c671fda 100644
> > --- a/arch/tile/Kconfig
> > +++ b/arch/tile/Kconfig
> > @@ -410,12 +410,6 @@ config TILE_USB
> >  	  Provides USB host adapter support for the built-in EHCI and OHCI
> >  	  interfaces on TILE-Gx chips.
> >  
> > -# USB OHCI needs the bounce pool since tilegx will often have more
> > -# than 4GB of memory, but we don't currently use the IOTLB to present
> > -# a 32-bit address to OHCI.  So we need to use a bounce pool instead.
> > -config NEED_BOUNCE_POOL
> > -	def_bool USB_OHCI_HCD
> > -
> >  source "drivers/pci/hotplug/Kconfig"
> >  
> >  endmenu
> > diff --git a/block/blk-core.c b/block/blk-core.c
> > index c973249..277134c 100644
> > --- a/block/blk-core.c
> > +++ b/block/blk-core.c
> > @@ -1474,6 +1474,11 @@ void blk_queue_bio(struct request_queue *q, struct bio *bio)
> >  	 */
> >  	blk_queue_bounce(q, &bio);
> >  
> > +	if (bio_integrity_enabled(bio) && bio_integrity_prep(bio)) {
> > +		bio_endio(bio, -EIO);
> > +		return;
> > +	}
> > +
> >  	if (bio->bi_rw & (REQ_FLUSH | REQ_FUA)) {
> >  		spin_lock_irq(q->queue_lock);
> >  		where = ELEVATOR_INSERT_FLUSH;
> > @@ -1714,9 +1719,6 @@ generic_make_request_checks(struct bio *bio)
> >  	 */
> >  	blk_partition_remap(bio);
> >  
> > -	if (bio_integrity_enabled(bio) && bio_integrity_prep(bio))
> > -		goto end_io;
> > -
>   Umm, why did you move this hunk?

I moved it so that the integrity data are generated against the contents of the
bounce buffer because the write paths don't wait for writeback to finish if the
snapshotting mode is enabled, which means (I think) that programs can wander in
and scribble on the original page in between bio_integrity_prep and
blk_queue_bounce.

> >  	if (bio_check_eod(bio, nr_sectors))
> >  		goto end_io;
> 
> > diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
> > index 780d4c6..0144fbb 100644
> > --- a/include/uapi/linux/fs.h
> > +++ b/include/uapi/linux/fs.h
> > @@ -69,6 +69,7 @@ struct inodes_stat_t {
> >  #define MS_REMOUNT	32	/* Alter flags of a mounted FS */
> >  #define MS_MANDLOCK	64	/* Allow mandatory locks on an FS */
> >  #define MS_DIRSYNC	128	/* Directory modifications are synchronous */
> > +#define MS_SNAP_STABLE	256	/* Snapshot pages during writeback, if needed */
> >  #define MS_NOATIME	1024	/* Do not update access times. */
> >  #define MS_NODIRATIME	2048	/* Do not update directory access times */
> >  #define MS_BIND		4096
>   Please don't mix MS_SNAP_STABLE flag among flags passed by mount(2)
> syscall. I think putting it at 1 << 27 might be acceptable. I remember
> Al Viro saying something along the lines that kernel internal superblock
> flags should be separated from those passed from userspace into a special
> superblock entry but that's a different story I guess.

Ok, I'll change it to 1<<27.  I'll add a comment stating that we're trying to
keep internal sb flags separate.  It looks like those last four flags are all
internal?

> > diff --git a/mm/Kconfig b/mm/Kconfig
> > index 278e3ab..7901d83 100644
> > --- a/mm/Kconfig
> > +++ b/mm/Kconfig
> > @@ -258,6 +258,19 @@ config BOUNCE
> >  	def_bool y
> >  	depends on BLOCK && MMU && (ZONE_DMA || HIGHMEM)
> >  
> > +# On the 'tile' arch, USB OHCI needs the bounce pool since tilegx will often
> > +# have more than 4GB of memory, but we don't currently use the IOTLB to present
> > +# a 32-bit address to OHCI.  So we need to use a bounce pool instead.
> > +#
> > +# We also use the bounce pool to provide stable page writes for jbd.  jbd
> > +# initiates buffer writeback without locking the page or setting PG_writeback,
> > +# and fixing that behavior (a second time; jbd2 doesn't have this problem) is
> > +# a major rework effort.  Instead, use the bounce buffer to snapshot pages
> > +# (until jbd goes away).  The only jbd user is ext3.
> > +config NEED_BOUNCE_POOL
> > +	bool
> > +	default y if (TILE && USB_OHCI_HCD) || (BLK_DEV_INTEGRITY && JBD)
> > +
> >  config NR_QUICK
> >  	int
> >  	depends on QUICKLIST
> > diff --git a/mm/bounce.c b/mm/bounce.c
> > index 0420867..a5b30f9 100644
> > --- a/mm/bounce.c
> > +++ b/mm/bounce.c
> > @@ -178,8 +178,44 @@ static void bounce_end_io_read_isa(struct bio *bio, int err)
> >  	__bounce_end_io_read(bio, isa_page_pool, err);
> >  }
> >  
> > +#ifdef CONFIG_NEED_BOUNCE_POOL
> > +static int must_snapshot_stable_pages(struct bio *bio)
> > +{
> > +	struct page *page;
> > +	struct backing_dev_info *bdi;
> > +	struct address_space *mapping;
> > +	struct bio_vec *from;
> > +	int i;
> > +
> > +	if (bio_data_dir(bio) != WRITE)
> > +		return 0;
> > +
> > +	/*
> > +	 * Based on the first page that has a valid mapping, decide whether or
> > +	 * not we have to employ bounce buffering to guarantee stable pages.
> > +	 */
> > +	bio_for_each_segment(from, bio, i) {
> > +		page = from->bv_page;
> > +		mapping = page_mapping(page);
> > +		if (!mapping)
> > +			continue;
> > +		bdi = mapping->backing_dev_info;
> > +		if (!bdi_cap_stable_pages_required(bdi))
> > +			return 0;
> > +		return mapping->host->i_sb->s_flags & MS_SNAP_STABLE;
> > +	}
> > +
> > +	return 0;
> > +}
>   How about using q->backing_dev_info for the
> bdi_cap_stable_pages_required() check? It will be a fast path and this check
> will be faster..

Ok.

--D
> 
> 								Honza
> -- 
> Jan Kara <jack@suse.cz>
> SUSE Labs, CR
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Martin K. Petersen - Jan. 17, 2013, 3:26 a.m.
>>>>> "Darrick" == Darrick J Wong <darrick.wong@oracle.com> writes:

>> Umm, why did you move this hunk?

Darrick> I moved it so that the integrity data are generated against the
Darrick> contents of the bounce buffer because the write paths don't
Darrick> wait for writeback to finish if the snapshotting mode is
Darrick> enabled, which means (I think) that programs can wander in and
Darrick> scribble on the original page in between bio_integrity_prep and
Darrick> blk_queue_bounce.

It's also important that bio_integrity_prep() is run after partition
remapping has taken place. But it looks like that's still the case after
your change.
Jan Kara - Jan. 17, 2013, 10:32 a.m.
On Wed 16-01-13 19:01:32, Darrick J. Wong wrote:
> > > diff --git a/block/blk-core.c b/block/blk-core.c
> > > index c973249..277134c 100644
> > > --- a/block/blk-core.c
> > > +++ b/block/blk-core.c
> > > @@ -1474,6 +1474,11 @@ void blk_queue_bio(struct request_queue *q, struct bio *bio)
> > >  	 */
> > >  	blk_queue_bounce(q, &bio);
> > >  
> > > +	if (bio_integrity_enabled(bio) && bio_integrity_prep(bio)) {
> > > +		bio_endio(bio, -EIO);
> > > +		return;
> > > +	}
> > > +
> > >  	if (bio->bi_rw & (REQ_FLUSH | REQ_FUA)) {
> > >  		spin_lock_irq(q->queue_lock);
> > >  		where = ELEVATOR_INSERT_FLUSH;
> > > @@ -1714,9 +1719,6 @@ generic_make_request_checks(struct bio *bio)
> > >  	 */
> > >  	blk_partition_remap(bio);
> > >  
> > > -	if (bio_integrity_enabled(bio) && bio_integrity_prep(bio))
> > > -		goto end_io;
> > > -
> >   Umm, why did you move this hunk?
> 
> I moved it so that the integrity data are generated against the contents of the
> bounce buffer because the write paths don't wait for writeback to finish if the
> snapshotting mode is enabled, which means (I think) that programs can wander in
> and scribble on the original page in between bio_integrity_prep and
> blk_queue_bounce.
  Ah, I see. OK.

> > >  	if (bio_check_eod(bio, nr_sectors))
> > >  		goto end_io;
> > 
> > > diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
> > > index 780d4c6..0144fbb 100644
> > > --- a/include/uapi/linux/fs.h
> > > +++ b/include/uapi/linux/fs.h
> > > @@ -69,6 +69,7 @@ struct inodes_stat_t {
> > >  #define MS_REMOUNT	32	/* Alter flags of a mounted FS */
> > >  #define MS_MANDLOCK	64	/* Allow mandatory locks on an FS */
> > >  #define MS_DIRSYNC	128	/* Directory modifications are synchronous */
> > > +#define MS_SNAP_STABLE	256	/* Snapshot pages during writeback, if needed */
> > >  #define MS_NOATIME	1024	/* Do not update access times. */
> > >  #define MS_NODIRATIME	2048	/* Do not update directory access times */
> > >  #define MS_BIND		4096
> >   Please don't mix MS_SNAP_STABLE flag among flags passed by mount(2)
> > syscall. I think putting it at 1 << 27 might be acceptable. I remember
> > Al Viro saying something along the lines that kernel internal superblock
> > flags should be separated from those passed from userspace into a special
> > superblock entry but that's a different story I guess.
> 
> Ok, I'll change it to 1<<27.  I'll add a comment stating that we're trying to
> keep internal sb flags separate.  It looks like those last four flags are all
> internal?
  Yes. Flags with low numbers are part of kernel ABI...

> > > diff --git a/mm/bounce.c b/mm/bounce.c
> > > index 0420867..a5b30f9 100644
> > > --- a/mm/bounce.c
> > > +++ b/mm/bounce.c
> > > @@ -178,8 +178,44 @@ static void bounce_end_io_read_isa(struct bio *bio, int err)
> > >  	__bounce_end_io_read(bio, isa_page_pool, err);
> > >  }
> > >  
> > > +#ifdef CONFIG_NEED_BOUNCE_POOL
> > > +static int must_snapshot_stable_pages(struct bio *bio)
> > > +{
> > > +	struct page *page;
> > > +	struct backing_dev_info *bdi;
> > > +	struct address_space *mapping;
> > > +	struct bio_vec *from;
> > > +	int i;
> > > +
> > > +	if (bio_data_dir(bio) != WRITE)
> > > +		return 0;
> > > +
> > > +	/*
> > > +	 * Based on the first page that has a valid mapping, decide whether or
> > > +	 * not we have to employ bounce buffering to guarantee stable pages.
> > > +	 */
> > > +	bio_for_each_segment(from, bio, i) {
> > > +		page = from->bv_page;
> > > +		mapping = page_mapping(page);
> > > +		if (!mapping)
> > > +			continue;
> > > +		bdi = mapping->backing_dev_info;
> > > +		if (!bdi_cap_stable_pages_required(bdi))
> > > +			return 0;
> > > +		return mapping->host->i_sb->s_flags & MS_SNAP_STABLE;
> > > +	}
> > > +
> > > +	return 0;
> > > +}
> >   How about using q->backing_dev_info for the
> > bdi_cap_stable_pages_required() check? It will be a fast path and this check
> > will be faster..
> Ok.
  And maybe I should have told explicitely that then you can move the check
before bio_for_each_segment() loop...

								Honza

Patch

diff --git a/arch/tile/Kconfig b/arch/tile/Kconfig
index 875d008..c671fda 100644
--- a/arch/tile/Kconfig
+++ b/arch/tile/Kconfig
@@ -410,12 +410,6 @@  config TILE_USB
 	  Provides USB host adapter support for the built-in EHCI and OHCI
 	  interfaces on TILE-Gx chips.
 
-# USB OHCI needs the bounce pool since tilegx will often have more
-# than 4GB of memory, but we don't currently use the IOTLB to present
-# a 32-bit address to OHCI.  So we need to use a bounce pool instead.
-config NEED_BOUNCE_POOL
-	def_bool USB_OHCI_HCD
-
 source "drivers/pci/hotplug/Kconfig"
 
 endmenu
diff --git a/block/blk-core.c b/block/blk-core.c
index c973249..277134c 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -1474,6 +1474,11 @@  void blk_queue_bio(struct request_queue *q, struct bio *bio)
 	 */
 	blk_queue_bounce(q, &bio);
 
+	if (bio_integrity_enabled(bio) && bio_integrity_prep(bio)) {
+		bio_endio(bio, -EIO);
+		return;
+	}
+
 	if (bio->bi_rw & (REQ_FLUSH | REQ_FUA)) {
 		spin_lock_irq(q->queue_lock);
 		where = ELEVATOR_INSERT_FLUSH;
@@ -1714,9 +1719,6 @@  generic_make_request_checks(struct bio *bio)
 	 */
 	blk_partition_remap(bio);
 
-	if (bio_integrity_enabled(bio) && bio_integrity_prep(bio))
-		goto end_io;
-
 	if (bio_check_eod(bio, nr_sectors))
 		goto end_io;
 
diff --git a/fs/ext3/super.c b/fs/ext3/super.c
index 6e50223..4ba2683 100644
--- a/fs/ext3/super.c
+++ b/fs/ext3/super.c
@@ -2065,6 +2065,7 @@  static int ext3_fill_super (struct super_block *sb, void *data, int silent)
 		test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
 		test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
 		"writeback");
+	sb->s_flags |= MS_SNAP_STABLE;
 
 	return 0;
 
diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
index 780d4c6..0144fbb 100644
--- a/include/uapi/linux/fs.h
+++ b/include/uapi/linux/fs.h
@@ -69,6 +69,7 @@  struct inodes_stat_t {
 #define MS_REMOUNT	32	/* Alter flags of a mounted FS */
 #define MS_MANDLOCK	64	/* Allow mandatory locks on an FS */
 #define MS_DIRSYNC	128	/* Directory modifications are synchronous */
+#define MS_SNAP_STABLE	256	/* Snapshot pages during writeback, if needed */
 #define MS_NOATIME	1024	/* Do not update access times. */
 #define MS_NODIRATIME	2048	/* Do not update directory access times */
 #define MS_BIND		4096
diff --git a/mm/Kconfig b/mm/Kconfig
index 278e3ab..7901d83 100644
--- a/mm/Kconfig
+++ b/mm/Kconfig
@@ -258,6 +258,19 @@  config BOUNCE
 	def_bool y
 	depends on BLOCK && MMU && (ZONE_DMA || HIGHMEM)
 
+# On the 'tile' arch, USB OHCI needs the bounce pool since tilegx will often
+# have more than 4GB of memory, but we don't currently use the IOTLB to present
+# a 32-bit address to OHCI.  So we need to use a bounce pool instead.
+#
+# We also use the bounce pool to provide stable page writes for jbd.  jbd
+# initiates buffer writeback without locking the page or setting PG_writeback,
+# and fixing that behavior (a second time; jbd2 doesn't have this problem) is
+# a major rework effort.  Instead, use the bounce buffer to snapshot pages
+# (until jbd goes away).  The only jbd user is ext3.
+config NEED_BOUNCE_POOL
+	bool
+	default y if (TILE && USB_OHCI_HCD) || (BLK_DEV_INTEGRITY && JBD)
+
 config NR_QUICK
 	int
 	depends on QUICKLIST
diff --git a/mm/bounce.c b/mm/bounce.c
index 0420867..a5b30f9 100644
--- a/mm/bounce.c
+++ b/mm/bounce.c
@@ -178,8 +178,44 @@  static void bounce_end_io_read_isa(struct bio *bio, int err)
 	__bounce_end_io_read(bio, isa_page_pool, err);
 }
 
+#ifdef CONFIG_NEED_BOUNCE_POOL
+static int must_snapshot_stable_pages(struct bio *bio)
+{
+	struct page *page;
+	struct backing_dev_info *bdi;
+	struct address_space *mapping;
+	struct bio_vec *from;
+	int i;
+
+	if (bio_data_dir(bio) != WRITE)
+		return 0;
+
+	/*
+	 * Based on the first page that has a valid mapping, decide whether or
+	 * not we have to employ bounce buffering to guarantee stable pages.
+	 */
+	bio_for_each_segment(from, bio, i) {
+		page = from->bv_page;
+		mapping = page_mapping(page);
+		if (!mapping)
+			continue;
+		bdi = mapping->backing_dev_info;
+		if (!bdi_cap_stable_pages_required(bdi))
+			return 0;
+		return mapping->host->i_sb->s_flags & MS_SNAP_STABLE;
+	}
+
+	return 0;
+}
+#else
+static int must_snapshot_stable_pages(struct bio *bio)
+{
+	return 0;
+}
+#endif /* CONFIG_NEED_BOUNCE_POOL */
+
 static void __blk_queue_bounce(struct request_queue *q, struct bio **bio_orig,
-			       mempool_t *pool)
+			       mempool_t *pool, int force)
 {
 	struct page *page;
 	struct bio *bio = NULL;
@@ -192,7 +228,7 @@  static void __blk_queue_bounce(struct request_queue *q, struct bio **bio_orig,
 		/*
 		 * is destination page below bounce pfn?
 		 */
-		if (page_to_pfn(page) <= queue_bounce_pfn(q))
+		if (page_to_pfn(page) <= queue_bounce_pfn(q) && !force)
 			continue;
 
 		/*
@@ -270,6 +306,7 @@  static void __blk_queue_bounce(struct request_queue *q, struct bio **bio_orig,
 
 void blk_queue_bounce(struct request_queue *q, struct bio **bio_orig)
 {
+	int must_bounce;
 	mempool_t *pool;
 
 	/*
@@ -278,13 +315,15 @@  void blk_queue_bounce(struct request_queue *q, struct bio **bio_orig)
 	if (!bio_has_data(*bio_orig))
 		return;
 
+	must_bounce = must_snapshot_stable_pages(*bio_orig);
+
 	/*
 	 * for non-isa bounce case, just check if the bounce pfn is equal
 	 * to or bigger than the highest pfn in the system -- in that case,
 	 * don't waste time iterating over bio segments
 	 */
 	if (!(q->bounce_gfp & GFP_DMA)) {
-		if (queue_bounce_pfn(q) >= blk_max_pfn)
+		if (queue_bounce_pfn(q) >= blk_max_pfn && !must_bounce)
 			return;
 		pool = page_pool;
 	} else {
@@ -295,7 +334,7 @@  void blk_queue_bounce(struct request_queue *q, struct bio **bio_orig)
 	/*
 	 * slow path
 	 */
-	__blk_queue_bounce(q, bio_orig, pool);
+	__blk_queue_bounce(q, bio_orig, pool, must_bounce);
 }
 
 EXPORT_SYMBOL(blk_queue_bounce);
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 9c5af4d..75a13f3 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -2305,6 +2305,10 @@  void wait_for_stable_page(struct page *page)
 
 	if (!bdi_cap_stable_pages_required(bdi))
 		return;
+#ifdef CONFIG_NEED_BOUNCE_POOL
+	if (mapping->host->i_sb->s_flags & MS_SNAP_STABLE)
+		return;
+#endif /* CONFIG_NEED_BOUNCE_POOL */
 
 	wait_on_page_writeback(page);
 }