[PATCH] block_dev: implement readpages() to optimize sequential read

[PATCH] block_dev: implement readpages() to optimize sequential read

[Akinobu Mita]

Sequential read from a block device is expected to be equal or faster
than from the file on a filesystem.  But it is not correct due to the
lack of effective readpages() in the address space operations for
block device.

This implements readpages() operation for block device by using
mpage_readpages() which can create multipage BIOs instead of BIOs for
each page.  It enables to reduce system CPU time consumption.

Install 1GB of RAM disk storage:

	# modprobe scsi_debug dev_size_mb=1024 delay=0

Sequential read from file on a filesystem:

	# mkfs.ext4 /dev/$DEV
	# mount /dev/$DEV /mnt
	# fio --name=t --size=512m --rw=read --filename=/mnt/file
	...
	  read : io=524288KB, bw=2133.4MB/s, iops=546133, runt=   240msec

Sequential read from a block device:
	# fio --name=t --size=512m --rw=read --filename=/dev/sdb
	...
(Without this commit)
	  read : io=524288KB, bw=1700.2MB/s, iops=435455, runt=   301msec

(With this commit)
	  read : io=524288KB, bw=2160.4MB/s, iops=553046, runt=   237msec

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: linux-fsdevel@vger.kernel.org
---
 fs/block_dev.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/fs/block_dev.c b/fs/block_dev.c
index 6d72746..e2f3ad08 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -304,6 +304,12 @@ static int blkdev_readpage(struct file * file, struct page * page)
 	return block_read_full_page(page, blkdev_get_block);
 }
 
+static int blkdev_readpages(struct file *file, struct address_space *mapping,
+			struct list_head *pages, unsigned nr_pages)
+{
+	return mpage_readpages(mapping, pages, nr_pages, blkdev_get_block);
+}
+
 static int blkdev_write_begin(struct file *file, struct address_space *mapping,
 			loff_t pos, unsigned len, unsigned flags,
 			struct page **pagep, void **fsdata)
@@ -1622,6 +1628,7 @@ static int blkdev_releasepage(struct page *page, gfp_t wait)
 
 static const struct address_space_operations def_blk_aops = {
 	.readpage	= blkdev_readpage,
+	.readpages	= blkdev_readpages,
 	.writepage	= blkdev_writepage,
 	.write_begin	= blkdev_write_begin,
 	.write_end	= blkdev_write_end,
-- 
1.9.1

Re: [PATCH] block_dev: implement readpages() to optimize sequential read

[Jeff Moyer]

Akinobu Mita <akinobu.mita@gmail.com> writes:

> Sequential read from a block device is expected to be equal or faster
> than from the file on a filesystem.  But it is not correct due to the
> lack of effective readpages() in the address space operations for
> block device.

Ah, a trip down memory lane.  ;-)  Here's a thread showing issues with
the last time this was proposed (by me, incidentally):

https://lkml.org/lkml/2009/6/2/480

At the very least, we need to see numbers on a real device, and see it
booted on something with >4k page size before taking this back in.

Cheers,
Jeff

Re: [PATCH] block_dev: implement readpages() to optimize sequential read

[Akinobu Mita]

2014-07-18 3:31 GMT+09:00 Jeff Moyer <jmoyer@redhat.com>:
> Akinobu Mita <akinobu.mita@gmail.com> writes:
>
>> Sequential read from a block device is expected to be equal or faster
>> than from the file on a filesystem.  But it is not correct due to the
>> lack of effective readpages() in the address space operations for
>> block device.
>
> Ah, a trip down memory lane.  ;-)  Here's a thread showing issues with
> the last time this was proposed (by me, incidentally):
>
> https://lkml.org/lkml/2009/6/2/480
>
> At the very least, we need to see numbers on a real device, and see it
> booted on something with >4k page size before taking this back in.

Thanks for the information.  First, I'll try to reproduce the >4k page
size issue with qemu with the architecture which can support 64k page.

Re: [PATCH] block_dev: implement readpages() to optimize sequential read

[Akinobu Mita]

2014-07-18 8:14 GMT+09:00 Akinobu Mita <akinobu.mita@gmail.com>:
> 2014-07-18 3:31 GMT+09:00 Jeff Moyer <jmoyer@redhat.com>:
>> Akinobu Mita <akinobu.mita@gmail.com> writes:
>>
>>> Sequential read from a block device is expected to be equal or faster
>>> than from the file on a filesystem.  But it is not correct due to the
>>> lack of effective readpages() in the address space operations for
>>> block device.
>>
>> Ah, a trip down memory lane.  ;-)  Here's a thread showing issues with
>> the last time this was proposed (by me, incidentally):
>>
>> https://lkml.org/lkml/2009/6/2/480
>>
>> At the very least, we need to see numbers on a real device, and see it
>> booted on something with >4k page size before taking this back in.
>
> Thanks for the information.  First, I'll try to reproduce the >4k page
> size issue with qemu with the architecture which can support 64k page.

I can reproduce the issue which attempts to access beyond end of
device, and it turns out that this can be reproducible on 4K page
system, too.

This problem only happens when reading a partition which is being
mounted by the filesystem at the same time.  The filesystem sets
the blocksize to a multiple of the logical block size by set_blocksize()
while it is mounted.  If the end of the partition is not aligned on
the new blocksize, reading the end of the partition causes access
beyond end of device.

So this patch is incomplete, and some guard check code is required in
mpage_readpages() or altenative solution is needed.