Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > linux.kernel > #1709261
| From | Minchan Kim <minchan@kernel.org> |
|---|---|
| Newsgroups | linux.kernel |
| Subject | [PATCH v2 3/7] fs: use on-stack-bio if backing device has BDI_CAP_SYNCHRONOUS capability |
| Date | 2017-08-11 07:20 +0200 |
| Message-ID | <udawy-7MZ-11@gated-at.bofh.it> (permalink) |
| References | <udawx-7MZ-3@gated-at.bofh.it> |
| Organization | linux.* mail to news gateway |
There is no need to use dynamic bio allocation for BDI_CAP_SYNCHRONOUS
devices. They can with on-stack bio without concern about waiting
bio allocation from mempool under heavy memory pressure.
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Minchan Kim <minchan@kernel.org>
---
Hi Mattew,
I didn't use sbvec[nr_pages] as you suggested[1] because I don't think
it's pointless in do_mpage_readpage which works per-page base as
I replied to you.
If I misunderstood something, please correct me.
[1] http://lkml.kernel.org/r/<20170808132904.GC31390@bombadil.infradead.org>
fs/mpage.c | 47 ++++++++++++++++++++++++++++++++++++++++-------
1 file changed, 40 insertions(+), 7 deletions(-)
diff --git a/fs/mpage.c b/fs/mpage.c
index 2e4c41ccb5c9..d3b777fdfd5a 100644
--- a/fs/mpage.c
+++ b/fs/mpage.c
@@ -162,6 +162,9 @@ do_mpage_readpage(struct bio *bio, struct page *page, unsigned nr_pages,
int fully_mapped = 1;
unsigned nblocks;
unsigned relative_block;
+ /* on-stack bio for synchronous devices */
+ struct bio sbio;
+ struct bio_vec sbvec;
if (page_has_buffers(page))
goto confused;
@@ -282,10 +285,22 @@ do_mpage_readpage(struct bio *bio, struct page *page, unsigned nr_pages,
page))
goto out;
}
- bio = mpage_alloc(bdev, blocks[0] << (blkbits - 9),
+
+ if (bdi_cap_synchronous_io(inode_to_bdi(inode))) {
+ bio = &sbio;
+ /* mpage_end_io calls bio_put unconditionally */
+ bio_get(&sbio);
+
+ bio_init(&sbio, &sbvec, 1);
+ sbio.bi_bdev = bdev;
+ sbio.bi_iter.bi_sector = blocks[0] << (blkbits - 9);
+ } else {
+
+ bio = mpage_alloc(bdev, blocks[0] << (blkbits - 9),
min_t(int, nr_pages, BIO_MAX_PAGES), gfp);
- if (bio == NULL)
- goto confused;
+ if (bio == NULL)
+ goto confused;
+ }
}
length = first_hole << blkbits;
@@ -302,6 +317,8 @@ do_mpage_readpage(struct bio *bio, struct page *page, unsigned nr_pages,
else
*last_block_in_bio = blocks[blocks_per_page - 1];
out:
+ if (bio == &sbio)
+ bio = mpage_bio_submit(REQ_OP_READ, 0, bio);
return bio;
confused:
@@ -492,6 +509,9 @@ static int __mpage_writepage(struct page *page, struct writeback_control *wbc,
loff_t i_size = i_size_read(inode);
int ret = 0;
int op_flags = wbc_to_write_flags(wbc);
+ /* on-stack-bio */
+ struct bio sbio;
+ struct bio_vec sbvec;
if (page_has_buffers(page)) {
struct buffer_head *head = page_buffers(page);
@@ -610,10 +630,21 @@ static int __mpage_writepage(struct page *page, struct writeback_control *wbc,
goto out;
}
}
- bio = mpage_alloc(bdev, blocks[0] << (blkbits - 9),
- BIO_MAX_PAGES, GFP_NOFS|__GFP_HIGH);
- if (bio == NULL)
- goto confused;
+
+ if (bdi_cap_synchronous_io(inode_to_bdi(inode))) {
+ bio = &sbio;
+ /* mpage_end_io calls bio_put unconditionally */
+ bio_get(&sbio);
+
+ bio_init(&sbio, &sbvec, 1);
+ sbio.bi_bdev = bdev;
+ sbio.bi_iter.bi_sector = blocks[0] << (blkbits - 9);
+ } else {
+ bio = mpage_alloc(bdev, blocks[0] << (blkbits - 9),
+ BIO_MAX_PAGES, GFP_NOFS|__GFP_HIGH);
+ if (bio == NULL)
+ goto confused;
+ }
wbc_init_bio(wbc, bio);
bio->bi_write_hint = inode->i_write_hint;
@@ -662,6 +693,8 @@ static int __mpage_writepage(struct page *page, struct writeback_control *wbc,
*/
mapping_set_error(mapping, ret);
out:
+ if (bio == &sbio)
+ bio = mpage_bio_submit(REQ_OP_WRITE, op_flags, bio);
mpd->bio = bio;
return ret;
}
--
2.7.4
Back to linux.kernel | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread
[PATCH v2 0/7] Replace rw_page with on-stack bio Minchan Kim <minchan@kernel.org> - 2017-08-11 07:20 +0200 [PATCH v2 3/7] fs: use on-stack-bio if backing device has BDI_CAP_SYNCHRONOUS capability Minchan Kim <minchan@kernel.org> - 2017-08-11 07:20 +0200 [PATCH v2 4/7] mm:swap: remove end_swap_bio_write argument Minchan Kim <minchan@kernel.org> - 2017-08-11 07:20 +0200 [PATCH v2 7/7] fs: remove rw_page Minchan Kim <minchan@kernel.org> - 2017-08-11 07:20 +0200 [PATCH v2 6/7] zram: remove zram_rw_page Minchan Kim <minchan@kernel.org> - 2017-08-11 07:20 +0200 [PATCH v2 5/7] mm:swap: use on-stack-bio for BDI_CAP_SYNCHRONOUS device Minchan Kim <minchan@kernel.org> - 2017-08-11 07:20 +0200 [PATCH v2 1/7] zram: set BDI_CAP_STABLE_WRITES once Minchan Kim <minchan@kernel.org> - 2017-08-11 07:20 +0200
csiph-web