From: "Darrick J. Wong" <djwong@kernel.org>
To: Dave Chinner <david@fromorbit.com>
Cc: linux-xfs@vger.kernel.org, hch@lst.de
Subject: Re: [PATCH 03/10] xfs: use alloc_pages_bulk_array() for buffers
Date: Thu, 27 May 2021 16:01:02 -0700 [thread overview]
Message-ID: <20210527230102.GD2402049@locust> (raw)
In-Reply-To: <20210527225951.GC2402049@locust>
On Thu, May 27, 2021 at 03:59:51PM -0700, Darrick J. Wong wrote:
> On Thu, May 27, 2021 at 08:47:15AM +1000, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@redhat.com>
> >
> > Because it's more efficient than allocating pages one at a time in a
> > loop.
> >
> > Signed-off-by: Dave Chinner <dchinner@redhat.com>
> > ---
> > fs/xfs/xfs_buf.c | 62 +++++++++++++++++++-----------------------------
> > 1 file changed, 24 insertions(+), 38 deletions(-)
> >
> > diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
> > index b1610115d401..8ca4add138c5 100644
> > --- a/fs/xfs/xfs_buf.c
> > +++ b/fs/xfs/xfs_buf.c
> > @@ -386,10 +386,7 @@ xfs_buf_alloc_pages(
> > xfs_buf_flags_t flags)
> > {
> > gfp_t gfp_mask = xb_to_gfp(flags);
> > - size_t size;
> > - size_t offset;
> > - size_t nbytes;
> > - int i;
> > + long filled = 0;
> > int error;
> >
> > /* Assure zeroed buffer for non-read cases. */
> > @@ -400,50 +397,39 @@ xfs_buf_alloc_pages(
> > if (unlikely(error))
> > return error;
> >
> > - offset = bp->b_offset;
> > bp->b_flags |= _XBF_PAGES;
> >
> > - for (i = 0; i < bp->b_page_count; i++) {
> > - struct page *page;
> > - uint retries = 0;
> > -retry:
> > - page = alloc_page(gfp_mask);
> > - if (unlikely(page == NULL)) {
> > - if (flags & XBF_READ_AHEAD) {
> > - bp->b_page_count = i;
> > - error = -ENOMEM;
> > - goto out_free_pages;
> > - }
> > + /*
> > + * Bulk filling of pages can take multiple calls. Not filling the entire
> > + * array is not an allocation failure, so don't back off if we get at
> > + * least one extra page.
> > + */
> > + for (;;) {
> > + long last = filled;
> >
> > - /*
> > - * This could deadlock.
> > - *
> > - * But until all the XFS lowlevel code is revamped to
> > - * handle buffer allocation failures we can't do much.
> > - */
> > - if (!(++retries % 100))
> > - xfs_err(NULL,
> > - "%s(%u) possible memory allocation deadlock in %s (mode:0x%x)",
> > - current->comm, current->pid,
> > - __func__, gfp_mask);
> > -
> > - XFS_STATS_INC(bp->b_mount, xb_page_retries);
> > - congestion_wait(BLK_RW_ASYNC, HZ/50);
> > - goto retry;
> > + filled = alloc_pages_bulk_array(gfp_mask, bp->b_page_count,
> > + bp->b_pages);
> > + if (filled == bp->b_page_count) {
> > + XFS_STATS_INC(bp->b_mount, xb_page_found);
> > + break;
> > }
> >
> > - XFS_STATS_INC(bp->b_mount, xb_page_found);
> > + if (filled != last)
> > + continue;
> >
> > - nbytes = min_t(size_t, size, PAGE_SIZE - offset);
> > - size -= nbytes;
> > - bp->b_pages[i] = page;
> > - offset = 0;
> > + if (flags & XBF_READ_AHEAD) {
> > + error = -ENOMEM;
> > + goto out_free_pages;
> > + }
> > +
> > + XFS_STATS_INC(bp->b_mount, xb_page_retries);
> > + congestion_wait(BLK_RW_ASYNC, HZ/50);
>
> Nit: spaces around operators ("HZ / 50").
>
> With that fixed,
> Reviewed-by: Darrick J. Wong <djwong@kernel.org>
>
> I have a question about _xfs_buf_get_pages:
Never mind, you fixed all this in the next patch, which my grep didn't
find. Question withdrawn.
--D
>
> STATIC int
> _xfs_buf_get_pages(
> struct xfs_buf *bp,
> int page_count)
> {
> /* Make sure that we have a page list */
> if (bp->b_pages == NULL) {
> bp->b_page_count = page_count;
> if (page_count <= XB_PAGES) {
> bp->b_pages = bp->b_page_array;
> } else {
> bp->b_pages = kmem_alloc(sizeof(struct page *) *
> page_count, KM_NOFS);
> if (bp->b_pages == NULL)
> return -ENOMEM;
> }
> memset(bp->b_pages, 0, sizeof(struct page *) * page_count);
> }
> return 0;
> }
>
> xfs_bufs are kmem_cache_zalloc'd, which means that b_page_array should
> be zeroed, right?
>
> And we could use kmem_zalloc for the pagecount > XB_PAGES case, which
> would make the memset necessary, wouldn't it?
>
> OFC that only holds if a buffer that fails the memory allocation is
> immediately fed to _xfs_buf_free_pages to null out b_pages, which I
> think is true...?
>
> --D
>
> > }
> > return 0;
> >
> > out_free_pages:
> > - for (i = 0; i < bp->b_page_count; i++)
> > - __free_page(bp->b_pages[i]);
> > + while (--filled >= 0)
> > + __free_page(bp->b_pages[filled]);
> > bp->b_flags &= ~_XBF_PAGES;
> > return error;
> > }
> > --
> > 2.31.1
> >
next prev parent reply other threads:[~2021-05-27 23:01 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-26 22:47 [PATCH 00/10] xfs: buffer bulk page allocation and cleanups Dave Chinner
2021-05-26 22:47 ` [PATCH 01/10] xfs: split up xfs_buf_allocate_memory Dave Chinner
2021-05-27 22:48 ` Darrick J. Wong
2021-05-27 23:10 ` Darrick J. Wong
2021-05-26 22:47 ` [PATCH 02/10] xfs: use xfs_buf_alloc_pages for uncached buffers Dave Chinner
2021-05-27 22:50 ` Darrick J. Wong
2021-05-26 22:47 ` [PATCH 03/10] xfs: use alloc_pages_bulk_array() for buffers Dave Chinner
2021-05-27 22:59 ` Darrick J. Wong
2021-05-27 23:01 ` Darrick J. Wong [this message]
2021-05-26 22:47 ` [PATCH 04/10] xfs: merge _xfs_buf_get_pages() Dave Chinner
2021-05-27 23:02 ` Darrick J. Wong
2021-05-26 22:47 ` [PATCH 05/10] xfs: move page freeing into _xfs_buf_free_pages() Dave Chinner
2021-05-27 23:03 ` Darrick J. Wong
2021-05-26 22:47 ` [PATCH 06/10] xfs: remove ->b_offset handling for page backed buffers Dave Chinner
2021-05-27 23:09 ` Darrick J. Wong
2021-06-01 1:46 ` Dave Chinner
2021-05-26 22:47 ` [PATCH 07/10] xfs: simplify the b_page_count calculation Dave Chinner
2021-05-27 23:15 ` Darrick J. Wong
2021-05-27 23:29 ` Dave Chinner
2021-05-26 22:47 ` [PATCH 08/10] xfs: get rid of xb_to_gfp() Dave Chinner
2021-05-27 23:12 ` Darrick J. Wong
2021-05-26 22:47 ` [PATCH 09/10] xfs: cleanup error handling in xfs_buf_get_map Dave Chinner
2021-05-27 23:16 ` Darrick J. Wong
2021-05-26 22:47 ` [PATCH 10/10] xfs: merge xfs_buf_allocate_memory Dave Chinner
2021-05-27 23:17 ` Darrick J. Wong
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210527230102.GD2402049@locust \
--to=djwong@kernel.org \
--cc=david@fromorbit.com \
--cc=hch@lst.de \
--cc=linux-xfs@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.