From: "Darrick J. Wong" <darrick.wong@oracle.com>
To: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Christoph Hellwig <hch@infradead.org>,
linux-xfs@vger.kernel.org, linux-fsdevel@vger.kernel.org
Subject: Re: [PATCH 2/2] iomap: Convert readahead to iomap_iter
Date: Tue, 11 Aug 2020 13:56:13 -0700 [thread overview]
Message-ID: <20200811205314.GF6107@magnolia> (raw)
In-Reply-To: <20200728173216.7184-3-willy@infradead.org>
On Tue, Jul 28, 2020 at 06:32:15PM +0100, Matthew Wilcox (Oracle) wrote:
> This approach removes at least two indirect function calls from the
> readahead path. Previous call chain (indirect function calls marked *):
>
> xfs_vm_readahead
> iomap_readahead
> iomap_apply
> xfs_read_iomap_begin [*]
> iomap_readahead_actor [*]
> iomap_readpage_actor
>
> New call chain:
>
> xfs_vm_readahead
> xfs_iomap_next_read
> iomi_advance
> iomap_readahead
> iomap_readpage_actor
I mostly like this, with a few comments...
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> ---
> fs/iomap/buffered-io.c | 82 ++++++++++++++----------------------------
> fs/xfs/xfs_aops.c | 9 ++++-
> fs/xfs/xfs_iomap.c | 15 ++++++++
> fs/xfs/xfs_iomap.h | 2 ++
> fs/zonefs/super.c | 20 ++++++++++-
> include/linux/iomap.h | 10 +++++-
> 6 files changed, 79 insertions(+), 59 deletions(-)
>
> diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
> index bcfc288dba3f..fff23ed6a682 100644
> --- a/fs/iomap/buffered-io.c
> +++ b/fs/iomap/buffered-io.c
> @@ -206,13 +206,6 @@ iomap_read_end_io(struct bio *bio)
> bio_put(bio);
> }
>
> -struct iomap_readpage_ctx {
> - struct page *cur_page;
> - bool cur_page_in_bio;
> - struct bio *bio;
> - struct readahead_control *rac;
> -};
> -
> static void
> iomap_read_inline_data(struct inode *inode, struct page *page,
> struct iomap *iomap)
> @@ -369,35 +362,10 @@ iomap_readpage(struct page *page, const struct iomap_ops *ops)
> }
> EXPORT_SYMBOL_GPL(iomap_readpage);
>
> -static loff_t
> -iomap_readahead_actor(struct inode *inode, loff_t pos, loff_t length,
> - void *data, struct iomap *iomap, struct iomap *srcmap)
> -{
> - struct iomap_readpage_ctx *ctx = data;
> - loff_t done, ret;
> -
> - for (done = 0; done < length; done += ret) {
> - if (ctx->cur_page && offset_in_page(pos + done) == 0) {
> - if (!ctx->cur_page_in_bio)
> - unlock_page(ctx->cur_page);
> - put_page(ctx->cur_page);
> - ctx->cur_page = NULL;
> - }
> - if (!ctx->cur_page) {
> - ctx->cur_page = readahead_page(ctx->rac);
> - ctx->cur_page_in_bio = false;
> - }
> - ret = iomap_readpage_actor(inode, pos + done, length - done,
> - ctx, iomap, srcmap);
> - }
> -
> - return done;
> -}
> -
> /**
> * iomap_readahead - Attempt to read pages from a file.
> + * @iomi: The iomap iterator for this operation.
> * @rac: Describes the pages to be read.
> - * @ops: The operations vector for the filesystem.
> *
> * This function is for filesystems to call to implement their readahead
> * address_space operation.
> @@ -409,35 +377,37 @@ iomap_readahead_actor(struct inode *inode, loff_t pos, loff_t length,
> * function is called with memalloc_nofs set, so allocations will not cause
> * the filesystem to be reentered.
> */
> -void iomap_readahead(struct readahead_control *rac, const struct iomap_ops *ops)
> +loff_t iomap_readahead(struct iomap_iter *iomi, struct iomap_readpage_ctx *ctx)
> {
> - struct inode *inode = rac->mapping->host;
> - loff_t pos = readahead_pos(rac);
> - loff_t length = readahead_length(rac);
> - struct iomap_readpage_ctx ctx = {
> - .rac = rac,
> - };
> -
> - trace_iomap_readahead(inode, readahead_count(rac));
> + loff_t done, ret, length = iomap_length(iomi);
>
> - while (length > 0) {
> - loff_t ret = iomap_apply(inode, pos, length, 0, ops,
> - &ctx, iomap_readahead_actor);
> - if (ret <= 0) {
> - WARN_ON_ONCE(ret == 0);
> - break;
> + for (done = 0; done < length; done += ret) {
> + if (ctx->cur_page && offset_in_page(iomi->pos + done) == 0) {
> + if (!ctx->cur_page_in_bio)
> + unlock_page(ctx->cur_page);
> + put_page(ctx->cur_page);
> + ctx->cur_page = NULL;
> }
> - pos += ret;
> - length -= ret;
> + if (!ctx->cur_page) {
> + ctx->cur_page = readahead_page(ctx->rac);
> + ctx->cur_page_in_bio = false;
> + }
> + ret = iomap_readpage_actor(iomi->inode, iomi->pos + done,
> + length - done, ctx,
> + &iomi->iomap, &iomi->srcmap);
> }
>
> - if (ctx.bio)
> - submit_bio(ctx.bio);
> - if (ctx.cur_page) {
> - if (!ctx.cur_page_in_bio)
> - unlock_page(ctx.cur_page);
> - put_page(ctx.cur_page);
> + if (iomi->len == done) {
> + if (ctx->bio)
> + submit_bio(ctx->bio);
> + if (ctx->cur_page) {
> + if (!ctx->cur_page_in_bio)
> + unlock_page(ctx->cur_page);
> + put_page(ctx->cur_page);
> + }
> }
> +
> + return done;
> }
> EXPORT_SYMBOL_GPL(iomap_readahead);
>
> diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
> index b35611882ff9..2884752e40e8 100644
> --- a/fs/xfs/xfs_aops.c
> +++ b/fs/xfs/xfs_aops.c
> @@ -625,7 +625,14 @@ STATIC void
> xfs_vm_readahead(
> struct readahead_control *rac)
> {
> - iomap_readahead(rac, &xfs_read_iomap_ops);
> + IOMAP_ITER(iomi, rac->mapping->host, readahead_pos(rac),
> + readahead_length(rac), 0);
> + struct iomap_readpage_ctx ctx = {
> + .rac = rac,
> + };
> +
> + while (iomap_iter(&iomi, xfs_iomap_next_read))
> + iomi.copied = iomap_readahead(&iomi, &ctx);
Why not have iomap_readahead set iomi.copied on its way out? The actor
function is supposed to set iomi.ret if an error happens, right?
Oh wait no, the actor function returns a positive copied value, or a
negative error code, and then it's up to the _next_read function to
notice if copied is negative, stuff it in ret, and then return false to
stop the iteration?
> }
>
> static int
> diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c
> index 0e3f62cde375..66f2fcaf136e 100644
> --- a/fs/xfs/xfs_iomap.c
> +++ b/fs/xfs/xfs_iomap.c
> @@ -1150,6 +1150,21 @@ const struct iomap_ops xfs_read_iomap_ops = {
> .iomap_begin = xfs_read_iomap_begin,
> };
>
> +int
> +xfs_iomap_next_read(
> + const struct iomap_iter *iomi,
> + struct iomap *iomap,
> + struct iomap *srcmap)
Aren't these last two parameters already in the iomap iter?
Are they passed separately to work around the pointer being const?
> +{
> + if (iomi->copied < 0)
> + return iomi->copied;
Is this boilerplate going to end up in every single iomap_next_t
function? If so, it should probably just go in iomap_iter prior to the
next() call, right?
I also wonder if these functions (and the typedef) ought to be called
iomap_iter_advance_t since that's what they do -- pick up the status
from the last round, and advance the iterator to the next mapping that
we want to process.
> + if (iomi->copied >= iomi->len)
> + return 0;
Er... if we copied more than we asked for, doesn't that imply something
bad just happened?
> +
> + return xfs_read_iomap_begin(iomi->inode, iomi->pos + iomi->copied,
> + iomi->len - iomi->copied, iomi->flags, iomap, srcmap);
Would be kinda nice if you could just pass the whole iomap_iter, but I
get that we're probably stuck with this until the entirety gets
converted.
--D
> +}
> +
> static int
> xfs_seek_iomap_begin(
> struct inode *inode,
> diff --git a/fs/xfs/xfs_iomap.h b/fs/xfs/xfs_iomap.h
> index 7d3703556d0e..1b1fa225e938 100644
> --- a/fs/xfs/xfs_iomap.h
> +++ b/fs/xfs/xfs_iomap.h
> @@ -46,4 +46,6 @@ extern const struct iomap_ops xfs_read_iomap_ops;
> extern const struct iomap_ops xfs_seek_iomap_ops;
> extern const struct iomap_ops xfs_xattr_iomap_ops;
>
> +int xfs_iomap_next_read(const struct iomap_iter *iomi, struct iomap *iomap,
> + struct iomap *srcmap);
> #endif /* __XFS_IOMAP_H__*/
> diff --git a/fs/zonefs/super.c b/fs/zonefs/super.c
> index 07bc42d62673..4842b85ce36d 100644
> --- a/fs/zonefs/super.c
> +++ b/fs/zonefs/super.c
> @@ -70,6 +70,17 @@ static int zonefs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
> return 0;
> }
>
> +static int zonefs_iomap_next(const struct iomap_iter *iomi,
> + struct iomap *iomap, struct iomap *srcmap)
> +{
> + if (iomi->copied < 0)
> + return iomi->copied;
> + if (iomi->copied >= iomi->len)
> + return 0;
> + return zonefs_iomap_begin(iomi->inode, iomi->pos + iomi->copied,
> + iomi->len - iomi->copied, iomi->flags, iomap, srcmap);
> +}
> +
> static const struct iomap_ops zonefs_iomap_ops = {
> .iomap_begin = zonefs_iomap_begin,
> };
> @@ -81,7 +92,14 @@ static int zonefs_readpage(struct file *unused, struct page *page)
>
> static void zonefs_readahead(struct readahead_control *rac)
> {
> - iomap_readahead(rac, &zonefs_iomap_ops);
> + IOMAP_ITER(iomi, rac->mapping->host, readahead_pos(rac),
> + readahead_length(rac), 0);
> + struct iomap_readpage_ctx ctx = {
> + .rac = rac,
> + };
> +
> + while (iomap_iter(&iomi, zonefs_iomap_next))
> + iomi.copied = iomap_readahead(&iomi, &ctx);
> }
>
> /*
> diff --git a/include/linux/iomap.h b/include/linux/iomap.h
> index fe58e68ec0c1..dd9bfed85c4f 100644
> --- a/include/linux/iomap.h
> +++ b/include/linux/iomap.h
> @@ -212,7 +212,6 @@ loff_t iomap_apply(struct inode *inode, loff_t pos, loff_t length,
> ssize_t iomap_file_buffered_write(struct kiocb *iocb, struct iov_iter *from,
> const struct iomap_ops *ops);
> int iomap_readpage(struct page *page, const struct iomap_ops *ops);
> -void iomap_readahead(struct readahead_control *, const struct iomap_ops *ops);
> int iomap_set_page_dirty(struct page *page);
> int iomap_is_partially_uptodate(struct page *page, unsigned long from,
> unsigned long count);
> @@ -299,6 +298,15 @@ int iomap_writepages(struct address_space *mapping,
> struct writeback_control *wbc, struct iomap_writepage_ctx *wpc,
> const struct iomap_writeback_ops *ops);
>
> +struct iomap_readpage_ctx {
> + struct page *cur_page;
> + bool cur_page_in_bio;
> + struct bio *bio;
> + struct readahead_control *rac;
> +};
> +
> +loff_t iomap_readahead(struct iomap_iter *, struct iomap_readpage_ctx *);
> +
> /*
> * Flags for direct I/O ->end_io:
> */
> --
> 2.27.0
>
next prev parent reply other threads:[~2020-08-11 20:56 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-07-28 17:32 [RFC 0/2] Avoid indirect function calls in iomap Matthew Wilcox (Oracle)
2020-07-28 17:32 ` [PATCH 1/2] iomap: Add iomap_iter Matthew Wilcox (Oracle)
2020-08-11 21:01 ` Darrick J. Wong
2020-08-11 21:32 ` Matthew Wilcox
2020-07-28 17:32 ` [PATCH 2/2] iomap: Convert readahead to iomap_iter Matthew Wilcox (Oracle)
2020-08-11 20:56 ` Darrick J. Wong [this message]
2020-08-11 22:31 ` Matthew Wilcox
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=20200811205314.GF6107@magnolia \
--to=darrick.wong@oracle.com \
--cc=hch@infradead.org \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-xfs@vger.kernel.org \
--cc=willy@infradead.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).