[PATCH v4 00/18] Refactor generic_file_buffered_read

[PATCH v4 00/18] Refactor generic_file_buffered_read

[Matthew Wilcox (Oracle)]

This is a combination of Christoph's work to refactor
generic_file_buffered_read() and some of my large-page support
which was disrupted by Kent's refactoring of generic_file_buffered_read.

v4:
 - Rebase on next-20210120
 - Fix compilation error with pagevec_init & reinit
 - Simplify filemap_range_uptodate() by passing pos instead of iocb
v3:
 - Fixed missing put_page for readahead with IOCB_NOIO (hch)
 - Fixed commit message for lock_page_for_iocb (nborisov)
v2:
 - Added pagevec conversion upfront and rebased other patches on top of
   it (me)
 - Limit page search by max pgoff_t rather than by number of pages (me)
 - Renamed mapping_get_read_thps() to filemap_get_read_batch() (hch/me)
 - Added doc for filemap_get_read_batch() (hch/me)
 - Removed 'first' parameter from filemap_update_page() (me)
 - Folded "Remove parameters from filemap_update_page()" into an earlier
   patch (hch)
 - Restructured filemap_update_page() error handling flow (hch/me)
 - Pass the pagevec to filemap_create_page() (hch)
 - Renamed 'find_page' label to 'retry' (hch)
 - Explicitly check for AOP_TRUNCATED_PAGE instead of assuming err > 0
   means retry (hch)
 - Move mark_page_accessed() and handling of i_size into main copy loop (me)


Christoph Hellwig (2):
  mm/filemap: Rename generic_file_buffered_read to filemap_read
  mm/filemap: Simplify generic_file_read_iter

Matthew Wilcox (Oracle) (16):
  mm/filemap: Rename generic_file_buffered_read subfunctions
  mm/filemap: Remove dynamically allocated array from filemap_read
  mm/filemap: Convert filemap_get_pages to take a pagevec
  mm/filemap: Use head pages in generic_file_buffered_read
  mm/filemap: Pass a sleep state to put_and_wait_on_page_locked
  mm/filemap: Support readpage splitting a page
  mm/filemap: Inline __wait_on_page_locked_async into caller
  mm/filemap: Don't call ->readpage if IOCB_WAITQ is set
  mm/filemap: Change filemap_read_page calling conventions
  mm/filemap: Change filemap_create_page calling conventions
  mm/filemap: Convert filemap_update_page to return an errno
  mm/filemap: Move the iocb checks into filemap_update_page
  mm/filemap: Add filemap_range_uptodate
  mm/filemap: Split filemap_readahead out of filemap_get_pages
  mm/filemap: Restructure filemap_get_pages
  mm/filemap: Don't relock the page after calling readpage

 fs/btrfs/file.c         |   2 +-
 include/linux/fs.h      |   4 +-
 include/linux/pagemap.h |   3 +-
 mm/filemap.c            | 565 ++++++++++++++++++----------------------
 mm/huge_memory.c        |   4 +-
 mm/migrate.c            |   4 +-
 6 files changed, 267 insertions(+), 315 deletions(-)

-- 
2.29.2

[PATCH v4 01/18] mm/filemap: Rename generic_file_buffered_read subfunctions

[Matthew Wilcox (Oracle)]

The recent split of generic_file_buffered_read() created some very
long function names which are hard to distinguish from each other.
Rename as follows:

generic_file_buffered_read_readpage -> filemap_read_page
generic_file_buffered_read_pagenotuptodate -> filemap_update_page
generic_file_buffered_read_no_cached_page -> filemap_create_page
generic_file_buffered_read_get_pages -> filemap_get_pages

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
---
 mm/filemap.c | 44 +++++++++++++++-----------------------------
 1 file changed, 15 insertions(+), 29 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index d90614f501daa..934e04f1760ef 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2168,11 +2168,8 @@ static int lock_page_for_iocb(struct kiocb *iocb, struct page *page)
 		return lock_page_killable(page);
 }
 
-static struct page *
-generic_file_buffered_read_readpage(struct kiocb *iocb,
-				    struct file *filp,
-				    struct address_space *mapping,
-				    struct page *page)
+static struct page *filemap_read_page(struct kiocb *iocb, struct file *filp,
+		struct address_space *mapping, struct page *page)
 {
 	struct file_ra_state *ra = &filp->f_ra;
 	int error;
@@ -2223,12 +2220,9 @@ generic_file_buffered_read_readpage(struct kiocb *iocb,
 	return page;
 }
 
-static struct page *
-generic_file_buffered_read_pagenotuptodate(struct kiocb *iocb,
-					   struct file *filp,
-					   struct iov_iter *iter,
-					   struct page *page,
-					   loff_t pos, loff_t count)
+static struct page *filemap_update_page(struct kiocb *iocb, struct file *filp,
+		struct iov_iter *iter, struct page *page, loff_t pos,
+		loff_t count)
 {
 	struct address_space *mapping = filp->f_mapping;
 	struct inode *inode = mapping->host;
@@ -2291,12 +2285,11 @@ generic_file_buffered_read_pagenotuptodate(struct kiocb *iocb,
 		return page;
 	}
 
-	return generic_file_buffered_read_readpage(iocb, filp, mapping, page);
+	return filemap_read_page(iocb, filp, mapping, page);
 }
 
-static struct page *
-generic_file_buffered_read_no_cached_page(struct kiocb *iocb,
-					  struct iov_iter *iter)
+static struct page *filemap_create_page(struct kiocb *iocb,
+		struct iov_iter *iter)
 {
 	struct file *filp = iocb->ki_filp;
 	struct address_space *mapping = filp->f_mapping;
@@ -2307,10 +2300,6 @@ generic_file_buffered_read_no_cached_page(struct kiocb *iocb,
 	if (iocb->ki_flags & IOCB_NOIO)
 		return ERR_PTR(-EAGAIN);
 
-	/*
-	 * Ok, it wasn't cached, so we need to create a new
-	 * page..
-	 */
 	page = page_cache_alloc(mapping);
 	if (!page)
 		return ERR_PTR(-ENOMEM);
@@ -2322,13 +2311,11 @@ generic_file_buffered_read_no_cached_page(struct kiocb *iocb,
 		return error != -EEXIST ? ERR_PTR(error) : NULL;
 	}
 
-	return generic_file_buffered_read_readpage(iocb, filp, mapping, page);
+	return filemap_read_page(iocb, filp, mapping, page);
 }
 
-static int generic_file_buffered_read_get_pages(struct kiocb *iocb,
-						struct iov_iter *iter,
-						struct page **pages,
-						unsigned int nr)
+static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
+		struct page **pages, unsigned int nr)
 {
 	struct file *filp = iocb->ki_filp;
 	struct address_space *mapping = filp->f_mapping;
@@ -2355,7 +2342,7 @@ static int generic_file_buffered_read_get_pages(struct kiocb *iocb,
 	if (nr_got)
 		goto got_pages;
 
-	pages[0] = generic_file_buffered_read_no_cached_page(iocb, iter);
+	pages[0] = filemap_create_page(iocb, iter);
 	err = PTR_ERR_OR_ZERO(pages[0]);
 	if (!IS_ERR_OR_NULL(pages[0]))
 		nr_got = 1;
@@ -2389,8 +2376,8 @@ static int generic_file_buffered_read_get_pages(struct kiocb *iocb,
 				break;
 			}
 
-			page = generic_file_buffered_read_pagenotuptodate(iocb,
-					filp, iter, page, pg_pos, pg_count);
+			page = filemap_update_page(iocb, filp, iter, page,
+					pg_pos, pg_count);
 			if (IS_ERR_OR_NULL(page)) {
 				for (j = i + 1; j < nr_got; j++)
 					put_page(pages[j]);
@@ -2466,8 +2453,7 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 			iocb->ki_flags |= IOCB_NOWAIT;
 
 		i = 0;
-		pg_nr = generic_file_buffered_read_get_pages(iocb, iter,
-							     pages, nr_pages);
+		pg_nr = filemap_get_pages(iocb, iter, pages, nr_pages);
 		if (pg_nr < 0) {
 			error = pg_nr;
 			break;
-- 
2.29.2

[PATCH v4 02/18] mm/filemap: Remove dynamically allocated array from filemap_read

[Matthew Wilcox (Oracle)]

Increasing the batch size runs into diminishing returns.  It's probably
better to make, eg, three calls to filemap_get_pages() than it is to
call into kmalloc().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 mm/filemap.c | 15 ++-------------
 1 file changed, 2 insertions(+), 13 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index 934e04f1760ef..5dec04c8e16b0 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2421,8 +2421,8 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 	struct file_ra_state *ra = &filp->f_ra;
 	struct address_space *mapping = filp->f_mapping;
 	struct inode *inode = mapping->host;
-	struct page *pages_onstack[PAGEVEC_SIZE], **pages = NULL;
-	unsigned int nr_pages = min_t(unsigned int, 512,
+	struct page *pages[PAGEVEC_SIZE];
+	unsigned int nr_pages = min_t(unsigned int, PAGEVEC_SIZE,
 			((iocb->ki_pos + iter->count + PAGE_SIZE - 1) >> PAGE_SHIFT) -
 			(iocb->ki_pos >> PAGE_SHIFT));
 	int i, pg_nr, error = 0;
@@ -2433,14 +2433,6 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 		return 0;
 	iov_iter_truncate(iter, inode->i_sb->s_maxbytes);
 
-	if (nr_pages > ARRAY_SIZE(pages_onstack))
-		pages = kmalloc_array(nr_pages, sizeof(void *), GFP_KERNEL);
-
-	if (!pages) {
-		pages = pages_onstack;
-		nr_pages = min_t(unsigned int, nr_pages, ARRAY_SIZE(pages_onstack));
-	}
-
 	do {
 		cond_resched();
 
@@ -2525,9 +2517,6 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 
 	file_accessed(filp);
 
-	if (pages != pages_onstack)
-		kfree(pages);
-
 	return written ? written : error;
 }
 EXPORT_SYMBOL_GPL(generic_file_buffered_read);
-- 
2.29.2

[PATCH v4 03/18] mm/filemap: Convert filemap_get_pages to take a pagevec

[Matthew Wilcox (Oracle)]

Using a pagevec lets us keep the pages and the number of pages together
which simplifies a lot of the calling conventions.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 mm/filemap.c | 82 ++++++++++++++++++++++++----------------------------
 1 file changed, 38 insertions(+), 44 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index 5dec04c8e16b0..9e1e2bbac8fc9 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2315,22 +2315,22 @@ static struct page *filemap_create_page(struct kiocb *iocb,
 }
 
 static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
-		struct page **pages, unsigned int nr)
+		struct pagevec *pvec)
 {
 	struct file *filp = iocb->ki_filp;
 	struct address_space *mapping = filp->f_mapping;
 	struct file_ra_state *ra = &filp->f_ra;
 	pgoff_t index = iocb->ki_pos >> PAGE_SHIFT;
 	pgoff_t last_index = (iocb->ki_pos + iter->count + PAGE_SIZE-1) >> PAGE_SHIFT;
-	int i, j, nr_got, err = 0;
+	unsigned int nr = min_t(unsigned long, last_index - index, PAGEVEC_SIZE);
+	int i, j, err = 0;
 
-	nr = min_t(unsigned long, last_index - index, nr);
 find_page:
 	if (fatal_signal_pending(current))
 		return -EINTR;
 
-	nr_got = find_get_pages_contig(mapping, index, nr, pages);
-	if (nr_got)
+	pvec->nr = find_get_pages_contig(mapping, index, nr, pvec->pages);
+	if (pvec->nr)
 		goto got_pages;
 
 	if (iocb->ki_flags & IOCB_NOIO)
@@ -2338,17 +2338,17 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 
 	page_cache_sync_readahead(mapping, ra, filp, index, last_index - index);
 
-	nr_got = find_get_pages_contig(mapping, index, nr, pages);
-	if (nr_got)
+	pvec->nr = find_get_pages_contig(mapping, index, nr, pvec->pages);
+	if (pvec->nr)
 		goto got_pages;
 
-	pages[0] = filemap_create_page(iocb, iter);
-	err = PTR_ERR_OR_ZERO(pages[0]);
-	if (!IS_ERR_OR_NULL(pages[0]))
-		nr_got = 1;
+	pvec->pages[0] = filemap_create_page(iocb, iter);
+	err = PTR_ERR_OR_ZERO(pvec->pages[0]);
+	if (!IS_ERR_OR_NULL(pvec->pages[0]))
+		pvec->nr = 1;
 got_pages:
-	for (i = 0; i < nr_got; i++) {
-		struct page *page = pages[i];
+	for (i = 0; i < pvec->nr; i++) {
+		struct page *page = pvec->pages[i];
 		pgoff_t pg_index = index + i;
 		loff_t pg_pos = max(iocb->ki_pos,
 				    (loff_t) pg_index << PAGE_SHIFT);
@@ -2356,9 +2356,9 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 
 		if (PageReadahead(page)) {
 			if (iocb->ki_flags & IOCB_NOIO) {
-				for (j = i; j < nr_got; j++)
-					put_page(pages[j]);
-				nr_got = i;
+				for (j = i; j < pvec->nr; j++)
+					put_page(pvec->pages[j]);
+				pvec->nr = i;
 				err = -EAGAIN;
 				break;
 			}
@@ -2369,9 +2369,9 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 		if (!PageUptodate(page)) {
 			if ((iocb->ki_flags & IOCB_NOWAIT) ||
 			    ((iocb->ki_flags & IOCB_WAITQ) && i)) {
-				for (j = i; j < nr_got; j++)
-					put_page(pages[j]);
-				nr_got = i;
+				for (j = i; j < pvec->nr; j++)
+					put_page(pvec->pages[j]);
+				pvec->nr = i;
 				err = -EAGAIN;
 				break;
 			}
@@ -2379,17 +2379,17 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 			page = filemap_update_page(iocb, filp, iter, page,
 					pg_pos, pg_count);
 			if (IS_ERR_OR_NULL(page)) {
-				for (j = i + 1; j < nr_got; j++)
-					put_page(pages[j]);
-				nr_got = i;
+				for (j = i + 1; j < pvec->nr; j++)
+					put_page(pvec->pages[j]);
+				pvec->nr = i;
 				err = PTR_ERR_OR_ZERO(page);
 				break;
 			}
 		}
 	}
 
-	if (likely(nr_got))
-		return nr_got;
+	if (likely(pvec->nr))
+		return 0;
 	if (err)
 		return err;
 	/*
@@ -2421,11 +2421,8 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 	struct file_ra_state *ra = &filp->f_ra;
 	struct address_space *mapping = filp->f_mapping;
 	struct inode *inode = mapping->host;
-	struct page *pages[PAGEVEC_SIZE];
-	unsigned int nr_pages = min_t(unsigned int, PAGEVEC_SIZE,
-			((iocb->ki_pos + iter->count + PAGE_SIZE - 1) >> PAGE_SHIFT) -
-			(iocb->ki_pos >> PAGE_SHIFT));
-	int i, pg_nr, error = 0;
+	struct pagevec pvec;
+	int i, error = 0;
 	bool writably_mapped;
 	loff_t isize, end_offset;
 
@@ -2444,12 +2441,9 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 		if ((iocb->ki_flags & IOCB_WAITQ) && written)
 			iocb->ki_flags |= IOCB_NOWAIT;
 
-		i = 0;
-		pg_nr = filemap_get_pages(iocb, iter, pages, nr_pages);
-		if (pg_nr < 0) {
-			error = pg_nr;
+		error = filemap_get_pages(iocb, iter, &pvec);
+		if (error < 0)
 			break;
-		}
 
 		/*
 		 * i_size must be checked after we know the pages are Uptodate.
@@ -2465,9 +2459,9 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 
 		end_offset = min_t(loff_t, isize, iocb->ki_pos + iter->count);
 
-		while ((iocb->ki_pos >> PAGE_SHIFT) + pg_nr >
+		while ((iocb->ki_pos >> PAGE_SHIFT) + pvec.nr >
 		       (end_offset + PAGE_SIZE - 1) >> PAGE_SHIFT)
-			put_page(pages[--pg_nr]);
+			put_page(pvec.pages[--pvec.nr]);
 
 		/*
 		 * Once we start copying data, we don't want to be touching any
@@ -2481,11 +2475,11 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 		 */
 		if (iocb->ki_pos >> PAGE_SHIFT !=
 		    ra->prev_pos >> PAGE_SHIFT)
-			mark_page_accessed(pages[0]);
-		for (i = 1; i < pg_nr; i++)
-			mark_page_accessed(pages[i]);
+			mark_page_accessed(pvec.pages[0]);
+		for (i = 1; i < pagevec_count(&pvec); i++)
+			mark_page_accessed(pvec.pages[i]);
 
-		for (i = 0; i < pg_nr; i++) {
+		for (i = 0; i < pagevec_count(&pvec); i++) {
 			unsigned int offset = iocb->ki_pos & ~PAGE_MASK;
 			unsigned int bytes = min_t(loff_t, end_offset - iocb->ki_pos,
 						   PAGE_SIZE - offset);
@@ -2497,9 +2491,9 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 			 * before reading the page on the kernel side.
 			 */
 			if (writably_mapped)
-				flush_dcache_page(pages[i]);
+				flush_dcache_page(pvec.pages[i]);
 
-			copied = copy_page_to_iter(pages[i], offset, bytes, iter);
+			copied = copy_page_to_iter(pvec.pages[i], offset, bytes, iter);
 
 			written += copied;
 			iocb->ki_pos += copied;
@@ -2511,8 +2505,8 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 			}
 		}
 put_pages:
-		for (i = 0; i < pg_nr; i++)
-			put_page(pages[i]);
+		for (i = 0; i < pagevec_count(&pvec); i++)
+			put_page(pvec.pages[i]);
 	} while (iov_iter_count(iter) && iocb->ki_pos < isize && !error);
 
 	file_accessed(filp);
-- 
2.29.2

[PATCH v4 04/18] mm/filemap: Use head pages in generic_file_buffered_read

[Matthew Wilcox (Oracle)]

Add filemap_get_read_batch() which returns the head pages which represent
a contiguous array of bytes in the file.  It also stops when encountering
a page marked as Readahead or !Uptodate (but does return that page)
so it can be handled appropriately by filemap_get_pages().  That lets
us remove the loop in filemap_get_pages() and check only the last page.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
 mm/filemap.c | 122 +++++++++++++++++++++++++++++++++++----------------
 1 file changed, 85 insertions(+), 37 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index 9e1e2bbac8fc9..3e6fae3208906 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2168,6 +2168,51 @@ static int lock_page_for_iocb(struct kiocb *iocb, struct page *page)
 		return lock_page_killable(page);
 }
 
+/*
+ * filemap_get_read_batch - Get a batch of pages for read
+ *
+ * Get a batch of pages which represent a contiguous range of bytes
+ * in the file.  No tail pages will be returned.  If @index is in the
+ * middle of a THP, the entire THP will be returned.  The last page in
+ * the batch may have Readahead set or be not Uptodate so that the
+ * caller can take the appropriate action.
+ */
+static void filemap_get_read_batch(struct address_space *mapping,
+		pgoff_t index, pgoff_t max, struct pagevec *pvec)
+{
+	XA_STATE(xas, &mapping->i_pages, index);
+	struct page *head;
+
+	rcu_read_lock();
+	for (head = xas_load(&xas); head; head = xas_next(&xas)) {
+		if (xas_retry(&xas, head))
+			continue;
+		if (xas.xa_index > max || xa_is_value(head))
+			break;
+		if (!page_cache_get_speculative(head))
+			goto retry;
+
+		/* Has the page moved or been split? */
+		if (unlikely(head != xas_reload(&xas)))
+			goto put_page;
+
+		if (!pagevec_add(pvec, head))
+			break;
+		if (!PageUptodate(head))
+			break;
+		if (PageReadahead(head))
+			break;
+		xas.xa_index = head->index + thp_nr_pages(head) - 1;
+		xas.xa_offset = (xas.xa_index >> xas.xa_shift) & XA_CHUNK_MASK;
+		continue;
+put_page:
+		put_page(head);
+retry:
+		xas_reset(&xas);
+	}
+	rcu_read_unlock();
+}
+
 static struct page *filemap_read_page(struct kiocb *iocb, struct file *filp,
 		struct address_space *mapping, struct page *page)
 {
@@ -2321,15 +2366,15 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 	struct address_space *mapping = filp->f_mapping;
 	struct file_ra_state *ra = &filp->f_ra;
 	pgoff_t index = iocb->ki_pos >> PAGE_SHIFT;
-	pgoff_t last_index = (iocb->ki_pos + iter->count + PAGE_SIZE-1) >> PAGE_SHIFT;
-	unsigned int nr = min_t(unsigned long, last_index - index, PAGEVEC_SIZE);
-	int i, j, err = 0;
+	pgoff_t last_index;
+	int err = 0;
 
+	last_index = DIV_ROUND_UP(iocb->ki_pos + iter->count, PAGE_SIZE);
 find_page:
 	if (fatal_signal_pending(current))
 		return -EINTR;
 
-	pvec->nr = find_get_pages_contig(mapping, index, nr, pvec->pages);
+	filemap_get_read_batch(mapping, index, last_index, pvec);
 	if (pvec->nr)
 		goto got_pages;
 
@@ -2338,29 +2383,30 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 
 	page_cache_sync_readahead(mapping, ra, filp, index, last_index - index);
 
-	pvec->nr = find_get_pages_contig(mapping, index, nr, pvec->pages);
+	filemap_get_read_batch(mapping, index, last_index, pvec);
 	if (pvec->nr)
 		goto got_pages;
 
 	pvec->pages[0] = filemap_create_page(iocb, iter);
 	err = PTR_ERR_OR_ZERO(pvec->pages[0]);
-	if (!IS_ERR_OR_NULL(pvec->pages[0]))
-		pvec->nr = 1;
+	if (IS_ERR_OR_NULL(pvec->pages[0]))
+		goto err;
+	pvec->nr = 1;
+	return 0;
 got_pages:
-	for (i = 0; i < pvec->nr; i++) {
-		struct page *page = pvec->pages[i];
-		pgoff_t pg_index = index + i;
+	{
+		struct page *page = pvec->pages[pvec->nr - 1];
+		pgoff_t pg_index = page->index;
 		loff_t pg_pos = max(iocb->ki_pos,
 				    (loff_t) pg_index << PAGE_SHIFT);
 		loff_t pg_count = iocb->ki_pos + iter->count - pg_pos;
 
 		if (PageReadahead(page)) {
 			if (iocb->ki_flags & IOCB_NOIO) {
-				for (j = i; j < pvec->nr; j++)
-					put_page(pvec->pages[j]);
-				pvec->nr = i;
+				put_page(page);
+				pvec->nr--;
 				err = -EAGAIN;
-				break;
+				goto err;
 			}
 			page_cache_async_readahead(mapping, ra, filp, page,
 					pg_index, last_index - pg_index);
@@ -2368,26 +2414,23 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 
 		if (!PageUptodate(page)) {
 			if ((iocb->ki_flags & IOCB_NOWAIT) ||
-			    ((iocb->ki_flags & IOCB_WAITQ) && i)) {
-				for (j = i; j < pvec->nr; j++)
-					put_page(pvec->pages[j]);
-				pvec->nr = i;
+			    ((iocb->ki_flags & IOCB_WAITQ) && pvec->nr > 1)) {
+				put_page(page);
+				pvec->nr--;
 				err = -EAGAIN;
-				break;
+				goto err;
 			}
 
 			page = filemap_update_page(iocb, filp, iter, page,
 					pg_pos, pg_count);
 			if (IS_ERR_OR_NULL(page)) {
-				for (j = i + 1; j < pvec->nr; j++)
-					put_page(pvec->pages[j]);
-				pvec->nr = i;
+				pvec->nr--;
 				err = PTR_ERR_OR_ZERO(page);
-				break;
 			}
 		}
 	}
 
+err:
 	if (likely(pvec->nr))
 		return 0;
 	if (err)
@@ -2429,6 +2472,7 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 	if (unlikely(iocb->ki_pos >= inode->i_sb->s_maxbytes))
 		return 0;
 	iov_iter_truncate(iter, inode->i_sb->s_maxbytes);
+	pagevec_init(&pvec);
 
 	do {
 		cond_resched();
@@ -2456,13 +2500,8 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 		isize = i_size_read(inode);
 		if (unlikely(iocb->ki_pos >= isize))
 			goto put_pages;
-
 		end_offset = min_t(loff_t, isize, iocb->ki_pos + iter->count);
 
-		while ((iocb->ki_pos >> PAGE_SHIFT) + pvec.nr >
-		       (end_offset + PAGE_SIZE - 1) >> PAGE_SHIFT)
-			put_page(pvec.pages[--pvec.nr]);
-
 		/*
 		 * Once we start copying data, we don't want to be touching any
 		 * cachelines that might be contended:
@@ -2476,24 +2515,32 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 		if (iocb->ki_pos >> PAGE_SHIFT !=
 		    ra->prev_pos >> PAGE_SHIFT)
 			mark_page_accessed(pvec.pages[0]);
-		for (i = 1; i < pagevec_count(&pvec); i++)
-			mark_page_accessed(pvec.pages[i]);
 
 		for (i = 0; i < pagevec_count(&pvec); i++) {
-			unsigned int offset = iocb->ki_pos & ~PAGE_MASK;
-			unsigned int bytes = min_t(loff_t, end_offset - iocb->ki_pos,
-						   PAGE_SIZE - offset);
-			unsigned int copied;
+			struct page *page = pvec.pages[i];
+			size_t page_size = thp_size(page);
+			size_t offset = iocb->ki_pos & (page_size - 1);
+			size_t bytes = min_t(loff_t, end_offset - iocb->ki_pos,
+					     page_size - offset);
+			size_t copied;
 
+			if (end_offset < page_offset(page))
+				break;
+			if (i > 0)
+				mark_page_accessed(page);
 			/*
 			 * If users can be writing to this page using arbitrary
 			 * virtual addresses, take care about potential aliasing
 			 * before reading the page on the kernel side.
 			 */
-			if (writably_mapped)
-				flush_dcache_page(pvec.pages[i]);
+			if (writably_mapped) {
+				int j;
+
+				for (j = 0; j < thp_nr_pages(page); j++)
+					flush_dcache_page(page + j);
+			}
 
-			copied = copy_page_to_iter(pvec.pages[i], offset, bytes, iter);
+			copied = copy_page_to_iter(page, offset, bytes, iter);
 
 			written += copied;
 			iocb->ki_pos += copied;
@@ -2507,6 +2554,7 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 put_pages:
 		for (i = 0; i < pagevec_count(&pvec); i++)
 			put_page(pvec.pages[i]);
+		pagevec_reinit(&pvec);
 	} while (iov_iter_count(iter) && iocb->ki_pos < isize && !error);
 
 	file_accessed(filp);
-- 
2.29.2

[PATCH v4 05/18] mm/filemap: Pass a sleep state to put_and_wait_on_page_locked

[Matthew Wilcox (Oracle)]

This is prep work for the next patch, but I think at least one of the
current callers would prefer a killable sleep to an uninterruptible one.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/pagemap.h | 3 +--
 mm/filemap.c            | 7 +++++--
 mm/huge_memory.c        | 4 ++--
 mm/migrate.c            | 4 ++--
 4 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index 932a260a29f20..55c3b299a9e47 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -679,8 +679,7 @@ static inline int wait_on_page_locked_killable(struct page *page)
 	return wait_on_page_bit_killable(compound_head(page), PG_locked);
 }
 
-extern void put_and_wait_on_page_locked(struct page *page);
-
+int put_and_wait_on_page_locked(struct page *page, int state);
 void wait_on_page_writeback(struct page *page);
 extern void end_page_writeback(struct page *page);
 void wait_for_stable_page(struct page *page);
diff --git a/mm/filemap.c b/mm/filemap.c
index 3e6fae3208906..c8ec47f3c3a17 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1383,20 +1383,23 @@ static int wait_on_page_locked_async(struct page *page,
 /**
  * put_and_wait_on_page_locked - Drop a reference and wait for it to be unlocked
  * @page: The page to wait for.
+ * @state: The sleep state (TASK_KILLABLE, TASK_UNINTERRUPTIBLE, etc).
  *
  * The caller should hold a reference on @page.  They expect the page to
  * become unlocked relatively soon, but do not wish to hold up migration
  * (for example) by holding the reference while waiting for the page to
  * come unlocked.  After this function returns, the caller should not
  * dereference @page.
+ *
+ * Return: 0 if the page was unlocked or -EINTR if interrupted by a signal.
  */
-void put_and_wait_on_page_locked(struct page *page)
+int put_and_wait_on_page_locked(struct page *page, int state)
 {
 	wait_queue_head_t *q;
 
 	page = compound_head(page);
 	q = page_waitqueue(page);
-	wait_on_page_bit_common(q, page, PG_locked, TASK_UNINTERRUPTIBLE, DROP);
+	return wait_on_page_bit_common(q, page, PG_locked, state, DROP);
 }
 
 /**
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 08a183f6c3ab0..948e61062660b 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -1428,7 +1428,7 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t pmd)
 		if (!get_page_unless_zero(page))
 			goto out_unlock;
 		spin_unlock(vmf->ptl);
-		put_and_wait_on_page_locked(page);
+		put_and_wait_on_page_locked(page, TASK_UNINTERRUPTIBLE);
 		goto out;
 	}
 
@@ -1464,7 +1464,7 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t pmd)
 		if (!get_page_unless_zero(page))
 			goto out_unlock;
 		spin_unlock(vmf->ptl);
-		put_and_wait_on_page_locked(page);
+		put_and_wait_on_page_locked(page, TASK_UNINTERRUPTIBLE);
 		goto out;
 	}
 
diff --git a/mm/migrate.c b/mm/migrate.c
index c1585ec29827c..4d71fd5fe856b 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -335,7 +335,7 @@ void __migration_entry_wait(struct mm_struct *mm, pte_t *ptep,
 	if (!get_page_unless_zero(page))
 		goto out;
 	pte_unmap_unlock(ptep, ptl);
-	put_and_wait_on_page_locked(page);
+	put_and_wait_on_page_locked(page, TASK_UNINTERRUPTIBLE);
 	return;
 out:
 	pte_unmap_unlock(ptep, ptl);
@@ -369,7 +369,7 @@ void pmd_migration_entry_wait(struct mm_struct *mm, pmd_t *pmd)
 	if (!get_page_unless_zero(page))
 		goto unlock;
 	spin_unlock(ptl);
-	put_and_wait_on_page_locked(page);
+	put_and_wait_on_page_locked(page, TASK_UNINTERRUPTIBLE);
 	return;
 unlock:
 	spin_unlock(ptl);
-- 
2.29.2

[PATCH v4 06/18] mm/filemap: Support readpage splitting a page

[Matthew Wilcox (Oracle)]

For page splitting to succeed, the thread asking to split the
page has to be the only one with a reference to the page.  Calling
wait_on_page_locked() while holding a reference to the page will
effectively prevent this from happening with sufficient threads waiting
on the same page.  Use put_and_wait_on_page_locked() to sleep without
holding a reference to the page, then retry the page lookup after the
page is unlocked.

Since we now get the page lock a little earlier in filemap_update_page(),
we can eliminate a number of duplicate checks.  The original intent
(commit ebded02788b5 ("avoid unnecessary calls to lock_page when waiting
for IO to complete during a read")) behind getting the page lock later
was to avoid re-locking the page after it has been brought uptodate by
another thread.  We still avoid that because we go through the normal
lookup path again after the winning thread has brought the page uptodate.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
 mm/filemap.c | 76 ++++++++++++++++------------------------------------
 1 file changed, 23 insertions(+), 53 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index c8ec47f3c3a17..abbfa44d7c18a 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1372,14 +1372,6 @@ static int __wait_on_page_locked_async(struct page *page,
 	return ret;
 }
 
-static int wait_on_page_locked_async(struct page *page,
-				     struct wait_page_queue *wait)
-{
-	if (!PageLocked(page))
-		return 0;
-	return __wait_on_page_locked_async(compound_head(page), wait, false);
-}
-
 /**
  * put_and_wait_on_page_locked - Drop a reference and wait for it to be unlocked
  * @page: The page to wait for.
@@ -2276,64 +2268,42 @@ static struct page *filemap_update_page(struct kiocb *iocb, struct file *filp,
 	struct inode *inode = mapping->host;
 	int error;
 
-	/*
-	 * See comment in do_read_cache_page on why
-	 * wait_on_page_locked is used to avoid unnecessarily
-	 * serialisations and why it's safe.
-	 */
 	if (iocb->ki_flags & IOCB_WAITQ) {
-		error = wait_on_page_locked_async(page,
-						iocb->ki_waitq);
+		error = lock_page_async(page, iocb->ki_waitq);
+		if (error) {
+			put_page(page);
+			return ERR_PTR(error);
+		}
 	} else {
-		error = wait_on_page_locked_killable(page);
-	}
-	if (unlikely(error)) {
-		put_page(page);
-		return ERR_PTR(error);
+		if (!trylock_page(page)) {
+			put_and_wait_on_page_locked(page, TASK_KILLABLE);
+			return NULL;
+		}
 	}
-	if (PageUptodate(page))
-		return page;
 
+	if (!page->mapping)
+		goto truncated;
+	if (PageUptodate(page))
+		goto uptodate;
 	if (inode->i_blkbits == PAGE_SHIFT ||
 			!mapping->a_ops->is_partially_uptodate)
-		goto page_not_up_to_date;
+		goto readpage;
 	/* pipes can't handle partially uptodate pages */
 	if (unlikely(iov_iter_is_pipe(iter)))
-		goto page_not_up_to_date;
-	if (!trylock_page(page))
-		goto page_not_up_to_date;
-	/* Did it get truncated before we got the lock? */
-	if (!page->mapping)
-		goto page_not_up_to_date_locked;
+		goto readpage;
 	if (!mapping->a_ops->is_partially_uptodate(page,
-				pos & ~PAGE_MASK, count))
-		goto page_not_up_to_date_locked;
+				pos & (thp_size(page) - 1), count))
+		goto readpage;
+uptodate:
 	unlock_page(page);
 	return page;
 
-page_not_up_to_date:
-	/* Get exclusive access to the page ... */
-	error = lock_page_for_iocb(iocb, page);
-	if (unlikely(error)) {
-		put_page(page);
-		return ERR_PTR(error);
-	}
-
-page_not_up_to_date_locked:
-	/* Did it get truncated before we got the lock? */
-	if (!page->mapping) {
-		unlock_page(page);
-		put_page(page);
-		return NULL;
-	}
-
-	/* Did somebody else fill it already? */
-	if (PageUptodate(page)) {
-		unlock_page(page);
-		return page;
-	}
-
+readpage:
 	return filemap_read_page(iocb, filp, mapping, page);
+truncated:
+	unlock_page(page);
+	put_page(page);
+	return NULL;
 }
 
 static struct page *filemap_create_page(struct kiocb *iocb,
-- 
2.29.2

[PATCH v4 07/18] mm/filemap: Inline __wait_on_page_locked_async into caller

[Matthew Wilcox (Oracle)]

The previous patch removed wait_on_page_locked_async(), so inline
__wait_on_page_locked_async into __lock_page_async().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
 mm/filemap.c | 53 ++++++++++++++++++++++------------------------------
 1 file changed, 22 insertions(+), 31 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index abbfa44d7c18a..076a97dcacf1e 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -1342,36 +1342,6 @@ int wait_on_page_bit_killable(struct page *page, int bit_nr)
 }
 EXPORT_SYMBOL(wait_on_page_bit_killable);
 
-static int __wait_on_page_locked_async(struct page *page,
-				       struct wait_page_queue *wait, bool set)
-{
-	struct wait_queue_head *q = page_waitqueue(page);
-	int ret = 0;
-
-	wait->page = page;
-	wait->bit_nr = PG_locked;
-
-	spin_lock_irq(&q->lock);
-	__add_wait_queue_entry_tail(q, &wait->wait);
-	SetPageWaiters(page);
-	if (set)
-		ret = !trylock_page(page);
-	else
-		ret = PageLocked(page);
-	/*
-	 * If we were succesful now, we know we're still on the
-	 * waitqueue as we're still under the lock. This means it's
-	 * safe to remove and return success, we know the callback
-	 * isn't going to trigger.
-	 */
-	if (!ret)
-		__remove_wait_queue(q, &wait->wait);
-	else
-		ret = -EIOCBQUEUED;
-	spin_unlock_irq(&q->lock);
-	return ret;
-}
-
 /**
  * put_and_wait_on_page_locked - Drop a reference and wait for it to be unlocked
  * @page: The page to wait for.
@@ -1539,7 +1509,28 @@ EXPORT_SYMBOL_GPL(__lock_page_killable);
 
 int __lock_page_async(struct page *page, struct wait_page_queue *wait)
 {
-	return __wait_on_page_locked_async(page, wait, true);
+	struct wait_queue_head *q = page_waitqueue(page);
+	int ret = 0;
+
+	wait->page = page;
+	wait->bit_nr = PG_locked;
+
+	spin_lock_irq(&q->lock);
+	__add_wait_queue_entry_tail(q, &wait->wait);
+	SetPageWaiters(page);
+	ret = !trylock_page(page);
+	/*
+	 * If we were succesful now, we know we're still on the
+	 * waitqueue as we're still under the lock. This means it's
+	 * safe to remove and return success, we know the callback
+	 * isn't going to trigger.
+	 */
+	if (!ret)
+		__remove_wait_queue(q, &wait->wait);
+	else
+		ret = -EIOCBQUEUED;
+	spin_unlock_irq(&q->lock);
+	return ret;
 }
 
 /*
-- 
2.29.2

[PATCH v4 08/18] mm/filemap: Don't call ->readpage if IOCB_WAITQ is set

[Matthew Wilcox (Oracle)]

The readpage operation can block in many (most?) filesystems, so we
should punt to a work queue instead of calling it.  This was the last
caller of lock_page_for_iocb(), so remove it.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
 mm/filemap.c | 14 ++------------
 1 file changed, 2 insertions(+), 12 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index 076a97dcacf1e..e904e53ae90d9 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2144,16 +2144,6 @@ static void shrink_readahead_size_eio(struct file_ra_state *ra)
 	ra->ra_pages /= 4;
 }
 
-static int lock_page_for_iocb(struct kiocb *iocb, struct page *page)
-{
-	if (iocb->ki_flags & IOCB_WAITQ)
-		return lock_page_async(page, iocb->ki_waitq);
-	else if (iocb->ki_flags & IOCB_NOWAIT)
-		return trylock_page(page) ? 0 : -EAGAIN;
-	else
-		return lock_page_killable(page);
-}
-
 /*
  * filemap_get_read_batch - Get a batch of pages for read
  *
@@ -2205,7 +2195,7 @@ static struct page *filemap_read_page(struct kiocb *iocb, struct file *filp,
 	struct file_ra_state *ra = &filp->f_ra;
 	int error;
 
-	if (iocb->ki_flags & (IOCB_NOIO | IOCB_NOWAIT)) {
+	if (iocb->ki_flags & (IOCB_NOIO | IOCB_NOWAIT | IOCB_WAITQ)) {
 		unlock_page(page);
 		put_page(page);
 		return ERR_PTR(-EAGAIN);
@@ -2226,7 +2216,7 @@ static struct page *filemap_read_page(struct kiocb *iocb, struct file *filp,
 	}
 
 	if (!PageUptodate(page)) {
-		error = lock_page_for_iocb(iocb, page);
+		error = lock_page_killable(page);
 		if (unlikely(error)) {
 			put_page(page);
 			return ERR_PTR(error);
-- 
2.29.2

[PATCH v4 09/18] mm/filemap: Change filemap_read_page calling conventions

[Matthew Wilcox (Oracle)]

Make this function more generic by passing the file instead of the iocb.
Check in the callers whether we should call readpage or not.  Also make
it return an errno / 0 / AOP_TRUNCATED_PAGE, and make calling put_page()
the caller's responsibility.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
---
 mm/filemap.c | 89 +++++++++++++++++++++++++---------------------------
 1 file changed, 42 insertions(+), 47 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index e904e53ae90d9..ff5c217057d4e 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2189,56 +2189,38 @@ static void filemap_get_read_batch(struct address_space *mapping,
 	rcu_read_unlock();
 }
 
-static struct page *filemap_read_page(struct kiocb *iocb, struct file *filp,
-		struct address_space *mapping, struct page *page)
+static int filemap_read_page(struct file *file, struct address_space *mapping,
+		struct page *page)
 {
-	struct file_ra_state *ra = &filp->f_ra;
 	int error;
 
-	if (iocb->ki_flags & (IOCB_NOIO | IOCB_NOWAIT | IOCB_WAITQ)) {
-		unlock_page(page);
-		put_page(page);
-		return ERR_PTR(-EAGAIN);
-	}
-
 	/*
-	 * A previous I/O error may have been due to temporary
-	 * failures, eg. multipath errors.
-	 * PG_error will be set again if readpage fails.
+	 * A previous I/O error may have been due to temporary failures,
+	 * eg. multipath errors.  PG_error will be set again if readpage
+	 * fails.
 	 */
 	ClearPageError(page);
 	/* Start the actual read. The read will unlock the page. */
-	error = mapping->a_ops->readpage(filp, page);
-
-	if (unlikely(error)) {
-		put_page(page);
-		return error != AOP_TRUNCATED_PAGE ? ERR_PTR(error) : NULL;
-	}
+	error = mapping->a_ops->readpage(file, page);
+	if (error)
+		return error;
+	if (PageUptodate(page))
+		return 0;
 
+	error = lock_page_killable(page);
+	if (error)
+		return error;
 	if (!PageUptodate(page)) {
-		error = lock_page_killable(page);
-		if (unlikely(error)) {
-			put_page(page);
-			return ERR_PTR(error);
-		}
-		if (!PageUptodate(page)) {
-			if (page->mapping == NULL) {
-				/*
-				 * invalidate_mapping_pages got it
-				 */
-				unlock_page(page);
-				put_page(page);
-				return NULL;
-			}
-			unlock_page(page);
-			shrink_readahead_size_eio(ra);
-			put_page(page);
-			return ERR_PTR(-EIO);
+		if (page->mapping == NULL) {
+			/* page truncated */
+			error = AOP_TRUNCATED_PAGE;
+		} else {
+			shrink_readahead_size_eio(&file->f_ra);
+			error = -EIO;
 		}
-		unlock_page(page);
 	}
-
-	return page;
+	unlock_page(page);
+	return error;
 }
 
 static struct page *filemap_update_page(struct kiocb *iocb, struct file *filp,
@@ -2280,7 +2262,18 @@ static struct page *filemap_update_page(struct kiocb *iocb, struct file *filp,
 	return page;
 
 readpage:
-	return filemap_read_page(iocb, filp, mapping, page);
+	if (iocb->ki_flags & (IOCB_NOIO | IOCB_NOWAIT | IOCB_WAITQ)) {
+		unlock_page(page);
+		put_page(page);
+		return ERR_PTR(-EAGAIN);
+	}
+	error = filemap_read_page(iocb->ki_filp, mapping, page);
+	if (!error)
+		return page;
+	put_page(page);
+	if (error == AOP_TRUNCATED_PAGE)
+		return NULL;
+	return ERR_PTR(error);
 truncated:
 	unlock_page(page);
 	put_page(page);
@@ -2296,7 +2289,7 @@ static struct page *filemap_create_page(struct kiocb *iocb,
 	struct page *page;
 	int error;
 
-	if (iocb->ki_flags & IOCB_NOIO)
+	if (iocb->ki_flags & (IOCB_NOIO | IOCB_NOWAIT | IOCB_WAITQ))
 		return ERR_PTR(-EAGAIN);
 
 	page = page_cache_alloc(mapping);
@@ -2305,12 +2298,14 @@ static struct page *filemap_create_page(struct kiocb *iocb,
 
 	error = add_to_page_cache_lru(page, mapping, index,
 				      mapping_gfp_constraint(mapping, GFP_KERNEL));
-	if (error) {
-		put_page(page);
-		return error != -EEXIST ? ERR_PTR(error) : NULL;
-	}
-
-	return filemap_read_page(iocb, filp, mapping, page);
+	if (!error)
+		error = filemap_read_page(iocb->ki_filp, mapping, page);
+	if (!error)
+		return page;
+	put_page(page);
+	if (error == -EEXIST || error == AOP_TRUNCATED_PAGE)
+		return NULL;
+	return ERR_PTR(error);
 }
 
 static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
-- 
2.29.2

[PATCH v4 10/18] mm/filemap: Change filemap_create_page calling conventions

[Matthew Wilcox (Oracle)]

By moving the iocb flag checks to the caller, we can pass the
file and the page index instead of the iocb.  It never needed the iter.
By passing the pagevec, we can return an errno (or AOP_TRUNCATED_PAGE)
instead of an ERR_PTR.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
 mm/filemap.c | 53 ++++++++++++++++++++++++++--------------------------
 1 file changed, 27 insertions(+), 26 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index ff5c217057d4e..602191fe6c53f 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2280,32 +2280,33 @@ static struct page *filemap_update_page(struct kiocb *iocb, struct file *filp,
 	return NULL;
 }
 
-static struct page *filemap_create_page(struct kiocb *iocb,
-		struct iov_iter *iter)
+static int filemap_create_page(struct file *file,
+		struct address_space *mapping, pgoff_t index,
+		struct pagevec *pvec)
 {
-	struct file *filp = iocb->ki_filp;
-	struct address_space *mapping = filp->f_mapping;
-	pgoff_t index = iocb->ki_pos >> PAGE_SHIFT;
 	struct page *page;
 	int error;
 
-	if (iocb->ki_flags & (IOCB_NOIO | IOCB_NOWAIT | IOCB_WAITQ))
-		return ERR_PTR(-EAGAIN);
-
 	page = page_cache_alloc(mapping);
 	if (!page)
-		return ERR_PTR(-ENOMEM);
+		return -ENOMEM;
 
 	error = add_to_page_cache_lru(page, mapping, index,
-				      mapping_gfp_constraint(mapping, GFP_KERNEL));
-	if (!error)
-		error = filemap_read_page(iocb->ki_filp, mapping, page);
-	if (!error)
-		return page;
+			mapping_gfp_constraint(mapping, GFP_KERNEL));
+	if (error == -EEXIST)
+		error = AOP_TRUNCATED_PAGE;
+	if (error)
+		goto error;
+
+	error = filemap_read_page(file, mapping, page);
+	if (error)
+		goto error;
+
+	pagevec_add(pvec, page);
+	return 0;
+error:
 	put_page(page);
-	if (error == -EEXIST || error == AOP_TRUNCATED_PAGE)
-		return NULL;
-	return ERR_PTR(error);
+	return error;
 }
 
 static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
@@ -2333,15 +2334,15 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 	page_cache_sync_readahead(mapping, ra, filp, index, last_index - index);
 
 	filemap_get_read_batch(mapping, index, last_index, pvec);
-	if (pvec->nr)
-		goto got_pages;
-
-	pvec->pages[0] = filemap_create_page(iocb, iter);
-	err = PTR_ERR_OR_ZERO(pvec->pages[0]);
-	if (IS_ERR_OR_NULL(pvec->pages[0]))
-		goto err;
-	pvec->nr = 1;
-	return 0;
+	if (!pagevec_count(pvec)) {
+		if (iocb->ki_flags & (IOCB_NOWAIT | IOCB_WAITQ))
+			return -EAGAIN;
+		err = filemap_create_page(filp, mapping,
+				iocb->ki_pos >> PAGE_SHIFT, pvec);
+		if (err == AOP_TRUNCATED_PAGE)
+			goto find_page;
+		return err;
+	}
 got_pages:
 	{
 		struct page *page = pvec->pages[pvec->nr - 1];
-- 
2.29.2

[PATCH v4 11/18] mm/filemap: Convert filemap_update_page to return an errno

[Matthew Wilcox (Oracle)]

Use AOP_TRUNCATED_PAGE to indicate that no error occurred, but the
page we looked up is no longer valid.  In this case, the reference
to the page will have been removed; if we hit any other error, the
caller will release the reference.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
---
 mm/filemap.c | 38 +++++++++++++++++---------------------
 1 file changed, 17 insertions(+), 21 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index 602191fe6c53f..adac12ed48415 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2223,24 +2223,21 @@ static int filemap_read_page(struct file *file, struct address_space *mapping,
 	return error;
 }
 
-static struct page *filemap_update_page(struct kiocb *iocb, struct file *filp,
-		struct iov_iter *iter, struct page *page, loff_t pos,
-		loff_t count)
+static int filemap_update_page(struct kiocb *iocb,
+		struct address_space *mapping, struct iov_iter *iter,
+		struct page *page, loff_t pos, loff_t count)
 {
-	struct address_space *mapping = filp->f_mapping;
 	struct inode *inode = mapping->host;
 	int error;
 
 	if (iocb->ki_flags & IOCB_WAITQ) {
 		error = lock_page_async(page, iocb->ki_waitq);
-		if (error) {
-			put_page(page);
-			return ERR_PTR(error);
-		}
+		if (error)
+			return error;
 	} else {
 		if (!trylock_page(page)) {
 			put_and_wait_on_page_locked(page, TASK_KILLABLE);
-			return NULL;
+			return AOP_TRUNCATED_PAGE;
 		}
 	}
 
@@ -2259,25 +2256,21 @@ static struct page *filemap_update_page(struct kiocb *iocb, struct file *filp,
 		goto readpage;
 uptodate:
 	unlock_page(page);
-	return page;
+	return 0;
 
 readpage:
 	if (iocb->ki_flags & (IOCB_NOIO | IOCB_NOWAIT | IOCB_WAITQ)) {
 		unlock_page(page);
-		put_page(page);
-		return ERR_PTR(-EAGAIN);
+		return -EAGAIN;
 	}
 	error = filemap_read_page(iocb->ki_filp, mapping, page);
-	if (!error)
-		return page;
-	put_page(page);
 	if (error == AOP_TRUNCATED_PAGE)
-		return NULL;
-	return ERR_PTR(error);
+		put_page(page);
+	return error;
 truncated:
 	unlock_page(page);
 	put_page(page);
-	return NULL;
+	return AOP_TRUNCATED_PAGE;
 }
 
 static int filemap_create_page(struct file *file,
@@ -2371,11 +2364,12 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 				goto err;
 			}
 
-			page = filemap_update_page(iocb, filp, iter, page,
+			err = filemap_update_page(iocb, mapping, iter, page,
 					pg_pos, pg_count);
-			if (IS_ERR_OR_NULL(page)) {
+			if (err) {
+				if (err < 0)
+					put_page(page);
 				pvec->nr--;
-				err = PTR_ERR_OR_ZERO(page);
 			}
 		}
 	}
@@ -2383,6 +2377,8 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 err:
 	if (likely(pvec->nr))
 		return 0;
+	if (err == AOP_TRUNCATED_PAGE)
+		goto find_page;
 	if (err)
 		return err;
 	/*
-- 
2.29.2

[PATCH v4 12/18] mm/filemap: Move the iocb checks into filemap_update_page

[Matthew Wilcox (Oracle)]

We don't need to give up when a non-blocking request sees a !Uptodate page.
We may be able to satisfy the read from a partially-uptodate page.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
 mm/filemap.c | 24 ++++++++++--------------
 1 file changed, 10 insertions(+), 14 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index adac12ed48415..141d5917770d6 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2230,15 +2230,16 @@ static int filemap_update_page(struct kiocb *iocb,
 	struct inode *inode = mapping->host;
 	int error;
 
-	if (iocb->ki_flags & IOCB_WAITQ) {
-		error = lock_page_async(page, iocb->ki_waitq);
-		if (error)
-			return error;
-	} else {
-		if (!trylock_page(page)) {
+	if (!trylock_page(page)) {
+		if (iocb->ki_flags & (IOCB_NOWAIT | IOCB_NOIO))
+			return -EAGAIN;
+		if (!(iocb->ki_flags & IOCB_WAITQ)) {
 			put_and_wait_on_page_locked(page, TASK_KILLABLE);
 			return AOP_TRUNCATED_PAGE;
 		}
+		error = __lock_page_async(page, iocb->ki_waitq);
+		if (error)
+			return error;
 	}
 
 	if (!page->mapping)
@@ -2356,14 +2357,9 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 		}
 
 		if (!PageUptodate(page)) {
-			if ((iocb->ki_flags & IOCB_NOWAIT) ||
-			    ((iocb->ki_flags & IOCB_WAITQ) && pvec->nr > 1)) {
-				put_page(page);
-				pvec->nr--;
-				err = -EAGAIN;
-				goto err;
-			}
-
+			if ((iocb->ki_flags & IOCB_WAITQ) &&
+			    pagevec_count(pvec) > 1)
+				iocb->ki_flags |= IOCB_NOWAIT;
 			err = filemap_update_page(iocb, mapping, iter, page,
 					pg_pos, pg_count);
 			if (err) {
-- 
2.29.2

[PATCH v4 13/18] mm/filemap: Add filemap_range_uptodate

[Matthew Wilcox (Oracle)]

Move the complicated condition and the calculations out of
filemap_update_page() into its own function.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
---
 mm/filemap.c | 68 ++++++++++++++++++++++++++++++----------------------
 1 file changed, 40 insertions(+), 28 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index 141d5917770d6..df755b86a0a7b 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2223,11 +2223,36 @@ static int filemap_read_page(struct file *file, struct address_space *mapping,
 	return error;
 }
 
+static bool filemap_range_uptodate(struct address_space *mapping,
+		loff_t pos, struct iov_iter *iter, struct page *page)
+{
+	int count;
+
+	if (PageUptodate(page))
+		return true;
+	/* pipes can't handle partially uptodate pages */
+	if (iov_iter_is_pipe(iter))
+		return false;
+	if (!mapping->a_ops->is_partially_uptodate)
+		return false;
+	if (mapping->host->i_blkbits >= (PAGE_SHIFT + thp_order(page)))
+		return false;
+
+	count = iter->count;
+	if (page_offset(page) > pos) {
+		count -= page_offset(page) - pos;
+		pos = 0;
+	} else {
+		pos -= page_offset(page);
+	}
+
+	return mapping->a_ops->is_partially_uptodate(page, pos, count);
+}
+
 static int filemap_update_page(struct kiocb *iocb,
 		struct address_space *mapping, struct iov_iter *iter,
-		struct page *page, loff_t pos, loff_t count)
+		struct page *page)
 {
-	struct inode *inode = mapping->host;
 	int error;
 
 	if (!trylock_page(page)) {
@@ -2244,34 +2269,25 @@ static int filemap_update_page(struct kiocb *iocb,
 
 	if (!page->mapping)
 		goto truncated;
-	if (PageUptodate(page))
-		goto uptodate;
-	if (inode->i_blkbits == PAGE_SHIFT ||
-			!mapping->a_ops->is_partially_uptodate)
-		goto readpage;
-	/* pipes can't handle partially uptodate pages */
-	if (unlikely(iov_iter_is_pipe(iter)))
-		goto readpage;
-	if (!mapping->a_ops->is_partially_uptodate(page,
-				pos & (thp_size(page) - 1), count))
-		goto readpage;
-uptodate:
-	unlock_page(page);
-	return 0;
 
-readpage:
-	if (iocb->ki_flags & (IOCB_NOIO | IOCB_NOWAIT | IOCB_WAITQ)) {
-		unlock_page(page);
-		return -EAGAIN;
-	}
+	error = 0;
+	if (filemap_range_uptodate(mapping, iocb->ki_pos, iter, page))
+		goto unlock;
+
+	error = -EAGAIN;
+	if (iocb->ki_flags & (IOCB_NOIO | IOCB_NOWAIT | IOCB_WAITQ))
+		goto unlock;
+
 	error = filemap_read_page(iocb->ki_filp, mapping, page);
 	if (error == AOP_TRUNCATED_PAGE)
 		put_page(page);
 	return error;
 truncated:
-	unlock_page(page);
+	error = AOP_TRUNCATED_PAGE;
 	put_page(page);
-	return AOP_TRUNCATED_PAGE;
+unlock:
+	unlock_page(page);
+	return error;
 }
 
 static int filemap_create_page(struct file *file,
@@ -2341,9 +2357,6 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 	{
 		struct page *page = pvec->pages[pvec->nr - 1];
 		pgoff_t pg_index = page->index;
-		loff_t pg_pos = max(iocb->ki_pos,
-				    (loff_t) pg_index << PAGE_SHIFT);
-		loff_t pg_count = iocb->ki_pos + iter->count - pg_pos;
 
 		if (PageReadahead(page)) {
 			if (iocb->ki_flags & IOCB_NOIO) {
@@ -2360,8 +2373,7 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 			if ((iocb->ki_flags & IOCB_WAITQ) &&
 			    pagevec_count(pvec) > 1)
 				iocb->ki_flags |= IOCB_NOWAIT;
-			err = filemap_update_page(iocb, mapping, iter, page,
-					pg_pos, pg_count);
+			err = filemap_update_page(iocb, mapping, iter, page);
 			if (err) {
 				if (err < 0)
 					put_page(page);
-- 
2.29.2

[PATCH v4 14/18] mm/filemap: Split filemap_readahead out of filemap_get_pages

[Matthew Wilcox (Oracle)]

This simplifies the error handling.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
---
 mm/filemap.c | 19 ++++++++++++++-----
 1 file changed, 14 insertions(+), 5 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index df755b86a0a7b..f1c5c7d7aae8e 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2319,6 +2319,17 @@ static int filemap_create_page(struct file *file,
 	return error;
 }
 
+static int filemap_readahead(struct kiocb *iocb, struct file *file,
+		struct address_space *mapping, struct page *page,
+		pgoff_t last_index)
+{
+	if (iocb->ki_flags & IOCB_NOIO)
+		return -EAGAIN;
+	page_cache_async_readahead(mapping, &file->f_ra, file, page,
+			page->index, last_index - page->index);
+	return 0;
+}
+
 static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 		struct pagevec *pvec)
 {
@@ -2356,17 +2367,15 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 got_pages:
 	{
 		struct page *page = pvec->pages[pvec->nr - 1];
-		pgoff_t pg_index = page->index;
 
 		if (PageReadahead(page)) {
-			if (iocb->ki_flags & IOCB_NOIO) {
+			err = filemap_readahead(iocb, filp, mapping, page,
+					last_index);
+			if (err) {
 				put_page(page);
 				pvec->nr--;
-				err = -EAGAIN;
 				goto err;
 			}
-			page_cache_async_readahead(mapping, ra, filp, page,
-					pg_index, last_index - pg_index);
 		}
 
 		if (!PageUptodate(page)) {
-- 
2.29.2

[PATCH v4 15/18] mm/filemap: Restructure filemap_get_pages

[Matthew Wilcox (Oracle)]

Remove the got_pages label, remove indentation, rename find_page to retry,
simplify error handling.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
---
 mm/filemap.c | 71 +++++++++++++++++++++-------------------------------
 1 file changed, 28 insertions(+), 43 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index f1c5c7d7aae8e..e09f0f1209a90 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2338,70 +2338,55 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 	struct file_ra_state *ra = &filp->f_ra;
 	pgoff_t index = iocb->ki_pos >> PAGE_SHIFT;
 	pgoff_t last_index;
+	struct page *page;
 	int err = 0;
 
 	last_index = DIV_ROUND_UP(iocb->ki_pos + iter->count, PAGE_SIZE);
-find_page:
+retry:
 	if (fatal_signal_pending(current))
 		return -EINTR;
 
-	filemap_get_read_batch(mapping, index, last_index, pvec);
-	if (pvec->nr)
-		goto got_pages;
-
-	if (iocb->ki_flags & IOCB_NOIO)
-		return -EAGAIN;
-
-	page_cache_sync_readahead(mapping, ra, filp, index, last_index - index);
-
 	filemap_get_read_batch(mapping, index, last_index, pvec);
+	if (!pagevec_count(pvec)) {
+		if (iocb->ki_flags & IOCB_NOIO)
+			return -EAGAIN;
+		page_cache_sync_readahead(mapping, ra, filp, index,
+				last_index - index);
+		filemap_get_read_batch(mapping, index, last_index, pvec);
+	}
 	if (!pagevec_count(pvec)) {
 		if (iocb->ki_flags & (IOCB_NOWAIT | IOCB_WAITQ))
 			return -EAGAIN;
 		err = filemap_create_page(filp, mapping,
 				iocb->ki_pos >> PAGE_SHIFT, pvec);
 		if (err == AOP_TRUNCATED_PAGE)
-			goto find_page;
+			goto retry;
 		return err;
 	}
-got_pages:
-	{
-		struct page *page = pvec->pages[pvec->nr - 1];
-
-		if (PageReadahead(page)) {
-			err = filemap_readahead(iocb, filp, mapping, page,
-					last_index);
-			if (err) {
-				put_page(page);
-				pvec->nr--;
-				goto err;
-			}
-		}
 
-		if (!PageUptodate(page)) {
-			if ((iocb->ki_flags & IOCB_WAITQ) &&
-			    pagevec_count(pvec) > 1)
-				iocb->ki_flags |= IOCB_NOWAIT;
-			err = filemap_update_page(iocb, mapping, iter, page);
-			if (err) {
-				if (err < 0)
-					put_page(page);
-				pvec->nr--;
-			}
-		}
+	page = pvec->pages[pagevec_count(pvec) - 1];
+	if (PageReadahead(page)) {
+		err = filemap_readahead(iocb, filp, mapping, page, last_index);
+		if (err)
+			goto err;
+	}
+	if (!PageUptodate(page)) {
+		if ((iocb->ki_flags & IOCB_WAITQ) && pagevec_count(pvec) > 1)
+			iocb->ki_flags |= IOCB_NOWAIT;
+		err = filemap_update_page(iocb, mapping, iter, page);
+		if (err)
+			goto err;
 	}
 
+	return 0;
 err:
-	if (likely(pvec->nr))
+	if (err < 0)
+		put_page(page);
+	if (likely(--pvec->nr))
 		return 0;
 	if (err == AOP_TRUNCATED_PAGE)
-		goto find_page;
-	if (err)
-		return err;
-	/*
-	 * No pages and no error means we raced and should retry:
-	 */
-	goto find_page;
+		goto retry;
+	return err;
 }
 
 /**
-- 
2.29.2

[PATCH v4 16/18] mm/filemap: Don't relock the page after calling readpage

[Matthew Wilcox (Oracle)]

We don't need to get the page lock again; we just need to wait for
the I/O to finish, so use wait_on_page_locked_killable() like the
other callers of ->readpage.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
---
 mm/filemap.c | 21 +++++++--------------
 1 file changed, 7 insertions(+), 14 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index e09f0f1209a90..5563bfe5e544b 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2204,23 +2204,16 @@ static int filemap_read_page(struct file *file, struct address_space *mapping,
 	error = mapping->a_ops->readpage(file, page);
 	if (error)
 		return error;
-	if (PageUptodate(page))
-		return 0;
 
-	error = lock_page_killable(page);
+	error = wait_on_page_locked_killable(page);
 	if (error)
 		return error;
-	if (!PageUptodate(page)) {
-		if (page->mapping == NULL) {
-			/* page truncated */
-			error = AOP_TRUNCATED_PAGE;
-		} else {
-			shrink_readahead_size_eio(&file->f_ra);
-			error = -EIO;
-		}
-	}
-	unlock_page(page);
-	return error;
+	if (PageUptodate(page))
+		return 0;
+	if (!page->mapping)	/* page truncated */
+		return AOP_TRUNCATED_PAGE;
+	shrink_readahead_size_eio(&file->f_ra);
+	return -EIO;
 }
 
 static bool filemap_range_uptodate(struct address_space *mapping,
-- 
2.29.2

[PATCH v4 17/18] mm/filemap: Rename generic_file_buffered_read to filemap_read

[Matthew Wilcox (Oracle)]

From: Christoph Hellwig <hch@lst.de>

Rename generic_file_buffered_read to match the naming of filemap_fault,
also update the written parameter to a more descriptive name and
improve the kerneldoc comment.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
---
 fs/btrfs/file.c    |  2 +-
 include/linux/fs.h |  4 ++--
 mm/filemap.c       | 35 ++++++++++++++++-------------------
 3 files changed, 19 insertions(+), 22 deletions(-)

diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 87355a38a6547..1a4913e1fd128 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -3633,7 +3633,7 @@ static ssize_t btrfs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
 			return ret;
 	}
 
-	return generic_file_buffered_read(iocb, to, ret);
+	return filemap_read(iocb, to, ret);
 }
 
 const struct file_operations btrfs_file_operations = {
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 8d559d43f2af9..a79f65607236a 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2948,8 +2948,8 @@ extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
 extern int generic_write_check_limits(struct file *file, loff_t pos,
 		loff_t *count);
 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out);
-extern ssize_t generic_file_buffered_read(struct kiocb *iocb,
-		struct iov_iter *to, ssize_t already_read);
+ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to,
+		ssize_t already_read);
 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
diff --git a/mm/filemap.c b/mm/filemap.c
index 5563bfe5e544b..50d87aff0ef86 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2383,23 +2383,20 @@ static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
 }
 
 /**
- * generic_file_buffered_read - generic file read routine
- * @iocb:	the iocb to read
- * @iter:	data destination
- * @written:	already copied
+ * filemap_read - Read data from the page cache.
+ * @iocb: The iocb to read.
+ * @iter: Destination for the data.
+ * @already_read: Number of bytes already read by the caller.
  *
- * This is a generic file read routine, and uses the
- * mapping->a_ops->readpage() function for the actual low-level stuff.
+ * Copies data from the page cache.  If the data is not currently present,
+ * uses the readahead and readpage address_space operations to fetch it.
  *
- * This is really ugly. But the goto's actually try to clarify some
- * of the logic when it comes to error handling etc.
- *
- * Return:
- * * total number of bytes copied, including those the were already @written
- * * negative error code if nothing was copied
+ * Return: Total number of bytes copied, including those already read by
+ * the caller.  If an error happens before any bytes are copied, returns
+ * a negative error number.
  */
-ssize_t generic_file_buffered_read(struct kiocb *iocb,
-		struct iov_iter *iter, ssize_t written)
+ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *iter,
+		ssize_t already_read)
 {
 	struct file *filp = iocb->ki_filp;
 	struct file_ra_state *ra = &filp->f_ra;
@@ -2423,7 +2420,7 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 		 * can no longer safely return -EIOCBQUEUED. Hence mark
 		 * an async read NOWAIT at that point.
 		 */
-		if ((iocb->ki_flags & IOCB_WAITQ) && written)
+		if ((iocb->ki_flags & IOCB_WAITQ) && already_read)
 			iocb->ki_flags |= IOCB_NOWAIT;
 
 		error = filemap_get_pages(iocb, iter, &pvec);
@@ -2483,7 +2480,7 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 
 			copied = copy_page_to_iter(page, offset, bytes, iter);
 
-			written += copied;
+			already_read += copied;
 			iocb->ki_pos += copied;
 			ra->prev_pos = iocb->ki_pos;
 
@@ -2500,9 +2497,9 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 
 	file_accessed(filp);
 
-	return written ? written : error;
+	return already_read ? already_read : error;
 }
-EXPORT_SYMBOL_GPL(generic_file_buffered_read);
+EXPORT_SYMBOL_GPL(filemap_read);
 
 /**
  * generic_file_read_iter - generic filesystem read routine
@@ -2576,7 +2573,7 @@ generic_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
 			goto out;
 	}
 
-	retval = generic_file_buffered_read(iocb, iter, retval);
+	retval = filemap_read(iocb, iter, retval);
 out:
 	return retval;
 }
-- 
2.29.2

[PATCH v4 18/18] mm/filemap: Simplify generic_file_read_iter

[Matthew Wilcox (Oracle)]

From: Christoph Hellwig <hch@lst.de>

Avoid the pointless goto out just for returning retval.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
---
 mm/filemap.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index 50d87aff0ef86..bcaaaf7f55ab7 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2529,7 +2529,7 @@ generic_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
 	ssize_t retval = 0;
 
 	if (!count)
-		goto out; /* skip atime */
+		return 0; /* skip atime */
 
 	if (iocb->ki_flags & IOCB_DIRECT) {
 		struct file *file = iocb->ki_filp;
@@ -2547,7 +2547,7 @@ generic_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
 						iocb->ki_pos,
 					        iocb->ki_pos + count - 1);
 			if (retval < 0)
-				goto out;
+				return retval;
 		}
 
 		file_accessed(file);
@@ -2570,12 +2570,10 @@ generic_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
 		 */
 		if (retval < 0 || !count || iocb->ki_pos >= size ||
 		    IS_DAX(inode))
-			goto out;
+			return retval;
 	}
 
-	retval = filemap_read(iocb, iter, retval);
-out:
-	return retval;
+	return filemap_read(iocb, iter, retval);
 }
 EXPORT_SYMBOL(generic_file_read_iter);
 
-- 
2.29.2

Re: [PATCH v4 01/18] mm/filemap: Rename generic_file_buffered_read subfunctions

[Christoph Hellwig]

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 02/18] mm/filemap: Remove dynamically allocated array from filemap_read

[Christoph Hellwig]

On Thu, Jan 21, 2021 at 04:16:00AM +0000, Matthew Wilcox (Oracle) wrote:
> Increasing the batch size runs into diminishing returns.  It's probably
> better to make, eg, three calls to filemap_get_pages() than it is to
> call into kmalloc().
> 
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 03/18] mm/filemap: Convert filemap_get_pages to take a pagevec

[Christoph Hellwig]

On Thu, Jan 21, 2021 at 04:16:01AM +0000, Matthew Wilcox (Oracle) wrote:
> Using a pagevec lets us keep the pages and the number of pages together
> which simplifies a lot of the calling conventions.
> 
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 09/18] mm/filemap: Change filemap_read_page calling conventions

[Christoph Hellwig]

On Thu, Jan 21, 2021 at 04:16:07AM +0000, Matthew Wilcox (Oracle) wrote:
> Make this function more generic by passing the file instead of the iocb.
> Check in the callers whether we should call readpage or not.  Also make
> it return an errno / 0 / AOP_TRUNCATED_PAGE, and make calling put_page()
> the caller's responsibility.

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 11/18] mm/filemap: Convert filemap_update_page to return an errno

[Christoph Hellwig]

On Thu, Jan 21, 2021 at 04:16:09AM +0000, Matthew Wilcox (Oracle) wrote:
> Use AOP_TRUNCATED_PAGE to indicate that no error occurred, but the
> page we looked up is no longer valid.  In this case, the reference
> to the page will have been removed; if we hit any other error, the
> caller will release the reference.

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 13/18] mm/filemap: Add filemap_range_uptodate

[Christoph Hellwig]

On Thu, Jan 21, 2021 at 04:16:11AM +0000, Matthew Wilcox (Oracle) wrote:
> Move the complicated condition and the calculations out of
> filemap_update_page() into its own function.

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 14/18] mm/filemap: Split filemap_readahead out of filemap_get_pages

[Christoph Hellwig]

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 15/18] mm/filemap: Restructure filemap_get_pages

[Christoph Hellwig]

On Thu, Jan 21, 2021 at 04:16:13AM +0000, Matthew Wilcox (Oracle) wrote:
> Remove the got_pages label, remove indentation, rename find_page to retry,
> simplify error handling.

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 16/18] mm/filemap: Don't relock the page after calling readpage

[Christoph Hellwig]

On Thu, Jan 21, 2021 at 04:16:14AM +0000, Matthew Wilcox (Oracle) wrote:
> We don't need to get the page lock again; we just need to wait for
> the I/O to finish, so use wait_on_page_locked_killable() like the
> other callers of ->readpage.

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v4 01/18] mm/filemap: Rename generic_file_buffered_read subfunctions

[Miaohe Lin]

Hi:
On 2021/1/21 12:15, Matthew Wilcox (Oracle) wrote:
> The recent split of generic_file_buffered_read() created some very
> long function names which are hard to distinguish from each other.
> Rename as follows:
> 
> generic_file_buffered_read_readpage -> filemap_read_page
> generic_file_buffered_read_pagenotuptodate -> filemap_update_page
> generic_file_buffered_read_no_cached_page -> filemap_create_page
> generic_file_buffered_read_get_pages -> filemap_get_pages
> > Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
> ---
>  mm/filemap.c | 44 +++++++++++++++-----------------------------
>  1 file changed, 15 insertions(+), 29 deletions(-)
> 
> diff --git a/mm/filemap.c b/mm/filemap.c
> index d90614f501daa..934e04f1760ef 100644
> --- a/mm/filemap.c
> +++ b/mm/filemap.c
> @@ -2168,11 +2168,8 @@ static int lock_page_for_iocb(struct kiocb *iocb, struct page *page)
>  		return lock_page_killable(page);
>  }
>  
> -static struct page *
> -generic_file_buffered_read_readpage(struct kiocb *iocb,
> -				    struct file *filp,
> -				    struct address_space *mapping,
> -				    struct page *page)
> +static struct page *filemap_read_page(struct kiocb *iocb, struct file *filp,
> +		struct address_space *mapping, struct page *page)
>  {
>  	struct file_ra_state *ra = &filp->f_ra;
>  	int error;
> @@ -2223,12 +2220,9 @@ generic_file_buffered_read_readpage(struct kiocb *iocb,
>  	return page;
>  }
>  
> -static struct page *
> -generic_file_buffered_read_pagenotuptodate(struct kiocb *iocb,
> -					   struct file *filp,
> -					   struct iov_iter *iter,
> -					   struct page *page,
> -					   loff_t pos, loff_t count)
> +static struct page *filemap_update_page(struct kiocb *iocb, struct file *filp,
> +		struct iov_iter *iter, struct page *page, loff_t pos,
> +		loff_t count)
>  {
>  	struct address_space *mapping = filp->f_mapping;
>  	struct inode *inode = mapping->host;
> @@ -2291,12 +2285,11 @@ generic_file_buffered_read_pagenotuptodate(struct kiocb *iocb,
>  		return page;
>  	}
>  
> -	return generic_file_buffered_read_readpage(iocb, filp, mapping, page);
> +	return filemap_read_page(iocb, filp, mapping, page);
>  }
>  
> -static struct page *
> -generic_file_buffered_read_no_cached_page(struct kiocb *iocb,
> -					  struct iov_iter *iter)
> +static struct page *filemap_create_page(struct kiocb *iocb,
> +		struct iov_iter *iter)
>  {
>  	struct file *filp = iocb->ki_filp;
>  	struct address_space *mapping = filp->f_mapping;
> @@ -2307,10 +2300,6 @@ generic_file_buffered_read_no_cached_page(struct kiocb *iocb,
>  	if (iocb->ki_flags & IOCB_NOIO)
>  		return ERR_PTR(-EAGAIN);
>  
> -	/*
> -	 * Ok, it wasn't cached, so we need to create a new
> -	 * page..
> -	 */
>  	page = page_cache_alloc(mapping);
>  	if (!page)
>  		return ERR_PTR(-ENOMEM);
> @@ -2322,13 +2311,11 @@ generic_file_buffered_read_no_cached_page(struct kiocb *iocb,
>  		return error != -EEXIST ? ERR_PTR(error) : NULL;
>  	}
>  
> -	return generic_file_buffered_read_readpage(iocb, filp, mapping, page);
> +	return filemap_read_page(iocb, filp, mapping, page);
>  }
>  
> -static int generic_file_buffered_read_get_pages(struct kiocb *iocb,
> -						struct iov_iter *iter,
> -						struct page **pages,
> -						unsigned int nr)
> +static int filemap_get_pages(struct kiocb *iocb, struct iov_iter *iter,
> +		struct page **pages, unsigned int nr)
>  {
>  	struct file *filp = iocb->ki_filp;
>  	struct address_space *mapping = filp->f_mapping;
> @@ -2355,7 +2342,7 @@ static int generic_file_buffered_read_get_pages(struct kiocb *iocb,
>  	if (nr_got)
>  		goto got_pages;
>  
> -	pages[0] = generic_file_buffered_read_no_cached_page(iocb, iter);
> +	pages[0] = filemap_create_page(iocb, iter);
>  	err = PTR_ERR_OR_ZERO(pages[0]);
>  	if (!IS_ERR_OR_NULL(pages[0]))
>  		nr_got = 1;
> @@ -2389,8 +2376,8 @@ static int generic_file_buffered_read_get_pages(struct kiocb *iocb,
>  				break;
>  			}
>  
> -			page = generic_file_buffered_read_pagenotuptodate(iocb,
> -					filp, iter, page, pg_pos, pg_count);
> +			page = filemap_update_page(iocb, filp, iter, page,
> +					pg_pos, pg_count);
>  			if (IS_ERR_OR_NULL(page)) {
>  				for (j = i + 1; j < nr_got; j++)
>  					put_page(pages[j]);
> @@ -2466,8 +2453,7 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
>  			iocb->ki_flags |= IOCB_NOWAIT;
>  
>  		i = 0;
> -		pg_nr = generic_file_buffered_read_get_pages(iocb, iter,
> -							     pages, nr_pages);
> +		pg_nr = filemap_get_pages(iocb, iter, pages, nr_pages);
>  		if (pg_nr < 0) {
>  			error = pg_nr;
>  			break;
> 

Looks good to me. Thanks.

Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>

Re: [PATCH v4 02/18] mm/filemap: Remove dynamically allocated array from filemap_read

[Miaohe Lin]

Hi:
On 2021/1/21 12:16, Matthew Wilcox (Oracle) wrote:
> Increasing the batch size runs into diminishing returns.  It's probably
> better to make, eg, three calls to filemap_get_pages() than it is to
> call into kmalloc().
> 

LGTM. Thanks.

Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>

> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> ---
>  mm/filemap.c | 15 ++-------------
>  1 file changed, 2 insertions(+), 13 deletions(-)
> 
> diff --git a/mm/filemap.c b/mm/filemap.c
> index 934e04f1760ef..5dec04c8e16b0 100644
> --- a/mm/filemap.c
> +++ b/mm/filemap.c
> @@ -2421,8 +2421,8 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
>  	struct file_ra_state *ra = &filp->f_ra;
>  	struct address_space *mapping = filp->f_mapping;
>  	struct inode *inode = mapping->host;
> -	struct page *pages_onstack[PAGEVEC_SIZE], **pages = NULL;
> -	unsigned int nr_pages = min_t(unsigned int, 512,
> +	struct page *pages[PAGEVEC_SIZE];
> +	unsigned int nr_pages = min_t(unsigned int, PAGEVEC_SIZE,
>  			((iocb->ki_pos + iter->count + PAGE_SIZE - 1) >> PAGE_SHIFT) -
>  			(iocb->ki_pos >> PAGE_SHIFT));
>  	int i, pg_nr, error = 0;
> @@ -2433,14 +2433,6 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
>  		return 0;
>  	iov_iter_truncate(iter, inode->i_sb->s_maxbytes);
>  
> -	if (nr_pages > ARRAY_SIZE(pages_onstack))
> -		pages = kmalloc_array(nr_pages, sizeof(void *), GFP_KERNEL);
> -
> -	if (!pages) {
> -		pages = pages_onstack;
> -		nr_pages = min_t(unsigned int, nr_pages, ARRAY_SIZE(pages_onstack));
> -	}
> -
>  	do {
>  		cond_resched();
>  
> @@ -2525,9 +2517,6 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
>  
>  	file_accessed(filp);
>  
> -	if (pages != pages_onstack)
> -		kfree(pages);
> -
>  	return written ? written : error;
>  }
>  EXPORT_SYMBOL_GPL(generic_file_buffered_read);
>