1 files changed, 203 insertions, 313 deletions

diff --git a/mm/readahead.c b/mm/readahead.c
index 9861e883fe5..39bf45d4332 100644
--- a/mm/readahead.c
+++ b/mm/readahead.c
@@ -21,8 +21,16 @@ void default_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
 }
 EXPORT_SYMBOL(default_unplug_io_fn);
 
+/*
+ * Convienent macros for min/max read-ahead pages.
+ * Note that MAX_RA_PAGES is rounded down, while MIN_RA_PAGES is rounded up.
+ * The latter is necessary for systems with large page size(i.e. 64k).
+ */
+#define MAX_RA_PAGES	(VM_MAX_READAHEAD*1024 / PAGE_CACHE_SIZE)
+#define MIN_RA_PAGES	DIV_ROUND_UP(VM_MIN_READAHEAD*1024, PAGE_CACHE_SIZE)
+
 struct backing_dev_info default_backing_dev_info = {
-	.ra_pages	= (VM_MAX_READAHEAD * 1024) / PAGE_CACHE_SIZE,
+	.ra_pages	= MAX_RA_PAGES,
 	.state		= 0,
 	.capabilities	= BDI_CAP_MAP_COPY,
 	.unplug_io_fn	= default_unplug_io_fn,
@@ -41,82 +49,6 @@ file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping)
 }
 EXPORT_SYMBOL_GPL(file_ra_state_init);
 
-/*
- * Return max readahead size for this inode in number-of-pages.
- */
-static inline unsigned long get_max_readahead(struct file_ra_state *ra)
-{
-	return ra->ra_pages;
-}
-
-static inline unsigned long get_min_readahead(struct file_ra_state *ra)
-{
-	return (VM_MIN_READAHEAD * 1024) / PAGE_CACHE_SIZE;
-}
-
-static inline void reset_ahead_window(struct file_ra_state *ra)
-{
-	/*
-	 * ... but preserve ahead_start + ahead_size value,
-	 * see 'recheck:' label in page_cache_readahead().
-	 * Note: We never use ->ahead_size as rvalue without
-	 * checking ->ahead_start != 0 first.
-	 */
-	ra->ahead_size += ra->ahead_start;
-	ra->ahead_start = 0;
-}
-
-static inline void ra_off(struct file_ra_state *ra)
-{
-	ra->start = 0;
-	ra->flags = 0;
-	ra->size = 0;
-	reset_ahead_window(ra);
-	return;
-}
-
-/*
- * Set the initial window size, round to next power of 2 and square
- * for small size, x 4 for medium, and x 2 for large
- * for 128k (32 page) max ra
- * 1-8 page = 32k initial, > 8 page = 128k initial
- */
-static unsigned long get_init_ra_size(unsigned long size, unsigned long max)
-{
-	unsigned long newsize = roundup_pow_of_two(size);
-
-	if (newsize <= max / 32)
-		newsize = newsize * 4;
-	else if (newsize <= max / 4)
-		newsize = newsize * 2;
-	else
-		newsize = max;
-	return newsize;
-}
-
-/*
- * Set the new window size, this is called only when I/O is to be submitted,
- * not for each call to readahead.  If a cache miss occured, reduce next I/O
- * size, else increase depending on how close to max we are.
- */
-static inline unsigned long get_next_ra_size(struct file_ra_state *ra)
-{
-	unsigned long max = get_max_readahead(ra);
-	unsigned long min = get_min_readahead(ra);
-	unsigned long cur = ra->size;
-	unsigned long newsize;
-
-	if (ra->flags & RA_FLAG_MISS) {
-		ra->flags &= ~RA_FLAG_MISS;
-		newsize = max((cur - 2), min);
-	} else if (cur < max / 16) {
-		newsize = 4 * cur;
-	} else {
-		newsize = 2 * cur;
-	}
-	return min(newsize, max);
-}
-
 #define list_to_page(head) (list_entry((head)->prev, struct page, lru))
 
 /**
@@ -193,66 +125,6 @@ out:
 }
 
 /*
- * Readahead design.
- *
- * The fields in struct file_ra_state represent the most-recently-executed
- * readahead attempt:
- *
- * start:	Page index at which we started the readahead
- * size:	Number of pages in that read
- *              Together, these form the "current window".
- *              Together, start and size represent the `readahead window'.
- * prev_index:  The page which the readahead algorithm most-recently inspected.
- *              It is mainly used to detect sequential file reading.
- *              If page_cache_readahead sees that it is again being called for
- *              a page which it just looked at, it can return immediately without
- *              making any state changes.
- * offset:      Offset in the prev_index where the last read ended - used for
- *              detection of sequential file reading.
- * ahead_start,
- * ahead_size:  Together, these form the "ahead window".
- * ra_pages:	The externally controlled max readahead for this fd.
- *
- * When readahead is in the off state (size == 0), readahead is disabled.
- * In this state, prev_index is used to detect the resumption of sequential I/O.
- *
- * The readahead code manages two windows - the "current" and the "ahead"
- * windows.  The intent is that while the application is walking the pages
- * in the current window, I/O is underway on the ahead window.  When the
- * current window is fully traversed, it is replaced by the ahead window
- * and the ahead window is invalidated.  When this copying happens, the
- * new current window's pages are probably still locked.  So
- * we submit a new batch of I/O immediately, creating a new ahead window.
- *
- * So:
- *
- *   ----|----------------|----------------|-----
- *       ^start           ^start+size
- *                        ^ahead_start     ^ahead_start+ahead_size
- *
- *         ^ When this page is read, we submit I/O for the
- *           ahead window.
- *
- * A `readahead hit' occurs when a read request is made against a page which is
- * the next sequential page. Ahead window calculations are done only when it
- * is time to submit a new IO.  The code ramps up the size agressively at first,
- * but slow down as it approaches max_readhead.
- *
- * Any seek/ramdom IO will result in readahead being turned off.  It will resume
- * at the first sequential access.
- *
- * There is a special-case: if the first page which the application tries to
- * read happens to be the first page of the file, it is assumed that a linear
- * read is about to happen and the window is immediately set to the initial size
- * based on I/O request size and the max_readahead.
- *
- * This function is to be called for every read request, rather than when
- * it is time to perform readahead.  It is called only once for the entire I/O
- * regardless of size unless readahead is unable to start enough I/O to satisfy
- * the request (I/O request > max_readahead).
- */
-
-/*
  * do_page_cache_readahead actually reads a chunk of disk.  It allocates all
  * the pages first, then submits them all for I/O. This avoids the very bad
  * behaviour which would occur if page allocations are causing VM writeback.
@@ -265,7 +137,8 @@ out:
  */
 static int
 __do_page_cache_readahead(struct address_space *mapping, struct file *filp,
-			pgoff_t offset, unsigned long nr_to_read)
+			pgoff_t offset, unsigned long nr_to_read,
+			unsigned long lookahead_size)
 {
 	struct inode *inode = mapping->host;
 	struct page *page;
@@ -278,7 +151,7 @@ __do_page_cache_readahead(struct address_space *mapping, struct file *filp,
 	if (isize == 0)
 		goto out;
 
- 	end_index = ((isize - 1) >> PAGE_CACHE_SHIFT);
+	end_index = ((isize - 1) >> PAGE_CACHE_SHIFT);
 
 	/*
 	 * Preallocate as many pages as we will need.
@@ -286,7 +159,7 @@ __do_page_cache_readahead(struct address_space *mapping, struct file *filp,
 	read_lock_irq(&mapping->tree_lock);
 	for (page_idx = 0; page_idx < nr_to_read; page_idx++) {
 		pgoff_t page_offset = offset + page_idx;
-		
+
 		if (page_offset > end_index)
 			break;
 
@@ -301,6 +174,8 @@ __do_page_cache_readahead(struct address_space *mapping, struct file *filp,
 			break;
 		page->index = page_offset;
 		list_add(&page->lru, &page_pool);
+		if (page_idx == nr_to_read - lookahead_size)
+			SetPageReadahead(page);
 		ret++;
 	}
 	read_unlock_irq(&mapping->tree_lock);
@@ -337,7 +212,7 @@ int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
 		if (this_chunk > nr_to_read)
 			this_chunk = nr_to_read;
 		err = __do_page_cache_readahead(mapping, filp,
-						offset, this_chunk);
+						offset, this_chunk, 0);
 		if (err < 0) {
 			ret = err;
 			break;
@@ -350,28 +225,6 @@ int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
 }
 
 /*
- * Check how effective readahead is being.  If the amount of started IO is
- * less than expected then the file is partly or fully in pagecache and
- * readahead isn't helping.
- *
- */
-static inline int check_ra_success(struct file_ra_state *ra,
-			unsigned long nr_to_read, unsigned long actual)
-{
-	if (actual == 0) {
-		ra->cache_hit += nr_to_read;
-		if (ra->cache_hit >= VM_MAX_CACHE_HIT) {
-			ra_off(ra);
-			ra->flags |= RA_FLAG_INCACHE;
-			return 0;
-		}
-	} else {
-		ra->cache_hit=0;
-	}
-	return 1;
-}
-
-/*
  * This version skips the IO if the queue is read-congested, and will tell the
  * block layer to abandon the readahead if request allocation would block.
  *
@@ -384,200 +237,237 @@ int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
 	if (bdi_read_congested(mapping->backing_dev_info))
 		return -1;
 
-	return __do_page_cache_readahead(mapping, filp, offset, nr_to_read);
+	return __do_page_cache_readahead(mapping, filp, offset, nr_to_read, 0);
 }
 
 /*
- * Read 'nr_to_read' pages starting at page 'offset'. If the flag 'block'
- * is set wait till the read completes.  Otherwise attempt to read without
- * blocking.
- * Returns 1 meaning 'success' if read is successful without switching off
- * readahead mode. Otherwise return failure.
+ * Given a desired number of PAGE_CACHE_SIZE readahead pages, return a
+ * sensible upper limit.
  */
-static int
-blockable_page_cache_readahead(struct address_space *mapping, struct file *filp,
-			pgoff_t offset, unsigned long nr_to_read,
-			struct file_ra_state *ra, int block)
+unsigned long max_sane_readahead(unsigned long nr)
+{
+	return min(nr, (node_page_state(numa_node_id(), NR_INACTIVE)
+		+ node_page_state(numa_node_id(), NR_FREE_PAGES)) / 2);
+}
+
+/*
+ * Submit IO for the read-ahead request in file_ra_state.
+ */
+static unsigned long ra_submit(struct file_ra_state *ra,
+		       struct address_space *mapping, struct file *filp)
 {
 	int actual;
 
-	if (!block && bdi_read_congested(mapping->backing_dev_info))
-		return 0;
+	actual = __do_page_cache_readahead(mapping, filp,
+					ra->start, ra->size, ra->async_size);
+
+	return actual;
+}
 
-	actual = __do_page_cache_readahead(mapping, filp, offset, nr_to_read);
+/*
+ * Set the initial window size, round to next power of 2 and square
+ * for small size, x 4 for medium, and x 2 for large
+ * for 128k (32 page) max ra
+ * 1-8 page = 32k initial, > 8 page = 128k initial
+ */
+static unsigned long get_init_ra_size(unsigned long size, unsigned long max)
+{
+	unsigned long newsize = roundup_pow_of_two(size);
 
-	return check_ra_success(ra, nr_to_read, actual);
+	if (newsize <= max / 32)
+		newsize = newsize * 4;
+	else if (newsize <= max / 4)
+		newsize = newsize * 2;
+	else
+		newsize = max;
+
+	return newsize;
 }
 
-static int make_ahead_window(struct address_space *mapping, struct file *filp,
-				struct file_ra_state *ra, int force)
+/*
+ *  Get the previous window size, ramp it up, and
+ *  return it as the new window size.
+ */
+static unsigned long get_next_ra_size(struct file_ra_state *ra,
+						unsigned long max)
 {
-	int block, ret;
-
-	ra->ahead_size = get_next_ra_size(ra);
-	ra->ahead_start = ra->start + ra->size;
-
-	block = force || (ra->prev_index >= ra->ahead_start);
-	ret = blockable_page_cache_readahead(mapping, filp,
-			ra->ahead_start, ra->ahead_size, ra, block);
-
-	if (!ret && !force) {
-		/* A read failure in blocking mode, implies pages are
-		 * all cached. So we can safely assume we have taken
-		 * care of all the pages requested in this call.
-		 * A read failure in non-blocking mode, implies we are
-		 * reading more pages than requested in this call.  So
-		 * we safely assume we have taken care of all the pages
-		 * requested in this call.
-		 *
-		 * Just reset the ahead window in case we failed due to
-		 * congestion.  The ahead window will any way be closed
-		 * in case we failed due to excessive page cache hits.
-		 */
-		reset_ahead_window(ra);
-	}
+	unsigned long cur = ra->size;
+	unsigned long newsize;
 
-	return ret;
+	if (cur < max / 16)
+		newsize = 4 * cur;
+	else
+		newsize = 2 * cur;
+
+	return min(newsize, max);
 }
 
-/**
- * page_cache_readahead - generic adaptive readahead
- * @mapping: address_space which holds the pagecache and I/O vectors
- * @ra: file_ra_state which holds the readahead state
- * @filp: passed on to ->readpage() and ->readpages()
- * @offset: start offset into @mapping, in PAGE_CACHE_SIZE units
- * @req_size: hint: total size of the read which the caller is performing in
- *            PAGE_CACHE_SIZE units
+/*
+ * On-demand readahead design.
+ *
+ * The fields in struct file_ra_state represent the most-recently-executed
+ * readahead attempt:
+ *
+ *                        |<----- async_size ---------|
+ *     |------------------- size -------------------->|
+ *     |==================#===========================|
+ *     ^start             ^page marked with PG_readahead
  *
- * page_cache_readahead() is the main function.  If performs the adaptive
- * readahead window size management and submits the readahead I/O.
+ * To overlap application thinking time and disk I/O time, we do
+ * `readahead pipelining': Do not wait until the application consumed all
+ * readahead pages and stalled on the missing page at readahead_index;
+ * Instead, submit an asynchronous readahead I/O as soon as there are
+ * only async_size pages left in the readahead window. Normally async_size
+ * will be equal to size, for maximum pipelining.
  *
- * Note that @filp is purely used for passing on to the ->readpage[s]()
- * handler: it may refer to a different file from @mapping (so we may not use
- * @filp->f_mapping or @filp->f_path.dentry->d_inode here).
- * Also, @ra may not be equal to &@filp->f_ra.
+ * In interleaved sequential reads, concurrent streams on the same fd can
+ * be invalidating each other's readahead state. So we flag the new readahead
+ * page at (start+size-async_size) with PG_readahead, and use it as readahead
+ * indicator. The flag won't be set on already cached pages, to avoid the
+ * readahead-for-nothing fuss, saving pointless page cache lookups.
+ *
+ * prev_index tracks the last visited page in the _previous_ read request.
+ * It should be maintained by the caller, and will be used for detecting
+ * small random reads. Note that the readahead algorithm checks loosely
+ * for sequential patterns. Hence interleaved reads might be served as
+ * sequential ones.
+ *
+ * There is a special-case: if the first page which the application tries to
+ * read happens to be the first page of the file, it is assumed that a linear
+ * read is about to happen and the window is immediately set to the initial size
+ * based on I/O request size and the max_readahead.
  *
+ * The code ramps up the readahead size aggressively at first, but slow down as
+ * it approaches max_readhead.
+ */
+
+/*
+ * A minimal readahead algorithm for trivial sequential/random reads.
  */
-unsigned long
-page_cache_readahead(struct address_space *mapping, struct file_ra_state *ra,
-		     struct file *filp, pgoff_t offset, unsigned long req_size)
+static unsigned long
+ondemand_readahead(struct address_space *mapping,
+		   struct file_ra_state *ra, struct file *filp,
+		   bool hit_readahead_marker, pgoff_t offset,
+		   unsigned long req_size)
 {
-	unsigned long max, newsize;
+	unsigned long max;	/* max readahead pages */
 	int sequential;
 
-	/*
-	 * We avoid doing extra work and bogusly perturbing the readahead
-	 * window expansion logic.
-	 */
-	if (offset == ra->prev_index && --req_size)
-		++offset;
-
-	/* Note that prev_index == -1 if it is a first read */
-	sequential = (offset == ra->prev_index + 1);
-	ra->prev_index = offset;
-	ra->prev_offset = 0;
-
-	max = get_max_readahead(ra);
-	newsize = min(req_size, max);
-
-	/* No readahead or sub-page sized read or file already in cache */
-	if (newsize == 0 || (ra->flags & RA_FLAG_INCACHE))
-		goto out;
-
-	ra->prev_index += newsize - 1;
+	max = ra->ra_pages;
+	sequential = (offset - ra->prev_index <= 1UL) || (req_size > max);
 
 	/*
-	 * Special case - first read at start of file. We'll assume it's
-	 * a whole-file read and grow the window fast.  Or detect first
-	 * sequential access
+	 * It's the expected callback offset, assume sequential access.
+	 * Ramp up sizes, and push forward the readahead window.
 	 */
-	if (sequential && ra->size == 0) {
-		ra->size = get_init_ra_size(newsize, max);
-		ra->start = offset;
-		if (!blockable_page_cache_readahead(mapping, filp, offset,
-							 ra->size, ra, 1))
-			goto out;
-
-		/*
-		 * If the request size is larger than our max readahead, we
-		 * at least want to be sure that we get 2 IOs in flight and
-		 * we know that we will definitly need the new I/O.
-		 * once we do this, subsequent calls should be able to overlap
-		 * IOs,* thus preventing stalls. so issue the ahead window
-		 * immediately.
-		 */
-		if (req_size >= max)
-			make_ahead_window(mapping, filp, ra, 1);
-
-		goto out;
+	if (offset && (offset == (ra->start + ra->size - ra->async_size) ||
+			offset == (ra->start + ra->size))) {
+		ra->start += ra->size;
+		ra->size = get_next_ra_size(ra, max);
+		ra->async_size = ra->size;
+		goto readit;
 	}
 
 	/*
-	 * Now handle the random case:
-	 * partial page reads and first access were handled above,
-	 * so this must be the next page otherwise it is random
+	 * Standalone, small read.
+	 * Read as is, and do not pollute the readahead state.
 	 */
-	if (!sequential) {
-		ra_off(ra);
-		blockable_page_cache_readahead(mapping, filp, offset,
-				 newsize, ra, 1);
-		goto out;
+	if (!hit_readahead_marker && !sequential) {
+		return __do_page_cache_readahead(mapping, filp,
+						offset, req_size, 0);
 	}
 
 	/*
-	 * If we get here we are doing sequential IO and this was not the first
-	 * occurence (ie we have an existing window)
+	 * It may be one of
+	 * 	- first read on start of file
+	 * 	- sequential cache miss
+	 * 	- oversize random read
+	 * Start readahead for it.
 	 */
-	if (ra->ahead_start == 0) {	 /* no ahead window yet */
-		if (!make_ahead_window(mapping, filp, ra, 0))
-			goto recheck;
-	}
+	ra->start = offset;
+	ra->size = get_init_ra_size(req_size, max);
+	ra->async_size = ra->size > req_size ? ra->size - req_size : ra->size;
 
 	/*
-	 * Already have an ahead window, check if we crossed into it.
-	 * If so, shift windows and issue a new ahead window.
-	 * Only return the #pages that are in the current window, so that
-	 * we get called back on the first page of the ahead window which
-	 * will allow us to submit more IO.
+	 * Hit on a marked page without valid readahead state.
+	 * E.g. interleaved reads.
+	 * Not knowing its readahead pos/size, bet on the minimal possible one.
 	 */
-	if (ra->prev_index >= ra->ahead_start) {
-		ra->start = ra->ahead_start;
-		ra->size = ra->ahead_size;
-		make_ahead_window(mapping, filp, ra, 0);
-recheck:
-		/* prev_index shouldn't overrun the ahead window */
-		ra->prev_index = min(ra->prev_index,
-			ra->ahead_start + ra->ahead_size - 1);
+	if (hit_readahead_marker) {
+		ra->start++;
+		ra->size = get_next_ra_size(ra, max);
 	}
 
-out:
-	return ra->prev_index + 1;
+readit:
+	return ra_submit(ra, mapping, filp);
 }
-EXPORT_SYMBOL_GPL(page_cache_readahead);
 
-/*
- * handle_ra_miss() is called when it is known that a page which should have
- * been present in the pagecache (we just did some readahead there) was in fact
- * not found.  This will happen if it was evicted by the VM (readahead
- * thrashing)
+/**
+ * page_cache_sync_readahead - generic file readahead
+ * @mapping: address_space which holds the pagecache and I/O vectors
+ * @ra: file_ra_state which holds the readahead state
+ * @filp: passed on to ->readpage() and ->readpages()
+ * @offset: start offset into @mapping, in pagecache page-sized units
+ * @req_size: hint: total size of the read which the caller is performing in
+ *            pagecache pages
  *
- * Turn on the cache miss flag in the RA struct, this will cause the RA code
- * to reduce the RA size on the next read.
+ * page_cache_sync_readahead() should be called when a cache miss happened:
+ * it will submit the read.  The readahead logic may decide to piggyback more
+ * pages onto the read request if access patterns suggest it will improve
+ * performance.
  */
-void handle_ra_miss(struct address_space *mapping,
-		struct file_ra_state *ra, pgoff_t offset)
+void page_cache_sync_readahead(struct address_space *mapping,
+			       struct file_ra_state *ra, struct file *filp,
+			       pgoff_t offset, unsigned long req_size)
 {
-	ra->flags |= RA_FLAG_MISS;
-	ra->flags &= ~RA_FLAG_INCACHE;
-	ra->cache_hit = 0;
+	/* no read-ahead */
+	if (!ra->ra_pages)
+		return;
+
+	/* do read-ahead */
+	ondemand_readahead(mapping, ra, filp, false, offset, req_size);
 }
+EXPORT_SYMBOL_GPL(page_cache_sync_readahead);
 
-/*
- * Given a desired number of PAGE_CACHE_SIZE readahead pages, return a
- * sensible upper limit.
- */
-unsigned long max_sane_readahead(unsigned long nr)
+/**
+ * page_cache_async_readahead - file readahead for marked pages
+ * @mapping: address_space which holds the pagecache and I/O vectors
+ * @ra: file_ra_state which holds the readahead state
+ * @filp: passed on to ->readpage() and ->readpages()
+ * @page: the page at @offset which has the PG_readahead flag set
+ * @offset: start offset into @mapping, in pagecache page-sized units
+ * @req_size: hint: total size of the read which the caller is performing in
+ *            pagecache pages
+ *
+ * page_cache_async_ondemand() should be called when a page is used which
+ * has the PG_readahead flag: this is a marker to suggest that the application
+ * has used up enough of the readahead window that we should start pulling in
+ * more pages. */
+void
+page_cache_async_readahead(struct address_space *mapping,
+			   struct file_ra_state *ra, struct file *filp,
+			   struct page *page, pgoff_t offset,
+			   unsigned long req_size)
 {
-	return min(nr, (node_page_state(numa_node_id(), NR_INACTIVE)
-		+ node_page_state(numa_node_id(), NR_FREE_PAGES)) / 2);
+	/* no read-ahead */
+	if (!ra->ra_pages)
+		return;
+
+	/*
+	 * Same bit is used for PG_readahead and PG_reclaim.
+	 */
+	if (PageWriteback(page))
+		return;
+
+	ClearPageReadahead(page);
+
+	/*
+	 * Defer asynchronous read-ahead on IO congestion.
+	 */
+	if (bdi_read_congested(mapping->backing_dev_info))
+		return;
+
+	/* do read-ahead */
+	ondemand_readahead(mapping, ra, filp, true, offset, req_size);
 }
+EXPORT_SYMBOL_GPL(page_cache_async_readahead);