Skip to contentSkip to conversation

squashfs enhance

RFC In review Minchan Kim 1 participant 8 messages Sep 16, 2013 – Sep 17, 2013 SquashFS file system

lore
Share & tools
lore
#0 lore raw

cover squashfs enhance

Our proudct have used squashfs for rootfs and it saves a few bucks per device. Super thanks, Squashfs! You were a perfect for us. But unfortunately, our device start to become complex so sometime we need better throughput for sequential I/O but current squashfs couldn't meet our usecase.

When I dive into the code, it has some problems.

1) Too many copy
2) Only a decompression stream buffer so that concurrent read stuck with that
3) short of readpages

This patchset try to solve above problems.

Just two patches are clean up so it shouldn't change any behavior. And functions factored out will be used for later patches. If they changes some behavior, it's not what I intended. :(

3rd patch removes cache usage for (not-fragemented, no-tail-end) data pages so that we can reduce memory copy.

4th patch supports multiple decompress stream buffer so concurrent read could be handled at the same time. When I tested experiment, It obviously reduces a half time.

5th patch try to implement asynchronous readahead functions so I found it can enhance about 35% with lots of I/O merging.

Any comments are welcome.
Thanks.

Minchan Kim (5):
squashfs: clean up squashfs_read_data
squashfs: clean up squashfs_readpage
squashfs: remove cache for normal data page
squashfs: support multiple decompress stream buffer
squashfs: support readpages

 fs/squashfs/block.c          |  245 +++++++++-----
 fs/squashfs/cache.c          |   16 +-
 fs/squashfs/decompressor.c   |  107 +++++-
 fs/squashfs/decompressor.h   |   27 +-
 fs/squashfs/file.c           |  738 ++++++++++++++++++++++++++++++++++++++----
 fs/squashfs/lzo_wrapper.c    |   12 +-
 fs/squashfs/squashfs.h       |   12 +-
 fs/squashfs/squashfs_fs_sb.h |   11 +-
 fs/squashfs/super.c          |   44 ++-
 fs/squashfs/xz_wrapper.c     |   20 +-
 fs/squashfs/zlib_wrapper.c   |   12 +-
 11 files changed, 1024 insertions(+), 220 deletions(-)
Signature · 1 line
1.7.9.5
#1 lore raw

RFC 1/5 squashfs: clean up squashfs_read_data

The squashfs_read_data functions has a role to read a block and decompress for datablock and metadata.
This patch cleans it up so it has squashfs_read_datablock and squashfs_meta_datablock and morever, squashfs_read_datablock has two part, one is just request I/O and other part is decompress.

Such clean up is useful for code readability and supporting readpages.

---
 fs/squashfs/block.c        |  244 +++++++++++++++++++++++++++++---------------
 fs/squashfs/cache.c        |   16 ++-
 fs/squashfs/decompressor.c |    2 +-
 fs/squashfs/squashfs.h     |    6 +-
 4 files changed, 176 insertions(+), 92 deletions(-)
fs/squashfs/block.c +159 -85
index 41d108e..f33c6ef 100644
@@ -77,99 +77,25 @@ static struct buffer_head *get_block_length(struct super_block *sb,
 }
 
 
-/*
- * Read and decompress a metadata block or datablock.  Length is non-zero
- * if a datablock is being read (the size is stored elsewhere in the
- * filesystem), otherwise the length is obtained from the first two bytes of
- * the metadata block.  A bit in the length field indicates if the block
- * is stored uncompressed in the filesystem (usually because compression
- * generated a larger block - this does occasionally happen with compression
- * algorithms).
- */
-int squashfs_read_data(struct super_block *sb, void **buffer, u64 index,
-			int length, u64 *next_index, int srclength, int pages)
-{
-	struct squashfs_sb_info *msblk = sb->s_fs_info;
-	struct buffer_head **bh;
-	int offset = index & ((1 << msblk->devblksize_log2) - 1);
-	u64 cur_index = index >> msblk->devblksize_log2;
-	int bytes, compressed, b = 0, k = 0, page = 0, avail;
-
-	bh = kcalloc(((srclength + msblk->devblksize - 1)
-		>> msblk->devblksize_log2) + 1, sizeof(*bh), GFP_KERNEL);
-	if (bh == NULL)
-		return -ENOMEM;
-
-	if (length) {
-		/*
-		 * Datablock.
-		 */
-		bytes = -offset;
-		compressed = SQUASHFS_COMPRESSED_BLOCK(length);
-		length = SQUASHFS_COMPRESSED_SIZE_BLOCK(length);
-		if (next_index)
-			*next_index = index + length;
-
-		TRACE("Block @ 0x%llx, %scompressed size %d, src size %d\n",
-			index, compressed ? "" : "un", length, srclength);
-
-		if (length < 0 || length > srclength ||
-				(index + length) > msblk->bytes_used)
-			goto read_failure;
-
-		for (b = 0; bytes < length; b++, cur_index++) {
-			bh[b] = sb_getblk(sb, cur_index);
-			if (bh[b] == NULL)
-				goto block_release;
-			bytes += msblk->devblksize;
-		}
-		ll_rw_block(READ, b, bh);
-	} else {
-		/*
-		 * Metadata block.
-		 */
-		if ((index + 2) > msblk->bytes_used)
-			goto read_failure;
-
-		bh[0] = get_block_length(sb, &cur_index, &offset, &length);
-		if (bh[0] == NULL)
-			goto read_failure;
-		b = 1;
-
-		bytes = msblk->devblksize - offset;
-		compressed = SQUASHFS_COMPRESSED(length);
-		length = SQUASHFS_COMPRESSED_SIZE(length);
-		if (next_index)
-			*next_index = index + length + 2;
-
-		TRACE("Block @ 0x%llx, %scompressed size %d\n", index,
-				compressed ? "" : "un", length);
 
-		if (length < 0 || length > srclength ||
-					(index + length) > msblk->bytes_used)
-			goto block_release;
-
-		for (; bytes < length; b++) {
-			bh[b] = sb_getblk(sb, ++cur_index);
-			if (bh[b] == NULL)
-				goto block_release;
-			bytes += msblk->devblksize;
-		}
-		ll_rw_block(READ, b - 1, bh + 1);
-	}
+int squashfs_decompress_block(struct squashfs_sb_info *msblk, int compressed,
+		void **buffer, struct buffer_head **bh, int nr_bh,
+		int offset, int length, int srclength, int pages)
+{
+	int k = 0;
 
 	if (compressed) {
-		length = squashfs_decompress(msblk, buffer, bh, b, offset,
-			 length, srclength, pages);
+		length = squashfs_decompress(msblk, buffer, bh, nr_bh,
+				offset, length, srclength, pages);
 		if (length < 0)
-			goto read_failure;
+			goto out;
 	} else {
 		/*
 		 * Block is uncompressed.
 		 */
-		int in, pg_offset = 0;
+		int bytes, in, avail, pg_offset = 0, page = 0;
 
-		for (bytes = length; k < b; k++) {
+		for (bytes = length; k < nr_bh; k++) {
 			in = min(bytes, msblk->devblksize - offset);
 			bytes -= in;
 			wait_on_buffer(bh[k]);
@@ -193,6 +119,154 @@ int squashfs_read_data(struct super_block *sb, void **buffer, u64 index,
 		}
 	}
 
+	return length;
+
+block_release:
+	for (; k < nr_bh; k++)
+		put_bh(bh[k]);
+out:
+	return length;
+}
+
+int squashfs_read_submit(struct super_block *sb, u64 index, int length,
+			int srclength, struct buffer_head **bh, int *nr_bh)
+{
+	struct squashfs_sb_info *msblk = sb->s_fs_info;
+	int offset = index & ((1 << msblk->devblksize_log2) - 1);
+	u64 cur_index = index >> msblk->devblksize_log2;
+	int bytes, b = 0, k = 0;
+
+	bytes = -offset;
+	if (length < 0 || length > srclength ||
+			(index + length) > msblk->bytes_used)
+		goto read_failure;
+
+	for (b = 0; bytes < length; b++, cur_index++) {
+		bh[b] = sb_getblk(sb, cur_index);
+		if (bh[b] == NULL)
+			goto block_release;
+		bytes += msblk->devblksize;
+	}
+
+	ll_rw_block(READ, b, bh);
+	*nr_bh = b;
+	return 0;
+
+block_release:
+	for (; k < b; k++)
+		put_bh(bh[k]);
+
+read_failure:
+	ERROR("squashfs_read_submit failed to read block 0x%llx\n",
+					(unsigned long long) index);
+	return -EIO;
+}
+
+/*
+ * Read and decompress a datablock. @length should be non-zero(the size is
+ * stored elsewhere in the filesystem). A bit in the length field indicates
+ * if the block is stored uncompressed in the filesystem (usually because
+ * compression generated a larger block - this does occasionally happen with
+ * compression algorithms).
+ */
+int squashfs_read_datablock(struct super_block *sb, void **buffer, u64 index,
+			int length, int srclength, int pages)
+{
+	struct squashfs_sb_info *msblk = sb->s_fs_info;
+	struct buffer_head **bh;
+	int offset = index & ((1 << msblk->devblksize_log2) - 1);
+	int compressed, ret, b = 0;
+
+	BUG_ON(!length);
+
+	bh = kcalloc(((srclength + msblk->devblksize - 1)
+		>> msblk->devblksize_log2) + 1, sizeof(*bh), GFP_KERNEL);
+	if (bh == NULL)
+		return -ENOMEM;
+
+	compressed = SQUASHFS_COMPRESSED_BLOCK(length);
+	length = SQUASHFS_COMPRESSED_SIZE_BLOCK(length);
+
+	ret = squashfs_read_submit(sb, index, length, srclength, bh, &b);
+	if (ret < 0) {
+		kfree(bh);
+		return ret;
+	}
+
+
+	TRACE("Data block @ 0x%llx, %scompressed size %d, src size %d\n",
+		index, compressed ? "" : "un", length, srclength);
+
+	length = squashfs_decompress_block(sb, compressed, buffer, bh,
+				b, offset, length, srclength, pages);
+	if (length < 0) {
+		ERROR("squashfs_read_datablock failed to read block 0x%llx\n",
+					(unsigned long long) index);
+		kfree(bh);
+		return -EIO;
+	}
+
+	kfree(bh);
+	return length;
+}
+
+/*
+ * Read and decompress a metadata block. @length is obtained from the first
+ * two bytes of the metadata block.  A bit in the length field indicates if
+ * the block is stored uncompressed in the filesystem (usually because
+ * compression generated a larger block - this does occasionally happen with
+ * compression algorithms).
+ */
+int squashfs_read_metablock(struct super_block *sb, void **buffer, u64 index,
+			int length, u64 *next_index, int srclength, int pages)
+{
+	struct squashfs_sb_info *msblk = sb->s_fs_info;
+	struct buffer_head **bh;
+	int offset = index & ((1 << msblk->devblksize_log2) - 1);
+	u64 cur_index = index >> msblk->devblksize_log2;
+	int bytes, compressed, b = 0, k = 0;
+
+	BUG_ON(length);
+
+	bh = kcalloc(((srclength + msblk->devblksize - 1)
+		>> msblk->devblksize_log2) + 1, sizeof(*bh), GFP_KERNEL);
+	if (bh == NULL)
+		return -ENOMEM;
+
+	if ((index + 2) > msblk->bytes_used)
+		goto read_failure;
+
+	bh[0] = get_block_length(sb, &cur_index, &offset, &length);
+	if (bh[0] == NULL)
+		goto read_failure;
+	b = 1;
+
+	bytes = msblk->devblksize - offset;
+	compressed = SQUASHFS_COMPRESSED(length);
+	length = SQUASHFS_COMPRESSED_SIZE(length);
+	if (next_index)
+		*next_index = index + length + 2;
+
+	TRACE("Meta block @ 0x%llx, %scompressed size %d\n", index,
+			compressed ? "" : "un", length);
+
+	if (length < 0 || length > srclength ||
+				(index + length) > msblk->bytes_used)
+		goto block_release;
+
+	for (; bytes < length; b++) {
+		bh[b] = sb_getblk(sb, ++cur_index);
+		if (bh[b] == NULL)
+			goto block_release;
+		bytes += msblk->devblksize;
+	}
+	ll_rw_block(READ, b - 1, bh + 1);
+
+	length = squashfs_decompress_block(msblk, compressed, buffer, bh, b,
+				offset, length, srclength, pages);
+	if (length < 0)
+		goto read_failure;
+
 	kfree(bh);
 	return length;
 
@@ -201,7 +275,7 @@ block_release:
 		put_bh(bh[k]);
 
 read_failure:
-	ERROR("squashfs_read_data failed to read block 0x%llx\n",
+	ERROR("squashfs_read_metablock failed to read block 0x%llx\n",
 					(unsigned long long) index);
 	kfree(bh);
 	return -EIO;
fs/squashfs/cache.c +11 -5
index af0b738..6e6a616 100644
@@ -119,9 +119,15 @@ struct squashfs_cache_entry *squashfs_cache_get(struct super_block *sb,
 			entry->error = 0;
 			spin_unlock(&cache->lock);
 
-			entry->length = squashfs_read_data(sb, entry->data,
-				block, length, &entry->next_index,
-				cache->block_size, cache->pages);
+			if (length)
+				entry->length = squashfs_read_datablock(sb,
+					entry->data, block, length,
+					cache->block_size, cache->pages);
+			else
+				entry->length = squashfs_read_metablock(sb,
+					entry->data, block, length,
+					&entry->next_index, cache->block_size,
+					cache->pages);
 
 			spin_lock(&cache->lock);
 
@@ -424,8 +430,8 @@ void *squashfs_read_table(struct super_block *sb, u64 block, int length)
 	for (i = 0; i < pages; i++, buffer += PAGE_CACHE_SIZE)
 		data[i] = buffer;
 
-	res = squashfs_read_data(sb, data, block, length |
-		SQUASHFS_COMPRESSED_BIT_BLOCK, NULL, length, pages);
+	res = squashfs_read_datablock(sb, data, block, length |
+		SQUASHFS_COMPRESSED_BIT_BLOCK, length, pages);
 
 	kfree(data);
 
fs/squashfs/decompressor.c +1 -1
index 3f6271d..e47453e 100644
@@ -97,7 +97,7 @@ void *squashfs_decompressor_init(struct super_block *sb, unsigned short flags)
 		if (buffer == NULL)
 			return ERR_PTR(-ENOMEM);
 
-		length = squashfs_read_data(sb, &buffer,
+		length = squashfs_read_metablock(sb, &buffer,
 			sizeof(struct squashfs_super_block), 0, NULL,
 			PAGE_CACHE_SIZE, 1);
 
fs/squashfs/squashfs.h +5 -1
index d126651..405baed 100644
@@ -28,8 +28,12 @@
 #define WARNING(s, args...)	pr_warning("SQUASHFS: "s, ## args)
 
 /* block.c */
-extern int squashfs_read_data(struct super_block *, void **, u64, int, u64 *,
+extern int squashfs_read_datablock(struct super_block *, void **, u64, int,
 				int, int);
+extern int squashfs_read_metablock(struct super_block *, void **, u64, int,
+				u64 *, int, int);
+extern int squashfs_read_submit(struct super_block *, u64, int, int,
+				struct buffer_head **, int *);
 
 /* cache.c */
 extern struct squashfs_cache *squashfs_cache_init(char *, int, int);
Signature · 1 line
1.7.9.5
#2 lore raw

RFC 2/5 squashfs: clean up squashfs_readpage

In reply to Minchan Kim

Now squashfs_readpage handles regular data, fragmented data and hole pags so it's rather complex. This patch cleans it up so it makes simple for review and upcoming readahread support. It shouldn't change any old behavior.

---
 fs/squashfs/file.c |  237 ++++++++++++++++++++++++++++++++++++++++------------
 1 file changed, 184 insertions(+), 53 deletions(-)
fs/squashfs/file.c +184 -53
index 8ca62c2..d4d472f 100644
@@ -370,88 +370,189 @@ static int read_blocklist(struct inode *inode, int index, u64 *block)
 	return le32_to_cpu(size);
 }
 
+static int squashfs_fragment_readpage(struct file *file, struct page *page)
+{
+	struct inode *inode = page->mapping->host;
+	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
+	int bytes, i, offset = 0;
+	struct squashfs_cache_entry *buffer = NULL;
+	void *pageaddr;
 
-static int squashfs_readpage(struct file *file, struct page *page)
+	int mask = (1 << (msblk->block_log - PAGE_CACHE_SHIFT)) - 1;
+	int start_index = page->index & ~mask;
+
+	TRACE("Entered squashfs_readpage, page index %lx, start block %llx\n",
+				page->index, squashfs_i(inode)->start);
+
+	/*
+	 * Datablock is stored inside a fragment (tail-end packed
+	 * block).
+	 */
+	buffer = squashfs_get_fragment(inode->i_sb,
+			squashfs_i(inode)->fragment_block,
+			squashfs_i(inode)->fragment_size);
+
+	if (buffer->error) {
+		ERROR("Unable to read page, block %llx, size %x\n",
+			squashfs_i(inode)->fragment_block,
+			squashfs_i(inode)->fragment_size);
+		squashfs_cache_put(buffer);
+		goto error_out;
+	}
+
+	bytes = i_size_read(inode) & (msblk->block_size - 1);
+	offset = squashfs_i(inode)->fragment_offset;
+
+	/*
+	 * Loop copying datablock into pages.  As the datablock likely covers
+	 * many PAGE_CACHE_SIZE pages (default block size is 128 KiB) explicitly
+	 * grab the pages from the page cache, except for the page that we've
+	 * been called to fill.
+	 */
+	for (i = start_index; bytes > 0; i++,
+			bytes -= PAGE_CACHE_SIZE, offset += PAGE_CACHE_SIZE) {
+		struct page *push_page;
+		int avail = min_t(int, bytes, PAGE_CACHE_SIZE);
+
+		TRACE("bytes %d, i %d, available_bytes %d\n", bytes, i, avail);
+
+		push_page = (i == page->index) ? page :
+			grab_cache_page_nowait(page->mapping, i);
+
+		if (!push_page)
+			continue;
+
+		if (PageUptodate(push_page))
+			goto skip_page;
+
+		pageaddr = kmap_atomic(push_page);
+		squashfs_copy_data(pageaddr, buffer, offset, avail);
+		memset(pageaddr + avail, 0, PAGE_CACHE_SIZE - avail);
+		kunmap_atomic(pageaddr);
+		flush_dcache_page(push_page);
+		SetPageUptodate(push_page);
+skip_page:
+		unlock_page(push_page);
+		if (i != page->index)
+			page_cache_release(push_page);
+	}
+
+	squashfs_cache_put(buffer);
+
+	return 0;
+
+error_out:
+	SetPageError(page);
+	pageaddr = kmap_atomic(page);
+	memset(pageaddr, 0, PAGE_CACHE_SIZE);
+	kunmap_atomic(pageaddr);
+	flush_dcache_page(page);
+	if (!PageError(page))
+		SetPageUptodate(page);
+	unlock_page(page);
+
+	return 0;
+}
+
+static int squashfs_hole_readpage(struct file *file, struct inode *inode,
+				int index, struct page *page)
+{
+	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
+	int bytes, i, offset = 0;
+	void *pageaddr;
+
+	int start_index = index << (msblk->block_log - PAGE_CACHE_SHIFT);
+	int file_end = i_size_read(inode) >> msblk->block_log;
+
+	bytes = index == file_end ?
+		(i_size_read(inode) & (msblk->block_size - 1)) :
+		 msblk->block_size;
+
+	/*
+	 * Loop copying datablock into pages.  As the datablock likely covers
+	 * many PAGE_CACHE_SIZE pages (default block size is 128 KiB) explicitly
+	 * grab the pages from the page cache, except for the page that we've
+	 * been called to fill.
+	 */
+	for (i = start_index; bytes > 0; i++,
+			bytes -= PAGE_CACHE_SIZE, offset += PAGE_CACHE_SIZE) {
+		struct page *push_page;
+
+		push_page = (page && i == page->index) ? page :
+			grab_cache_page_nowait(page->mapping, i);
+
+		if (!push_page)
+			continue;
+
+		if (PageUptodate(push_page))
+			goto skip_page;
+
+		pageaddr = kmap_atomic(push_page);
+		memset(pageaddr, 0, PAGE_CACHE_SIZE);
+		kunmap_atomic(pageaddr);
+		flush_dcache_page(push_page);
+		SetPageUptodate(push_page);
+skip_page:
+		unlock_page(push_page);
+		if (page && i == page->index)
+			continue;
+		page_cache_release(push_page);
+	}
+
+	return 0;
+}
+
+static int squashfs_regular_readpage(struct file *file, struct page *page)
 {
+	u64 block = 0;
+	int bsize;
 	struct inode *inode = page->mapping->host;
 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
-	int bytes, i, offset = 0, sparse = 0;
+	int bytes, i, offset = 0;
 	struct squashfs_cache_entry *buffer = NULL;
 	void *pageaddr;
 
 	int mask = (1 << (msblk->block_log - PAGE_CACHE_SHIFT)) - 1;
 	int index = page->index >> (msblk->block_log - PAGE_CACHE_SHIFT);
 	int start_index = page->index & ~mask;
-	int end_index = start_index | mask;
-	int file_end = i_size_read(inode) >> msblk->block_log;
 
 	TRACE("Entered squashfs_readpage, page index %lx, start block %llx\n",
 				page->index, squashfs_i(inode)->start);
 
-	if (page->index >= ((i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
-					PAGE_CACHE_SHIFT))
-		goto out;
-
-	if (index < file_end || squashfs_i(inode)->fragment_block ==
-					SQUASHFS_INVALID_BLK) {
-		/*
-		 * Reading a datablock from disk.  Need to read block list
-		 * to get location and block size.
-		 */
-		u64 block = 0;
-		int bsize = read_blocklist(inode, index, &block);
-		if (bsize < 0)
-			goto error_out;
+	/*
+	 * Reading a datablock from disk.  Need to read block list
+	 * to get location and block size.
+	 */
+	bsize = read_blocklist(inode, index, &block);
+	if (bsize < 0)
+		goto error_out;
 
-		if (bsize == 0) { /* hole */
-			bytes = index == file_end ?
-				(i_size_read(inode) & (msblk->block_size - 1)) :
-				 msblk->block_size;
-			sparse = 1;
-		} else {
-			/*
-			 * Read and decompress datablock.
-			 */
-			buffer = squashfs_get_datablock(inode->i_sb,
-								block, bsize);
-			if (buffer->error) {
-				ERROR("Unable to read page, block %llx, size %x"
-					"\n", block, bsize);
-				squashfs_cache_put(buffer);
-				goto error_out;
-			}
-			bytes = buffer->length;
-		}
+	if (bsize == 0) { /* hole */
+		return squashfs_hole_readpage(file, inode, index, page);
 	} else {
 		/*
-		 * Datablock is stored inside a fragment (tail-end packed
-		 * block).
+		 * Read and decompress datablock.
 		 */
-		buffer = squashfs_get_fragment(inode->i_sb,
-				squashfs_i(inode)->fragment_block,
-				squashfs_i(inode)->fragment_size);
-
+		buffer = squashfs_get_datablock(inode->i_sb,
+							block, bsize);
 		if (buffer->error) {
 			ERROR("Unable to read page, block %llx, size %x\n",
-				squashfs_i(inode)->fragment_block,
-				squashfs_i(inode)->fragment_size);
+				block, bsize);
 			squashfs_cache_put(buffer);
 			goto error_out;
 		}
-		bytes = i_size_read(inode) & (msblk->block_size - 1);
-		offset = squashfs_i(inode)->fragment_offset;
+		bytes = buffer->length;
 	}
-
 	/*
 	 * Loop copying datablock into pages.  As the datablock likely covers
 	 * many PAGE_CACHE_SIZE pages (default block size is 128 KiB) explicitly
 	 * grab the pages from the page cache, except for the page that we've
 	 * been called to fill.
 	 */
-	for (i = start_index; i <= end_index && bytes > 0; i++,
+	for (i = start_index; bytes > 0; i++,
 			bytes -= PAGE_CACHE_SIZE, offset += PAGE_CACHE_SIZE) {
 		struct page *push_page;
-		int avail = sparse ? 0 : min_t(int, bytes, PAGE_CACHE_SIZE);
+		int avail = min_t(int, bytes, PAGE_CACHE_SIZE);
 
 		TRACE("bytes %d, i %d, available_bytes %d\n", bytes, i, avail);
 
@@ -476,14 +577,12 @@ skip_page:
 			page_cache_release(push_page);
 	}
 
-	if (!sparse)
-		squashfs_cache_put(buffer);
+	squashfs_cache_put(buffer);
 
 	return 0;
 
 error_out:
 	SetPageError(page);
-out:
 	pageaddr = kmap_atomic(page);
 	memset(pageaddr, 0, PAGE_CACHE_SIZE);
 	kunmap_atomic(pageaddr);
@@ -495,6 +594,38 @@ out:
 	return 0;
 }
 
+static int squashfs_readpage(struct file *file, struct page *page)
+{
+	struct inode *inode = page->mapping->host;
+	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
+	void *pageaddr;
+
+	int index = page->index >> (msblk->block_log - PAGE_CACHE_SHIFT);
+	int file_end = i_size_read(inode) >> msblk->block_log;
+
+	TRACE("Entered squashfs_readpage, page index %lx, start block %llx\n",
+				page->index, squashfs_i(inode)->start);
+
+	if (page->index >= ((i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
+					PAGE_CACHE_SHIFT))
+		goto out;
+
+	if (index < file_end || squashfs_i(inode)->fragment_block ==
+					SQUASHFS_INVALID_BLK)
+		return squashfs_regular_readpage(file, page);
+	else
+		return squashfs_fragment_readpage(file, page);
+out:
+	pageaddr = kmap_atomic(page);
+	memset(pageaddr, 0, PAGE_CACHE_SIZE);
+	kunmap_atomic(pageaddr);
+	flush_dcache_page(page);
+	if (!PageError(page))
+		SetPageUptodate(page);
+	unlock_page(page);
+
+	return 0;
+}
 
 const struct address_space_operations squashfs_aops = {
 	.readpage = squashfs_readpage
Signature · 1 line
1.7.9.5
#3 lore raw

RFC 3/5 squashfs: remove cache for normal data page

In reply to Minchan Kim

Sqsuashfs have used cache for normal data pages but it's pointless because MM already has cache layer and squashfs adds extra pages into MM's page cache when it reads a page from compressed block.

This patch removes cache usage for normal data pages so it could remove unnecessary one copy(ie, from cache to page cache) and decouple normal data page path from squashfs cache layer.

---
 fs/squashfs/file.c |  117 +++++++++++++++++++++++++++++++++-------------------
 1 file changed, 75 insertions(+), 42 deletions(-)
fs/squashfs/file.c +75 -42
index d4d472f..36508c3 100644
@@ -505,11 +505,14 @@ skip_page:
 static int squashfs_regular_readpage(struct file *file, struct page *page)
 {
 	u64 block = 0;
-	int bsize;
-	struct inode *inode = page->mapping->host;
+	int bsize, i, data_len, pages, nr_pages = 0;
+	struct address_space *mapping = page->mapping;
+	struct inode *inode = mapping->host;
 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
-	int bytes, i, offset = 0;
-	struct squashfs_cache_entry *buffer = NULL;
+	gfp_t gfp_mask;
+	struct page *push_page;
+	struct page **page_array;
+	void **buffer = NULL;
 	void *pageaddr;
 
 	int mask = (1 << (msblk->block_log - PAGE_CACHE_SHIFT)) - 1;
@@ -519,6 +522,8 @@ static int squashfs_regular_readpage(struct file *file, struct page *page)
 	TRACE("Entered squashfs_readpage, page index %lx, start block %llx\n",
 				page->index, squashfs_i(inode)->start);
 
+	pages = msblk->block_size >> PAGE_CACHE_SHIFT;
+	pages = pages ? pages : 1;
 	/*
 	 * Reading a datablock from disk.  Need to read block list
 	 * to get location and block size.
@@ -527,60 +532,88 @@ static int squashfs_regular_readpage(struct file *file, struct page *page)
 	if (bsize < 0)
 		goto error_out;
 
-	if (bsize == 0) { /* hole */
+	if (bsize == 0)
 		return squashfs_hole_readpage(file, inode, index, page);
-	} else {
-		/*
-		 * Read and decompress datablock.
-		 */
-		buffer = squashfs_get_datablock(inode->i_sb,
-							block, bsize);
-		if (buffer->error) {
-			ERROR("Unable to read page, block %llx, size %x\n",
-				block, bsize);
-			squashfs_cache_put(buffer);
-			goto error_out;
-		}
-		bytes = buffer->length;
-	}
+
 	/*
-	 * Loop copying datablock into pages.  As the datablock likely covers
-	 * many PAGE_CACHE_SIZE pages (default block size is 128 KiB) explicitly
-	 * grab the pages from the page cache, except for the page that we've
-	 * been called to fill.
+	 * Read and decompress data block
 	 */
-	for (i = start_index; bytes > 0; i++,
-			bytes -= PAGE_CACHE_SIZE, offset += PAGE_CACHE_SIZE) {
-		struct page *push_page;
-		int avail = min_t(int, bytes, PAGE_CACHE_SIZE);
+	gfp_mask = mapping_gfp_mask(mapping);
+	buffer = kcalloc(1 << (msblk->block_log - PAGE_CACHE_SHIFT),
+				sizeof(void *), GFP_KERNEL);
+	if (!buffer)
+		return -ENOMEM;
 
-		TRACE("bytes %d, i %d, available_bytes %d\n", bytes, i, avail);
+	page_array = kcalloc(1 << (msblk->block_log - PAGE_CACHE_SHIFT),
+				sizeof(struct page *), GFP_KERNEL);
 
-		push_page = (i == page->index) ? page :
-			grab_cache_page_nowait(page->mapping, i);
+	if (!page_array)
+		goto release_buffer;
 
+	/* alloc buffer pages */
+	for (i = 0; i < pages; i++) {
+		if (page->index == start_index + i)
+			push_page = page;
+		else
+			push_page = __page_cache_alloc(gfp_mask);
 		if (!push_page)
-			continue;
+			goto release_page_array;
+		nr_pages++;
+		buffer[i] = kmap(push_page);
+		page_array[i] = push_page;
+	}
 
-		if (PageUptodate(push_page))
-			goto skip_page;
+	data_len = squashfs_read_datablock(inode->i_sb, buffer,
+			block, bsize, msblk->block_size, pages);
 
-		pageaddr = kmap_atomic(push_page);
-		squashfs_copy_data(pageaddr, buffer, offset, avail);
-		memset(pageaddr + avail, 0, PAGE_CACHE_SIZE - avail);
-		kunmap_atomic(pageaddr);
+	if (data_len < 0) {
+		ERROR("Unable to read page, block %llx, size %x\n",
+			block, bsize);
+		for (i = 0; i < nr_pages; i++) {
+			kunmap(page_array[i]);
+			page_cache_release(page_array[i]);
+		}
+		kfree(buffer);
+		kfree(page_array);
+		goto error_out;
+	}
+
+	for (i = 0; i < pages; i++) {
+		push_page = page_array[i];
 		flush_dcache_page(push_page);
 		SetPageUptodate(push_page);
-skip_page:
-		unlock_page(push_page);
-		if (i != page->index)
+		kunmap(page_array[i]);
+		if (page->index == start_index + i) {
+			unlock_page(push_page);
+			continue;
+		}
+
+		if (add_to_page_cache_lru(push_page, mapping,
+			start_index + i, gfp_mask)) {
 			page_cache_release(push_page);
-	}
+			continue;
+		}
 
-	squashfs_cache_put(buffer);
+		unlock_page(push_page);
+		page_cache_release(push_page);
+	}
 
+	kfree(page_array);
+	kfree(buffer);
 	return 0;
 
+release_page_array:
+	for (i = 0; i < nr_pages; i++) {
+		kunmap(page_array[i]);
+		page_cache_release(page_array[i]);
+	}
+
+	kfree(page_array);
+
+release_buffer:
+	kfree(buffer);
+	return -ENOMEM;
+
 error_out:
 	SetPageError(page);
 	pageaddr = kmap_atomic(page);
Signature · 1 line
1.7.9.5
#4 lore raw

RFC 4/5 squashfs: support multiple decompress stream buffer

In reply to Minchan Kim

Now squashfs have used for only one stream buffer for decompression so it hurts concurrent read performance due to locking lock of getting stream buffer.

When file system mount, the number of stream buffer is started from num_online_cpus() and grows up to num_online_cpus() * 2M / block_size * 2. The rationale is MM does readahead chunk into 2M unit to prevent too much memory pin and while one request is waitting, we should request another chunk. That's why I multiply by 2.

If it reveals too much memory problem, we can add shrinker routine.

I did test following as

Two 1G file dd read

dd if=test/test1.dat of=/dev/null &
dd if=test/test2.dat of=/dev/null &

old : 60sec -> new : 30 sec

---
 fs/squashfs/block.c          |    9 ++--
 fs/squashfs/decompressor.c   |  105 ++++++++++++++++++++++++++++++++++++++----
 fs/squashfs/decompressor.h   |   27 +++++------
 fs/squashfs/lzo_wrapper.c    |   12 ++---
 fs/squashfs/squashfs.h       |    3 +-
 fs/squashfs/squashfs_fs_sb.h |    7 ++-
 fs/squashfs/super.c          |   40 ++++++++++++----
 fs/squashfs/xz_wrapper.c     |   20 ++++----
 fs/squashfs/zlib_wrapper.c   |   12 ++---
 9 files changed, 168 insertions(+), 67 deletions(-)
fs/squashfs/block.c +5 -4
index f33c6ef..d41bac8 100644
@@ -78,14 +78,14 @@ static struct buffer_head *get_block_length(struct super_block *sb,
 
 
 
-int squashfs_decompress_block(struct squashfs_sb_info *msblk, int compressed,
+int squashfs_decompress_block(struct super_block *sb, int compressed,
 		void **buffer, struct buffer_head **bh, int nr_bh,
 		int offset, int length, int srclength, int pages)
 {
 	int k = 0;
 
 	if (compressed) {
-		length = squashfs_decompress(msblk, buffer, bh, nr_bh,
+		length = squashfs_decompress(sb, buffer, bh, nr_bh,
 				offset, length, srclength, pages);
 		if (length < 0)
 			goto out;
@@ -93,6 +93,7 @@ int squashfs_decompress_block(struct squashfs_sb_info *msblk, int compressed,
 		/*
 		 * Block is uncompressed.
 		 */
+		struct squashfs_sb_info *msblk = sb->s_fs_info;
 		int bytes, in, avail, pg_offset = 0, page = 0;
 
 		for (bytes = length; k < nr_bh; k++) {
@@ -262,8 +263,8 @@ int squashfs_read_metablock(struct super_block *sb, void **buffer, u64 index,
 	}
 	ll_rw_block(READ, b - 1, bh + 1);
 
-	length = squashfs_decompress_block(msblk, compressed, buffer, bh, b,
-				offset, length, srclength, pages);
+	length = squashfs_decompress_block(sb, compressed, buffer, bh,
+				b, offset, length, srclength, pages);
 	if (length < 0)
 		goto read_failure;
 
fs/squashfs/decompressor.c +97 -8

144 lines not on this page · raw message

fs/squashfs/decompressor.h +12 -15
index 330073e..207c810 100644
@@ -26,27 +26,24 @@
 struct squashfs_decompressor {
 	void	*(*init)(struct squashfs_sb_info *, void *, int);
 	void	(*free)(void *);
-	int	(*decompress)(struct squashfs_sb_info *, void **,
+	int	(*decompress)(struct squashfs_sb_info *, void*, void **,
 		struct buffer_head **, int, int, int, int, int);
 	int	id;
 	char	*name;
 	int	supported;
 };
 
-static inline void squashfs_decompressor_free(struct squashfs_sb_info *msblk,
-	void *s)
-{
-	if (msblk->decompressor)
-		msblk->decompressor->free(s);
-}
-
-static inline int squashfs_decompress(struct squashfs_sb_info *msblk,
-	void **buffer, struct buffer_head **bh, int b, int offset, int length,
-	int srclength, int pages)
-{
-	return msblk->decompressor->decompress(msblk, buffer, bh, b, offset,
-		length, srclength, pages);
-}
+struct squashfs_decomp_strm {
+	void *strm;
+	struct list_head list;
+};
+
+int squashfs_decompress(struct super_block *sb, void **buffer,
+			struct buffer_head **bh, int b, int offset, int length,
+			int srclength, int pages);
+
+void squashfs_decompressor_free(struct squashfs_sb_info *msblk,
+	struct squashfs_decomp_strm *stream);
 
 #ifdef CONFIG_SQUASHFS_XZ
 extern const struct squashfs_decompressor squashfs_xz_comp_ops;
fs/squashfs/lzo_wrapper.c +4 -8
index 00f4dfc..e1bf135 100644
@@ -74,17 +74,15 @@ static void lzo_free(void *strm)
 }
 
 
-static int lzo_uncompress(struct squashfs_sb_info *msblk, void **buffer,
-	struct buffer_head **bh, int b, int offset, int length, int srclength,
-	int pages)
+static int lzo_uncompress(struct squashfs_sb_info *msblk, void *strm,
+		void **buffer, struct buffer_head **bh, int b, int offset,
+		int length, int srclength, int pages)
 {
-	struct squashfs_lzo *stream = msblk->stream;
+	struct squashfs_lzo *stream = strm;
 	void *buff = stream->input;
 	int avail, i, bytes = length, res;
 	size_t out_len = srclength;
 
-	mutex_lock(&msblk->read_data_mutex);
-
 	for (i = 0; i < b; i++) {
 		wait_on_buffer(bh[i]);
 		if (!buffer_uptodate(bh[i]))
@@ -111,7 +109,6 @@ static int lzo_uncompress(struct squashfs_sb_info *msblk, void **buffer,
 		bytes -= avail;
 	}
 
-	mutex_unlock(&msblk->read_data_mutex);
 	return res;
 
 block_release:
@@ -119,7 +116,6 @@ block_release:
 		put_bh(bh[i]);
 
 failed:
-	mutex_unlock(&msblk->read_data_mutex);
 
 	ERROR("lzo decompression failed, data probably corrupt\n");
 	return -EIO;
fs/squashfs/squashfs.h +2 -1
index 405baed..4bb1f90 100644
@@ -52,7 +52,8 @@ extern void *squashfs_read_table(struct super_block *, u64, int);
 
 /* decompressor.c */
 extern const struct squashfs_decompressor *squashfs_lookup_decompressor(int);
-extern void *squashfs_decompressor_init(struct super_block *, unsigned short);
+extern struct squashfs_decomp_strm *squashfs_decompressor_init(
+				struct super_block *);
 
 /* export.c */
 extern __le64 *squashfs_read_inode_lookup_table(struct super_block *, u64, u64,
fs/squashfs/squashfs_fs_sb.h +5 -2
index 52934a2..0a209e6 100644
@@ -63,10 +63,10 @@ struct squashfs_sb_info {
 	__le64					*id_table;
 	__le64					*fragment_index;
 	__le64					*xattr_id_table;
-	struct mutex				read_data_mutex;
+	struct mutex                            comp_strm_mutex;
 	struct mutex				meta_index_mutex;
 	struct meta_index			*meta_index;
-	void					*stream;
+	struct list_head			strm_list;
 	__le64					*inode_lookup_table;
 	u64					inode_table;
 	u64					directory_table;
@@ -76,5 +76,8 @@ struct squashfs_sb_info {
 	long long				bytes_used;
 	unsigned int				inodes;
 	int					xattr_ids;
+	wait_queue_head_t			decomp_wait_queue;
+	int					nr_avail_decomp;
+	unsigned short				flags;
 };
 #endif
fs/squashfs/super.c +31 -9
index 60553a9..70aa9c4 100644
@@ -84,7 +84,7 @@ static int squashfs_fill_super(struct super_block *sb, void *data, int silent)
 	unsigned short flags;
 	unsigned int fragments;
 	u64 lookup_table_start, xattr_id_table_start, next_table;
-	int err;
+	int err, i;
 
 	TRACE("Entered squashfs_fill_superblock\n");
 
@@ -98,7 +98,9 @@ static int squashfs_fill_super(struct super_block *sb, void *data, int silent)
 	msblk->devblksize = sb_min_blocksize(sb, SQUASHFS_DEVBLK_SIZE);
 	msblk->devblksize_log2 = ffz(~msblk->devblksize);
 
-	mutex_init(&msblk->read_data_mutex);
+	INIT_LIST_HEAD(&msblk->strm_list);
+	mutex_init(&msblk->comp_strm_mutex);
+	init_waitqueue_head(&msblk->decomp_wait_queue);
 	mutex_init(&msblk->meta_index_mutex);
 
 	/*
@@ -176,6 +178,7 @@ static int squashfs_fill_super(struct super_block *sb, void *data, int silent)
 	msblk->directory_table = le64_to_cpu(sblk->directory_table_start);
 	msblk->inodes = le32_to_cpu(sblk->inodes);
 	flags = le16_to_cpu(sblk->flags);
+	msblk->flags = flags;
 
 	TRACE("Found valid superblock on %s\n", bdevname(sb->s_bdev, b));
 	TRACE("Inodes are %scompressed\n", SQUASHFS_UNCOMPRESSED_INODES(flags)
@@ -212,11 +215,16 @@ static int squashfs_fill_super(struct super_block *sb, void *data, int silent)
 		goto failed_mount;
 	}
 
-	msblk->stream = squashfs_decompressor_init(sb, flags);
-	if (IS_ERR(msblk->stream)) {
-		err = PTR_ERR(msblk->stream);
-		msblk->stream = NULL;
-		goto failed_mount;
+	/* Allocate mutliple decompressor */
+	for (i = 0; i < num_online_cpus(); i++) {
+		struct squashfs_decomp_strm *decomp_strm;
+		decomp_strm = squashfs_decompressor_init(sb);
+		if (IS_ERR(decomp_strm)) {
+			err = PTR_ERR(decomp_strm);
+			goto failed_mount;
+		}
+		list_add(&decomp_strm->list, &msblk->strm_list);
+		msblk->nr_avail_decomp++;
 	}
 
 	/* Handle xattrs */
@@ -336,7 +344,14 @@ failed_mount:
 	squashfs_cache_delete(msblk->block_cache);
 	squashfs_cache_delete(msblk->fragment_cache);
 	squashfs_cache_delete(msblk->read_page);
-	squashfs_decompressor_free(msblk, msblk->stream);
+	while (!list_empty(&msblk->strm_list)) {
+		struct squashfs_decomp_strm *decomp_strm =
+			list_entry(msblk->strm_list.prev,
+					struct squashfs_decomp_strm, list);
+		list_del(&decomp_strm->list);
+		squashfs_decompressor_free(msblk, decomp_strm);
+	}
+	msblk->nr_avail_decomp = 0;
 	kfree(msblk->inode_lookup_table);
 	kfree(msblk->fragment_index);
 	kfree(msblk->id_table);
@@ -383,7 +398,14 @@ static void squashfs_put_super(struct super_block *sb)
 		squashfs_cache_delete(sbi->block_cache);
 		squashfs_cache_delete(sbi->fragment_cache);
 		squashfs_cache_delete(sbi->read_page);
-		squashfs_decompressor_free(sbi, sbi->stream);
+		while (!list_empty(&sbi->strm_list)) {
+			struct squashfs_decomp_strm *decomp_strm =
+				list_entry(sbi->strm_list.prev,
+					struct squashfs_decomp_strm, list);
+			list_del(&decomp_strm->list);
+			squashfs_decompressor_free(sbi, decomp_strm);
+		}
+		sbi->nr_avail_decomp = 0;
 		kfree(sbi->id_table);
 		kfree(sbi->fragment_index);
 		kfree(sbi->meta_index);
fs/squashfs/xz_wrapper.c +8 -12
index 1760b7d1..98b8bb5 100644
@@ -103,15 +103,13 @@ static void squashfs_xz_free(void *strm)
 }
 
 
-static int squashfs_xz_uncompress(struct squashfs_sb_info *msblk, void **buffer,
-	struct buffer_head **bh, int b, int offset, int length, int srclength,
-	int pages)
+static int squashfs_xz_uncompress(struct squashfs_sb_info *msblk, void *strm,
+		void **buffer, struct buffer_head **bh, int b, int offset,
+		int length, int srclength, int pages)
 {
 	enum xz_ret xz_err;
 	int avail, total = 0, k = 0, page = 0;
-	struct squashfs_xz *stream = msblk->stream;
-
-	mutex_lock(&msblk->read_data_mutex);
+	struct squashfs_xz *stream = strm;
 
 	xz_dec_reset(stream->state);
 	stream->buf.in_pos = 0;
@@ -126,7 +124,7 @@ static int squashfs_xz_uncompress(struct squashfs_sb_info *msblk, void **buffer,
 			length -= avail;
 			wait_on_buffer(bh[k]);
 			if (!buffer_uptodate(bh[k]))
-				goto release_mutex;
+				goto out;
 
 			stream->buf.in = bh[k]->b_data + offset;
 			stream->buf.in_size = avail;
@@ -149,20 +147,18 @@ static int squashfs_xz_uncompress(struct squashfs_sb_info *msblk, void **buffer,
 
 	if (xz_err != XZ_STREAM_END) {
 		ERROR("xz_dec_run error, data probably corrupt\n");
-		goto release_mutex;
+		goto out;
 	}
 
 	if (k < b) {
 		ERROR("xz_uncompress error, input remaining\n");
-		goto release_mutex;
+		goto out;
 	}
 
 	total += stream->buf.out_pos;
-	mutex_unlock(&msblk->read_data_mutex);
 	return total;
 
-release_mutex:
-	mutex_unlock(&msblk->read_data_mutex);
+out:
 
 	for (; k < b; k++)
 		put_bh(bh[k]);
fs/squashfs/zlib_wrapper.c +4 -8
index 55d918f..82c1a70 100644
@@ -61,15 +61,13 @@ static void zlib_free(void *strm)
 }
 
 
-static int zlib_uncompress(struct squashfs_sb_info *msblk, void **buffer,
-	struct buffer_head **bh, int b, int offset, int length, int srclength,
-	int pages)
+static int zlib_uncompress(struct squashfs_sb_info *msblk, void *strm,
+		void **buffer, struct buffer_head **bh, int b, int offset,
+		int length, int srclength, int pages)
 {
 	int zlib_err, zlib_init = 0;
 	int k = 0, page = 0;
-	z_stream *stream = msblk->stream;
-
-	mutex_lock(&msblk->read_data_mutex);
+	z_stream *stream = strm;
 
 	stream->avail_out = 0;
 	stream->avail_in = 0;
@@ -126,11 +124,9 @@ static int zlib_uncompress(struct squashfs_sb_info *msblk, void **buffer,
 	}
 
 	length = stream->total_out;
-	mutex_unlock(&msblk->read_data_mutex);
 	return length;
 
 release_mutex:
-	mutex_unlock(&msblk->read_data_mutex);
 
 	for (; k < b; k++)
 		put_bh(bh[k]);
Signature · 1 line
1.7.9.5
#5 lore raw

RFC 5/5 squashfs: support readpages

In reply to Minchan Kim

This patch supports squashfs_readpages so it can do readahead pages without unplugging I/O scheduler. With blktrace, I confirmed following test.

2 compression ratio 1G file(ie, 500M consumed by real storage) sequential read with fadvise(SEQUENTIAL) hint and tune some knobs for block device and read_ahead_kb.

Old :

 Reads Queued:      524289,   524289KiB  Writes Queued:           0,        0KiB
 Read Dispatches:     4096,   524289KiB  Write Dispatches:        0,        0KiB
 Reads Requeued:         0               Writes Requeued:         0
 Reads Completed:     4096,   524289KiB  Writes Completed:        0,        0KiB
 Read Merges:       520193,   520193KiB  Write Merges:            0,        0KiB
 PC Reads Queued:        0,        0KiB  PC Writes Queued:        0,        0KiB
 PC Read Disp.:         10,        0KiB  PC Write Disp.:          0,        0KiB
 PC Reads Req.:          0               PC Writes Req.:          0
 PC Reads Compl.:        0               PC Writes Compl.:        0
 IO unplugs:          4096               Timer unplugs:           0

New :

 Reads Queued:      524289,   524289KiB  Writes Queued:           0,        0KiB
 Read Dispatches:      633,   524289KiB  Write Dispatches:        0,        0KiB
 Reads Requeued:         0               Writes Requeued:         0
 Reads Completed:      633,   524289KiB  Writes Completed:        0,        0KiB
 Read Merges:       523656,   523656KiB  Write Merges:            0,        0KiB
 PC Reads Queued:        0,        0KiB  PC Writes Queued:        0,        0KiB
 PC Read Disp.:          7,        0KiB  PC Write Disp.:          0,        0KiB
 PC Reads Req.:          0               PC Writes Req.:          0
 PC Reads Compl.:        0               PC Writes Compl.:        0
 IO unplugs:            31               Timer unplugs:           0

IOW, lots of I/O were merged before issuing. Of course, I can confirm that with blktrace.

old:

  8,34   0     1933     0.014381550  4957  Q   R 4197628 + 2 [seqread]
  8,34   0     1934     0.014381629  4957  M   R 4197628 + 2 [seqread]
  8,32   0     1935     0.014381821  4957  A   R 4197630 + 2 <- (8,34) 1278
  8,34   0     1936     0.014381919  4957  Q   R 4197630 + 2 [seqread]
  8,34   0     1937     0.014381998  4957  M   R 4197630 + 2 [seqread]
  8,32   0     1938     0.014382131  4957  A   R 4197632 + 2 <- (8,34) 1280
  8,34   0     1939     0.014382203  4957  Q   R 4197632 + 2 [seqread]
  8,34   0     1940     0.014382281  4957  M   R 4197632 + 2 [seqread]
  8,34   0     1941     0.014382873  4957  I   R 4197378 + 256 [seqread]
  8,34   0        0     0.014383649     0  m   N cfq4957S  / insert_request
  8,34   0        0     0.014384609     0  m   N cfq4957S  / dispatch_insert
  8,34   0        0     0.014385132     0  m   N cfq4957S  / dispatched a request
  8,34   0        0     0.014385460     0  m   N cfq4957S  / activate rq, drv=1
  8,34   0     1942     0.014385581  4957  D   R 4197378 + 256 [seqread]

new:

  8,34   0    98321     0.352583517  5101  M   R 4261888 + 2 [seqread]
  8,34   0    98322     0.353008833  5101  I   R 4230404 + 4096 [seqread]
  8,34   0        0     0.353012357     0  m   N cfq5101SN / insert_request
  8,34   0    98323     0.353013872  5101  I   R 4234500 + 4096 [seqread]
  8,34   0        0     0.353014218     0  m   N cfq5101SN / insert_request
  8,34   0    98324     0.353014553  5101  I   R 4238596 + 4096 [seqread]
  8,34   0        0     0.353014802     0  m   N cfq5101SN / insert_request
  8,34   0    98325     0.353014992  5101  I   R 4242692 + 4096 [seqread]
  8,34   0        0     0.353015315     0  m   N cfq5101SN / insert_request

Voila, it's a result by that.
elapsed time, old : 17.5 sec new : 11.5 sec

A drawback from my mind is magic value 3ms for waiting more I/O so that it can make delay at least 3ms although the I/O was done earlier. The reason I selected that value is that it was a vaule kblockd had used for a long time to plug/unplug for scheduler until we introduced explicit blk_[start|finish]_plug. If it's really problem, we can add a hook into plug and flush the I/O if someone is waiting the I/O.

---
 fs/squashfs/block.c          |    2 -
 fs/squashfs/file.c           |  462 +++++++++++++++++++++++++++++++++++++++++-
 fs/squashfs/squashfs.h       |    3 +
 fs/squashfs/squashfs_fs_sb.h |    4 +
 fs/squashfs/super.c          |    4 +-
 5 files changed, 465 insertions(+), 10 deletions(-)
fs/squashfs/block.c +0 -2
index d41bac8..e8bf200 100644
@@ -76,8 +76,6 @@ static struct buffer_head *get_block_length(struct super_block *sb,
 	return bh;
 }
 
-
-
 int squashfs_decompress_block(struct super_block *sb, int compressed,
 		void **buffer, struct buffer_head **bh, int nr_bh,
 		int offset, int length, int srclength, int pages)
fs/squashfs/file.c +455 -7

497 lines not on this page · raw message

fs/squashfs/squashfs.h +3 -0
index 4bb1f90..2de96a7 100644
@@ -34,6 +34,8 @@ extern int squashfs_read_metablock(struct super_block *, void **, u64, int,
 				u64 *, int, int);
 extern int squashfs_read_submit(struct super_block *, u64, int, int,
 				struct buffer_head **, int *);
+extern int squashfs_decompress_block(struct super_block *, int, void **,
+			struct buffer_head **, int, int, int, int, int);
 
 /* cache.c */
 extern struct squashfs_cache *squashfs_cache_init(char *, int, int);
@@ -89,6 +91,7 @@ extern const struct export_operations squashfs_export_ops;
 
 /* file.c */
 extern const struct address_space_operations squashfs_aops;
+extern void squashfs_decomp_work(struct work_struct *work);
 
 /* inode.c */
 extern const struct inode_operations squashfs_inode_ops;
fs/squashfs/squashfs_fs_sb.h +4 -0
index 0a209e6..bdcc0f2 100644
@@ -79,5 +79,9 @@ struct squashfs_sb_info {
 	wait_queue_head_t			decomp_wait_queue;
 	int					nr_avail_decomp;
 	unsigned short				flags;
+
+	struct delayed_work			delay_work;
+	spinlock_t				decomp_lock;
+	struct list_head			decomp_list;
 };
 #endif
fs/squashfs/super.c +3 -1
index 70aa9c4..78d12d6 100644
@@ -73,7 +73,6 @@ static const struct squashfs_decompressor *supported_squashfs_filesystem(short
 	return decompressor;
 }
 
-
 static int squashfs_fill_super(struct super_block *sb, void *data, int silent)
 {
 	struct squashfs_sb_info *msblk;
@@ -98,6 +97,9 @@ static int squashfs_fill_super(struct super_block *sb, void *data, int silent)
 	msblk->devblksize = sb_min_blocksize(sb, SQUASHFS_DEVBLK_SIZE);
 	msblk->devblksize_log2 = ffz(~msblk->devblksize);
 
+	spin_lock_init(&msblk->decomp_lock);
+	INIT_DELAYED_WORK(&msblk->delay_work, squashfs_decomp_work);
+	INIT_LIST_HEAD(&msblk->decomp_list);
 	INIT_LIST_HEAD(&msblk->strm_list);
 	mutex_init(&msblk->comp_strm_mutex);
 	init_waitqueue_head(&msblk->decomp_wait_queue);
Signature · 1 line
1.7.9.5
#6 lore raw

Re: [RFC 5/5] squashfs: support readpages

This patch had a bug so I send fixed below patchset. Thanks.

Quoted 1 line

From ccbf1322b6322bc34bb3e6f75a27f4fecf84ed02 Mon Sep 17 00:00:00 2001

From: Minchan Kim <[email protected]>
Date: Mon, 16 Sep 2013 14:28:49 +0900
Subject: [RFC 5/5] squashfs: support readpages

This patch supports squashfs_readpages so it can do readahead pages without unplugging I/O scheduler. With blktrace, I confirmed following test.

2 compression ratio 1G file(ie, 500M consumed by real storage) sequential read with fadvise(SEQUENTIAL) hint and tune some knobs for block device and read_ahead_kb.

Old :

 Reads Queued:      524289,   524289KiB  Writes Queued:           0,        0KiB
 Read Dispatches:     4096,   524289KiB  Write Dispatches:        0,        0KiB
 Reads Requeued:         0               Writes Requeued:         0
 Reads Completed:     4096,   524289KiB  Writes Completed:        0,        0KiB
 Read Merges:       520193,   520193KiB  Write Merges:            0,        0KiB
 PC Reads Queued:        0,        0KiB  PC Writes Queued:        0,        0KiB
 PC Read Disp.:         10,        0KiB  PC Write Disp.:          0,        0KiB
 PC Reads Req.:          0               PC Writes Req.:          0
 PC Reads Compl.:        0               PC Writes Compl.:        0
 IO unplugs:          4096               Timer unplugs:           0

New :

 Reads Queued:      524289,   524289KiB  Writes Queued:           0,        0KiB
 Read Dispatches:      633,   524289KiB  Write Dispatches:        0,        0KiB
 Reads Requeued:         0               Writes Requeued:         0
 Reads Completed:      633,   524289KiB  Writes Completed:        0,        0KiB
 Read Merges:       523656,   523656KiB  Write Merges:            0,        0KiB
 PC Reads Queued:        0,        0KiB  PC Writes Queued:        0,        0KiB
 PC Read Disp.:          7,        0KiB  PC Write Disp.:          0,        0KiB
 PC Reads Req.:          0               PC Writes Req.:          0
 PC Reads Compl.:        0               PC Writes Compl.:        0
 IO unplugs:            31               Timer unplugs:           0

IOW, lots of I/O were merged before issuing. Of course, I can confirm that with blktrace.

old:

  8,34   0     1933     0.014381550  4957  Q   R 4197628 + 2 [seqread]
  8,34   0     1934     0.014381629  4957  M   R 4197628 + 2 [seqread]
  8,32   0     1935     0.014381821  4957  A   R 4197630 + 2 <- (8,34) 1278
  8,34   0     1936     0.014381919  4957  Q   R 4197630 + 2 [seqread]
  8,34   0     1937     0.014381998  4957  M   R 4197630 + 2 [seqread]
  8,32   0     1938     0.014382131  4957  A   R 4197632 + 2 <- (8,34) 1280
  8,34   0     1939     0.014382203  4957  Q   R 4197632 + 2 [seqread]
  8,34   0     1940     0.014382281  4957  M   R 4197632 + 2 [seqread]
  8,34   0     1941     0.014382873  4957  I   R 4197378 + 256 [seqread]
  8,34   0        0     0.014383649     0  m   N cfq4957S  / insert_request
  8,34   0        0     0.014384609     0  m   N cfq4957S  / dispatch_insert
  8,34   0        0     0.014385132     0  m   N cfq4957S  / dispatched a request
  8,34   0        0     0.014385460     0  m   N cfq4957S  / activate rq, drv=1
  8,34   0     1942     0.014385581  4957  D   R 4197378 + 256 [seqread]

new:

  8,34   0    98321     0.352583517  5101  M   R 4261888 + 2 [seqread]
  8,34   0    98322     0.353008833  5101  I   R 4230404 + 4096 [seqread]
  8,34   0        0     0.353012357     0  m   N cfq5101SN / insert_request
  8,34   0    98323     0.353013872  5101  I   R 4234500 + 4096 [seqread]
  8,34   0        0     0.353014218     0  m   N cfq5101SN / insert_request
  8,34   0    98324     0.353014553  5101  I   R 4238596 + 4096 [seqread]
  8,34   0        0     0.353014802     0  m   N cfq5101SN / insert_request
  8,34   0    98325     0.353014992  5101  I   R 4242692 + 4096 [seqread]
  8,34   0        0     0.353015315     0  m   N cfq5101SN / insert_request

Voila, it's a result by that.
elapsed time, old : 17.5 sec new : 11.5 sec

A drawback from my mind is magic value 3ms for waiting more I/O so that it can make delay at least 3ms although the I/O was done earlier. The reason I selected that value is that it was a vaule kblockd had used for a long time to plug/unplug for scheduler until we introduced explicit blk_[start|finish]_plug. If it's really problem, we can add a hook into plug and flush the I/O if someone is waiting the I/O.

---
 fs/squashfs/block.c          |   2 -
 fs/squashfs/file.c           | 447 ++++++++++++++++++++++++++++++++++++++++++-
 fs/squashfs/squashfs.h       |   3 +
 fs/squashfs/squashfs_fs_sb.h |   4 +
 fs/squashfs/super.c          |   4 +-
 5 files changed, 450 insertions(+), 10 deletions(-)
fs/squashfs/block.c +0 -2
index d41bac8..e8bf200 100644
@@ -76,8 +76,6 @@ static struct buffer_head *get_block_length(struct super_block *sb,
 	return bh;
 }
 
-
-
 int squashfs_decompress_block(struct super_block *sb, int compressed,
 		void **buffer, struct buffer_head **bh, int nr_bh,
 		int offset, int length, int srclength, int pages)
fs/squashfs/file.c +440 -7

482 lines not on this page · raw message

fs/squashfs/squashfs.h +3 -0
index 4bb1f90..2de96a7 100644
@@ -34,6 +34,8 @@ extern int squashfs_read_metablock(struct super_block *, void **, u64, int,
 				u64 *, int, int);
 extern int squashfs_read_submit(struct super_block *, u64, int, int,
 				struct buffer_head **, int *);
+extern int squashfs_decompress_block(struct super_block *, int, void **,
+			struct buffer_head **, int, int, int, int, int);
 
 /* cache.c */
 extern struct squashfs_cache *squashfs_cache_init(char *, int, int);
@@ -89,6 +91,7 @@ extern const struct export_operations squashfs_export_ops;
 
 /* file.c */
 extern const struct address_space_operations squashfs_aops;
+extern void squashfs_decomp_work(struct work_struct *work);
 
 /* inode.c */
 extern const struct inode_operations squashfs_inode_ops;
fs/squashfs/squashfs_fs_sb.h +4 -0
index 0a209e6..bdcc0f2 100644
@@ -79,5 +79,9 @@ struct squashfs_sb_info {
 	wait_queue_head_t			decomp_wait_queue;
 	int					nr_avail_decomp;
 	unsigned short				flags;
+
+	struct delayed_work			delay_work;
+	spinlock_t				decomp_lock;
+	struct list_head			decomp_list;
 };
 #endif
fs/squashfs/super.c +3 -1
index 70aa9c4..78d12d6 100644
@@ -73,7 +73,6 @@ static const struct squashfs_decompressor *supported_squashfs_filesystem(short
 	return decompressor;
 }
 
-
 static int squashfs_fill_super(struct super_block *sb, void *data, int silent)
 {
 	struct squashfs_sb_info *msblk;
@@ -98,6 +97,9 @@ static int squashfs_fill_super(struct super_block *sb, void *data, int silent)
 	msblk->devblksize = sb_min_blocksize(sb, SQUASHFS_DEVBLK_SIZE);
 	msblk->devblksize_log2 = ffz(~msblk->devblksize);
 
+	spin_lock_init(&msblk->decomp_lock);
+	INIT_DELAYED_WORK(&msblk->delay_work, squashfs_decomp_work);
+	INIT_LIST_HEAD(&msblk->decomp_list);
 	INIT_LIST_HEAD(&msblk->strm_list);
 	mutex_init(&msblk->comp_strm_mutex);
 	init_waitqueue_head(&msblk->decomp_wait_queue);
Signature · 5 lines
1.8.2.1

-- 
Kind regards,
Minchan Kim
#7 lore raw

Re: [RFC 5/5] squashfs: support readpages

And I found regression from iozone random read test. This patch fixes it.

Quoted 1 line

From 546a541c6e29b321e4908ebe0f8aac506eb5b3af Mon Sep 17 00:00:00 2001

From: Minchan Kim <[email protected]>
Date: Tue, 17 Sep 2013 10:20:56 +0900
Subject: [PATCH] squahsfs: read synchronously when readahead window is small

Normally, MM does readahead with small window size so in that case, our plugging logic can hurt performance.
If readahead window is smaller than compress block size, let's do old synchronous read.

---
 fs/squashfs/file.c |   18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
fs/squashfs/file.c +18 -0
index b98965c..2537eeb 100644
@@ -1007,6 +1007,24 @@ static int squashfs_readpages(struct file *file, struct address_space *mapping,
 	start_bindex = hpage->index >> (msblk->block_log - PAGE_CACHE_SHIFT);
 	last_bindex = tpage->index >> (msblk->block_log - PAGE_CACHE_SHIFT);
 
+	/*
+	 * Normally, MM readahread window is smaller than our compressed
+	 * block size. In that case, plugging could hurt performance so
+	 * let's do synchronous read in that case.
+	 */
+	if (start_bindex == last_bindex) {
+		list_del(&hpage->lru);
+		if (add_to_page_cache_lru(hpage, mapping, hpage->index,
+				GFP_KERNEL)) {
+			page_cache_release(hpage);
+			return 0;
+		}
+
+		ret = squashfs_readpage(file, hpage);
+		page_cache_release(hpage);
+		return ret;
+	}
+
 	if (last_bindex >= (i_size_read(inode) >> msblk->block_log))
 		return 0;
 
Signature · 1 line
1.7.9.5
Quoted 662 lines · Minchan Kim

This patch had a bug so I send fixed below patchset. Thanks.

Quoted 1 line

From ccbf1322b6322bc34bb3e6f75a27f4fecf84ed02 Mon Sep 17 00:00:00 2001

From: Minchan Kim <[email protected]>
Date: Mon, 16 Sep 2013 14:28:49 +0900
Subject: [RFC 5/5] squashfs: support readpages

This patch supports squashfs_readpages so it can do readahead pages without unplugging I/O scheduler. With blktrace, I confirmed following test.

2 compression ratio 1G file(ie, 500M consumed by real storage) sequential read with fadvise(SEQUENTIAL) hint and tune some knobs for block device and read_ahead_kb.

Old :

 Reads Queued:      524289,   524289KiB  Writes Queued:           0,        0KiB
 Read Dispatches:     4096,   524289KiB  Write Dispatches:        0,        0KiB
 Reads Requeued:         0               Writes Requeued:         0
 Reads Completed:     4096,   524289KiB  Writes Completed:        0,        0KiB
 Read Merges:       520193,   520193KiB  Write Merges:            0,        0KiB
 PC Reads Queued:        0,        0KiB  PC Writes Queued:        0,        0KiB
 PC Read Disp.:         10,        0KiB  PC Write Disp.:          0,        0KiB
 PC Reads Req.:          0               PC Writes Req.:          0
 PC Reads Compl.:        0               PC Writes Compl.:        0
 IO unplugs:          4096               Timer unplugs:           0

New :

 Reads Queued:      524289,   524289KiB  Writes Queued:           0,        0KiB
 Read Dispatches:      633,   524289KiB  Write Dispatches:        0,        0KiB
 Reads Requeued:         0               Writes Requeued:         0
 Reads Completed:      633,   524289KiB  Writes Completed:        0,        0KiB
 Read Merges:       523656,   523656KiB  Write Merges:            0,        0KiB
 PC Reads Queued:        0,        0KiB  PC Writes Queued:        0,        0KiB
 PC Read Disp.:          7,        0KiB  PC Write Disp.:          0,        0KiB
 PC Reads Req.:          0               PC Writes Req.:          0
 PC Reads Compl.:        0               PC Writes Compl.:        0
 IO unplugs:            31               Timer unplugs:           0

IOW, lots of I/O were merged before issuing. Of course, I can confirm that with blktrace.

old:

  8,34   0     1933     0.014381550  4957  Q   R 4197628 + 2 [seqread]
  8,34   0     1934     0.014381629  4957  M   R 4197628 + 2 [seqread]
  8,32   0     1935     0.014381821  4957  A   R 4197630 + 2 <- (8,34) 1278
  8,34   0     1936     0.014381919  4957  Q   R 4197630 + 2 [seqread]
  8,34   0     1937     0.014381998  4957  M   R 4197630 + 2 [seqread]
  8,32   0     1938     0.014382131  4957  A   R 4197632 + 2 <- (8,34) 1280
  8,34   0     1939     0.014382203  4957  Q   R 4197632 + 2 [seqread]
  8,34   0     1940     0.014382281  4957  M   R 4197632 + 2 [seqread]
  8,34   0     1941     0.014382873  4957  I   R 4197378 + 256 [seqread]
  8,34   0        0     0.014383649     0  m   N cfq4957S  / insert_request
  8,34   0        0     0.014384609     0  m   N cfq4957S  / dispatch_insert
  8,34   0        0     0.014385132     0  m   N cfq4957S  / dispatched a request
  8,34   0        0     0.014385460     0  m   N cfq4957S  / activate rq, drv=1
  8,34   0     1942     0.014385581  4957  D   R 4197378 + 256 [seqread]

new:

  8,34   0    98321     0.352583517  5101  M   R 4261888 + 2 [seqread]
  8,34   0    98322     0.353008833  5101  I   R 4230404 + 4096 [seqread]
  8,34   0        0     0.353012357     0  m   N cfq5101SN / insert_request
  8,34   0    98323     0.353013872  5101  I   R 4234500 + 4096 [seqread]
  8,34   0        0     0.353014218     0  m   N cfq5101SN / insert_request
  8,34   0    98324     0.353014553  5101  I   R 4238596 + 4096 [seqread]
  8,34   0        0     0.353014802     0  m   N cfq5101SN / insert_request
  8,34   0    98325     0.353014992  5101  I   R 4242692 + 4096 [seqread]
  8,34   0        0     0.353015315     0  m   N cfq5101SN / insert_request

Voila, it's a result by that.
elapsed time, old : 17.5 sec new : 11.5 sec

A drawback from my mind is magic value 3ms for waiting more I/O so that it can make delay at least 3ms although the I/O was done earlier. The reason I selected that value is that it was a vaule kblockd had used for a long time to plug/unplug for scheduler until we introduced explicit blk_[start|finish]_plug. If it's really problem, we can add a hook into plug and flush the I/O if someone is waiting the I/O.

---
 fs/squashfs/block.c          |   2 -
 fs/squashfs/file.c           | 447 ++++++++++++++++++++++++++++++++++++++++++-
 fs/squashfs/squashfs.h       |   3 +
 fs/squashfs/squashfs_fs_sb.h |   4 +
 fs/squashfs/super.c          |   4 +-
 5 files changed, 450 insertions(+), 10 deletions(-)
fs/squashfs/block.c +0 -2
index d41bac8..e8bf200 100644
@@ -76,8 +76,6 @@ static struct buffer_head *get_block_length(struct super_block *sb,
 	return bh;
 }
 
-
-
 int squashfs_decompress_block(struct super_block *sb, int compressed,
 		void **buffer, struct buffer_head **bh, int nr_bh,
 		int offset, int length, int srclength, int pages)
fs/squashfs/file.c +440 -7
index 36508c3..b98965c 100644
@@ -47,6 +47,7 @@
 #include <linux/string.h>
 #include <linux/pagemap.h>
 #include <linux/mutex.h>
+#include <linux/swap.h>
 
 #include "squashfs_fs.h"
 #include "squashfs_fs_sb.h"
@@ -454,8 +455,55 @@ error_out:
 	return 0;
 }
 
-static int squashfs_hole_readpage(struct file *file, struct inode *inode,
-				int index, struct page *page)
+static int squashfs_hole_readpages(struct inode *inode, int index,
+				struct list_head *page_list)
+{
+	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
+	int bytes, i, offset = 0;
+	void *pageaddr;
+	struct page *push_page;
+
+	int start_index = index << (msblk->block_log - PAGE_CACHE_SHIFT);
+	int file_end = i_size_read(inode) >> msblk->block_log;
+
+	bytes = index == file_end ?
+		(i_size_read(inode) & (msblk->block_size - 1)) :
+		 msblk->block_size;
+
+	/*
+	 * Loop copying datablock into pages.  As the datablock likely covers
+	 * many PAGE_CACHE_SIZE pages (default block size is 128 KiB) explicitly
+	 * grab the pages from the page cache, except for the page that we've
+	 * been called to fill.
+	 */
+	for (i = start_index; bytes > 0; i++,
+			bytes -= PAGE_CACHE_SIZE, offset += PAGE_CACHE_SIZE) {
+
+		push_page = list_entry(page_list->prev, struct page, lru);
+		list_del(&push_page->lru);
+
+		pageaddr = kmap_atomic(push_page);
+		memset(pageaddr, 0, PAGE_CACHE_SIZE);
+		kunmap_atomic(pageaddr);
+		flush_dcache_page(push_page);
+		SetPageUptodate(push_page);
+
+		lru_cache_add_file(push_page);
+		unlock_page(push_page);
+		page_cache_release(push_page);
+	}
+
+	while (!list_empty(page_list)) {
+		push_page = list_entry(page_list->prev, struct page, lru);
+		list_del(&push_page->lru);
+		page_cache_release(push_page);
+	}
+
+	return 0;
+}
+
+static int squashfs_hole_readpage(struct inode *inode, int index,
+				struct page *page)
 {
 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
 	int bytes, i, offset = 0;
@@ -478,8 +526,8 @@ static int squashfs_hole_readpage(struct file *file, struct inode *inode,
 			bytes -= PAGE_CACHE_SIZE, offset += PAGE_CACHE_SIZE) {
 		struct page *push_page;
 
-		push_page = (page && i == page->index) ? page :
-			grab_cache_page_nowait(page->mapping, i);
+		push_page = (i == page->index) ? page :
+			grab_cache_page_nowait(inode->i_mapping, i);
 
 		if (!push_page)
 			continue;
@@ -494,7 +542,7 @@ static int squashfs_hole_readpage(struct file *file, struct inode *inode,
 		SetPageUptodate(push_page);
 skip_page:
 		unlock_page(push_page);
-		if (page && i == page->index)
+		if (i == page->index)
 			continue;
 		page_cache_release(push_page);
 	}
@@ -533,7 +581,7 @@ static int squashfs_regular_readpage(struct file *file, struct page *page)
 		goto error_out;
 
 	if (bsize == 0)
-		return squashfs_hole_readpage(file, inode, index, page);
+		return squashfs_hole_readpage(inode, index, page);
 
 	/*
 	 * Read and decompress data block
@@ -660,6 +708,391 @@ out:
 	return 0;
 }
 
+static int squashfs_ra_readblock(struct inode *inode, int index,
+			struct buffer_head **bh, int *nr_bh,
+			int *block_size, u64 *block)
+{
+	int bsize;
+	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
+	int ret, b;
+
+	/*
+	 * Reading a datablock from disk.  Need to read block list
+	 * to get location and block size.
+	 */
+	*block_size = read_blocklist(inode, index, block);
+	if (*block_size < 0)
+		return 1;
+
+	if (*block_size == 0)
+		return 0;
+
+	bsize = SQUASHFS_COMPRESSED_SIZE_BLOCK(*block_size);
+	ret = squashfs_read_submit(inode->i_sb, *block, bsize,
+				msblk->block_size, bh, &b);
+	if (ret < 0)
+		return ret;
+
+	*nr_bh = b;
+	return 0;
+}
+
+struct squashfs_ra_private {
+	struct buffer_head **bh;
+	int nr_bh;
+	int block_size;
+	u64 block;
+	void **buffer;
+	struct page **page_array;
+};
+
+/* Caller should free buffer head */
+static int squashfs_ra_read_submit(struct inode *inode, int bindex,
+			struct squashfs_ra_private *ra_priv)
+{
+	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
+	int ret;
+	struct buffer_head **bh;
+	int nr_bh = -1, bsize;
+	u64 block;
+	bh = kcalloc(((msblk->block_size + msblk->devblksize - 1)
+				>> msblk->devblksize_log2) + 1,
+				sizeof(*bh), GFP_KERNEL);
+	if (!bh)
+		return -ENOMEM;
+
+	ra_priv->bh = bh;
+	ret = squashfs_ra_readblock(inode, bindex, bh, &nr_bh,
+					&bsize, &block);
+	if (ret != 0)
+		goto release_bh;
+
+	ra_priv->nr_bh = nr_bh;
+	ra_priv->block_size = bsize;
+	ra_priv->block = block;
+
+	return 0;
+
+release_bh:
+	kfree(ra_priv->bh);
+	return ret;
+}
+
+static int squashfs_ra_decompress(struct inode *inode, int bindex,
+			struct squashfs_ra_private *ra_priv, gfp_t gfp_mask,
+			struct list_head *page_list)
+{
+	int j;
+	int ret = -ENOMEM;
+	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
+	int pages_per_block;
+
+	void **buffer;
+	struct buffer_head **bh;
+	int block_size, nr_bh;
+	u64 block;
+	struct page **page_array, *page;
+
+	int compressed, offset, length;
+
+	pages_per_block = msblk->block_size >> PAGE_CACHE_SHIFT;
+	pages_per_block = pages_per_block ? pages_per_block : 1;
+
+	bh = ra_priv->bh;
+	block_size = ra_priv->block_size;
+	nr_bh = ra_priv->nr_bh;
+	block = ra_priv->block;
+
+	if (block_size == 0)
+		return squashfs_hole_readpages(inode, bindex, page_list);
+
+	buffer = kcalloc(1 << (msblk->block_log - PAGE_CACHE_SHIFT),
+					sizeof(void *), GFP_KERNEL);
+	if (!buffer)
+		goto out;
+	ra_priv->buffer = buffer;
+
+	page_array = kcalloc(1 << (msblk->block_log - PAGE_CACHE_SHIFT),
+					sizeof(struct page *), GFP_KERNEL);
+	if (!page_array)
+		goto release_buffer;
+	ra_priv->page_array = page_array;
+
+	/* alloc buffer pages */
+	for (j = 0; j < pages_per_block; j++) {
+		page = list_entry(page_list->prev, struct page, lru);
+		list_del(&page->lru);
+		buffer[j] = kmap(page);
+		page_array[j] = page;
+	}
+
+	compressed = SQUASHFS_COMPRESSED_BLOCK(block_size);
+	length = SQUASHFS_COMPRESSED_SIZE_BLOCK(block_size);
+
+	offset = block & ((1 << msblk->devblksize_log2) - 1);
+	length = squashfs_decompress_block(inode->i_sb, compressed, buffer,
+				bh, nr_bh, offset, length, msblk->block_size,
+				pages_per_block);
+
+	for (j = 0; j < pages_per_block; j++) {
+		page = page_array[j];
+		flush_dcache_page(page);
+		SetPageUptodate(page);
+		lru_cache_add_file(page);
+		if (!PageLocked(page))
+			printk("unlock page %p j %d\n", page, j);
+		unlock_page(page);
+	}
+
+	ret = 0;
+
+release_buffer:
+	if (ra_priv->buffer) {
+		for (j = 0; j < pages_per_block; j++) {
+			if (!ra_priv->buffer[j])
+				break;
+			kunmap(ra_priv->page_array[j]);
+			page_cache_release(ra_priv->page_array[j]);
+		}
+		kfree(ra_priv->page_array);
+	}
+
+	kfree(ra_priv->buffer);
+out:
+	return ret;
+}
+
+/*
+ * Fill hole pages between page_index and last_page_index
+ * Return 0 if function is successful, otherwise, return 1
+ * All pages are locked and added into page cache so that caller should
+ * add them into LRU and unlock.
+ */
+int hole_plugging(struct squashfs_sb_info *msblk, struct list_head *page_list,
+				int start_bindex, int last_bindex,
+				struct address_space *mapping)
+{
+	struct page *page, *tmp_page, *hpage, *tpage;
+	int page_index, last_page_index;
+	gfp_t gfp_mask = mapping_gfp_mask(mapping);
+
+	page_index = start_bindex << (msblk->block_log - PAGE_CACHE_SHIFT);
+	last_page_index = ((last_bindex + 1) <<
+			(msblk->block_log - PAGE_CACHE_SHIFT));
+	
+	hpage = list_entry(page_list->prev, struct page, lru);
+	for (; page_index < hpage->index; page_index++) {
+		struct page *new_page = page_cache_alloc_readahead(mapping);
+		if (!new_page)
+			return 1;
+		new_page->index = page_index;
+		list_add(&new_page->lru, &hpage->lru);
+	}
+
+	tpage = list_entry(page_list->next, struct page, lru);
+	page_index = tpage->index + 1;
+	for (; page_index < last_page_index; page_index++) {
+		struct page *new_page = page_cache_alloc_readahead(mapping);
+		if (!new_page)
+			return 1;
+		new_page->index = page_index;
+		list_add(&new_page->lru, page_list);
+	}
+
+	list_for_each_entry_reverse(page, page_list, lru)
+		if (add_to_page_cache(page, mapping, page->index, gfp_mask))
+			goto remove_pagecache;
+	return 0;
+
+remove_pagecache:
+	list_for_each_entry_reverse(tmp_page, page_list, lru) {
+		if (tmp_page == page)
+			break;
+		delete_from_page_cache(tmp_page);
+		unlock_page(tmp_page);
+	}
+	return 1;
+}
+
+struct decomp_work {
+	struct inode *inode;
+	int bindex;
+	struct list_head list;
+	struct list_head pages;
+	struct squashfs_ra_private *ra_priv;
+	gfp_t gfp_mask;
+};
+
+void squashfs_decomp_work(struct work_struct *work)
+{
+	struct inode *inode;
+	struct list_head *pages;
+	struct squashfs_ra_private *ra_priv;
+	struct decomp_work *decomp_work;
+	gfp_t gfp_mask;
+	int bindex;
+	struct squashfs_sb_info *msblk =
+		container_of(work, struct squashfs_sb_info, delay_work.work);
+
+	for (;;) {
+		decomp_work = NULL;
+		spin_lock(&msblk->decomp_lock);
+		if (!list_empty(&msblk->decomp_list)) {
+			decomp_work = list_entry(msblk->decomp_list.prev,
+					struct decomp_work, list);
+			list_del(&decomp_work->list);
+		}
+		spin_unlock(&msblk->decomp_lock);
+		if (!decomp_work)
+			return;
+
+		inode = decomp_work->inode;
+		bindex = decomp_work->bindex;
+		ra_priv = decomp_work->ra_priv;
+		gfp_mask = decomp_work->gfp_mask;
+		pages = &decomp_work->pages;
+
+		if (squashfs_ra_decompress(inode, bindex, ra_priv,
+						gfp_mask, pages)) {
+			struct page *page, *tmp;
+			list_for_each_entry_safe_reverse(page, tmp,
+							pages, lru) {
+				list_del(&page->lru);
+				delete_from_page_cache(page);
+				unlock_page(page);
+				page_cache_release(page);
+			}
+		}
+
+		kfree(ra_priv->bh);
+		kfree(ra_priv);
+		kfree(decomp_work);
+	}
+}
+
+static int move_pages(struct list_head *page_list, struct list_head *pages,
+				int nr)
+{
+	int moved_pages = 0;
+	struct page *page, *tmp_page;
+	list_for_each_entry_safe_reverse(page, tmp_page,
+						page_list, lru) {
+		list_move(&page->lru, pages);
+		moved_pages++;
+		if (moved_pages == nr)
+			break;
+	}
+
+	return moved_pages;
+}
+
+static int squashfs_readpages(struct file *file, struct address_space *mapping,
+		struct list_head *page_list, unsigned nr_pages)
+{
+	struct inode *inode = mapping->host;
+	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
+	struct page *hpage, *tpage, *page;
+	struct decomp_work **work;
+	int start_bindex, last_bindex;
+	struct squashfs_ra_private **ra_priv;
+	int pages_per_block, i, ret = -ENOMEM;
+	gfp_t gfp_mask;
+	int nr_blocks;
+
+	gfp_mask = mapping_gfp_mask(mapping);
+
+	hpage = list_entry(page_list->prev, struct page, lru);
+	tpage = list_entry(page_list->next, struct page, lru);
+
+	start_bindex = hpage->index >> (msblk->block_log - PAGE_CACHE_SHIFT);
+	last_bindex = tpage->index >> (msblk->block_log - PAGE_CACHE_SHIFT);
+
+	if (last_bindex >= (i_size_read(inode) >> msblk->block_log))
+		return 0;
+
+	/* fill with pages for readahead  */
+	if (hole_plugging(msblk, page_list, start_bindex,
+					last_bindex, mapping)) {
+		ret = 0;
+		goto out;
+	}
+
+	nr_blocks = last_bindex - start_bindex + 1;
+	ra_priv = kcalloc(nr_blocks, sizeof(*ra_priv), GFP_KERNEL);
+	if (!ra_priv)
+		goto remove_pagecache;
+
+	for (i = 0; i < nr_blocks; i++) {
+		ra_priv[i] = kmalloc(sizeof(**ra_priv),	GFP_KERNEL);
+		if (ra_priv[i] == NULL)
+			goto release_ra_priv;
+	}
+
+	work = kcalloc(nr_blocks, sizeof(*work), GFP_KERNEL);
+	if (!work)
+		goto release_ra_priv;
+
+	for (i = 0; i < nr_blocks; i++) {
+		work[i] = kmalloc(sizeof(**work), GFP_KERNEL);
+		if (!work[i])
+			goto release_work;
+	}
+
+	for (i = 0; i < nr_blocks; i++) {
+		ret = squashfs_ra_read_submit(inode, start_bindex + i ,
+							ra_priv[i]);
+		if (ret)
+			goto release_ra_priv;
+	}
+
+	ret = 0;
+
+	queue_delayed_work(system_unbound_wq, &msblk->delay_work,
+		msecs_to_jiffies(3));
+
+	pages_per_block = msblk->block_size >> PAGE_CACHE_SHIFT;
+	pages_per_block = pages_per_block ? pages_per_block : 1;
+
+	for (i = 0; i < nr_blocks; i++) {
+		struct decomp_work *decomp_work = work[i];
+
+		INIT_LIST_HEAD(&decomp_work->pages);
+		decomp_work->bindex = start_bindex + i;
+		decomp_work->ra_priv = ra_priv[i];
+		decomp_work->gfp_mask = gfp_mask;
+		decomp_work->inode = inode;
+
+		move_pages(page_list, &decomp_work->pages,
+			pages_per_block);
+
+		spin_lock(&msblk->decomp_lock);
+		list_add(&decomp_work->list, &msblk->decomp_list);
+		spin_unlock(&msblk->decomp_lock);
+	}
+
+	kfree(ra_priv);
+	kfree(work);
+
+	return ret;
+
+release_work:
+	for (i = 0; i < nr_blocks; i++)
+		kfree(work[i]);
+	kfree(work);
+release_ra_priv:
+	for (i = 0; i < nr_blocks; i++)
+		kfree(ra_priv[i]);
+	kfree(ra_priv);
+remove_pagecache:
+	list_for_each_entry_reverse(page, page_list, lru) {
+		delete_from_page_cache(page);
+		unlock_page(page);
+	}
+out:
+	return ret;
+}
+
 const struct address_space_operations squashfs_aops = {
-	.readpage = squashfs_readpage
+	.readpage = squashfs_readpage,
+	.readpages = squashfs_readpages,
 };
fs/squashfs/squashfs.h +3 -0
index 4bb1f90..2de96a7 100644
@@ -34,6 +34,8 @@ extern int squashfs_read_metablock(struct super_block *, void **, u64, int,
 				u64 *, int, int);
 extern int squashfs_read_submit(struct super_block *, u64, int, int,
 				struct buffer_head **, int *);
+extern int squashfs_decompress_block(struct super_block *, int, void **,
+			struct buffer_head **, int, int, int, int, int);
 
 /* cache.c */
 extern struct squashfs_cache *squashfs_cache_init(char *, int, int);
@@ -89,6 +91,7 @@ extern const struct export_operations squashfs_export_ops;
 
 /* file.c */
 extern const struct address_space_operations squashfs_aops;
+extern void squashfs_decomp_work(struct work_struct *work);
 
 /* inode.c */
 extern const struct inode_operations squashfs_inode_ops;
fs/squashfs/squashfs_fs_sb.h +4 -0
index 0a209e6..bdcc0f2 100644
@@ -79,5 +79,9 @@ struct squashfs_sb_info {
 	wait_queue_head_t			decomp_wait_queue;
 	int					nr_avail_decomp;
 	unsigned short				flags;
+
+	struct delayed_work			delay_work;
+	spinlock_t				decomp_lock;
+	struct list_head			decomp_list;
 };
 #endif
fs/squashfs/super.c +3 -1
index 70aa9c4..78d12d6 100644
@@ -73,7 +73,6 @@ static const struct squashfs_decompressor *supported_squashfs_filesystem(short
 	return decompressor;
 }
 
-
 static int squashfs_fill_super(struct super_block *sb, void *data, int silent)
 {
 	struct squashfs_sb_info *msblk;
@@ -98,6 +97,9 @@ static int squashfs_fill_super(struct super_block *sb, void *data, int silent)
 	msblk->devblksize = sb_min_blocksize(sb, SQUASHFS_DEVBLK_SIZE);
 	msblk->devblksize_log2 = ffz(~msblk->devblksize);
 
+	spin_lock_init(&msblk->decomp_lock);
+	INIT_DELAYED_WORK(&msblk->delay_work, squashfs_decomp_work);
+	INIT_LIST_HEAD(&msblk->decomp_list);
 	INIT_LIST_HEAD(&msblk->strm_list);
 	mutex_init(&msblk->comp_strm_mutex);
 	init_waitqueue_head(&msblk->decomp_wait_queue);
Signature · 10 lines
1.8.2.1

-- 
Kind regards,
Minchan Kim
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Signature · 2 lines
Kind regards,
Minchan Kim