Skip to contentSkip to conversation

squashfs: fix some regressions introduced in the readahead code

PATCH Applied Phillip Lougher 3 participants 6 messages Oct 20, 2022 – Oct 22, 2022 1 message not in the archive

Subsystems (3)
lore
Share & tools
lore

Phillip Lougher

author maintainer
#0 lore raw

cover squashfs: fix some regressions introduced in the readahead code

Hi,

This patch-set fixes 3 regressions introduced by the recent readahead code. The first regression is causing "snaps" to randomly fail after a couple of hours or days, which how the regression came to light.

----------------------------------------------------------------
Phillip Lougher (3):
squashfs: fix read regression introduced in readahead code
squashfs: fix extending readahead beyond end of file
squashfs: fix buffer release race condition in readahead code

 fs/squashfs/file.c       | 23 ++++++++++++++---------
 fs/squashfs/page_actor.c |  3 +++
 fs/squashfs/page_actor.h |  6 +++++-
 3 files changed, 22 insertions(+), 10 deletions(-)
author maintainer
#1 lore raw

PATCH 1/3 squashfs: fix read regression introduced in readahead code

If a file isn't a whole multiple of the page size, the last page will have trailing bytes unfilled.

There was a mistake in the readahead code which did this. In particular it incorrectly assumed that the last page in the readahead page array (page[nr_pages - 1]) will always contain the last page in the block, which if we're at file end, will be the page that needs to be zero filled.

But the readahead code may not return the last page in the block, which means it is unmapped and will be skipped by the decompressors (a temporary buffer used).

In this case the zero filling code will zero out the wrong page, leading to data corruption.

Fix this by by extending the "page actor" to return the last page if present, or NULL if a temporary buffer was used.

---
 fs/squashfs/file.c       | 7 ++++---
 fs/squashfs/page_actor.c | 3 +++
 fs/squashfs/page_actor.h | 6 +++++-
 3 files changed, 12 insertions(+), 4 deletions(-)
fs/squashfs/file.c +4 -3
index e56510964b22..e526eb7a1658 100644
@@ -557,6 +557,7 @@ static void squashfs_readahead(struct readahead_control *ractl)
 		int res, bsize;
 		u64 block = 0;
 		unsigned int expected;
+		struct page *last_page;
 
 		nr_pages = __readahead_batch(ractl, pages, max_pages);
 		if (!nr_pages)
@@ -593,15 +594,15 @@ static void squashfs_readahead(struct readahead_control *ractl)
 
 		res = squashfs_read_data(inode->i_sb, block, bsize, NULL, actor);
 
-		squashfs_page_actor_free(actor);
+		last_page = squashfs_page_actor_free(actor);
 
 		if (res == expected) {
 			int bytes;
 
 			/* Last page (if present) may have trailing bytes not filled */
 			bytes = res % PAGE_SIZE;
-			if (pages[nr_pages - 1]->index == file_end && bytes)
-				memzero_page(pages[nr_pages - 1], bytes,
+			if (index == file_end && bytes && last_page)
+				memzero_page(last_page, bytes,
 					     PAGE_SIZE - bytes);
 
 			for (i = 0; i < nr_pages; i++) {
fs/squashfs/page_actor.c +3 -0
index 54b93bf4a25c..81af6c4ca115 100644
@@ -71,11 +71,13 @@ static void *handle_next_page(struct squashfs_page_actor *actor)
 			(actor->next_index != actor->page[actor->next_page]->index)) {
 		actor->next_index++;
 		actor->returned_pages++;
+		actor->last_page = NULL;
 		return actor->alloc_buffer ? actor->tmp_buffer : ERR_PTR(-ENOMEM);
 	}
 
 	actor->next_index++;
 	actor->returned_pages++;
+	actor->last_page = actor->page[actor->next_page];
 	return actor->pageaddr = kmap_local_page(actor->page[actor->next_page++]);
 }
 
@@ -125,6 +127,7 @@ struct squashfs_page_actor *squashfs_page_actor_init_special(struct squashfs_sb_
 	actor->returned_pages = 0;
 	actor->next_index = page[0]->index & ~((1 << (msblk->block_log - PAGE_SHIFT)) - 1);
 	actor->pageaddr = NULL;
+	actor->last_page = NULL;
 	actor->alloc_buffer = msblk->decompressor->alloc_buffer;
 	actor->squashfs_first_page = direct_first_page;
 	actor->squashfs_next_page = direct_next_page;
fs/squashfs/page_actor.h +5 -1
index 95ffbb543d91..97d4983559b1 100644
@@ -16,6 +16,7 @@ struct squashfs_page_actor {
 	void    *(*squashfs_first_page)(struct squashfs_page_actor *);
 	void    *(*squashfs_next_page)(struct squashfs_page_actor *);
 	void    (*squashfs_finish_page)(struct squashfs_page_actor *);
+	struct page *last_page;
 	int	pages;
 	int	length;
 	int	next_page;
@@ -29,10 +30,13 @@ extern struct squashfs_page_actor *squashfs_page_actor_init(void **buffer,
 extern struct squashfs_page_actor *squashfs_page_actor_init_special(
 				struct squashfs_sb_info *msblk,
 				struct page **page, int pages, int length);
-static inline void squashfs_page_actor_free(struct squashfs_page_actor *actor)
+static inline struct page *squashfs_page_actor_free(struct squashfs_page_actor *actor)
 {
+	struct page *last_page = actor->last_page;
+
 	kfree(actor->tmp_buffer);
 	kfree(actor);
+	return last_page;
 }
 static inline void *squashfs_first_page(struct squashfs_page_actor *actor)
 {
Signature · 1 line
2.35.1
author maintainer
#2 lore raw

PATCH 2/3 squashfs: fix extending readahead beyond end of file

In reply to Phillip Lougher

The readahead code will try to extend readahead to the entire size of the Squashfs data block.

But, it didn't take into account that the last block at the end of the file may not be a whole block. In this case, the code would extend readahead to beyond the end of the file, leaving trailing pages.

Fix this by only requesting the expected number of pages.

Fixes 8fc78b6fe24c ("squashfs: implement readahead")
---
 fs/squashfs/file.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
fs/squashfs/file.c +7 -4
index e526eb7a1658..f0afd4d6fd30 100644
@@ -559,6 +559,12 @@ static void squashfs_readahead(struct readahead_control *ractl)
 		unsigned int expected;
 		struct page *last_page;
 
+		expected = start >> msblk->block_log == file_end ?
+			   (i_size_read(inode) & (msblk->block_size - 1)) :
+			    msblk->block_size;
+
+		max_pages = (expected + PAGE_SIZE - 1) >> PAGE_SHIFT;
+
 		nr_pages = __readahead_batch(ractl, pages, max_pages);
 		if (!nr_pages)
 			break;
@@ -567,13 +573,10 @@ static void squashfs_readahead(struct readahead_control *ractl)
 			goto skip_pages;
 
 		index = pages[0]->index >> shift;
+
 		if ((pages[nr_pages - 1]->index >> shift) != index)
 			goto skip_pages;
 
-		expected = index == file_end ?
-			   (i_size_read(inode) & (msblk->block_size - 1)) :
-			    msblk->block_size;
-
 		if (index == file_end && squashfs_i(inode)->fragment_block !=
 						SQUASHFS_INVALID_BLK) {
 			res = squashfs_readahead_fragment(pages, nr_pages,
Signature · 1 line
2.35.1
author maintainer
#3 lore raw

PATCH 3/3 squashfs: fix buffer release race condition in readahead code

In reply to Phillip Lougher

Fix a buffer release race condition, where the error value was used after release.

Fixes b09a7a036d20 ("squashfs: support reading fragments in readahead call")
---
 fs/squashfs/file.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
fs/squashfs/file.c +3 -2
index f0afd4d6fd30..8ba8c4c50770 100644
@@ -506,8 +506,9 @@ static int squashfs_readahead_fragment(struct page **page,
 		squashfs_i(inode)->fragment_size);
 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
 	unsigned int n, mask = (1 << (msblk->block_log - PAGE_SHIFT)) - 1;
+	int error = buffer->error;
 
-	if (buffer->error)
+	if (error)
 		goto out;
 
 	expected += squashfs_i(inode)->fragment_offset;
@@ -529,7 +530,7 @@ static int squashfs_readahead_fragment(struct page **page,
 
 out:
 	squashfs_cache_put(buffer);
-	return buffer->error;
+	return error;
 }
 
 static void squashfs_readahead(struct readahead_control *ractl)
Signature · 1 line
2.35.1
Tested-by on the v1 series
#4 lore raw
In reply to Phillip Lougher
Quoted 12 lines · Phillip Lougher

Hi,

This patch-set fixes 3 regressions introduced by the recent readahead code. The first regression is causing "snaps" to randomly fail after a couple of hours or days, which how the regression came to light.

----------------------------------------------------------------
Phillip Lougher (3):
squashfs: fix read regression introduced in readahead code
squashfs: fix extending readahead beyond end of file
squashfs: fix buffer release race condition in readahead code

Quoted 4 lines
 fs/squashfs/file.c       | 23 ++++++++++++++---------
 fs/squashfs/page_actor.c |  3 +++
 fs/squashfs/page_actor.h |  6 +++++-
 3 files changed, 22 insertions(+), 10 deletions(-)

The regression goes away with this series, thanks.

For the series,

Signature · 1 line
An old man doll... just what I always wanted! - Clara

A message that is not in the archive (sent off-list or not yet received): [email protected]

#6 lore raw
Quoted 3 lines · On Sat, 22 Oct 2022 03:38:01 +0200 Mirsad Goran Todorovac…

Please consider crediting Marc Miltenberger, who was first to reproduce and confirm that the
bug was not in my build but in mainline kernel.

Thanks, Marc. I've added your Reported-by: to my copies of the three patches.