[RFC v3 0/2] Fix gfs2 readahead deadlocks

[RFC v3 0/2] Fix gfs2 readahead deadlocks

[Andreas Gruenbacher]

In the previous version, we could end up calling ->readpage without
checking IOCB_NOIO.

Andreas Gruenbacher (2):
  fs: Add IOCB_NOIO flag for generic_file_read_iter
  gfs2: Rework read and page fault locking

 fs/gfs2/aops.c     | 45 +--------------------------------------
 fs/gfs2/file.c     | 52 ++++++++++++++++++++++++++++++++++++++++++++--
 include/linux/fs.h |  1 +
 mm/filemap.c       | 23 ++++++++++++++++++--
 4 files changed, 73 insertions(+), 48 deletions(-)

-- 
2.26.2

[RFC v3 1/2] fs: Add IOCB_NOIO flag for generic_file_read_iter

[Andreas Gruenbacher]

Add an IOCB_NOIO flag that indicates to generic_file_read_iter that it
shouldn't trigger any filesystem I/O for the actual request or for
readahead.  This allows to do tentative reads out of the page cache as
some filesystems allow, and to take the appropriate locks and retry the
reads only if the requested pages are not cached.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
---
 include/linux/fs.h |  1 +
 mm/filemap.c       | 23 +++++++++++++++++++++--
 2 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/include/linux/fs.h b/include/linux/fs.h
index 3f881a892ea7..1ab2ea19e883 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -315,6 +315,7 @@ enum rw_hint {
 #define IOCB_SYNC		(1 << 5)
 #define IOCB_WRITE		(1 << 6)
 #define IOCB_NOWAIT		(1 << 7)
+#define IOCB_NOIO		(1 << 8)
 
 struct kiocb {
 	struct file		*ki_filp;
diff --git a/mm/filemap.c b/mm/filemap.c
index f0ae9a6308cb..385759c4ce4b 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2028,7 +2028,7 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 
 		page = find_get_page(mapping, index);
 		if (!page) {
-			if (iocb->ki_flags & IOCB_NOWAIT)
+			if (iocb->ki_flags & (IOCB_NOWAIT | IOCB_NOIO))
 				goto would_block;
 			page_cache_sync_readahead(mapping,
 					ra, filp,
@@ -2038,6 +2038,10 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 				goto no_cached_page;
 		}
 		if (PageReadahead(page)) {
+			if (iocb->ki_flags & IOCB_NOIO) {
+				put_page(page);
+				goto out;
+			}
 			page_cache_async_readahead(mapping,
 					ra, filp, page,
 					index, last_index - index);
@@ -2160,6 +2164,11 @@ ssize_t generic_file_buffered_read(struct kiocb *iocb,
 		}
 
 readpage:
+		if (iocb->ki_flags & IOCB_NOIO) {
+			unlock_page(page);
+			put_page(page);
+			goto would_block;
+		}
 		/*
 		 * A previous I/O error may have been due to temporary
 		 * failures, eg. multipath errors.
@@ -2249,9 +2258,19 @@ EXPORT_SYMBOL_GPL(generic_file_buffered_read);
  *
  * This is the "read_iter()" routine for all filesystems
  * that can use the page cache directly.
+ *
+ * The IOCB_NOWAIT flag in iocb->ki_flags indicates that -EAGAIN shall
+ * be returned when no data can be read without waiting for I/O requests
+ * to complete; it doesn't prevent readahead.
+ *
+ * The IOCB_NOIO flag in iocb->ki_flags indicates that no new I/O
+ * requests shall be made for the read or for readahead.  When no data
+ * can be read, -EAGAIN shall be returned.  When readahead would be
+ * triggered, a partial, possibly empty read shall be returned.
+ *
  * Return:
  * * number of bytes copied, even for partial reads
- * * negative error code if nothing was read
+ * * negative error code (or 0 if IOCB_NOIO) if nothing was read
  */
 ssize_t
 generic_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
-- 
2.26.2

Re: [RFC v3 1/2] fs: Add IOCB_NOIO flag for generic_file_read_iter

[Jens Axboe]

On 7/7/20 8:44 AM, Andreas Gruenbacher wrote:
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index 3f881a892ea7..1ab2ea19e883 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -315,6 +315,7 @@ enum rw_hint {
>  #define IOCB_SYNC		(1 << 5)
>  #define IOCB_WRITE		(1 << 6)
>  #define IOCB_NOWAIT		(1 << 7)
> +#define IOCB_NOIO		(1 << 8)

Just to make this even more trivial in terms of merge conflicts, could
you do 1 << 9 instead?

-- 
Jens Axboe

Re: [RFC v3 1/2] fs: Add IOCB_NOIO flag for generic_file_read_iter

[Andreas Grünbacher]

Am Di., 7. Juli 2020 um 20:40 Uhr schrieb Jens Axboe <axboe@kernel.dk>:
> On 7/7/20 8:44 AM, Andreas Gruenbacher wrote:
> > diff --git a/include/linux/fs.h b/include/linux/fs.h
> > index 3f881a892ea7..1ab2ea19e883 100644
> > --- a/include/linux/fs.h
> > +++ b/include/linux/fs.h
> > @@ -315,6 +315,7 @@ enum rw_hint {
> >  #define IOCB_SYNC            (1 << 5)
> >  #define IOCB_WRITE           (1 << 6)
> >  #define IOCB_NOWAIT          (1 << 7)
> > +#define IOCB_NOIO            (1 << 8)
>
> Just to make this even more trivial in terms of merge conflicts, could
> you do 1 << 9 instead?

Sure.

Andreas

[RFC v3 2/2] gfs2: Rework read and page fault locking

[Andreas Gruenbacher]

So far, gfs2 has taken the inode glocks inside the ->readpage and
->readahead address space operations.  Since commit d4388340ae0b ("fs:
convert mpage_readpages to mpage_readahead"), gfs2_readahead is passed
the pages to read ahead locked.  With that, the current holder of the
inode glock may be trying to lock one of those pages while
gfs2_readahead is trying to take the inode glock, resulting in a
deadlock.

Fix that by moving the lock taking to the higher-level ->read_iter file
and ->fault vm operations.  This also gets rid of an ugly lock inversion
workaround in gfs2_readpage.

The cache consistency model of filesystems like gfs2 is such that if
data is found in the page cache, the data is up to date and can be used
without taking any filesystem locks.  If a page is not cached,
filesystem locks must be taken before populating the page cache.

To avoid taking the inode glock when the data is already cached,
gfs2_file_read_iter first tries to read the data with the IOCB_NOIO flag
set.  If that fails, the inode glock is taken and the operation is
retried with the IOCB_NOIO flag cleared.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
---
 fs/gfs2/aops.c | 45 +------------------------------------------
 fs/gfs2/file.c | 52 ++++++++++++++++++++++++++++++++++++++++++++++++--
 2 files changed, 51 insertions(+), 46 deletions(-)

diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c
index 72c9560f4467..68cd700a2719 100644
--- a/fs/gfs2/aops.c
+++ b/fs/gfs2/aops.c
@@ -468,21 +468,10 @@ static int stuffed_readpage(struct gfs2_inode *ip, struct page *page)
 }
 
 
-/**
- * __gfs2_readpage - readpage
- * @file: The file to read a page for
- * @page: The page to read
- *
- * This is the core of gfs2's readpage. It's used by the internal file
- * reading code as in that case we already hold the glock. Also it's
- * called by gfs2_readpage() once the required lock has been granted.
- */
-
 static int __gfs2_readpage(void *file, struct page *page)
 {
 	struct gfs2_inode *ip = GFS2_I(page->mapping->host);
 	struct gfs2_sbd *sdp = GFS2_SB(page->mapping->host);
-
 	int error;
 
 	if (i_blocksize(page->mapping->host) == PAGE_SIZE &&
@@ -505,36 +494,11 @@ static int __gfs2_readpage(void *file, struct page *page)
  * gfs2_readpage - read a page of a file
  * @file: The file to read
  * @page: The page of the file
- *
- * This deals with the locking required. We have to unlock and
- * relock the page in order to get the locking in the right
- * order.
  */
 
 static int gfs2_readpage(struct file *file, struct page *page)
 {
-	struct address_space *mapping = page->mapping;
-	struct gfs2_inode *ip = GFS2_I(mapping->host);
-	struct gfs2_holder gh;
-	int error;
-
-	unlock_page(page);
-	gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &gh);
-	error = gfs2_glock_nq(&gh);
-	if (unlikely(error))
-		goto out;
-	error = AOP_TRUNCATED_PAGE;
-	lock_page(page);
-	if (page->mapping == mapping && !PageUptodate(page))
-		error = __gfs2_readpage(file, page);
-	else
-		unlock_page(page);
-	gfs2_glock_dq(&gh);
-out:
-	gfs2_holder_uninit(&gh);
-	if (error && error != AOP_TRUNCATED_PAGE)
-		lock_page(page);
-	return error;
+	return __gfs2_readpage(file, page);
 }
 
 /**
@@ -598,16 +562,9 @@ static void gfs2_readahead(struct readahead_control *rac)
 {
 	struct inode *inode = rac->mapping->host;
 	struct gfs2_inode *ip = GFS2_I(inode);
-	struct gfs2_holder gh;
 
-	gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &gh);
-	if (gfs2_glock_nq(&gh))
-		goto out_uninit;
 	if (!gfs2_is_stuffed(ip))
 		mpage_readahead(rac, gfs2_block_map);
-	gfs2_glock_dq(&gh);
-out_uninit:
-	gfs2_holder_uninit(&gh);
 }
 
 /**
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index fe305e4bfd37..bebde537ac8c 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -558,8 +558,29 @@ static vm_fault_t gfs2_page_mkwrite(struct vm_fault *vmf)
 	return block_page_mkwrite_return(ret);
 }
 
+static vm_fault_t gfs2_fault(struct vm_fault *vmf)
+{
+	struct inode *inode = file_inode(vmf->vma->vm_file);
+	struct gfs2_inode *ip = GFS2_I(inode);
+	struct gfs2_holder gh;
+	vm_fault_t ret;
+	int err;
+
+	gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &gh);
+	err = gfs2_glock_nq(&gh);
+	if (err) {
+		ret = block_page_mkwrite_return(err);
+		goto out_uninit;
+	}
+	ret = filemap_fault(vmf);
+	gfs2_glock_dq(&gh);
+out_uninit:
+	gfs2_holder_uninit(&gh);
+	return ret;
+}
+
 static const struct vm_operations_struct gfs2_vm_ops = {
-	.fault = filemap_fault,
+	.fault = gfs2_fault,
 	.map_pages = filemap_map_pages,
 	.page_mkwrite = gfs2_page_mkwrite,
 };
@@ -824,6 +845,9 @@ static ssize_t gfs2_file_direct_write(struct kiocb *iocb, struct iov_iter *from)
 
 static ssize_t gfs2_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
 {
+	struct gfs2_inode *ip;
+	struct gfs2_holder gh;
+	size_t written = 0;
 	ssize_t ret;
 
 	if (iocb->ki_flags & IOCB_DIRECT) {
@@ -832,7 +856,31 @@ static ssize_t gfs2_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
 			return ret;
 		iocb->ki_flags &= ~IOCB_DIRECT;
 	}
-	return generic_file_read_iter(iocb, to);
+	iocb->ki_flags |= IOCB_NOIO;
+	ret = generic_file_read_iter(iocb, to);
+	iocb->ki_flags &= ~IOCB_NOIO;
+	if (ret >= 0) {
+		if (!iov_iter_count(to))
+			return ret;
+		written = ret;
+	} else {
+		if (ret != -EAGAIN)
+			return ret;
+		if (iocb->ki_flags & IOCB_NOWAIT)
+			return ret;
+	}
+	ip = GFS2_I(iocb->ki_filp->f_mapping->host);
+	gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &gh);
+	ret = gfs2_glock_nq(&gh);
+	if (ret)
+		goto out_uninit;
+	ret = generic_file_read_iter(iocb, to);
+	if (ret > 0)
+		written += ret;
+	gfs2_glock_dq(&gh);
+out_uninit:
+	gfs2_holder_uninit(&gh);
+	return written ? written : ret;
 }
 
 /**
-- 
2.26.2