Jan Kara
2010-Aug-24 12:28 UTC
[Ocfs2-devel] [PATCH] ocfs2: Fix deadlock when allocating page
We cannot call grab_cache_page() when holding filesystem locks or with
a transaction started as grab_cache_page() calls page allocation with
GFP_KERNEL flag and thus page reclaim can recurse back into the filesystem
causing deadlocks or various assertion failures. We have to use
find_or_create_page() instead and pass it GFP_NOFS as we do with other
allocations.
Signed-off-by: Jan Kara <jack at suse.cz>
---
fs/ocfs2/alloc.c | 2 +-
fs/ocfs2/file.c | 2 +-
fs/ocfs2/refcounttree.c | 5 +++--
3 files changed, 5 insertions(+), 4 deletions(-)
diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index 215e12c..592fae5 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -6672,7 +6672,7 @@ int ocfs2_grab_pages(struct inode *inode, loff_t start,
loff_t end,
last_page_bytes = PAGE_ALIGN(end);
index = start >> PAGE_CACHE_SHIFT;
do {
- pages[numpages] = grab_cache_page(mapping, index);
+ pages[numpages] = find_or_create_page(mapping, index, GFP_NOFS);
if (!pages[numpages]) {
ret = -ENOMEM;
mlog_errno(ret);
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index 81296b4..b28cce8 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -774,7 +774,7 @@ static int ocfs2_write_zero_page(struct inode *inode, u64
abs_from,
BUG_ON(abs_to > (((u64)index + 1) << PAGE_CACHE_SHIFT));
BUG_ON(abs_from & (inode->i_blkbits - 1));
- page = grab_cache_page(mapping, index);
+ page = find_or_create_page(mapping, index, GFP_NOFS);
if (!page) {
ret = -ENOMEM;
mlog_errno(ret);
diff --git a/fs/ocfs2/refcounttree.c b/fs/ocfs2/refcounttree.c
index 73a11cc..0afeda8 100644
--- a/fs/ocfs2/refcounttree.c
+++ b/fs/ocfs2/refcounttree.c
@@ -2960,7 +2960,7 @@ static int ocfs2_duplicate_clusters_by_page(handle_t
*handle,
if (map_end & (PAGE_CACHE_SIZE - 1))
to = map_end & (PAGE_CACHE_SIZE - 1);
- page = grab_cache_page(mapping, page_index);
+ page = find_or_create_page(mapping, page_index, GFP_NOFS);
/*
* In case PAGE_CACHE_SIZE <= CLUSTER_SIZE, This page
@@ -3179,7 +3179,8 @@ static int ocfs2_cow_sync_writeback(struct super_block
*sb,
if (map_end > end)
map_end = end;
- page = grab_cache_page(context->inode->i_mapping, page_index);
+ page = find_or_create_page(context->inode->i_mapping,
+ page_index, GFP_NOFS);
BUG_ON(!page);
wait_on_page_writeback(page);
--
1.6.4.2
Sunil Mushran
2010-Aug-24 13:12 UTC
[Ocfs2-devel] [PATCH] ocfs2: Fix deadlock when allocating page
Thanks. I think we have few open bugzillas for this issue. On Aug 24, 2010, at 5:28 AM, Jan Kara <jack at suse.cz> wrote:> We cannot call grab_cache_page() when holding filesystem locks or with > a transaction started as grab_cache_page() calls page allocation with > GFP_KERNEL flag and thus page reclaim can recurse back into the filesystem > causing deadlocks or various assertion failures. We have to use > find_or_create_page() instead and pass it GFP_NOFS as we do with other > allocations. > > Signed-off-by: Jan Kara <jack at suse.cz> > --- > fs/ocfs2/alloc.c | 2 +- > fs/ocfs2/file.c | 2 +- > fs/ocfs2/refcounttree.c | 5 +++-- > 3 files changed, 5 insertions(+), 4 deletions(-) > > diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c > index 215e12c..592fae5 100644 > --- a/fs/ocfs2/alloc.c > +++ b/fs/ocfs2/alloc.c > @@ -6672,7 +6672,7 @@ int ocfs2_grab_pages(struct inode *inode, loff_t start, loff_t end, > last_page_bytes = PAGE_ALIGN(end); > index = start >> PAGE_CACHE_SHIFT; > do { > - pages[numpages] = grab_cache_page(mapping, index); > + pages[numpages] = find_or_create_page(mapping, index, GFP_NOFS); > if (!pages[numpages]) { > ret = -ENOMEM; > mlog_errno(ret); > diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c > index 81296b4..b28cce8 100644 > --- a/fs/ocfs2/file.c > +++ b/fs/ocfs2/file.c > @@ -774,7 +774,7 @@ static int ocfs2_write_zero_page(struct inode *inode, u64 abs_from, > BUG_ON(abs_to > (((u64)index + 1) << PAGE_CACHE_SHIFT)); > BUG_ON(abs_from & (inode->i_blkbits - 1)); > > - page = grab_cache_page(mapping, index); > + page = find_or_create_page(mapping, index, GFP_NOFS); > if (!page) { > ret = -ENOMEM; > mlog_errno(ret); > diff --git a/fs/ocfs2/refcounttree.c b/fs/ocfs2/refcounttree.c > index 73a11cc..0afeda8 100644 > --- a/fs/ocfs2/refcounttree.c > +++ b/fs/ocfs2/refcounttree.c > @@ -2960,7 +2960,7 @@ static int ocfs2_duplicate_clusters_by_page(handle_t *handle, > if (map_end & (PAGE_CACHE_SIZE - 1)) > to = map_end & (PAGE_CACHE_SIZE - 1); > > - page = grab_cache_page(mapping, page_index); > + page = find_or_create_page(mapping, page_index, GFP_NOFS); > > /* > * In case PAGE_CACHE_SIZE <= CLUSTER_SIZE, This page > @@ -3179,7 +3179,8 @@ static int ocfs2_cow_sync_writeback(struct super_block *sb, > if (map_end > end) > map_end = end; > > - page = grab_cache_page(context->inode->i_mapping, page_index); > + page = find_or_create_page(context->inode->i_mapping, > + page_index, GFP_NOFS); > BUG_ON(!page); > > wait_on_page_writeback(page); > -- > 1.6.4.2 > > > _______________________________________________ > Ocfs2-devel mailing list > Ocfs2-devel at oss.oracle.com > http://oss.oracle.com/mailman/listinfo/ocfs2-devel
Mark Fasheh
2010-Aug-24 16:34 UTC
[Ocfs2-devel] [PATCH] ocfs2: Fix deadlock when allocating page
On Tue, Aug 24, 2010 at 02:28:03PM +0200, Jan Kara wrote:> We cannot call grab_cache_page() when holding filesystem locks or with > a transaction started as grab_cache_page() calls page allocation with > GFP_KERNEL flag and thus page reclaim can recurse back into the filesystem > causing deadlocks or various assertion failures. We have to use > find_or_create_page() instead and pass it GFP_NOFS as we do with other > allocations. > > Signed-off-by: Jan Kara <jack at suse.cz>Acked-by: Mark Fasheh <mfasheh at suse.com> --Mark -- Mark Fasheh