Commit 20f829999c38 has moved the inode glock taking from gfs2_readpage and gfs2_readahead into gfs2_file_read_iter and gfs2_fault. In gfs2_fault, we didn't take into account that page faults can occur while holding the inode glock, for example,
gfs2_file_read_iter [grabs inode glock] generic_file_read_iter filemap_read copy_page_to_iter gfs2_fault [tries to grab inode glock again]
gfs2_file_write_iter iomap_file_buffered_write iomap_apply iomap_ops->iomap_begin [grabs inode glock] iomap_write_actor iov_iter_fault_in_readable gfs2_fault [tries to grab inode glock again]
Fix that by checking if we're holding the inode glock already.
Reported-by: Jan Kara jack@suse.cz Fixes: 20f829999c38 ("gfs2: Rework read and page fault locking") Cc: stable@vger.kernel.org # v5.8+ Signed-off-by: Andreas Gruenbacher agruenba@redhat.com --- fs/gfs2/file.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-)
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c index 3ebc9af39a04..658fed79d65a 100644 --- a/fs/gfs2/file.c +++ b/fs/gfs2/file.c @@ -538,18 +538,22 @@ 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); + bool recursive = gfs2_glock_is_locked_by_me(ip->i_gl); 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; + if (likely(!recursive)) { + err = gfs2_glock_nq(&gh); + if (err) { + ret = block_page_mkwrite_return(err); + goto out_uninit; + } } ret = filemap_fault(vmf); - gfs2_glock_dq(&gh); + if (likely(!recursive)) + gfs2_glock_dq(&gh); out_uninit: gfs2_holder_uninit(&gh); return ret;