]> git.proxmox.com Git - mirror_ubuntu-kernels.git/commitdiff
nilfs2: fix general protection fault in nilfs_lookup_dirty_data_buffers()
authorRyusuke Konishi <konishi.ryusuke@gmail.com>
Sat, 5 Aug 2023 13:20:38 +0000 (22:20 +0900)
committerStefan Bader <stefan.bader@canonical.com>
Mon, 30 Oct 2023 11:00:33 +0000 (12:00 +0100)
BugLink: https://bugs.launchpad.net/bugs/2039884
commit f83913f8c5b882a312e72b7669762f8a5c9385e4 upstream.

A syzbot stress test reported that create_empty_buffers() called from
nilfs_lookup_dirty_data_buffers() can cause a general protection fault.

Analysis using its reproducer revealed that the back reference "mapping"
from a page/folio has been changed to NULL after dirty page/folio gang
lookup in nilfs_lookup_dirty_data_buffers().

Fix this issue by excluding pages/folios from being collected if, after
acquiring a lock on each page/folio, its back reference "mapping" differs
from the pointer to the address space struct that held the page/folio.

Link: https://lkml.kernel.org/r/20230805132038.6435-1-konishi.ryusuke@gmail.com
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Reported-by: syzbot+0ad741797f4565e7e2d2@syzkaller.appspotmail.com
Closes: https://lkml.kernel.org/r/0000000000002930a705fc32b231@google.com
Tested-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Kamal Mostafa <kamal@canonical.com>
Signed-off-by: Stefan Bader <stefan.bader@canonical.com>
fs/nilfs2/segment.c

index 329c13ca0a77298c03ef219c789db4f9ef93cc54..ff7527b12a4f1e7ac63299953bb488f8aae5f030 100644 (file)
@@ -725,6 +725,11 @@ static size_t nilfs_lookup_dirty_data_buffers(struct inode *inode,
                struct page *page = pvec.pages[i];
 
                lock_page(page);
+               if (unlikely(page->mapping != mapping)) {
+                       /* Exclude pages removed from the address space */
+                       unlock_page(page);
+                       continue;
+               }
                if (!page_has_buffers(page))
                        create_empty_buffers(page, i_blocksize(inode), 0);
                unlock_page(page);