f2fs: fix to avoid use-after-free in f2fs_write_multi_pages()

In compress cluster, if physical block number is less than logic
page number, race condition will cause use-after-free issue as
described below:

- f2fs_write_compressed_pages
 - fio.page = cic->rpages[0];
 - f2fs_outplace_write_data
					- f2fs_compress_write_end_io
					 - kfree(cic->rpages);
					 - kfree(cic);
 - fio.page = cic->rpages[1];

f2fs_write_multi_pages+0xfd0/0x1a98
f2fs_write_data_pages+0x74c/0xb5c
do_writepages+0x64/0x108
__writeback_single_inode+0xdc/0x4b8
writeback_sb_inodes+0x4d0/0xa68
__writeback_inodes_wb+0x88/0x178
wb_writeback+0x1f0/0x424
wb_workfn+0x2f4/0x574
process_one_work+0x210/0x48c
worker_thread+0x2e8/0x44c
kthread+0x110/0x120
ret_from_fork+0x10/0x18

Fixes: 4c8ff7095b ("f2fs: support data compression")
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This commit is contained in:
Chao Yu 2020-02-28 18:08:46 +08:00 committed by Jaegeuk Kim
parent 06c7540fd2
commit 95978caa13

View File

@ -848,7 +848,7 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
blkaddr = datablock_addr(dn.inode, dn.node_page,
dn.ofs_in_node);
fio.page = cic->rpages[i];
fio.page = cc->rpages[i];
fio.old_blkaddr = blkaddr;
/* cluster header */