On 2018/1/10 10:43, Daeho Jeong wrote:
> When committing inmem pages is successful, we revoke already committed
> blocks in __revoke_inmem_pages() and finally replace the committed
> ones with the old blocks using f2fs_replace_block(). However, if
> the committed block was newly created one, the address of the old
> block is NEW_ADDR and __f2fs_replace_block() cannot handle NEW_ADDR
> as new_blkaddr properly and a kernel panic occurrs.
Good catch!
>
> Signed-off-by: Daeho Jeong <daeho.je...@samsung.com>
> Tested-by: Shu Tan <shu....@samsung.com>
> ---
> fs/f2fs/segment.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> index c117e09..463f420 100644
> --- a/fs/f2fs/segment.c
> +++ b/fs/f2fs/segment.c
> @@ -248,7 +248,11 @@ static int __revoke_inmem_pages(struct inode *inode,
> goto next;
> }
> get_node_info(sbi, dn.nid, &ni);
> - f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
> + if (cur->old_addr == NEW_ADDR) {
> + invalidate_blocks(sbi, dn.data_blkaddr);
> + f2fs_update_data_blkaddr(&dn, NULL_ADDR);
Original intention here is to recover status to the timing before
committing atomic write. As at that timing blkaddr in dnode should be
cur->old_addr(NEW_ADDR), so we need to change to call:
f2fs_update_data_blkaddr(&dn, NEW_ADDR);
Otherwise, metadata will become inconsistent, because blkaddr value is
NULL_ADDR means that current block is not preallocated, but
total_valid_block_count has already been updated. Right?
Thanks,
> + } else
> + f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
> cur->old_addr, ni.version, true, true);
> f2fs_put_dnode(&dn);
> }
>
