| Message ID | 504874D8.60001@cn.fujitsu.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Thu, Sep 06, 2012 at 04:03:04AM -0600, Miao Xie wrote: > When we delete a inode, we will remove all the delayed items including delayed > inode update, and then truncate all the relative metadata. If there is lots of > metadata, we will end the current transaction, and start a new transaction to > truncate the left metadata. In this way, we will leave a inode item that its > link counter is > 0, and also may leave some directory index items in fs/file tree > after the current transaction ends. In other words, the metadata in this fs/file tree > is inconsistent. If we create a snapshot for this tree now, we will find a inode with > corrupted metadata in the new snapshot, and we won't continue to drop the left metadata, > because its link counter is not 0. > > We fix this problem by updating the inode item before the current transaction ends. > > Signed-off-by: Miao Xie <miaox@cn.fujitsu.com> > --- > Changelog v1 -> v4: > - Update the comment of the truncation in the btrfs_evict_inode() > - Fix enospc problem of the inode update This isn't the right way to do the enospc fix, we need to do btrfs_start_transaction(root, 1); and then change the trans->block_rsv to our reserve for the truncate and then set it back to the trans rsv for the update that way we don't run out of space because we used our reservation for the truncate. Just update this patch and send it along and I'll include it. Thanks, Josef -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, 6 Sep 2012 09:09:14 -0400, Josef Bacik wrote: > On Thu, Sep 06, 2012 at 04:03:04AM -0600, Miao Xie wrote: >> When we delete a inode, we will remove all the delayed items including delayed >> inode update, and then truncate all the relative metadata. If there is lots of >> metadata, we will end the current transaction, and start a new transaction to >> truncate the left metadata. In this way, we will leave a inode item that its >> link counter is > 0, and also may leave some directory index items in fs/file tree >> after the current transaction ends. In other words, the metadata in this fs/file tree >> is inconsistent. If we create a snapshot for this tree now, we will find a inode with >> corrupted metadata in the new snapshot, and we won't continue to drop the left metadata, >> because its link counter is not 0. >> >> We fix this problem by updating the inode item before the current transaction ends. >> >> Signed-off-by: Miao Xie <miaox@cn.fujitsu.com> >> --- >> Changelog v1 -> v4: >> - Update the comment of the truncation in the btrfs_evict_inode() >> - Fix enospc problem of the inode update > > This isn't the right way to do the enospc fix, we need to do > > btrfs_start_transaction(root, 1); and then change the trans->block_rsv to our > reserve for the truncate and then set it back to the trans rsv for the update > that way we don't run out of space because we used our reservation for the > truncate. Just update this patch and send it along and I'll include it. > Thanks, btrfs_start_transaction() will cause the deadlock problem just as I said in comment, the reason is: start transaction | v reserve meta-data space | v flush delay allocation -> iput inode -> evict inode ^ | | v wait for delay allocation flush <- reserve meta-data space So we may introduce a special starting-transaction function which can reserve the space without flush. I'll make a patch with this way. Thanks Miao -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c index eb768c4..8f2d1bf 100644 --- a/fs/btrfs/delayed-inode.c +++ b/fs/btrfs/delayed-inode.c @@ -650,7 +650,8 @@ static int btrfs_delayed_inode_reserve_metadata( * we're accounted for. */ if (!src_rsv || (!trans->bytes_reserved && - src_rsv->type != BTRFS_BLOCK_RSV_DELALLOC)) { + src_rsv->type != BTRFS_BLOCK_RSV_DELALLOC && + src_rsv->type != BTRFS_BLOCK_RSV_TEMP)) { ret = btrfs_block_rsv_add_noflush(root, dst_rsv, num_bytes); /* * Since we're under a transaction reserve_metadata_bytes could diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index d494c11..709f5b9 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3738,7 +3738,7 @@ void btrfs_evict_inode(struct inode *inode) struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_block_rsv *rsv, *global_rsv; - u64 min_size = btrfs_calc_trunc_metadata_size(root, 1); + u64 min_size; unsigned long nr; int ret; @@ -3772,21 +3772,23 @@ void btrfs_evict_inode(struct inode *inode) btrfs_orphan_del(NULL, inode); goto no_delete; } + + min_size = btrfs_calc_trunc_metadata_size(root, 1); + min_size += btrfs_calc_trans_metadata_size(root, 1); rsv->size = min_size; global_rsv = &root->fs_info->global_block_rsv; btrfs_i_size_write(inode, 0); /* - * This is a bit simpler than btrfs_truncate since - * - * 1) We've already reserved our space for our orphan item in the - * unlink. - * 2) We're going to delete the inode item, so we don't need to update - * it at all. + * This is a bit simpler than btrfs_truncate since we've already + * reserved our space for our orphan item in the unlink, so we just + * need to reserve some slack space in case we add bytes and update + * inode item when doing the truncate. * - * So we just need to reserve some slack space in case we add bytes when - * doing the truncate. + * The differentiation is we can not reserve the space for the inode + * update when starting the transaction because it may cause + * the deadlock. */ while (1) { ret = btrfs_block_rsv_refill_noflush(root, rsv, min_size); @@ -3820,6 +3822,9 @@ void btrfs_evict_inode(struct inode *inode) if (ret != -EAGAIN) break; + ret = btrfs_update_inode(trans, root, inode); + BUG_ON(ret); + nr = trans->blocks_used; btrfs_end_transaction(trans, root); trans = NULL;
When we delete a inode, we will remove all the delayed items including delayed inode update, and then truncate all the relative metadata. If there is lots of metadata, we will end the current transaction, and start a new transaction to truncate the left metadata. In this way, we will leave a inode item that its link counter is > 0, and also may leave some directory index items in fs/file tree after the current transaction ends. In other words, the metadata in this fs/file tree is inconsistent. If we create a snapshot for this tree now, we will find a inode with corrupted metadata in the new snapshot, and we won't continue to drop the left metadata, because its link counter is not 0. We fix this problem by updating the inode item before the current transaction ends. Signed-off-by: Miao Xie <miaox@cn.fujitsu.com> --- Changelog v1 -> v4: - Update the comment of the truncation in the btrfs_evict_inode() - Fix enospc problem of the inode update --- fs/btrfs/delayed-inode.c | 3 ++- fs/btrfs/inode.c | 23 ++++++++++++++--------- 2 files changed, 16 insertions(+), 10 deletions(-)