On 2018年06月26日 16:46, Misono Tomohiro wrote:
> On 2018/06/26 16:09, Nikolay Borisov wrote:
>> Commit 5d23515be669 ("btrfs: Move qgroup rescan on quota enable to
>> btrfs_quota_enable") not only resulted in an easier to follow code but
>> it also introduced a subtle bug. It changed the timing when the initial
>> transaction rescan was happening - before the commit it would happen
>> after transaction commit had occured but after the commit it might happen
>> before the transaction was committed. This results in failure to
>> correctly rescan the quota since there could be data which is still not
>> committed on disk.
>>
>> This patch aims to fix this by movign the transaction creation/commit
>> inside btrfs_quota_enable, which allows to schedule the quota commit
>> after the transaction has been committed.
>>
>> Fixes: 5d23515be669 ("btrfs: Move qgroup rescan on quota enable to
>> btrfs_quota_enable")
>> Reported-by: Misono Tomohiro <misono.tomoh...@jp.fujitsu.com>
>> Signed-off-by: Nikolay Borisov <nbori...@suse.com>
>> ---
>> Hi Misono,
>>
>> Care to test the following patch ? If you say it's ok I will do a similar
>> one
>> for the btrfs_quota_disable function. This will also allow me to get rid of
>> the extra err variable in btrfs_ioctl_quota_ctl. Additionally I think the
>> number of blocks (2) passed to the transaction for reservation might be
>> wrong.
>
> Hi,
>
> The patch does not removes start_transaction() from btrfs_ioctl_quota_ctl(),
> so this does not work but I understand your approach (continue to below).
>
>>
>> fs/btrfs/ioctl.c | 2 +-
>> fs/btrfs/qgroup.c | 17 ++++++++++++++---
>> fs/btrfs/qgroup.h | 3 +--
>> 3 files changed, 16 insertions(+), 6 deletions(-)
>>
>> diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
>> index a399750b9e41..bf99d7aae3ae 100644
>> --- a/fs/btrfs/ioctl.c
>> +++ b/fs/btrfs/ioctl.c
>> @@ -5161,7 +5161,7 @@ static long btrfs_ioctl_quota_ctl(struct file *file,
>> void __user *arg)
>>
>> switch (sa->cmd) {
>> case BTRFS_QUOTA_CTL_ENABLE:
>> - ret = btrfs_quota_enable(trans, fs_info);
>> + ret = btrfs_quota_enable(fs_info);
>> break;
>> case BTRFS_QUOTA_CTL_DISABLE:
>> ret = btrfs_quota_disable(trans, fs_info);
>> diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c
>> index 1874a6d2e6f5..91bb7e97c0d0 100644
>> --- a/fs/btrfs/qgroup.c
>> +++ b/fs/btrfs/qgroup.c
>> @@ -875,8 +875,7 @@ static int btrfs_clean_quota_tree(struct
>> btrfs_trans_handle *trans,
>> return ret;
>> }
>>
>> -int btrfs_quota_enable(struct btrfs_trans_handle *trans,
>> - struct btrfs_fs_info *fs_info)
>> +int btrfs_quota_enable(struct btrfs_fs_info *fs_info)
>> {
>> struct btrfs_root *quota_root;
>> struct btrfs_root *tree_root = fs_info->tree_root;
>> @@ -886,6 +885,7 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
>> struct btrfs_key key;
>> struct btrfs_key found_key;
>> struct btrfs_qgroup *qgroup = NULL;
>> + struct btrfs_trans_handle *trans = NULL;
>> int ret = 0;
>> int slot;
>>
>> @@ -893,6 +893,12 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
>> if (fs_info->quota_root)
>> goto out;
>>
>> + trans = btrfs_start_transaction(tree_root, 2);
>
> (Should we use fs_root for quota?)
>
>> + if (IS_ERR(trans)) {
>> + ret = PTR_ERR(trans);
>> + goto out;
>> + }
>> +
>> fs_info->qgroup_ulist = ulist_alloc(GFP_KERNEL);
>> if (!fs_info->qgroup_ulist) {
>> ret = -ENOMEM;
>> @@ -987,6 +993,11 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
>> fs_info->quota_root = quota_root;
>> set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
>> spin_unlock(&fs_info->qgroup_lock);
>> +
>> + ret = btrfs_commit_transaction(trans);
>> + if (ret)
>> + goto out_free_path;
>> +
>> ret = qgroup_rescan_init(fs_info, 0, 1);
>
> However, I'm not sure if this approach completely works well when some files
> are
> concurrently written while quota is being enabled.
> Since before the commit 5d23515be669, quota_rescan_init() is called during
> transaction
> commit, but now quota_rescan_init() is called outside of transacation.
> So, is there still a slight possibility that the same problem occurs here?
This is the tricky part of btrfs quota rescan.
For rescan, it only scans commit root (even before the large quota
rework), and records where the current scanning location.
Then, qgroup code does a trick, if new dirty extents is after current
scanning location, which means later rescan would scan that extent
later, so it skips the accounting and let rescan to handle it.
Currently since qgroup only accounts extent at transaction commit time,
the only possible cause of problems should be the timing of
@qgroup_rescan_progress initialization.
I think we should hold a trans handle when setting
@qgroup_rescan_progress, but I may need extra investigation into the race.
Thanks,
Qu
>
> (I don't completely understands how quota works yet , so correct me if I'm
> wrong.)
>
>> if (!ret) {
>> qgroup_rescan_zero_tracking(fs_info);
>> @@ -3061,7 +3072,7 @@ static int __btrfs_qgroup_release_data(struct inode
>> *inode,
>> if (free && reserved)
>> return qgroup_free_reserved_data(inode, reserved, start, len);
>> extent_changeset_init(&changeset);
>> - ret = clear_record_extent_bits(&BTRFS_I(inode)->io_tree, start,
>> + ret = clear_record_extent_bits(&BTRFS_I(inode)->io_tree, start,
>> start + len -1, EXTENT_QGROUP_RESERVED, &changeset);
>> if (ret < 0)
>> goto out;
>> diff --git a/fs/btrfs/qgroup.h b/fs/btrfs/qgroup.h
>> index d60dd06445ce..3900efab0e70 100644
>> --- a/fs/btrfs/qgroup.h
>> +++ b/fs/btrfs/qgroup.h
>> @@ -141,8 +141,7 @@ struct btrfs_qgroup {
>> #define QGROUP_RELEASE (1<<1)
>> #define QGROUP_FREE (1<<2)
>>
>> -int btrfs_quota_enable(struct btrfs_trans_handle *trans,
>> - struct btrfs_fs_info *fs_info);
>> +int btrfs_quota_enable(struct btrfs_fs_info *fs_info);
>> int btrfs_quota_disable(struct btrfs_trans_handle *trans,
>> struct btrfs_fs_info *fs_info);
>> int btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info);
>>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majord...@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
