| Message ID | 1365161897-19448-2-git-send-email-list.btrfs@jan-o-sch.net (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Fri, Apr 05, 2013 at 05:38:16AM -0600, Jan Schmidt wrote: > If qgroup tracking is out of sync, a rescan operation can be started to > iterate the extent tree and recalculate all qgroup tracking data. > > Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net> > --- > fs/btrfs/ctree.h | 17 ++- > fs/btrfs/disk-io.c | 6 + > fs/btrfs/ioctl.c | 83 +++++++++++--- > fs/btrfs/qgroup.c | 271 +++++++++++++++++++++++++++++++++++++++++-- > include/uapi/linux/btrfs.h | 11 ++- > 5 files changed, 354 insertions(+), 34 deletions(-) > > diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h > index 0d82922..bd4e2a7 100644 > --- a/fs/btrfs/ctree.h > +++ b/fs/btrfs/ctree.h > @@ -1019,9 +1019,9 @@ struct btrfs_block_group_item { > */ > #define BTRFS_QGROUP_STATUS_FLAG_ON (1ULL << 0) > /* > - * SCANNING is set during the initialization phase > + * RESCAN is set during the initialization phase > */ > -#define BTRFS_QGROUP_STATUS_FLAG_SCANNING (1ULL << 1) > +#define BTRFS_QGROUP_STATUS_FLAG_RESCAN (1ULL << 1) > /* > * Some qgroup entries are known to be out of date, > * either because the configuration has changed in a way that > @@ -1050,7 +1050,7 @@ struct btrfs_qgroup_status_item { > * only used during scanning to record the progress > * of the scan. It contains a logical address > */ > - __le64 scan; > + __le64 rescan; > } __attribute__ ((__packed__)); > > struct btrfs_qgroup_info_item { > @@ -1587,6 +1587,11 @@ struct btrfs_fs_info { > /* used by btrfs_qgroup_record_ref for an efficient tree traversal */ > u64 qgroup_seq; > > + /* qgroup rescan items */ > + struct mutex qgroup_rescan_lock; /* protects the progress item */ > + struct btrfs_key qgroup_rescan_progress; > + struct btrfs_workers qgroup_rescan_workers; > + > /* filesystem state */ > unsigned long fs_state; > > @@ -2864,8 +2869,8 @@ BTRFS_SETGET_FUNCS(qgroup_status_version, struct btrfs_qgroup_status_item, > version, 64); > BTRFS_SETGET_FUNCS(qgroup_status_flags, struct btrfs_qgroup_status_item, > flags, 64); > -BTRFS_SETGET_FUNCS(qgroup_status_scan, struct btrfs_qgroup_status_item, > - scan, 64); > +BTRFS_SETGET_FUNCS(qgroup_status_rescan, struct btrfs_qgroup_status_item, > + rescan, 64); > > /* btrfs_qgroup_info_item */ > BTRFS_SETGET_FUNCS(qgroup_info_generation, struct btrfs_qgroup_info_item, > @@ -3784,7 +3789,7 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans, > struct btrfs_fs_info *fs_info); > int btrfs_quota_disable(struct btrfs_trans_handle *trans, > struct btrfs_fs_info *fs_info); > -int btrfs_quota_rescan(struct btrfs_fs_info *fs_info); > +int btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info); > int btrfs_add_qgroup_relation(struct btrfs_trans_handle *trans, > struct btrfs_fs_info *fs_info, u64 src, u64 dst); > int btrfs_del_qgroup_relation(struct btrfs_trans_handle *trans, > diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c > index 6d19a0a..60d15fe 100644 > --- a/fs/btrfs/disk-io.c > +++ b/fs/btrfs/disk-io.c > @@ -2192,6 +2192,7 @@ int open_ctree(struct super_block *sb, > fs_info->qgroup_seq = 1; > fs_info->quota_enabled = 0; > fs_info->pending_quota_state = 0; > + mutex_init(&fs_info->qgroup_rescan_lock); > > btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); > btrfs_init_free_cluster(&fs_info->data_alloc_cluster); > @@ -2394,6 +2395,8 @@ int open_ctree(struct super_block *sb, > btrfs_init_workers(&fs_info->readahead_workers, "readahead", > fs_info->thread_pool_size, > &fs_info->generic_worker); > + btrfs_init_workers(&fs_info->qgroup_rescan_workers, "qgroup-rescan", 1, > + &fs_info->generic_worker); > > /* > * endios are largely parallel and should have a very > @@ -2428,6 +2431,7 @@ int open_ctree(struct super_block *sb, > ret |= btrfs_start_workers(&fs_info->caching_workers); > ret |= btrfs_start_workers(&fs_info->readahead_workers); > ret |= btrfs_start_workers(&fs_info->flush_workers); > + ret |= btrfs_start_workers(&fs_info->qgroup_rescan_workers); > if (ret) { > err = -ENOMEM; > goto fail_sb_buffer; > @@ -2773,6 +2777,7 @@ fail_sb_buffer: > btrfs_stop_workers(&fs_info->delayed_workers); > btrfs_stop_workers(&fs_info->caching_workers); > btrfs_stop_workers(&fs_info->flush_workers); > + btrfs_stop_workers(&fs_info->qgroup_rescan_workers); > fail_alloc: > fail_iput: > btrfs_mapping_tree_free(&fs_info->mapping_tree); > @@ -3463,6 +3468,7 @@ int close_ctree(struct btrfs_root *root) > btrfs_stop_workers(&fs_info->caching_workers); > btrfs_stop_workers(&fs_info->readahead_workers); > btrfs_stop_workers(&fs_info->flush_workers); > + btrfs_stop_workers(&fs_info->qgroup_rescan_workers); > > #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY > if (btrfs_test_opt(root, CHECK_INTEGRITY)) > diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c > index 898c572..a4681fb 100644 > --- a/fs/btrfs/ioctl.c > +++ b/fs/btrfs/ioctl.c > @@ -3693,12 +3693,10 @@ static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) > goto drop_write; > } > > - if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) { > - trans = btrfs_start_transaction(root, 2); > - if (IS_ERR(trans)) { > - ret = PTR_ERR(trans); > - goto out; > - } > + trans = btrfs_start_transaction(root, 2); > + if (IS_ERR(trans)) { > + ret = PTR_ERR(trans); > + goto out; > } > > switch (sa->cmd) { > @@ -3708,9 +3706,6 @@ static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) > case BTRFS_QUOTA_CTL_DISABLE: > ret = btrfs_quota_disable(trans, root->fs_info); > break; > - case BTRFS_QUOTA_CTL_RESCAN: > - ret = btrfs_quota_rescan(root->fs_info); > - break; > default: > ret = -EINVAL; > break; > @@ -3719,11 +3714,9 @@ static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) > if (copy_to_user(arg, sa, sizeof(*sa))) > ret = -EFAULT; > > - if (trans) { > - err = btrfs_commit_transaction(trans, root); > - if (err && !ret) > - ret = err; > - } > + err = btrfs_commit_transaction(trans, root); > + if (err && !ret) > + ret = err; > out: > kfree(sa); > drop_write: > @@ -3877,6 +3870,64 @@ drop_write: > return ret; > } > > +static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) > +{ > + struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; > + struct btrfs_ioctl_quota_rescan_args *qsa; > + int ret; > + > + if (!capable(CAP_SYS_ADMIN)) > + return -EPERM; > + > + ret = mnt_want_write_file(file); > + if (ret) > + return ret; > + > + qsa = memdup_user(arg, sizeof(*qsa)); > + if (IS_ERR(qsa)) { > + ret = PTR_ERR(qsa); > + goto drop_write; > + } > + > + if (qsa->flags) { > + ret = -EINVAL; > + goto out; > + } > + > + ret = btrfs_qgroup_rescan(root->fs_info); > + > +out: > + kfree(qsa); > +drop_write: > + mnt_drop_write_file(file); > + return ret; > +} > + > +static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) > +{ > + struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; > + struct btrfs_ioctl_quota_rescan_args *qsa; > + int ret = 0; > + > + if (!capable(CAP_SYS_ADMIN)) > + return -EPERM; > + > + qsa = kzalloc(sizeof(*qsa), GFP_NOFS); > + if (IS_ERR(qsa)) > + return PTR_ERR(qsa); > + IIRC kzalloc returns NULL if something was wrong, not an ERR_PTR(). 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 Fri, Apr 05, 2013 at 09:05:03AM -0400, Josef Bacik wrote: > On Fri, Apr 05, 2013 at 05:38:16AM -0600, Jan Schmidt wrote: > > +static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) > > +{ > > + struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; > > + struct btrfs_ioctl_quota_rescan_args *qsa; > > + int ret = 0; > > + > > + if (!capable(CAP_SYS_ADMIN)) > > + return -EPERM; > > + > > + qsa = kzalloc(sizeof(*qsa), GFP_NOFS); > > + if (IS_ERR(qsa)) > > + return PTR_ERR(qsa); > > + > > IIRC kzalloc returns NULL if something was wrong, not an ERR_PTR(). Thanks, And we can get rid of kzalloc/kfree entirely, btrfs_ioctl_quota_rescan_args is only 2x u64. david -- 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 Fri, Apr 05, 2013 at 01:38:16PM +0200, Jan Schmidt wrote: > + if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { I was wondering if merging qgroup_flags with fs_state would make sense to you. There are currently 3 bits for qgroups, and it's a whole filesystem state anyway. > + pr_debug("current progress key (%llu %u %llu), search_slot ret %d\n", > + fs_info->qgroup_rescan_progress.objectid, > + fs_info->qgroup_rescan_progress.type, > + fs_info->qgroup_rescan_progress.offset, ret); Please use (unsigned long long) for u64 types (objectid, offset). > --- a/include/uapi/linux/btrfs.h > +++ b/include/uapi/linux/btrfs.h > @@ -376,12 +376,17 @@ struct btrfs_ioctl_get_dev_stats { > > #define BTRFS_QUOTA_CTL_ENABLE 1 > #define BTRFS_QUOTA_CTL_DISABLE 2 > -#define BTRFS_QUOTA_CTL_RESCAN 3 > +/* 3 has formerly been reserved for BTRFS_QUOTA_CTL_RESCAN */ It's not clear if 3 can be reused or not. If yes, there's no need for the comment, if not, a #define should be left in place to capture the value. > struct btrfs_ioctl_quota_ctl_args { > __u64 cmd; > __u64 status; > }; > > +struct btrfs_ioctl_quota_rescan_args { > + __u64 flags; > + __u64 progress; Like I'm always worried about future etensions, adding a few reserved[] does not hurt. david -- 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 Wed, April 10, 2013 at 18:47 (+0200), David Sterba wrote: > On Fri, Apr 05, 2013 at 01:38:16PM +0200, Jan Schmidt wrote: >> + if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { > > I was wondering if merging qgroup_flags with fs_state would make sense > to you. There are currently 3 bits for qgroups, and it's a whole > filesystem state anyway. Well, yes, it can be done. I don't think it saves enough to justify the effort, but feel free if you like. It's definitely not the purpose of this patch set :-) > >> + pr_debug("current progress key (%llu %u %llu), search_slot ret %d\n", >> + fs_info->qgroup_rescan_progress.objectid, >> + fs_info->qgroup_rescan_progress.type, >> + fs_info->qgroup_rescan_progress.offset, ret); > > Please use (unsigned long long) for u64 types (objectid, offset). > >> --- a/include/uapi/linux/btrfs.h >> +++ b/include/uapi/linux/btrfs.h >> @@ -376,12 +376,17 @@ struct btrfs_ioctl_get_dev_stats { >> >> #define BTRFS_QUOTA_CTL_ENABLE 1 >> #define BTRFS_QUOTA_CTL_DISABLE 2 >> -#define BTRFS_QUOTA_CTL_RESCAN 3 >> +/* 3 has formerly been reserved for BTRFS_QUOTA_CTL_RESCAN */ > > It's not clear if 3 can be reused or not. If yes, there's no need for > the comment, if not, a #define should be left in place to capture the > value. It shouldn't. I just wanted to 1) make sure there are no users left and 2) point out to the next rescan code editor that this is not the flag he should be looking for. My new approach to satisfy both goals now is to leave the #define in there and add __UNUSED to the end of the name. >> struct btrfs_ioctl_quota_ctl_args { >> __u64 cmd; >> __u64 status; >> }; >> >> +struct btrfs_ioctl_quota_rescan_args { >> + __u64 flags; >> + __u64 progress; > > Like I'm always worried about future etensions, adding a few reserved[] > does not hurt. Thanks, v2 to come. -Jan -- 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 Mon, April 15, 2013 at 07:44 (+0200), Jan Schmidt wrote:
> Thanks, v2 to come.
Uh, but not immediately. I didn't get tracking of "exclusive" right. That will
need some time to fix and test.
-Jan
--
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
Hello Jan, > On Mon, April 15, 2013 at 07:44 (+0200), Jan Schmidt wrote: >> Thanks, v2 to come. > > Uh, but not immediately. I didn't get tracking of "exclusive" right. That will > need some time to fix and test. 'exclusive' adds the complexity of btrfs qgroup. So if you send V2. I'd like you add more lines in changelog. Besides, i have a question in my mind.(I have not seen you code).. When qgroup rescan will happen? 1> when quota is enabled 2> if a new qgroup relations is created, rescan should happen? 2> user call qgroup rescan.. Thanks, Wang > > -Jan > -- > 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 > > -- 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 Mon, April 15, 2013 at 08:08 (+0200), Wang Shilong wrote: > Hello Jan, > >> On Mon, April 15, 2013 at 07:44 (+0200), Jan Schmidt wrote: >>> Thanks, v2 to come. >> >> Uh, but not immediately. I didn't get tracking of "exclusive" right. That will >> need some time to fix and test. > > > 'exclusive' adds the complexity of btrfs qgroup. > So if you send V2. I'd like you add more lines in changelog. Yes, the commit message will be longer as you requested previously. This does not include a complete description on how "exclusive" works. The qgroup pdf explains that. > Besides, i have a question in my mind.(I have not seen you code).. > When qgroup rescan will happen? > > 1> when quota is enabled That's what the second patch does, yes. Your patches should be merged in a way that we first create the level 0 qgroups for all subvolumes and then start the rescan, obviously. > 2> if a new qgroup relations is created, rescan should happen? With your patches, there will be no subvolume qgroups missing. For the higher level groups, one needs expert knowledge anyway. I think it's best to leave that decision to the administrator configuring those qgroups. > 2> user call qgroup rescan.. Of course, yes. -Jan -- 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
Hello Jan, > On Mon, April 15, 2013 at 08:08 (+0200), Wang Shilong wrote: >> Hello Jan, >> >>> On Mon, April 15, 2013 at 07:44 (+0200), Jan Schmidt wrote: >>>> Thanks, v2 to come. >>> Uh, but not immediately. I didn't get tracking of "exclusive" right. That will >>> need some time to fix and test. >> >> 'exclusive' adds the complexity of btrfs qgroup. >> So if you send V2. I'd like you add more lines in changelog. > > Yes, the commit message will be longer as you requested previously. This does > not include a complete description on how "exclusive" works. The qgroup pdf > explains that. Yeah, changelog really helps for newbies(like me ^_^) > >> Besides, i have a question in my mind.(I have not seen you code).. >> When qgroup rescan will happen? >> >> 1> when quota is enabled > > That's what the second patch does, yes. Your patches should be merged in a way > that we first create the level 0 qgroups for all subvolumes and then start the > rescan, obviously. > >> 2> if a new qgroup relations is created, rescan should happen? > > With your patches, there will be no subvolume qgroups missing. For the higher > level groups, one needs expert knowledge anyway. I think it's best to leave that > decision to the administrator configuring those qgroups. IMO, it is better that qgroup rescan automatically if a qgroup relation is added. Thank, Wang > >> 2> user call qgroup rescan.. > > Of course, yes. > > -Jan > -- > 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 > > -- 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 15.04.2013 09:19, Wang Shilong wrote: > Hello Jan, > >> On Mon, April 15, 2013 at 08:08 (+0200), Wang Shilong wrote: > >>> Hello Jan, >>> >>>> On Mon, April 15, 2013 at 07:44 (+0200), Jan Schmidt wrote: >>>>> Thanks, v2 to come. >>>> Uh, but not immediately. I didn't get tracking of "exclusive" right. That will >>>> need some time to fix and test. >>> >>> 'exclusive' adds the complexity of btrfs qgroup. >>> So if you send V2. I'd like you add more lines in changelog. >> >> Yes, the commit message will be longer as you requested previously. This does >> not include a complete description on how "exclusive" works. The qgroup pdf >> explains that. > > > Yeah, changelog really helps for newbies(like me ^_^) > >> >>> Besides, i have a question in my mind.(I have not seen you code).. >>> When qgroup rescan will happen? >>> >>> 1> when quota is enabled >> >> That's what the second patch does, yes. Your patches should be merged in a way >> that we first create the level 0 qgroups for all subvolumes and then start the >> rescan, obviously. >> >>> 2> if a new qgroup relations is created, rescan should happen? >> >> With your patches, there will be no subvolume qgroups missing. For the higher >> level groups, one needs expert knowledge anyway. I think it's best to leave that >> decision to the administrator configuring those qgroups. > > > IMO, it is better that qgroup rescan automatically if a qgroup relation > is added. I think we should leave it as it is, for several reasons: a) the administrator might want to add multiple configurations. Only after the last one a rescan makes sense. b) the config changes might not make a rescan necessary. For example, you could prepare a configuration for newly created or not-yet-existent qgroups. In this case, a rescan of the complete filesystem might be unnecessary. A rescan might even be harmful, as during the rescan the quota is wrong and users can go over quota. Also, a rescan adds significant load to the system. -Arne > > Thank, > Wang > >> >>> 2> user call qgroup rescan.. >> >> Of course, yes. >> >> -Jan >> -- >> 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 >> >> > > > > -- > 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 -- 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/ctree.h b/fs/btrfs/ctree.h index 0d82922..bd4e2a7 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1019,9 +1019,9 @@ struct btrfs_block_group_item { */ #define BTRFS_QGROUP_STATUS_FLAG_ON (1ULL << 0) /* - * SCANNING is set during the initialization phase + * RESCAN is set during the initialization phase */ -#define BTRFS_QGROUP_STATUS_FLAG_SCANNING (1ULL << 1) +#define BTRFS_QGROUP_STATUS_FLAG_RESCAN (1ULL << 1) /* * Some qgroup entries are known to be out of date, * either because the configuration has changed in a way that @@ -1050,7 +1050,7 @@ struct btrfs_qgroup_status_item { * only used during scanning to record the progress * of the scan. It contains a logical address */ - __le64 scan; + __le64 rescan; } __attribute__ ((__packed__)); struct btrfs_qgroup_info_item { @@ -1587,6 +1587,11 @@ struct btrfs_fs_info { /* used by btrfs_qgroup_record_ref for an efficient tree traversal */ u64 qgroup_seq; + /* qgroup rescan items */ + struct mutex qgroup_rescan_lock; /* protects the progress item */ + struct btrfs_key qgroup_rescan_progress; + struct btrfs_workers qgroup_rescan_workers; + /* filesystem state */ unsigned long fs_state; @@ -2864,8 +2869,8 @@ BTRFS_SETGET_FUNCS(qgroup_status_version, struct btrfs_qgroup_status_item, version, 64); BTRFS_SETGET_FUNCS(qgroup_status_flags, struct btrfs_qgroup_status_item, flags, 64); -BTRFS_SETGET_FUNCS(qgroup_status_scan, struct btrfs_qgroup_status_item, - scan, 64); +BTRFS_SETGET_FUNCS(qgroup_status_rescan, struct btrfs_qgroup_status_item, + rescan, 64); /* btrfs_qgroup_info_item */ BTRFS_SETGET_FUNCS(qgroup_info_generation, struct btrfs_qgroup_info_item, @@ -3784,7 +3789,7 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info); int btrfs_quota_disable(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info); -int btrfs_quota_rescan(struct btrfs_fs_info *fs_info); +int btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info); int btrfs_add_qgroup_relation(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info, u64 src, u64 dst); int btrfs_del_qgroup_relation(struct btrfs_trans_handle *trans, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 6d19a0a..60d15fe 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -2192,6 +2192,7 @@ int open_ctree(struct super_block *sb, fs_info->qgroup_seq = 1; fs_info->quota_enabled = 0; fs_info->pending_quota_state = 0; + mutex_init(&fs_info->qgroup_rescan_lock); btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); btrfs_init_free_cluster(&fs_info->data_alloc_cluster); @@ -2394,6 +2395,8 @@ int open_ctree(struct super_block *sb, btrfs_init_workers(&fs_info->readahead_workers, "readahead", fs_info->thread_pool_size, &fs_info->generic_worker); + btrfs_init_workers(&fs_info->qgroup_rescan_workers, "qgroup-rescan", 1, + &fs_info->generic_worker); /* * endios are largely parallel and should have a very @@ -2428,6 +2431,7 @@ int open_ctree(struct super_block *sb, ret |= btrfs_start_workers(&fs_info->caching_workers); ret |= btrfs_start_workers(&fs_info->readahead_workers); ret |= btrfs_start_workers(&fs_info->flush_workers); + ret |= btrfs_start_workers(&fs_info->qgroup_rescan_workers); if (ret) { err = -ENOMEM; goto fail_sb_buffer; @@ -2773,6 +2777,7 @@ fail_sb_buffer: btrfs_stop_workers(&fs_info->delayed_workers); btrfs_stop_workers(&fs_info->caching_workers); btrfs_stop_workers(&fs_info->flush_workers); + btrfs_stop_workers(&fs_info->qgroup_rescan_workers); fail_alloc: fail_iput: btrfs_mapping_tree_free(&fs_info->mapping_tree); @@ -3463,6 +3468,7 @@ int close_ctree(struct btrfs_root *root) btrfs_stop_workers(&fs_info->caching_workers); btrfs_stop_workers(&fs_info->readahead_workers); btrfs_stop_workers(&fs_info->flush_workers); + btrfs_stop_workers(&fs_info->qgroup_rescan_workers); #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY if (btrfs_test_opt(root, CHECK_INTEGRITY)) diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 898c572..a4681fb 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -3693,12 +3693,10 @@ static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) goto drop_write; } - if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) { - trans = btrfs_start_transaction(root, 2); - if (IS_ERR(trans)) { - ret = PTR_ERR(trans); - goto out; - } + trans = btrfs_start_transaction(root, 2); + if (IS_ERR(trans)) { + ret = PTR_ERR(trans); + goto out; } switch (sa->cmd) { @@ -3708,9 +3706,6 @@ static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) case BTRFS_QUOTA_CTL_DISABLE: ret = btrfs_quota_disable(trans, root->fs_info); break; - case BTRFS_QUOTA_CTL_RESCAN: - ret = btrfs_quota_rescan(root->fs_info); - break; default: ret = -EINVAL; break; @@ -3719,11 +3714,9 @@ static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg) if (copy_to_user(arg, sa, sizeof(*sa))) ret = -EFAULT; - if (trans) { - err = btrfs_commit_transaction(trans, root); - if (err && !ret) - ret = err; - } + err = btrfs_commit_transaction(trans, root); + if (err && !ret) + ret = err; out: kfree(sa); drop_write: @@ -3877,6 +3870,64 @@ drop_write: return ret; } +static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg) +{ + struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; + struct btrfs_ioctl_quota_rescan_args *qsa; + int ret; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + + ret = mnt_want_write_file(file); + if (ret) + return ret; + + qsa = memdup_user(arg, sizeof(*qsa)); + if (IS_ERR(qsa)) { + ret = PTR_ERR(qsa); + goto drop_write; + } + + if (qsa->flags) { + ret = -EINVAL; + goto out; + } + + ret = btrfs_qgroup_rescan(root->fs_info); + +out: + kfree(qsa); +drop_write: + mnt_drop_write_file(file); + return ret; +} + +static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg) +{ + struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root; + struct btrfs_ioctl_quota_rescan_args *qsa; + int ret = 0; + + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + + qsa = kzalloc(sizeof(*qsa), GFP_NOFS); + if (IS_ERR(qsa)) + return PTR_ERR(qsa); + + if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { + qsa->flags = 1; + qsa->progress = root->fs_info->qgroup_rescan_progress.objectid; + } + + if (copy_to_user(arg, qsa, sizeof(*qsa))) + ret = -EFAULT; + + kfree(qsa); + return ret; +} + static long btrfs_ioctl_set_received_subvol(struct file *file, void __user *arg) { @@ -4115,6 +4166,10 @@ long btrfs_ioctl(struct file *file, unsigned int return btrfs_ioctl_qgroup_create(file, argp); case BTRFS_IOC_QGROUP_LIMIT: return btrfs_ioctl_qgroup_limit(file, argp); + case BTRFS_IOC_QUOTA_RESCAN: + return btrfs_ioctl_quota_rescan(file, argp); + case BTRFS_IOC_QUOTA_RESCAN_STATUS: + return btrfs_ioctl_quota_rescan_status(file, argp); case BTRFS_IOC_DEV_REPLACE: return btrfs_ioctl_dev_replace(root, argp); case BTRFS_IOC_GET_FSLABEL: diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c index b44124d..e0d9cf8 100644 --- a/fs/btrfs/qgroup.c +++ b/fs/btrfs/qgroup.c @@ -31,13 +31,13 @@ #include "locking.h" #include "ulist.h" #include "backref.h" +#include "extent_io.h" /* TODO XXX FIXME * - subvol delete -> delete when ref goes to 0? delete limits also? * - reorganize keys * - compressed * - sync - * - rescan * - copy also limits on subvol creation * - limit * - caches fuer ulists @@ -98,6 +98,14 @@ struct btrfs_qgroup_list { struct btrfs_qgroup *member; }; +struct qgroup_rescan { + struct btrfs_work work; + struct btrfs_fs_info *fs_info; +}; + +static void qgroup_rescan_start(struct btrfs_fs_info *fs_info, + struct qgroup_rescan *qscan); + /* must be called with qgroup_lock held */ static struct btrfs_qgroup *find_qgroup_rb(struct btrfs_fs_info *fs_info, u64 qgroupid) @@ -298,7 +306,20 @@ int btrfs_read_qgroup_config(struct btrfs_fs_info *fs_info) } fs_info->qgroup_flags = btrfs_qgroup_status_flags(l, ptr); - /* FIXME read scan element */ + fs_info->qgroup_rescan_progress.objectid = + btrfs_qgroup_status_rescan(l, ptr); + if (fs_info->qgroup_flags & + BTRFS_QGROUP_STATUS_FLAG_RESCAN) { + struct qgroup_rescan *qscan = + kmalloc(sizeof(*qscan), GFP_NOFS); + if (!qscan) { + ret = -ENOMEM; + goto out; + } + fs_info->qgroup_rescan_progress.type = 0; + fs_info->qgroup_rescan_progress.offset = 0; + qgroup_rescan_start(fs_info, qscan); + } goto next1; } @@ -719,9 +740,12 @@ static int update_qgroup_status_item(struct btrfs_trans_handle *trans, l = path->nodes[0]; slot = path->slots[0]; ptr = btrfs_item_ptr(l, slot, struct btrfs_qgroup_status_item); + spin_lock(&fs_info->qgroup_lock); btrfs_set_qgroup_status_flags(l, ptr, fs_info->qgroup_flags); btrfs_set_qgroup_status_generation(l, ptr, trans->transid); - /* XXX scan */ + btrfs_set_qgroup_status_rescan(l, ptr, + fs_info->qgroup_rescan_progress.objectid); + spin_unlock(&fs_info->qgroup_lock); btrfs_mark_buffer_dirty(l); @@ -830,7 +854,7 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans, fs_info->qgroup_flags = BTRFS_QGROUP_STATUS_FLAG_ON | BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT; btrfs_set_qgroup_status_flags(leaf, ptr, fs_info->qgroup_flags); - btrfs_set_qgroup_status_scan(leaf, ptr, 0); + btrfs_set_qgroup_status_rescan(leaf, ptr, 0); btrfs_mark_buffer_dirty(leaf); @@ -894,9 +918,218 @@ out: return ret; } -int btrfs_quota_rescan(struct btrfs_fs_info *fs_info) +static void qgroup_dirty(struct btrfs_fs_info *fs_info, + struct btrfs_qgroup *qgroup) +{ + if (list_empty(&qgroup->dirty)) + list_add(&qgroup->dirty, &fs_info->dirty_qgroups); +} + +/* + * returns < 0 on error, 0 when more leafs are to be scanned. + * returns 1 when done, 2 when done and FLAG_INCONSISTENT was cleared. + */ +static int +qgroup_rescan_leaf(struct qgroup_rescan *qscan, struct btrfs_path *path, + struct btrfs_trans_handle *trans, struct ulist *tmp, + struct extent_buffer *scratch_leaf) +{ + struct btrfs_key found; + struct btrfs_fs_info *fs_info = qscan->fs_info; + struct ulist *roots = NULL; + struct ulist_node *unode; + struct ulist_iterator uiter; + struct seq_list tree_mod_seq_elem = {}; + int slot; + int ret; + + path->leave_spinning = 1; + mutex_lock(&fs_info->qgroup_rescan_lock); + ret = btrfs_search_slot_for_read(fs_info->extent_root, + &fs_info->qgroup_rescan_progress, + path, 1, 0); + if (ret == 1) { + fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN; + if (fs_info->qgroup_flags & + BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT) { + fs_info->qgroup_flags &= + ~BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT; + ret = 2; + } + } else if (ret == 0) { + btrfs_item_key_to_cpu(path->nodes[0], &found, + btrfs_header_nritems(path->nodes[0]) - 1); + fs_info->qgroup_rescan_progress.objectid = found.objectid + 1; + } + mutex_unlock(&fs_info->qgroup_rescan_lock); + + pr_debug("current progress key (%llu %u %llu), search_slot ret %d\n", + fs_info->qgroup_rescan_progress.objectid, + fs_info->qgroup_rescan_progress.type, + fs_info->qgroup_rescan_progress.offset, ret); + + if (ret) { + btrfs_release_path(path); + return ret; + } + + btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem); + memcpy(scratch_leaf, path->nodes[0], sizeof(*scratch_leaf)); + slot = path->slots[0]; + btrfs_release_path(path); + + for (; slot < btrfs_header_nritems(scratch_leaf); ++slot) { + btrfs_item_key_to_cpu(scratch_leaf, &found, slot); + if (found.type != BTRFS_EXTENT_ITEM_KEY) + continue; + ret = btrfs_find_all_roots(trans, fs_info, found.objectid, + tree_mod_seq_elem.seq, &roots); + if (ret < 0) + break; + spin_lock(&fs_info->qgroup_lock); + ULIST_ITER_INIT(&uiter); + while ((unode = ulist_next(roots, &uiter))) { + struct btrfs_qgroup *qg; + struct ulist_node *tmp_unode; + struct ulist_iterator tmp_uiter; + + qg = find_qgroup_rb(fs_info, unode->val); + if (!qg) + continue; + + ulist_reinit(tmp); + ulist_add(tmp, qg->qgroupid, (uintptr_t)qg, GFP_ATOMIC); + ULIST_ITER_INIT(&tmp_uiter); + while ((tmp_unode = ulist_next(tmp, &tmp_uiter))) { + struct btrfs_qgroup_list *glist; + + qg = (struct btrfs_qgroup *)(uintptr_t) + tmp_unode->aux; + + qg->rfer += found.offset; + qg->rfer_cmpr += found.offset; + if (roots->nnodes == 1) { + qg->excl += found.offset; + qg->excl_cmpr += found.offset; + } + qgroup_dirty(fs_info, qg); + + list_for_each_entry(glist, &qg->groups, + next_group) { + ulist_add(tmp, glist->group->qgroupid, + (uintptr_t)glist->group, + GFP_ATOMIC); + } + } + } + spin_unlock(&fs_info->qgroup_lock); + ulist_free(roots); + } + + btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem); + + return ret; +} + +static void btrfs_qgroup_rescan_worker(struct btrfs_work *work) { - /* FIXME */ + struct qgroup_rescan *qscan = container_of(work, struct qgroup_rescan, + work); + struct btrfs_path *path; + struct btrfs_trans_handle *trans = NULL; + struct btrfs_fs_info *fs_info = qscan->fs_info; + struct ulist *tmp = NULL; + struct extent_buffer *scratch_leaf = NULL; + int err = -ENOMEM; + + path = btrfs_alloc_path(); + if (!path) + goto out; + tmp = ulist_alloc(GFP_NOFS); + if (!tmp) + goto out; + scratch_leaf = kmalloc(sizeof(*scratch_leaf), GFP_NOFS); + if (!scratch_leaf) + goto out; + + err = 0; + while (!err) { + trans = btrfs_start_transaction(fs_info->fs_root, 0); + if (IS_ERR(trans)) { + err = PTR_ERR(trans); + break; + } + err = qgroup_rescan_leaf(qscan, path, trans, tmp, scratch_leaf); + if (err > 0) + btrfs_commit_transaction(trans, fs_info->fs_root); + else + btrfs_end_transaction(trans, fs_info->fs_root); + } + +out: + kfree(scratch_leaf); + ulist_free(tmp); + btrfs_free_path(path); + kfree(qscan); + + if (err < 0) + pr_err("btrfs: qgroup scan failed with %d\n", err); + else + pr_info("btrfs: qgroup scan completed%s\n", + err == 2 ? " (inconsistency flag cleared)" : ""); +} + +static void +qgroup_rescan_start(struct btrfs_fs_info *fs_info, struct qgroup_rescan *qscan) +{ + qscan->work.func = btrfs_qgroup_rescan_worker; + qscan->fs_info = fs_info; + + pr_info("btrfs: qgroup scan started\n"); + btrfs_queue_worker(&fs_info->qgroup_rescan_workers, &qscan->work); +} + +int +btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info) +{ + int ret = 0; + struct rb_node *n; + struct btrfs_qgroup *qgroup; + struct qgroup_rescan *qscan = kmalloc(sizeof(*qscan), GFP_NOFS); + + if (!qscan) + return -ENOMEM; + + mutex_lock(&fs_info->qgroup_rescan_lock); + spin_lock(&fs_info->qgroup_lock); + if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) + ret = -EINPROGRESS; + else if (!(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_ON)) + ret = -EINVAL; + if (ret) { + spin_unlock(&fs_info->qgroup_lock); + mutex_unlock(&fs_info->qgroup_rescan_lock); + kfree(qscan); + return ret; + } + + fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_RESCAN; + memset(&fs_info->qgroup_rescan_progress, 0, + sizeof(fs_info->qgroup_rescan_progress)); + + /* clear all current qgroup tracking information */ + for (n = rb_first(&fs_info->qgroup_tree); n; n = rb_next(n)) { + qgroup = rb_entry(n, struct btrfs_qgroup, node); + qgroup->rfer = 0; + qgroup->rfer_cmpr = 0; + qgroup->excl = 0; + qgroup->excl_cmpr = 0; + } + spin_unlock(&fs_info->qgroup_lock); + mutex_unlock(&fs_info->qgroup_rescan_lock); + + qgroup_rescan_start(fs_info, qscan); + return 0; } @@ -1045,13 +1278,6 @@ unlock: return ret; } -static void qgroup_dirty(struct btrfs_fs_info *fs_info, - struct btrfs_qgroup *qgroup) -{ - if (list_empty(&qgroup->dirty)) - list_add(&qgroup->dirty, &fs_info->dirty_qgroups); -} - /* * btrfs_qgroup_record_ref is called when the ref is added or deleted. it puts * the modification into a list that's later used by btrfs_end_transaction to @@ -1142,6 +1368,15 @@ int btrfs_qgroup_account_ref(struct btrfs_trans_handle *trans, BUG(); } + mutex_lock(&fs_info->qgroup_rescan_lock); + if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { + if (fs_info->qgroup_rescan_progress.objectid <= node->bytenr) { + mutex_unlock(&fs_info->qgroup_rescan_lock); + return 0; + } + } + mutex_unlock(&fs_info->qgroup_rescan_lock); + /* * the delayed ref sequence number we pass depends on the direction of * the operation. for add operations, we pass (node->seq - 1) to skip @@ -1155,7 +1390,17 @@ int btrfs_qgroup_account_ref(struct btrfs_trans_handle *trans, if (ret < 0) return ret; + mutex_lock(&fs_info->qgroup_rescan_lock); spin_lock(&fs_info->qgroup_lock); + if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { + if (fs_info->qgroup_rescan_progress.objectid <= node->bytenr) { + ret = 0; + mutex_unlock(&fs_info->qgroup_rescan_lock); + goto unlock; + } + } + mutex_unlock(&fs_info->qgroup_rescan_lock); + quota_root = fs_info->quota_root; if (!quota_root) goto unlock; diff --git a/include/uapi/linux/btrfs.h b/include/uapi/linux/btrfs.h index fa3a5f9..eff1ab9 100644 --- a/include/uapi/linux/btrfs.h +++ b/include/uapi/linux/btrfs.h @@ -376,12 +376,17 @@ struct btrfs_ioctl_get_dev_stats { #define BTRFS_QUOTA_CTL_ENABLE 1 #define BTRFS_QUOTA_CTL_DISABLE 2 -#define BTRFS_QUOTA_CTL_RESCAN 3 +/* 3 has formerly been reserved for BTRFS_QUOTA_CTL_RESCAN */ struct btrfs_ioctl_quota_ctl_args { __u64 cmd; __u64 status; }; +struct btrfs_ioctl_quota_rescan_args { + __u64 flags; + __u64 progress; +}; + struct btrfs_ioctl_qgroup_assign_args { __u64 assign; __u64 src; @@ -502,6 +507,10 @@ struct btrfs_ioctl_send_args { struct btrfs_ioctl_qgroup_create_args) #define BTRFS_IOC_QGROUP_LIMIT _IOR(BTRFS_IOCTL_MAGIC, 43, \ struct btrfs_ioctl_qgroup_limit_args) +#define BTRFS_IOC_QUOTA_RESCAN _IOW(BTRFS_IOCTL_MAGIC, 44, \ + struct btrfs_ioctl_quota_rescan_args) +#define BTRFS_IOC_QUOTA_RESCAN_STATUS _IOR(BTRFS_IOCTL_MAGIC, 45, \ + struct btrfs_ioctl_quota_rescan_args) #define BTRFS_IOC_GET_FSLABEL _IOR(BTRFS_IOCTL_MAGIC, 49, \ char[BTRFS_LABEL_SIZE]) #define BTRFS_IOC_SET_FSLABEL _IOW(BTRFS_IOCTL_MAGIC, 50, \
If qgroup tracking is out of sync, a rescan operation can be started to iterate the extent tree and recalculate all qgroup tracking data. Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net> --- fs/btrfs/ctree.h | 17 ++- fs/btrfs/disk-io.c | 6 + fs/btrfs/ioctl.c | 83 +++++++++++--- fs/btrfs/qgroup.c | 271 +++++++++++++++++++++++++++++++++++++++++-- include/uapi/linux/btrfs.h | 11 ++- 5 files changed, 354 insertions(+), 34 deletions(-)