Skip to contentSkip to conversation

btrfs, sysfs cleanup and add dev_state

PATCH RESEND In review Anand Jain 2 participants 11 messages Feb 3, 2020 – Feb 12, 2020 Btrfs file system

lore
Share & tools
lore

Anand Jain

author
#0 lore raw

PATCH RESEND v5 4/4 btrfs, sysfs cleanup and add dev_state

Resend:
Patch 4/4 is already integrated without the cleanup and preparatory patches (1,2,3)/4. So I am resending those 3 patches. And the changes in patch 4/4 as in misc-next is imported into patch v5 4/4 here. Patch 4/4 has the details of the changes.

Patch 1/4 is a cleanup patch.
Patch 2/4 adds the kobject devinfo which is a preparatory to patch 4/4. Patch 3/4 was sent before, no functional changes, change log is updated. Patch 4/4 creates the attribute dev_state.

Anand Jain (4):
btrfs: sysfs, use btrfs_sysfs_remove_fsid in fail return in add_fsid
btrfs: sysfs, add UUID/devinfo kobject
btrfs: sysfs, rename device_link add,remove functions
btrfs: sysfs, add devid/dev_state kobject and device attribute

 fs/btrfs/dev-replace.c |   4 +-
 fs/btrfs/sysfs.c       | 156 +++++++++++++++++++++++++++++++++++++++----------
 fs/btrfs/sysfs.h       |   4 +-
 fs/btrfs/volumes.c     |   8 +--
 fs/btrfs/volumes.h     |   3 +
 5 files changed, 137 insertions(+), 38 deletions(-)
Signature · 1 line
1.8.3.1
author
#1 lore raw

PATCH RESEND 1/4 btrfs: sysfs, use btrfs_sysfs_remove_fsid in fail return in add_fsid

We have one simple function btrfs_sysfs_remove_fsid() to cleanup kobjects added by btrfs_sysfs_add_fsid() and calls kobject_put() and kobject_delete() only if the kobject is initialized or not null. So use this function while retreating in the function btrfs_sysfs_add_fsid().

One difference though, earlier we did not call kobject_del() during retreat in btrfs_sysfs_add_fsid() and I did experiment to figureout if that's an error or warning, however I didn't notice any such issues with or without kobject_del() not being called.

So this patch is just for cleanup.

---
 fs/btrfs/sysfs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/btrfs/sysfs.c +1 -1
index 55e4ed1af29c..8def038dc2bd 100644
@@ -1256,7 +1256,7 @@ int btrfs_sysfs_add_fsid(struct btrfs_fs_devices *fs_devs)
 	if (!fs_devs->devices_kobj) {
 		btrfs_err(fs_devs->fs_info,
 			  "failed to init sysfs device interface");
-		kobject_put(&fs_devs->fsid_kobj);
+		btrfs_sysfs_remove_fsid(fs_devs);
 		return -ENOMEM;
 	}
 
Signature · 1 line
2.23.0
author
#2 lore raw

PATCH RESEND 2/4 btrfs: sysfs, add UUID/devinfo kobject

In reply to Anand Jain

Preparatory patch creates kobject /sys/fs/btrfs/UUID/devinfo to hold btrfs_fs_devices::dev_state attribute.

This is being added in the mount context, that means we don't see UUID/devinfo before the mount (yet).

---
 fs/btrfs/sysfs.c   | 15 +++++++++++++++
 fs/btrfs/volumes.h |  1 +
 2 files changed, 16 insertions(+)
fs/btrfs/sysfs.c +15 -0
index 8def038dc2bd..58ef2c04e5be 100644
@@ -901,6 +901,12 @@ static int addrm_unknown_feature_attrs(struct btrfs_fs_info *fs_info, bool add)
 
 static void __btrfs_sysfs_remove_fsid(struct btrfs_fs_devices *fs_devs)
 {
+	if (fs_devs->devinfo_kobj) {
+		kobject_del(fs_devs->devinfo_kobj);
+		kobject_put(fs_devs->devinfo_kobj);
+		fs_devs->devinfo_kobj = NULL;
+	}
+
 	if (fs_devs->devices_kobj) {
 		kobject_del(fs_devs->devices_kobj);
 		kobject_put(fs_devs->devices_kobj);
@@ -1260,6 +1266,15 @@ int btrfs_sysfs_add_fsid(struct btrfs_fs_devices *fs_devs)
 		return -ENOMEM;
 	}
 
+	fs_devs->devinfo_kobj = kobject_create_and_add("devinfo",
+						       &fs_devs->fsid_kobj);
+	if (!fs_devs->devinfo_kobj) {
+		btrfs_err(fs_devs->fs_info,
+			  "failed to init sysfs devinfo kobject");
+		btrfs_sysfs_remove_fsid(fs_devs);
+		return -ENOMEM;
+	}
+
 	return 0;
 }
 
fs/btrfs/volumes.h +1 -0
index 9c7d4fe5c39a..98535f1e208e 100644
@@ -254,6 +254,7 @@ struct btrfs_fs_devices {
 	/* sysfs kobjects */
 	struct kobject fsid_kobj;
 	struct kobject *devices_kobj;
+	struct kobject *devinfo_kobj;
 	struct completion kobj_unregister;
 };
 
Signature · 1 line
2.23.0
author
#3 lore raw

PATCH RESEND v2 3/4 btrfs: sysfs, rename device_link add,remove functions

In reply to Anand Jain

In preparation to add btrfs_device::dev_state attribute in
/sys/fs/btrfs/UUID/devinfo

Rename btrfs_sysfs_add_device_link() and btrfs_sysfs_rm_device_link() to btrfs_sysfs_add_devices_attr() and btrfs_sysfs_remove_devices_attr() as these functions is going to create device attribute rather than just the link to the disk. No functional changes.

---
v2: Update change log. Change ..UUID/devices to ..UUID/devinfo

 fs/btrfs/dev-replace.c |  4 ++--
 fs/btrfs/sysfs.c       | 12 ++++++------
 fs/btrfs/sysfs.h       |  4 ++--
 fs/btrfs/volumes.c     |  8 ++++----
 4 files changed, 14 insertions(+), 14 deletions(-)
fs/btrfs/dev-replace.c +2 -2
index b6a07fa93b53..13a4ebd04a7a 100644
@@ -512,7 +512,7 @@ static int btrfs_dev_replace_start(struct btrfs_fs_info *fs_info,
 	atomic64_set(&dev_replace->num_uncorrectable_read_errors, 0);
 	up_write(&dev_replace->rwsem);
 
-	ret = btrfs_sysfs_add_device_link(tgt_device->fs_devices, tgt_device);
+	ret = btrfs_sysfs_add_devices_attr(tgt_device->fs_devices, tgt_device);
 	if (ret)
 		btrfs_err(fs_info, "kobj add dev failed %d", ret);
 
@@ -743,7 +743,7 @@ static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
 	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
 
 	/* replace the sysfs entry */
-	btrfs_sysfs_rm_device_link(fs_info->fs_devices, src_device);
+	btrfs_sysfs_remove_devices_attr(fs_info->fs_devices, src_device);
 	btrfs_rm_dev_replace_free_srcdev(src_device);
 
 	/* write back the superblocks */
fs/btrfs/sysfs.c +6 -6
index 58ef2c04e5be..10b9bc551330 100644
@@ -960,7 +960,7 @@ void btrfs_sysfs_remove_mounted(struct btrfs_fs_info *fs_info)
 	addrm_unknown_feature_attrs(fs_info, false);
 	sysfs_remove_group(&fs_info->fs_devices->fsid_kobj, &btrfs_feature_attr_group);
 	sysfs_remove_files(&fs_info->fs_devices->fsid_kobj, btrfs_attrs);
-	btrfs_sysfs_rm_device_link(fs_info->fs_devices, NULL);
+	btrfs_sysfs_remove_devices_attr(fs_info->fs_devices, NULL);
 }
 
 static const char * const btrfs_feature_set_names[FEAT_MAX] = {
@@ -1149,7 +1149,7 @@ int btrfs_sysfs_add_space_info_type(struct btrfs_fs_info *fs_info,
 
 /* when one_device is NULL, it removes all device links */
 
-int btrfs_sysfs_rm_device_link(struct btrfs_fs_devices *fs_devices,
+int btrfs_sysfs_remove_devices_attr(struct btrfs_fs_devices *fs_devices,
 		struct btrfs_device *one_device)
 {
 	struct hd_struct *disk;
@@ -1181,8 +1181,8 @@ int btrfs_sysfs_rm_device_link(struct btrfs_fs_devices *fs_devices,
 	return 0;
 }
 
-int btrfs_sysfs_add_device_link(struct btrfs_fs_devices *fs_devices,
-				struct btrfs_device *one_device)
+int btrfs_sysfs_add_devices_attr(struct btrfs_fs_devices *fs_devices,
+				 struct btrfs_device *one_device)
 {
 	int error = 0;
 	struct btrfs_device *dev;
@@ -1286,13 +1286,13 @@ int btrfs_sysfs_add_mounted(struct btrfs_fs_info *fs_info)
 
 	btrfs_set_fs_info_ptr(fs_info);
 
-	error = btrfs_sysfs_add_device_link(fs_devs, NULL);
+	error = btrfs_sysfs_add_devices_attr(fs_devs, NULL);
 	if (error)
 		return error;
 
 	error = sysfs_create_files(fsid_kobj, btrfs_attrs);
 	if (error) {
-		btrfs_sysfs_rm_device_link(fs_devs, NULL);
+		btrfs_sysfs_remove_devices_attr(fs_devs, NULL);
 		return error;
 	}
 
fs/btrfs/sysfs.h +2 -2
index 3d27b39eaf94..9d97b3c8db4e 100644
@@ -14,9 +14,9 @@ enum btrfs_feature_set {
 
 char *btrfs_printable_features(enum btrfs_feature_set set, u64 flags);
 const char * const btrfs_feature_set_name(enum btrfs_feature_set set);
-int btrfs_sysfs_add_device_link(struct btrfs_fs_devices *fs_devices,
+int btrfs_sysfs_add_devices_attr(struct btrfs_fs_devices *fs_devices,
 		struct btrfs_device *one_device);
-int btrfs_sysfs_rm_device_link(struct btrfs_fs_devices *fs_devices,
+int btrfs_sysfs_remove_devices_attr(struct btrfs_fs_devices *fs_devices,
                 struct btrfs_device *one_device);
 int btrfs_sysfs_add_fsid(struct btrfs_fs_devices *fs_devs);
 void btrfs_sysfs_remove_fsid(struct btrfs_fs_devices *fs_devs);
fs/btrfs/volumes.c +4 -4
index 9cfc668f91f4..45dd254afb26 100644
@@ -2054,7 +2054,7 @@ int btrfs_rm_device(struct btrfs_fs_info *fs_info, const char *device_path,
 	if (device->bdev) {
 		cur_devices->open_devices--;
 		/* remove sysfs entry */
-		btrfs_sysfs_rm_device_link(fs_devices, device);
+		btrfs_sysfs_remove_devices_attr(fs_devices, device);
 	}
 
 	num_devices = btrfs_super_num_devices(fs_info->super_copy) - 1;
@@ -2174,7 +2174,7 @@ void btrfs_destroy_dev_replace_tgtdev(struct btrfs_device *tgtdev)
 
 	mutex_lock(&fs_devices->device_list_mutex);
 
-	btrfs_sysfs_rm_device_link(fs_devices, tgtdev);
+	btrfs_sysfs_remove_devices_attr(fs_devices, tgtdev);
 
 	if (tgtdev->bdev)
 		fs_devices->open_devices--;
@@ -2522,7 +2522,7 @@ int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *device_path
 				    orig_super_num_devices + 1);
 
 	/* add sysfs device entry */
-	btrfs_sysfs_add_device_link(fs_devices, device);
+	btrfs_sysfs_add_devices_attr(fs_devices, device);
 
 	/*
 	 * we've got more storage, clear any full flags on the space
@@ -2590,7 +2590,7 @@ int btrfs_init_new_device(struct btrfs_fs_info *fs_info, const char *device_path
 	return ret;
 
 error_sysfs:
-	btrfs_sysfs_rm_device_link(fs_devices, device);
+	btrfs_sysfs_remove_devices_attr(fs_devices, device);
 	mutex_lock(&fs_info->fs_devices->device_list_mutex);
 	mutex_lock(&fs_info->chunk_mutex);
 	list_del_rcu(&device->dev_list);
Signature · 1 line
2.23.0
author
#4 lore raw

PATCH v5 4/4 btrfs: sysfs, add devid/dev_state kobject and device attribute

In reply to Anand Jain

New sysfs attributes that track the filesystem status of devices, stored in the per-filesystem directory in /sys/fs/btrfs/FSID/devinfo . There's a directory for each device, with name corresponding to the numerical device id.

  in_fs_metadata    - device is in the list of fs metadata
  missing           - device is missing (no device node or block device)
  replace_target    - device is target of replace
  writeable         - writes from fs are allowed

These attributes reflect the state of the device::dev_state and created at mount time.

Sample output:

  $ pwd
   /sys/fs/btrfs/6e1961f1-5918-4ecc-a22f-948897b409f7/devinfo/1/
  $ ls
    in_fs_metadata  missing  replace_target  writeable
  $ cat missing
    0

The output from these attributes are 0 or 1. 0 indicates unset and 1 indicates set. These attributes are readonly.

It is observed that the device delete thread and sysfs read thread will not race because the delete thread calls sysfs kobject_put() which in turn waits for existing sysfs read to complete.

Note for device replace devid swap:

During the replace the target device temporarily assumes devid 0 before assigning the devid of the soruce device.

In btrfs_dev_replace_finishing() we remove source sysfs devid using the function btrfs_sysfs_remove_devices_attr(), so after that call kobject_rename() to update the devid in the sysfs. This adds and calls btrfs_sysfs_update_devid() helper function to update the device id.

---
v5: squash [PATCH] btrfs: update devid after replace
import changes as in misc-next
changelog
rename btrfs_sysfs_xx to btrfs_devinfo_xx
reorder devinfo attributes
relocate btrfs_sysfs_update_devid with in sysfs.c add device in the title
rename btrfs_sysfs_missing_show to btrfs_devinfo_missing_show

v4:
after patch
[PATCH v5 2/2] btrfs: reset device back to allocation state when removing in misc-next, the device::devid_kobj remains stale, fix it by using release.

v3:
Use optional groupid devid in BTRFS_ATTR(), it was blank in v2.

V2:
Make the devinfo attribute to carry one parameter, so now instead of dev_state attribute, we create in_fs_metadata, writeable, missing and replace_target attributes.

---
 fs/btrfs/dev-replace.c |   1 +
 fs/btrfs/sysfs.c       | 155 +++++++++++++++++++++++++++++++++++------
 fs/btrfs/sysfs.h       |   1 +
 fs/btrfs/volumes.h     |   4 ++
 4 files changed, 138 insertions(+), 23 deletions(-)
fs/btrfs/dev-replace.c +1 -0
index 13a4ebd04a7a..131d23de5f64 100644
@@ -744,6 +744,7 @@ static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
 
 	/* replace the sysfs entry */
 	btrfs_sysfs_remove_devices_attr(fs_info->fs_devices, src_device);
+	btrfs_sysfs_update_devid(tgt_device);
 	btrfs_rm_dev_replace_free_srcdev(src_device);
 
 	/* write back the superblocks */
fs/btrfs/sysfs.c +132 -23

189 lines not on this page · raw message

fs/btrfs/sysfs.h +1 -0
index 9d97b3c8db4e..ccf33eaf9e59 100644
@@ -34,5 +34,6 @@ void btrfs_sysfs_add_block_group_type(struct btrfs_block_group *cache);
 int btrfs_sysfs_add_space_info_type(struct btrfs_fs_info *fs_info,
 				    struct btrfs_space_info *space_info);
 void btrfs_sysfs_remove_space_info(struct btrfs_space_info *space_info);
+void btrfs_sysfs_update_devid(struct btrfs_device *device);
 
 #endif
fs/btrfs/volumes.h +4 -0
index 98535f1e208e..309cda477589 100644
@@ -136,6 +136,10 @@ struct btrfs_device {
 	atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
 
 	struct extent_io_tree alloc_state;
+
+	struct completion kobj_unregister;
+	/* For sysfs/FSID/devinfo/devid/ */
+	struct kobject devid_kobj;
 };
 
 /*
Signature · 1 line
2.23.0
#5 lore raw
In reply to Anand Jain
Quoted 5 lines · Anand Jain

Resend:
Patch 4/4 is already integrated without the cleanup and preparatory patches (1,2,3)/4. So I am resending those 3 patches. And the changes in patch 4/4 as in misc-next is imported into patch v5 4/4 here. Patch 4/4 has the details of the changes.

All the patches have been merged and now are in linus/master branch. I haven't reordered the patches so they're not in a group and the preparatory patches are about 10 commits above v5.5-rc7 which is the beginning of the branch.

author
#6 lore raw
Quoted 8 lines · David Sterba
Quoted 5 lines · Anand Jain

Resend:
Patch 4/4 is already integrated without the cleanup and preparatory patches (1,2,3)/4. So I am resending those 3 patches. And the changes in patch 4/4 as in misc-next is imported into patch v5 4/4 here. Patch 4/4 has the details of the changes.

All the patches have been merged and now are in linus/master branch.

David,

Quoted 3 lines

I haven't reordered the patches so they're not in a group and the preparatory patches are about 10 commits above v5.5-rc7 which is the beginning of the branch.

It will cause problem during bisect, without the preparatory patches the devid kobject lands under devices instead of under devinfo, that's how I noticed that preparatory patches are missing.

Thanks, Anand

#7 lore raw
Quoted 15 lines · Anand Jain
Quoted 8 lines · David Sterba
Quoted 5 lines · Anand Jain

Resend:
Patch 4/4 is already integrated without the cleanup and preparatory patches (1,2,3)/4. So I am resending those 3 patches. And the changes in patch 4/4 as in misc-next is imported into patch v5 4/4 here. Patch 4/4 has the details of the changes.

All the patches have been merged and now are in linus/master branch.

David,

Quoted 8 lines
Quoted 3 lines

I haven't reordered the patches so they're not in a group and the preparatory patches are about 10 commits above v5.5-rc7 which is the beginning of the branch.

It will cause problem during bisect, without the preparatory patches the devid kobject lands under devices instead of under devinfo, that's how I noticed that preparatory patches are missing.

I've tested current master and the device ids are indeed placed under device/ and there is no devinfo/, so this needs to be fixed.

Preparatory patches I was referring to are:

bc036bb33524 btrfs: sysfs, merge btrfs_sysfs_add devices_kobj and fsid be2cf92e0a2f btrfs: sysfs, rename btrfs_sysfs_add_device() c6761a9ed329 btrfs: sysfs, btrfs_sysfs_add_fsid() drop unused argument parent b5501504cb6a btrfs: sysfs, rename devices kobject holder to devices_kobj

so I'll have another look at what you sent.

author
#8 lore raw
Quoted 28 lines · David Sterba
Quoted 15 lines · Anand Jain
Quoted 8 lines · David Sterba
Quoted 5 lines · Anand Jain

Resend:
Patch 4/4 is already integrated without the cleanup and preparatory patches (1,2,3)/4. So I am resending those 3 patches. And the changes in patch 4/4 as in misc-next is imported into patch v5 4/4 here. Patch 4/4 has the details of the changes.

All the patches have been merged and now are in linus/master branch.

David,

Quoted 8 lines
Quoted 3 lines

I haven't reordered the patches so they're not in a group and the preparatory patches are about 10 commits above v5.5-rc7 which is the beginning of the branch.

It will cause problem during bisect, without the preparatory patches the devid kobject lands under devices instead of under devinfo, that's how I noticed that preparatory patches are missing.

I've tested current master and the device ids are indeed placed under device/ and there is no devinfo/, so this needs to be fixed.

Quoted 9 lines

Preparatory patches I was referring to are:

bc036bb33524 btrfs: sysfs, merge btrfs_sysfs_add devices_kobj and fsid be2cf92e0a2f btrfs: sysfs, rename btrfs_sysfs_add_device() c6761a9ed329 btrfs: sysfs, btrfs_sysfs_add_fsid() drop unused argument parent b5501504cb6a btrfs: sysfs, rename devices kobject holder to devices_kobj

so I'll have another look at what you sent.

David,

Busy? Can I ping on this?

Those aren't the preparatory patches. The patches sent here applies nicely on misc-next with git revert 668e48af7a94

Thanks, Anand

#9 lore raw
Quoted 10 lines · Anand Jain
Quoted 10 lines

I've tested current master and the device ids are indeed placed under device/ and there is no devinfo/, so this needs to be fixed.

Preparatory patches I was referring to are:

bc036bb33524 btrfs: sysfs, merge btrfs_sysfs_add devices_kobj and fsid be2cf92e0a2f btrfs: sysfs, rename btrfs_sysfs_add_device() c6761a9ed329 btrfs: sysfs, btrfs_sysfs_add_fsid() drop unused argument parent b5501504cb6a btrfs: sysfs, rename devices kobject holder to devices_kobj

Quoted 8 lines
Quoted 2 lines

so I'll have another look at what you sent.

David,

Busy? Can I ping on this?

Those aren't the preparatory patches. The patches sent here applies nicely on misc-next with git revert 668e48af7a94

Do you realize that patch 668e48af7a94 had been merged to linus/master at the time you sent the patchset a week ago? How can I do git revert? That was in the middle of merge window, 2 weeks after code freeze. If I don't merge a patch intentionally or by mistake, then you send me an incremental change.

Please send update with this patch order:

* "btrfs: sysfs, add UUID/devinfo kobject"
* fixup of the device id files created under devices
* the other cleanups