Skip to contentSkip to conversation

btrfs: sysfs: export dev stats in devinfo directory

PATCH In review David Sterba 4 participants 4 messages Jun 11, 2021 – Jun 14, 2021 Btrfs file system

lore
Share & tools
lore

David Sterba

author maintainer
#0 lore raw

PATCH v2 1/1 btrfs: sysfs: export dev stats in devinfo directory

The device stats can be read by ioctl, wrapped by command 'btrfs device stats'. Provide another source where to read the information in /sys/fs/btrfs/FSID/devinfo/DEVID/stats . The format is a list of 'key value' pairs one per line, which is common in other stat files. The names are the same as used in other device stat outputs.

The stats are all in one file as it's the snapshot of all available stats. The 'one value per file' is not very suitable here. The stats should be valid right after the stats item is read from disk, shortly after initializing the device.

In case the stats are not yet valid, print just 'invalid' as the file contents.

---

v2:
- print 'invalid' separtely and not among the values
- rename file name to 'error_stats' to leave 'stats' available for any other kind of stats we'd like in the future

 fs/btrfs/sysfs.c | 29 +++++++++++++++++++++++++++++
 1 file changed, 29 insertions(+)
fs/btrfs/sysfs.c +29 -0
index 4b508938e728..ebde1d09e686 100644
@@ -1495,7 +1495,36 @@ static ssize_t btrfs_devinfo_writeable_show(struct kobject *kobj,
 }
 BTRFS_ATTR(devid, writeable, btrfs_devinfo_writeable_show);
 
+static ssize_t btrfs_devinfo_error_stats_show(struct kobject *kobj,
+		struct kobj_attribute *a, char *buf)
+{
+	struct btrfs_device *device = container_of(kobj, struct btrfs_device,
+						   devid_kobj);
+
+	if (!device->dev_stats_valid)
+		return scnprintf(buf, PAGE_SIZE, "invalid\n");
+
+	/*
+	 * Print all at once so we get a snapshot of all values from the same
+	 * time. Keep them in sync and in order of definition of
+	 * btrfs_dev_stat_values.
+	 */
+	return scnprintf(buf, PAGE_SIZE,
+		"write_errs %d\n"
+		"read_errs %d\n"
+		"flush_errs %d\n"
+		"corruption_errs %d\n"
+		"generation_errs %d\n",
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_WRITE_ERRS),
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_READ_ERRS),
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_FLUSH_ERRS),
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_CORRUPTION_ERRS),
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_GENERATION_ERRS));
+}
+BTRFS_ATTR(devid, error_stats, btrfs_devinfo_error_stats_show);
+
 static struct attribute *devid_attrs[] = {
+	BTRFS_ATTR_PTR(devid, error_stats),
 	BTRFS_ATTR_PTR(devid, in_fs_metadata),
 	BTRFS_ATTR_PTR(devid, missing),
 	BTRFS_ATTR_PTR(devid, replace_target),
Signature · 1 line
2.31.1
#1 lore raw
Quoted 3 lines · David Sterba

The device stats can be read by ioctl, wrapped by command 'btrfs device stats'. Provide another source where to read the information in /sys/fs/btrfs/FSID/devinfo/DEVID/stats . The format is a list of

                                   ^error_stats :)

Otherwise a very welcome improvement, thank you!

cheers
Holger

Reviewed-by on v2 1/1
#2 lore raw
In reply to David Sterba
Quoted 26 lines

The stats are all in one file as it's the snapshot of all available stats. The 'one value per file' is not very suitable here. The stats should be valid right after the stats item is read from disk, shortly after initializing the device.

In case the stats are not yet valid, print just 'invalid' as the file contents.

---

v2:
- print 'invalid' separtely and not among the values
- rename file name to 'error_stats' to leave 'stats' available for any other kind of stats we'd like in the future

  fs/btrfs/sysfs.c | 29 +++++++++++++++++++++++++++++
  1 file changed, 29 insertions(+)
fs/btrfs/sysfs.c +0 -0
index 4b508938e728..ebde1d09e686 100644
@@ -1495,7 +1495,36 @@ static ssize_t btrfs_devinfo_writeable_show(struct kobject *kobj,
  }

fs/btrfs/sysfs.c static ssize_t btrfs_devinfo_writeable_show(struct kobject *kobj, · line 1505

Quoted 10 lines
fs/btrfs/sysfs.c +8 -0
  BTRFS_ATTR(devid, writeable, btrfs_devinfo_writeable_show);
 
+static ssize_t btrfs_devinfo_error_stats_show(struct kobject *kobj,
+		struct kobj_attribute *a, char *buf)
+{
+	struct btrfs_device *device = container_of(kobj, struct btrfs_device,
+						   devid_kobj);
+
+	if (!device->dev_stats_valid)
+		return scnprintf(buf, PAGE_SIZE, "invalid\n");

and

Nit:
Instead of invalid, IMO, not yet valid is closer to what this error implies. And matches with the ioctl Warning message.

7743 btrfs_warn(fs_info, "get dev_stats failed, not yet valid");

Quoted 16 lines
fs/btrfs/sysfs.c +15 -0
+
+	/*
+	 * Print all at once so we get a snapshot of all values from the same
+	 * time. Keep them in sync and in order of definition of
+	 * btrfs_dev_stat_values.
+	 */
+	return scnprintf(buf, PAGE_SIZE,
+		"write_errs %d\n"
+		"read_errs %d\n"
+		"flush_errs %d\n"
+		"corruption_errs %d\n"
+		"generation_errs %d\n",
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_WRITE_ERRS),
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_READ_ERRS),
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_FLUSH_ERRS),

fs/btrfs/sysfs.c static ssize_t btrfs_devinfo_writeable_show(struct kobject *kobj, · line 1530

Quoted 10 lines
fs/btrfs/sysfs.c +6 -0
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_CORRUPTION_ERRS),
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_GENERATION_ERRS));
+}
+BTRFS_ATTR(devid, error_stats, btrfs_devinfo_error_stats_show);
+
  static struct attribute *devid_attrs[] = {
+	BTRFS_ATTR_PTR(devid, error_stats),
  	BTRFS_ATTR_PTR(devid, in_fs_metadata),
  	BTRFS_ATTR_PTR(devid, missing),
  	BTRFS_ATTR_PTR(devid, replace_target),

write_errs 0
read_errs 0
flush_errs 0
corruption_errs 1
generation_errs 0

Another option was to print all errors in a single line. A single line will help continuous monitoring of the error_stats. But it is not a big deal.

Thanks, Anand

#3 lore raw
Quoted 39 lines · Anand Jain
Quoted 26 lines

The stats are all in one file as it's the snapshot of all available stats. The 'one value per file' is not very suitable here. The stats should be valid right after the stats item is read from disk, shortly after initializing the device.

In case the stats are not yet valid, print just 'invalid' as the file contents.

---

v2:
- print 'invalid' separtely and not among the values
- rename file name to 'error_stats' to leave 'stats' available for any other kind of stats we'd like in the future

  fs/btrfs/sysfs.c | 29 +++++++++++++++++++++++++++++
  1 file changed, 29 insertions(+)
fs/btrfs/sysfs.c +0 -0
index 4b508938e728..ebde1d09e686 100644
@@ -1495,7 +1495,36 @@ static ssize_t btrfs_devinfo_writeable_show(struct kobject *kobj,
  }

fs/btrfs/sysfs.c static ssize_t btrfs_devinfo_writeable_show(struct kobject *kobj, · line 1505

Quoted 10 lines
fs/btrfs/sysfs.c +8 -0
  BTRFS_ATTR(devid, writeable, btrfs_devinfo_writeable_show);
 
+static ssize_t btrfs_devinfo_error_stats_show(struct kobject *kobj,
+		struct kobj_attribute *a, char *buf)
+{
+	struct btrfs_device *device = container_of(kobj, struct btrfs_device,
+						   devid_kobj);
+
+	if (!device->dev_stats_valid)
+		return scnprintf(buf, PAGE_SIZE, "invalid\n");

and

Quoted 5 lines

Nit:
Instead of invalid, IMO, not yet valid is closer to what this error implies. And matches with the ioctl Warning message.

7743 btrfs_warn(fs_info, "get dev_stats failed, not yet valid");

Yeah I was not sure about that, 'invalid' does not have the same meaning. I'd like something that's short and easily parseable so no long sentences or multi-word value like 'not yet valid'.

It's trivial to change if somebody has a better idea or convince me with some variation on not-yet-valid.

Quoted 35 lines
Quoted 16 lines
fs/btrfs/sysfs.c +15 -0
+
+	/*
+	 * Print all at once so we get a snapshot of all values from the same
+	 * time. Keep them in sync and in order of definition of
+	 * btrfs_dev_stat_values.
+	 */
+	return scnprintf(buf, PAGE_SIZE,
+		"write_errs %d\n"
+		"read_errs %d\n"
+		"flush_errs %d\n"
+		"corruption_errs %d\n"
+		"generation_errs %d\n",
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_WRITE_ERRS),
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_READ_ERRS),
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_FLUSH_ERRS),

fs/btrfs/sysfs.c static ssize_t btrfs_devinfo_writeable_show(struct kobject *kobj, · line 1530

Quoted 10 lines
fs/btrfs/sysfs.c +6 -0
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_CORRUPTION_ERRS),
+		btrfs_dev_stat_read(device, BTRFS_DEV_STAT_GENERATION_ERRS));
+}
+BTRFS_ATTR(devid, error_stats, btrfs_devinfo_error_stats_show);
+
  static struct attribute *devid_attrs[] = {
+	BTRFS_ATTR_PTR(devid, error_stats),
  	BTRFS_ATTR_PTR(devid, in_fs_metadata),
  	BTRFS_ATTR_PTR(devid, missing),
  	BTRFS_ATTR_PTR(devid, replace_target),

write_errs 0
read_errs 0
flush_errs 0
corruption_errs 1
generation_errs 0

Quoted 3 lines

Another option was to print all errors in a single line. A single line will help continuous monitoring of the error_stats. But it is not a big deal.

I think that's just a matter of transforming the data to the convenient form. Line based stats values is more common and grep-friendly. For a monitoring tool/scripts that want it in one line it's perhaps a simple

cat devinfo/error_sats | tr '\n' ' '

Converting from one line to multiple lines is not that trivial.