[PATCH v10 resend 0/3] readmirror feature (read_policy sysfs and in-memory only approach)

[PATCH v10 resend 0/3] readmirror feature (read_policy sysfs and in-memory only approach)

[Anand Jain]

Resend and based on misc-next (as dependency on genhd.c is gone).

v10: patch1/3 (fix btrfs_strmatch) and patch 3/3 (use scnprintf instead of snprintf)
     and add rb.

v9: C coding style fixes in 1/3 and 3/3

v8:
Separate the sysfs framework and the %pid read_policy into a separate
patchset here, so that the new read policies can be in its own patch set.

A latency based read_policy is being prepared to send it in a separate
patchset as it depends on a few changes in the block layer as well.

__ Original email: __

v7:
Fix switch's fall through warning. Changle logs updates where necessary.

v6:
Patch 4/5 - If there is no change in device's read prefer then don't log
Patch 4/5 - Add pid to the logs
Patch 5/5 - If there isn't read preferred device in the chunk don't reset
read policy to default, instead just use stripe 0. As this is in
the read path it avoids going through the device list to find
read preferred device. So inline to this drop to check if there
is read preferred device before setting read policy to device.

v5:
Worked on review comments as received in its previous version.
Please refer to individual patches for the specific changes.
Introduces the new read_policy 'device'.

v4:
Rename readmirror attribute to read_policy. Drop separate kobj for
readmirror instead create read_policy attribute in fsid kobj.
merge v2:2/3 and v2:3/3 into v4:2/2. Patch titles have changed.
 
v3:
v2:
Mainly fixes the fs_devices::readmirror declaration type from atomic_t
to u8. (Thanks Josef).

v1:
As of now we use only %pid method to read stripped mirrored data. So
application's process id determines the stripe id to be read. This type
of routing typically helps in a system with many small independent
applications tying to read random data. On the other hand the %pid
based read IO distribution policy is inefficient if there is a single
application trying to read large data as because the overall disk
bandwidth would remains under utilized.

One type of readmirror policy isn't good enough and other choices are
routing the IO based on device's waitqueue or manual when we have a
read-preferred device or a readmirror policy based on the target storage
caching. So this patch-set introduces a framework where we could add more
readmirror policies.

This policy is a filesystem wide policy as of now, and though the
readmirror policy at the subvolume level is a novel approach as it
provides maximum flexibility in the data center, but as of now its not
practical to implement such a granularity as you can't really ensure
reflinked extents will be read from the stripe of its desire and so
there will be more limitations and it can be assessed separately.

The approach in this patch-set is sys interface with in-memory policy.
And does not add any new readmirror type in this set, which can be add
once we are ok with the framework. Also the default policy remains %pid.

Previous works:
----------------------------------------------------------------------
There were few RFCs [1] before, mainly to figure out storage
(or in memory only) for the readmirror policy and the interface needed.

[1]
https://www.mail-archive.com/linux-btrfs@vger.kernel.org/msg86368.html

https://lore.kernel.org/linux-btrfs/20190826090438.7044-1-anand.jain@oracle.com/

https://lore.kernel.org/linux-btrfs/5fcf9c23-89b5-b167-1f80-a0f4ac107d0b@oracle.com/

https://patchwork.kernel.org/cover/10859213/

Mount -o:
In the first trial it was attempted to use the mount -o option to carry
the readmirror policy, this is good for debugging which can make sure
even the mount thread metadata tree blocks are read from the disk desired.
It was very effective in testing radi1/raid10 write-holes.

Extended attribute:
As extended attribute is associated with the inode, to implement this
there is bit of extended attribute abuse or else makes it mandatory to
mount the rootid 5. Its messy unless readmirror policy is applied at the
subvol level which is not possible as of now. 

An item type:
The proposed patch was to create an item to hold the readmirror policy,
it makes sense when compared to the abusive extended attribute approach
but introduces a new item and so no backward compatibility.
-----------------------------------------------------------------------

Anand Jain (3):
  btrfs: add btrfs_strmatch helper
  btrfs: create read policy framework
  btrfs: create read policy sysfs attribute, pid

 fs/btrfs/sysfs.c   | 67 ++++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/volumes.c | 15 ++++++++++-
 fs/btrfs/volumes.h | 14 ++++++++++
 3 files changed, 95 insertions(+), 1 deletion(-)

-- 
2.25.1

[PATCH v10 1/3] btrfs: add btrfs_strmatch helper

[Anand Jain]

Add a generic helper to match the golden-string in the given-string,
and ignore the leading and trailing whitespaces if any.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Suggested-by: David Sterba <dsterba@suse.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
---
v10: return bool instead of int.
     drop unnecessary local variable and ( ) with in if.
v9: use Josef suggested C coding style, using single if statement.
v5: born
 fs/btrfs/sysfs.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c
index fcd6c7a9bbd1..5955379d3d9e 100644
--- a/fs/btrfs/sysfs.c
+++ b/fs/btrfs/sysfs.c
@@ -888,6 +888,24 @@ static ssize_t btrfs_generation_show(struct kobject *kobj,
 }
 BTRFS_ATTR(, generation, btrfs_generation_show);
 
+/*
+ * Match the %golden in the %given. Ignore the leading and trailing whitespaces
+ * if any.
+ */
+static bool btrfs_strmatch(const char *given, const char *golden)
+{
+	size_t len = strlen(golden);
+
+	/* skip leading whitespace */
+	given = skip_spaces(given);
+
+	if (strncmp(given, golden, len) == 0 &&
+	    !strlen(skip_spaces(given + len)))
+		return true;
+
+	return false;
+}
+
 static const struct attribute *btrfs_attrs[] = {
 	BTRFS_ATTR_PTR(, label),
 	BTRFS_ATTR_PTR(, nodesize),
-- 
2.25.1

Re: [PATCH v10 1/3] btrfs: add btrfs_strmatch helper

[David Sterba]

On Wed, Oct 28, 2020 at 09:14:45PM +0800, Anand Jain wrote:
> Add a generic helper to match the golden-string in the given-string,
> and ignore the leading and trailing whitespaces if any.
> 
> Signed-off-by: Anand Jain <anand.jain@oracle.com>
> Suggested-by: David Sterba <dsterba@suse.com>
> Reviewed-by: Josef Bacik <josef@toxicpanda.com>
> ---
> v10: return bool instead of int.
>      drop unnecessary local variable and ( ) with in if.
> v9: use Josef suggested C coding style, using single if statement.
> v5: born
>  fs/btrfs/sysfs.c | 18 ++++++++++++++++++
>  1 file changed, 18 insertions(+)
> 
> diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c
> index fcd6c7a9bbd1..5955379d3d9e 100644
> --- a/fs/btrfs/sysfs.c
> +++ b/fs/btrfs/sysfs.c
> @@ -888,6 +888,24 @@ static ssize_t btrfs_generation_show(struct kobject *kobj,
>  }
>  BTRFS_ATTR(, generation, btrfs_generation_show);
>  
> +/*
> + * Match the %golden in the %given. Ignore the leading and trailing whitespaces
> + * if any.
> + */
> +static bool btrfs_strmatch(const char *given, const char *golden)
> +{
> +	size_t len = strlen(golden);
> +
> +	/* skip leading whitespace */

Comments should start with uppercase unless it's an identifier.

> +	given = skip_spaces(given);
> +
> +	if (strncmp(given, golden, len) == 0 &&
> +	    !strlen(skip_spaces(given + len)))

You had strlen() == 0 instead of !strlen in v9, the == 0 is better here.

I'll fix that locally and push to misc-next so we can proceed with the
actual policies.

[PATCH v10 2/3] btrfs: create read policy framework

[Anand Jain]

As of now, we use the %pid method to read striped mirrored data, which means
process id determines the stripe id to read. This type of routing
typically helps in a system with many small independent processes tying
to read random data. On the other hand, the %pid based read IO policy is
inefficient because if there is a single process trying to read a large
file, the overall disk bandwidth remains under-utilized.

So this patch introduces a read policy framework so that we could add more
read policies, such as IO routing based on the device's wait-queue or manual
when we have a read-preferred device or a policy based on the target
storage caching.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
---
v10: -
v9: -
v8: use fallthrough;
v7: Fix missing /* fall through */ in the switch
    Removed Reviewed-by: Josef Bacik <josef@toxicpanda.com>
v6:-
v5: Title renamed from:- btrfs: add read_policy framework
    Change log updated.
    Unnecessary comment dropped, added more where necessary.
    Optimize code in the switch remove duplicate code.
    Define BTRFS_READ_POLICY_DEFAULT dropped.
    Rename enum btrfs_read_policy_type to enum btrfs_read_policy.
    Rename BTRFS_READ_BY_PID to BTRFS_READ_POLICY_PID.
    (As its mainly renames. Reviewed-by retained).
v4: -
v3: Declare fs_devices::readmirror as enum btrfs_readmirror_policy_type
v2: Declare fs_devices::readmirror as u8 instead of atomic_t 
    A small change in comment and change log wordings.

 fs/btrfs/volumes.c | 15 ++++++++++++++-
 fs/btrfs/volumes.h | 14 ++++++++++++++
 2 files changed, 28 insertions(+), 1 deletion(-)

diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index b18e026424a6..6bf487626f23 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -1223,6 +1223,7 @@ static int open_fs_devices(struct btrfs_fs_devices *fs_devices,
 	fs_devices->latest_bdev = latest_dev->bdev;
 	fs_devices->total_rw_bytes = 0;
 	fs_devices->chunk_alloc_policy = BTRFS_CHUNK_ALLOC_REGULAR;
+	fs_devices->read_policy = BTRFS_READ_POLICY_PID;
 
 	return 0;
 }
@@ -5485,7 +5486,19 @@ static int find_live_mirror(struct btrfs_fs_info *fs_info,
 	else
 		num_stripes = map->num_stripes;
 
-	preferred_mirror = first + current->pid % num_stripes;
+	switch (fs_info->fs_devices->read_policy) {
+	default:
+		/*
+		 * Shouldn't happen, just warn and use pid instead of failing.
+		 */
+		btrfs_warn_rl(fs_info,
+			      "unknown read_policy type %u, fallback to pid",
+			      fs_info->fs_devices->read_policy);
+		fallthrough;
+	case BTRFS_READ_POLICY_PID:
+		preferred_mirror = first + current->pid % num_stripes;
+		break;
+	}
 
 	if (dev_replace_is_ongoing &&
 	    fs_info->dev_replace.cont_reading_from_srcdev_mode ==
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index 232f02bd214f..97f075516696 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -211,6 +211,15 @@ enum btrfs_chunk_allocation_policy {
 	BTRFS_CHUNK_ALLOC_REGULAR,
 };
 
+/*
+ * Read policies for the mirrored block groups, read picks the stripe based
+ * on these policies.
+ */
+enum btrfs_read_policy {
+	BTRFS_READ_POLICY_PID,
+	BTRFS_NR_READ_POLICY,
+};
+
 struct btrfs_fs_devices {
 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
 	u8 metadata_uuid[BTRFS_FSID_SIZE];
@@ -264,6 +273,11 @@ struct btrfs_fs_devices {
 	struct completion kobj_unregister;
 
 	enum btrfs_chunk_allocation_policy chunk_alloc_policy;
+
+	/*
+	 * policy used to read the mirrored stripes
+	 */
+	enum btrfs_read_policy read_policy;
 };
 
 #define BTRFS_BIO_INLINE_CSUM_SIZE	64
-- 
2.25.1

Re: [PATCH v10 2/3] btrfs: create read policy framework

[David Sterba]

On Wed, Oct 28, 2020 at 09:14:46PM +0800, Anand Jain wrote:
> @@ -5485,7 +5486,19 @@ static int find_live_mirror(struct btrfs_fs_info *fs_info,
>  	else
>  		num_stripes = map->num_stripes;
>  
> -	preferred_mirror = first + current->pid % num_stripes;
> +	switch (fs_info->fs_devices->read_policy) {
> +	default:
> +		/*
> +		 * Shouldn't happen, just warn and use pid instead of failing.
> +		 */
> +		btrfs_warn_rl(fs_info,
> +			      "unknown read_policy type %u, fallback to pid",
> +			      fs_info->fs_devices->read_policy);

So this would flood the system log with useless messages once this
happens. We should really reset it to pid or whatever default we choose.
Fixed.

> +		fallthrough;
> +	case BTRFS_READ_POLICY_PID:
> +		preferred_mirror = first + current->pid % num_stripes;

This is in the original code, but this should really use ( ) to make the
precedence clear.

[PATCH v10 3/3] btrfs: create read policy sysfs attribute, pid

[Anand Jain]

Add

 /sys/fs/btrfs/UUID/read_policy

attribute so that the read policy for the raid1, raid1c34 and raid10 can
be tuned.

When this attribute is read, it shall show all available policies, with
active policy being with in [ ]. The read_policy attribute can be written
using one of the items listed in there.

For example:
  $cat /sys/fs/btrfs/UUID/read_policy
  [pid]
  $echo pid > /sys/fs/btrfs/UUID/read_policy

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
---
v10: use scnprintf instead of snprintf
v9: fix C coding style, static const char*
v5: 
  Title rename: old: btrfs: sysfs, add read_policy attribute
  Uses the btrfs_strmatch() helper (BTRFS_READ_POLICY_NAME_MAX dropped).
  Use the table for the policy names.
  Rename len to ret.
  Use a simple logic to prefix space in btrfs_read_policy_show()
  Reviewed-by: Josef Bacik <josef@toxicpanda.com> dropped.

v4:-
v3: rename [by_pid] to [pid]
v2: v2: check input len before strip and kstrdup

 fs/btrfs/sysfs.c | 49 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)

diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c
index 5955379d3d9e..4dbf90ff088a 100644
--- a/fs/btrfs/sysfs.c
+++ b/fs/btrfs/sysfs.c
@@ -906,6 +906,54 @@ static bool btrfs_strmatch(const char *given, const char *golden)
 	return false;
 }
 
+static const char * const btrfs_read_policy_name[] = { "pid" };
+
+static ssize_t btrfs_read_policy_show(struct kobject *kobj,
+				      struct kobj_attribute *a, char *buf)
+{
+	int i;
+	ssize_t ret = 0;
+	struct btrfs_fs_devices *fs_devices = to_fs_devs(kobj);
+
+	for (i = 0; i < BTRFS_NR_READ_POLICY; i++) {
+		if (fs_devices->read_policy == i)
+			ret += scnprintf(buf + ret, PAGE_SIZE - ret, "%s[%s]",
+					 (ret == 0 ? "" : " "),
+					 btrfs_read_policy_name[i]);
+		else
+			ret += scnprintf(buf + ret, PAGE_SIZE - ret, "%s%s",
+					 (ret == 0 ? "" : " "),
+					 btrfs_read_policy_name[i]);
+	}
+
+	ret += scnprintf(buf + ret, PAGE_SIZE - ret, "\n");
+
+	return ret;
+}
+
+static ssize_t btrfs_read_policy_store(struct kobject *kobj,
+				       struct kobj_attribute *a,
+				       const char *buf, size_t len)
+{
+	int i;
+	struct btrfs_fs_devices *fs_devices = to_fs_devs(kobj);
+
+	for (i = 0; i < BTRFS_NR_READ_POLICY; i++) {
+		if (btrfs_strmatch(buf, btrfs_read_policy_name[i])) {
+			if (i != fs_devices->read_policy) {
+				fs_devices->read_policy = i;
+				btrfs_info(fs_devices->fs_info,
+					   "read policy set to '%s'",
+					   btrfs_read_policy_name[i]);
+			}
+			return len;
+		}
+	}
+
+	return -EINVAL;
+}
+BTRFS_ATTR_RW(, read_policy, btrfs_read_policy_show, btrfs_read_policy_store);
+
 static const struct attribute *btrfs_attrs[] = {
 	BTRFS_ATTR_PTR(, label),
 	BTRFS_ATTR_PTR(, nodesize),
@@ -916,6 +964,7 @@ static const struct attribute *btrfs_attrs[] = {
 	BTRFS_ATTR_PTR(, checksum),
 	BTRFS_ATTR_PTR(, exclusive_operation),
 	BTRFS_ATTR_PTR(, generation),
+	BTRFS_ATTR_PTR(, read_policy),
 	NULL,
 };
 
-- 
2.25.1

Re: [PATCH v10 3/3] btrfs: create read policy sysfs attribute, pid

[David Sterba]

On Wed, Oct 28, 2020 at 09:14:47PM +0800, Anand Jain wrote:
> +static ssize_t btrfs_read_policy_store(struct kobject *kobj,
> +				       struct kobj_attribute *a,
> +				       const char *buf, size_t len)
> +{
> +	int i;
> +	struct btrfs_fs_devices *fs_devices = to_fs_devs(kobj);
> +
> +	for (i = 0; i < BTRFS_NR_READ_POLICY; i++) {
> +		if (btrfs_strmatch(buf, btrfs_read_policy_name[i])) {

Does sysfs guarantee that the buf is nul terminated string or that it
contains a null at all? Because if not, the skip_space step of strmatch
could run out of the buffer.

Re: [PATCH v10 3/3] btrfs: create read policy sysfs attribute, pid

[Anand Jain]

On 30/10/20 12:45 am, David Sterba wrote:
> On Wed, Oct 28, 2020 at 09:14:47PM +0800, Anand Jain wrote:
>> +static ssize_t btrfs_read_policy_store(struct kobject *kobj,
>> +				       struct kobj_attribute *a,
>> +				       const char *buf, size_t len)
>> +{
>> +	int i;
>> +	struct btrfs_fs_devices *fs_devices = to_fs_devs(kobj);
>> +
>> +	for (i = 0; i < BTRFS_NR_READ_POLICY; i++) {
>> +		if (btrfs_strmatch(buf, btrfs_read_policy_name[i])) {
> 
> Does sysfs guarantee that the buf is nul terminated string or that it
> contains a null at all? Because if not, the skip_space step of strmatch
> could run out of the buffer.
> 

It does

[  173.555507]  ? btrfs_read_policy_store+0x3e/0x12d [btrfs]
[  173.555541]  ? kobj_attr_store+0x16/0x30
[  173.555562]  ? sysfs_kf_write+0x54/0x80
[  173.555582]  ? kernfs_fop_write+0xfa/0x290
[  173.555611]  ? vfs_write+0xee/0x2f0
[  173.555641]  ? ksys_write+0x80/0x170
[  173.555671]  ? __x64_sys_write+0x1e/0x30
[  173.555692]  ? do_syscall_64+0x4b/0x80
[  173.555708]  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9


static ssize_t kernfs_fop_write(struct file *file, const char __user 
*user_buf,
                                 size_t count, loff_t *ppos)
{
::
         buf[len] = '\0';        /* guarantee string termination */


Thanks, Anand

Re: [PATCH v10 3/3] btrfs: create read policy sysfs attribute, pid

[David Sterba]

On Fri, Oct 30, 2020 at 03:30:10AM +0800, Anand Jain wrote:
> 
> 
> On 30/10/20 12:45 am, David Sterba wrote:
> > On Wed, Oct 28, 2020 at 09:14:47PM +0800, Anand Jain wrote:
> >> +static ssize_t btrfs_read_policy_store(struct kobject *kobj,
> >> +				       struct kobj_attribute *a,
> >> +				       const char *buf, size_t len)
> >> +{
> >> +	int i;
> >> +	struct btrfs_fs_devices *fs_devices = to_fs_devs(kobj);
> >> +
> >> +	for (i = 0; i < BTRFS_NR_READ_POLICY; i++) {
> >> +		if (btrfs_strmatch(buf, btrfs_read_policy_name[i])) {
> > 
> > Does sysfs guarantee that the buf is nul terminated string or that it
> > contains a null at all? Because if not, the skip_space step of strmatch
> > could run out of the buffer.
> > 
> 
> It does
> 
> [  173.555507]  ? btrfs_read_policy_store+0x3e/0x12d [btrfs]
> [  173.555541]  ? kobj_attr_store+0x16/0x30
> [  173.555562]  ? sysfs_kf_write+0x54/0x80
> [  173.555582]  ? kernfs_fop_write+0xfa/0x290
> [  173.555611]  ? vfs_write+0xee/0x2f0
> [  173.555641]  ? ksys_write+0x80/0x170
> [  173.555671]  ? __x64_sys_write+0x1e/0x30
> [  173.555692]  ? do_syscall_64+0x4b/0x80
> [  173.555708]  ? entry_SYSCALL_64_after_hwframe+0x44/0xa9
> 
> 
> static ssize_t kernfs_fop_write(struct file *file, const char __user 
> *user_buf,
>                                  size_t count, loff_t *ppos)
> {
> ::
>          buf[len] = '\0';        /* guarantee string termination */

Ok thanks.