* [PATCH 0/2][V11] btrfs-progs: allocation_hint disk property
@ 2022-01-26 20:32 Goffredo Baroncelli
2022-01-26 20:32 ` [PATCH 1/2] btrfs-progs: new "allocation_hint" property Goffredo Baroncelli
2022-01-26 20:32 ` [PATCH 2/2] Update man page for allocator_hint property Goffredo Baroncelli
0 siblings, 2 replies; 3+ messages in thread
From: Goffredo Baroncelli @ 2022-01-26 20:32 UTC (permalink / raw)
To: linux-btrfs
Cc: Zygo Blaxell, Josef Bacik, David Sterba, Sinnamohideen Shafeeq,
Paul Jones, Boris Burkov, Goffredo Baroncelli
From: Goffredo Baroncelli <kreijack@inwind.it>
This patches set is the userspace portion of the serie
"[PATCH V11] btrfs: allocation_hint".
Look this patches set for further information.
Changelog
V11:
- find the correct /sys/fs/btrfs/<UUID> directory using the
devinfo/major_minor sysfs file.
- update the man page to the .rst file format
G.Baroncelli
Goffredo Baroncelli (2):
btrfs-progs: new "allocation_hint" property.
Update man page for allocator_hint property.
Documentation/btrfs-property.rst | 21 +++
cmds/property.c | 253 +++++++++++++++++++++++++++++++
kernel-shared/ctree.h | 13 ++
3 files changed, 287 insertions(+)
--
2.34.1
^ permalink raw reply [flat|nested] 3+ messages in thread
* [PATCH 1/2] btrfs-progs: new "allocation_hint" property.
2022-01-26 20:32 [PATCH 0/2][V11] btrfs-progs: allocation_hint disk property Goffredo Baroncelli
@ 2022-01-26 20:32 ` Goffredo Baroncelli
2022-01-26 20:32 ` [PATCH 2/2] Update man page for allocator_hint property Goffredo Baroncelli
1 sibling, 0 replies; 3+ messages in thread
From: Goffredo Baroncelli @ 2022-01-26 20:32 UTC (permalink / raw)
To: linux-btrfs
Cc: Zygo Blaxell, Josef Bacik, David Sterba, Sinnamohideen Shafeeq,
Paul Jones, Boris Burkov, Goffredo Baroncelli
From: Goffredo Baroncelli <kreijack@inwind.it>
Handle the property allocation_hint of a btrfs device. Below
an example of use:
$ # set a new value
$ sudo btrfs property set /dev/vde allocation_hint DATA_ONLY
$ # get the current value
$ sudo btrfs property get /dev/vde allocation_hint
devid=4, path=/dev/vde: allocation_hint=DATA_ONLY
The following values are availables:
- DATA_ONLY
- DATA_PREFERRED (default)
- METADATA_PREFERRED
- METADATA_ONLY
Root privileges are required.
Signed-off-by: Goffredo Baroncelli <kreijack@inwind.it>
---
cmds/property.c | 253 ++++++++++++++++++++++++++++++++++++++++++
kernel-shared/ctree.h | 13 +++
2 files changed, 266 insertions(+)
diff --git a/cmds/property.c b/cmds/property.c
index b3ccc0ff..a409f4e9 100644
--- a/cmds/property.c
+++ b/cmds/property.c
@@ -22,6 +22,7 @@
#include <sys/ioctl.h>
#include <sys/stat.h>
#include <sys/xattr.h>
+#include <sys/sysmacros.h>
#include <uuid/uuid.h>
#include <btrfsutil.h>
#include "cmds/commands.h"
@@ -232,6 +233,252 @@ out:
return ret;
}
+/*
+ * @major, @minor -> device to find
+ * @uuid -> uuid of the filesystem (out)
+ * @devid -> devid (out)
+ *
+ * return -> -ENOTSUP operation not supported (i.e. btrfs doesn't
+ * support allocation_hint)
+ * -> -ENODEV operation supported, but device not mounted
+ * -> -EACCES error accessing sysfs
+ * -> 0 ok
+ */
+#define BTRFSYSFS "/sys/fs/btrfs/"
+static int btrfs_find_devid_uuid_by_major_minor(int major,
+ int minor,
+ u64 *devid,
+ char *uuid) {
+
+ DIR *dp = NULL, *dp2 = NULL;
+ char path[200]; /* should be enough for
+ * /sys/btrfs/<uuid>/devices/<devid>/major_minor
+ */
+ int ret;
+ struct dirent *de;
+ int l;
+
+ strcpy(path, BTRFSYSFS);
+ dp = opendir(path);
+ if (!dp) {
+ ret = -EACCES;
+ goto quit;
+ }
+
+ while ((de = readdir(dp)) != NULL) {
+ if (strlen(de->d_name) != 36)
+ continue;
+
+ strcpy(path + sizeof(BTRFSYSFS) - 1, de->d_name);
+ strcat(path, "/devinfo/");
+
+ dp2 = opendir(path);
+ if (!dp2) {
+ ret = -EACCES;
+ goto quit;
+ }
+
+ l = strlen(path);
+ while ((de = readdir(dp2)) != NULL) {
+ FILE *fp;
+ char buf[100];
+ char *endptr;
+ int r, maj, mnr;
+
+ errno = 0;
+ *devid = strtoull(de->d_name, &endptr, 0);
+
+ /* check for invalid devid */
+ if (errno || *endptr != 0)
+ continue;
+
+
+ strcpy(path + l, de->d_name);
+ strcat(path, "/major_minor");
+
+ fp = fopen(path, "r");
+ if (!fp) {
+ if (errno == -ENOENT)
+ ret = -ENOTSUP;
+ else
+ ret = -EACCES;
+ goto quit;
+ }
+ r = fread(buf, 1, sizeof(buf) - 2, fp);
+ buf[r] = 0;
+ fclose(fp);
+
+ if (!strcmp(buf, "N/A\n"))
+ continue;
+
+ r = sscanf(buf, "%d:%d", &maj, &mnr);
+ if (r != 2) {
+ ret = -EACCES;
+ goto quit;
+ }
+
+ if (maj == major && minor == mnr) {
+ strncpy(uuid,
+ path + sizeof(BTRFSYSFS) - 1,
+ 36);
+ uuid[36] = 0;
+ ret = 0;
+ goto quit;
+ }
+ }
+ }
+
+ ret = -ENODEV;
+quit:
+ if (dp)
+ closedir(dp);
+ if (dp2)
+ closedir(dp2);
+ return ret;
+}
+
+/*
+ * @dev -> device to find
+ * @uuid -> uuid of the filesystem (out)
+ * @devid -> devid (out)
+ *
+ * return -> -ENOTSUP operation not supported (i.e. btrfs doesn't
+ * support allocation_hint)
+ * -> -ENODEV operation supported, but device not mounted
+ * -> -EACCES error accessing sysfs
+ * -> 0 ok
+ */
+static int btrfs_find_devid_uuid_by_dev(const char *dev,
+ u64 *devid,
+ char *uuid)
+{
+ struct stat st;
+ int r;
+
+ r = stat(dev, &st);
+ if (r < 0 && errno == -ENODEV)
+ return -ENODEV;
+ if (r < 0)
+ return -EACCES;
+
+ return btrfs_find_devid_uuid_by_major_minor(major(st.st_rdev),
+ minor(st.st_rdev),
+ devid, uuid);
+}
+
+static struct ull_charp_pair_t {
+ u64 value;
+ const char *descr;
+} allocation_hint_description[] = {
+ {BTRFS_DEV_ALLOCATION_HINT_METADATA_PREFERRED, "METADATA_PREFERRED"},
+ {BTRFS_DEV_ALLOCATION_HINT_METADATA_ONLY, "METADATA_ONLY"},
+ {BTRFS_DEV_ALLOCATION_HINT_DATA_PREFERRED, "DATA_PREFERRED"},
+ {BTRFS_DEV_ALLOCATION_HINT_DATA_ONLY, "DATA_ONLY"},
+ {0, NULL}
+};
+
+static int prop_allocation_hint(enum prop_object_type type,
+ const char *object,
+ const char *name,
+ const char *val,
+ bool force)
+{
+ int ret, fd2 = -1;
+ u64 devid;
+ char sysfs_file[PATH_MAX];
+ int i;
+ u64 v;
+ char buf[1024];
+
+ strcpy(sysfs_file, BTRFSYSFS);
+ ret = btrfs_find_devid_uuid_by_dev(object, &devid,
+ sysfs_file + sizeof(BTRFSYSFS) - 1);
+
+ if (ret)
+ goto out;
+
+ sprintf(sysfs_file + strlen(sysfs_file),
+ "/devinfo/%llu/allocation_hint", devid);
+
+ if (!val) {
+ /* READ */
+ fd2 = open(sysfs_file, O_RDONLY);
+ if (fd2 < 0 && errno == ENOENT) {
+ /* older kernel doesn't have allocation_hint; return nothing */
+ ret = 0;
+ goto out;
+ } else if (fd2 < 0) {
+ error("'allocation_hint' property not available or accessible.");
+ ret = -errno;
+ goto out;
+ }
+
+ ret = read(fd2, buf, sizeof(buf) - 1);
+ if (ret < 0) {
+ error("Unable to read the 'allocation_hint' property.");
+ ret = -errno;
+ goto out;
+ }
+
+ buf[sizeof(buf) - 1] = 0;
+ v = strtoull(buf, NULL, 0);
+
+ for (i = 0 ; allocation_hint_description[i].descr ; i++)
+ if (v == allocation_hint_description[i].value)
+ break;
+
+ if (allocation_hint_description[i].descr)
+ printf("allocation_hint=%s\n",
+ allocation_hint_description[i].descr);
+ else
+ printf("allocation_hint=unknown:%llu\n", v);
+ ret = 0;
+ } else {
+ /* WRITE */
+ for (i = 0 ; allocation_hint_description[i].descr ; i++)
+ if (!strcmp(val, allocation_hint_description[i].descr))
+ break;
+
+ if (allocation_hint_description[i].descr) {
+ v = allocation_hint_description[i].value;
+ } else if (sscanf(val, "%llu", &v) != 1) {
+ error("Invalid value '%s'\n", val);
+ ret = -3;
+ goto out;
+ }
+ if (v & ~BTRFS_DEV_ALLOCATION_HINT_MASK) {
+ error("Invalid value '%s'\n", val);
+ ret = -3;
+ goto out;
+ }
+
+ fd2 = open(sysfs_file, O_RDWR);
+ if (fd2 < 0) {
+ error("'allocation_hint' property not available or accessible for updating.");
+ ret = -errno;
+ goto out;
+ }
+
+ sprintf(buf, "%llu", v);
+
+ ret = write(fd2, buf, strlen(buf));
+
+ if (ret != strlen(buf)) {
+ error("Unable to update 'allocation_hint' property.");
+ ret = -errno;
+ goto out;
+ }
+
+ }
+
+ ret = 0;
+out:
+ if (fd2 >= 0)
+ close(fd2);
+
+ return ret;
+}
+
const struct prop_handler prop_handlers[] = {
{
.name ="ro",
@@ -254,6 +501,12 @@ const struct prop_handler prop_handlers[] = {
.types = prop_object_inode,
.handler = prop_compression
},
+ {
+ .name = "allocation_hint",
+ .desc = "hint to store the data/metadata chunks",
+ .types = prop_object_dev,
+ .handler = prop_allocation_hint
+ },
{NULL, NULL, 0, 0, NULL}
};
diff --git a/kernel-shared/ctree.h b/kernel-shared/ctree.h
index ab2aaed6..628539c0 100644
--- a/kernel-shared/ctree.h
+++ b/kernel-shared/ctree.h
@@ -216,6 +216,19 @@ struct btrfs_mapping_tree {
struct cache_tree cache_tree;
};
+/* btrfs chunk allocation hints */
+#define BTRFS_DEV_ALLOCATION_HINT_BIT_COUNT 2
+#define BTRFS_DEV_ALLOCATION_HINT_MASK ((1ULL << \
+ BTRFS_DEV_ALLOCATION_HINT_BIT_COUNT) - 1)
+/* preferred metadata chunk, but data chunk allowed */
+#define BTRFS_DEV_ALLOCATION_HINT_METADATA_PREFERRED (1ULL)
+/* only metadata chunk are allowed */
+#define BTRFS_DEV_ALLOCATION_HINT_METADATA_ONLY (2ULL)
+/* only data chunk allowed */
+#define BTRFS_DEV_ALLOCATION_HINT_DATA_ONLY (3ULL)
+/* preferred data chunk, but metadata chunk allowed */
+#define BTRFS_DEV_ALLOCATION_HINT_DATA_PREFERRED (0ULL)
+
#define BTRFS_UUID_SIZE 16
struct btrfs_dev_item {
/* the internal btrfs device id */
--
2.34.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* [PATCH 2/2] Update man page for allocator_hint property.
2022-01-26 20:32 [PATCH 0/2][V11] btrfs-progs: allocation_hint disk property Goffredo Baroncelli
2022-01-26 20:32 ` [PATCH 1/2] btrfs-progs: new "allocation_hint" property Goffredo Baroncelli
@ 2022-01-26 20:32 ` Goffredo Baroncelli
1 sibling, 0 replies; 3+ messages in thread
From: Goffredo Baroncelli @ 2022-01-26 20:32 UTC (permalink / raw)
To: linux-btrfs
Cc: Zygo Blaxell, Josef Bacik, David Sterba, Sinnamohideen Shafeeq,
Paul Jones, Boris Burkov, Goffredo Baroncelli
From: Goffredo Baroncelli <kreijack@inwind.it>
Update the man page of the btrfs property subcommand to show the use
of the device property "allocation_hint".
Signed-off-by: Goffredo Baroncelli <kreijack@inwind.it>
---
Documentation/btrfs-property.rst | 21 +++++++++++++++++++++
1 file changed, 21 insertions(+)
diff --git a/Documentation/btrfs-property.rst b/Documentation/btrfs-property.rst
index 5896faa2..155839fe 100644
--- a/Documentation/btrfs-property.rst
+++ b/Documentation/btrfs-property.rst
@@ -48,6 +48,27 @@ get [-t <type>] <object> [<name>]
compression
compression algorithm set for an inode, possible values: *lzo*, *zlib*, *zstd*.
To disable compression use "" (empty string), *no* or *none*.
+ allocation_hint
+ a device property that instructs how and when the allocator should use a
+ block device.
+ Possible values are:
+
+ METADATA_PREFERRED
+ the device has an higher priority when a new metadata
+ chunk is allocated. Data chunk is allowed only if there is no other
+ possibility.
+ METADATA_ONLY
+ the device is used only for metadata chunk.
+ Data chunk is never allowed.
+ DATA_PREFERRED (default)
+ the device has an higher priority when a new data
+ chunk is allocated. Metadata chunk is allowed only if there is no other
+ possibility.
+ DATA_ONLY
+ the device is used only for data chunk. Metadata chunk is never allowed.
+
+ The default is 'DATA_PREFERRED'; if all the disks have this setting the
+ allocator uses all of them with the same priority.
list [-t <type>] <object>
Lists available properties with their descriptions for the given object.
--
2.34.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
end of thread, other threads:[~2022-01-26 20:32 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-01-26 20:32 [PATCH 0/2][V11] btrfs-progs: allocation_hint disk property Goffredo Baroncelli
2022-01-26 20:32 ` [PATCH 1/2] btrfs-progs: new "allocation_hint" property Goffredo Baroncelli
2022-01-26 20:32 ` [PATCH 2/2] Update man page for allocator_hint property Goffredo Baroncelli
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).