btrfs: beef-up btrfs support with a limited understanding of quota

With this change we understand more than just leaf quota groups for
btrfs file systems. Specifically:

- When we create a subvolume we can now optionally add the new subvolume
  to all qgroups its parent subvolume was member of too. Alternatively
  it is also possible to insert an intermediary quota group between the
  parent's qgroups and the subvolume's leaf qgroup, which is useful for
  a concept of "subtree" qgroups, that contain a subvolume and all its
  children.

- The remove logic for subvolumes has been updated to optionally remove
  any leaf qgroups or "subtree" qgroups, following the logic above.

- The snapshot logic for subvolumes has been updated to replicate the
  original qgroup setup of the source, if it follows the "subtree"
  design described above. It will not cover qgroup setups that introduce
  arbitrary qgroups, especially those orthogonal to the subvolume
  hierarchy.

This also tries to be more graceful when setting up /var/lib/machines as
btrfs. For example, if mkfs.btrfs is missing we don't even try to set it
up as loopback device.

Fixes #1559
Fixes #1129

2 files changed, 66 insertions, 20 deletions

diff --git a/src/shared/machine-image.c b/src/shared/machine-image.c
index 9c1e4d5e13..8ed3ad7f44 100644
--- a/src/shared/machine-image.c
+++ b/src/shared/machine-image.c
@@ -176,11 +176,10 @@ static int image_make(
                                 return r;
                         if (r) {
                                 BtrfsSubvolInfo info;
-                                BtrfsQuotaInfo quota;
 
                                 /* It's a btrfs subvolume */
 
-                                r = btrfs_subvol_get_info_fd(fd, &info);
+                                r = btrfs_subvol_get_info_fd(fd, 0, &info);
                                 if (r < 0)
                                         return r;
 
@@ -195,13 +194,17 @@ static int image_make(
                                 if (r < 0)
                                         return r;
 
-                                r = btrfs_subvol_get_quota_fd(fd, &quota);
-                                if (r >= 0) {
-                                        (*ret)->usage = quota.referenced;
-                                        (*ret)->usage_exclusive = quota.exclusive;
+                                if (btrfs_quota_scan_ongoing(fd) == 0) {
+                                        BtrfsQuotaInfo quota;
 
-                                        (*ret)->limit = quota.referenced_max;
-                                        (*ret)->limit_exclusive = quota.exclusive_max;
+                                        r = btrfs_subvol_get_subtree_quota_fd(fd, 0, &quota);
+                                        if (r >= 0) {
+                                                (*ret)->usage = quota.referenced;
+                                                (*ret)->usage_exclusive = quota.exclusive;
+
+                                                (*ret)->limit = quota.referenced_max;
+                                                (*ret)->limit_exclusive = quota.exclusive_max;
+                                        }
                                 }
 
                                 return 1;
@@ -397,7 +400,7 @@ int image_remove(Image *i) {
         switch (i->type) {
 
         case IMAGE_SUBVOLUME:
-                r = btrfs_subvol_remove(i->path, true);
+                r = btrfs_subvol_remove(i->path, BTRFS_REMOVE_RECURSIVE|BTRFS_REMOVE_QUOTA);
                 if (r < 0)
                         return r;
                 break;
@@ -587,7 +590,12 @@ int image_clone(Image *i, const char *new_name, bool read_only) {
         case IMAGE_DIRECTORY:
                 new_path = strjoina("/var/lib/machines/", new_name);
 
-                r = btrfs_subvol_snapshot(i->path, new_path, (read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | BTRFS_SNAPSHOT_FALLBACK_COPY | BTRFS_SNAPSHOT_RECURSIVE);
+                r = btrfs_subvol_snapshot(i->path, new_path, (read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | BTRFS_SNAPSHOT_FALLBACK_COPY | BTRFS_SNAPSHOT_RECURSIVE | BTRFS_SNAPSHOT_QUOTA);
+
+                /* Enable "subtree" quotas for the copy, if we didn't
+                 * copy any quota from the source. */
+                (void) btrfs_subvol_auto_qgroup(i->path, 0, true);
+
                 break;
 
         case IMAGE_RAW:
@@ -629,6 +637,10 @@ int image_read_only(Image *i, bool b) {
         switch (i->type) {
 
         case IMAGE_SUBVOLUME:
+
+                /* Note that we set the flag only on the top-level
+                 * subvolume of the image. */
+
                 r = btrfs_subvol_set_read_only(i->path, b);
                 if (r < 0)
                         return r;
@@ -729,7 +741,14 @@ int image_set_limit(Image *i, uint64_t referenced_max) {
         if (i->type != IMAGE_SUBVOLUME)
                 return -EOPNOTSUPP;
 
-        return btrfs_quota_limit(i->path, referenced_max);
+        /* We set the quota both for the subvolume as well as for the
+         * subtree. The latter is mostly for historical reasons, since
+         * we didn't use to have a concept of subtree quota, and hence
+         * only modified the subvolume quota. */
+
+        (void) btrfs_qgroup_set_limit(i->path, 0, referenced_max);
+        (void) btrfs_subvol_auto_qgroup(i->path, 0, true);
+        return btrfs_subvol_set_subtree_quota_limit(i->path, 0, referenced_max);
 }
 
 int image_name_lock(const char *name, int operation, LockFile *ret) {
diff --git a/src/shared/machine-pool.c b/src/shared/machine-pool.c
index 8af78f47d5..1da7d0815f 100644
--- a/src/shared/machine-pool.c
+++ b/src/shared/machine-pool.c
@@ -170,7 +170,7 @@ int setup_machine_directory(uint64_t size, sd_bus_error *error) {
         };
         _cleanup_close_ int fd = -1, control = -1, loop = -1;
         _cleanup_free_ char* loopdev = NULL;
-        char tmpdir[] = "/tmp/import-mount.XXXXXX", *mntdir = NULL;
+        char tmpdir[] = "/tmp/machine-pool.XXXXXX", *mntdir = NULL;
         bool tmpdir_made = false, mntdir_made = false, mntdir_mounted = false;
         char buf[FORMAT_BYTES_MAX];
         int r, nr = -1;
@@ -194,14 +194,35 @@ int setup_machine_directory(uint64_t size, sd_bus_error *error) {
 
                 r = btrfs_quota_enable("/var/lib/machines", true);
                 if (r < 0)
-                        log_warning_errno(r, "Failed to enable quota, ignoring: %m");
+                        log_warning_errno(r, "Failed to enable quota for /var/lib/machines, ignoring: %m");
 
+                r = btrfs_subvol_auto_qgroup("/var/lib/machines", 0, true);
+                if (r < 0)
+                        log_warning_errno(r, "Failed to set up default quota hierarchy for /var/lib/machines, ignoring: %m");
+
+                return 1;
+        }
+
+        if (path_is_mount_point("/var/lib/machines", AT_SYMLINK_FOLLOW) > 0) {
+                log_debug("/var/lib/machines is already a mount point, not creating loopback file for it.");
                 return 0;
         }
 
-        if (path_is_mount_point("/var/lib/machines", AT_SYMLINK_FOLLOW) > 0 ||
-            dir_is_empty("/var/lib/machines") == 0)
-                return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "/var/lib/machines is not a btrfs file system. Operation is not supported on legacy file systems.");
+        r = dir_is_populated("/var/lib/machines");
+        if (r < 0 && r != -ENOENT)
+                return r;
+        if (r > 0) {
+                log_debug("/var/log/machines is already populated, not creating loopback file for it.");
+                return 0;
+        }
+
+        r = mkfs_exists("btrfs");
+        if (r == -ENOENT) {
+                log_debug("mkfs.btrfs is missing, cannot create loopback file for /var/lib/machines.");
+                return 0;
+        }
+        if (r < 0)
+                return r;
 
         fd = setup_machine_raw(size, error);
         if (fd < 0)
@@ -266,6 +287,10 @@ int setup_machine_directory(uint64_t size, sd_bus_error *error) {
         if (r < 0)
                 log_warning_errno(r, "Failed to enable quota, ignoring: %m");
 
+        r = btrfs_subvol_auto_qgroup(mntdir, 0, true);
+        if (r < 0)
+                log_warning_errno(r, "Failed to set up default quota hierarchy, ignoring: %m");
+
         if (chmod(mntdir, 0700) < 0) {
                 r = sd_bus_error_set_errnof(error, errno, "Failed to fix owner: %m");
                 goto fail;
@@ -286,7 +311,7 @@ int setup_machine_directory(uint64_t size, sd_bus_error *error) {
         (void) rmdir(mntdir);
         (void) rmdir(tmpdir);
 
-        return 0;
+        return 1;
 
 fail:
         if (mntdir_mounted)
@@ -370,9 +395,11 @@ int grow_machine_directory(void) {
         if (r <= 0)
                 return r;
 
-        r = btrfs_quota_limit("/var/lib/machines", new_size);
-        if (r < 0)
-                return r;
+        /* Also bump the quota, of both the subvolume leaf qgroup, as
+         * well as of any subtree quota group by the same id but a
+         * higher level, if it exists. */
+        (void) btrfs_qgroup_set_limit("/var/lib/machines", 0, new_size);
+        (void) btrfs_subvol_set_subtree_quota_limit("/var/lib/machines", 0, new_size);
 
         log_info("Grew /var/lib/machines btrfs loopback file system to %s.", format_bytes(buf, sizeof(buf), new_size));
         return 1;