Add disk quota support for btrfs

Signed-off-by: Zhu Guihua <zhugh.fnst@cn.fujitsu.com>

daemon/graphdriver/btrfs/btrfs.go

@@ -19,12 +19,15 @@ import (
 	"os"
 	"path"
 	"path/filepath"
+	"strings"
 	"syscall"
 	"unsafe"
 
 	"github.com/docker/docker/daemon/graphdriver"
 	"github.com/docker/docker/pkg/idtools"
 	"github.com/docker/docker/pkg/mount"
+	"github.com/docker/docker/pkg/parsers"
+	"github.com/docker/go-units"
 	"github.com/opencontainers/runc/libcontainer/label"
 )
 
@@ -32,6 +35,16 @@ func init() {
 	graphdriver.Register("btrfs", Init)
 }
 
+var (
+	quotaEnabled  = false
+	userDiskQuota = false
+)
+
+type btrfsOptions struct {
+	minSpace uint64
+	size     uint64
+}
+
 // Init returns a new BTRFS driver.
 // An error is returned if BTRFS is not supported.
 func Init(home string, options []string, uidMaps, gidMaps []idtools.IDMap) (graphdriver.Driver, error) {
@@ -57,21 +70,58 @@ func Init(home string, options []string, uidMaps, gidMaps []idtools.IDMap) (grap
 		return nil, err
 	}
 
+	opt, err := parseOptions(options)
+	if err != nil {
+		return nil, err
+	}
+
+	if userDiskQuota {
+		if err := subvolEnableQuota(home); err != nil {
+			return nil, err
+		}
+		quotaEnabled = true
+	}
+
 	driver := &Driver{
 		home:    home,
 		uidMaps: uidMaps,
 		gidMaps: gidMaps,
+		options: opt,
 	}
 
 	return graphdriver.NewNaiveDiffDriver(driver, uidMaps, gidMaps), nil
 }
 
+func parseOptions(opt []string) (btrfsOptions, error) {
+	var options btrfsOptions
+	for _, option := range opt {
+		key, val, err := parsers.ParseKeyValueOpt(option)
+		if err != nil {
+			return options, err
+		}
+		key = strings.ToLower(key)
+		switch key {
+		case "btrfs.min_space":
+			minSpace, err := units.RAMInBytes(val)
+			if err != nil {
+				return options, err
+			}
+			userDiskQuota = true
+			options.minSpace = uint64(minSpace)
+		default:
+			return options, fmt.Errorf("Unknown option %s", key)
+		}
+	}
+	return options, nil
+}
+
 // Driver contains information about the filesystem mounted.
 type Driver struct {
 	//root of the file system
 	home    string
 	uidMaps []idtools.IDMap
 	gidMaps []idtools.IDMap
+	options btrfsOptions
 }
 
 // String prints the name of the driver (btrfs).
@@ -100,6 +150,12 @@ func (d *Driver) GetMetadata(id string) (map[string]string, error) {
 
 // Cleanup unmounts the home directory.
 func (d *Driver) Cleanup() error {
+	if quotaEnabled {
+		if err := subvolDisableQuota(d.home); err != nil {
+			return err
+		}
+	}
+
 	return mount.Unmount(d.home)
 }
 
@@ -238,6 +294,78 @@ func subvolDelete(dirpath, name string) error {
 	return nil
 }
 
+func subvolEnableQuota(path string) error {
+	dir, err := openDir(path)
+	if err != nil {
+		return err
+	}
+	defer closeDir(dir)
+
+	var args C.struct_btrfs_ioctl_quota_ctl_args
+	args.cmd = C.BTRFS_QUOTA_CTL_ENABLE
+	_, _, errno := syscall.Syscall(syscall.SYS_IOCTL, getDirFd(dir), C.BTRFS_IOC_QUOTA_CTL,
+		uintptr(unsafe.Pointer(&args)))
+	if errno != 0 {
+		return fmt.Errorf("Failed to enable btrfs quota for %s: %v", dir, errno.Error())
+	}
+
+	return nil
+}
+
+func subvolDisableQuota(path string) error {
+	dir, err := openDir(path)
+	if err != nil {
+		return err
+	}
+	defer closeDir(dir)
+
+	var args C.struct_btrfs_ioctl_quota_ctl_args
+	args.cmd = C.BTRFS_QUOTA_CTL_DISABLE
+	_, _, errno := syscall.Syscall(syscall.SYS_IOCTL, getDirFd(dir), C.BTRFS_IOC_QUOTA_CTL,
+		uintptr(unsafe.Pointer(&args)))
+	if errno != 0 {
+		return fmt.Errorf("Failed to disable btrfs quota for %s: %v", dir, errno.Error())
+	}
+
+	return nil
+}
+
+func subvolRescanQuota(path string) error {
+	dir, err := openDir(path)
+	if err != nil {
+		return err
+	}
+	defer closeDir(dir)
+
+	var args C.struct_btrfs_ioctl_quota_rescan_args
+	_, _, errno := syscall.Syscall(syscall.SYS_IOCTL, getDirFd(dir), C.BTRFS_IOC_QUOTA_RESCAN_WAIT,
+		uintptr(unsafe.Pointer(&args)))
+	if errno != 0 {
+		return fmt.Errorf("Failed to rescan btrfs quota for %s: %v", dir, errno.Error())
+	}
+
+	return nil
+}
+
+func subvolLimitQgroup(path string, size uint64) error {
+	dir, err := openDir(path)
+	if err != nil {
+		return err
+	}
+	defer closeDir(dir)
+
+	var args C.struct_btrfs_ioctl_qgroup_limit_args
+	args.lim.max_referenced = C.__u64(size)
+	args.lim.flags = C.BTRFS_QGROUP_LIMIT_MAX_RFER
+	_, _, errno := syscall.Syscall(syscall.SYS_IOCTL, getDirFd(dir), C.BTRFS_IOC_QGROUP_LIMIT,
+		uintptr(unsafe.Pointer(&args)))
+	if errno != 0 {
+		return fmt.Errorf("Failed to limit qgroup for %s: %v", dir, errno.Error())
+	}
+
+	return nil
+}
+
 func (d *Driver) subvolumesDir() string {
 	return path.Join(d.home, "subvolumes")
 }
@@ -254,11 +382,6 @@ func (d *Driver) CreateReadWrite(id, parent, mountLabel string, storageOpt map[s
 
 // Create the filesystem with given id.
 func (d *Driver) Create(id, parent, mountLabel string, storageOpt map[string]string) error {
-
-	if len(storageOpt) != 0 {
-		return fmt.Errorf("--storage-opt is not supported for btrfs")
-	}
-
 	subvolumes := path.Join(d.home, "subvolumes")
 	rootUID, rootGID, err := idtools.GetRootUIDGID(d.uidMaps, d.gidMaps)
 	if err != nil {
@@ -285,6 +408,16 @@ func (d *Driver) Create(id, parent, mountLabel string, storageOpt map[string]str
 		}
 	}
 
+	if _, ok := storageOpt["size"]; ok {
+		driver := &Driver{}
+		if err := d.parseStorageOpt(storageOpt, driver); err != nil {
+			return err
+		}
+		if err := d.setStorageSize(path.Join(subvolumes, id), driver); err != nil {
+			return err
+		}
+	}
+
 	// if we have a remapped root (user namespaces enabled), change the created snapshot
 	// dir ownership to match
 	if rootUID != 0 || rootGID != 0 {
@@ -296,6 +429,49 @@ func (d *Driver) Create(id, parent, mountLabel string, storageOpt map[string]str
 	return label.Relabel(path.Join(subvolumes, id), mountLabel, false)
 }
 
+// Parse btrfs storage options
+func (d *Driver) parseStorageOpt(storageOpt map[string]string, driver *Driver) error {
+	// Read size to change the subvolume disk quota per container
+	for key, val := range storageOpt {
+		key := strings.ToLower(key)
+		switch key {
+		case "size":
+			size, err := units.RAMInBytes(val)
+			if err != nil {
+				return err
+			}
+			driver.options.size = uint64(size)
+		default:
+			return fmt.Errorf("Unknown option %s", key)
+		}
+	}
+
+	return nil
+}
+
+// Set btrfs storage size
+func (d *Driver) setStorageSize(dir string, driver *Driver) error {
+	if driver.options.size <= 0 {
+		return fmt.Errorf("btrfs: invalid storage size: %s", units.HumanSize(float64(driver.options.size)))
+	}
+	if d.options.minSpace > 0 && driver.options.size < d.options.minSpace {
+		return fmt.Errorf("btrfs: storage size cannot be less than %s", units.HumanSize(float64(d.options.minSpace)))
+	}
+
+	if !quotaEnabled {
+		if err := subvolEnableQuota(d.home); err != nil {
+			return err
+		}
+		quotaEnabled = true
+	}
+
+	if err := subvolLimitQgroup(dir, driver.options.size); err != nil {
+		return err
+	}
+
+	return nil
+}
+
 // Remove the filesystem with given id.
 func (d *Driver) Remove(id string) error {
 	dir := d.subvolumesDirID(id)
@@ -308,6 +484,9 @@ func (d *Driver) Remove(id string) error {
 	if err := os.RemoveAll(dir); err != nil && !os.IsNotExist(err) {
 		return err
 	}
+	if err := subvolRescanQuota(d.home); err != nil {
+		return err
+	}
 	return nil
 }
 

docs/reference/commandline/dockerd.md

@@ -185,8 +185,10 @@ Linux kernel as of [3.18.0](https://lkml.org/lkml/2014/10/26/137). Call
 ### Storage driver options
 
 Particular storage-driver can be configured with options specified with
-`--storage-opt` flags. Options for `devicemapper` are prefixed with `dm` and
-options for `zfs` start with `zfs`.
+`--storage-opt` flags. Options for `devicemapper` are prefixed with `dm`,
+options for `zfs` start with `zfs` and options for `btrfs` start with `btrfs`.
+
+#### Devicemapper options
 
 *  `dm.thinpooldev`
 
@@ -470,7 +472,7 @@ options for `zfs` start with `zfs`.
     $ dockerd --storage-opt dm.min_free_space=10%
     ```
 
-Currently supported options of `zfs`:
+#### ZFS options
 
 * `zfs.fsname`
 
@@ -482,6 +484,18 @@ Currently supported options of `zfs`:
 
         $ dockerd -s zfs --storage-opt zfs.fsname=zroot/docker
 
+#### Btrfs options
+
+* `btrfs.min_space`
+
+    Specifies the mininum size to use when creating the subvolume which is used
+    for containers. If user uses disk quota for btrfs when creating or running
+    a container with **--storage-opt size** option, docker should ensure the
+    **size** cannot be smaller than **btrfs.min_space**.
+
+    Example use:
+        $ docker daemon -s btrfs --storage-opt btrfs.min_space=10G
+
 ## Docker runtime execution options
 
 The Docker daemon relies on a

man/docker-create.1.md

@@ -333,6 +333,7 @@ unit, `b` is used. Set LIMIT to `-1` to enable unlimited swap.
    $ docker create -it --storage-opt size=120G fedora /bin/bash
 
    This (size) will allow to set the container rootfs size to 120G at creation time. User cannot pass a size less than the Default BaseFS Size.
+   This option is only available for the `devicemapper`, `btrfs` and `zfs` graphrivers.
   
 **--stop-signal**=*SIGTERM*
   Signal to stop a container. Default is SIGTERM.

man/docker-run.1.md

@@ -484,6 +484,7 @@ its root filesystem mounted as read only prohibiting any writes.
    $ docker run -it --storage-opt size=120G fedora /bin/bash
 
    This (size) will allow to set the container rootfs size to 120G at creation time. User cannot pass a size less than the Default BaseFS Size.
+   This option is only available for the `devicemapper`, `btrfs` and `zfs` graphrivers.
 
 **--stop-signal**=*SIGTERM*
   Signal to stop a container. Default is SIGTERM.

man/dockerd.8.md

@@ -247,9 +247,9 @@ backends use operating system level technologies and can be
 configured.
 
 Specify options to the storage backend with **--storage-opt** flags. The
-backends that currently take options are *devicemapper* and *zfs*.
-Options for *devicemapper* are prefixed with *dm* and options for *zfs*
-start with *zfs*.
+backends that currently take options are *devicemapper*, *zfs* and *btrfs*.
+Options for *devicemapper* are prefixed with *dm*, options for *zfs*
+start with *zfs* and options for *btrfs* start with *btrfs*.
 
 Specifically for devicemapper, the default is a "loopback" model which
 requires no pre-configuration, but is extremely inefficient.  Do not
@@ -511,6 +511,17 @@ By default docker will pick up the zfs filesystem where docker graph
 
 Example use: `dockerd -s zfs --storage-opt zfs.fsname=zroot/docker`
 
+## Btrfs options
+
+#### btrfs.min_space
+
+Specifies the mininum size to use when creating the subvolume which is used
+for containers. If user uses disk quota for btrfs when creating or running
+a container with **--storage-opt size** option, docker should ensure the
+**size** cannot be smaller than **btrfs.min_space**.
+
+Example use: `docker daemon -s btrfs --storage-opt btrfs.min_space=10G`
+
 # CLUSTER STORE OPTIONS
 
 The daemon uses libkv to advertise