linux-fsdevel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v5 0/8] Add quota support to UBIFS
@ 2021-01-22 15:15 Sascha Hauer
  2021-01-22 15:15 ` [PATCH 1/8] quota: Allow to pass mount path to quotactl Sascha Hauer
                   ` (7 more replies)
  0 siblings, 8 replies; 23+ messages in thread
From: Sascha Hauer @ 2021-01-22 15:15 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: Richard Weinberger, linux-mtd, kernel, Jan Kara, Sascha Hauer

This series adds quota support to UBIFS.

I realized that counting xattr inodes for quota is quite broken, so I
removed it this time. The problem is that the code used to add the xattr
inodes to the owners of the xattr inodes which are not necessarily the
owner of the files. When root added xattrs to files he doesn't own or
when a file was chowned then the inode count became wrong.

One point that has lead to discussions last time was the integration of
the mountpoint used for quotactl rather than the path to a block device
(which doesn't exist for UBIFS). Last time Al suggested that getting a
super_block from a path should be in core code. I was about to integrate
Jans approach of introducing several variants of a hold_super_()
function when I realized that the variants of get_super_* took the
opposite direction and are now open coded in quota code. I took a
similar approach of open coding this which brings us close to where we
once were. Let's see how welcomed this is.

This series follows a very simple approach to quota: Neither the quota
limits nor the quota usage are ever written to the medium. The quota
usage is reconstructed from the filesystem during mount time. The quota
limits must be set by the user each time after mount. This is probably
not very convenient for systems that are used interactively, but UBIFS
is targetted to embedded systems and here running a script after mount
shouldn't be a problem. This of course isn't the way quota was thought
to be, but I believe this is a good compromise for a feature that I predict
is only rarely used on UBIFS. The big upside of this approach is that
no on-disk format changes are required and thus we can't get any
broken/corrupt filesystems because of quota support. Reconstructing the
quota data each time during mount has an noticable but I think for many
cases acceptable time overhead. I mounted a ~56MiB rootfs with 1920 files
which takes around 0.7s longer when quota is enabled.

As UBIFS works on mtd there is no block_device involved. The quotactl
system call requires a path to a block device as argument. To overcome
this we add support for passing the mount point instead. This is done
with a new Q_PATH flag to the quotactl syscall indicating that the special
argument belongs to the mount path rather than a path to the block device
file

The UBIFS quota support itself is based on a series by Dongsheng Yang
posted here:
http://lists.infradead.org/pipermail/linux-mtd/2015-September/061812.html
This part hasn't changed much, except that the code for reading and writing
quota files has been dropped.

changes since v4:
- Rebase on v5.11-rc5
- Drop quota inode counting for xattrs

changes since v3:
- implement ubifs_dqblk_find() using ubifs_dqblk_find_next()
- Fix copy/paste errors in flag settings
- drop unnecessary inline declaration
- add ubifs_assert() for catching wrong usage of setflags()
- add helper function for projid checking
- Add a feature flag for supporting projid

changes since v2:
- Rebase on Jans quota-without-inode series
- Use recently introduced vfs_ioc_fssetxattr_check() and simple_fill_fsxattr()
- fix project quota support (was broken in v2 due to upstream changes in UBIFS)
- check for illegal renames due to different project id

Changes since v1:
- Introduce Q_PATH flag to make passing a mountpath explicit
- Do not mess with fs layer as suggested by Al Viro
- create separate usrquota, grpquota and prjquota options rather than just
  a single quota option
- register a UBIFS specific quota_format and use dquot_enable()
- drop "quota: Only module_put the format when existing" which is no
  longer necesary

Sascha Hauer (8):
  quota: Allow to pass mount path to quotactl
  ubifs: move checks and preparation into setflags()
  ubifs: Add support for FS_IOC_FS[SG]ETXATTR ioctls
  ubifs: do not ubifs_inode() on potentially NULL pointer
  ubifs: Factor out ubifs_set_feature_flag()
  ubifs: Add support for project id
  ubifs: export get_znode
  ubifs: Add quota support

 Documentation/filesystems/ubifs.rst |   6 +
 fs/quota/quota.c                    |  66 ++-
 fs/ubifs/Makefile                   |   1 +
 fs/ubifs/dir.c                      | 139 +++++--
 fs/ubifs/file.c                     |  43 ++
 fs/ubifs/ioctl.c                    | 221 ++++++++--
 fs/ubifs/journal.c                  |   4 +-
 fs/ubifs/quota.c                    | 606 ++++++++++++++++++++++++++++
 fs/ubifs/sb.c                       |  29 +-
 fs/ubifs/super.c                    |  87 +++-
 fs/ubifs/tnc.c                      |  32 +-
 fs/ubifs/ubifs-media.h              |  10 +-
 fs/ubifs/ubifs.h                    |  44 ++
 fs/ubifs/xattr.c                    |   5 +-
 include/uapi/linux/quota.h          |   2 +
 15 files changed, 1198 insertions(+), 97 deletions(-)
 create mode 100644 fs/ubifs/quota.c

-- 
2.20.1


^ permalink raw reply	[flat|nested] 23+ messages in thread

* [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-22 15:15 [PATCH v5 0/8] Add quota support to UBIFS Sascha Hauer
@ 2021-01-22 15:15 ` Sascha Hauer
  2021-01-22 17:16   ` Christoph Hellwig
  2021-01-22 15:15 ` [PATCH 2/8] ubifs: move checks and preparation into setflags() Sascha Hauer
                   ` (6 subsequent siblings)
  7 siblings, 1 reply; 23+ messages in thread
From: Sascha Hauer @ 2021-01-22 15:15 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: Richard Weinberger, linux-mtd, kernel, Jan Kara, Sascha Hauer

This patch introduces the Q_PATH flag to the quotactl cmd argument.
When given, the path given in the special argument to quotactl will
be the mount path where the filesystem is mounted, instead of a path
to the block device.
This is necessary for filesystems which do not have a block device as
backing store. Particularly this is done for upcoming UBIFS support.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 fs/quota/quota.c           | 66 ++++++++++++++++++++++++++++----------
 include/uapi/linux/quota.h |  1 +
 2 files changed, 50 insertions(+), 17 deletions(-)

diff --git a/fs/quota/quota.c b/fs/quota/quota.c
index 6d16b2be5ac4..f653b27a9a4e 100644
--- a/fs/quota/quota.c
+++ b/fs/quota/quota.c
@@ -17,6 +17,7 @@
 #include <linux/capability.h>
 #include <linux/quotaops.h>
 #include <linux/types.h>
+#include <linux/mount.h>
 #include <linux/writeback.h>
 #include <linux/nospec.h>
 #include "compat.h"
@@ -859,25 +860,10 @@ static bool quotactl_cmd_onoff(int cmd)
 		 (cmd == Q_XQUOTAON) || (cmd == Q_XQUOTAOFF);
 }
 
-/*
- * look up a superblock on which quota ops will be performed
- * - use the name of a block device to find the superblock thereon
- */
-static struct super_block *quotactl_block(const char __user *special, int cmd)
+static struct super_block *quotactl_sb(dev_t dev, int cmd)
 {
-#ifdef CONFIG_BLOCK
 	struct super_block *sb;
-	struct filename *tmp = getname(special);
 	bool excl = false, thawed = false;
-	int error;
-	dev_t dev;
-
-	if (IS_ERR(tmp))
-		return ERR_CAST(tmp);
-	error = lookup_bdev(tmp->name, &dev);
-	putname(tmp);
-	if (error)
-		return ERR_PTR(error);
 
 	if (quotactl_cmd_onoff(cmd)) {
 		excl = true;
@@ -901,12 +887,50 @@ static struct super_block *quotactl_block(const char __user *special, int cmd)
 		goto retry;
 	}
 	return sb;
+}
+
+/*
+ * look up a superblock on which quota ops will be performed
+ * - use the name of a block device to find the superblock thereon
+ */
+static struct super_block *quotactl_block(const char __user *special, int cmd)
+{
+#ifdef CONFIG_BLOCK
+	struct filename *tmp = getname(special);
+	int error;
+	dev_t dev;
 
+	if (IS_ERR(tmp))
+		return ERR_CAST(tmp);
+	error = lookup_bdev(tmp->name, &dev);
+	putname(tmp);
+	if (error)
+		return ERR_PTR(error);
+
+	return quotactl_sb(dev, cmd);
 #else
 	return ERR_PTR(-ENODEV);
 #endif
 }
 
+static struct super_block *quotactl_path(const char __user *special, int cmd)
+{
+	struct super_block *sb;
+	struct path path;
+	int error;
+
+	error = user_path_at(AT_FDCWD, special, LOOKUP_FOLLOW | LOOKUP_AUTOMOUNT,
+			   &path);
+	if (error)
+		return ERR_PTR(error);
+
+	sb = quotactl_sb(path.mnt->mnt_sb->s_dev, cmd);
+
+	path_put(&path);
+
+	return sb;
+}
+
 /*
  * This is the system call interface. This communicates with
  * the user-level programs. Currently this only supports diskquota
@@ -920,6 +944,7 @@ SYSCALL_DEFINE4(quotactl, unsigned int, cmd, const char __user *, special,
 	struct super_block *sb = NULL;
 	struct path path, *pathp = NULL;
 	int ret;
+	bool q_path;
 
 	cmds = cmd >> SUBCMDSHIFT;
 	type = cmd & SUBCMDMASK;
@@ -927,6 +952,9 @@ SYSCALL_DEFINE4(quotactl, unsigned int, cmd, const char __user *, special,
 	if (type >= MAXQUOTAS)
 		return -EINVAL;
 
+	q_path = cmds & Q_PATH;
+	cmds &= ~Q_PATH;
+
 	/*
 	 * As a special case Q_SYNC can be called without a specific device.
 	 * It will iterate all superblocks that have quota enabled and call
@@ -951,7 +979,11 @@ SYSCALL_DEFINE4(quotactl, unsigned int, cmd, const char __user *, special,
 			pathp = &path;
 	}
 
-	sb = quotactl_block(special, cmds);
+	if (q_path)
+		sb = quotactl_path(special, cmds);
+	else
+		sb = quotactl_block(special, cmds);
+
 	if (IS_ERR(sb)) {
 		ret = PTR_ERR(sb);
 		goto out;
diff --git a/include/uapi/linux/quota.h b/include/uapi/linux/quota.h
index f17c9636a859..e1787c0df601 100644
--- a/include/uapi/linux/quota.h
+++ b/include/uapi/linux/quota.h
@@ -71,6 +71,7 @@
 #define Q_GETQUOTA 0x800007	/* get user quota structure */
 #define Q_SETQUOTA 0x800008	/* set user quota structure */
 #define Q_GETNEXTQUOTA 0x800009	/* get disk limits and usage >= ID */
+#define Q_PATH     0x400000	/* quotactl special arg contains mount path */
 
 /* Quota format type IDs */
 #define	QFMT_VFS_OLD 1
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 23+ messages in thread

* [PATCH 2/8] ubifs: move checks and preparation into setflags()
  2021-01-22 15:15 [PATCH v5 0/8] Add quota support to UBIFS Sascha Hauer
  2021-01-22 15:15 ` [PATCH 1/8] quota: Allow to pass mount path to quotactl Sascha Hauer
@ 2021-01-22 15:15 ` Sascha Hauer
  2021-01-22 15:15 ` [PATCH 3/8] ubifs: Add support for FS_IOC_FS[SG]ETXATTR ioctls Sascha Hauer
                   ` (5 subsequent siblings)
  7 siblings, 0 replies; 23+ messages in thread
From: Sascha Hauer @ 2021-01-22 15:15 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: Richard Weinberger, linux-mtd, kernel, Jan Kara, Sascha Hauer

setflags() can be reused for upcoming FS_IOC_FS[SG]ETXATTR ioctl support.
In preparation for that move the checks and preparation into that
function so we can reuse them as well.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 fs/ubifs/ioctl.c | 46 +++++++++++++++++++++++++++-------------------
 1 file changed, 27 insertions(+), 19 deletions(-)

diff --git a/fs/ubifs/ioctl.c b/fs/ubifs/ioctl.c
index 4363d85a3fd4..b64a9cec7d47 100644
--- a/fs/ubifs/ioctl.c
+++ b/fs/ubifs/ioctl.c
@@ -101,18 +101,35 @@ static int ubifs2ioctl(int ubifs_flags)
 	return ioctl_flags;
 }
 
-static int setflags(struct inode *inode, int flags)
+static int setflags(struct file *file, int flags)
 {
 	int oldflags, err, release;
+	struct inode *inode = file_inode(file);
 	struct ubifs_inode *ui = ubifs_inode(inode);
 	struct ubifs_info *c = inode->i_sb->s_fs_info;
 	struct ubifs_budget_req req = { .dirtied_ino = 1,
 					.dirtied_ino_d = ui->data_len };
 
-	err = ubifs_budget_space(c, &req);
+	if (IS_RDONLY(inode))
+		return -EROFS;
+
+	if (!inode_owner_or_capable(inode))
+		return -EACCES;
+
+	/*
+	 * Make sure the file-system is read-write and make sure it
+	 * will not become read-only while we are changing the flags.
+	 */
+	err = mnt_want_write_file(file);
 	if (err)
 		return err;
 
+	dbg_gen("set flags: %#x, i_flags %#x", flags, inode->i_flags);
+
+	err = ubifs_budget_space(c, &req);
+	if (err)
+		goto out_drop;
+
 	mutex_lock(&ui->ui_mutex);
 	oldflags = ubifs2ioctl(ui->flags);
 	err = vfs_ioc_setflags_prepare(inode, oldflags, flags);
@@ -131,11 +148,17 @@ static int setflags(struct inode *inode, int flags)
 		ubifs_release_budget(c, &req);
 	if (IS_SYNC(inode))
 		err = write_inode_now(inode, 1);
+
+	mnt_drop_write_file(file);
+
 	return err;
 
 out_unlock:
 	mutex_unlock(&ui->ui_mutex);
 	ubifs_release_budget(c, &req);
+out_drop:
+	mnt_drop_write_file(file);
+
 	return err;
 }
 
@@ -152,12 +175,6 @@ long ubifs_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 		return put_user(flags, (int __user *) arg);
 
 	case FS_IOC_SETFLAGS: {
-		if (IS_RDONLY(inode))
-			return -EROFS;
-
-		if (!inode_owner_or_capable(inode))
-			return -EACCES;
-
 		if (get_user(flags, (int __user *) arg))
 			return -EFAULT;
 
@@ -168,17 +185,8 @@ long ubifs_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 		if (!S_ISDIR(inode->i_mode))
 			flags &= ~FS_DIRSYNC_FL;
 
-		/*
-		 * Make sure the file-system is read-write and make sure it
-		 * will not become read-only while we are changing the flags.
-		 */
-		err = mnt_want_write_file(file);
-		if (err)
-			return err;
-		dbg_gen("set flags: %#x, i_flags %#x", flags, inode->i_flags);
-		err = setflags(inode, flags);
-		mnt_drop_write_file(file);
-		return err;
+
+		return setflags(file, flags);
 	}
 	case FS_IOC_SET_ENCRYPTION_POLICY: {
 		struct ubifs_info *c = inode->i_sb->s_fs_info;
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 23+ messages in thread

* [PATCH 3/8] ubifs: Add support for FS_IOC_FS[SG]ETXATTR ioctls
  2021-01-22 15:15 [PATCH v5 0/8] Add quota support to UBIFS Sascha Hauer
  2021-01-22 15:15 ` [PATCH 1/8] quota: Allow to pass mount path to quotactl Sascha Hauer
  2021-01-22 15:15 ` [PATCH 2/8] ubifs: move checks and preparation into setflags() Sascha Hauer
@ 2021-01-22 15:15 ` Sascha Hauer
  2021-01-22 17:18   ` Christoph Hellwig
  2021-01-22 15:15 ` [PATCH 4/8] ubifs: do not ubifs_inode() on potentially NULL pointer Sascha Hauer
                   ` (4 subsequent siblings)
  7 siblings, 1 reply; 23+ messages in thread
From: Sascha Hauer @ 2021-01-22 15:15 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: Richard Weinberger, linux-mtd, kernel, Jan Kara, Sascha Hauer

The FS_IOC_FS[SG]ETXATTR ioctls are an alternative to FS_IOC_[GS]ETFLAGS
with additional features. This patch adds support for these ioctls.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 fs/ubifs/ioctl.c | 107 ++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 102 insertions(+), 5 deletions(-)

diff --git a/fs/ubifs/ioctl.c b/fs/ubifs/ioctl.c
index b64a9cec7d47..5977edd4d185 100644
--- a/fs/ubifs/ioctl.c
+++ b/fs/ubifs/ioctl.c
@@ -101,9 +101,46 @@ static int ubifs2ioctl(int ubifs_flags)
 	return ioctl_flags;
 }
 
-static int setflags(struct file *file, int flags)
+/* Transfer xflags flags to internal */
+static unsigned long ubifs_xflags_to_iflags(__u32 xflags)
 {
-	int oldflags, err, release;
+	unsigned long iflags = 0;
+
+	if (xflags & FS_XFLAG_SYNC)
+		iflags |= UBIFS_SYNC_FL;
+	if (xflags & FS_XFLAG_IMMUTABLE)
+		iflags |= UBIFS_IMMUTABLE_FL;
+	if (xflags & FS_XFLAG_APPEND)
+		iflags |= UBIFS_APPEND_FL;
+
+        return iflags;
+}
+
+/* Transfer internal flags to xflags */
+static __u32 ubifs_iflags_to_xflags(unsigned long flags)
+{
+	__u32 xflags = 0;
+
+	if (flags & UBIFS_SYNC_FL)
+		xflags |= FS_XFLAG_SYNC;
+	if (flags & UBIFS_IMMUTABLE_FL)
+		xflags |= FS_XFLAG_IMMUTABLE;
+	if (flags & UBIFS_APPEND_FL)
+		xflags |= FS_XFLAG_APPEND;
+
+        return xflags;
+}
+
+static void ubifs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
+{
+	struct ubifs_inode *ui = ubifs_inode(inode);
+
+	simple_fill_fsxattr(fa, ubifs_iflags_to_xflags(ui->flags));
+}
+
+static int setflags(struct file *file, int flags, struct fsxattr *fa)
+{
+	int ubi_flags, oldflags, err, release;
 	struct inode *inode = file_inode(file);
 	struct ubifs_inode *ui = ubifs_inode(inode);
 	struct ubifs_info *c = inode->i_sb->s_fs_info;
@@ -116,6 +153,13 @@ static int setflags(struct file *file, int flags)
 	if (!inode_owner_or_capable(inode))
 		return -EACCES;
 
+	ubifs_assert(c, !(flags && fa));
+
+	if (fa)
+		ubi_flags = ubifs_xflags_to_iflags(fa->fsx_xflags);
+	else
+		ubi_flags = ioctl2ubifs(flags);
+
 	/*
 	 * Make sure the file-system is read-write and make sure it
 	 * will not become read-only while we are changing the flags.
@@ -132,12 +176,24 @@ static int setflags(struct file *file, int flags)
 
 	mutex_lock(&ui->ui_mutex);
 	oldflags = ubifs2ioctl(ui->flags);
-	err = vfs_ioc_setflags_prepare(inode, oldflags, flags);
+	err = vfs_ioc_setflags_prepare(inode, oldflags, ubifs2ioctl(ubi_flags));
 	if (err)
 		goto out_unlock;
 
 	ui->flags &= ~ioctl2ubifs(UBIFS_SETTABLE_IOCTL_FLAGS);
-	ui->flags |= ioctl2ubifs(flags);
+
+	if (fa) {
+		struct fsxattr old_fa;
+
+		ubifs_fill_fsxattr(inode, &old_fa);
+
+		err = vfs_ioc_fssetxattr_check(inode, &old_fa, fa);
+		if (err)
+			goto out_unlock;
+	}
+
+	ui->flags |= ubi_flags;
+
 	ubifs_set_inode_flags(inode);
 	inode->i_ctime = current_time(inode);
 	release = ui->dirty;
@@ -162,6 +218,41 @@ static int setflags(struct file *file, int flags)
 	return err;
 }
 
+static int ubifs_ioc_fsgetxattr(struct file *file, void __user *arg)
+{
+	struct inode *inode = file_inode(file);
+	struct fsxattr fa;
+
+	ubifs_fill_fsxattr(inode, &fa);
+
+	if (copy_to_user(arg, &fa, sizeof(fa)))
+		return -EFAULT;
+
+	return 0;
+}
+
+static int check_xflags(unsigned int flags)
+{
+	if (flags & ~(FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))
+		return -EOPNOTSUPP;
+	return 0;
+}
+
+static int ubifs_ioc_fssetxattr(struct file *file, const void __user *arg)
+{
+	struct fsxattr fa;
+	int err;
+
+	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
+		return -EFAULT;
+
+	err = check_xflags(fa.fsx_xflags);
+	if (err)
+		return err;
+
+	return setflags(file, 0, &fa);
+}
+
 long ubifs_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 {
 	int flags, err;
@@ -186,7 +277,7 @@ long ubifs_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 			flags &= ~FS_DIRSYNC_FL;
 
 
-		return setflags(file, flags);
+		return setflags(file, flags, NULL);
 	}
 	case FS_IOC_SET_ENCRYPTION_POLICY: {
 		struct ubifs_info *c = inode->i_sb->s_fs_info;
@@ -218,6 +309,12 @@ long ubifs_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 	case FS_IOC_GET_ENCRYPTION_NONCE:
 		return fscrypt_ioctl_get_nonce(file, (void __user *)arg);
 
+	case FS_IOC_FSGETXATTR:
+		return ubifs_ioc_fsgetxattr(file, (void __user *)arg);
+
+	case FS_IOC_FSSETXATTR:
+		return ubifs_ioc_fssetxattr(file, (const void __user *)arg);
+
 	default:
 		return -ENOTTY;
 	}
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 23+ messages in thread

* [PATCH 4/8] ubifs: do not ubifs_inode() on potentially NULL pointer
  2021-01-22 15:15 [PATCH v5 0/8] Add quota support to UBIFS Sascha Hauer
                   ` (2 preceding siblings ...)
  2021-01-22 15:15 ` [PATCH 3/8] ubifs: Add support for FS_IOC_FS[SG]ETXATTR ioctls Sascha Hauer
@ 2021-01-22 15:15 ` Sascha Hauer
  2021-01-22 15:15 ` [PATCH 5/8] ubifs: Factor out ubifs_set_feature_flag() Sascha Hauer
                   ` (3 subsequent siblings)
  7 siblings, 0 replies; 23+ messages in thread
From: Sascha Hauer @ 2021-01-22 15:15 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: Richard Weinberger, linux-mtd, kernel, Jan Kara, Sascha Hauer

new_inode() may return NULL, so only derefence the return inode when
non NULL. This is merely a cleanup as calling ubifs_inode() on a NULL
pointer doesn't do any harm, only using the return value would.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 fs/ubifs/dir.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/ubifs/dir.c b/fs/ubifs/dir.c
index 9a6b8660425a..cc1d7f5085ab 100644
--- a/fs/ubifs/dir.c
+++ b/fs/ubifs/dir.c
@@ -82,10 +82,10 @@ struct inode *ubifs_new_inode(struct ubifs_info *c, struct inode *dir,
 	bool encrypted = false;
 
 	inode = new_inode(c->vfs_sb);
-	ui = ubifs_inode(inode);
 	if (!inode)
 		return ERR_PTR(-ENOMEM);
 
+	ui = ubifs_inode(inode);
 	/*
 	 * Set 'S_NOCMTIME' to prevent VFS form updating [mc]time of inodes and
 	 * marking them dirty in file write path (see 'file_update_time()').
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 23+ messages in thread

* [PATCH 5/8] ubifs: Factor out ubifs_set_feature_flag()
  2021-01-22 15:15 [PATCH v5 0/8] Add quota support to UBIFS Sascha Hauer
                   ` (3 preceding siblings ...)
  2021-01-22 15:15 ` [PATCH 4/8] ubifs: do not ubifs_inode() on potentially NULL pointer Sascha Hauer
@ 2021-01-22 15:15 ` Sascha Hauer
  2021-01-22 15:15 ` [PATCH 6/8] ubifs: Add support for project id Sascha Hauer
                   ` (2 subsequent siblings)
  7 siblings, 0 replies; 23+ messages in thread
From: Sascha Hauer @ 2021-01-22 15:15 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: Richard Weinberger, linux-mtd, kernel, Jan Kara, Sascha Hauer

The code setting a feature flag can be reused for upcoming projid
support. Factor out a function to share the code.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 fs/ubifs/sb.c | 24 +++++++++++++++---------
 1 file changed, 15 insertions(+), 9 deletions(-)

diff --git a/fs/ubifs/sb.c b/fs/ubifs/sb.c
index c160f718c288..87466836fcfc 100644
--- a/fs/ubifs/sb.c
+++ b/fs/ubifs/sb.c
@@ -924,28 +924,34 @@ int ubifs_fixup_free_space(struct ubifs_info *c)
 	return err;
 }
 
-int ubifs_enable_encryption(struct ubifs_info *c)
+static int ubifs_set_feature_flag(struct ubifs_info *c, unsigned int flag)
 {
-	int err;
 	struct ubifs_sb_node *sup = c->sup_node;
 
-	if (!IS_ENABLED(CONFIG_FS_ENCRYPTION))
-		return -EOPNOTSUPP;
-
-	if (c->encrypted)
+	if ((sup->flags & cpu_to_le32(flag)) == cpu_to_le32(flag))
 		return 0;
 
 	if (c->ro_mount || c->ro_media)
 		return -EROFS;
 
 	if (c->fmt_version < 5) {
-		ubifs_err(c, "on-flash format version 5 is needed for encryption");
+		ubifs_err(c, "on-flash format version 5 is needed for feature flags");
 		return -EINVAL;
 	}
 
-	sup->flags |= cpu_to_le32(UBIFS_FLG_ENCRYPTION);
+	sup->flags |= cpu_to_le32(flag);
+
+	return ubifs_write_sb_node(c, sup);
+}
+
+int ubifs_enable_encryption(struct ubifs_info *c)
+{
+	int err;
+
+	if (!IS_ENABLED(CONFIG_FS_ENCRYPTION))
+		return -EOPNOTSUPP;
 
-	err = ubifs_write_sb_node(c, sup);
+	err = ubifs_set_feature_flag(c, UBIFS_FLG_ENCRYPTION);
 	if (!err)
 		c->encrypted = 1;
 
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 23+ messages in thread

* [PATCH 6/8] ubifs: Add support for project id
  2021-01-22 15:15 [PATCH v5 0/8] Add quota support to UBIFS Sascha Hauer
                   ` (4 preceding siblings ...)
  2021-01-22 15:15 ` [PATCH 5/8] ubifs: Factor out ubifs_set_feature_flag() Sascha Hauer
@ 2021-01-22 15:15 ` Sascha Hauer
  2021-01-22 15:15 ` [PATCH 7/8] ubifs: export get_znode Sascha Hauer
  2021-01-22 15:15 ` [PATCH 8/8] ubifs: Add quota support Sascha Hauer
  7 siblings, 0 replies; 23+ messages in thread
From: Sascha Hauer @ 2021-01-22 15:15 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: Richard Weinberger, linux-mtd, kernel, Jan Kara, Sascha Hauer

The project id is necessary for quota project id support. This adds
support for the project id to UBIFS as well as support for the
FS_PROJINHERIT_FL flag.

This includes a change for the UBIFS on-disk format. struct
ubifs_ino_node gains a project id number and a UBIFS_PROJINHERIT_FL
flag. A feature flag is added to prevent older UBIFS implementations
from mounting images with project ids. The feature flag is set when
a projid is set for the first time.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 fs/ubifs/dir.c         | 36 ++++++++++++++++++++++++++++++++++--
 fs/ubifs/ioctl.c       | 42 ++++++++++++++++++++++++++++++++++++++++--
 fs/ubifs/journal.c     |  2 +-
 fs/ubifs/sb.c          |  5 +++++
 fs/ubifs/super.c       |  1 +
 fs/ubifs/ubifs-media.h | 10 +++++++---
 fs/ubifs/ubifs.h       |  5 +++++
 7 files changed, 93 insertions(+), 8 deletions(-)

diff --git a/fs/ubifs/dir.c b/fs/ubifs/dir.c
index cc1d7f5085ab..e724fc91e2d1 100644
--- a/fs/ubifs/dir.c
+++ b/fs/ubifs/dir.c
@@ -56,7 +56,8 @@ static int inherit_flags(const struct inode *dir, umode_t mode)
 		 */
 		return 0;
 
-	flags = ui->flags & (UBIFS_COMPR_FL | UBIFS_SYNC_FL | UBIFS_DIRSYNC_FL);
+	flags = ui->flags & (UBIFS_COMPR_FL | UBIFS_SYNC_FL | UBIFS_DIRSYNC_FL |
+			     UBIFS_PROJINHERIT_FL);
 	if (!S_ISDIR(mode))
 		/* The "DIRSYNC" flag only applies to directories */
 		flags &= ~UBIFS_DIRSYNC_FL;
@@ -105,6 +106,11 @@ struct inode *ubifs_new_inode(struct ubifs_info *c, struct inode *dir,
 		goto out_iput;
 	}
 
+	if (ubifs_inode(dir)->flags & UBIFS_PROJINHERIT_FL)
+		ui->projid = ubifs_inode(dir)->projid;
+	else
+		ui->projid = make_kprojid(&init_user_ns, UBIFS_DEF_PROJID);
+
 	switch (mode & S_IFMT) {
 	case S_IFREG:
 		inode->i_mapping->a_ops = &ubifs_file_address_operations;
@@ -709,6 +715,9 @@ static int ubifs_link(struct dentry *old_dentry, struct inode *dir,
 	ubifs_assert(c, inode_is_locked(dir));
 	ubifs_assert(c, inode_is_locked(inode));
 
+	if (!projid_eq(dir_ui->projid, ui->projid))
+                return -EXDEV;
+
 	err = fscrypt_prepare_link(old_dentry, dir, dentry);
 	if (err)
 		return err;
@@ -1542,6 +1551,20 @@ static int ubifs_xrename(struct inode *old_dir, struct dentry *old_dentry,
 	return err;
 }
 
+static int ubifs_rename_check_projid(struct inode *dir, struct inode *inode)
+{
+	struct ubifs_inode *ud = ubifs_inode(dir);
+	struct ubifs_inode *ui = ubifs_inode(inode);
+
+	if (!(ud->flags & UBIFS_PROJINHERIT_FL))
+		return 0;
+
+	if (projid_eq(ud->projid, ui->projid))
+		return 0;
+
+	return -EXDEV;
+}
+
 static int ubifs_rename(struct inode *old_dir, struct dentry *old_dentry,
 			struct inode *new_dir, struct dentry *new_dentry,
 			unsigned int flags)
@@ -1560,8 +1583,17 @@ static int ubifs_rename(struct inode *old_dir, struct dentry *old_dentry,
 	if (err)
 		return err;
 
-	if (flags & RENAME_EXCHANGE)
+	err = ubifs_rename_check_projid(new_dir, old_dentry->d_inode);
+	if (err)
+		return err;
+
+	if (flags & RENAME_EXCHANGE) {
+		err = ubifs_rename_check_projid(old_dir, new_dentry->d_inode);
+		if (err)
+			return err;
+
 		return ubifs_xrename(old_dir, old_dentry, new_dir, new_dentry);
+	}
 
 	return do_rename(old_dir, old_dentry, new_dir, new_dentry, flags);
 }
diff --git a/fs/ubifs/ioctl.c b/fs/ubifs/ioctl.c
index 5977edd4d185..a9a79935cc13 100644
--- a/fs/ubifs/ioctl.c
+++ b/fs/ubifs/ioctl.c
@@ -19,7 +19,7 @@
 /* Need to be kept consistent with checked flags in ioctl2ubifs() */
 #define UBIFS_SETTABLE_IOCTL_FLAGS \
 	(FS_COMPR_FL | FS_SYNC_FL | FS_APPEND_FL | \
-	 FS_IMMUTABLE_FL | FS_DIRSYNC_FL)
+	 FS_IMMUTABLE_FL | FS_DIRSYNC_FL | FS_PROJINHERIT_FL)
 
 /* Need to be kept consistent with checked flags in ubifs2ioctl() */
 #define UBIFS_GETTABLE_IOCTL_FLAGS \
@@ -70,6 +70,8 @@ static int ioctl2ubifs(int ioctl_flags)
 		ubifs_flags |= UBIFS_IMMUTABLE_FL;
 	if (ioctl_flags & FS_DIRSYNC_FL)
 		ubifs_flags |= UBIFS_DIRSYNC_FL;
+	if (ioctl_flags & FS_PROJINHERIT_FL)
+		ubifs_flags |= UBIFS_PROJINHERIT_FL;
 
 	return ubifs_flags;
 }
@@ -97,6 +99,8 @@ static int ubifs2ioctl(int ubifs_flags)
 		ioctl_flags |= FS_DIRSYNC_FL;
 	if (ubifs_flags & UBIFS_CRYPT_FL)
 		ioctl_flags |= FS_ENCRYPT_FL;
+	if (ubifs_flags & UBIFS_PROJINHERIT_FL)
+		ioctl_flags |= FS_PROJINHERIT_FL;
 
 	return ioctl_flags;
 }
@@ -112,6 +116,8 @@ static unsigned long ubifs_xflags_to_iflags(__u32 xflags)
 		iflags |= UBIFS_IMMUTABLE_FL;
 	if (xflags & FS_XFLAG_APPEND)
 		iflags |= UBIFS_APPEND_FL;
+	if (xflags & FS_XFLAG_PROJINHERIT)
+		iflags |= UBIFS_PROJINHERIT_FL;
 
         return iflags;
 }
@@ -127,15 +133,34 @@ static __u32 ubifs_iflags_to_xflags(unsigned long flags)
 		xflags |= FS_XFLAG_IMMUTABLE;
 	if (flags & UBIFS_APPEND_FL)
 		xflags |= FS_XFLAG_APPEND;
+	if (flags & UBIFS_PROJINHERIT_FL)
+		xflags |= FS_XFLAG_PROJINHERIT;
 
         return xflags;
 }
 
+static int ubifs_ioc_setproject(struct file *file, __u32 projid)
+{
+	struct inode *inode = file_inode(file);
+	struct ubifs_inode *ui = ubifs_inode(inode);
+	kprojid_t kprojid;
+
+	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
+	if (projid_eq(kprojid, ui->projid))
+		return 0;
+
+	ui->projid = kprojid;
+
+	return 0;
+}
+
 static void ubifs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
 {
 	struct ubifs_inode *ui = ubifs_inode(inode);
 
 	simple_fill_fsxattr(fa, ubifs_iflags_to_xflags(ui->flags));
+
+	fa->fsx_projid = (__u32)from_kprojid(&init_user_ns, ui->projid);
 }
 
 static int setflags(struct file *file, int flags, struct fsxattr *fa)
@@ -146,6 +171,7 @@ static int setflags(struct file *file, int flags, struct fsxattr *fa)
 	struct ubifs_info *c = inode->i_sb->s_fs_info;
 	struct ubifs_budget_req req = { .dirtied_ino = 1,
 					.dirtied_ino_d = ui->data_len };
+	__u32 projid = UBIFS_DEF_PROJID;
 
 	if (IS_RDONLY(inode))
 		return -EROFS;
@@ -190,9 +216,20 @@ static int setflags(struct file *file, int flags, struct fsxattr *fa)
 		err = vfs_ioc_fssetxattr_check(inode, &old_fa, fa);
 		if (err)
 			goto out_unlock;
+
+		projid = fa->fsx_projid;
+
+		err = ubifs_ioc_setproject(file, projid);
+		if (err)
+			goto out_unlock;
 	}
 
 	ui->flags |= ubi_flags;
+	if (ui->flags & UBIFS_FLG_PROJID || projid != UBIFS_DEF_PROJID) {
+		err = ubifs_enable_projid(c);
+		if (err)
+			goto out_unlock;
+	}
 
 	ubifs_set_inode_flags(inode);
 	inode->i_ctime = current_time(inode);
@@ -233,7 +270,8 @@ static int ubifs_ioc_fsgetxattr(struct file *file, void __user *arg)
 
 static int check_xflags(unsigned int flags)
 {
-	if (flags & ~(FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))
+	if (flags & ~(FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND |
+		      FS_XFLAG_PROJINHERIT))
 		return -EOPNOTSUPP;
 	return 0;
 }
diff --git a/fs/ubifs/journal.c b/fs/ubifs/journal.c
index 03410ae0813a..2463d417bf6c 100644
--- a/fs/ubifs/journal.c
+++ b/fs/ubifs/journal.c
@@ -54,7 +54,6 @@
  */
 static inline void zero_ino_node_unused(struct ubifs_ino_node *ino)
 {
-	memset(ino->padding1, 0, 4);
 	memset(ino->padding2, 0, 26);
 }
 
@@ -469,6 +468,7 @@ static void pack_inode(struct ubifs_info *c, struct ubifs_ino_node *ino,
 	ino->xattr_cnt   = cpu_to_le32(ui->xattr_cnt);
 	ino->xattr_size  = cpu_to_le32(ui->xattr_size);
 	ino->xattr_names = cpu_to_le32(ui->xattr_names);
+	ino->projid = cpu_to_le32(from_kprojid(&init_user_ns, ui->projid));
 	zero_ino_node_unused(ino);
 
 	/*
diff --git a/fs/ubifs/sb.c b/fs/ubifs/sb.c
index 87466836fcfc..6b673db23fd4 100644
--- a/fs/ubifs/sb.c
+++ b/fs/ubifs/sb.c
@@ -957,3 +957,8 @@ int ubifs_enable_encryption(struct ubifs_info *c)
 
 	return err;
 }
+
+int ubifs_enable_projid(struct ubifs_info *c)
+{
+	return ubifs_set_feature_flag(c, UBIFS_FLG_PROJID);
+}
diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c
index 138b9426c6c1..11a43f17032e 100644
--- a/fs/ubifs/super.c
+++ b/fs/ubifs/super.c
@@ -159,6 +159,7 @@ struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
 	ui->xattr_size  = le32_to_cpu(ino->xattr_size);
 	ui->xattr_names = le32_to_cpu(ino->xattr_names);
 	ui->synced_i_size = ui->ui_size = inode->i_size;
+	ui->projid = make_kprojid(&init_user_ns, le32_to_cpu(ino->projid));
 
 	ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
 
diff --git a/fs/ubifs/ubifs-media.h b/fs/ubifs/ubifs-media.h
index 3c9792cbb6ff..59077e7fddbc 100644
--- a/fs/ubifs/ubifs-media.h
+++ b/fs/ubifs/ubifs-media.h
@@ -316,6 +316,7 @@ enum {
  * UBIFS_DIRSYNC_FL: I/O on this directory inode has to be synchronous
  * UBIFS_XATTR_FL: this inode is the inode for an extended attribute value
  * UBIFS_CRYPT_FL: use encryption for this inode
+ * UBIFS_PROJINHERIT_FL: Create with parents projid
  *
  * Note, these are on-flash flags which correspond to ioctl flags
  * (@FS_COMPR_FL, etc). They have the same values now, but generally, do not
@@ -329,6 +330,7 @@ enum {
 	UBIFS_DIRSYNC_FL   = 0x10,
 	UBIFS_XATTR_FL     = 0x20,
 	UBIFS_CRYPT_FL     = 0x40,
+	UBIFS_PROJINHERIT_FL = 0x80,
 };
 
 /* Inode flag bits used by UBIFS */
@@ -427,6 +429,7 @@ enum {
  *			  support 64bit cookies for lookups by hash
  * UBIFS_FLG_ENCRYPTION: this filesystem contains encrypted files
  * UBIFS_FLG_AUTHENTICATION: this filesystem contains hashes for authentication
+ * UBIFS_FLG_PROJID: this filesystem has project id support enabled
  */
 enum {
 	UBIFS_FLG_BIGLPT = 0x02,
@@ -434,11 +437,12 @@ enum {
 	UBIFS_FLG_DOUBLE_HASH = 0x08,
 	UBIFS_FLG_ENCRYPTION = 0x10,
 	UBIFS_FLG_AUTHENTICATION = 0x20,
+	UBIFS_FLG_PROJID = 0x40,
 };
 
 #define UBIFS_FLG_MASK (UBIFS_FLG_BIGLPT | UBIFS_FLG_SPACE_FIXUP | \
 		UBIFS_FLG_DOUBLE_HASH | UBIFS_FLG_ENCRYPTION | \
-		UBIFS_FLG_AUTHENTICATION)
+		UBIFS_FLG_AUTHENTICATION | UBIFS_FLG_PROJID)
 
 /**
  * struct ubifs_ch - common header node.
@@ -497,7 +501,7 @@ union ubifs_dev_desc {
  * @data_len: inode data length
  * @xattr_cnt: count of extended attributes this inode has
  * @xattr_size: summarized size of all extended attributes in bytes
- * @padding1: reserved for future, zeroes
+ * @projid: Quota project id
  * @xattr_names: sum of lengths of all extended attribute names belonging to
  *               this inode
  * @compr_type: compression type used for this inode
@@ -531,7 +535,7 @@ struct ubifs_ino_node {
 	__le32 data_len;
 	__le32 xattr_cnt;
 	__le32 xattr_size;
-	__u8 padding1[4]; /* Watch 'zero_ino_node_unused()' if changing! */
+	__le32 projid;
 	__le32 xattr_names;
 	__le16 compr_type;
 	__u8 padding2[26]; /* Watch 'zero_ino_node_unused()' if changing! */
diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h
index fc2cdde3b549..fc7e05718fea 100644
--- a/fs/ubifs/ubifs.h
+++ b/fs/ubifs/ubifs.h
@@ -155,6 +155,8 @@
 #define UBIFS_HMAC_ARR_SZ 0
 #endif
 
+#define UBIFS_DEF_PROJID 0
+
 /*
  * Lockdep classes for UBIFS inode @ui_mutex.
  */
@@ -362,6 +364,7 @@ struct ubifs_gced_idx_leb {
  *                 inodes
  * @ui_size: inode size used by UBIFS when writing to flash
  * @flags: inode flags (@UBIFS_COMPR_FL, etc)
+ * @projid: The project id of this inode
  * @compr_type: default compression type used for this inode
  * @last_page_read: page number of last page read (for bulk read)
  * @read_in_a_row: number of consecutive pages read in a row (for bulk read)
@@ -413,6 +416,7 @@ struct ubifs_inode {
 	loff_t synced_i_size;
 	loff_t ui_size;
 	int flags;
+	kprojid_t projid;
 	pgoff_t last_page_read;
 	pgoff_t read_in_a_row;
 	int data_len;
@@ -1902,6 +1906,7 @@ int ubifs_read_superblock(struct ubifs_info *c);
 int ubifs_write_sb_node(struct ubifs_info *c, struct ubifs_sb_node *sup);
 int ubifs_fixup_free_space(struct ubifs_info *c);
 int ubifs_enable_encryption(struct ubifs_info *c);
+int ubifs_enable_projid(struct ubifs_info *c);
 
 /* replay.c */
 int ubifs_validate_entry(struct ubifs_info *c,
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 23+ messages in thread

* [PATCH 7/8] ubifs: export get_znode
  2021-01-22 15:15 [PATCH v5 0/8] Add quota support to UBIFS Sascha Hauer
                   ` (5 preceding siblings ...)
  2021-01-22 15:15 ` [PATCH 6/8] ubifs: Add support for project id Sascha Hauer
@ 2021-01-22 15:15 ` Sascha Hauer
  2021-01-22 17:20   ` Christoph Hellwig
  2021-01-22 15:15 ` [PATCH 8/8] ubifs: Add quota support Sascha Hauer
  7 siblings, 1 reply; 23+ messages in thread
From: Sascha Hauer @ 2021-01-22 15:15 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: Richard Weinberger, linux-mtd, kernel, Jan Kara, Sascha Hauer

get_znode will be needed by upcoming UBIFS quota support. Rename it to
ubifs_get_znode and export it.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 fs/ubifs/tnc.c   | 32 ++++++++++++++++----------------
 fs/ubifs/ubifs.h |  2 ++
 2 files changed, 18 insertions(+), 16 deletions(-)

diff --git a/fs/ubifs/tnc.c b/fs/ubifs/tnc.c
index 488f3da7a6c6..b30cd47a7e4c 100644
--- a/fs/ubifs/tnc.c
+++ b/fs/ubifs/tnc.c
@@ -570,15 +570,15 @@ static int matches_name(struct ubifs_info *c, struct ubifs_zbranch *zbr,
 }
 
 /**
- * get_znode - get a TNC znode that may not be loaded yet.
+ * ubifs_get_znode - get a TNC znode that may not be loaded yet.
  * @c: UBIFS file-system description object
  * @znode: parent znode
  * @n: znode branch slot number
  *
  * This function returns the znode or a negative error code.
  */
-static struct ubifs_znode *get_znode(struct ubifs_info *c,
-				     struct ubifs_znode *znode, int n)
+struct ubifs_znode *ubifs_get_znode(struct ubifs_info *c,
+				    struct ubifs_znode *znode, int n)
 {
 	struct ubifs_zbranch *zbr;
 
@@ -618,11 +618,11 @@ static int tnc_next(struct ubifs_info *c, struct ubifs_znode **zn, int *n)
 		nn = znode->iip + 1;
 		znode = zp;
 		if (nn < znode->child_cnt) {
-			znode = get_znode(c, znode, nn);
+			znode = ubifs_get_znode(c, znode, nn);
 			if (IS_ERR(znode))
 				return PTR_ERR(znode);
 			while (znode->level != 0) {
-				znode = get_znode(c, znode, 0);
+				znode = ubifs_get_znode(c, znode, 0);
 				if (IS_ERR(znode))
 					return PTR_ERR(znode);
 			}
@@ -662,12 +662,12 @@ static int tnc_prev(struct ubifs_info *c, struct ubifs_znode **zn, int *n)
 		nn = znode->iip - 1;
 		znode = zp;
 		if (nn >= 0) {
-			znode = get_znode(c, znode, nn);
+			znode = ubifs_get_znode(c, znode, nn);
 			if (IS_ERR(znode))
 				return PTR_ERR(znode);
 			while (znode->level != 0) {
 				nn = znode->child_cnt - 1;
-				znode = get_znode(c, znode, nn);
+				znode = ubifs_get_znode(c, znode, nn);
 				if (IS_ERR(znode))
 					return PTR_ERR(znode);
 			}
@@ -2571,7 +2571,7 @@ static int tnc_delete(struct ubifs_info *c, struct ubifs_znode *znode, int n)
 		while (znode->child_cnt == 1 && znode->level != 0) {
 			zp = znode;
 			zbr = &znode->zbranch[0];
-			znode = get_znode(c, znode, 0);
+			znode = ubifs_get_znode(c, znode, 0);
 			if (IS_ERR(znode))
 				return PTR_ERR(znode);
 			znode = dirty_cow_znode(c, zbr);
@@ -3098,12 +3098,12 @@ static struct ubifs_znode *left_znode(struct ubifs_info *c,
 			return NULL;
 		if (n >= 0) {
 			/* Now go down the rightmost branch to 'level' */
-			znode = get_znode(c, znode, n);
+			znode = ubifs_get_znode(c, znode, n);
 			if (IS_ERR(znode))
 				return znode;
 			while (znode->level != level) {
 				n = znode->child_cnt - 1;
-				znode = get_znode(c, znode, n);
+				znode = ubifs_get_znode(c, znode, n);
 				if (IS_ERR(znode))
 					return znode;
 			}
@@ -3135,11 +3135,11 @@ static struct ubifs_znode *right_znode(struct ubifs_info *c,
 			return NULL;
 		if (n < znode->child_cnt) {
 			/* Now go down the leftmost branch to 'level' */
-			znode = get_znode(c, znode, n);
+			znode = ubifs_get_znode(c, znode, n);
 			if (IS_ERR(znode))
 				return znode;
 			while (znode->level != level) {
-				znode = get_znode(c, znode, 0);
+				znode = ubifs_get_znode(c, znode, 0);
 				if (IS_ERR(znode))
 					return znode;
 			}
@@ -3224,13 +3224,13 @@ static struct ubifs_znode *lookup_znode(struct ubifs_info *c,
 		}
 		if (znode->level == level + 1)
 			break;
-		znode = get_znode(c, znode, n);
+		znode = ubifs_get_znode(c, znode, n);
 		if (IS_ERR(znode))
 			return znode;
 	}
 	/* Check if the child is the one we are looking for */
 	if (znode->zbranch[n].lnum == lnum && znode->zbranch[n].offs == offs)
-		return get_znode(c, znode, n);
+		return ubifs_get_znode(c, znode, n);
 	/* If the key is unique, there is nowhere else to look */
 	if (!is_hash_key(c, key))
 		return NULL;
@@ -3256,7 +3256,7 @@ static struct ubifs_znode *lookup_znode(struct ubifs_info *c,
 		/* Check it */
 		if (znode->zbranch[n].lnum == lnum &&
 		    znode->zbranch[n].offs == offs)
-			return get_znode(c, znode, n);
+			return ubifs_get_znode(c, znode, n);
 		/* Stop if the key is less than the one we are looking for */
 		if (keys_cmp(c, &znode->zbranch[n].key, key) < 0)
 			break;
@@ -3278,7 +3278,7 @@ static struct ubifs_znode *lookup_znode(struct ubifs_info *c,
 		/* Check it */
 		if (znode->zbranch[n].lnum == lnum &&
 		    znode->zbranch[n].offs == offs)
-			return get_znode(c, znode, n);
+			return ubifs_get_znode(c, znode, n);
 		/* Stop if the key is greater than the one we are looking for */
 		if (keys_cmp(c, &znode->zbranch[n].key, key) > 0)
 			break;
diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h
index fc7e05718fea..a29e5c3dd1fb 100644
--- a/fs/ubifs/ubifs.h
+++ b/fs/ubifs/ubifs.h
@@ -1820,6 +1820,8 @@ int ubifs_find_dirty_idx_leb(struct ubifs_info *c);
 int ubifs_save_dirty_idx_lnums(struct ubifs_info *c);
 
 /* tnc.c */
+struct ubifs_znode *ubifs_get_znode(struct ubifs_info *c,
+				    struct ubifs_znode *znode, int n);
 int ubifs_lookup_level0(struct ubifs_info *c, const union ubifs_key *key,
 			struct ubifs_znode **zn, int *n);
 int ubifs_tnc_lookup_nm(struct ubifs_info *c, const union ubifs_key *key,
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 23+ messages in thread

* [PATCH 8/8] ubifs: Add quota support
  2021-01-22 15:15 [PATCH v5 0/8] Add quota support to UBIFS Sascha Hauer
                   ` (6 preceding siblings ...)
  2021-01-22 15:15 ` [PATCH 7/8] ubifs: export get_znode Sascha Hauer
@ 2021-01-22 15:15 ` Sascha Hauer
  7 siblings, 0 replies; 23+ messages in thread
From: Sascha Hauer @ 2021-01-22 15:15 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: Richard Weinberger, linux-mtd, kernel, Jan Kara, Sascha Hauer,
	Dongsheng Yang

This introduces poor man's quota support for UBIFS. Unlike other
implementations we never store anything on the flash. This has two
big advantages:

- No possible regressions with a changed on-disk format
- no quota files can get out of sync.

There are downsides as well:

- During mount the whole index must be scanned which takes some time
- The quota limits must be set manually each time a filesystem is mounted.

UBIFS is targetted for embedded systems and quota limits are likely not
changed interactively, so having to restore the quota limits with a
script shouldn't be a big deal. The mount time penalty is a price we
must pay, but for that we get a simple and straight forward
implementation for this rather rarely used feature.

The quota data itself is stored in a red-black tree in memory. It is
implemented as a quota format. When enabled with the "quota" mount
option all three quota types (user, group, project) are enabled.

The quota integration into UBIFS is taken from a series posted earlier
by Dongsheng Yang. Like the earlier series we only account regular files
for quota. All others are counted in the number of files, but do not
require any quota space.

Signed-off-by: Dongsheng Yang <yangds.fnst@cn.fujitsu.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 Documentation/filesystems/ubifs.rst |   6 +
 fs/ubifs/Makefile                   |   1 +
 fs/ubifs/dir.c                      | 103 +++--
 fs/ubifs/file.c                     |  43 ++
 fs/ubifs/ioctl.c                    |  32 ++
 fs/ubifs/journal.c                  |   2 +
 fs/ubifs/quota.c                    | 606 ++++++++++++++++++++++++++++
 fs/ubifs/super.c                    |  86 +++-
 fs/ubifs/ubifs.h                    |  37 ++
 fs/ubifs/xattr.c                    |   5 +-
 include/uapi/linux/quota.h          |   1 +
 11 files changed, 896 insertions(+), 26 deletions(-)
 create mode 100644 fs/ubifs/quota.c

diff --git a/Documentation/filesystems/ubifs.rst b/Documentation/filesystems/ubifs.rst
index e6ee99762534..999b815afb38 100644
--- a/Documentation/filesystems/ubifs.rst
+++ b/Documentation/filesystems/ubifs.rst
@@ -105,6 +105,12 @@ auth_key=		specify the key used for authenticating the filesystem.
 auth_hash_name=		The hash algorithm used for authentication. Used for
 			both hashing and for creating HMACs. Typical values
 			include "sha256" or "sha512"
+usrquota		Enable user disk quota support
+			(requires CONFIG_QUOTA).
+grpquota		Enable group disk quota support
+			(requires CONFIG_QUOTA).
+prjquota		Enable project disk quota support
+			(requires CONFIG_QUOTA).
 ====================	=======================================================
 
 
diff --git a/fs/ubifs/Makefile b/fs/ubifs/Makefile
index 5c4b845754a7..07f657f5ea8f 100644
--- a/fs/ubifs/Makefile
+++ b/fs/ubifs/Makefile
@@ -9,3 +9,4 @@ ubifs-y += misc.o
 ubifs-$(CONFIG_FS_ENCRYPTION) += crypto.o
 ubifs-$(CONFIG_UBIFS_FS_XATTR) += xattr.o
 ubifs-$(CONFIG_UBIFS_FS_AUTHENTICATION) += auth.o
+ubifs-$(CONFIG_QUOTA) += quota.o
diff --git a/fs/ubifs/dir.c b/fs/ubifs/dir.c
index e724fc91e2d1..335b5e8e139c 100644
--- a/fs/ubifs/dir.c
+++ b/fs/ubifs/dir.c
@@ -64,28 +64,13 @@ static int inherit_flags(const struct inode *dir, umode_t mode)
 	return flags;
 }
 
-/**
- * ubifs_new_inode - allocate new UBIFS inode object.
- * @c: UBIFS file-system description object
- * @dir: parent directory inode
- * @mode: inode mode flags
- *
- * This function finds an unused inode number, allocates new inode and
- * initializes it. Returns new inode in case of success and an error code in
- * case of failure.
- */
-struct inode *ubifs_new_inode(struct ubifs_info *c, struct inode *dir,
-			      umode_t mode)
+static int ubifs_init_inode(struct ubifs_info *c, struct inode *inode,
+			    struct inode *dir, umode_t mode)
 {
 	int err;
-	struct inode *inode;
 	struct ubifs_inode *ui;
 	bool encrypted = false;
 
-	inode = new_inode(c->vfs_sb);
-	if (!inode)
-		return ERR_PTR(-ENOMEM);
-
 	ui = ubifs_inode(inode);
 	/*
 	 * Set 'S_NOCMTIME' to prevent VFS form updating [mc]time of inodes and
@@ -103,7 +88,7 @@ struct inode *ubifs_new_inode(struct ubifs_info *c, struct inode *dir,
 	err = fscrypt_prepare_new_inode(dir, inode, &encrypted);
 	if (err) {
 		ubifs_err(c, "fscrypt_prepare_new_inode failed: %i", err);
-		goto out_iput;
+		return err;
 	}
 
 	if (ubifs_inode(dir)->flags & UBIFS_PROJINHERIT_FL)
@@ -111,6 +96,11 @@ struct inode *ubifs_new_inode(struct ubifs_info *c, struct inode *dir,
 	else
 		ui->projid = make_kprojid(&init_user_ns, UBIFS_DEF_PROJID);
 
+	dquot_initialize(inode);
+	err = dquot_alloc_inode(inode);
+	if (err)
+		return err;
+
 	switch (mode & S_IFMT) {
 	case S_IFREG:
 		inode->i_mapping->a_ops = &ubifs_file_address_operations;
@@ -150,7 +140,7 @@ struct inode *ubifs_new_inode(struct ubifs_info *c, struct inode *dir,
 			spin_unlock(&c->cnt_lock);
 			ubifs_err(c, "out of inode numbers");
 			err = -EINVAL;
-			goto out_iput;
+			goto out_drop;
 		}
 		ubifs_warn(c, "running out of inode numbers (current %lu, max %u)",
 			   (unsigned long)c->highest_inum, INUM_WATERMARK);
@@ -171,16 +161,79 @@ struct inode *ubifs_new_inode(struct ubifs_info *c, struct inode *dir,
 		err = fscrypt_set_context(inode, NULL);
 		if (err) {
 			ubifs_err(c, "fscrypt_set_context failed: %i", err);
-			goto out_iput;
+			goto out_bad;
 		}
 	}
 
-	return inode;
+	return 0;
 
-out_iput:
+out_bad:
 	make_bad_inode(inode);
-	iput(inode);
-	return ERR_PTR(err);
+
+	dquot_free_inode(inode);
+out_drop:
+	dquot_drop(inode);
+
+	return err;
+}
+
+/**
+ * ubifs_new_inode - allocate new UBIFS inode object.
+ * @c: UBIFS file-system description object
+ * @dir: parent directory inode
+ * @mode: inode mode flags
+ *
+ * This function finds an unused inode number, allocates new inode and
+ * initializes it. Returns new inode in case of success and an error code in
+ * case of failure.
+ */
+struct inode *ubifs_new_inode(struct ubifs_info *c, struct inode *dir,
+			      umode_t mode)
+{
+	struct inode *inode;
+	int err;
+
+	inode = new_inode(c->vfs_sb);
+	if (!inode)
+		return ERR_PTR(-ENOMEM);
+
+	err = ubifs_init_inode(c, inode, dir, mode);
+	if (err) {
+		iput(inode);
+		return ERR_PTR(err);
+	}
+
+	return inode;
+}
+
+/**
+ * ubifs_new_xattr_inode - allocate new UBIFS inode object to be used as xattr
+ * @c: UBIFS file-system description object
+ * @dir: parent directory inode
+ * @mode: inode mode flags
+ *
+ * like ubifs_new_inode, but allocates an inode without quota accounting to be
+ * used for xattrs.
+ */
+struct inode *ubifs_new_xattr_inode(struct ubifs_info *c, struct inode *dir,
+			      umode_t mode)
+{
+	struct inode *inode;
+	int err;
+
+	inode = new_inode(c->vfs_sb);
+	if (!inode)
+		return ERR_PTR(-ENOMEM);
+
+	inode->i_flags |= S_NOQUOTA;
+
+	err = ubifs_init_inode(c, inode, dir, mode);
+	if (err) {
+		iput(inode);
+		return ERR_PTR(err);
+	}
+
+	return inode;
 }
 
 static int dbg_check_name(const struct ubifs_info *c,
@@ -780,6 +833,8 @@ static int ubifs_unlink(struct inode *dir, struct dentry *dentry)
 	unsigned int saved_nlink = inode->i_nlink;
 	struct fscrypt_name nm;
 
+	dquot_initialize(inode);
+
 	/*
 	 * Budget request settings: deletion direntry, deletion inode (+1 for
 	 * @dirtied_ino), changing the parent directory inode. If budgeting
diff --git a/fs/ubifs/file.c b/fs/ubifs/file.c
index 2bc7780d2963..d74134692d75 100644
--- a/fs/ubifs/file.c
+++ b/fs/ubifs/file.c
@@ -427,8 +427,10 @@ static int ubifs_write_begin(struct file *file, struct address_space *mapping,
 	struct ubifs_inode *ui = ubifs_inode(inode);
 	pgoff_t index = pos >> PAGE_SHIFT;
 	int err, appending = !!(pos + len > inode->i_size);
+	int quota_size = 0;
 	int skipped_read = 0;
 	struct page *page;
+	int ret = 0;
 
 	ubifs_assert(c, ubifs_inode(inode)->ui_size == inode->i_size);
 	ubifs_assert(c, !c->ro_media && !c->ro_mount);
@@ -436,6 +438,16 @@ static int ubifs_write_begin(struct file *file, struct address_space *mapping,
 	if (unlikely(c->ro_error))
 		return -EROFS;
 
+	quota_size = ((pos + len) - inode->i_size);
+	if (quota_size < 0)
+		quota_size = 0;
+	if (S_ISREG(inode->i_mode)) {
+		dquot_initialize(inode);
+		ret = dquot_alloc_space_nodirty(inode, quota_size);
+		if (unlikely(ret))
+			return ret;
+	}
+
 	/* Try out the fast-path part first */
 	page = grab_cache_page_write_begin(mapping, index, flags);
 	if (unlikely(!page))
@@ -541,6 +553,7 @@ static int ubifs_write_end(struct file *file, struct address_space *mapping,
 	struct ubifs_inode *ui = ubifs_inode(inode);
 	struct ubifs_info *c = inode->i_sb->s_fs_info;
 	loff_t end_pos = pos + len;
+	int quota_size = 0;
 	int appending = !!(end_pos > inode->i_size);
 
 	dbg_gen("ino %lu, pos %llu, pg %lu, len %u, copied %d, i_size %lld",
@@ -559,6 +572,11 @@ static int ubifs_write_end(struct file *file, struct address_space *mapping,
 		dbg_gen("copied %d instead of %d, read page and repeat",
 			copied, len);
 		cancel_budget(c, page, ui, appending);
+		quota_size = ((pos + len) - inode->i_size);
+		if (quota_size < 0)
+			quota_size = 0;
+		if (S_ISREG(inode->i_mode))
+			dquot_free_space_nodirty(inode, quota_size);
 		ClearPageChecked(page);
 
 		/*
@@ -1111,6 +1129,7 @@ static int do_truncation(struct ubifs_info *c, struct inode *inode,
 	int err;
 	struct ubifs_budget_req req;
 	loff_t old_size = inode->i_size, new_size = attr->ia_size;
+	loff_t quota_size = (old_size - new_size);
 	int offset = new_size & (UBIFS_BLOCK_SIZE - 1), budgeted = 1;
 	struct ubifs_inode *ui = ubifs_inode(inode);
 
@@ -1190,6 +1209,10 @@ static int do_truncation(struct ubifs_info *c, struct inode *inode,
 	do_attr_changes(inode, attr);
 	err = ubifs_jnl_truncate(c, inode, old_size, new_size);
 	mutex_unlock(&ui->ui_mutex);
+	if (quota_size < 0)
+		quota_size = 0;
+	if (S_ISREG(inode->i_mode))
+		dquot_free_space_nodirty(inode, quota_size);
 
 out_budg:
 	if (budgeted)
@@ -1226,6 +1249,17 @@ static int do_setattr(struct ubifs_info *c, struct inode *inode,
 
 	if (attr->ia_valid & ATTR_SIZE) {
 		dbg_gen("size %lld -> %lld", inode->i_size, new_size);
+		if (S_ISREG(inode->i_mode)) {
+			if (new_size > inode->i_size) {
+				err = dquot_alloc_space_nodirty(inode, new_size - inode->i_size);
+				if (err) {
+					ubifs_release_budget(c, &req);
+					return err;
+				}
+			} else {
+				dquot_free_space_nodirty(inode, inode->i_size - new_size);
+			}
+		}
 		truncate_setsize(inode, new_size);
 	}
 
@@ -1277,6 +1311,15 @@ int ubifs_setattr(struct dentry *dentry, struct iattr *attr)
 	if (err)
 		return err;
 
+	if (is_quota_modification(inode, attr))
+		dquot_initialize(inode);
+	if ((attr->ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)) ||
+	    (attr->ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid))) {
+		err = dquot_transfer(inode, attr);
+		if (err)
+			return err;
+	}
+
 	if ((attr->ia_valid & ATTR_SIZE) && attr->ia_size < inode->i_size)
 		/* Truncation to a smaller size */
 		err = do_truncation(c, inode, attr);
diff --git a/fs/ubifs/ioctl.c b/fs/ubifs/ioctl.c
index a9a79935cc13..f63af4c2e194 100644
--- a/fs/ubifs/ioctl.c
+++ b/fs/ubifs/ioctl.c
@@ -139,16 +139,48 @@ static __u32 ubifs_iflags_to_xflags(unsigned long flags)
         return xflags;
 }
 
+#ifdef CONFIG_QUOTA
+static int ubifs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
+{
+	struct dquot *transfer_to[MAXQUOTAS] = {};
+	struct ubifs_info *c = inode->i_sb->s_fs_info;
+	struct super_block *sb = c->vfs_sb;
+	int err = 0;
+
+	err = dquot_initialize(inode);
+	if (err)
+		return err;
+
+	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
+	if (!IS_ERR(transfer_to[PRJQUOTA])) {
+		err = __dquot_transfer(inode, transfer_to);
+		dqput(transfer_to[PRJQUOTA]);
+	}
+
+	return err;
+}
+#else
+static int ubifs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
+{
+	return 0;
+}
+#endif
+
 static int ubifs_ioc_setproject(struct file *file, __u32 projid)
 {
 	struct inode *inode = file_inode(file);
 	struct ubifs_inode *ui = ubifs_inode(inode);
 	kprojid_t kprojid;
+	int err;
 
 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
 	if (projid_eq(kprojid, ui->projid))
 		return 0;
 
+	err = ubifs_transfer_project_quota(inode, kprojid);
+	if (err)
+		return err;
+
 	ui->projid = kprojid;
 
 	return 0;
diff --git a/fs/ubifs/journal.c b/fs/ubifs/journal.c
index 2463d417bf6c..b16855259977 100644
--- a/fs/ubifs/journal.c
+++ b/fs/ubifs/journal.c
@@ -631,6 +631,8 @@ int ubifs_jnl_update(struct ubifs_info *c, const struct inode *dir,
 		}
 		ui->del_cmtno = c->cmt_no;
 		orphan_added = 1;
+		if (S_ISREG(inode->i_mode))
+			dquot_free_space_nodirty((struct inode *)inode, inode->i_size);
 	}
 
 	err = write_head(c, BASEHD, dent, len, &lnum, &dent_offs, sync);
diff --git a/fs/ubifs/quota.c b/fs/ubifs/quota.c
new file mode 100644
index 000000000000..bebefb846b1b
--- /dev/null
+++ b/fs/ubifs/quota.c
@@ -0,0 +1,606 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * This file is part of UBIFS.
+ *
+ * Copyright (C) 2019 Pengutronix, Sascha Hauer <s.hauer@pengutronix.de>
+ */
+#include "ubifs.h"
+
+struct ubifs_dqblk {
+	struct mem_dqblk dqblk;
+	struct rb_node rb;
+	struct kqid kqid;
+};
+
+/**
+ * ubifs_dqblk_insert - find qid in tree
+ * @c: UBIFS file-system description object
+ * @new: The new entry to insert
+ *
+ * This inserts a new entry to the dqblk tree which is sorted by qids. Caller
+ * must make sure this entry doesn't exist already.
+ */
+static void ubifs_dqblk_insert(struct ubifs_info *c, struct ubifs_dqblk *new)
+{
+	struct rb_root *root = &c->dqblk_tree[new->kqid.type];
+	struct rb_node **link = &root->rb_node, *parent = NULL;
+	struct ubifs_dqblk *ud;
+
+	/* Go to the bottom of the tree */
+	while (*link) {
+		parent = *link;
+
+		ud = rb_entry(parent, struct ubifs_dqblk, rb);
+
+		if (qid_lt(new->kqid, ud->kqid))
+			link = &(*link)->rb_left;
+		else
+			link = &(*link)->rb_right;
+	}
+
+	/* Put the new node there */
+	rb_link_node(&new->rb, parent, link);
+	rb_insert_color(&new->rb, root);
+}
+
+/**
+ * ubifs_dqblk_find_next - find the next qid
+ * @c: UBIFS file-system description object
+ * @qid: The qid to look for
+ *
+ * Find the next dqblk entry with a qid that is bigger or equally big than the
+ * given qid. Returns the next dqblk entry if found or NULL if no dqblk exists
+ * with a qid that is at least equally big.
+ */
+static struct ubifs_dqblk *ubifs_dqblk_find_next(struct ubifs_info *c,
+						 struct kqid qid)
+{
+	struct rb_node *node = c->dqblk_tree[qid.type].rb_node;
+	struct ubifs_dqblk *next = NULL;
+
+	while (node) {
+		struct ubifs_dqblk *ud = rb_entry(node, struct ubifs_dqblk, rb);
+
+		if (qid_eq(qid, ud->kqid))
+			return ud;
+
+		if (qid_lt(qid, ud->kqid)) {
+			next = ud;
+			node = node->rb_left;
+		} else {
+			node = node->rb_right;
+		}
+	}
+
+	return next;
+}
+
+/**
+ * ubifs_dqblk_find - find qid in tree
+ * @c: UBIFS file-system description object
+ * @qid: The qid to look for
+ *
+ * This walks the dqblk tree and searches a given qid. Returns the dqblk entry
+ * when found or NULL otherwise.
+ */
+static struct ubifs_dqblk *ubifs_dqblk_find(struct ubifs_info *c,
+					    struct kqid qid)
+{
+	struct ubifs_dqblk *next;
+
+	next = ubifs_dqblk_find_next(c, qid);
+
+	if (next && qid_eq(qid, next->kqid))
+		return next;
+
+	return NULL;
+}
+
+/**
+ * ubifs_dqblk_get - get dqblk entry for given qid
+ * @c: UBIFS file-system description object
+ * @qid: The qid to look for
+ *
+ * This searches the given qid in the dqblk tree and returns it if found. If not,
+ * a new dqblk entry is created, inserted into the dqblk tree and returned.
+ */
+static struct ubifs_dqblk *ubifs_dqblk_get(struct ubifs_info *c,
+					   struct kqid qid)
+{
+	struct ubifs_dqblk *ud;
+
+	ud = ubifs_dqblk_find(c, qid);
+	if (ud)
+		return ud;
+
+	ud = kzalloc(sizeof(*ud), GFP_KERNEL);
+	if (!ud)
+		return NULL;
+
+	ud->kqid = qid;
+
+	ubifs_dqblk_insert(c, ud);
+
+	return ud;
+}
+
+/**
+ * ubifs_check_quota_file - check if quota file is valid for this format
+ * @sb: The superblock
+ * @type: Quota type
+ *
+ * We currently do not store any quota file on flash. It's completely in RAM and
+ * thus always valid.
+ */
+static int ubifs_check_quota_file(struct super_block *sb, int type)
+{
+	return 1;
+}
+
+/**
+ * ubifs_read_file_info - read quota file info
+ * @sb: The superblock
+ * @type: Quota type
+ *
+ * We currently do not store any quota info on flash. Just fill in default
+ * values.
+ */
+static int ubifs_read_file_info(struct super_block *sb, int type)
+{
+	struct quota_info *dqopt = sb_dqopt(sb);
+	struct mem_dqinfo *info = &dqopt->info[type];
+
+	down_read(&dqopt->dqio_sem);
+
+	/*
+	 * Used space is stored as unsigned 64-bit value in bytes but
+	 * quota core supports only signed 64-bit values so use that
+	 * as a limit
+	 */
+	info->dqi_max_spc_limit = 0x7fffffffffffffffLL; /* 2^63-1 */
+	info->dqi_max_ino_limit = 0x7fffffffffffffffLL;
+	info->dqi_bgrace = 0;
+	info->dqi_igrace = 0;
+	info->dqi_flags = 0;
+
+	up_read(&dqopt->dqio_sem);
+
+	return 0;
+}
+
+/**
+ * ubifs_write_file_info - write quota file to device
+ * @sb: The superblock
+ * @type: Quota type
+ *
+ * We currently do not store any quota file on flash. Nothing to do here.
+ */
+static int ubifs_write_file_info(struct super_block *sb, int type)
+{
+	return 0;
+}
+
+/**
+ * ubifs_read_dquot - read dquot from device
+ * @dquot: The dquot entry to read
+ *
+ * For us this means finding the entry in the dquot tree.
+ */
+static int ubifs_read_dquot(struct dquot *dquot)
+{
+	struct quota_info *dqopt = sb_dqopt(dquot->dq_sb);
+	struct ubifs_info *c = dquot->dq_sb->s_fs_info;
+	struct ubifs_dqblk *ud;
+	int ret = 0;
+
+	down_read(&dqopt->dqio_sem);
+
+	ud = ubifs_dqblk_find(c, dquot->dq_id);
+	if (ud)
+		dquot->dq_dqb = ud->dqblk;
+	else
+		ret = -ENOENT;
+
+	up_read(&dqopt->dqio_sem);
+
+	return 0;
+}
+
+/**
+ * ubifs_write_dquot - write dquot to device
+ * @dquot: The dquot entry to read
+ *
+ * For us this means finding or creating the entry in the dquot tree and setting
+ * it to the dquots contents.
+ */
+static int ubifs_write_dquot(struct dquot *dquot)
+{
+	struct quota_info *dqopt = sb_dqopt(dquot->dq_sb);
+	struct ubifs_info *c = dquot->dq_sb->s_fs_info;
+	int ret = 0;
+	struct ubifs_dqblk *ud;
+
+	down_write(&dqopt->dqio_sem);
+
+	ud = ubifs_dqblk_get(c, dquot->dq_id);
+	if (!ud) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	ud->dqblk = dquot->dq_dqb;
+
+out:
+	up_write(&dqopt->dqio_sem);
+
+	return ret;
+}
+
+/**
+ * ubifs_release_dquot - release memory associated to a dquot entry
+ * @dquot: The dquot entry
+ *
+ * Nothing to do here, we didn't allocate anything. Our data is freed at unmount
+ * time.
+ */
+static int ubifs_release_dquot(struct dquot *dquot)
+{
+	return 0;
+}
+
+/**
+ * ubifs_free_file_info - free memory allocated during reading the file info
+ * @sb: The superblock
+ * @type: Quota type
+ *
+ * Nothing to do here.
+ */
+static int ubifs_free_file_info(struct super_block *sb, int type)
+{
+	return 0;
+}
+
+static int ubifs_get_next_id(struct super_block *sb, struct kqid *qid)
+{
+	struct quota_info *dqopt = sb_dqopt(sb);
+	struct ubifs_info *c = sb->s_fs_info;
+	struct ubifs_dqblk *ud;
+	int ret = 0;
+
+	down_read(&dqopt->dqio_sem);
+
+	ud = ubifs_dqblk_find_next(c, *qid);
+	if (!ud) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	*qid = ud->kqid;
+out:
+	up_read(&dqopt->dqio_sem);
+
+	return ret;
+}
+
+static const struct quota_format_ops ubifs_format_ops = {
+	.check_quota_file	= ubifs_check_quota_file,
+	.read_file_info		= ubifs_read_file_info,
+	.write_file_info	= ubifs_write_file_info,
+	.free_file_info		= ubifs_free_file_info,
+	.read_dqblk		= ubifs_read_dquot,
+	.commit_dqblk		= ubifs_write_dquot,
+	.release_dqblk		= ubifs_release_dquot,
+	.get_next_id		= ubifs_get_next_id,
+};
+
+/**
+ * add_inode_quota - read quota informations for an inode
+ * @c: UBIFS file-system description object
+ * @zbr: The zbranch to read the inode from
+ *
+ * This reads the inode as referred to by @zbr from the medium and
+ * extracts the quota relevant informations. Returns 0 for success
+ * or a negative error code otherwise.
+ */
+static int add_inode_quota(struct ubifs_info *c, struct ubifs_zbranch *zbr)
+{
+	struct ubifs_ino_node *ino;
+	int err, type = key_type(c, &zbr->key);
+	struct kqid qid[MAXQUOTAS];
+	kprojid_t kprojid;
+	kgid_t kgid;
+	kuid_t kuid;
+	int i;
+
+	if (type != UBIFS_INO_KEY) {
+		ubifs_err(c, "%s called on non INO node %d", __func__, type);
+		return 0;
+	}
+
+	if (zbr->len < UBIFS_CH_SZ) {
+		ubifs_err(c, "bad leaf length %d (LEB %d:%d)",
+				zbr->len, zbr->lnum, zbr->offs);
+		return -EINVAL;
+	}
+
+	ino = kmalloc(zbr->len, GFP_NOFS);
+	if (!ino)
+		return -ENOMEM;
+
+	err = ubifs_tnc_read_node(c, zbr, ino);
+	if (err) {
+		ubifs_err(c, "cannot read leaf node at LEB %d:%d, error %d",
+				zbr->lnum, zbr->offs, err);
+		goto out_free;
+	}
+
+	if (ino->flags & UBIFS_XATTR_FL)
+		goto out_free;
+
+	kuid = make_kuid(&init_user_ns, le32_to_cpu(ino->uid));
+	qid[USRQUOTA] = make_kqid_uid(kuid);
+
+	kgid = make_kgid(&init_user_ns, le32_to_cpu(ino->gid));
+	qid[GRPQUOTA] = make_kqid_gid(kgid);
+
+	kprojid = make_kprojid(&init_user_ns, le32_to_cpu(ino->projid));
+	qid[PRJQUOTA] = make_kqid_projid(kprojid);
+
+	for (i = 0; i < UBIFS_MAXQUOTAS; i++) {
+		struct ubifs_dqblk *ud;
+
+		if (!(c->quota_enable & (1 << type)))
+			continue;
+
+		ud = ubifs_dqblk_get(c, qid[i]);
+		if (!ud) {
+			err = -ENOMEM;
+			goto out_free;
+		}
+
+		if (S_ISREG(le32_to_cpu(ino->mode)))
+			ud->dqblk.dqb_curspace += le64_to_cpu(ino->size);
+		ud->dqblk.dqb_curinodes++;
+	}
+
+	err = 0;
+
+out_free:
+	kfree(ino);
+
+	return err;
+}
+
+/**
+ * get_next_parent_inum - get the next inum
+ * @c: UBIFS file-system description object
+ * @znode: The znode to start searching from
+ * @inum: The next inum returned here
+ *
+ * Helper function for ubifs_quota_walk_index(). If we know the next inum we
+ * find up the tree is still the same as we just handled we do not have to walk
+ * down the tree but can skip these branches. Returns 0 for success or -ENOENT
+ * when we can't go further to the upper-right but hit the root node instead.
+ */
+static int get_next_parent_inum(struct ubifs_info *c, struct ubifs_znode *znode,
+				ino_t *inum)
+{
+	struct ubifs_znode *zp;
+
+	while (1) {
+		int nn;
+
+		zp = znode->parent;
+		if (!zp)
+			return -ENOENT;
+
+		nn = znode->iip + 1;
+		znode = zp;
+
+		if (nn < znode->child_cnt) {
+			*inum = key_inum(c, &znode->zbranch[nn].key);
+			return 0;
+		}
+	}
+}
+
+/**
+ * ubifs_quota_walk_index - Collect quota info
+ * @c: UBIFS file-system description object
+ *
+ * This function walks the whole index and collects the quota usage information
+ * for all inodes. Returns 0 for success or a negative error code otherwise.
+ */
+static int ubifs_quota_walk_index(struct ubifs_info *c)
+{
+	int err;
+	struct ubifs_znode *znode;
+	int i, nn = 0;
+	ino_t inum = 0, pinum;
+
+	if (!c->zroot.znode) {
+		c->zroot.znode = ubifs_load_znode(c, &c->zroot, NULL, 0);
+		if (IS_ERR(c->zroot.znode)) {
+			err = PTR_ERR(c->zroot.znode);
+			c->zroot.znode = NULL;
+			return err;
+		}
+	}
+
+	znode = c->zroot.znode;
+
+	while (1) {
+		while (znode->level != 0) {
+			znode = ubifs_get_znode(c, znode, nn);
+			if (IS_ERR(znode))
+				return PTR_ERR(znode);
+			nn = 0;
+		}
+
+		for (i = 0; i < znode->child_cnt; i++) {
+			if (inum == key_inum(c, &znode->zbranch[i].key))
+				continue;
+
+			add_inode_quota(c, &znode->zbranch[i]);
+			inum = key_inum(c, &znode->zbranch[i].key);
+		}
+
+		while (1) {
+			struct ubifs_znode *zp;
+
+			zp = znode->parent;
+			if (!zp)
+				return 0;
+
+			nn = znode->iip + 1;
+			znode = zp;
+
+			err = get_next_parent_inum(c, znode, &pinum);
+
+			/*
+			 * Optimization: When the next parent node is still for
+			 * the inode we just handled we can skip parsing our
+			 * children.
+			 */
+			if (!err && pinum == inum)
+				continue;
+
+			/*
+			 * Optimization: We can skip scanning all child nodes
+			 * which are for the same inode as we already looked at.
+			 */
+			while (nn < znode->child_cnt - 1 &&
+			       inum == key_inum(c, &znode->zbranch[nn + 1].key))
+				nn++;
+
+			if (nn < znode->child_cnt)
+				break;
+		}
+	}
+}
+
+/**
+ * ubifs_enable_quotas - enable quota
+ * @c: UBIFS file-system description object
+ *
+ * Enable usage tracking for all quota types.
+ */
+int ubifs_enable_quotas(struct ubifs_info *c)
+{
+	struct super_block *sb = c->vfs_sb;
+	struct quota_info *dqopt = sb_dqopt(sb);
+	int type, ret;
+
+	if (!c->quota_enable)
+		return 0;
+
+	dqopt->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
+
+	for (type = 0; type < UBIFS_MAXQUOTAS; type++) {
+		if (!(c->quota_enable & (1 << type)))
+			continue;
+
+		ret = dquot_load_quota_sb(sb, type, QFMT_UBIFS,
+				   DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
+		if (ret)
+			goto out_quota_off;
+
+		c->dqblk_tree[type] = RB_ROOT;
+	}
+
+	return ubifs_quota_walk_index(c);
+
+out_quota_off:
+	ubifs_disable_quotas(c);
+
+	return ret;
+}
+
+/**
+ * ubifs_disable_quotas - disable quota
+ * @c: UBIFS file-system description object
+ *
+ * Disable quota for UBFIS.
+ */
+void ubifs_disable_quotas(struct ubifs_info *c)
+{
+	struct rb_node *n;
+	struct ubifs_dqblk *ud;
+	int type;
+
+	if (!c->quota_enable)
+		return;
+
+	dquot_disable(c->vfs_sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
+
+	for (type = 0; type < UBIFS_MAXQUOTAS; type++) {
+		while ((n = rb_first(&c->dqblk_tree[type]))) {
+			ud = rb_entry(n, struct ubifs_dqblk, rb);
+			rb_erase(n, &c->dqblk_tree[type]);
+			kfree(ud);
+		}
+	}
+}
+
+static int ubifs_get_projid(struct inode *inode, kprojid_t *projid)
+{
+	struct ubifs_inode *ui = ubifs_inode(inode);
+
+	*projid = ui->projid;
+
+	return 0;
+}
+
+static const struct dquot_operations ubifs_dquot_operations = {
+	.write_dquot	= dquot_commit,
+	.acquire_dquot	= dquot_acquire,
+	.release_dquot	= dquot_release,
+	.mark_dirty	= dquot_mark_dquot_dirty,
+	.write_info	= dquot_commit_info,
+	.alloc_dquot	= dquot_alloc,
+	.destroy_dquot	= dquot_destroy,
+	.get_next_id	= dquot_get_next_id,
+	.get_projid	= ubifs_get_projid,
+};
+
+static const struct quotactl_ops ubifs_quotactl_ops = {
+	.get_state	= dquot_get_state,
+	.set_info	= dquot_set_dqinfo,
+	.get_dqblk	= dquot_get_dqblk,
+	.set_dqblk	= dquot_set_dqblk,
+	.get_nextdqblk	= dquot_get_next_dqblk,
+};
+
+static struct quota_format_type ubifs_quota_format = {
+	.qf_fmt_id	= QFMT_UBIFS,
+	.qf_ops		= &ubifs_format_ops,
+	.qf_owner	= THIS_MODULE
+};
+
+int ubifs_register_quota_format(void)
+{
+	return register_quota_format(&ubifs_quota_format);
+}
+
+void ubifs_unregister_quota_format(void)
+{
+	return unregister_quota_format(&ubifs_quota_format);
+}
+
+/**
+ * ubifs_init_quota - init quota ops for UBIFS
+ * @c: UBIFS file-system description object
+ *
+ * This inits the quota operations for UBIFS
+ */
+void ubifs_init_quota(struct ubifs_info *c)
+{
+	struct super_block *sb = c->vfs_sb;
+
+	if (!c->quota_enable)
+		return;
+
+	sb->dq_op = &ubifs_dquot_operations;
+	sb->s_qcop = &ubifs_quotactl_ops;
+	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
+}
diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c
index 11a43f17032e..b3be9b947cfa 100644
--- a/fs/ubifs/super.c
+++ b/fs/ubifs/super.c
@@ -169,6 +169,7 @@ struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
 
 	switch (inode->i_mode & S_IFMT) {
 	case S_IFREG:
+		inode_set_bytes(inode, ino->size);
 		inode->i_mapping->a_ops = &ubifs_file_address_operations;
 		inode->i_op = &ubifs_file_inode_operations;
 		inode->i_fop = &ubifs_file_operations;
@@ -372,6 +373,7 @@ static void ubifs_evict_inode(struct inode *inode)
 	if (is_bad_inode(inode))
 		goto out;
 
+	dquot_initialize(inode);
 	ui->ui_size = inode->i_size = 0;
 	err = ubifs_jnl_delete_inode(c, inode);
 	if (err)
@@ -381,7 +383,7 @@ static void ubifs_evict_inode(struct inode *inode)
 		 */
 		ubifs_err(c, "can't delete inode %lu, error %d",
 			  inode->i_ino, err);
-
+	dquot_free_inode(inode);
 out:
 	if (ui->dirty)
 		ubifs_release_dirty_inode_budget(c, ui);
@@ -391,6 +393,7 @@ static void ubifs_evict_inode(struct inode *inode)
 		smp_wmb();
 	}
 done:
+	dquot_drop(inode);
 	clear_inode(inode);
 	fscrypt_put_encryption_info(inode);
 }
@@ -453,6 +456,13 @@ static int ubifs_show_options(struct seq_file *s, struct dentry *root)
 	else if (c->mount_opts.chk_data_crc == 1)
 		seq_puts(s, ",no_chk_data_crc");
 
+	if (c->quota_enable & (1 << USRQUOTA))
+		seq_puts(s, ",usrquota");
+	if (c->quota_enable & (1 << GRPQUOTA))
+		seq_puts(s, ",grpquota");
+	if (c->quota_enable & (1 << PRJQUOTA))
+		seq_puts(s, ",prjquota");
+
 	if (c->mount_opts.override_compr) {
 		seq_printf(s, ",compr=%s",
 			   ubifs_compr_name(c, c->mount_opts.compr_type));
@@ -945,6 +955,29 @@ static int check_volume_empty(struct ubifs_info *c)
 	return 0;
 }
 
+#ifdef CONFIG_QUOTA
+static struct dquot **ubifs_get_dquots(struct inode *inode)
+{
+	return ubifs_inode(inode)->i_dquot;
+}
+
+static ssize_t ubifs_quota_read(struct super_block *sb, int type, char *data,
+				size_t len, loff_t off)
+{
+	BUG();
+
+	return 0;
+}
+
+static ssize_t ubifs_quota_write(struct super_block *sb, int type, const char *data,
+				 size_t len, loff_t off)
+{
+	BUG();
+
+	return 0;
+}
+#endif
+
 /*
  * UBIFS mount options.
  *
@@ -968,6 +1001,9 @@ enum {
 	Opt_chk_data_crc,
 	Opt_no_chk_data_crc,
 	Opt_override_compr,
+	Opt_usrquota,
+	Opt_grpquota,
+	Opt_prjquota,
 	Opt_assert,
 	Opt_auth_key,
 	Opt_auth_hash_name,
@@ -983,6 +1019,9 @@ static const match_table_t tokens = {
 	{Opt_chk_data_crc, "chk_data_crc"},
 	{Opt_no_chk_data_crc, "no_chk_data_crc"},
 	{Opt_override_compr, "compr=%s"},
+	{Opt_usrquota, "usrquota"},
+	{Opt_grpquota, "grpquota"},
+	{Opt_prjquota, "prjquota"},
 	{Opt_auth_key, "auth_key=%s"},
 	{Opt_auth_hash_name, "auth_hash_name=%s"},
 	{Opt_ignore, "ubi=%s"},
@@ -1110,6 +1149,23 @@ static int ubifs_parse_options(struct ubifs_info *c, char *options,
 			kfree(act);
 			break;
 		}
+#ifdef CONFIG_QUOTA
+		case Opt_usrquota:
+			c->quota_enable |= 1 << USRQUOTA;
+			break;
+		case Opt_grpquota:
+			c->quota_enable |= 1 << GRPQUOTA;
+			break;
+		case Opt_prjquota:
+			c->quota_enable |= 1 << PRJQUOTA;
+			break;
+#else
+		case Opt_usrquota:
+		case Opt_grpquota:
+		case Opt_prjquota:
+			ubifs_err(c, "quota operations not supported");
+			break;
+#endif
 		case Opt_auth_key:
 			if (!is_remount) {
 				c->auth_key_name = kstrdup(args[0].from,
@@ -1839,6 +1895,7 @@ static int ubifs_remount_rw(struct ubifs_info *c)
 		 */
 		err = dbg_check_space_info(c);
 	}
+	c->vfs_sb->s_flags &= ~SB_RDONLY;
 
 	mutex_unlock(&c->umount_mutex);
 	return err;
@@ -1908,6 +1965,7 @@ static void ubifs_remount_ro(struct ubifs_info *c)
 	err = dbg_check_space_info(c);
 	if (err)
 		ubifs_ro_mode(c, err);
+	c->vfs_sb->s_flags |= SB_RDONLY;
 	mutex_unlock(&c->umount_mutex);
 }
 
@@ -1918,6 +1976,8 @@ static void ubifs_put_super(struct super_block *sb)
 
 	ubifs_msg(c, "un-mount UBI device %d", c->vi.ubi_num);
 
+	ubifs_disable_quotas(c);
+
 	/*
 	 * The following asserts are only valid if there has not been a failure
 	 * of the media. For example, there will be dirty inodes if we failed
@@ -2015,11 +2075,17 @@ static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
 		err = ubifs_remount_rw(c);
 		if (err)
 			return err;
+		err = dquot_resume(sb, -1);
+		if (err)
+			return err;
 	} else if (!c->ro_mount && (*flags & SB_RDONLY)) {
 		if (c->ro_error) {
 			ubifs_msg(c, "cannot re-mount R/O due to prior errors");
 			return -EROFS;
 		}
+		err = dquot_suspend(sb, -1);
+		if (err)
+			return err;
 		ubifs_remount_ro(c);
 	}
 
@@ -2051,6 +2117,11 @@ const struct super_operations ubifs_super_operations = {
 	.remount_fs    = ubifs_remount_fs,
 	.show_options  = ubifs_show_options,
 	.sync_fs       = ubifs_sync_fs,
+#ifdef CONFIG_QUOTA
+	.get_dquots    = ubifs_get_dquots,
+	.quota_read    = ubifs_quota_read,
+	.quota_write   = ubifs_quota_write,
+#endif
 };
 
 /**
@@ -2211,6 +2282,8 @@ static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
 	sb->s_xattr = ubifs_xattr_handlers;
 	fscrypt_set_ops(sb, &ubifs_crypt_operations);
 
+	ubifs_init_quota(c);
+
 	mutex_lock(&c->umount_mutex);
 	err = mount_ubifs(c);
 	if (err) {
@@ -2225,6 +2298,10 @@ static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
 		goto out_umount;
 	}
 
+	err = ubifs_enable_quotas(c);
+	if (err)
+		goto out_umount;
+
 	sb->s_root = d_make_root(root);
 	if (!sb->s_root) {
 		err = -ENOMEM;
@@ -2432,6 +2509,10 @@ static int __init ubifs_init(void)
 
 	dbg_debugfs_init();
 
+	err = ubifs_register_quota_format();
+	if (err)
+		goto out_compressors;
+
 	err = register_filesystem(&ubifs_fs_type);
 	if (err) {
 		pr_err("UBIFS error (pid %d): cannot register file system, error %d",
@@ -2442,6 +2523,8 @@ static int __init ubifs_init(void)
 
 out_dbg:
 	dbg_debugfs_exit();
+	ubifs_unregister_quota_format();
+out_compressors:
 	ubifs_compressors_exit();
 out_shrinker:
 	unregister_shrinker(&ubifs_shrinker_info);
@@ -2467,6 +2550,7 @@ static void __exit ubifs_exit(void)
 	 */
 	rcu_barrier();
 	kmem_cache_destroy(ubifs_inode_slab);
+	ubifs_unregister_quota_format();
 	unregister_filesystem(&ubifs_fs_type);
 }
 module_exit(ubifs_exit);
diff --git a/fs/ubifs/ubifs.h b/fs/ubifs/ubifs.h
index a29e5c3dd1fb..aefb30fa1d27 100644
--- a/fs/ubifs/ubifs.h
+++ b/fs/ubifs/ubifs.h
@@ -27,6 +27,7 @@
 #include <linux/security.h>
 #include <linux/xattr.h>
 #include <linux/random.h>
+#include <linux/quotaops.h>
 #include <crypto/hash_info.h>
 #include <crypto/hash.h>
 #include <crypto/algapi.h>
@@ -157,6 +158,8 @@
 
 #define UBIFS_DEF_PROJID 0
 
+#define UBIFS_MAXQUOTAS 3
+
 /*
  * Lockdep classes for UBIFS inode @ui_mutex.
  */
@@ -416,6 +419,9 @@ struct ubifs_inode {
 	loff_t synced_i_size;
 	loff_t ui_size;
 	int flags;
+#ifdef CONFIG_QUOTA
+	struct dquot *i_dquot[UBIFS_MAXQUOTAS];
+#endif
 	kprojid_t projid;
 	pgoff_t last_page_read;
 	pgoff_t read_in_a_row;
@@ -1045,6 +1051,7 @@ struct ubifs_debug_info;
  * @rw_incompat: the media is not R/W compatible
  * @assert_action: action to take when a ubifs_assert() fails
  * @authenticated: flag indigating the FS is mounted in authenticated mode
+ * @quota_enable: If true, quota is enabled on this filesystem
  *
  * @tnc_mutex: protects the Tree Node Cache (TNC), @zroot, @cnext, @enext, and
  *             @calc_idx_sz
@@ -1386,6 +1393,9 @@ struct ubifs_info {
 	struct ubi_device_info di;
 	struct ubi_volume_info vi;
 
+	unsigned int quota_enable;
+	struct rb_root dqblk_tree[UBIFS_MAXQUOTAS];
+
 	struct rb_root orph_tree;
 	struct list_head orph_list;
 	struct list_head orph_new;
@@ -2105,6 +2115,33 @@ int ubifs_decrypt(const struct inode *inode, struct ubifs_data_node *dn,
 
 extern const struct fscrypt_operations ubifs_crypt_operations;
 
+/* quota.c */
+#ifdef CONFIG_QUOTA
+int ubifs_register_quota_format(void);
+void ubifs_unregister_quota_format(void);
+void ubifs_init_quota(struct ubifs_info *c);
+int ubifs_enable_quotas(struct ubifs_info *c);
+void ubifs_disable_quotas(struct ubifs_info *c);
+#else
+static inline int ubifs_register_quota_format(void)
+{
+	return 0;
+}
+static inline void ubifs_unregister_quota_format(void)
+{
+}
+static inline void ubifs_init_quota(struct ubifs_info *c)
+{
+}
+static inline int ubifs_enable_quotas(struct ubifs_info *c)
+{
+	return 0;
+}
+static inline void ubifs_disable_quotas(struct ubifs_info *c)
+{
+}
+#endif
+
 /* Normal UBIFS messages */
 __printf(2, 3)
 void ubifs_msg(const struct ubifs_info *c, const char *fmt, ...);
diff --git a/fs/ubifs/xattr.c b/fs/ubifs/xattr.c
index a0b9b349efe6..54cfe28c25b0 100644
--- a/fs/ubifs/xattr.c
+++ b/fs/ubifs/xattr.c
@@ -64,6 +64,9 @@ enum {
 static const struct inode_operations empty_iops;
 static const struct file_operations empty_fops;
 
+struct inode *ubifs_new_xattr_inode(struct ubifs_info *c, struct inode *dir,
+			      umode_t mode);
+
 /**
  * create_xattr - create an extended attribute.
  * @c: UBIFS file-system description object
@@ -110,7 +113,7 @@ static int create_xattr(struct ubifs_info *c, struct inode *host,
 	if (err)
 		return err;
 
-	inode = ubifs_new_inode(c, host, S_IFREG | S_IRWXUGO);
+	inode = ubifs_new_xattr_inode(c, host, S_IFREG | S_IRWXUGO);
 	if (IS_ERR(inode)) {
 		err = PTR_ERR(inode);
 		goto out_budg;
diff --git a/include/uapi/linux/quota.h b/include/uapi/linux/quota.h
index e1787c0df601..7082868ca225 100644
--- a/include/uapi/linux/quota.h
+++ b/include/uapi/linux/quota.h
@@ -78,6 +78,7 @@
 #define	QFMT_VFS_V0 2
 #define QFMT_OCFS2 3
 #define	QFMT_VFS_V1 4
+#define QFMT_UBIFS 5
 
 /* Size of block in which space limits are passed through the quota
  * interface */
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-22 15:15 ` [PATCH 1/8] quota: Allow to pass mount path to quotactl Sascha Hauer
@ 2021-01-22 17:16   ` Christoph Hellwig
  2021-01-25  8:38     ` Sascha Hauer
  0 siblings, 1 reply; 23+ messages in thread
From: Christoph Hellwig @ 2021-01-22 17:16 UTC (permalink / raw)
  To: Sascha Hauer
  Cc: linux-fsdevel, Richard Weinberger, linux-mtd, kernel, Jan Kara

On Fri, Jan 22, 2021 at 04:15:29PM +0100, Sascha Hauer wrote:
> This patch introduces the Q_PATH flag to the quotactl cmd argument.
> When given, the path given in the special argument to quotactl will
> be the mount path where the filesystem is mounted, instead of a path
> to the block device.
> This is necessary for filesystems which do not have a block device as
> backing store. Particularly this is done for upcoming UBIFS support.
> 
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

I hate overloading quotactl even more.  Why not add a new quotactl_path
syscall instead?

> +static struct super_block *quotactl_sb(dev_t dev, int cmd)
>  {
>  	struct super_block *sb;
>  	bool excl = false, thawed = false;
>  
>  	if (quotactl_cmd_onoff(cmd)) {
>  		excl = true;
> @@ -901,12 +887,50 @@ static struct super_block *quotactl_block(const char __user *special, int cmd)
>  		goto retry;
>  	}
>  	return sb;
> +}
> +
> +/*
> + * look up a superblock on which quota ops will be performed
> + * - use the name of a block device to find the superblock thereon
> + */
> +static struct super_block *quotactl_block(const char __user *special, int cmd)
> +{
> +#ifdef CONFIG_BLOCK
> +	struct filename *tmp = getname(special);
> +	int error;
> +	dev_t dev;
>  
> +	if (IS_ERR(tmp))
> +		return ERR_CAST(tmp);
> +	error = lookup_bdev(tmp->name, &dev);
> +	putname(tmp);
> +	if (error)
> +		return ERR_PTR(error);
> +
> +	return quotactl_sb(dev, cmd);
>  #else
>  	return ERR_PTR(-ENODEV);
>  #endif

Normal kernel style would be to keep the ifdef entirely outside the
function.

> +static struct super_block *quotactl_path(const char __user *special, int cmd)
> +{
> +	struct super_block *sb;
> +	struct path path;
> +	int error;
> +
> +	error = user_path_at(AT_FDCWD, special, LOOKUP_FOLLOW | LOOKUP_AUTOMOUNT,
> +			   &path);

This adds an overly long line.

> +	if (error)
> +		return ERR_PTR(error);
> +
> +	sb = quotactl_sb(path.mnt->mnt_sb->s_dev, cmd);

I think quotactl_sb should take the superblock directly.  This will
need a little refactoring of user_get_super, but will lead to much
better logic.

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 3/8] ubifs: Add support for FS_IOC_FS[SG]ETXATTR ioctls
  2021-01-22 15:15 ` [PATCH 3/8] ubifs: Add support for FS_IOC_FS[SG]ETXATTR ioctls Sascha Hauer
@ 2021-01-22 17:18   ` Christoph Hellwig
  0 siblings, 0 replies; 23+ messages in thread
From: Christoph Hellwig @ 2021-01-22 17:18 UTC (permalink / raw)
  To: Sascha Hauer
  Cc: linux-fsdevel, Richard Weinberger, linux-mtd, kernel, Jan Kara

On Fri, Jan 22, 2021 at 04:15:31PM +0100, Sascha Hauer wrote:
> +	if (xflags & FS_XFLAG_APPEND)
> +		iflags |= UBIFS_APPEND_FL;
> +
> +        return iflags;

This uses spaces for indentation

> +	if (flags & UBIFS_SYNC_FL)
> +		xflags |= FS_XFLAG_SYNC;
> +	if (flags & UBIFS_IMMUTABLE_FL)
> +		xflags |= FS_XFLAG_IMMUTABLE;
> +	if (flags & UBIFS_APPEND_FL)
> +		xflags |= FS_XFLAG_APPEND;
> +
> +        return xflags;

Same here.

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 7/8] ubifs: export get_znode
  2021-01-22 15:15 ` [PATCH 7/8] ubifs: export get_znode Sascha Hauer
@ 2021-01-22 17:20   ` Christoph Hellwig
  0 siblings, 0 replies; 23+ messages in thread
From: Christoph Hellwig @ 2021-01-22 17:20 UTC (permalink / raw)
  To: Sascha Hauer
  Cc: linux-fsdevel, Richard Weinberger, linux-mtd, kernel, Jan Kara

On Fri, Jan 22, 2021 at 04:15:35PM +0100, Sascha Hauer wrote:
> get_znode will be needed by upcoming UBIFS quota support. Rename it to
> ubifs_get_znode and export it.

I wonder why'd you want to export ubifs functionality, but it only
gains global instead of static scope, so this looks ok to me.

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-22 17:16   ` Christoph Hellwig
@ 2021-01-25  8:38     ` Sascha Hauer
  2021-01-25 15:45       ` Jan Kara
  0 siblings, 1 reply; 23+ messages in thread
From: Sascha Hauer @ 2021-01-25  8:38 UTC (permalink / raw)
  To: Christoph Hellwig
  Cc: linux-fsdevel, Richard Weinberger, linux-mtd, kernel, Jan Kara

On Fri, Jan 22, 2021 at 05:16:58PM +0000, Christoph Hellwig wrote:
> On Fri, Jan 22, 2021 at 04:15:29PM +0100, Sascha Hauer wrote:
> > This patch introduces the Q_PATH flag to the quotactl cmd argument.
> > When given, the path given in the special argument to quotactl will
> > be the mount path where the filesystem is mounted, instead of a path
> > to the block device.
> > This is necessary for filesystems which do not have a block device as
> > backing store. Particularly this is done for upcoming UBIFS support.
> > 
> > Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> 
> I hate overloading quotactl even more.  Why not add a new quotactl_path
> syscall instead?

We can probably do that. Honza, what do you think?

> 
> > +static struct super_block *quotactl_sb(dev_t dev, int cmd)
> >  {
> >  	struct super_block *sb;
> >  	bool excl = false, thawed = false;
> >  
> >  	if (quotactl_cmd_onoff(cmd)) {
> >  		excl = true;
> > @@ -901,12 +887,50 @@ static struct super_block *quotactl_block(const char __user *special, int cmd)
> >  		goto retry;
> >  	}
> >  	return sb;
> > +}
> > +
> > +/*
> > + * look up a superblock on which quota ops will be performed
> > + * - use the name of a block device to find the superblock thereon
> > + */
> > +static struct super_block *quotactl_block(const char __user *special, int cmd)
> > +{
> > +#ifdef CONFIG_BLOCK
> > +	struct filename *tmp = getname(special);
> > +	int error;
> > +	dev_t dev;
> >  
> > +	if (IS_ERR(tmp))
> > +		return ERR_CAST(tmp);
> > +	error = lookup_bdev(tmp->name, &dev);
> > +	putname(tmp);
> > +	if (error)
> > +		return ERR_PTR(error);
> > +
> > +	return quotactl_sb(dev, cmd);
> >  #else
> >  	return ERR_PTR(-ENODEV);
> >  #endif
> 
> Normal kernel style would be to keep the ifdef entirely outside the
> function.

It has been like this before, so changing this should be done in an
extra patch. Not sure if it's worth it though.

> 
> > +static struct super_block *quotactl_path(const char __user *special, int cmd)
> > +{
> > +	struct super_block *sb;
> > +	struct path path;
> > +	int error;
> > +
> > +	error = user_path_at(AT_FDCWD, special, LOOKUP_FOLLOW | LOOKUP_AUTOMOUNT,
> > +			   &path);
> 
> This adds an overly long line.

ok, will change.

> 
> > +	if (error)
> > +		return ERR_PTR(error);
> > +
> > +	sb = quotactl_sb(path.mnt->mnt_sb->s_dev, cmd);
> 
> I think quotactl_sb should take the superblock directly.  This will
> need a little refactoring of user_get_super, but will lead to much
> better logic.

What do you mean by "take"? Take the superblock as an argument to
quotactl_sb() or take a reference to the superblock?
Sorry, I don't really get where you aiming at.

Sascha

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-25  8:38     ` Sascha Hauer
@ 2021-01-25 15:45       ` Jan Kara
  2021-01-25 20:42         ` Christoph Hellwig
  2021-01-26 10:45         ` Sascha Hauer
  0 siblings, 2 replies; 23+ messages in thread
From: Jan Kara @ 2021-01-25 15:45 UTC (permalink / raw)
  To: Sascha Hauer
  Cc: Christoph Hellwig, linux-fsdevel, Richard Weinberger, linux-mtd,
	kernel, Jan Kara

On Mon 25-01-21 09:38:54, Sascha Hauer wrote:
> On Fri, Jan 22, 2021 at 05:16:58PM +0000, Christoph Hellwig wrote:
> > On Fri, Jan 22, 2021 at 04:15:29PM +0100, Sascha Hauer wrote:
> > > This patch introduces the Q_PATH flag to the quotactl cmd argument.
> > > When given, the path given in the special argument to quotactl will
> > > be the mount path where the filesystem is mounted, instead of a path
> > > to the block device.
> > > This is necessary for filesystems which do not have a block device as
> > > backing store. Particularly this is done for upcoming UBIFS support.
> > > 
> > > Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> > 
> > I hate overloading quotactl even more.  Why not add a new quotactl_path
> > syscall instead?
> 
> We can probably do that. Honza, what do you think?

Hum, yes, probably it would be cleaner to add a new syscall for this so
that we don't overload quotactl(2). I just didn't think of this.

> > > +static struct super_block *quotactl_path(const char __user *special, int cmd)
> > > +{
> > > +	struct super_block *sb;
> > > +	struct path path;
> > > +	int error;
> > > +
> > > +	error = user_path_at(AT_FDCWD, special, LOOKUP_FOLLOW | LOOKUP_AUTOMOUNT,
> > > +			   &path);
> > 
> > This adds an overly long line.
> 
> ok, will change.
> 
> > 
> > > +	if (error)
> > > +		return ERR_PTR(error);
> > > +
> > > +	sb = quotactl_sb(path.mnt->mnt_sb->s_dev, cmd);
> > 
> > I think quotactl_sb should take the superblock directly.  This will
> > need a little refactoring of user_get_super, but will lead to much
> > better logic.
> 
> What do you mean by "take"? Take the superblock as an argument to
> quotactl_sb() or take a reference to the superblock?
> Sorry, I don't really get where you aiming at.

I think Christoph was pointing at the fact it is suboptimal to search for
superblock by device number when you already have a pointer to it.  And I
guess he was suggesting we could pass 'sb' pointer to quotactl_sb() when we
already have it. Although to be honest, I'm not sure how Christoph imagines
the refactoring of user_get_super() he mentions - when we have a path
looked up through user_path(), that pins the superblock the path is on so
it cannot be unmounted. So perhaps quotactl_sb() can done like:

	...
retry:
	if (passed_sb) {
		sb = passed_sb;
		sb->s_count++;
		if (excl)
			down_write(&sb->s_umount);
		else
			down_read(&sb->s_umount);
	} else {
		sb = user_get_super(dev, excl);
		if (!sb)
			return ERR_PTR(-ENODEV);
	}
	...


								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-25 15:45       ` Jan Kara
@ 2021-01-25 20:42         ` Christoph Hellwig
  2021-01-26 13:17           ` Jan Kara
  2021-01-26 10:45         ` Sascha Hauer
  1 sibling, 1 reply; 23+ messages in thread
From: Christoph Hellwig @ 2021-01-25 20:42 UTC (permalink / raw)
  To: Jan Kara
  Cc: Sascha Hauer, Christoph Hellwig, linux-fsdevel,
	Richard Weinberger, linux-mtd, kernel, Jan Kara

On Mon, Jan 25, 2021 at 04:45:07PM +0100, Jan Kara wrote:
> > What do you mean by "take"? Take the superblock as an argument to
> > quotactl_sb() or take a reference to the superblock?
> > Sorry, I don't really get where you aiming at.
> 
> I think Christoph was pointing at the fact it is suboptimal to search for
> superblock by device number when you already have a pointer to it.  And I
> guess he was suggesting we could pass 'sb' pointer to quotactl_sb() when we
> already have it. Although to be honest, I'm not sure how Christoph imagines
> the refactoring of user_get_super() he mentions - when we have a path
> looked up through user_path(), that pins the superblock the path is on so
> it cannot be unmounted. So perhaps quotactl_sb() can done like:

I don't think we need a quotactl_sb at all, do_quotactl is in fact a
pretty good abstraction as-is.

For the path based one we just need to factor out a little helper
to set excl and thaw and then call it like:

	sb = path.dentry->d_inode->i_sb;

	if (excl)
		down_write(&sb->s_umount);
	else
		down_read(&sb->s_umount);
	if (thawed && sb->s_writers.frozen != SB_UNFROZEN)
		ret = -EBUSY;
	else
		ret = do_quotactl(sb, type, cmds, id, addr, &path);
	if (excl)
		up_write(&sb->s_umount);
	else
		up_read(&sb->s_umount);

as there is no good reason to bring over the somewhat strange wait until
unfrozen semantics to a new syscall.

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-25 15:45       ` Jan Kara
  2021-01-25 20:42         ` Christoph Hellwig
@ 2021-01-26 10:45         ` Sascha Hauer
  2021-01-27 14:46           ` Jan Kara
  1 sibling, 1 reply; 23+ messages in thread
From: Sascha Hauer @ 2021-01-26 10:45 UTC (permalink / raw)
  To: Jan Kara
  Cc: Christoph Hellwig, linux-fsdevel, Richard Weinberger, linux-mtd,
	kernel, Jan Kara

On Mon, Jan 25, 2021 at 04:45:07PM +0100, Jan Kara wrote:
> On Mon 25-01-21 09:38:54, Sascha Hauer wrote:
> > On Fri, Jan 22, 2021 at 05:16:58PM +0000, Christoph Hellwig wrote:
> > > On Fri, Jan 22, 2021 at 04:15:29PM +0100, Sascha Hauer wrote:
> > > > This patch introduces the Q_PATH flag to the quotactl cmd argument.
> > > > When given, the path given in the special argument to quotactl will
> > > > be the mount path where the filesystem is mounted, instead of a path
> > > > to the block device.
> > > > This is necessary for filesystems which do not have a block device as
> > > > backing store. Particularly this is done for upcoming UBIFS support.
> > > > 
> > > > Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> > > 
> > > I hate overloading quotactl even more.  Why not add a new quotactl_path
> > > syscall instead?
> > 
> > We can probably do that. Honza, what do you think?
> 
> Hum, yes, probably it would be cleaner to add a new syscall for this so
> that we don't overload quotactl(2). I just didn't think of this.

How should the semantics of that new syscall look like?

The easiest and most obvious way would be to do it like the quotactl(2)
and just replace the special argument with a path:

int quotactl_path(int cmd, const char *path, int id, caddr_t addr);

If we try adding a new syscall then we could completely redefine the API
and avoid the shortcomings of the original quotactl(2) if there are any.
Can you foresee the discussions we end up in? I am afraid I am opening a
can of worms here.
OTOH there might be value in keeping the new syscall compatible to the
existing one, but I don't know how much this argument counts.

Sascha

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-25 20:42         ` Christoph Hellwig
@ 2021-01-26 13:17           ` Jan Kara
  2021-01-26 13:34             ` Christoph Hellwig
  0 siblings, 1 reply; 23+ messages in thread
From: Jan Kara @ 2021-01-26 13:17 UTC (permalink / raw)
  To: Christoph Hellwig
  Cc: Jan Kara, Sascha Hauer, linux-fsdevel, Richard Weinberger,
	linux-mtd, kernel, Jan Kara

On Mon 25-01-21 20:42:49, Christoph Hellwig wrote:
> On Mon, Jan 25, 2021 at 04:45:07PM +0100, Jan Kara wrote:
> > > What do you mean by "take"? Take the superblock as an argument to
> > > quotactl_sb() or take a reference to the superblock?
> > > Sorry, I don't really get where you aiming at.
> > 
> > I think Christoph was pointing at the fact it is suboptimal to search for
> > superblock by device number when you already have a pointer to it.  And I
> > guess he was suggesting we could pass 'sb' pointer to quotactl_sb() when we
> > already have it. Although to be honest, I'm not sure how Christoph imagines
> > the refactoring of user_get_super() he mentions - when we have a path
> > looked up through user_path(), that pins the superblock the path is on so
> > it cannot be unmounted. So perhaps quotactl_sb() can done like:
> 
> I don't think we need a quotactl_sb at all, do_quotactl is in fact a
> pretty good abstraction as-is.
> 
> For the path based one we just need to factor out a little helper
> to set excl and thaw and then call it like:
> 
> 	sb = path.dentry->d_inode->i_sb;
> 
> 	if (excl)
> 		down_write(&sb->s_umount);
> 	else
> 		down_read(&sb->s_umount);
> 	if (thawed && sb->s_writers.frozen != SB_UNFROZEN)
> 		ret = -EBUSY;
> 	else
> 		ret = do_quotactl(sb, type, cmds, id, addr, &path);
> 	if (excl)
> 		up_write(&sb->s_umount);
> 	else
> 		up_read(&sb->s_umount);
> 
> as there is no good reason to bring over the somewhat strange wait until
> unfrozen semantics to a new syscall.

Well, I don't think that "wait until unfrozen" is that strange e.g. for
Q_SETQUOTA - it behaves like setxattr() or any other filesystem
modification operation. And IMO it is desirable that filesystem freezing is
transparent for operations like these. For stuff like Q_QUOTAON, I agree
that returning EBUSY makes sense but then I'm not convinced it's really
simpler or more useful behavior...

								Honza

-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-26 13:17           ` Jan Kara
@ 2021-01-26 13:34             ` Christoph Hellwig
  2021-01-26 16:18               ` Jan Kara
  0 siblings, 1 reply; 23+ messages in thread
From: Christoph Hellwig @ 2021-01-26 13:34 UTC (permalink / raw)
  To: Jan Kara
  Cc: Christoph Hellwig, Sascha Hauer, linux-fsdevel,
	Richard Weinberger, linux-mtd, kernel, Jan Kara

On Tue, Jan 26, 2021 at 02:17:52PM +0100, Jan Kara wrote:
> Well, I don't think that "wait until unfrozen" is that strange e.g. for
> Q_SETQUOTA - it behaves like setxattr() or any other filesystem
> modification operation. And IMO it is desirable that filesystem freezing is
> transparent for operations like these. For stuff like Q_QUOTAON, I agree
> that returning EBUSY makes sense but then I'm not convinced it's really
> simpler or more useful behavior...

If we want it to behave like other syscalls we'll just need to throw in
a mnt_want_write/mnt_drop_write pair.  Than it behaves exactly like other
syscalls.

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-26 13:34             ` Christoph Hellwig
@ 2021-01-26 16:18               ` Jan Kara
  2021-01-27 14:05                 ` Sascha Hauer
  0 siblings, 1 reply; 23+ messages in thread
From: Jan Kara @ 2021-01-26 16:18 UTC (permalink / raw)
  To: Christoph Hellwig
  Cc: Jan Kara, Sascha Hauer, linux-fsdevel, Richard Weinberger,
	linux-mtd, kernel, Jan Kara

On Tue 26-01-21 13:34:06, Christoph Hellwig wrote:
> On Tue, Jan 26, 2021 at 02:17:52PM +0100, Jan Kara wrote:
> > Well, I don't think that "wait until unfrozen" is that strange e.g. for
> > Q_SETQUOTA - it behaves like setxattr() or any other filesystem
> > modification operation. And IMO it is desirable that filesystem freezing is
> > transparent for operations like these. For stuff like Q_QUOTAON, I agree
> > that returning EBUSY makes sense but then I'm not convinced it's really
> > simpler or more useful behavior...
> 
> If we want it to behave like other syscalls we'll just need to throw in
> a mnt_want_write/mnt_drop_write pair.  Than it behaves exactly like other
> syscalls.

Right, we could do that. I'd just note that the "wait until unfrozen" and
holding of sb->s_umount semaphore is equivalent to
mnt_want_write/mnt_drop_write pair. But I agree
mnt_want_write/mnt_drop_write is easier to understand and there's no reason
not to use it. So I'm for that simplification in the new syscall.

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-26 16:18               ` Jan Kara
@ 2021-01-27 14:05                 ` Sascha Hauer
  2021-01-27 14:19                   ` Jan Kara
  0 siblings, 1 reply; 23+ messages in thread
From: Sascha Hauer @ 2021-01-27 14:05 UTC (permalink / raw)
  To: Jan Kara
  Cc: Christoph Hellwig, linux-fsdevel, Richard Weinberger, linux-mtd,
	kernel, Jan Kara

On Tue, Jan 26, 2021 at 05:18:29PM +0100, Jan Kara wrote:
> On Tue 26-01-21 13:34:06, Christoph Hellwig wrote:
> > On Tue, Jan 26, 2021 at 02:17:52PM +0100, Jan Kara wrote:
> > > Well, I don't think that "wait until unfrozen" is that strange e.g. for
> > > Q_SETQUOTA - it behaves like setxattr() or any other filesystem
> > > modification operation. And IMO it is desirable that filesystem freezing is
> > > transparent for operations like these. For stuff like Q_QUOTAON, I agree
> > > that returning EBUSY makes sense but then I'm not convinced it's really
> > > simpler or more useful behavior...
> > 
> > If we want it to behave like other syscalls we'll just need to throw in
> > a mnt_want_write/mnt_drop_write pair.  Than it behaves exactly like other
> > syscalls.
> 
> Right, we could do that. I'd just note that the "wait until unfrozen" and
> holding of sb->s_umount semaphore is equivalent to
> mnt_want_write/mnt_drop_write pair. But I agree
> mnt_want_write/mnt_drop_write is easier to understand and there's no reason
> not to use it. So I'm for that simplification in the new syscall.

Due to the user_path_at() to the mountpoint the fs won't go away, so I
guess for non-exclusive, non-write quota command I don't need any
additional locking. For non-exclusive, write commands I'll need a
mnt_want_write/mnt_drop_write pair. What about the exclusive, write
commands (Q_QUOTAON/Q_QUOTAOFF)?

Sascha

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-27 14:05                 ` Sascha Hauer
@ 2021-01-27 14:19                   ` Jan Kara
  0 siblings, 0 replies; 23+ messages in thread
From: Jan Kara @ 2021-01-27 14:19 UTC (permalink / raw)
  To: Sascha Hauer
  Cc: Jan Kara, Christoph Hellwig, linux-fsdevel, Richard Weinberger,
	linux-mtd, kernel, Jan Kara

On Wed 27-01-21 15:05:10, Sascha Hauer wrote:
> On Tue, Jan 26, 2021 at 05:18:29PM +0100, Jan Kara wrote:
> > On Tue 26-01-21 13:34:06, Christoph Hellwig wrote:
> > > On Tue, Jan 26, 2021 at 02:17:52PM +0100, Jan Kara wrote:
> > > > Well, I don't think that "wait until unfrozen" is that strange e.g. for
> > > > Q_SETQUOTA - it behaves like setxattr() or any other filesystem
> > > > modification operation. And IMO it is desirable that filesystem freezing is
> > > > transparent for operations like these. For stuff like Q_QUOTAON, I agree
> > > > that returning EBUSY makes sense but then I'm not convinced it's really
> > > > simpler or more useful behavior...
> > > 
> > > If we want it to behave like other syscalls we'll just need to throw in
> > > a mnt_want_write/mnt_drop_write pair.  Than it behaves exactly like other
> > > syscalls.
> > 
> > Right, we could do that. I'd just note that the "wait until unfrozen" and
> > holding of sb->s_umount semaphore is equivalent to
> > mnt_want_write/mnt_drop_write pair. But I agree
> > mnt_want_write/mnt_drop_write is easier to understand and there's no reason
> > not to use it. So I'm for that simplification in the new syscall.
> 
> Due to the user_path_at() to the mountpoint the fs won't go away, so I
> guess for non-exclusive, non-write quota command I don't need any
> additional locking.

They still need s_umount in read mode - to serialize against quotaon /
quotaoff (happening either through quotactl or filesystem remount).

> For non-exclusive, write commands I'll need a
> mnt_want_write/mnt_drop_write pair. What about the exclusive, write
> commands (Q_QUOTAON/Q_QUOTAOFF)?

And these still need s_umount in write mode for similar reasons. So the
only thing you really save is the "wait-for-unfreeze" code.

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-26 10:45         ` Sascha Hauer
@ 2021-01-27 14:46           ` Jan Kara
  2021-01-27 16:25             ` Christoph Hellwig
  0 siblings, 1 reply; 23+ messages in thread
From: Jan Kara @ 2021-01-27 14:46 UTC (permalink / raw)
  To: Sascha Hauer
  Cc: Jan Kara, Christoph Hellwig, linux-fsdevel, Richard Weinberger,
	linux-mtd, kernel, Jan Kara, linux-api

On Tue 26-01-21 11:45:57, Sascha Hauer wrote:
> On Mon, Jan 25, 2021 at 04:45:07PM +0100, Jan Kara wrote:
> > On Mon 25-01-21 09:38:54, Sascha Hauer wrote:
> > > On Fri, Jan 22, 2021 at 05:16:58PM +0000, Christoph Hellwig wrote:
> > > > On Fri, Jan 22, 2021 at 04:15:29PM +0100, Sascha Hauer wrote:
> > > > > This patch introduces the Q_PATH flag to the quotactl cmd argument.
> > > > > When given, the path given in the special argument to quotactl will
> > > > > be the mount path where the filesystem is mounted, instead of a path
> > > > > to the block device.
> > > > > This is necessary for filesystems which do not have a block device as
> > > > > backing store. Particularly this is done for upcoming UBIFS support.
> > > > > 
> > > > > Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> > > > 
> > > > I hate overloading quotactl even more.  Why not add a new quotactl_path
> > > > syscall instead?
> > > 
> > > We can probably do that. Honza, what do you think?
> > 
> > Hum, yes, probably it would be cleaner to add a new syscall for this so
> > that we don't overload quotactl(2). I just didn't think of this.
> 
> How should the semantics of that new syscall look like?
> 
> The easiest and most obvious way would be to do it like the quotactl(2)
> and just replace the special argument with a path:
> 
> int quotactl_path(int cmd, const char *path, int id, caddr_t addr);

Yes, that's what I meant.

> If we try adding a new syscall then we could completely redefine the API
> and avoid the shortcomings of the original quotactl(2) if there are any.
> Can you foresee the discussions we end up in? I am afraid I am opening a
> can of worms here.
> OTOH there might be value in keeping the new syscall compatible to the
> existing one, but I don't know how much this argument counts.

That's a good question but also a can of worms as you write :). One obvious
problem with quotactl() is that's it's ioctl-like interface. So we have
several different operations mixed into a single syscall. Currently there
are these operations:

#define Q_SYNC     0x800001     /* sync disk copy of a filesystems quotas */
#define Q_QUOTAON  0x800002     /* turn quotas on */
#define Q_QUOTAOFF 0x800003     /* turn quotas off */
#define Q_GETFMT   0x800004     /* get quota format used on given filesystem */
#define Q_GETINFO  0x800005     /* get information about quota files */
#define Q_SETINFO  0x800006     /* set information about quota files */
#define Q_GETQUOTA 0x800007     /* get user quota structure */
#define Q_SETQUOTA 0x800008     /* set user quota structure */
#define Q_GETNEXTQUOTA 0x800009 /* get disk limits and usage >= ID */
<plus their XFS variants>

In a puristic world they'd be 9 different syscalls ... or somewhat less
because Q_GETNEXTQUOTA is a superset of Q_GETQUOTA, we could drop Q_SYNC
and Q_GETFMT because they have dubious value these days so we'd be left
with 6. I don't have a strong opinion whether 6 syscalls are worth the
cleanliness or whether we should go with just one new quotactl_path()
syscall. I've CCed linux-api in case other people have opinion.

Anyway, even if we go with single quotactl_path() syscall we should remove
the duplication between VFS and XFS quotactls when we are creating a new
syscall. Thoughts?

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

^ permalink raw reply	[flat|nested] 23+ messages in thread

* Re: [PATCH 1/8] quota: Allow to pass mount path to quotactl
  2021-01-27 14:46           ` Jan Kara
@ 2021-01-27 16:25             ` Christoph Hellwig
  0 siblings, 0 replies; 23+ messages in thread
From: Christoph Hellwig @ 2021-01-27 16:25 UTC (permalink / raw)
  To: Jan Kara
  Cc: Sascha Hauer, Christoph Hellwig, linux-fsdevel,
	Richard Weinberger, linux-mtd, kernel, Jan Kara, linux-api

On Wed, Jan 27, 2021 at 03:46:46PM +0100, Jan Kara wrote:
> In a puristic world they'd be 9 different syscalls ... or somewhat less
> because Q_GETNEXTQUOTA is a superset of Q_GETQUOTA, we could drop Q_SYNC
> and Q_GETFMT because they have dubious value these days so we'd be left
> with 6. I don't have a strong opinion whether 6 syscalls are worth the
> cleanliness or whether we should go with just one new quotactl_path()
> syscall. I've CCed linux-api in case other people have opinion.
> 
> Anyway, even if we go with single quotactl_path() syscall we should remove
> the duplication between VFS and XFS quotactls when we are creating a new
> syscall. Thoughts?

I thunk the multiplexer is just fine.  We don't really need Q_SYNC
for a new syscall.  For XFS vs classic VFS quota we probably don't
need to duplicate the two data structure, but we need to make sure we
catch the superset of the information if we want to disable the old
ones.

So I suspect just supporting evrything excpt for the global Q_SYNC
and reusing do_quotactl as-is is the most maintainable and easiest
to understand way forward.

^ permalink raw reply	[flat|nested] 23+ messages in thread

end of thread, other threads:[~2021-01-27 16:27 UTC | newest]

Thread overview: 23+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-01-22 15:15 [PATCH v5 0/8] Add quota support to UBIFS Sascha Hauer
2021-01-22 15:15 ` [PATCH 1/8] quota: Allow to pass mount path to quotactl Sascha Hauer
2021-01-22 17:16   ` Christoph Hellwig
2021-01-25  8:38     ` Sascha Hauer
2021-01-25 15:45       ` Jan Kara
2021-01-25 20:42         ` Christoph Hellwig
2021-01-26 13:17           ` Jan Kara
2021-01-26 13:34             ` Christoph Hellwig
2021-01-26 16:18               ` Jan Kara
2021-01-27 14:05                 ` Sascha Hauer
2021-01-27 14:19                   ` Jan Kara
2021-01-26 10:45         ` Sascha Hauer
2021-01-27 14:46           ` Jan Kara
2021-01-27 16:25             ` Christoph Hellwig
2021-01-22 15:15 ` [PATCH 2/8] ubifs: move checks and preparation into setflags() Sascha Hauer
2021-01-22 15:15 ` [PATCH 3/8] ubifs: Add support for FS_IOC_FS[SG]ETXATTR ioctls Sascha Hauer
2021-01-22 17:18   ` Christoph Hellwig
2021-01-22 15:15 ` [PATCH 4/8] ubifs: do not ubifs_inode() on potentially NULL pointer Sascha Hauer
2021-01-22 15:15 ` [PATCH 5/8] ubifs: Factor out ubifs_set_feature_flag() Sascha Hauer
2021-01-22 15:15 ` [PATCH 6/8] ubifs: Add support for project id Sascha Hauer
2021-01-22 15:15 ` [PATCH 7/8] ubifs: export get_znode Sascha Hauer
2021-01-22 17:20   ` Christoph Hellwig
2021-01-22 15:15 ` [PATCH 8/8] ubifs: Add quota support Sascha Hauer

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).