[PATCH RFC v2 0/2] O_NOCMTIME protected by...

[PATCH RFC v2 0/2] O_NOCMTIME protected by generic mount option

From: Zach Brown <hidden>
Date: 2015-05-15 21:23:46

Here's a current draft of what is now the O_NOCMTIME series.  It
implements the frequent suggestion to gate unprivileged O_NOCMTIME use
with a mount option.

This method has the advantage of being entirely runtime.  There's no
persistence that'd require updating all the tools that deal with each
file system's format.  It's also requested by writers as they open.
Writes to the file that know nothing of O_NOCMTIME will behave as
usual.

Another suggested method is to use inode attributes: require root to
set +nocmtime on a dir and inherit it down subdirs to new files.  This
nicely solves the unprivilieged use problem without having to fiddle
with mount options but it requires touching all systems that support
it and would prevent cmtime updates on all writes to the inode.  I
have a patch series that starts on this but haven't taken it very far.

Sage is working on spinning up some hardware to test the various dirty
inode avoidance methods at load and should have numbers soon.  That'll
tell us if lazytime isn't good enough and if any of this is worth the
trouble.

- z

[PATCH RFC v2 1/2] vfs: add generic nocmtime mount flag

From: Zach Brown <hidden>
Date: 2015-05-15 21:23:47

Add the infrastructure to support a generic nocmtime mount flag.  Like
MS_NOATIME/MNT_NOATIME this can be used to support the mount option in
file systems without having to touch each file system.

This will be used to provide a priviledged indication that unpriviledged
apps can safely use O_NOCMTIME to prevent cmtime updates without harm.

Signed-off-by: Zach Brown <redacted>
---
 fs/namespace.c          | 2 ++
 fs/proc_namespace.c     | 1 +
 fs/statfs.c             | 2 ++
 include/linux/mount.h   | 1 +
 include/linux/statfs.h  | 1 +
 include/uapi/linux/fs.h | 1 +
 6 files changed, 8 insertions(+)

diff --git a/fs/namespace.c b/fs/namespace.c
index 1b9e111..48be1f9 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c

@@ -2652,6 +2652,8 @@ long do_mount(const char *dev_name, const char __user *dir_name,
 		mnt_flags |= MNT_NOEXEC;
 	if (flags & MS_NOATIME)
 		mnt_flags |= MNT_NOATIME;
+	if (flags & MS_NOCMTIME)
+		mnt_flags |= MNT_NOCMTIME;
 	if (flags & MS_NODIRATIME)
 		mnt_flags |= MNT_NODIRATIME;
 	if (flags & MS_STRICTATIME)

diff --git a/fs/proc_namespace.c b/fs/proc_namespace.c
index 8db932d..49d7839 100644
--- a/fs/proc_namespace.c
+++ b/fs/proc_namespace.c

@@ -64,6 +64,7 @@ static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt)
 		{ MNT_NODEV, ",nodev" },
 		{ MNT_NOEXEC, ",noexec" },
 		{ MNT_NOATIME, ",noatime" },
+		{ MNT_NOCMTIME, ",nocmtime" },
 		{ MNT_NODIRATIME, ",nodiratime" },
 		{ MNT_RELATIME, ",relatime" },
 		{ 0, NULL }

diff --git a/fs/statfs.c b/fs/statfs.c
index 083dc0a..43d3de2 100644
--- a/fs/statfs.c
+++ b/fs/statfs.c

@@ -23,6 +23,8 @@ static int flags_by_mnt(int mnt_flags)
 		flags |= ST_NOEXEC;
 	if (mnt_flags & MNT_NOATIME)
 		flags |= ST_NOATIME;
+	if (mnt_flags & MNT_NOCMTIME)
+		flags |= ST_NOCMTIME;
 	if (mnt_flags & MNT_NODIRATIME)
 		flags |= ST_NODIRATIME;
 	if (mnt_flags & MNT_RELATIME)

diff --git a/include/linux/mount.h b/include/linux/mount.h
index f822c3c..deb458f 100644
--- a/include/linux/mount.h
+++ b/include/linux/mount.h

@@ -28,6 +28,7 @@ struct mnt_namespace;
 #define MNT_NODIRATIME	0x10
 #define MNT_RELATIME	0x20
 #define MNT_READONLY	0x40	/* does the user want this to be r/o? */
+#define MNT_NOCMTIME	0x80	/* allow O_NOCMTIME to stop cmtime updates */
 
 #define MNT_SHRINKABLE	0x100
 #define MNT_WRITE_HOLD	0x200

diff --git a/include/linux/statfs.h b/include/linux/statfs.h
index 0166d32..bde224e 100644
--- a/include/linux/statfs.h
+++ b/include/linux/statfs.h

@@ -39,5 +39,6 @@ struct kstatfs {
 #define ST_NOATIME	0x0400	/* do not update access times */
 #define ST_NODIRATIME	0x0800	/* do not update directory access times */
 #define ST_RELATIME	0x1000	/* update atime relative to mtime/ctime */
+#define ST_NOCMTIME	0x2000	/* allow O_NOCMTIME to stop cmtime updates */
 
 #endif

diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
index 9b964a5..af1131e 100644
--- a/include/uapi/linux/fs.h
+++ b/include/uapi/linux/fs.h

@@ -91,6 +91,7 @@ struct inodes_stat_t {
 #define MS_I_VERSION	(1<<23) /* Update inode I_version field */
 #define MS_STRICTATIME	(1<<24) /* Always perform atime updates */
 #define MS_LAZYTIME	(1<<25) /* Update the on-disk [acm]times lazily */
+#define MS_NOCMTIME	(1<<26) /* allow O_NOCMTIME to stop cmtime updates */
 
 /* These sb flags are internal to the kernel */
 #define MS_NOSEC	(1<<28)

-- 
2.1.0

`h`	back out one level
`j`	next message in thread
`k`	previous message in thread
`l`	drill in
`Esc`	close help / fold thread tree
`?`	toggle this help