| Message ID | 20210121131959.646623-3-christian.brauner@ubuntu.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | idmapped mounts | expand |
On Thu, Jan 21, 2021 at 02:19:21PM +0100, Christian Brauner wrote: > Add simple helpers to make it easy to map kuids into and from idmapped > mounts. We provide simple wrappers that filesystems can use to e.g. > initialize inodes similar to i_{uid,gid}_read() and i_{uid,gid}_write(). > Accessing an inode through an idmapped mount maps the i_uid and i_gid of > the inode to the mount's user namespace. If the fsids are used to > initialize inodes they are unmapped according to the mount's user > namespace. Passing the initial user namespace to these helpers makes > them a nop and so any non-idmapped paths will not be impacted. > > Link: https://lore.kernel.org/r/20210112220124.837960-9-christian.brauner@ubuntu.com > Cc: David Howells <dhowells@redhat.com> > Cc: Al Viro <viro@zeniv.linux.org.uk> > Cc: linux-fsdevel@vger.kernel.org > Reviewed-by: Christoph Hellwig <hch@lst.de> > Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com> > --- > /* v2 */ > - Christoph Hellwig <hch@lst.de>: > - Get rid of the ifdefs and the config option that hid idmapped mounts. > > /* v3 */ > unchanged > > /* v4 */ > - Serge Hallyn <serge@hallyn.com>: > - Use "mnt_userns" to refer to a vfsmount's userns everywhere to make > terminology consistent. > > /* v5 */ > unchanged > base-commit: 7c53f6b671f4aba70ff15e1b05148b10d58c2837 > > /* v6 */ > unchanged > base-commit: 19c329f6808995b142b3966301f217c831e7cf31 > --- > include/linux/fs.h | 47 ++++++++++++++++++++++++++++++++++++++++++++++ > 1 file changed, 47 insertions(+) > > diff --git a/include/linux/fs.h b/include/linux/fs.h > index fd0b80e6361d..3165998e2294 100644 > --- a/include/linux/fs.h > +++ b/include/linux/fs.h > @@ -40,6 +40,7 @@ > #include <linux/build_bug.h> > #include <linux/stddef.h> > #include <linux/mount.h> > +#include <linux/cred.h> > > #include <asm/byteorder.h> > #include <uapi/linux/fs.h> > @@ -1573,6 +1574,52 @@ static inline void i_gid_write(struct inode *inode, gid_t gid) > inode->i_gid = make_kgid(inode->i_sb->s_user_ns, gid); > } > > +static inline kuid_t kuid_into_mnt(struct user_namespace *mnt_userns, > + kuid_t kuid) > +{ > + return make_kuid(mnt_userns, __kuid_val(kuid)); > +} > + Hi Christian, I am having little trouble w.r.t function names and trying to figure out whether they are mapping id down or up. For example, kuid_into_mnt() ultimately calls map_id_down(). That is, id visible inside user namespace is mapped to host (if observer is in init_user_ns, IIUC). But fsuid_into_mnt() ultimately calls map_id_up(). That's take a kuid and map it into the user_namespace. So both the helpers end with into_mnt() but one maps id down and other maps id up. I found this confusing and was wondering how should I visualize it. So thought of asking you. Is this intentional or can naming be improved so that *_into_mnt() means one thing (Either map_id_up() or map_id_down()). And vice-a-versa for *_from_mnt(). Thanks Vivek > +static inline kgid_t kgid_into_mnt(struct user_namespace *mnt_userns, > + kgid_t kgid) > +{ > + return make_kgid(mnt_userns, __kgid_val(kgid)); > +} > + > +static inline kuid_t i_uid_into_mnt(struct user_namespace *mnt_userns, > + const struct inode *inode) > +{ > + return kuid_into_mnt(mnt_userns, inode->i_uid); > +} > + > +static inline kgid_t i_gid_into_mnt(struct user_namespace *mnt_userns, > + const struct inode *inode) > +{ > + return kgid_into_mnt(mnt_userns, inode->i_gid); > +} > + > +static inline kuid_t kuid_from_mnt(struct user_namespace *mnt_userns, > + kuid_t kuid) > +{ > + return KUIDT_INIT(from_kuid(mnt_userns, kuid)); > +} > + > +static inline kgid_t kgid_from_mnt(struct user_namespace *mnt_userns, > + kgid_t kgid) > +{ > + return KGIDT_INIT(from_kgid(mnt_userns, kgid)); > +} > + > +static inline kuid_t fsuid_into_mnt(struct user_namespace *mnt_userns) > +{ > + return kuid_from_mnt(mnt_userns, current_fsuid()); > +} > + > +static inline kgid_t fsgid_into_mnt(struct user_namespace *mnt_userns) > +{ > + return kgid_from_mnt(mnt_userns, current_fsgid()); > +} > + > extern struct timespec64 current_time(struct inode *inode); > > /* > -- > 2.30.0 >
diff --git a/include/linux/fs.h b/include/linux/fs.h index fd0b80e6361d..3165998e2294 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -40,6 +40,7 @@ #include <linux/build_bug.h> #include <linux/stddef.h> #include <linux/mount.h> +#include <linux/cred.h> #include <asm/byteorder.h> #include <uapi/linux/fs.h> @@ -1573,6 +1574,52 @@ static inline void i_gid_write(struct inode *inode, gid_t gid) inode->i_gid = make_kgid(inode->i_sb->s_user_ns, gid); } +static inline kuid_t kuid_into_mnt(struct user_namespace *mnt_userns, + kuid_t kuid) +{ + return make_kuid(mnt_userns, __kuid_val(kuid)); +} + +static inline kgid_t kgid_into_mnt(struct user_namespace *mnt_userns, + kgid_t kgid) +{ + return make_kgid(mnt_userns, __kgid_val(kgid)); +} + +static inline kuid_t i_uid_into_mnt(struct user_namespace *mnt_userns, + const struct inode *inode) +{ + return kuid_into_mnt(mnt_userns, inode->i_uid); +} + +static inline kgid_t i_gid_into_mnt(struct user_namespace *mnt_userns, + const struct inode *inode) +{ + return kgid_into_mnt(mnt_userns, inode->i_gid); +} + +static inline kuid_t kuid_from_mnt(struct user_namespace *mnt_userns, + kuid_t kuid) +{ + return KUIDT_INIT(from_kuid(mnt_userns, kuid)); +} + +static inline kgid_t kgid_from_mnt(struct user_namespace *mnt_userns, + kgid_t kgid) +{ + return KGIDT_INIT(from_kgid(mnt_userns, kgid)); +} + +static inline kuid_t fsuid_into_mnt(struct user_namespace *mnt_userns) +{ + return kuid_from_mnt(mnt_userns, current_fsuid()); +} + +static inline kgid_t fsgid_into_mnt(struct user_namespace *mnt_userns) +{ + return kgid_from_mnt(mnt_userns, current_fsgid()); +} + extern struct timespec64 current_time(struct inode *inode); /*
Add simple helpers to make it easy to map kuids into and from idmapped mounts. We provide simple wrappers that filesystems can use to e.g. initialize inodes similar to i_{uid,gid}_read() and i_{uid,gid}_write(). Accessing an inode through an idmapped mount maps the i_uid and i_gid of the inode to the mount's user namespace. If the fsids are used to initialize inodes they are unmapped according to the mount's user namespace. Passing the initial user namespace to these helpers makes them a nop and so any non-idmapped paths will not be impacted. Link: https://lore.kernel.org/r/20210112220124.837960-9-christian.brauner@ubuntu.com Cc: David Howells <dhowells@redhat.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: linux-fsdevel@vger.kernel.org Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com> --- /* v2 */ - Christoph Hellwig <hch@lst.de>: - Get rid of the ifdefs and the config option that hid idmapped mounts. /* v3 */ unchanged /* v4 */ - Serge Hallyn <serge@hallyn.com>: - Use "mnt_userns" to refer to a vfsmount's userns everywhere to make terminology consistent. /* v5 */ unchanged base-commit: 7c53f6b671f4aba70ff15e1b05148b10d58c2837 /* v6 */ unchanged base-commit: 19c329f6808995b142b3966301f217c831e7cf31 --- include/linux/fs.h | 47 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 47 insertions(+)