| Message ID | 20241106-statmount-v2-2-93ba2aad38d1@kernel.org (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | fs: allow statmount to fetch the sb->s_subtype field | expand |
On Wed 06-11-24 14:53:06, Jeff Layton wrote: > /proc/self/mountinfo displays the devicename for the mount, but > statmount() doesn't yet have a way to return it. Add a new > STATMOUNT_MNT_DEVNAME flag, claim the 32-bit __spare1 field to hold the > offset into the str[] array. STATMOUNT_MNT_DEVNAME will only be set in > the return mask if there is a device string. > > Signed-off-by: Jeff Layton <jlayton@kernel.org> Just one question below: > @@ -5078,6 +5091,12 @@ static int statmount_string(struct kstatmount *s, u64 flag) > if (seq->count == sm->fs_subtype) > return 0; > break; > + case STATMOUNT_MNT_DEVNAME: > + sm->mnt_devname = seq->count; > + ret = statmount_mnt_devname(s, seq); > + if (seq->count == sm->mnt_devname) Why this odd check? Why don't you rather do: if (ret) ? > + return ret; > + break; > default: > WARN_ON_ONCE(true); > return -EINVAL; Honza
On Thu, 2024-11-07 at 10:40 +0100, Jan Kara wrote: > On Wed 06-11-24 14:53:06, Jeff Layton wrote: > > /proc/self/mountinfo displays the devicename for the mount, but > > statmount() doesn't yet have a way to return it. Add a new > > STATMOUNT_MNT_DEVNAME flag, claim the 32-bit __spare1 field to hold the > > offset into the str[] array. STATMOUNT_MNT_DEVNAME will only be set in > > the return mask if there is a device string. > > > > Signed-off-by: Jeff Layton <jlayton@kernel.org> > > Just one question below: > > > @@ -5078,6 +5091,12 @@ static int statmount_string(struct kstatmount *s, u64 flag) > > if (seq->count == sm->fs_subtype) > > return 0; > > break; > > + case STATMOUNT_MNT_DEVNAME: > > + sm->mnt_devname = seq->count; > > + ret = statmount_mnt_devname(s, seq); > > + if (seq->count == sm->mnt_devname) > > Why this odd check? Why don't you rather do: > if (ret) > ? > statmount_mnt_devname() can return without emitting anything to the seq if ->show_devname and r->mnt_devname are both NULL. In that case, we don't want statmount_string() to return an error, but we also don't want to do any further manipulation of the seq. So, the above handles either the case where show_devname returned an error and the case where there was nothing to emit. I did consider having statmount_mnt_devname() return -ENOBUFS if there was nothing to emit, and then handle that in the caller, but checking to see whether the seq has advanced seemed like a cleaner solution. I can add a comment to that effect since it does look a bit confusing. > > + return ret; > > + break; > > default: > > WARN_ON_ONCE(true); > > return -EINVAL; > > Honza >
On Thu 07-11-24 06:15:40, Jeff Layton wrote: > On Thu, 2024-11-07 at 10:40 +0100, Jan Kara wrote: > > On Wed 06-11-24 14:53:06, Jeff Layton wrote: > > > /proc/self/mountinfo displays the devicename for the mount, but > > > statmount() doesn't yet have a way to return it. Add a new > > > STATMOUNT_MNT_DEVNAME flag, claim the 32-bit __spare1 field to hold the > > > offset into the str[] array. STATMOUNT_MNT_DEVNAME will only be set in > > > the return mask if there is a device string. > > > > > > Signed-off-by: Jeff Layton <jlayton@kernel.org> > > > > Just one question below: > > > > > @@ -5078,6 +5091,12 @@ static int statmount_string(struct kstatmount *s, u64 flag) > > > if (seq->count == sm->fs_subtype) > > > return 0; > > > break; > > > + case STATMOUNT_MNT_DEVNAME: > > > + sm->mnt_devname = seq->count; > > > + ret = statmount_mnt_devname(s, seq); > > > + if (seq->count == sm->mnt_devname) > > > > Why this odd check? Why don't you rather do: > > if (ret) > > ? > > > > statmount_mnt_devname() can return without emitting anything to the seq > if ->show_devname and r->mnt_devname are both NULL. In that case, we > don't want statmount_string() to return an error, but we also don't > want to do any further manipulation of the seq. So, the above handles > either the case where show_devname returned an error and the case where > there was nothing to emit. > > I did consider having statmount_mnt_devname() return -ENOBUFS if there > was nothing to emit, and then handle that in the caller, but checking > to see whether the seq has advanced seemed like a cleaner solution. OK, but don't we want to emit an empty string - i.e., just \0 character in case r->mnt_devname is NULL and there's no ->show_devname? Because now we literaly emit nothing and it's going to be very confusing for userspace to parse this when this happens in the middle of other strings in the seq. And emitting \0 is exactly what will happen if we don't do anything special in STATMOUNT_MNT_DEVNAME case... Honza
On Thu, 2024-11-07 at 12:57 +0100, Jan Kara wrote: > On Thu 07-11-24 06:15:40, Jeff Layton wrote: > > On Thu, 2024-11-07 at 10:40 +0100, Jan Kara wrote: > > > On Wed 06-11-24 14:53:06, Jeff Layton wrote: > > > > /proc/self/mountinfo displays the devicename for the mount, but > > > > statmount() doesn't yet have a way to return it. Add a new > > > > STATMOUNT_MNT_DEVNAME flag, claim the 32-bit __spare1 field to hold the > > > > offset into the str[] array. STATMOUNT_MNT_DEVNAME will only be set in > > > > the return mask if there is a device string. > > > > > > > > Signed-off-by: Jeff Layton <jlayton@kernel.org> > > > > > > Just one question below: > > > > > > > @@ -5078,6 +5091,12 @@ static int statmount_string(struct kstatmount *s, u64 flag) > > > > if (seq->count == sm->fs_subtype) > > > > return 0; > > > > break; > > > > + case STATMOUNT_MNT_DEVNAME: > > > > + sm->mnt_devname = seq->count; > > > > + ret = statmount_mnt_devname(s, seq); > > > > + if (seq->count == sm->mnt_devname) > > > > > > Why this odd check? Why don't you rather do: > > > if (ret) > > > ? > > > > > > > statmount_mnt_devname() can return without emitting anything to the seq > > if ->show_devname and r->mnt_devname are both NULL. In that case, we > > don't want statmount_string() to return an error, but we also don't > > want to do any further manipulation of the seq. So, the above handles > > either the case where show_devname returned an error and the case where > > there was nothing to emit. > > > > I did consider having statmount_mnt_devname() return -ENOBUFS if there > > was nothing to emit, and then handle that in the caller, but checking > > to see whether the seq has advanced seemed like a cleaner solution. > > OK, but don't we want to emit an empty string - i.e., just \0 character in > case r->mnt_devname is NULL and there's no ->show_devname? > Yes. I think it's better to just not set STATMOUNT_MNT_DEVNAME at all if it's not present than to emit an empty string. > Because now we > literaly emit nothing and it's going to be very confusing for userspace to > parse this when this happens in the middle of other strings in the seq. > And emitting \0 is exactly what will happen if we don't do anything special > in STATMOUNT_MNT_DEVNAME case... > That's why I'm having statmount_string() return immediately if no string was emitted. That avoids the setting of the flag in the mask and the unneeded NULL termination in the buffer. The caller will need to check whether STATMOUNT_MNT_DEVNAME is set in the returned mask. If it's not set, then no string was emitted and the sm->mnt_devname index is invalid. In that case, we will have emitted nothing into the buffer, so the other strings shouldn't be affected. FWIW, here's the comment I added over that if statement: + /* + * If nothing was emitted, return immediately to avoid + * setting the flag and NULL terminating the buffer. + */ + if (seq->count == sm->mnt_devname) + return ret;
On Wed, 6 Nov 2024 at 20:53, Jeff Layton <jlayton@kernel.org> wrote: > > /proc/self/mountinfo displays the devicename for the mount, but > statmount() doesn't yet have a way to return it. Add a new > STATMOUNT_MNT_DEVNAME flag, claim the 32-bit __spare1 field to hold the > offset into the str[] array. STATMOUNT_MNT_DEVNAME will only be set in > the return mask if there is a device string. > > Signed-off-by: Jeff Layton <jlayton@kernel.org> > --- > fs/namespace.c | 25 ++++++++++++++++++++++++- > include/uapi/linux/mount.h | 3 ++- > 2 files changed, 26 insertions(+), 2 deletions(-) > > diff --git a/fs/namespace.c b/fs/namespace.c > index 52ab892088f08ad71647eff533dd6f3025bbae03..d4ed2cb5de12c86b4da58626441e072fc109b2ff 100644 > --- a/fs/namespace.c > +++ b/fs/namespace.c > @@ -5014,6 +5014,19 @@ static void statmount_fs_subtype(struct kstatmount *s, struct seq_file *seq) > seq_puts(seq, sb->s_subtype); > } > > +static int statmount_mnt_devname(struct kstatmount *s, struct seq_file *seq) > +{ > + struct super_block *sb = s->mnt->mnt_sb; > + struct mount *r = real_mount(s->mnt); > + > + if (sb->s_op->show_devname) > + return sb->s_op->show_devname(seq, s->mnt->mnt_root); I think the resulting string should be unescaped just like statmount_mnt_root(). The same goes for the option strings, which went in last cycle. I see no reason to require users of this interface to implement unescaping themselves. Others beside libmount probably won't do it and will be surprised when encountering escaped strings because they are rare. Thanks, Miklos
On Thu, 2024-11-07 at 14:50 +0100, Miklos Szeredi wrote: > On Wed, 6 Nov 2024 at 20:53, Jeff Layton <jlayton@kernel.org> wrote: > > > > /proc/self/mountinfo displays the devicename for the mount, but > > statmount() doesn't yet have a way to return it. Add a new > > STATMOUNT_MNT_DEVNAME flag, claim the 32-bit __spare1 field to hold the > > offset into the str[] array. STATMOUNT_MNT_DEVNAME will only be set in > > the return mask if there is a device string. > > > > Signed-off-by: Jeff Layton <jlayton@kernel.org> > > --- > > fs/namespace.c | 25 ++++++++++++++++++++++++- > > include/uapi/linux/mount.h | 3 ++- > > 2 files changed, 26 insertions(+), 2 deletions(-) > > > > diff --git a/fs/namespace.c b/fs/namespace.c > > index 52ab892088f08ad71647eff533dd6f3025bbae03..d4ed2cb5de12c86b4da58626441e072fc109b2ff 100644 > > --- a/fs/namespace.c > > +++ b/fs/namespace.c > > @@ -5014,6 +5014,19 @@ static void statmount_fs_subtype(struct kstatmount *s, struct seq_file *seq) > > seq_puts(seq, sb->s_subtype); > > } > > > > +static int statmount_mnt_devname(struct kstatmount *s, struct seq_file *seq) > > +{ > > + struct super_block *sb = s->mnt->mnt_sb; > > + struct mount *r = real_mount(s->mnt); > > + > > + if (sb->s_op->show_devname) > > + return sb->s_op->show_devname(seq, s->mnt->mnt_root); > > I think the resulting string should be unescaped just like statmount_mnt_root(). > > The same goes for the option strings, which went in last cycle. > > I see no reason to require users of this interface to implement > unescaping themselves. Others beside libmount probably won't do it > and will be surprised when encountering escaped strings because they > are rare. Good point. I'll fix that up for the next version. Thanks,
diff --git a/fs/namespace.c b/fs/namespace.c index 52ab892088f08ad71647eff533dd6f3025bbae03..d4ed2cb5de12c86b4da58626441e072fc109b2ff 100644 --- a/fs/namespace.c +++ b/fs/namespace.c @@ -5014,6 +5014,19 @@ static void statmount_fs_subtype(struct kstatmount *s, struct seq_file *seq) seq_puts(seq, sb->s_subtype); } +static int statmount_mnt_devname(struct kstatmount *s, struct seq_file *seq) +{ + struct super_block *sb = s->mnt->mnt_sb; + struct mount *r = real_mount(s->mnt); + + if (sb->s_op->show_devname) + return sb->s_op->show_devname(seq, s->mnt->mnt_root); + + if (r->mnt_devname) + seq_puts(seq, r->mnt_devname); + return 0; +} + static void statmount_mnt_ns_id(struct kstatmount *s, struct mnt_namespace *ns) { s->sm.mask |= STATMOUNT_MNT_NS_ID; @@ -5078,6 +5091,12 @@ static int statmount_string(struct kstatmount *s, u64 flag) if (seq->count == sm->fs_subtype) return 0; break; + case STATMOUNT_MNT_DEVNAME: + sm->mnt_devname = seq->count; + ret = statmount_mnt_devname(s, seq); + if (seq->count == sm->mnt_devname) + return ret; + break; default: WARN_ON_ONCE(true); return -EINVAL; @@ -5220,6 +5239,9 @@ static int do_statmount(struct kstatmount *s, u64 mnt_id, u64 mnt_ns_id, if (!err && s->mask & STATMOUNT_FS_SUBTYPE) err = statmount_string(s, STATMOUNT_FS_SUBTYPE); + if (!err && s->mask & STATMOUNT_MNT_DEVNAME) + err = statmount_string(s, STATMOUNT_MNT_DEVNAME); + if (!err && s->mask & STATMOUNT_MNT_NS_ID) statmount_mnt_ns_id(s, ns); @@ -5241,7 +5263,8 @@ static inline bool retry_statmount(const long ret, size_t *seq_size) } #define STATMOUNT_STRING_REQ (STATMOUNT_MNT_ROOT | STATMOUNT_MNT_POINT | \ - STATMOUNT_FS_TYPE | STATMOUNT_MNT_OPTS | STATMOUNT_FS_SUBTYPE) + STATMOUNT_FS_TYPE | STATMOUNT_MNT_OPTS | \ + STATMOUNT_FS_SUBTYPE | STATMOUNT_MNT_DEVNAME) static int prepare_kstatmount(struct kstatmount *ks, struct mnt_id_req *kreq, struct statmount __user *buf, size_t bufsize, diff --git a/include/uapi/linux/mount.h b/include/uapi/linux/mount.h index fa206fb56b3b25cf80f7d430e1b6bab19c3220e4..0f9ea2748f6376b5e71ba969598fc5b641a2c77f 100644 --- a/include/uapi/linux/mount.h +++ b/include/uapi/linux/mount.h @@ -174,7 +174,7 @@ struct statmount { __u32 mnt_point; /* [str] Mountpoint relative to current root */ __u64 mnt_ns_id; /* ID of the mount namespace */ __u32 fs_subtype; /* [str] Subtype of fs_type (if any) */ - __u32 __spare1[1]; + __u32 mnt_devname; /* [str] Device string for the mount */ __u64 __spare2[48]; char str[]; /* Variable size part containing strings */ }; @@ -210,6 +210,7 @@ struct mnt_id_req { #define STATMOUNT_MNT_NS_ID 0x00000040U /* Want/got mnt_ns_id */ #define STATMOUNT_MNT_OPTS 0x00000080U /* Want/got mnt_opts */ #define STATMOUNT_FS_SUBTYPE 0x00000100U /* Want/got subtype */ +#define STATMOUNT_MNT_DEVNAME 0x00000200U /* Want/got devname */ /* * Special @mnt_id values that can be passed to listmount
/proc/self/mountinfo displays the devicename for the mount, but statmount() doesn't yet have a way to return it. Add a new STATMOUNT_MNT_DEVNAME flag, claim the 32-bit __spare1 field to hold the offset into the str[] array. STATMOUNT_MNT_DEVNAME will only be set in the return mask if there is a device string. Signed-off-by: Jeff Layton <jlayton@kernel.org> --- fs/namespace.c | 25 ++++++++++++++++++++++++- include/uapi/linux/mount.h | 3 ++- 2 files changed, 26 insertions(+), 2 deletions(-)