On 2025-09-11, Christian Brauner <[email protected]> wrote: > On Thu, Sep 11, 2025 at 02:49:49AM +1000, Aleksa Sarai wrote: > > On 2025-09-10, Christian Brauner <[email protected]> wrote: > > > The mount namespace has supported id retrieval for a while already. > > > Add support for the other types as well. > > > > > > Signed-off-by: Christian Brauner <[email protected]> > > > --- > > > fs/nsfs.c | 74 > > > +++++++++++++++++++++++++++++++++++++++-------- > > > include/uapi/linux/nsfs.h | 12 ++++++-- > > > 2 files changed, 72 insertions(+), 14 deletions(-) > > > > > > diff --git a/fs/nsfs.c b/fs/nsfs.c > > > index 3c6fcf652633..527480e67fd1 100644 > > > --- a/fs/nsfs.c > > > +++ b/fs/nsfs.c > > > @@ -173,6 +173,13 @@ static bool nsfs_ioctl_valid(unsigned int cmd) > > > case NS_GET_NSTYPE: > > > case NS_GET_OWNER_UID: > > > case NS_GET_MNTNS_ID: > > > + case NS_GET_NETNS_ID: > > > + case NS_GET_CGROUPNS_ID: > > > + case NS_GET_IPCNS_ID: > > > + case NS_GET_UTSNS_ID: > > > + case NS_GET_PIDNS_ID: > > > + case NS_GET_TIMENS_ID: > > > + case NS_GET_USERNS_ID: > > > case NS_GET_PID_FROM_PIDNS: > > > case NS_GET_TGID_FROM_PIDNS: > > > case NS_GET_PID_IN_PIDNS: > > > @@ -226,18 +233,6 @@ static long ns_ioctl(struct file *filp, unsigned int > > > ioctl, > > > argp = (uid_t __user *) arg; > > > uid = from_kuid_munged(current_user_ns(), user_ns->owner); > > > return put_user(uid, argp); > > > - case NS_GET_MNTNS_ID: { > > > - __u64 __user *idp; > > > - __u64 id; > > > - > > > - if (ns->ops->type != CLONE_NEWNS) > > > - return -EINVAL; > > > - > > > - mnt_ns = container_of(ns, struct mnt_namespace, ns); > > > - idp = (__u64 __user *)arg; > > > - id = mnt_ns->ns.ns_id; > > > - return put_user(id, idp); > > > - } > > > case NS_GET_PID_FROM_PIDNS: > > > fallthrough; > > > case NS_GET_TGID_FROM_PIDNS: > > > @@ -283,6 +278,61 @@ static long ns_ioctl(struct file *filp, unsigned int > > > ioctl, > > > ret = -ESRCH; > > > return ret; > > > } > > > + case NS_GET_MNTNS_ID: > > > + fallthrough; > > > + case NS_GET_NETNS_ID: > > > + fallthrough; > > > + case NS_GET_CGROUPNS_ID: > > > + fallthrough; > > > + case NS_GET_IPCNS_ID: > > > + fallthrough; > > > + case NS_GET_UTSNS_ID: > > > + fallthrough; > > > + case NS_GET_PIDNS_ID: > > > + fallthrough; > > > + case NS_GET_TIMENS_ID: > > > + fallthrough; > > > + case NS_GET_USERNS_ID: { > > > + __u64 __user *idp; > > > + __u64 id; > > > + int expected_type; > > > + > > > + switch (ioctl) { > > > + case NS_GET_MNTNS_ID: > > > + expected_type = CLONE_NEWNS; > > > + break; > > > + case NS_GET_NETNS_ID: > > > + expected_type = CLONE_NEWNET; > > > + break; > > > + case NS_GET_CGROUPNS_ID: > > > + expected_type = CLONE_NEWCGROUP; > > > + break; > > > + case NS_GET_IPCNS_ID: > > > + expected_type = CLONE_NEWIPC; > > > + break; > > > + case NS_GET_UTSNS_ID: > > > + expected_type = CLONE_NEWUTS; > > > + break; > > > + case NS_GET_PIDNS_ID: > > > + expected_type = CLONE_NEWPID; > > > + break; > > > + case NS_GET_TIMENS_ID: > > > + expected_type = CLONE_NEWTIME; > > > + break; > > > + case NS_GET_USERNS_ID: > > > + expected_type = CLONE_NEWUSER; > > > + break; > > > + default: > > > + return -EINVAL; > > > + } > > > + > > > + if (ns->ops->type != expected_type) > > > + return -EINVAL; > > > > While I get that having this be per-ns-type lets programs avoid being > > tricked into thinking that one namespace ID is actually another > > namespace, it feels a bit ugly to have to add a new ioctl for every new > > namespace. > > > > If we added a way to get the CLONE_* flag for a namespace (NS_GET_TYPE) > > That exists afaict: NS_GET_NSTYPE.
D'oh, yeah that's all you need.
> > we could have just NS_GET_ID. Of course, we would have to trust
> > userspace to do the right thing...
>
> So NS_GET_ID can just return the id and be done with it. If userspace
> wants to know what type it is they can issue a separate ioctl. But since
> the id space is shared all ids of all namespaces can be compared with
> each other reliably. So really for comparision you wouldn't need to
> care. IOW, yes.
Ah, I didn't realise they're all in the same id-space -- in that case it
makes even more sense to just have a single NS_GET_ID IMHO.
> > > +
> > > + idp = (__u64 __user *)arg;
> > > + id = ns->ns_id;
> > > + return put_user(id, idp);
> > > + }
> > > }
> > >
> > > /* extensible ioctls */
> > > diff --git a/include/uapi/linux/nsfs.h b/include/uapi/linux/nsfs.h
> > > index 97d8d80d139f..f7c21840cc09 100644
> > > --- a/include/uapi/linux/nsfs.h
> > > +++ b/include/uapi/linux/nsfs.h
> > > @@ -16,8 +16,6 @@
> > > #define NS_GET_NSTYPE _IO(NSIO, 0x3)
> > > /* Get owner UID (in the caller's user namespace) for a user namespace */
> > > #define NS_GET_OWNER_UID _IO(NSIO, 0x4)
> > > -/* Get the id for a mount namespace */
> > > -#define NS_GET_MNTNS_ID _IOR(NSIO, 0x5, __u64)
> > > /* Translate pid from target pid namespace into the caller's pid
> > > namespace. */
> > > #define NS_GET_PID_FROM_PIDNS _IOR(NSIO, 0x6, int)
> > > /* Return thread-group leader id of pid in the callers pid namespace. */
> > > @@ -42,6 +40,16 @@ struct mnt_ns_info {
> > > /* Get previous namespace. */
> > > #define NS_MNT_GET_PREV _IOR(NSIO, 12, struct mnt_ns_info)
> > >
> > > +/* Retrieve namespace identifiers. */
> > > +#define NS_GET_MNTNS_ID _IOR(NSIO, 5, __u64)
> > > +#define NS_GET_NETNS_ID _IOR(NSIO, 13, __u64)
> > > +#define NS_GET_CGROUPNS_ID _IOR(NSIO, 14, __u64)
> > > +#define NS_GET_IPCNS_ID _IOR(NSIO, 15, __u64)
> > > +#define NS_GET_UTSNS_ID _IOR(NSIO, 16, __u64)
> > > +#define NS_GET_PIDNS_ID _IOR(NSIO, 17, __u64)
> > > +#define NS_GET_TIMENS_ID _IOR(NSIO, 18, __u64)
> > > +#define NS_GET_USERNS_ID _IOR(NSIO, 19, __u64)
> > > +
> > > enum init_ns_ino {
> > > IPC_NS_INIT_INO = 0xEFFFFFFFU,
> > > UTS_NS_INIT_INO = 0xEFFFFFFEU,
> > >
> > > --
> > > 2.47.3
> > >
> >
> > --
> > Aleksa Sarai
> > Senior Software Engineer (Containers)
> > SUSE Linux GmbH
> > https://www.cyphar.com/
>
>
--
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
https://www.cyphar.com/
signature.asc
Description: PGP signature
