From mboxrd@z Thu Jan 1 00:00:00 1970 From: Kirill Tkhai Subject: [PATCH v4] pid_ns: Introduce ioctl to set vector of ns_last_pid's on ns hierarhy Date: Fri, 28 Apr 2017 14:58:44 +0300 Message-ID: <149338066474.18594.2768059055849440055.stgit@localhost.localdomain> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Return-path: Sender: linux-fsdevel-owner@vger.kernel.org To: serge@hallyn.com, ebiederm@xmission.com, agruenba@redhat.com, gregkh@linuxfoundation.org, linux-kernel@vger.kernel.org, oleg@redhat.com, paul@paul-moore.com, ktkhai@virtuozzo.com, viro@zeniv.linux.org.uk, avagin@openvz.org, linux-api@vger.kernel.org, linux-fsdevel@vger.kernel.org, mtk.manpages@gmail.com, akpm@linux-foundation.org, luto@amacapital.net, gorcunov@openvz.org, mingo@kernel.org, keescook@chromium.org List-Id: linux-api@vger.kernel.org On implementing of nested pid namespaces support in CRIU (checkpoint-restore in userspace tool) we run into the situation, that it's impossible to create a task with specific NSpid effectively. After commit 49f4d8b93ccf "pidns: Capture the user namespace and filter ns_last_pid" it is impossible to set ns_last_pid on any pid namespace, except task's active pid_ns (before the commit it was possible to write to pid_ns_for_children). Thus, if a restored task in a container has more than one pid_ns levels, the restorer code must have a task helper for every pid namespace of the task's pid_ns hierarhy. This is a big problem, because of communication with a helper for every pid_ns in the hierarchy is not cheap. It's not performance-good as it implies many helpers wakeups to create a single task (independently, how you communicate with the helpers). This patch tries to decide the problem. It introduces a new pid_ns ioctl(NS_SET_LAST_PID_VEC), which allows to write a vector of last pids on pid_ns hierarchy. The vector is passed as array of pids in struct ns_ioc_pid_vec, written in reverse order. The first number corresponds to the opened namespace ns_last_pid, the second is to its parent, etc. So, if you have the pid namespaces hierarchy like: pid_ns1 (grand father) | v pid_ns2 (father) | v pid_ns3 (child) and the pid_ns3 is open, then the corresponding vector will be {last_ns_pid3, last_ns_pid2, last_ns_pid1}. This vector may be short and it may contain less levels. For example, {last_ns_pid3, last_ns_pid2} or even {last_ns_pid3}, in dependence of which levels you want to populate. v4: Declare struct ns_ioc_pid_vec directly instead of include uapi file. Make the interface independent of CONFIG_CHECKPOINT_RESTORE. Include linux/types.h in nsfs.h for pid_t. Get all vectors at once. Make checks atomical. v3: Use __u32 in uapi instead of unsigned int. v2: Kill pid_ns->child_reaper check as it's impossible to have such a pid namespace file open. Use generic namespaces ioctl() number. Pass pids as array, not as a string. Signed-off-by: Kirill Tkhai --- fs/nsfs.c | 5 +++++ include/linux/pid_namespace.h | 9 ++++++++- include/uapi/linux/nsfs.h | 8 ++++++++ kernel/pid_namespace.c | 36 ++++++++++++++++++++++++++++++++++++ 4 files changed, 57 insertions(+), 1 deletion(-) diff --git a/fs/nsfs.c b/fs/nsfs.c index 323f492e0822..f669a1552003 100644 --- a/fs/nsfs.c +++ b/fs/nsfs.c @@ -6,6 +6,7 @@ #include #include #include +#include #include #include @@ -186,6 +187,10 @@ 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_SET_LAST_PID_VEC: + if (ns->ops->type != CLONE_NEWPID) + return -EINVAL; + return pidns_set_last_pid_vec(ns, (void *)arg); default: return -ENOTTY; } diff --git a/include/linux/pid_namespace.h b/include/linux/pid_namespace.h index c2a989dee876..661fad08bf8c 100644 --- a/include/linux/pid_namespace.h +++ b/include/linux/pid_namespace.h @@ -54,6 +54,7 @@ struct pid_namespace { struct ns_common ns; }; +struct ns_ioc_pid_vec; extern struct pid_namespace init_pid_ns; #define PIDNS_HASH_ADDING (1U << 31) @@ -71,7 +72,8 @@ extern struct pid_namespace *copy_pid_ns(unsigned long flags, extern void zap_pid_ns_processes(struct pid_namespace *pid_ns); extern int reboot_pid_ns(struct pid_namespace *pid_ns, int cmd); extern void put_pid_ns(struct pid_namespace *ns); - +extern long pidns_set_last_pid_vec(struct ns_common *ns, + struct ns_ioc_pid_vec __user *vec); #else /* !CONFIG_PID_NS */ #include @@ -101,6 +103,11 @@ static inline int reboot_pid_ns(struct pid_namespace *pid_ns, int cmd) { return 0; } +static inline long pidns_set_last_pid_vec(struct ns_common *ns, + struct ns_ioc_pid_vec __user *vec) +{ + return -ENOTTY; +} #endif /* CONFIG_PID_NS */ extern struct pid_namespace *task_active_pid_ns(struct task_struct *tsk); diff --git a/include/uapi/linux/nsfs.h b/include/uapi/linux/nsfs.h index 1a3ca79f466b..9d320276eafe 100644 --- a/include/uapi/linux/nsfs.h +++ b/include/uapi/linux/nsfs.h @@ -2,6 +2,7 @@ #define __LINUX_NSFS_H #include +#include #define NSIO 0xb7 @@ -14,5 +15,12 @@ #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) +/* Set a vector of ns_last_pid for a pid namespace stack */ +#define NS_SET_LAST_PID_VEC _IO(NSIO, 0x5) + +struct ns_ioc_pid_vec { + __u32 nr; + pid_t pid[0]; +}; #endif /* __LINUX_NSFS_H */ diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c index de461aa0bf9a..f2aaed6ce0ac 100644 --- a/kernel/pid_namespace.c +++ b/kernel/pid_namespace.c @@ -21,6 +21,7 @@ #include #include #include +#include struct pid_cache { int nr_ids; @@ -428,6 +429,41 @@ static struct ns_common *pidns_get_parent(struct ns_common *ns) return &get_pid_ns(pid_ns)->ns; } +long pidns_set_last_pid_vec(struct ns_common *ns, + struct ns_ioc_pid_vec __user *vec) +{ + struct pid_namespace *pid_ns = to_pid_ns(ns), *top; + pid_t pid[MAX_PID_NS_LEVEL]; + u32 i, nr; + + BUILD_BUG_ON(sizeof(pid_t) * MAX_PID_NS_LEVEL > 128); + if (get_user(nr, &vec->nr)) + return -EFAULT; + if (nr > MAX_PID_NS_LEVEL || nr < 1) + return -EINVAL; + if (copy_from_user(pid, &vec->pid[0], nr * sizeof(pid_t)) != 0) + return -EFAULT; + + top = pid_ns; + for (i = 0; i < nr-1; i++) { + top = top->parent; + if (!top || pid[i] < 0 || pid[i] > pid_max) + return -EINVAL; + } + if (!ns_capable(top->user_ns, CAP_SYS_ADMIN)) + return -EPERM; + if (pid[nr-1] < 0 || pid[nr-1] > pid_max) + return -EINVAL; + + for (i = 0; i < nr; i++) { + /* Write directly: see the comment in pid_ns_ctl_handler() */ + pid_ns->last_pid = pid[i]; + pid_ns = pid_ns->parent; + } + + return 0; +} + static struct user_namespace *pidns_owner(struct ns_common *ns) { return to_pid_ns(ns)->user_ns;