LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Andrei Vagin <avagin@virtuozzo.com>
To: David Howells <dhowells@redhat.com>
Cc: viro@zeniv.linux.org.uk, linux-nfs@vger.kernel.org,
	linux-kernel@vger.kernel.org,
	linux-security-module@vger.kernel.org,
	linux-fsdevel@vger.kernel.org, linux-afs@lists.infradead.org
Subject: Re: [12/24] proc: Add fs_context support to procfs [ver #7]
Date: Mon, 25 Jun 2018 23:13:22 -0700	[thread overview]
Message-ID: <20180626061320.GA12548@outlook.office365.com> (raw)
In-Reply-To: <20180619033450.GA11639@outlook.office365.com>

On Mon, Jun 18, 2018 at 08:34:50PM -0700, Andrei Vagin wrote:
> Hi David,
> 
> We run CRIU tests for vfs/for-next, and today a few of these test failed. I
> found that the problem appears after this patch..
> 
> >  int pid_ns_prepare_proc(struct pid_namespace *ns)
> >  {
> > +	struct proc_fs_context *ctx;
> > +	struct fs_context *fc;
> >  	struct vfsmount *mnt;
> > +	int ret;
> > +
> > +	fc = vfs_new_fs_context(&proc_fs_type, NULL, 0,
> > +				FS_CONTEXT_FOR_KERNEL_MOUNT);
> > +	if (IS_ERR(fc))
> > +		return PTR_ERR(fc);
> > +
> > +	ctx = container_of(fc, struct proc_fs_context, fc);
> > +	if (ctx->pid_ns != ns) {
> > +		put_pid_ns(ctx->pid_ns);
> > +		get_pid_ns(ns);
> > +		ctx->pid_ns = ns;
> > +	}
> > +
> > +	ret = vfs_get_tree(fc);
> > +	if (ret < 0) {
> > +		put_fs_context(fc);
> > +		return ret;
> > +	}
> >  
> > -	mnt = kern_mount_data(&proc_fs_type, ns, 0);

Here ns->user_ns and get_current_cred()->user_ns are not always equal

> > +	mnt = vfs_create_mount(fc);
> > +	put_fs_context(fc);
> >  	if (IS_ERR(mnt))
> >  		return PTR_ERR(mnt);
> >  

> #define _GNU_SOURCE
> #include <sys/types.h>
> #include <sched.h>
> #include <unistd.h>
> #include <stdio.h>
> #include <sys/mount.h>
> #include <sys/wait.h>
> #include <sys/stat.h>
> #include <fcntl.h>
> #include <stdlib.h>
> #include <grp.h>
> #include <linux/limits.h>
> 
> 
> #define NS_STACK_SIZE	4096
> 
> #define __stack_aligned__	__attribute__((aligned(16)))
> 
> /* All arguments should be above stack, because it grows down */
> struct ns_exec_args {
> 	char stack[NS_STACK_SIZE] __stack_aligned__;
> 	char stack_ptr[0];
> 	int pfd[2];
> };
> 
> static int ns_exec(void *_arg)
> {
> 	struct ns_exec_args *args = (struct ns_exec_args *) _arg;
> 	int ret;
> 
> 	close(args->pfd[1]);
> 	if (read(args->pfd[0], &ret, sizeof(ret)) != sizeof(ret))
> 		return -1;
> 
> 	setsid();
> 
> 	if (setuid(0) || setgid(0) || setgroups(0, NULL)) {
> 		fprintf(stderr, "set*id failed: %m\n");
> 		return -1;
> 	}
> 
> 	if (mount("proc", "/mnt", "proc", MS_MGC_VAL | MS_NOSUID | MS_NOEXEC | MS_NODEV, NULL)) {
> 		fprintf(stderr, "mount(/proc) failed: %m\n");
> 		return -1;
> 	}
> 
> 	return 0;
> }
> 
> #define UID_MAP "0 100000 100000\n100000 200000 50000"
> #define GID_MAP "0 400000 50000\n50000 500000 100000"
> int main()
> {
> 	pid_t pid;
> 	int ret, status;
> 	struct ns_exec_args args;
> 	int flags;
> 	char pname[PATH_MAX];
> 	int fd, pfd[2];
> 
> 	if (pipe(pfd))
> 		return 1;
> 
> 	args.pfd[0] = pfd[0];
> 	args.pfd[1] = pfd[1];
> 
> 	flags = CLONE_NEWPID | CLONE_NEWNS | CLONE_NEWUTS |
> 		CLONE_NEWNET | CLONE_NEWIPC | CLONE_NEWUSER | SIGCHLD;
> 
> 	pid = clone(ns_exec, args.stack_ptr, flags, &args);
> 	if (pid < 0) {
> 		fprintf(stderr, "clone() failed: %m\n");
> 		exit(1);
> 	}
> 
> 
> 	snprintf(pname, sizeof(pname), "/proc/%d/uid_map", pid);
> 	fd = open(pname, O_WRONLY);
> 	if (fd < 0) {
> 		fprintf(stderr, "open(%s): %m\n", pname);
> 		exit(1);
> 	}
> 	if (write(fd, UID_MAP, sizeof(UID_MAP)) < 0) {
> 		fprintf(stderr, "write(" UID_MAP "): %m\n");
> 		exit(1);
> 	}
> 	close(fd);
> 
> 	snprintf(pname, sizeof(pname), "/proc/%d/gid_map", pid);
> 	fd = open(pname, O_WRONLY);
> 	if (fd < 0) {
> 		fprintf(stderr, "open(%s): %m\n", pname);
> 		exit(1);
> 	}
> 	if (write(fd, GID_MAP, sizeof(GID_MAP)) < 0) {
> 		fprintf(stderr, "write(" GID_MAP "): %m\n");
> 		exit(1);
> 	}
> 	close(fd);
> 
> 	if (write(pfd[1], &ret, sizeof(ret)) != sizeof(ret))
> 		return 1;
> 
> 	if (waitpid(pid, &status, 0) != pid)
> 		return 1;
> 	if (status)
> 		return 1;
> 
> 	return 0;
> }


  reply	other threads:[~2018-06-26  6:13 UTC|newest]

Thread overview: 40+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-04-19 13:31 [PATCH 00/24] VFS: Introduce filesystem context " David Howells
2018-04-19 13:31 ` [PATCH 01/24] vfs: Undo an overly zealous MS_RDONLY -> SB_RDONLY conversion " David Howells
2018-04-19 13:31 ` [PATCH 02/24] VFS: Suppress MS_* flag defs within the kernel unless explicitly enabled " David Howells
2018-04-19 13:31 ` [PATCH 03/24] VFS: Introduce the structs and doc for a filesystem context " David Howells
2018-04-23  3:36   ` Randy Dunlap
2018-05-01 14:29   ` David Howells
2018-05-01 15:31     ` Randy Dunlap
2018-04-19 13:31 ` [PATCH 04/24] VFS: Add LSM hooks for " David Howells
2018-04-19 20:32   ` Paul Moore
2018-04-20 15:35   ` David Howells
2018-04-23 13:25     ` Stephen Smalley
2018-04-24 15:22     ` David Howells
2018-04-25 14:07       ` Stephen Smalley
2018-04-19 13:31 ` [PATCH 05/24] apparmor: Implement security hooks for the new mount API " David Howells
2018-05-04  0:10   ` John Johansen
2018-05-11 12:20   ` David Howells
2018-04-19 13:31 ` [PATCH 06/24] tomoyo: " David Howells
2018-04-19 13:31 ` [PATCH 07/24] smack: Implement filesystem context security hooks " David Howells
2018-04-19 13:31 ` [PATCH 08/24] VFS: Require specification of size of mount data for internal mounts " David Howells
2018-04-19 13:32 ` [PATCH 09/24] VFS: Implement a filesystem superblock creation/configuration context " David Howells
2018-04-19 13:32 ` [PATCH 10/24] VFS: Remove unused code after filesystem context changes " David Howells
2018-04-19 13:32 ` [PATCH 11/24] procfs: Move proc_fill_super() to fs/proc/root.c " David Howells
2018-04-19 13:32 ` [PATCH 12/24] proc: Add fs_context support to procfs " David Howells
2018-06-19  3:34   ` [12/24] " Andrei Vagin
2018-06-26  6:13     ` Andrei Vagin [this message]
2018-06-26  7:27       ` Andrei Vagin
2018-06-26  8:57       ` David Howells
2018-06-28  5:50         ` Andrei Vagin
2018-04-19 13:32 ` [PATCH 13/24] ipc: Convert mqueue fs to fs_context " David Howells
2018-04-19 13:32 ` [PATCH 14/24] cpuset: Use " David Howells
2018-04-19 13:32 ` [PATCH 15/24] kernfs, sysfs, cgroup, intel_rdt: Support " David Howells
2018-04-19 13:33 ` [PATCH 16/24] hugetlbfs: Convert to " David Howells
2018-04-19 13:33 ` [PATCH 17/24] VFS: Remove kern_mount_data() " David Howells
2018-04-19 13:33 ` [PATCH 18/24] VFS: Implement fsopen() to prepare for a mount " David Howells
2018-04-19 13:33 ` [PATCH 19/24] VFS: Implement fsmount() to effect a pre-configured " David Howells
2018-04-19 13:33 ` [PATCH 20/24] afs: Fix server record deletion " David Howells
2018-04-19 13:33 ` [PATCH 21/24] net: Export get_proc_net() " David Howells
2018-04-19 13:33 ` [PATCH 22/24] afs: Add fs_context support " David Howells
2018-04-19 13:33 ` [PATCH 23/24] afs: Implement namespacing " David Howells
2018-04-19 13:33 ` [PATCH 24/24] afs: Use fs_context to pass parameters over automount " David Howells

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20180626061320.GA12548@outlook.office365.com \
    --to=avagin@virtuozzo.com \
    --cc=dhowells@redhat.com \
    --cc=linux-afs@lists.infradead.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-nfs@vger.kernel.org \
    --cc=linux-security-module@vger.kernel.org \
    --cc=viro@zeniv.linux.org.uk \
    --subject='Re: [12/24] proc: Add fs_context support to procfs [ver #7]' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).