linux-kernel - Re: [patch] clone.2: Add EACCES with CLONE_INTO

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20210830073801.tbbxhnw3dg5saxt7@wittgenstein>
Date:   Mon, 30 Aug 2021 09:38:01 +0200
From:   Christian Brauner <christian.brauner@...ntu.com>
To:     Andrew Wock <ajwock@...il.com>
Cc:     linux-man@...r.kernel.org, alx.manpages@...il.com,
        mtk.manpages@...il.com, christian@...uner.io,
        linux-kernel@...r.kernel.org
Subject: Re: [patch] clone.2: Add EACCES with CLONE_INTO_CGROUP + clone3 to
 ERRORS

On Sun, Aug 29, 2021 at 03:57:06PM -0400, Andrew Wock wrote:
> Resending because it's my first time mailing the lkml and I used html.
> Reattempting w/ gmail's plaintext mode.  I apologise if this is
> reaching you twice.
> 
> I noticed that clone3 can send the EACCES errno after I wrote a
> program that used clone3 with the CLONE_INTO_CGROUP flag.  To me, it's
> important to know what kind of failure occurred if the clone3 fails,
> so I was glad that a unique errno is set for this case, but it wasn't
> documented on the clone man page.

In essence, any error that could occur during regular fs-based migration
at write-time can also occur during CLONE_INTO_CGROUP. The clone3()
manpage just has the inverse of that above statement:

"Note that all of the usual restrictions (described in cgroups(7)) on
placing a process into a version 2 cgroup apply."

> 
> I've attached a patch and a test program.
> 
> Test program is attached as clone3_doc.c.  Create
> /sys/fs/cgroup/not-allowed as root, then run the program.  It should
> set errno to EACCES.

This is a manpage update, right? In that case it's not necessarily
needed to Cc lkml aka linux-kernel@... itself.

For the content:
Acked-by: Christian Brauner <christian.brauner@...ntu.com>

(I have no idea what patch format Michael will accept so I can't really
ack that. :))

> 
> Thanks,
> Andrew Wock

> #include <stdio.h>
> #include <errno.h>
> #include <stdlib.h>
> #include <string.h>
> #include <signal.h>
> #include <fcntl.h>
> 
> #include <linux/sched.h>    /* Definition of struct clone_args */
> #include <sched.h>          /* Definition of CLONE_* constants */
> #include <sys/syscall.h>    /* Definition of SYS_* constants */
> #include <unistd.h>
> 
> /*
>  * Preconditions:
>  * - /sys/fs/cgroup/not-allowed is a real cgroup.
>  * - You are not root and do not have write permissions to
>  *   /sys/fs/cgroup/not-allowed/cgroup.procs
>  */
> int main() {
>   pid_t pid;
>   int fd;
>   struct clone_args cl_args = {0};
>   char *cgPath = "/sys/fs/cgroup/not-allowed";
> 
>   fd = open(cgPath, O_RDONLY);
>   if (fd == -1) {
>     fprintf(stderr, "Could not open cgroup %s: %s\n", cgPath, strerror(errno));
>     exit(1);
>   }
> 
>   cl_args.exit_signal = SIGCHLD;
>   cl_args.flags = CLONE_INTO_CGROUP;
>   cl_args.cgroup = fd;
>   pid = syscall(SYS_clone3, &cl_args, sizeof(cl_args));
>   if (pid == -1) {
>     if (errno == EACCES) {
>       printf("EACCES detected\n");
>       exit(0);
>     }
>     fprintf(stderr, "Could not clone into cgroup: %s\n", strerror(errno));
>   } else if (pid == 0) {
>     fprintf(stderr, "Are you root, or do you have write access to %s?\n", cgPath);
>   }
>   exit(1);
> }