linux-kernel - Re: [PATCH -next] fork.c: Move check of clone NEWIPC and SYSVSEM to copy

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]

Date:   Fri, 1 Dec 2017 09:19:21 +0100
From:   Michal Hocko <mhocko@...nel.org>
To:     Marcos Paulo de Souza <marcos.souza.org@...il.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>,
        Ingo Molnar <mingo@...nel.org>, Rik van Riel <riel@...hat.com>,
        Stephen Rothwell <sfr@...b.auug.org.au>,
        "Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>,
        Jiri Olsa <jolsa@...nel.org>,
        Hari Bathini <hbathini@...ux.vnet.ibm.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Arnaldo Carvalho de Melo <acme@...hat.com>,
        "Eric W. Biederman" <ebiederm@...ssion.com>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH -next] fork.c: Move check of clone NEWIPC and SYSVSEM to
 copy_process

On Wed 29-11-17 22:33:43, Marcos Paulo de Souza wrote:
> On Thu, Nov 30, 2017 at 11:04:06AM +0100, Michal Hocko wrote:
> > CC Eric
> > 
> > On Sun 26-11-17 14:06:52, Marcos Paulo de Souza wrote:
> > > Currently this check for CLONE_NEWIPC with CLONE_SYSVSEM is done inside
> > > copy_namespaces, resulting in a handful of error paths being executed if
> > > these flags were used together. So, move this check to the beginning of
> > > copy_process, exiting earlier if the condition is true.
> > > 
> > > This move is safe because copy_namespaces is called just from
> > > copy_process function.
> 
> This change is introduced right below the point where clone_flags is already
> checking for inconsistencies in namespace flags[1], and returns EINVAL when
> conflicting flags are informed together.
> 
> In this case, it's easier to return early when conflicting flags are informed at
> the beginning, so moving a namespace check to where namespaces are already being
> sanitized makes sense.

What doe easier mean in this context?

> If the code stays where it is now, and a user calls clone
> syscalls informing CLONE_NEWIPC | CLONE_SYSVSEM, the code will need to undo a
> lot of work before returning the same EINVAL[2].

And why do we care about that error path? Does it trigger that often?

Do not take me wrong. I am not saying the patch is incorrect or wrong.
The changelog lacks the justification and explanation _why_ do we care
at all. A rarely failing path reorganization doesn't sound too
interesting to me. I cannot judge the cleanup aspect of this though but
from a very naive diffstat POV it doesn't look like a huge win to me.
 
> [1] https://elixir.free-electrons.com/linux/latest/source/kernel/fork.c#L1552
> [2] https://elixir.free-electrons.com/linux/latest/source/kernel/fork.c#L1953
> 
> > 
> > I am not familiar with the code all that much but the justification is
> > not clear to me. Thesea re namespace related flags so why should we pull
> > them out of copy_namespaces. I do not see any simplifications in the
> > error code paths or something like that.
> > 
> > > Signed-off-by: Marcos Paulo de Souza <marcos.souza.org@...il.com>
> > > ---
> > >  kernel/fork.c    | 11 +++++++++++
> > >  kernel/nsproxy.c | 11 -----------
> > >  2 files changed, 11 insertions(+), 11 deletions(-)
> > > 
> > > diff --git a/kernel/fork.c b/kernel/fork.c
> > > index 2113e252cb9d..691f9ba135fc 100644
> > > --- a/kernel/fork.c
> > > +++ b/kernel/fork.c
> > > @@ -1600,6 +1600,17 @@ static __latent_entropy struct task_struct *copy_process(
> > >  		return ERR_PTR(-EINVAL);
> > >  
> > >  	/*
> > > +	 * CLONE_NEWIPC must detach from the undolist: after switching
> > > +	 * to a new ipc namespace, the semaphore arrays from the old
> > > +	 * namespace are unreachable.  In clone parlance, CLONE_SYSVSEM
> > > +	 * means share undolist with parent, so we must forbid using
> > > +	 * it along with CLONE_NEWIPC.
> > > +	 */
> > > +	if ((clone_flags & (CLONE_NEWIPC | CLONE_SYSVSEM)) ==
> > > +		(CLONE_NEWIPC | CLONE_SYSVSEM))
> > > +		return ERR_PTR(-EINVAL);
> > > +
> > > +	/*
> > >  	 * Thread groups must share signals as well, and detached threads
> > >  	 * can only be started up within the thread group.
> > >  	 */
> > > diff --git a/kernel/nsproxy.c b/kernel/nsproxy.c
> > > index f6c5d330059a..30882727dff5 100644
> > > --- a/kernel/nsproxy.c
> > > +++ b/kernel/nsproxy.c
> > > @@ -151,17 +151,6 @@ int copy_namespaces(unsigned long flags, struct task_struct *tsk)
> > >  	if (!ns_capable(user_ns, CAP_SYS_ADMIN))
> > >  		return -EPERM;
> > >  
> > > -	/*
> > > -	 * CLONE_NEWIPC must detach from the undolist: after switching
> > > -	 * to a new ipc namespace, the semaphore arrays from the old
> > > -	 * namespace are unreachable.  In clone parlance, CLONE_SYSVSEM
> > > -	 * means share undolist with parent, so we must forbid using
> > > -	 * it along with CLONE_NEWIPC.
> > > -	 */
> > > -	if ((flags & (CLONE_NEWIPC | CLONE_SYSVSEM)) ==
> > > -		(CLONE_NEWIPC | CLONE_SYSVSEM)) 
> > > -		return -EINVAL;
> > > -
> > >  	new_ns = create_new_namespaces(flags, tsk, user_ns, tsk->fs);
> > >  	if (IS_ERR(new_ns))
> > >  		return  PTR_ERR(new_ns);
> > > -- 
> > > 2.13.6
> > > 
> > 
> > -- 
> > Michal Hocko
> > SUSE Labs
> 
> -- 
> Thanks,
> 	Marcos

-- 
Michal Hocko
SUSE Labs