linux-kernel - Re: [PATCH v5 3.1.0-rc4-tip 3/26] Uprobes: register/unregister probes.

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

Date:	Thu, 6 Oct 2011 12:21:25 +0530
From:	Srikar Dronamraju <srikar@...ux.vnet.ibm.com>
To:	Oleg Nesterov <oleg@...hat.com>
Cc:	Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...e.hu>,
	Steven Rostedt <rostedt@...dmis.org>,
	Linux-mm <linux-mm@...ck.org>,
	Arnaldo Carvalho de Melo <acme@...radead.org>,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	Jonathan Corbet <corbet@....net>,
	Hugh Dickins <hughd@...gle.com>,
	Christoph Hellwig <hch@...radead.org>,
	Masami Hiramatsu <masami.hiramatsu.pt@...achi.com>,
	Thomas Gleixner <tglx@...utronix.de>,
	Andi Kleen <andi@...stfloor.org>,
	LKML <linux-kernel@...r.kernel.org>,
	Jim Keniston <jkenisto@...ux.vnet.ibm.com>,
	Roland McGrath <roland@...k.frob.com>,
	Ananth N Mavinakayanahalli <ananth@...ibm.com>,
	Andrew Morton <akpm@...ux-foundation.org>
Subject: Re: [PATCH v5 3.1.0-rc4-tip 3/26]   Uprobes: register/unregister
 probes.

* Oleg Nesterov <oleg@...hat.com> [2011-10-05 20:50:08]:

> On 10/05, Srikar Dronamraju wrote:
> >
> > Agree. Infact I encountered this problem last week and had fixed it.
> > In mycase, I had mapped the file read and write while trying to insert
> > probes.
> > The changed code looks like this
> >
> > 	if (!vma)
> > 		return NULL;
> 
> This is unneeded, vma_prio_tree_foreach() stops when vma_prio_tree_next()
> returns NULL. IOW, you can never see vma == NULL.

Agree.

> 
> > 	if (!valid_vma(vma))
> > 		continue;
> 
> Yes.
> 
> > > > +	mutex_lock(&inode->i_mutex);
> > > > +	uprobe = alloc_uprobe(inode, offset);
> > >
> > > Looks like, alloc_uprobe() doesn't need ->i_mutex.
> >
> >
> > Actually this was pointed out by you in the last review.
> > https://lkml.org/lkml/2011/7/24/91
> 
> OOPS ;) may be deserves a comment...

will add a comment.

> 
> > > > +void unregister_uprobe(struct inode *inode, loff_t offset,
> > > > +				struct uprobe_consumer *consumer)
> > > > +{
> > > > +	struct uprobe *uprobe;
> > > > +
> > > > +	inode = igrab(inode);
> > > > +	if (!inode || !consumer)
> > > > +		return;
> > > > +
> > > > +	if (offset > inode->i_size)
> > > > +		return;
> > > > +
> > > > +	uprobe = find_uprobe(inode, offset);
> > > > +	if (!uprobe)
> > > > +		return;
> > > > +
> > > > +	if (!del_consumer(uprobe, consumer)) {
> > > > +		put_uprobe(uprobe);
> > > > +		return;
> > > > +	}
> > > > +
> > > > +	mutex_lock(&inode->i_mutex);
> > > > +	if (!uprobe->consumers)
> > > > +		__unregister_uprobe(inode, offset, uprobe);
> > >
> > > It seemes that del_consumer() should be done under ->i_mutex. If it
> > > removes the last consumer, we can race with register_uprobe() which
> > > takes ->i_mutex before us and does another __register_uprobe(), no?
> >
> > We should still be okay, because we check for the consumers before we
> > do the actual unregister in form of __unregister_uprobe.
> > since the consumer is again added by the time we get the lock, we dont
> > do the actual unregistration and go as if del_consumer deleted one
> > consumer but not the last.
> 
> Yes, but I meant in this case register_uprobe() does the unnecessary
> __register_uprobe() because it sees ->consumers == NULL (add_consumer()
> returns NULL).

yes we might be doing an unnecessary __register_uprobe() but because it
raced with unregister_uprobe() and got the lock, we would avoid doing a 
__unregister_uprobe().  

However I am okay to move the lock before del_consumer(). Please let me
know how you prefer this.

> 
> I guess this is probably harmless because of is_bkpt_insn/-EEXIST
> logic, but still.
> 

Agree.

> 
> Btw. __register_uprobe() does
> 
> 		ret = install_breakpoint(mm, uprobe, vma, vi->vaddr);
> 		if (ret && (ret != -ESRCH || ret != -EEXIST)) {
> 			up_read(&mm->mmap_sem);
> 			mmput(mm);
> 			break;
> 		}
> 		ret = 0;
> 		up_read(&mm->mmap_sem);
> 		mmput(mm);
> 
> Yes, this is cosmetic, but why do we duplicate up_read/mmput ?
> 
> Up to you, but
> 
> 		ret = install_breakpoint(mm, uprobe, vma, vi->vaddr);
> 		up_read(&mm->mmap_sem);
> 		mmput(mm);
> 
> 		if (ret) {
> 			if (ret != -ESRCH && ret != -EEXIST)
> 				break;
> 			ret = 0;
> 		}
> 
> Looks a bit simpler.

Okay, will do.

> 
> Oh, wait. I just noticed that the original code does
> 
> 	(ret != -ESRCH || ret != -EEXIST)
> 
> this expression is always true ;)

Right, will correct this.
> 
> Oleg.
> 

-- 
Thanks and Regards
Srikar
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/