linux-kernel - Re: [PATCHv4 0/3] Kernel Live Patching

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:	Wed, 26 Nov 2014 09:27:59 -0600
From:	Josh Poimboeuf <jpoimboe@...hat.com>
To:	Jiri Kosina <jkosina@...e.cz>
Cc:	Masami Hiramatsu <masami.hiramatsu.pt@...achi.com>,
	Seth Jennings <sjenning@...hat.com>,
	Vojtech Pavlik <vojtech@...e.cz>,
	Steven Rostedt <rostedt@...dmis.org>,
	Petr Mladek <pmladek@...e.cz>, Miroslav Benes <mbenes@...e.cz>,
	Christoph Hellwig <hch@...radead.org>,
	Greg KH <gregkh@...uxfoundation.org>,
	Andy Lutomirski <luto@...capital.net>,
	live-patching@...r.kernel.org, x86@...nel.org, kpatch@...hat.com,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCHv4 0/3] Kernel Live Patching

On Wed, Nov 26, 2014 at 10:18:24AM +0100, Jiri Kosina wrote:
> On Wed, 26 Nov 2014, Masami Hiramatsu wrote:
> 
> > > Note to Steve:
> > > Masami's IPMODIFY patch is heading for -next via your tree.  Once it arrives,
> > > I'll rebase and make the change to set IPMODIFY.  Do not pull this for -next
> > > yet.  This version (v4) is for review and gathering acks.
> > 
> > BTW, as we discussed IPMODIFY is an exclusive flag. So if we allocate 
> > ftrace_ops for each function in each patch, it could be conflict each 
> > other.
> 
> Yup, this corresponds to what Petr brought up yesterday. There are cases 
> where all solutions (kpatch, kgraft, klp) would allocate multiple 
> ftrace_ops for a single function entry (think of patching one function 
> multiple times in a row).
> 
> So it's not as easy as just setting the flag.
> 
> > Maybe we need to have another ops hashtable to find such conflict and 
> > new handler to handle it.
> 
> If I understand your proposal correctly, that would sound like a hackish 
> workaround, trying to basically trick the IPMODIFY flag semantics you just 
> implemented :)

I think Masami may be proposing something similar to what we do in
kpatch today.  We have a single ftrace_ops and handler which is used for
all functions.  The handler accesses a global hash of kpatch_func
structs which is indexed by the original function's IP address.

It actually works out pretty well because it nicely encapsulates the
knowledge about which functions are patched in a single place.  And it
makes it easy to track function versions (for incremental patching and
rollback).

> What I'd propose instead is to make sure that we always have 
> just a ftrace_ops per function entry, and only update the pointers there 
> as necessary. Fortunately we can do the switch atomically, by making use 
> of ->private.

But how would you update multiple functions atomically, to enforce
per-thread consistency?

-- 
Josh
--
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/