linux-kernel - Re: [PATCH] Convert filldir[64]() from __put_user() to unsafe_put

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]

Message-ID: <20191011001104.GJ26530@ZenIV.linux.org.uk>
Date:   Fri, 11 Oct 2019 01:11:04 +0100
From:   Al Viro <viro@...iv.linux.org.uk>
To:     Linus Torvalds <torvalds@...ux-foundation.org>
Cc:     Guenter Roeck <linux@...ck-us.net>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        linux-fsdevel <linux-fsdevel@...r.kernel.org>
Subject: Re: [PATCH] Convert filldir[64]() from __put_user() to
 unsafe_put_user()

On Thu, Oct 10, 2019 at 03:12:49PM -0700, Linus Torvalds wrote:
> On Thu, Oct 10, 2019 at 12:55 PM Al Viro <viro@...iv.linux.org.uk> wrote:
> >
> > Anyway, another question you way: what do you think of try/catch approaches
> > to __get_user() blocks, like e.g. restore_sigcontext() is doing?
> 
> I'd rather have them converted to our unsafe_get/put_user() instead.
> 
> We don't generate great code for the "get" case (because of how gcc
> doesn't allow us to mix "asm goto" and outputs), but I really despise
> the x86-specific "{get,put}_user_ex()" machinery. It's not actually
> doing a real try/catch at all, and will just keep taking faults if one
> happens.
> 
> But I've not gotten around to rewriting those disgusting sequences to
> the unsafe_get/put_user() model. I did look at it, and it requires
> some changes exactly *because* the _ex() functions are broken and
> continue, but also because the current code ends up also doing other
> things inside the try/catch region that you're not supposed to do in a
> user_access_begin/end() region .

Hmm...  Which one was that?  AFAICS, we have
	do_sys_vm86: only get_user_ex()
	restore_sigcontext(): get_user_ex(), set_user_gs()
	ia32_restore_sigcontext(): get_user_ex()

So at least get_user_try/get_user_ex/get_user_catch should be killable.
The other side...
	save_v86_state(): put_user_ex()
	setup_sigcontext(): put_user_ex()
	__setup_rt_frame(): put_user_ex(), static_cpu_has()
	another one in __setup_rt_frame(): put_user_ex()
	x32_setup_rt_frame(): put_user_ex()
	ia32_setup_sigcontext(): put_user_ex()
	ia32_setup_frame(): put_user_ex()
	another one in ia32_setup_frame(): put_user_ex(), static_cpu_has()

IDGI...  Is static_cpu_has() not allowed in there?  Looks like it's all inlines
and doesn't do any potentially risky memory accesses...  What am I missing?

As for the try/catch model...  How about
	if (!user_access_begin())
		sod off
	...
	unsafe_get_user(..., l);
	...
	unsafe_get_user_nojump();
	...
	unsafe_get_user_nojump();
	...
	if (user_access_did_fail())
		goto l;

	user_access_end()
	...
	return 0;
l:
	...
	user_access_end()
	return -EFAULT;

making it clear that we are delaying the check for failures until it's
more convenient.  And *not* trying to trick C parser into enforcing
anything - let objtool do it and to hell with do { and } while (0) in
magic macros.  Could be mixed with the normal unsafe_..._user() without
any problems, AFAICS...