linux-kernel - Re: [PATCH] fs/proc/kcore.c: Omit kernel text area for hardened usercopy feature

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, 7 Sep 2016 09:32:57 +0200
From:   Jiri Olsa <jolsa@...hat.com>
To:     Kees Cook <keescook@...omium.org>
Cc:     Andi Kleen <andi@...stfloor.org>, Jiri Olsa <jolsa@...nel.org>,
        lkml <linux-kernel@...r.kernel.org>,
        Ingo Molnar <mingo@...nel.org>,
        Adrian Hunter <adrian.hunter@...el.com>,
        KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>,
        Linus Torvalds <torvalds@...ux-foundation.org>
Subject: Re: [PATCH] fs/proc/kcore.c: Omit kernel text area for hardened
 usercopy feature

On Tue, Sep 06, 2016 at 01:56:40PM -0400, Kees Cook wrote:

SNIP

> >  static __must_check __always_inline int
> > diff --git a/fs/proc/kcore.c b/fs/proc/kcore.c
> > index a939f5ed7f89..c7a22a8a157e 100644
> > --- a/fs/proc/kcore.c
> > +++ b/fs/proc/kcore.c
> > @@ -516,7 +516,7 @@ read_kcore(struct file *file, char __user *buffer, size_t buflen, loff_t *fpos)
> >                         if (kern_addr_valid(start)) {
> >                                 unsigned long n;
> >
> > -                               n = copy_to_user(buffer, (char *)start, tsz);
> > +                               n = copy_to_user_nocheck(buffer, (char *)start, tsz);
> >                                 /*
> >                                  * We cannot distinguish between fault on source
> >                                  * and fault on destination. When this happens
> 
> This patch is x86-specific (but ARCH_PROC_KCORE_TEXT is on multiple
> architectures), which I don't think we want. Instead, let's get the
> usercopy helper code centralized (Al Viro is looking at this already),
> and then we can design arch-agnostic methods to handle this.
> 
> In the meantime, how about continuing to use a bounce buffer like
> already done in the vmalloc_or_module_addr() case immediately above?

ok, sounds good.. so something like below? untested

thanks,
jirka


---
diff --git a/fs/proc/kcore.c b/fs/proc/kcore.c
index a939f5ed7f89..de07c273f725 100644
--- a/fs/proc/kcore.c
+++ b/fs/proc/kcore.c
@@ -515,8 +515,20 @@ read_kcore(struct file *file, char __user *buffer, size_t buflen, loff_t *fpos)
 		} else {
 			if (kern_addr_valid(start)) {
 				unsigned long n;
+				char *buf;
 
-				n = copy_to_user(buffer, (char *)start, tsz);
+				buf = kzalloc(tsz, GFP_KERNEL);
+				if (!buf)
+					return -ENOMEM;
+
+				/*
+				 * Using bounce buffer to bypass the hardened
+				 * user copy kernel text checks.
+				 */
+				memcpy(buf, (char *) start, tsz);
+
+				n = copy_to_user(buffer, buf, tsz);
+				kfree(buf);
 				/*
 				 * We cannot distinguish between fault on source
 				 * and fault on destination. When this happens