linux-kernel - Re: [PATCH v2] x86/mm/fault: Allow stack access below %rsp

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: <7788f826-1217-8f69-3622-c11b6e08efb1@redhat.com>
Date:   Tue, 6 Nov 2018 15:14:15 -0500
From:   Waiman Long <longman@...hat.com>
To:     Dave Hansen <dave.hansen@...ux.intel.com>,
        Andy Lutomirski <luto@...nel.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>
Cc:     "H. Peter Anvin" <hpa@...or.com>, x86@...nel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] x86/mm/fault: Allow stack access below %rsp

On 11/06/2018 03:12 PM, Waiman Long wrote:
> The current x86 page fault handler allows stack access below the stack
> pointer if it is no more than 64k+256 bytes. Any access beyond the 64k+
> limit will cause a segmentation fault.
>
> The gcc -fstack-check option generates code to probe the stack for
> large stack allocation to see if the stack is accessible. The newer gcc
> does that while updating the %rsp simultaneously. Older gcc's like gcc4
> doesn't do that. As a result, an application compiled with an old gcc
> and the -fstack-check option may fail to start at all.
>
> % cat test.c
> int main() {
> 	char tmp[1024*128];
> 	printf("### ok\n");
> 	return 0;
> }
> % gcc -fstack-check -g -o test test.c
> % ./test
> Segmentation fault
>
> The old binary was working in older kernels where expand_stack() was
> somehow called before the check. But it is not working in newer kernels.
> Besides, the 64k+ limit check is kind of crude and will not catch a
> lot of mistakes that userspace applications may be misbehaving anyway.
> I think the kernel isn't the right place for this kind of tests. We
> should leave it to userspace instrumentation tools to perform them.
>
> The 64k+ limit check is now removed to just let expand_stack() decide
> if a segmentation fault should happen, when the RLIMIT_STACK limit is
> exceeded, for example.
>
> Signed-off-by: Waiman Long <longman@...hat.com>
> ---
>  arch/x86/mm/fault.c | 12 ------------
>  1 file changed, 12 deletions(-)
>
> diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
> index 71d4b9d..29525cf 100644
> --- a/arch/x86/mm/fault.c
> +++ b/arch/x86/mm/fault.c
> @@ -1380,18 +1380,6 @@ void do_user_addr_fault(struct pt_regs *regs,
>  		bad_area(regs, sw_error_code, address);
>  		return;
>  	}
> -	if (sw_error_code & X86_PF_USER) {
> -		/*
> -		 * Accessing the stack below %sp is always a bug.
> -		 * The large cushion allows instructions like enter
> -		 * and pusha to work. ("enter $65535, $31" pushes
> -		 * 32 pointers and then decrements %sp by 65535.)
> -		 */
> -		if (unlikely(address + 65536 + 32 * sizeof(unsigned long) < regs->sp)) {
> -			bad_area(regs, sw_error_code, address);
> -			return;
> -		}
> -	}
>  	if (unlikely(expand_stack(vma, address))) {
>  		bad_area(regs, sw_error_code, address);
>  		return;

This v2 patch has no code change. I just updated the commit log to
capture some of the conversion that I had with the reviewers.

Cheers,
Longman