linux-kernel - Re: [PATCH v2 5/8] x86/speculation: Add basic support for IBPB

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: <20180121190145.uuk3xizxejckth5s@pd.tnic>
Date:   Sun, 21 Jan 2018 20:01:45 +0100
From:   Borislav Petkov <bp@...en8.de>
To:     KarimAllah Ahmed <karahmed@...zon.com>
Cc:     David Woodhouse <dwmw@...zon.co.uk>, arjan@...ux.intel.com,
        tglx@...utronix.de, karahmed@...zon.de, x86@...nel.org,
        linux-kernel@...r.kernel.org, tim.c.chen@...ux.intel.com,
        peterz@...radead.org, pbonzini@...hat.com, ak@...ux.intel.com,
        torvalds@...ux-foundation.org, gregkh@...ux-foundation.org
Subject: Re: [PATCH v2 5/8] x86/speculation: Add basic support for IBPB

On Sun, Jan 21, 2018 at 07:29:43PM +0100, KarimAllah Ahmed wrote:
> Because static_cpu_has is an indirect branch which will cause speculation
> and
> we have to avoid that.

How so?

The JMP_NOSPEC macro protects against JMP <reg> jumps but the
static_cpu_has() macros all add JMPs with an immediate offset from the
next instruction and I wouldn't call them indirect JMPs as there are no
registers to speculate on there.

IOW, before alternatives, the patch site of static_cpu_has() looks like this:

# 151 "./arch/x86/include/asm/cpufeature.h" 1
        1: jmp 6f

and that 6f label is:

6:
 testb $1,boot_cpu_data+50(%rip)        #, MEM[(const char *)&boot_cpu_data + 50B]
 jnz .L707      #
 jmp .L706      #

i.e., we basically do if (boot_cpu_has(..)).

If the feature is not present, same patch site turns into:

4: jmp .L706    #
5:

after patching. Which is a label after the whole thing. That is not an
indrect jump through a register either.

If the feature is present, the patch site becomes:

	NOP	- added by the patching

# ./arch/x86/include/asm/msr.h:105:     asm volatile("1: wrmsr\n"
        .loc 18 105 0
        movl    $73, %ecx       #, tmp138
        movl    $1, %eax        #, tmp139
        xorl    %edx, %edx      # tmp140
#APP
# 105 "./arch/x86/include/asm/msr.h" 1
        1: wrmsr
2:

so execution runs directly into the MSR write and the JMP is gone.

So I don't see indirect branches anywhere...

-- 
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.