linux-kernel - Re: [PATCH] x86/sev: Don't use cc_platform

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] [day] [month] [year] [list]

Message-ID: <c6d89f21-b914-a7c1-d34c-dd083185d18b@intel.com>
Date:   Wed, 24 Aug 2022 11:53:49 -0700
From:   Dave Hansen <dave.hansen@...el.com>
To:     Borislav Petkov <bp@...en8.de>
Cc:     Tom Lendacky <thomas.lendacky@....com>,
        linux-kernel@...r.kernel.org, x86@...nel.org,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        "H. Peter Anvin" <hpa@...or.com>,
        Asish Kalra <ashish.kalra@....com>,
        Michael Roth <michael.roth@....com>,
        Joerg Roedel <jroedel@...e.de>,
        "Kirill A. Shutemov" <kirill@...temov.name>,
        stable@...r.kernel.org, Sean Christopherson <seanjc@...gle.com>
Subject: Re: [PATCH] x86/sev: Don't use cc_platform_has() for early SEV-SNP
 calls

On 8/24/22 11:48, Borislav Petkov wrote:
> On Wed, Aug 24, 2022 at 11:43:10AM -0700, Dave Hansen wrote:
>> So, we don't have *ANY* control over where the compiler uses jump
>> tables.  The kernel just happened to add some code that uses them, fell
>> over, and this adds a hack to get booting again.
>>
>> Isn't this a bigger problem?
> I had the same question already. Was thinking of maybe disabling
> the compiler from producing jump tables in the ident-mapped code.
> Tom's argument is that that might prevent the compiler from doing
> optimizations but I haven't talked to compiler folks whether those
> optimizations are even worth the effort.
> 
> Regardless, the potential problem is limited:
> 
> "# (jump-tables are implicitly disabled by RETPOLINE)"

Ahh, I missed the connection with retpoline.  The ubiquity of
RETPOLINE=y probably means we'll see more of these issues because people
won't find them unless they're building and running weirdo configurations.

> i.e., only RETPOLINE=n builds for now which should be a minority?
> 
> I guess when this explodes somewhere else again, we will have to
> generalize a fix.

Yep.  It also reminds me to add RETPOLINE=n build to my tests.