linux-kernel - Re: [PATCH v10 01/28] x86/fpu/xstate: Fix the state copy function to the XSTATE buffer

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: <87lf3c6hj6.ffs@tglx>
Date:   Fri, 01 Oct 2021 14:44:45 +0200
From:   Thomas Gleixner <tglx@...utronix.de>
To:     "Chang S. Bae" <chang.seok.bae@...el.com>, bp@...e.de,
        luto@...nel.org, mingo@...nel.org, x86@...nel.org
Cc:     len.brown@...el.com, lenb@...nel.org, dave.hansen@...el.com,
        thiago.macieira@...el.com, jing2.liu@...el.com,
        ravi.v.shankar@...el.com, linux-kernel@...r.kernel.org,
        chang.seok.bae@...el.com
Subject: Re: [PATCH v10 01/28] x86/fpu/xstate: Fix the state copy function
 to the XSTATE buffer

Chang,

On Wed, Aug 25 2021 at 08:53, Chang S. Bae wrote:

> Harden copy_uabi_to_xstate() so that it can handle the case where
> __raw_xsave() returns NULL.

That's an implementation detail, but does not explain why this can
happen and what this patch is about.

> This does not happen in practice today, but theoretically could happen
> in the future.

So what does the patch "fix"? When the subject says "Fix..." then I'm
expecting a bug in the code to be fixed.

There is none because the use case which can trip over this does not
exist today. You are adding it later.

Subject: .....: Prepare copy_uabi_to_xstate() to handle dynamic features

or something like that along with a reasonable explanation.

But in a later patch you add in the very same function:

> +    hdr.xfeatures &= fpu->state_mask;

which prevents that already because __raw_xsave_addr() is not invoked
for the zeroed bits in hdr.xfeatures:

>  		if (hdr.xfeatures & mask) {
>  			void *dst = __raw_xsave_addr(xsave, i);

Confused.

I'm not against the change per se, but I'm not accepting changelogs
which make no sense at all. News at 11.

Thanks,

        tglx