Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <af726924-4eb1-aa2c-319f-0a67003ef37f@intel.com>
Date:   Fri, 20 May 2022 10:52:40 -0700
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.8.1
Subject: Re: [PATCHv2 2/3] x86/tdx: Clarify RIP adjustments in #VE handler
Content-Language: en-US
To:     "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
        tglx@linutronix.de, mingo@redhat.com, bp@alien8.de,
        luto@kernel.org, peterz@infradead.org
Cc:     ak@linux.intel.com, dan.j.williams@intel.com, david@redhat.com,
        hpa@zytor.com, linux-kernel@vger.kernel.org,
        sathyanarayanan.kuppuswamy@linux.intel.com, seanjc@google.com,
        thomas.lendacky@amd.com, x86@kernel.org
References: <20220520031316.47722-1-kirill.shutemov@linux.intel.com>
 <20220520031316.47722-3-kirill.shutemov@linux.intel.com>
From:   Dave Hansen <dave.hansen@intel.com>
In-Reply-To: <20220520031316.47722-3-kirill.shutemov@linux.intel.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 5/19/22 20:13, Kirill A. Shutemov wrote:
> +	/*
> +	 * If the #VE happened due to instruction execution, GET_VEINFO
> +	 * provides info on the instruction in out.r10.
> +	 *
> +	 * For #VE due to EPT violation, info in out.r10 is not usable and
> +	 * kernel has to decode instruction manually to find out its length.
> +	 */
> +	if (ve->exit_reason != EXIT_REASON_EPT_VIOLATION) {
> +		ve->instr_len   = lower_32_bits(out.r10);
> +		ve->instr_info  = upper_32_bits(out.r10);
> +	} else {
> +		ve->instr_len   = 0;
> +		ve->instr_info  = 0;
> +	}
>  }

This is _better_, but I still don't like it.  It still hides the magic
down in here and if someone screws it up, they'll silently get a loop of
#VE's.  If we stick the logic a helper like:

int ve_instr_len(struct ve_info *ve)
{
	if (WARN_ON_ONCE(ve->exit_reason != EXIT_REASON_EPT_VIOLATION))
		return 0;

	return ve->instr_len;
}

and then folks consume the instruction length with that, we get a splat
right where and when it goes wrong.

BTW, how do we know that all non-EPT_VIOLATION exits reasons are
instruction execution?