Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII;
 format=flowed
Content-Transfer-Encoding: 7bit
Date:   Wed, 11 Nov 2020 12:46:30 +0000
From:   Marc Zyngier <maz@kernel.org>
To:     David Brazdil <dbrazdil@google.com>
Cc:     kvmarm@lists.cs.columbia.edu, linux-arm-kernel@lists.infradead.org,
        linux-kernel@vger.kernel.org, James Morse <james.morse@arm.com>,
        Julien Thierry <julien.thierry.kdev@gmail.com>,
        Suzuki K Poulose <suzuki.poulose@arm.com>,
        Catalin Marinas <catalin.marinas@arm.com>,
        Will Deacon <will@kernel.org>, Dennis Zhou <dennis@kernel.org>,
        Tejun Heo <tj@kernel.org>, Christoph Lameter <cl@linux.com>,
        Mark Rutland <mark.rutland@arm.com>,
        Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>,
        Quentin Perret <qperret@google.com>,
        Andrew Scull <ascull@google.com>,
        Andrew Walbran <qwandor@google.com>, kernel-team@android.com
Subject: Re: [PATCH v1 06/24] kvm: arm64: Support per_cpu_ptr in nVHE hyp code
In-Reply-To: <20201111123222.ooevu6nryu3ncphv@google.com>
References: <20201109113233.9012-1-dbrazdil@google.com>
 <20201109113233.9012-7-dbrazdil@google.com>
 <f28c9a67759cb04157e888b3a71b2ce2@kernel.org>
 <20201111123222.ooevu6nryu3ncphv@google.com>
User-Agent: Roundcube Webmail/1.4.9
Message-ID: <52b47a4284aaff2e198385ef9acbe01f@kernel.org>
Precedence: bulk

On 2020-11-11 12:32, David Brazdil wrote:
>> > +
>> > +	cpu_base_array = kern_hyp_va(&kvm_arm_hyp_percpu_base[0]);
>> 
>> There is no guarantee that this will not generate a PC relative
>> addressing, resulting in kern_hyp_va() being applied twice.
>> 
>> Consider using hyp_symbol_addr() instead, which always does the right
>> by forcing a PC relative addressing and not subsequently mangling
>> the address.
>> 
>> > +	this_cpu_base = kern_hyp_va(cpu_base_array[cpu]);
>> > +	return this_cpu_base - (unsigned long)&__per_cpu_start;
>> 
>> And this is the opposite case: if the compiler generates an absolute
>> address, you're toast. Yes, this is just as unlikely, but hey...
>> Same remedy should apply.
> 
> Good point, and I'll probably keep forgetting about this in the future. 
> Now
> that all .hyp.text is only executed under hyp page tables, should we 
> start
> thinking about fixing up the relocations?

Why not, if you can deal with the hypervisor text being mapped at a 
random
location, and make sure that the kernel doesn't process the relocations
for you. This would certainly save us a lot of runtime offsetting (which
I'm adding to in a separate series).

         M.
-- 
Jazz is not dead. It just smells funny...