linux-kernel - Re: [PATCH 18/18] arm64: apple: Add initial Mac Mini 2020 (M1) devicetree

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

Message-ID: <8d660b1f-cb80-d16c-14e4-2a1c7f5229d1@marcan.st>
Date:   Mon, 8 Feb 2021 23:53:52 +0900
From:   Hector Martin <marcan@...can.st>
To:     Marc Zyngier <maz@...nel.org>
Cc:     soc@...nel.org, linux-arm-kernel@...ts.infradead.org,
        robh+dt@...nel.org, Arnd Bergmann <arnd@...nel.org>,
        linux-kernel@...r.kernel.org, devicetree@...r.kernel.org,
        Olof Johansson <olof@...om.net>
Subject: Re: [PATCH 18/18] arm64: apple: Add initial Mac Mini 2020 (M1)
 devicetree

On 08/02/2021 21.27, Marc Zyngier wrote:
>> +	timer {
>> +		compatible = "arm,armv8-timer";
>> +		interrupt-parent = <&aic>;
>> +		interrupts = <AIC_FIQ 0 IRQ_TYPE_LEVEL_HIGH>,
>> +				<AIC_FIQ 0 IRQ_TYPE_LEVEL_HIGH>,
>> +				<AIC_FIQ 1 IRQ_TYPE_LEVEL_HIGH>,
>> +				<AIC_FIQ 0 IRQ_TYPE_LEVEL_HIGH>;
> 
> This unfortunately doesn't match the binding, which doesn't cater
> for systems without a secure physical timer, nor allows the description
> of the EL2 virtual timer.
> 
> You should also have *different* interrupts for EL1 and EL2 timers,
> although this is all a lie...

Well, we do - now that I confirmed all 4 timers work properly, the AIC 
driver should provide all 4. And ideally I find those EL1 timer mask 
bits and implement them in the aic driver too (for only the virt timers 
that have them and of course need them).

I just found the code in arm_arch_timer that forwards all this stuff to 
the kvm code, so it all makes sense now; if I can wire that up properly, 
heck, KVM might even just work here.

> 
> Looking at the only similar case, XGene lies about the secure timer
> (it doesn't have any), and of course doesn't have an EL2 virtual
> timer (ARMv8.0 only).
> 
> A sensible course of action could be to update the binding to at least:
> 
> - tell the kernel that there is no secure physical timer (and that
>     the interrupt should be ignored)
> - introduce a 5th possible interrupt for the EL2 virtual timer.

Sounds like I should be introducing interrupt-names support into this 
driver and using that, so we can just not specify IRQs that don't exist, 
instead of the hack with dummies. Falling back to indexes of course, to 
keep DT compat. i.e.

const char *names = {"phys-secure", "phys", "virt", "hyp-phys", "hyp-virt"};

bool has_names = of_property_read_bool(..., "interrupt-names");

for (each irq)
	if (has_names) foo = of_irq_get_byname(..., names[i])
	else foo = of_irq_get(..., i)

That said, is there a use case for the EL2 virtual timer? The driver 
always uses the EL2 physical timer with VHE right now. I guess it's 
worth describing it in the binding and dts, even if the driver never 
selects it...?

-- 
Hector Martin (marcan@...can.st)
Public Key: https://mrcn.st/pub