Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH v7 1/3] irqchip/gic-v3-its: add ability to save/restore
 ITS state
To:     "dbasehore ." <dbasehore@chromium.org>
Cc:     Mark Rutland <mark.rutland@arm.com>,
        linux-kernel <linux-kernel@vger.kernel.org>,
        Soby Mathew <Soby.Mathew@arm.com>,
        Sudeep Holla <sudeep.holla@arm.com>,
        devicetree@vger.kernel.org, robh+dt@kernel.org,
        Linux-pm mailing list <linux-pm@vger.kernel.org>,
        "Wysocki, Rafael J" <rafael.j.wysocki@intel.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Brian Norris <briannorris@chromium.org>
References: <20180301054820.42847-1-dbasehore@chromium.org>
 <20180301054820.42847-2-dbasehore@chromium.org>
 <20180301114144.nmwu4qtslvj7ogj5@lakrids.cambridge.arm.com>
 <0a74f87a-027a-df7d-ca2d-fc164ec9cdf7@arm.com>
 <CAGAzgsqLgujex--HCo3dRGvOPXG0zu-RwkbuGhgNhBTc3sURJA@mail.gmail.com>
From:   Marc Zyngier <marc.zyngier@arm.com>
Organization: ARM Ltd
Message-ID: <6a5b5723-85bb-e46f-f621-eab17a01f1f0@arm.com>
Date:   Wed, 14 Mar 2018 10:22:30 +0000
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.6.0
MIME-Version: 1.0
In-Reply-To: <CAGAzgsqLgujex--HCo3dRGvOPXG0zu-RwkbuGhgNhBTc3sURJA@mail.gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-GB
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 02/03/18 02:08, dbasehore . wrote:
> On Thu, Mar 1, 2018 at 4:29 AM, Marc Zyngier <marc.zyngier@arm.com> wrote:
>> Hi Mark,
>>
>> On 01/03/18 11:41, Mark Rutland wrote:
>>> On Wed, Feb 28, 2018 at 09:48:18PM -0800, Derek Basehore wrote:
>>>> Some platforms power off GIC logic in suspend, so we need to
>>>> save/restore state. The distributor and redistributor registers need
>>>> to be handled in platform code due to access permissions on those
>>>> registers, but the ITS registers can be restored in the kernel.
>>>>
>>>> Signed-off-by: Derek Basehore <dbasehore@chromium.org>
>>>
>>> How much state do we have to save/restore?
>>>
>>> Given we can apparently read all this state, couldn't we *always* save
>>> the state, then upon resume detect if the state has been lost, restoring
>>> it if so?
>>>
>>> That way, we don't need a property in FW tables for DT or ACPI.
>>
>> That's a good point. I guess that we could just compare the saved
>> GITS_CTLR register and restore the full state only if the ITS comes back
>> as disabled.
>>
>> I'm just a bit worried that it makes it an implicit convention between
>> kernel an FW, which could change in funny ways. Importantly, the PSCI
>> spec says states FW should restore *the whole state*. Obviously, it
>> cannot to that on HW that doesn't allow you to read out the state, hence
>> the DT flag that outlines the departure from the expected behaviour.
>>
>> I'm happy to go either way, but then I have the feeling that we should
>> go back to quirking it on the actual implementation (GIC500 in this
>> case) if we're to from the property.
>>
>>>
>>> [...]
>>>
>>>> @@ -3261,6 +3363,9 @@ static int __init its_probe_one(struct resource *res,
>>>>              ctlr |= GITS_CTLR_ImDe;
>>>>      writel_relaxed(ctlr, its->base + GITS_CTLR);
>>>>
>>>> +    if (fwnode_property_present(handle, "reset-on-suspend"))
>>>> +            its->flags |= ITS_FLAGS_SAVE_SUSPEND_STATE;
>>>
>>> Does this allow this property on an ACPI system?
>>>
>>> If we need this on ACPI, we need a spec update to handle this properly,
>>> and shouldn't use device properties like this.
>>
>> Well spotted. I guess that dropping the property would fix that
>> altogether, assuming we feel that the above is safe enough.
>>
>> Thoughts?
> 
> I'm fine changing it to get rid of the devicetree property.
> 
> What's the reason for quirking the behavior though? It's not that much
> code + data and nothing else relies on the state of the ITS getting
> disabled across suspend/resume. Even if something did, we'd have to
> resolve it with this feature anyways.

The reason we do this is to cope with GIC500 having the collection state
in registers instead of memory. If we didn't have this extraordinary
misfeature, FW could do a full save/restore of the ITS, and we wouldn't
have to do anything (which is what the driver currently expects).

A middle ground approach is to limit the feature to systems where
GITS_TYPER.HCC is non-zero instead of limiting it to GIC500. Pretty easy
to fix. This should have the same effect, as GIC500 is the only
implementation I'm aware of with HCC!=0.

Given that we're already at -rc5 and that I'd like to queue things for
4.17, I've made this change myself and queued patches 1 and 3 here[1].

Can you please have a look at let me know if that works for you?

Thanks,

	M.

[1] git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git
irq/irqchip-next
-- 
Jazz is not dead. It just smells funny...