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;
Subject: Re: AHCI SATA Runtime PM
To:     Alexander Monakov <amonakov@ispras.ru>,
        Jens Axboe <axboe@kernel.dk>
Cc:     linux-kernel@vger.kernel.org, linux-pm@vger.kernel.org,
        Zhang Rui <rui.zhang@intel.com>, linux-ide@vger.kernel.org
References: <alpine.LNX.2.20.13.2103041814050.15170@monopod.intra.ispras.ru>
From:   Hans de Goede <hdegoede@redhat.com>
Message-ID: <f74cdf09-c08f-5cff-e046-2729aeb703be@redhat.com>
Date:   Wed, 17 Mar 2021 16:02:06 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.7.0
MIME-Version: 1.0
In-Reply-To: <alpine.LNX.2.20.13.2103041814050.15170@monopod.intra.ispras.ru>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi,

On 3/4/21 4:34 PM, Alexander Monakov wrote:
> Hello, Hans, Linux PM folks,
> 
> I'm looking for clarification regarding this patch discussion:
> 
> https://patchwork.kernel.org/project/linux-pm/patch/20180420101834.15783-2-0v3rdr0n3@gmail.com/
> 
> Hans said,
> 
>> Ah, so the AHCI code has runtime pm enabled by default (so there
>> is another pm_runtime_allow() somewhere, but then disables it for
>> unused ports to make hotpluging something into those ports work.
> 
> I have a laptop where two unused AHCI SATA controllers are present (but
> obviously nothing can be hotplugged into them). Apparently due to the above,
> they do not enter runtime autosuspend.
> 
> The problem is, these "ATA port" nodes don't seem to participate in udev
> hierarchy, so it's unclear how I'm supposed to automatically re-enable
> runtime PM for them. For PCI device nodes, I have
> 
> ACTION=="add", SUBSYSTEM=="pci", TEST=="power/control", ATTR{power/control}="auto"
> 
> but ata1/uevent is empty, so there's no obvious way to write the corresponding
> UDev rule.

You can enable runtime-pm from udev using the following rule:

ACTION=="add", SUBSYSTEM=="ata_port", ATTR{../../power/control}="auto"

> Prior to discovering the above patch discussion, I have filed a bug:
> https://bugzilla.kernel.org/show_bug.cgi?id=211837
> 
> Does the above correctly reflect how AHCI PM is supposed to be?

This is a complicated question, I just checked and according to:
https://www.intel.com/content/www/us/en/io/serial-ata/serial-ata-ahci-spec-rev1-3-1.html

AHCI 1.3 compliant controllers should raise a PME signal on hotplug
changes even when in D3. So perhaps the solution is to catch this signal
and then wakeup the controller; and remove the pm_runtime_forbid(dev); call
from ata_tport_add().

I don't have much experience with dealing with the PME stuff, so you will
need to do some research there, or perhaps someone else can help?

This seems to be more of a sata/ahci question then a generic linux-pm question
though, so I have added Jebns and linux-ide to the To / Cc.

> If so, what is
> the proper way to enable runtime PM for unused ATA ports?

Regards,

Hans