From mboxrd@z Thu Jan 1 00:00:00 1970 From: Zhang Rui Subject: Re: [PATCH] ACPI: evaluate _PS3 when entering D3 Cold Date: Sun, 01 Apr 2012 16:01:54 +0800 Message-ID: <1333267314.2387.122.camel@rui.sh.intel.com> References: <1333217910-29579-1-git-send-email-aaron.lu@amd.com> <1333263819.2387.94.camel@rui.sh.intel.com> <20120401153434.GA3268@localhost.amd.com> <201204010947.27234.rjw@sisk.pl> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: In-Reply-To: <201204010947.27234.rjw@sisk.pl> Sender: linux-kernel-owner@vger.kernel.org To: "Rafael J. Wysocki" Cc: Aaron Lu , Lin Ming , Len Brown , linux-acpi@vger.kernel.org, linux-pm@vger.kernel.org, linux-kernel@vger.kernel.org, Andiry Xu , Alex He , "Moore, Robert" List-Id: linux-acpi@vger.kernel.org On =E6=97=A5, 2012-04-01 at 09:47 +0200, Rafael J. Wysocki wrote: > On Sunday, April 01, 2012, Aaron Lu wrote: > > Hi, > >=20 > > On Sun, Apr 01, 2012 at 03:03:39PM +0800, Zhang Rui wrote: > > > First of all, I agree that we must evaluate _PS3 when setting dev= ice to > > > either D3_HOT or D3_COLD. > > Good. > >=20 > > >=20 > > > And here is my understanding about D3/D3_HOT/D3_COLD, > > >=20 > > > if _PR3 exists, it means the devices supports both D3_HOT and D3_= COLD. > > Agree. > >=20 > > >=20 > > > if only _PS3 exists, we can only say that the state after evaluat= ing > > > _PS3 is D3, it could either be D3_HOT or D3_COLD, and this is dev= ice > > > specific, which in your case, is D3_COLD. > > I prefer Rafeal's definition, let's just *assume* the device is at = D3 > > cold after its _PS3 is executed. Unless it has _PR3, in which case,= we > > have a chance to put the device into D3 hot instead. > >=20 > > >=20 > > > BTW, here is the description of _S0W in ACPI spec, > > > If OSPM has not indicated that it supports _PR3 through the OSPM > > > Platform-Wide Capabilities (see Section 6.2.10.2), then the value= "3" > > > corresponds to D3. If it has indicated _PR3 support, the value "3= " > > > represents D3hot and the value "4" represents D3cold. > > >=20 > > > So IMO, the _S0W should return 3 in AMD's implementation as it do= es not > > > have _PR3. > > OK, sounds like a firmware bug. > > Thanks for identifying this. >=20 > I don't think this is a bug. It actually may return either 3 or 4, b= ecause > there is no difference between them if there's no _PR3 (i.e. the acti= on to > carry out by software would only be different if _PR3 were present). >=20 I mean, surely that software should handle this case. But this is still a violation of ACPI spec, as the device has only one D3 state, instead of D3_HOT/D3_COLD,, thus _S0W should return 3 instead= =2E thanks, rui > > > > And the ACPI does have some words like: > > > >=20 > > > > ------ > > > > Platform/drivers must assume that the device will have power co= mpletely > > > > removed when the device is place into =E2=80=9CD3=E2=80=9D via = _PS3 > > > > ------ > > > >=20 > > > I think this means OS can not access device any more after evalua= ting > > > _PS3, and it should re-enumerate the device when transiting back = to D0. > > >=20 > > > > This is in section 7.2.11: _PR3. > > > >=20 > > > > >=20 > > > > > Another problem: > > > > >=20 > > > > > With your patch, both D3hot and D3cold will evaluate _PS3, ri= ght? > > > > >=20 > > > > Yes. > > > >=20 > > > > > Will it have problem on AMD platform if you try to put ODD in= to D3hot > > > > > state? _PS3 is evaluated, so it actually enters D3Cold state. > > > >=20 > > > > There is no D3 hot support for this device(from the firmware's > > > > perspective), either it is at D0(via _PS0), or it will be at D3= cold(via > > > > _PS3). > > > >=20 > > > I was trying to make a cleanup of the D3/D3_HOT/D3_COLD support i= n > > > Linux, and this gives me some clue. > > This is great, and I would like to help as much as I can. > >=20 > > >=20 > > > How about this? > > >=20 > > > We should use the term "D3" in general in Linux. > > > Without _PR3, OS should *assume* that the power is removed, altho= ugh it > > > may be not true. > > > With _PR3, OS can *assure* that the power is removed, because it = knows > > > how to remove thw power (evaluating _PR3._OFF). > > >=20 > > > So the difference is that OS need to make sure whether to evaluat= e > > > _PR3._OFF when _PR3 exists. For example, a device has _PR3, but _= S0W > > > returns 3, OS should not evaluate _PR3._OFF when the device sleep= s with > > > remote wakeup support. > > >=20 > > > what do you think? > > I agree with Rafeal's ideas. >=20 > Good. :-) >=20 > Thanks, > Rafael