linux-kernel - Re: [PATCH v2] ACPI: D3 states cleanup

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>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Sat, 21 Apr 2012 01:36:36 +0000
From:	"Lu, Aaron" <Aaron.Lu@....com>
To:	"Rafael J. Wysocki" <rjw@...k.pl>
CC:	Lin Ming <ming.m.lin@...el.com>, Len Brown <lenb@...nel.org>,
	linux-acpi <linux-acpi@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	Zhang Rui <rui.zhang@...el.com>,
	"Huang, Ying" <ying.huang@...el.com>
Subject: Re: [PATCH v2] ACPI: D3 states cleanup


在 2012-4-20，下午7:08，"Rafael J. Wysocki" <rjw@...k.pl> 写道：

> On Friday, April 20, 2012, Aaron Lu wrote:
>> Hi,
>> 
>> On Fri, Apr 20, 2012 at 01:37:35PM +0800, Lin Ming wrote:
>>>>>>> There are two ACPI D3 states defined now:
>>>>>>> ACPI_STATE_D3 and ACPI_STATE_D3_COLD.
>>>>>>> 
>>>>>>> But the uses of these states are not clear/correct in some code.
>>>>>>> For example, some code refer ACPI_STATE_D3 as D3hot and others refer
>>>>>>> it as D3cold.
>>>>>>> 
>>>>>>> This patch introduces ACPI_STATE_D3_HOT to refer to ACPI D3hot state.
>>>>>>> And changes ACPI_STATE_D3 to refer to ACPI D3cold state only.
>>>>>>> Also redefines ACPI_STATE_D3_COLD the same as ACPI_STATE_D3.
>>>>>>> 
>>>>>> 
>>>>>> With this patch now, if a device has _PS3, then we will set its D3hot
>>>>>> flag valid. This doesn't feel right to me, since per our discussion the
>>>>>> other day, we should assume _PS3 will put the device into D3cold.
>>>>>> 
>>>>>> Or do you mean: if _PS3 is available, then both D3hot and D3cold is
>>>>>> valid from the perspective of acpi, it is the individual driver's
>>>>>> responsibility to decide which state is actually valid and will be used.
>>>>> 
>>>>> Right.
>>>>> 
>>>>> ACPI_STATE_D3(same as ACPI_STATE_D3_COLD now) is always valid.
>>>>> 
>>>> 
>>>> I mean, if _PS3 is available, can we say D3hot is valid?
>>> 
>>> Yes.
>>> 
>> 
>> OK, now I'm confused...
>> 
>> First, let me try to clarify the meaning of acpi power state's valid
>> flag.
>> 
>> By valid, I suppose it means the device can be in that state, instead of
>> we have a way to program this device to go into that state.
>> 
>> e.g. D0 is valid means the device can be in D0 state, and D3_cold is
>> valid means the device can be in D3_cold state. We unconditionally set
>> these two states as valid, because we know every device supports these
>> two states. But we might not be able to put the device into that state
>> in software, since we might not have _PS0 or _PS3 control methods for it.
>> 
>> And if we do have the _PSx or _PRx control methods, we knows we have a
>> way to put the device into that state, and hence the device should be
>> able to support that power state, so we will set that state as valid too.
>> 
>> Is this correct?
>> 
>> For D3hot, obviously not all device supports this state, so we will need
>> to figure it out through the acpi table.
>> I remembered Rafael said the following words the other day in a reply to
>> my evaluate_ps3_when_entering_d3_cold_patch:
>> 
>> -----------------------------------------------------------------------
>> I'd rather say that with _PR3 we have the opportunity to avoid removing
>> power completely from the device.  In other words, D3_hot is supported (and
>> it is supported _only_ in that case).
>> -----------------------------------------------------------------------
>> 
>> So I think here is a problem, that if a device has only _PS3, why should
>> we say D3hot is supported? Is there a reason for this that I missed?
> 
> OK, I agree.  We need to special case the situation in which _PR3 is not
> present, but _PS3 is.  IOW, we should do something like this in the loop in
> acpi_bus_get_power_flags():
> 
> 
>        /* State is valid if we have some power control */
>        if (ps->resources.count
>            || (ps->flags.explicit_set && i < ACPI_STATE_D3))
>            ps->flags.valid = 1;
> 

Looks good to me, thanks.

-Aaron