[PATCH/RFC 0/7] SATA ACPI suspend/resume support

[PATCH/RFC 0/7] SATA ACPI suspend/resume support

[Randy Dunlap]

This is an update to the SATA suspend/resume patches
to use ACPI methods (specifically _SDD and _GTF).
This patch series applies to 2.6.15-rc5-git3.

It boots but needs lots of testing & review/comments,
as well as adding a call to write taskfiles on
COMRESET event(s) (at least on async. loss of signal) and
freeing ata_device->gtf_address on module unload.

The patchset begins with the libata suspend/resume patch from
Jens Axboe and builds on that.

---
~Randy

Re: [PATCH/RFC 0/7] SATA ACPI suspend/resume support

[Jeff Garzik]

Randy Dunlap wrote:
> This is an update to the SATA suspend/resume patches
> to use ACPI methods (specifically _SDD and _GTF).
> This patch series applies to 2.6.15-rc5-git3.
> 
> It boots but needs lots of testing & review/comments,
> as well as adding a call to write taskfiles on
> COMRESET event(s) (at least on async. loss of signal) and
> freeing ata_device->gtf_address on module unload.
> 
> The patchset begins with the libata suspend/resume patch from
> Jens Axboe and builds on that.

overall:

* we do need some way of sending BIOS settings to the devices, so I 
grudgingly accept this, in concept.

* needs a command line arg that disables ACPI use.  libata=noacpi[, 
option...] would be my suggestion.

* currently assumes that to_pci_dev(struct device *) is valid for all 
cases.  so, need some way to limit calling the ATA ACPI functions, 
depending on runtime hardware configuration.

* at least one endian bug spotted

Re: [PATCH/RFC 0/7] SATA ACPI suspend/resume support

[Mark Lord]

Randy Dunlap wrote:
> This is an update to the SATA suspend/resume patches
> to use ACPI methods (specifically _SDD and _GTF).
> This patch series applies to 2.6.15-rc5-git3.
...

Randy,

Jeff is obviously a bit too overworked,
so I'll save him the trouble by THANKING YOU
for working on this!

Keep at it, and it'll make it in.
Lots of us need it.  LONG overdue.

Cheers!

Re: [PATCH/RFC 0/7] SATA ACPI suspend/resume support

[Randy.Dunlap]

On Wed, 14 Dec 2005 22:04:27 -0500 Mark Lord wrote:

> Randy Dunlap wrote:
> > This is an update to the SATA suspend/resume patches
> > to use ACPI methods (specifically _SDD and _GTF).
> > This patch series applies to 2.6.15-rc5-git3.
> ...
> 
> Randy,
> 
> Jeff is obviously a bit too overworked,

Yes, clearly, but he does good stuff.

> so I'll save him the trouble by THANKING YOU
> for working on this!
> 
> Keep at it, and it'll make it in.
> Lots of us need it.  LONG overdue.
> 
> Cheers!

Thanks, I'll make the changes that he suggested and resend them.

---
~Randy

Re: [PATCH/RFC 0/7] SATA ACPI suspend/resume support

[Randy.Dunlap]

On Wed, 14 Dec 2005 17:29:02 -0500 Jeff Garzik wrote:

> Randy Dunlap wrote:
> > This is an update to the SATA suspend/resume patches
> > to use ACPI methods (specifically _SDD and _GTF).
> > This patch series applies to 2.6.15-rc5-git3.
> > 
> > It boots but needs lots of testing & review/comments,
> > as well as adding a call to write taskfiles on
> > COMRESET event(s) (at least on async. loss of signal) and
> > freeing ata_device->gtf_address on module unload.
> > 
> > The patchset begins with the libata suspend/resume patch from
> > Jens Axboe and builds on that.
> 
> overall:
> 
> * we do need some way of sending BIOS settings to the devices, so I 
> grudgingly accept this, in concept.
> 
> * needs a command line arg that disables ACPI use.  libata=noacpi[, 
> option...] would be my suggestion.

Jeff, please clarify on this module option.

A module_param() that parallels atapi_enabled (like 'noacpi')
won't allow for additional ",option" values after it.  It would
be a simple int.

Are you suggesting a string parameter (charp) and
in-module parsing of a comma-separated option string?
If so, why?  I don't think that is necessary.

> * currently assumes that to_pci_dev(struct device *) is valid for all 
> cases.  so, need some way to limit calling the ATA ACPI functions, 
> depending on runtime hardware configuration.
> 
> * at least one endian bug spotted

---
~Randy

Re: [PATCH/RFC 0/7] SATA ACPI suspend/resume support

[Jae-hyeon Park]

This patch set does not work on my ThinkPad X41 Tablet.  I applied the
seven patches to kernel 2.6.15-rc5-git3.  I added "acpi_sleep=s3_bios"
to the kernel command line.

Let me describe how I tested in the following.  I suspend my laptop
with the following commands:

# init 1
# sync
# echo -n mem > /sys/power/state

The screen is cleared and the following is displayed:

Stopping tasks: ==========|
acpi_bus-0201 [57] bus_set_power         : Device is not power manageable
PM: Entering mem sleep

I suspect that the line "acpi_bus ..." indicates an error relevant to
this wakeup problem, although I don't know anything about kernel
hacking.  After a second, the laptop is suspended.  If I press the Fn
key to wake up the machine, it prints messages including:

...
acpi_bus-0201 [60] bus_set_power         : Device is not power manageable
...
ATA: abnormal status 0x80 on port 0x1F7
ATA: abnormal status 0x80 on port 0x1F7
ATA: abnormal status 0x80 on port 0x1F7

After a few tens of seconds, the screen before sleep is restored
showing the commands that I gave.  At this time, the hard disk LED is
off.

After I try to read the hard disk, for example by

# ls /usr/src

the hard disk LED stays on.  After a while the following message is
repeatedly printed:

ata1: command 0xc8 timeout, stat 0xd0 host_stat 0x1
ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ 0xb/47/00
ata1: status=0xd0 { Busy }
sd 0:0:0:0: SCSI error: return code = 0x8000002
sda: Current: sense key: Aborted Command
    Additional sense: Scsi parity error
Info fld=28f6944
end_request: I/O error, dev sda, sector 29379684
ATA: abnormal status 0xD0 on port 0x1F7
ATA: abnormal status 0xD0 on port 0x1F7
ATA: abnormal status 0xD0 on port 0x1F7

ata1: command 0xca timeout, stat 0xd0 host_stat 0x1
ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ 0xb/47/00
ata1: status=0xd0 { Busy }
sd 0:0:0:0: SCSI error: return code = 0x8000002
sda: Current: sense key: Aborted Command
    Additional sense: Scsi parity error
Info fld=28f6944
end_request: I/O error, dev sda, sector 29379692
ATA: abnormal status 0xD0 on port 0x1F7
ATA: abnormal status 0xD0 on port 0x1F7
ATA: abnormal status 0xD0 on port 0x1F7

It is confusing that the symptom after an unsuccessful wakeup
basically appears to be the same as the one described in the posting,
http://marc.theaimsgroup.com/?l=linux-kernel&m=111409309804068&w=2
FYI, software suspend 2 works okay on this machine with SATA resume
patch.

Any ideas?  Thanks in advance.

Jae-hyeon

Re: [PATCH/RFC 0/7] SATA ACPI suspend/resume support

[Randy Dunlap]

On 20 Dec 2005 00:18:24 +0900
Jae-hyeon Park <jhpark@tuhep.phys.tohoku.ac.jp> wrote:

> This patch set does not work on my ThinkPad X41 Tablet.  I applied the
> seven patches to kernel 2.6.15-rc5-git3.  I added "acpi_sleep=s3_bios"
> to the kernel command line.
> 
> Let me describe how I tested in the following.  I suspend my laptop
> with the following commands:
> 
> # init 1
> # sync
> # echo -n mem > /sys/power/state
> 
> The screen is cleared and the following is displayed:
> 
> Stopping tasks: ==========|
> acpi_bus-0201 [57] bus_set_power         : Device is not power manageable
> PM: Entering mem sleep

I've asked for some clarification on this message, including
how serious (problematic) it is to the suspend process,
but have received no answer yet (on the acpi-devel mailing
list).  It appears to mostly be a debugging information message.

> I suspect that the line "acpi_bus ..." indicates an error relevant to
> this wakeup problem, although I don't know anything about kernel
> hacking.  After a second, the laptop is suspended.  If I press the Fn
> key to wake up the machine, it prints messages including:
> 
> ...
> acpi_bus-0201 [60] bus_set_power         : Device is not power manageable
> ...
> ATA: abnormal status 0x80 on port 0x1F7
> ATA: abnormal status 0x80 on port 0x1F7
> ATA: abnormal status 0x80 on port 0x1F7
> 
> After a few tens of seconds, the screen before sleep is restored
> showing the commands that I gave.  At this time, the hard disk LED is
> off.
> 
> After I try to read the hard disk, for example by
> 
> # ls /usr/src
> 
> the hard disk LED stays on.  After a while the following message is
> repeatedly printed:
> 
> ata1: command 0xc8 timeout, stat 0xd0 host_stat 0x1
> ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ 0xb/47/00
> ata1: status=0xd0 { Busy }
> sd 0:0:0:0: SCSI error: return code = 0x8000002
> sda: Current: sense key: Aborted Command
>     Additional sense: Scsi parity error
> Info fld=28f6944
> end_request: I/O error, dev sda, sector 29379684
> ATA: abnormal status 0xD0 on port 0x1F7
> ATA: abnormal status 0xD0 on port 0x1F7
> ATA: abnormal status 0xD0 on port 0x1F7
> 
> ata1: command 0xca timeout, stat 0xd0 host_stat 0x1
> ata1: translated ATA stat/err 0xd0/00 to SCSI SK/ASC/ASCQ 0xb/47/00
> ata1: status=0xd0 { Busy }
> sd 0:0:0:0: SCSI error: return code = 0x8000002
> sda: Current: sense key: Aborted Command
>     Additional sense: Scsi parity error
> Info fld=28f6944
> end_request: I/O error, dev sda, sector 29379692
> ATA: abnormal status 0xD0 on port 0x1F7
> ATA: abnormal status 0xD0 on port 0x1F7
> ATA: abnormal status 0xD0 on port 0x1F7
> 
> It is confusing that the symptom after an unsuccessful wakeup
> basically appears to be the same as the one described in the posting,
> http://marc.theaimsgroup.com/?l=linux-kernel&m=111409309804068&w=2
> FYI, software suspend 2 works okay on this machine with SATA resume
> patch.
> 
> Any ideas?  Thanks in advance.

I've read your mail several times now and don't have any real
ideas for you, but I wanted to reply so that you wouldn't feel
like you were being ignored.

I'll have a new patch series in another day or so.  Please try
it again and let me/us know how it goes.

Thanks,
---
~Randy

Re: [PATCH/RFC 0/7] SATA ACPI suspend/resume support

[Jae-hyeon Park]

Randy Dunlap <randy_d_dunlap@linux.intel.com> writes:

> > The screen is cleared and the following is displayed:
> > 
> > Stopping tasks: ==========|
> > acpi_bus-0201 [57] bus_set_power         : Device is not power manageable
> > PM: Entering mem sleep
> 
> I've asked for some clarification on this message, including
> how serious (problematic) it is to the suspend process,
> but have received no answer yet (on the acpi-devel mailing
> list).  It appears to mostly be a debugging information message.

I hope the following information is useful.
I changed part of drivers/acpi/bus.c

        if (!device->flags.power_manageable) {
                ACPI_DEBUG_PRINT((ACPI_DB_WARN,
                                  "Device is not power manageable\n"));
                return_VALUE(-ENODEV);
        }

to

        if (!device->flags.power_manageable) {
                ACPI_DEBUG_PRINT((ACPI_DB_WARN,
                                  "Device %s is not power manageable\n",
                                  device->kobj.name));
                return_VALUE(-ENODEV);
        }
        
And the kobj.name included in the warning was:

acpi_bus-0202 [57] bus_set_power         : Device IDE0 is not power manageable

Looks like this is related to a hard drive controller?

Jae-hyeon

Re: [PATCH/RFC 0/7] SATA ACPI suspend/resume support

[Randy Dunlap]

On 22 Dec 2005 21:52:12 +0900
Jae-hyeon Park <jhpark@tuhep.phys.tohoku.ac.jp> wrote:

> Randy Dunlap <randy_d_dunlap@linux.intel.com> writes:
> 
> > > The screen is cleared and the following is displayed:
> > > 
> > > Stopping tasks: ==========|
> > > acpi_bus-0201 [57] bus_set_power         : Device is not power manageable
> > > PM: Entering mem sleep
> > 
> > I've asked for some clarification on this message, including
> > how serious (problematic) it is to the suspend process,
> > but have received no answer yet (on the acpi-devel mailing
> > list).  It appears to mostly be a debugging information message.
> 
> I hope the following information is useful.
> I changed part of drivers/acpi/bus.c
> 
>         if (!device->flags.power_manageable) {
>                 ACPI_DEBUG_PRINT((ACPI_DB_WARN,
>                                   "Device is not power manageable\n"));
>                 return_VALUE(-ENODEV);
>         }
> 
> to
> 
>         if (!device->flags.power_manageable) {
>                 ACPI_DEBUG_PRINT((ACPI_DB_WARN,
>                                   "Device %s is not power manageable\n",
>                                   device->kobj.name));
>                 return_VALUE(-ENODEV);
>         }
>         
> And the kobj.name included in the warning was:
> 
> acpi_bus-0202 [57] bus_set_power         : Device IDE0 is not power manageable
> 
> Looks like this is related to a hard drive controller?

Yes, but from looking at the code, it still doesn't look
like a critical error to me, just a warning.
Nevertheless, we should get your patch merged IMO.

Thanks,
---
~Randy

Re: [PATCH/RFC 0/7] SATA ACPI suspend/resume support

[Jae-hyeon Park]

Randy Dunlap <randy_d_dunlap@linux.intel.com> writes:

> On 20 Dec 2005 00:18:24 +0900
> Jae-hyeon Park <jhpark@tuhep.phys.tohoku.ac.jp> wrote:
> 
> > This patch set does not work on my ThinkPad X41 Tablet.  I applied the
> > seven patches to kernel 2.6.15-rc5-git3.  I added "acpi_sleep=s3_bios"
> > to the kernel command line.
> > 
> 
> I've read your mail several times now and don't have any real
> ideas for you, but I wanted to reply so that you wouldn't feel
> like you were being ignored.
> 

I found that ACPI sleep/wakeup works if I (compile and) insert the
snd_intel8x0m module, which I don't need normally.  If I have rmmod-ed
snd_intel8x0m, wakeup fails.  As you said, the message "device is not
power manageable" was not a critical error.

Thanks,

Jae-hyeon