RE: Suspend-to-RAM on Sony Vaios

RE: Suspend-to-RAM on Sony Vaios

[Pallipadi, Venkatesh]

>-----Original Message-----
>From: acpi-devel-admin-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org 
>[mailto:acpi-devel-admin-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org] On Behalf Of 
>ole.rohne-vJEk5272eHo@public.gmane.org
>Sent: Wednesday, November 03, 2004 3:32 AM
>To: Andi Kleen
>Cc: Pavel Machek; Emmanuel Thom??; Jon Valvatne; 
>acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
>Subject: Re: [ACPI] Suspend-to-RAM on Sony Vaios
>
>"Later in an emulator" is arguably safer than real mode. But the
>emulation needs to be done from kernel at the proper point in the
>resume sequence. Leaving it to the user-mode suspend/resume script
>is not an option.
>

Other options here are:
- to have the usermode emulator and call it from 
Proper point in kernel using call_usermodehelper().
- or have the emulator in the kernel itself.

-Venki


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_idU88&alloc_id\x12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[Andi Kleen]

On Wed, Nov 03, 2004 at 06:45:53AM -0800, Pallipadi, Venkatesh wrote:
> Other options here are:
> - to have the usermode emulator and call it from 
> Proper point in kernel using call_usermodehelper().

I don't think proper point is a big issue anyways. The 
only kernel user that really cares about video state is the
console.

But you would need to be careful to not schedule
the X server or a DRI client before it has run.
You would need a kind of "single process" mode.

I suspect doing it in assembly early would be cleaner.

-Andi


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[ole.rohne-vJEk5272eHo]

Andi> I don't think proper point is a big issue anyways. The only
Andi> kernel user that really cares about video state is the console.

And as most modern graphics adapters are bus master capable, how do
you make sure it is not hanging on the bus request at the moment you
re-enable BM arbitration?

Speaking as someone who's tried extensively to get the "usermode,
anytime"-approach to work, what really turned me off was not being
able to get the console working.

Andi> But you would need to be careful to not schedule
Andi> the X server or a DRI client before it has run.
Andi> You would need a kind of "single process" mode.

Andi> I suspect doing it in assembly early would be cleaner.

:-)

Ole



-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[Andi Kleen]

On Wed, Nov 03, 2004 at 06:13:09PM +0100, ole.rohne-vJEk5272eHo@public.gmane.org wrote:
> Andi> I don't think proper point is a big issue anyways. The only
> Andi> kernel user that really cares about video state is the console.
> 
> And as most modern graphics adapters are bus master capable, how do
> you make sure it is not hanging on the bus request at the moment you
> re-enable BM arbitration?

The machine is comming fresh out of suspend. Nobody should be doing
bus mastering. 
> 
> Speaking as someone who's tried extensively to get the "usermode,
> anytime"-approach to work, what really turned me off was not being
> able to get the console working.

When you want to restore a graphics mode in the X server you don't
have a console anyways, so I don't see the problem here.

earlyprintk to serial or vga if you're in text mode should always work.

-Andi


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[Matthew Garrett]

On Wed, 2004-11-03 at 18:21 +0100, Andi Kleen wrote:

> earlyprintk to serial or vga if you're in text mode should always work.

On most machines, vga textmode won't work until the card is POSTed
again.

-- 
Matthew Garrett | mjg59-1xO5oi07KQx4cg9Nei1l7Q@public.gmane.org



-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[Andi Kleen]

On Wed, Nov 03, 2004 at 06:01:38PM +0000, Matthew Garrett wrote:
> On Wed, 2004-11-03 at 18:21 +0100, Andi Kleen wrote:
> 
> > earlyprintk to serial or vga if you're in text mode should always work.
> 
> On most machines, vga textmode won't work until the card is POSTed
> again.

But serial does.  

-Andi


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[Matthew Garrett]

On Thu, 2004-11-04 at 05:15 +0100, Andi Kleen wrote:
> On Wed, Nov 03, 2004 at 06:01:38PM +0000, Matthew Garrett wrote:
> > On Wed, 2004-11-03 at 18:21 +0100, Andi Kleen wrote:
> > 
> > > earlyprintk to serial or vga if you're in text mode should always work.
> > 
> > On most machines, vga textmode won't work until the card is POSTed
> > again.
> 
> But serial does.  

Few modern laptops have serial ports. 

-- 
Matthew Garrett | mjg59-1xO5oi07KQx4cg9Nei1l7Q@public.gmane.org



-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[Andi Kleen]

On Thu, Nov 04, 2004 at 10:39:14AM +0000, Matthew Garrett wrote:
> On Thu, 2004-11-04 at 05:15 +0100, Andi Kleen wrote:
> > On Wed, Nov 03, 2004 at 06:01:38PM +0000, Matthew Garrett wrote:
> > > On Wed, 2004-11-03 at 18:21 +0100, Andi Kleen wrote:
> > > 
> > > > earlyprintk to serial or vga if you're in text mode should always work.
> > > 
> > > On most machines, vga textmode won't work until the card is POSTed
> > > again.
> > 
> > But serial does.  
> 
> Few modern laptops have serial ports. 

Yes, that's an issue. USB EHCI has a special polled debug port mode
for this, but so far nobody has written a nice driver for it.

But anyways, as said VGA doesn't help you to debug VGA POST.
You always need some other debug channel.

I heard other people use LED on parport hacks, which probably works.

Anyways, this thread doesn't seem to serve any useful purpose
anymore so I won't post to it anymore.

-Andi


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[ole.rohne-vJEk5272eHo]

Andi> I heard other people use LED on parport hacks, which probably works.

Modern laptops don't have parports either...

Anyway, my real objections to purely userspace POSTing are:

(1) it is hard to get it to work reliably (some method work with X11
    but not with text mode console, other methods work without DRM etc)
(2) the previous point goes to show that hardware sourcery belongs in
    kernel space (or before), as this is the only way to ensure proper
    ordering, bus mastering disable etc

Andi> Anyways, this thread doesn't seem to serve any useful purpose
Andi> anymore so I won't post to it anymore.

???

Ole




-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Re: Suspend-to-RAM on Sony Vaios

[Oisín Mac Fhearaí]

On Thursday 04 November 2004 13:49, ole.rohne-vJEk5272eHo@public.gmane.org wrote:
> Andi> I heard other people use LED on parport hacks, which probably works.
>
> Modern laptops don't have parports either...

What about the PS/2 port? I don't think I've seen any machines without one of 
these.

Oisín


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Re: Suspend-to-RAM on Sony Vaios

[Matthew Garrett]

On Thu, 2004-11-04 at 22:16 +0000, Oisín Mac Fhearaí wrote:
> On Thursday 04 November 2004 13:49, ole.rohne-vJEk5272eHo@public.gmane.org wrote:
> > Andi> I heard other people use LED on parport hacks, which probably works.
> >
> > Modern laptops don't have parports either...
> 
> What about the PS/2 port? I don't think I've seen any machines without one of 
> these.

My current laptop has 2 USB sockets, VGA, modem, ethernet and sound.

-- 
Matthew Garrett | mjg59-1xO5oi07KQx4cg9Nei1l7Q@public.gmane.org



-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_idU88&alloc_id\x12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[ole.rohne-vJEk5272eHo]

>> earlyprintk to serial or vga if you're in text mode should always work.

Matthew> On most machines, vga textmode won't work until the card is
Matthew> POSTed again.

And if the POSTing is left to X, it never works again. This is because
the X console switching code restores what it thinks is the proper PCI
config for text mode.

Ole




-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[ole.rohne-vJEk5272eHo]

Andi> On Wed, Nov 03, 2004 at 06:13:09PM +0100, ole.rohne-vJEk5272eHo@public.gmane.org wrote:
Andi> I don't think proper point is a big issue anyways. The only
Andi> kernel user that really cares about video state is the console.

>> And as most modern graphics adapters are bus master capable, how do
>> you make sure it is not hanging on the bus request at the moment you
>> re-enable BM arbitration?

Andi> The machine is comming fresh out of suspend. Nobody should be doing
Andi> bus mastering. 

There is absolutely no guarantee that the machine comes "fresh out of
suspend"! Here is my recollection of what the ACPI specs says: The
machine resumes with interrupts and BM disabled, it is the
responsibility of the OS to make sure all devices are "quiescent"
before reenabling interrupts and BM.

Once again I'd lament the fact that most of the device suspend/resume
code is not done from specs, but rather empirically based on overly
forgiving firmware implementations. Specifically, the code seems to
assume the hardware as it comes out of a boot, but there is no such
guarantee.

>> Speaking as someone who's tried extensively to get the "usermode,
>> anytime"-approach to work, what really turned me off was not being
>> able to get the console working.

Andi> When you want to restore a graphics mode in the X server you
Andi> don't have a console anyways, so I don't see the problem here.

If you decide you don't care about the console there is no
problem. But please don't advertise that as fully working S3
suspend/resume.

Andi> early printk to serial or vga if you're in text mode should
Andi> always work.

You mean "should always work" as in "not hanging the machine"?
Maybe...  It certainly doesn't "work" in the sense of displaying text
on the screen.

Ole




-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[Andi Kleen]

On Thu, Nov 04, 2004 at 09:24:29AM +0100, ole.rohne-vJEk5272eHo@public.gmane.org wrote:
> There is absolutely no guarantee that the machine comes "fresh out of
> suspend"! Here is my recollection of what the ACPI specs says: The

I don't see who should cause busmaster in flight operations
before suspend wakeup.

> code is not done from specs, but rather empirically based on overly
> forgiving firmware implementations. Specifically, the code seems to
> assume the hardware as it comes out of a boot, but there is no such
> guarantee.

Yes, that is the big bug of the ACPI specification. They should
have required the BIOS to post all hardware at least to the 
same state as it happens after initial power on. But Windows
chose to do it differently and now it is too late.

I think the best we can do is to use the BIOS infrastructure to
do this BIOS POST manually while still in real mode.

> 
> Andi> early printk to serial or vga if you're in text mode should
> Andi> always work.
> 
> You mean "should always work" as in "not hanging the machine"?
> Maybe...  It certainly doesn't "work" in the sense of displaying text
> on the screen.

Of course you cannot display anything on the screen when the video
card is not posted yet. I feel you're intentionally trying to misundertand
me.  If you want to debug the POST process you need some other debug
channel, like serial output or the usb debug console.

-Andi


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[Pavel Machek]

Hi!

> > Andi> early printk to serial or vga if you're in text mode should
> > Andi> always work.
> > 
> > You mean "should always work" as in "not hanging the machine"?
> > Maybe...  It certainly doesn't "work" in the sense of displaying text
> > on the screen.
> 
> Of course you cannot display anything on the screen when the video
> card is not posted yet. I feel you're intentionally trying to misundertand
> me.  If you want to debug the POST process you need some other debug
> channel, like serial output or the usb debug console.

This is quite hard to do. Bios does not init video card, and it does
initialize other hardware, either. Serials are not available on many
machines, and initializing usb from real mode would be way too
hard....
								Pavel
-- 
People were complaining that M$ turns users into beta-testers...
...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Suspend-to-RAM on Sony Vaios

[ole.rohne-vJEk5272eHo]

>> There is absolutely no guarantee that the machine comes "fresh out of
>> suspend"! Here is my recollection of what the ACPI specs says: The

Andi> I don't see who should cause busmaster in flight operations
Andi> before suspend wakeup.

I'm supposing non-initialized hardware could physically hang on the
bus master request line.

Ole



-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click