EC SCI model?

EC SCI model?

[Nate Lawson]

I am a FreeBSD developer but this is a more general question with ACPI.  I
looked through the spec (especially 13 Embedded Controller) and I have a
question regarding the IBF/OBF stuff.  Testing has shown that I get
multiple callbacks of my EcGpeHandler() routine even when IBF and OBF have
not changed.  Isn't the GPE only supposed to trigger when they change (and
only once)?  I've gotten more reliability by waiting for the GPE, checking
the status and if it isn't what I'm waiting for, go back to sleep.

Also, what is the procedure for sleeping while holding the ACPI Global
Lock?  I need to sleep while waiting for the GPE to trigger between each
phase of read and write (i.e. command, address, data).  I'm not using
burst mode but may consider implementing it to make this easier and avoid
the sleep altogether.

Thanks,
-Nate


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

RE: EC SCI model?

[Grover, Andrew]

> From: Nate Lawson [mailto:nate-Y6VGUYTwhu0@public.gmane.org] 

> I am a FreeBSD developer but this is a more general question 
> with ACPI.  I
> looked through the spec (especially 13 Embedded Controller) 
> and I have a
> question regarding the IBF/OBF stuff.  Testing has shown that I get
> multiple callbacks of my EcGpeHandler() routine even when IBF 
> and OBF have
> not changed.  Isn't the GPE only supposed to trigger when 
> they change (and
> only once)?  I've gotten more reliability by waiting for the 
> GPE, checking
> the status and if it isn't what I'm waiting for, go back to sleep.

Hi Nate,

Please feel free to take a look at the Linux EC driver. (It's GPLed but
you should still be able to use it as a reference w/o contamination.)

We ended up not being fully interrupt-driven, because of the very
problem you mentioned, IIRC. Once the 3-phase ec read/write sequence has
begun, we busy-wait for IBF/OBF. This is of course not ideal but it
seems to work. Perhaps Paul would like to comment more, after he gets
back from vacation. ;-)

Regards -- Andy


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

RE: EC SCI model?

[Diefenbaugh, Paul S]

I'd echo Andy's recommendation to check out the Linux code.  We went
through quite a bit of wrenching before getting this to work right on
all of the various ECs - noting that most are incapable of handing
EC_SCIs in a purely interrupt-driven fashion...
___________________________________________ 
Paul Diefenbaugh 
Intel Corporation, MPG | MPA | System Software Architecture


-----Original Message-----
From: Grover, Andrew 
Sent: Tuesday, June 24, 2003 6:02 PM
To: 'Nate Lawson'; acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
Cc: Diefenbaugh, Paul S
Subject: RE: [ACPI] EC SCI model?

> From: Nate Lawson [mailto:nate-Y6VGUYTwhu0@public.gmane.org] 

> I am a FreeBSD developer but this is a more general question 
> with ACPI.  I
> looked through the spec (especially 13 Embedded Controller) 
> and I have a
> question regarding the IBF/OBF stuff.  Testing has shown that I get
> multiple callbacks of my EcGpeHandler() routine even when IBF 
> and OBF have
> not changed.  Isn't the GPE only supposed to trigger when 
> they change (and
> only once)?  I've gotten more reliability by waiting for the 
> GPE, checking
> the status and if it isn't what I'm waiting for, go back to sleep.

Hi Nate,

Please feel free to take a look at the Linux EC driver. (It's GPLed but
you should still be able to use it as a reference w/o contamination.)

We ended up not being fully interrupt-driven, because of the very
problem you mentioned, IIRC. Once the 3-phase ec read/write sequence has
begun, we busy-wait for IBF/OBF. This is of course not ideal but it
seems to work. Perhaps Paul would like to comment more, after he gets
back from vacation. ;-)

Regards -- Andy


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: EC SCI model?

[Ducrot Bruno]

On Tue, Jun 24, 2003 at 04:40:23PM -0700, Nate Lawson wrote:
> I am a FreeBSD developer but this is a more general question with ACPI.  I
> looked through the spec (especially 13 Embedded Controller) and I have a
> question regarding the IBF/OBF stuff.  Testing has shown that I get
> multiple callbacks of my EcGpeHandler() routine even when IBF and OBF have
> not changed.  Isn't the GPE only supposed to trigger when they change (and
> only once)?  I've gotten more reliability by waiting for the GPE, checking
> the status and if it isn't what I'm waiting for, go back to sleep.
> 
> Also, what is the procedure for sleeping while holding the ACPI Global
> Lock?  I need to sleep while waiting for the GPE to trigger between each
> phase of read and write (i.e. command, address, data).  I'm not using
> burst mode but may consider implementing it to make this easier and avoid
> the sleep altogether.
> 

Look at the NetBSD implementation of the EC.  It look stable for me, and do not
have the GPL issue...

-- 
Ducrot Bruno

--  Which is worse:  ignorance or apathy?
--  Don't know.  Don't care.


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

RE: EC SCI model?

[Nate Lawson]

On Tue, 24 Jun 2003, Diefenbaugh, Paul S wrote:
> I'd echo Andy's recommendation to check out the Linux code.  We went
> through quite a bit of wrenching before getting this to work right on
> all of the various ECs - noting that most are incapable of handing
> EC_SCIs in a purely interrupt-driven fashion...
> ___________________________________________
> Paul Diefenbaugh
> Intel Corporation, MPG | MPA | System Software Architecture

I ran into the same problem.  My laptop is reporting events as
level-triggered.  I'll examine other implementations as I rework things.

Thanks,
-Nate


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: EC SCI model?

[Ducrot Bruno]

Hi Paul, pleased to read you again on this list.

On Tue, Jun 24, 2003 at 10:54:59PM -0700, Diefenbaugh, Paul S wrote:
> I'd echo Andy's recommendation to check out the Linux code.  We went
> through quite a bit of wrenching before getting this to work right on
> all of the various ECs - noting that most are incapable of handing
> EC_SCIs in a purely interrupt-driven fashion...

When you done those tests, it was before the fix which forced the
SCI interrupt to be correctly initialized if not present in firmware
routing last year?
If yes, could you re-consider to redo the tests?

Cheers,

-- 
Ducrot Bruno

--  Which is worse:  ignorance or apathy?
--  Don't know.  Don't care.


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: EC SCI model?

[Nate Lawson]

On Thu, 26 Jun 2003, Ducrot Bruno wrote:
> On Tue, Jun 24, 2003 at 10:54:59PM -0700, Diefenbaugh, Paul S wrote:
> > I'd echo Andy's recommendation to check out the Linux code.  We went
> > through quite a bit of wrenching before getting this to work right on
> > all of the various ECs - noting that most are incapable of handing
> > EC_SCIs in a purely interrupt-driven fashion...
>
> When you done those tests, it was before the fix which forced the
> SCI interrupt to be correctly initialized if not present in firmware
> routing last year?
> If yes, could you re-consider to redo the tests?

Any sign of problems with burst mode?  I'm considering whether to enable
it for our EC support.

-Nate


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

RE: EC SCI model?

[Diefenbaugh, Paul S]

> When you done those tests, it was before the fix which forced the
> SCI interrupt to be correctly initialized if not present in firmware
> routing last year?

Our original implementation was done on the BeOS several years ago; then
ported to Linux.  I believe the problem is unrelated to the SCI issue
you've mentioned (an EC hardware bug).

___________________________________________ 
Paul Diefenbaugh 
Intel Corporation, MPG | MPA | System Software Architecture


-----Original Message-----
From: Ducrot Bruno [mailto:ducrot-kk6yZipjEM5g9hUCZPvPmw@public.gmane.org] 
Sent: Thursday, June 26, 2003 4:26 AM
To: Diefenbaugh, Paul S
Cc: Grover, Andrew; Nate Lawson; acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
Subject: Re: [ACPI] EC SCI model?

Hi Paul, pleased to read you again on this list.

On Tue, Jun 24, 2003 at 10:54:59PM -0700, Diefenbaugh, Paul S wrote:
> I'd echo Andy's recommendation to check out the Linux code.  We went
> through quite a bit of wrenching before getting this to work right on
> all of the various ECs - noting that most are incapable of handing
> EC_SCIs in a purely interrupt-driven fashion...

When you done those tests, it was before the fix which forced the
SCI interrupt to be correctly initialized if not present in firmware
routing last year?
If yes, could you re-consider to redo the tests?

Cheers,

-- 
Ducrot Bruno

--  Which is worse:  ignorance or apathy?
--  Don't know.  Don't care.


-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100006ave/direct;at.asp_061203_01/01

Re: EC SCI model?

[Ducrot Bruno]

On Mon, Jun 30, 2003 at 02:16:16PM -0700, Diefenbaugh, Paul S wrote:
> > When you done those tests, it was before the fix which forced the
> > SCI interrupt to be correctly initialized if not present in firmware
> > routing last year?
> 
> Our original implementation was done on the BeOS several years ago; then
> ported to Linux.  I believe the problem is unrelated to the SCI issue
> you've mentioned (an EC hardware bug).
> 

Is it possible then to get a fully interrupt-driven EC driver, but
fallback to polling via acpi blacklist?


-- 
Ducrot Bruno

--  Which is worse:  ignorance or apathy?
--  Don't know.  Don't care.


-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100006ave/direct;at.asp_061203_01/01

RE: EC SCI model?

[Diefenbaugh, Paul S]

> Is it possible then to get a fully interrupt-driven EC driver, 
> but fallback to polling via acpi blacklist?

Yes - we have seen a few good implementations ... but the side-effect of
a bad one is an interrupt storm that essentially hangs the host system.
Not necessarily the most user-friendly means for determining if/when a
system should be blacklisted... ;~)

___________________________________________ 
Paul Diefenbaugh 
Intel Corporation, MPG | MPA | System Software Architecture


-----Original Message-----
From: Ducrot Bruno [mailto:ducrot-kk6yZipjEM5g9hUCZPvPmw@public.gmane.org] 
Sent: Tuesday, July 01, 2003 5:46 AM
To: Diefenbaugh, Paul S
Cc: Grover, Andrew; Nate Lawson; acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
Subject: Re: [ACPI] EC SCI model?

On Mon, Jun 30, 2003 at 02:16:16PM -0700, Diefenbaugh, Paul S wrote:
> > When you done those tests, it was before the fix which forced the
> > SCI interrupt to be correctly initialized if not present in firmware
> > routing last year?
> 
> Our original implementation was done on the BeOS several years ago;
then
> ported to Linux.  I believe the problem is unrelated to the SCI issue
> you've mentioned (an EC hardware bug).
> 

Is it possible then to get a fully interrupt-driven EC driver, but
fallback to polling via acpi blacklist?


-- 
Ducrot Bruno

--  Which is worse:  ignorance or apathy?
--  Don't know.  Don't care.


-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100006ave/direct;at.asp_061203_01/01