2.6.28-rc1: rapid hotkeys break on eeepc again :(

DISCUSSION Alan Jenkins 5 participants 18 messages Oct 28, 2008 – Nov 3, 2008 ACPI

Share & tools

Alan Jenkins

author

2008-10-28 23:49 UTC #0 lore raw

Similar but different to the original symptoms of bugzilla #11089 (which is closed, fixed by the specific patch I tested). Hotkeys give up the ghost after being held down for a while, or pressed rapidly. However they don't get "out of sync" as in that bug. I've already tried the acpi-test tree, but that didn't fix it.

Bad: v2.6.28-rc1
Good: v2.6.27-89-g3156d1e

I can't see any non-trivial changes to the EC or eeepc-laptop drivers in that range, so it looks like I have to do this the hard way.

Bisecting: 3857 revisions left to test after this

Ow.

Alan

Alexey Starikovskiy

2008-10-29 00:03 UTC #1 lore raw

Please check if changing msleep to udelay helps.

Quoted 22 lines · Alan Jenkins

Similar but different to the original symptoms of bugzilla #11089 (which is closed, fixed by the specific patch I tested). Hotkeys give up the ghost after being held down for a while, or pressed rapidly. However they don't get "out of sync" as in that bug. I've already tried the acpi-test tree, but that didn't fix it.
Bad: v2.6.28-rc1
Good: v2.6.27-89-g3156d1e
I can't see any non-trivial changes to the EC or eeepc-laptop drivers in that range, so it looks like I have to do this the hard way.
Bisecting: 3857 revisions left to test after this
Ow.
Alan
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Alan Jenkins

author

2008-10-29 09:44 UTC #2 lore raw

Quoted 7 lines · Alexey Starikovskiy

Please check if changing msleep to udelay helps.
Quoted 5 lines · Alan Jenkins
Similar but different to the original symptoms of bugzilla #11089 (which is closed, fixed by the specific patch I tested). Hotkeys give up the ghost after being held down for a while, or pressed rapidly. However they don't get "out of sync" as in that bug. I've already tried the acpi-test tree, but that didn't fix it.

That was in acpi-test, unless I misread the changelog. It's the main
reason I tried that, I should have mentioned it, sorry. So no, it
didn't help.

Thanks
Alan

Alexey Starikovskiy

2008-10-29 14:12 UTC #3 lore raw

Quoted 11 lines · Alan Jenkins

Quoted 7 lines · Alexey Starikovskiy
Please check if changing msleep to udelay helps.
Quoted 5 lines · Alan Jenkins
Similar but different to the original symptoms of bugzilla #11089 (which is closed, fixed by the specific patch I tested). Hotkeys give up the ghost after being held down for a while, or pressed rapidly. However they don't get "out of sync" as in that bug. I've already tried the acpi-test tree, but that didn't fix it.
That was in acpi-test, unless I misread the changelog. It's the main
reason I tried that, I should have mentioned it, sorry. So no, it
didn't help.

Any luck with bisect?

Alan Jenkins

author

2008-10-29 14:27 UTC #4 lore raw

Quoted 10 lines · Alexey Starikovskiy

Quoted 9 lines · Alan Jenkins
Quoted 8 lines · Alexey Starikovskiy
Please check if changing msleep to udelay helps.
Quoted 6 lines · Alan Jenkins
Similar but different to the original symptoms of bugzilla #11089 (which
is closed, fixed by the specific patch I tested). Hotkeys give up the ghost after being held down for a while, or pressed rapidly. However they don't get "out of sync" as in that bug. I've already tried the acpi-test tree, but that didn't fix it.

Quoted 4 lines

Quoted 3 lines
That was in acpi-test, unless I misread the changelog. It's the main
reason I tried that, I should have mentioned it, sorry. So no, it
didn't help.
Any luck with bisect?

It didn't seem to like what I gave it. I ended up at a commit which lacked the EC rewrite - when as far as I'm concerned, I'm tracking a regression from after that point.

So I tried some manual fiddling, and I'm now starting a bisect between two different points on mainline. That's got to work :). It looks like it broke sometime after the main merge from the acpi tree.

$ git-describe
v2.6.27-7166-g765426e
$ git-bisect good
$ git-bisect bad v2.6.28-rc1
Bisecting: 215 revisions left to test after this

Alan Jenkins

author

2008-10-29 20:27 UTC #5 lore raw

Quoted 31 lines · Alan Jenkins

Quoted 14 lines · Alexey Starikovskiy
Quoted 12 lines · Alan Jenkins
Quoted 10 lines · Alexey Starikovskiy
Please check if changing msleep to udelay helps.
Quoted 7 lines · Alan Jenkins
Similar but different to the original symptoms of bugzilla #11089 (which
is closed, fixed by the specific patch I tested). Hotkeys give up the ghost after being held down for a while, or pressed rapidly. However they don't get "out of sync" as in that bug. I've already tried the acpi-test tree, but that didn't fix it.
Quoted 6 lines
Quoted 4 lines
That was in acpi-test, unless I misread the changelog. It's the main
reason I tried that, I should have mentioned it, sorry. So no, it
didn't help.
Any luck with bisect?
It didn't seem to like what I gave it. I ended up at a commit which lacked the EC rewrite - when as far as I'm concerned, I'm tracking a regression from after that point.
So I tried some manual fiddling, and I'm now starting a bisect between two different points on mainline. That's got to work :). It looks like it broke sometime after the main merge from the acpi tree.

Quoted 6 lines

$ git-describe
v2.6.27-7166-g765426e
$ git-bisect good
$ git-bisect bad v2.6.28-rc1
Bisecting: 215 revisions left to test after this

Well, that didn't work.

I wasn't quite methodical enough. It depends on my doing something... it might be the lid switch, or using the sound, or network, or switching from battery to AC, or removing the battery...

So it needn't be a regression from the patches I tested before. More likely it's a problem with certain action(s) that I've only performed recently.

Alan

Alexey Starikovskiy

2008-10-29 20:34 UTC #6 lore raw

Quoted 1 line · Alan Jenkins

Well, that didn't work.

Pity...
Could you produce a debug log from EC? Probably it's a good time to open a new bug report...

Regards,
Alex.

Alan Jenkins

author

2008-10-31 10:26 UTC #7 lore raw

Bisection result: eeepc hotk breakage

Quoted 6 lines · Alexey Starikovskiy

Quoted 1 line · Alan Jenkins
Well, that didn't work.
Pity...
Could you produce a debug log from EC? Probably it's a good time to open a new bug report...

<http://bugzilla.kernel.org/show_bug.cgi?id=11917>
"Asus Eee PC hotkeys stop working after prolonged usage"

I stuck at it, found a reliable test and bisected successfully. 2.6.28-rc2 is broken, but I can fix it by reverting the bad commit. (There were some conflicts, I fixed up the dock driver manually and avoided building the rest).

27663c5855b10af9ec67bc7dfba001426ba21222 is first bad commit commit 27663c5855b10af9ec67bc7dfba001426ba21222

Author: Matthew Wilcox <[email protected]>
Date:   Fri Oct 10 02:22:59 2008 -0400

ACPI: Change acpi_evaluate_integer to support 64-bit on 32-bit kernels

As of version 2.0, ACPI can return 64-bit integers. The current acpi_evaluate_integer only supports 64-bit integers on 64-bit platforms. Change the argument to take a pointer to an acpi_integer so we support 64-bit integers on all platforms.

lenb: replaced use of "acpi_integer" with "unsigned long long"
lenb: fixed bug in acpi_thermal_trips_update()

Signed-off-by Matthew Wilcox <[email protected]>

Signed-off-by Len Brown <[email protected]>

Alexey Starikovskiy

2008-10-31 13:51 UTC #8 lore raw

Re: Bisection result: eeepc hotk breakage

Quoted 7 lines · Alan Jenkins

Quoted 5 lines · Alexey Starikovskiy
Quoted 1 line · Alan Jenkins
Well, that didn't work.
Pity...
Could you produce a debug log from EC? Probably it's a good time to open a new bug report...

Quoted 5 lines

<http://bugzilla.kernel.org/show_bug.cgi?id=11917>
"Asus Eee PC hotkeys stop working after prolonged usage"
I stuck at it, found a reliable test and bisected successfully. 2.6.28-rc2 is broken, but I can fix it by reverting the bad commit. (There were some conflicts, I fixed up the dock driver manually and avoided building the rest).

Alan, to make things clear -- reverting following commit makes EC work without a problem?

Thanks,
Alex.

Quoted 20 lines

27663c5855b10af9ec67bc7dfba001426ba21222 is first bad commit commit 27663c5855b10af9ec67bc7dfba001426ba21222
Author: Matthew Wilcox <[email protected]>
Date:   Fri Oct 10 02:22:59 2008 -0400
ACPI: Change acpi_evaluate_integer to support 64-bit on 32-bit kernels
As of version 2.0, ACPI can return 64-bit integers. The current acpi_evaluate_integer only supports 64-bit integers on 64-bit platforms. Change the argument to take a pointer to an acpi_integer so we support 64-bit integers on all platforms.
lenb: replaced use of "acpi_integer" with "unsigned long long"
lenb: fixed bug in acpi_thermal_trips_update()
Signed-off-by Matthew Wilcox <[email protected]>
Signed-off-by Len Brown <[email protected]>

Alan Jenkins

author

2008-10-31 13:53 UTC #9 lore raw

Re: Bisection result: eeepc hotk breakage

Quoted 12 lines · Alexey Starikovskiy

Quoted 10 lines · Alan Jenkins
Quoted 5 lines · Alexey Starikovskiy
Quoted 1 line · Alan Jenkins
Well, that didn't work.
Pity...
Could you produce a debug log from EC? Probably it's a good time to open a new bug report...
<http://bugzilla.kernel.org/show_bug.cgi?id=11917>
"Asus Eee PC hotkeys stop working after prolonged usage"

Quoted 8 lines

Quoted 4 lines
I stuck at it, found a reliable test and bisected successfully. 2.6.28-rc2 is broken, but I can fix it by reverting the bad commit. (There were some conflicts, I fixed up the dock driver manually and avoided building the rest).
Alan, to make things clear -- reverting following commit makes EC work without
a problem?

Yes, that's right.

Quoted 23 lines

Quoted 22 lines
27663c5855b10af9ec67bc7dfba001426ba21222 is first bad commit commit 27663c5855b10af9ec67bc7dfba001426ba21222
Author: Matthew Wilcox <[email protected]>
Date:   Fri Oct 10 02:22:59 2008 -0400
ACPI: Change acpi_evaluate_integer to support 64-bit on 32-bit
kernels
As of version 2.0, ACPI can return 64-bit integers. The current acpi_evaluate_integer only supports 64-bit integers on 64-bit
platforms.
Change the argument to take a pointer to an acpi_integer so we
support
64-bit integers on all platforms.
lenb: replaced use of "acpi_integer" with "unsigned long long"
lenb: fixed bug in acpi_thermal_trips_update()
Signed-off-by Matthew Wilcox <[email protected]>
Signed-off-by Len Brown <[email protected]>

Alexey Starikovskiy

2008-10-31 14:15 UTC #10 lore raw

Re: Bisection result: eeepc hotk breakage

Quoted 15 lines · Alan Jenkins

Quoted 14 lines · Alexey Starikovskiy
Quoted 13 lines · Alan Jenkins
Quoted 5 lines · Alexey Starikovskiy
Quoted 1 line · Alan Jenkins
Well, that didn't work.
Pity...
Could you produce a debug log from EC? Probably it's a good time to open a new bug report...
<http://bugzilla.kernel.org/show_bug.cgi?id=11917>
"Asus Eee PC hotkeys stop working after prolonged usage"
I stuck at it, found a reliable test and bisected successfully. 2.6.28-rc2 is broken, but I can fix it by reverting the bad commit. (There were some conflicts, I fixed up the dock driver manually and avoided building the rest).

Quoted 7 lines

Quoted 4 lines
Alan, to make things clear -- reverting following commit makes EC work without
a problem?
Yes, that's right.

Great!
What kernel version is that -- 32 or 64?

Quoted 24 lines

Quoted 23 lines
Quoted 23 lines
27663c5855b10af9ec67bc7dfba001426ba21222 is first bad commit commit 27663c5855b10af9ec67bc7dfba001426ba21222
Author: Matthew Wilcox <[email protected]>
Date:   Fri Oct 10 02:22:59 2008 -0400
ACPI: Change acpi_evaluate_integer to support 64-bit on 32-bit
kernels
As of version 2.0, ACPI can return 64-bit integers. The current acpi_evaluate_integer only supports 64-bit integers on 64-bit
platforms.
Change the argument to take a pointer to an acpi_integer so we
support
64-bit integers on all platforms.
lenb: replaced use of "acpi_integer" with "unsigned long long"
lenb: fixed bug in acpi_thermal_trips_update()
Signed-off-by Matthew Wilcox <[email protected]>
Signed-off-by Len Brown <[email protected]>

Alan Jenkins

author

2008-10-31 14:16 UTC #11 lore raw

Re: Bisection result: eeepc hotk breakage

Quoted 22 lines · Alexey Starikovskiy

Quoted 21 lines · Alan Jenkins
Quoted 18 lines · Alexey Starikovskiy
Quoted 6 lines · Alan Jenkins
Quoted 5 lines · Alexey Starikovskiy
Quoted 1 line · Alan Jenkins
Well, that didn't work.
Pity...
Could you produce a debug log from EC? Probably it's a good time to open a new bug report...
Quoted 7 lines
<http://bugzilla.kernel.org/show_bug.cgi?id=11917>
"Asus Eee PC hotkeys stop working after prolonged usage"
I stuck at it, found a reliable test and bisected successfully. 2.6.28-rc2 is broken, but I can fix it by reverting the bad commit. (There were some conflicts, I fixed up the dock driver manually and avoided building the rest).
Alan, to make things clear -- reverting following commit makes EC work without
a problem?

Quoted 3 lines

Quoted 1 line
Yes, that's right.
Great!
What kernel version is that -- 32 or 64?

32 bit, my EeePC doesn't do 64.

Quoted 25 lines

Quoted 24 lines
Quoted 23 lines
Quoted 23 lines
27663c5855b10af9ec67bc7dfba001426ba21222 is first bad commit commit 27663c5855b10af9ec67bc7dfba001426ba21222
Author: Matthew Wilcox <[email protected]>
Date:   Fri Oct 10 02:22:59 2008 -0400
ACPI: Change acpi_evaluate_integer to support 64-bit on 32-bit
kernels
As of version 2.0, ACPI can return 64-bit integers. The current acpi_evaluate_integer only supports 64-bit integers on 64-bit
platforms.
Change the argument to take a pointer to an acpi_integer so we
support
64-bit integers on all platforms.
lenb: replaced use of "acpi_integer" with "unsigned long long"
lenb: fixed bug in acpi_thermal_trips_update()
Signed-off-by Matthew Wilcox <[email protected]>
Signed-off-by Len Brown <[email protected]>

Alexey Starikovskiy

2008-10-31 14:54 UTC #12 lore raw

Re: Bisection result: eeepc hotk breakage

Quoted 22 lines · Alan Jenkins

Quoted 21 lines · Alexey Starikovskiy
Quoted 20 lines · Alan Jenkins
Quoted 18 lines · Alexey Starikovskiy
Quoted 6 lines · Alan Jenkins
Quoted 5 lines · Alexey Starikovskiy
Quoted 1 line · Alan Jenkins
Well, that didn't work.
Pity...
Could you produce a debug log from EC? Probably it's a good time to open a new bug report...
Quoted 7 lines
<http://bugzilla.kernel.org/show_bug.cgi?id=11917>
"Asus Eee PC hotkeys stop working after prolonged usage"
I stuck at it, found a reliable test and bisected successfully. 2.6.28-rc2 is broken, but I can fix it by reverting the bad commit. (There were some conflicts, I fixed up the dock driver manually and avoided building the rest).
Alan, to make things clear -- reverting following commit makes EC work without
a problem?

Quoted 5 lines

Quoted 3 lines
Quoted 1 line
Yes, that's right.
Great!
What kernel version is that -- 32 or 64?
32 bit, my EeePC doesn't do 64.

is the attached patch enough to fix ?

drivers/acpi/ec.c +1 -0

index ef42316..e0f1534 100644
@@ -751,6 +751,7 @@ ec_parse_device(acpi_handle handle, u32 Level, void *context, void **retval)
 		return status;
 	ec->gpe = tmp;
 	/* Use the global lock for all EC transactions? */
+	tmp = 0;
 	acpi_evaluate_integer(handle, "_GLK", NULL, &tmp);
 	ec->global_lock = tmp;
 	ec->handle = handle;

Alan Jenkins

author

2008-10-31 15:47 UTC #13 lore raw

Re: Bisection result: eeepc hotk breakage

Quoted 24 lines · Alexey Starikovskiy

Quoted 23 lines · Alan Jenkins
Quoted 22 lines · Alexey Starikovskiy
Quoted 21 lines · Alan Jenkins
Quoted 19 lines · Alexey Starikovskiy
Quoted 6 lines · Alan Jenkins
Quoted 5 lines · Alexey Starikovskiy
Quoted 1 line · Alan Jenkins
Well, that didn't work.
Pity...
Could you produce a debug log from EC? Probably it's a good time to open a new bug report...
Quoted 7 lines
<http://bugzilla.kernel.org/show_bug.cgi?id=11917>
"Asus Eee PC hotkeys stop working after prolonged usage"
I stuck at it, found a reliable test and bisected successfully. 2.6.28-rc2 is broken, but I can fix it by reverting the bad commit. (There were some conflicts, I fixed up the dock driver manually and avoided building the rest).
Alan, to make things clear -- reverting following commit makes EC work
without
a problem?

Quoted 7 lines

Quoted 5 lines
Quoted 3 lines
Quoted 1 line
Yes, that's right.
Great!
What kernel version is that -- 32 or 64?
32 bit, my EeePC doesn't do 64.
is the attached patch enough to fix ?

Nice. Yes, that fixes it, thanks!

I checked the commit diff for any similar potential errors. Most of the changes are "obviously correct". However, there are changes in thermal.c which look similar to the one that caused the problem in the EC driver.

I think someone knowledgeable needs to audit the changes to the thermal driver, in case the Thinkpads start melting again :-P. The commit#
again: 27663c5855b10af9ec67bc7dfba001426ba21222.

Alan

Alexey Starikovskiy

2008-10-31 16:07 UTC #14 lore raw

Re: Bisection result: eeepc hotk breakage

Quoted 3 lines · Alan Jenkins

Nice. Yes, that fixes it, thanks!

great, I've updated #11917.

Quoted 10 lines

I checked the commit diff for any similar potential errors. Most of the changes are "obviously correct". However, there are changes in thermal.c which look similar to the one that caused the problem in the EC driver.
I think someone knowledgeable needs to audit the changes to the thermal driver, in case the Thinkpads start melting again :-P. The commit#
again: 27663c5855b10af9ec67bc7dfba001426ba21222.

no, everything is fine in thermal -- they check status after every call.

Thanks again,
Alex.

Zhao Yakui

2008-11-03 00:56 UTC #15 lore raw

Re: Bisection result: eeepc hotk breakage

In reply to Alexey Starikovskiy

Quoted 18 lines · Alexey Starikovskiy

Quoted 17 lines · Alan Jenkins
Quoted 16 lines · Alexey Starikovskiy
Quoted 15 lines · Alan Jenkins
Quoted 14 lines · Alexey Starikovskiy
Quoted 13 lines · Alan Jenkins
Quoted 5 lines · Alexey Starikovskiy
Quoted 1 line · Alan Jenkins
Well, that didn't work.
Pity...
Could you produce a debug log from EC? Probably it's a good time to open a new bug report...
<http://bugzilla.kernel.org/show_bug.cgi?id=11917>
"Asus Eee PC hotkeys stop working after prolonged usage"
I stuck at it, found a reliable test and bisected successfully. 2.6.28-rc2 is broken, but I can fix it by reverting the bad commit. (There were some conflicts, I fixed up the dock driver manually and avoided building the rest).

Quoted 8 lines

Quoted 8 lines
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
Alan, to make things clear -- reverting following commit makes EC work without
a problem?
Yes, that's right.
Great!
What kernel version is that -- 32 or 64?

Thanks for the patch.
It seems that the code path in the function of ec_parse_device is
incorrect.
Whether the ec->global_lock is assigned to a value should be based on
the returned value of acpi_evaluate_integer(handle, "_GLK", NULL, &tmp).
If AE_OK is returned, the ec->global_lock is assigned to true.
Otherwise it should be False.
Is this OK?

Thanks.

Quoted 4 lines

Quoted 2 lines
32 bit, my EeePC doesn't do 64.
is the attached patch enough to fix ?

Rafael J. Wysocki

2008-10-31 14:54 UTC #16 lore raw

Re: Bisection result: eeepc hotk breakage

In reply to Alan Jenkins

Quoted 17 lines · Alan Jenkins

Quoted 16 lines · Alexey Starikovskiy
Quoted 14 lines · Alan Jenkins
Quoted 5 lines · Alexey Starikovskiy
Quoted 1 line · Alan Jenkins
Well, that didn't work.
Pity...
Could you produce a debug log from EC? Probably it's a good time to open a new bug report...
<http://bugzilla.kernel.org/show_bug.cgi?id=11917>
"Asus Eee PC hotkeys stop working after prolonged usage"
I stuck at it, found a reliable test and bisected successfully. 2.6.28-rc2 is broken, but I can fix it by reverting the bad commit. (There were some conflicts, I fixed up the dock driver manually and avoided building the rest).

Quoted 7 lines

Quoted 4 lines
Alan, to make things clear -- reverting following commit makes EC work without
a problem?
Yes, that's right.

Hm, so in fact it _is_ a regression from 2.6.27 (this commit could not be present in that kernel).

Thanks,
Rafael

Quoted 28 lines

Quoted 15 lines
Quoted 15 lines
27663c5855b10af9ec67bc7dfba001426ba21222 is first bad commit commit 27663c5855b10af9ec67bc7dfba001426ba21222
Author: Matthew Wilcox <[email protected]>
Date:   Fri Oct 10 02:22:59 2008 -0400
ACPI: Change acpi_evaluate_integer to support 64-bit on 32-bit
kernels
As of version 2.0, ACPI can return 64-bit integers. The current acpi_evaluate_integer only supports 64-bit integers on 64-bit
platforms.
Change the argument to take a pointer to an acpi_integer so we
support
64-bit integers on all platforms.
Quoted 6 lines
Quoted 5 lines
lenb: replaced use of "acpi_integer" with "unsigned long long"
lenb: fixed bug in acpi_thermal_trips_update()
Signed-off-by Matthew Wilcox <[email protected]>
Signed-off-by Len Brown <[email protected]>
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Alan Jenkins

author

2008-10-31 15:06 UTC #17 lore raw

Re: Bisection result: eeepc hotk breakage

Quoted 29 lines · Rafael J. Wysocki

Quoted 27 lines · Alan Jenkins
Quoted 25 lines · Alexey Starikovskiy
Quoted 10 lines · Alan Jenkins
Quoted 8 lines · Alexey Starikovskiy
Quoted 2 lines · Alan Jenkins
Well, that didn't work.
Pity...
Could you produce a debug log from EC? Probably it's a good time to open a new bug report...
Quoted 8 lines
<http://bugzilla.kernel.org/show_bug.cgi?id=11917>
"Asus Eee PC hotkeys stop working after prolonged usage"
I stuck at it, found a reliable test and bisected successfully. 2.6.28-rc2 is broken, but I can fix it by reverting the bad commit. (There were some conflicts, I fixed up the dock driver manually and avoided building the rest).
Alan, to make things clear -- reverting following commit makes EC work without
a problem?

Quoted 6 lines

Quoted 2 lines
Yes, that's right.
Hm, so in fact it _is_ a regression from 2.6.27 (this commit could not be present in that kernel).

Yes, I suppose it is distinct from the previous EC regression.