Skip to contentSkip to conversation

PM branch rebased to 2.6.29... for real this time

DISCUSSION Kevin Hilman 4 participants 8 messages Mar 25, 2009 – Apr 3, 2009

lore
Share & tools
lore
#0 lore raw

Hello,

The previous rebase was actually to 2.6.29-rc8. Now that 2.6.29 is out, I've rebased the PM brach onto linux-omap HEAD just after the 2.6.29 merge.

Minimal retention and off-mode on Beagle and RX51.

Kevin

#1 lore raw
Quoted 7 lines

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Kevin Hilman
Sent: Thursday, March 26, 2009 4:26 AM
To: [email protected]
Subject: PM branch rebased to 2.6.29... for real this time

Quoted 7 lines

Hello,

The previous rebase was actually to 2.6.29-rc8. Now that 2.6.29 is out, I've rebased the PM brach onto linux-omap HEAD just after the 2.6.29 merge.

Minimal retention and off-mode on Beagle and RX51.

Another problem that I found on OMAP3EVM:

When I compiled in CPUidle and CPUfreq (over omap3_evm_defconfig), the kernel did not boot-up. The last few statements are:

<3>clock: dpll5_ck failed transition to 'locked'
clock: dpll5_ck failed transition to 'locked'
<6>Disabling unused clock "dpll4_m6x2_ck"
Disabling unused clock "dpll4_m6x2_ck"
<6>Disabling unused clock "dpll3_m3x2_ck"
Disabling unused clock "dpll3_m3x2_ck"
<6>Disabling unused clock "sys_clkout1"
Disabling unused clock "sys_clkout1"

The PC is at the WFI statement.

Tried other combinations as well:

1) only CPUidle enabled - okay.
2) only CPUfreq enabled - not okay.

Best regards,
Sanjeev

Quoted 9 lines

Kevin
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#2 lore raw
Quoted 13 lines · Sanjeev
Quoted 13 lines

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Kevin Hilman
Sent: Thursday, March 26, 2009 4:26 AM
To: [email protected]
Subject: PM branch rebased to 2.6.29... for real this time

Hello,

The previous rebase was actually to 2.6.29-rc8. Now that 2.6.29 is out, I've rebased the PM brach onto linux-omap HEAD just after the 2.6.29 merge.

Quoted 6 lines
Quoted 1 line

Minimal retention and off-mode on Beagle and RX51.

Another problem that I found on OMAP3EVM:

When I compiled in CPUidle and CPUfreq (over omap3_evm_defconfig), the kernel did not boot-up. The last few statements are:

There are known problems with CPUfreq on top of the new clock notifiers series.

I am waiting for Rajendra to re-send his series which removes the virt clocks on top of the latest PM branch.

Quoted 9 lines

<3>clock: dpll5_ck failed transition to 'locked'
clock: dpll5_ck failed transition to 'locked'
<6>Disabling unused clock "dpll4_m6x2_ck"
Disabling unused clock "dpll4_m6x2_ck"
<6>Disabling unused clock "dpll3_m3x2_ck"
Disabling unused clock "dpll3_m3x2_ck"
<6>Disabling unused clock "sys_clkout1"
Disabling unused clock "sys_clkout1"

Quoted 7 lines

The PC is at the WFI statement.

Tried other combinations as well:

1) only CPUidle enabled - okay.
2) only CPUfreq enabled - not okay.

Sounds to me like CPUfreq is changing frequencies during bootup. Did you select ondemand as the default CPUfreq governor? If so, can you try with performance as the default governor.

If you're already using performance, then u-boot is setting a slower speed and CPUfreq may decide to change it during boot. If so, can you try the userspace governor as the default governor. This should prevent any automatic CPUFreq changes during bootup.

Kevin

#3 lore raw
Quoted 48 lines

-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Wednesday, April 01, 2009 10:42 AM
To: Premi, Sanjeev
Cc: [email protected]; Nayak, Rajendra
Subject: Re: PM branch rebased to 2.6.29... for real this time

Quoted 13 lines · Sanjeev
Quoted 13 lines

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Kevin Hilman
Sent: Thursday, March 26, 2009 4:26 AM
To: [email protected]
Subject: PM branch rebased to 2.6.29... for real this time

Hello,

The previous rebase was actually to 2.6.29-rc8. Now that 2.6.29 is out, I've rebased the PM brach onto linux-omap HEAD just after the 2.6.29 merge.

Quoted 6 lines
Quoted 1 line

Minimal retention and off-mode on Beagle and RX51.

Another problem that I found on OMAP3EVM:

When I compiled in CPUidle and CPUfreq (over omap3_evm_defconfig), the kernel did not boot-up. The last few statements are:

There are known problems with CPUfreq on top of the new clock notifiers series.

I am waiting for Rajendra to re-send his series which removes the virt clocks on top of the latest PM branch.

Quoted 11 lines

<3>clock: dpll5_ck failed transition to 'locked'
clock: dpll5_ck failed transition to 'locked'
<6>Disabling unused clock "dpll4_m6x2_ck"
Disabling unused clock "dpll4_m6x2_ck"
<6>Disabling unused clock "dpll3_m3x2_ck"
Disabling unused clock "dpll3_m3x2_ck"
<6>Disabling unused clock "sys_clkout1"
Disabling unused clock "sys_clkout1"

The PC is at the WFI statement.

Quoted 8 lines
Quoted 5 lines

Tried other combinations as well:

1) only CPUidle enabled - okay.
2) only CPUfreq enabled - not okay.

Sounds to me like CPUfreq is changing frequencies during bootup. Did you select ondemand as the default CPUfreq governor?

[sp] Yes. This is what I feel too. Only I was not clear why the process gets stuck at WFI. Haven't been able to debug further. So far...

Quoted 1 line

If so, can you try with performance as the default governor.

[sp] It was performance governor only.

Quoted 2 lines

If you're already using performance, then u-boot is setting a slower speed and CPUfreq may decide to change it during boot.

[sp] That is the case.

Quoted 10 lines

If so, can you
try the userspace governor as the default governor. This should prevent any automatic CPUFreq changes during bootup.

Kevin

#4 lore raw

"Premi, Sanjeev" <[email protected]> writes:

[...]

Quoted 6 lines
Quoted 2 lines

Sounds to me like CPUfreq is changing frequencies during bootup. Did you select ondemand as the default CPUfreq governor?

[sp] Yes. This is what I feel too. Only I was not clear why the process gets stuck at WFI. Haven't been able to debug further. So far...

Quoted 9 lines
Quoted 1 line

If so, can you try with performance as the default governor.

[sp] It was performance governor only.

Quoted 2 lines

If you're already using performance, then u-boot is setting a slower speed and CPUfreq may decide to change it during boot.

[sp] That is the case.

Can you try setting the default governor to userspace so that no DVFS changes will happen during boot?

Kevin

#5 lore raw

Hi Kevin, Sanjeev,

I'm having the same problem.

when clk_set_rate() is invoked in omap_cpu_init() as shown below:

	clk_set_rate(mpu_clk, policy->cpuinfo.max_freq * 1000);

It acquires the clocks mutex by the next statement.

	mutex_lock(&clocks_mutex);

But after invoking arch_clock->clk_set_rate() in the middle of clk_set_rate(), clock_set_rate() is invoked again. So deadlock occurs due to mutex_lock() in clock_set_rate().

I think setting the default governor can not help this deadlock.

Regards,
Kyuwon

Quoted 29 lines · Kevin Hilman

"Premi, Sanjeev" <[email protected]> writes:

[...]

Quoted 6 lines
Quoted 2 lines

Sounds to me like CPUfreq is changing frequencies during bootup. Did you select ondemand as the default CPUfreq governor?

[sp] Yes. This is what I feel too. Only I was not clear why the process gets stuck at WFI. Haven't been able to debug further. So far...

Quoted 9 lines
Quoted 1 line

If so, can you try with performance as the default governor.

[sp] It was performance governor only.

Quoted 2 lines

If you're already using performance, then u-boot is setting a slower speed and CPUfreq may decide to change it during boot.

[sp] That is the case.

Can you try setting the default governor to userspace so that no DVFS changes will happen during boot?

Kevin
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Signature · 1 line
Kyuwon
#6 lore raw

Yes, this locking problem requires Rajendra's patchset which needs to be updated for the PM branch.

Kevin

Quoted 51 lines · Kim Kyuwon

Hi Kevin, Sanjeev,

I'm having the same problem.

when clk_set_rate() is invoked in omap_cpu_init() as shown below:

	clk_set_rate(mpu_clk, policy->cpuinfo.max_freq * 1000);

It acquires the clocks mutex by the next statement.

	mutex_lock(&clocks_mutex);

But after invoking arch_clock->clk_set_rate() in the middle of clk_set_rate(), clock_set_rate() is invoked again. So deadlock occurs due to mutex_lock() in clock_set_rate().

I think setting the default governor can not help this deadlock.

Regards,
Kyuwon

Quoted 25 lines · Kevin Hilman

"Premi, Sanjeev" <[email protected]> writes:

[...]

Quoted 5 lines
Quoted 2 lines

Sounds to me like CPUfreq is changing frequencies during bootup. Did you select ondemand as the default CPUfreq governor?

[sp] Yes. This is what I feel too. Only I was not clear why the process gets stuck at WFI. Haven't been able to debug further. So far...

Quoted 7 lines
Quoted 1 line

If so, can you try with performance as the default governor.

[sp] It was performance governor only.

Quoted 2 lines

If you're already using performance, then u-boot is setting a slower speed and CPUfreq may decide to change it during boot.

[sp] That is the case.

Can you try setting the default governor to userspace so that no DVFS changes will happen during boot?

Kevin
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#7 lore raw
Quoted 10 lines

-----Original Message-----
From: Kevin Hilman [mailto:[email protected]]
Sent: Thursday, April 02, 2009 9:37 PM
To: Kim Kyuwon
Cc: Premi, Sanjeev; [email protected]; Nayak, Rajendra
Subject: Re: PM branch rebased to 2.6.29... for real this time

Yes, this locking problem requires Rajendra's patchset which needs to be
updated for the PM branch.

I will send the updated patch set on the latest PM branch maybe later today.

Quoted 64 lines

Kevin

Quoted 6 lines · Kim Kyuwon

Hi Kevin, Sanjeev,

I'm having the same problem.

when clk_set_rate() is invoked in omap_cpu_init() as shown below:

Quoted 8 lines
	clk_set_rate(mpu_clk, policy->cpuinfo.max_freq * 1000);

It acquires the clocks mutex by the next statement.

	mutex_lock(&clocks_mutex);

But after invoking arch_clock->clk_set_rate() in the middle of clk_set_rate(), clock_set_rate() is invoked again. So

deadlock occurs

Quoted 14 lines

due to mutex_lock() in clock_set_rate().

I think setting the default governor can not help this deadlock.

Regards,
Kyuwon

Quoted 5 lines · Kevin Hilman

"Premi, Sanjeev" <[email protected]> writes:

[...]

Quoted 1 line
Quoted 1 line

Sounds to me like CPUfreq is changing frequencies during

bootup. Did

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 1 line

you select ondemand as the default CPUfreq governor?

[sp] Yes. This is what I feel too. Only I was not clear

why the process

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines

gets stuck at WFI. Haven't been able to debug further. So far...

Quoted 1 line

If so, can you try with performance as the default governor.

[sp] It was performance governor only.

Quoted 1 line

If you're already using performance, then u-boot is

setting a slower

Quoted 4 lines
Quoted 4 lines
Quoted 3 lines
Quoted 1 line

speed and CPUfreq may decide to change it during boot.

[sp] That is the case.

Can you try setting the default governor to userspace so

that no DVFS

Quoted 5 lines
Quoted 5 lines

changes will happen during boot?

Kevin
--
To unsubscribe from this list: send the line "unsubscribe

linux-omap" in

Quoted 6 lines
Quoted 3 lines

the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html