Skip to contentSkip to conversation

Suspend to RAM broken on BeagleBoard ?

DISCUSSION Thomas Petazzoni 4 participants 10 messages Sep 10, 2010 – Sep 13, 2010

lore
Share & tools
lore
#0 lore raw

Hello,

I've tested the current linux-omap Git tree and Kevin's pm-core branch on a BeagleBoard using the omap3_defconfig, and when trying a suspend to RAM, I get a "Class driver suspend failed for cpu0" :

# echo mem > /sys/power/state 
[   93.395050] PM: Syncing filesystems ... done.
[   93.410827] Freezing user space processes ... (elapsed 0.01 seconds) done.
[   93.434570] Freezing remaining freezable tasks ... (elapsed 0.01 seconds) done.
[   93.458557] Suspending console(s) (use no_console_suspend to debug)
[   93.590728] PM: suspend of devices complete after 119.629 msecs
[   93.593902] PM: late suspend of devices complete after 3.112 msecs
[   93.594421] Class driver suspend failed for cpu0
[   93.595153] PM: early resume of devices complete after 0.518 msecs
[   93.887603] PM: resume of devices complete after 292.052 msecs
[   93.922851] Restarting tasks ... done.

Is this a known issue ?

I also have a problem with USB support, both on linux-omap and Kevin's pm-core branch. My USB Ethernet adapter, supported by the asix driver, isn't detected anymore. With a pm git tree of a couple of weeks ago, it was working. This is again with just a plain omap3_defconfig.

Thanks!

Thomas

Signature · 4 lines
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com
#1 lore raw

Hi,

Quoted 4 lines · Thomas Petazzoni

I also have a problem with USB support, both on linux-omap and Kevin's pm-core branch. My USB Ethernet adapter, supported by the asix driver, isn't detected anymore. With a pm git tree of a couple of weeks ago, it was working. This is again with just a plain omap3_defconfig.

is it attached to EHCI/OHCI or musb port ?

Signature · 1 line
balbi
#2 lore raw
Quoted 8 lines · On Fri, 10 Sep 2010 20:00:53 +0300 Felipe Balbi <[email protected]> wrote:
Quoted 5 lines · Thomas Petazzoni

I also have a problem with USB support, both on linux-omap and Kevin's pm-core branch. My USB Ethernet adapter, supported by the asix driver, isn't detected anymore. With a pm git tree of a couple of weeks ago, it was working. This is again with just a plain omap3_defconfig.

is it attached to EHCI/OHCI or musb port ?

It's a normal USB Ethernet adapter that is attached to the EHCI/OCHI port of the BeagleBoard. I have a revision C3.

With older Git versions from the PM tree, it used to work, even if it wasn't 100% stable. But at least it was detected, and I was able to nfs mount the root filesystem. I don't know if I can easily bisect this, as I think the PM branches are constantly rebased.

Thanks,

Thomas

Signature · 4 lines
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com
#4 lore raw
In reply to Thomas Petazzoni
Quoted 6 lines · Thomas Petazzoni

Hello,

I've tested the current linux-omap Git tree and Kevin's pm-core branch on a BeagleBoard using the omap3_defconfig, and when trying a suspend to RAM, I get a "Class driver suspend failed for cpu0" :

Quoted 8 lines
# echo mem > /sys/power/state 
[   93.395050] PM: Syncing filesystems ... done.
[   93.410827] Freezing user space processes ... (elapsed 0.01 seconds) done.
[   93.434570] Freezing remaining freezable tasks ... (elapsed 0.01 seconds) done.
[   93.458557] Suspending console(s) (use no_console_suspend to debug)
[   93.590728] PM: suspend of devices complete after 119.629 msecs
[   93.593902] PM: late suspend of devices complete after 3.112 msecs
[   93.594421] Class driver suspend failed for cpu0

It's the CPUfreq driver that's failing to suspend.

Quoted 5 lines
[   93.595153] PM: early resume of devices complete after 0.518 msecs
[   93.887603] PM: resume of devices complete after 292.052 msecs
[   93.922851] Restarting tasks ... done.

Is this a known issue ?

Yes, CPUfreq is not supported in l-o master or pm-core (only in the full pm branch.) If you Kconfig out CPUfreq, this failure will go away.

[...]

Kevin

#5 lore raw
Quoted 1 line · On Fri, 10 Sep 2010 14:54:23 -0700 Kevin Hilman <[email protected]>…

It's the CPUfreq driver that's failing to suspend.

Ok.

Quoted 3 lines

Yes, CPUfreq is not supported in l-o master or pm-core (only in the full pm branch.) If you Kconfig out CPUfreq, this failure will go away.

Aah, yes, I forgot about this. I know it's temporary, but wouldn't it be good to disable CPUfreq in omap3_defconfig in this case ?

BTW, I'm now seeing:

[   37.830596] Powerdomain (core_pwrdm) didn't enter target state 1
[   37.830627] Powerdomain (dss_pwrdm) didn't enter target state 1

when the system is woken up.

Concerning the USB issue, no idea ?

Thanks,

Thomas

Signature · 4 lines
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com
#3 lore raw
In reply to Thomas Petazzoni

Hi,

Quoted 11 lines · Thomas Petazzoni
Quoted 8 lines · Thomas Petazzoni
Quoted 5 lines

I also have a problem with USB support, both on linux-omap and Kevin's pm-core branch. My USB Ethernet adapter, supported by the asix driver, isn't detected anymore. With a pm git tree of a couple of weeks ago, it was working. This is again with just a plain omap3_defconfig.

is it attached to EHCI/OHCI or musb port ?

It's a normal USB Ethernet adapter that is attached to the EHCI/OCHI port of the BeagleBoard. I have a revision C3.

Would you have any debugging dumps to share ? What shows up on dmesg after you attach the device ? (please clear the message ring buffer before attaching the device with dmesg -c).

Signature · 1 line
balbi
#6 lore raw
In reply to Thomas Petazzoni
Quoted 5 lines · Thomas Petazzoni
Quoted 1 line · On Fri, 10 Sep 2010 14:54:23 -0700 Kevin Hilman <[email protected]>…

It's the CPUfreq driver that's failing to suspend.

Quoted 8 lines

Ok.

Quoted 3 lines

Yes, CPUfreq is not supported in l-o master or pm-core (only in the full pm branch.) If you Kconfig out CPUfreq, this failure will go away.

Aah, yes, I forgot about this. I know it's temporary, but wouldn't it be good to disable CPUfreq in omap3_defconfig in this case ?

Yes.

Quoted 6 lines

BTW, I'm now seeing:

[   37.830596] Powerdomain (core_pwrdm) didn't enter target state 1
[   37.830627] Powerdomain (dss_pwrdm) didn't enter target state 1

when the system is woken up.

Looks like DSS has not been initialized, so is not hitting RET.

Do you have the DSS driver enabled and built-in? Without this (and until we have DSS converted to hwmod) the DSS HW is not in a state that can hit the low power mode. If the DSS driver is allowed to initialize, even without a panel, it should set the DSS HW in a state that will then allow RET.

Kevin

#7 lore raw

Hello,

Quoted 7 lines · On Mon, 13 Sep 2010 07:56:54 -0700 Kevin Hilman <[email protected]>…

Looks like DSS has not been initialized, so is not hitting RET.

Do you have the DSS driver enabled and built-in? Without this (and until we have DSS converted to hwmod) the DSS HW is not in a state that can hit the low power mode. If the DSS driver is allowed to initialize, even without a panel, it should set the DSS HW in a state that will then allow RET.

Indeed:

$ grep DSS .config
# CONFIG_OMAP2_DSS is not set

It's not enabled simply because it isn't enabled by omap3_defconfig. So maybe it should also be fixed for this ?

I have zero knowledge of DSS, but shouldn't the kernel be able to suspend properly even if the DSS driver is disabled ? But maybe it's too complicated to have the DSS init sequence required to allow the system to hit RET outside the DSS driver. Thoughts ?

Regards,

Thomas

Signature · 4 lines
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com
#8 lore raw
Quoted 4 lines

I have zero knowledge of DSS, but shouldn't the kernel be able to suspend properly even if the DSS driver is disabled ? But maybe it's too complicated to have the DSS init sequence required to allow the system to hit RET outside the DSS driver. Thoughts ?

I'm not sure if it's supposed to work from reset state but there are chances DSS is left running by your bootloader.

#9 lore raw
In reply to Thomas Petazzoni
Quoted 15 lines · Thomas Petazzoni

Hello,

Quoted 7 lines · On Mon, 13 Sep 2010 07:56:54 -0700 Kevin Hilman <[email protected]>…

Looks like DSS has not been initialized, so is not hitting RET.

Do you have the DSS driver enabled and built-in? Without this (and until we have DSS converted to hwmod) the DSS HW is not in a state that can hit the low power mode. If the DSS driver is allowed to initialize, even without a panel, it should set the DSS HW in a state that will then allow RET.

Indeed:

Quoted 8 lines
$ grep DSS .config
# CONFIG_OMAP2_DSS is not set

It's not enabled simply because it isn't enabled by omap3_defconfig. So maybe it should also be fixed for this ?

I have zero knowledge of DSS, but shouldn't the kernel be able to suspend properly even if the DSS driver is disabled ?

Yes, it should, and it's almost there.

Quoted 2 lines

But maybe it's too complicated to have the DSS init sequence required to allow the system to hit RET outside the DSS driver. Thoughts ?

One of the multiple motivations for moving to the omap_hwmod model of describing all the HW IP blocks in a central location is so that we have a common init and reset for every on-chip device.

Once an omap_hwmod exists for a device, during init, the hwmod core will ensure that that block is properly initialized, reset and in a state that will not prevent the chip from entering low-power sleep. This will happen in the hwmod core, and is independent of the existence of a driver for that device, so will solve problems like this one.

Currently, we have a hack here and a hack there for common devices that the bootloader blindy leaves in an state that prevents low-power states, but this is clearly a hack and does not scale. omap_hwmod is the solution to scaling this to every IP block.

Kevin