Skip to contentSkip to conversation

Broken cpuidle on PM branch?

DISCUSSION Amit Kucheria 3 participants 7 messages Oct 12, 2009 – Oct 14, 2009

lore
Share & tools
lore
#0 lore raw

Hi,

I am testing twl4030 script optimisations on the current PM branch. But I am seeing the board (RX51) freeze when CPU_IDLE is enabled in the config.

Is it known to work on other boards?

Regards,
Amit

Signature · 3 lines
-------------------------------------------------------------------------
Amit Kucheria,				 Kernel Developer, Verdurent
-------------------------------------------------------------------------
#1 lore raw

Hello Amit,

Quoted 6 lines · ext Amit Kucheria

Hi,

I am testing twl4030 script optimisations on the current PM branch. But I am seeing the board (RX51) freeze when CPU_IDLE is enabled in the config.

Is it known to work on other boards?

I'm actually seeing this same behavior. But the board actually does not freeze. If you keep a key pressed of send a sysrq through serial line then you eventually get it back.

It seams to be something related to serial driver and wakeup ?

Quoted 11 lines

Regards,
Amit

Signature · 7 lines
-------------------------------------------------------------------------
Amit Kucheria,				 Kernel Developer, Verdurent
-------------------------------------------------------------------------
--
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
Eduardo Valentin
#2 lore raw
Quoted 8 lines · Eduardo Valentin

Hello Amit,

Quoted 5 lines · ext Amit Kucheria

Hi,

I am testing twl4030 script optimisations on the current PM branch. But I am seeing the board (RX51) freeze when CPU_IDLE is enabled in the config.

Quoted 8 lines
Quoted 1 line

Is it known to work on other boards?

I'm actually seeing this same behavior. But the board actually does not freeze. If you keep a key pressed of send a sysrq through serial line then you eventually get it back.

It seams to be something related to serial driver and wakeup ?

Do you see this problem after doing

  # echo 1 > /debug/pm_debug/sleep_while_idle

? There is a known problem with the UART inactivity timer that I'm debugging currently.

Kevin

Quoted 18 lines
Quoted 11 lines

Regards,
Amit

Signature · 7 lines
-------------------------------------------------------------------------
Amit Kucheria,				 Kernel Developer, Verdurent
-------------------------------------------------------------------------
--
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 · 5 lines
Eduardo Valentin
--
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
#3 lore raw
Quoted 16 lines · ext Kevin Hilman
Quoted 14 lines · Eduardo Valentin

Hello Amit,

Quoted 6 lines · ext Amit Kucheria

Hi,

I am testing twl4030 script optimisations on the current PM branch. But I am seeing the board (RX51) freeze when CPU_IDLE is enabled in the config.

Is it known to work on other boards?

I'm actually seeing this same behavior. But the board actually does not freeze. If you keep a key pressed of send a sysrq through serial line then you eventually get it back.

Quoted 7 lines
Quoted 2 lines

It seams to be something related to serial driver and wakeup ?

Do you see this problem after doing

  # echo 1 > /debug/pm_debug/sleep_while_idle

No I don't. The problem is gone it seams.

similar effect happens if we

# echo 0 > /sys/devices/platform/serial8250.2/sleep_timeout

Of course, this would prevent device go to sleep I guess.

Quoted 2 lines

? There is a known problem with the UART inactivity timer that I'm debugging currently.

Ok.

Quoted 22 lines

Kevin

Quoted 18 lines
Quoted 11 lines

Regards,
Amit

Signature · 7 lines
-------------------------------------------------------------------------
Amit Kucheria,				 Kernel Developer, Verdurent
-------------------------------------------------------------------------
--
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 · 5 lines
Eduardo Valentin
--
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
Eduardo Valentin
#4 lore raw
Quoted 25 lines · Eduardo Valentin
Quoted 16 lines · ext Kevin Hilman
Quoted 14 lines · Eduardo Valentin

Hello Amit,

Quoted 6 lines · ext Amit Kucheria

Hi,

I am testing twl4030 script optimisations on the current PM branch. But I am seeing the board (RX51) freeze when CPU_IDLE is enabled in the config.

Is it known to work on other boards?

I'm actually seeing this same behavior. But the board actually does not freeze. If you keep a key pressed of send a sysrq through serial line then you eventually get it back.

Quoted 7 lines
Quoted 2 lines

It seams to be something related to serial driver and wakeup ?

Do you see this problem after doing

  # echo 1 > /debug/pm_debug/sleep_while_idle
Quoted 8 lines

No I don't. The problem is gone it seams.

similar effect happens if we

# echo 0 > /sys/devices/platform/serial8250.2/sleep_timeout

Of course, this would prevent device go to sleep I guess.

Yes, there is a problem with UART3 (in PER) going idle independently of CORE/MPU that I have not yet found.

Kevin

#5 lore raw
Quoted 31 lines · ext Kevin Hilman
Quoted 29 lines · Eduardo Valentin
Quoted 16 lines · ext Kevin Hilman
Quoted 14 lines · Eduardo Valentin

Hello Amit,

Quoted 6 lines · ext Amit Kucheria

Hi,

I am testing twl4030 script optimisations on the current PM branch. But I am seeing the board (RX51) freeze when CPU_IDLE is enabled in the config.

Is it known to work on other boards?

I'm actually seeing this same behavior. But the board actually does not freeze. If you keep a key pressed of send a sysrq through serial line then you eventually get it back.

Quoted 7 lines
Quoted 2 lines

It seams to be something related to serial driver and wakeup ?

Do you see this problem after doing

  # echo 1 > /debug/pm_debug/sleep_while_idle

No I don't. The problem is gone it seams.

similar effect happens if we

Quoted 7 lines
Quoted 4 lines
# echo 0 > /sys/devices/platform/serial8250.2/sleep_timeout

Of course, this would prevent device go to sleep I guess.

Yes, there is a problem with UART3 (in PER) going idle independently of CORE/MPU that I have not yet found.

yeah, one possible fix would be preventing it if core is going to on @@ -514,7 +514,9 @@ void omap_sram_idle(void)

        /* PER */
        if (per_next_state < PWRDM_POWER_ON) {

- omap_uart_prepare_idle(2);
+ if (core_next_state != PWRDM_POWER_ON)
+ omap_uart_prepare_idle(2);
+

Needs to see what is the impact of this approach though

Quoted 2 lines

Kevin

Signature · 1 line
Eduardo Valentin
#6 lore raw
In reply to Amit Kucheria
Quoted 6 lines · Amit Kucheria

Hi,

I am testing twl4030 script optimisations on the current PM branch. But I am seeing the board (RX51) freeze when CPU_IDLE is enabled in the config.

Is it known to work on other boards?

Yes, and it works for me on RX51 as well, output of the maemo powertop below.

For a known working defconfig, start from omap3_pm_defconfig and change the low-level debug UART from UART1 to UART3.

There is a known UART bug in the current PM branch where if you don't 'echo 1 > /debug/pm_debug/sleep_while_idle', the UART eventually locks up when the inactivity timer expires. I'm still looking into this one.

Kevin

/ # powertop
Powertop 1.13.1
status: Unable to send message: Connection refused Mounting debugfs...Success
Sleeping for 11 seconds before sampling Collecting data for 30 seconds
Sample interval was 00m 30s 20569us

C# | Ratio | Avg/dura | Frequency | Ratio
--------+--------+----------+-----------+--------------

     C0 |  83.4% |          |   600 MHz |   0.0%                                
     C1 |  16.6% |  832.4ms |   550 MHz |   0.0%                                
     C2 |   0.0% |          |   500 MHz | 100.0%                                
     C3 |   0.0% |          |   250 MHz |   0.0%                                
     C4 |   0.0% |          |   125 MHz |   0.0%                                

IRQ# | Activity | Type | Name
--------+------------+----------------+---------------------------

     37 |         27 |           INTC | gp                                      
     11 |         23 |           INTC | prcm                                    
     12 |          4 |           INTC | DMA                                     
    225 |          2 |           GPIO | omap2-onenand                           

PID# | Activity | Name | Function Entry (Expire)
--------+------------+----------------+---------------------------

      0 |         16 |  <kernel core> | hrtimer_start (tick_sched_timer)        
     93 |          5 |    bdi-default | bdi_forker_task (process_timeout)       
    483 |          5 |flush-ubifs_0_0 | schedule_timeout_interruptible (process)
      0 |          4 |  <kernel core> | arm_supers_timer (sync_supers_timer_fn) 
      5 |          3 |       events/0 | queue_delayed_work (delayed_work_timer_)
    483 |          1 |flush-ubifs_0_0 | hrtimer_start_range_ns (wbuf_timer_call)
    483 |          1 |flush-ubifs_0_0 | hrtimer_start_range_ns (wbuf_timer_call)
    484 |          1 |       powertop | hrtimer_start_range_ns (hrtimer_wakeup) 

Power domain activity breakdown
Domain | % of time spent in states
--------+---------+---------+---------+---------+----------

usbhost |OFF:   0%|RET: 100%|INA:   0%| ON:   0%| now:(RET)                     
    sgx |OFF: 100%|RET:   0%|INA:   0%| ON:   0%| now:(OFF)                     
    per |OFF:   0%|RET:  83%|INA:   0%| ON:  16%| now:(ON)                      
    dss |OFF:   0%|RET: 100%|INA:   0%| ON:   0%| now:(RET)                     
    cam |OFF:   0%|RET: 100%|INA:   0%| ON:   0%| now:(RET)                     
   core |OFF:   0%|RET:  83%|INA:   0%| ON:  16%| now:(ON)                      
   neon |OFF:   0%|RET:  83%|INA:   0%| ON:  16%| now:(ON)                      
    mpu |OFF:   0%|RET:  83%|INA:   0%| ON:  16%| now:(ON)                      
   iva2 |OFF:   0%|RET: 100%|INA:   0%| ON:   0%| now:(RET)                     

Clock activity breakdown at end of period
Domain | Active clocks
--------+---------------+---------------+------------------

   core |          SDRC |      OMAPCTRL |         UART1                         
        |         UART2 |                                                       
  core3 |       USBTLL                                                          
   wkup |          GPT1 |       32KSYNC |         GPIO1                         
        |                                                                       
  ckgen |          CORE |          PERI |           96M                         
        |           48M |           12M |           54M                         
        |      EMU_CORE |                                                       
    per |         UART3 |         GPIO2 |         GPIO3                         
        |         GPIO4 |         GPIO5 |         GPIO6                         
        |                                                                       
Total wakeups    92,   3.1/s | IRQ   56,   1.9/s | Timers   36,   1.2/s         
HW wakeups       23,   0.8/s |     Real gp_timers expired   27,   0.9/s         

Dismounting debufs...Success
/ #