MobileRead Forums - View Single Post

knc1 · 03-08-2016, 05:19 PM

Keep in mind, it might be the kernel itself that is the battery hog.
It would be interesting to see what version/build date 5.7.3 is using:

Code:

[root@kindle bin]# uname -a
Linux kindle 3.0.35-lab126 #1 PREEMPT Wed Apr 15 08:08:43 PDT 2015 armv7l GNU/Linux

Which is a very early PW-3 (5.6.1.1) build.

And top shows that user 'framework' running 'cvm' is using 98.8% of the cpu - -
With the screen saver showing and supposedly running in the 'low power' state.

I noticed that some time ago but gave it little thought because that device is always on external power.

Hmmm......

Code:

[root@kindle bin]# cat /proc/loadavg
1.25 1.30 1.24 3/349 6417

Now that is just wrong for the state it is supposed to be in right now (it should be only active enough to respond to the serial port UART - this isn't my first rodeo).

Note: the first three numbers are what top displays as "Load Adverage:" - which used to be the average number of tasks waiting in the ready queue for three time periods.
At least with the 2.6.x series kernels, I don't know what it is in the 3.0.y series kernels, it may be the price of tea in China.

The mystery deepens:

Code:

[root@kindle bin]# cat /proc/softirqs
                    CPU0       
          HI:          0
       TIMER:  175869613
      NET_TX:        254
      NET_RX:       3016
       BLOCK:          0
BLOCK_IOPOLL:          0
     TASKLET:   44937475
       SCHED:          0
     HRTIMER:        376
         RCU:  187274637

[root@kindle bin]# cat /proc/uptime
1837228.26 68387.13 0 0

[root@kindle bin]# cat /sys/power/wakeup_count
0

[root@kindle bin]# cat /sys/power/device_suspend_time_threshold
off

None of that is what I would expect from a device that is 'sleeping' or at least in a heavy dose.

Another change that I just noticed:
/proc/config.gz is enabled in this kernel build (the kernel's build-time configuration file) - Lab126 did not used to do that for us.

Ah, so -

Code:

[root@kindle bin]# zcat /proc/config.gz | less

- - - - -
#
# CPU Power Management
#

#
# CPU Frequency scaling
#
- - - - -
# CONFIG_CPU_IDLE is not set
- - - - -

This 5.6.1.1 kernel build will never detect when the cpu is idle.
At least I think that is the situation - I haven't studied kernel config options for a long time.

Edit:
Close:
http://cateee.net/lkddb/web-lkddb/CPU_IDLE.html
And:
http://lists.infradead.org/pipermail...ry/232741.html

Somebody with 5.7.xxx please report the same things I listed above (I included the commands with each).

03-08-2016, 05:19 PM	#12
knc1 Going Viral Posts: 17,212 Karma: 18210809 Join Date: Feb 2012 Location: Central Texas Device: No K1, PW2, KV, KOA	Keep in mind, it might be the kernel itself that is the battery hog. It would be interesting to see what version/build date 5.7.3 is using: Code: [root@kindle bin]# uname -a Linux kindle 3.0.35-lab126 #1 PREEMPT Wed Apr 15 08:08:43 PDT 2015 armv7l GNU/Linux Which is a very early PW-3 (5.6.1.1) build. And top shows that user 'framework' running 'cvm' is using 98.8% of the cpu - - With the screen saver showing and supposedly running in the 'low power' state. I noticed that some time ago but gave it little thought because that device is always on external power. Hmmm...... Code: [root@kindle bin]# cat /proc/loadavg 1.25 1.30 1.24 3/349 6417 Now that is just wrong for the state it is supposed to be in right now (it should be only active enough to respond to the serial port UART - this isn't my first rodeo). Note: the first three numbers are what top displays as "Load Adverage:" - which used to be the average number of tasks waiting in the ready queue for three time periods. At least with the 2.6.x series kernels, I don't know what it is in the 3.0.y series kernels, it may be the price of tea in China. The mystery deepens: Code: [root@kindle bin]# cat /proc/softirqs CPU0 HI: 0 TIMER: 175869613 NET_TX: 254 NET_RX: 3016 BLOCK: 0 BLOCK_IOPOLL: 0 TASKLET: 44937475 SCHED: 0 HRTIMER: 376 RCU: 187274637 [root@kindle bin]# cat /proc/uptime 1837228.26 68387.13 0 0 [root@kindle bin]# cat /sys/power/wakeup_count 0 [root@kindle bin]# cat /sys/power/device_suspend_time_threshold off None of that is what I would expect from a device that is 'sleeping' or at least in a heavy dose. Another change that I just noticed: /proc/config.gz is enabled in this kernel build (the kernel's build-time configuration file) - Lab126 did not used to do that for us. Ah, so - Code: [root@kindle bin]# zcat /proc/config.gz \| less - - - - - # # CPU Power Management # # # CPU Frequency scaling # - - - - - # CONFIG_CPU_IDLE is not set - - - - - This 5.6.1.1 kernel build will never detect when the cpu is idle. At least I think that is the situation - I haven't studied kernel config options for a long time. Edit: Close: http://cateee.net/lkddb/web-lkddb/CPU_IDLE.html And: http://lists.infradead.org/pipermail...ry/232741.html Somebody with 5.7.xxx please report the same things I listed above (I included the commands with each). Last edited by knc1; 03-08-2016 at 08:49 PM.