cpu.Info() returns error when clock speed not available

[tmm1]

I'm getting the following error from cpu.Info():

open /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_max_freq: no such file or directory

on an ARM device that has no clock speed entry in/proc/cpuinfo:

$ cat /proc/cpuinfo
Processor	: Marvell PJ4Bv7 Processor rev 2 (v7l)
processor	: 0
BogoMIPS	: 1332.01

processor	: 1
BogoMIPS	: 1332.01

Features	: swp half thumb fastmult vfp edsp vfpv3 tls
CPU implementer	: 0x56
CPU architecture: 7
CPU variant	: 0x2
CPU part	: 0x584
CPU revision	: 2

Hardware	: Marvell Armada XP Development Board
Revision	: 0000
Serial	: 0000000000000000

and no cpufreq node available under /sys/.../cpu:

$ ls /sys/devices/system/cpu/cpu0/
cpuidle online topology

In this case, I think it would be useful to return the known information about the CPUs, even if the Mhz field is set to 0.

[tmm1]

This specific device is a Synology branded RS815 NAS. The admin web UI for the device shows a clock speed of 1.33ghz, so perhaps gopsutil should fallback to the BogoMIPS entry in /proc/cpuinfo when no other clock speed information is available?

[tmm1]

FYI I found another device running linux 3.2.26 where /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_max_freq is reported in Hz instead of KHz.

[schmichael]

From the little bit of work I've done on ARM it seems that missing cpufreq is common and likely never going to be fully covered.

So on systems lacking cpufreq it might be worth falling back to dmidecode -s processor-frequency which works in the two test cases I have available.

There are 3 significant downsides to this:

1. dmidecode is rarely if ever installed by default on systems
2. dmidecode seems to require root (or at least read access to pseudo-files that are only readable by root on my system)
3. dmidecode's man page has a big warning about how untrustworthy it is:

More often than not, information contained in the DMI tables is inaccurate, incomplete or simply wrong.

Still seems worth trying in a fallback.

[shirou]

@tmm1 Sorry for very very late response. I created #335 which just ignore error. Does this resolve your problem?

@schmichael Thank you for researching. I agree your downside view about dmidecode, especially it needs root.

[schmichael]

@shirou Seems reasonable to me. Could I open a new issue (and maybe a PR someday soon) for adding a dmidecode fallback? I know it's of extremely limited usefulness, so feel free to say "no thanks."

[tmm1]

@shirou Thanks, that resolves my issue. I also have some other changes locally which you might find useful:

fancybits/gopsutil@8abfc8b parse bogomips values
fancybits/gopsutil@d3bf1ea use processor name as model fallback
fancybits/gopsutil@70e6762 handle older kernels which list cpu speed in hz

@schmichael Thanks for the dmidecode tip. Unfortunately none of the ARM systems which I have seen have that tool installed, and most don't allow me to run things as root. So this particular workaround would not be very useful to me, atleast.

[shirou]

@tmm1 sorry, when I close #335, also closed this issue.
And thank you for your commit. according to your valuable commits, I did two commit. I didn't include parse bogomips values commit, because Bogomips is too linux specific.

3499dec
a63f1ae
I have wanted to make a PR in order to note your name, but I failed push branch. very sad. sorry. But I really thank you for these.

[tmm1]

@shirou Thank you. No problem, I am glad they are included in master now.