In brief: I'm using Ubuntu on a netbook (Intel Atom) Zareason Terra HD. The version from 10.10 to the newest (11.04 and 11.10) there is a serious power regression. The problem is hardware related, according to PowerTop, Ubuntu 10.10 use about 6.5W with kernel 2.6.35. The newest version of Ubuntu, i.e. from kernel 2.6.38 to 3.3 (tested in mainline) use 7,5W or more. The main cause, according to powertop, is my 3G internal modem "Dispositivo USB: Globetrotter HSDPA" that are always at 100% of usage, draining about 1W (or more), even if I turn it off by network manager or from bios. 2.6.35 kernel doesn't have this problem. This issue reduces my battery life of about 1h. The only way is to turn it off by powertop, enabling power management for the device. I'm reporting here this bug as requested on Launchpad, where you can see the complete bug report, hardware and powertop details: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/875801 (here you can see another probably power management hardware related bug to "Device Codec audio hwC0D0: Realtek", and a possible optimization operable activating power management for /dev/sda by powertop. If needed I will open a separate bug report for theese.) Ask me for any more information or test. Thanks
if this device is disabled for both releases, does powertop result match?
I tried disabling Network Manager connection and disabling the device from the BIOS. In both cases powertop shows that the device is used at 100%. Enabling powersave mode for this device (the only way to shut it down) from powertop, powertop shows: 2.6.38 - 8 W (Ubuntu 11.04) 3.0 - 7,33 W (Ubuntu 11.10) 3.3 (daily #201201290405) - 7.7 W (Note that 3.0 as already applied the patch by Matthew Garrett for ASPM, officially introduced in 3.3 kernel)
I'm investigating the issue. Changing the /sys/bus/usb/devices/usb4/4-1/power/control from "on" to "auto", have the right effect to auto-suspend the HDSPA modem. This will save about 500 mW. More investigations revealed that in Ubuntu 12.04, this device has the maximum power usage setted to 500mW (found with the command "lsusb -v", at the line "MaxPower"). So suspending this device has less effect on power consumption compared to previous version of Ubuntu (~ 1W). This mean that there is something else that had raised its power usage. (this can be related only to Ubuntu and not Linux) However, the changes at the file /sys/bus/usb/devices/usb4/4-1/power/control will reset at the next reboot of the netbook. 1) So, how can I make that changes be permanent? 2) If these are not permanent, this means that some other process at the boot will automatically check the devices and will sets the power properties to them. Can someone tell me which process do this control (because is obviously bugged)? If someone can help me with these questions, I can continue to investigate this issue. Thanks. (P.S. Kernel 3.5 also have this issue)
Created attachment 79211 [details] Device Descriptor
If you can reproduce a regression using an upsream kernel by simply booting the new kernel, then it is a kernel bug. However, this one looks like it may be related to Ubuntu user-land policy...
Because I have no time to compile a vanilla kernel, I tried a Debian precompiled kernel. I used the 3.5.5 linux kernel (taken from here: http://packages.debian.org/experimental/linux-image-3.5-trunk-686-pae, I don't know "how much upstream" is this kernel). The results is interesting anyway: powertop doesn't shows any more the HSDPA Modem consuming power. The device is completely off. So, I think, the bug is related to some patch applied from canonical developers to Ubuntu mainline kernel. Despite this, the overall system power consumption is not improved: the net interface wlan0 (iwlwifi) is now draining a lot of power, about 1.5 W. Enabling powersave mode for wlan0, overall system power consumption drops fron about 9 W to about 7.10 W (a good result). So, I think the problem is related to distributions patch. Correct me if I'm wrong, or if this test is not much reliable.
(In reply to comment #6) > Despite this, the overall system power consumption is not improved: > the net interface wlan0 (iwlwifi) is now draining a lot of power, about 1.5 > W. > Enabling powersave mode for wlan0, overall system power consumption drops > fron > about 9 W to about 7.10 W (a good result). > > So, I think the problem is related to distributions patch. > Correct me if I'm wrong, or if this test is not much reliable. bug closed. please feel free to re-open it if you can reproduce the problem again in a vanilla kernel.