Canonical's Kate Stewart set a milestone for correcting the ASPM power issue
by Ubuntu 11.10 Beta 1. Ubuntu 11.10 Beta will be released today, but it will not fix the Linux 2.6.38 power regression
that's caused by a change in PCI-E Active State Power Management
A month ago I wrote, "A few hours ago, Kate Stewart marked LP bug #760131 as being a milestone candidate for Ubuntu 11.10 Beta 1. This bug is for the main power regression introduced in the Linux 2.6.38 kernel as caused by PCI Express Active-State Power Management changes. There hasn't been a "solution" upstream in either the Linux 3.0 or 3.1 kernels yet since this is a tough problem. I'm not sure what Canonical is planning to do to "fix" the situation (considering their overall lack of low-level technical contributions particularly in the kernel area) besides possibly forcing PCI-E ASPM or just postponing the fix to a later milestone."
Sure enough, a fix has been delayed. Martin Pitt has changed the milestone target for triaging from Ubuntu 11.10 Beta 1 to Ubuntu 11.10 Beta 2. That's still extremely optimistic. The second beta is due out on the 22nd of September, but still it's unlikely to see a PCI Express ASPM change within Ubuntu 11.10. The 15th of September also marks the kernel freeze for Ubuntu 11.10, so unless they come up with some sort of workaround (such as forcing PCI-E ASPM on my default, but that could crash some systems), there won't be any fixes for the Oneiric Ocelot release.
Heck, we'll only likely see a real fix in Ubuntu 12.04 LTS if the upstream kernel developers address the problem by coming up with a new way
to more accurately determine when PCI Express Active-State Power Management is supported or not. This likely will come down to testing hardware/driver combinations and having each driver set the ASPM supported bit flag when it's known to work safely, since many system BIOSes aren't advertising their ASPM support when in fact the system can handle this power management feature.
There's no real improvement with the Linux 3.1 kernel
and I haven't seen anything queuing up yet for the Linux 3.2 kernel. In fact, with Intel graphics the power consumption can go up in Linux 3.1
. There's also other Linux power management regressions still outstanding.
The tracking of the ASPM power management regression can be done via LaunchPad Bug #760131
, "Power consumption raised significantly in natty."