[Karmic]Please enable PCI express ASPM (powersave)

[Erik Andrén]

Binary package hint: linux-image

In order to enjoy a large power save in pci express systems the CONFIG_PCIEASPM=y must be set.
Currently it's disabled.

Please activate this option it in the jaunty kernel.

[Erik Andrén (erik-andren)]

I measured the powersavings on my system (Clevo M720R) by activating this feature to be about 0.5 W

[Erik Andrén (erik-andren)]

Still an issue in karmic.

[Mike (mike-fdb)]

Also need to:
echo "powersave" > /sys/module/pcie_aspm/parameters/policy

[Martin Pitt (pitti)]

Once the kernel gets changed, I can add this to pm-utils-powersave-policy.

[Martin Pitt (pitti)]

Kernel team, would that be something appropriate for lucid?

[Leann Ogasawara (leannogasawara)]

Indeed it seems this is only enabled in Lucid for powerpc-smp:

ogasawara@emiko:~/ubuntu-lucid/debian.master/config$ grep -rn "CONFIG_PCIEASPM" *
config.common.ports:2325:# CONFIG_PCIEASPM_DEBUG is not set
config.common.ubuntu:2793:# CONFIG_PCIEASPM is not set
ia64/config.common.ia64:162:# CONFIG_PCIEASPM is not set
powerpc/config.flavour.powerpc:30:# CONFIG_PCIEASPM is not set
powerpc/config.flavour.powerpc-smp:31:CONFIG_PCIEASPM=y
powerpc/config.flavour.powerpc64-smp:31:# CONFIG_PCIEASPM is not set

However, examining this config option a bit closer it seems this is noted as being experimental. We typically don't enable experimental options. I'll try to raise this for discussion to get additional input. Thanks.

config PCIEASPM
        bool "PCI Express ASPM support(Experimental)"
        depends on PCI && EXPERIMENTAL && PCIEPORTBUS
        default n
        help
          This enables PCI Express ASPM (Active State Power Management) and
          Clock Power Management. ASPM supports state L0/L0s/L1.

[Mike (mike-fdb)]

I've tried to google for problems coused by CONFIG_PCIEASPM, and found very few of them:
http://permalink.gmane.org/gmane.linux.debian.devel.bugs.general/680670
https://bugzilla.redhat.com/show_bug.cgi?id=455200
and, maybe, http://permalink.gmane.org/gmane.linux.debian.devel.bugs.general/680670

It might be a good idea to turn option ON for beta1 and, if problems arise turn OFF in beta2.

[Martin Pitt (pitti)]

Seems a bit late for lucid now.

[Launchpad Janitor (janitor)]

This bug was fixed in the package linux - 2.6.34-4.11

---------------
linux (2.6.34-4.11) maverick; urgency=low

  [ Amit Kucheria ]

  * SAUCE: omap: remove calls to usb_nop_xceiv_register from board files
  * [Config] Add support for OMAP-mainline flavour

  [ Andy Whitcroft ]

  * SAUCE: powerpc: fix compile error when ptrace.h is included from
    userspace
    - LP: #583733

  [ Chase Douglas ]

  * Revert "SAUCE: Don't register vga16fb framebuffer if other framebuffers
    are present"
  * Revert "SAUCE: Disable function tracing after hitting __schedule_bug"
  * Revert "SAUCE: drm/i915: don't change DRM configuration when releasing
    load detect pipe"

  [ Kees Cook ]

  * SAUCE: fs: block cross-uid sticky symlinks
  * SAUCE: fs: block hardlinks to non-accessible sources

  [ Koen Kooi ]

  * SAUCE: board-omap3-beagle: add DSS2 support

  [ Leann Ogasawara ]

  * Revert "staging/go7007 -- disable"
  * Revert "[Config] staging/winbond -- disable"
  * Revert "Disable 4MB page tables for Atom, work around errata AAE44"
  * Revert "SAUCE: sync before umount to reduce time taken by ext4 umount"
  * Revert "SAUCE: Enable an e1000e Intel Corporation 82567 Gigabit
    controller"
  * Revert "SAUCE: Fix MODULE_IMPORT/MODULE_EXPORT"
  * Revert "SAUCE: Created MODULE_EXPORT/MODULE_IMPORT macros"
  * Revert "SAUCE: input/mouse/alps: Do not call psmouse_reset() for alps"
  * Revert "SAUCE: r8169: disable TSO by default for RTL8111/8168B
    chipsets."
  * Revert "[Upstream] b43: Declare all possible firmware files."
  * Revert "add Breaks: against hardy lvm2"
  * Revert "SAUCE: Guest OS does not recognize a lun with non zero target
    id on Vmware ESX Server"
  * Revert "SAUCE: Catch nonsense keycodes and silently ignore"
  * [Config] Enable CONFIG_ECRYPT_FS=y for ports
  * [Config] Enable CONFIG_USB=y for armel and sparc
  * [Config] Enable CONFIG_SCSI=y for ia64 and sparc
  * [Config] Enable CONFIG_RFKILL=y for ports
  * [Config] Enable CONFIG_ATH9K_DEBUGFS=y
  * [Config] Enable CONFIG_IWMC3200TOP_DEBUGFS=y
  * [Config] Enable CONFIG_RCU_FAST_NO_HZ=y
  * [Config] Enable CONFIG_IWLWIFI_DEVICE_TRACING=y
  * [Config] Enable CONFIG_LIBERTAS_MESH=y
  * [Config] Enable CONFIG_MMC_RICOH_MMC=y
  * [Config] CONFIG_RT2800USB_UNKNOWN=y
  * [Config] Enable CONFIG_VGA_SWITCHEROO=y
  * [Config] Enable CONFIG_CEPH_FS=m
  * [Config] Enable CONFIG_CRYPTO_PCRYPT=m
  * [Config] Enable CONFIG_EEEPC_WMI=m
  * [Config] Enable CONFIG_RT2800PCI=m
  * [Config] Enable CONFIG_SCSI_HPSA=m
  * [Config] Enable CONFIG_VHOST_NET=m
  * [Config] Disable CONFIG_SND_HDA_INPUT_BEEP_MODE by default
    - LP: #582350
  * [Config] Disable CONFIG_SOUND_OSS* and CONFIG_SND_*OSS
    - LP: #579300
  * [Config] Enable CONFIG_PCIEASPM=y
    - LP: #333990
  * [Config] updateconfigs for OMAP flavour

  [ Loïc Minier ]

  * Enable perf tools on armel

  [ Tim Gardner ]

  * SAUCE: Updated ndiswrapper to 1.56
    - LP: #582555
  * [Config] Added virtual flavour
  * [Config] Remove support for sub-flavours
  * [Config] Removed amd64 preempt flavour
  * [Config] updateconfigs, updateportsconfigs after flavour munging
 -- Leann Ogasawara <email address hidden> Tue, 25 May 2010 09:34:5...

[Mike (mike-fdb)]

Please have a look at comment #3 before closing this bug.

[Chase Douglas (chasedouglas)]

It seems to me that we should look into a policy addition to pm-utils-powersave-policy when on battery for:

echo "powersave" > /sys/module/pcie_aspm/parameters/policy

I will take a look at integrating this with other changes for maverick.

[AttilaN (attila123456)]

With CONFIG_PCIEASPM=y in the maverick kernel my CPU fan doesn't spin down even at the lowest temperatures. It's noticeably louder than without this option.
Dell Inspiron e1505.

[Mike (mike-fdb)]

AttilaN, please test your system with different ASPM params, to make sure it is ASPM responsible for.
http://www.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/6-Beta/html/Power_Management_Guide/ASPM.html

[AttilaN (attila123456)]

I see the same behavior with all 3 states [default, performance, powersave]. Only including pcie_aspm=off on the kernel command line helps.

[Mike (mike-fdb)]

Try to contact Dell support for BIOS update.
If it doesn't help, file a bug in kernel bugzilla.

[Chase Douglas (chasedouglas)]

The new pm-utils-1.4.1-1 in Maverick includes a power script to set the pcie_aspm policy to powersave when on battery and to default when on ac.