[coreboot] kernel boot time down to ~1.5 second on the x60 with i915.fastboot success

[Charles Devereaux]

Hello

Here is the dmesg, and the kernel .config. Kernel boot comes down to 1.54s
indeed during a simple test with just the bzimage (I have to setup again an
optimized distribution for these tests)

The command-line I now suggest is :
i915.fastboot=1 i915.lvds_downclock=1 i915.i915_enable_rc6=7 nohz=on
nmi_watchdog=0 pcie_aspm=force thinkpad_acpi.force_load=1
thinkpad_acpi.brightness_enable=0 thinkpad_acpi.fan_control=1
snd-hda-intel.index=0 snd_hda_intel.power_save=10
snd_hda_intel.model=thinkpad snd-hda-intel.probe_mask=0x103
snd-pcsp.index=1 btusb.reset=1

i915.i915_enable_fbc=1 is not safe : it causes screen corruption on resume

PS: I was told that these 2 patches allowed fastboot with a "close enough"
mode, even if timings didn't match:

http://lists.freedesktop.org/archives/intel-gfx/2014-October/053501.html
http://lists.freedesktop.org/archives/intel-gfx/2014-October/053502.html

I haven't used them though, so I can't comment on that, but they are a more
elegant solution for sure :-)


On Fri, Dec 5, 2014 at 2:44 AM, Charles Devereaux <coreboot at guylhem.net>
wrote:

> Hello
>
> i915.fastboot can be made to work on the x60, shaving about 0.8s from boot
> (and removing the annoying blinking screen)
>
> Since coreboot doesn't select the same video mode the kernel does, the
> easiest way to make that work is to comment out request for a mode change
> when the modes are different: in
> linux-3.14.25/drivers/gpu/drm/i915/intel_display.c comment out on line
> 9879: config->mode_changed = true; so that it looks like:
>         if (set->mode && !drm_mode_equal(set->mode, &set->crtc->mode)) {
>                 DRM_DEBUG_KMS("modes are different, full mode set\n");
>                 drm_mode_debug_printmodeline(&set->crtc->mode);
>                 drm_mode_debug_printmodeline(set->mode);
> //              config->mode_changed = true;
>         }
>
> It's hackish, but the results are worth it!
>
> It was tested with a vanilla kernel 3.14.25 with the following kernel
> options:
> i915.lvds_downclock=1 i915.semaphores=1 i915.i915_enable_rc6=7
> i915.i915_enable_fbc=1 nohz=on nmi_watchdog=0 pcie_aspm=force
> thinkpad_acpi.force_load=1 thinkpad_acpi.brightness_enable=0
> thinkpad_acpi.fan_control=1 snd-hda-intel.index=0
> snd_hda_intel.power_save=10 snd_hda_intel.model=thinkpad
> snd-hda-intel.probe_mask=0x103 snd-pcsp.index=1 btusb.reset=1
> i915.fastboot=1
>
> I'm also using the latest libreboot without ACPI brightness (coreboot head
> 8ffc085e1affaabbe3dca8ac6a89346b71dfc02e with patches 6725 7048 5345 7050
> 7051) with CONFIG_MAINBOARD_DO_NATIVE_VGA_INIT and
> CONFIG_FRAMEBUFFER_KEEP_VESA_MODE.
>
> Grub.cfg has:
> set gfxpayload=keep
> terminal_output --append gfxterm
>
> The first results I posted on the list were a kernel boot in about 2.2s. I
> should now be around 1.5s. I will prepare a static kernel and post the
> .config and dmesg for  precise measurements in case anyone is interested in
> replicating the results.
>
> I believe further improvements could be made to go below the psychological
> 1 second mark:  now that I have received my 128 MB SPI chips (25Q128FVSG :
> cf
> http://www.ebay.com/itm/New-WINBOND-IC-Chip-FLASH-128M-BIT-25Q128FVSG-W25Q128FVSG-SOP8-Package-/301006935484)
> , I'll see if fitting a bzImage is helpful.
>
> Charles
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.coreboot.org/pipermail/coreboot/attachments/20141205/e745c4c7/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dmesg.fastboot
Type: application/octet-stream
Size: 48307 bytes
Desc: not available
URL: <http://www.coreboot.org/pipermail/coreboot/attachments/20141205/e745c4c7/attachment-0002.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: x60.config
Type: application/octet-stream
Size: 110205 bytes
Desc: not available
URL: <http://www.coreboot.org/pipermail/coreboot/attachments/20141205/e745c4c7/attachment-0003.obj>