Needs builtin lpia modules to improve boot time

Bug #240938 reported by Loïc Minier
8
Affects Status Importance Assigned to Milestone
linux (Ubuntu)
Fix Released
Undecided
Amit Kucheria
Hardy
Fix Released
Undecided
Amit Kucheria

Bug Description

to reach our boot time goal for lpia, we need to compile some modules in the kernel instead of loadable objects. This gains us some seconds because we don't block in the big kernel load and load kernel modules sequentially, instead drivers are inited concurrently.

This is critical for our hardy lpia kernel and we don't want to fork it for UME to maintain it in the next 18 months.

TEST CASE: boot lpia system with bootchart installed; boot time is ~36s with previous kernel ~33s with new kernel (builtin modules).

Revision history for this message
Loïc Minier (lool) wrote :

I believe this is fixed in the intrepid git tree.

Changed in linux:
status: New → Fix Committed
Revision history for this message
Loïc Minier (lool) wrote :

This is fixed in amit's tree and I believe the hardy-proposed tree

Changed in linux:
assignee: nobody → amitk
status: New → Fix Committed
Revision history for this message
Tim Gardner (timg-tpi) wrote :

SRU Justification

Impact: LPIA boot times are too long.

Patch Description: Change the LPIA kernel config to include some devices in the kernel image in order to avoid a module probe.

Patch: http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-hardy.git;a=commit;h=07beb8a86f410ae80432489a28275de7cfa4ef10

TEST Case: See bug description

Loïc Minier (lool)
description: updated
Revision history for this message
Martin Pitt (pitti) wrote :

Accepted into -proposed, please test and give feedback here. Please see https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

Loïc Minier (lool)
Changed in linux:
assignee: nobody → amitk
Amit Kucheria (amitk)
Changed in linux:
status: Fix Committed → Fix Released
status: Fix Committed → Fix Released
Revision history for this message
Chris Gregan (cgregan) wrote :

Tested and confirmed. Boot time stands at 35 seconds according to my latest boot chart. See attached

Revision history for this message
Amit Kucheria (amitk) wrote :

Tested boot time at 34s and 36s.

Revision history for this message
Amit Kucheria (amitk) wrote :

"Don't release SRU bugs"

Changed in linux:
status: Fix Released → Fix Committed
status: Fix Released → Fix Committed
Revision history for this message
Martin Pitt (pitti) wrote :

I test-booted the -19.34 kernel on i386 and amd64, and confirm that it works well. Since there are no code changes, a shallow test (miscompilations, etc.) suffices here.

Revision history for this message
Martin Pitt (pitti) wrote :

Copied to hardy-updates.

Changed in linux:
status: Fix Committed → Fix Released
Revision history for this message
Dave Miller (dave-dsm) wrote :

are we really all going to download updated kernels, with headers and all, just to shave 3 secs off the boot time? am i missing something?

Revision history for this message
Alex Bennée (ajbennee) wrote :

I have to echo Dave Millers comments. There have been a lot of kernel updates recently and it seems rather Windowsy to force everyone to reboot just for some boot time saving. In future it might be worth keeping "fixes" like this rolled up until the next security/bug fix before pushing out to distro.

Revision history for this message
Gergely Janossy (marcabru) wrote :

Maybe I'm too lame, but is it ok to install a new kernel+headers+modules every week, so that the old kernels still remains bootable? Every new kernel pack is ~100meg on my hard disk and if I don't uninstall it (a novice user should not do this), in a year the hard disk will be full.

Revision history for this message
Loïc Minier (lool) wrote :

Dave, Alex, and others: sorry for the additional download and reboot; this additional -19 is unfortunate, and these changes were supposed to be present in the first -19 upload, but weren't (due to a misunderstanding). I hope the facts that the ABI wasn't bumped and that the two -19 updates were close to each other compensate, at least people waiting for 8.04.1 will only see one. Over the next cycle (intrepid) we will use a separate source package for lpia kernels, so all arches wont need to be updated at the same time and you wont have to pull such updates.

Gergely, if the latest kernel works for you, you don't need to keep the older ones and you can purge them. It's not easy to do so automatically for you: we don't want to remove the current kernel during an upgrade because, well, you're running it; also the new kernel might not boot at all and you need a way to recover your system in this case. However, the kernel team is working on building solutions to this problem; one building block is being able to boot the last known-working kernel.

See blog post by Ben Collins: http://blog.phunnypharm.org/2008/06/keeping-last-successfully-booted-kernel.html

The spec on kernel removals is at: https://wiki.ubuntu.com/KernelTeam/removing-old-kernels

Revision history for this message
Gergely Janossy (marcabru) wrote :

Loïc, thanks.
OFFTOPIC: As far as I can understand from the link, there is a solution for automatic removing of the unnecessary kernels on the way to intrepid, and for hardy one should apt-get remove old kernels manually.

Revision history for this message
Steve Beattie (sbeattie) wrote :

My understanding is this was fixed during the intrepid development cycle, closing the devel task. If this is incorrect, please re-open.

Changed in linux:
status: Fix Committed → Fix Released
Revision history for this message
Loïc Minier (lool) wrote :

I think it was indeed, thanks.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.