Posted Jul 22, 2004 17:18 UTC (Thu) by xav (guest, #18536)
[Link]
Your fault for not being present at the Kernel Summit to explain how swsusp2 works :)
Software suspend: and then there were two
Posted Jul 26, 2004 8:12 UTC (Mon) by NCunningham (guest, #6457)
[Link]
:>
I'm thinking about preparing a paper for the Aussie Linux conference next year.
Software suspend: and then there were two
Posted Jul 22, 2004 17:53 UTC (Thu) by Ross (subscriber, #4065)
[Link]
I'm curious about the integration plans. Are you planning to merge with Pat's changes and then with the mainline kernel or will one of the patch sets be merged with mainline and then the other? And are there any conflicting changes which won't be able to exist in the same code base? Inquiring minds want to know ;)
Software suspend
Posted Jul 24, 2004 22:17 UTC (Sat) by Duncan (guest, #6647)
[Link]
From the coverage, it looks to me like the two current mainline implementations will merge first, since that code is already cooked and the patches ready to integrate. That would also bring an immediate benefit in the form of simplification of the current two implementations into one, lessening code bloat and simplifying the task of later integration of swsuspend2.
Of course, that's all in how I read the LWN coverage. I could be reading it wrong.
As for swsuspend2, I read that its the only one that does SMP, which interests me since I've a dual Opteron system here as my desktop. (Yes, it's fast! I've put that speed to good use running Gentoo, stage-one compiled from source, and just freshly updated to KDE 3.3.0-beta2 today. =:^) It'd be nice to be able to reliably suspend it and shut the fans and the like off, since it's in my bedroom, and it IS a dual Opteron, with all the cooling requirements that means. <g>
I recently discovered, to my GREAT surprise, that hardware suspend to RAM seems to work on this system -- with one caveat -- upon recovery, my time is all screwed up, as it doesn't advance with the hwclock over suspend and naturally the software time doesn't advance either. The /real/ complicating factor there is that I've been running NTP, and it would /really/ skew its compensation factor if it tried to do a time sync in /that/ state!
Anyway, now that I know hardware suspend to RAM /can/ work, when I get time, I'm hoping to throw an NTP service shutdown script into the suspend sequence, and an NTPClient fast-sync script into the resume, b4 restarting the NTP service itself.
That said, I'd STILL like to be able to suspend to disk, and be able to entirely power off the system, and swsuspend2 seems my only chance at that, at this point, tho I've been hesitant to look into it as it'd be my first off-main patch, and I generally catch all the full releases and part of the rcs, and an out-of-tree patch would mean either slowing down on that, or potentially large amounts of work to keep it merged, particularly given the rumored intrusiveness of the patch. I'd be /very/ happy to see it merged into mainline, however, /particularly/ if it works for my dual Opteron!
Duncan
Software suspend
Posted Jul 26, 2004 8:11 UTC (Mon) by NCunningham (guest, #6457)
[Link]
X64-64 isn't supported yet, but it's not far away.
Regards,
Nigel
Software suspend: and then there were two
Posted Jul 26, 2004 8:10 UTC (Mon) by NCunningham (guest, #6457)
[Link]
Hi.
Duncan's reply is right. The two in-kernel versions are being merged first. There are already patches in the -mm tree.
After the dust has settled, I'll push my patches.
Right now I'm preparing the last feature changes to suspend2 before merging: adding support for resuming an initrd and making suspend modular. This will allow encrypted storage of an image (the encryption can be set up from the initrd), and will also allow at least some upgrades to suspend to be done without rebooting the system. The system I'm running right now has LZF, GZIP and swapwriter support compiled as modules. I just need to get the core done.
Regards,
Nigel
Software suspend: and then there were two
Posted Jul 23, 2004 12:57 UTC (Fri) by erich (subscriber, #7127)
[Link]
I'd really love to see swsusp2 in the mainline kernel. It rocks. I've been playing around with a different init concept - based on a minit extension i wrote - that would even allow stopping services not needed when the system is suspended (freeing memory and thus making suspend faster and easier). Unfortunately this isn't in a state of being published; i hope i could find some time during summer break, but i fear i won't.