tuxonice-kernel-old
tuxonice-kernel-old copied to clipboard
Hibernation fails with seeking memory with 32GB ram
I have a problem with my 32G of ram laptop. When I try to hibernate with ~42% or more of memory filled it shows looking to freeing 70MB (or more) and seem busy waiting. If I press ESC to cancel the ui shows cancelling, but nothing happens. The cancelling is never done, though the ui remains responsive. I'm trying to hibernate to a 32G empty swap partition using lzo compression, so it should have no issue finding space. Currently using a custom compiled tuxonice-sources-4.4.5 kernel from the Gentoo repositories
I have upgraded the kernel to 4.5.2 and even tried the filewriter to not use swap at all, but the problem persists. It now seems to magicly free some memory sometimes when seeking to free a small amount (<100MB), but it still gets stuck when seeking for a larger amount
What kind of memory does it try to clear / save on? Isn't the whole exercise of checking if space is available before an actual hibernation useless when using compression? Is it possible to simply remove the check and bail out / cancel, when space runs out during storing?
A workaround could be to use a 64GB swap disk/file.
I tried that, but that doesn't seem to change anything, it's like there is a hard limit somewhere in a check. I tried inspecting the source code, but I couldn't make heads or tails of it. I could try to help out, if you have some pointers where checks are done.
it seems unwilling to start with over 16G of memory use
I have the same issue. 8 GB RAM, 15.8 GB swap partition. I use ubuntu 16.04 and I installed tux on ice from https://launchpad.net/~tuxonice ppa. How can I try newer kernel versions with patches to this issue? What can I do if my system will fail to start with these patches? Excuse me, I'm very new to kernel hacking. I would be very glad if you provide link to documentation
Sorry for my slow response, guys. I have a number of projects on the go, but will try to get TuxOnIce some attention soon.
Some pointers into the source code would already be great ;)