Gentoo Archives: gentoo-user

From: Joshua Murphy <poisonbl@×××××.com>
To: gentoo-user@l.g.o
Subject: Re: [gentoo-user] hibernate fails, "s2disk" not found
Date: Thu, 30 Oct 2008 07:25:48
Message-Id: c30988c30810300025n1f1e3547o6e9d2f1e2b2dee83@mail.gmail.com
In Reply to: Re: [gentoo-user] hibernate fails, "s2disk" not found by Walter Dnes
1 On Thu, Oct 30, 2008 at 12:33 AM, Walter Dnes <waltdnes@××××××××.org> wrote:
2 > On Wed, Oct 29, 2008 at 12:27:39PM +0100, Pint?r Tibor wrote
3 >>> [d530][root][~] emerge -pv s2disk
4 >>> These are the packages that would be merged, in order:
5 >>> Calculating dependencies |
6 >>> emerge: there are no ebuilds to satisfy "s2disk".
7 >>> Now what?
8 >>
9 >> emerge -pv suspend
10 >
11 > First, I had to keyword "=sys-power/suspend-0.8 ~x86" in
12 > package.keywords. I set up suspend.conf like so...
13 >
14 > snapshot device = /dev/snapshot
15 > resume device = /dev/sda6
16 > #image size = 350000000
17 > #suspend loglevel = 2
18 > compute checksum = y
19 > #compress = y
20 > #encrypt = y
21 > #early writeout = y
22 > #splash = y
23 >
24 > Here is my disk layout
25 >
26 > [d530][root][~] fdisk -l
27 >
28 > Disk /dev/sda: 500.1 GB, 500107862016 bytes
29 > 255 heads, 63 sectors/track, 60801 cylinders
30 > Units = cylinders of 16065 * 512 = 8225280 bytes
31 > Disk identifier: 0xd0000000
32 >
33 > Device Boot Start End Blocks Id System
34 > /dev/sda1 1 60801 488384001 5 Extended
35 > /dev/sda5 1 62 497952 83 Linux
36 > /dev/sda6 63 549 3911796 82 Linux swap / Solaris
37 > /dev/sda7 550 60801 483974158+ 83 Linux
38 >
39 > No, it's not LVM. / is half-a-gig, followed by swap, followed by the
40 > rest of the drive. I use multiple bindmounts to make things look
41 > normal. When I tried "sync", followed by "hibernate" it shut down, but
42 > when I powered back up with the power button, here's what happened...
43 >
44 > - on the reboot, it complained about the superblock "last access" being
45 > in the future (the half-gig partition is ext2)
46 >
47 > - it "fixed" the access date
48 >
49 > - complained that the hard drive was "dirty", i.e. not properly shut
50 > down
51 >
52 > - rebooted
53 >
54 > - played back a whole bunch of disk transactions on /dev/sda7
55 > (reiserfs). Did i mention I ran "sync" before "hibernate"?
56 >
57 > - it did the rest of the ordinary boot process.
58 >
59 > - it did *NOT* restore anything from the previous session. Do I have to
60 > explicitly set something to tell it to restore a previous session?
61 > Gentoo-wiki is stll down.
62 >
63 > --
64 > Walter Dnes <waltdnes@××××××××.org>
65
66 I've not touched hibernation, but from what I've gathered in 5 mins on
67 google, I've a few probable guesses. Firstly, the image of the system
68 set aside by s2disk leaves the system's partitions in a 'dirty' state
69 simply because they're technically still in use. Your kernel needs to
70 be told to look at /dev/sda6 for resuming, appending resume=/dev/sda6
71 (might be resume=swap:/dev/sda6, not sure) in your grub/lilo config
72 should handle that part, though I've a vague recollection of something
73 about a default resume partition last time I ran through menuconfig
74 too, so you may even be able to avoid that. When it boots, rather than
75 running through init and the usual, it needs access to the 'resume'
76 executable, which has to be usable *before* the system's current
77 mounts are reestablished... which means initrd or initramfs. If you
78 already have both of those things (kernel knowing what the resume
79 partition is and having access to 'resume' where it wants it) as they
80 should be... I'm at a loss.
81
82 If you go to:
83 http://www.google.com/search?q=HOWTO+Userspace+Software+Suspend+(uswsusp)+-+Gentoo+Linux+Wiki
84 and then hit 'cached' under the first hit, it'll take you to google's
85 last copy of the page.
86
87 Also, as a side note, from http://www.freewebs.com/gkiagia/hibernate.html --
88 "WARNING: Never boot the suspended system with another kernel than the
89 one that you used to suspend and never try to mount suspended
90 partitions from another linux system, such as a live cd, otherwise you
91 will probably have data loss."
92
93 --
94 Poison [BLX]
95 Joshua M. Murphy