[PATCH] linux-boot: Resume from hibernation after pre-boot.

DoneSubmitted by Jack Hill.
Details
3 participants
  • Jack Hill
  • Ludovic Courtès
  • Tobias Geerinckx-Rice
Owner
unassigned
Severity
normal
J
J
Jack Hill wrote on 9 Jul 2021 04:48
(address . guix-patches@gnu.org)(name . Tobias Geerinckx-Rice)(address . me@tobias.gr)
20210709024853.1170-1-jackhill@jackhill.us
* gnu/build/linux-boot.scm (boot-system): Call resume-if-hibernated
after pre-mount.
---

This change allows resuming from a swap device that needs device-mapper
to be set up first. In my case I tested swap on lvm on dm-crypt. I
believe this change to be safe since pre-mount shouldn't mount or
otherwise touch the filesystems. It it did, there would be data
corruption, so special care should be taken when reviewing.

gnu/build/linux-boot.scm | 18 +++++++++---------
1 file changed, 9 insertions(+), 9 deletions(-)

Toggle diff (38 lines)
diff --git a/gnu/build/linux-boot.scm b/gnu/build/linux-boot.scm
index 461df9ff46..778ee99690 100644
--- a/gnu/build/linux-boot.scm
+++ b/gnu/build/linux-boot.scm
@@ -563,15 +563,6 @@ upon error."
         (load-linux-modules-from-directory linux-modules
                                            linux-module-directory)
 
-        (unless (or (member "hibernate=noresume" args)
-                    ;; Also handle the equivalent old-style argument.
-                    ;; See Documentation/admin-guide/kernel-parameters.txt.
-                    (member "noresume" args))
-          ;; Try to resume immediately after loading (storage) modules
-          ;; but before any on-disk file systems have been mounted.
-          (false-if-exception           ; failure is not fatal
-           (resume-if-hibernated (find-long-option "resume" args))))
-
         (when keymap-file
           (let ((status (system* "loadkeys" keymap-file)))
             (unless (zero? status)
@@ -606,6 +597,15 @@ upon error."
           (unless (pre-mount)
             (error "pre-mount actions failed")))
 
+        (unless (or (member "hibernate=noresume" args)
+                    ;; Also handle the equivalent old-style argument.
+                    ;; See Documentation/admin-guide/kernel-parameters.txt.
+                    (member "noresume" args))
+          ;; Try to resume immediately after loading (storage) modules
+          ;; but before any on-disk file systems have been mounted.
+          (false-if-exception           ; failure is not fatal
+           (resume-if-hibernated (find-long-option "resume" args))))
+
         (setenv "EXT2FS_NO_MTAB_OK" "1")
 
         (if root-device
-- 
2.32.0
T
T
Tobias Geerinckx-Rice wrote on 9 Jul 2021 05:21
(name . Jack Hill)(address . jackhill@jackhill.us)(address . guix-patches@gnu.org)
87a6mwgnvt.fsf@nckx
Jack Hill 写道:
Toggle quote (4 lines)
> * gnu/build/linux-boot.scm (boot-system): Call
> resume-if-hibernated
> after pre-mount.

All right! \o/ Whilst I'm glad it didn't eat your data, it was
still just an off-the-cuff suggestion on IRC & I'd like to take
the weekend to think about the whole picture.

Kind regards,

T G-R
-----BEGIN PGP SIGNATURE-----

iIMEARYKACsWIQT12iAyS4c9C3o4dnINsP+IT1VteQUCYOfEFg0cbWVAdG9iaWFz
LmdyAAoJEA2w/4hPVW155PUA/0XvlC1lk4repufb6af41nb9zjPn9RAKxVMGmg2A
FwHmAQC3qkBVZa3EecQH3lcldu0+LLGodZyfUR8pdjMZ2WAxBg==
=qcha
-----END PGP SIGNATURE-----

L
L
Ludovic Courtès wrote on 5 Sep 14:51 +0200
Re: bug#57583: Guix cannot resume after hibernation
(name . Josselin Poiret)(address . dev@jpoiret.xyz)
87h71m3rx1.fsf@gnu.org
Hi,

Josselin Poiret <dev@jpoiret.xyz> skribis:

Toggle quote (9 lines)
> Yusuf Talha via Bug reports for GNU Guix <bug-guix@gnu.org> writes:
>
>> I haven't encountered this problem on any other distros.  You probably realized that I encrypted my disk.  Hibernation wasn't working while I was using Guix without encryption a few months ago either.  So I don't think LUKS is the problem here.
>
>>From what I can tell, resuming from an encrypted partition isn't
> currently supported. This could potentially be done, though it's not my
> priority right now (and testing these kinds of features is dangerous,
> you could make a wrong move and incur heavy data loss).

I believe this is fixed by this:


Jack, do I get it right?

The patch in question fell through the cracks, but it actually LGTM.

Thanks,
Ludo’.
T
T
Tobias Geerinckx-Rice wrote on 5 Sep 15:05 +0200
(name . Ludovic Courtès)(address . ludo@gnu.org)
87wnaiezj2.fsf@nckx
Jack, Yusuf, Ludo',

Yusuf: you could try pulling, reconfiguring, hibernating, and
resuming, if you're brave-ish.

Ludovic Courtès 写道:
Toggle quote (3 lines)
> I believe this is fixed by this: […] The patch in question fell
> through the cracks

It should and it had, sorry.

I *dislike* the fact that this makes the resume path no longer
trivially obviously correct — because we now actually do something
potentially useful (and hence, dangerous) during it — but that's
unavoidable I suppose. It's certainly not the fault of this
patch.

Pushed as ec16f88522041dc285a35705ff9ee95672b78143, thanks Jack!

Kind regards,

T G-R
-----BEGIN PGP SIGNATURE-----

iIMEARYKACsWIQT12iAyS4c9C3o4dnINsP+IT1VteQUCYxX1oQ0cbWVAdG9iaWFz
LmdyAAoJEA2w/4hPVW150cYA/3ZS6cUFlwn+9Q7MDVcWwHPe1Kbi4BHjtXrw5BZP
ZS/fAQCId35XEGp8/W1LizS/imQg14KfTLyfPyddJ6Xs1hV2Dw==
=uYzh
-----END PGP SIGNATURE-----

Closed
?
Your comment

This issue is archived.

To comment on this conversation send email to 49475@debbugs.gnu.org