Re: [PATCH initramfs-tools 0/9] Fix resume device configuration

[Ben Hutchings]

On Fri, 2017-04-21 at 11:02 +0200, Martin Steigerwald wrote:
[...]
> Currently I don´t have uswsusp installed, but pm-utils is 
> still there. I am not sure whether systemd uses it or does it all by itself 
> tough.

I think 'systemctl hibernate' invokes in-kernel swsusp by default.  I
don't know what pm-utils does.

[...]
> I never managed to get around actually looking whats going on here, but I now 
> found that I do not have any resume setting in
> 
> shambhala:/etc/initramfs-tools> grep -ir "RESUME" .
> shambhala:/etc/initramfs-tools#1>

So initramfs-tools will automatically select a swap partition and put
its UUID in the initramfs...

> Instead I have it here:
> 
> shambhala:/etc/default> grep -i resume grub
> GRUB_CMDLINE_LINUX_DEFAULT="resume=/dev/sda2 init=/bin/systemd"

...but the command line parameter overrides that configuration, anyway.

> shambhala:~> swapon -s
> Dateiname   Typ Größe   Benutzt 
> Priorität
> /dev/sda2   partition   4000180 0   -1
> 
> shambhala:~> lsblk /dev/sda2
> NAME MAJ:MIN RM  SIZE RO TYPE MOUNTPOINT
> sda2   8:20  3,8G  0 part [SWAP]
> shambhala:~> LANG=C swapon -s
> FilenameTypeSizeUsed
> Priority
> /dev/sda2   partition   4000180 0   -1
> 
> 
> Okay, I will upgrade to the new initramfs-tools and look whether it fixes the 
> issue.

I think that the changes in 0.128 - to wait for the resume device to
appear - may fix this problem.  The changes I've posted here shouldn't
make any difference for your configuration.

> If not, I may try setting RESUME var to first "auto" and if not working 
> still then the device in question to see whether that helps. I welcome any 
> other ideas and I am willing to report back what I find.

RESUME=auto isn't supported in a released version.

Ben.

-- 
Ben Hutchings
I'm not a reverse psychological virus.  Please don't copy me into your
sig.



signature.asc
Description: This is a digitally signed message part

Re: [PATCH 3/3] builddeb: add make fastdeb-pkg target

[Ben Hutchings]

On Fri, 2017-04-21 at 18:57 +0900, Masahiro Yamada wrote:
[...]
> In hindsight, it seems more straightforward to me
> to build the binary packages from the source package.
> But, I know this refactoring will need lots of changes, which may
> break something.
> 
> Once we support this fastdeb-pkg target,
> we will not be able to convert the script in that way.
> (because we do not have a standard way to generate a particular binary
> package from *.dsc)
[...]

We kind of do, though it's a relatively new feature:
https://wiki.debian.org/BuildProfileSpec

Ben.

-- 
Ben Hutchings
I'm not a reverse psychological virus.  Please don't copy me into your
sig.


signature.asc
Description: This is a digitally signed message part

Bug#860900: AST device driver better support on Power

[Frederic Bonnard]

Source: linux
Source-Version: 4.9.18-1
Tags: patch

--

Hi,
many Power (especially P9) machines use AST graphic cards.
There were many improvements and bug fixes recently on this side and it
would be nice to have those in futur 4.9.x Stretch kernels.
Some of those patches have already been taken in the linux lts 4.9
kernel, namely 4.9.15 :
- [arm64,powerpc*,x86] drm/ast: Fix test for VGA enabled
- [arm64,powerpc*,x86] drm/ast: Call open_key before enable_mmio in POST 
code
- [arm64,powerpc*,x86] drm/ast: Fix AST2400 POST failure without BMC FW or 
VBIOS

Since then and still in 4.9.24, there are no other patch for this. Those
that would be nice to have are in this list (ordered from 1st to last to
be applied) :

[not included] 6c971c09f38704513c426ba6515f22fb3d6c87d5 : drm/ast: Fixed system 
hanged if disable P2A
[included] 3856081eede297b617560b85e948cfb00bb395ec : drm/ast: Fix AST2400 
POST failure without BMC FW or VBIOS
[not included] 71f677a91046599ece96ebab21df956ce909c456 : drm/ast: Handle 
configuration without P2A bridge
[not included] 22acdbb1bdbd56cc9939e18516dfcf214a9d835b : drm/ast: const'ify 
mode setting tables
[not included] cf2f6bd402470caed4e4d75b783d2d48277a3568 : drm/ast: Remove 
spurrious include
[not included] 6475a7cce61967fca4dd793b60acf5a7dc70bc9a : drm/ast: Fix 
calculation of MCLK
[not included] 9f93c8b3c08f8c456aad86fd05caa6a1688320ff : drm/ast: Base support 
for AST2500
[not included] bad09da6deab157440a3f0dd5e6c03cdc60c0a91 : drm/ast: Fixed vram 
size incorrect issue on POWER
[not included] b368e53aae7694e0fb2bdec9667c9acc5108b629 : drm/ast: Factor 
mmc_test code in POST code
[not included] d95618ef0a05236a2642d1a829fffd90de88e5f0 : drm/ast: Rename 
ast_init_dram_2300 to ast_post_chip_2300
[not included] 94fdc2a86a927b409d9a09ba28bcb08e34f3ac8d : drm/ast: POST code 
for the new AST2500
[included] 905f21a49d388de3e99438235f3301cabf0c0ef4 : drm/ast: Fix test for 
VGA enabled
[included] 9bb92f51558f2ef5f56c257bdcea0588f31d857e : drm/ast: Call 
open_key before enable_mmio in POST code

only "[not included]" ones should be needed as 3 of them are already in
4.9.18-1 ; the latter should cause some minor fuzz.

Thanks,

F.


pgps2805kUTmc.pgp
Description: PGP signature

Re: [PATCH initramfs-tools 0/9] Fix resume device configuration

[Michael Prokop]

* Ben Hutchings [Fri Apr 21, 2017 at 04:33:09AM +0100]:

> The change in version 0.128 to wait for the resume device to appear
> uncovered a number of systems for which the resume device is not
> properly configured.  In fact, systems with swap partitions that
> are never available at boot could not be configured correctly,
> other than by adding 'noresume' to the kernel command line!

> While working on that, I found and fixed a couple of other
> longstanding bugs in resume device selection.
[...]
> If no-one finds issues with this these changes then I'm hoping
> to make one more release for "stretch" with just these.

The changes LGTM (I didn't test/verify them though).
Thanks, Ben.

regards,
-mika-


signature.asc
Description: Digital signature

Re: [PATCH 3/3] builddeb: add make fastdeb-pkg target

[Masahiro Yamada]

Hi Riku,


2017-04-20 19:53 GMT+09:00 Riku Voipio :
> Hi,
>
> Thanks for taking time to review.
>
> On 18 April 2017 at 17:31, Masahiro Yamada
>  wrote:
>> Before expanding this even more,
>> please help me understand this script.
>>
>> Commit 3716001b implemented source package build
>> at the end of the builddeb script.
>> However, the binary packages are still built in its own way.
>> (So, debian/rules is not used in general workflows,
>> so we need to test it explicitly somehow.)
>
> Historic reasons.. since the builddeb already existed, the easiest way
> to implement debian/rules was to make a skeleton debian/rules that
> calls builddeb indrectly.
>
>> Why do not we create a source package first,
>> then build binary packages from it?
>
> I wouldn't do a source package per se, but we could create a debian/
> directory tree and the call dpkg-buildpackage. This would be a more
> substantial overhaul that might break some expected behavior of
> deb-pkg targets. If people think its worth, I can give it a shot.


In hindsight, it seems more straightforward to me
to build the binary packages from the source package.
But, I know this refactoring will need lots of changes, which may
break something.

Once we support this fastdeb-pkg target,
we will not be able to convert the script in that way.
(because we do not have a standard way to generate a particular binary
package from *.dsc)
That's why I asked the question, and I want to be careful.



>> rpm-pkg does that way.
>> (generate a spec-file, then run rpmbuild)
>
> speaking of the mkspec script, it would be nice to refactor it to here
> documents over endless lines of echo.

As always, cleaning patches are welcome.



-- 
Best Regards
Masahiro Yamada

Re: [PATCH v2 1/3] builddeb: Update a few outdated and hardcoded strings

[Masahiro Yamada]

Hi Riku,


The merge windows will be open shortly,
so I'd like to apply the first two.

2017-03-31 22:09 GMT+09:00  :
> From: Riku Voipio 
>
> The builddeb script has some hardcoded references to Linux version 2.6
> which is ancient. Drop Provides as the virtual packages provided are not
> useful anymore. Leave the Provides for linux-kernel-headers, as someone
> might still be referring to it.
>
> While at it, updated copyright date and drop Standards-Version:


Could you add some comments about
"why" for the Standard-Version removal?

I know the answer is in the v2:
https://patchwork.kernel.org/patch/9532275/
but I'd like you to record it in git-log.

With the comment improved, I will pick up this soon.

Thanks!


-- 
Best Regards
Masahiro Yamada

Re: [PATCH initramfs-tools 0/9] Fix resume device configuration

[Martin Steigerwald]

Dear Ben.

Ben Hutchings - 21.04.17, 04:33:
> The change in version 0.128 to wait for the resume device to appear
> uncovered a number of systems for which the resume device is not
> properly configured.  In fact, systems with swap partitions that
> are never available at boot could not be configured correctly,
> other than by adding 'noresume' to the kernel command line!
> 
> While working on that, I found and fixed a couple of other
> longstanding bugs in resume device selection.
> 
> This patch series:
> 
> - Fixes the sorting of swap partitions
> - Makes the RESUME variable work like every other configuration
>   variable, and documents it
> - Adds support for RESUME=none (disable resume) and RESUME=auto
>   (explicitly request automatic selection)
> - Adds warning and informational messages where the resume device
>   configuration is automatically fixed-up

Thats interesting. I have an old ThinkPad T42 above my hi-fi equipment as a 
media player and it stopped resuming from hibernation after some update quite 
some time ago.  But instead it just boots as if there would be no hibernation 
image. On hibernation it actually does write the hibernation image – I use in 
kernel hibernation, not userspace software suspend, but I think I tried using 
that one as well. Currently I don´t have uswsusp installed, but pm-utils is 
still there. I am not sure whether systemd uses it or does it all by itself 
tough. I configured in-kernel suspend there as well:

shambhala:/etc/pm> cat ./config.d/sleepmodule.conf
SLEEP_MODULE=kernel

Currently initramfs-tools is at 0.127 on this not so regularily updated 
laptop.


I never managed to get around actually looking whats going on here, but I now 
found that I do not have any resume setting in

shambhala:/etc/initramfs-tools> grep -ir "RESUME" .
shambhala:/etc/initramfs-tools#1>

Instead I have it here:

shambhala:/etc/default> grep -i resume grub
GRUB_CMDLINE_LINUX_DEFAULT="resume=/dev/sda2 init=/bin/systemd"


shambhala:~> swapon -s
Dateiname   Typ Größe   Benutzt 
Priorität
/dev/sda2   partition   4000180 0   -1

shambhala:~> lsblk /dev/sda2
NAME MAJ:MIN RM  SIZE RO TYPE MOUNTPOINT
sda2   8:20  3,8G  0 part [SWAP]
shambhala:~> LANG=C swapon -s
FilenameTypeSizeUsed
Priority
/dev/sda2   partition   4000180 0   -1


Okay, I will upgrade to the new initramfs-tools and look whether it fixes the 
issue. If not, I may try setting RESUME var to first "auto" and if not working 
still then the device in question to see whether that helps. I welcome any 
other ideas and I am willing to report back what I find.

Ciao,
-- 
Martin

signature.asc
Description: This is a digitally signed message part.