[helium-dev] xserver-xorg-video-intel optional in Stretch

johnraff · 2017-05-15 05:14:50

This issue came up on GitHub about the helium-dev netinstall: https://github.com/BunsenLabs/bunsen-ne … /issues/16

The user's computer failed to start X till xserver-xorg-video-intel was installed. On Jessie, that is a hard, but indirect, dependency of xorg which is on our install list. In Stretch, it's a Recommend so isn't installed by default. The package now says:

Debian wrote:

The use of this driver is discouraged if your hw is new enough (ca. 2007 and newer). You can try uninstalling this driver and let the server use it's builtin modesetting driver instead.

So... should we add that package separately to our install list for helium, or not?

Head_on_a_Stick · 2017-05-15 06:56:59

johnraff wrote:

not

^ This.

The Intel DDX driver is absolutely not needed.

nobody · 2017-05-15 09:11:45

Head_on_a_Stick wrote:

johnraff wrote:
not
^ This.
The Intel DDX driver is absolutely not needed.

Obviously it is needed for some Intel GMA (as they were called back then). We should investigate if it is possible for a xorg config file to match all newer chipsets (maybe starting with Sandy Bridge) with the new driver and all older chipsets with the old driver. Perhaps it would be easier to configure the separate Intel driver as the default and override that choice for the newer chipsets to be the modeset driver? Or is it possible to have it installed and let X prefer the modeset driver, and only fall back to the Intel driver when the modeset one fails?

This is a somewhat sad state of affairs as for the first time in probably a decade, Debian Xorg wouldn't be plug&play on Intel…

johnraff · 2017-05-15 09:32:48

twoion wrote:

This is a somewhat sad state of affairs...

Yes, indeed. Clearly for the GitHub user who reported, "absolutely not needed" didn't quite cover it.

We should investigate if it is possible for a xorg config file to match all newer chipsets (maybe starting with Sandy Bridge) with the new driver and all older chipsets with the old driver. Perhaps it would be easier to configure the separate Intel driver as the default and override that choice for the newer chipsets to be the modeset driver? Or is it possible to have it installed and let X prefer the modeset driver, and only fall back to the Intel driver when the modeset one fails?

It would be great if we were able to continue support for older intel cards.

Is there an easy way to test for the video card at install time? If so, then the Debian installer we use allows for arbitary commands (including apt installs) to be run during the install process - we could install the intel driver if a card needing it was detected. A run-time solution like your suggestion of a multi-option xorg.conf would of course be more flexible.

Unia · 2017-05-15 09:34:38

Does having it installed for those who do not need it cause a drop in performance? I'm aware of the situation but I haven't done any testing myself and the internet is saying contradicting things as always.

Head_on_a_Stick · 2017-05-15 19:02:36

johnraff wrote:

"absolutely not needed" didn't quite cover it

Yes, sorry about that, my mistake :8

twoion wrote:

This is a somewhat sad state of affairs as for the first time in probably a decade, Debian Xorg wouldn't be plug&play on Intel…

+1, I am a little shocked to be honest but I suppose the removal advice was for hardware that is "ca. 2007 and newer"

Unia wrote:

Does having it installed for those who do not need it cause a drop in performance?

If anything, having the Intel DDX driver installed will make systems slightly faster, at least if my old Phoronix benchmarks were anything to go by.

The problem with xf86-video-intel is the buggy nature of the driver and the complete lack of upstream progress from Intel.

The modesetting DDX driver is being actively developed and it is the future, much like libinput & Wayland.

johnraff · 2017-05-16 01:28:28

On reflection, it might not be such a dire situation. xserver-xorg-video-intel is still a Recommend of xserver-xorg-video-all, which means, according to Debian policy:

The Recommends field should list packages that would be found together with this one in all but unusual installations.

So it's expected that the intel driver would usually be installed. (BL installs from its package list without recommends, but the Debian default is to install recommends.)

Another option might be to install the package, and wait till the post-install script (bl-welcome) to detect post-2007 hardware and then offer to remove it.

kafran · 2017-05-16 02:24:15

Do I really need this for my Atom N450 netbook? Its not that old, rsrs. Maybe it was something wrong with xorg conf.

Last edited by kafran (2017-05-16 02:37:08)

Head_on_a_Stick · 2017-05-16 06:18:08

kafran wrote:

Maybe it was something wrong with xorg conf.

You should not need an Xorg configuration file, did you create one yourself?

If you did, does LightDM start if you remove both the xorg.conf file and xserver-xorg-video-intel?

kafran · 2017-05-16 12:29:39

Head_on_a_Stick wrote:

kafran wrote:
Maybe it was something wrong with xorg conf.
You should not need an Xorg configuration file, did you create one yourself?
If you did, does LightDM start if you remove both the xorg.conf file and xserver-xorg-video-intel?

No, I did not. I followed exactly the installation instructions on github page.

johnraff · 2017-05-18 02:53:08

I have confirmed this issue on old intel hardware.
Install via helium-dev netinstall - installation proceeds normally, but after reboot fails to start X. Ttys are OK.
Install xserver-xorg-video-intel and X starts normally.
(Screen brightness keys work, as do volume keys, but the volume notification bar does not appear. That's probably another issue...)

Hardware:
Panasonic Let's note "toughbook" CF-T5KW9AXS (2006)
https://panasonic.biz/cns/pc/support/pr … st.html#t5
https://panasonic.biz/cns/pc/doc/catali … t5c06d.pdf
Graphics: 945GMS Express Integrated Graphics Controller

Debian Wiki suggests the Intel driver:
https://wiki.debian.org/GraphicsCard#Intel
https://wiki.debian.org/KernelModesetti … l_GfxCards
although a poster on the second page does suggest some possible tweaks to make modesetting work.

@HoaS I'm happy to post any logs etc that might help with finding our best policy on this.

Head_on_a_Stick · 2017-05-18 06:20:21

johnraff wrote:

best policy on this

I would say that we have to include xserver-xorg-video-intel to provide a working graphical environment for all users.

It's a real shame as I would have much preferred to drop that package but I suppose it could be a bl-welcome modification?

johnraff · 2017-05-22 01:31:50

^Yes it is a shame - it seems to be a buggy package and lacking maintainance - but at the moment it looks like the easiest way to get X working with old hardware.

It would be good to know exactly what the critical difference is that allows the modesetting driver to work without xserver-xorg-video-intel. If there was some test that bl-welcome could use, then yes a default install followed by a prompt to remove the #$%& thing on first run, if the hardware passes that test, might be a way to handle it for the moment.

kafran · 2017-05-25 09:05:38

johnraff wrote:

(...) but the volume notification bar does not appear. That's probably another issue (...)

Mine have this issue too, I didn't noticed.

I'm on a HP mini 210-1040BR

Hardware information:

System:    Host: 1040BR Kernel: 4.11.0-2.4-liquorix-amd64 x86_64 (64 bit)
           Desktop: Openbox 3.6.1 Distro: BunsenLabs GNU/Linux 8.7 (Hydrogen)
Machine:   Device: laptop System: Hewlett-Packard product: HP Mini 210-1000
           Mobo: Hewlett-Packard model: 3660 v: 48.26
           BIOS: Hewlett-Packard v: F.24 date: 03/18/2011
Battery    BAT0: charge: 39.0 Wh 95.3% condition: 41.0/56.0 Wh (73%)
CPU:       Single core Intel Atom N450 (-HT-) cache: 512 KB 
           clock speeds: max: 1666 MHz 1: 1666 MHz 2: 1666 MHz
Graphics:  Card: Intel Atom Processor D4xx/D5xx/N4xx/N5xx Integrated Graphics Controller
           Display Server: X.Org 1.19.2 drivers: intel (unloaded: modesetting,fbdev,vesa)
           Resolution: 1024x600@59.99hz
           GLX Renderer: Mesa DRI Intel Pineview M GLX Version: 2.1 Mesa 13.0.6
Audio:     Card Intel NM10/ICH7 Family High Definition Audio Controller
           driver: snd_hda_intel
           Sound: Advanced Linux Sound Architecture v: k4.11.0-2.4-liquorix-amd64
Network:   Card-1: Realtek RTL8101/2/6E PCI Express Fast/Gigabit Ethernet controller
           driver: r8169
           IF: eno1 state: down mac: 68:b5:99:4d:d4:6e
           Card-2: Broadcom Limited BCM4312 802.11b/g LP-PHY
           driver: b43-pci-bridge
           IF: wlan0 state: up mac: f0:7b:cb:98:02:b8
Drives:    HDD Total Size: 250.1GB (10.0% used)
           ID-1: /dev/sda model: ST9250410AS size: 250.1GB
Partition: ID-1: / size: 15G used: 4.0G (28%) fs: ext4 dev: /dev/dm-2
           ID-2: /boot size: 474M used: 71M (16%) fs: ext2 dev: /dev/sda1
           ID-3: /home size: 206G used: 12G (6%) fs: ext4 dev: /dev/dm-3
           ID-4: swap-1 size: 8.59GB used: 0.00GB (0%) fs: swap dev: /dev/dm-1
Sensors:   System Temperatures: cpu: 45.0C
Info:      Processes: 168 Uptime: 7 min Memory: 424.9/1991.4MB
           Client: Shell (bash) inxi: 2.3.5

kafran · 2017-06-02 21:49:15

What's up guys. Some news about this topic.

After installing Bunsenlabs Helium-dev on my netbook (an old hardware), I decided to migrate from Mint XFCE to Bunsenlabs Helium-dev on my main computer too (a not so old one). I had no problem with the modesetting driver to start X, but a new problem for me: screen tearing!

With the intel driver I had this on /usr/share/X11/xorg.conf.d/20-intel.conf:

Section "Device"
   Identifier  "Intel Graphics"
   Driver      "intel"
   Option      "AccelMethod" "SNA"
   Option      "TearFree"    "true"
EndSection

After looking all over the internet how to skip the screen tearing on intel driver with modesetting I found this https://whirm.eu/posts/fix-for-xorgs-mo … r-tearing/

The autor says to run compton with "--backend=glx" or "--backend=glx --glx-swap-method=buffer-age"

Starting compton with "--backend=glx" fixed my problem. I think there is no harm to make this default on Bunsenlabs for a tearing free desktop.

Here my hardware info:

System:    Host: i5447 Kernel: 4.11.0-3.3-liquorix-amd64 x86_64 (64 bit)
           Desktop: Openbox 3.6.1 Distro: Debian GNU/Linux 9 (stretch)
Machine:   Device: portable System: Dell product: Inspiron 5447 v: A05
           Mobo: Dell model: 0V6MJM v: A00 UEFI: Dell v: A05 date: 08/12/2014
Battery    BAT1: charge: 31.5 Wh 100.0% condition: 31.5/38.4 Wh (82%)
CPU:       Dual core Intel Core i5-4210U (-HT-MCP-) cache: 3072 KB 
           clock speeds: max: 2401 MHz 1: 1400 MHz 2: 768 MHz 3: 768 MHz
           4: 1700 MHz
Graphics:  Card: Intel Haswell-ULT Integrated Graphics Controller
           Display Server: X.Org 1.19.2 drivers: modesetting (unloaded: fbdev,vesa)
           Resolution: 1366x768@60.00hz
           GLX Renderer: Mesa DRI Intel Haswell Mobile
           GLX Version: 3.0 Mesa 13.0.6
Audio:     Card-1 Intel 8 Series HD Audio Controller driver: snd_hda_intel
           Card-2 Intel Haswell-ULT HD Audio Controller driver: snd_hda_intel
           Sound: Advanced Linux Sound Architecture v: k4.11.0-3.3-liquorix-amd64
Network:   Card-1: Realtek RTL8101/2/6E PCI Express Fast/Gigabit Ethernet controller
           driver: r8169
           IF: enp1s0 state: down mac: ec:f4:bb:f4:de:53
           Card-2: Qualcomm Atheros QCA9565 / AR9565 Wireless Network Adapter
           driver: ath9k
           IF: wlp2s0 state: up mac: 90:48:9a:ef:7b:a5
           Card-3: Atheros
           IF: null-if-id state: N/A speed: N/A duplex: N/A mac: N/A
Drives:    HDD Total Size: 1000.2GB (46.5% used)
           ID-1: /dev/sda model: ST1000LM024_HN size: 1000.2GB
Partition: ID-1: / size: 30G used: 3.5G (12%) fs: ext4 dev: /dev/dm-2
           ID-2: /boot size: 473M used: 72M (16%) fs: ext2 dev: /dev/sda2
           ID-3: /home size: 872G used: 416G (51%) fs: ext4 dev: /dev/dm-3
           ID-4: swap-1 size: 16.11GB used: 0.00GB (0%) fs: swap dev: /dev/dm-1
Sensors:   System Temperatures: cpu: 46.0C mobo: N/A
           Fan Speeds (in rpm): cpu: 2800
Info:      Processes: 175 Uptime: 37 min Memory: 812.2/7867.0MB
           Client: Shell (bash) inxi: 2.3.5

Last edited by kafran (2017-06-02 21:50:42)

Head_on_a_Stick · 2017-06-02 21:58:21

kafran wrote:

The autor says to run compton with "--backend=glx" or "--backend=glx --glx-swap-method=buffer-age"
Starting compton with "--backend=glx" fixed my problem. I think there is no harm to make this default on Bunsenlabs for a tearing free desktop.

This has come up before and I think that the compton options are hardware-dependent and may not work on all devices.

Also, many find the "drag lag" induced by the vsync to be intolerable.

EDIT: thanks for the suggestion!

Last edited by Head_on_a_Stick (2017-06-02 21:59:36)

kafran · 2017-06-03 00:10:31

Head_on_a_Stick wrote:

This has come up before

Really? It was hard to find this information. Maybe I should had looked here first.

Head_on_a_Stick wrote:

and I think that the compton options are hardware-dependent and may not work on all devices.

Ahw. Ok. Well, I'm still finding myself with these new Xorg drivers. For example, now I need to figure out how to configure my touchpad with libinput. I already find out how to activate "tap to click". Now I'm thinking the double click too slow.

johnraff · 2017-06-03 00:28:51

Head_on_a_Stick wrote:

I think that the compton options are hardware-dependent and may not work on all devices.

In the Hydrogen bl-compositor script we check glxinfo before applying vsync:

EXECXCOMP='compton -b'
if glxinfo | egrep -iq 'direct rendering: yes'; then
    EXECXCOMP+=' --vsync opengl'
fi

Is there some similar test we could do before applying '--backend=glx'?

kafran · 2017-06-03 12:15:41

johnraff wrote:

Is there some similar test we could do before applying '--backend=glx'?

The test I did was to watch this video https://launchpadlibrarian.net/27475515 … _60fps.mp4

Well, at least an improvement would be "bl-compositor" receiving compton arguments or have this as a configurable option so the user do no need to change the "bl-compositor" program but "~/.conf/openbox/autostart"

Last edited by kafran (2017-06-03 12:17:04)

brontosaurusrex · 2017-06-03 13:34:09

compton -cCG --backend glx --paint-on-overlay < is working fine with legacy nvidia drivers (macpro 2009).

inxi -G
Graphics:  Card: NVIDIA G96 [GeForce 9500 GT]
           Display Server: X.Org 1.19.2 driver: nvidia Resolution: 1920x1080@60.00hz, 1920x1200@59.95hz
           GLX Renderer: GeForce GT 120/PCIe/SSE2 GLX Version: 3.3.0 NVIDIA 340.102

just saying.

Last edited by brontosaurusrex (2017-06-03 13:35:07)

#1 2017-05-15 05:14:50

[helium-dev] xserver-xorg-video-intel optional in Stretch

#2 2017-05-15 06:56:59

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#3 2017-05-15 09:11:45

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#4 2017-05-15 09:32:48

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#5 2017-05-15 09:34:38

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#6 2017-05-15 19:02:36

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#7 2017-05-16 01:28:28

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#8 2017-05-16 02:24:15

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#9 2017-05-16 06:18:08

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#10 2017-05-16 12:29:39

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#11 2017-05-18 02:53:08

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#12 2017-05-18 06:20:21

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#13 2017-05-22 01:31:50

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#14 2017-05-25 09:05:38

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#15 2017-06-02 21:49:15

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#16 2017-06-02 21:58:21

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#17 2017-06-03 00:10:31

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#18 2017-06-03 00:28:51

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#19 2017-06-03 12:15:41

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#20 2017-06-03 13:34:09

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

Board footer