Poulsbo aka Intel GMA500 drivers PKGBUILD

skottish · 2010-07-17 04:43:10

imraro wrote:

Not good news ((

Intel doesn't have an army of engineers to work on Linux graphics drivers; I'd be surprised if they have five part-timers.

I have a laptop in my possession that has a major Intel chip that it took more than 8 years to get even a stable driver. Poulsbo on Linux was a reaction to commercial demands (Dell) and it wasn't helpful to F/OSS. Poulsbo is a failure on Linux and crappy on Windows at best.

AdamW, with Fedora I believe, has done remarkable work trying to make this functional. But, even such a great developer has been forced to stop making this a priority. The world can't wait for Intel to do anything for this chip.

sergeykish · 2010-07-19 13:58:57

Ubuntu users have working drivers for a long time. I've waited for a Fedora (they use Xorg 1.8) packages from http://www.happyassassin.net, but it takes too long. So I take AUR packages and updated them with http://code.google.com/p/gma500/.

It works on latest kernel (kernel 2.6.34.1-1) and latest Xorg (xorg-server 1.8.1.902-1). Acceleration is not working yet (Option "ShadowFB" "true").

Now I'm interested in hardware acceleration and publishing it on AUR.

jerryluc · 2010-07-19 15:55:28

sergeykish wrote:

Ubuntu users have working drivers for a long time. I've waited for a Fedora (they use Xorg 1.8) packages from http://www.happyassassin.net, but it takes too long. So I take AUR packages and updated them with http://code.google.com/p/gma500/.
It works on latest kernel (kernel 2.6.34.1-1) and latest Xorg (xorg-server 1.8.1.902-1). Acceleration is not working yet (Option "ShadowFB" "true").
Now I'm interested in hardware acceleration and publishing it on AUR.

If you have working packages using AUR and other code, why don´y you share with the rest of us?

sergeykish · 2010-07-20 08:05:35

Actually I want to share. Yesterday I asked AUR packages maintainer chris.griffith to transfer ownership or apply PKGBUILDs.

Another idea is to create new packages "psb-kmods-new", "xorg-x11-drv-psb-new" etc. Or maybe publish them on pastebin and give a link.

Wolfie276 · 2010-07-21 01:23:38

I got an issue while compiling IEGD on arch, my Makefile is as follow:

KERNELVER ?= $(shell uname -r)
KERNELDIR ?= /lib/modules/$(KERNELVER)/build
INSTALLDIR ?= /lib/modules/$(KERNELVER)/kernel/drivers/char/agp
PWD ?= $(realpath .)

EXTRA_CFLAGS += -I$(PWD)/include
EXTRA_CFLAGS += -I$(PWD)/../linux-$(KERNELVER)/drivers/char/agp
EXTRA_CFLAGS += -I/lib/modules/$(KERNELVER)/build/include/drm

all: clean modules

modules:
    @echo $(PWD)
    @$(MAKE) -C $(KERNELDIR) M=$(PWD) modules

install:
    install -o root -g root -m 755 -d $(INSTALLDIR)
    install -o root -g root -m 744 iegd_mod.ko $(INSTALLDIR)
    /sbin/depmod -a

uninstall:
    rm -rf $(INSTALLDIR)/iegd_mod.ko
    /sbin/depmod -a

clean:
    @rm -f *.o iegd*.ko iegd*.mod.c iegd*.o agp/*.o drm/*.o Module.symvers
    @rm -fr .intel* .tmp* .*.cmd agp/.*.cmd drm/.*.cmd

obj-m := iegd_mod.o
iegd_mod-objs := agp/pci.o agp/global.o agp/drv_alm.o agp/drv_nap.o agp/drv_plb.o agp/drv_cmn.o agp/drv_gn4.o drm/iegd_drv.o drm/iegd_interface.o drm/iegd_interface_265.o drm/iegd_interface_2611.o drm/iegd_interface_2615.o drm/iegd_interface_2624.o drm/psb_irq.o

And the error code is as follows:

/mnt/USB/IEGD_10_3_Linux/IKM/agp/drv_alm.c:67:2: warning: initialization from incompatible pointer type
/mnt/USB/IEGD_10_3_Linux/IKM/agp/drv_alm.c: In function iegd_alm_insert_entries:
/mnt/USB/IEGD_10_3_Linux/IKM/agp/drv_alm.c:396:10: error: struct agp_memory has no member named pages
make[2]: *** [/mnt/USB/IEGD_10_3_Linux/IKM/agp/drv_alm.o] Error 1
make[1] *** [_module_/mnt/USB/IEGD_10_3_Linux/IKM] Error 2
make[1] *** Leaving directory '/usr/src/linux-2.6.27-lts'
make: *** [modules] Error 2

The kernel version is 2.6.27-lts as you probably figured already.

EDIT: I feel like an idiot for not trying to delete the IEGD folder, then copy a fresh one. Problem solved.

Last edited by Wolfie276 (2010-07-22 02:24:47)

chris.griffith · 2010-07-23 00:11:36

Hi gang,

I have been dormant on this issue for very long time cause I was waiting for intel to do something with IEGD or the rumored "New Gallium Poulsbo" driver. I have been actively watching all sources for updates. I really did not think the Ubuntu community was gonna be able to update the psb drivers cause the binary blob was compiled against older system. Apparently, they have something working well enough that even fedora has packages.

@sergeykish: I got your request, and just have one question. If I or you update the drivers in the AUR, then users with older system will not be able to use the old PKGBUILDS, correct? Is this acceptable?

on a side note, I am so completely furiated with intel and how they have left us out to dry. I have been patiently waiting since November for some workable solution. Everyone I talk to at intel says a similar story, it is not our fault or our problem. How hard could it be to release new binary blobs?

Last edited by chris.griffith (2010-07-23 00:12:24)

Wolfie276 · 2010-07-23 00:16:34

Its not Intel's fault, its PowerVR's fault in the end. Intel however bought the rights to distribute the chip, hiring someone with little access to information about the chip.

Last edited by Wolfie276 (2010-07-23 01:39:28)

skottish · 2010-07-23 01:09:55

Wolfie276 wrote:

Its not Intel's fault, its PowerVR's fault in the end. Intel however bought the rights to distribute the chip, hiring someone with little access to information about the chip.

Intel signed a contract that they weren't prepared to fulfill, so they contracted the work out to PowerVR knowing full well that they were going to use other chips in the near future. This is, in fact, Intel's fault.

Wolfie276 · 2010-07-23 01:39:12

Intel signed a contract that they weren't prepared to fulfill, so they contracted the work out to PowerVR knowing full well that they were going to use other chips in the near future. This is, in fact, Intel's fault.

Ahh, OK. They want money as well...

EDIT:

Fine, I admit I am stuck!

I'm running 2.6.27-lts.

If I run make without copying agp.h into include, or /root/IEGD_10_3_Linux/linux-2.6.27-lts/drivers/char/agp it complains that agp.h doesn't exist, like this:

make[1]: Entering directory '/usr/src/linux-2.6.27-lts'
  CC [M]  /root/IEGD_10_3_Linux/IKM/agp/pci.o
/root/IEGD_10_3_Linux/IKM/agp/pci.c:28:17: fatal error: agp.h: No such file or directory
compilation terminated.
make[2]: *** [/root/IEGD_10_3_Linux/IKM/agp/drv_alm.o] Error 1
make[1] *** [_module_/root/IEGD_10_3_Linux/IKM] Error 2
make[1] *** Leaving directory '/usr/src/linux-2.6.27-lts'
make: *** [modules] Error 2

If i put agp.h in /root/IEGD_10_3_Linux/IKM/include or /root/IEGD_10_3_Linux/linux-2.6.27-lts/drivers/char/agp it complains about:

/root/IEGD_10_3_Linux/IKM/agp/drv_alm.c:67:2: warning: initialization from incompatible pointer type
/root/IEGD_10_3_Linux/IKM/agp/drv_alm.c: In function iegd_alm_insert_entries:
/root/IEGD_10_3_Linux/IKM/agp/drv_alm.c:396:10: error: struct agp_memory has no member named pages
make[2]: *** [/root/IEGD_10_3_Linux/IKM/agp/drv_alm.o] Error 1
make[1] *** [_module_/root/IEGD_10_3_Linux/IKM] Error 2
make[1] *** Leaving directory '/usr/src/linux-2.6.27-lts'
make: *** [modules] Error 2

Any suggestions about what the hell I can do?

EDIT2: I should mention the 2.6.27-lts is from archboot.

Last edited by Wolfie276 (2010-07-23 01:43:00)

canolucas · 2010-07-24 13:49:42

@chris.griffith: i think it is acceptable to put the new PKGBUILDs in AUR, regardless they won't work with older packages/dependencies. This is what happens with almost every package anyways.. new packages usually require new dependecies, and I see nothing wrong with that. and, by the way, the current PKGBUILDs do not work with mainline core, they are broken and out of date, so, in my opinion, the update sergeykish is ofering won't hurt. It will, in any case, make many archers happy (including me!).
just my 2 cents ..

Last edited by canolucas (2010-07-24 13:56:04)

sergeykish · 2010-07-25 12:16:15

Updated the AUR.

Bad poulsbo carma - the day I got working driver my girlfriend said "Sorry, I don't love you", netbooks HDD said "Sorry, I'm broken" and PM said "We are in the trouble, should work as much as can".

So PKGBUILDs uploaded without clean test. It will be tested only few days later with a new SSD. I made as few changes as can. PKGBUILDs requirements is not changed yet. And I wait for a response. It should work with latest kernel and xorg versions 1.7 and 1.8. My system is up to date, 2D works fine.

If startx says something like "there is no method in Xpsb.so" please remove xpsb-glx's Xpsb.so. And update its PKGBUILD.

Don't forget about xorgs
> Option "ShadowFB" "true"

sergeykish · 2010-07-25 12:35:50

Chris, thanks for psb packages! It is the great work and helped me to use Arch on a netbook about year. They was also easy to fix - up to date packages required only few patches.

Older system users should not update software. Hm... maybe I made a mistake when leaved requirements without changes. Should I set requirements to own (the only I tested)? There I can RTFM?

canolucas · 2010-07-25 23:54:13

@sergeykish: it says there is no kernel config available, so i think we need kernel26-headers as a dependency in psb-kmod's PKGBUILD? after installing this package it compiled fine. the other dependencies look ok.

after some testing, 2d accel works great, but the 3d component fails. if anyone want some logs, tell me and i'll send them.

Last edited by canolucas (2010-07-26 01:14:25)

ypnos · 2010-07-26 07:28:12

does vaapi work, too?

sergeykish · 2010-07-26 08:00:56

@canolucas: thanks, fixed build dependecies. Ubuntu have 3D, so it is possible.

@ypnos: vaapi works with 3D.

Jonhoo · 2010-07-27 22:47:41

Just tried to install the driver on a stock fully updated kernel, and X now refuses to start with the error:

Fatal server error:
[    25.523] xf86MapVidMem: Could not mmap framebuffer (0x7f6a3010,0x800000) (Operation not permitted)

Any ideas what might cause this?
Current xorg.conf:

Section "Device"
    Option "ShadowFB" "true"
    Identified "Card0"
    Driver "psb"
    VendorName "Intel Corporation"
    BoardName "System Controller Hub (SCH Poulsbo) Graphics Controller"
    BusID "PCI:0:2:0"
EndSection

Also, as a side-note, might be useful to put libdrm_poulsbo as provides libdrm?

Edit: From what I can see here (https://bbs.archlinux.org/viewtopic.php … 51#p610051) there might be a problem with GRUB loading a driver already.. Problem is that I am using GRUB2, and thus don't have a vga= line, but GRUB_GFXMODE instead.. Ideas?

Edit2: After doing a bit of research, seems this is a known error (http://www.happyassassin.net/2009/05/13 … fedora-10/). Solution was simply to add "mem=1500MB" to grub boot line. For grub2, add it after quiet in the line that reads "GRUB_CMDLINE_LINUX_DEFAULT"

Last edited by Jonhoo (2010-07-27 23:00:58)

synorgy · 2010-07-28 01:28:52

I just installed on a fully up to date 1101HA. At first I was getting some error regarding sdvo not being detected and X was failing to start. Today I rebooted and am now in X. I'm afraid that my problems with X have to do with what appears to be either a hardware problem that the linux kernel some how exacerbates or a buggy ACPI implementation (since there have been some changes with regards to eee's and their acpi in the kernel lately). I'm waiting to see how xorg reacts when this bug kicks in.

All in all - Thanks for the patched build, sergeykish. Seems to work okay here so far

I'll make a report on stability when I get a better idea of how things are going.

synorgy · 2010-07-28 02:03:09

...or not. Though I don't think it has anything to do with the driver. As soon as the bug I referred to in my post above showed up X froze and while I could get to a TTY, I couldn't get back into X. The outward error I got was that there was no usable configuration for my screen, but a look into Xorg.0.log showed that Xorg and the driver were having problems finding the monitor to be initialized. Nothing seemed to help at this point - I made sure my xorg.conf looked right and that uvesafb worked but could think of nothing else to try. I'll have to hunt this bug down somehow, but I honestly have no idea where to start. :3

canolucas · 2010-07-28 05:30:53

@synorgy: here are the working device and screen sections in my xorg.conf

Section "Device"
Identifier "Card0"
Driver "psb"
VendorName "Intel Corporation"
BoardName "System Controller Hub (SCH Poulsbo) Graphics Controller"
BusID "PCI:0:2:0"
Option "AccelMethod" "EXA"
Option "MigrationHeuristic" "greedy"
Option "NoDDC"
Option "IgnoreACPI" "yes"
Option "ShadowFB" "true"
EndSection
Section "Screen"
Identifier "Screen0"
Device "Card0"
Monitor "Monitor0"
EndSection

NOTES:
* i also removed xf86-video-fbdev, v86d and 915resolution-static packages
* and removed v86d and 915resolution hooks in /etc/mkinitcpio.conf and built again the initrd's (pacman -S kernel26), as i don't need them anymore to get 1366x786 res

Last edited by canolucas (2010-07-28 07:20:14)

synorgy · 2010-07-28 12:36:08

@canolucas - I'll try these values when I get back from work tonight, but I'm not sure they'll change anything. As I mentioned, it seems like a problem with more than justthe GMA500 on this netbook. It seems to be some sort of ACPI issue as it expresses itself through 1) jumping my reported CPU temp up to 80C 2) no longer reporting any battery information and 3) borking X if the psb driver is in use. Uvesafb allows me to continue using Xorg when this happens and Windows doesn't seem to have this issue at all, so it looks as if something has changed in the past few kernel releases. When I get some time I plan on compiling some vanilla kernel pre-2.6.30 and seeing if the problem goes away. If so, I'll move through major versions until I find the regression (provided there is a regression and not some pebkac). Unfortunately, this happens just as we're at our busiest at one of my jobs so I haven't had any time to myself lately. I'll report back when I find out more

kest · 2010-07-29 07:01:56

Hi all. somebody can make a PKGBUILD for IEGD?

ighea · 2010-07-29 13:06:43

kest wrote:

Hi all. somebody can make a PKGBUILD for IEGD?

http://www.ihku.biz/joggler/iegd/ these may help.

jerryluc · 2010-07-30 10:32:24

Thanks for the updated PKGBUILDs, sergeykish!

"oohh, I can watch smooth youtube videos on my aspire one!"

Just a little question: When i remove libdrm, will it break libgl and mesa witch depends on libdrm? Or will libdrm-poulsbo provide new files so it's only a dependency problem?

synorgy · 2010-07-30 11:50:46

It provides new files, though they aren't as high a version (I think libdrm is at 2.4.2 and libdrm_poulsbo is 2.3.1)

pascalp · 2010-07-31 12:27:03

Hi,

Does anyone have an idea about this error:

[ 34.575] (II) PSB(0): Debug: MMIO virtual address is 0xb7365000
[ 34.575] (--) PSB(0): Mapped PCI MMIO at physical address 0xd8080000
with size 512 kiB
[ 34.575] (EE) PSB(0): the stolenBase is:0x3f800000
[ 34.575] (--) PSB(0): Detected 8060 kiB of "stolen" memory set aside as video RAM.
[ 34.575] (EE) PSB(0): screnIndex is:0;fbPhys is:0x3f800000; fbsize is:0x007df000
[ 34.579] (--) PSB(0): Mapped graphics aperture at physical address 0x3f800000
with size 7 MiB
[ 34.579] (II) PSB(0): Debug: DRM device init
[ 34.580] (II) PSB(0): Poulsbo MemClock 533, CoreClock 200
[ 34.580] (II) PSB(0): Poulsbo Latencies 324 744 210 450
[ 34.580] (II) PSB(0): sku_value is 0x00800000, sku_bSDVOEnable is 1, sku_bMaxResEnableInt is 0

Arch Linux

#251 2010-07-17 04:43:10

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#252 2010-07-19 13:58:57

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#253 2010-07-19 15:55:28

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#254 2010-07-20 08:05:35

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#255 2010-07-21 01:23:38

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#256 2010-07-23 00:11:36

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#257 2010-07-23 00:16:34

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#258 2010-07-23 01:09:55

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#259 2010-07-23 01:39:12

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#260 2010-07-24 13:49:42

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#261 2010-07-25 12:16:15

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#262 2010-07-25 12:35:50

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#263 2010-07-25 23:54:13

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#264 2010-07-26 07:28:12

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#265 2010-07-26 08:00:56

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#266 2010-07-27 22:47:41

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#267 2010-07-28 01:28:52

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#268 2010-07-28 02:03:09

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#269 2010-07-28 05:30:53

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#270 2010-07-28 12:36:08

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#271 2010-07-29 07:01:56

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#272 2010-07-29 13:06:43

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#273 2010-07-30 10:32:24

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#274 2010-07-30 11:50:46

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

#275 2010-07-31 12:27:03

Re: Poulsbo aka Intel GMA500 drivers PKGBUILD

Board footer