1612 replies to this topic

[chopped]

try installing the kexts i posted above with kext helper.

[Cristianbam]

So guys, I have a Toshiba Satellite L300 with an Mobile Intel® GMA 4500M graphics card. Will these kexts work for me?

[VampireMI2]

Hello! everybody thanks for hard work
i have HP CQ40 install Leopard 10.5.7 retail with vanilla kernel
Video GMA x4500mhd (2a42:8086)
sound IDT -->VoodooHDA 0.2.2 Work!
Ethernet work!
Wlan Broadcom Work!
Bluetooth Work!
Touchpad Work!
every thing work only Video Card

When I follow Tiennek with patch 20090704.2

kext are load
resolution work! (1280x800) very good
but mouse issue
no mouse cursor but i can click and work

What can i do
Pleases help me

[Benjan]

chopped, on Jul 11 2009, 03:19 PM, said:

try installing the kexts i posted above with kext helper.

Thanks, but I tried exactly as u said, the same result,  still got black(or dark gray) screen after blue one.

[coldequation]

Family specific matching doesn't mean much. I get that and my display works fine. You'll note that in your log, it tries to get the kexts to load multiple times.

The key lines are between:

Jul  8 15:42:30 shahzebds-2081cto kernel[0]: nstruction_state_L2_disable instruction_state_L2_fill_buffers_disable instruction_L1_queue_disable instruction_state_invalidate indirect_load_enable

and

Jul  8 15:42:36 shahzebds-2081cto kernel[0]: display: Not usable

In there, it's apparent that certain registers and addresses are not set correctly, so the kext is not
usable (even if it shows up a loaded). The dumped SVG registers look pretty much the same as Tienneke's.

Oh, well.

chopped, on Jul 10 2009, 07:22 PM, said:

here are the kexts that loaded for me, and here is my startup log the startup log is interesting because i assume that i get a family specific matching fails for the video.
Archive.zip is the kexts.

[iPoco]

Tienneke, on Jul 8 2009, 07:00 AM, said:

Can you post a picture? (screenshot won't show it.)

Here it is. You can't really see much from it but anyways. And also on the left side there is a rectangle boxish thing. That stays 5cm below the mouse.

Found something interesting. On some reboots my screen looks like the green has been set too high and then if I go into the System Preferences > Display and change the resolution it returns to normal color but the lines remain.

iPoco

Attached Files

•  photo.jpg   89.36K   91 downloads

[chopped]

here a list of everything in s/l/e

Attached Files

•  extentionslist.txt   17.54K   24 downloads

[coldequation]

Status update.

I'm including these updates to document things for myself, other developers, and the future. Much of what follows will be gobbledygook, so just ignore it, if you're not a geek.

I've been trying to decode the Intel manuals, and I think I'm getting closer to understanding what is happening. The crucial lines are found in the system logs of people who don't use the -res option of patch4500.

Recall that using the -res option deletes the 2A42AppleIntelGMAX3100.kext, which means it won't load. But it's not helpful in trying to figure out *why* it's failing. The crucial lines are these:

Jul  3 01:38:22 localhost kernel[0]: ESR:  Instruction error
Jul  3 01:38:22 localhost kernel[0]: ESR: Page Table Error
Jul  3 01:38:22 localhost kernel[0]: PGTBL_ERR: MI_PTERROR_INVALID_TLB_MISS
Jul  3 01:38:22 localhost kernel[0]: PGTBL_ERR source: Reserved

The Intel doc says that:

Invalid TLB Miss:  An unexpected TLB miss (detected at GTT request time) was encountered (e.g., during Display/Overlay/Sprite access).

Recall that I highlighted three differences: the GTT, stolen memory, and *overlays*.

Once I figure out all the acronym soup, and why this error happens, I think I can come up with a way of injecting changed values, *before* the display kexts load up.

To test this, we can try using reggie_se in the CHUD developer tools to change thinks in the memory mapped registers. Then we try loading the kext on the command line and see what happens.

Hopefully, I'll have some commands to try out next week. Enjoy your Sundays, everyone.

-u

[coldequation]

Cursor issues:

The Intel doc might suggest why Lenovo users can't see the cursor:

Quote

PIPEACONF—Pipe A Configuration Register
Memory Offset Address: 70008h
Default: 00000000h
Normal Access: Read/Write double buffered
Cursor Planes Off: This bit when set will cause all enabled cursor planes that are assigned to this
pipe to be disabled by overriding the current setting of the plane enable bit, at the next VBLANK.  
Timing signals continue as they were but the cursor(s) no longer appear on the screen. Setting the bit back to a zero will then allow the cursor planes to resume on the following VBLANK.
0 = Normal Operation
1 = Planes assigned to this pipe are disabled.

This is bit 18. I'll double check by testing with reggie_se on my GMA950, but it sounds like we can set this bit (the register is marked as read/write) to enable the cursor.

The problem might also be in a different register:

Quote

CURACNTR—Cursor A Control Register
This register, and all other cursor registers will remain in their holding register (readable) after a
write. The holding registers are transferred into the active registers on the asserting edge of
Vertical Blank only after a write cycle to the base address register has completed.  
Memory Offset Address: 70080h
Default: 00000000h
Normal Access: Read/Write
Cursor is disabled. This is the default after reset.  When the cursor register value
changes from enabled to disabled, the cursor will stop fetching data at the following
VBLANK event.  
The cursor enable can be overridden by the pipe cursor disable bit.  The value of these
bits do not change when disabled by the pipe cursor disable bit.
Bits 5 and 2-0.

I can try to make my machine behave like the lenovos, and then (hopefully!) set it back. I'll try this Mon or Tuesday.

-u

[Trauma!]

Hey Coldequation,

you are doing a great job I hope some other experimented devs like you will join the thread !!

Keep up the good work m8, I'm steel around if you need me to test or try things out

Latter then...

PS : About AppleIntelGMAX3100.kext loading ... it seems that chopper has delete all the related bundles and plugins, wich explain why he can load the driver. How ever it could be interesting to compare IOregs dumps from X4500 user with AppleIntelGMAX3100.kext loaded without the plugins.

[qualchrone]

after upgrading to 10.5.7 and installing trauma's, I finally got native resolution
yet, my lcd got darker so powerless that I was bare to see

anyone can help?

[nomi81]

Hello guys any way to fix my problem, I even downloaded the kexts by trauma and also tried patch and deleted the kext which is without FB but still my display goes to blue screen and then blank. I am using Dell Latitude E6400. Please do help to me.

At least just only native resolution.

Thanks

[oliveirapaulo]

Hey guys...

I've a Sony VAIO VGN-FW245J... This notebook uses an Intel GMA X4500MHD...

I've hex-edited AppleIntelGMAX3100 and AppleIntelGMAX3100FB binaries and the corresponding Info.plist files... Leopard starts up but I've an unusable display (looks like a light gray --- see it below)...



After that I connected using VNC and using System Profiler I could see that the two kexts were loaded.

Then I've the idea of connecting my Panasonic Plasma 42" PZ80 via the VGA out from notebook. After clicking 'Detect Displays' about 30 seconds later (in this time I observed my notebook blanking screen 3 or 4 times --- perhaps trying to find an usable resolution) my Plasma TV shows what you can see in picture 2... Yes... 800x600 resolution only (I tried higher ones with no luck) and no artifacts/corruption!!!



Unfortunately no CI/QE and I've noted that all known resolutions (from 640x480 to 1920x1200) are listed in 'System Preferences' for the built-in notebook display (perhaps because Leo couldn't get a valid list from the hardware --- incorrect addresses???)...

I've disassembled AppleIntelGMAX3100 and AppleIntelGMAX3100FB files and I'll try to find operations with addresses and registers on GM965 chipset, according to the Intel specs and datasheets.

Patience and research...

[coldequation]

Thanks! BTW, thanks for all your help! Also, what you found WRT to chopper makes sense. I posted earlier what Slice said about the plugins-- they are crucial to getting QE/Ci to work.

I just wish I had a 4500 to work on-- it would make things way easier.

Trauma!, on Jul 12 2009, 04:58 AM, said:

Hey Coldequation,

you are doing a great job I hope some other experimented devs like you will join the thread !!

Keep up the good work m8, I'm steel around if you need me to test or try things out

Latter then...

PS : About AppleIntelGMAX3100.kext loading ... it seems that chopper has delete all the related bundles and plugins, wich explain why he can load the driver. How ever it could be interesting to compare IOregs dumps from X4500 user with AppleIntelGMAX3100.kext loaded without the plugins.

Excellent work, oliveirapaulo.

@oliveirapaulo. Please see my post that looked at the BSD source diffs. The main issues are the GTT, Pipe, Overlay and whether stolen memory is used or not. That will, of course affect the size and offsets of memory used.

I also posted some details from Slice that tell what each of the plugins do, so the problem may be with a plugin that is trying to load, which then fails, causing the kext not to load. Since Trauma found that chopper can load the 24A2AppleIntelGMAX3100.kext *without* the plugins, you may want to investigate that issue. Of course, if 24A2AppleIntelGMAX3100 sets up something badly for a plugin, which then fails, that could be the issue too!

BTW, this morning I am able to look at the memory and registers of my GMA950 with reggie_se. You might want to review the intel docs, and try reading/writing registers with that to see if you can effect any changes to the 4500's behavior.

@nomi81: please use Tienneke's Genius Bar thread to debug issues with native resolution. Unfortunately, and especially since I do not hae a 4500, I can't help much. You can also see if Tienneke or Trauma is on the irc channel previously posted.

@qualchrone: I found some registers that specifically refer to backlight control, as well as one that deals with legacy backlight control. Later in the week-- remind me if I forget, I'll try to suggest some things you can try. Right now, I'm trying to see if I can fix the cursor issues that lenovo users are having. Before I get to your brightness issue, you will need to install the CHUD package from XCODE Tools on the Retail OSX DVD or Download it at http://connect.apple.com/ (you'll need an account there).

oliveirapaulo, on Jul 12 2009, 06:12 AM, said:

Hey guys...

I've a Sony VAIO VGN-FW245J... This notebook uses an Intel GMA X4500MHD...

Then I've the idea of connecting my Panasonic Plasma 42" PZ80 via the VGA out from notebook. After clicking 'Detect Displays' about 30 seconds later (in this time I observed my notebook blanking screen 3 or 4 times --- perhaps trying to find an usable resolution) my Plasma TV shows what you can see in picture 2... Yes... 800x600 resolution only (I tried higher ones with no luck) and no artifacts/corruption!!!

Unfortunately no CI/QE and I've noted that all known resolutions (from 640x480 to 1920x1200) are listed in 'System Preferences' for the built-in notebook display (perhaps because Leo couldn't get a valid list from the hardware --- incorrect addresses???)...

I've disassembled AppleIntelGMAX3100 and AppleIntelGMAX3100FB files and I'll try to find operations with addresses and registers on GM965 chipset, according to the Intel specs and datasheets.

Patience and research...

[coldequation]

I NEED YOUR HELP. See below.

I'd like to see if we can sort out the problems that lenovo users are having with their vanishing cursors in native resolution. So I'll need help from two groups of people: those of you who have native resolution with a working cursor, and lenovo users who are having cursor problems. Lenovo users: this may be painful, as you MUST do this in native resolution mode (meaning you won't be able to see the cursor!).

Here are the tests to try:

1. First install CHUD tools from your retail Apple DVD, or download from  http://connect.apple.com/ (you need to have an account there).

2. Open up a terminal window. Type sudo -s, and type your password.

3. Next, if you have a working cursor, type this command:

./cursorCheck.sh good "Gateway MX8738"

but replace the Gateway with your make and model in quotes.

If your cursor is not working, type this command:

./cursorCheck.sh bad "Gateway MX8738"

but replace the Gateway with your make and model in quotes.

4. Post the resulting output (it will be in a file named lenovo-test.txt)

I am not sure what the result should be on a 4500, but I'll compare the files and see if we can then use reggie_se to set the correct values.

I've attached the cursorCheck.sh script as a zip file.

Thanks,
-u

Attached Files

•  cursorCheck.sh.zip   1.09K   18 downloads

[iPoco]

Here you go.

I was wondering about my lines issue. Could the type of display influence it? There were three types and I was wondering if this could change anything?

iPoco

Attached Files

•  lenovo_test.txt   553bytes   23 downloads

[coldequation]

Thanks. I can't guess about your lines issue. Have you tried to see if if happens on an external display?

ipoco, on Jul 12 2009, 09:12 AM, said:

Here you go.

I was wondering about my lines issue. Could the type of display influence it? There were three types and I was wondering if this could change anything?

iPoco

[Tienneke]

coldequation, on Jul 12 2009, 01:50 PM, said:

I've attached the cursorCheck.sh script as a zip file.

Thanks,
-u

I've attached my results here Hope you've got something interesting

Attached Files

•  lenovo_test.txt   554bytes   20 downloads

[chopped]

im a little confused at the working cursor vs non working cursor. even if the cursor is visible once we increase the size, its considered non working because its not visible at normal size right?

[iPoco]

Yup.