Intel(R) GMA 950 doesn't work with Intel Extreme driver and with Vesa driver.

[walera]

My notebook Quanta SW1 with Intel 82945GM chipset and integrated Intel GMA 950 graphics. I try install different versions of Haiku from Alpha 2 to hrev46656 but by use intel_extreme driver (default loading from installation CD) I see colors noice on LCD screen. If I choice fail-safe video mode (I try all modes) I see black screen only but I hear loading from CD. Resolution of integrated LCD screen is 1280x800. I make my tests with external D-SUB monitor also but result was identical.. I try loading from Zeta 1.21 CD also. Zeta doesn't work in VESA mode also but can work in monochrome B/W mode. I make listdev command under Zeta (see attachment).

[walera]

Color noise by use intel_extreme driver

[pulkomandy]

This looks like the intel_extreme driver didn't manage to allocate enough memory for the framebuffer. A syslog would be useful, if you manage to extract it.

[walera]

Syslog by default loading

[walera]

Syslog by loading in VESA mode 1024x768 32 bit

[walera]

Replying to pulkomandy:

This looks like the intel_extreme driver didn't manage to allocate enough memory for the framebuffer.

But I see problem in VESA driver also.

A syslog would be useful, if you manage to extract it.

I make bootable USB flash and extract syslogs in two loading modes: default loading (intel_extreme driver) and by loading in VESA mode. See attachment.

[pulkomandy]

Things look fine on the video driver initialization and mode setting side. However I noticed this:

2263	KERN: add_memory_type_range(5414, 0xc0000000, 0x10000000, 0)
2264	KERN: set MTRRs to:
2265	KERN:   mtrr:  0: base: 0xc0000000, size: 0x40000000, type: 0
2266	KERN: remove_memory_type_range(86, 0xc0000000, 0x300000, 1)
2267	KERN: set MTRRs to:
2268	KERN:   mtrr:  0: base: 0xc0000000, size: 0x40000000, type: 0
2269	KERN: add_memory_type_range(5414, 0xc0000000, 0x10000000, 1)
2270	KERN: set MTRRs to:
2271	KERN:   mtrr:  0: base: 0xcf691000, size:     0x1000, type: 0
2272	KERN:   mtrr:  1: base: 0xcf697000, size:     0x1000, type: 0
2273	KERN:   mtrr:  2: base: 0xcf698000, size:     0x1000, type: 0
2274	KERN:   mtrr:  3: base: 0xd0000000, size: 0x10000000, type: 0
2275	KERN:   mtrr:  4: base: 0xe0000000, size: 0x20000000, type: 0
2276	KERN:   mtrr:  5: base: 0xc0000000, size: 0x40000000, type: 1
2277	KERN: loaded driver /boot/system/add-ons/kernel/drivers/dev/net/pegasus
2278	KERN: add_memory_type_range(5422, 0x90000, 0x70000, 0)
2279	KERN: set MTRRs to:
2280	KERN:   mtrr:  0: base:    0x90000, size:    0x10000, type: 0
2281	KERN:   mtrr:  1: base:    0xa0000, size:    0x20000, type: 0
2282	KERN:   mtrr:  2: base:    0xc0000, size:    0x40000, type: 0
2283	KERN:   mtrr:  3: base: 0xd0000000, size: 0x10000000, type: 0
2284	KERN:   mtrr:  4: base: 0xe0000000, size: 0x20000000, type: 0
2285	KERN:   mtrr:  5: base: 0xc0000000, size: 0x40000000, type: 1
2286	KERN: update_mtrrs(): Succeeded setting MTRRs after ignoring uncacheable ranges up to size 0x1000.
2287	KERN: remove_memory_type_range(5422, 0x90000, 0x70000, 0)
2288	KERN: set MTRRs to:
2289	KERN:   mtrr:  0: base: 0xcf691000, size:     0x1000, type: 0
2290	KERN:   mtrr:  1: base: 0xcf697000, size:     0x1000, type: 0
2291	KERN:   mtrr:  2: base: 0xcf698000, size:     0x1000, type: 0
2292	KERN:   mtrr:  3: base: 0xd0000000, size: 0x10000000, type: 0
2293	KERN:   mtrr:  4: base: 0xe0000000, size: 0x20000000, type: 0
2294	KERN:   mtrr:  5: base: 0xc0000000, size: 0x40000000, type: 1

There seem to be not enough MTRRs to set the complete memory caching map somewhere during video init, then the problem goes away as a range is unmapped.

Is this a 64-bit machine? I think in 64-bit mode, a better system can be used for setting this, avoiding the limitation. If it isn't, maybe we'll have to fix our MTRR allocation code again.

[bonefish]

The MTRR code is fine. It can ignore uncacheable ranges, since that property will still be enforced via the PTE flags.

[walera]

Replying to pulkomandy:

Is this a 64-bit machine? I think in 64-bit mode, a better system can be used for setting this, avoiding the limitation. If it isn't, maybe we'll have to fix our MTRR allocation code again.

This machine support 64 bit OS. CPU - Intel® Core™2 Duo Processor T7200, RAM - 4 Gb. Win7 64 bit work on this computer w/o any problems.

I try last x86_64 nightly image but with identical result..

I make new test - remove 2 Gb RAM module. After it Haiku loaded w/o any problems. See new syslog in attachment.

[walera]

Syslog by loading with installed RAM 2Gb

[waddlesplash]

There have been a lot of fixes to the Intel video driver in the past 3 years. Could you please retest?

[hey68you]

color noise and left margin added

[hey68you]

color noise and left margin added

[hey68you]

color noise and left margin added

[hey68you]

screen-shot color noise and left margin added

[hey68you]

I'm having similar issue with a left margin added (and right side of screen is out-of-bounds/hidden).

I'm on post R1/beta hrev 52397 Samsung SA300/SA350 24" Intel GMA (i945G)

Please see attachments and screenshot attachment

device Display controller (VGA compatible controller, VGA controller) [3|0|0]
  vendor 8086: Intel Corporation
  device 2772: 82945G/GZ Integrated Graphics Controller

[diver]

You seem to have a different problem though? This ticket is about black screen on boot.

[hey68you]

But I do have the color noise :)

In any case, I think you (on IRC) once pointed me to an different open ticket with a similar issue... please let me know if you can find it and/or should I open a separate new bug?

[waddlesplash]

Please retest after hrev53040.

[waddlesplash]

Color noise is indeed a separate ticket, so if there is at least output, this ticket is fixed.