Bug in Vanilla Kernel 2.6.32.2, PCI subsystem?
 

Dear linuxquestions.org-members,

I tried to get the x.org intel driver to get working on my sister's laptop (Medion E5218). The installed linux distribution is Slackware64 13.0. After having installed everything, everything except WLAN and x.org did not work. Because there were some errors like "DMA: Out of SW-IOMMU space" errors I upgraded to the vanilla kernel 2.6.32.2.

The original kernel was 2.6.29.6 and x.org's intel driver worked well, except when terminated the X-server I obtained a flickering screen.
I hoped that this issue would be solved together with the kernel update.

But now, under kernel 2.6.32.2, the intel driver does not work at all: I only get a blank, black screen. But the system is still working and I can login via ssh. I observed that X was taking 100%-CPU time. Thus I started gdb and attached to the running X-server process. After interrupting the X-server via SIGINT, gdb shows that the intel driver (intel_drv.so) waits for a flag indicating that the LCD screen is turned on, but it never detects this (If you set the variable manually via gdb, the server will startup but nothing will be displayed).

The next step of my investigation was to find out the reason for this strage behaviour. By comparing with the working 2.6.29.6 kernel, I found that the IOMEM of the intel graphics card is located at 0xc0000000, while kernel 2.6.32.2 reports the same resource to be located at 0xbc000000. This seems a little bit odd to me.

My question is whether it is usual that the address of PCI-resources can change (at will)?

/proc/iomem (for kernel 2.6.29.6)
/proc/iomem (for kernel 2.6.32.2):
Additional information:
Graphics Card: Intel 4500 MHD (PCI 00:02.0)
Total Memory: 4 Gb (but using mem=2G does not change anything).

Thank you for your answers.