Created attachment 434028 [details]
/var/log/messages

You could try pci=nocrs:

Jul 23 18:01:56 snotra kernel: PCI: Using host bridge windows from ACPI; if necessary, use "pci=nocrs" and report a bug

Adding pci=nocrs did not change anything

Font corruption with matroxfb was mentioned on LKML recently:

http://permalink.gmane.org/gmane.linux.kernel/1015771

You should be able to blacklist matroxfb and use vesa or vga modes. We may just have to disable building the matroxfb driver entirely unless someone can figure out why it broke.

Fix, from: http://marc.info/?l=linux-kernel&m=128086050905521&q=raw


Fix incorrect use of memcpy_toio() in matroxfb that broke in 2.6.34.

Signed-off-by: Ondrej Zary <linux@rainbow-software.org>

--- linux-2.6.35-rc2/drivers/video/matrox/matroxfb_base.h
+++ linux-2.6.35-rc3/drivers/video/matrox/matroxfb_base.h
@@ -151,13 +151,13 @@ static inline void mga_writel(vaddr_t va
 static inline void mga_memcpy_toio(vaddr_t va, const void* src, int len) {
 #if defined(__alpha__) || defined(__i386__) || defined(__x86_64__)
 	/*
-	 * memcpy_toio works for us if:
+	 * iowrite32_rep works for us if:
 	 *  (1) Copies data as 32bit quantities, not byte after byte,
 	 *  (2) Performs LE ordered stores, and
 	 *  (3) It copes with unaligned source (destination is guaranteed to be page
 	 *      aligned and length is guaranteed to be multiple of 4).
 	 */
-	memcpy_toio(va.vaddr, src, len);
+	iowrite32_rep(va.vaddr, src, len >> 2);
 #else
         u_int32_t __iomem* addr = va.vaddr;

Fix will be in 2.6.34.2-33

Confirmed fixed with 2.6.34.2-33 great work everyone....

This bug never hit a released kernel, so closing it now.