<html> <head> <base href="https://bugs.freedesktop.org/" /> </head> <body> <div> <a class="bz_bug_link bz_status_NEW " title="NEW --- - [Ubuntu 12.04.1 LTS 32-bit] Unity GUI will hang during bootup with ATI Technologies Radeon X600 (RV380) and VIA Technologies PT880 Ultra chipset" href="https://bugs.freedesktop.org/show_bug.cgi?id=55784#c38">Comment # 38</a> on <a class="bz_bug_link bz_status_NEW " title="NEW --- - [Ubuntu 12.04.1 LTS 32-bit] Unity GUI will hang during bootup with ATI Technologies Radeon X600 (RV380) and VIA Technologies PT880 Ultra chipset" href="https://bugs.freedesktop.org/show_bug.cgi?id=55784">bug 55784</a> from <a class="email" href="mailto:mypersonalmailbox1@mail.com" title="mypersonalmailbox1@mail.com">mypersonalmailbox1@mail.com</a> <pre>(In reply to <a href="show_bug.cgi?id=55784#c37">comment #37</a>) Hi Alex, My replies are embedded below. > (In reply to <a href="show_bug.cgi?id=55784#c36">comment #36</a>) > As I understand it, this board is some kind of hybrid AGP and PCIE > combination. I'm not sure how well this hw was validated by the > manufacturer, especially with newer asics. The fact that all PCIE cards > seem to fail seems to indicate some sort of chipset problem. > > Yes, PT880 Ultra is a little weird PCI Express/AGP hybrid chipset. ULi also had such chipset as well (used HyperTransport to link multiple ULi ASICs), but with AGP 3.0 and full PCI Express x16 support. I own one such board (from ASRock), but it doesn't POST probably because the mainboard got fried by a bad power supply. I picked up that ASRock board from E-waste collection bin so I shouldn't expect much, either. A friend of mine who used to work for Philips Semiconductors once told me that if a PCI device cannot pass PCI-SIG's compatibility testing, one has almost no chance of selling the part to mainstream computer/board manufacturers. VIA Technologies claims that PT880 Ultra is listed under PCI-SIG's integrator's list. <a href="http://www.via.com.tw/en/products/chipsets/p4-series/pt880_ultra/index.jsp">http://www.via.com.tw/en/products/chipsets/p4-series/pt880_ultra/index.jsp</a> <a href="http://www.pcisig.com/developers/compliance_program/integrators_list/pcie">http://www.pcisig.com/developers/compliance_program/integrators_list/pcie</a> Just a few hours ago, I added XFX Radeon HD 5450 PCIe graphics card to ASUS P5VD1-X. As expected, it couldn't boot Ubuntu 12.04 LTS 32-bit. It hangs up after displaying Ubuntu 12.04 LTS Unity GUI reddish wallpaper (BTW, mouse cursor still can move around the screen correctly.), but I was able to boot Ubuntu 10.04.4 LTS 32-bit from a DVD. I tested the ACPI S3 State resume, and it resumes reliably. So I know many developers have negative opinions of VIA Technologies chipset, but at least with Ubuntu 10.04.4 LTS 32-bit, it works perfectly. Somebody broke the code in Ubuntu 12.04 LTS. > I'm not sure. You could check the maintainers file in the kernel. > > I'm not familiar with WinDbg. Since you get the hang when the GUI starts > up, that would indicate that the problem is related to acceleration, > probably DMA. You can try adding: > Option "NoAccel" "True" > to the device section of your xorg.conf to disable acceleration. If that > works, then you can further narrow down the problem. > > Off hand I would guess that perhaps the chipset has problems with MSIs or > DMA. You can try booting with: > pci=nomsi > on the kernel command line in grub to disable MSIs. If that fixes it, > you'll need to add a pci quirk to the kernel to disable MSIs on that chipset. > > If the board has problems supporting >32 bit DMA mask, you can try forcing > the DMA mask to 32 in the driver: > diff --git a/drivers/gpu/drm/radeon/radeon_device.c > b/drivers/gpu/drm/radeon/radeon_device.c > index e2f5f88..98945e4 100644 > --- a/drivers/gpu/drm/radeon/radeon_device.c > +++ b/drivers/gpu/drm/radeon/radeon_device.c > @@ -1043,6 +1043,7 @@ int radeon_device_init(struct radeon_device *rdev, > if ((rdev->flags & RADEON_IS_PCI) && > (rdev->family <= CHIP_RS740)) > rdev->need_dma32 = true; > + rdev->need_dma32 = true; > > dma_bits = rdev->need_dma32 ? 32 : 40; > r = pci_set_dma_mask(rdev->pdev, DMA_BIT_MASK(dma_bits)); > > > If that works, you'd need to add a PCI quirk for the chipset to limit the > PCI DMA mask to 32 bits. Regards, fpgahardwareengineer</pre> </div> <hr> You are receiving this mail because: <ul> <li>You are the assignee for the bug.</li> </ul> </body> </html>