| Summary: | Xorg does not deal well with aperture < memory | ||
|---|---|---|---|
| Product: | xorg | Reporter: | Saleem Abdulrasool <compnerd> |
| Component: | Driver/Radeon | Assignee: | Xorg Project Team <xorg-team> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | high | CC: | benh |
| Version: | git | ||
| Hardware: | x86 (IA32) | ||
| OS: | Linux (All) | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
|
Description
Saleem Abdulrasool
2005-11-07 18:34:54 UTC
Actually, it's the opposite :) HDP_APER_CNTL does set apertures to overlap which causes the problem :) We need to detect that case where CONFIG_MEMSIZE > CONFIG_APER_SIZE and set HDP_APER_CNTL to 1 in this case, thus effectively using both apertures to map the framebuffer. However, we also need to figure out what's up with more recent cards & HDP stuff which might work a bit differently, and we need to fix various assumptions done about CONFIG_APER_SIZE when setting MC_FB_LOCATION among others (including in the DRM). The PCI BAR is actually 2*CONFIG_APER_SIZE, not CONFIG_APER_SIZE, and a setup like this one will make us use both apertures. Changing to Drivers/Radeon, the problem is not EXA specific. EXA shows it right away due to the way it uses VRAM, but it _will_ happen with XAA or anything else that starts using VRAM above the 64M point. Ben, this is fixed in the xf86-video-ati 6.5.8 release or CVS, right? Yes. On some cards by setting HOST_PATH_CNTL properly, on others by limiting the accessible video RAM, depending on the card's type and version Thanks Ben, closing. |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.