https://bugzilla.kernel.org/show_bug.cgi?id=29502

Summary: Change to legacy algo to preferring lower post dividers breaks certain modelines Product: Drivers Version: 2.5 Kernel Version: 2.6.38-rc5 Platform: All OS/Version: Linux Tree: Mainline Status: NEW Severity: normal Priority: P1 Component: Video(DRI - non Intel) AssignedTo: drivers_video-dri@kernel-bugs.osdl.org ReportedBy: bitbytebit@gmail.com Regression: Yes

There's an issue with some legacy chips plus low dotclocks in the 2.6.38-rc5 kernel, This is the change/patch that broke them. Reverting this patch, they work as they did before...

"Also, switch the legacy algo back to preferring lower post dividers."

@@ -849,7 +951,7 @@ void radeon_compute_pll(struct radeon_pll *pll, max_fractional_feed_div = pll->max_frac_feedback_div; }

- for (post_div = max_post_div; post_div >= min_post_div; --post_div) { + for (post_div = min_post_div; post_div <= max_post_div; ++post_div) { uint32_t ref_div;

if ((pll->flags & RADEON_PLL_NO_ODD_POST_DIV) && (post_div & 1))

A person has a ATI Radeon 9200SE 5964 (AGP)and with 2.6.38-rc4-git4+ it works while with 2.6.38-rc5 there are certain (not all) modelines with low dotclocks that end up being totally out of sync.

Here's an example of a non-working modeline for him with the newer kernel, vs. the older one step back before the patch with the above change...

# toki 256x224@59.61 15.6774Khz ModeLine "256x224x59.61" 5.518455 256 272 304 352 224 235 238 263 -HSync -VSync

There's a lot of other ones thought that can work with the same lower dotclocks, so I'm not sure why that is, he tested that change with the legacy pll computation reversed back to how it was and now works again with modelines like the above.

Thanks, Chris