In MergedFB mode, a transient vertical bar is seen during LCD/DFP screen panning when cursor moves downward hitting bottom. The bar goes across the whole height of the screen, with a width of approximately 60 pixels on a 1680 wide screen. The bar disappears as soon as cursor movement stops. The problem is not observed when * In single mode (non-MergedFB mode); or * on CRT (analog) monitor; or * software cursor is used. Version of driver: xf86-video-ati-6.6.192.
At what points exactly does the effect start and stop? It sounds like it could be due to the HW cursor being positioned outside of the viewport.
(In reply to comment #1) > At what points exactly does the effect start and stop? It sounds like it could > be due to the HW cursor being positioned outside of the viewport. > When cursor moves downwards to pan the screen. The vertical bar stars when cursor hits the bottom of the screen, dispears when cursor stops moving.
(In reply to comment #2) > The vertical bar stars when cursor hits the bottom of the screen The cursor image or the hot spot (which is located where in the image)?
(In reply to comment #3) > (In reply to comment #2) > > The vertical bar stars when cursor hits the bottom of the screen > > The cursor image or the hot spot (which is located where in the image)? > The vertical bar starts when the hot spot (the tip of the arrow) hits the bottom of the screen. Horizontally the verical bar is positioned where the cursor is. More info: (1) You have to move cursor downward fast to reproduce the problem. (2) The vertical bar appears only on LCD/DFP screen, not on CRT screen, even in MergedFB mode the latter has the same image and the same cursor movement pattern is observed on both screens.
(In reply to comment #4) > The vertical bar starts when the hot spot (the tip of the arrow) hits the > bottom of the screen. Horizontally the verical bar is positioned where > the cursor is. More info: > > (1) You have to move cursor downward fast to reproduce the problem. Then it's probably indeed related to the cursor being outside of the CRTC viewport. Sounds like the MergedFB cursor code doesn't disable the HW cursor when it should.
Created attachment 10995 [details] [review] Patch to fix the problem In RADEONSetCursorPositionMerged(), while total_y2 is the vertial boundary condition (VDisplay - 1) of CRT2 for cursor of Scrn2, total_y1 could be (VDisplay - 1) of CRT1 or CRT2, depending on how they are merged. The patch ensures total_y1 be (VDisplay - 1) of CRT1.
Alex, can you review this?
Yes the patch looks fine. Feel free to commit it.
committed to 6.7 branch
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.