Re: Problem with Radeon flat panel support

[Ani Joshi <ajoshi@shell.unixbox.com>]

Gregorio,

Look in radeon_get_pllinfo() and see the code which calculates the
ref_div, and hard code that to 67.  It seems you are getting 12 for that
calculated value, which is wrong (I've never seen any 1st generation
mac radeon (AGP or PCI) with a ref_clock of that value).  Try that and
lemme know if that fixes anything.

Other then that, all the other debug info seems correct.  You may want to
try 2.4.18-pre5 (when it comes out) or when 2.4.18 when its finished as
I've updated the driver recently w/ regards to flat panel.


ani


On 22 Jan 2002, Gregorio Gervasio Jr. wrote:

>
>         I have a Samsung flat panel connected to the DVI port of a
> Radeon Mac Edition (PCI version).  The system is a PowerWave/120 which
> boots Linux using BootX.  The flat panel worked fine (only) at
> 1280x1024@60 using 2.4.15-pre6-ben0 but not with 2.4.18-pre3-ben0.
> The difference seems to be that the new radeonfb now tries to program
> the flat panel registers whereas the old one preserved most of the
> registers (which presumably MacOS has initialized).  As far as I can
> tell, the register values are:
>
>                       initial   2.4.15-pre6  2.4.18-pre3
>                                   (works)       (fails)
> ----------------------------------------------------------
> fp_crtc_h_total_disp  00000032    00000032     009f00d6 *
> fp_crtc_v_total_disp  00000029    00000029     03ff0429 *
> fp_h_sync_strt_wid    00120008    00120008     001204fa *
> fp_v_sync_strt_wid    00030000    00030000     00030400 *
> fp_horz_stretch       009f0000    009f0000     009f0000
> fp_vert_stretch       003ff000    003ff000     003ff000
> fp_gen_ctl            040311cc    000311cd     000311cd
> ----------------------------------------------------------
>
> With debug enabled, the messages from radeonfb during boot seem okay:
>
> radeonfb_pci_register BEGIN
> PCI: Enabling device 00:0d.0 (0086 -> 0087)
> radeonfb: ref_clk=2700, ref_div=12, xclk=16615 from OF
> radeonfb: probed DDR SGRAM 32768k videoram
> radeonfb: detected DFP panel size: 1280x1024
> hStart = 1280, hEnd = 1424, hTotal = 1720
> vStart = 1025, vEnd = 1028, vTotal = 1066
> h_total_disp = 0x9f00d6^I   hsync_strt_wid = 0x1204fa
> v_total_disp = 0x3ff0429^I   vsync_strt_wid = 0x30400
> post div = 0x2
> fb_div = 0x60
> ppll_div_3 = 0x10060
> ron = 1792, roff = 22064
> vclk_freq = 10800, per = 788
> Console: switching to colour frame buffer device 160x64
> radeonfb: ATI Radeon QD  DDR SGRAM 32 MB
> radeonfb: CRT port DFP monitor connected
> radeonfb_pci_register END
>
>
> Any ideas on what may be going wrong?
>
> For now, I can get the old behavior if I fool radeonfb into thinking
> that the monitor is a regular CRT instead of a flat panel (thus
> skipping all the FP-specific code).
> --
> Gregorio Gervasio, Jr.
> gtgj@pacbell.net
>
>


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/