From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jakub Bogusz Subject: tdfxfb patch - fixes and enhancements Date: Mon, 9 Dec 2002 22:35:48 +0100 Sender: linux-fbdev-devel-admin@lists.sourceforge.net Message-ID: <20021209213548.GC8048@satan.blackhosts> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="tsOsTdHNUZQcU9Ye" Return-path: Received: from pg6.warszawa.cvx.ppp.tpnet.pl ([213.76.102.6] helo=satan.blackhosts) by sc8-sf-list1.sourceforge.net with esmtp (Cipher TLSv1:DES-CBC3-SHA:168) (Exim 3.31-VA-mm2 #1 (Debian)) id 18LWOD-0006KU-00 for ; Mon, 09 Dec 2002 14:28:53 -0800 Received: from satan.blackhosts (localhost [127.0.0.1]) by satan.blackhosts (8.12.6/8.12.6) with ESMTP id gB9MYXmM021081 for ; Mon, 9 Dec 2002 23:34:58 +0100 Received: (from qboosh@localhost) by satan.blackhosts (8.12.6/8.12.6/Submit) id gB9LZmDZ019880 for linux-fbdev-devel@lists.sourceforge.net; Mon, 9 Dec 2002 22:35:48 +0100 Content-Disposition: inline Errors-To: linux-fbdev-devel-admin@lists.sourceforge.net List-Help: List-Post: List-Subscribe: , List-Id: List-Unsubscribe: , List-Archive: To: linux-fbdev-devel@lists.sourceforge.net --tsOsTdHNUZQcU9Ye Content-Type: text/plain; charset=us-ascii Content-Disposition: inline (note that I'm not subscriber, so please Cc answers to me) Is this list proper place for tdfxfb-related things, or should I post to some other address (linux-kernel?)? I don't see tdfxfb in MAINTAINERS file. 1. hardware cursor (seems to be Voodoo4, maybe Voodoo5 too, specific). On my Voodoo4 4500 I got some (mostly white) junk in bottom 48 (of 64) lines of hardware cursor. After setting hardware cursor base 1 page lower (alternatively speaking, rounding it to even page boundary), cursor is displayed correctly. 2. 16/24/32bpp modes should be defined as TRUECOLOR, not DIRECTCOLOR. It's because Voodoo (or at least tdfxfb driver) doesn't support colour mapping in high- and true-colour modes. This seems to be already fixed in Linux 2.5 tree. It caused displaying Tux logo in wrong colours on 16/24/32bpp tdfxfb. 3. (partial) double-scan modes support. Double-scan modes were silently not supported (flag was set only in internal data structure, but not in chipset); with my patch, it's possible to set such mode (but without hardware cursor - because it's handled in normal Y coordinate, not doubled). 4. interlaced modes support. As far as I could find - Voodoo Banshee doesn't support interlace, but Voodoo3 (at least some models) and higher do. With patch, it works on my Voodoo4 4500 (refresh rate seems to be fixed at 50Hz (for TV Out, which I don't have?), but it's still useful for displaying some high resolution picture for a while). Attached patch was made againt Linux 2.4.20, and requires some changes for 2.5 tree (BTW, it seems that hardware cursor support is gone in tdfxfb for 2.5?). -- Jakub Bogusz http://www.cs.net.pl/~qboosh/ PLD Linux http://www.pld.org.pl/ --tsOsTdHNUZQcU9Ye Content-Type: text/plain; charset=us-ascii Content-Disposition: attachment; filename="linux-tdfxfb-fixes.patch" This patch contains: * fix for hardware cursor (Voodoo4 4500 specific problem??? bottom 48 lines contained random data if cursor start address was not aligned down to even page boundary...) * fix for logo colours (I'm not sure about hardware, but at least this driver doesn't support colour lookup tables in 16/24/32bpp, so visual must be TRUECOLOR, not DIRECTCOLOR) * interlace support (works on Voodoo4 (50 Hz only?), problably Voodoo[35] too) * partial doublescan support (hardware cursor still doesn't work correctly, but such modes doesn't just turn monitor out-of-sync) -- Jakub Bogusz --- linux-2.4.20/drivers/video/tdfxfb.c.orig Mon Feb 25 20:38:07 2002 +++ linux-2.4.20/drivers/video/tdfxfb.c Mon Dec 2 23:19:54 2002 @@ -208,6 +208,7 @@ #define VGAINIT1_MASK 0x1fffff #define VIDCFG_VIDPROC_ENABLE BIT(0) #define VIDCFG_CURS_X11 BIT(1) +#define VIDCFG_INTERLACE BIT(3) #define VIDCFG_HALF_MODE BIT(4) #define VIDCFG_DESK_ENABLE BIT(7) #define VIDCFG_CLUT_BYPASS BIT(10) @@ -238,6 +239,7 @@ #define TDFXF_VSYNC_ACT_LOW 0x08 #define TDFXF_LINE_DOUBLE 0x10 #define TDFXF_VIDEO_ENABLE 0x20 +#define TDFXF_INTERLACE 0x40 #define TDFXF_HSYNC_MASK 0x03 #define TDFXF_VSYNC_MASK 0x0c @@ -1321,10 +1323,17 @@ hbs = hd; hbe = ht; - vd = par->vdispend - 1; - vs = par->vsyncsta - 1; - ve = par->vsyncend - 1; - vt = par->vtotal - 2; + if (par->video & TDFXF_LINE_DOUBLE) { + vd = (par->vdispend << 1) - 1; + vs = (par->vsyncsta << 1) - 1; + ve = (par->vsyncend << 1) - 1; + vt = (par->vtotal << 1) - 2; + } else { + vd = par->vdispend - 1; + vs = par->vsyncsta - 1; + ve = par->vsyncend - 1; + vt = par->vtotal - 2; + } vbs = vd; vbe = vt; @@ -1429,9 +1438,6 @@ VGAINIT0_EXTSHIFTOUT; reg.vgainit1 = tdfx_inl(VGAINIT1) & 0x1fffff; - fb_info.cursor.enable=reg.vidcfg | VIDCFG_HWCURSOR_ENABLE; - fb_info.cursor.disable=reg.vidcfg; - reg.stride = par->width*cpp; reg.cursloc = 0; @@ -1450,9 +1456,20 @@ reg.gfxpll = do_calc_pll(..., &fout); #endif - reg.screensize = par->width | (par->height << 12); - reg.vidcfg &= ~VIDCFG_HALF_MODE; + if (par->video & TDFXF_LINE_DOUBLE) { + reg.screensize = par->width | (par->height << 13); + reg.vidcfg |= VIDCFG_HALF_MODE; + reg.crt[0x09] |= 0x80; + } else { + reg.screensize = par->width | (par->height << 12); + reg.vidcfg &= ~VIDCFG_HALF_MODE; + } + if (par->video & TDFXF_INTERLACE) + reg.vidcfg |= VIDCFG_INTERLACE; + fb_info.cursor.enable=reg.vidcfg | VIDCFG_HWCURSOR_ENABLE; + fb_info.cursor.disable=reg.vidcfg; + reg.miscinit0 = tdfx_inl(MISCINIT0); #if defined(__BIG_ENDIAN) @@ -1496,11 +1513,6 @@ return -EINVAL; } - if((var->vmode & FB_VMODE_MASK) == FB_VMODE_INTERLACED) { - DPRINTK("interlace not supported\n"); - return -EINVAL; - } - if(var->xoffset) { DPRINTK("xoffset not supported\n"); return -EINVAL; @@ -1516,9 +1528,10 @@ return -EINVAL; } - /* fixme: does Voodoo3 support interlace? Banshee doesn't */ - if((var->vmode & FB_VMODE_MASK) == FB_VMODE_INTERLACED) { - DPRINTK("interlace not supported\n"); + /* Banshee doesn't support interlace, but Voodoo4 and probably Voodoo3 do. */ + if(((var->vmode & FB_VMODE_MASK) == FB_VMODE_INTERLACED) + && (i->dev == PCI_DEVICE_ID_3DFX_BANSHEE)) { + DPRINTK("interlace not supported on Banshee\n"); return -EINVAL; } @@ -1578,6 +1591,8 @@ par->video |= TDFXF_VSYNC_ACT_LOW; if((var->vmode & FB_VMODE_MASK) == FB_VMODE_DOUBLE) par->video |= TDFXF_LINE_DOUBLE; + else if((var->vmode & FB_VMODE_MASK) == FB_VMODE_INTERLACED) + par->video |= TDFXF_INTERLACE; if(var->activate == FB_ACTIVATE_NOW) par->video |= TDFXF_VIDEO_ENABLE; } @@ -1639,6 +1654,8 @@ v.sync |= FB_SYNC_VERT_HIGH_ACT; if(par->video & TDFXF_LINE_DOUBLE) v.vmode = FB_VMODE_DOUBLE; + else if(par->video & TDFXF_INTERLACE) + v.vmode = FB_VMODE_INTERLACED; *var = v; return 0; } @@ -1672,7 +1689,7 @@ fix->line_length = par->lpitch; fix->visual = (par->bpp == 8) ? FB_VISUAL_PSEUDOCOLOR - : FB_VISUAL_DIRECTCOLOR; + : FB_VISUAL_TRUECOLOR; fix->xpanstep = 0; fix->ypanstep = nopan ? 0 : 1; @@ -2386,7 +2403,9 @@ static void tdfxfb_hwcursor_init(void) { unsigned int start; - start = (fb_info.bufbase_size-1024) & PAGE_MASK; + start = (fb_info.bufbase_size-1024) & (PAGE_MASK << 1); + /* even page boundary - on Voodoo4 4500 bottom 48 lines + * contained trash when just page boundary was used... */ fb_info.bufbase_size=start; fb_info.cursor.cursorimage=fb_info.bufbase_size; printk("tdfxfb: reserving 1024 bytes for the hwcursor at %p\n", --tsOsTdHNUZQcU9Ye-- ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf