S3 trio config

S3 trio config

[Paul Mackerras]

We have this difference between drivers/video/Config in linuxppc_2_4
compared to the official tree:

diff -urN linux-2.4.19-pre9/drivers/video/Config.in linuxppc_2_4/drivers/video/Config.in
--- linux-2.4.19-pre9/drivers/video/Config.in	Tue May 28 22:39:29 2002
+++ linuxppc_2_4/drivers/video/Config.in	Sat Jun  1 14:40:58 2002
@@ -73,7 +73,7 @@
       dep_bool '  Apple "valkyrie" display support' CONFIG_FB_VALKYRIE $CONFIG_ALL_PPC
       bool '  Chips 65550 display support' CONFIG_FB_CT65550
       bool '  IMS Twin Turbo display support' CONFIG_FB_IMSTT
-      bool '  S3 Trio display support' CONFIG_FB_S3TRIO
+      dep_bool '  S3 Trio display support' CONFIG_FB_S3TRIO $CONFIG_ALL_PPC
       tristate '  VGA 16-color graphics console' CONFIG_FB_VGA16
    fi
    if [ "$CONFIG_PARISC" = "y" ]; then

Does anyone know why we have this?  Is the S3 Trio display really only
used on PReP or CHRP machines, not on any PCs?

Paul.

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

Re: S3 trio config

[Paul Mackerras]

I wrote:

> Does anyone know why we have this?  Is the S3 Trio display really only
> used on PReP or CHRP machines, not on any PCs?

I should have said, not on any other PPC systems, of course, since
that code was in a CONFIG_PPC section.

Paul.

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

Re: S3 trio config

[Leigh Brown]

[-- Attachment #1: Type: text/plain, Size: 1307 bytes --]

On Sat, 2002-06-01 at 09:39, Geert Uytterhoeven wrote:
> On Sat, 1 Jun 2002, Paul Mackerras wrote:
> > I wrote:
> > > Does anyone know why we have this?  Is the S3 Trio display really only
> > > used on PReP or CHRP machines, not on any PCs?
> >
> > I should have said, not on any other PPC systems, of course, since
> > that code was in a CONFIG_PPC section.
>
> S3triofb relies on Open Firmware, so it can work on PowerMac/CHRP only. AFAIK
> it was used on CHRP only.

Well, S3triofb hasn't even compiled for ages.  I've attached a patch
that allows it to compile, and removes the dependency on OF.  However,
it still requires some sort of firmware to initialise the card.  I've
added a little hack so that it ignores the card if it wasn't initialised
by the firmware, since the driver doesn't have the code to initialise it
properly.

It works on the 7043-140 using a PReP kernel, and has been tested by
quite a few people.  Unfortunately, I no longer have an S3Trio card (and
to be honest, I know very little about graphics cards).

IMHO it's worth applying the patch as a working (just) driver is a
better starting point than a non-compiling one.

If there are other PPC systems that bring up the S3Trio in 640x480x8
mode then I suppose the driver will be applicable to them also....

Cheers,

Leigh.


[-- Attachment #2: S3trio.patch.final --]
[-- Type: text/x-patch, Size: 8490 bytes --]

diff -ur linuxppc_2_4_devel/drivers/video/S3triofb.c linuxppc_2_4_devel-s3/drivers/video/S3triofb.c
--- linuxppc_2_4_devel/drivers/video/S3triofb.c	Sat Jun  1 22:27:27 2002
+++ linuxppc_2_4_devel-s3/drivers/video/S3triofb.c	Sun Jun  2 10:48:37 2002
@@ -17,8 +17,7 @@
  */
 
 /*
-	Bugs : + OF dependencies should be removed.
-               + This driver should be merged with the CyberVision driver. The
+	Bugs : + This driver should be merged with the CyberVision driver. The
                  CyberVision is a Zorro III implementation of the S3Trio64 chip.
 
 */
@@ -37,9 +36,6 @@
 #include <linux/fb.h>
 #include <linux/init.h>
 #include <linux/selection.h>
-#include <asm/io.h>
-#include <asm/prom.h>
-#include <asm/pci-bridge.h>
 #include <linux/pci.h>
 #ifdef CONFIG_FB_COMPAT_XPMAC
 #include <asm/vc_ioctl.h>
@@ -61,6 +57,7 @@
 #define IO_OUT16VAL(v, r)       (((v) << 8) | (r))
 
 static int currcon = 0;
+static int disabled;
 static struct display disp;
 static struct fb_info fb_info;
 static struct { u_char red, green, blue, pad; } palette[256];
@@ -68,14 +65,20 @@
 static char *s3trio_base;
 
 static struct fb_fix_screeninfo fb_fix;
-static struct fb_var_screeninfo fb_var = { 0, };
-
+static struct fb_var_screeninfo S3triofb_default_var = {
+	640, 480, 640, 480, 0, 0, 8, 0,
+	{0, 8, 0}, {0, 8, 0}, {0, 8, 0}, {0, 0, 0},
+	0, 0, -1, FB_ACCELF_TEXT, 39722, 40, 24, 32, 11, 96, 2,
+	FB_SYNC_COMP_HIGH_ACT|FB_SYNC_VERT_HIGH_ACT,
+	FB_VMODE_NONINTERLACED
+};
+static struct fb_var_screeninfo fb_var = { 0,};
 
     /*
      *  Interface used by the world
      */
 
-static void __init s3triofb_of_init(struct device_node *dp);
+static void __init s3triofb_pci_init(struct pci_dev *dp);
 static int s3trio_get_fix(struct fb_fix_screeninfo *fix, int con,
 			  struct fb_info *info);
 static int s3trio_get_var(struct fb_var_screeninfo *var, int con,
@@ -260,16 +263,43 @@
 }
 
 
+#ifndef MODULE
+
+int __init s3triofb_setup(char *options) {
+	char *this_opt;
+
+	if (!options || !*options)
+		return 0;
+
+	while ((this_opt = strsep(&options, ",")) != NULL) {
+		if (!*this_opt)
+			continue;
+
+		if (!strcmp(this_opt, "disabled"))
+			disabled = 1;
+	}
+	return 0;
+}
+
+
+
 int __init s3triofb_init(void)
 {
-	struct device_node *dp;
+	struct pci_dev *dp = NULL;
 
-	dp = find_devices("S3Trio");
-	if (dp != 0)
-	    s3triofb_of_init(dp);
-	return 0;
+	if (disabled)
+		return -ENXIO;
+
+	dp = pci_find_device(PCI_VENDOR_ID_S3, PCI_DEVICE_ID_S3_TRIO, dp);
+	if ((dp != 0) && ((dp->class >> 16) == PCI_BASE_CLASS_DISPLAY)) {
+		s3triofb_pci_init(dp);
+		return 0;
+	} else
+		return -ENODEV;
 }
 
+#endif
+
 void __init s3trio_resetaccel(void){
 
 
@@ -313,111 +343,69 @@
 	outw( MF_PIX_CONTROL | MFA_SRC_FOREGR_MIX,  0xbee8);
 }
 
-int __init s3trio_init(struct device_node *dp){
-
-    u_char bus, dev;
-    unsigned int t32;
-    unsigned short cmd;
-
-	pci_device_loc(dp,&bus,&dev);
-                pcibios_read_config_dword(bus, dev, PCI_VENDOR_ID, &t32);
-                if(t32 == (PCI_DEVICE_ID_S3_TRIO << 16) + PCI_VENDOR_ID_S3) {
-                        pcibios_read_config_dword(bus, dev, PCI_BASE_ADDRESS_0, &t32);
-                        pcibios_read_config_dword(bus, dev, PCI_BASE_ADDRESS_1, &t32);
-			pcibios_read_config_word(bus, dev, PCI_COMMAND,&cmd);
-
-			pcibios_write_config_word(bus, dev, PCI_COMMAND, PCI_COMMAND_IO | PCI_COMMAND_MEMORY);
-
-			pcibios_write_config_dword(bus, dev, PCI_BASE_ADDRESS_0,0xffffffff);
-                        pcibios_read_config_dword(bus, dev, PCI_BASE_ADDRESS_0, &t32);
-
-/* This is a gross hack as OF only maps enough memory for the framebuffer and
-   we want to use MMIO too. We should find out which chunk of address space
-   we can use here */
-			pcibios_write_config_dword(bus,dev,PCI_BASE_ADDRESS_0,0xc6000000);
-
-			/* unlock s3 */
-
-			outb(0x01, 0x3C3);
-
-			outb(inb(0x03CC) | 1, 0x3c2);
-
-			outw(IO_OUT16VAL(0x48, 0x38),0x03D4);
-			outw(IO_OUT16VAL(0xA0, 0x39),0x03D4);
-			outb(0x33,0x3d4);
-			outw(IO_OUT16VAL((inb(0x3d5) & ~(0x2 | 0x10 |  0x40)) |
-					  0x20, 0x33), 0x3d4);
-
-			outw(IO_OUT16VAL(0x6, 0x8), 0x3c4);
-
-			/* switch to MMIO only mode */
-
-			outb(0x58, 0x3d4);
-			outw(IO_OUT16VAL(inb(0x3d5) | 3 | 0x10, 0x58), 0x3d4);
-			outw(IO_OUT16VAL(8, 0x53), 0x3d4);
-
-			/* switch off I/O accesses */
-
-#if 0
-			pcibios_write_config_word(bus, dev, PCI_COMMAND,
-				        PCI_COMMAND_IO | PCI_COMMAND_MEMORY);
-#endif
-			return 1;
-                }
+int __init s3trio_init(struct pci_dev *dp)
+{
+	/* unlock s3 */
+	outb(0x01, 0x3C3);
+	outb(inb(0x03CC) | 1, 0x3c2);
+	outw(IO_OUT16VAL(0x48, 0x38),0x03D4);
+	outw(IO_OUT16VAL(0xA0, 0x39),0x03D4);
+	outb(0x33,0x3d4);
+	outw(IO_OUT16VAL((inb(0x3d5) & ~(0x2 | 0x10 |  0x40)) |
+			  0x20, 0x33), 0x3d4);
+
+	outw(IO_OUT16VAL(0x6, 0x8), 0x3c4);
+	printk("S3trio: unlocked s3\n");
+
+	/* switch to MMIO only mode */
+	outb(0x58, 0x3d4);
+	outw(IO_OUT16VAL(inb(0x3d5) | 3 | 0x10, 0x58), 0x3d4);
+	outw(IO_OUT16VAL(8, 0x53), 0x3d4);
+	printk("S3trio: switched to mmio only mode\n");
 
-	return 0;
+	return 1;
 }
 
 
     /*
      *  Initialisation
-     *  We heavily rely on OF for the moment. This needs fixing.
      */
 
-static void __init s3triofb_of_init(struct device_node *dp)
+static void __init s3triofb_pci_init(struct pci_dev *dp)
 {
-    int i, *pp, len;
+    int i;
     unsigned long address, size;
     u_long *CursorBase;
+    u16 cmd;
 
-    strncat(s3trio_name, dp->name, sizeof(s3trio_name));
-    s3trio_name[sizeof(s3trio_name)-1] = '\0';
     strcpy(fb_fix.id, s3trio_name);
 
-    if((pp = (int *)get_property(dp, "vendor-id", &len)) != NULL
-	&& *pp!=PCI_VENDOR_ID_S3) {
-	printk("%s: can't find S3 Trio board\n", dp->full_name);
-	return;
-    }
+    fb_var = S3triofb_default_var;
+    fb_fix.line_length = fb_var.xres_virtual;
+    fb_fix.smem_len = fb_fix.line_length*fb_var.yres;
 
-    if((pp = (int *)get_property(dp, "device-id", &len)) != NULL
-	&& *pp!=PCI_DEVICE_ID_S3_TRIO) {
-	printk("%s: can't find S3 Trio board\n", dp->full_name);
-	return;
+    /* This driver cannot cope if the firmware has not initialised the
+     * device.  So, if the device isn't enabled we simply return
+     */
+    pci_read_config_word(dp, PCI_COMMAND, &cmd);
+    if (!(cmd & PCI_COMMAND_MEMORY)) {
+	    printk(KERN_NOTICE "S3trio: card was not initialised by firmware\n");
+	    return;
     }
 
-    if ((pp = (int *)get_property(dp, "depth", &len)) != NULL
-	&& len == sizeof(int) && *pp != 8) {
-	printk("%s: can't use depth = %d\n", dp->full_name, *pp);
-	return;
+    /* Enable it anyway */
+    if (pci_enable_device(dp)) {
+	    printk(KERN_ERR "S3trio: failed to enable PCI device\n");
+	    return;
     }
-    if ((pp = (int *)get_property(dp, "width", &len)) != NULL
-	&& len == sizeof(int))
-	fb_var.xres = fb_var.xres_virtual = *pp;
-    if ((pp = (int *)get_property(dp, "height", &len)) != NULL
-	&& len == sizeof(int))
-	fb_var.yres = fb_var.yres_virtual = *pp;
-    if ((pp = (int *)get_property(dp, "linebytes", &len)) != NULL
-	&& len == sizeof(int))
-	fb_fix.line_length = *pp;
-    else
-	fb_fix.line_length = fb_var.xres_virtual;
-    fb_fix.smem_len = fb_fix.line_length*fb_var.yres;
 
-    address = 0xc6000000;
-    size = 64*1024*1024;
-    if (!request_mem_region(address, size, "S3triofb"))
+    /* There is only one memory region and it covers the mmio and fb areas */
+    address = pci_resource_start(dp, 0);
+    size    = pci_resource_len(dp, 0); /* size = 64*1024*1024; */
+    if (!request_mem_region(address, size, "S3triofb")) {
+	printk("S3trio: failed to allocate memory region\n");
 	return;
+    }
 
     s3trio_init(dp);
     s3trio_base = ioremap(address, size);
@@ -552,8 +540,7 @@
 #endif
     disp.scrollmode = fb_var.accel_flags & FB_ACCELF_TEXT ? 0 : SCROLL_YREDRAW;
 
-    strcpy(fb_info.modename, "Trio64 ");
-    strncat(fb_info.modename, dp->full_name, sizeof(fb_info.modename));
+    strcpy(fb_info.modename, "Trio64");
     fb_info.node = -1;
     fb_info.fbops = &s3trio_ops;
 #if 0
@@ -591,7 +578,7 @@
 	return;
 
     printk("fb%d: S3 Trio frame buffer device on %s\n",
-	   GET_FB_IDX(fb_info.node), dp->full_name);
+	   GET_FB_IDX(fb_info.node), dp->name);
 }
 
 

Re: S3 trio config

[Geert Uytterhoeven]

On 2 Jun 2002, Leigh Brown wrote:
> On Sat, 2002-06-01 at 09:39, Geert Uytterhoeven wrote:
> > On Sat, 1 Jun 2002, Paul Mackerras wrote:
> > > I wrote:
> > > > Does anyone know why we have this?  Is the S3 Trio display really only
> > > > used on PReP or CHRP machines, not on any PCs?
> > >
> > > I should have said, not on any other PPC systems, of course, since
> > > that code was in a CONFIG_PPC section.
> >
> > S3triofb relies on Open Firmware, so it can work on PowerMac/CHRP only. AFAIK
> > it was used on CHRP only.
>
> Well, S3triofb hasn't even compiled for ages.  I've attached a patch
> that allows it to compile, and removes the dependency on OF.  However,
> it still requires some sort of firmware to initialise the card.  I've
> added a little hack so that it ignores the card if it wasn't initialised
> by the firmware, since the driver doesn't have the code to initialise it
> properly.
>
> It works on the 7043-140 using a PReP kernel, and has been tested by
> quite a few people.  Unfortunately, I no longer have an S3Trio card (and
> to be honest, I know very little about graphics cards).
>
> IMHO it's worth applying the patch as a working (just) driver is a
> better starting point than a non-compiling one.

Yes, that's better. Thanks!

> If there are other PPC systems that bring up the S3Trio in 640x480x8
> mode then I suppose the driver will be applicable to them also....

Peter: does OF initialize your board to 640x480, or is a different resolution
possible as well?

OF on my LongTrail (Peter's is newer) never did anything with a Trio64.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds


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

Re: S3 trio config

[Tom Rini]

On Sat, Jun 01, 2002 at 03:15:07PM +1000, Paul Mackerras wrote:

> We have this difference between drivers/video/Config in linuxppc_2_4
> compared to the official tree:
>
> diff -urN linux-2.4.19-pre9/drivers/video/Config.in linuxppc_2_4/drivers/video/Config.in
> --- linux-2.4.19-pre9/drivers/video/Config.in	Tue May 28 22:39:29 2002
> +++ linuxppc_2_4/drivers/video/Config.in	Sat Jun  1 14:40:58 2002
> @@ -73,7 +73,7 @@
>        dep_bool '  Apple "valkyrie" display support' CONFIG_FB_VALKYRIE $CONFIG_ALL_PPC
>        bool '  Chips 65550 display support' CONFIG_FB_CT65550
>        bool '  IMS Twin Turbo display support' CONFIG_FB_IMSTT
> -      bool '  S3 Trio display support' CONFIG_FB_S3TRIO
> +      dep_bool '  S3 Trio display support' CONFIG_FB_S3TRIO $CONFIG_ALL_PPC
>        tristate '  VGA 16-color graphics console' CONFIG_FB_VGA16
>     fi
>     if [ "$CONFIG_PARISC" = "y" ]; then
>
> Does anyone know why we have this?  Is the S3 Trio display really only
> used on PReP or CHRP machines, not on any PCs?

IIRC, when I did this the S3Triofb code was full of OF'isms, so it's
really only useful on some CHRP machines.  And just checking, it's still
got:
int __init s3triofb_init(void)
{
        struct device_node *dp;

        dp = find_devices("S3Trio");
        if (dp != 0)
            s3triofb_of_init(dp);
        return 0;
}

So it certainly won't link on !CONFIG_ALL_PPC.

But more importantly, that needs to be cleaned up a bit to be more like:
if [ "$CONFIG_PPC32" = "y" ]; then
   if [ "$CONFIG_ALL_PPC" = "y" ]; then
      ... apple drivers ...
      ... S3Trio ...
   fi
   ... other drivers ...
fi

--
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

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

Re: S3 trio config

[Tom Rini]

On Sun, Jun 02, 2002 at 11:38:01AM +0100, Leigh Brown wrote:
> On Sat, 2002-06-01 at 09:39, Geert Uytterhoeven wrote:
> > On Sat, 1 Jun 2002, Paul Mackerras wrote:
> > > I wrote:
> > > > Does anyone know why we have this?  Is the S3 Trio display really only
> > > > used on PReP or CHRP machines, not on any PCs?
> > >
> > > I should have said, not on any other PPC systems, of course, since
> > > that code was in a CONFIG_PPC section.
> >
> > S3triofb relies on Open Firmware, so it can work on PowerMac/CHRP only. AFAIK
> > it was used on CHRP only.
>
> Well, S3triofb hasn't even compiled for ages.  I've attached a patch
> that allows it to compile, and removes the dependency on OF.  However,
> it still requires some sort of firmware to initialise the card.  I've
> added a little hack so that it ignores the card if it wasn't initialised
> by the firmware, since the driver doesn't have the code to initialise it
> properly.

Okay, I've applied this patch and done a bit more work so that it
_might_ work as a module.  Geert, who should all of this goto in the end
since it doesn't seem like this driver has had an active maintainer for
a while..

--
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

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

Re: S3 trio config

[Geert Uytterhoeven]

On Tue, 4 Jun 2002, Tom Rini wrote:
> On Sun, Jun 02, 2002 at 11:38:01AM +0100, Leigh Brown wrote:
> > On Sat, 2002-06-01 at 09:39, Geert Uytterhoeven wrote:
> > > On Sat, 1 Jun 2002, Paul Mackerras wrote:
> > > > I wrote:
> > > > > Does anyone know why we have this?  Is the S3 Trio display really only
> > > > > used on PReP or CHRP machines, not on any PCs?
> > > >
> > > > I should have said, not on any other PPC systems, of course, since
> > > > that code was in a CONFIG_PPC section.
> > >
> > > S3triofb relies on Open Firmware, so it can work on PowerMac/CHRP only. AFAIK
> > > it was used on CHRP only.
> >
> > Well, S3triofb hasn't even compiled for ages.  I've attached a patch
> > that allows it to compile, and removes the dependency on OF.  However,
> > it still requires some sort of firmware to initialise the card.  I've
> > added a little hack so that it ignores the card if it wasn't initialised
> > by the firmware, since the driver doesn't have the code to initialise it
> > properly.
>
> Okay, I've applied this patch and done a bit more work so that it
> _might_ work as a module.  Geert, who should all of this goto in the end
> since it doesn't seem like this driver has had an active maintainer for
> a while..

Originally the driver was written by Peter De Schrijver, but he seems to have
been very silent last week.
Since it's PPC related, I think the `generic PPC pusher' can take care of it.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds


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

Re: S3 trio config

[Leigh Brown]

This little patch allows it to compile (non-module) for me:

--- linuxppc_2_4_devel/drivers/video/S3triofb.c	Wed Jun  5 06:01:45 2002
+++ linuxppc_2_4_devel-240/drivers/video/S3triofb.c	Wed Jun  5 10:00:01 2002
@@ -851,4 +851,4 @@
 module_init(s3triofb_init);
 MODULE_LICENSE("GPL");
 #endif
-module_exit(s3trio_exit);
+module_exit(s3triofb_exit);


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

Re: S3 trio config

[Jeff Garzik]

Here's some code people can play with, too:
http://gtf.org/garzik/kernel/files/s3fb/

There is a cyberfb attempt to get working on PCI (ppc, x86, ...) and
also an attempt to get a brand new s3fb going for PCI.  Both need
debugging :( but in theory one could replace S3triofb

    Jeff


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