[PATCH] mach64: detect the dot clock divider correctly on sparc

STALE2891d

8 messages, 4 authors, 2018-10-08 · open the first message on its own page

[PATCH] mach64: detect the dot clock divider correctly on sparc

From: Mikulas Patocka <mpatocka@redhat.com>
Date: 2018-08-17 19:19:37

On Sun Ultra 5, it happens that the dot clock is not set up properly for
some videomodes. For example, if we set the videomode "r1024x768x60" in
the firmware, Linux would incorrectly set a videomode with refresh rate
180Hz when booting (suprisingly, my LCD monitor can display it, although
display quality is very low).

The reason is this: Older mach64 cards set the divider in the register
VCLK_POST_DIV. The register has four 2-bit fields (the field that is
actually used is specified in the lowest two bits of the register
CLOCK_CNTL). The 2 bits select divider "1, 2, 4, 8". On newer mach64 cards,
there's another bit added - the top four bits of PLL_EXT_CNTL extend the
divider selection, so we have possible dividers "1, 2, 4, 8, 3, 5, 6, 12".
The Linux driver clears the top four bits of PLL_EXT_CNTL and never sets
them, so it can work regardless if the card supports them. However, the
sparc64 firmware may set these extended dividers during boot - and the
mach64 driver detects incorrect dot clock in this case.

This patch makes the driver read the additional divider bit from
PLL_EXT_CNTL and calculate the initial refresh rate properly.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@vger.kernel.org

---
 drivers/video/fbdev/aty/atyfb.h      |    3 ++-
 drivers/video/fbdev/aty/atyfb_base.c |    7 ++++---
 drivers/video/fbdev/aty/mach64_ct.c  |   10 +++++-----
 3 files changed, 11 insertions(+), 9 deletions(-)

Index: linux-stable/drivers/video/fbdev/aty/atyfb.h
=================================--- linux-stable.orig/drivers/video/fbdev/aty/atyfb.h	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/atyfb.h	2018-08-13 21:17:14.000000000 +0200
@@ -333,6 +333,8 @@ extern const struct aty_pll_ops aty_pll_
 extern void aty_set_pll_ct(const struct fb_info *info, const union aty_pll *pll);
 extern u8 aty_ld_pll_ct(int offset, const struct atyfb_par *par);
 
+extern const u8 aty_postdividers[8];
+
 
     /*
      *  Hardware cursor support
@@ -359,7 +361,6 @@ static inline void wait_for_idle(struct
 
 extern void aty_reset_engine(const struct atyfb_par *par);
 extern void aty_init_engine(struct atyfb_par *par, struct fb_info *info);
-extern u8   aty_ld_pll_ct(int offset, const struct atyfb_par *par);
 
 void atyfb_copyarea(struct fb_info *info, const struct fb_copyarea *area);
 void atyfb_fillrect(struct fb_info *info, const struct fb_fillrect *rect);
Index: linux-stable/drivers/video/fbdev/aty/atyfb_base.c
=================================--- linux-stable.orig/drivers/video/fbdev/aty/atyfb_base.c	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/atyfb_base.c	2018-08-13 21:22:23.000000000 +0200
@@ -3087,17 +3087,18 @@ static int atyfb_setup_sparc(struct pci_
 		/*
 		 * PLL Reference Divider M:
 		 */
-		M = pll_regs[2];
+		M = pll_regs[PLL_REF_DIV];
 
 		/*
 		 * PLL Feedback Divider N (Dependent on CLOCK_CNTL):
 		 */
-		N = pll_regs[7 + (clock_cntl & 3)];
+		N = pll_regs[VCLK0_FB_DIV + (clock_cntl & 3)];
 
 		/*
 		 * PLL Post Divider P (Dependent on CLOCK_CNTL):
 		 */
-		P = 1 << (pll_regs[6] >> ((clock_cntl & 3) << 1));
+		P = aty_postdividers[((pll_regs[VCLK_POST_DIV] >> ((clock_cntl & 3) << 1)) & 3) |
+		                     ((pll_regs[PLL_EXT_CNTL] >> (2 + (clock_cntl & 3))) & 4)];
 
 		/*
 		 * PLL Divider Q:
Index: linux-stable/drivers/video/fbdev/aty/mach64_ct.c
=================================--- linux-stable.orig/drivers/video/fbdev/aty/mach64_ct.c	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/mach64_ct.c	2018-08-13 21:35:32.000000000 +0200
@@ -115,7 +115,7 @@ static void aty_st_pll_ct(int offset, u8
  */
 
 #define Maximum_DSP_PRECISION 7
-static u8 postdividers[] = {1,2,4,8,3};
+const u8 aty_postdividers[8] = {1,2,4,8,3,5,6,12};
 
 static int aty_dsp_gt(const struct fb_info *info, u32 bpp, struct pll_ct *pll)
 {
@@ -222,7 +222,7 @@ static int aty_valid_pll_ct(const struct
 		pll->vclk_post_div += (q <  64*8);
 		pll->vclk_post_div += (q <  32*8);
 	}
-	pll->vclk_post_div_real = postdividers[pll->vclk_post_div];
+	pll->vclk_post_div_real = aty_postdividers[pll->vclk_post_div];
 	//    pll->vclk_post_div <<= 6;
 	pll->vclk_fb_div = q * pll->vclk_post_div_real / 8;
 	pllvclk = (1000000 * 2 * pll->vclk_fb_div) /
@@ -513,7 +513,7 @@ static int aty_init_pll_ct(const struct
 		u8 mclk_fb_div, pll_ext_cntl;
 		pll->ct.pll_ref_div = aty_ld_pll_ct(PLL_REF_DIV, par);
 		pll_ext_cntl = aty_ld_pll_ct(PLL_EXT_CNTL, par);
-		pll->ct.xclk_post_div_real = postdividers[pll_ext_cntl & 0x07];
+		pll->ct.xclk_post_div_real = aty_postdividers[pll_ext_cntl & 0x07];
 		mclk_fb_div = aty_ld_pll_ct(MCLK_FB_DIV, par);
 		if (pll_ext_cntl & PLL_MFB_TIMES_4_2B)
 			mclk_fb_div <<= 1;
@@ -535,7 +535,7 @@ static int aty_init_pll_ct(const struct
 		xpost_div += (q <  64*8);
 		xpost_div += (q <  32*8);
 	}
-	pll->ct.xclk_post_div_real = postdividers[xpost_div];
+	pll->ct.xclk_post_div_real = aty_postdividers[xpost_div];
 	pll->ct.mclk_fb_div = q * pll->ct.xclk_post_div_real / 8;
 
 #ifdef CONFIG_PPC
@@ -584,7 +584,7 @@ static int aty_init_pll_ct(const struct
 			mpost_div += (q <  64*8);
 			mpost_div += (q <  32*8);
 		}
-		sclk_post_div_real = postdividers[mpost_div];
+		sclk_post_div_real = aty_postdividers[mpost_div];
 		pll->ct.sclk_fb_div = q * sclk_post_div_real / 8;
 		pll->ct.spll_cntl2 = mpost_div << 4;
 #ifdef DEBUG

Re: [PATCH] mach64: detect the dot clock divider correctly on sparc

From: David Miller <davem@davemloft.net>
Date: 2018-08-17 19:35:45

From: Mikulas Patocka <mpatocka@redhat.com>
Date: Fri, 17 Aug 2018 15:19:37 -0400 (EDT)
On Sun Ultra 5, it happens that the dot clock is not set up properly for
some videomodes. For example, if we set the videomode "r1024x768x60" in
the firmware, Linux would incorrectly set a videomode with refresh rate
180Hz when booting (suprisingly, my LCD monitor can display it, although
display quality is very low).

The reason is this: Older mach64 cards set the divider in the register
VCLK_POST_DIV. The register has four 2-bit fields (the field that is
actually used is specified in the lowest two bits of the register
CLOCK_CNTL). The 2 bits select divider "1, 2, 4, 8". On newer mach64 cards,
there's another bit added - the top four bits of PLL_EXT_CNTL extend the
divider selection, so we have possible dividers "1, 2, 4, 8, 3, 5, 6, 12".
The Linux driver clears the top four bits of PLL_EXT_CNTL and never sets
them, so it can work regardless if the card supports them. However, the
sparc64 firmware may set these extended dividers during boot - and the
mach64 driver detects incorrect dot clock in this case.

This patch makes the driver read the additional divider bit from
PLL_EXT_CNTL and calculate the initial refresh rate properly.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@vger.kernel.org
Acked-by: David S. Miller <davem@davemloft.net>

Re: [PATCH] mach64: detect the dot clock divider correctly on sparc

From: Ville Syrjälä <syrjala@sci.fi>
Date: 2018-08-19 10:37:09

On Fri, Aug 17, 2018 at 03:19:37PM -0400, Mikulas Patocka wrote:
quoted hunk
On Sun Ultra 5, it happens that the dot clock is not set up properly for
some videomodes. For example, if we set the videomode "r1024x768x60" in
the firmware, Linux would incorrectly set a videomode with refresh rate
180Hz when booting (suprisingly, my LCD monitor can display it, although
display quality is very low).

The reason is this: Older mach64 cards set the divider in the register
VCLK_POST_DIV. The register has four 2-bit fields (the field that is
actually used is specified in the lowest two bits of the register
CLOCK_CNTL). The 2 bits select divider "1, 2, 4, 8". On newer mach64 cards,
there's another bit added - the top four bits of PLL_EXT_CNTL extend the
divider selection, so we have possible dividers "1, 2, 4, 8, 3, 5, 6, 12".
The Linux driver clears the top four bits of PLL_EXT_CNTL and never sets
them, so it can work regardless if the card supports them. However, the
sparc64 firmware may set these extended dividers during boot - and the
mach64 driver detects incorrect dot clock in this case.

This patch makes the driver read the additional divider bit from
PLL_EXT_CNTL and calculate the initial refresh rate properly.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@vger.kernel.org

---
 drivers/video/fbdev/aty/atyfb.h      |    3 ++-
 drivers/video/fbdev/aty/atyfb_base.c |    7 ++++---
 drivers/video/fbdev/aty/mach64_ct.c  |   10 +++++-----
 3 files changed, 11 insertions(+), 9 deletions(-)

Index: linux-stable/drivers/video/fbdev/aty/atyfb.h
=================================> --- linux-stable.orig/drivers/video/fbdev/aty/atyfb.h	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/atyfb.h	2018-08-13 21:17:14.000000000 +0200
@@ -333,6 +333,8 @@ extern const struct aty_pll_ops aty_pll_
 extern void aty_set_pll_ct(const struct fb_info *info, const union aty_pll *pll);
 extern u8 aty_ld_pll_ct(int offset, const struct atyfb_par *par);
 
+extern const u8 aty_postdividers[8];
+
 
     /*
      *  Hardware cursor support
@@ -359,7 +361,6 @@ static inline void wait_for_idle(struct
 
 extern void aty_reset_engine(const struct atyfb_par *par);
 extern void aty_init_engine(struct atyfb_par *par, struct fb_info *info);
-extern u8   aty_ld_pll_ct(int offset, const struct atyfb_par *par);
 
 void atyfb_copyarea(struct fb_info *info, const struct fb_copyarea *area);
 void atyfb_fillrect(struct fb_info *info, const struct fb_fillrect *rect);
Index: linux-stable/drivers/video/fbdev/aty/atyfb_base.c
=================================> --- linux-stable.orig/drivers/video/fbdev/aty/atyfb_base.c	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/atyfb_base.c	2018-08-13 21:22:23.000000000 +0200
@@ -3087,17 +3087,18 @@ static int atyfb_setup_sparc(struct pci_
 		/*
 		 * PLL Reference Divider M:
 		 */
-		M = pll_regs[2];
+		M = pll_regs[PLL_REF_DIV];
 
 		/*
 		 * PLL Feedback Divider N (Dependent on CLOCK_CNTL):
 		 */
-		N = pll_regs[7 + (clock_cntl & 3)];
+		N = pll_regs[VCLK0_FB_DIV + (clock_cntl & 3)];
 
 		/*
 		 * PLL Post Divider P (Dependent on CLOCK_CNTL):
 		 */
-		P = 1 << (pll_regs[6] >> ((clock_cntl & 3) << 1));
+		P = aty_postdividers[((pll_regs[VCLK_POST_DIV] >> ((clock_cntl & 3) << 1)) & 3) |
+		                     ((pll_regs[PLL_EXT_CNTL] >> (2 + (clock_cntl & 3))) & 4)];
 
 		/*
 		 * PLL Divider Q:
Index: linux-stable/drivers/video/fbdev/aty/mach64_ct.c
=================================> --- linux-stable.orig/drivers/video/fbdev/aty/mach64_ct.c	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/mach64_ct.c	2018-08-13 21:35:32.000000000 +0200
@@ -115,7 +115,7 @@ static void aty_st_pll_ct(int offset, u8
  */
 
 #define Maximum_DSP_PRECISION 7
-static u8 postdividers[] = {1,2,4,8,3};
+const u8 aty_postdividers[8] = {1,2,4,8,3,5,6,12};
Not sure about the '5'. That setting is marked as reserved in my docs.
So I guess it doesn't really matter what number we put there.

Reviewed-by: Ville Syrjälä <syrjala@sci.fi>
quoted hunk
 
 static int aty_dsp_gt(const struct fb_info *info, u32 bpp, struct pll_ct *pll)
 {
@@ -222,7 +222,7 @@ static int aty_valid_pll_ct(const struct
 		pll->vclk_post_div += (q <  64*8);
 		pll->vclk_post_div += (q <  32*8);
 	}
-	pll->vclk_post_div_real = postdividers[pll->vclk_post_div];
+	pll->vclk_post_div_real = aty_postdividers[pll->vclk_post_div];
 	//    pll->vclk_post_div <<= 6;
 	pll->vclk_fb_div = q * pll->vclk_post_div_real / 8;
 	pllvclk = (1000000 * 2 * pll->vclk_fb_div) /
@@ -513,7 +513,7 @@ static int aty_init_pll_ct(const struct
 		u8 mclk_fb_div, pll_ext_cntl;
 		pll->ct.pll_ref_div = aty_ld_pll_ct(PLL_REF_DIV, par);
 		pll_ext_cntl = aty_ld_pll_ct(PLL_EXT_CNTL, par);
-		pll->ct.xclk_post_div_real = postdividers[pll_ext_cntl & 0x07];
+		pll->ct.xclk_post_div_real = aty_postdividers[pll_ext_cntl & 0x07];
 		mclk_fb_div = aty_ld_pll_ct(MCLK_FB_DIV, par);
 		if (pll_ext_cntl & PLL_MFB_TIMES_4_2B)
 			mclk_fb_div <<= 1;
@@ -535,7 +535,7 @@ static int aty_init_pll_ct(const struct
 		xpost_div += (q <  64*8);
 		xpost_div += (q <  32*8);
 	}
-	pll->ct.xclk_post_div_real = postdividers[xpost_div];
+	pll->ct.xclk_post_div_real = aty_postdividers[xpost_div];
 	pll->ct.mclk_fb_div = q * pll->ct.xclk_post_div_real / 8;
 
 #ifdef CONFIG_PPC
@@ -584,7 +584,7 @@ static int aty_init_pll_ct(const struct
 			mpost_div += (q <  64*8);
 			mpost_div += (q <  32*8);
 		}
-		sclk_post_div_real = postdividers[mpost_div];
+		sclk_post_div_real = aty_postdividers[mpost_div];
 		pll->ct.sclk_fb_div = q * sclk_post_div_real / 8;
 		pll->ct.spll_cntl2 = mpost_div << 4;
 #ifdef DEBUG
-- 
Ville Syrjälä
syrjala@sci.fi
http://www.sci.fi/~syrjala/

Re: [PATCH] mach64: detect the dot clock divider correctly on sparc

From: Mikulas Patocka <mpatocka@redhat.com>
Date: 2018-08-23 21:17:24


On Sun, 19 Aug 2018, Ville Syrjälä wrote:
On Fri, Aug 17, 2018 at 03:19:37PM -0400, Mikulas Patocka wrote:
quoted
On Sun Ultra 5, it happens that the dot clock is not set up properly for
some videomodes. For example, if we set the videomode "r1024x768x60" in
the firmware, Linux would incorrectly set a videomode with refresh rate
180Hz when booting (suprisingly, my LCD monitor can display it, although
display quality is very low).

The reason is this: Older mach64 cards set the divider in the register
VCLK_POST_DIV. The register has four 2-bit fields (the field that is
actually used is specified in the lowest two bits of the register
CLOCK_CNTL). The 2 bits select divider "1, 2, 4, 8". On newer mach64 cards,
there's another bit added - the top four bits of PLL_EXT_CNTL extend the
divider selection, so we have possible dividers "1, 2, 4, 8, 3, 5, 6, 12".
The Linux driver clears the top four bits of PLL_EXT_CNTL and never sets
them, so it can work regardless if the card supports them. However, the
sparc64 firmware may set these extended dividers during boot - and the
mach64 driver detects incorrect dot clock in this case.

This patch makes the driver read the additional divider bit from
PLL_EXT_CNTL and calculate the initial refresh rate properly.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@vger.kernel.org

---
 drivers/video/fbdev/aty/atyfb.h      |    3 ++-
 drivers/video/fbdev/aty/atyfb_base.c |    7 ++++---
 drivers/video/fbdev/aty/mach64_ct.c  |   10 +++++-----
 3 files changed, 11 insertions(+), 9 deletions(-)

Index: linux-stable/drivers/video/fbdev/aty/atyfb.h
===================================================================
--- linux-stable.orig/drivers/video/fbdev/aty/atyfb.h	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/atyfb.h	2018-08-13 21:17:14.000000000 +0200
@@ -333,6 +333,8 @@ extern const struct aty_pll_ops aty_pll_
 extern void aty_set_pll_ct(const struct fb_info *info, const union aty_pll *pll);
 extern u8 aty_ld_pll_ct(int offset, const struct atyfb_par *par);
 
+extern const u8 aty_postdividers[8];
+
 
     /*
      *  Hardware cursor support
@@ -359,7 +361,6 @@ static inline void wait_for_idle(struct
 
 extern void aty_reset_engine(const struct atyfb_par *par);
 extern void aty_init_engine(struct atyfb_par *par, struct fb_info *info);
-extern u8   aty_ld_pll_ct(int offset, const struct atyfb_par *par);
 
 void atyfb_copyarea(struct fb_info *info, const struct fb_copyarea *area);
 void atyfb_fillrect(struct fb_info *info, const struct fb_fillrect *rect);
Index: linux-stable/drivers/video/fbdev/aty/atyfb_base.c
===================================================================
--- linux-stable.orig/drivers/video/fbdev/aty/atyfb_base.c	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/atyfb_base.c	2018-08-13 21:22:23.000000000 +0200
@@ -3087,17 +3087,18 @@ static int atyfb_setup_sparc(struct pci_
 		/*
 		 * PLL Reference Divider M:
 		 */
-		M = pll_regs[2];
+		M = pll_regs[PLL_REF_DIV];
 
 		/*
 		 * PLL Feedback Divider N (Dependent on CLOCK_CNTL):
 		 */
-		N = pll_regs[7 + (clock_cntl & 3)];
+		N = pll_regs[VCLK0_FB_DIV + (clock_cntl & 3)];
 
 		/*
 		 * PLL Post Divider P (Dependent on CLOCK_CNTL):
 		 */
-		P = 1 << (pll_regs[6] >> ((clock_cntl & 3) << 1));
+		P = aty_postdividers[((pll_regs[VCLK_POST_DIV] >> ((clock_cntl & 3) << 1)) & 3) |
+		                     ((pll_regs[PLL_EXT_CNTL] >> (2 + (clock_cntl & 3))) & 4)];
 
 		/*
 		 * PLL Divider Q:
Index: linux-stable/drivers/video/fbdev/aty/mach64_ct.c
===================================================================
--- linux-stable.orig/drivers/video/fbdev/aty/mach64_ct.c	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/mach64_ct.c	2018-08-13 21:35:32.000000000 +0200
@@ -115,7 +115,7 @@ static void aty_st_pll_ct(int offset, u8
  */
 
 #define Maximum_DSP_PRECISION 7
-static u8 postdividers[] = {1,2,4,8,3};
+const u8 aty_postdividers[8] = {1,2,4,8,3,5,6,12};
Not sure about the '5'. That setting is marked as reserved in my docs.
So I guess it doesn't really matter what number we put there.
I downloaded documentation here: 
http://ftp.nluug.nl/ftp/pub/NetBSD/misc/cegger/hw_manuals/ati/mach64_and_ragepro/

The RAGE XL/XC Register Reference Guide - at page 4-77 it says "select 
alternate post dividers", but doesn't say what they are.
At page 4-89 it lists post dividers 1,2,4,8,3,5,6,12. It's hard to say if 
they also apply to the PLL_EXT_CNTL register or not.

At lest the divider 3 is valid because the Sparc firmware sets it and it 
works as expected.

Mikulas
Reviewed-by: Ville Syrjälä <syrjala@sci.fi>

Re: [PATCH] mach64: detect the dot clock divider correctly on sparc

From: Ville Syrjälä <syrjala@sci.fi>
Date: 2018-08-24 11:59:07

On Thu, Aug 23, 2018 at 05:17:20PM -0400, Mikulas Patocka wrote:

On Sun, 19 Aug 2018, Ville Syrjälä wrote:
quoted
On Fri, Aug 17, 2018 at 03:19:37PM -0400, Mikulas Patocka wrote:
quoted
On Sun Ultra 5, it happens that the dot clock is not set up properly for
some videomodes. For example, if we set the videomode "r1024x768x60" in
the firmware, Linux would incorrectly set a videomode with refresh rate
180Hz when booting (suprisingly, my LCD monitor can display it, although
display quality is very low).

The reason is this: Older mach64 cards set the divider in the register
VCLK_POST_DIV. The register has four 2-bit fields (the field that is
actually used is specified in the lowest two bits of the register
CLOCK_CNTL). The 2 bits select divider "1, 2, 4, 8". On newer mach64 cards,
there's another bit added - the top four bits of PLL_EXT_CNTL extend the
divider selection, so we have possible dividers "1, 2, 4, 8, 3, 5, 6, 12".
The Linux driver clears the top four bits of PLL_EXT_CNTL and never sets
them, so it can work regardless if the card supports them. However, the
sparc64 firmware may set these extended dividers during boot - and the
mach64 driver detects incorrect dot clock in this case.

This patch makes the driver read the additional divider bit from
PLL_EXT_CNTL and calculate the initial refresh rate properly.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@vger.kernel.org

---
 drivers/video/fbdev/aty/atyfb.h      |    3 ++-
 drivers/video/fbdev/aty/atyfb_base.c |    7 ++++---
 drivers/video/fbdev/aty/mach64_ct.c  |   10 +++++-----
 3 files changed, 11 insertions(+), 9 deletions(-)

Index: linux-stable/drivers/video/fbdev/aty/atyfb.h
=================================> > > --- linux-stable.orig/drivers/video/fbdev/aty/atyfb.h	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/atyfb.h	2018-08-13 21:17:14.000000000 +0200
@@ -333,6 +333,8 @@ extern const struct aty_pll_ops aty_pll_
 extern void aty_set_pll_ct(const struct fb_info *info, const union aty_pll *pll);
 extern u8 aty_ld_pll_ct(int offset, const struct atyfb_par *par);
 
+extern const u8 aty_postdividers[8];
+
 
     /*
      *  Hardware cursor support
@@ -359,7 +361,6 @@ static inline void wait_for_idle(struct
 
 extern void aty_reset_engine(const struct atyfb_par *par);
 extern void aty_init_engine(struct atyfb_par *par, struct fb_info *info);
-extern u8   aty_ld_pll_ct(int offset, const struct atyfb_par *par);
 
 void atyfb_copyarea(struct fb_info *info, const struct fb_copyarea *area);
 void atyfb_fillrect(struct fb_info *info, const struct fb_fillrect *rect);
Index: linux-stable/drivers/video/fbdev/aty/atyfb_base.c
=================================> > > --- linux-stable.orig/drivers/video/fbdev/aty/atyfb_base.c	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/atyfb_base.c	2018-08-13 21:22:23.000000000 +0200
@@ -3087,17 +3087,18 @@ static int atyfb_setup_sparc(struct pci_
 		/*
 		 * PLL Reference Divider M:
 		 */
-		M = pll_regs[2];
+		M = pll_regs[PLL_REF_DIV];
 
 		/*
 		 * PLL Feedback Divider N (Dependent on CLOCK_CNTL):
 		 */
-		N = pll_regs[7 + (clock_cntl & 3)];
+		N = pll_regs[VCLK0_FB_DIV + (clock_cntl & 3)];
 
 		/*
 		 * PLL Post Divider P (Dependent on CLOCK_CNTL):
 		 */
-		P = 1 << (pll_regs[6] >> ((clock_cntl & 3) << 1));
+		P = aty_postdividers[((pll_regs[VCLK_POST_DIV] >> ((clock_cntl & 3) << 1)) & 3) |
+		                     ((pll_regs[PLL_EXT_CNTL] >> (2 + (clock_cntl & 3))) & 4)];
 
 		/*
 		 * PLL Divider Q:
Index: linux-stable/drivers/video/fbdev/aty/mach64_ct.c
=================================> > > --- linux-stable.orig/drivers/video/fbdev/aty/mach64_ct.c	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/mach64_ct.c	2018-08-13 21:35:32.000000000 +0200
@@ -115,7 +115,7 @@ static void aty_st_pll_ct(int offset, u8
  */
 
 #define Maximum_DSP_PRECISION 7
-static u8 postdividers[] = {1,2,4,8,3};
+const u8 aty_postdividers[8] = {1,2,4,8,3,5,6,12};
Not sure about the '5'. That setting is marked as reserved in my docs.
So I guess it doesn't really matter what number we put there.
I downloaded documentation here: 
http://ftp.nluug.nl/ftp/pub/NetBSD/misc/cegger/hw_manuals/ati/mach64_and_ragepro/

The RAGE XL/XC Register Reference Guide - at page 4-77 it says "select 
alternate post dividers", but doesn't say what they are.
At page 4-89 it lists post dividers 1,2,4,8,3,5,6,12. It's hard to say if 
they also apply to the PLL_EXT_CNTL register or not.
Oh, I missed the V2CLK reg. Indeed I see the /5 there in my docs as
well.

For the other clocks the Rage II docs I have have the following table:
ALT_VCLKx_POST VCLKx_POST Post Divider
0 00 /1
0 01 /2
0 10 /4
0 11 /8
1 00 /3
1 01 reserved
1 10 /6
1 11 /12

Curiously that table is nowhere to be found in the docs for any of the
more recent chips.

But yeah, given the V2CLK thing /5 seems like a perfectly good 
value to use here.

-- 
Ville Syrjälä
syrjala@sci.fi
http://www.sci.fi/~syrjala/

Re: [PATCH] mach64: detect the dot clock divider correctly on sparc

From: David Miller <davem@davemloft.net>
Date: 2018-10-08 06:06:35

From: Ville Syrjälä <syrjala@sci.fi>
Date: Fri, 24 Aug 2018 14:58:52 +0300
But yeah, given the V2CLK thing /5 seems like a perfectly good 
value to use here.
Patch applied.

Re: [PATCH] mach64: detect the dot clock divider correctly on sparc

From: Bartlomiej Zolnierkiewicz <hidden>
Date: 2018-10-08 10:27:13

On 08/19/2018 12:36 PM, Ville Syrj채l채 wrote:
On Fri, Aug 17, 2018 at 03:19:37PM -0400, Mikulas Patocka wrote:
quoted
On Sun Ultra 5, it happens that the dot clock is not set up properly for
some videomodes. For example, if we set the videomode "r1024x768x60" in
the firmware, Linux would incorrectly set a videomode with refresh rate
180Hz when booting (suprisingly, my LCD monitor can display it, although
display quality is very low).

The reason is this: Older mach64 cards set the divider in the register
VCLK_POST_DIV. The register has four 2-bit fields (the field that is
actually used is specified in the lowest two bits of the register
CLOCK_CNTL). The 2 bits select divider "1, 2, 4, 8". On newer mach64 cards,
there's another bit added - the top four bits of PLL_EXT_CNTL extend the
divider selection, so we have possible dividers "1, 2, 4, 8, 3, 5, 6, 12".
The Linux driver clears the top four bits of PLL_EXT_CNTL and never sets
them, so it can work regardless if the card supports them. However, the
sparc64 firmware may set these extended dividers during boot - and the
mach64 driver detects incorrect dot clock in this case.

This patch makes the driver read the additional divider bit from
PLL_EXT_CNTL and calculate the initial refresh rate properly.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@vger.kernel.org

---
 drivers/video/fbdev/aty/atyfb.h      |    3 ++-
 drivers/video/fbdev/aty/atyfb_base.c |    7 ++++---
 drivers/video/fbdev/aty/mach64_ct.c  |   10 +++++-----
 3 files changed, 11 insertions(+), 9 deletions(-)

Index: linux-stable/drivers/video/fbdev/aty/atyfb.h
=================================>> --- linux-stable.orig/drivers/video/fbdev/aty/atyfb.h	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/atyfb.h	2018-08-13 21:17:14.000000000 +0200
@@ -333,6 +333,8 @@ extern const struct aty_pll_ops aty_pll_
 extern void aty_set_pll_ct(const struct fb_info *info, const union aty_pll *pll);
 extern u8 aty_ld_pll_ct(int offset, const struct atyfb_par *par);
 
+extern const u8 aty_postdividers[8];
+
 
     /*
      *  Hardware cursor support
@@ -359,7 +361,6 @@ static inline void wait_for_idle(struct
 
 extern void aty_reset_engine(const struct atyfb_par *par);
 extern void aty_init_engine(struct atyfb_par *par, struct fb_info *info);
-extern u8   aty_ld_pll_ct(int offset, const struct atyfb_par *par);
 
 void atyfb_copyarea(struct fb_info *info, const struct fb_copyarea *area);
 void atyfb_fillrect(struct fb_info *info, const struct fb_fillrect *rect);
Index: linux-stable/drivers/video/fbdev/aty/atyfb_base.c
=================================>> --- linux-stable.orig/drivers/video/fbdev/aty/atyfb_base.c	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/atyfb_base.c	2018-08-13 21:22:23.000000000 +0200
@@ -3087,17 +3087,18 @@ static int atyfb_setup_sparc(struct pci_
 		/*
 		 * PLL Reference Divider M:
 		 */
-		M = pll_regs[2];
+		M = pll_regs[PLL_REF_DIV];
 
 		/*
 		 * PLL Feedback Divider N (Dependent on CLOCK_CNTL):
 		 */
-		N = pll_regs[7 + (clock_cntl & 3)];
+		N = pll_regs[VCLK0_FB_DIV + (clock_cntl & 3)];
 
 		/*
 		 * PLL Post Divider P (Dependent on CLOCK_CNTL):
 		 */
-		P = 1 << (pll_regs[6] >> ((clock_cntl & 3) << 1));
+		P = aty_postdividers[((pll_regs[VCLK_POST_DIV] >> ((clock_cntl & 3) << 1)) & 3) |
+		                     ((pll_regs[PLL_EXT_CNTL] >> (2 + (clock_cntl & 3))) & 4)];
 
 		/*
 		 * PLL Divider Q:
Index: linux-stable/drivers/video/fbdev/aty/mach64_ct.c
=================================>> --- linux-stable.orig/drivers/video/fbdev/aty/mach64_ct.c	2018-08-13 21:12:11.000000000 +0200
+++ linux-stable/drivers/video/fbdev/aty/mach64_ct.c	2018-08-13 21:35:32.000000000 +0200
@@ -115,7 +115,7 @@ static void aty_st_pll_ct(int offset, u8
  */
 
 #define Maximum_DSP_PRECISION 7
-static u8 postdividers[] = {1,2,4,8,3};
+const u8 aty_postdividers[8] = {1,2,4,8,3,5,6,12};
Not sure about the '5'. That setting is marked as reserved in my docs.
So I guess it doesn't really matter what number we put there.

Reviewed-by: Ville Syrj채l채 <syrjala@sci.fi>
Patch queued for 4.20, thanks.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

Re: [PATCH] mach64: detect the dot clock divider correctly on sparc

From: Bartlomiej Zolnierkiewicz <hidden>
Date: 2018-10-08 10:31:46


On 10/08/2018 07:56 AM, David Miller wrote:
From: Ville Syrjälä <syrjala@sci.fi>
Date: Fri, 24 Aug 2018 14:58:52 +0300
quoted
But yeah, given the V2CLK thing /5 seems like a perfectly good 
value to use here.
Patch applied.
Fine with me if you also need it in sparc tree.

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help