Skip to contentSkip to conversation

ARM : omap3 : fix wrong container_of in clock36xx.c

PATCH New Jean-Philippe Francois 3 participants 10 messages May 30, 2013 – Jun 10, 2013

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 ARM : omap3 : fix wrong container_of in clock36xx.c

omap36xx_pwrdn_clk_enable_with_hsdiv_restore expects the parent hw of the clock to be a clk_hw_omap. However, looking at cclock3xxx_data.c, all concerned clock have parent defined as clk_divider. Instead of using container_of to eventually get to the register and directly mess with the divider, change freq via clk_set_rate, and let the clock framework toggle the divider value.
Tested with 3.9 on dm3730.

arch/arm/mach-omap2/clock36xx.c +12 -17
Index: b/arch/arm/mach-omap2/clock36xx.c
===================================================================
@@ -39,30 +39,25 @@
  */
 int omap36xx_pwrdn_clk_enable_with_hsdiv_restore(struct clk_hw *clk)
 {
-	struct clk_hw_omap *parent;
-	struct clk_hw *parent_hw;
-	u32 dummy_v, orig_v, clksel_shift;
 	int ret;
 
 	/* Clear PWRDN bit of HSDIVIDER */
 	ret = omap2_dflt_clk_enable(clk);
 
-	parent_hw = __clk_get_hw(__clk_get_parent(clk->clk));
-	parent = to_clk_hw_omap(parent_hw);
-
-	/* Restore the dividers */
+	/* kick parent's clksel register after toggling PWRDN bit */
 	if (!ret) {
-		clksel_shift = __ffs(parent->clksel_mask);
-		orig_v = __raw_readl(parent->clksel_reg);
-		dummy_v = orig_v;
-
-		/* Write any other value different from the Read value */
-		dummy_v ^= (1 << clksel_shift);
-		__raw_writel(dummy_v, parent->clksel_reg);
-
-		/* Write the original divider */
-		__raw_writel(orig_v, parent->clksel_reg);
+		struct clk *parent = clk_get_parent(clk->clk);
+		unsigned long parent_rate = clk_get_rate(parent);
+		ret = clk_set_rate(parent, parent_rate/2);
+		if(ret)
+			goto badfreq;
+		ret = clk_set_rate(parent, parent_rate);
+		if(ret)
+			goto badfreq;
 	}
+	return ret;
 
+ badfreq :
+	omap2_dflt_clk_disable(clk);
 	return ret;
 }
#1 lore raw

PATCH v2 1/1 ARM : omap3 : fix wrong container_of in clock36xx.c

Quoting Jean-Philippe Francois (2013-05-30 01:50:27)

Quoted 8 lines

omap36xx_pwrdn_clk_enable_with_hsdiv_restore expects the parent hw of the clock to be a clk_hw_omap. However, looking at cclock3xxx_data.c, all concerned clock have parent defined as clk_divider. Instead of using container_of to eventually get to the register and directly mess with the divider, change freq via clk_set_rate, and let the clock framework toggle the divider value.
Tested with 3.9 on dm3730.

Did you use git-format-patch to create this patch? Its a bit nicer to use that or if you just use diff then use "diff -up" or "diff -uprN" (taken from Documentation/SubmittingPatches.txt).

Also did you test this to make sure it works? I don't mean a boot test, but actually disabling/re-enabling an HSDIVIDER on 3630? The previous code just programmed the clksel field to 1, and this code divides the rate by 2, then restores it. I just used that as an example in my previous email and it needs to be verified that it works (though it should if I remember this errata correctly).

If that testing is done and everything looks good then please add:

Quoted 48 lines

Index: b/arch/arm/mach-omap2/clock36xx.c
===================================================================

arch/arm/mach-omap2/clock36xx.c +12 -17
@@ -39,30 +39,25 @@
  */
 int omap36xx_pwrdn_clk_enable_with_hsdiv_restore(struct clk_hw *clk)
 {
-       struct clk_hw_omap *parent;
-       struct clk_hw *parent_hw;
-       u32 dummy_v, orig_v, clksel_shift;
        int ret;
 
        /* Clear PWRDN bit of HSDIVIDER */
        ret = omap2_dflt_clk_enable(clk);
 
-       parent_hw = __clk_get_hw(__clk_get_parent(clk->clk));
-       parent = to_clk_hw_omap(parent_hw);
-
-       /* Restore the dividers */
+       /* kick parent's clksel register after toggling PWRDN bit */
        if (!ret) {
-               clksel_shift = __ffs(parent->clksel_mask);
-               orig_v = __raw_readl(parent->clksel_reg);
-               dummy_v = orig_v;
-
-               /* Write any other value different from the Read value */
-               dummy_v ^= (1 << clksel_shift);
-               __raw_writel(dummy_v, parent->clksel_reg);
-
-               /* Write the original divider */
-               __raw_writel(orig_v, parent->clksel_reg);
+               struct clk *parent = clk_get_parent(clk->clk);
+               unsigned long parent_rate = clk_get_rate(parent);
+               ret = clk_set_rate(parent, parent_rate/2);
+               if(ret)
+                       goto badfreq;
+               ret = clk_set_rate(parent, parent_rate);
+               if(ret)
+                       goto badfreq;
        }
+       return ret;
 
+ badfreq :
+       omap2_dflt_clk_disable(clk);
        return ret;
 }
#2 lore raw

PATCH v2 1/1 ARM : omap3 : fix wrong container_of in clock36xx.c

2013/5/31 Mike Turquette <[email protected]>:

Quoted 8 lines

Quoting Jean-Philippe Francois (2013-05-30 01:50:27)

Quoted 7 lines

omap36xx_pwrdn_clk_enable_with_hsdiv_restore expects the parent hw of the clock to be a clk_hw_omap. However, looking at cclock3xxx_data.c, all concerned clock have parent defined as clk_divider. Instead of using container_of to eventually get to the register and directly mess with the divider, change freq via clk_set_rate, and let the clock framework toggle the divider value.
Tested with 3.9 on dm3730.

Quoted 6 lines
Quoted 1 line

Did you use git-format-patch to create this patch? Its a bit nicer to use that or if you just use diff then use "diff -up" or "diff -uprN" (taken from Documentation/SubmittingPatches.txt).

It is easier for my build system to work with tarball + quilt patches, so I am not working with git. I will look into the alternative you provided.

Quoted 7 lines

Also did you test this to make sure it works? I don't mean a boot test, but actually disabling/re-enabling an HSDIVIDER on 3630? The previous code just programmed the clksel field to 1, and this code divides the rate by 2, then restores it. I just used that as an example in my previous email and it needs to be verified that it works (though it should if I remember this errata correctly).

Yes, it is exactly what happens on my board when using the camera, because the sensor clock is provided by the SoC. So this patch works for this particular clock.
If the asked frequency is the lower limit of the frequency range, then asking for half the
frequency won't change the divider, but I think it is not a problem in practice.

Quoted 4 lines

If that testing is done and everything looks good then please add:

How should I proceed ? Should I just add the acked by below, or should I resend the patch ?

Jean-Philippe Fran?ois

Quoted 48 lines
Quoted 48 lines

Index: b/arch/arm/mach-omap2/clock36xx.c
===================================================================

arch/arm/mach-omap2/clock36xx.c +12 -17
@@ -39,30 +39,25 @@
  */
 int omap36xx_pwrdn_clk_enable_with_hsdiv_restore(struct clk_hw *clk)
 {
-       struct clk_hw_omap *parent;
-       struct clk_hw *parent_hw;
-       u32 dummy_v, orig_v, clksel_shift;
        int ret;
 
        /* Clear PWRDN bit of HSDIVIDER */
        ret = omap2_dflt_clk_enable(clk);
 
-       parent_hw = __clk_get_hw(__clk_get_parent(clk->clk));
-       parent = to_clk_hw_omap(parent_hw);
-
-       /* Restore the dividers */
+       /* kick parent's clksel register after toggling PWRDN bit */
        if (!ret) {
-               clksel_shift = __ffs(parent->clksel_mask);
-               orig_v = __raw_readl(parent->clksel_reg);
-               dummy_v = orig_v;
-
-               /* Write any other value different from the Read value */
-               dummy_v ^= (1 << clksel_shift);
-               __raw_writel(dummy_v, parent->clksel_reg);
-
-               /* Write the original divider */
-               __raw_writel(orig_v, parent->clksel_reg);
+               struct clk *parent = clk_get_parent(clk->clk);
+               unsigned long parent_rate = clk_get_rate(parent);
+               ret = clk_set_rate(parent, parent_rate/2);
+               if(ret)
+                       goto badfreq;
+               ret = clk_set_rate(parent, parent_rate);
+               if(ret)
+                       goto badfreq;
        }
+       return ret;
 
+ badfreq :
+       omap2_dflt_clk_disable(clk);
        return ret;
 }
maintainer
#3 lore raw

PATCH v2 1/1 ARM : omap3 : fix wrong container_of in clock36xx.c

Quoted 2 lines · jean-philippe francois

How should I proceed ? Should I just add the acked by below, or should I resend the patch ?

It's been fixed up locally here. Mike, please speak up if I shouldn't apply your ack.

- Paul

maintainer
#4 lore raw

PATCH v2 1/1 ARM : omap3 : fix wrong container_of in clock36xx.c

In reply to Jean-Philippe Francois
Quoted 8 lines · Jean-Philippe Francois

omap36xx_pwrdn_clk_enable_with_hsdiv_restore expects the parent hw of the clock to be a clk_hw_omap. However, looking at cclock3xxx_data.c, all concerned clock have parent defined as clk_divider. Instead of using container_of to eventually get to the register and directly mess with the divider, change freq via clk_set_rate, and let the clock framework toggle the divider value.
Tested with 3.9 on dm3730.

Queued for 3.10-rc fixes with Mike's ack, after fixing the checkpatch.pl errors listed below. Please don't forget to run scripts/checkpatch.pl on future patches -- Documentation/SubmittingPatches has more info about this.

ERROR: space required before the open parenthesis '(' #59: FILE: arch/arm/mach-omap2/clock36xx.c:52:
+ if(ret)

ERROR: space required before the open parenthesis '(' #62: FILE: arch/arm/mach-omap2/clock36xx.c:55:
+ if(ret)

WARNING: please, no spaces at the start of a line
#67: FILE: arch/arm/mach-omap2/clock36xx.c:60:
+ badfreq :$

ERROR: space prohibited before that ':' (ctx:WxE)
#67: FILE: arch/arm/mach-omap2/clock36xx.c:60:
+ badfreq :

          ^

Updated patch follows.

- Paul

From: Jean-Philippe Francois <[email protected]>
Date: Thu, 30 May 2013 10:50:27 +0200
Subject: [PATCH] ARM: omap3: clock: fix wrong container_of in clock36xx.c
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

omap36xx_pwrdn_clk_enable_with_hsdiv_restore expects the parent hw of the clock to be a clk_hw_omap. However, looking at cclock3xxx_data.c, all concerned clock have parent defined as clk_divider. Instead of using container_of to eventually get to the register and directly mess with the divider, change freq via clk_set_rate, and let the clock framework toggle the divider value.
Tested with 3.9 on dm3730.

---
 arch/arm/mach-omap2/clock36xx.c |   29 ++++++++++++-----------------
 1 file changed, 12 insertions(+), 17 deletions(-)
arch/arm/mach-omap2/clock36xx.c +12 -17
index 8f3bf4e..7f69fbb 100644
@@ -39,30 +39,25 @@
  */
 int omap36xx_pwrdn_clk_enable_with_hsdiv_restore(struct clk_hw *clk)
 {
-	struct clk_hw_omap *parent;
-	struct clk_hw *parent_hw;
-	u32 dummy_v, orig_v, clksel_shift;
 	int ret;
 
 	/* Clear PWRDN bit of HSDIVIDER */
 	ret = omap2_dflt_clk_enable(clk);
 
-	parent_hw = __clk_get_hw(__clk_get_parent(clk->clk));
-	parent = to_clk_hw_omap(parent_hw);
-
-	/* Restore the dividers */
+	/* kick parent's clksel register after toggling PWRDN bit */
 	if (!ret) {
-		clksel_shift = __ffs(parent->clksel_mask);
-		orig_v = __raw_readl(parent->clksel_reg);
-		dummy_v = orig_v;
-
-		/* Write any other value different from the Read value */
-		dummy_v ^= (1 << clksel_shift);
-		__raw_writel(dummy_v, parent->clksel_reg);
-
-		/* Write the original divider */
-		__raw_writel(orig_v, parent->clksel_reg);
+		struct clk *parent = clk_get_parent(clk->clk);
+		unsigned long parent_rate = clk_get_rate(parent);
+		ret = clk_set_rate(parent, parent_rate/2);
+		if (ret)
+			goto badfreq;
+		ret = clk_set_rate(parent, parent_rate);
+		if (ret)
+			goto badfreq;
 	}
+	return ret;
 
+badfreq:
+	omap2_dflt_clk_disable(clk);
 	return ret;
 }
Signature · 1 line
1.7.10.4
maintainer
#5 lore raw

PATCH v2 1/1 ARM : omap3 : fix wrong container_of in clock36xx.c

In reply to Jean-Philippe Francois

Hi

Quoted 8 lines · Jean-Philippe Francois

omap36xx_pwrdn_clk_enable_with_hsdiv_restore expects the parent hw of the clock to be a clk_hw_omap. However, looking at cclock3xxx_data.c, all concerned clock have parent defined as clk_divider. Instead of using container_of to eventually get to the register and directly mess with the divider, change freq via clk_set_rate, and let the clock framework toggle the divider value.
Tested with 3.9 on dm3730.

Tested this patch before applying, and noticed that it causes the retention dynamic idle power management test to fail here on the 3730beaglexm:

http://www.pwsan.com/omap/testlogs/fixes_b_v3.10-rc/20130605113443/pm/3730beaglexm/3730beaglexm_log.txt

Not sure at this point if this is caused by the patch, or if the patch is just unmasking another bug.

Jean, Mike, Rajendra, care to take a look at this?

- Paul

#6 lore raw

PATCH v2 1/1 ARM : omap3 : fix wrong container_of in clock36xx.c

2013/6/5 Paul Walmsley <[email protected]>:

Quoted 17 lines

Hi

Quoted 8 lines · Jean-Philippe Francois

omap36xx_pwrdn_clk_enable_with_hsdiv_restore expects the parent hw of the clock to be a clk_hw_omap. However, looking at cclock3xxx_data.c, all concerned clock have parent defined as clk_divider. Instead of using container_of to eventually get to the register and directly mess with the divider, change freq via clk_set_rate, and let the clock framework toggle the divider value.
Tested with 3.9 on dm3730.

Tested this patch before applying, and noticed that it causes the retention dynamic idle power management test to fail here on the 3730beaglexm:

Quoted 7 lines

http://www.pwsan.com/omap/testlogs/fixes_b_v3.10-rc/20130605113443/pm/3730beaglexm/3730beaglexm_log.txt

Not sure at this point if this is caused by the patch, or if the patch is just unmasking another bug.

Jean, Mike, Rajendra, care to take a look at this?

I am booting with nohlt on my board, because I otherwise have memory corruption error.
I never looked into it because power consumption is currently a non issue for us, and the board
will always be working anyway.

However I have a beaglexm too, so I can probably help with testing.

Does the first version [1] of the patch, that only touch the MSB of the divider also trigger the
bug ?

[1] https://patchwork.kernel.org/patch/2609681/

Jean-Philippe

Quoted 6 lines

- Paul
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo at vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

maintainer
#7 lore raw

PATCH v2 1/1 ARM : omap3 : fix wrong container_of in clock36xx.c

Quoted 5 lines · jean-philippe francois

Does the first version [1] of the patch, that only touch the MSB of the divider also trigger the
bug ?

[1] https://patchwork.kernel.org/patch/2609681/

That one passes the PM test here. Will take this one for v3.10-rc fixes instead, since it fixes a known DSS problem and the original code wasn't correct.

Jean, could you work with Mike to come up with a better approach for v3.11 or v3.12?

- Paul

#8 lore raw

PATCH v2 1/1 ARM : omap3 : fix wrong container_of in clock36xx.c

2013/6/6 Paul Walmsley <[email protected]>:

Quoted 6 lines · jean-philippe francois
Quoted 5 lines

Does the first version [1] of the patch, that only touch the MSB of the divider also trigger the
bug ?

Quoted 9 lines
Quoted 1 line

[1] https://patchwork.kernel.org/patch/2609681/

That one passes the PM test here. Will take this one for v3.10-rc fixes instead, since it fixes a known DSS problem and the original code wasn't correct.

Jean, could you work with Mike to come up with a better approach for v3.11 or v3.12?

Hi,

I am ok to work on it, however could you explain to me what is the expected output of your PM tests, and how to reproduce them ? The board I used to test the clock frequency was correct suffered from heavy handed oscilloscope probing and is currently out of order :(

Jean-Philippe Fran?ois.

Quoted 6 lines

- Paul
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo at vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

maintainer
#9 lore raw

PATCH v2 1/1 ARM : omap3 : fix wrong container_of in clock36xx.c

Hello Jean-Philippe,

Quoted 2 lines · jean-philippe francois

I am ok to work on it, however could you explain to me what is the expected output of your PM tests, and how to reproduce them ?

Here is the expected output for my ancient 3730 ES1.0 Beagle-XM:

http://www.pwsan.com/omap/testlogs/test_v3.10-rc5/20130609031534/pm/3730beaglexm/3730beaglexm_log.txt

When I ran that same test on the second version of your patch, the system froze during the retention dynamic idle test:

----------------

%% Start retention dynamic idle UART wakeup test

echo 3000 > /sys/devices/platform/omap_uart.0/power/autosuspend_delay_ms root at beagleboard:~#
root at beagleboard:~# echo 3000 > /sys/devices/platform/omap_uart.1/power/autosuspend_delay_ms root at beagleboard:~#
root at beagleboard:~# echo 3000 > /sys/devices/platform/omap_uart.2/power/autosuspend_delay_ms root at beagleboard:~#
root at beagleboard:~# echo 3000 > /sys/devices/platform/omap_uart.3/power/autosuspend_delay_ms root at beagleboard:~#
root at beagleboard:~#

-------------------

That is where the log stopped.

Sorry to hear about your board - I can certainly sympathize,

regards,

- Paul