Patchwork [U-Boot] ARM: MX51: PLL errata workaround

login
register
mail settings
Submitter David Jander
Date July 14, 2011, 7:11 a.m.
Message ID <1310627513-17461-1-git-send-email-david@protonic.nl>
Download mbox | patch
Permalink /patch/104636/
State Accepted
Delegated to: Stefano Babic
Headers show

Comments

David Jander - July 14, 2011, 7:11 a.m.
This is a port of the official PLL errata workaround from Freescale to
mainline u-boot.
The PLL's in the i.MX51 processor can go out of lock due to a metastable
condition in an analog flip-flop when used at high frequencies.
This workaround implements an undocumented feature in the PLL (dither
mode), which causes the effect of this failure to be much lower (in terms
of frequency deviation), avoiding system failure, or at least decreasing
the likelihood of system failure.

Signed-off-by: David Jander <david@protonic.nl>
---
 arch/arm/cpu/armv7/mx5/lowlevel_init.S   |   38 ++++++++++++++++++++++++++++++
 arch/arm/include/asm/arch-mx5/imx-regs.h |    5 ++++
 doc/README.imx5                          |   18 ++++++++++++++
 3 files changed, 61 insertions(+), 0 deletions(-)
 create mode 100644 doc/README.imx5
Stefano Babic - July 14, 2011, 8:30 a.m.
On 07/14/2011 09:11 AM, David Jander wrote:
> This is a port of the official PLL errata workaround from Freescale to
> mainline u-boot.
> The PLL's in the i.MX51 processor can go out of lock due to a metastable
> condition in an analog flip-flop when used at high frequencies.
> This workaround implements an undocumented feature in the PLL (dither
> mode), which causes the effect of this failure to be much lower (in terms
> of frequency deviation), avoiding system failure, or at least decreasing
> the likelihood of system failure.
> 
> Signed-off-by: David Jander <david@protonic.nl>

Hi David,

do you have now also an official Errata number from Freescale to be
added to your documentation ?

Best regards,
Stefano Babic
David Jander - July 14, 2011, 10:13 a.m.
Hi Stefano,

On Thu, 14 Jul 2011 10:30:11 +0200
Stefano Babic <sbabic@denx.de> wrote:

> On 07/14/2011 09:11 AM, David Jander wrote:
> > This is a port of the official PLL errata workaround from Freescale to
> > mainline u-boot.
> > The PLL's in the i.MX51 processor can go out of lock due to a metastable
> > condition in an analog flip-flop when used at high frequencies.
> > This workaround implements an undocumented feature in the PLL (dither
> > mode), which causes the effect of this failure to be much lower (in terms
> > of frequency deviation), avoiding system failure, or at least decreasing
> > the likelihood of system failure.
> > 
> > Signed-off-by: David Jander <david@protonic.nl>
> 
> Hi David,
> 
> do you have now also an official Errata number from Freescale to be
> added to your documentation ?

Freescale promised an official errata a week ago, but released the errata just
yesterday (hadn't seen it until now).
The number is ENGcm12051.
Do you mind including it in the commit yourself, or do you want me to re-post?

Best regards,
David Jander - July 14, 2011, 10:23 a.m.
On Thu, 14 Jul 2011 10:30:11 +0200
Stefano Babic <sbabic@denx.de> wrote:

> On 07/14/2011 09:11 AM, David Jander wrote:
> > This is a port of the official PLL errata workaround from Freescale to
> > mainline u-boot.
> > The PLL's in the i.MX51 processor can go out of lock due to a metastable
> > condition in an analog flip-flop when used at high frequencies.
> > This workaround implements an undocumented feature in the PLL (dither
> > mode), which causes the effect of this failure to be much lower (in terms
> > of frequency deviation), avoiding system failure, or at least decreasing
> > the likelihood of system failure.
> > 
> > Signed-off-by: David Jander <david@protonic.nl>
> 
> Hi David,
> 
> do you have now also an official Errata number from Freescale to be
> added to your documentation ?

Needless to say that this supersedes my patch sent back in May, 26th... which
btw did not help much, since _all_ PLL's are affected by this problem. Not
only PLL1.

Best regards,
Stefano Babic - July 14, 2011, 10:37 a.m.
On 07/14/2011 12:13 PM, David Jander wrote:

Hi David,

> Freescale promised an official errata a week ago, but released the errata just
> yesterday (hadn't seen it until now).
> The number is ENGcm12051.
> Do you mind including it in the commit yourself, or do you want me to re-post?

Well, it is enough - I can include it in the commit message, if there
will be no further comments. From my point of view the patch is ok. The
fact that the patch can be disabled will probably increase the number of
testers.

Acked-by : Stefano Babic <sbabic@denx.de>

Best regards,
Stefano Babic
Stefano Babic - July 14, 2011, 10:41 a.m.
On 07/14/2011 12:23 PM, David Jander wrote:
>> Hi David,
>>
>> do you have now also an official Errata number from Freescale to be
>> added to your documentation ?
> 
> Needless to say that this supersedes my patch sent back in May, 26th... which
> btw did not help much, since _all_ PLL's are affected by this problem. Not
> only PLL1.

Yes. I have set the old patch as superseeded in patchwork

Best regards,
Stefano Babic

Patch

diff --git a/arch/arm/cpu/armv7/mx5/lowlevel_init.S b/arch/arm/cpu/armv7/mx5/lowlevel_init.S
index 96ebfe2..ee4150d 100644
--- a/arch/arm/cpu/armv7/mx5/lowlevel_init.S
+++ b/arch/arm/cpu/armv7/mx5/lowlevel_init.S
@@ -117,6 +117,35 @@ 
 	beq 1b
 .endm
 
+.macro setup_pll_errata pll, freq
+	ldr r2, =\pll
+	mov r1, #0x0
+	str r1, [r2, #PLL_DP_CONFIG] /* Disable auto-restart AREN bit */
+	ldr r1, =0x00001236
+	str r1, [r2, #PLL_DP_CTL]    /* Restart PLL with PLM=1 */
+1:	ldr r1, [r2, #PLL_DP_CTL]    /* Wait for lock */
+	ands r1, r1, #0x1
+	beq 1b
+
+	ldr r5, \freq
+	str r5, [r2, #PLL_DP_MFN]    /* Modify MFN value */
+	str r5, [r2, #PLL_DP_HFS_MFN]
+
+	mov r1, #0x1
+	str r1, [r2, #PLL_DP_CONFIG] /* Reload MFN value */
+
+2:	ldr r1, [r2, #PLL_DP_CONFIG]
+	tst r1, #1
+	bne 2b
+
+	ldr r1, =100		     /* Wait at least 4 us */
+3:	subs r1, r1, #1
+	bge 3b
+
+	mov r1, #0x2
+	str r1, [r2, #PLL_DP_CONFIG] /* Enable auto-restart AREN bit */
+.endm
+
 .macro init_clock
 	ldr r0, =CCM_BASE_ADDR
 
@@ -153,7 +182,12 @@ 
 	mov r1, #0x4
 	str r1, [r0, #CLKCTL_CCSR]
 
+#if defined(CONFIG_MX51_PLL_ERRATA)
+	setup_pll PLL1_BASE_ADDR, 864
+	setup_pll_errata PLL1_BASE_ADDR, W_DP_MFN_800_DIT
+#else
 	setup_pll PLL1_BASE_ADDR, 800
+#endif
 
 #if defined(CONFIG_MX51)
 	setup_pll PLL3_BASE_ADDR, 665
@@ -283,6 +317,10 @@  lowlevel_init:
 	mov pc,lr
 
 /* Board level setting value */
+W_DP_OP_864:              .word DP_OP_864
+W_DP_MFD_864:             .word DP_MFD_864
+W_DP_MFN_864:             .word DP_MFN_864
+W_DP_MFN_800_DIT:         .word DP_MFN_800_DIT
 W_DP_OP_800:              .word DP_OP_800
 W_DP_MFD_800:             .word DP_MFD_800
 W_DP_MFN_800:             .word DP_MFN_800
diff --git a/arch/arm/include/asm/arch-mx5/imx-regs.h b/arch/arm/include/asm/arch-mx5/imx-regs.h
index d7f83c5..0b1caf0 100644
--- a/arch/arm/include/asm/arch-mx5/imx-regs.h
+++ b/arch/arm/include/asm/arch-mx5/imx-regs.h
@@ -235,6 +235,11 @@ 
 
 /* Assuming 24MHz input clock with doubler ON */
 /*                            MFI         PDF */
+#define DP_OP_864	((8 << 4) + ((1 - 1)  << 0))
+#define DP_MFD_864	(180 - 1) /* PL Dither mode */
+#define DP_MFN_864	180
+#define DP_MFN_800_DIT	60 /* PL Dither mode */
+
 #define DP_OP_850	((8 << 4) + ((1 - 1)  << 0))
 #define DP_MFD_850	(48 - 1)
 #define DP_MFN_850	41
diff --git a/doc/README.imx5 b/doc/README.imx5
new file mode 100644
index 0000000..e9b2c88
--- /dev/null
+++ b/doc/README.imx5
@@ -0,0 +1,18 @@ 
+U-Boot for Freescale i.MX5x
+
+This file contains information for the port of U-Boot to the Freescale
+i.MX5x SoCs.
+
+1. CONFIGURATION OPTIONS/SETTINGS
+---------------------------------
+
+1.1 CONFIG_MX51_PLL_ERRATA: Workaround for i.MX51 PLL errata.
+    This option should be enabled by all boards using the i.MX51 silicon
+    version up until (including) 3.0 running at 800MHz.
+    The PLL's in the i.MX51 processor can go out of lock due to a metastable
+    condition in an analog flip-flop when used at high frequencies.
+    This workaround implements an undocumented feature in the PLL (dither
+    mode), which causes the effect of this failure to be much lower (in terms
+    of frequency deviation), avoiding system failure, or at least decreasing
+    the likelihood of system failure.
+