[PATCH 0/2] Meson8b: fixes for the cpu_scale_div clock

[PATCH 0/2] Meson8b: fixes for the cpu_scale_div clock

[Martin Blumenstingl]

While trying to add support for the TWD timer I found that our CPU clock
calculation (when running off the "cpu_scale_div clock) is incorrect.

The main problem was the cpu_scale_div clock:
it's divider table has an off-by-one error. the old formula was:
  parent_rate / 2 * register_value
however, testing shows that the correct formula is:
  parent_rate / 2 * (register value + 1)
See the commit message of patch #1 for a complete description of the
problem and it's history.

While looking at the "cpu_scale_div" clock I also found that it's
register width was also off-by-one, so this is fixed in a separate patch
as well.

Both patches are not critical because I haven't seen a case where u-boot
uses cpu_scale_div for the CPU clock. It's only a problem when playing
with the CPU clock in u-boot (by writing registers manually) or as soon
as we support CPU frequency scaling.


Martin Blumenstingl (2):
  clk: meson: meson8b: fix incorrect divider mapping in cpu_scale_table
  clk: meson: meson8b: fix the width of the cpu_scale_div clock

 drivers/clk/meson/meson8b.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

-- 
2.19.0

[PATCH 1/2] clk: meson: meson8b: fix incorrect divider mapping in cpu_scale_table

[Martin Blumenstingl]

The public S805 datasheet only mentions that
HHI_SYS_CPU_CLK_CNTL1[20:29] contains a divider called "cpu_scale_div".
Unfortunately it does not mention how to use the register contents.

The Amlogic 3.10 GPL kernel sources are using the following code to
calculate the CPU clock based on that register (taken from
arch/arm/mach-meson8/clock.c in the 3.10 Amlogic kernel, shortened to
make it easier to read):
N = (aml_read_reg32(P_HHI_SYS_CPU_CLK_CNTL1) >> 20) & 0x3FF;
if (sel == 3) /* use cpu_scale_div */
  div = 2 * N;
else
  div = ... /* not relevant for this example */
cpu_clk = parent_clk / div;

This suggests that the formula is: parent_rate / 2 * register_value
However, running perf (which can measure the CPU clock rate thanks to
the ARM PMU) shows that this formula is not correct.
This can be reproduced with the following steps:
1. boot into u-boot
2. let the CPU clock run off the XTAL clock:
   mw.l 0xC110419C 0x30 1
3. set the cpu_scale_div register:
   to value 0x1: mw.l 0xC110415C 0x801016A2 1
   to value 0x2: mw.l 0xC110415C 0x802016A2 1
   to value 0x5: mw.l 0xC110415C 0x805016A2 1
4. let the CPU clock run off cpu_scale_div:
   mw.l 0xC110419C 0xbd 1
5. boot Linux
6. run: perf stat -aB stress --cpu 4 --timeout 10
7. check the "cycles" value

I get the following results depending on the cpu_scale_div value:
- (cpu_in_sel - this is the input clock for cpu_scale_div - runs at
   1.2GHz)
- 0x1 = 300MHz
- 0x2 = 200MHz
- 0x5 = 100MHz

This means that the actual formula to calculate the output of the
cpu_scale_div clock is: parent_rate / 2 * (register value + 1).

The register value 0x0 is reserved. When letting the CPU clock run off
the cpu_scale_div while the value is 0x0 the whole board hangs (even in
u-boot).

I also verified this with the TWD timer: when adding this to the .dts
without specifying it's clock it will auto-detect the PERIPH (which is
the input clock of the TWD) clock rate (and the result is shown in the
kernel log). On Meson8, Meson8b and Meson8m2 the PERIPH clock is CPUCLK
divided by 4. This also matched for all three test-cases from above (in
all cases the TWD timer clock rate was approx. one fourth of the CPU
clock rate).

A small note regarding the "fixes" tag: the original issue seems to
exist virtually since forever. Even commit 28b9fcd016126e ("clk:
meson8b: Add support for Meson8b clocks") seems to handle this wrong. I
still decided to use commit 251b6fd38bcb9c ("clk: meson: rework meson8b
cpu clock") because this is the first commit which gets the CPU hiearchy
correct and thus it's the first commit where the cpu_scale_div register
is used correctly (apart from the bug in the cpu_scale_table).

Fixes: 251b6fd38bcb9c ("clk: meson: rework meson8b cpu clock")
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
---
 drivers/clk/meson/meson8b.c | 15 ++++++++-------
 1 file changed, 8 insertions(+), 7 deletions(-)

diff --git a/drivers/clk/meson/meson8b.c b/drivers/clk/meson/meson8b.c
index e7fe66f61d7b..d3f4e11d106d 100644
--- a/drivers/clk/meson/meson8b.c
+++ b/drivers/clk/meson/meson8b.c
@@ -579,13 +579,14 @@ static struct clk_fixed_factor meson8b_cpu_div3 = {
 };
 
 static const struct clk_div_table cpu_scale_table[] = {
-	{ .val = 2, .div = 4 },
-	{ .val = 3, .div = 6 },
-	{ .val = 4, .div = 8 },
-	{ .val = 5, .div = 10 },
-	{ .val = 6, .div = 12 },
-	{ .val = 7, .div = 14 },
-	{ .val = 8, .div = 16 },
+	{ .val = 1, .div = 4 },
+	{ .val = 2, .div = 6 },
+	{ .val = 3, .div = 8 },
+	{ .val = 4, .div = 10 },
+	{ .val = 5, .div = 12 },
+	{ .val = 6, .div = 14 },
+	{ .val = 7, .div = 16 },
+	{ .val = 8, .div = 18 },
 	{ /* sentinel */ },
 };
 
-- 
2.19.0

[PATCH 2/2] clk: meson: meson8b: fix the width of the cpu_scale_div clock

[Martin Blumenstingl]

According to the public S805 datasheet HHI_SYS_CPU_CLK_CNTL1[29:20] is
the register for the CPU scale_div clock. This matches the code in
Amlogic's 3.10 GPL kernel sources:
N = (aml_read_reg32(P_HHI_SYS_CPU_CLK_CNTL1) >> 20) & 0x3FF;

This means that the divider register is 10 bit wide instead of 9 bits.
So far this is not a problem since all u-boot versions I have seen are
not using the cpu_scale_div clock at all (instead they are configuring
the CPU clock to run off cpu_in_sel directly).

The fixes tag points to the latest rework of the CPU clocks. However,
even before the rework it was wrong. Commit 7a29a869434e8b ("clk: meson:
Add support for Meson clock controller") defines MESON_N_WIDTH as 9 (in
drivers/clk/meson/clk-cpu.c). But since the old clk-cpu implementation
this only carries the fixes tag for the CPU clock rewordk.

Fixes: 251b6fd38bcb9c ("clk: meson: rework meson8b cpu clock")
Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
---
 drivers/clk/meson/meson8b.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/clk/meson/meson8b.c b/drivers/clk/meson/meson8b.c
index d3f4e11d106d..8658a662b10e 100644
--- a/drivers/clk/meson/meson8b.c
+++ b/drivers/clk/meson/meson8b.c
@@ -594,7 +594,7 @@ static struct clk_regmap meson8b_cpu_scale_div = {
 	.data = &(struct clk_regmap_div_data){
 		.offset =  HHI_SYS_CPU_CLK_CNTL1,
 		.shift = 20,
-		.width = 9,
+		.width = 10,
 		.table = cpu_scale_table,
 		.flags = CLK_DIVIDER_ALLOW_ZERO,
 	},
-- 
2.19.0

[PATCH 0/2] Meson8b: fixes for the cpu_scale_div clock

[Neil Armstrong]

On 27/09/2018 10:59, Martin Blumenstingl wrote:
> While trying to add support for the TWD timer I found that our CPU clock
> calculation (when running off the "cpu_scale_div clock) is incorrect.
> 
> The main problem was the cpu_scale_div clock:
> it's divider table has an off-by-one error. the old formula was:
>   parent_rate / 2 * register_value
> however, testing shows that the correct formula is:
>   parent_rate / 2 * (register value + 1)
> See the commit message of patch #1 for a complete description of the
> problem and it's history.
> 
> While looking at the "cpu_scale_div" clock I also found that it's
> register width was also off-by-one, so this is fixed in a separate patch
> as well.
> 
> Both patches are not critical because I haven't seen a case where u-boot
> uses cpu_scale_div for the CPU clock. It's only a problem when playing
> with the CPU clock in u-boot (by writing registers manually) or as soon
> as we support CPU frequency scaling.
> 
> 
> Martin Blumenstingl (2):
>   clk: meson: meson8b: fix incorrect divider mapping in cpu_scale_table
>   clk: meson: meson8b: fix the width of the cpu_scale_div clock
> 
>  drivers/clk/meson/meson8b.c | 17 +++++++++--------
>  1 file changed, 9 insertions(+), 8 deletions(-)
> 

Pushed to fixes/drivers for 4.20

Neil

[PATCH 0/2] Meson8b: fixes for the cpu_scale_div clock

[Neil Armstrong]

On 08/11/2018 15:16, Neil Armstrong wrote:
> On 27/09/2018 10:59, Martin Blumenstingl wrote:
>> While trying to add support for the TWD timer I found that our CPU clock
>> calculation (when running off the "cpu_scale_div clock) is incorrect.
>>
>> The main problem was the cpu_scale_div clock:
>> it's divider table has an off-by-one error. the old formula was:
>>   parent_rate / 2 * register_value
>> however, testing shows that the correct formula is:
>>   parent_rate / 2 * (register value + 1)
>> See the commit message of patch #1 for a complete description of the
>> problem and it's history.
>>
>> While looking at the "cpu_scale_div" clock I also found that it's
>> register width was also off-by-one, so this is fixed in a separate patch
>> as well.
>>
>> Both patches are not critical because I haven't seen a case where u-boot
>> uses cpu_scale_div for the CPU clock. It's only a problem when playing
>> with the CPU clock in u-boot (by writing registers manually) or as soon
>> as we support CPU frequency scaling.
>>
>>
>> Martin Blumenstingl (2):
>>   clk: meson: meson8b: fix incorrect divider mapping in cpu_scale_table
>>   clk: meson: meson8b: fix the width of the cpu_scale_div clock
>>
>>  drivers/clk/meson/meson8b.c | 17 +++++++++--------
>>  1 file changed, 9 insertions(+), 8 deletions(-)
>>
> 
> Pushed to fixes/drivers for 4.20
> 
> Neil
> 

Finally, pushed to next/drivers to since these are not critical and to
satisfy the dependencies of "Meson8b: make the CPU clock mutable" patchset.