build/patch/kernel/mvebu-default/420-rtc-trimming-support.patch
Aditya Prayoga e261c6f828 Move mvebu DEFAULT, NEXT and DEV branch to next kernel (LTS) and U-boot #1426 (#1487)
* Initial Mvebu RFC https://github.com/armbian/build/issues/1426

Signed-off-by: Igor Pecovnik <igor.pecovnik@gmail.com>

* mvebu: add missing patches

Signed-off-by: Igor Pecovnik <igor.pecovnik@gmail.com>

* mvebu: change making u-boot targets to standard way, adjust patches and config

Signed-off-by: Igor Pecovnik <igor.pecovnik@gmail.com>

* helios4: set default branch to use U-Boot 2018.11

Switch over to U-Boot 2018.11 that has been used for some time in next
branch.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: helios4: Enable DEV branch

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* u-boot: Add RTC support on Clearfog and Helios4

Added DM driver for mvebu RTC and enable it on Clearfog and Helios4
configuration.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* add boot-marvell.cmd backward compatibility

The patches added missing variable that used on boot-marvell.cmd and
also adjust the some memory addresses to prevent crash due to usage of
fdt_high and initrd_high.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* helios4: Added SPI NOR flash target

Build bootable SPI NOR flash image.
Change the boot order to USB -> SATA -> MMC

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* Restore SPI support on U-Boot 2019.04

* mvebu: kernel: Added Wake-On-GPIO and WoL support

The patch set was missing during transition.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* [#1429] SolidRun's ARMADA A388 SOM U-Boot ODT Update

Old versions of U-Boot did not configure correctly the ODT on data
signals of DDR RAM on SolidRun's ARMADA A388 SOMs.

According to SolidRun Knowledge Base, the changes already pushed to
mainline U-Boot. But then it was overwritten when Marvell DDR Training
Tool updated

[URL]
https://developer.solid-run.com/knowledge-base/armada-38x-som-u-boot-odt-update/

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* [#1429] mvebu: u-boot: Add revision id for Armada 38x B0

Added patch for SolidRun U-Boot v2018.01 and
for Helios4 U-Boot v2018.11

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* clearfog: Added SPI NOR flash target

Build bootable SPI NOR flash image.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: clearfog: DEV branch use mainline U-Boot

Also move clearfog base patch into its own board folder.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: enable U-Boot uart target

Normal MMC image can be used for uart boot using following command:

./tools/kwboot -b u-boot-spl.kwb /dev/ttyUSBX

But on Helios4, the SPL failed to continue the booting process if ECC is
enabled, so disable it.

Since the usage of uart boot is more for rescue/debug, disable autoboot.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: NEXT branch use mainline U-Boot

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: helios4: fix fancontrol related bug during buster testing

- On kernel 4.19, cpu thermal sensor changed the name from
armada_thermal into f10e4078.thermal. Added this new name to udev rules
- Since DEFAULT branch now use kernel 4.14, update fancontrol
configuration
- Load lm75 kernel module
- On kernel 4.19, cpu temp reading about 20 degree C lower, update
fancontrol configuration.

[URL]
https://forum.armbian.com/topic/10214-clearfogpro-possible-change-in-temperature-reporting-between-414next-and-419dev

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: helios4: Override vendor provided fancontrol unit

systemd emit following message on dmesg

systemd[1]: /lib/systemd/system/fancontrol.service:9: PIDFile=
references path below legacy directory /var/run/, updating
/var/run/fancontrol.pid \xe2\x86\x92 /run/fancontrol.pid; please update
the unit file accordingly.

Override and change the value in the unit file to remove the message.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: DEV branch use its own u-boot patch folder

The patches are copied over from u-boot-mvebu-next

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: u-boot: Make clearfog model distinction more obvious

While at it, also change SerDes LANE4 into USB 3.0 on Clearfog Base.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* lib: Use apt-get install instead of dpkg on install_deb_chroot()

dpkg -i does not install dependencies required by the package.
This is needed if the BSP package requires other package that is not
installed during debootstrap.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* config: mvebu: helios4: Move various tweak to family_tweak_bsp()

Various tweak in family_tweaks_s() only applied to SD card image.
Move it to family_tweaks_bsp() so it will also included on the BSP
package and applied to existing user.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* config: mvebu: helios4: Add /etc/modules to BSP

On kernel 4.19, user need to modify the /etc/modules to add lm75 kernel
module. Pack the file into BSP so user no longer needed to modify it.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: kernel: Make zbud as built-in module

To remove the following error:
[    1.705485] zswap: default zpool zbud not available
[    1.705488] zswap: pool creation failed

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* bootscripts: mvebu: Add default value for spi_workaround

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: kernel: Backport armada_thermal changes to 4.14 (#1452)

On kernel 4.19, armada_thermal driver has been fixed to address
Marvell's Errata #132698 (The changes first appear on LK 4.16). The
result is temperature reading is around 20 degree Celsius lower.

Currently armbian-motd apply -20C tweak for both LK 4.14 and LK 4.19
which is incorrect. Instead of adding some logic on what condition to
apply the tweak, it is better to remove the tweak and patch the kernel
instead.

Revert commit b3dd4e9 ("[ mvebu ] Put back Armada temperature tweak in
motd")
which is part of #1421 solution.

[URL]
https://forum.armbian.com/topic/10214-clearfogpro-possible-change-in-temperature-reporting-between-414next-and-419dev/

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: helios4: unified fancontrol config

Since LK 4.14 on DEFAULT branch already patched and the temp reading is
same as LK 4.19 on NEXT branch, it is no longer needed to separate
fancontrol configuration file.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* u-boot: helios4: Remove rev id patch

The patch is already applied in helios4 repo, no need to have it in
armbian.
This revert helios4 part of commit 7411c55

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* u-boot: clearfog: enable PCIe support and PCIe reset

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* u-boot: clearfog: add boot-marvell.cmd backward compatibility

The patches added missing variable that used on boot-marvell.cmd and
also adjust the some memory addresses to prevent crash due to usage of
fdt_high and initrd_high.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu: helios4: tweak regarding temperature setting

Make fan speed similar compared to pre-patched armada-thermal. Target
PWM value around 70 during idle.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* config: sources: clearfog to use u-boot 2018.01 for NEXT branch

This changes also affect Helios4. Moved the shared U-Boot source setting
back to Helios4 for NEXT branch.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* config: boards: build Stretch image for Clearfog and Helios4

Also remove DEV from Helios4 CLI_TARGET

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* kernel: mvebu-next: Disable access to SPI Flash

User need to set spi_workaround=yes to enable SPI Flash access and lost
access to internal SATA.

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu-next: adjust kernel config

* mvebu-dev: bump to 5.2 and adjust kernel configuraion. Tested for building.

* Adjust kernel config, add AUFS

Signed-off-by: Igor Pecovnik <igor.pecovnik@gmail.com>

* mvebu-next: Adjust kernel config, add debug GPIO

Signed-off-by: Aditya Prayoga <aditya@kobol.io>

* mvebu-dev: separate Clearfog Base U-boot configuration file and patch

Signed-off-by: Aditya Prayoga <aditya@kobol.io>
2019-07-31 12:51:00 +02:00

143 lines
4.5 KiB
Diff

commit f94ffbc2c2a4128c4412bb483d0807722dfb682b
Author: Russell King <rmk+kernel@armlinux.org.uk>
Date: Fri Sep 29 11:23:31 2017 +0100
rtc: armada38x: add support for trimming the RTC
Add support for trimming the RTC using the offset mechanism. This RTC
supports two modes: low update mode and high update mode. Low update
mode has finer precision than high update mode, so we use the low mode
where possible.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Alexandre Belloni <alexandre.belloni@free-electrons.com>
--- a/drivers/rtc/rtc-armada38x.c
+++ b/drivers/rtc/rtc-armada38x.c
@@ -28,6 +28,8 @@
#define RTC_IRQ_AL_EN BIT(0)
#define RTC_IRQ_FREQ_EN BIT(1)
#define RTC_IRQ_FREQ_1HZ BIT(2)
+#define RTC_CCR 0x18
+#define RTC_CCR_MODE BIT(15)
#define RTC_TIME 0xC
#define RTC_ALARM1 0x10
@@ -343,18 +345,117 @@ static irqreturn_t armada38x_rtc_alarm_i
return IRQ_HANDLED;
}
+/*
+ * The information given in the Armada 388 functional spec is complex.
+ * They give two different formulas for calculating the offset value,
+ * but when considering "Offset" as an 8-bit signed integer, they both
+ * reduce down to (we shall rename "Offset" as "val" here):
+ *
+ * val = (f_ideal / f_measured - 1) / resolution where f_ideal = 32768
+ *
+ * Converting to time, f = 1/t:
+ * val = (t_measured / t_ideal - 1) / resolution where t_ideal = 1/32768
+ *
+ * => t_measured / t_ideal = val * resolution + 1
+ *
+ * "offset" in the RTC interface is defined as:
+ * t = t0 * (1 + offset * 1e-9)
+ * where t is the desired period, t0 is the measured period with a zero
+ * offset, which is t_measured above. With t0 = t_measured and t = t_ideal,
+ * offset = (t_ideal / t_measured - 1) / 1e-9
+ *
+ * => t_ideal / t_measured = offset * 1e-9 + 1
+ *
+ * so:
+ *
+ * offset * 1e-9 + 1 = 1 / (val * resolution + 1)
+ *
+ * We want "resolution" to be an integer, so resolution = R * 1e-9, giving
+ * offset = 1e18 / (val * R + 1e9) - 1e9
+ * val = (1e18 / (offset + 1e9) - 1e9) / R
+ * with a common transformation:
+ * f(x) = 1e18 / (x + 1e9) - 1e9
+ * offset = f(val * R)
+ * val = f(offset) / R
+ *
+ * Armada 38x supports two modes, fine mode (954ppb) and coarse mode (3815ppb).
+ */
+static long armada38x_ppb_convert(long ppb)
+{
+ long div = ppb + 1000000000L;
+
+ return div_s64(1000000000000000000LL + div / 2, div) - 1000000000L;
+}
+
+static int armada38x_rtc_read_offset(struct device *dev, long *offset)
+{
+ struct armada38x_rtc *rtc = dev_get_drvdata(dev);
+ unsigned long ccr, flags;
+ long ppb_cor;
+
+ spin_lock_irqsave(&rtc->lock, flags);
+ ccr = rtc->data->read_rtc_reg(rtc, RTC_CCR);
+ spin_unlock_irqrestore(&rtc->lock, flags);
+
+ ppb_cor = (ccr & RTC_CCR_MODE ? 3815 : 954) * (s8)ccr;
+ /* ppb_cor + 1000000000L can never be zero */
+ *offset = armada38x_ppb_convert(ppb_cor);
+
+ return 0;
+}
+
+static int armada38x_rtc_set_offset(struct device *dev, long offset)
+{
+ struct armada38x_rtc *rtc = dev_get_drvdata(dev);
+ unsigned long ccr = 0;
+ long ppb_cor, off;
+
+ /*
+ * The maximum ppb_cor is -128 * 3815 .. 127 * 3815, but we
+ * need to clamp the input. This equates to -484270 .. 488558.
+ * Not only is this to stop out of range "off" but also to
+ * avoid the division by zero in armada38x_ppb_convert().
+ */
+ offset = clamp(offset, -484270L, 488558L);
+
+ ppb_cor = armada38x_ppb_convert(offset);
+
+ /*
+ * Use low update mode where possible, which gives a better
+ * resolution of correction.
+ */
+ off = DIV_ROUND_CLOSEST(ppb_cor, 954);
+ if (off > 127 || off < -128) {
+ ccr = RTC_CCR_MODE;
+ off = DIV_ROUND_CLOSEST(ppb_cor, 3815);
+ }
+
+ /*
+ * Armada 388 requires a bit pattern in bits 14..8 depending on
+ * the sign bit: { 0, ~S, S, S, S, S, S }
+ */
+ ccr |= (off & 0x3fff) ^ 0x2000;
+ rtc_delayed_write(ccr, rtc, RTC_CCR);
+
+ return 0;
+}
+
static const struct rtc_class_ops armada38x_rtc_ops = {
.read_time = armada38x_rtc_read_time,
.set_time = armada38x_rtc_set_time,
.read_alarm = armada38x_rtc_read_alarm,
.set_alarm = armada38x_rtc_set_alarm,
.alarm_irq_enable = armada38x_rtc_alarm_irq_enable,
+ .read_offset = armada38x_rtc_read_offset,
+ .set_offset = armada38x_rtc_set_offset,
};
static const struct rtc_class_ops armada38x_rtc_ops_noirq = {
.read_time = armada38x_rtc_read_time,
.set_time = armada38x_rtc_set_time,
.read_alarm = armada38x_rtc_read_alarm,
+ .read_offset = armada38x_rtc_read_offset,
+ .set_offset = armada38x_rtc_set_offset,
};
static const struct armada38x_rtc_data armada38x_data = {