[U-Boot] [PATCH] mxs: fixed battery boot on imx233-olinuxino-micro
This patch makes it possible to boot from battery on olinuxino Boards from Olimex.
We should not set DOUBLE_FETS in HW_POWER_MINPWR, it makes my system reset with in seconds while running on battery power.
Also mxs_power_enable_4p2() should not be called if running from battery. It switches VDDD and VDDA to LinReg mode and than turns off the VDDIO DCDC supply. At this point the system resets. If we run on battery this is no good idea because the LinRegs are chained behind VDDIO and battery voltage seems not enough to power the system from LinRegs. Also is the power system already running on DCDC after BootROM hands over to u-boot, so no sense in switching back to LinRegs. --- arch/arm/cpu/arm926ejs/mxs/spl_power_init.c | 6 ------ 1 file changed, 6 deletions(-)
diff --git a/arch/arm/cpu/arm926ejs/mxs/spl_power_init.c b/arch/arm/cpu/arm926ejs/mxs/spl_power_init.c index d25019a..0cbb6c4 100644 --- a/arch/arm/cpu/arm926ejs/mxs/spl_power_init.c +++ b/arch/arm/cpu/arm926ejs/mxs/spl_power_init.c @@ -237,9 +237,6 @@ static void mxs_src_power_init(void) POWER_LOOPCTRL_RCSCALE_THRESH | POWER_LOOPCTRL_EN_RCSCALE_8X);
- clrsetbits_le32(&power_regs->hw_power_minpwr, - POWER_MINPWR_HALFFETS, POWER_MINPWR_DOUBLE_FETS); - /* 5V to battery handoff ... FIXME */ setbits_le32(&power_regs->hw_power_5vctrl, POWER_5VCTRL_DCDC_XFER); early_delay(30); @@ -631,8 +628,6 @@ static void mxs_batt_boot(void)
writel(POWER_CTRL_ENIRQ_DCDC4P2_BO, &power_regs->hw_power_ctrl_clr);
- clrsetbits_le32(&power_regs->hw_power_minpwr, - POWER_MINPWR_HALFFETS, POWER_MINPWR_DOUBLE_FETS);
mxs_power_set_linreg();
@@ -655,7 +650,6 @@ static void mxs_batt_boot(void) POWER_5VCTRL_CHARGE_4P2_ILIMIT_MASK, 0x8 << POWER_5VCTRL_CHARGE_4P2_ILIMIT_OFFSET);
- mxs_power_enable_4p2(); }
/**
On Wednesday, May 21, 2014 at 04:01:55 AM, Peter Schumann wrote:
This patch makes it possible to boot from battery on olinuxino Boards from Olimex.
We should not set DOUBLE_FETS in HW_POWER_MINPWR
Please explain how you arrived to this conclusion.
, it makes my system reset with in seconds while running on battery power.
Also mxs_power_enable_4p2() should not be called if running from battery. It switches VDDD and VDDA to LinReg mode and than turns off the VDDIO DCDC supply. At this point the system resets.
This has been tested on i.MX28 , can you please check what the differences are between these two chips and why it works only on one of them ?
If we run on battery this is no good idea because the LinRegs are chained behind VDDIO and battery voltage seems not enough to power the system from LinRegs. Also is the power system already running on DCDC after BootROM hands over to u-boot, so no sense in switching back to LinRegs.
The BootROM never hands over to U-Boot, so this really makes no sense. Can you please explain ? The BootROM handles over to U-Boot SPL, which configures the power block.
Also, I find it a very bad idea to depend on the BootROM to actually start the DCDC converter. I don't think we can universally say the DCDC converter is running when entering U-Boot SPL. I also don't think we can even depend on the power block configuration when exiting the BootROM -- for example in JTAG Boot Mode, the BootROM halts early in the boot process and will likely not configure anything with regards to the power block.
[...]
THanks!
Best regards, Marek Vasut
Am 24.05.2014 12:44, schrieb Marek Vasut:
Please explain how you arrived to this conclusion.
Hello Marek,
I started writing an powerblock driver for the i.MX233 to use the imx233 olimex board with a battery. Short time (instant up to 90 seconds) after the 5v to battery handoff the system reseted. The system was running on DCDC prior to the handoff. I suspected that some of the power rails where droping and trigger hardware brownouts. But this was not the case. The DCDC got shutdown and some ms after that the rails begun to drop. I have some scope pic of this happening i it is of interest.
I than tried out some random bits in the MINPWR register and the reset problem was gone. Than I track down it to this 1 bit causing the instability.
This has been tested on i.MX28 , can you please check what the differences are between these two chips and why it works only on one of them ?
From the block diagram in the AN4199 pdf i am quite sure that it wont
work this way on the i.MX28 either. The VDDIO Linreg is sourced from the 5V rail. So if we run from battery only we can not deactivate VDDIO from DCDC by setting DISABLE_FET.
The BootROM never hands over to U-Boot, so this really makes no sense. Can you please explain ? The BootROM handles over to U-Boot SPL, which configures the power block.
Sorry I meant the U-Boot SPL.I not that familiar with U-Boot architecture.
Also, I find it a very bad idea to depend on the BootROM to actually start the DCDC converter. I don't think we can universally say the DCDC converter is running when entering U-Boot SPL. I also don't think we can even depend on the power block configuration when exiting the BootROM -- for example in JTAG Boot Mode, the BootROM halts early in the boot process and will likely not configure anything with regards to the power block.
I am with you at the point that proper initialization is very important and we should not rely on BootROM.
If the system is running on battery only, we should not switch the VDDIO regulator to Linreg mode because there are no 5V to source it. In mxs_power_enable_4p2() all Regulators get switched to Linreg mode and that kills the power to hole system and there by is triggering an brownout reset.
The mxs_power_enable_4p2() function can only work if we assume there is 5V present.
I hope I got clearer way this will not work in a battery only scenario.
Best regards, Peter Schumann
On Wednesday, May 21, 2014 at 04:01:55 AM, Peter Schumann wrote:
This patch makes it possible to boot from battery on olinuxino Boards from Olimex.
We should not set DOUBLE_FETS in HW_POWER_MINPWR, it makes my system reset with in seconds while running on battery power.
Also mxs_power_enable_4p2() should not be called if running from battery. It switches VDDD and VDDA to LinReg mode and than turns off the VDDIO DCDC supply. At this point the system resets. If we run on battery this is no good idea because the LinRegs are chained behind VDDIO and battery voltage seems not enough to power the system from LinRegs. Also is the power system already running on DCDC after BootROM hands over to u-boot, so no sense in switching back to LinRegs.
Sorry for the late reply, I'm somewhat congested recently. In the meantime, I was not able to test this yet. But you're missing Signed-off-by: line, can you please re-send this with a proper SoB line [1] ?
[1] http://www.denx.de/wiki/U-Boot/Patches#General_Patch_Submission_Rules
Thanks
Best regards, Marek Vasut
On Friday, June 06, 2014 at 12:18:09 AM, Marek Vasut wrote:
On Wednesday, May 21, 2014 at 04:01:55 AM, Peter Schumann wrote:
This patch makes it possible to boot from battery on olinuxino Boards from Olimex.
We should not set DOUBLE_FETS in HW_POWER_MINPWR, it makes my system reset with in seconds while running on battery power.
Also mxs_power_enable_4p2() should not be called if running from battery. It switches VDDD and VDDA to LinReg mode and than turns off the VDDIO DCDC supply. At this point the system resets. If we run on battery this is no good idea because the LinRegs are chained behind VDDIO and battery voltage seems not enough to power the system from LinRegs. Also is the power system already running on DCDC after BootROM hands over to u-boot, so no sense in switching back to LinRegs.
Sorry for the late reply, I'm somewhat congested recently. In the meantime, I was not able to test this yet. But you're missing Signed-off-by: line, can you please re-send this with a proper SoB line [1] ?
[1] http://www.denx.de/wiki/U-Boot/Patches#General_Patch_Submission_Rules
Bump?
Best regards, Marek Vasut
On Friday, June 13, 2014 at 09:36:51 AM, Peter Schumann wrote:
Bump?
Best regards, Marek Vasut
Hi Marek,
I am occupied with work at the moment. I will get it done at the weekend.
Thanks!
Best regards, Marek Vasut
This patch makes it possible to boot from battery on olinuxino Boards from Olimex.
We should not set DOUBLE_FETS in HW_POWER_MINPWR, it makes my system reset with in seconds while running on battery power.
Also mxs_power_enable_4p2() should not be called if running from battery. It switches VDDD and VDDA to LinReg mode and than turns off the VDDIO DCDC supply. At this point the system resets. If we run on battery this is no good idea because the LinRegs are chained behind VDDIO and battery voltage seems not enough to power the system from LinRegs. Also is the power system already running on DCDC after BootROM hands over to u-boot, so no sense in switching back to LinRegs.
Signed-off-by: Peter Schumann peter.schumann@jeepxj.de --- arch/arm/cpu/arm926ejs/mxs/spl_power_init.c | 6 ------ 1 file changed, 6 deletions(-)
diff --git a/arch/arm/cpu/arm926ejs/mxs/spl_power_init.c b/arch/arm/cpu/arm926ejs/mxs/spl_power_init.c index d25019a..0cbb6c4 100644 --- a/arch/arm/cpu/arm926ejs/mxs/spl_power_init.c +++ b/arch/arm/cpu/arm926ejs/mxs/spl_power_init.c @@ -237,9 +237,6 @@ static void mxs_src_power_init(void) POWER_LOOPCTRL_RCSCALE_THRESH | POWER_LOOPCTRL_EN_RCSCALE_8X);
- clrsetbits_le32(&power_regs->hw_power_minpwr, - POWER_MINPWR_HALFFETS, POWER_MINPWR_DOUBLE_FETS); - /* 5V to battery handoff ... FIXME */ setbits_le32(&power_regs->hw_power_5vctrl, POWER_5VCTRL_DCDC_XFER); early_delay(30); @@ -631,8 +628,6 @@ static void mxs_batt_boot(void)
writel(POWER_CTRL_ENIRQ_DCDC4P2_BO, &power_regs->hw_power_ctrl_clr);
- clrsetbits_le32(&power_regs->hw_power_minpwr, - POWER_MINPWR_HALFFETS, POWER_MINPWR_DOUBLE_FETS);
mxs_power_set_linreg();
@@ -655,7 +650,6 @@ static void mxs_batt_boot(void) POWER_5VCTRL_CHARGE_4P2_ILIMIT_MASK, 0x8 << POWER_5VCTRL_CHARGE_4P2_ILIMIT_OFFSET);
- mxs_power_enable_4p2(); }
/**
On Sunday, June 15, 2014 at 03:11:56 PM, Peter Schumann wrote:
This patch makes it possible to boot from battery on olinuxino Boards from Olimex.
We should not set DOUBLE_FETS in HW_POWER_MINPWR, it makes my system reset with in seconds while running on battery power.
Also mxs_power_enable_4p2() should not be called if running from battery. It switches VDDD and VDDA to LinReg mode and than turns off the VDDIO DCDC supply. At this point the system resets. If we run on battery this is no good idea because the LinRegs are chained behind VDDIO and battery voltage seems not enough to power the system from LinRegs. Also is the power system already running on DCDC after BootROM hands over to u-boot, so no sense in switching back to LinRegs.
Signed-off-by: Peter Schumann peter.schumann@jeepxj.de
+CC Fabio, I'd like to see his take on this too.
Best regards, Marek Vasut
On Tuesday, June 17, 2014 at 12:18:55 AM, Marek Vasut wrote:
On Sunday, June 15, 2014 at 03:11:56 PM, Peter Schumann wrote:
This patch makes it possible to boot from battery on olinuxino Boards from Olimex.
We should not set DOUBLE_FETS in HW_POWER_MINPWR, it makes my system reset with in seconds while running on battery power.
Also mxs_power_enable_4p2() should not be called if running from battery. It switches VDDD and VDDA to LinReg mode and than turns off the VDDIO DCDC supply. At this point the system resets. If we run on battery this is no good idea because the LinRegs are chained behind VDDIO and battery voltage seems not enough to power the system from LinRegs. Also is the power system already running on DCDC after BootROM hands over to u-boot, so no sense in switching back to LinRegs.
Signed-off-by: Peter Schumann peter.schumann@jeepxj.de
+CC Fabio, I'd like to see his take on this too.
Reviewed-by: Marek Vasut marex@denx.de
I'd like to get this merged after 2014.07 though, so it can get some proper testing. We're too late in the -rc cycle anyway ...
Best regards, Marek Vasut
Hi peter,
On 17/06/2014 00:19, Marek Vasut wrote:
On Tuesday, June 17, 2014 at 12:18:55 AM, Marek Vasut wrote:
On Sunday, June 15, 2014 at 03:11:56 PM, Peter Schumann wrote:
This patch makes it possible to boot from battery on olinuxino Boards from Olimex.
We should not set DOUBLE_FETS in HW_POWER_MINPWR, it makes my system reset with in seconds while running on battery power.
Also mxs_power_enable_4p2() should not be called if running from battery. It switches VDDD and VDDA to LinReg mode and than turns off the VDDIO DCDC supply. At this point the system resets. If we run on battery this is no good idea because the LinRegs are chained behind VDDIO and battery voltage seems not enough to power the system from LinRegs. Also is the power system already running on DCDC after BootROM hands over to u-boot, so no sense in switching back to LinRegs.
Signed-off-by: Peter Schumann peter.schumann@jeepxj.de
+CC Fabio, I'd like to see his take on this too.
Reviewed-by: Marek Vasut marex@denx.de
I'd like to get this merged after 2014.07 though, so it can get some proper testing. We're too late in the -rc cycle anyway ...
Agree. Reason is that this change is for all boards, and without a proper testing I will avoid to merge it in the release.
Best regards, Stefano Babic
Am 17.06.2014 18:00, schrieb Stefano Babic:
Hi peter, ... On 17/06/2014 00:19, Marek Vasut wrote: Agree. Reason is that this change is for all boards, and without a proper testing I will avoid to merge it in the release.
Best regards, Stefano Babic
Hi Stefano,
I've only the imx233 olimex board for testing. The DOUBLE_FETS switch in HW_POWER_MINPWR is some thing were maybe problems emerge.
I have very little load on the powerrails (imx233 + DRAM), so maybe with more current demand DOUBLE_FETS is working OK and even more may require it to be set.
The DOUBLE_FETS and HALF_FETS switches control the switching MOSFET size in the DCDC Converter. My assumption is that if we have a to big FET setting and to low current demand we trigger some failsafe mechanism. Because the required "on time" of the FET is smaller than the minimum possible "on time" of then DCDC FET control logic.
I don't have an easy way to test this assumption without an adjustable load.
The other change regarding mxs_power_enable_4p2() should be relatively safe to apply.
If I can help with some testing let me know.
Best regards, Peter Schumann
On Wednesday, June 18, 2014 at 08:17:05 PM, Peter Schumann wrote:
Am 17.06.2014 18:00, schrieb Stefano Babic:
Hi peter, ... On 17/06/2014 00:19, Marek Vasut wrote: Agree. Reason is that this change is for all boards, and without a proper testing I will avoid to merge it in the release.
Best regards, Stefano Babic
Hi Stefano,
I've only the imx233 olimex board for testing. The DOUBLE_FETS switch in HW_POWER_MINPWR is some thing were maybe problems emerge.
I have very little load on the powerrails (imx233 + DRAM), so maybe with more current demand DOUBLE_FETS is working OK and even more may require it to be set.
If that was the case, the problem would manifest even when _not_ booting from battery-only supply, correct?
The DOUBLE_FETS and HALF_FETS switches control the switching MOSFET size in the DCDC Converter. My assumption is that if we have a to big FET setting and to low current demand we trigger some failsafe mechanism. Because the required "on time" of the FET is smaller than the minimum possible "on time" of then DCDC FET control logic.
I don't have an easy way to test this assumption without an adjustable load.
The other change regarding mxs_power_enable_4p2() should be relatively safe to apply.
This breaks booting from battery on MX28 for some reason. The DC converter is not enabled at that point.
Best regards, Marek Vasut
Hi Peter, hi Marek,
On 19/06/2014 09:12, Marek Vasut wrote:
On Wednesday, June 18, 2014 at 08:17:05 PM, Peter Schumann wrote:
Am 17.06.2014 18:00, schrieb Stefano Babic:
Hi peter, ... On 17/06/2014 00:19, Marek Vasut wrote: Agree. Reason is that this change is for all boards, and without a proper testing I will avoid to merge it in the release.
Best regards, Stefano Babic
Hi Stefano,
I've only the imx233 olimex board for testing. The DOUBLE_FETS switch in HW_POWER_MINPWR is some thing were maybe problems emerge.
I have very little load on the powerrails (imx233 + DRAM), so maybe with more current demand DOUBLE_FETS is working OK and even more may require it to be set.
If that was the case, the problem would manifest even when _not_ booting from battery-only supply, correct?
The DOUBLE_FETS and HALF_FETS switches control the switching MOSFET size in the DCDC Converter. My assumption is that if we have a to big FET setting and to low current demand we trigger some failsafe mechanism. Because the required "on time" of the FET is smaller than the minimum possible "on time" of then DCDC FET control logic.
I don't have an easy way to test this assumption without an adjustable load.
The other change regarding mxs_power_enable_4p2() should be relatively safe to apply.
This breaks booting from battery on MX28 for some reason. The DC converter is not enabled at that point.
As I have not (yet) forgotten this patch: current status is that mx28 is broken after applying this patch. I set the status as "Changes requested" and it will be not applied.
Best regards, Stefano Babic
participants (3)
-
Marek Vasut -
Peter Schumann -
Stefano Babic