Skip to contentSkip to conversation

i2c issues with efikamx and mainline uboot

U-Boot Konstantinos Margaritis 4 participants 6 messages Nov 19, 2013 – Nov 21, 2013

lore
Share & tools
lore
#0 lore raw

(Disclaimer: u-boot newbie)

Hi all,

With Marek's huge help (thanks Marek!) I've tried to merge his tree [1] that provides lcd support for the efikamx/efikasb with recent mainline uboot. It boots, but I've hit some wall now and couldn't get passed it, so I thought I'd ask for some expert advice, Marek suggested I mail here and CC himself and Stefano as well.

So, I've migrated the structures to the iomux v3 coding conventions, added some missing defines for iomuxes used, etc. I've forked the code in github [2] and am doing all the changes there, in the hope that I can at some point submit it back to mainline. I also want to to enable multi-host USB support for the platform so that tftp booting will be at last possible, but that's going to happen later, I want to be done with LCD first.

The result boots, but in short it fails to initialise the LCD (for the efikamx, didn't test efikasb yet). I noticed that amongst many changes, in drivers/i2c/mxc_i2c.c, i2c_imx_bus_busy() in Marek's tree () got sort of replaced with wait_for_sr_state(), which does the same thing but is more generic -or so I understand, please correct me if I'm wrong in my interpretation.

So what's the problem? Well, I2CR_MSTA (master mode) is set, but I2SR_IBB fails to get set. In every case I get this result in the I2SR register:

I2SR = 81 -> I2SR_ICF | I2SR_RX_NO_AK, ie.

which according to the iMX515 Reference Manual corresponds to:

"Transfer is complete, and set by the falling edge of the ninth clock of a byte transfer." + "A ?No acknowledge? signal was detected at the ninth clock."

so a previous answer didn't send an acknowledge or rather the opposite it sent a NACK.

I'm at a loss here, attaching a boot log where I added some printfs around. I'd appreciate any pointer as to what I'm doing wrong.

Thanks

Konstantinos

[1] git://git.denx.de/u-boot-marex.git branch: efikasb2
[2] https://github.com/markos/u-boot branch: efikamx
-------------- next part --------------
A non-text attachment was scrubbed...
Name: efikamx-uboot-lcd.log
Type: application/octet-stream
Size: 51228 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20131119/17e9e6ab/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20131119/17e9e6ab/attachment.pgp>

#1 lore raw

Dear Konstantinos Margaritis,

Quoted 37 lines

(Disclaimer: u-boot newbie)

Hi all,

With Marek's huge help (thanks Marek!) I've tried to merge his tree [1] that provides lcd support for the efikamx/efikasb with recent mainline uboot. It boots, but I've hit some wall now and couldn't get passed it, so I thought I'd ask for some expert advice, Marek suggested I mail here and CC himself and Stefano as well.

So, I've migrated the structures to the iomux v3 coding conventions, added some missing defines for iomuxes used, etc. I've forked the code in github [2] and am doing all the changes there, in the hope that I can at some point submit it back to mainline. I also want to to enable multi-host USB support for the platform so that tftp booting will be at last possible, but that's going to happen later, I want to be done with LCD first.

The result boots, but in short it fails to initialise the LCD (for the efikamx, didn't test efikasb yet). I noticed that amongst many changes, in drivers/i2c/mxc_i2c.c, i2c_imx_bus_busy() in Marek's tree () got sort of replaced with wait_for_sr_state(), which does the same thing but is more generic -or so I understand, please correct me if I'm wrong in my interpretation.

So what's the problem? Well, I2CR_MSTA (master mode) is set, but I2SR_IBB fails to get set. In every case I get this result in the I2SR register:

I2SR = 81 -> I2SR_ICF | I2SR_RX_NO_AK, ie.

which according to the iMX515 Reference Manual corresponds to:

"Transfer is complete, and set by the falling edge of the ninth clock of a byte transfer." + "A ?No acknowledge? signal was detected at the ninth clock."

Quoted 5 lines

so a previous answer didn't send an acknowledge or rather the opposite it sent a NACK.

I'm at a loss here, attaching a boot log where I added some printfs around. I'd appreciate any pointer as to what I'm doing wrong.

Honestly, I have a feeling the MX51 I2C controller might be broken. Isn't there maybe something in the ERRATA for MX51 (I know it's HUGE) ?

Best regards,
Marek Vasut

#2 lore raw

Dear Marek Vasut,

Quoted 42 lines · Marek Vasut

Dear Konstantinos Margaritis,

Quoted 40 lines

(Disclaimer: u-boot newbie)

Hi all,

With Marek's huge help (thanks Marek!) I've tried to merge his tree [1] that provides lcd support for the efikamx/efikasb with recent mainline uboot. It boots, but I've hit some wall now and couldn't get passed it, so I thought I'd ask for some expert advice, Marek suggested I mail here and CC himself and Stefano as well.

So, I've migrated the structures to the iomux v3 coding conventions, added some missing defines for iomuxes used, etc. I've forked the code in github [2] and am doing all the changes there, in the hope that I can at some point submit it back to mainline. I also want to to enable multi-host USB support for the platform so that tftp booting will be at last possible, but that's going to happen later, I want to be done with LCD first.

The result boots, but in short it fails to initialise the LCD (for the efikamx, didn't test efikasb yet). I noticed that amongst many changes, in drivers/i2c/mxc_i2c.c, i2c_imx_bus_busy() in Marek's tree () got sort of replaced with wait_for_sr_state(), which does the same thing but is more generic -or so I understand, please correct me if I'm wrong in my interpretation.

So what's the problem? Well, I2CR_MSTA (master mode) is set, but I2SR_IBB fails to get set. In every case I get this result in the I2SR register:

I2SR = 81 -> I2SR_ICF | I2SR_RX_NO_AK, ie.

which according to the iMX515 Reference Manual corresponds to:

"Transfer is complete, and set by the falling edge of the ninth clock of a byte transfer." + "A ?No acknowledge? signal was detected at the ninth clock."

so a previous answer didn't send an acknowledge or rather the opposite it sent a NACK.

Quoted 6 lines
Quoted 2 lines

I'm at a loss here, attaching a boot log where I added some printfs around. I'd appreciate any pointer as to what I'm doing wrong.

Honestly, I have a feeling the MX51 I2C controller might be broken. Isn't there
maybe something in the ERRATA for MX51 (I know it's HUGE) ?

I use the i.MX515, and the I?C worked fine at some point in mainline U-Boot. If it does not work for you now, it's either because of a software regression, or because of some board-specific issue, e.g. pads. SION may have to be set for I?C.

Best regards,
Beno?t

#3 lore raw
Quoted 4 lines · On Tue, 19 Nov 2013 18:02:19 +0100 (CET) Beno?t Th?baudeau…

I use the i.MX515, and the I?C worked fine at some point in mainline U-Boot. If it does not work for you now, it's either because of a software regression, or because of some board-specific issue, e.g. pads. SION may have to be set for I?C.

Hi Beno?t,

By "SION may have to be set for I?C", do you mean that IOMUX_CONFIG_SION might have to be defined and ORed in the MX51 pads in arch/arm/include/asm/arch-mx5/iomux-mx51.h just is it is set in arch/arm/include/asm/arch-mx5/iomux-mx53.h, for example:

MX53_PAD_KEY_COL3__I2C2_SCL = IOMUX_PAD(0x364, 0x03C, 4 | IOMUX_CONFIG_SION, 0x81C, 0, NO_PAD_CTRL),

How do I know in which ones do I have to set SION? I've added some pad defines in the mx51 file, needed for LCD bringup, and I hope I did it the right way -at least I followed the RM as close as I could and Marek was of tremendous help here, but I didn't even know about that SION thing while doing that. I'd gladly try it out, before following Fabio's advice and try to merge the patches to the u-boot-imx tree -tried already but they don't apply cleanly, I'll have to do it when I get more free time, most likely during the weekend.

Regards

Konstantinos
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20131121/f72684af/attachment.pgp>

#4 lore raw

Dear Konstantinos Margaritis,

Quoted 9 lines · Konstantinos Margaritis
Quoted 4 lines · On Tue, 19 Nov 2013 18:02:19 +0100 (CET) Beno?t Th?baudeau…

I use the i.MX515, and the I?C worked fine at some point in mainline U-Boot. If it does not work for you now, it's either because of a software regression, or because of some board-specific issue, e.g. pads. SION may have to be set for I?C.

Hi Beno?t,

Quoted 7 lines

By "SION may have to be set for I?C", do you mean that IOMUX_CONFIG_SION might have to be defined and ORed in the MX51 pads in arch/arm/include/asm/arch-mx5/iomux-mx51.h just is it is set in arch/arm/include/asm/arch-mx5/iomux-mx53.h, for example:

MX53_PAD_KEY_COL3__I2C2_SCL = IOMUX_PAD(0x364, 0x03C, 4 | IOMUX_CONFIG_SION, 0x81C, 0, NO_PAD_CTRL),

Correct.

Quoted 8 lines

How do I know in which ones do I have to set SION? I've added some pad defines in the mx51 file, needed for LCD bringup, and I hope I did it the right way -at least I followed the RM as close as I could and Marek was of tremendous help here, but I didn't even know about that SION thing while doing that. I'd gladly try it out, before following Fabio's advice and try to merge the patches to the u-boot-imx tree -tried already but they don't apply cleanly, I'll have to do it when I get more free time, most likely during the weekend.

The normal rule is that SION is required whenever a pin is in input and output at the same time. E.g., this is the case when you want to read the actual level on a GPIO pin in output mode. This is also the case for both I?C pins (SDA and SCL):
- SDA is either the master or the slave output depending on the time slot,
- SCL may be stretched by the slave if it is too slow compared to the master requests.
In both cases, the pad may set some level, but the IP also needs to read back the actual level in order to work correctly. Also make sure that SDA and SCL are configured in open drain mode, and that there is a pull-up (either enabled in the SoC pad, or as an external resistor on the board) on both.

Best regards,
Beno?t

#5 lore raw
In reply to Konstantinos Margaritis
Quoted 9 lines · Konstantinos Margaritis

(Disclaimer: u-boot newbie)

Hi all,

With Marek's huge help (thanks Marek!) I've tried to merge his tree [1] that provides lcd support for the efikamx/efikasb with recent mainline uboot. It boots, but I've hit some wall now and couldn't get passed it, so I thought I'd ask for some expert advice, Marek suggested I mail here and CC himself and Stefano as well.

I see that you use '#define CONFIG_I2C_MXC', which is deprecated. Now we use '#define CONFIG_SYS_I2C_MXC' instead.

It would be better if you could use Stefano's u-boot-imx tree to base your work, and post the patches to the list so that we can comment.

Regards,

Fabio Estevam