ARM: OMAP3: Beagle: Fix USB Host on beagle xM Ax/Bx

On Beagle xM Rev. Ax/Bx, the USB power enable GPIO logic is reversed when compared to other revisions i.e. it is active high instead of active low.

Use the beagle_config.usb_pwr_level flag correctly so that the power regulator can be configured at runtime.

Signed-off-by Roger Quadros <[email protected]>

---
 arch/arm/mach-omap2/board-omap3beagle.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

arch/arm/mach-omap2/board-omap3beagle.c +3 -3

index 5382215..21136b2 100644
@@ -112,13 +112,13 @@ static u8 omap3_beagle_version;
  */
 static struct {
 	int mmc1_gpio_wp;
-	int usb_pwr_level;
+	bool usb_pwr_level;	/* 0 - Active Low, 1 - Active High */
 	int dvi_pd_gpio;
 	int usr_button_gpio;
 	int mmc_caps;
 } beagle_config = {
 	.mmc1_gpio_wp = -EINVAL,
-	.usb_pwr_level = GPIOF_OUT_INIT_LOW,
+	.usb_pwr_level = 0,
 	.dvi_pd_gpio = -EINVAL,
 	.usr_button_gpio = 4,
 	.mmc_caps = MMC_CAP_4_BIT_DATA | MMC_CAP_8_BIT_DATA,

@@ -178,7 +178,7 @@ static void __init omap3_beagle_init_rev(void)
 	case 0:
 		printk(KERN_INFO "OMAP3 Beagle Rev: xM Ax/Bx\n");
 		omap3_beagle_version = OMAP3BEAGLE_BOARD_XM;
-		beagle_config.usb_pwr_level = GPIOF_OUT_INIT_HIGH;
+		beagle_config.usb_pwr_level = 1;
 		beagle_config.mmc_caps &= ~MMC_CAP_8_BIT_DATA;
 		break;
 	case 2:

Signature · 1 line

1.7.4.1

On Beagle xM Rev. Ax/Bx, the USB power enable GPIO logic is reversed when compared to other revisions i.e. it is active high instead of active low.

Use the beagle_config.usb_pwr_level flag correctly so that the power regulator can be configured at runtime.

Signed-off-by Roger Quadros <[email protected]>

Roger,

Quoted 8 lines · Roger Quadros

On Beagle xM Rev. Ax/Bx, the USB power enable GPIO logic is reversed when compared to other revisions i.e. it is active high instead of active low.

Use the beagle_config.usb_pwr_level flag correctly so that the power regulator can be configured at runtime.

Signed-off-by Roger Quadros <[email protected]>

I'm currently trying to use an SMSC95xx USB ethernet adapter on the OMAP3 Beagleboard (old one "Beagle Rev C1/C2/C3") via the OMAP EHCI controller. And am testing this by trying to mount the rootfs via NFS. What I see is that the NFS mounting stalls/hangs most of the time quite early in the mount process:

[    6.425689] VFS: Mounted root (nfs filesystem) on device 0:12.
[    6.435485] devtmpfs: mounted
[    6.440002] Freeing unused kernel memory: 384K (c077a000 - c07da000)
INIT: version 2.88 booting
Starting udev
[  188.311309] nfs: server 10.0.0.152 not responding, still trying
[  188.317687] nfs: server 10.0.0.152 not responding, still trying

...

Only very seldom the rootfs can be mounted successfully to the prompt.

I'm using latest kernel.org for this (git ID d2b4a646 also tested with v3.10 release). This is with DT-enabled booting and without-DT (same problem). I also tested some older Linux kernel versions and it fails there as well. Only v3.2.40 seems to be able to mount this rootfs via NFS reliably (approx. 10 times success).

Is this a known issue on the Beagleboard? Do you have any ideas what might cause such an issue (unreliable USB connection, hangup of NFS mounting)? Any hints/links appretiated.

Hi Roger,

Quoted 11 lines · Roger Quadros

Quoted 5 lines

I'm using latest kernel.org for this (git ID d2b4a646 also tested with v3.10 release). This is with DT-enabled booting and without-DT (same problem). I also tested some older Linux kernel versions and it fails there as well. Only v3.2.40 seems to be able to mount this rootfs via NFS reliably (approx. 10 times success).

I can't really tell at the moment where the problem is.

Did you compile in all the necessary drivers required for NFS booting? In the beagle case that would mean NOP_USB_XCEIV, USB_EHCI_HCD, USB_EHCI_HCD_OMAP and the smsc95xx driver.

Yes, sure. Otherwise no USB/NFS mount would have been possible.

Quoted 2 lines

If you can paste your full kernel boot log we can check if all these show up before the kernel tries to look up for the NFS server.

Please find it attached (one log for the NFS mount working and one for failing - both with exact the same kernel/dtb). Also my current .config.

Hi Roger,

Quoted 3 lines

I have been too lazy to setup NFS :). But I will give it a shot soon to see what's going wrong.

Really appreciated. :) Please let me know if you need anything else from me to reproduce this issue.

Stefan,

I tried it out today and it worked 5/5 times. I used a bare bones debian root filesystem for
the NFS root.

Could it be possible that the boot scripts on your filesystem are somehow shutting down
the ethernet interface? Might be worth checking on a fresh filesystem.

From your log I can see, that you are using a slightly different board

Hi Roger,

Quoted 5 lines · Roger Quadros

Stefan,

I tried it out today and it worked 5/5 times. I used a bare bones debian root filesystem for
the NFS root.

Thanks for testing.

Quoted 3 lines

Could it be possible that the boot scripts on your filesystem are somehow shutting down
the ethernet interface? Might be worth checking on a fresh filesystem.

I'm pretty sure that its not a problem with the filesystem. I'm using exactly this filesystem for multiple other ARM based boards as well. And all other boards (most of them with buildin ethernet controller instead of USB connected) have no problem mounting/booting from it.

BTW: The filesystem is from the ELDK 5.3 [1][2]. You could download/install it yourself.

From your log I can see, that you are using a slightly different board than I am ("Beagle xM Rev C" vs. "Beagle Rev C1/C2/C3"). So its a

different OMAP variant. Could this perhaps make a difference? Do you by chance have the "old" Beagleboard (non-xM) at hand for another test?

Quoted 8 lines

Quoted 3 lines

Could it be possible that the boot scripts on your filesystem are somehow shutting down
the ethernet interface? Might be worth checking on a fresh filesystem.

I'm pretty sure that its not a problem with the filesystem. I'm using exactly this filesystem for multiple other ARM based boards as well. And all other boards (most of them with buildin ethernet controller instead of USB connected) have no problem mounting/booting from it.

OK, then it might be something to do with beagle.
Is pandaboard working fine for you?

Quoted 2 lines

BTW: The filesystem is from the ELDK 5.3 [1][2]. You could download/install it yourself.

OK. Will give it a try.

Quoted 3 lines

From your log I can see, that you are using a slightly different board than I am ("Beagle xM Rev C" vs. "Beagle Rev C1/C2/C3"). So its a

My board is a C2.

Quoted 2 lines

different OMAP variant. Could this perhaps make a difference? Do you by chance have the "old" Beagleboard (non-xM) at hand for another test?

Sure. I have the old beagleboard with me. Let me know what to test.

My u-boot script is below and boot log is attached.

Quoted 1 line · Roger Quadros

My u-boot script is below and boot log is attached.

BTW: I noticed a git ID (8e1cacd) in your kernel that I can't identify. Do you have any patches locally applied to the kernel.org version that might be helpful with this USB issue I'm seeing?

Quoted 3 lines

BTW: I noticed a git ID (8e1cacd) in your kernel that I can't identify. Do you have any patches locally applied to the kernel.org version that might be helpful with this USB issue I'm seeing?

Yes I had tested it on a more recent kernel with some local patches applied that I've recently sent for review.

But to cross check with your setup I've retested it with commit d2b4a646 with the 2 below patches on top

https://lkml.org/lkml/2013/6/20/327
https://lkml.org/lkml/2013/6/20/339

I've tried it again with the ELDK 5.3 root but still can't get it to fail.

I think we should match our u-boot as well. Could you please let me know your u-boot commit and uEnv.txt?

My network setup is pretty simple, beagle-xm is connected directly to my laptop's ethernet port. No network switch in between.

When you get the problem do you cold boot the board or warm boot it? I've been always cold booting it.

Quoted 11 lines · Stefan Roese

Quoted 10 lines · Roger Quadros

Quoted 3 lines

BTW: I noticed a git ID (8e1cacd) in your kernel that I can't identify. Do you have any patches locally applied to the kernel.org version that might be helpful with this USB issue I'm seeing?

Yes I had tested it on a more recent kernel with some local patches applied that I've recently sent for review.

But to cross check with your setup I've retested it with commit d2b4a646 with the 2 below patches on top

Quoted 6 lines

Quoted 2 lines

https://lkml.org/lkml/2013/6/20/327
https://lkml.org/lkml/2013/6/20/339

Thanks. I have exactly those 2 patches applies as well. But on a different git commit. And with some local (unrelated) patches as well. I'll revert to exactly your version tomorrow and test again.

Okay. I changed to exactly this version: d2b4a646 with those 2 patches on top.

Quoted 8 lines

Quoted 4 lines

I've tried it again with the ELDK 5.3 root but still can't get it to fail.

I think we should match our u-boot as well. Could you please let me know your u-boot commit and uEnv.txt?

Good idea. I'm currently using a "dirty" local version. I'll try to switch to an officially available version tomorrow. And send you the version/git id.

I updated my U-Boot to the following version: Based on mainline git commit ID 225fd8c. Only this patch added:

http://patchwork.ozlabs.org/patch/256702/

This is needed for USB to work in U-Boot. Otherwise my external USB SMSC95xx will not be detected in U-Boot.

Quoted 7 lines

Quoted 3 lines

Good idea. I'm currently using a "dirty" local version. I'll try to switch to an officially available version tomorrow. And send you the version/git id.

I updated my U-Boot to the following version: Based on mainline git commit ID 225fd8c. Only this patch added:

Quoted 4 lines

http://patchwork.ozlabs.org/patch/256702/

This is needed for USB to work in U-Boot. Otherwise my external USB SMSC95xx will not be detected in U-Boot.

I can only test on beagle-xm (which is also rev. C2) with integrated SMSC ethernet chip as I don't have an external USB-ethernet adapter.

The above commit doesn't work well for beagle-xm,

so my u-boot is on tag v2013.04
and you don't need that additional patch.

Can you please retest on it? Thanks.

You will need to set "usbethaddr" before starting USB else SMSC will not work. e.g.

	setenv usbethaddr A6:66:4D:B0:69:04

Also could you please increase the Linux console message verbosity to debug (8), so that all kernel messages are visible in the log? If it spits out something more than what you sent last time then do send it to me.

Let's first sort out the beagle-xm problem, then we can see about old beagle.

Roger,

Quoted 12 lines · Roger Quadros

Quoted 11 lines

Quoted 3 lines

Good idea. I'm currently using a "dirty" local version. I'll try to switch to an officially available version tomorrow. And send you the version/git id.

I updated my U-Boot to the following version: Based on mainline git commit ID 225fd8c. Only this patch added:

http://patchwork.ozlabs.org/patch/256702/

This is needed for USB to work in U-Boot. Otherwise my external USB SMSC95xx will not be detected in U-Boot.

Quoted 2 lines

I can only test on beagle-xm (which is also rev. C2) with integrated SMSC ethernet chip as I don't have an external USB-ethernet adapter.

Too bad. This difference in boards makes the tests a bit less meaningful. Is there really no USB-ethernet adapter somewhere in your department (perhaps a non-SMSC95xx might serve as well?)? I might send/order you one if necessary. ;)

Quoted 2 lines

The above commit doesn't work well for beagle-xm,

Why is this? What does happen? Does U-Boot not boot up to the prompt? Please send me a log.

Quoted 5 lines

so my u-boot is on tag v2013.04
and you don't need that additional patch.

Can you please retest on it? Thanks.

Done, please see below.

Quoted 3 lines

You will need to set "usbethaddr" before starting USB else SMSC will not work. e.g.

	setenv usbethaddr A6:66:4D:B0:69:04

I don't have a beagle-xm, only my beable. And my U-Boot network setup is already working, thanks.

Quoted 4 lines

Also could you please increase the Linux console message verbosity to debug (8), so that all kernel messages are visible in the log? If it spits out something more than what you sent last time then do send it to me.

I added "ignore_loglevel" to the cmdline. But I cannot see anything interesting that has not been there before with "debug" as well.

Quoted 7 lines

Quoted 2 lines

I can only test on beagle-xm (which is also rev. C2) with integrated SMSC ethernet chip as I don't have an external USB-ethernet adapter.

Too bad. This difference in boards makes the tests a bit less meaningful. Is there really no USB-ethernet adapter somewhere in your department (perhaps a non-SMSC95xx might serve as well?)? I might send/order you one if necessary. ;)

Thanks for the offer :). I'll just get one for myself by today/tomorrow. The one available here uses this driver "drivers/net/usb/asix.c"

Quoted 5 lines

Quoted 2 lines

The above commit doesn't work well for beagle-xm,

Why is this? What does happen? Does U-Boot not boot up to the prompt? Please send me a log.

The last time it had failed I remember seeing a lot of IO pad configuration errors. But I'm no longer able to reproduce that problem. Maybe it was just a messed up build. In case I observe it again I'll let you know.

Quoted 12 lines

Quoted 5 lines

so my u-boot is on tag v2013.04
and you don't need that additional patch.

Can you please retest on it? Thanks.

Done, please see below.

Quoted 3 lines

You will need to set "usbethaddr" before starting USB else SMSC will not work. e.g.

	setenv usbethaddr A6:66:4D:B0:69:04

Quoted 2 lines

I don't have a beagle-xm, only my beable. And my U-Boot network setup is already working, thanks.

Right. you won't need that for beagle. It was only for beagle-xm or panda with on board usb-ethernet.

Quoted 8 lines · Roger Quadros

Quoted 7 lines

Quoted 2 lines

I can only test on beagle-xm (which is also rev. C2) with integrated SMSC ethernet chip as I don't have an external USB-ethernet adapter.

Too bad. This difference in boards makes the tests a bit less meaningful. Is there really no USB-ethernet adapter somewhere in your department (perhaps a non-SMSC95xx might serve as well?)? I might send/order you one if necessary. ;)

Quoted 2 lines

Thanks for the offer :). I'll just get one for myself by today/tomorrow. The one available here uses this driver "drivers/net/usb/asix.c"

Great.

Quoted 9 lines · Stefan Roese

Quoted 8 lines · Roger Quadros

Quoted 7 lines

Quoted 2 lines

I can only test on beagle-xm (which is also rev. C2) with integrated SMSC ethernet chip as I don't have an external USB-ethernet adapter.

Too bad. This difference in boards makes the tests a bit less meaningful. Is there really no USB-ethernet adapter somewhere in your department (perhaps a non-SMSC95xx might serve as well?)? I might send/order you one if necessary. ;)

Quoted 5 lines

Quoted 2 lines

Thanks for the offer :). I'll just get one for myself by today/tomorrow. The one available here uses this driver "drivers/net/usb/asix.c"

Great.

OK. Finally managed to reproduce the problem.

Don't know the root cause yet. Will need to investigate.

Hi Roger,

Quoted 16 lines · Roger Quadros

Quoted 14 lines · Stefan Roese

Quoted 10 lines · Roger Quadros

Quoted 7 lines

Quoted 2 lines

I can only test on beagle-xm (which is also rev. C2) with integrated SMSC ethernet chip as I don't have an external USB-ethernet adapter.

Too bad. This difference in boards makes the tests a bit less meaningful. Is there really no USB-ethernet adapter somewhere in your department (perhaps a non-SMSC95xx might serve as well?)? I might send/order you one if necessary. ;)

Thanks for the offer :). I'll just get one for myself by today/tomorrow. The one available here uses this driver "drivers/net/usb/asix.c"

Great.

Quoted 1 line

OK. Finally managed to reproduce the problem.

Ahh, great. Thanks for the effort.

Quoted 1 line

Don't know the root cause yet. Will need to investigate.

Thanks. Just let me know if I can be of any assistance (testing etc).