Stress testing a 2.6.28/rev c2 - further tests required

Hello,

Tonight, I've tried to experiment on the USB problems and understand
the kernel version discrepancies.
Using Robert Nelson kernels with an identical mmc card, uboot and
xloader between the tests:

- tried to do a stress test installing debian with a 2.6.28,

First it wasn't working on the MMC card I wanted to use (not detected,
so I didn't get the "guided" partman menu entry mentionned in the wiki
- took me a while to realise it was the 512 Mb mmc card not being
detected - got it just for the test since I had read sdhc were giving
the problem!!!)

However I could do a netcat, using a cdc ethernet device hooked to my
apple usb keyboard, itself hooked on the EHCI port. The results were
quite unexpected- the transfert didn't succeed, and after that the
keyboard experienced a strange bug (I had to press 2 times the same
key to get it to be displayed. There was no "reset" command but I
didn't play with stty and it was working fine before)

- So tried to boot a 2.6.29 on a debian rootfs created separately :
now the very same apple usb keyboard (which includes a 2.0 usb hub) is
not even detected on the USB EHCI port

If I try to plug it on the USB OTG with a mini a-a adapter, is not
working either - at least complains about "powered hub" status of the
apple keybord. The devices that were plugged in the USB OTG port
during the 2.6.28 test are not working either.

With such issues at hand, I didn't bother trying the stess test. I'll
prepare a better test protocol first because there are quite a lot of
problems as soon as you leave the angstrom way.

Even if I really need some sleep now, that's quite strange, and it
looks like both a software problem (2.6.29/2.6.28 differences on USB)
and an hardware problem (netcat shouldn't corrupt the usb keyboard)

Did anyone get similar results ?

Guylhem

Hi Guylhem,

Here's a couple extra data points for you...

While testing/validiating the elinux wiki debian/ubuntu installs
guides this week on RevB5 & RevC2 hardware for 2.6.29-oer31. I found
2.6.29-oer31 would boot up my RevB5 board with a non working usb-otg
port. After some trial and error it turned out this was due to the
RevC u-boot validation image i had installed on it. I reflashed the
latest Angstrom u-boot* image from the dev tree and haven't seen any
usb issues since.

Otherwise up until the end of this week, i was running 2.6.28-oer17
24/7 on a RevB5 & RevC2

RevB5: u-boot:2009(2) :

OS: Debian Squeeze
MMC Card: boot.scr & uImage based on
http://www.rcn-ee.com/deb/kernel/beagle/squeeze/v2.6.28-79d042a-oer17/linux-image-2.6.28-oer17_1.0squeeze_armel.deb
USB-Harddrive: Debian root filesystem
Typical Usage: Downloads 75Mb, spends 2.5 hours building kernel deb's,
upload final deb, repeat...

Usb Devices:
Bus 002 Device 003: ID 1058:0901 Western Digital Technologies, Inc.
MyBook External HDD
Bus 002 Device 002: ID 05e3:0606 Genesys Logic, Inc. USB 2.0 Hub /
D-Link DUB-H4 USB 2.0 Hub
Bus 002 Device 004: ID 9710:7830 MosChip Semiconductor MCS7830 Ethernet
Bus 002 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub

RevC2: uboot (RevC validation image)
OS: Fedora 10
MMC Card: Cross Compiled uImage
http://www.rcn-ee.com/deb/kernel/CC-v2.6.28-79d042a-oer17
USB-Harddrive: Fedora root filesystem
Typical Usage. rpmbuild 24/7, attempting to build all java packages...

Usb Devices- OTG port. (haven't tested the ehci port)
Currently down, but same as above, but with a 500Gb Western Digital HDD..

*2.6.29-oer31
Prebuilt: http://www.rcn-ee.com/deb/kernel/CC-beagle-v2.6.29-58cf2f1-oer31
Patch: http://rcn-ee.homeip.net:81/dl/omap/beagle/v2.6.29-58cf2f1-oer31/v2.6.29-58cf2f1-oer31.diff
Defconfig: http://rcn-ee.homeip.net:81/dl/omap/beagle/v2.6.29-58cf2f1-oer31/defconfig
Basedon: http://cgit.openembedded.net/cgit.cgi?url=openembedded/commit/&id=2747c78c7c545a1188cfb255603f20e8ebf67475

u-boot*
Prebuilt: http://rcn-ee.homeip.net:81/dl/omap/uboot/u-boot-beagleboard-2009.03+r22+gitr2dea1db2a3b7c12ed70bbf8ee50755089c5e5170-r22.bin

u-boot(2)*
Prebuilt: http://rcn-ee.homeip.net:81/dl/omap/uboot/u-boot-beagle-04feb2009.bin

Regards,

Hello

While testing/validiating the elinux wiki debian/ubuntu installs
guides this week on RevB5 & RevC2 hardware for 2.6.29-oer31. I found
2.6.29-oer31 would boot up my RevB5 board with a non working usb-otg
port. After some trial and error it turned out this was due to the
RevC u-boot validation image i had installed on it. I reflashed the
latest Angstrom u-boot* image from the dev tree and haven't seen any
usb issues since.

Got http://rcn-ee.homeip.net:81/dl/omap/uboot/u-boot-beagleboard-2009.03+r22+gitr2dea1db2a3b7c12ed70bbf8ee50755089c5e5170-r22.bin

I now have:

X-Loader 1.41
MD5 (u-boot-revc2.bin) = 4e481709ae19b9a4515089c1a8909072
version says:
U-Bot 2009.03 (Apr 21 2009 - 20:27:07)

Yet the EHCI port is not detecting the keyboard that is detected with 2.6.28

Even better, using the revc2 validation image from
http://code.google.com/p/beagleboard/wiki/BeagleboardRevCValidation,
the apple keyboard on ehci is just working fine

If I take the very same environment and just replace the kernel by a
2.6.29 uImage made from linux-image-2.6.29-oer31_1.0lenny_armel.deb,
it doesn't work.

It's a kernel software problem! I'll check try to compare the
defconfigs and see what has changed in the patches.

zcat /proc/config.gz|grep EHCI
CONFIG_USB_ARCH_HAS_EHCI=y
CONFIG_USB_EHCI_HCD=y
CONFIG_OMAP_EHCI_PHY_MODE=y
# CONFIG_OMAP_EHCI_TLL_MODE is not set
CONFIG_USB_EHCI_ROOT_HUB_TT=y
# CONFIG_USB_EHCI_TT_NEWSCHED is not set

Otherwise up until the end of this week, i was running 2.6.28-oer17
24/7 on a RevB5 & RevC2

There's a problem with my mmc card, but at least EHCI is fine with it.
I'll try to find and backport the mmc patch.

BTW The exact message I get when I connect the keyboard to the OTG
port: "can't connect bus-powered hub to this port"
Somehow the OTG port is by itself considered as a bus-powered hub, so
I can't put another hub on it.

Of course, there's no power source on the apple keyboard internal hub
(!) - but it's not a big drain. IMHO there should be an option
somewhere to allow it.

Got http://rcn-ee.homeip.net:81/dl/omap/uboot/u-boot-beagleboard-2009.03+r22+gitr2dea1db2a3b7c12ed70bbf8ee50755089c5e5170-r22.bin

I now have:

X-Loader 1.41
MD5 (u-boot-revc2.bin) = 4e481709ae19b9a4515089c1a8909072
version says:
U-Bot 2009.03 (Apr 21 2009 - 20:27:07)

Yet the EHCI port is not detecting the keyboard that is detected with 2.6.28

Even better, using the revc2 validation image from
http://code.google.com/p/beagleboard/wiki/BeagleboardRevCValidation,
the apple keyboard on ehci is just working fine

If I take the very same environment and just replace the kernel by a
2.6.29 uImage made from linux-image-2.6.29-oer31_1.0lenny_armel.deb,
it doesn't work.

PS. that's the exact same as the CC-beagle* image, just built in a
lenny chroot...

Last time i tried using the ehci port on Rev C it didn't work..
(2.6.28's..) But the otg port worked fine so i just used that. If you
search the group, there's been a couple more of the same reports, but
no one is really sure what the cause is. I'll run a couple more tests
today..

There's a problem with my mmc card, but at least EHCI is fine with it.
I'll try to find and backport the mmc patch.

2.6.28-oer17 doesn't have Frans mmc patch (it came later, oer23, might)

2.6.29-oer31 has Frans mmc patch..

BTW The exact message I get when I connect the keyboard to the OTG
port: "can't connect bus-powered hub to this port"
Somehow the OTG port is by itself considered as a bus-powered hub, so
I can't put another hub on it.

Of course, there's no power source on the apple keyboard internal hub
(!) - but it's not a big drain. IMHO there should be an option
somewhere to allow it.

even if it worked with the evaluation kernel, just for testing I'd
really recommend sticking a powered usb 2.0 hub between the beagle and
any usb device. (otg or ehci)

Regards,

Last time i tried using the ehci port on Rev C it didn't work..
(2.6.28's..) But the otg port worked fine so i just used that. If you
search the group, there's been a couple more of the same reports, but
no one is really sure what the cause is. I'll run a couple more tests
today..

2.6.28-oer17(1) RevC2 EHCI - USB2.0 Hub - [WD usb harddrive, usb-ethernet]

boot log: http://pastebin.ca/1403445

[voodoo@fedora-arm ~]$ cat /proc/cpuinfo
Processor : ARMv7 Processor rev 3 (v7l)
BogoMIPS : 486.31
Features : swp half thumb fastmult vfp edsp
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x1
CPU part : 0xc08
CPU revision : 3

Hardware : OMAP3 Beagle Board
Revision : 0020
Serial : 0000000000000000

[voodoo@fedora-arm ~]$ lsusb
Bus 002 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
Bus 001 Device 004: ID 9710:7830 MosChip Semiconductor MCS7830 Ethernet
Bus 001 Device 003: ID 1058:1100 Western Digital Technologies, Inc.
Bus 001 Device 002: ID 05e3:0608 Genesys Logic, Inc. USB-2.0 4-Port HUB
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub

And with the USB-OTG port, same setup...

http://en.pastebin.ca/1403459

1: http://www.rcn-ee.com/deb/kernel/CC-v2.6.28-79d042a-oer17

Regards,

Now with 2.6.29-oer31 this is a different story...

EHCI: can't find my usb "/dev/sda1"
http://en.pastebin.ca/1403524

OTG: perfect...
http://en.pastebin.ca/1403525

Haven't found a working solution yet for 2.6.29...

btw kinda cool and repeatable with 2.6.28-oer17

OTG:
bash-3.2# hdparm -tT /dev/sda
/dev/sda:
Timing cached reads: 230 MB in 2.01 seconds = 114.41 MB/sec
Timing buffered disk reads: 20 MB in 3.08 seconds = 6.50 MB/sec

EHCI:
bash-3.2# hdparm -tT /dev/sda
/dev/sda:
Timing cached reads: 262 MB in 2.00 seconds = 130.77 MB/sec
Timing buffered disk reads: 64 MB in 3.05 seconds = 21.01 MB/sec

Regards,

Anyone else, try:
http://www.rcn-ee.com/deb/kernel/CC-beagle-v2.6.29-58cf2f1-oer31.1

the ehci.patch had become corrupted in git, this now boots on my RevC2.

Regards,

Did you try to run the Chameleon man running test? I'm wondering if
you get the same effect with USB when running that demo (I've RMA'ed
my board, awaiting repairs)...

Hello
Hello

Currently uploading a 4 Gb image from a HD on the EHCI port to a CDC
Ethernet on the OTG port.

Then I'll try doing it again at LAN speeds (uploading and downloading)

If there's a rare error, it may show up. So far, nothing. In the past,
I've had packet strangeness when plugging/unplugging a USB cable. I'll
try that too.

Guylhem

Hi,

The Chameleon man is a graphics demo that is included in the demo
image of openembedded. Just click on the start menu->applications-

demo's.

Hope to hear from you soon...

Regards, Eelco