Hi
I'm trying to use my snowball for development testing. It seems that having kernel command line specify ip=dhpc is evil as that makes all of the userspace network configuration not to happen.
When this happens ifup scripts are not executed and ntpdate is never invoked. I have not debugged this correctly yet but that seems to be the case for me.
Thanks ZK
On Tue, Mar 13, 2012, Zygmunt Krynicki wrote:
I'm trying to use my snowball for development testing. It seems that having kernel command line specify ip=dhpc is evil as that makes all of the userspace network configuration not to happen.
When this happens ifup scripts are not executed and ntpdate is never invoked. I have not debugged this correctly yet but that seems to be the case for me.
ip=dhcp is typically meant for NFS root, or sometimes when you're just starting an initrd and would like the kernel to configure the network interface for you, but it's usually a bad idea for anything else; would be best to stop passing this and use the network setup of the rootfs.
W dniu 13.03.2012 14:35, Loïc Minier pisze:
On Tue, Mar 13, 2012, Zygmunt Krynicki wrote:
I'm trying to use my snowball for development testing. It seems that having kernel command line specify ip=dhpc is evil as that makes all of the userspace network configuration not to happen.
When this happens ifup scripts are not executed and ntpdate is never invoked. I have not debugged this correctly yet but that seems to be the case for me.
ip=dhcp is typically meant for NFS root, or sometimes when you're just starting an initrd and would like the kernel to configure the network interface for you, but it's usually a bad idea for anything else; would be best to stop passing this and use the network setup of the rootfs.
Right. I've found that this option is somehow provided by default on my V1 (no typo here) snowball with the current desktop image. It would be prudent to check if any other boards are affected.
On Tue, Mar 13, 2012 at 2:37 PM, Zygmunt Krynicki < zygmunt.krynicki@linaro.org> wrote:
W dniu 13.03.2012 14:35, Loïc Minier pisze:
On Tue, Mar 13, 2012, Zygmunt Krynicki wrote:
I'm trying to use my snowball for development testing. It seems that having kernel command line specify ip=dhpc is evil as that makes all of the userspace network configuration not to happen.
When this happens ifup scripts are not executed and ntpdate is never invoked. I have not debugged this correctly yet but that seems to be the case for me.
ip=dhcp is typically meant for NFS root, or sometimes when you're just starting an initrd and would like the kernel to configure the network interface for you, but it's usually a bad idea for anything else; would be best to stop passing this and use the network setup of the rootfs.
Right. I've found that this option is somehow provided by default on my V1 (no typo here) snowball with the current desktop image. It would be prudent to check if any other boards are affected.
That's part of the hwpack meta info, no? If so, bark at Ricardo's tree (CCed) :). I agree that adding ip=dhcp by default feels incorrect, but I don't know the background...
W dniu 13.03.2012 14:49, Alexander Sack pisze:
On Tue, Mar 13, 2012 at 2:37 PM, Zygmunt Krynicki <zygmunt.krynicki@linaro.org mailto:zygmunt.krynicki@linaro.org> wrote:
W dniu 13.03.2012 14:35, Loïc Minier pisze: On Tue, Mar 13, 2012, Zygmunt Krynicki wrote: I'm trying to use my snowball for development testing. It seems that having kernel command line specify ip=dhpc is evil as that makes all of the userspace network configuration not to happen. When this happens ifup scripts are not executed and ntpdate is never invoked. I have not debugged this correctly yet but that seems to be the case for me. ip=dhcp is typically meant for NFS root, or sometimes when you're just starting an initrd and would like the kernel to configure the network interface for you, but it's usually a bad idea for anything else; would be best to stop passing this and use the network setup of the rootfs. Right. I've found that this option is somehow provided by default on my V1 (no typo here) snowball with the current desktop image. It would be prudent to check if any other boards are affected.That's part of the hwpack meta info, no? If so, bark at Ricardo's tree (CCed) :). I agree that adding ip=dhcp by default feels incorrect, but I don't know the background...
I'll try to figure out where it comes from. If anyone wants to help have a look at this:
https://pastebin.linaro.org/433/
On Tue, Mar 13, 2012 at 11:20 AM, Zygmunt Krynicki zygmunt.krynicki@linaro.org wrote:
W dniu 13.03.2012 14:49, Alexander Sack pisze:
On Tue, Mar 13, 2012 at 2:37 PM, Zygmunt Krynicki <zygmunt.krynicki@linaro.org mailto:zygmunt.krynicki@linaro.org> wrote:
W dniu 13.03.2012 14:35, Loïc Minier pisze:
On Tue, Mar 13, 2012, Zygmunt Krynicki wrote:
I'm trying to use my snowball for development testing. It seems that having kernel command line specify ip=dhpc is evil as that makes all of the userspace network configuration not to happen.
When this happens ifup scripts are not executed and ntpdate is never invoked. I have not debugged this correctly yet but that seems to be the case for me.
ip=dhcp is typically meant for NFS root, or sometimes when you're just starting an initrd and would like the kernel to configure the network interface for you, but it's usually a bad idea for anything else; would be best to stop passing this and use the network setup of the rootfs.
Right. I've found that this option is somehow provided by default on my V1 (no typo here) snowball with the current desktop image. It would be prudent to check if any other boards are affected.
That's part of the hwpack meta info, no? If so, bark at Ricardo's tree (CCed) :). I agree that adding ip=dhcp by default feels incorrect, but I don't know the background...
I'll try to figure out where it comes from. If anyone wants to help have a look at this:
From your /proc/cmdline it's clear that u-boot/kernel didn't load your
boot script.
Can you check your u-boot's env and also give the full dmesg output?
For newer u-boot I believe you could try latest u-boot-linaro, as John is working on doing the proper support for Snowball this cycle. Check with him to see if it'd also work on a V1.
Thanks,
On Tue, Mar 13, 2012, Alexander Sack wrote:
That's part of the hwpack meta info, no? If so, bark at Ricardo's tree (CCed) :). I agree that adding ip=dhcp by default feels incorrect, but I don't know the background...
It might also be set in the default u-boot environment, or the absolute default cmdline is the on the kernel config.
W dniu 13.03.2012 15:47, Loïc Minier pisze:
On Tue, Mar 13, 2012, Alexander Sack wrote:
That's part of the hwpack meta info, no? If so, bark at Ricardo's tree (CCed) :). I agree that adding ip=dhcp by default feels incorrect, but I don't know the background...
It might also be set in the default u-boot environment, or the absolute default cmdline is the on the kernel config.
This is what the bootloader tells me:
U8500 $ version
U-Boot 2009.11-dirty (jun 06 2011 - 12:30:05)
Can I upgrade this bootloader to anything on V1 board?
On Tue, Mar 13, 2012 at 3:47 PM, Loïc Minier loic.minier@linaro.org wrote:
On Tue, Mar 13, 2012, Alexander Sack wrote:
That's part of the hwpack meta info, no? If so, bark at Ricardo's tree (CCed) :). I agree that adding ip=dhcp by default feels incorrect, but I don't know the background...
It might also be set in the default u-boot environment, or the absolute default cmdline is the on the kernel config.
I tried to dive, but couldn't spot a file that contains "boot arguments" for the snowball hwpacks in the linaro-images project... so I guess it's coming from the uboot/kernel default. (where in the hwpack config would I see meta info about extra cmdline anyway?)
Zyga: guess, file a bug and target against linaro-ubuntu and https://launchpad.net/igloocommunity... Thanks!
Ricardo: I think adding a test case for /proc/cmdline to our monthly sign off plan makes sense. Not sure how to make that easy for QA engineer/automation to catch though.
On Tue, Mar 13, 2012, Alexander Sack wrote:
I tried to dive, but couldn't spot a file that contains "boot arguments" for the snowball hwpacks in the linaro-images project... so I guess it's coming from the uboot/kernel default. (where in the hwpack config would I see meta info about extra cmdline anyway?)
I guess if you didn't find it in a text file such as the manifest, then it comes from u-boot or the kernel defaults indeed.
Zygmunt, is it pulling this from your SD image, or from the emmc image that shipped with the board? On Mar 13, 2012 10:23 AM, "Loïc Minier" loic.minier@linaro.org wrote:
On Tue, Mar 13, 2012, Alexander Sack wrote:
I tried to dive, but couldn't spot a file that contains "boot arguments" for the snowball hwpacks in the linaro-images project... so I guess it's coming from the uboot/kernel default. (where in the hwpack config would I see meta info about extra cmdline anyway?)
I guess if you didn't find it in a text file such as the manifest, then it comes from u-boot or the kernel defaults indeed.
-- Loďc Minier
W dniu 13.03.2012 17:24, Paul Larson pisze:
Zygmunt, is it pulling this from your SD image, or from the emmc image that shipped with the board?
I've pasted the contents of all filesystems earlier in this thread. I believe it is coming from the uboot binary or the kernel. u-boot is on the emmc and is "old" 2011.09. The kernel is current and is loaded from SD
The snowball boot is a bit weird, and since you have a very old board (V1) I suspect you won't be able to reflash the emmc with anything sensible. As far as I'm aware, STE dropped support for emmc flash for any board < V3.
Dave
Dave Pigott Validation Engineer T: +44 1223 45 00 24 | M +44 7940 45 93 44 Linaro.org │ Open source software for ARM SoCs Follow Linaro: Facebook | Twitter | Blog
On 13 Mar 2012, at 17:42, Zygmunt Krynicki wrote:
W dniu 13.03.2012 17:24, Paul Larson pisze:
Zygmunt, is it pulling this from your SD image, or from the emmc image that shipped with the board?
I've pasted the contents of all filesystems earlier in this thread. I believe it is coming from the uboot binary or the kernel. u-boot is on the emmc and is "old" 2011.09. The kernel is current and is loaded from SD
-- Zygmunt Krynicki Linaro Validation Team
linaro-validation mailing list linaro-validation@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-validation
On Tue, Mar 13, 2012 at 11:55 AM, Alexander Sack asac@linaro.org wrote:
On Tue, Mar 13, 2012 at 3:47 PM, Loïc Minier loic.minier@linaro.org wrote:
On Tue, Mar 13, 2012, Alexander Sack wrote:
That's part of the hwpack meta info, no? If so, bark at Ricardo's tree (CCed) :). I agree that adding ip=dhcp by default feels incorrect, but I don't know the background...
It might also be set in the default u-boot environment, or the absolute default cmdline is the on the kernel config.
I tried to dive, but couldn't spot a file that contains "boot arguments" for the snowball hwpacks in the linaro-images project... so I guess it's coming from the uboot/kernel default. (where in the hwpack config would I see meta info about extra cmdline anyway?)
Zyga: guess, file a bug and target against linaro-ubuntu and https://launchpad.net/igloocommunity... Thanks!
Ricardo: I think adding a test case for /proc/cmdline to our monthly sign off plan makes sense. Not sure how to make that easy for QA engineer/automation to catch though.
Not so sure if we should be checking /proc/cmdline directly, as usually what makes sense is to test if the image is not affected by a broken kernel argument or similar (like broken video decode or gles by lack of proper parameters).
Cheers,
linaro-validation@lists.linaro.org
-
Alexander Sack -
Dave Pigott -
Loïc Minier -
Paul Larson -
Ricardo Salveti -
Zygmunt Krynicki