Hi,
I've tried to use Linaro 11.08 nano image with vexpress 11.08 hwpack but the board hangs in the init script and I never get the shell prompt.
if I use the linaro 11.07 nano image with the 11.08 hwpack, the vexpress board boot successfully. Did someone else face the same problem ?
Regards, Vincent
On Mon, 2011-09-26 at 11:44 +0200, Vincent Guittot wrote:
I've tried to use Linaro 11.08 nano image with vexpress 11.08 hwpack but the board hangs in the init script and I never get the shell prompt.
I've used the pre-built 11.08 nano image [1] without problem. (Kernel booting from NOR, no initrd, filesystem image on a USB drive).
What kind of media is your root filesystem on? And how far is it getting (i.e. what's the last part of the serial output)?
Hi
On 26 September 2011 13:19, Jon Medhurst (Tixy) jon.medhurst@linaro.org wrote:
On Mon, 2011-09-26 at 11:44 +0200, Vincent Guittot wrote:
I've tried to use Linaro 11.08 nano image with vexpress 11.08 hwpack but the board hangs in the init script and I never get the shell prompt.
I've used the pre-built 11.08 nano image [1] without problem. (Kernel booting from NOR, no initrd, filesystem image on a USB drive).
What kind of media is your root filesystem on? And how far is it getting (i.e. what's the last part of the serial output)?
My kernel and initrd is on the embedded flash and my rootfs on the MMC.
hereafter is the end of my serial output:
[ 6.842961] Freeing init memory: 240K Loading, please wait... [ 7.131406] udevd[540]: starting version 173 Begin: Loading essential drivers ... done. Begin: Running /scripts/init-premount ... done. Begin: Mounting root file system ... Begin: Running /scripts/local-top ... done. [ 7.311414] atkbd serio0: keyboard reset failed on mb:kmi0 [ 8.591256] atkbd serio1: keyboard reset failed on mb:kmi1 Begin: Running /scripts/local-premount ... done. [ 40.620703] EXT3-fs: barriers not enabled [ 40.816689] kjournald starting. Commit interval 5 seconds [ 41.000219] EXT3-fs (mmcblk0p2): using internal journal [ 41.015948] EXT3-fs (mmcblk0p2): recovery complete [ 41.031153] EXT3-fs (mmcblk0p2): mounted filesystem with ordered data mode Begin: Running /scripts/local-bottom ... done. done. Begin: Running /scripts/init-bottom ... done.
Vincent
-- Tixy
[1] http://releases.linaro.org/images/linaro-n/nano/11.08/vexpress-nano.img.gz
On Mon, 2011-09-26 at 13:28 +0200, Vincent Guittot wrote:
Hi
On 26 September 2011 13:19, Jon Medhurst (Tixy) jon.medhurst@linaro.org wrote:
On Mon, 2011-09-26 at 11:44 +0200, Vincent Guittot wrote:
I've tried to use Linaro 11.08 nano image with vexpress 11.08 hwpack but the board hangs in the init script and I never get the shell prompt.
I've used the pre-built 11.08 nano image [1] without problem. (Kernel booting from NOR, no initrd, filesystem image on a USB drive).
What kind of media is your root filesystem on? And how far is it getting (i.e. what's the last part of the serial output)?
My kernel and initrd is on the embedded flash and my rootfs on the MMC.
I've heard that due to bugs all high bandwidth devices are flakey. Specifically, MMC and SATA media are unreliable and the best solution is to try USB.
I have just copied my 11.08 image to MMC and tried booting off that. I found that it does seem to hang but after 5 minutes or so it eventually gets to the command prompt. Once booted, using ssh to copy a large file gave a transfer rate of about 30kB/s.
It's a bit odd though if you find a 11.07 image works and an 11.08 image does on the same kernel and MMC card.
I've heard that due to bugs all high bandwidth devices are flakey.
They are getting better and better though :-)
Specifically, MMC and SATA media are unreliable and the best solution is to try USB.
With the latest official firmware (and IO FPGA) from the VE CD 3.0:
https://silver.arm.com/browse/VE030
MMC with the default card clock of ~500kHz is rock-solid :-) Personally I actually use it with mmci.fmax=12000000, which makes it rather fast. The downside is that I can't use any USB devices at the same time (well, you can do some tricks with IRQ core affinity to make it work). So 6000000 (6MHz) is a safe compromise.
I have just copied my 11.08 image to MMC and tried booting off that. I found that it does seem to hang but after 5 minutes or so it eventually gets to the command prompt. Once booted, using ssh to copy a large file gave a transfer rate of about 30kB/s.
With the increased MMC card clock it becomes really usable, but then you may face an issue with the Ethernet ;-)
https://bugs.launchpad.net/linux-linaro/+bug/673820
Which can be fixed with even newer IO FPGA files, as mentioned in the bug log.
Cheers!
Paweł
On Mon, 2011-09-26 at 14:28 +0100, Pawel Moll wrote:
Specifically, MMC and SATA media are unreliable and the best solution is to try USB.
With the latest official firmware (and IO FPGA) from the VE CD 3.0:
https://silver.arm.com/browse/VE030
MMC with the default card clock of ~500kHz is rock-solid :-)
That probably explains the performance I was seeing then. ;-)
It's good to hear the problems are getting fixed. I was worried when I heard the issues were hardware problems with no workarounds, didn't realise the hardware was FPGA.
Specifically, MMC and SATA media are unreliable and the best solution is to try USB.
BTW, what do you mean by "SATA media" here?
PCI Express doesn't work (broken test chip, not the motherboard itself), so you can't use PCI SATA Host Controller and there is no HC on the motherboard :-)
Cheers!
Paweł
On Mon, 2011-09-26 at 15:04 +0100, Pawel Moll wrote:
Specifically, MMC and SATA media are unreliable and the best solution is to try USB.
BTW, what do you mean by "SATA media" here?
PCI Express doesn't work (broken test chip, not the motherboard itself), so you can't use PCI SATA Host Controller and there is no HC on the motherboard :-)
I was repeating what I thought I had heard a couple of weeks ago when I was told that USB was the only reliable interface to use for a filesystem.
So I guess it must be no SATA because of broken PCI, no MMC because of FPGA bugs (now fixed), no CF card because of (?).
So I guess it must be no SATA because of broken PCI,
True.
no MMC because of FPGA bugs (now fixed),
Work-arounded ;-) (hopefully really fixed in near future).
no CF card because of (?).
CF usability depends on the card. Most of the old cards Just Work (TM). New cards tend to have some problem with writing to them (it happens 30x slower than it should). It's on my TODO list, so sooner or later it will happen :-) Ah, and you need PATA_PLATFORM enabled.
* Ethernet
Was buggered, is fixed now :-)
Cheers!
Paweł
+++ Pawel Moll [2011-09-26 15:48 +0100]:
So I guess it must be no SATA because of broken PCI,
True.
no MMC because of FPGA bugs (now fixed),
Work-arounded ;-) (hopefully really fixed in near future).
no CF card because of (?).
CF usability depends on the card. Most of the old cards Just Work (TM). New cards tend to have some problem with writing to them (it happens 30x slower than it should). It's on my TODO list, so sooner or later it will happen :-) Ah, and you need PATA_PLATFORM enabled.
- Ethernet
Was buggered, is fixed now :-)
When did it become un-buggered? Tim Klein here is trying to get said machine to boot over TFTP with NFS root and it just hangs at init. Which rootfs version and kernel are known to work?
Wookey
- Ethernet
Was buggered, is fixed now :-)
When did it become un-buggered?
According to this bug log:
https://bugs.launchpad.net/linux-linaro/+bug/673820
(my own entry ;-) it was 2011-08-31.
Tim Klein here is trying to get said machine to boot over TFTP with NFS root and it just hangs at init.
How do you know it hangs? :-) What I mean it the fact it seems to be hung after the kernel initialization is done (which is what I think you mean) suggests a problem with the boot scripts (upstart or whatever). I've observed similar behaviour with 11.06 or 11.07 when using my own kernel (no initrd) - setting CONFIG_DEVTMPFS_MOUNT=y helped and I couldn't be bothered with debugging upstart :-)
Which rootfs version and kernel are known to work?
It was a hardware issue, not a kernel/rootfs one. ARM Support has a fix and can provide to engineers on case-by-case basis. It will be also included in the next VE CD release.
Cheers!
Paweł
On Mon, Sep 26, 2011 at 03:48:09PM +0100, Pawel Moll wrote:
So I guess it must be no SATA because of broken PCI,
True.
no MMC because of FPGA bugs (now fixed),
Work-arounded ;-) (hopefully really fixed in near future).
no CF card because of (?).
CF usability depends on the card. Most of the old cards Just Work (TM). New cards tend to have some problem with writing to them (it happens 30x slower than it should). It's on my TODO list, so sooner or later it will happen :-) Ah, and you need PATA_PLATFORM enabled.
- Ethernet
Was buggered, is fixed now :-)
Cheers!
Paweł
I think that in practice USB storage tends to be fastest and most reliable. This might change in the future, but I've found that for now a USB card reader, USB stick or a real disk attached via USB works pretty well.
Admittedly though, I haven't been testing the other mechanisms too much lately.
Cheers ---Dave
-
Dave Martin -
Jon Medhurst (Tixy) -
Pawel Moll -
Vincent Guittot -
Wookey