Hi Tom and Zhiwei
I think I managed to replicate your problem(s). It seems tmpfs is not
being mounted correctly, which also explains your ntpd problems. The
kernel then falls-back in a failsafe mode with root mounted ro. I
suspect you have an old kernel without tmpfs support (see my email on
01 December, http://www.mail-archive.com/[email protected]/msg01058.html
with full list of updates). Try:
http://casper.berkeley.edu/svn/trunk/roach/sw/binaries/linux/uImage-20091006-mmcfix
.
Also, the usbboot string was still bad. I've updated the wiki. You want:
setenv usbboot setenv bootargs console=ttyS0,115200 mtdparts=\$\
{partitions\} rootdelay=8 root='/dev/sda1 rw'\; bootm 0xfc000000
The fact that your bootargs weren't being honoured gave-away the error
in the usbboot string and the give-away with the kernel error was this
error message:
mount: wrong fs type, bad option, bad superblock on tmpfs,
missing codepage or other error
In some cases useful info is found in syslog - try
dmesg | tail or so
Apologies for the ethaddr=xx. You are correct, there should not be an
equals sign. Also, apologies wrt the confusion with the setenv
bootargs command. The proposed command would have changed the default
bootargs, but the usbboot macro overwrites it. If you want to boot
with the defaults, do a "run boot". To explain...
If you look at your environment (printenv), it should look something
like this:
bootdelay=10
baudrate=115200
loads_echo=
preboot=run initboot
hostname=roach
netdev=eth0
init_eeprom=echo Programming EEPROM for configuration H;imw 0x52 0.1
87; imw 0x52 1.1 78; imw 0x52 2.1 52; imw 0x51
bit=report clear;check serial-output echo serial check;check cpld-
probe roachcpld;mw.l 0x201000 0 400; check mem-zt
kernel_addr=0xfc000000
partitions=physmap-flash.0:1792k(linux),2...@0x1c0000(fdt),
81...@0x200000(root),546...@0xa00000(usr),2...@0x3f60000
boot=bootm ${kernel_addr}
bootargs=console=ttyS0,115200 mtdparts=physmap-flash.0:1792k(linux),
2...@0x1c0000(fdt),81...@0x200000(root),54656k2
yget=loady 0x200000
update=protect off 0xfffa0000 0xffffffff;era 0xfffa0000
0xffffffff;cp.b 0x200000 0xfffa0000 ${filesize};protect onf
soloboot=setenv bootargs console=ttyS0,115200 mtdparts=${partitions}
root=/dev/mtdblock2; bootm 0xfc000000
usbboot=setenv bootargs console=ttyS0,115200 mtdparts=${partitions}
rootdelay=8 root=/dev/sda1; bootm 0xfc000000
netboot=dhcp 0x400000; setenv bootargs console=ttyS0,115200 mtdparts=$
{partitions} root=${rootpath} ip=dhcp; bootm0
newuboot=cp.b 0xfffa0000 0x200000 0x60000; run yget update
newkernel=run yget; era 0xfc000000 0xfc1bffff; cp.b 0x200000
0xfc000000 0x1c0000
newroot=run yget; era 0xfc200000 0xfc9fffff; cp.b 0x200000 0xfc200000
${filesize}
initboot=echo; echo type run netboot to boot via dhcp+tftp+nfs; echo
type run soloboot to run from flash without
clearenv=erase 0xfff60000 0xfff9ffff
ethaddr=02:00:00:02:02:17
ethact=ppc_4xx_eth0
bootcmd=run netboot
mem=524264k
ver=U-Boot 2008.10-svn2226 (Nov 13 2009 - 09:27:45)
The ROACH uboot boot procedure is something like this (and you'll see
this macros in the printenv listing above):
1) uboot runs "preboot".
2) "preboot" is a macro which runs "initboot".
3) "initboot" is another macro which brings up the boot options
list. ROACH then waits "bootdelay" time and then runs the "bootcmd"
macro.
4) "bootcmd" defaults to booting off flash at 0xfc000000. You can
change this default to any one of four boot options by doing a "setenv
bootcmd XXXXX":
The four boot options (usb, mmc/sd, network, flash) are all also
macros. They generally change environment variables as required and
then execute a boot command (eg from onboard flash at 0xfc000000),
with boot arguments from the "bootargs" uboot environment variable.
Hope this sorts you out.
Jason
On 15 Jan 2010, at 23:36, Tom Downes wrote:
Jason:
I continue to be baffled. I tried your instructions, including the
clearenv, except for 3 changes:
1. You had "setenv ethaddr=XX:XX..." when there should be no equal
sign.
2. I did not default to usbboot.
3. I modified your setenv line to be preceded by "setenv usbboot" so
that when I run usbboot it follows those instructions. I did
saveenv the changes to memory.
It continues to mount the old USB stick rw and the new USB stick
(same model so far as I can tell) ro. I am running 2.6.25-svn2338
and etch from 11/30 on the new stick. I don't know what filesystem
is on the old stick.
Here is the output of mount and the contents of fstab after boot on
the old USB:
# mount
rootfs on / type rootfs (rw)
/dev/root on / type ext2 (rw,errors=continue)
proc on /proc type proc (rw)
devpts on /dev/pts type devpts (rw,noexec,mode=620)
tmpfs on /var/tmp type tmpfs (rw,size=8192k)
usbfs on /proc/bus/usb type usbfs (rw,nodev,noexec)
/dev/mtdblock3 on /mnt/flash type jffs2 (rw)
# less /etc/fstab
proc /proc proc defaults 0 0
devpts /dev/pts devpts mode=620,noexec 0 0
tmpfs /var/tmp tmpfs size=8M 0 0
usbfs /proc/bus/usb usbfs noexec,nodev 0 0
/dev/mtdblock3 /mnt/flash jffs2 defaults 0 0
And the new USB. Enablign noatime makes sense to me, but it is
different from the old USB. Otherwise it is only the "ro" that is
different for root and the various other filesystems are a bit
different.
# mount
rootfs on / type rootfs (rw)
/dev/root on / type ext2 (ro,noatime,errors=continue)
proc on /proc type proc (rw)
sysfs on /sys type sysfs (rw)
devpts on /dev/pts type devpts (rw,noexec,mode=620)
usbfs on /proc/bus/usb type usbfs (rw,nodev,noexec)
/dev/mtdblock3 on /mnt/flash type jffs2 (rw)
tmpfs0 on /var/tmp type tmpfs (rw,size=32768k)
tmpfs1 on /tmp type tmpfs (rw,size=16384k)
# less /etc/fstab
proc /proc proc defaults 0 0
sysfs /sys sysfs defaults 0 0
devpts /dev/pts devpts mode=620,noexec 0 0
usbfs /proc/bus/usb usbfs noexec,nodev 0 0
/dev/mtdblock3 /mnt/flash jffs2 defaults 0 0
I have tried the following (while mounted ro) on the new USB drive
and rebooted, but no luck.
tune2fs -i 0 /dev/sda1
tune2fs -c 0 /dev/sda1
tune2fs -e continue /dev/sda1
One thing to note is that the rootdelay=8 argument is not always
respected. Sometimes it waits for a couple seconds (not 8) and
others no time at all.
I strongly suspect something very simple with the kernel respecting
something about the ext2 filesystem. For each USB stick I get
warnings that the drive has not been checked no matter how many
times I run e2fsck with combinations of -f and -p.
Tom
On 15 Jan 2010, at 08:08, Tom Downes wrote:
Thanks - I will try the environment variable changes less booting to
USB by default until I actually see it work. I see that you changed
the wiki.
I have successfully remounted it in the past - the problem is that
a number of the init.d scripts, including if-up and ntpd fail due to
not being able to write to disk. So it's not really a usable system
until you go back and basically enter runlevel 5 manually. This is
true when using the latest stock etch distribution with 0
modifications. I think the default behavior of USB booting ought to
be re-examined in that context.
USB drives with write protect switches are reasonably common and
could provide the same functionality.
Thanks to all the list for the help.
Tom
On Thu, Jan 14, 2010 at 9:28 PM, Jason Manley <[email protected]
> wrote:
I think the key error is that you weren't escaping your new commands.
But I suggest you start from scratch. Erase any changes you may have
made and return to defaults (this will restore your apparent missing
'partitions' variable, which automounts the onboard flash memory):
* run clearenv
* reset
* setenv ethaddr=02:00:00:xx:xx:xx
(where xx:xx:xx is your board's serial number)
* saveenv
* reset
Now your board is back to the way it was when shipped from the
factory.
To tell it to boot to usb by default...
* setenv bootcmd run usbboot
To set the default kernel boot arguments (according to the wiki)...
* setenv bootargs console=ttyS0,115200 mtdparts=\$\{partitions\}
rootdelay=8 root='/dev/sda1 rw'\; bootm 0xfc000000
* saveenv
I think the problem might be that the command on the wiki doesn't
escape the semi colons, dollar signs or curly-braces, so uboot
interprets these as separate commands.
However, Marc at KAT suggests that you rather start with it mounted
RO (the default) and then remount it after the kernel's started. He
considers this a "safer" option.
To do this manually, once booted into linux, do
* mount -o remount,rw /
Apparently you can make that change automatic on boot in fstab,
though I haven't tried this myself.
Jason
On 15 Jan 2010, at 03:22, Mark Wagner wrote:
Hi Tom,
Have any of these permutations worked for you? I've gotten it to
boot with the command I sent you earlier, but Jason also mentioned
to me awhile ago that editing the fstab file would work too, but I
haven't tried that yet.
Mark
On Thu, Jan 14, 2010 at 3:23 PM, Tom Downes <[email protected]>
wrote:
Mark:
I think the problem might be more basic. I have not found clear
instructions on the precise command that one is supposed to run to
change the usbboot environment variable. It appears to be csh-like
environment that one is greeted with when you break the automatic
boot process.
I think the problem that I was having before was that when I ran:
setenv usbboot bootargs console=ttyS0,115200 mtdparts=${partitions}
rootdelay=8 root=/dev/sda1 rw; bootm fc000000
it interpreted "bootm fc000000" as a 2nd command rather than part of
the setenv. It also seems like there should be a second setenv
after usbboot so that usbboot is not so much an environment variable
itself as a shorthand for setting the bootargs environment variable
to values appropriate for booting by USB.
So... I have tried a number of permutations of the below code,
putting in single/double quotes all around (to manipulate whether
the $ is interpreted literally or whether its value is taken in),
including where you suggest and around ${partitions}. Putting
single or double quotes around the whole thing allows the setenv
commad to be interpreted as a single command rather than one setenv
followed by a bootm.
Variation of what I have tried:
setenv usbboot 'setenv bootargs console=ttyS0,115200 mtdparts=$
{partitions} rootdelay=8 root=/dev/sda1 rw; bootm fc000000'
Sorry for being obtuse, but I'm not sure this booting environment is
really csh/tcsh or just some extremely basic CLI.
Tom
On Thu, Jan 14, 2010 at 2:46 PM, Mark Wagner <[email protected]
> wrote:
Hi Tom,
I needed to change root=/dev/sda1 rw to root='/dev/sda1 rw' (add
qoutes) in the command below to get the usb filesystem mounted read/
write.
Mark
On Thu, Jan 14, 2010 at 2:39 PM, Tom Downes <[email protected]>
wrote:
Yes, it boots read-only even trying to set usbboot as below.
Though, as I said, I cannot set mtdparts=${partitions}, I have to
remove the dollar sign. If that is important then I'm not sure what
to do because if I leave the dollar sign in it immediately boots and
not to USB.
setenv usbboot bootargs console=ttyS0,115200 mtdparts=${partitions}
rootdelay=8 root=/dev/sda1 rw; bootm fc000000
I'll look into things more this afternoon.
Tom
On Tue, Jan 12, 2010 at 8:22 AM, Zhiwei Liu <[email protected]>
wrote:
Tom,
Do you mean it still mounts the new filesystem read-only even if you
change the usbboot environment to "bootargs console=ttyS0,115200
mtdparts=${partitions}
rootdelay=8 root=/dev/sda1 rw; bootm fc000000"?
I wish I could play with the environment a little bit, but I made it
to boot off the USB stick automatically.
I can not interrupt the standard boot process at uboot. I'm still
struggling to bring it back to the standard
boot process, It seems like I have to reload the uboot to overwrite
the 'bootargs' environment.
FYR, I also had some warnings when booting.
--------------------------------------------------
EXT2-fs warning: mounting unchecked fs, running e2fsck is recommended
VFS: Mounted root (ext2 filesystem).
Freeing unused kernel memory: 136k init
INIT: version 2.86 booting
warning: can't open /etc/mtab: No such file or directory
mount: wrong fs type, bad option, bad superblock on tmpfs,
missing codepage or other error
In some cases useful info is found in syslog - try
dmesg | tail or so
---------------------------------------------------
I don't know how to fix that.
Zhiwei
On Tue, Jan 5, 2010 at 5:18 PM, Tom Downes <[email protected]>
wrote:
Zhiwei (and others):
We are having the same problem as you upon upgrading to the 20091006
kernel and the 20091130 etch file system on USB. We have an older
filesystem (don't recall the date) on one USB stick and the new one on
another. Both USB sticks are formatted ext2 and are the same
brand/model.
For some reason, when booting by "run usbboot" it will mount the old
filesystem read-write but the new filesystem read-only. Doing a
remount fixes the problem, but of course the boot process has had a
number of failures along the way.
This is after changing the usbboot environment variable to what you
list, but when I try to do it with the dollar sign for partitions it
immediately runs. So if I remove that, it will work. Is that really
an environment variable? In any case, it works for one system, but
not the other so I'm not sure it's the issue precisely.
setenv usbboot bootargs console=ttyS0,115200 mtdparts=${partitions}
rootdelay=8 root=/dev/sda1 rw; bootm fc000000
saveenv
run usbboot
(minus the dollar sign for mtdparts)
For both USB sticks it will complain about the file systems not having
been checked. I have gone through and run "tune2fs -c 0 -i 0" on the
new stick and done a manual force check with "e2fsck -fp". On the
ROACH and on a true desktop. Yet it still gives the error. My
impression is that sometimes the kernel will mount an ext2 fs
read-only if it doesn't think that it has been checked. I would think
that this is the source of the problem were it not for the fact that
the same kernel mounts one disk read-write and the other read-only.
I usually get the date set correctly along the way and do soft reboots
so that the clock doesn't get too far off. Otherwise I would be
concerned about automatic checks for a given number of days (despite
my use of tune2fs).
In both cases it also complains about the jffs2 filesystem not
mounting properly at the very end of boot.
I'm a bit baffled...
Tom
