Skip to contentSkip to conversation

Request For Help - Embedded x86, Kernel 2.6.33 boots, Mounts JFFS2 rootfs, Stops at init

DISCUSSION Graeme Russ 3 participants 6 messages Apr 28, 2010 – Apr 30, 2010

lore
Share & tools
lore
#0 lore raw

Hi All,

I've been working for a while now getting the x86 U-Boot port up to speed to boot a linux kernel

I now have a root fs (build from Linux-From-Scratch and trimmed to <15M using qemu) but all attempts to get any serial console output past launching of init have failed.

I have taken the root fs which successfully boots using qemu and created a jffs2 image and flashed it onto one of the on-board flash chips. U-Boot can successfully 'ls' the root fs

My bootags are:

root=/dev/mtdblock1 rw rootfstype=jffs2 console=ttyS0,115200

as Linux can mount the root partition:

[    1.447819] Serial: 8250/16550 driver, 6 ports, IRQ sharing disabled
[    1.450027] serial8250: ttyS0 at I/O 0x3f8 (irq = 4) is a 16550A
[    1.452421] serial8250: ttyS1 at I/O 0x2f8 (irq = 3) is a 16550A
[    1.505824] brd: module loaded
[    1.507350] Uniform Multi-Platform E-IDE driver
[    1.509245] ide-gd driver 1.18
[    1.512852] Serck eNET flash device: 0x1000000 at 0x10000000
[    1.523853] eNET_Flash_0: Found 1 x16 devices at 0x0 in 16-bit bank
[    1.524828]  Intel/Sharp Extended Query Table at 0x0031
[    1.526492]  Intel/Sharp Extended Query Table at 0x0031
[    1.527492] Using buffer write method
[    1.528242] cfi_cmdset_0001: Erase suspend on write enabled
[    1.529242] Serck eNET flash device: 0x1000000 at 0x11000000
[    1.539872] eNET_Flash_1: Found 1 x16 devices at 0x0 in 16-bit bank
[    1.540951]  Intel/Sharp Extended Query Table at 0x0031
[    1.543632]  Intel/Sharp Extended Query Table at 0x0031
[    1.544426] Using buffer write method
[    1.545495] cfi_cmdset_0001: Erase suspend on write enabled
[    1.547452] Serck eNET flash device: 0x80000 at 0x38000000
[    1.549664] Found: AMD AM29LV040B
[    1.550136] eNET_Boot_Flash: Found 1 x8 devices at 0x0 in 8-bit bank
[    1.551136] number of JEDEC chips: 1
[    1.551975] Adding Device 0
[    1.556819] Partitioning Device 1
[    1.558150] Creating 1 MTD partitions on "eNET_Flash_1":
[    1.559150] 0x000000000000-0x000001000000 : "rootfs"
[    1.563816] Adding Device 2
[    1.568821] 8139too Fast Ethernet driver 0.9.28
[    1.573803] eth0: RealTek RTL8139 at 0xc8896000, 00:40:f4:6e:67:ac, IRQ 10
[    1.578796] eth1: RealTek RTL8139 at 0xc889a100, 00:40:f4:6e:67:ac, IRQ 9
[    1.582822] TCP cubic registered
[    1.584516] Using IPI Shortcut mode
[    1.901864] VFS: Mounted root (jffs2 filesystem) on device 31:1.
[    1.902926] Freeing unused kernel memory: 204k freed
[    1.909817] Write protecting the kernel text: 1924k
[    1.911896] Write protecting the kernel read-only data: 596k
[    1.915842] kernel_execve(/sbin/init)

(I've added a printk in kernel_execve() which gives me the last line)

I initially had hot-plug support compiled into the kernel and was getting a lot of kernel_execve(/sbin/hotplug) messages as well (plus a udev:
starting version 151 message which appear 100+ seconds and the above messages), but I have since removed that and used mknod to create /dev/ttyS0 directly

I have also tried the following test script:

#!/bin/bash
stty -F /dev/ttyS0 115200
cat hello > /dev/ttyS0

but still no luck

Does anyone have any advise on how I can debug this problem?

Regards,

Graeme

#1 lore raw
Quoted 19 lines · Graeme Russ

Hi All,

I've been working for a while now getting the x86 U-Boot port up to speed to boot a linux kernel

I now have a root fs (build from Linux-From-Scratch and trimmed to <15M using qemu) but all attempts to get any serial console output past launching of init have failed.

I have taken the root fs which successfully boots using qemu and created a jffs2 image and flashed it onto one of the on-board flash chips. U-Boot can successfully 'ls' the root fs

My bootags are:

root=/dev/mtdblock1 rw rootfstype=jffs2 console=ttyS0,115200

as Linux can mount the root partition:

[    1.447819] Serial: 8250/16550 driver, 6 ports, IRQ sharing disabled

[ snip ]

Quoted 13 lines
[    1.911896] Write protecting the kernel read-only data: 596k
[    1.915842] kernel_execve(/sbin/init)

(I've added a printk in kernel_execve() which gives me the last line)

I initially had hot-plug support compiled into the kernel and was getting a lot of kernel_execve(/sbin/hotplug) messages as well (plus a udev:
starting version 151 message which appear 100+ seconds and the above messages), but I have since removed that and used mknod to create /dev/ttyS0 directly

I have also tried the following test script:

Quoted 7 lines

#!/bin/bash
stty -F /dev/ttyS0 115200
cat hello > /dev/ttyS0

but still no luck

Does anyone have any advise on how I can debug this problem?

Have you tried with init=/bin/sh?
Also, make sure you have /dev/null and /dev/console nodes.

Quoted 9 lines

Regards,

Graeme
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 2 lines
Sincerely yours,
Mike.
#2 lore raw
In reply to Graeme Russ
Quoted 4 lines · Graeme Russ

I've been working for a while now getting the x86 U-Boot port up to speed to boot a linux kernel

...ewww x86.

Quoted 15 lines

[snipped]

[    1.911896] Write protecting the kernel read-only data: 596k
[    1.915842] kernel_execve(/sbin/init)

(I've added a printk in kernel_execve() which gives me the last line)

I initially had hot-plug support compiled into the kernel and was getting a lot of kernel_execve(/sbin/hotplug) messages as well (plus a udev:
starting version 151 message which appear 100+ seconds and the above messages), but I have since removed that and used mknod to create /dev/ttyS0 directly

I have also tried the following test script:

Quoted 8 lines

#!/bin/bash
stty -F /dev/ttyS0 115200
cat hello > /dev/ttyS0

but still no luck

Does anyone have any advise on how I can debug this problem?

What does you /dev directory look like? You need to have at least the following:
----
mknod console c 5 1
mknod null c 1 3
mknod ttyS0 c 4 64
----

Cheers

Signature · 2 lines
Alexander Clouter
.sigmonster says: Your love life will be... interesting.
#3 lore raw
Quoted 6 lines · Alexander Clouter
Quoted 4 lines · Graeme Russ

I've been working for a while now getting the x86 U-Boot port up to speed to boot a linux kernel

...ewww x86.

Yeah, nothing like a challange ;)

Quoted 24 lines
Quoted 16 lines

[snipped]

[    1.911896] Write protecting the kernel read-only data: 596k
[    1.915842] kernel_execve(/sbin/init)

(I've added a printk in kernel_execve() which gives me the last line)

I initially had hot-plug support compiled into the kernel and was getting a lot of kernel_execve(/sbin/hotplug) messages as well (plus a udev:
starting version 151 message which appear 100+ seconds and the above messages), but I have since removed that and used mknod to create /dev/ttyS0 directly

I have also tried the following test script:

Quoted 8 lines

#!/bin/bash
stty -F /dev/ttyS0 115200
cat hello > /dev/ttyS0

but still no luck

Does anyone have any advise on how I can debug this problem?

Quoted 8 lines

What does you /dev directory look like? You need to have at least the following:
----
mknod console c 5 1
mknod null c 1 3
mknod ttyS0 c 4 64
----

tick, tick, tick

A little further background (should have included this last night but it was late and I thought the problem must be trivial)...

I had originally mounted this root fs using NFS with (very) limited success. After mounting and performing a few NFS requests (observed using wireshark) the NFS activity would simply stop with 'server not responding' messages. I turned on debugging in the 8139too and NFS drivers. With debugging output enabled I was still getting 'server not responding' messages, but (eventually) the boot sequence would continue. This even resulted in logs in /var/log of the NFS share for the device (filled with raw network packet data).

But no console...

What if I init with the following:

#!/bin/bash
echo hello > /hello.txt

and then rebooted and checked the root fs in U-Boot? How long will it take for hello.txt to be flushed? Can I force the flush?

Regards,

Graeme

#4 lore raw
Quoted 17 lines · Graeme Russ
Quoted 7 lines · Alexander Clouter
Quoted 3 lines · Graeme Russ

I've been working for a while now getting the x86 U-Boot port up to speed to boot a linux kernel

...ewww x86.

Quoted 1 line
Quoted 8 lines

What does you /dev directory look like? You need to have at least the following:
----
mknod console c 5 1
mknod null c 1 3
mknod ttyS0 c 4 64
----

Quoted 2 lines

tick, tick, tick

Success!!!

I feel _really_ stupid about the cause of the problem.... I hadn't set up the interrupt mapping of the serial ports in my board initialisation routines in U-Boot. Once I mapped IRQ4 to UART0 it all sprang to life ;)

Big thanks to everyone for helping and a big thanks to all the linux and GNU developers for creating such a wonderful 'toy' ;)

Graeme

#5 lore raw

Re: Request For Help - Embedded x86, Kernel 2.6.33 boots, Mounts ?JFFS2 rootfs, Stops at init

In reply to Graeme Russ
Quoted 16 lines · Graeme Russ
Quoted 7 lines · Alexander Clouter

What does you /dev directory look like? You need to have at least the following:
----
mknod console c 5 1
mknod null c 1 3
mknod ttyS0 c 4 64
----

tick, tick, tick

A little further background (should have included this last night but it was late and I thought the problem must be trivial)...

Quoted 9 lines

I had originally mounted this root fs using NFS with (very) limited success. After mounting and performing a few NFS requests (observed using wireshark) the NFS activity would simply stop with 'server not responding' messages. I turned on debugging in the 8139too and NFS drivers. With debugging output enabled I was still getting 'server not responding' messages, but (eventually) the boot sequence would continue. This even resulted in logs in /var/log of the NFS share for the device (filled with raw network packet data).

I got this behaviour when setting up a PXE booting workstation network registration system. My issue was that the IP changed and was running running TCP based NFS; I think the kernel continues to try to use the old IP, same socket, but as it is no longer there...Game Over. Depending on how your boot up worked, at the PXE booting phase (or just where the kernel uses DHCP to get a lease) things work, however then you get to the network configuring part of the boot up, dhclient uses a different identifier so you get a different IP.

Well...that's what happened to us and the simpliest fix was to move to UDP for NFS.

Quoted 2 lines

But no console...

...instead of LFS, have you looked at buildroot[1]?

Quoted 5 lines

What if I init with the following:

#!/bin/bash
echo hello > /hello.txt

....put a 'while (true); do sleep 1; done' on the end of that. Although you should see a 'attempted to kill init' without it there.

Quoted 3 lines

and then rebooted and checked the root fs in U-Boot? How long will it take for hello.txt to be flushed? Can I force the flush?

Have a look at:

http://www.spinics.net/lists/mips/msg37361.html

Especially my 'readelf' bit, I'm guessing you might have forgotten a library or two in your keenness to trim your root mountpoint.

Cheers

[1] http://buildroot.uclibc.org/

Signature · 2 lines
Alexander Clouter
.sigmonster says: Time to be aggressive.  Go after a tattooed Virgo.