2.6.31 new release

[calamari]

I also updated my plug with 2.6.31 and got the following warning:

------------[ cut here ]------------
WARNING: at arch/arm/kernel/process.c:171 cpu_idle+0x80/0xc0()
Modules linked in: dm_crypt dm_mod snd_usb_audio snd_pcm snd_timer snd_page_alloc snd_usb_lib snd_rawmidi snd_seq_device snd_hwdep snd soundcore
[<c0031a78>] (unwind_backtrace+0x0/0xe0) from [<c00423f8>] (warn_slowpath_common+0x50/0x68)
[<c00423f8>] (warn_slowpath_common+0x50/0x68) from [<c002d8d0>] (cpu_idle+0x80/0xc0)
[<c002d8d0>] (cpu_idle+0x80/0xc0) from [<c00089ec>] (start_kernel+0x268/0x2c0)
[<c00089ec>] (start_kernel+0x268/0x2c0) from [<00008034>] (0x8034)
---[ end trace a3132aad8b495ecd ]---

I saw a thread related to 2.6.31_rc7 relating to this warning, has anyone else got this warning?

For the release candidates this was easy to fix using a patch as described here, using the following patch:
http://patchwork.kernel.org/patch/40871/

[rooster]

Yes I saw the patch but thought it was integrated to 2.6.31...

[glee_hokie]

2.6.31 is now available.

cbxbiker61, what toolchain did you use to compile the kernel?

I'd like to make a custom configured kernel.  I'm starting from the kernel.org 2.6.31 sources, added the patches that you provide and am using your .config that you have kindly posted.  I'd like to be sure I'm using the same toolchain as well.

Thanks!

[sillis]

I received this error message while trying to get new kernel to work on my sheevaplug:

root@debian:/home/sillis/temp# sh README-2.6.31 --rootkernel
README-2.6.31: 29: Syntax error: "(" unexpected

Current kernel is 2.6.30-rc8.

Has anyone else experienced problems using this readme file?

[DamonHD]

Because you're trying to run it under /bin/sh rather than bash.

Just do ./README...

after giving it execute permisions with chmod u+x README...

Rgds

Damon

[sillis]

Because you're trying to run it under /bin/sh rather than bash.

Just do ./README...

after giving it execute permisions with chmod u+x README...

Rgds

Damon

Oh hell, i feel so stupid now, Thanks Damon.

I dumped the kernel update using the readme file and proceeded to just copy paste uimage to /boot and kernelheaders&modules to right folders. At the end of boot process i got the following message.

insmod: can't read '/boot/fat.ko': No such file or directory

Maybe because i didnt use the readme file ?

but it seems everything is working now, uname -a reports update was a success..
Linux debian 2.6.31 #2 PREEMPT Wed Sep 16 00:55:42 MDT 2009 armv5tel GNU/Linux

[DamonHD]

Don't go feeling stupid without letting me get ahead of you in the "stupid" queue!  %-P

Actually I think it would be slightly better to have the script execute under 'sh' anyway which runs everywhere and in much less memory, and is somewhat less bizarre IMHO, but it's a very minor point.  It's great that these scripts have been put together for us to use!

Rgds

Damon

[calamari]

Boots fine.  Seems robust.  dm_crypt still broken.

Just compiled the orion 2.6.31 trunk version and dm_crypt / cryptsetup seems to work fine. I haven't tried cbxbiker61's version yet.
In case anybody else wants to give this alternative version a try: http://photon123.dyndns.org/sheeva/README-2.6.31-06466-g4c203e00

[theblop]

Just compiled the orion 2.6.31 trunk version and dm_crypt / cryptsetup seems to work fine. I haven't tried cbxbiker61's version yet.
In case anybody else wants to give this alternative version a try: http://photon123.dyndns.org/sheeva/README-2.6.31-06466-g4c203e00

Can you do a quick benchmark on a dm_crypt usb HD if you have one?

With a 2.6.29 kernel I get about 5MB/s running a 500MB write/read test on a dm_crypted USB hard disk (33MB/s without dm_crypt):

Code:

# write test
sync && dd if=/dev/zero of=./test bs=1000k count=500
# read test
sync && dd if=./test of=/dev/null && rm ./test

I wonder how much faster dm_crypt would be if it used the sheevaplug's encryption chipset.

[calamari]

Can you do a quick benchmark on a dm_crypt usb HD if you have one?

Sure, however I only have a USB stick with an encrypted partition connected right now. I ran the read/write tests a few times, the maximum I got was the following:

Code:

# write test
sync && dd if=/dev/zero of=./test bs=1000k count=500
500+0 records in
500+0 records out
512000000 bytes (512 MB) copied, 67.3066 s, 7.6 MB/s

# read test
sync && dd if=./test of=/dev/null && rm ./test
1000000+0 records in
1000000+0 records out
512000000 bytes (512 MB) copied, 61.8155 s, 8.3 MB/s

I wonder how much faster dm_crypt would be if it used the sheevaplug's encryption chipset.

Note that for the tests above the mv_cesa module was loaded. Yet I have no idea whether it's actually used for my cipher or not.

Code:

lsmod
Module                  Size  Used by
ipv6                  263160  10
dm_crypt               13820  1
dm_mod                 67868  3 dm_crypt
mv_cesa                 6856  1

[theblop]

Code:

# write test
sync && dd if=/dev/zero of=./test bs=1000k count=500
500+0 records in
500+0 records out
512000000 bytes (512 MB) copied, 67.3066 s, 7.6 MB/s

# read test
sync && dd if=./test of=/dev/null && rm ./test
1000000+0 records in
1000000+0 records out
512000000 bytes (512 MB) copied, 61.8155 s, 8.3 MB/s

Thanks, it looks promising! I'll try compiling the 2.6.31 orion kernel and see what I get with my hard disk.

[racalac]

Hi cbxbiker61,
Thanks for your work on pre-built Sheeva kernels. I've been using your 2.6.30.6 release with usb-serial happily for awhile. I'd like to humbly request configuring CONFIG_USB_HIDDEV with your next release. This would allow monitoring of APC Uninterruptable Power Supplies and battery status. I'm using the Sheevaplug and an Arduino for a home monitoring system that I'd like to be able to report a mains power loss

Thanks very much!

[CrunchBytes]

Hi cbxbiker61,

Just a quick note to say 'Sir, you are a star!!'.

Your latest pre-built kernel fixed my 'tun' issue with the standard install.

Fantastic!

Thanks,

Richard

[DamonHD]

Seconded on the star status...

[theblop]

Code:

# write test
sync && dd if=/dev/zero of=./test bs=1000k count=500
500+0 records in
500+0 records out
512000000 bytes (512 MB) copied, 67.3066 s, 7.6 MB/s

# read test
sync && dd if=./test of=/dev/null && rm ./test
1000000+0 records in
1000000+0 records out
512000000 bytes (512 MB) copied, 61.8155 s, 8.3 MB/s

Thanks, it looks promising! I'll try compiling the 2.6.31 orion kernel and see what I get with my hard disk.

Ok I compiled 2.6.31 from the latest orion tree with CONFIG_CRYPTO_DEV_MV_CESA=y and got about twice the speed:

Code:

write 500 MB:
512000000 bytes (512 MB) copied, 57.3278 s, 8.9 MB/s
read:
512000000 bytes (512 MB) copied, 45.0345 s, 11.4 MB/s

About 3 times slower as without encryption, but it'll do for my backups!