hippo, I&#39;m starting a new thread to keep track of this... I think 2.6.12 looks better than .16 - it hasn&#39;t failed on me yet. the patches you gave me were a bit messy though, I think maybe some things were missing. I also get these warnings on every single file compiled: In file included from /share/nios2/tmp/linux-2.6.12/include/linux/sched.h:17, from /share/nios2/tmp/linux-2.6.12/include/linux/module.h:10, from /share/nios2/tmp/linux-2.6.12/init/version.c:10: /share/nios2/tmp/linux-2.6.12/include/linux/nodemask.h: In function `__first_unset_node&#39;: /share/nios2/tmp/linux-2.6.12/include/linux/nodemask.h:250: warning: passing arg 1 of `find_next_zero_bit&#39; discards qualifiers from pointer target type Page allocation failures is one thing, but I&#39;m more worried about the kernel hanging completely like .16 was doing. I also have a feeling it&#39;s the busybox version. I noticed that in 1.1.1, the file utilities were a bit messed up. For example, here is the output from a normal &#39;df -h&#39;: # busybox_old df -h Filesystem Size Used Available Use% Mounted on /dev/ram0 235.0k 14.0k 221.0k 6% /var /dev/ram1 235.0k 13.0k 222.0k 6% /tmp /dev/hda1 14.5M 12.3M 1.5M 89% /mnt/cf# Here is the output of 1.1.1 df -h: # df -h Filesystem Size Used Available Use% Mounted on /dev/ram0 235.0k 14.0k 221.0k 6% (null) /dev/ram1 235.0k 13.0k 222.0k 6% (null) /dev/hda1 14.5M 12.3M 1.5M 89% (null)# All the filesystem utilities, like mke2fs, du, e2fsck, are weird like this. I will keep playing around with 2.6.12 and see what happens.

--- Quote Start --- originally posted by jdhar@Apr 13 2006, 10:01 AM hippo, i&#39;m starting a new thread to keep track of this... i think 2.6.12 looks better than .16 - it hasn&#39;t failed on me yet. the patches you gave me were a bit messy though, i think maybe some things were missing. i also get these warnings on every single file compiled: in file included from /share/nios2/tmp/linux-2.6.12/include/linux/sched.h:17, from /share/nios2/tmp/linux-2.6.12/include/linux/module.h:10, from /share/nios2/tmp/linux-2.6.12/init/version.c:10: /share/nios2/tmp/linux-2.6.12/include/linux/nodemask.h: in function `__first_unset_node&#39;: /share/nios2/tmp/linux-2.6.12/include/linux/nodemask.h:250: warning: passing arg 1 of `find_next_zero_bit&#39; discards qualifiers from pointer target type page allocation failures is one thing, but i&#39;m more worried about the kernel hanging completely like .16 was doing. i also have a feeling it&#39;s the busybox version. i noticed that in 1.1.1, the file utilities were a bit messed up. for example, here is the output from a normal &#39;df -h&#39;: # busybox_old df -h filesystem size used available use% mounted on /dev/ram0 235.0k 14.0k 221.0k 6% /var /dev/ram1 235.0k 13.0k 222.0k 6% /tmp /dev/hda1 14.5m 12.3m 1.5m 89% /mnt/cf# here is the output of 1.1.1 df -h: # df -h filesystem size used available use% mounted on /dev/ram0 235.0k 14.0k 221.0k 6% (null) /dev/ram1 235.0k 13.0k 222.0k 6% (null) /dev/hda1 14.5m 12.3m 1.5m 89% (null)# all the filesystem utilities, like mke2fs, du, e2fsck, are weird like this. i will keep playing around with 2.6.12 and see what happens. <div align='right'><{post_snapback}> (index.php?act=findpost&pid=14295) --- quote end --- --- Quote End --- Nice to get a new track. The busybox of 20060320 is very close to the official release 1.1.1 . I didn&#39;t find problem like yours. I wonder if you are editing on Windows, so that CRLF may cause problems to patch. Try "patch -l" to ignore white space. I will check the 2.6.12 patch files again.

hippo, I do mostly everything in linux now, butp atching especially; I don&#39;t patch in windows at all - especially after our very first conversation http://forum.niosforum.com/work2/style_emoticons/<#EMO_DIR#>/smile.gif Either way, i have the kernel patched up and it&#39;s running. So far, CF, JTAG and OpenEth is up, and Im&#39; getting my MMC in there. Once I have that much, I will be able to tell if it&#39;s more stable. This kernel isn&#39;t much different than .11, so I&#39;m not sure if this is much progress http://forum.niosforum.com/work2/style_emoticons/<#EMO_DIR#>/sad.gif

--- Quote Start --- originally posted by jdhar@Apr 13 2006, 11:23 AM hippo, i do mostly everything in linux now, butp atching especially; i don&#39;t patch in windows at all - especially after our very first conversation http://forum.niosforum.com/work2/style_emoticons/<#emo_dir#>/smile.gif either way, i have the kernel patched up and it&#39;s running. so far, cf, jtag and openeth is up, and im&#39; getting my mmc in there. once i have that much, i will be able to tell if it&#39;s more stable. this kernel isn&#39;t much different than .11, so i&#39;m not sure if this is much progress http://forum.niosforum.com/work2/style_emoticons/<#emo_dir#>/sad.gif <div align='right'><{post_snapback}> (index.php?act=findpost&pid=14302) --- quote end --- --- Quote End --- The 2.6.12 is the first move toward newer kernels. There is no big different really. But some internal data structures and api were changed during each kernel revision. It is very likely that something failed or become obsolete. Some codes ported by Microtronix were not updated and broke eventually. I am going to read more books on kernel internal. I have to learn more. http://forum.niosforum.com/work2/style_emoticons/<#EMO_DIR#>/smile.gif

argh, between designing new hardware and writing these drivers, I wish I had more time to help out on the core side http://forum.niosforum.com/work2/style_emoticons/<#EMO_DIR#>/sad.gif Anyway, I have put the first &#39;semi-stable&#39; version up: http://via.dynalias.org/snapshots/ (http://via.dynalias.org/snapshots/) I still need to move the MTD driver over, and do some more testing. But if all goes well, I will leave 2.6.11 behind. These two kernels are similar enough that any drivers can be easily ported between the two.

hippo, can you put a patch up for zImage? It&#39;s integrated in your other patch, so i can&#39;t apply it.

linux 2.6.12 instability | Altera Community

26 Replies

Altera_Forum
Honored Contributor
20 years ago
Just another tip,

I won't compile kernel in place, not in the source tree.
I use "make .... O=~/board_1/kernel-branch " for each board and each version of kernel,
the source tree must be clean first, with "make .... clean" and "make .... mrproper" .

And each board can have a copy of rootfs of different apps setup. ~/board_1/rootfs

I build kernel in the source tree for the buildroot guide,
only because it would be easier for newbie.
Altera_Forum
Honored Contributor
20 years ago
That's a good patch system, I didn't know you had set it up that way - could I get the patch set for .14 and .16?

When I noticed the dd issue, I was using a Avalon system with pipelining enabled. When I addeda DMA component to the system, it locked everything up when I would try and access certain registers. It's definitely an Avalon bug, and I have my FAE looking into it. Anyway, I'm not using pipelining now, so I'm seeing if it makes a difference with this issue.
Altera_Forum
Honored Contributor
20 years ago
oh god, I would be in big trouble if I compiled directly in the tree! I have over 20 different configurations for my hardware, so it would be hell with 20 different copies of the source http://forum.niosforum.com/work2/style_emoticons/<#EMO_DIR#>/smile.gif

I have a separate folder with a makefile that just handles all the cross-compiling stuff and pointing to the right source tree.

http://via.dynalias.org/svn/uclinux/kernel...skeleton/trunk/ (http://via.dynalias.org/svn/uclinux/kernel_build/skeleton/trunk/) (just in case anyone is interested)
Altera_Forum
Honored Contributor
20 years ago
--- Quote Start ---
originally posted by jdhar@Apr 13 2006, 02:03 PM
oh god, i would be in big trouble if i compiled directly in the tree! i have over 20 different configurations for my hardware, so it would be hell with 20 different copies of the source http://forum.niosforum.com/work2/style_emoticons/<#emo_dir#>/smile.gif

i have a separate folder with a makefile that just handles all the cross-compiling stuff and pointing to the right source tree.

http://via.dynalias.org/svn/uclinux/kernel...skeleton/trunk/ (http://via.dynalias.org/svn/uclinux/kernel_build/skeleton/trunk/) (just in case anyone is interested)
<div align='right'><{post_snapback}> (index.php?act=findpost&pid=14315)
--- quote end ---

--- Quote End ---

Ya, you knew it already. http://forum.niosforum.com/work2/style_emoticons/<#EMO_DIR#>/smile.gif
Altera_Forum
Honored Contributor
20 years ago
kernel logs (v0.2): http://via.dynalias.org/snapshots/linux-2.6.12-0.2.tar.bz2 (http://via.dynalias.org/snapshots/linux-2.6.12-0.2.tar.bz2)

------------------------------------------------------------------------
r908 | jdhar | 2006-04-13 01:37:07 -0400 (Thu, 13 Apr 2006) | 1 line

v0.2 - moved over I2C Opencores and EPCS MTD Driver. zImage compressed image support also added

------------------------------------------------------------------------
r903 | jdhar | 2006-04-12 23:42:54 -0400 (Wed, 12 Apr 2006) | 1 line

v0.1 - 2.6.12, minimally patched with MMC 0.1.4, open_eth, JTAG UART, CF. No MTD driver yet
Altera_Forum
Honored Contributor
20 years ago
--- Quote Start ---
originally posted by jdhar@Apr 13 2006, 02:01 PM
that's a good patch system, i didn't know you had set it up that way - could i get the patch set for .14 and .16?

when i noticed the dd issue, i was using a avalon system with pipelining enabled. when i addeda dma component to the system, it locked everything up when i would try and access certain registers. it's definitely an avalon bug, and i have my fae looking into it. anyway, i'm not using pipelining now, so i'm seeing if it makes a difference with this issue.
<div align='right'><{post_snapback}> (index.php?act=findpost&pid=14314)
--- quote end ---

--- Quote End ---

You need to be very careful with DMA, the original 2.6.11 port of Microtronix did not handle DMA cache. And it is very important for PCI. Anyways, I fix them and my PCI is working well.

Pipelineing is quite new. I had asked for it long ago. But it is still not usable for me up to now.

Those patches are too big to post here. I have only 1MB quota here. Please give me an upload point to my private mail. I will prepare them when I have time.
Altera_Forum
Honored Contributor
20 years ago
The bug was weird actually..

if I had a simple system with just a DMA core tied to an SPI Device (I was trying to do DMA with my MMC core), and pipelining enabled, it would completely locku p the bus if I tried to even read an SPI Register without DMA enabled. I thought it was uClinux, so I used a simpel hello world in NIOS II IDE to read an SPI Register and it still locked up the system. I think they still have issues with pipelining, but only in more complex systems (I have 6 masters DMAing into DDR).

No rush for those patches, I'll send an upload point. Thx.
Altera_Forum
Honored Contributor
20 years ago
--- Quote Start ---
originally posted by jdhar@Apr 13 2006, 02:01 PM
that's a good patch system, i didn't know you had set it up that way - could i get the patch set for .14 and .16?

when i noticed the dd issue, i was using a avalon system with pipelining enabled. when i addeda dma component to the system, it locked everything up when i would try and access certain registers. it's definitely an avalon bug, and i have my fae looking into it. anyway, i'm not using pipelining now, so i'm seeing if it makes a difference with this issue.
<div align='right'><{post_snapback}> (index.php?act=findpost&pid=14314)
--- quote end ---

--- Quote End ---

patches for 2.6.12-2.6.16 uploaded.
note the patches for the struct flip changed in serial/jtag uart rx in 2.6.16 .
Altera_Forum
Honored Contributor
20 years ago
--- Quote Start ---
originally posted by hippo+apr 13 2006, 10:30 pm--><div class='quotetop'>quote (hippo @ apr 13 2006, 10:30 pm)</div>
--- quote start ---
<!--quotebegin-jdhar@Apr 13 2006, 02:01 PM
that's a good patch system, i didn't know you had set it up that way - could i get the patch set for .14 and .16?

when i noticed the dd issue, i was using a avalon system with pipelining enabled. when i addeda dma component to the system, it locked everything up when i would try and access certain registers. it's definitely an avalon bug, and i have my fae looking into it. anyway, i'm not using pipelining now, so i'm seeing if it makes a difference with this issue.
<div align='right'><{post_snapback}> (index.php?act=findpost&pid=14314)
--- quote end ---

--- Quote End ---

patches for 2.6.12-2.6.16 uploaded.
note the patches for the struct flip changed in serial/jtag uart rx in 2.6.16 .
<div align='right'><{post_snapback}> (index.php?act=findpost&pid=14338)</div>
[/b]
--- Quote End ---

altpci patch posted,
http://forum.niosforum.com/forum/index.php...ype=post&id=140 (http://forum.niosforum.com/forum/index.php?act=attach&type=post&id=140)

but the patch work for 2.6.15 and 2.6.16 .

it can not work in 2.6.11, the file arch/nios2nommu/mm/dma-noncoherent.c
should be back ported from the one from arch/mips/mm .
I don't have time to do it tonight. help yourself, please.
Altera_Forum
Honored Contributor
20 years ago
great work hippo ; i'm getting .14 in there now, and then i will try .16.

Forum Discussion

linux 2.6.12 instability

26 Replies

Recent Discussions

Run-time errors using EMIF in Agilex 5 NIOS-V design

Nios V/g load-writeback hazard

Ashling RISC Free IDE fails to download ELF file

Dual Port Nios V BSP error

set_output_delay explained for dummies