Skip to contentSkip to conversation

Kernel crash on yosemite

DISCUSSION TheNop 3 participants 9 messages Feb 7, 2005 – Feb 8, 2005 MIPS

lore
Share & tools
lore

TheNop

author
#0 lore raw

Hi ,

This is my configuration:
~ yosemite board with RM9000 chip revision 1.1
~ BusyBox 1.0
~ Kernel 2.6.8.1

When I try to copy a large file (~ 3,5 Mb) within the NFS file system the kernel crashs without any output on the console. It could be a problem with the titan_ge driver, but I have no idee how to solve the problem.

What can I do?

Best regards
TheNop

#1 lore raw
Quoted 11 lines · TheNop

This is my configuration:
~ yosemite board with RM9000 chip revision 1.1
~ BusyBox 1.0
~ Kernel 2.6.8.1

When I try to copy a large file (~ 3,5 Mb) within the NFS file system the kernel crashs without any output on the console.
It could be a problem with the titan_ge driver, but I have no idee how to solve the problem.

What can I do?

There have been various fixes to the network driver since then so I recommend you upgrade your kernel. One problem you're going to encounter with recent kernels is that they only support the Titan 1.2 part which I think are the ones in volume production.

Ralf

#2 lore raw
Quoted 16 lines · Ralf Baechle
Quoted 11 lines · TheNop

This is my configuration:
~ yosemite board with RM9000 chip revision 1.1
~ BusyBox 1.0
~ Kernel 2.6.8.1

When I try to copy a large file (~ 3,5 Mb) within the NFS file system the kernel crashs without any output on the console. It could be a problem with the titan_ge driver, but I have no idee how to solve the problem.

What can I do?

Quoted 9 lines

There have been various fixes to the network driver since then so I recommend you upgrade your kernel. One problem you're going to encounter with recent kernels is that they only support the Titan 1.2 part which I think are the ones in volume production.

Ralf

However, adding support for Titan 1.0 and 1.1 in the GE driver should be fairly straight forward.

Thanks
Manish Lachwani

#3 lore raw
Quoted 6 lines · Manish Lachwani
Quoted 3 lines

recommend you upgrade your kernel. One problem you're going to encounter with recent kernels is that they only support the Titan 1.2 part which I think are the ones in volume production.

However, adding support for Titan 1.0 and 1.1 in the GE driver should be fairly straight forward.

True, the CVS log should provide reasonable information about that. What can't be sanely retrofitted is SMP for < 1.2.

Ralf

#4 lore raw
Quoted 9 lines · Ralf Baechle
Quoted 5 lines · Manish Lachwani
Quoted 5 lines

recommend you upgrade your kernel. One problem you're going to encounter with recent kernels is that they only support the Titan 1.2 part which I think are the ones in volume production.

Quoted 10 lines
Quoted 3 lines

However, adding support for Titan 1.0 and 1.1 in the GE driver should be fairly straight forward.

True, the CVS log should provide reasonable information about that. What can't be sanely retrofitted is SMP for < 1.2.

Ralf

I completely agree. In any case, I dont think the CVS sources ever supported SMP for Titan < 1.2, correct?

Thanks
Manish Lachwani

#5 lore raw
Quoted 2 lines · Manish Lachwani

I completely agree. In any case, I dont think the CVS sources ever supported SMP for Titan < 1.2, correct?

No, that would have either meant taking a heavy performance penalty or hacking mm in very ugly ways. And due to the probably low numbers of the early silicon and the availability of later revisions the reason for such hacks was low anyway.

Ralf

#6 lore raw
Quoted 11 lines · Ralf Baechle
Quoted 3 lines · Manish Lachwani

I completely agree. In any case, I dont think the CVS sources ever supported SMP for Titan < 1.2, correct?

No, that would have either meant taking a heavy performance penalty or hacking mm in very ugly ways.

How would I not know this? After all, I was the one who actually put this code together to get SMP to work on Titan 1.0 and 1.1, remember? ;) And you are right, there is no reason to put such hacks.

Manish Lachwani

Quoted 8 lines

And due to the probably low numbers of
the early silicon and the availability of later revisions the reason for such hacks was low anyway.

Ralf

author
#7 lore raw

Hi,

I'm sorry, but all the discussion on this thread did'n help me to solve my problem. Since more than a month I'm waitig to get a yosemite board with 1.2 version. This still stops my development, when I can't use the yosemite with 1.1 version.

In another therad we discuss to create a patch (slowpath) for the titan version 1.1. the patch I apply to the the newer kernel sources (2.6.11.rc1) but, I have the same problems. So I think the problem could be the slowpath thing.

Have anybody an idee how to figure out what could be the problem when I make "cp foo foo1" (foo is a lage file ~3,5 Mb)?

Best regards
TheNop

Quoted 31 lines · Manish Lachwani
Quoted 10 lines · Ralf Baechle
Quoted 2 lines · Manish Lachwani

I completely agree. In any case, I dont think the CVS sources ever supported SMP for Titan < 1.2, correct?

No, that would have either meant taking a heavy performance penalty or hacking mm in very ugly ways.

How would I not know this? After all, I was the one who actually put this code together to get SMP to work on Titan 1.0 and 1.1, remember? ;) And you are right, there is no reason to put such hacks.

Manish Lachwani

Quoted 8 lines

And due to the probably low numbers of
the early silicon and the availability of later revisions the reason for such hacks was low anyway.

Ralf

#8 lore raw
Quoted 5 lines · TheNop

I'm sorry, but all the discussion on this thread did'n help me to solve my problem. Since more than a month I'm waitig to get a yosemite board with 1.2 version. This still stops my development, when I can't use the yosemite with 1.1 version.

Quoted 7 lines

In another therad we discuss to create a patch (slowpath) for the titan version 1.1. the patch I apply to the the newer kernel sources (2.6.11.rc1) but, I have the same problems. So I think the problem could be the slowpath thing.

Have anybody an idee how to figure out what could be the problem when I make "cp foo foo1" (foo is a lage file ~3,5 Mb)?

A network driver bug in the 2.6 Titan driver which I recently fixes was that the allocated receive and transmit rings were allocated overlapping which was causing havoc.

Yosemite w/ Titan 1.1 was suffering from unexplained system freezes for me; a more recent board with Titan 1.2 is very stable.

Ralf