Pointer to info on decoding oops messages

Pointer to info on decoding oops messages

[Ian Kent]

The subject says it all.

Any suggestions please.

-- 

   ,-._|\    Ian Kent
  /      \   Perth, Western Australia
  *_.--._/   E-mail: raven@themaw.net
        v    Web: http://themaw.net/

Re: Pointer to info on decoding oops messages

[Randy.Dunlap]

On Wed, 26 Nov 2003 21:41:18 +0800 (WST) Ian Kent <raven@themaw.net> wrote:

| 
| The subject says it all.
| 
| Any suggestions please.

Unless the instruction pointer register is scrogged, you get
some registers, some stack data, the code bytes at the next
instruction pointer, and some backtrace symbols.
These are mostly self-explanatory.

However, I'm guessing that you want help with the _cause_ of
an oops, not just with decoding the oops messages.

Anyway, here's where I pointed out some basic info recently:
http://marc.theaimsgroup.com/?l=linux-kernel&m=106433590230743&w=2
and
http://marc.theaimsgroup.com/?l=linux-kernel&m=106434159306658&w=2

--
~Randy
MOTD:  Always include version info.

Re: Pointer to info on decoding oops messages

[Pat LaVarre]

> From:http://marc.theaimsgroup.com/?l=linux-kernel&m=106433590230743
>
> List: linux-kernel
> Subject: Re: How to understand an oops?
> From: "Randy.Dunlap" <rddunlap () osdl ! org>
> Date: 2003-09-23 16:36:07
>
> > How can I find other code/modules from which
> > skb_drop_fraglist is called and used?
>
> Use grep (or cscope, but that would be
> overkill in this case).
> I found it only in net/core/skbuff.c.

Me, at my desk, I reach net/core/skbuff.c more quickly via such gratis
web services as:

http://lxr.linux.no/ident?v=2.6.0-test7;i=skb_drop_fraglist

[[[But I'm so painfully new that as yet I have no good way to insert
that comment into that relevant linux-kernel thread.]]]

Pat LaVarre
http://members.aol.com/ppaatt/

Re: Pointer to info on decoding oops messages

[Ian Kent]

On Wed, 26 Nov 2003, Randy.Dunlap wrote:

> On Wed, 26 Nov 2003 21:41:18 +0800 (WST) Ian Kent <raven@themaw.net> wrote:
>
> |
> | The subject says it all.
> |
> | Any suggestions please.
>
> Unless the instruction pointer register is scrogged, you get
> some registers, some stack data, the code bytes at the next
> instruction pointer, and some backtrace symbols.
> These are mostly self-explanatory.

Umm .. yes but ....

>
> However, I'm guessing that you want help with the _cause_ of
> an oops, not just with decoding the oops messages.

Well yes.

What I really need is a way to determine if my poorly coded changes are
causing the problem or if I am dealing with a bug in other code.

>
> Anyway, here's where I pointed out some basic info recently:
> http://marc.theaimsgroup.com/?l=linux-kernel&m=106433590230743&w=2
> and
> http://marc.theaimsgroup.com/?l=linux-kernel&m=106434159306658&w=2

Thanks. Will look it up.

-- 

   ,-._|\    Ian Kent
  /      \   Perth, Western Australia
  *_.--._/   E-mail: raven@themaw.net
        v    Web: http://themaw.net/

Re: Pointer to info on decoding oops messages

[Randy.Dunlap]

On Thu, 27 Nov 2003 09:36:16 +0800 (WST) Ian Kent <raven@themaw.net> wrote:

| On Wed, 26 Nov 2003, Randy.Dunlap wrote:
| 
| > On Wed, 26 Nov 2003 21:41:18 +0800 (WST) Ian Kent <raven@themaw.net> wrote:
| >
| > |
| > | The subject says it all.
| > |
| > | Any suggestions please.
| >
| > Unless the instruction pointer register is scrogged, you get
| > some registers, some stack data, the code bytes at the next
| > instruction pointer, and some backtrace symbols.
| > These are mostly self-explanatory.
| 
| Umm .. yes but ....
| 
| >
| > However, I'm guessing that you want help with the _cause_ of
| > an oops, not just with decoding the oops messages.
| 
| Well yes.
| 
| What I really need is a way to determine if my poorly coded changes are
| causing the problem or if I am dealing with a bug in other code.

Have you posted any of your oopsen?  You can try that.
Sometimes that will get you some good feedback, sometimes not....

| >
| > Anyway, here's where I pointed out some basic info recently:
| > http://marc.theaimsgroup.com/?l=linux-kernel&m=106433590230743&w=2
| > and
| > http://marc.theaimsgroup.com/?l=linux-kernel&m=106434159306658&w=2
| 
| Thanks. Will look it up.

--
~Randy

Re: Pointer to info on decoding oops messages

[Ian Kent]

On Sun, 30 Nov 2003, Randy.Dunlap wrote:

>
> Have you posted any of your oopsen?  You can try that.
> Sometimes that will get you some good feedback, sometimes not....
>

Have identified my error and I am moving forward again.

Thanks all. I managed to gain a little more knowledge about kernel
debuging from the responses. In the end it was old faithful 'printk' that
caught the nasty creature.

-- 

   ,-._|\    Ian Kent
  /      \   Perth, Western Australia
  *_.--._/   E-mail: raven@themaw.net
        v    Web: http://themaw.net/

Re: Pointer to info on decoding oops messages

[Ian Kent]

On Wed, 26 Nov 2003, Ian Kent wrote:

>
> The subject says it all.
>

Once upon a time, in a galaxy far away I could get an assembly listing
with original source lines interspersed. This would allow me to quickly
identify the address (source line) causing a problem from stack dumps.

Anyone know how this can be acheived with gcc and the kernel source.

-- 

   ,-._|\    Ian Kent
  /      \   Perth, Western Australia
  *_.--._/   E-mail: raven@themaw.net
        v    Web: http://themaw.net/

Re: Pointer to info on decoding oops messages

[Randy.Dunlap]

On Thu, 27 Nov 2003 09:54:05 +0800 (WST) Ian Kent <raven@themaw.net> wrote:

| On Wed, 26 Nov 2003, Ian Kent wrote:
| 
| >
| > The subject says it all.
| >
| 
| Once upon a time, in a galaxy far away I could get an assembly listing
| with original source lines interspersed. This would allow me to quickly
| identify the address (source line) causing a problem from stack dumps.
| 
| Anyone know how this can be acheived with gcc and the kernel source.

[See MOTD, below.]

In 2.6.x, you can do this:

  make drivers/usb/core/usb.lst

This doesn't work in 2.4 (at least not in 2.4.22, where I
tested it).  You can always use objdump to list an object file,
but it doesn't provide mixed source/object lines.

There was a question similar to this on the gcc mailing list a few
weeks ago.  One answer posted was this (from Jim Wilson):

  If you are using GNU as, then see the GNU as docs for the -a option 
  which produces listings file.  Try for instance
	gcc -O -g -Wa,-ahls file.c

so maybe you change some CFLAGS or AFLAGS etc. in Makefile and have
'as' generate a listing for you.

--
~Randy
MOTD:  Always include version info.

Re: Pointer to info on decoding oops messages

[Ian Kent]

On Mon, 1 Dec 2003, Randy.Dunlap wrote:

>
> [See MOTD, below.]
>

Oops.

Point taken.

-- 

   ,-._|\    Ian Kent
  /      \   Perth, Western Australia
  *_.--._/   E-mail: raven@themaw.net
        v    Web: http://themaw.net/