[Xenomai-help] Problem compiling patched kernel: multiple definitions of mcount

[Xenomai-help] Problem compiling patched kernel: multiple definitions of mcount

[James Doebbler]

Hello,
 I am trying to patch and install a kernel using Xenomai (on Ubuntu
8.10).  I am using a vanilla linux-2.6.27 kernel and the latest
xenomai version (2.4.6.1).  I had some earlier problems that I
resolved by using the latest adeos patch (now I see that has been
fixed in 2.4.6.1), but I've been running into another problem.  I get
an error when compiling due to multiple definitions of mcount in the
entry_32.S and mcount_32.S files.  I looked at the code and decided
that disabling function tracing (CONFIG_FTRACE) in the kernel config
would eliminate the kernel function tracing and the ipipe function
tracing from stepping on each others' toes.  The kernel seems to be
compiling okay now (still in progress).  I just wanted to ask if this
is a bug or expected, and if the action I took was appropriate.  Do I
lose anything by disabling kernel function tracing?

Thanks,
James

Re: [Xenomai-help] Problem compiling patched kernel: multiple definitions of mcount

[Jan Kiszka]

James Doebbler wrote:
> Hello,
>  I am trying to patch and install a kernel using Xenomai (on Ubuntu
> 8.10).  I am using a vanilla linux-2.6.27 kernel and the latest
> xenomai version (2.4.6.1).  I had some earlier problems that I
> resolved by using the latest adeos patch (now I see that has been
> fixed in 2.4.6.1), but I've been running into another problem.  I get
> an error when compiling due to multiple definitions of mcount in the
> entry_32.S and mcount_32.S files.  I looked at the code and decided
> that disabling function tracing (CONFIG_FTRACE) in the kernel config
> would eliminate the kernel function tracing and the ipipe function
> tracing from stepping on each others' toes.  The kernel seems to be
> compiling okay now (still in progress).  I just wanted to ask if this
> is a bug or expected, and if the action I took was appropriate.  Do I
> lose anything by disabling kernel function tracing?

OK, that experience confirms my first glance on the latest patch /wrt
FTRACE integration: unfinished. The solution will be using the FTRACE
mcount functions for the ipipe tracer - should be feasible, except for
dynamic FRTRACE, ie. the thing that killed puppies and e1000 NICs. :)

Hope I'll find some time to look into this, but I can't promise when.
For now, if you want FTRACE (probably not, when you don't know it yet),
disable CONFIG_IPIPE_TRACE.

Jan

-- 
Siemens AG, Corporate Technology, CT SE 2 ES-OS
Corporate Competence Center Embedded Linux

Re: [Xenomai-help] Problem compiling patched kernel: multiple definitions of mcount

[James Doebbler]

Thanks for the reply, that clears things up nicely.  The rest of the
kernel finished compiling just fine, so I'll just continue on without
FTRACE for now.

James

On Thu, Nov 13, 2008 at 10:51 AM, Jan Kiszka <jan.kiszka@domain.hid> wrote:
> James Doebbler wrote:
>> Hello,
>>  I am trying to patch and install a kernel using Xenomai (on Ubuntu
>> 8.10).  I am using a vanilla linux-2.6.27 kernel and the latest
>> xenomai version (2.4.6.1).  I had some earlier problems that I
>> resolved by using the latest adeos patch (now I see that has been
>> fixed in 2.4.6.1), but I've been running into another problem.  I get
>> an error when compiling due to multiple definitions of mcount in the
>> entry_32.S and mcount_32.S files.  I looked at the code and decided
>> that disabling function tracing (CONFIG_FTRACE) in the kernel config
>> would eliminate the kernel function tracing and the ipipe function
>> tracing from stepping on each others' toes.  The kernel seems to be
>> compiling okay now (still in progress).  I just wanted to ask if this
>> is a bug or expected, and if the action I took was appropriate.  Do I
>> lose anything by disabling kernel function tracing?
>
> OK, that experience confirms my first glance on the latest patch /wrt
> FTRACE integration: unfinished. The solution will be using the FTRACE
> mcount functions for the ipipe tracer - should be feasible, except for
> dynamic FRTRACE, ie. the thing that killed puppies and e1000 NICs. :)
>
> Hope I'll find some time to look into this, but I can't promise when.
> For now, if you want FTRACE (probably not, when you don't know it yet),
> disable CONFIG_IPIPE_TRACE.
>
> Jan
>
> --
> Siemens AG, Corporate Technology, CT SE 2 ES-OS
> Corporate Competence Center Embedded Linux
>

Re: [Xenomai-help] Problem compiling patched kernel: multiple definitions of mcount

[Philippe Gerum]

Jan Kiszka wrote:
> James Doebbler wrote:
>> Hello,
>>  I am trying to patch and install a kernel using Xenomai (on Ubuntu
>> 8.10).  I am using a vanilla linux-2.6.27 kernel and the latest
>> xenomai version (2.4.6.1).  I had some earlier problems that I
>> resolved by using the latest adeos patch (now I see that has been
>> fixed in 2.4.6.1), but I've been running into another problem.  I get
>> an error when compiling due to multiple definitions of mcount in the
>> entry_32.S and mcount_32.S files.  I looked at the code and decided
>> that disabling function tracing (CONFIG_FTRACE) in the kernel config
>> would eliminate the kernel function tracing and the ipipe function
>> tracing from stepping on each others' toes.  The kernel seems to be
>> compiling okay now (still in progress).  I just wanted to ask if this
>> is a bug or expected, and if the action I took was appropriate.  Do I
>> lose anything by disabling kernel function tracing?
> 
> OK, that experience confirms my first glance on the latest patch /wrt
> FTRACE integration: unfinished.

Not quite. Not addressed would be more appropriate.

 The solution will be using the FTRACE
> mcount functions for the ipipe tracer - should be feasible, except for
> dynamic FRTRACE, ie. the thing that killed puppies and e1000 NICs. :)
> 
> Hope I'll find some time to look into this, but I can't promise when.
> For now, if you want FTRACE (probably not, when you don't know it yet),
> disable CONFIG_IPIPE_TRACE.
> 
> Jan
> 


-- 
Philippe.