Inequivalent aliases when stracing tst-cputimer1.

Inequivalent aliases when stracing tst-cputimer1.

[Carlos O'Donell]

James, Dave,

While running some glibc tests on 2.6.39-rc3a+ I noticed the following:
~~~
...
[ 7649.424000] INEQUIVALENT ALIASES 0x13000 and 0x14000 in file tst-cputimer1
[ 7649.568000] INEQUIVALENT ALIASES 0x13000 and 0x14000 in file tst-cputimer1
[ 7649.720000] INEQUIVALENT ALIASES 0x13000 and 0x14000 in file tst-cputimer1
...
~~~

This is a binutils head, building glibc head, with gcc 4.4.4.

This *only* happens when I strace the application.

While the application is being strace'd it continually dumps the same
message about inequivalent aliases.

Any idea what's going on here?

How can I fix it? :-)

Cheers,
Carlos.

Re: Inequivalent aliases when stracing tst-cputimer1.

[John David Anglin]

On 7/3/2011 10:39 PM, Carlos O'Donell wrote:
> James, Dave,
>
> While running some glibc tests on 2.6.39-rc3a+ I noticed the following:
> ~~~
> ...
> [ 7649.424000] INEQUIVALENT ALIASES 0x13000 and 0x14000 in file tst-cputimer1
> [ 7649.568000] INEQUIVALENT ALIASES 0x13000 and 0x14000 in file tst-cputimer1
> [ 7649.720000] INEQUIVALENT ALIASES 0x13000 and 0x14000 in file tst-cputimer1
> ...
> ~~~
>
> This is a binutils head, building glibc head, with gcc 4.4.4.
>
> This *only* happens when I strace the application.
>
> While the application is being strace'd it continually dumps the same
> message about inequivalent aliases.
>
> Any idea what's going on here?
>
> How can I fix it? :-)

You need to update binutils to get new linker (actually, it's the 
default linker script that's changed).
The new linker script starts the data section on a page boundary.

It's not GCC version dependent.  However, newer GCC versions have 
improved delay slot handling.

It's caused by overlapping mappings of text and data which you can see 
by dumping the maps file
for the process.  I think I may have back ported the fix.  If you build 
binutils head, you will have to
configure with --disable-initfini-array to build glibc.

I don't know all the circumstances under which the message appears.  It 
always occurs on a core
dump.  It occurs when a page needs to be flushed and there are two or 
more inequivalent aliases
to the page.

It's probably harmless on A500.  It affects PA8800/PA8900.

Dave

Re: Inequivalent aliases when stracing tst-cputimer1.

[John David Anglin]

On 7/4/2011 9:28 AM, John David Anglin wrote:
> On 7/3/2011 10:39 PM, Carlos O'Donell wrote:
>> This is a binutils head, building glibc head, with gcc 4.4.4. You 
>> need to update binutils to get new linker (actually, it's the default 
>> linker script that's changed).
>> The new linker script starts the data section on a page boundary.

Depending on how GCC is configured, it may or may not use your binutils 
head linker.  gcc -v should show
what's going on.

Dave

Re: Inequivalent aliases when stracing tst-cputimer1.

[Carlos O'Donell]

On Mon, Jul 4, 2011 at 10:21 AM, John David Anglin <dave.anglin@bell.ne=
t> wrote:
> On 7/4/2011 9:28 AM, John David Anglin wrote:
>>
>> On 7/3/2011 10:39 PM, Carlos O'Donell wrote:
>>>
>>> This is a binutils head, building glibc head, with gcc 4.4.4. You n=
eed to
>>> update binutils to get new linker (actually, it's the default linke=
r script
>>> that's changed).
>>> The new linker script starts the data section on a page boundary.
>
> Depending on how GCC is configured, it may or may not use your binuti=
ls head
> linker. =A0gcc -v should show
> what's going on.

You are correct, that was my mistake. I had rebuilt this particular
test without using binutils head, once I rebuilt with the newer linker
script, the two load segments have the correct alignment and the
application no longer exhibits these inequivalent aliases under
strace.

Thanks, that's one problem fixed.

Cheers,
Carlos.
--
To unsubscribe from this list: send the line "unsubscribe linux-parisc"=
 in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html