[Xenomai] Rework of mlockall and dlopen handling

[Xenomai] Rework of mlockall and dlopen handling

[Jan Kiszka]

Hi,

I've coded down what we discussed and pushed it into 'for-upstream'.
This does not yet remove redundant mlockall from included userspace
code or change the docs, but the core mechanisms are ready. Patch list
below summarizes the steps.

Jan


Jan Kiszka (4):
      Invoke mlockall on every skin library initialization
      posix: Control auto-shadowing via environment variable
      Remove mlockall alert handler
      Drop --enable-dlopen, disable optimized TLS support by default

 configure.in               |   70 +++----------------------------------------
 include/asm-generic/bind.h |   10 ------
 src/skins/common/bind.c    |   24 ++++-----------
 src/skins/posix/init.c     |   52 +++++++++++---------------------
 src/skins/psos+/init.c     |   15 +---------
 src/skins/uitron/init.c    |   15 +---------
 6 files changed, 31 insertions(+), 155 deletions(-)

Re: [Xenomai] Rework of mlockall and dlopen handling

[Gilles Chanteperdrix]

On 04/23/2013 05:33 PM, Jan Kiszka wrote:

> Hi,
> 
> I've coded down what we discussed and pushed it into 'for-upstream'.
> This does not yet remove redundant mlockall from included userspace
> code or change the docs, but the core mechanisms are ready. Patch list
> below summarizes the steps.


I would have been a bit less extreme, and simply inverted the default
value of --enable-dlopen-skins and --enable-*-auto-mlockall, but what
you did is fine by me.

Maybe we could keep --disable-dlopen-skins though?


-- 
                                                                Gilles.

Re: [Xenomai] Rework of mlockall and dlopen handling

[Jan Kiszka]

On 2013-04-23 20:30, Gilles Chanteperdrix wrote:
> On 04/23/2013 05:33 PM, Jan Kiszka wrote:
> 
>> Hi,
>>
>> I've coded down what we discussed and pushed it into 'for-upstream'.
>> This does not yet remove redundant mlockall from included userspace
>> code or change the docs, but the core mechanisms are ready. Patch list
>> below summarizes the steps.
> 
> 
> I would have been a bit less extreme, and simply inverted the default
> value of --enable-dlopen-skins and --enable-*-auto-mlockall, but what
> you did is fine by me.
> 
> Maybe we could keep --disable-dlopen-skins though?

...and translate it to --with-__thread? Or what should be the semantics?

Jan

-- 
Siemens AG, Corporate Technology, CT RTC ITP SDP-DE
Corporate Competence Center Embedded Linux

Re: [Xenomai] Rework of mlockall and dlopen handling

[Gilles Chanteperdrix]

On 04/23/2013 08:36 PM, Jan Kiszka wrote:

> On 2013-04-23 20:30, Gilles Chanteperdrix wrote:
>> On 04/23/2013 05:33 PM, Jan Kiszka wrote:
>>
>>> Hi,
>>>
>>> I've coded down what we discussed and pushed it into 'for-upstream'.
>>> This does not yet remove redundant mlockall from included userspace
>>> code or change the docs, but the core mechanisms are ready. Patch list
>>> below summarizes the steps.
>>
>>
>> I would have been a bit less extreme, and simply inverted the default
>> value of --enable-dlopen-skins and --enable-*-auto-mlockall, but what
>> you did is fine by me.
>>
>> Maybe we could keep --disable-dlopen-skins though?
> 
> ...and translate it to --with-__thread? Or what should be the semantics?


Yes, you are right, it is useless.

-- 
                                                                Gilles.