[Xenomai-core] [git pull] RTDM fixes and reworks

[Xenomai-core] [git pull] RTDM fixes and reworks

[Jan Kiszka]

The following changes since commit 113ea4d56e8b215cb56ae7673013163ea5a5987d:
  Gilles Chanteperdrix (1):
        switchtest: increase stack sizes

are available in the git repository at:

  git://git.xenomai.org/xenomai-jki.git for-upstream


This is the first of likely three rounds of patches for 2.5.x. It mostly
deals with introducing rtdm_rt_capable to replace the
__xn_exec_conforming marks of RTDM syscalls and with deprecating RT
open/socket/close for RTDM devices. The latter affects quite a few
devices in our tree, so I'm CC'ing the maintainers to check if they see
any problems in dropping RT creation/closure at this point.

Besides some smaller fixes and an irqbench refactoring, this series also
contains a solution for the confusions about incompatible rttest<n>
devices. It simply avoids collisions be using different device names for
the different device subclasses.

Please review and merge if no concerns remain.

BTW, the other two rounds I mentioned will deal with /proc overflow
fixes and avoiding RTDM context polling on close (I still need to finish
my unit test, therefore this delay). Those should follow over the
weekend.


Jan Kiszka (16):
      RTDM: Add rtdm_rt_capable() service
      RTDM: Bump API version and document changes
      Revert "rtdm: tag syscalls as conforming"
      RTDM: Deprecate open_rt, socket_rt, and close_rt
      RTIPC: Fix memory leak on failing socket creation
      RTIPC: Drop support for RT socket creation/deletion
      RTCAN: Drop support for socket creation/deletion in real-time
      16550A: Drop support for device opening/closing in real-time
      Analogy: Drop support for opening/closing in real-time
      RTDM: Instrument rtdm_context_lock/unlock to detect misuses
      RTDM: Add rtdm_context_put()
      Fix historic msleep wrapping
      irqbench: Fix x86-64 build
      irqbench: Refactor user space helpers
      irqbench: Auto-detect bases of port address and IRQ values
      rttest: Resolved test device conflicts via separate name spaces

Wolfgang Mauerer (1):
      RTDM: Fix potential NULL pointer dereference

 configure.in                          |    2 +-
 include/asm-generic/wrappers.h        |    2 +-
 include/rtdm/rtdm.h                   |    4 +-
 include/rtdm/rtdm_driver.h            |   27 ++-
 include/rtdm/rttesting.h              |   10 +-
 ksrc/drivers/analogy/rtdm_interface.c |    2 -
 ksrc/drivers/can/rtcan_raw.c          |    2 -
 ksrc/drivers/ipc/rtipc.c              |   13 +-
 ksrc/drivers/serial/16550A.c          |   51 +---
 ksrc/drivers/testing/irqbench.c       |    4 +-
 ksrc/drivers/testing/switchtest.c     |    4 +-
 ksrc/drivers/testing/timerbench.c     |    4 +-
 ksrc/skins/rtdm/API.CHANGES           |    9 +
 ksrc/skins/rtdm/core.c                |   41 +++-
 ksrc/skins/rtdm/device.c              |   13 +-
 ksrc/skins/rtdm/drvlib.c              |   25 ++
 ksrc/skins/rtdm/syscall.c             |   10 +-
 src/testsuite/irqbench/Makefile.am    |    2 +-
 src/testsuite/irqbench/irqbench.c     |  480 ++++++++++++++++-----------------
 src/testsuite/irqbench/irqloop.c      |  304 +++++++++++----------
 src/testsuite/latency/latency.c       |    2 +-
 src/testsuite/switchtest/switchtest.c |   50 +---
 22 files changed, 554 insertions(+), 507 deletions(-)

Re: [Xenomai-core] [git pull] RTDM fixes and reworks

[Gilles Chanteperdrix]

Jan Kiszka wrote:
> The following changes since commit 113ea4d56e8b215cb56ae7673013163ea5a5987d:
>   Gilles Chanteperdrix (1):
>         switchtest: increase stack sizes
> 
> are available in the git repository at:
> 
>   git://git.xenomai.org/xenomai-jki.git for-upstream
> 
> 
> This is the first of likely three rounds of patches for 2.5.x. It mostly
> deals with introducing rtdm_rt_capable to replace the
> __xn_exec_conforming marks of RTDM syscalls and with deprecating RT
> open/socket/close for RTDM devices. The latter affects quite a few
> devices in our tree, so I'm CC'ing the maintainers to check if they see
> any problems in dropping RT creation/closure at this point.
> 
> Besides some smaller fixes and an irqbench refactoring, this series also
> contains a solution for the confusions about incompatible rttest<n>
> devices. It simply avoids collisions be using different device names for
> the different device subclasses.

switchtest already used another solution for that problem: the different
devices use different ioctls. So, IMO we already had a solution for that
issue.

-- 
					    Gilles.

Re: [Xenomai-core] [git pull] RTDM fixes and reworks

[Jan Kiszka]

Gilles Chanteperdrix wrote:
> Jan Kiszka wrote:
>> The following changes since commit 113ea4d56e8b215cb56ae7673013163ea5a5987d:
>>   Gilles Chanteperdrix (1):
>>         switchtest: increase stack sizes
>>
>> are available in the git repository at:
>>
>>   git://git.xenomai.org/xenomai-jki.git for-upstream
>>
>>
>> This is the first of likely three rounds of patches for 2.5.x. It mostly
>> deals with introducing rtdm_rt_capable to replace the
>> __xn_exec_conforming marks of RTDM syscalls and with deprecating RT
>> open/socket/close for RTDM devices. The latter affects quite a few
>> devices in our tree, so I'm CC'ing the maintainers to check if they see
>> any problems in dropping RT creation/closure at this point.
>>
>> Besides some smaller fixes and an irqbench refactoring, this series also
>> contains a solution for the confusions about incompatible rttest<n>
>> devices. It simply avoids collisions be using different device names for
>> the different device subclasses.
> 
> switchtest already used another solution for that problem: the different
> devices use different ioctls. So, IMO we already had a solution for that
> issue.

Look at the diffstat of switchtest. I wanted to fix the issue for the
rest without replicating code.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Xenomai-core] [git pull] RTDM fixes and reworks

[Gilles Chanteperdrix]

Jan Kiszka wrote:
> Gilles Chanteperdrix wrote:
>> Jan Kiszka wrote:
>>> The following changes since commit 113ea4d56e8b215cb56ae7673013163ea5a5987d:
>>>   Gilles Chanteperdrix (1):
>>>         switchtest: increase stack sizes
>>>
>>> are available in the git repository at:
>>>
>>>   git://git.xenomai.org/xenomai-jki.git for-upstream
>>>
>>>
>>> This is the first of likely three rounds of patches for 2.5.x. It mostly
>>> deals with introducing rtdm_rt_capable to replace the
>>> __xn_exec_conforming marks of RTDM syscalls and with deprecating RT
>>> open/socket/close for RTDM devices. The latter affects quite a few
>>> devices in our tree, so I'm CC'ing the maintainers to check if they see
>>> any problems in dropping RT creation/closure at this point.
>>>
>>> Besides some smaller fixes and an irqbench refactoring, this series also
>>> contains a solution for the confusions about incompatible rttest<n>
>>> devices. It simply avoids collisions be using different device names for
>>> the different device subclasses.
>> switchtest already used another solution for that problem: the different
>> devices use different ioctls. So, IMO we already had a solution for that
>> issue.
> 
> Look at the diffstat of switchtest. I wanted to fix the issue for the
> rest without replicating code.

I would agree with you for patches in xenomai-head. But in that case,
this is an ABI breakage.

-- 
					    Gilles.

Re: [Xenomai-core] [git pull] RTDM fixes and reworks

[Jan Kiszka]

Gilles Chanteperdrix wrote:
> Jan Kiszka wrote:
>> Gilles Chanteperdrix wrote:
>>> Jan Kiszka wrote:
>>>> The following changes since commit 113ea4d56e8b215cb56ae7673013163ea5a5987d:
>>>>   Gilles Chanteperdrix (1):
>>>>         switchtest: increase stack sizes
>>>>
>>>> are available in the git repository at:
>>>>
>>>>   git://git.xenomai.org/xenomai-jki.git for-upstream
>>>>
>>>>
>>>> This is the first of likely three rounds of patches for 2.5.x. It mostly
>>>> deals with introducing rtdm_rt_capable to replace the
>>>> __xn_exec_conforming marks of RTDM syscalls and with deprecating RT
>>>> open/socket/close for RTDM devices. The latter affects quite a few
>>>> devices in our tree, so I'm CC'ing the maintainers to check if they see
>>>> any problems in dropping RT creation/closure at this point.
>>>>
>>>> Besides some smaller fixes and an irqbench refactoring, this series also
>>>> contains a solution for the confusions about incompatible rttest<n>
>>>> devices. It simply avoids collisions be using different device names for
>>>> the different device subclasses.
>>> switchtest already used another solution for that problem: the different
>>> devices use different ioctls. So, IMO we already had a solution for that
>>> issue.
>> Look at the diffstat of switchtest. I wanted to fix the issue for the
>> rest without replicating code.
> 
> I would agree with you for patches in xenomai-head. But in that case,
> this is an ABI breakage.

It's internal testing stuff.

If you think anyone out there may have written his own back- or
front-end, I could make all test devices additionally register under the
old rttest<n> as well. I just don't think it's worth the effort.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Xenomai-core] [git pull] RTDM fixes and reworks

[Gilles Chanteperdrix]

Jan Kiszka wrote:
> Gilles Chanteperdrix wrote:
>> Jan Kiszka wrote:
>>> Gilles Chanteperdrix wrote:
>>>> Jan Kiszka wrote:
>>>>> The following changes since commit 113ea4d56e8b215cb56ae7673013163ea5a5987d:
>>>>>   Gilles Chanteperdrix (1):
>>>>>         switchtest: increase stack sizes
>>>>>
>>>>> are available in the git repository at:
>>>>>
>>>>>   git://git.xenomai.org/xenomai-jki.git for-upstream
>>>>>
>>>>>
>>>>> This is the first of likely three rounds of patches for 2.5.x. It mostly
>>>>> deals with introducing rtdm_rt_capable to replace the
>>>>> __xn_exec_conforming marks of RTDM syscalls and with deprecating RT
>>>>> open/socket/close for RTDM devices. The latter affects quite a few
>>>>> devices in our tree, so I'm CC'ing the maintainers to check if they see
>>>>> any problems in dropping RT creation/closure at this point.
>>>>>
>>>>> Besides some smaller fixes and an irqbench refactoring, this series also
>>>>> contains a solution for the confusions about incompatible rttest<n>
>>>>> devices. It simply avoids collisions be using different device names for
>>>>> the different device subclasses.
>>>> switchtest already used another solution for that problem: the different
>>>> devices use different ioctls. So, IMO we already had a solution for that
>>>> issue.
>>> Look at the diffstat of switchtest. I wanted to fix the issue for the
>>> rest without replicating code.
>> I would agree with you for patches in xenomai-head. But in that case,
>> this is an ABI breakage.
> 
> It's internal testing stuff.
> 
> If you think anyone out there may have written his own back- or
> front-end, I could make all test devices additionally register under the
> old rttest<n> as well. I just don't think it's worth the effort.

No, the point is anyone out there with a 2.5.0 user-space Xenomai should
be able to run it with Xenomai 2.5.3. So, there should be at least a
Kconfig option to keep the old names.


-- 
					    Gilles.

Re: [Xenomai-core] [git pull] RTDM fixes and reworks

[Jan Kiszka]

Gilles Chanteperdrix wrote:
> Jan Kiszka wrote:
>> Gilles Chanteperdrix wrote:
>>> Jan Kiszka wrote:
>>>> Gilles Chanteperdrix wrote:
>>>>> Jan Kiszka wrote:
>>>>>> The following changes since commit 113ea4d56e8b215cb56ae7673013163ea5a5987d:
>>>>>>   Gilles Chanteperdrix (1):
>>>>>>         switchtest: increase stack sizes
>>>>>>
>>>>>> are available in the git repository at:
>>>>>>
>>>>>>   git://git.xenomai.org/xenomai-jki.git for-upstream
>>>>>>
>>>>>>
>>>>>> This is the first of likely three rounds of patches for 2.5.x. It mostly
>>>>>> deals with introducing rtdm_rt_capable to replace the
>>>>>> __xn_exec_conforming marks of RTDM syscalls and with deprecating RT
>>>>>> open/socket/close for RTDM devices. The latter affects quite a few
>>>>>> devices in our tree, so I'm CC'ing the maintainers to check if they see
>>>>>> any problems in dropping RT creation/closure at this point.
>>>>>>
>>>>>> Besides some smaller fixes and an irqbench refactoring, this series also
>>>>>> contains a solution for the confusions about incompatible rttest<n>
>>>>>> devices. It simply avoids collisions be using different device names for
>>>>>> the different device subclasses.
>>>>> switchtest already used another solution for that problem: the different
>>>>> devices use different ioctls. So, IMO we already had a solution for that
>>>>> issue.
>>>> Look at the diffstat of switchtest. I wanted to fix the issue for the
>>>> rest without replicating code.
>>> I would agree with you for patches in xenomai-head. But in that case,
>>> this is an ABI breakage.
>> It's internal testing stuff.
>>
>> If you think anyone out there may have written his own back- or
>> front-end, I could make all test devices additionally register under the
>> old rttest<n> as well. I just don't think it's worth the effort.
> 
> No, the point is anyone out there with a 2.5.0 user-space Xenomai should
> be able to run it with Xenomai 2.5.3. So, there should be at least a
> Kconfig option to keep the old names.

OK, that's fine. Will add a switch to register devices with the old
names instead.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Xenomai-core] [git pull] RTDM fixes and reworks

[Gilles Chanteperdrix]

Jan Kiszka wrote:
> Gilles Chanteperdrix wrote:
>> Jan Kiszka wrote:
>>> Gilles Chanteperdrix wrote:
>>>> Jan Kiszka wrote:
>>>>> Gilles Chanteperdrix wrote:
>>>>>> Jan Kiszka wrote:
>>>>>>> The following changes since commit 113ea4d56e8b215cb56ae7673013163ea5a5987d:
>>>>>>>   Gilles Chanteperdrix (1):
>>>>>>>         switchtest: increase stack sizes
>>>>>>>
>>>>>>> are available in the git repository at:
>>>>>>>
>>>>>>>   git://git.xenomai.org/xenomai-jki.git for-upstream
>>>>>>>
>>>>>>>
>>>>>>> This is the first of likely three rounds of patches for 2.5.x. It mostly
>>>>>>> deals with introducing rtdm_rt_capable to replace the
>>>>>>> __xn_exec_conforming marks of RTDM syscalls and with deprecating RT
>>>>>>> open/socket/close for RTDM devices. The latter affects quite a few
>>>>>>> devices in our tree, so I'm CC'ing the maintainers to check if they see
>>>>>>> any problems in dropping RT creation/closure at this point.
>>>>>>>
>>>>>>> Besides some smaller fixes and an irqbench refactoring, this series also
>>>>>>> contains a solution for the confusions about incompatible rttest<n>
>>>>>>> devices. It simply avoids collisions be using different device names for
>>>>>>> the different device subclasses.
>>>>>> switchtest already used another solution for that problem: the different
>>>>>> devices use different ioctls. So, IMO we already had a solution for that
>>>>>> issue.
>>>>> Look at the diffstat of switchtest. I wanted to fix the issue for the
>>>>> rest without replicating code.
>>>> I would agree with you for patches in xenomai-head. But in that case,
>>>> this is an ABI breakage.
>>> It's internal testing stuff.
>>>
>>> If you think anyone out there may have written his own back- or
>>> front-end, I could make all test devices additionally register under the
>>> old rttest<n> as well. I just don't think it's worth the effort.
>> No, the point is anyone out there with a 2.5.0 user-space Xenomai should
>> be able to run it with Xenomai 2.5.3. So, there should be at least a
>> Kconfig option to keep the old names.
> 
> OK, that's fine. Will add a switch to register devices with the old
> names instead.

We need the other way around too: the new tools should be able to fall
back to the old name scheme (so in particular, we should keep the
switchtest portion, because it works better with the old name scheme).

-- 
					    Gilles.