Hi,
I have attached two patches:
1) The extracted RTDM code from devel 2117, for stable 2124
(etherlabmaster-1.5-2124-a_rtdm_from_2117.patch)
2) In addition to the above, some bug fixes and changes to allow
multiple domains
(etherlabmaster-1.5-2124-c_rtdm_full.patch)
A summary of the changes I have made:
- added a EC_RTDM_MAX_DOMAINS define to specify the maximum domains per
rtdm connection
- changed CstructMstrAttach by changing the domainIndex to a domainCnt
and domainIndices array
- previously the CstructMstrAttach.masterIndex parameter was ignored,
this is now used
- the following methods now also take an index parameter:
ecrt_rtdm_domain_process
ecrt_rtdm_domain_queue
ecrt_rtdm_domain
- added methods:
ecrt_rtdm_domain_process_all
ecrt_rtdm_domain_queue_all
ecrt_rtdm_domain_states
- I have corrected the typo queque -> queue
- There are a few fixes to remove compiler warnings
- I have updated both the RTAI and Xenomi examples (but have not
compiled
the Xenomi examples)
Andreas,
I have not had any problems with locking issues at the moment.
My system is RTAI 3.8.1 on Linux 2.6.29.5.
(please also ignore my previous email, I created the patch against the
wrong directory.)
Regards,
Graeme.
-----Original Message-----
From: Andreas Stewering-Bone [mailto:[email protected]]
Sent: Friday, 28 October 2011 19:51
To: Graeme Foot
Subject: Re: RTDM patch 2117 -> 2124
Hello Graeme,
Am 28.10.2011 00:49, schrieb Graeme Foot:
Hi,
I have attached the patch that extracts the RTDM (RTAI& Xenomai) code
from devel branch (2117) for stable branch 2124.
This patch does not include any of my changes as I am still working on
them.
I have also come across a few bugs which I will fix:
ec_rtdm.h
- A couple of incorrectly named constants in rt_dev_ioctl calls:
EC_RTDM_MASTER_MONITOR_QUEQUE ->
EC_RTDM_MASTER_SYNC_MONITOR_QUEUE
EC_RTDM_MASTER_MONITOR_PROCESS ->
EC_RTDM_MASTER_SYNC_MONITOR_PROCESS
rtdm/module.c
- The MasterAttach function does not use the masterindex parameter.
Instead it just uses the internal masterno value (which is the device
name postfix number).
Once I've made my changes would you prefer a delta to the above patch
or a complete RTDM patch?
A complete patch would be better.
I'm using RTAI and yes I will develop against the stable branch.
On my system 2.6.32.11 and RTAI 3.8.1 I noticed some locking problems.
What is your experience under RTAI?
Under Xenomai it runs fine.
One question about code requirements, do you allow gcc compound
statement macros? eg:
I think you can use it because Linux Kernel compilation will happen
under gcc.
#define fn(idx, arg) \
({ data_t data; \
data.idx = idx; \
int ret = fn2(data); \
arg = data.arg; \
ret; })
The reason I ask is the ecrt_rtdm_domain_state function will require
two parameters, domainIndex (in) and domainState (out) which I will need
to combine into one data structure.
If I can't use a compound statement I will need to use an inline
function, but this will require ecrt.h to be included. Or I could just
require the user to fill in the data structure.
Do you want me to post these to the forum as well?
It is your decision.
But we would prefer it.
Best regards
Andreas
Regards,
Graeme.
