[etherlab-users] rtdm sdo_request patch files

[Graeme Foot]

Hi,

I have attached two patch files which add the following sdo functions to
the rtdm interface:
- ecrt_rtdm_sdo_request_timeout
- ecrt_rtdm_sdo_request_data
- ecrt_rtdm_sdo_request_data_size
- ecrt_rtdm_sdo_request_state
- ecrt_rtdm_sdo_request_read
- ecrt_rtdm_sdo_request_write


Attachments:
"etherlabmaster-1.5-2124-c_rtdm_full.patch" is a complete rtdm patch
file for 2124.
"etherlabmaster-1.5-2124-b_rtdm_changes_2.patch" is an update to my
previous patches with just the sdo additions.


Other changes:
- I renamed the struct "CstructMstrAttach" to "ec_rtdm_master_attach_t"
so that it conforms better to the ethercat coding standards.


Usage:
1) create the sdo request object as usual in non-realtime

  sdoReq = ecrt_slave_config_create_sdo_request(slaveConfig, 
                 index, subIndex, dataSize);

Note rtdm as you don't want memory allocations during realtime threads.
Also, if you are reading a variable length sdo, set "dataSize" to the
maximum memory required (the rtdm sdo read call will use as much as
required as long as it fits, but will not realloc new memory if it
doesn't).

2) to write an sdo value...

  2a) set up write:

    void *data = ecrt_rtdm_sdo_request_data(rtdmFD, sdoReq);
    EC_WRITE_S32(data, value);
    ecrt_rtdm_sdo_request_write(rtdmFD, sdoReq);

  2b) wait for write complete:

    switch (ecrt_rtdm_sdo_request_state(rtdmFD, sdoReq)) 
    {
      case EC_REQUEST_SUCCESS : ... success logic ...; break;
      case EC_REQUEST_ERROR :   ... error logic ...; break;
      default : break;
    }

3) to read an sdo value...

  3a) set up read:

    ecrt_rtdm_sdo_request_read(rtdmFD, sdoReq);

  3b) wait for read complete:

    switch (ecrt_rtdm_sdo_request_state(rtdmFD, sdoReq)) 
    {
      case EC_REQUEST_SUCCESS : 
      {
        void *data = ecrt_rtdm_sdo_request_data(rtdmFD, sdoReq);
        value = EC_READ_S32(data);
        ... success logic ...; 
      } break;
      case EC_REQUEST_ERROR : ... error logic ...; break;
      default : break;
    }


Let me know if there's something I've forgotten to mention.


Regards,
Graeme.


-----Original Message-----
From: Graeme Foot 
Sent: Wednesday, 11 January 2012 10:44
To: etherlab-users at etherlab.org
Cc: 'George Broz'
Subject: RE: [etherlab-users] API questions (plaintext repost)

Hi,

Re: RTDM patch for 2124

I did the update to RTDM to get the multiple domains going.  I haven't
heard the latest status either.


As for SDO's via RTDM, I was about to add some of them to the RTDM api,
in the next couple of weeks probably.

SDO requests take a bit of time to process and you don't want to block
the realtime thread, so you have a couple of options with SDO's via a
realtime thread:
1) pass off a request to a helper thread to get/set SDO information then
poll for a response.
2) make the SDO methods available via RTDM and set up a state machine to
monitor the SDO's progress.


For my application I'm thinking #2 is the slightly better option.


Graeme.


-----Original Message-----
From: etherlab-users-bounces at etherlab.org
[mailto:etherlab-users-bounces at etherlab.org] On Behalf Of George Broz
Sent: Tuesday, 10 January 2012 23:16
To: etherlab-users at etherlab.org
Subject: [etherlab-users] API questions (plaintext repost)

Hello,
 
I'm running a recent version (month or two old) of the IgH master stack
(it supports the e1000e driver on Linux 2.6.37.6) on Xenomai 2.5.6. I
had a handful of questions.
 
The information returned by ecrt_master_state() (slaves responding,
al_states, and link_up) returns stale information if the first slave is
unplugged after a slave network is brought up. The master has to be
restarted (/etc/init.d/ethercat restart) before the slave count is
accurate (and even then link_up reads incorrectly). Typical behavior or
a bug?
 
When do I need to use ecrt_master_callbacks()? I've read the docs and
source. An internal mutex protects the datagram queue. If I'm doing PDO
exchange (receive/process/queue/send) in one thread and want to
piggyback short SDO commands from another thread onto the cyclic packet,
do I need to use ecrt_master_callbacks() to coordinate? The source
already adds fsm packets asynchronously with respect to the cyclic
thread.
 
The state-machine implementation of the IgH stack is a pleasure to
follow. It would be helpful to have access to the state flags of these
state machines to integrate better with application code. Any plans for
that? 
 
In October, 2011 a patch (2124) was submitted for multi-domain support
through the RTDM interface. Can anyone comment on its status? I'd be
happy to help test this for the "default, (Xenomai-enabled)" branch. Are
there any plans to include RTDM and Xenomai support in the "stable"
branch?
 
The RTDM interface doesn't include support for SDOs b/c *configuration*
is not needed during realtime. But often you need to *query* for
additional information during realtime, e.g. the detailed cause of a
raised error bit in the PDO. AFAIK, without an RTDM interface, this
would drive the calling Xenomai task into secondary (Linux) mode
introducing other issues... for another newsgroup.
 
Thanks in advance for any answers/comments,

--George Broz
Moog, Inc. Industrial Group
_______________________________________________
etherlab-users mailing list
etherlab-users at etherlab.org
http://lists.etherlab.org/mailman/listinfo/etherlab-users
-------------- next part --------------
A non-text attachment was scrubbed...
Name: etherlabmaster-1.5-2124-b_rtdm_changes_2.patch
Type: application/octet-stream
Size: 22642 bytes
Desc: etherlabmaster-1.5-2124-b_rtdm_changes_2.patch
URL: <http://lists.etherlab.org/pipermail/etherlab-users/attachments/20120117/87238af8/attachment-0008.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: etherlabmaster-1.5-2124-c_rtdm_full.patch
Type: application/octet-stream
Size: 152213 bytes
Desc: etherlabmaster-1.5-2124-c_rtdm_full.patch
URL: <http://lists.etherlab.org/pipermail/etherlab-users/attachments/20120117/87238af8/attachment-0009.obj>