Working RTAI RTDM module
LXRT exmaple application to be tested

Working Xenomai Posix example

Work on Xenomai Posix interface, to be tested
Further work on RTAI rtdm interface needed

Fix application time for RTDM interface

First working RTDM Version,
to be reviewed,
further testing and optimation needed

Thanks to Moehwald GmbH B.Benner for Basic Code

Add drivers for kernel 2.6.33
Thanks to
Dipl.-Inform. J?rgen Kunz
Technische Universit?t Darmstadt
for e100, e1000, r8169 driver

Merge, adding 8139too for 2.6.32

Added e1000 driver for 2.6.28.

Changed accidentally used & operator to &&.

Add Driver 8139too for kernel 2.6.32

Merged.

fixed kernel oops bug in ec_fsm_coe_down_response (Blame the C Compiler)

merged

Reduced frame size when sending to mailbox, uses 2 datagrams. Enabled with configure option --enable-mboxframesize

fixed race condition when comparing datagram state
EC_DATAGRAM_QUEUED must be compared before EC_DATAGRAM_SENT because it might change in between

Added tag Stable for changeset ce2fae4d3c4f

merged

Fixed segmented sdo dictionary upload (only first segment contains list type)

lock io_mutex when unqueueing datagrams; unqueue datagrams from send-receive queue.

Minor changes (consistent variable naming)

Datagram queue race fixed: insert datagrams into master-fsm_queue with fsm_queue list_head

Optionally compile with trace_printk support

properly deactivate slaves if there is no eoe slave

always use rt_mutex instead of semaphore for mutual exclusion on kernels >= 2.6.24 (removed --enable-mutex build option)

releasing of refcounted requests fixed: do not call EC_SLAVE_DBG without valid slave

e1000 on 2.6.33: handle link-down

SoE request wait_event() deadlock fixed: use wait_event_interruptible and refcount ec_master_soe_request_t objects with kref

Register read/write wait_event() deadlock fixed: use wait_event_interruptible and refcount ec_reg_request_t objects with kref

Sii write wait_event() deadlock fixed: use wait_event_interruptible and refcount ec_master_sii_write_request_t objects with kref

FoE wait_event() deadlock fixed: refcount ec_master_foe_request_t objects with kref

use rt_mutex instead of semaphore for mutual exclusion (use --enable-mutex)

SDO wait_event() deadlock fixed: refcount ec_master_sdo_request_t objects with kref

Failed FoE transfer: notify foe_queue instead of sdo_queue

fixed warning due to 64bit size_t

ecrt_master_callbacks: replaced send/receive callback mechanism with locking mechanism; removed ecrt_master_send_ext

fixed 'ethercat eoe' command: interchanged rx/tx bytes/rates

renamed ext_injection_timeout -> fsm_injection_timeout

removed now obsolete ext_datagram_queu

removed eoe thread: eoe processing now done by Idle/Operation thread

use ext_queue_sem for fsm datagram queue (instead of io_sem)

replaced injection_seq mechanism with fsm datagram queue

Increment injection_seq on slave state changes

ec_master_queue_external_datagram: fixed missing up(io_sem)

merge

Be more tolerant when selecting slaves and domains.

Fixed / improved config selection.

Assume increasing ranges if second number is omitted.

merge.

Implemented ranges for slaves, configs and domains.

Added ecrt_master_reset() method.

Inplace I/O bug: correct domain tx-size when tx-size of input and output fmmu is different.

fixed compiling without EC_HAVE_CYCLES (dc_jiffies_app_start_time)

Minor 64bit improvements (use %zu when formatting size_t values in printk)

fixed ioctl codes, increased ioctl version magic

Merged

Re-enabled waiting for clock synchronity before setting dc offsets. (changeset b0dcdfbd4238)

fixed race when calculating DC offsets and start times due to ec_slave_config_attach: removed obsolete request to change slave state to OP.
The request to change to OP is done by ec_master_request_op after activating the master.

renamed dc_cycles_app_time -> dc_cycles_app_start_time & dc_jiffies_app_time -> dc_jiffies_app_start_time

Idle thread considers fsm step prepared by Op-thread (Otherwise some state changes may get lost)

use app_start_time for initial DC offset calculation because app_time might be overwritten while calculating the offsets.

Use ESC port order (0-3-1-2) for bus topology and DC delay calculation

hg ignored debuild.log

Added e1000 port for 2.6.31

merged