LinCAN sources go through big white-space cleanup. Signed-off-by: Pavel Pisa <pisa@cmp.felk.cvut.cz>
The LinCAN driver license unified according to DCE FEE CTU head and superiors request. This is horribly long time pending request but I hold it to simplify merging of contributed changes. But it cannot be hold longer when new release is in preparation. Excuse me for inconvenience.
The first round of I/O space pointers separation. The new 2.6.x Linux kernels separates I/O space addresses from kernel and user space ones. They are defined as (void __iomem *). To follow this rule and suppress some warnings LinCAN defines type can_ioptr_t now which allows to keep code compatible between 2.4.x and 2.6.x kernels. These changes exposes more other inconsistencies in some boards support code which should be corrected one by one later. The code requires testing and checking on more hardware.
Elimination of superfluous chips header files
To prevent future name collisions "chip_t" changed to "canchip_t"
Changed filling of the chip specific operations. This enables simpler fine tuning for specific boards which needs modified functions. ENS CPC-PCI requires special interrupt handling for example.
LinCAN version updated to 0.3
Added prefix sja to all sjaXX_YY register bit-fields definitions. Changed to solve ARM-linux kernel and LinCAN name collision.
Changed data size from byte to unsigned for write_register indirect call This is required for the support of the cards connected to the wider bus architecture. This should not have negative impact, because char has been propagated to int by C call convention anyway. The global inline canobj_write_reg function has been left intact (byte only), because it could better optimize for static call case.
LinCAN can be compiled in mode with RT-Linux chip worker threads now. Setup tested to work with Virtual and PCM3680 boards now.
Header-files cleanup and CAN queue edges and ends locking reimplemented. The code has been updated and tested with 2.6.0-test9 as well. Default asynchronous close does not wait for transmission of the messages left in the Tx FIFOs. ENDS dispose is delayed in such case. Virtual chip driver can model bus latency if nonzero baudrate is selected.
LinCAN driver major structured comments and documentation update
CAN driver infrastructure redesign to LinCAN-0.2 version
The original version of Arnaud Westenberg Linux CAN-bus driver can-0.7.1 (http://home.wanadoo.nl/arnaud/)