[Freeipmi-devel] Several KCS fixes and new bug??

[Albert Chu]

I submitted a few kcs interface fixes:

kcs_write & kcs_write_interruptible: Unlock the semaphore on EBUSY error.

kcs_write_interruptible & kcs_cmd don't output sludge on an EAGAIN error.  

One issue I came upon is the following when the user is heavily writing
to the BMC.

User attempts to write command with kcs_write
EBUSY error occurs at some point, kcs_write returns -1.
User attempts to write new command with kcs_write.

Data written during the first kcs_write attempt may still be lingering
in the BMC's buffers, and will subsequently mess up the 2nd write's
attempt.  I noticed many C7h (data length invalid) errors when this
occured, but some C1h (invalid command) and C2h (command invalid for
given LUN) errors occurred as well.  

I attempted to do:

ibf_clear
clear_obj
kcs_abort

before ever issueing a kcs-write.  This helped a lot, but did not
eliminate all errors.  

This is definitely a bug.  But I'm not sure if this bug is solvable.  As
far as I can tell, there is nothing in the KCS interface that gives us a
way to deal with this.  This bug may simply be something that must be
dealt with from user land.

Al

--
Albert Chu
address@hidden
Lawrence Livermore National Laboratory