Re: [pok-devel] ARINC - partition

[Julien Delange <julien@xxxxxxxxx>]

Le Mon, Feb 13, 2012 at 01:57:33PM +0100, Matias Hastaran :
> I look forward to hearing from you about the patch,

Dear Mathias,

Thanks for submitting this patch, this is a good step to ensure the
ARINC653 compatibility, especially because we support just a subset of
the standard. However, I have some comments :
1. The kernel would not be specific to any existing partitioning
   standard. Instead, we should keep the kernel generic and as small as
   possible.
2. We should not duplicate structures and especially members between
   different functions. This would make difficult to keep system
   consistency (the same requirement/property in two structures could be
   quite confusing and difficult to ensure that parameters have the same
   value everytime). In particular, the period/duration/identifier
   can be retrieved from the existing pok_partition_t structure.
3. Instead of making a new structure in the kernel, it would be better
   to extend the pok_partition_t structure by:
     a. Adding a start condition
	 b. Adding the lock level
4. One these parameters are added, you can add new syscall to get
   atomically the period, duration, lock_level, operating_mode and start
   condition. Then, the arinc653 layer in libpok can fill the
   ARINC653-dedicated  structure by performing appropriate syscall to
   the kernel.
   
But we have to keep in mind that we have to be standard-agnostic on the
kernel-side and avoid any specific function against a standard. As a
result, all ARINC653 functions should be kept in the libpok/ library,
and nothing related to arinc should reside in the kernel.

Sorry for this long list of comments and this first shot is quite
promising and it could be useful to keep the initial guidelines,
especially for potential regular committers :-)

Best,