On Mon, 13 Jun 2005, Tom Duffy wrote:
On Mon, 2005-06-13 at 17:10 -0400, James Lentini wrote:
This looks good. I've checked it in as revision 2595.
Thanks.
Eventually, we could move the list up into the dapl_common structure.
Gross. I agree with Christoph.
You find the dapl_common structure gross from a style point of view,
right? I agree that some of the fields currently in dapl_common are
not actually necessary for every DAPL object (e.g. a dapl_ia does not
need an owner_ia pointer). I see these as implementation errors, but I
still see value in using a common structure in principal. Is there a
performance problem that I am missing?
The one oddity is that all the objects are put on the IA list except
CRs which are placed on an SP. I think it would be worth reflecting
this in the list's member name. Instead of calling it a ia_list_entry
as it is now, a name like parent_list or owner_list would be more
appropriate.
Eventually, wouldn't we get rid of ia_list_entry? Cause each struct has
its own list entry (name "list", which seems to be the convention), and
the IA has the heads of each list.
That would be the other alternative.
Right now, dapl_common has an ia_list_entry member. For each DAPL
object, this list entry that gets linked on a list head stored in the
parent IA. There are two exceptions to this rule:
- a CR's ia_list_entry is linked onto its parent SP's cr list.
- an IA's ia_list_entry is unused. It has a separate
hca_ia_list_entry that it links onto its parent HCA's ia list
Obviously this is confusing; when I looked at it yesterday I didn't
realize that an IA's ia_list_entry was unused.
My suggestion yesterday was to replace dapl_common's ia_list_entry
member with a list_head (and remove the list_head's in the individual
dapl object).
The above option is moot if we remove dapl_common all together.
Your wording is a bit confusing cause you say, "all the objects are put
on the IA list". Shouldn't that read, "each of the objects is put on
its respective list owned by the IA" or some such?
Your right, my wording was poor. You description is correct.
-tduffy
_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
