Re: [PATCH rdma-rc 6/7] IB/core: Fix incorrect array allocation

[Leon Romanovsky <leon-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>]

[-- Attachment #1: Type: text/plain, Size: 1580 bytes --]

On Tue, Jun 07, 2016 at 09:43:29AM -0400, Doug Ledford wrote:
> On 6/7/2016 3:16 AM, Leon Romanovsky wrote:
> > On Mon, Jun 06, 2016 at 07:53:43PM -0400, Doug Ledford wrote:
> >> On 6/4/2016 8:15 AM, Leon Romanovsky wrote:
> >>> From: Mark Bloch <markb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
> >>>
> >>> In a attribute group struct, attrs should point to a NULL
> >>> terminated list. Which means we need to allocate one more
> >>> slot in the array.
> >>
> >> This patch is both right and wrong.  You're right with the intent (that
> >> there should be a total of 2 extra entries in the array size, one for
> >> the NULL termination and one for the lifespan entry), but wrong with the
> >> mechanics (unless I've missed something).  We already have two extra
> >> entries because sizeof(*hsag) includes our first counter, so just
> >> num_counters is actually enough to NULL terminate the list, and + 1
> >> gives us lifespan plus a NULL terminate spot.  The comment could be
> >> cleaned up to make this more clear though, so I'll do that.
> > 
> > Thanks Doug.
> > 
> 
> I was wrong with this BTW (I tested to make sure).  Even though the
> struct includes the name for the array, because it's 0 length, it
> reserves 0 space.  I thought it would reserve one u64.  In any case, I
> ended up writing a fix of my own since I had already marked your patch
> as not needed in patchworks.  I listed Mark as the reporter in that patch.

Since I know that you are testing the patches, and your "I'll do that."
was enough for me.

Thanks.

> 




[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]