[Openib-windows] races on __destrot_obj function

[Fabian Tillier]

Hi Yossi,

On 7/11/06, Yossi Leybovich <sleybo at mellanox.co.il> wrote:
>
>
> > -----Original Message-----
> > From: ftillier.sst at gmail.com [mailto:ftillier.sst at gmail.com]
> > On Behalf Of Fabian Tillier
> > Sent: Tuesday, July 11, 2006 7:29 PM
> > To: Yossi Leybovich
> > Cc: openib-windows at openib.org
> > Subject: Re: [Openib-windows] races on __destrot_obj function
> >
> > Hi Yossi,
> >
> > On 7/11/06, Yossi Leybovich <sleybo at mellanox.co.il> wrote:
> > > Attached is patch that solve this problem
> > >
> > > It includes:
> > >
> > > 1. fix to cl_obj to return value if the parent port is not in
> > > CL_INITIALIZED 2. fix to insert/insert_locked function to
> > check return
> > > value 3. add ref count mechanism to help debug ref count
> > problems 4.
> > > not call port_resume in endpt cleanup but in port destroy
> > >
> > > I tested this patch with our test and it pass
> > >
> > > Pls review
> >
> > Comments about the patch below.  It would be good IMO to have
> > this split up, similar to how large patches are done in Linux
> > - so each of the above enumerations (1-4) would be a sparate
> > patch.  It just makes it a bit easier to review/apply, so
> > that it's not all-or-nothing.
> >
>
> I can separate the complib patch from the IPoIB patch .(part 1)
> But it would take a lot of time to remove all the debug utils I added
> from the code to create one patch then reinsert it and remove the former
> changes and create another patch.
>
> You right there few typos in the patch
> 1. IPOIB_DBG_OBJ sould be 1<<13
> 2. there is one place where I had extra braces
> 3. ipoib_port_ref/deref should use atomic functions
>
> Are you going to fix and apply or you want me to do this

I'll fix it and send it to you for review/testing, and then you can
apply it if you think it looks good.

- Fab