[Openib-windows] IBAL fixes for memfree driver

[Leonid Keller]

Hi Fab,
I've checked the error codes with FW guys and it turned that you were
right:
RDD/EEC codes are not in use now.
So left are ony:
	IB_WCS_BAD_RESP_ERR,
 	IB_WCS_LOCAL_ACCESS_ERR,
 	IB_WCS_GENERAL_ERR

Regarding IB_WCS_LOCAL_ACCESS_ERR:
According to IB spec, it must be two different syndroms, so the FW
returns them.
I don't think, that we need to "simplify" the spec for user, mapping
them to one error code:
If the he doesn't matter, he won't check it.

> -----Original Message-----
> From: ftillier.sst at gmail.com [mailto:ftillier.sst at gmail.com] 
> On Behalf Of Fabian Tillier
> Sent: Wednesday, March 15, 2006 7:46 PM
> To: Leonid Keller
> Cc: openib-windows at openib.org
> Subject: Re: [Openib-windows] IBAL fixes for memfree driver
> 
> Hi Leonid,
> 
> A few more questions about the new WCS codes.
> 
> On 3/7/06, Leonid Keller <leonid at mellanox.co.il> wrote:
> >
> > Index: ib_types.h
> > ===================================================================
> > @@ -8764,6 +8750,17 @@
> >   IB_WCS_RNR_RETRY_ERR,
> >   IB_WCS_TIMEOUT_RETRY_ERR,
> >   IB_WCS_REM_INVALID_REQ_ERR,
> > + IB_WCS_LOCAL_EEC_OP_ERR,
> 
> Can we map this to IB_WCS_LOCAL_OP_ERR?  Why do we need a 
> dedicated error code for EECs?  Is LOCAL_EEC_OP_ERROR only 
> possible for work requests posted the send queue of an RD QP 
> (i.e. not the receive queue)?  If so, can the user get both 
> LOCAL_OP_ERR for the send queue of an RD QP and 
> LOCAL_EEC_OP_ERR for an EEC used on that QP?
> 
> > + IB_WCS_BAD_RESP_ERR,
> > + IB_WCS_LOCAL_ACCESS_ERR,
> 
> How is this different from IB_WCS_LOCAL_PROTECTION_ERR?  Why 
> is the RDMA Write with Immediate data a special case?  Why 
> does the user need to distinguish this from the error code?
> 
> > + IB_WCS_REM_INV_REQ_ERR,
> 
> How is this different from IB_WCS_REM_INVALID_REQ_ERR?
> 
> > + IB_WCS_LOCAL_RDD_VIOL_ERR,
> 
> Can we map this to GENERAL_ERR or FATAL_ERR, at least until 
> we add RDD support?  None of the EEC and RD errors should 
> occur until then anyway, so having a generic error code for 
> now would work better.
> 
> > + IB_WCS_REM_ABORT_ERR,
> > + IB_WCS_INV_EECN_ERR,
> 
> Again, map to GENERAL or FATAL?
> 
> > + IB_WCS_INV_EEC_STATE_ERR,
> 
> Same.
> 
> > + IB_WCS_FATAL_ERR,
> > + IB_WCS_RESP_TIMEOUT_ERR,
> 
> How is this different from IB_WCS_TIMEOUT_RETRY_ERR?  The 
> documentation below doesn't give any information.
> 
> Thanks,
> 
> - Fab
> 
> 
> 
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: ib_types_new.txt
URL: <http://lists.openfabrics.org/pipermail/ofw/attachments/20060322/02e866ec/attachment.txt>