On Tuesday 16 October 2007 05:20, Roland Dreier wrote:
> int mthca_arbel_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags)
> {
> struct mthca_cq *cq = to_mcq(ibcq);
> - __be32 doorbell[2];
> - u32 sn;
> - __be32 ci;
> -
> - sn = cq->arm_sn & 3;
> - ci = cpu_to_be32(cq->cons_index);
> + __be32 db_rec[2];
> + u32 dbhi;
> + u32 sn = cq->arm_sn & 3;
>
> - doorbell[0] = ci;
> - doorbell[1] = cpu_to_be32((cq->cqn << 8) | (2 << 5) | (sn << 3) |
> - ((flags & IB_CQ_SOLICITED_MASK) ==
> - IB_CQ_SOLICITED ? 1 : 2));
> + db_rec[0] = cpu_to_be32(cq->cons_index);
> + db_rec[1] = cpu_to_be32((cq->cqn << 8) | (2 << 5) | (sn << 3) |
> + ((flags & IB_CQ_SOLICITED_MASK) ==
> + IB_CQ_SOLICITED ? 1 : 2));
>
> - mthca_write_db_rec(doorbell, cq->arm_db);
> + mthca_write_db_rec(db_rec, cq->arm_db);
>
Patch looks good, but don't you have the same 64-bit alignment problem in
mthca_write_db_rec() ?
- Jack
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
