On 9 Sep 2025, at 12:54, Paolo Abeni pabeni@redhat.com wrote:
On 9/3/25 6:31 PM, Håkon Bugge wrote:
We need to increment i_fastreg_wrs before we bail out from rds_ib_post_reg_frmr().
Elaborating a bit more on the `why` could help the review.
Yes, it's lean. Let me expand on it.
Fixes: 1659185fb4d0 ("RDS: IB: Support Fastreg MR (FRMR) memory registration mode") Fixes: 3a2886cca703 ("net/rds: Keep track of and wait for FRWR segments in use upon shutdown") Signed-off-by: Håkon Bugge haakon.bugge@oracle.com
[...] @@ -178,9 +181,11 @@ static int rds_ib_post_reg_frmr(struct rds_ib_mr *ibmr)
- being accessed while registration is still pending.
*/ wait_event(frmr->fr_reg_done, !frmr->fr_reg);
out:
- return ret;
+out_inc:
- atomic_inc(&ibmr->ic->i_fastreg_wrs);
The existing error path on ib_post_send() is left untouched. I think it would be cleaner and less error prone to let it use the above label, too.
I agree, overlooked that. I'll send a v3 tomorrow.
Thanks for the review, Håkon
/P