On Tue, Oct 21, 2014 at 12:06 AM, Or Gerlitz <gerlitz.or@xxxxxxxxx> wrote: > On Mon, Oct 20, 2014 at 9:11 PM, Minh Duc Tran <MinhDuc.Tran@xxxxxxxxxx> wrote: >>>> +++ b/drivers/infiniband/ulp/iser/iser_verbs.c >>>> @@ -114,6 +114,9 @@ static int iser_create_device_ib_res(struct iser_device *device) >>>> if (IS_ERR(device->pd)) >>>> goto pd_err; >>>> >>>> + max_cqe = (dev_attr->max_cqe < ISER_MAX_CQ_LEN) ? >>>> + dev_attr->max_cqe : ISER_MAX_CQ_LEN; >>>> + >> >>>If I was the ocrdma maintainer I would say load and clear: NO, please. >>>Your current offering supports 32 CQs per device, and this means that on 32 core node you will be able to run only iSER, no other ULP. >>>Generally, this approach is wrong, causing bad user experience and we will not do that. >> >> Hi Or, >> >> All 32 CQs will never get used. Here are the reasons: >> 1) current #define ISER_MAX_CQ 4 >> 2) Your proposed TODO "#1. change the number of CQs to be min(num_cpus, 1/2 of what the device can support)" >> This would limits the max CQs device supports to 1/2 anyway. >> >> If you don't agree, I like to get your suggestions of what I should do in this case for CQs that have 8k of entries each. So I was wrong, sorry... you were talking on CQEs not CQs. I think Sagi has some proposal on how to harden the code that deals with setting the size of CQs, let me talk to him. -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
