Re: [PATCH 1/1] IB/isert: add module param to set sg_tablesize for IO cmd

[Max Gurtovoy <mgurtovoy@nvidia.com>]

On 10/14/2020 1:21 AM, Sagi Grimberg wrote:
>
>
> On 10/11/20 2:06 AM, Max Gurtovoy wrote:
>> From: Max Gurtovoy <maxg@mellanox.com>
>>
>> Currently, iser target support max IO size of 16MiB by default. For some
>> adapters, allocating this amount of resources might reduce the total
>> number of possible connections that can be created. For those adapters,
>> it's preferred to reduce the max IO size to be able to create more
>> connections. Since there is no handshake procedure for max IO size in
>> iser protocol, set the default max IO size to 1MiB and add a module
>> parameter for enabling the option to control it for suitable adapters.
>>
>> Reported-by: Krishnamraju Eraparaju <krishna2@chelsio.com>
>> Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
>
> You can add a "Fixes:" tag to help it get to stable.

But the previous commit is not buggy. It reveals some resource lacking 
in some HCAs.

I can add it if needed.

>
>> ---
>>   drivers/infiniband/ulp/isert/ib_isert.c | 27 ++++++++++++++++++++++++-
>>   drivers/infiniband/ulp/isert/ib_isert.h |  6 ++++++
>>   2 files changed, 32 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/infiniband/ulp/isert/ib_isert.c 
>> b/drivers/infiniband/ulp/isert/ib_isert.c
>> index 695f701dc43d..5a47f1bbca96 100644
>> --- a/drivers/infiniband/ulp/isert/ib_isert.c
>> +++ b/drivers/infiniband/ulp/isert/ib_isert.c
>> @@ -28,6 +28,18 @@ static int isert_debug_level;
>>   module_param_named(debug_level, isert_debug_level, int, 0644);
>>   MODULE_PARM_DESC(debug_level, "Enable debug tracing if > 0 
>> (default:0)");
>>   +static int isert_sg_tablesize_set(const char *val,
>> +                  const struct kernel_param *kp);
>> +static const struct kernel_param_ops sg_tablesize_ops = {
>> +    .set = isert_sg_tablesize_set,
>> +    .get = param_get_int,
>> +};
>> +
>> +static int isert_sg_tablesize = ISCSI_ISER_SG_TABLESIZE;
>> +module_param_cb(sg_tablesize, &sg_tablesize_ops, 
>> &isert_sg_tablesize, 0644);
>> +MODULE_PARM_DESC(sg_tablesize,
>> +         "Number of gather/scatter entries in a single scsi command, 
>> should >= 128 (default: 256, max: 4096)");
>> +
>>   static DEFINE_MUTEX(device_list_mutex);
>>   static LIST_HEAD(device_list);
>>   static struct workqueue_struct *isert_comp_wq;
>> @@ -47,6 +59,19 @@ static void isert_send_done(struct ib_cq *cq, 
>> struct ib_wc *wc);
>>   static void isert_login_recv_done(struct ib_cq *cq, struct ib_wc *wc);
>>   static void isert_login_send_done(struct ib_cq *cq, struct ib_wc *wc);
>>   +static int isert_sg_tablesize_set(const char *val, const struct 
>> kernel_param *kp)
>> +{
>> +    int n = 0, ret;
>> +
>> +    ret = kstrtoint(val, 10, &n);
>> +    if (ret != 0 || n < ISCSI_ISER_MIN_SG_TABLESIZE ||
>> +        n > ISCSI_ISER_MAX_SG_TABLESIZE)
>> +        return -EINVAL;
>> +
>> +    return param_set_int(val, kp);
>> +}
>> +
>> +
>>   static inline bool
>>   isert_prot_cmd(struct isert_conn *conn, struct se_cmd *cmd)
>>   {
>> @@ -101,7 +126,7 @@ isert_create_qp(struct isert_conn *isert_conn,
>>       attr.cap.max_send_wr = ISERT_QP_MAX_REQ_DTOS + 1;
>>       attr.cap.max_recv_wr = ISERT_QP_MAX_RECV_DTOS + 1;
>>       factor = rdma_rw_mr_factor(device->ib_device, cma_id->port_num,
>> -                   ISCSI_ISER_MAX_SG_TABLESIZE);
>> +                   isert_sg_tablesize);
>>       attr.cap.max_rdma_ctxs = ISCSI_DEF_XMIT_CMDS_MAX * factor;
>>       attr.cap.max_send_sge = device->ib_device->attrs.max_send_sge;
>>       attr.cap.max_recv_sge = 1;
>> diff --git a/drivers/infiniband/ulp/isert/ib_isert.h 
>> b/drivers/infiniband/ulp/isert/ib_isert.h
>> index 7fee4a65e181..90ef215bf755 100644
>> --- a/drivers/infiniband/ulp/isert/ib_isert.h
>> +++ b/drivers/infiniband/ulp/isert/ib_isert.h
>> @@ -65,6 +65,12 @@
>>    */
>>   #define ISER_RX_SIZE        (ISCSI_DEF_MAX_RECV_SEG_LEN + 1024)
>>   +/* Default I/O size is 1MB */
>> +#define ISCSI_ISER_SG_TABLESIZE 256
>
> Maybe name it ISCSI_ISER_DEF_SG_TABLESIZE

ok


>
> Otherwise,
>
> Reviewed-by: Sagi Grimberg <sagi@grimberg.me>