Re: [PATCH v1 iproute2-next 1/4] rdma: add helper rd_sendrecv_msg()

[Steve Wise <swise@opengridcomputing.com>]

On 2/26/2019 1:16 PM, Leon Romanovsky wrote:
> On Tue, Feb 26, 2019 at 11:19:12AM -0600, Steve Wise wrote:
>> On 2/23/2019 3:31 AM, Leon Romanovsky wrote:
>>> On Sat, Feb 23, 2019 at 11:26:15AM +0200, Leon Romanovsky wrote:
>>>> On Thu, Feb 21, 2019 at 08:19:03AM -0800, Steve Wise wrote:
>>>>> This function sends the constructed netlink message and then
>>>>> receives the response, displaying any error text.
>>>>>
>>>>> Change 'rdma dev set' to use it.
>>>>>
>>>>> Signed-off-by: Steve Wise <swise@opengridcomputing.com>
>>>>> ---
>>>>>  rdma/dev.c   |  2 +-
>>>>>  rdma/rdma.h  |  1 +
>>>>>  rdma/utils.c | 21 +++++++++++++++++++++
>>>>>  3 files changed, 23 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/rdma/dev.c b/rdma/dev.c
>>>>> index 60ff4b31e320..d2949c378f08 100644
>>>>> --- a/rdma/dev.c
>>>>> +++ b/rdma/dev.c
>>>>> @@ -273,7 +273,7 @@ static int dev_set_name(struct rd *rd)
>>>>>  	mnl_attr_put_u32(rd->nlh, RDMA_NLDEV_ATTR_DEV_INDEX, rd->dev_idx);
>>>>>  	mnl_attr_put_strz(rd->nlh, RDMA_NLDEV_ATTR_DEV_NAME, rd_argv(rd));
>>>>>
>>>>> -	return rd_send_msg(rd);
>>>>> +	return rd_sendrecv_msg(rd, seq);
>>>>>  }
>>>>>
>>>>>  static int dev_one_set(struct rd *rd)
>>>>> diff --git a/rdma/rdma.h b/rdma/rdma.h
>>>>> index 547bb5749a39..20be2f12c4f8 100644
>>>>> --- a/rdma/rdma.h
>>>>> +++ b/rdma/rdma.h
>>>>> @@ -115,6 +115,7 @@ bool rd_check_is_key_exist(struct rd *rd, const char *key);
>>>>>   */
>>>>>  int rd_send_msg(struct rd *rd);
>>>>>  int rd_recv_msg(struct rd *rd, mnl_cb_t callback, void *data, uint32_t seq);
>>>>> +int rd_sendrecv_msg(struct rd *rd, unsigned int seq);
>>>>>  void rd_prepare_msg(struct rd *rd, uint32_t cmd, uint32_t *seq, uint16_t flags);
>>>>>  int rd_dev_init_cb(const struct nlmsghdr *nlh, void *data);
>>>>>  int rd_attr_cb(const struct nlattr *attr, void *data);
>>>>> diff --git a/rdma/utils.c b/rdma/utils.c
>>>>> index 069d44fece10..a6f2826c9605 100644
>>>>> --- a/rdma/utils.c
>>>>> +++ b/rdma/utils.c
>>>>> @@ -664,6 +664,27 @@ int rd_recv_msg(struct rd *rd, mnl_cb_t callback, void *data, unsigned int seq)
>>>>>  	return ret;
>>>>>  }
>>>>>
>>>>> +static int null_cb(const struct nlmsghdr *nlh, void *data)
>>>>> +{
>>>>> +	return MNL_CB_OK;
>>>>> +}
>>>>> +
>>>>> +int rd_sendrecv_msg(struct rd *rd, unsigned int seq)
>>>>> +{
>>>>> +	int ret;
>>>>> +
>>>>> +	ret = rd_send_msg(rd);
>>>>> +	if (ret) {
>>>>> +		perror(NULL);
>>>> This is more or less already done in rd_send_msg() and that function
>>>> prints something in case of execution error. So the missing piece
>>>> is to update rd_recv_msg(), so all places will "magically" print errors
>>>> and not only dev_set_name().
>>>>
>>>>> +		goto out;
>>>>> +	}
>>>>> +	ret = rd_recv_msg(rd, null_cb, rd, seq);
>>> Will this "null_cb" work for all send/recv flows or only in flows where
>>> response can be error only?
>>
>> Only those flows where no nl attributes are expected to be returned.
>>
>>
>>> Will we need this recv_msg if we implement
>>> extack support?
>>
>> I'm not sure how extack works.  Do you know?
> I can't say that :)


We can change things if/when we support extack.

Stevo.