[PATCH V2 8/8] nvme-core: fix nvme_submit_sync_cmd() args

[Chaitanya Kulkarni]

On 3/27/23 17:43, Christoph Hellwig wrote:
> On Sun, Mar 26, 2023 at 11:04:18PM -0700, Chaitanya Kulkarni wrote:
>> 4 files changed, 112 insertions(+), 38 deletions(-)
> This:
>
>   - adds a shit ton of code

agree, but it is similar what we do for nvme_command structure
all over in the host side core and fabrics code nothing new :-
nvme_identify_ns_descs()
nvme_identify_ns()
nvme_ns_info_from_id_cs_indep()
nvme_init_ms()
nvme_get_log()
nvme_init_non_mdts_limits()
nvme_scan_ns_list()
nvme_identify_ctrl()
nvmf_reg_read32()
nvmf_reg_read64()
nvmf_reg_write32()
nvmf_connect_admin_queue()
nvmf_connect_io_queue()

>   - makes it harder to use the function

we pass nvme_command struct similar way with initializing
its members, this patch is doing the same without inventing
anything new similar to block layer code (mentioned in commit
log) and host side core/fabrics code for the nvme_command
mentioned above.

>   - creates tons of churn
>
> Why?

in general there is even opposition to have 6 args when I sent out
a patch-series for __nvme_submit_sync_cmd() it has 8!

currently it has long list of arguments and I think people have
complained about it already on the mailing list, if someone wants
to add more argument(s) what is the right way of doing that
since this patch is not acceptable ? please explain ...

if you still think keeping long list of argument is a better way
to go then feel free to drop this no biggy ..

-ck