Re: [PATCH liburing 1/1] man: add io_uring_register_region.3

[Pavel Begunkov <asml.silence@gmail.com>]

On 1/15/26 14:14, Jens Axboe wrote:
> On 1/15/26 6:06 AM, Pavel Begunkov wrote:
>> On 1/14/26 23:53, Jens Axboe wrote:
>> ...
>>>>>> And the compiler is smart enough to optimise it out since
>>>>>> it's all on stack.
>>>>>
>>>>> Not sure I follow these emails. For the normal case,
>>>>> io_validate_ext_arg() copies in the args via a normal user copy, which
>>>>> depending on options and the arch (or even sub-arch, amd more expensive)
>>>>> is more or less expensive.
>>>>
>>>> In the end, after prep that is still just a move instruction, e.g.
>>>> for x86. And it loads into a register and stores it into ext_arg,
>>>> just like with registration. User copy needs to prepare page fault
>>>> handling / etc., which could be costly (e.g. I see stac + lfence
>>>> in asm), but that's not exactly about avoiding copies.
>>>
>>> Those are implementation details. The user copy is stac/clac, and then
>>> the loads. This is what makes it more expensive. I don't want to be
>>> writing about stac/clac in the man page, that's irrelevant to the user.
>>
>> Confused why would you be thinking about putting that into the
>> man page. I'm saying that it claims copy avoidance, but there is
>> no difference in the number of copies. It's also uncomfortable
>> that it's in a commit with my name attached, while the change
>> wouldn't fall under the "language edits" note.
> 
> Sheesh let's turn down the sensitivity. If you want it changed, send a

Not sure what kind of sensitivity you mean, but no worries, there
wasn't any. I still believe, however, that technicalities like in
this case matter.

> patch. I'm trying to phrase it in such a way that it makes sense to
> people without getting into too much detail. It avoids copying from
> USERSPACE, which is the expensive part.
> 
> I think we've spent enough time on this detail at this point, no?

-- 
Pavel Begunkov