Re: [RFC] what to do with IOCB_DSYNC?

[Jens Axboe <axboe@kernel.dk>]

On 5/26/22 5:19 PM, Al Viro wrote:
> On Sun, May 22, 2022 at 07:04:36PM +0000, Al Viro wrote:
> 
>> copy_page_{from,to}_iter() iovec side is needed not for the sake of
>> optimizations - if you look at the generic variant, you'll see that it's
>> basically "kmap_local_page + copy_{to,from}_iter() for the contents +
>> kunmap_local_page", which obviously relies upon the copy_{to,from}_iter()
>> being non-blocking.  So iovec part of it genuinely needs to differ from
>> the generic variant; it's just that on top of that it had been (badly)
>> microoptimized.  So were iterators, but that got at least somewhat cleaned
>> up a while ago.  And no, turning that into indirect calls ended up with
>> arseloads of overhead, more's the pity...
> 
> Actually, I take that back - the depth of kmap stack won't be increased
> by more than one compared to mainline, so we can switch iovec and ubuf
> variants to generic.
> 
> See #new.iov_iter; I do like the diffstat - -110 lines of code, with new
> flavour added...
> 
> Completely untested, though.

That's nice looking indeed. Not just dropping the code, but reducing the
unwieldly number of branches in there. A bit pressed for time these
days, but I'll try and resurrect some testing mid next week.

-- 
Jens Axboe