Re: [PATCH v11 13/13] block: apply COR-filter to block-stream jobs

[Andrey Shinkevich]

On 16.10.2020 18:45, Vladimir Sementsov-Ogievskiy wrote:
> 15.10.2020 20:16, Andrey Shinkevich wrote:
> > On 14.10.2020 19:24, Max Reitz wrote:
> > > On 12.10.20 19:43, Andrey Shinkevich wrote:
> >
> > [...]
> >
> > > > ---
> > > >   block/stream.c             | 93 
> > > > +++++++++++++++++++++++++++++-----------------
> > > >   tests/qemu-iotests/030     | 51 +++----------------------
> > > >   tests/qemu-iotests/030.out |  4 +-
> > > >   tests/qemu-iotests/141.out |  2 +-
> > > >   tests/qemu-iotests/245     | 19 +++++++---
> > > >   5 files changed, 81 insertions(+), 88 deletions(-)
> > >
> > > Looks like stream_run() could be a bit streamlined now (the allocation
> > > checking should be unnecessary, unconditionally calling
> > > stream_populate() should be sufficient), but not necessary now.
> >
> > That is what I had kept in my mind when I tackled this patch. But 
> > there is an underwater reef to streamline. Namely, how the 
> > block-stream job gets known about a long unallocated tail to exit the 
> > loop earlier in the stream_run(). Shall we return the '-EOF' or 
> > another error code from the cor_co_preadv_part() to be handled by the 
> > stream_run()? Any other suggestions, if any, will be appreciated.
>
> Just calling read CHUNK by CHUNK may be less efficient than 
> is_allocated()-driven loop: you may end up with splitting regions 
> unaligned to CHUNK-granularity, which would not be splitted with 
> is_allocated()-driven loop. Current loop allows chunks unaligned to CHUNK.

The cor_co_preadv_part() will check for the end of a file in the next 
version. So, the unalignment is not going to be the issue.

Andrey

> So, I think, it's better to keep is_allocated() logic as is for now.