Re: [PATCH v2 0/7] vmsplice: fix some problems in my previous vmsplice patchset

"David Hildenbrand (Arm)" <[email protected]>:
> I think we concluded that we cannot rip out vmsplice that way at this point, and
> I suspect that Christian will drop that topic branch from -next after -rc1.

I think my patches still have a chance.

On fuse regression: I return EINVAL for particular combination of
flags used by fuse. This causes fuse to fail-back to non-vmsplice
code path. I did Debian code search, and I found none significant
packages, which use same combination of options.

So I think I was able to deal with fuse regression.

On CRIU named fifo "Not supported" regression: it is handled.

On CRIU major performance regression: it is NOT handled. But I still
think my approach is right. (See cover letter for details.)

(I wrote about all these in cover letter for this v2 patchset.)

So all regressions found so far (except for CRIU major performance
regression) are handled.

Other option is to introduce some deprecation period (as
suggested by Andrei Vagin). I can do this, if needed.

-- 
Askar Safin