Re: fallocate / posix_fallocate for new WAL file creation (etc...)

On 2013-05-30 06:55:16 -0400, Greg Smith wrote:
> On 5/30/13 6:49 AM, Robert Haas wrote:
> >On Wed, May 29, 2013 at 10:42 AM, Andres Freund <andres(at)2ndquadrant(dot)com> wrote:
> >>So we don't even know whether we can read. I think that means we need to
> >>zero the file anyway...
> >
> >Surely this is undue pessimism.
>
> There have been many occasions where I've found the Linux kernel defining
> support for POSIX behavior with a NOP stub that basically says "we should
> make this work one day". I don't know whether the fallocate code is one of
> those or a fully implemented call. Based on that history, until I see a
> reader that validates the resulting files are good I have to assume they're
> not.

That argument in contrast I find not very convincing though. What was
the last incidence of such a system call that did not just error out
with ENOTSUPP or such?

The linux fallocate call is fully specified for this behaviour and got
added 2.6.23, there wasn't a stub before, so I am far less worried about
it than about the underspecifiedness of posix_fallocate(). Also, if some
system call doesn't follow its documented specifications it's not fully
our problem anymore. If we rely on undocumented behaviour though...

Greetings,

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services