Re: [Bug-tar] Re: [patch] O_NOATIME support

[Joerg Schilling]

Ian Turner <address@hidden> wrote:

> > Should't a clean Linux implementation offer a related [f]pathconf() call
> > for this feature?
>
> Maybe. But the Linux implementors were just trying to grab the libc 
> interface, 
> which doesn't say anything about fpathconf().

Well, the interface itself does not seem to be well designed.

Something like ioctl(fd, _FIOSATIME, &atv) (as on Solaris) seems to be more
closer to the UNIX & POSIX philosophy....

BTW: star uses ioctl(fd, _FIOSATIME, &atv) since 10 years.

> Probably the interface to find out if O_NOATIME works will be simple: If you 
> do open(), and O_NOATIME is not OK for this filesystem, then open() will just 
> return EINVAL. I'm OK with adding an fpathconf() facility too, but the big 
> work is going to be just teaching the kernel which filesystems work and which 
> don't.

How do you like to be sure if you are on a Linux version that just ignores
O_NOATIME and thus never returns EINVAL? This does not look to be a good idea
unless the O_NOATIME flag would have been implemented this way from the 
very beginning.

Jörg

-- 
 EMail:address@hidden (home) Jörg Schilling D-13353 Berlin
       address@hidden           (uni)  
       address@hidden   (work) Blog: http://schily.blogspot.com/
 URL:  http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily