I would be REALLY careful with the realtime device, up until recently it was
guaranteed to write to the wrong device, this is one reason the realtime option
is hidden. A recent change (in the cvs tree) means it possibly write to the
correct device, but there is no real guarantee of that yet. How much testing
have you done?
Steve
p.s. I do see Irix realtime calls in lmdd, but these would need changing for
linux - have you done this?
> I would like to begin testing xfs with lids (www.lids.org) and rtlinux (ww
> w.rtlinux.com) to see if the inclusion of lids will play nicely with xfs, and
> if both methods of creating acls will be interchangeable (i.e, can lidsadm m
> odify chacl created acl's) and rtlinux, to see if the realtime section's perf
> ormance could be made truly 'realtime'. I'd like other's thoughts on this, a
> s I am not a programmer, and I can barely read code, so
> as to tell whether or not these projects would or should work together.
> The realtime section could also be marked at least 'experimental' in the cvs
> tree, since it does build quite well, and having to add 'CONFIG_XFS_RT' to .c
> onfig by hand would be a lot easier if it would just be made a regular
> (albeit experimental) option. It does work. lmdd is the only program that I
> have been able to use to open inodes with a realtime flag though (thanks stev
> e lord).
>
> Any input on this would be greatly appreciated.
>
> Thanks,
> -alan
>
> P.S: I know that GRIO is not a real possibility for 2.4, but is it being thou
> ght of as a 2.5 kernel subject at least?
|