Re: ACL patch

2020-05-15 Thread Christos Zoulas
> On May 15, 2020, at 5:53 PM, Andrew Doran wrote: > > > One problem with the VV_HASACLS thing is that VOP_ACCESS() is called with an > LK_SHARED lock and you can't modify v_vflag with that. > > It looks like you have to explicitly enable ACLs before the file system is > mounted with tunefs,

Re: ACL patch

2020-05-15 Thread Andrew Doran
weekend to take a look but > >> it > >> was gone. The best place for a flag would be cache_enter_id() since it > >> deals with the necessary synchronisation. I'll see about adding a flag > >> now. > >> In addition to the places it's called now it

Re: ACL patch

2020-05-14 Thread Christos Zoulas
now. >> In addition to the places it's called now it would need to be called >> whenever an inode gains an ACL. > >Ok done should just need to mark the cached ID as invalid when the inode has >/ gains an ACL. Ok, I made some changes using this in my latest acl patch. Can you please check? https://www.netbsd.org/~christos/acl.diff Best, christos

Re: ACL patch

2020-05-13 Thread Christos Zoulas
In article <20200512235040.gd9...@homeworld.netbsd.org>, Andrew Doran wrote: >On Tue, May 12, 2020 at 10:37:27PM +, Andrew Doran wrote: >> On Sun, May 10, 2020 at 09:48:10AM -0400, Christos Zoulas wrote: >> > >> > > Maybe I missed it but I didn't see a way for cache_lookup_linked() and >> >

Re: ACL patch

2020-05-12 Thread Andrew Doran
On Tue, May 12, 2020 at 10:37:27PM +, Andrew Doran wrote: > On Sun, May 10, 2020 at 09:48:10AM -0400, Christos Zoulas wrote: > > > > > Maybe I missed it but I didn't see a way for cache_lookup_linked() and > > > cache_have_id() to know a vnode has ACLs. The presence of ACLs means > > >

Re: ACL patch

2020-05-12 Thread Andrew Doran
On Sun, May 10, 2020 at 09:48:10AM -0400, Christos Zoulas wrote: > > > Maybe I missed it but I didn't see a way for cache_lookup_linked() and > > cache_have_id() to know a vnode has ACLs. The presence of ACLs means those > > routines can't do their imitation of VOP_ACCESS() and need to fail so

Re: ACL patch

2020-05-10 Thread Christos Zoulas
> Maybe I missed it but I didn't see a way for cache_lookup_linked() and > cache_have_id() to know a vnode has ACLs. The presence of ACLs means those > routines can't do their imitation of VOP_ACCESS() and need to fail so that > the lookup is handled by VOP_LOOKUP(). To handle that on a

Re: ACL patch

2020-05-05 Thread Christos Zoulas
> On May 4, 2020, at 7:51 PM, Andrew Doran wrote: >> Thanks for reviewing. > > +++ sbin/tunefs/tunefs.8 > ... > +.It Fl n Cm enable | disable > +Turn on/off the administrative NFSv4 ACL enable flag. Fixed, this was before when: -a = posix alls -n =

Re: ACL patch

2020-05-04 Thread Andrew Doran
On Sat, May 02, 2020 at 07:10:29PM -0400, Christos Zoulas wrote: > Hi, > > The following patch ports the FreeBSD FFS ACLS (both posix1e and nfs4) to > NetBSD. Comments? I will let this sit for a week or so and then commit it > if I don't hear screams. > > [it is ~24K lines, so not posted

ACL patch

2020-05-02 Thread Christos Zoulas
Hi, The following patch ports the FreeBSD FFS ACLS (both posix1e and nfs4) to NetBSD. Comments? I will let this sit for a week or so and then commit it if I don't hear screams. [it is ~24K lines, so not posted inline] https://www.netbsd.org/~christos/acl.diff Best, christos