Re: [RFC:PATCH 4/4] autofs4 - add new packet type for v5 communications

[Ian Kent <raven@themaw.net>]

On Fri, 17 Feb 2006, Andrew Morton wrote:

> Ian Kent <raven@themaw.net> wrote:
> >
> > On Fri, 17 Feb 2006, Andrew Morton wrote:
> > 
> > > Ian Kent <raven@themaw.net> wrote:
> > > >
> > > > +/* autofs v5 common packet struct */
> > > >  +struct autofs_v5_packet {
> > > >  +	struct autofs_packet_hdr hdr;
> > > >  +	autofs_wqt_t wait_queue_token;
> > > >  +	__u32 dev;
> > > >  +	__u64 ino;
> > > >  +	uid_t uid;
> > > >  +	gid_t gid;
> > > >  +	pid_t pid;
> > > >  +	pid_t tgid;
> > > >  +	int len;
> > > >  +	char name[NAME_MAX+1];
> > > >  +};
> > > 
> > > Is this known to work with 32-bit userspace on 64-bit kernels?
> > > 
> > > In particular, perhaps the ?id_t's should become a type of known size and
> > > alignment (u32 or u64)?
> > > 
> > 
> > Yes. I take your point.
> > 
> > I used this for some time on my Ultra 2, which has this type of arch, 
> > without problem. I increased the ino field from 32 to 64 bits since that 
> > time and haven't since tested it.
> > 
> > I'm happy to change them to 64 bit if you believe it will avoid potential 
> > problems?
> > 
> 
> This stuff always makes my head spin, but certainly using u64 throughout
> would be the safest approach.
> 

OK. I'll change it, test it tomorrow and post the resulting patch.

Ian