Re: [Qemu-devel] [PATCH 3/4] 9pfs: stat_to_qid: use device as input to qid.path

[Greg Kurz <groug@kaod.org>]

On Fri, 16 Feb 2018 11:20:33 +0100
Antonios Motakis <antonios.motakis@huawei.com> wrote:

<snip>
> 
> I'm not that sure that we can make the assumption that all entries in a 
> dir will share the st_dev,

The assumption stands for any entry that is not a directory actually.
But indeed a directory could be a mount point, and have a different
st_dev.

> I think we have to check it for each entry...
> 

Only if the entry might be a directory, ie,

dent->d_type == DT_DIR || dent->d_type == DT_UNKNOWN

> >  
> >> Also, POSIX permits returning specific errno codes that aren't otherwise
> >> listed for a syscall if the usual semantics of that errno code are
> >> indeed the reason for the failure (you should prefer to fail with errno
> >> codes documented by POSIX where possible, but POSIX does not limit you
> >> to just that set).
> >>  
> > Ok, then ENFILE wouldn't be that bad in the end.
> >
> > Thanks for your POSIX expertise :)  
> Will keep that one then!
> 
> >
> > Cheers,
> >
> > --
> > Greg  
>