Re: fcntl(F_GETPATH) support

[Christos Zoulas <christos%zoulas.com@localhost>]

> On Sep 15, 2019, at 12:43 PM, Mateusz Guzik <mjguzik%gmail.com@localhost> wrote:
> 
> On 9/14/19, Christos Zoulas <christos%zoulas.com@localhost> wrote:
>> 
>> Comments?
>> 
>> +	error = vnode_to_path(kpath, MAXPATHLEN, fp->f_vnode, l, l->l_proc);
> 
> What motivates this change?
> 
> I think it is a little problematic in that namecache resolution is not
> guaranteed to work. If the consumer does not check for failure or fallback
> to something else the user is going to experience "once in the blue moon"
> failures.
> 
> For example clang likes to get full paths to names. On Linux it can
> trivially do it thanks to dentry cache. On other systems there is a
> compilation check for F_GETPATH. If none of this is present it goes for
> realpath, which is slover but basically works. If F_GETPATH is present,
> it is always used and there is no fallback if it fails. Then you risk
> setting yourself up for a weird compilation failure, which may not even
> repeat itself after you retry.
> 
> All in all I don't think this should be added unless namecache becomes
> reliable. The above reasoning is why I did not add it to FreeBSD.
> 

I think it is quite reliable because all the file descriptors would be recently
opened and therefore be in the cache.  One would need to DoS the cache
cause eviction. If that turns out to be false, we can make the namecache
reliable, or withdraw it.

christos