Re: [Xen-devel] [PATCH 2 of 3] blktap3/libxl: Check whether blktap3 is available

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Thanos Makatos writes ("RE: [Xen-devel] [PATCH 2 of 3] blktap3/libxl: Check 
whether blktap3 is  available"):
> Tapback watches XenStore, so it doesn't need the control socket to
> function.

So if all you're doing is seeing whether it's available, surely you
can have it write something in a fixed location in xenstore on startup ?

> However, the control socket will be necessary for doing
> more elaborate stuff (e.g. controlling a tapdisk in a thread-safe
> manner).

I'm not sure I understand this comment.  What is not thread-safe about
controlling it via xenstore ?

> I understand the problems of using a processes' name in
> this case, but since the control socket is not absolutely necessary
> at this point (it's actually not yet implemented in the tapback
> daemon), could we leave this as a future improvement?

On the whole I think it would be better not to introduce bugs on the
vague promise of fixing them later "as a future improvement" ...

Thanks,
Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel