Re: [Xen-devel] [PATCH 2/2] libxl: treat "dying" domains as destroyed

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Ian Campbell writes ("Re: [Xen-devel] [PATCH 2/2] libxl: treat "dying" domains 
as destroyed"):
> This seems to do what I would expect, I tried with "xl reboot", "xl
> shutdown" & "xl destroy" as well as in-guest shutdown (which is no
> different in reality from "xl shutdown"). No leaks and in each case it
> rebooted, destroyed etc the domain as expected.

Yes, great, thanks.

> > +static void domain_death_occurred(libxl__egc *egc,
> > +                                  libxl_evgen_domain_death **evg_upd,
> > +                                  const char *why) {
> > +    /* Removes **evg from the list and advances *evg_upd to the next
> > +     * entry.  Call sites in _xswatch_callback must use "continue". */
> 
> There's no **evg in this context? ITYM *evg or just evg?

I meant **evg_upd.  The actual struct is removed from the list, after
all.

> Also it's not clear which list this talks about since there is
> death_list and death_reported as well as, presumably, a list of current
> domain infos somewhere. Did you mean "moves evg from death_list to
> death_reported list and advances *evg_upd ...."?

Yes.

> > +    EGC_GC;
> > +    libxl_evgen_domain_death *const evg = *evg_upd;
> > +
> > +    LIBXL__LOG(CTX, LIBXL__LOG_DEBUG, " death %s", why);
> 
> LIBXL__LOG puts a space between the boilerplate and the message so you
> end up with two spaces before "death" here.
> 
> Also the resultant messages are:
>       " death destroyed"
>       " death missing"
>       " death dying"
> in context that is:
>         libxl: debug: libxl.c:696:domain_death_occurred:  death dying
> 
> Which could do with a noun or something. 

I'll remove the word "death" (and the extra space).

Thanks,
Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel