Re: [PATCH v2 1/1] util/async-teardown: wire up query-command-line-options

[Claudio Imbrenda <imbrenda@linux.ibm.com>]

On Mon, 20 Mar 2023 17:05:07 +0100
Markus Armbruster <armbru@redhat.com> wrote:

> Thomas Huth <thuth@redhat.com> writes:
> 
> > On 20/03/2023 16.31, Markus Armbruster wrote:  
> >> Claudio Imbrenda <imbrenda@linux.ibm.com> writes:
> >>   
> >>> The recently introduced -async-teardown commandline option was not
> >>> wired up properly and did not show up in the output of the QMP command
> >>> query-command-line-options. This means that libvirt will have no way to
> >>> discover whether the feature is supported.  
> >> 
> >> There was nothing improper in its wiring.  The issue is that
> >> query-command-line-options is junk.  See my recent post
> >> 
> >>      Subject: query-command-line-options (was: [PATCH 1/7] qemu: capabilities: Introduce QEMU_CAPS_MACHINE_ACPI)
> >>      Date: Tue, 07 Mar 2023 10:40:23 +0100
> >>      Message-ID: <87jzzsc320.fsf_-_@pond.sub.org>
> >>   
> >>> This patch fixes the issue by correctly wiring up the commandline
> >>> option so that it appears in the output of query-command-line-options.
> >>>
> >>> Reported-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
> >>> Fixes: c891c24b1a ("os-posix: asynchronous teardown for shutdown on Linux")
> >>> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> >>> ---
> >>>   util/async-teardown.c | 17 +++++++++++++++++
> >>>   1 file changed, 17 insertions(+)
> >>>
> >>> diff --git a/util/async-teardown.c b/util/async-teardown.c
> >>> index 62cdeb0f20..c9b9a3cdb2 100644
> >>> --- a/util/async-teardown.c
> >>> +++ b/util/async-teardown.c
> >>> @@ -12,6 +12,9 @@
> >>>    */
> >>>   
> >>>   #include "qemu/osdep.h"
> >>> +#include "qemu/config-file.h"
> >>> +#include "qemu/option.h"
> >>> +#include "qemu/module.h"
> >>>   #include <dirent.h>
> >>>   #include <sys/prctl.h>
> >>>   #include <sched.h>
> >>> @@ -144,3 +147,17 @@ void init_async_teardown(void)
> >>>       clone(async_teardown_fn, new_stack_for_clone(), CLONE_VM, NULL);
> >>>       sigprocmask(SIG_SETMASK, &old_signals, NULL);
> >>>   }
> >>> +
> >>> +static QemuOptsList qemu_async_teardown_opts = {
> >>> +    .name = "async-teardown",
> >>> +    .head = QTAILQ_HEAD_INITIALIZER(qemu_async_teardown_opts.head),
> >>> +    .desc = {
> >>> +        { /* end of list */ }
> >>> +    },
> >>> +};
> >>> +
> >>> +static void register_async_teardown(void)
> >>> +{
> >>> +    qemu_add_opts(&qemu_async_teardown_opts);
> >>> +}
> >>> +opts_init(register_async_teardown);  
> >> 
> >> Now it *is* improperly wired up :)
> >> 
> >> You're defining new QemuOpts config group "async-teardown" with
> >> arbitrary option parameters, but don't actually use it for parsing or
> >> recording the option.  I figure because you can't: there is no option
> >> argument to parse and record, which is what QemuOpts is designed to do.
> >> 
> >> If you need the feature to be visible in query-command-line-options, you
> >> should make it an option parameter (a KEY, not a GROUP), preferably of
> >> an existing group / option.  
> >
> > Would it make sense to add it e.g. to "-action" instead, i.e. something like 
> > "-action teardown=async" ?  
> 
> I believe the new parameter "teardown" would be visible in
> query-command-line-options.
> 
> How well does it fit -action?

I guess it can be shoehorned in. generally action is about stuff that
happens in/to the guest, while in this case it's about how qemu will
perform the teardown of its address space once it terminates.

the important parts are: this is an OS-specific option (Linux), and it
needs to be parsed and enabled before sandboxing (otherwise clone(2)
might not work)