If domain_soft_reset_cb can't rename the save file, it doesn't call
initiate_domain_create() and calls domcreate_complete().

Skipping initiate_domain_create() means dcs->console_wait is
uninitialized and all 0s.

We have:
  domcreate_complete()
    libxl__xswait_stop()
      libxl__ev_xswatch_deregister().

The uninitialized slotnum 0 is considered valid (-1 is the invalid
sentinel), so the NULL pointer path to passed to xs_unwatch() which
segfaults.

libxl__ev_xswatch_deregister:watch w=0x12bc250 wpath=(null) token=0/0: deregister slotnum=0

Ensure dcs->console_xswait is minimally initialized by calling
libxl__xswait_init() in do_domain_soft_reset().

Also add a check for dcs->console_xswait.path being NULL in
domcreate_complete() to avoid the segfault.

Signed-off-by: Jason Andryuk <jandryuk@gmail.com>
---
The NULL check in domcreate_complete isn't needed when the xswait is
initialized, but it could catch other occurances.

 tools/libs/light/libxl_create.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/tools/libs/light/libxl_create.c b/tools/libs/light/libxl_create.c
index 15ed021f41..631caa416d 100644
--- a/tools/libs/light/libxl_create.c
+++ b/tools/libs/light/libxl_create.c
@@ -1970,7 +1970,8 @@ static void domcreate_complete(libxl__egc *egc,
     libxl_domain_config *const d_config = dcs->guest_config;
     libxl_domain_config *d_config_saved = &dcs->guest_config_saved;
 
-    libxl__xswait_stop(gc, &dcs->console_xswait);
+    if (dcs->console_xswait.path)
+        libxl__xswait_stop(gc, &dcs->console_xswait);
 
     libxl__domain_build_state_dispose(&dcs->build_state);
 
@@ -2176,6 +2177,10 @@ static int do_domain_soft_reset(libxl_ctx *ctx,
                               aop_console_how);
     cdcs->domid_out = &domid_out;
 
+    /* Initialize in case we end up in domcreate_complete without calling
+     * initiate_domain_create. */
+    libxl__xswait_init(&cdcs->dcs.console_xswait);
+
     dom_path = libxl__xs_get_dompath(gc, domid);
     if (!dom_path) {
         LOGD(ERROR, domid, "failed to read domain path");
-- 
2.35.1

On Wed, Mar 30, 2022 at 02:17:27PM -0400, Jason Andryuk wrote:
> diff --git a/tools/libs/light/libxl_create.c b/tools/libs/light/libxl_create.c
> index 15ed021f41..631caa416d 100644
> --- a/tools/libs/light/libxl_create.c
> +++ b/tools/libs/light/libxl_create.c
> @@ -1970,7 +1970,8 @@ static void domcreate_complete(libxl__egc *egc,
>      libxl_domain_config *const d_config = dcs->guest_config;
>      libxl_domain_config *d_config_saved = &dcs->guest_config_saved;
>  
> -    libxl__xswait_stop(gc, &dcs->console_xswait);
> +    if (dcs->console_xswait.path)
> +        libxl__xswait_stop(gc, &dcs->console_xswait);

libxl__xswait_stop() needs to be called here. It's a function that can
be called several time without ill effect, as long as `console_xswait`
is initialised properly (by calling libxl__xswait_init() like you did
below.

>      libxl__domain_build_state_dispose(&dcs->build_state);
>  
> @@ -2176,6 +2177,10 @@ static int do_domain_soft_reset(libxl_ctx *ctx,
>                                aop_console_how);
>      cdcs->domid_out = &domid_out;
>  
> +    /* Initialize in case we end up in domcreate_complete without calling
> +     * initiate_domain_create. */

That comment isn't needed. This is just part of the initialisation of
`cdcs->dcs`.

> +    libxl__xswait_init(&cdcs->dcs.console_xswait);
> +
>      dom_path = libxl__xs_get_dompath(gc, domid);
>      if (!dom_path) {
>          LOGD(ERROR, domid, "failed to read domain path");


As part of the patch, could you also move the libxl__xswait_init() call
in initiate_domain_create() to do_domain_create()? That would avoid
`console_xswait` been initialised twice, and do_domain_create() is the
function that initialise `dcs` before calling initiate_domain_create().

Thanks,

-- 
Anthony PERARD

On Thu, Mar 31, 2022 at 10:10 AM Anthony PERARD
<anthony.perard@citrix.com> wrote:
>
> On Wed, Mar 30, 2022 at 02:17:27PM -0400, Jason Andryuk wrote:
> > diff --git a/tools/libs/light/libxl_create.c b/tools/libs/light/libxl_create.c
> > index 15ed021f41..631caa416d 100644
> > --- a/tools/libs/light/libxl_create.c
> > +++ b/tools/libs/light/libxl_create.c
> > @@ -1970,7 +1970,8 @@ static void domcreate_complete(libxl__egc *egc,
> >      libxl_domain_config *const d_config = dcs->guest_config;
> >      libxl_domain_config *d_config_saved = &dcs->guest_config_saved;
> >
> > -    libxl__xswait_stop(gc, &dcs->console_xswait);
> > +    if (dcs->console_xswait.path)
> > +        libxl__xswait_stop(gc, &dcs->console_xswait);
>
> libxl__xswait_stop() needs to be called here. It's a function that can
> be called several time without ill effect, as long as `console_xswait`
> is initialised properly (by calling libxl__xswait_init() like you did
> below.
>
> >      libxl__domain_build_state_dispose(&dcs->build_state);
> >
> > @@ -2176,6 +2177,10 @@ static int do_domain_soft_reset(libxl_ctx *ctx,
> >                                aop_console_how);
> >      cdcs->domid_out = &domid_out;
> >
> > +    /* Initialize in case we end up in domcreate_complete without calling
> > +     * initiate_domain_create. */
>
> That comment isn't needed. This is just part of the initialisation of
> `cdcs->dcs`.
>
> > +    libxl__xswait_init(&cdcs->dcs.console_xswait);
> > +
> >      dom_path = libxl__xs_get_dompath(gc, domid);
> >      if (!dom_path) {
> >          LOGD(ERROR, domid, "failed to read domain path");
>
>
> As part of the patch, could you also move the libxl__xswait_init() call
> in initiate_domain_create() to do_domain_create()? That would avoid
> `console_xswait` been initialised twice, and do_domain_create() is the
> function that initialise `dcs` before calling initiate_domain_create().

Sure, that all sounds good.  Thanks!

Regards,
Jason