Re: [Xen-devel] [PATCH] libxc: zero-initialize structures in macros

[Wei Liu <wei.liu2@xxxxxxxxxx>]

On Fri, Sep 02, 2016 at 11:22:04AM -0600, Tamas K Lengyel wrote:
> On Fri, Sep 2, 2016 at 11:18 AM, Andrew Cooper
> <andrew.cooper3@xxxxxxxxxx> wrote:
> > On 02/09/16 17:39, Tamas K Lengyel wrote:
> >> While debugging applications built on top of libxc with Valgrind we get a 
> >> lot
> >> of complaining about relying on uninitialized values allocated in libxc.
> >> While these warnings are safe to ignore, zero-initializing the structures
> >> reduces Valgrind clutter a lot and aids in spotting real bugs.
> >>
> >> Signed-off-by: Tamas K Lengyel <tamas.lengyel@xxxxxxxxxxxx>
> >> ---
> >> Cc: Ian Jackson <ian.jackson@xxxxxxxxxxxxx>
> >> Cc: Wei Liu <wei.liu2@xxxxxxxxxx>
> >> ---
> >>  tools/libxc/xc_private.h | 10 +++++-----
> >>  1 file changed, 5 insertions(+), 5 deletions(-)
> >>
> >> diff --git a/tools/libxc/xc_private.h b/tools/libxc/xc_private.h
> >> index 75b761c..4e9073b 100644
> >> --- a/tools/libxc/xc_private.h
> >> +++ b/tools/libxc/xc_private.h
> >> @@ -59,11 +59,11 @@ struct iovec {
> >>  #include <sys/uio.h>
> >>  #endif
> >>
> >> -#define DECLARE_DOMCTL struct xen_domctl domctl
> >> -#define DECLARE_SYSCTL struct xen_sysctl sysctl
> >> -#define DECLARE_PHYSDEV_OP struct physdev_op physdev_op
> >> -#define DECLARE_FLASK_OP struct xen_flask_op op
> >> -#define DECLARE_PLATFORM_OP struct xen_platform_op platform_op
> >> +#define DECLARE_DOMCTL struct xen_domctl domctl = {0}
> >> +#define DECLARE_SYSCTL struct xen_sysctl sysctl = {0}
> >> +#define DECLARE_PHYSDEV_OP struct physdev_op physdev_op = {0}
> >> +#define DECLARE_FLASK_OP struct xen_flask_op op = {0}
> >> +#define DECLARE_PLATFORM_OP struct xen_platform_op platform_op = {0}
> >
> > I specifically took those out in the past, because it hides real
> > problems from Valgrind.
> >
> > Instead, I would recommend removing these wrappers entirely.  They serve
> > no useful purpose.
> >
> > Taking a random example of xc_get_pfn_type_batch(), it would be rather
> > more efficient to write
> >
> > ...
> >     DECLARE_HYPERCALL_BOUNCE(arr, sizeof(*arr) * num,
> > XC_HYPERCALL_BUFFER_BOUNCE_BOTH);
> >     struct xen_domctl domctl = {
> >         .cmd = XEN_DOMCTL_getpageframeinfo3,
> >         .domain = dom,
> >         .u.getpageframeinfo3.num = num,
> >     };
> > ...
> >
> > as it permits the compiler more freedom in how xen_domctl gets
> > constructed, as well as being able to plainly see exactly what is done
> > to the memory.
> >
> 
> Yea I don't really see much point using these macros as they are
> either and the one you propose certainly would make more sense.
> 

One reason I can think of why we would want those macros is that we
don't want to change all locations when the code fragment changes. But I
don't see how code segment is going to change for the macros under
discussion.

All in all, I don't object to eliminating those macros. Ian?

Wei.

> Tamas

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