* Xin Li <xin@xxxxxxxxx> wrote:
> On 5/15/2025 8:29 AM, Ingo Molnar wrote:
> >
> > * Xin Li (Intel) <xin@xxxxxxxxx> wrote:
> >
> > > xen_read_msr_safe() currently passes an uninitialized argument err to
> > > xen_do_read_msr(). But as xen_do_read_msr() may not set the argument,
> > > xen_read_msr_safe() could return err with an unpredictable value.
> > >
> > > To ensure correctness, initialize err to 0 (representing success)
> > > in xen_read_msr_safe().
> > >
> > > Because xen_read_msr_safe() is essentially a wrapper of xen_do_read_msr(),
> > > the latter should be responsible for initializing the value of *err to 0.
> > > Thus initialize *err to 0 in xen_do_read_msr().
> > >
> > > Fixes: 502ad6e5a619 ("x86/msr: Change the function type of native_read_msr_safe()")
> > > Reported-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> > > Closes: https://lore.kernel.org/xen-devel/aBxNI_Q0-MhtBSZG@stanley.mountain/
> > > Signed-off-by: Xin Li (Intel) <xin@xxxxxxxxx>
> > > ---
> > > arch/x86/xen/enlighten_pv.c | 5 ++++-
> > > 1 file changed, 4 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/arch/x86/xen/enlighten_pv.c b/arch/x86/xen/enlighten_pv.c
> > > index 3be38350f044..01f1d441347e 100644
> > > --- a/arch/x86/xen/enlighten_pv.c
> > > +++ b/arch/x86/xen/enlighten_pv.c
> > > @@ -1091,6 +1091,9 @@ static u64 xen_do_read_msr(u32 msr, int *err)
> > > {
> > > u64 val = 0; /* Avoid uninitialized value for safe variant. */
> > > + if (err)
> > > + *err = 0;
> > > +
> > > if (pmu_msr_chk_emulated(msr, &val, true))
> > > return val;
> > > @@ -1162,7 +1165,7 @@ static void xen_do_write_msr(u32 msr, u64 val, int *err)
> > > static int xen_read_msr_safe(u32 msr, u64 *val)
> > > {
> > > - int err;
> > > + int err = 0;
> > > *val = xen_do_read_msr(msr, &err);
> > > return err;
> >
> > So why not initialize 'err' with 0 in both callers, xen_read_msr_safe()
> > and xen_read_msr(), and avoid all the initialization trouble in
> > xen_do_read_msr()?
>
> Yeah, I should make the change in xen_read_msr() too.
>
> However xen_do_read_msr() should be implemented in a defensive way to
> set *err properly as it's part of its return value. Actually it was so,
> but one of my previous cleanup patch removed it because err is no longer
> passed to pmu_msr_chk_emulated().
Maybe. It's up to Juergen though.
Thanks,
Ingo
