On Mon, 11 May 2020 11:07:06 +0200
Cédric Le Goater <clg@kaod.org> wrote:
> On 5/7/20 7:26 PM, Greg Kurz wrote:
> > It is the job of the ppc_radix64_get_fully_qualified_addr() function
> > which is called at the beginning of ppc_radix64_xlate() to set both
> > lpid *and* pid. It doesn't buy us anything to initialize them first.
> >
> > Worse, a bug in ppc_radix64_get_fully_qualified_addr(), eg. failing to
> > set either lpid or pid, would be undetectable by static analysis tools
> > like coverity.
> >
> > Signed-off-by: Greg Kurz <groug@kaod.org>
> > ---
> > target/ppc/mmu-radix64.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/target/ppc/mmu-radix64.c b/target/ppc/mmu-radix64.c
> > index c76879f65b78..5e2d912ee346 100644
> > --- a/target/ppc/mmu-radix64.c
> > +++ b/target/ppc/mmu-radix64.c
> > @@ -433,7 +433,7 @@ static int ppc_radix64_xlate(PowerPCCPU *cpu, vaddr eaddr, int rwx,
> > bool cause_excp)
> > {
> > CPUPPCState *env = &cpu->env;
> > - uint64_t lpid = 0, pid = 0;
> > + uint64_t lpid, pid;
> > ppc_v3_pate_t pate;
> > int psize, prot;
> > hwaddr g_raddr;
> >
>
> I am seeing this failure with gcc version 9.3.1 20200408 (Red Hat 9.3.1-2) (GCC)
>
> target/ppc/mmu-radix64.c: In function ‘ppc_radix64_xlate’:
> target/ppc/mmu-radix64.c:314:12: error: ‘pid’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
> 314 | offset = pid * sizeof(struct prtb_entry);
> | ~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> target/ppc/mmu-radix64.c:439:20: note: ‘pid’ was declared here
> 439 | uint64_t lpid, pid;
> | ^~~
> target/ppc/mmu-radix64.c:458:14: error: ‘lpid’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
> 458 | if (!ppc64_v3_get_pate(cpu, lpid, &pate)) {
> | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> CC ppc64-softmmu/target/ppc/fpu_helper.o
>
>
> This seems like a compiler optimization issue.
>
Ah... it seems that gcc is trying to be smart but it doesn't realize
that ppc_radix64_get_fully_qualified_addr() doesn't have any path
that leaves lpid or pid unset... :-\ Adding a default: case in both
switch statements is enough to silent gcc.
I guess it may be easier for David if I post a v2 of the entire series
that addresses all the comments.
Thanks!
> C.