On Thu, Aug 22, 2019 at 04:55:14PM -0700, Nadav Amit wrote:
> > On Aug 22, 2019, at 4:50 PM, Sean Christopherson <sean.j.christopherson@xxxxxxxxx> wrote:
> >
> > Ensure the fw_cfg overrides are parsed prior consuming any of said
> > overrides. fwcfg_get_u() treats zero as a valid overide value, which
> > is slightly problematic since the overrides are in the .bss and thus
> > initialized to zero.
> >
> > Add a limit check when indexing fw_override so that future code doesn't
> > spontaneously explode.
> >
> > Cc: Nadav Amit <nadav.amit@xxxxxxxxx>
> > Fixes: 03b1e4570f967 ("x86: Support environments without test-devices")
> > Signed-off-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
> > ---
> > lib/x86/fwcfg.c | 10 ++++++++--
> > lib/x86/fwcfg.h | 2 --
> > x86/cstart64.S | 2 --
> > 3 files changed, 8 insertions(+), 6 deletions(-)
> >
> > diff --git a/lib/x86/fwcfg.c b/lib/x86/fwcfg.c
> > index d8d797f..06ef62c 100644
> > --- a/lib/x86/fwcfg.c
> > +++ b/lib/x86/fwcfg.c
> > @@ -5,10 +5,11 @@
> > static struct spinlock lock;
> >
> > static long fw_override[FW_CFG_MAX_ENTRY];
> > +static bool fw_override_done;
> >
> > bool no_test_device;
> >
> > -void read_cfg_override(void)
> > +static void read_cfg_override(void)
> > {
> > const char *str;
> > int i;
> > @@ -26,6 +27,8 @@ void read_cfg_override(void)
> >
> > if ((str = getenv("TEST_DEVICE")))
> > no_test_device = !atol(str);
> > +
> > + fw_override_done = true;
> > }
> >
> > static uint64_t fwcfg_get_u(uint16_t index, int bytes)
> > @@ -34,7 +37,10 @@ static uint64_t fwcfg_get_u(uint16_t index, int bytes)
> > uint8_t b;
> > int i;
> >
> > - if (fw_override[index] >= 0)
> > + if (!fw_override_done)
> > + read_cfg_override();
> > +
> > + if (index < FW_CFG_MAX_ENTRY && fw_override[index] >= 0)
> > return fw_override[index];
>
> How did that happen? I remember I tested this code with KVM..
It only breaks 32-bit KVM.
