[PATCH v2 2/4] lib: sbi: Add the bound check for events during config match

[Atish Patra]

On Wed, Jul 20, 2022 at 3:30 AM Andrew Jones <ajones at ventanamicro.com> wrote:
>
> On Tue, Jul 19, 2022 at 03:46:13PM -0700, Atish Patra wrote:
> > Currently, there is no sanity check for firmware event code. We don't see
> > any issue as Linux kernel driver does a bound check on firmware events
> > already. However, OpenSBI can't assume sane supervisor mode software
> > always. Thus, an invalid event idx can cause a buffer overflow error.
> > For hardware events, the match will fail for invalid event code anyways.
> > However, a search is unecessary if event code is invalid.
> >
> > Add a event ID validation function to solve the issue.
> >
> > Signed-off-by: Atish Patra <atishp at rivosinc.com>
> > ---
> >  lib/sbi/sbi_pmu.c | 39 +++++++++++++++++++++++++++++++++++++--
> >  1 file changed, 37 insertions(+), 2 deletions(-)
> >
> > diff --git a/lib/sbi/sbi_pmu.c b/lib/sbi/sbi_pmu.c
> > index 31631a2fab80..1170cba43b3d 100644
> > --- a/lib/sbi/sbi_pmu.c
> > +++ b/lib/sbi/sbi_pmu.c
> > @@ -107,6 +107,37 @@ static bool pmu_event_select_overlap(struct sbi_pmu_hw_event *evt,
> >       return FALSE;
> >  }
> >
> > +static int pmu_event_validate(unsigned long event_idx)
> > +{
> > +     uint32_t event_idx_type = get_cidx_type(event_idx);
> > +     uint32_t event_idx_code = get_cidx_code(event_idx);
> > +     uint32_t event_idx_code_max = -1;
> > +
> > +     switch(event_idx_type) {
> > +     case SBI_PMU_EVENT_TYPE_HW:
> > +             event_idx_code_max = SBI_PMU_HW_GENERAL_MAX;
> > +             break;
> > +     case SBI_PMU_EVENT_TYPE_FW:
> > +             event_idx_code_max = SBI_PMU_FW_MAX;
> > +             break;
> > +     case SBI_PMU_EVENT_TYPE_HW_CACHE:
> > +             event_idx_code_max = ((SBI_PMU_HW_CACHE_MAX & 0xFFF) << 0x03) |
> > +                                    (SBI_PMU_HW_CACHE_OP_MAX & 0x3) << 0x01 |
> > +                                     0x01;
>
> We need to check these fields independently, otherwise event_idx_code
> could be something like 6, which is less than event_idx_code_max (63),
> but represents a code with an invalid op_id.
>

Yes. I did not add those additional checks initially as this is just
an optimization for hw & hw cache events.
For an invalid event id, config_match will never find a suitable
counter anyways and return failure.

However, I get your point. We should just check independently for
completeness and avoid confusion.
I will fix it in v3.

> > +             break;
> > +     case SBI_PMU_EVENT_TYPE_HW_RAW:
> > +             event_idx_code_max = 1; // event_idx.code should be zero
> > +             break;
> > +     default:
> > +             return SBI_EINVAL;
> > +     }
> > +
> > +     if (event_idx_code < event_idx_code_max)
> > +             return event_idx_type;
> > +     else
>
> nit: no need for the 'else'
>

Thanks. Will fix it.

> > +             return SBI_EINVAL;
> > +}
> > +
> >  static int pmu_ctr_validate(uint32_t cidx, uint32_t *event_idx_code)
> >  {
> >       uint32_t event_idx_val;
> > @@ -614,12 +645,16 @@ int sbi_pmu_ctr_cfg_match(unsigned long cidx_base, unsigned long cidx_mask,
> >  {
> >       int ctr_idx = SBI_ENOTSUPP;
> >       u32 hartid = current_hartid();
> > -     int event_type = get_cidx_type(event_idx);
> > +     int event_type;
> >       struct sbi_pmu_fw_event *fevent;
> >       uint32_t fw_evt_code;
> >
> >       /* Do a basic sanity check of counter base & mask */
> > -     if ((cidx_base + sbi_fls(cidx_mask)) >= total_ctrs || event_type >= SBI_PMU_EVENT_TYPE_MAX)
> > +     if ((cidx_base + sbi_fls(cidx_mask)) >= total_ctrs)
> > +             return SBI_EINVAL;
> > +
> > +     event_type = pmu_event_validate(event_idx);
> > +     if (event_type < 0)
> >               return SBI_EINVAL;
> >
> >       if (flags & SBI_PMU_CFG_FLAG_SKIP_MATCH) {
> > --
> > 2.25.1
> >
>
> Thanks,
> drew
>
> --
> opensbi mailing list
> opensbi at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/opensbi



-- 
Regards,
Atish