Re: [PATCH -next v3] drivers: base: cacheinfo: fix shared_cpu_map

[Yong Xuan Wang]

Hi Sudeep,

> On Wed, Jan 4, 2023 at 6:59 PM Sudeep Holla <sudeep.holla@xxxxxxx> wrote:
>
> On Wed, Dec 28, 2022 at 03:24:19AM +0000, Yong-Xuan Wang wrote:
> > The cacheinfo sets up the shared_cpu_map by checking whether the caches
> > with the same index are shared between CPUs. However, this will trigger
> > slab-out-of-bounds access if the CPUs do not have the same cache hierarchy.
> > Another problem is the mismatched shared_cpu_map when the shared cache does
> > not have the same index between CPUs.
> >
> > CPU0  I       D       L3
> > index 0       1       2       x
> >       ^       ^       ^       ^
> > index 0       1       2       3
> > CPU1  I       D       L2      L3
> >
> > This patch checks each cache is shared with all caches on other CPUs.
> >
>
> Just curious to know if this is just Qemu config or a real platform.
> I had intentionally not supported this to just to get to know when such
> h/w appears in the real world 😁.
>

We are trying to build such kind of config in QEMU.

> > Reviewed-by: Pierre Gondois <pierre.gondois@xxxxxxx>
> > Signed-off-by: Yong-Xuan Wang <yongxuan.wang@xxxxxxxxxx>
> > ---
> >  drivers/base/cacheinfo.c | 25 +++++++++++++++----------
> >  1 file changed, 15 insertions(+), 10 deletions(-)
> >
> > diff --git a/drivers/base/cacheinfo.c b/drivers/base/cacheinfo.c
> > index 950b22cdb5f7..dfa804bcf3cc 100644
> > --- a/drivers/base/cacheinfo.c
> > +++ b/drivers/base/cacheinfo.c
> > @@ -256,7 +256,7 @@ static int cache_shared_cpu_map_setup(unsigned int cpu)
> >  {
> >       struct cpu_cacheinfo *this_cpu_ci = get_cpu_cacheinfo(cpu);
> >       struct cacheinfo *this_leaf, *sib_leaf;
> > -     unsigned int index;
> > +     unsigned int index, sib_index;
> >       int ret = 0;
> >
> >       if (this_cpu_ci->cpu_map_populated)
> > @@ -284,11 +284,12 @@ static int cache_shared_cpu_map_setup(unsigned int cpu)
> >
> >                       if (i == cpu || !sib_cpu_ci->info_list)
> >                               continue;/* skip if itself or no cacheinfo */
> > -
> > -                     sib_leaf = per_cpu_cacheinfo_idx(i, index);
> > -                     if (cache_leaves_are_shared(this_leaf, sib_leaf)) {
> > -                             cpumask_set_cpu(cpu, &sib_leaf->shared_cpu_map);
> > -                             cpumask_set_cpu(i, &this_leaf->shared_cpu_map);
> > +                     for (sib_index = 0; sib_index < cache_leaves(i); sib_index++) {
> > +                             sib_leaf = per_cpu_cacheinfo_idx(i, sib_index);
> > +                             if (cache_leaves_are_shared(this_leaf, sib_leaf)) {
> > +                                     cpumask_set_cpu(cpu, &sib_leaf->shared_cpu_map);
> > +                                     cpumask_set_cpu(i, &this_leaf->shared_cpu_map);
>
> Does it make sense to break here once we match as it is unlikely to match
> with any other indices ?
>

Yeah. We can break here once we find the shared instance. I'll send a
new version to fix it.
Thank you!

> > +                             }
> >                       }
> >               }
> >               /* record the maximum cache line size */
> > @@ -302,7 +303,7 @@ static int cache_shared_cpu_map_setup(unsigned int cpu)
> >  static void cache_shared_cpu_map_remove(unsigned int cpu)
> >  {
> >       struct cacheinfo *this_leaf, *sib_leaf;
> > -     unsigned int sibling, index;
> > +     unsigned int sibling, index, sib_index;
> >
> >       for (index = 0; index < cache_leaves(cpu); index++) {
> >               this_leaf = per_cpu_cacheinfo_idx(cpu, index);
> > @@ -313,9 +314,13 @@ static void cache_shared_cpu_map_remove(unsigned int cpu)
> >                       if (sibling == cpu || !sib_cpu_ci->info_list)
> >                               continue;/* skip if itself or no cacheinfo */
> >
> > -                     sib_leaf = per_cpu_cacheinfo_idx(sibling, index);
> > -                     cpumask_clear_cpu(cpu, &sib_leaf->shared_cpu_map);
> > -                     cpumask_clear_cpu(sibling, &this_leaf->shared_cpu_map);
> > +                     for (sib_index = 0; sib_index < cache_leaves(sibling); sib_index++) {
> > +                             sib_leaf = per_cpu_cacheinfo_idx(sibling, sib_index);
> > +                             if (cache_leaves_are_shared(this_leaf, sib_leaf)) {
> > +                                     cpumask_clear_cpu(cpu, &sib_leaf->shared_cpu_map);
> > +                                     cpumask_clear_cpu(sibling, &this_leaf->shared_cpu_map);
>
> Same comment as above.
>
> --
> Regards,
> Sudeep

Regards,
Yong-Xuan