Re: [PATCH] spapr_numa.c: fixes in spapr_numa_FORM2_write_rtas_tables()

[BALATON Zoltan]

On Wed, 22 Sep 2021, Daniel Henrique Barboza wrote:
> On 9/22/21 06:51, BALATON Zoltan wrote:
> > On Wed, 22 Sep 2021, Greg Kurz wrote:
> > > On Tue, 21 Sep 2021 16:43:47 -0300
> > > Daniel Henrique Barboza <danielhb413@gmail.com> wrote:
> > >
> > > > This patch has a handful of modifications for the recent added
> > > > FORM2 support:
> > > >
> > > > - there is no particular reason for both 'lookup_index_table' and
> > > > 'distance_table' to be allocated in the heap, since their sizes are
> > > > known right at the start of the function. Use static allocation in
> > > > them to spare a couple of g_new0() calls;
> > > >
> > > > - to not allocate more than the necessary size in 'distance_table'. At
> > > > this moment the array is oversized due to allocating uint32_t for all
> > > > elements, when most of them fits in an uint8_t;
> > > >
> > > > - create a NUMA_LOCAL_DISTANCE macro to avoid hardcoding the local
> > > > distance value.
> > >
> > > Not needed. A notion of minimal distance, which is obviously
> > > synonymous to local, already exists in the "sysemu/numa.h"
> > > header :
> > >
> > > #define NUMA_DISTANCE_MIN         10
> > >
> > > > Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
> > > > ---
> > > >  hw/ppc/spapr_numa.c | 35 +++++++++++++++++++----------------
> > > >  1 file changed, 19 insertions(+), 16 deletions(-)
> > > >
> > > > diff --git a/hw/ppc/spapr_numa.c b/hw/ppc/spapr_numa.c
> > > > index 58d5dc7084..039a0439c6 100644
> > > > --- a/hw/ppc/spapr_numa.c
> > > > +++ b/hw/ppc/spapr_numa.c
> > > > @@ -19,6 +19,9 @@
> > > >  /* Moved from hw/ppc/spapr_pci_nvlink2.c */
> > > >  #define SPAPR_GPU_NUMA_ID           (cpu_to_be32(1))
> > > >
> > > > +/* Macro to avoid hardcoding the local distance value */
> > > > +#define NUMA_LOCAL_DISTANCE         10
> > > > +
> > > >  /*
> > > >   * Retrieves max_dist_ref_points of the current NUMA affinity.
> > > >   */
> > > > @@ -500,17 +503,21 @@ static void 
> > > > spapr_numa_FORM2_write_rtas_tables(SpaprMachineState *spapr,
> > > >      MachineState *ms = MACHINE(spapr);
> > > >      NodeInfo *numa_info = ms->numa_state->nodes;
> > > >      int nb_numa_nodes = ms->numa_state->num_nodes;
> > > > +    /* Lookup index table has an extra uint32_t with its length */
> > > > +    uint32_t lookup_index_table[nb_numa_nodes + 1];
> > > >      int distance_table_entries = nb_numa_nodes * nb_numa_nodes;
> > > > -    g_autofree uint32_t *lookup_index_table = NULL;
> > > > -    g_autofree uint32_t *distance_table = NULL;
> > > > -    int src, dst, i, distance_table_size;
> > > > -    uint8_t *node_distances;
> > > > +    /*
> > > > +     * Distance table is an uint8_t array with a leading uint32_t
> > > > +     * containing its length.
> > > > +     */
> > > > +    uint8_t distance_table[distance_table_entries + 4];
> > > > +    uint32_t *distance_table_length;
> > > > +    int src, dst, i;
> > > >
> > > >      /*
> > > >       * ibm,numa-lookup-index-table: array with length and a
> > > >       * list of NUMA ids present in the guest.
> > > >       */
> > > > -    lookup_index_table = g_new0(uint32_t, nb_numa_nodes + 1);
> > > >      lookup_index_table[0] = cpu_to_be32(nb_numa_nodes);
> > > >
> > > >      for (i = 0; i < nb_numa_nodes; i++) {
> > > > @@ -518,8 +525,7 @@ static void 
> > > > spapr_numa_FORM2_write_rtas_tables(SpaprMachineState *spapr,
> > > >      }
> > > >
> > > >      _FDT(fdt_setprop(fdt, rtas, "ibm,numa-lookup-index-table",
> > > > -                     lookup_index_table,
> > > > -                     (nb_numa_nodes + 1) * sizeof(uint32_t)));
> > > > +                     lookup_index_table, sizeof(lookup_index_table)));
> > > >
> > > >      /*
> > > >       * ibm,numa-distance-table: contains all node distances. First
> > > > @@ -531,11 +537,10 @@ static void 
> > > > spapr_numa_FORM2_write_rtas_tables(SpaprMachineState *spapr,
> > > >       * array because NUMA ids can be sparse (node 0 is the first,
> > > >       * node 8 is the second ...).
> > > >       */
> > > > -    distance_table = g_new0(uint32_t, distance_table_entries + 1);
> > > > -    distance_table[0] = cpu_to_be32(distance_table_entries);
> > > > +    distance_table_length = (uint32_t *)distance_table;
> > > > +    distance_table_length[0] = cpu_to_be32(distance_table_entries);
> > > >
> > > > -    node_distances = (uint8_t *)&distance_table[1];
> > > > -    i = 0;
> > > > +    i = 4;
> > >
> > > A comment reminding why we're doing that wouldn't hurt, e.g.
> > >
> > > /* Skip the array size (uint32_t) */
> >
> > Then maybe instead of (or in addition to) a comment you could write 
> > sizeof(uint32_t) or sizeof(distance_rable[0]) instead of constant 4 to make 
> > this more explicit.
>
> distance_table is an uint8_t array. sizeof(distance_table[0]) would return 1.

Yes, sorry I was looking at uint32_t *distance_table_length; instead of 
the array definition.

> Doing i = sizeof(uint32_t) demands the reader to realize "this works because 
> it is
> skipping an uint32_t in an uint8_t array and sizeof(uint8_t) is 1".
>
> I think it's clearer to just be explicit in the comment:
>
>
> /* First 4 uint8_t contains the uint32_t array length */

That explains it better (although a bit confusing, if you need the length 
why don't you have a struct with the length and the array instead of 
sroring it in the array when that's a different type, unless this is some 
data structure defined that way, I don't know what this is all about). But 
I don't really care, adding this comment is fine.

Regards,
BALATON Zoltan