From: Greg Kurz <groug@kaod.org>
To: "Philippe Mathieu-Daudé" <philmd@redhat.com>
Cc: Daniel Henrique Barboza <danielhb413@gmail.com>,
qemu-ppc@nongnu.org, qemu-devel@nongnu.org,
david@gibson.dropbear.id.au
Subject: Re: [PATCH] spapr_numa.c: fixes in spapr_numa_FORM2_write_rtas_tables()
Date: Wed, 22 Sep 2021 13:52:50 +0200 [thread overview]
Message-ID: <20210922135250.31a2d4d4@bahia.huguette> (raw)
In-Reply-To: <78cc89be-4e4b-c23a-e40a-b41865037008@redhat.com>
On Wed, 22 Sep 2021 13:17:32 +0200
Philippe Mathieu-Daudé <philmd@redhat.com> wrote:
> On 9/21/21 21:43, Daniel Henrique Barboza 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.
> >
> > 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;
>
> This should have be of ptrdiff_t type.
>
Why ? I don't see pointer subtraction in the code.
> > + /*
> > + * Distance table is an uint8_t array with a leading uint32_t
> > + * containing its length.
> > + */
> > + uint8_t distance_table[distance_table_entries + 4];
>
> The previous code seems better by using the heap, now we have
> to worry about stack overflow...
>
Indeed the size of this array could be up to 16k + 4. I guess
Philippe's point make sense. David's request was to use uint8_t
instead of uin32t_t, not to drop g_new0(). Please revert to
using the heap.
lookup_index_table[] can _only_ grow up to 516 bytes, which is
less problematic, but it doesn't hurt either to allocate it
on the heap. Not changing that would make this patch simpler.
> > + uint32_t *distance_table_length;
>
> Please drop, ...
>
> > + 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);
>
> ... and use instead:
>
> stl_be_p(distance_table, distance_table_entries);
>
+1
> > - node_distances = (uint8_t *)&distance_table[1];
> > - i = 0;
> > + i = 4;
> >
> > for (src = 0; src < nb_numa_nodes; src++) {
> > for (dst = 0; dst < nb_numa_nodes; dst++) {
> > @@ -546,18 +551,16 @@ static void spapr_numa_FORM2_write_rtas_tables(SpaprMachineState *spapr,
> > * adding the numa_info to retrieve distance info from.
> > */
> > if (src == dst) {
> > - node_distances[i++] = 10;
> > + distance_table[i++] = NUMA_LOCAL_DISTANCE;
> > continue;
> > }
> >
> > - node_distances[i++] = numa_info[src].distance[dst];
> > + distance_table[i++] = numa_info[src].distance[dst];
> > }
> > }
> >
> > - distance_table_size = distance_table_entries * sizeof(uint8_t) +
> > - sizeof(uint32_t);
> > _FDT(fdt_setprop(fdt, rtas, "ibm,numa-distance-table",
> > - distance_table, distance_table_size));
> > + distance_table, sizeof(distance_table)));
> > }
> >
> > /*
> >
>
next prev parent reply other threads:[~2021-09-22 11:53 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-09-21 19:43 [PATCH] spapr_numa.c: fixes in spapr_numa_FORM2_write_rtas_tables() Daniel Henrique Barboza
2021-09-22 3:08 ` David Gibson
2021-09-22 8:26 ` Greg Kurz
2021-09-22 9:51 ` BALATON Zoltan
2021-09-22 11:00 ` Daniel Henrique Barboza
2021-09-22 11:10 ` BALATON Zoltan
2021-09-22 11:17 ` Philippe Mathieu-Daudé
2021-09-22 11:50 ` Daniel Henrique Barboza
2021-09-22 11:52 ` Greg Kurz [this message]
2021-09-22 12:35 ` Philippe Mathieu-Daudé
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210922135250.31a2d4d4@bahia.huguette \
--to=groug@kaod.org \
--cc=danielhb413@gmail.com \
--cc=david@gibson.dropbear.id.au \
--cc=philmd@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-ppc@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).