From: Saurabh Sengar <ssengar@xxxxxxxxxxxxxxxxxxx>
On a x86 system under test with 1780 CPUs, topology_span_sane() takes
around 8 seconds cumulatively for all the iterations. It is an expensive
operation which does the sanity of non-NUMA topology masks.
CPU topology is not something which changes very frequently hence make
this check optional for the systems where the topology is trusted and
need faster bootup.
Restrict this to sched_verbose kernel cmdline option so that this penalty
can be avoided for the systems who want to avoid it.
Cc: stable@xxxxxxxxxxxxxxx
Fixes: ccf74128d66c ("sched/topology: Assert non-NUMA topology masks don't (partially) overlap")
Signed-off-by: Saurabh Sengar <ssengar@xxxxxxxxxxxxxxxxxxx>
Co-developed-by: Naman Jain <namjain@xxxxxxxxxxxxxxxxxxx>
Signed-off-by: Naman Jain <namjain@xxxxxxxxxxxxxxxxxxx>
---
Changes since v2:
https://lore.kernel.org/all/1731922777-7121-1-git-send-email-ssengar@xxxxxxxxxxxxxxxxxxx/
- Use sched_debug() instead of using sched_debug_verbose
variable directly (addressing Prateek's comment)
Changes since v1:
https://lore.kernel.org/all/1729619853-2597-1-git-send-email-ssengar@xxxxxxxxxxxxxxxxxxx/
- Use kernel cmdline param instead of compile time flag.
Adding a link to the other patch which is under review.
https://lore.kernel.org/lkml/20241031200431.182443-1-steve.wahl@xxxxxxx/
Above patch tries to optimize the topology sanity check, whereas this
patch makes it optional. We believe both patches can coexist, as even
with optimization, there will still be some performance overhead for
this check.
---
kernel/sched/topology.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c
index c49aea8c1025..b030c1a2121f 100644
--- a/kernel/sched/topology.c
+++ b/kernel/sched/topology.c
@@ -2359,6 +2359,13 @@ static bool topology_span_sane(struct sched_domain_topology_level *tl,
{
int i = cpu + 1;
+ /* Skip the topology sanity check for non-debug, as it is a time-consuming operatin */
+ if (!sched_debug()) {
+ pr_info_once("%s: Skipping topology span sanity check. Use `sched_verbose` boot parameter to enable it.\n",
+ __func__);
+ return true;
+ }
+
/* NUMA levels are allowed to overlap */
if (tl->flags & SDTL_OVERLAP)
return true;
base-commit: 00f3246adeeacbda0bd0b303604e46eb59c32e6e
--
2.43.0
