| Message ID | 20220718192844.1805158-16-yury.norov@gmail.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | Introduce DEBUG_BITMAP config option and bitmap_check_params() | expand |
On Mon, Jul 18, 2022 at 12:28:43PM -0700, Yury Norov wrote: > kernel/sched/topology.c | 10 ++++++++-- > 1 file changed, 8 insertions(+), 2 deletions(-) > > diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c > index 05b6c2ad90b9..ad32d0a43424 100644 > --- a/kernel/sched/topology.c > +++ b/kernel/sched/topology.c > @@ -2211,6 +2211,8 @@ static struct sched_domain *build_sched_domain(struct sched_domain_topology_leve > static bool topology_span_sane(struct sched_domain_topology_level *tl, > const struct cpumask *cpu_map, int cpu) > { > + const struct cpumask *mc = tl->mask(cpu); > + const struct cpumask *mi; > int i; > > /* NUMA levels are allowed to overlap */ > @@ -2226,14 +2228,18 @@ static bool topology_span_sane(struct sched_domain_topology_level *tl, > for_each_cpu(i, cpu_map) { > if (i == cpu) > continue; > + > + mi = tl->mask(i); > + if (mi == mc) > + continue; > + > /* > * We should 'and' all those masks with 'cpu_map' to exactly > * match the topology we're about to build, but that can only > * remove CPUs, which only lessens our ability to detect > * overlaps > */ > - if (!cpumask_equal(tl->mask(cpu), tl->mask(i)) && > - cpumask_intersects(tl->mask(cpu), tl->mask(i))) > + if (!cpumask_equal(mc, mi) && cpumask_intersects(mc, mi)) > return false; > } This is once again a super slow path; but I don't suppose you're making the code worse in this case.
diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c index 05b6c2ad90b9..ad32d0a43424 100644 --- a/kernel/sched/topology.c +++ b/kernel/sched/topology.c @@ -2211,6 +2211,8 @@ static struct sched_domain *build_sched_domain(struct sched_domain_topology_leve static bool topology_span_sane(struct sched_domain_topology_level *tl, const struct cpumask *cpu_map, int cpu) { + const struct cpumask *mc = tl->mask(cpu); + const struct cpumask *mi; int i; /* NUMA levels are allowed to overlap */ @@ -2226,14 +2228,18 @@ static bool topology_span_sane(struct sched_domain_topology_level *tl, for_each_cpu(i, cpu_map) { if (i == cpu) continue; + + mi = tl->mask(i); + if (mi == mc) + continue; + /* * We should 'and' all those masks with 'cpu_map' to exactly * match the topology we're about to build, but that can only * remove CPUs, which only lessens our ability to detect * overlaps */ - if (!cpumask_equal(tl->mask(cpu), tl->mask(i)) && - cpumask_intersects(tl->mask(cpu), tl->mask(i))) + if (!cpumask_equal(mc, mi) && cpumask_intersects(mc, mi)) return false; }
topology_span_sane() checks if cpu == i before calling cpumask_equal(tl->mask(cpu), tl->mask(i)). However, tl->mask(cpu) and tl->mask(i) may point to the same cpumask even if i != cpu. Fix the check accordingly. While here, move tl->mask(cpu) out of the loop, and make the in-loop code calculating tl->mask(i) only once. Catched with CONFIG_DEBUG_BITMAP: [ 0.867917] Call Trace: [ 0.868209] <TASK> [ 0.868471] build_sched_domains+0x36f/0x1a40 [ 0.868576] sched_init_smp+0x44/0xba [ 0.869012] ? mtrr_aps_init+0x84/0xa0 [ 0.869465] kernel_init_freeable+0x12e/0x26e [ 0.869982] ? rest_init+0xd0/0xd0 [ 0.870406] kernel_init+0x16/0x120 [ 0.870821] ret_from_fork+0x22/0x30 [ 0.871244] </TASK> [ 0.871502] ---[ end trace 0000000000000000 ]--- [ 0.872040] b1: ffffffffb1fd3480 [ 0.872041] b2: ffffffffb1fd3480 [ 0.872041] b3: 0 [ 0.872042] nbits: 256 [ 0.872042] start: 0 [ 0.872042] off: 0 [ 0.872043] Bitmap: parameters check failed [ 0.872043] include/linux/bitmap.h [427]: bitmap_equal Signed-off-by: Yury Norov <yury.norov@gmail.com> --- kernel/sched/topology.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)