* [PATCH 1/3] sched: topology: add input parameter for sched_domain_flags_f()
2022-03-29 9:15 [PATCH 0/3] support for describing cache topology from DT Qing Wang
@ 2022-03-29 9:15 ` Qing Wang
2022-03-29 9:23 ` Peter Zijlstra
2022-03-29 9:15 ` [PATCH 2/3] arch_topology: support for describing cache topology from DT Qing Wang
2022-03-29 9:15 ` [PATCH 3/3] arm64: add arm64 default topology Qing Wang
2 siblings, 1 reply; 5+ messages in thread
From: Qing Wang @ 2022-03-29 9:15 UTC (permalink / raw)
To: Catalin Marinas, Will Deacon, Michael Ellerman,
Benjamin Herrenschmidt, Paul Mackerras, Thomas Gleixner,
Ingo Molnar, Borislav Petkov, x86, H. Peter Anvin, Sudeep Holla,
Greg Kroah-Hartman, Rafael J. Wysocki, Peter Zijlstra, Juri Lelli,
Vincent Guittot, Dietmar Eggemann, Steven Rostedt, Ben Segall,
Mel Gorman, Daniel Bristot de Oliveira, linux-arm-kernel,
linux-kernel, linuxppc-dev
Cc: Wang Qing
From: Wang Qing <wangqing@vivo.com>
sched_domain_flags_f() are statically set now, but actually, we can get a lot
of necessary information based on the cpu_map. e.g. we can know whether its
cache is shared.
Allows custom extension without affecting current.
Signed-off-by: Wang Qing <wangqing@vivo.com>
---
arch/powerpc/kernel/smp.c | 4 ++--
arch/x86/kernel/smpboot.c | 8 ++++----
include/linux/sched/topology.h | 10 +++++-----
kernel/sched/topology.c | 2 +-
4 files changed, 12 insertions(+), 12 deletions(-)
diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index de0f6f0..e503d23
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -1000,7 +1000,7 @@ static bool shared_caches;
#ifdef CONFIG_SCHED_SMT
/* cpumask of CPUs with asymmetric SMT dependency */
-static int powerpc_smt_flags(void)
+static int powerpc_smt_flags(const struct cpumask *cpu_map)
{
int flags = SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
@@ -1018,7 +1018,7 @@ static int powerpc_smt_flags(void)
* since the migrated task remains cache hot. We want to take advantage of this
* at the scheduler level so an extra topology level is required.
*/
-static int powerpc_shared_cache_flags(void)
+static int powerpc_shared_cache_flags(const struct cpumask *cpu_map)
{
return SD_SHARE_PKG_RESOURCES;
}
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 2ef1477..c005a8e
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -535,25 +535,25 @@ static bool match_llc(struct cpuinfo_x86 *c, struct cpuinfo_x86 *o)
#if defined(CONFIG_SCHED_SMT) || defined(CONFIG_SCHED_CLUSTER) || defined(CONFIG_SCHED_MC)
-static inline int x86_sched_itmt_flags(void)
+static inline int x86_sched_itmt_flags(const struct cpumask *cpu_map)
{
return sysctl_sched_itmt_enabled ? SD_ASYM_PACKING : 0;
}
#ifdef CONFIG_SCHED_MC
-static int x86_core_flags(void)
+static int x86_core_flags(const struct cpumask *cpu_map)
{
return cpu_core_flags() | x86_sched_itmt_flags();
}
#endif
#ifdef CONFIG_SCHED_SMT
-static int x86_smt_flags(void)
+static int x86_smt_flags(const struct cpumask *cpu_map)
{
return cpu_smt_flags() | x86_sched_itmt_flags();
}
#endif
#ifdef CONFIG_SCHED_CLUSTER
-static int x86_cluster_flags(void)
+static int x86_cluster_flags(const struct cpumask *cpu_map)
{
return cpu_cluster_flags() | x86_sched_itmt_flags();
}
diff --git a/include/linux/sched/topology.h b/include/linux/sched/topology.h
index 56cffe4..6aa985a
--- a/include/linux/sched/topology.h
+++ b/include/linux/sched/topology.h
@@ -36,28 +36,28 @@ extern const struct sd_flag_debug sd_flag_debug[];
#endif
#ifdef CONFIG_SCHED_SMT
-static inline int cpu_smt_flags(void)
+static inline int cpu_smt_flags(const struct cpumask *cpu_map)
{
return SD_SHARE_CPUCAPACITY | SD_SHARE_PKG_RESOURCES;
}
#endif
#ifdef CONFIG_SCHED_CLUSTER
-static inline int cpu_cluster_flags(void)
+static inline int cpu_cluster_flags(const struct cpumask *cpu_map)
{
return SD_SHARE_PKG_RESOURCES;
}
#endif
#ifdef CONFIG_SCHED_MC
-static inline int cpu_core_flags(void)
+static inline int cpu_core_flags(const struct cpumask *cpu_map)
{
return SD_SHARE_PKG_RESOURCES;
}
#endif
#ifdef CONFIG_NUMA
-static inline int cpu_numa_flags(void)
+static inline int cpu_numa_flags(const struct cpumask *cpu_map)
{
return SD_NUMA;
}
@@ -180,7 +180,7 @@ void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms);
bool cpus_share_cache(int this_cpu, int that_cpu);
typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
-typedef int (*sched_domain_flags_f)(void);
+typedef int (*sched_domain_flags_f)(const struct cpumask *cpu_map);
#define SDTL_OVERLAP 0x01
diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c
index 05b6c2a..34dfec4
--- a/kernel/sched/topology.c
+++ b/kernel/sched/topology.c
@@ -1556,7 +1556,7 @@ sd_init(struct sched_domain_topology_level *tl,
sd_weight = cpumask_weight(tl->mask(cpu));
if (tl->sd_flags)
- sd_flags = (*tl->sd_flags)();
+ sd_flags = (*tl->sd_flags)(tl->mask(cpu));
if (WARN_ONCE(sd_flags & ~TOPOLOGY_SD_FLAGS,
"wrong sd_flags in topology description\n"))
sd_flags &= TOPOLOGY_SD_FLAGS;
--
2.7.4
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
^ permalink raw reply related [flat|nested] 5+ messages in thread* Re: [PATCH 1/3] sched: topology: add input parameter for sched_domain_flags_f()
2022-03-29 9:15 ` [PATCH 1/3] sched: topology: add input parameter for sched_domain_flags_f() Qing Wang
@ 2022-03-29 9:23 ` Peter Zijlstra
0 siblings, 0 replies; 5+ messages in thread
From: Peter Zijlstra @ 2022-03-29 9:23 UTC (permalink / raw)
To: Qing Wang
Cc: Catalin Marinas, Will Deacon, Michael Ellerman,
Benjamin Herrenschmidt, Paul Mackerras, Thomas Gleixner,
Ingo Molnar, Borislav Petkov, x86, H. Peter Anvin, Sudeep Holla,
Greg Kroah-Hartman, Rafael J. Wysocki, Juri Lelli,
Vincent Guittot, Dietmar Eggemann, Steven Rostedt, Ben Segall,
Mel Gorman, Daniel Bristot de Oliveira, linux-arm-kernel,
linux-kernel, linuxppc-dev
On Tue, Mar 29, 2022 at 02:15:19AM -0700, Qing Wang wrote:
> From: Wang Qing <wangqing@vivo.com>
>
> sched_domain_flags_f() are statically set now, but actually, we can get a lot
> of necessary information based on the cpu_map. e.g. we can know whether its
> cache is shared.
>
> Allows custom extension without affecting current.
Still NAK
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 2/3] arch_topology: support for describing cache topology from DT
2022-03-29 9:15 [PATCH 0/3] support for describing cache topology from DT Qing Wang
2022-03-29 9:15 ` [PATCH 1/3] sched: topology: add input parameter for sched_domain_flags_f() Qing Wang
@ 2022-03-29 9:15 ` Qing Wang
2022-03-29 9:15 ` [PATCH 3/3] arm64: add arm64 default topology Qing Wang
2 siblings, 0 replies; 5+ messages in thread
From: Qing Wang @ 2022-03-29 9:15 UTC (permalink / raw)
To: Catalin Marinas, Will Deacon, Michael Ellerman,
Benjamin Herrenschmidt, Paul Mackerras, Thomas Gleixner,
Ingo Molnar, Borislav Petkov, x86, H. Peter Anvin, Sudeep Holla,
Greg Kroah-Hartman, Rafael J. Wysocki, Peter Zijlstra, Juri Lelli,
Vincent Guittot, Dietmar Eggemann, Steven Rostedt, Ben Segall,
Mel Gorman, Daniel Bristot de Oliveira, linux-arm-kernel,
linux-kernel, linuxppc-dev
Cc: Wang Qing
From: Wang Qing <wangqing@vivo.com>
When ACPI is not enabled, we can get cache topolopy from DT like:
* cpu0: cpu@000 {
* next-level-cache = <&L2_1>;
* L2_1: l2-cache {
* compatible = "cache";
* next-level-cache = <&L3_1>;
* };
* L3_1: l3-cache {
* compatible = "cache";
* };
* };
*
* cpu1: cpu@001 {
* next-level-cache = <&L2_1>;
* cpu-idle-states = <&clusteroff_l &mcusysoff
* &system_mem &system_pll &system_bus
* &s2idle>;
* };
* cpu2: cpu@002 {
* L2_2: l2-cache {
* compatible = "cache";
* next-level-cache = <&L3_1>;
* };
* };
*
* cpu3: cpu@003 {
* next-level-cache = <&L2_2>;
* };
cache_topology hold the pointer describing "next-level-cache",
it can describe the cache topology of every level.
Signed-off-by: Wang Qing <wangqing@vivo.com>
---
drivers/base/arch_topology.c | 89 ++++++++++++++++++++++++++++++++++++++++++-
include/linux/arch_topology.h | 4 ++
2 files changed, 92 insertions(+), 1 deletion(-)
diff --git a/drivers/base/arch_topology.c b/drivers/base/arch_topology.c
index 1d6636e..41e0301
--- a/drivers/base/arch_topology.c
+++ b/drivers/base/arch_topology.c
@@ -647,6 +647,92 @@ static int __init parse_dt_topology(void)
}
#endif
+
+/*
+ * cpu cache topology table
+ */
+#define MAX_CACHE_LEVEL 7
+struct device_node *cache_topology[NR_CPUS][MAX_CACHE_LEVEL];
+
+void init_cpu_cache_topology(void)
+{
+ struct device_node *node_cpu, *node_cache;
+ int cpu;
+ int level = 0;
+
+ for_each_possible_cpu(cpu) {
+ node_cpu = of_get_cpu_node(cpu, NULL);
+ if (!node_cpu)
+ continue;
+
+ level = 0;
+ node_cache = node_cpu;
+ while (level < MAX_CACHE_LEVEL) {
+ node_cache = of_parse_phandle(node_cache, "next-level-cache", 0);
+ if (!node_cache)
+ break;
+
+ cache_topology[cpu][level++] = node_cache;
+ }
+ of_node_put(node_cpu);
+ }
+}
+
+/*
+ * private means only shared within cpu_mask
+ * Returns -1 if not described int DT.
+ */
+int cpu_share_private_cache(const struct cpumask *cpu_mask)
+{
+ int cache_level, cpu_id;
+ struct cpumask cache_mask;
+ int cpu = cpumask_first(cpu_mask);
+
+ for (cache_level = 0; cache_level < MAX_CACHE_LEVEL; cache_level++) {
+ if (!cache_topology[cpu][cache_level])
+ return -1;
+
+ cpumask_clear(&cache_mask);
+ for (cpu_id = 0; cpu_id < NR_CPUS; cpu_id++) {
+ if (cache_topology[cpu][cache_level] == cache_topology[cpu_id][cache_level])
+ cpumask_set_cpu(cpu_id, &cache_mask);
+ }
+
+ if (cpumask_equal(cpu_mask, &cache_mask))
+ return 1;
+ }
+
+ return 0;
+}
+
+bool cpu_share_llc(int cpu1, int cpu2)
+{
+ int cache_level;
+
+ for (cache_level = MAX_CACHE_LEVEL - 1; cache_level > 0; cache_level--) {
+ if (!cache_topology[cpu1][cache_level])
+ continue;
+
+ if (cache_topology[cpu1][cache_level] == cache_topology[cpu2][cache_level])
+ return true;
+
+ return false;
+ }
+
+ return false;
+}
+
+bool cpu_share_l2c(int cpu1, int cpu2)
+{
+ if (!cache_topology[cpu1][0])
+ return false;
+
+ if (cache_topology[cpu1][0] == cache_topology[cpu2][0])
+ return true;
+
+ return false;
+}
+
/*
* cpu topology table
*/
@@ -684,7 +770,8 @@ void update_siblings_masks(unsigned int cpuid)
for_each_online_cpu(cpu) {
cpu_topo = &cpu_topology[cpu];
- if (cpuid_topo->llc_id == cpu_topo->llc_id) {
+ if ((cpuid_topo->llc_id != -1 && cpuid_topo->llc_id == cpu_topo->llc_id)
+ || (cpuid_topo->llc_id == -1 && cpu_share_llc(cpu, cpuid))) {
cpumask_set_cpu(cpu, &cpuid_topo->llc_sibling);
cpumask_set_cpu(cpuid, &cpu_topo->llc_sibling);
}
diff --git a/include/linux/arch_topology.h b/include/linux/arch_topology.h
index 58cbe18..a402ff6
--- a/include/linux/arch_topology.h
+++ b/include/linux/arch_topology.h
@@ -86,6 +86,10 @@ extern struct cpu_topology cpu_topology[NR_CPUS];
#define topology_cluster_cpumask(cpu) (&cpu_topology[cpu].cluster_sibling)
#define topology_llc_cpumask(cpu) (&cpu_topology[cpu].llc_sibling)
void init_cpu_topology(void);
+void init_cpu_cache_topology(void);
+int cpu_share_private_cache(const struct cpumask *cpu_mask);
+bool cpu_share_llc(int cpu1, int cpu2);
+bool cpu_share_l2c(int cpu1, int cpu2);
void store_cpu_topology(unsigned int cpuid);
const struct cpumask *cpu_coregroup_mask(int cpu);
const struct cpumask *cpu_clustergroup_mask(int cpu);
--
2.7.4
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
^ permalink raw reply related [flat|nested] 5+ messages in thread* [PATCH 3/3] arm64: add arm64 default topology
2022-03-29 9:15 [PATCH 0/3] support for describing cache topology from DT Qing Wang
2022-03-29 9:15 ` [PATCH 1/3] sched: topology: add input parameter for sched_domain_flags_f() Qing Wang
2022-03-29 9:15 ` [PATCH 2/3] arch_topology: support for describing cache topology from DT Qing Wang
@ 2022-03-29 9:15 ` Qing Wang
2 siblings, 0 replies; 5+ messages in thread
From: Qing Wang @ 2022-03-29 9:15 UTC (permalink / raw)
To: Catalin Marinas, Will Deacon, Michael Ellerman,
Benjamin Herrenschmidt, Paul Mackerras, Thomas Gleixner,
Ingo Molnar, Borislav Petkov, x86, H. Peter Anvin, Sudeep Holla,
Greg Kroah-Hartman, Rafael J. Wysocki, Peter Zijlstra, Juri Lelli,
Vincent Guittot, Dietmar Eggemann, Steven Rostedt, Ben Segall,
Mel Gorman, Daniel Bristot de Oliveira, linux-arm-kernel,
linux-kernel, linuxppc-dev
Cc: Wang Qing
From: Wang Qing <wangqing@vivo.com>
default_topology does not fit arm64, especially CPU and cache topology.
Add arm64_topology, so we can do more based on CONFIG_GENERIC_ARCH_TOPOLOGY.
arm64_xxx_flags() prefer to get the cache attribute from DT.
Signed-off-by: Wang Qing <wangqing@vivo.com>
---
arch/arm64/kernel/smp.c | 56 +++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 56 insertions(+)
diff --git a/arch/arm64/kernel/smp.c b/arch/arm64/kernel/smp.c
index 27df5c1..d245012
--- a/arch/arm64/kernel/smp.c
+++ b/arch/arm64/kernel/smp.c
@@ -715,6 +715,60 @@ void __init smp_init_cpus(void)
}
}
+#ifdef CONFIG_SCHED_CLUSTER
+static int arm64_cluster_flags(const struct cpumask *cpu_map)
+{
+ int flag = cpu_cluster_flags();
+ int ret = cpu_share_private_cache(cpu_map);
+ if (ret == 1)
+ flag |= SD_SHARE_PKG_RESOURCES;
+ else if (ret == 0)
+ flag &= ~SD_SHARE_PKG_RESOURCES;
+
+ return flag;
+}
+#endif
+
+#ifdef CONFIG_SCHED_MC
+static int arm64_core_flags(const struct cpumask *cpu_map)
+{
+ int flag = cpu_core_flags();
+ int ret = cpu_share_private_cache(cpu_map);
+ if (ret == 1)
+ flag |= SD_SHARE_PKG_RESOURCES;
+ else if (ret == 0)
+ flag &= ~SD_SHARE_PKG_RESOURCES;
+
+ return flag;
+}
+#endif
+
+static int arm64_die_flags(const struct cpumask *cpu_map)
+{
+ int flag = 0;
+ int ret = cpu_share_private_cache(cpu_map);
+ if (ret == 1)
+ flag |= SD_SHARE_PKG_RESOURCES;
+ else if (ret == 0)
+ flag &= ~SD_SHARE_PKG_RESOURCES;
+
+ return flag;
+}
+
+static struct sched_domain_topology_level arm64_topology[] = {
+#ifdef CONFIG_SCHED_SMT
+ { cpu_smt_mask, cpu_smt_flags, SD_INIT_NAME(SMT) },
+#endif
+#ifdef CONFIG_SCHED_CLUSTER
+ { cpu_clustergroup_mask, arm64_cluster_flags, SD_INIT_NAME(CLS) },
+#endif
+#ifdef CONFIG_SCHED_MC
+ { cpu_coregroup_mask, arm64_core_flags, SD_INIT_NAME(MC) },
+#endif
+ { cpu_cpu_mask, arm64_die_flags, SD_INIT_NAME(DIE) },
+ { NULL, },
+};
+
void __init smp_prepare_cpus(unsigned int max_cpus)
{
const struct cpu_operations *ops;
@@ -723,6 +777,8 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
unsigned int this_cpu;
init_cpu_topology();
+ init_cpu_cache_topology();
+ set_sched_topology(arm64_topology);
this_cpu = smp_processor_id();
store_cpu_topology(this_cpu);
--
2.7.4
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
^ permalink raw reply related [flat|nested] 5+ messages in thread