diff options
author | Heiko Carstens <heiko.carstens@de.ibm.com> | 2020-02-19 13:29:15 +0100 |
---|---|---|
committer | Vasily Gorbik <gor@linux.ibm.com> | 2020-02-27 16:02:21 +0100 |
commit | 701dc81e7412daaf3c5bf4bc55d35c8b1525112a (patch) | |
tree | 1f986fd4e3826553d5d6dce323dc95a756d5ce37 /arch | |
parent | 4a559cd15dbc79958fa9b18ad4e8afe4a0bf4744 (diff) |
s390/mm: remove fake numa support
It turned out that fake numa support is rather useless on s390, since
there are no scenarios where there is any performance or other benefit
when used.
However it does provide maintenance cost and breaks from time to time.
Therefore remove it.
CONFIG_NUMA is still supported with a very small backend and only one
node. This way userspace applications which require NUMA interfaces
continue to work.
Note that NODES_SHIFT is set to 1 (= 2 nodes) instead of 0 (= 1 node),
since there is quite a bit of kernel code which assumes that more than
one node is possible if CONFIG_NUMA is enabled.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Diffstat (limited to 'arch')
-rw-r--r-- | arch/s390/Kconfig | 61 | ||||
-rw-r--r-- | arch/s390/include/asm/numa.h | 13 | ||||
-rw-r--r-- | arch/s390/include/asm/topology.h | 9 | ||||
-rw-r--r-- | arch/s390/kernel/setup.c | 1 | ||||
-rw-r--r-- | arch/s390/kernel/topology.c | 2 | ||||
-rw-r--r-- | arch/s390/numa/Makefile | 2 | ||||
-rw-r--r-- | arch/s390/numa/mode_emu.c | 577 | ||||
-rw-r--r-- | arch/s390/numa/numa.c | 147 | ||||
-rw-r--r-- | arch/s390/numa/numa_mode.h | 25 | ||||
-rw-r--r-- | arch/s390/numa/toptree.c | 351 | ||||
-rw-r--r-- | arch/s390/numa/toptree.h | 61 |
11 files changed, 19 insertions, 1230 deletions
diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig index 8abe77536d9d..6b1f715dd8bb 100644 --- a/arch/s390/Kconfig +++ b/arch/s390/Kconfig @@ -450,14 +450,6 @@ config NR_CPUS config HOTPLUG_CPU def_bool y -# Some NUMA nodes have memory ranges that span -# other nodes. Even though a pfn is valid and -# between a node's start and end pfns, it may not -# reside on that node. See memmap_init_zone() -# for details. <- They meant memory holes! -config NODES_SPAN_OTHER_NODES - def_bool NUMA - config NUMA bool "NUMA support" depends on SCHED_TOPOLOGY @@ -467,58 +459,9 @@ config NUMA This option adds NUMA support to the kernel. - An operation mode can be selected by appending - numa=<method> to the kernel command line. - - The default behaviour is identical to appending numa=plain to - the command line. This will create just one node with all - available memory and all CPUs in it. - config NODES_SHIFT - int "Maximum NUMA nodes (as a power of 2)" - range 1 10 - depends on NUMA - default "4" - help - Specify the maximum number of NUMA nodes available on the target - system. Increases memory reserved to accommodate various tables. - -menu "Select NUMA modes" - depends on NUMA - -config NUMA_EMU - bool "NUMA emulation" - default y - help - Numa emulation mode will split the available system memory into - equal chunks which then are distributed over the configured number - of nodes in a round-robin manner. - - The number of fake nodes is limited by the number of available memory - chunks (i.e. memory size / fake size) and the number of supported - nodes in the kernel. - - The CPUs are assigned to the nodes in a way that partially respects - the original machine topology (if supported by the machine). - Fair distribution of the CPUs is not guaranteed. - -config EMU_SIZE - hex "NUMA emulation memory chunk size" - default 0x10000000 - range 0x400000 0x100000000 - depends on NUMA_EMU - help - Select the default size by which the memory is chopped and then - assigned to emulated NUMA nodes. - - This can be overridden by specifying - - emu_size=<n> - - on the kernel command line where also suffixes K, M, G, and T are - supported. - -endmenu + int + default "1" config SCHED_SMT def_bool n diff --git a/arch/s390/include/asm/numa.h b/arch/s390/include/asm/numa.h index 35f8cbe7e5bb..23cd5d1b734b 100644 --- a/arch/s390/include/asm/numa.h +++ b/arch/s390/include/asm/numa.h @@ -13,24 +13,13 @@ #ifdef CONFIG_NUMA #include <linux/numa.h> -#include <linux/cpumask.h> void numa_setup(void); -int numa_pfn_to_nid(unsigned long pfn); -int __node_distance(int a, int b); -void numa_update_cpu_topology(void); - -extern cpumask_t node_to_cpumask_map[MAX_NUMNODES]; -extern int numa_debug_enabled; #else static inline void numa_setup(void) { } -static inline void numa_update_cpu_topology(void) { } -static inline int numa_pfn_to_nid(unsigned long pfn) -{ - return 0; -} #endif /* CONFIG_NUMA */ + #endif /* _ASM_S390_NUMA_H */ diff --git a/arch/s390/include/asm/topology.h b/arch/s390/include/asm/topology.h index cca406fdbe51..bd3417185e30 100644 --- a/arch/s390/include/asm/topology.h +++ b/arch/s390/include/asm/topology.h @@ -16,7 +16,6 @@ struct cpu_topology_s390 { unsigned short socket_id; unsigned short book_id; unsigned short drawer_id; - unsigned short node_id; unsigned short dedicated : 1; cpumask_t thread_mask; cpumask_t core_mask; @@ -71,19 +70,23 @@ static inline void topology_expect_change(void) { } #define cpu_to_node cpu_to_node static inline int cpu_to_node(int cpu) { - return cpu_topology[cpu].node_id; + return 0; } /* Returns a pointer to the cpumask of CPUs on node 'node'. */ #define cpumask_of_node cpumask_of_node static inline const struct cpumask *cpumask_of_node(int node) { - return &node_to_cpumask_map[node]; + return cpu_possible_mask; } #define pcibus_to_node(bus) __pcibus_to_node(bus) #define node_distance(a, b) __node_distance(a, b) +static inline int __node_distance(int a, int b) +{ + return 0; +} #else /* !CONFIG_NUMA */ diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c index b2c2f75860e8..1158a63a8e0e 100644 --- a/arch/s390/kernel/setup.c +++ b/arch/s390/kernel/setup.c @@ -790,6 +790,7 @@ static void __init memblock_add_mem_detect_info(void) memblock_physmem_add(start, end - start); } memblock_set_bottom_up(false); + memblock_set_node(0, ULONG_MAX, &memblock.memory, 0); memblock_dump_all(); } diff --git a/arch/s390/kernel/topology.c b/arch/s390/kernel/topology.c index 3627953007ed..c189f5d996ff 100644 --- a/arch/s390/kernel/topology.c +++ b/arch/s390/kernel/topology.c @@ -26,7 +26,6 @@ #include <linux/nodemask.h> #include <linux/node.h> #include <asm/sysinfo.h> -#include <asm/numa.h> #define PTF_HORIZONTAL (0UL) #define PTF_VERTICAL (1UL) @@ -267,7 +266,6 @@ static void update_cpu_masks(void) cpumask_set_cpu(cpu, &cpus_with_topology); } } - numa_update_cpu_topology(); } void store_topology(struct sysinfo_15_1_x *info) diff --git a/arch/s390/numa/Makefile b/arch/s390/numa/Makefile index 66c2dff74895..c89d26f4f77d 100644 --- a/arch/s390/numa/Makefile +++ b/arch/s390/numa/Makefile @@ -1,4 +1,2 @@ # SPDX-License-Identifier: GPL-2.0 obj-y += numa.o -obj-y += toptree.o -obj-$(CONFIG_NUMA_EMU) += mode_emu.o diff --git a/arch/s390/numa/mode_emu.c b/arch/s390/numa/mode_emu.c deleted file mode 100644 index 72d742bb2d17..000000000000 --- a/arch/s390/numa/mode_emu.c +++ /dev/null @@ -1,577 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 -/* - * NUMA support for s390 - * - * NUMA emulation (aka fake NUMA) distributes the available memory to nodes - * without using real topology information about the physical memory of the - * machine. - * - * It distributes the available CPUs to nodes while respecting the original - * machine topology information. This is done by trying to avoid to separate - * CPUs which reside on the same book or even on the same MC. - * - * Because the current Linux scheduler code requires a stable cpu to node - * mapping, cores are pinned to nodes when the first CPU thread is set online. - * - * Copyright IBM Corp. 2015 - */ - -#define KMSG_COMPONENT "numa_emu" -#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt - -#include <linux/kernel.h> -#include <linux/cpumask.h> -#include <linux/memblock.h> -#include <linux/node.h> -#include <linux/memory.h> -#include <linux/slab.h> -#include <asm/smp.h> -#include <asm/topology.h> -#include "numa_mode.h" -#include "toptree.h" - -/* Distances between the different system components */ -#define DIST_EMPTY 0 -#define DIST_CORE 1 -#define DIST_MC 2 -#define DIST_BOOK 3 -#define DIST_DRAWER 4 -#define DIST_MAX 5 - -/* Node distance reported to common code */ -#define EMU_NODE_DIST 10 - -/* Node ID for free (not yet pinned) cores */ -#define NODE_ID_FREE -1 - -/* Different levels of toptree */ -enum toptree_level {CORE, MC, BOOK, DRAWER, NODE, TOPOLOGY}; - -/* The two toptree IDs */ -enum {TOPTREE_ID_PHYS, TOPTREE_ID_NUMA}; - -/* Number of NUMA nodes */ -static int emu_nodes = 1; -/* NUMA stripe size */ -static unsigned long emu_size; - -/* - * Node to core pinning information updates are protected by - * "sched_domains_mutex". - */ -static struct { - s32 to_node_id[CONFIG_NR_CPUS]; /* Pinned core to node mapping */ - int total; /* Total number of pinned cores */ - int per_node_target; /* Cores per node without extra cores */ - int per_node[MAX_NUMNODES]; /* Number of cores pinned to node */ -} *emu_cores; - -/* - * Pin a core to a node - */ -static void pin_core_to_node(int core_id, int node_id) -{ - if (emu_cores->to_node_id[core_id] == NODE_ID_FREE) { - emu_cores->per_node[node_id]++; - emu_cores->to_node_id[core_id] = node_id; - emu_cores->total++; - } else { - WARN_ON(emu_cores->to_node_id[core_id] != node_id); - } -} - -/* - * Number of pinned cores of a node - */ -static int cores_pinned(struct toptree *node) -{ - return emu_cores->per_node[node->id]; -} - -/* - * ID of the node where the core is pinned (or NODE_ID_FREE) - */ -static int core_pinned_to_node_id(struct toptree *core) -{ - return emu_cores->to_node_id[core->id]; -} - -/* - * Number of cores in the tree that are not yet pinned - */ -static int cores_free(struct toptree *tree) -{ - struct toptree *core; - int count = 0; - - toptree_for_each(core, tree, CORE) { - if (core_pinned_to_node_id(core) == NODE_ID_FREE) - count++; - } - return count; -} - -/* - * Return node of core - */ -static struct toptree *core_node(struct toptree *core) -{ - return core->parent->parent->parent->parent; -} - -/* - * Return drawer of core - */ -static struct toptree *core_drawer(struct toptree *core) -{ - return core->parent->parent->parent; -} - -/* - * Return book of core - */ -static struct toptree *core_book(struct toptree *core) -{ - return core->parent->parent; -} - -/* - * Return mc of core - */ -static struct toptree *core_mc(struct toptree *core) -{ - return core->parent; -} - -/* - * Distance between two cores - */ -static int dist_core_to_core(struct toptree *core1, struct toptree *core2) -{ - if (core_drawer(core1)->id != core_drawer(core2)->id) - return DIST_DRAWER; - if (core_book(core1)->id != core_book(core2)->id) - return DIST_BOOK; - if (core_mc(core1)->id != core_mc(core2)->id) - return DIST_MC; - /* Same core or sibling on same MC */ - return DIST_CORE; -} - -/* - * Distance of a node to a core - */ -static int dist_node_to_core(struct toptree *node, struct toptree *core) -{ - struct toptree *core_node; - int dist_min = DIST_MAX; - - toptree_for_each(core_node, node, CORE) - dist_min = min(dist_min, dist_core_to_core(core_node, core)); - return dist_min == DIST_MAX ? DIST_EMPTY : dist_min; -} - -/* - * Unify will delete empty nodes, therefore recreate nodes. - */ -static void toptree_unify_tree(struct toptree *tree) -{ - int nid; - - toptree_unify(tree); - for (nid = 0; nid < emu_nodes; nid++) - toptree_get_child(tree, nid); -} - -/* - * Find the best/nearest node for a given core and ensure that no node - * gets more than "emu_cores->per_node_target + extra" cores. - */ -static struct toptree *node_for_core(struct toptree *numa, struct toptree *core, - int extra) -{ - struct toptree *node, *node_best = NULL; - int dist_cur, dist_best, cores_target; - - cores_target = emu_cores->per_node_target + extra; - dist_best = DIST_MAX; - node_best = NULL; - toptree_for_each(node, numa, NODE) { - /* Already pinned cores must use their nodes */ - if (core_pinned_to_node_id(core) == node->id) { - node_best = node; - break; - } - /* Skip nodes that already have enough cores */ - if (cores_pinned(node) >= cores_target) - continue; - dist_cur = dist_node_to_core(node, core); - if (dist_cur < dist_best) { - dist_best = dist_cur; - node_best = node; - } - } - return node_best; -} - -/* - * Find the best node for each core with respect to "extra" core count - */ -static void toptree_to_numa_single(struct toptree *numa, struct toptree *phys, - int extra) -{ - struct toptree *node, *core, *tmp; - - toptree_for_each_safe(core, tmp, phys, CORE) { - node = node_for_core(numa, core, extra); - if (!node) - return; - toptree_move(core, node); - pin_core_to_node(core->id, node->id); - } -} - -/* - * Move structures of given level to specified NUMA node - */ -static void move_level_to_numa_node(struct toptree *node, struct toptree *phys, - enum toptree_level level, bool perfect) -{ - int cores_free, cores_target = emu_cores->per_node_target; - struct toptree *cur, *tmp; - - toptree_for_each_safe(cur, tmp, phys, level) { - cores_free = cores_target - toptree_count(node, CORE); - if (perfect) { - if (cores_free == toptree_count(cur, CORE)) - toptree_move(cur, node); - } else { - if (cores_free >= toptree_count(cur, CORE)) - toptree_move(cur, node); - } - } -} - -/* - * Move structures of a given level to NUMA nodes. If "perfect" is specified - * move only perfectly fitting structures. Otherwise move also smaller - * than needed structures. - */ -static void move_level_to_numa(struct toptree *numa, struct toptree *phys, - enum toptree_level level, bool perfect) -{ - struct toptree *node; - - toptree_for_each(node, numa, NODE) - move_level_to_numa_node(node, phys, level, perfect); -} - -/* - * For the first run try to move the big structures - */ -static void toptree_to_numa_first(struct toptree *numa, struct toptree *phys) -{ - struct toptree *core; - - /* Always try to move perfectly fitting structures first */ - move_level_to_numa(numa, phys, DRAWER, true); - move_level_to_numa(numa, phys, DRAWER, false); - move_level_to_numa(numa, phys, BOOK, true); - move_level_to_numa(numa, phys, BOOK, false); - move_level_to_numa(numa, phys, MC, true); - move_level_to_numa(numa, phys, MC, false); - /* Now pin all the moved cores */ - toptree_for_each(core, numa, CORE) - pin_core_to_node(core->id, core_node(core)->id); -} - -/* - * Allocate new topology and create required nodes - */ -static struct toptree *toptree_new(int id, int nodes) -{ - struct toptree *tree; - int nid; - - tree = toptree_alloc(TOPOLOGY, id); - if (!tree) - goto fail; - for (nid = 0; nid < nodes; nid++) { - if (!toptree_get_child(tree, nid)) - goto fail; - } - return tree; -fail: - panic("NUMA emulation could not allocate topology"); -} - -/* - * Allocate and initialize core to node mapping - */ -static void __ref create_core_to_node_map(void) -{ - int i; - - emu_cores = memblock_alloc(sizeof(*emu_cores), 8); - if (!emu_cores) - panic("%s: Failed to allocate %zu bytes align=0x%x\n", - __func__, sizeof(*emu_cores), 8); - for (i = 0; i < ARRAY_SIZE(emu_cores->to_node_id); i++) - emu_cores->to_node_id[i] = NODE_ID_FREE; -} - -/* - * Move cores from physical topology into NUMA target topology - * and try to keep as much of the physical topology as possible. - */ -static struct toptree *toptree_to_numa(struct toptree *phys) -{ - static int first = 1; - struct toptree *numa; - int cores_total; - - cores_total = emu_cores->total + cores_free(phys); - emu_cores->per_node_target = cores_total / emu_nodes; - numa = toptree_new(TOPTREE_ID_NUMA, emu_nodes); - if (first) { - toptree_to_numa_first(numa, phys); - first = 0; - } - toptree_to_numa_single(numa, phys, 0); - toptree_to_numa_single(numa, phys, 1); - toptree_unify_tree(numa); - - WARN_ON(cpumask_weight(&phys->mask)); - return numa; -} - -/* - * Create a toptree out of the physical topology that we got from the hypervisor - */ -static struct toptree *toptree_from_topology(void) -{ - struct toptree *phys, *node, *drawer, *book, *mc, *core; - struct cpu_topology_s390 *top; - int cpu; - - phys = toptree_new(TOPTREE_ID_PHYS, 1); - - for_each_cpu(cpu, &cpus_with_topology) { - top = &cpu_topology[cpu]; - node = toptree_get_child(phys, 0); - drawer = toptree_get_child(node, top->drawer_id); - book = toptree_get_child(drawer, top->book_id); - mc = toptree_get_child(book, top->socket_id); - core = toptree_get_child(mc, smp_get_base_cpu(cpu)); - if (!drawer || !book || !mc || !core) - panic("NUMA emulation could not allocate memory"); - cpumask_set_cpu(cpu, &core->mask); - toptree_update_mask(mc); - } - return phys; -} - -/* - * Add toptree core to topology and create correct CPU masks - */ -static void topology_add_core(struct toptree *core) -{ - struct cpu_topology_s390 *top; - int cpu; - - for_each_cpu(cpu, &core->mask) { - top = &cpu_topology[cpu]; - cpumask_copy(&top->thread_mask, &core->mask); - cpumask_copy(&top->core_mask, &core_mc(core)->mask); - cpumask_copy(&top->book_mask, &core_book(core)->mask); - cpumask_copy(&top->drawer_mask, &core_drawer(core)->mask); - cpumask_set_cpu(cpu, &node_to_cpumask_map[core_node(core)->id]); - top->node_id = core_node(core)->id; - } -} - -/* - * Apply toptree to topology and create CPU masks - */ -static void toptree_to_topology(struct toptree *numa) -{ - struct toptree *core; - int i; - - /* Clear all node masks */ - for (i = 0; i < MAX_NUMNODES; i++) - cpumask_clear(&node_to_cpumask_map[i]); - - /* Rebuild all masks */ - toptree_for_each(core, numa, CORE) - topology_add_core(core); -} - -/* - * Show the node to core mapping - */ -static void print_node_to_core_map(void) -{ - int nid, cid; - - if (!numa_debug_enabled) - return; - printk(KERN_DEBUG "NUMA node to core mapping\n"); - for (nid = 0; nid < emu_nodes; nid++) { - printk(KERN_DEBUG " node %3d: ", nid); - for (cid = 0; cid < ARRAY_SIZE(emu_cores->to_node_id); cid++) { - if (emu_cores->to_node_id[cid] == nid) - printk(KERN_CONT "%d ", cid); - } - printk(KERN_CONT "\n"); - } -} - -static void pin_all_possible_cpus(void) -{ - int core_id, node_id, cpu; - static int initialized; - - if (initialized) - return; - print_node_to_core_map(); - node_id = 0; - for_each_possible_cpu(cpu) { - core_id = smp_get_base_cpu(cpu); - if (emu_cores->to_node_id[core_id] != NODE_ID_FREE) - continue; - pin_core_to_node(core_id, node_id); - cpu_topology[cpu].node_id = node_id; - node_id = (node_id + 1) % emu_nodes; - } - print_node_to_core_map(); - initialized = 1; -} - -/* - * Transfer physical topology into a NUMA topology and modify CPU masks - * according to the NUMA topology. - * - * Must be called with "sched_domains_mutex" lock held. - */ -static void emu_update_cpu_topology(void) -{ - struct toptree *phys, *numa; - - if (emu_cores == NULL) - create_core_to_node_map(); - phys = toptree_from_topology(); - numa = toptree_to_numa(phys); - toptree_free(phys); - toptree_to_topology(numa); - toptree_free(numa); - pin_all_possible_cpus(); -} - -/* - * If emu_size is not set, use CONFIG_EMU_SIZE. Then round to minimum - * alignment (needed for memory hotplug). - */ -static unsigned long emu_setup_size_adjust(unsigned long size) -{ - unsigned long size_new; - - size = size ? : CONFIG_EMU_SIZE; - size_new = roundup(size, memory_block_size_bytes()); - if (size_new == size) - return size; - pr_warn("Increasing memory stripe size from %ld MB to %ld MB\n", - size >> 20, size_new >> 20); - return size_new; -} - -/* - * If we have not enough memory for the specified nodes, reduce the node count. - */ -static int emu_setup_nodes_adjust(int nodes) -{ - int nodes_max; - - nodes_max = memblock.memory.total_size / emu_size; - nodes_max = max(nodes_max, 1); - if (nodes_max >= nodes) - return nodes; - pr_warn("Not enough memory for %d nodes, reducing node count\n", nodes); - return nodes_max; -} - -/* - * Early emu setup - */ -static void emu_setup(void) -{ - int nid; - - emu_size = emu_setup_size_adjust(emu_size); - emu_nodes = emu_setup_nodes_adjust(emu_nodes); - for (nid = 0; nid < emu_nodes; nid++) - node_set(nid, node_possible_map); - pr_info("Creating %d nodes with memory stripe size %ld MB\n", - emu_nodes, emu_size >> 20); -} - -/* - * Return node id for given page number - */ -static int emu_pfn_to_nid(unsigned long pfn) -{ - return (pfn / (emu_size >> PAGE_SHIFT)) % emu_nodes; -} - -/* - * Return stripe size - */ -static unsigned long emu_align(void) -{ - return emu_size; -} - -/* - * Return distance between two nodes - */ -static int emu_distance(int node1, int node2) -{ - return (node1 != node2) * EMU_NODE_DIST; -} - -/* - * Define callbacks for generic s390 NUMA infrastructure - */ -const struct numa_mode numa_mode_emu = { - .name = "emu", - .setup = emu_setup, - .update_cpu_topology = emu_update_cpu_topology, - .__pfn_to_nid = emu_pfn_to_nid, - .align = emu_align, - .distance = emu_distance, -}; - -/* - * Kernel parameter: emu_nodes=<n> - */ -static int __init early_parse_emu_nodes(char *p) -{ - int count; - - if (!p || kstrtoint(p, 0, &count) != 0 || count <= 0) - return 0; - emu_nodes = min(count, MAX_NUMNODES); - return 0; -} -early_param("emu_nodes", early_parse_emu_nodes); - -/* - * Kernel parameter: emu_size=[<n>[k|M|G|T]] - */ -static int __init early_parse_emu_size(char *p) -{ - if (p) - emu_size = memparse(p, NULL); - return 0; -} -early_param("emu_size", early_parse_emu_size); diff --git a/arch/s390/numa/numa.c b/arch/s390/numa/numa.c index d2910fa834c8..51c5a9f6e525 100644 --- a/arch/s390/numa/numa.c +++ b/arch/s390/numa/numa.c @@ -7,165 +7,36 @@ * Copyright IBM Corp. 2015 */ -#define KMSG_COMPONENT "numa" -#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt - #include <linux/kernel.h> #include <linux/mmzone.h> #include <linux/cpumask.h> #include <linux/memblock.h> -#include <linux/slab.h> #include <linux/node.h> - #include <asm/numa.h> -#include "numa_mode.h" -pg_data_t *node_data[MAX_NUMNODES]; +struct pglist_data *node_data[MAX_NUMNODES]; EXPORT_SYMBOL(node_data); -cpumask_t node_to_cpumask_map[MAX_NUMNODES]; -EXPORT_SYMBOL(node_to_cpumask_map); - -static void plain_setup(void) -{ - node_set(0, node_possible_map); -} - -const struct numa_mode numa_mode_plain = { - .name = "plain", - .setup = plain_setup, -}; - -static const struct numa_mode *mode = &numa_mode_plain; - -int numa_pfn_to_nid(unsigned long pfn) -{ - return mode->__pfn_to_nid ? mode->__pfn_to_nid(pfn) : 0; -} - -void numa_update_cpu_topology(void) -{ - if (mode->update_cpu_topology) - mode->update_cpu_topology(); -} - -int __node_distance(int a, int b) -{ - return mode->distance ? mode->distance(a, b) : 0; -} -EXPORT_SYMBOL(__node_distance); - -int numa_debug_enabled; - -/* - * numa_setup_memory() - Assign bootmem to nodes - * - * The memory is first added to memblock without any respect to nodes. - * This is fixed before remaining memblock memory is handed over to the - * buddy allocator. - * An important side effect is that large bootmem allocations might easily - * cross node boundaries, which can be needed for large allocations with - * smaller memory stripes in each node (i.e. when using NUMA emulation). - * - * Memory defines nodes: - * Therefore this routine also sets the nodes online with memory. - */ -static void __init numa_setup_memory(void) +void __init numa_setup(void) { - unsigned long cur_base, align, end_of_dram; - int nid = 0; - - end_of_dram = memblock_end_of_DRAM(); - align = mode->align ? mode->align() : ULONG_MAX; - - /* - * Step through all available memory and assign it to the nodes - * indicated by the mode implementation. - * All nodes which are seen here will be set online. - */ - cur_base = 0; - do { - nid = numa_pfn_to_nid(PFN_DOWN(cur_base)); - node_set_online(nid); - memblock_set_node(cur_base, align, &memblock.memory, nid); - cur_base += align; - } while (cur_base < end_of_dram); + int nid; - /* Allocate and fill out node_data */ + nodes_clear(node_possible_map); + node_set(0, node_possible_map); + node_set_online(0); for (nid = 0; nid < MAX_NUMNODES; nid++) { NODE_DATA(nid) = memblock_alloc(sizeof(pg_data_t), 8); if (!NODE_DATA(nid)) panic("%s: Failed to allocate %zu bytes align=0x%x\n", __func__, sizeof(pg_data_t), 8); } - - for_each_online_node(nid) { - unsigned long start_pfn, end_pfn; - unsigned long t_start, t_end; - int i; - - start_pfn = ULONG_MAX; - end_pfn = 0; - for_each_mem_pfn_range(i, nid, &t_start, &t_end, NULL) { - if (t_start < start_pfn) - start_pfn = t_start; - if (t_end > end_pfn) - end_pfn = t_end; - } - NODE_DATA(nid)->node_spanned_pages = end_pfn - start_pfn; - NODE_DATA(nid)->node_id = nid; - } -} - -/* - * numa_setup() - Earliest initialization - * - * Assign the mode and call the mode's setup routine. - */ -void __init numa_setup(void) -{ - pr_info("NUMA mode: %s\n", mode->name); - nodes_clear(node_possible_map); - /* Initially attach all possible CPUs to node 0. */ - cpumask_copy(&node_to_cpumask_map[0], cpu_possible_mask); - if (mode->setup) - mode->setup(); - numa_setup_memory(); - memblock_dump_all(); + NODE_DATA(0)->node_spanned_pages = memblock_end_of_DRAM() >> PAGE_SHIFT; + NODE_DATA(0)->node_id = 0; } -/* - * numa_init_late() - Initialization initcall - * - * Register NUMA nodes. - */ static int __init numa_init_late(void) { - int nid; - - for_each_online_node(nid) - register_one_node(nid); + register_one_node(0); return 0; } arch_initcall(numa_init_late); - -static int __init parse_debug(char *parm) -{ - numa_debug_enabled = 1; - return 0; -} -early_param("numa_debug", parse_debug); - -static int __init parse_numa(char *parm) -{ - if (!parm) - return 1; - if (strcmp(parm, numa_mode_plain.name) == 0) - mode = &numa_mode_plain; -#ifdef CONFIG_NUMA_EMU - if (strcmp(parm, numa_mode_emu.name) == 0) - mode = &numa_mode_emu; -#endif - return 0; -} -early_param("numa", parse_numa); diff --git a/arch/s390/numa/numa_mode.h b/arch/s390/numa/numa_mode.h deleted file mode 100644 index dfd3e2784081..000000000000 --- a/arch/s390/numa/numa_mode.h +++ /dev/null @@ -1,25 +0,0 @@ -/* SPDX-License-Identifier: GPL-2.0 */ -/* - * NUMA support for s390 - * - * Define declarations used for communication between NUMA mode - * implementations and NUMA core functionality. - * - * Copyright IBM Corp. 2015 - */ -#ifndef __S390_NUMA_MODE_H -#define __S390_NUMA_MODE_H - -struct numa_mode { - char *name; /* Name of mode */ - void (*setup)(void); /* Initizalize mode */ - void (*update_cpu_topology)(void); /* Called by topology code */ - int (*__pfn_to_nid)(unsigned long pfn); /* PFN to node ID */ - unsigned long (*align)(void); /* Minimum node alignment */ - int (*distance)(int a, int b); /* Distance between two nodes */ -}; - -extern const struct numa_mode numa_mode_plain; -extern const struct numa_mode numa_mode_emu; - -#endif /* __S390_NUMA_MODE_H */ diff --git a/arch/s390/numa/toptree.c b/arch/s390/numa/toptree.c deleted file mode 100644 index 71a608cd4f61..000000000000 --- a/arch/s390/numa/toptree.c +++ /dev/null @@ -1,351 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 -/* - * NUMA support for s390 - * - * A tree structure used for machine topology mangling - * - * Copyright IBM Corp. 2015 - */ - -#include <linux/kernel.h> -#include <linux/memblock.h> -#include <linux/cpumask.h> -#include <linux/list.h> -#include <linux/list_sort.h> -#include <linux/slab.h> -#include <asm/numa.h> - -#include "toptree.h" - -/** - * toptree_alloc - Allocate and initialize a new tree node. - * @level: The node's vertical level; level 0 contains the leaves. - * @id: ID number, explicitly not unique beyond scope of node's siblings - * - * Allocate a new tree node and initialize it. - * - * RETURNS: - * Pointer to the new tree node or NULL on error - */ -struct toptree __ref *toptree_alloc(int level, int id) -{ - struct toptree *res; - - if (slab_is_available()) - res = kzalloc(sizeof(*res), GFP_KERNEL); - else - res = memblock_alloc(sizeof(*res), 8); - if (!res) - return res; - - INIT_LIST_HEAD(&res->children); - INIT_LIST_HEAD(&res->sibling); - cpumask_clear(&res->mask); - res->level = level; - res->id = id; - return res; -} - -/** - * toptree_remove - Remove a tree node from a tree - * @cand: Pointer to the node to remove - * - * The node is detached from its parent node. The parent node's - * masks will be updated to reflect the loss of the child. - */ -static void toptree_remove(struct toptree *cand) -{ - struct toptree *oldparent; - - list_del_init(&cand->sibling); - oldparent = cand->parent; - cand->parent = NULL; - toptree_update_mask(oldparent); -} - -/** - * toptree_free - discard a tree node - * @cand: Pointer to the tree node to discard - * - * Checks if @cand is attached to a parent node. Detaches it - * cleanly using toptree_remove. Possible children are freed - * recursively. In the end @cand itself is freed. - */ -void __ref toptree_free(struct toptree *cand) -{ - struct toptree *child, *tmp; - - if (cand->parent) - toptree_remove(cand); - toptree_for_each_child_safe(child, tmp, cand) - toptree_free(child); - if (slab_is_available()) - kfree(cand); - else - memblock_free_early((unsigned long)cand, sizeof(*cand)); -} - -/** - * toptree_update_mask - Update node bitmasks - * @cand: Pointer to a tree node - * - * The node's cpumask will be updated by combining all children's - * masks. Then toptree_update_mask is called recursively for the - * parent if applicable. - * - * NOTE: - * This must not be called on leaves. If called on a leaf, its - * CPU mask is cleared and lost. - */ -void toptree_update_mask(struct toptree *cand) -{ - struct toptree *child; - - cpumask_clear(&cand->mask); - list_for_each_entry(child, &cand->children, sibling) - cpumask_or(&cand->mask, &cand->mask, &child->mask); - if (cand->parent) - toptree_update_mask(cand->parent); -} - -/** - * toptree_insert - Insert a tree node into tree - * @cand: Pointer to the node to insert - * @target: Pointer to the node to which @cand will added as a child - * - * Insert a tree node into a tree. Masks w |