* [PATCHv2 0/2] HMAT memroy hotplug support
@ 2019-04-15 15:16 Keith Busch
2019-04-15 15:16 ` Keith Busch
` (2 more replies)
0 siblings, 3 replies; 8+ messages in thread
From: Keith Busch @ 2019-04-15 15:16 UTC (permalink / raw)
To: linux-kernel, linux-acpi, linux-mm
Cc: Rafael Wysocki, Dave Hansen, Dan Williams, Brice Goglin, Keith Busch
This is much the same as the original version, accept there's an initial
patch to add hotplug support for memory caching. The first version only
provided the attributes for locality and performance.
The second patch adds the hotplug support via memory notifier. The
difference from v1 is the added lock, ensuring onlining multiple regions
is single threaded to prevent duplicate hmat registration races.
Keith Busch (2):
hmat: Register memory-side cache after parsing
hmat: Register attributes for memory hot add
drivers/acpi/hmat/hmat.c | 108 +++++++++++++++++++++++++++++++++++------------
1 file changed, 82 insertions(+), 26 deletions(-)
--
2.14.4
^ permalink raw reply [flat|nested] 8+ messages in thread* [PATCHv2 0/2] HMAT memroy hotplug support 2019-04-15 15:16 [PATCHv2 0/2] HMAT memroy hotplug support Keith Busch @ 2019-04-15 15:16 ` Keith Busch 2019-04-15 15:16 ` [PATCHv2 1/2] hmat: Register memory-side cache after parsing Keith Busch 2019-04-15 15:16 ` [PATCHv2 2/2] hmat: Register attributes for memory hot add Keith Busch 2 siblings, 0 replies; 8+ messages in thread From: Keith Busch @ 2019-04-15 15:16 UTC (permalink / raw) To: linux-kernel, linux-acpi, linux-mm Cc: Rafael Wysocki, Dave Hansen, Dan Williams, Brice Goglin, Keith Busch This is much the same as the original version, accept there's an initial patch to add hotplug support for memory caching. The first version only provided the attributes for locality and performance. The second patch adds the hotplug support via memory notifier. The difference from v1 is the added lock, ensuring onlining multiple regions is single threaded to prevent duplicate hmat registration races. Keith Busch (2): hmat: Register memory-side cache after parsing hmat: Register attributes for memory hot add drivers/acpi/hmat/hmat.c | 108 +++++++++++++++++++++++++++++++++++------------ 1 file changed, 82 insertions(+), 26 deletions(-) -- 2.14.4 ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCHv2 1/2] hmat: Register memory-side cache after parsing 2019-04-15 15:16 [PATCHv2 0/2] HMAT memroy hotplug support Keith Busch 2019-04-15 15:16 ` Keith Busch @ 2019-04-15 15:16 ` Keith Busch 2019-04-15 15:16 ` Keith Busch 2019-04-15 15:16 ` [PATCHv2 2/2] hmat: Register attributes for memory hot add Keith Busch 2 siblings, 1 reply; 8+ messages in thread From: Keith Busch @ 2019-04-15 15:16 UTC (permalink / raw) To: linux-kernel, linux-acpi, linux-mm Cc: Rafael Wysocki, Dave Hansen, Dan Williams, Brice Goglin, Keith Busch Instead of registering the hmat cache attributes in line with parsing the table, save the attributes in the memory target and register them after parsing completes. This will make it easier to register the attributes later when hot add is supported. Signed-off-by: Keith Busch <keith.busch@intel.com> --- drivers/acpi/hmat/hmat.c | 48 +++++++++++++++++++++++++++++++++--------------- 1 file changed, 33 insertions(+), 15 deletions(-) diff --git a/drivers/acpi/hmat/hmat.c b/drivers/acpi/hmat/hmat.c index b7824a0309f7..bdb167c026ff 100644 --- a/drivers/acpi/hmat/hmat.c +++ b/drivers/acpi/hmat/hmat.c @@ -41,6 +41,7 @@ struct memory_target { unsigned int memory_pxm; unsigned int processor_pxm; struct node_hmem_attrs hmem_attrs; + struct node_cache_attrs cache_attrs; }; struct memory_initiator { @@ -314,7 +315,7 @@ static __init int hmat_parse_cache(union acpi_subtable_headers *header, const unsigned long end) { struct acpi_hmat_cache *cache = (void *)header; - struct node_cache_attrs cache_attrs; + struct memory_target *target; u32 attrs; if (cache->header.length < sizeof(*cache)) { @@ -328,37 +329,40 @@ static __init int hmat_parse_cache(union acpi_subtable_headers *header, cache->memory_PD, cache->cache_size, attrs, cache->number_of_SMBIOShandles); - cache_attrs.size = cache->cache_size; - cache_attrs.level = (attrs & ACPI_HMAT_CACHE_LEVEL) >> 4; - cache_attrs.line_size = (attrs & ACPI_HMAT_CACHE_LINE_SIZE) >> 16; + target = find_mem_target(cache->memory_PD); + if (!target) + return 0; + + target->cache_attrs.size = cache->cache_size; + target->cache_attrs.level = (attrs & ACPI_HMAT_CACHE_LEVEL) >> 4; + target->cache_attrs.line_size = (attrs & ACPI_HMAT_CACHE_LINE_SIZE) >> 16; switch ((attrs & ACPI_HMAT_CACHE_ASSOCIATIVITY) >> 8) { case ACPI_HMAT_CA_DIRECT_MAPPED: - cache_attrs.indexing = NODE_CACHE_DIRECT_MAP; + target->cache_attrs.indexing = NODE_CACHE_DIRECT_MAP; break; case ACPI_HMAT_CA_COMPLEX_CACHE_INDEXING: - cache_attrs.indexing = NODE_CACHE_INDEXED; + target->cache_attrs.indexing = NODE_CACHE_INDEXED; break; case ACPI_HMAT_CA_NONE: default: - cache_attrs.indexing = NODE_CACHE_OTHER; + target->cache_attrs.indexing = NODE_CACHE_OTHER; break; } switch ((attrs & ACPI_HMAT_WRITE_POLICY) >> 12) { case ACPI_HMAT_CP_WB: - cache_attrs.write_policy = NODE_CACHE_WRITE_BACK; + target->cache_attrs.write_policy = NODE_CACHE_WRITE_BACK; break; case ACPI_HMAT_CP_WT: - cache_attrs.write_policy = NODE_CACHE_WRITE_THROUGH; + target->cache_attrs.write_policy = NODE_CACHE_WRITE_THROUGH; break; case ACPI_HMAT_CP_NONE: default: - cache_attrs.write_policy = NODE_CACHE_WRITE_OTHER; + target->cache_attrs.write_policy = NODE_CACHE_WRITE_OTHER; break; } - node_add_cache(pxm_to_node(cache->memory_PD), &cache_attrs); return 0; } @@ -577,20 +581,34 @@ static __init void hmat_register_target_initiators(struct memory_target *target) } } +static __init void hmat_register_target_cache(struct memory_target *target) +{ + unsigned mem_nid = pxm_to_node(target->memory_pxm); + node_add_cache(mem_nid, &target->cache_attrs); +} + static __init void hmat_register_target_perf(struct memory_target *target) { unsigned mem_nid = pxm_to_node(target->memory_pxm); node_set_perf_attrs(mem_nid, &target->hmem_attrs, 0); } +static __init void hmat_register_target(struct memory_target *target) +{ + if (!node_online(pxm_to_node(target->memory_pxm))) + return; + + hmat_register_target_initiators(target); + hmat_register_target_cache(target); + hmat_register_target_perf(target); +} + static __init void hmat_register_targets(void) { struct memory_target *target; - list_for_each_entry(target, &targets, node) { - hmat_register_target_initiators(target); - hmat_register_target_perf(target); - } + list_for_each_entry(target, &targets, node) + hmat_register_target(target); } static __init void hmat_free_structures(void) -- 2.14.4 ^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCHv2 1/2] hmat: Register memory-side cache after parsing 2019-04-15 15:16 ` [PATCHv2 1/2] hmat: Register memory-side cache after parsing Keith Busch @ 2019-04-15 15:16 ` Keith Busch 0 siblings, 0 replies; 8+ messages in thread From: Keith Busch @ 2019-04-15 15:16 UTC (permalink / raw) To: linux-kernel, linux-acpi, linux-mm Cc: Rafael Wysocki, Dave Hansen, Dan Williams, Brice Goglin, Keith Busch Instead of registering the hmat cache attributes in line with parsing the table, save the attributes in the memory target and register them after parsing completes. This will make it easier to register the attributes later when hot add is supported. Signed-off-by: Keith Busch <keith.busch@intel.com> --- drivers/acpi/hmat/hmat.c | 48 +++++++++++++++++++++++++++++++++--------------- 1 file changed, 33 insertions(+), 15 deletions(-) diff --git a/drivers/acpi/hmat/hmat.c b/drivers/acpi/hmat/hmat.c index b7824a0309f7..bdb167c026ff 100644 --- a/drivers/acpi/hmat/hmat.c +++ b/drivers/acpi/hmat/hmat.c @@ -41,6 +41,7 @@ struct memory_target { unsigned int memory_pxm; unsigned int processor_pxm; struct node_hmem_attrs hmem_attrs; + struct node_cache_attrs cache_attrs; }; struct memory_initiator { @@ -314,7 +315,7 @@ static __init int hmat_parse_cache(union acpi_subtable_headers *header, const unsigned long end) { struct acpi_hmat_cache *cache = (void *)header; - struct node_cache_attrs cache_attrs; + struct memory_target *target; u32 attrs; if (cache->header.length < sizeof(*cache)) { @@ -328,37 +329,40 @@ static __init int hmat_parse_cache(union acpi_subtable_headers *header, cache->memory_PD, cache->cache_size, attrs, cache->number_of_SMBIOShandles); - cache_attrs.size = cache->cache_size; - cache_attrs.level = (attrs & ACPI_HMAT_CACHE_LEVEL) >> 4; - cache_attrs.line_size = (attrs & ACPI_HMAT_CACHE_LINE_SIZE) >> 16; + target = find_mem_target(cache->memory_PD); + if (!target) + return 0; + + target->cache_attrs.size = cache->cache_size; + target->cache_attrs.level = (attrs & ACPI_HMAT_CACHE_LEVEL) >> 4; + target->cache_attrs.line_size = (attrs & ACPI_HMAT_CACHE_LINE_SIZE) >> 16; switch ((attrs & ACPI_HMAT_CACHE_ASSOCIATIVITY) >> 8) { case ACPI_HMAT_CA_DIRECT_MAPPED: - cache_attrs.indexing = NODE_CACHE_DIRECT_MAP; + target->cache_attrs.indexing = NODE_CACHE_DIRECT_MAP; break; case ACPI_HMAT_CA_COMPLEX_CACHE_INDEXING: - cache_attrs.indexing = NODE_CACHE_INDEXED; + target->cache_attrs.indexing = NODE_CACHE_INDEXED; break; case ACPI_HMAT_CA_NONE: default: - cache_attrs.indexing = NODE_CACHE_OTHER; + target->cache_attrs.indexing = NODE_CACHE_OTHER; break; } switch ((attrs & ACPI_HMAT_WRITE_POLICY) >> 12) { case ACPI_HMAT_CP_WB: - cache_attrs.write_policy = NODE_CACHE_WRITE_BACK; + target->cache_attrs.write_policy = NODE_CACHE_WRITE_BACK; break; case ACPI_HMAT_CP_WT: - cache_attrs.write_policy = NODE_CACHE_WRITE_THROUGH; + target->cache_attrs.write_policy = NODE_CACHE_WRITE_THROUGH; break; case ACPI_HMAT_CP_NONE: default: - cache_attrs.write_policy = NODE_CACHE_WRITE_OTHER; + target->cache_attrs.write_policy = NODE_CACHE_WRITE_OTHER; break; } - node_add_cache(pxm_to_node(cache->memory_PD), &cache_attrs); return 0; } @@ -577,20 +581,34 @@ static __init void hmat_register_target_initiators(struct memory_target *target) } } +static __init void hmat_register_target_cache(struct memory_target *target) +{ + unsigned mem_nid = pxm_to_node(target->memory_pxm); + node_add_cache(mem_nid, &target->cache_attrs); +} + static __init void hmat_register_target_perf(struct memory_target *target) { unsigned mem_nid = pxm_to_node(target->memory_pxm); node_set_perf_attrs(mem_nid, &target->hmem_attrs, 0); } +static __init void hmat_register_target(struct memory_target *target) +{ + if (!node_online(pxm_to_node(target->memory_pxm))) + return; + + hmat_register_target_initiators(target); + hmat_register_target_cache(target); + hmat_register_target_perf(target); +} + static __init void hmat_register_targets(void) { struct memory_target *target; - list_for_each_entry(target, &targets, node) { - hmat_register_target_initiators(target); - hmat_register_target_perf(target); - } + list_for_each_entry(target, &targets, node) + hmat_register_target(target); } static __init void hmat_free_structures(void) -- 2.14.4 ^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCHv2 2/2] hmat: Register attributes for memory hot add 2019-04-15 15:16 [PATCHv2 0/2] HMAT memroy hotplug support Keith Busch 2019-04-15 15:16 ` Keith Busch 2019-04-15 15:16 ` [PATCHv2 1/2] hmat: Register memory-side cache after parsing Keith Busch @ 2019-04-15 15:16 ` Keith Busch 2019-04-15 15:16 ` Keith Busch [not found] ` <9f130b73-e5ae-0529-69a1-28bd2ca29581@inria.fr> 2 siblings, 2 replies; 8+ messages in thread From: Keith Busch @ 2019-04-15 15:16 UTC (permalink / raw) To: linux-kernel, linux-acpi, linux-mm Cc: Rafael Wysocki, Dave Hansen, Dan Williams, Brice Goglin, Keith Busch Some memory nodes described in HMAT may not be online at the time the we parse the subtables. Should the node be set to online later, as can happen when using PMEM as RAM after boot, the nodes will be missing their initiator links and performance attributes. Register a memory notifier callback and register the memory attributes the first time its node is brought online if it wasn't registered, ensuring a node's attributes may be registered only once. Reported-by: Brice Goglin <Brice.Goglin@inria.fr> Signed-off-by: Keith Busch <keith.busch@intel.com> --- drivers/acpi/hmat/hmat.c | 72 ++++++++++++++++++++++++++++++++++++------------ 1 file changed, 55 insertions(+), 17 deletions(-) diff --git a/drivers/acpi/hmat/hmat.c b/drivers/acpi/hmat/hmat.c index bdb167c026ff..4fcfad6c2181 100644 --- a/drivers/acpi/hmat/hmat.c +++ b/drivers/acpi/hmat/hmat.c @@ -14,14 +14,18 @@ #include <linux/init.h> #include <linux/list.h> #include <linux/list_sort.h> +#include <linux/memory.h> +#include <linux/mutex.h> #include <linux/node.h> #include <linux/sysfs.h> -static __initdata u8 hmat_revision; +static u8 hmat_revision; -static __initdata LIST_HEAD(targets); -static __initdata LIST_HEAD(initiators); -static __initdata LIST_HEAD(localities); +static LIST_HEAD(targets); +static LIST_HEAD(initiators); +static LIST_HEAD(localities); + +static DEFINE_MUTEX(target_lock); /* * The defined enum order is used to prioritize attributes to break ties when @@ -42,6 +46,7 @@ struct memory_target { unsigned int processor_pxm; struct node_hmem_attrs hmem_attrs; struct node_cache_attrs cache_attrs; + bool registered; }; struct memory_initiator { @@ -54,7 +59,7 @@ struct memory_locality { struct acpi_hmat_locality *hmat_loc; }; -static __init struct memory_initiator *find_mem_initiator(unsigned int cpu_pxm) +static struct memory_initiator *find_mem_initiator(unsigned int cpu_pxm) { struct memory_initiator *initiator; @@ -64,7 +69,7 @@ static __init struct memory_initiator *find_mem_initiator(unsigned int cpu_pxm) return NULL; } -static __init struct memory_target *find_mem_target(unsigned int mem_pxm) +static struct memory_target *find_mem_target(unsigned int mem_pxm) { struct memory_target *target; @@ -149,7 +154,7 @@ static __init const char *hmat_data_type_suffix(u8 type) } } -static __init u32 hmat_normalize(u16 entry, u64 base, u8 type) +static u32 hmat_normalize(u16 entry, u64 base, u8 type) { u32 value; @@ -184,7 +189,7 @@ static __init u32 hmat_normalize(u16 entry, u64 base, u8 type) return value; } -static __init void hmat_update_target_access(struct memory_target *target, +static void hmat_update_target_access(struct memory_target *target, u8 type, u32 value) { switch (type) { @@ -439,7 +444,7 @@ static __init int srat_parse_mem_affinity(union acpi_subtable_headers *header, return 0; } -static __init u32 hmat_initiator_perf(struct memory_target *target, +static u32 hmat_initiator_perf(struct memory_target *target, struct memory_initiator *initiator, struct acpi_hmat_locality *hmat_loc) { @@ -477,7 +482,7 @@ static __init u32 hmat_initiator_perf(struct memory_target *target, hmat_loc->data_type); } -static __init bool hmat_update_best(u8 type, u32 value, u32 *best) +static bool hmat_update_best(u8 type, u32 value, u32 *best) { bool updated = false; @@ -521,7 +526,7 @@ static int initiator_cmp(void *priv, struct list_head *a, struct list_head *b) return ia->processor_pxm - ib->processor_pxm; } -static __init void hmat_register_target_initiators(struct memory_target *target) +static void hmat_register_target_initiators(struct memory_target *target) { static DECLARE_BITMAP(p_nodes, MAX_NUMNODES); struct memory_initiator *initiator; @@ -581,13 +586,13 @@ static __init void hmat_register_target_initiators(struct memory_target *target) } } -static __init void hmat_register_target_cache(struct memory_target *target) +static void hmat_register_target_cache(struct memory_target *target) { unsigned mem_nid = pxm_to_node(target->memory_pxm); node_add_cache(mem_nid, &target->cache_attrs); } -static __init void hmat_register_target_perf(struct memory_target *target) +static void hmat_register_target_perf(struct memory_target *target) { unsigned mem_nid = pxm_to_node(target->memory_pxm); node_set_perf_attrs(mem_nid, &target->hmem_attrs, 0); @@ -598,12 +603,17 @@ static __init void hmat_register_target(struct memory_target *target) if (!node_online(pxm_to_node(target->memory_pxm))) return; - hmat_register_target_initiators(target); - hmat_register_target_cache(target); - hmat_register_target_perf(target); + mutex_lock(&target_lock); + if (!target->registered) { + hmat_register_target_initiators(target); + hmat_register_target_cache(target); + hmat_register_target_perf(target); + target->registered = true; + } + mutex_unlock(&target_lock); } -static __init void hmat_register_targets(void) +static void hmat_register_targets(void) { struct memory_target *target; @@ -611,6 +621,30 @@ static __init void hmat_register_targets(void) hmat_register_target(target); } +static int hmat_callback(struct notifier_block *self, + unsigned long action, void *arg) +{ + struct memory_target *target; + struct memory_notify *mnb = arg; + int pxm, nid = mnb->status_change_nid; + + if (nid == NUMA_NO_NODE || action != MEM_ONLINE) + return NOTIFY_OK; + + pxm = node_to_pxm(nid); + target = find_mem_target(pxm); + if (!target) + return NOTIFY_OK; + + hmat_register_target(target); + return NOTIFY_OK; +} + +static struct notifier_block hmat_callback_nb = { + .notifier_call = hmat_callback, + .priority = 2, +}; + static __init void hmat_free_structures(void) { struct memory_target *target, *tnext; @@ -676,6 +710,10 @@ static __init int hmat_init(void) } } hmat_register_targets(); + + /* Keep the table and structures if the notifier may use them */ + if (!register_hotmemory_notifier(&hmat_callback_nb)) + return 0; out_put: hmat_free_structures(); acpi_put_table(tbl); -- 2.14.4 ^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCHv2 2/2] hmat: Register attributes for memory hot add 2019-04-15 15:16 ` [PATCHv2 2/2] hmat: Register attributes for memory hot add Keith Busch @ 2019-04-15 15:16 ` Keith Busch [not found] ` <9f130b73-e5ae-0529-69a1-28bd2ca29581@inria.fr> 1 sibling, 0 replies; 8+ messages in thread From: Keith Busch @ 2019-04-15 15:16 UTC (permalink / raw) To: linux-kernel, linux-acpi, linux-mm Cc: Rafael Wysocki, Dave Hansen, Dan Williams, Brice Goglin, Keith Busch Some memory nodes described in HMAT may not be online at the time the we parse the subtables. Should the node be set to online later, as can happen when using PMEM as RAM after boot, the nodes will be missing their initiator links and performance attributes. Register a memory notifier callback and register the memory attributes the first time its node is brought online if it wasn't registered, ensuring a node's attributes may be registered only once. Reported-by: Brice Goglin <Brice.Goglin@inria.fr> Signed-off-by: Keith Busch <keith.busch@intel.com> --- drivers/acpi/hmat/hmat.c | 72 ++++++++++++++++++++++++++++++++++++------------ 1 file changed, 55 insertions(+), 17 deletions(-) diff --git a/drivers/acpi/hmat/hmat.c b/drivers/acpi/hmat/hmat.c index bdb167c026ff..4fcfad6c2181 100644 --- a/drivers/acpi/hmat/hmat.c +++ b/drivers/acpi/hmat/hmat.c @@ -14,14 +14,18 @@ #include <linux/init.h> #include <linux/list.h> #include <linux/list_sort.h> +#include <linux/memory.h> +#include <linux/mutex.h> #include <linux/node.h> #include <linux/sysfs.h> -static __initdata u8 hmat_revision; +static u8 hmat_revision; -static __initdata LIST_HEAD(targets); -static __initdata LIST_HEAD(initiators); -static __initdata LIST_HEAD(localities); +static LIST_HEAD(targets); +static LIST_HEAD(initiators); +static LIST_HEAD(localities); + +static DEFINE_MUTEX(target_lock); /* * The defined enum order is used to prioritize attributes to break ties when @@ -42,6 +46,7 @@ struct memory_target { unsigned int processor_pxm; struct node_hmem_attrs hmem_attrs; struct node_cache_attrs cache_attrs; + bool registered; }; struct memory_initiator { @@ -54,7 +59,7 @@ struct memory_locality { struct acpi_hmat_locality *hmat_loc; }; -static __init struct memory_initiator *find_mem_initiator(unsigned int cpu_pxm) +static struct memory_initiator *find_mem_initiator(unsigned int cpu_pxm) { struct memory_initiator *initiator; @@ -64,7 +69,7 @@ static __init struct memory_initiator *find_mem_initiator(unsigned int cpu_pxm) return NULL; } -static __init struct memory_target *find_mem_target(unsigned int mem_pxm) +static struct memory_target *find_mem_target(unsigned int mem_pxm) { struct memory_target *target; @@ -149,7 +154,7 @@ static __init const char *hmat_data_type_suffix(u8 type) } } -static __init u32 hmat_normalize(u16 entry, u64 base, u8 type) +static u32 hmat_normalize(u16 entry, u64 base, u8 type) { u32 value; @@ -184,7 +189,7 @@ static __init u32 hmat_normalize(u16 entry, u64 base, u8 type) return value; } -static __init void hmat_update_target_access(struct memory_target *target, +static void hmat_update_target_access(struct memory_target *target, u8 type, u32 value) { switch (type) { @@ -439,7 +444,7 @@ static __init int srat_parse_mem_affinity(union acpi_subtable_headers *header, return 0; } -static __init u32 hmat_initiator_perf(struct memory_target *target, +static u32 hmat_initiator_perf(struct memory_target *target, struct memory_initiator *initiator, struct acpi_hmat_locality *hmat_loc) { @@ -477,7 +482,7 @@ static __init u32 hmat_initiator_perf(struct memory_target *target, hmat_loc->data_type); } -static __init bool hmat_update_best(u8 type, u32 value, u32 *best) +static bool hmat_update_best(u8 type, u32 value, u32 *best) { bool updated = false; @@ -521,7 +526,7 @@ static int initiator_cmp(void *priv, struct list_head *a, struct list_head *b) return ia->processor_pxm - ib->processor_pxm; } -static __init void hmat_register_target_initiators(struct memory_target *target) +static void hmat_register_target_initiators(struct memory_target *target) { static DECLARE_BITMAP(p_nodes, MAX_NUMNODES); struct memory_initiator *initiator; @@ -581,13 +586,13 @@ static __init void hmat_register_target_initiators(struct memory_target *target) } } -static __init void hmat_register_target_cache(struct memory_target *target) +static void hmat_register_target_cache(struct memory_target *target) { unsigned mem_nid = pxm_to_node(target->memory_pxm); node_add_cache(mem_nid, &target->cache_attrs); } -static __init void hmat_register_target_perf(struct memory_target *target) +static void hmat_register_target_perf(struct memory_target *target) { unsigned mem_nid = pxm_to_node(target->memory_pxm); node_set_perf_attrs(mem_nid, &target->hmem_attrs, 0); @@ -598,12 +603,17 @@ static __init void hmat_register_target(struct memory_target *target) if (!node_online(pxm_to_node(target->memory_pxm))) return; - hmat_register_target_initiators(target); - hmat_register_target_cache(target); - hmat_register_target_perf(target); + mutex_lock(&target_lock); + if (!target->registered) { + hmat_register_target_initiators(target); + hmat_register_target_cache(target); + hmat_register_target_perf(target); + target->registered = true; + } + mutex_unlock(&target_lock); } -static __init void hmat_register_targets(void) +static void hmat_register_targets(void) { struct memory_target *target; @@ -611,6 +621,30 @@ static __init void hmat_register_targets(void) hmat_register_target(target); } +static int hmat_callback(struct notifier_block *self, + unsigned long action, void *arg) +{ + struct memory_target *target; + struct memory_notify *mnb = arg; + int pxm, nid = mnb->status_change_nid; + + if (nid == NUMA_NO_NODE || action != MEM_ONLINE) + return NOTIFY_OK; + + pxm = node_to_pxm(nid); + target = find_mem_target(pxm); + if (!target) + return NOTIFY_OK; + + hmat_register_target(target); + return NOTIFY_OK; +} + +static struct notifier_block hmat_callback_nb = { + .notifier_call = hmat_callback, + .priority = 2, +}; + static __init void hmat_free_structures(void) { struct memory_target *target, *tnext; @@ -676,6 +710,10 @@ static __init int hmat_init(void) } } hmat_register_targets(); + + /* Keep the table and structures if the notifier may use them */ + if (!register_hotmemory_notifier(&hmat_callback_nb)) + return 0; out_put: hmat_free_structures(); acpi_put_table(tbl); -- 2.14.4 ^ permalink raw reply related [flat|nested] 8+ messages in thread
[parent not found: <9f130b73-e5ae-0529-69a1-28bd2ca29581@inria.fr>]
* Re: [PATCHv2 2/2] hmat: Register attributes for memory hot add [not found] ` <9f130b73-e5ae-0529-69a1-28bd2ca29581@inria.fr> @ 2019-04-16 15:01 ` Keith Busch 2019-04-16 15:01 ` Keith Busch 0 siblings, 1 reply; 8+ messages in thread From: Keith Busch @ 2019-04-16 15:01 UTC (permalink / raw) To: Brice Goglin Cc: linux-kernel, linux-acpi, linux-mm, Rafael Wysocki, Dave Hansen, Dan Williams On Tue, Apr 16, 2019 at 04:55:21PM +0200, Brice Goglin wrote: > Hello Keith > > Several issues: > > * We always get a memory_side_cache, even if nothing was found in ACPI. > You should at least ignore the cache if size==0? > > * Your code seems to only work with a single level of cache, since > there's a single cache_attrs entry in each target structure. > > * I was getting a section mismatch warning and a crash on PMEM node > hotplug until I applied the patch below. > > WARNING: vmlinux.o(.text+0x47d3f7): Section mismatch in reference from the function hmat_callback() to the function .init.text:hmat_register_target() > The function hmat_callback() references > the function __init hmat_register_target(). > This is often because hmat_callback lacks a __init > annotation or the annotation of hmat_register_target is wrong. > > Thanks > > Brice Oh, thanks for the notice. I'll add multi-level and no-caches into my test, as I had it fixed to one. Will need to respin this series. ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCHv2 2/2] hmat: Register attributes for memory hot add 2019-04-16 15:01 ` Keith Busch @ 2019-04-16 15:01 ` Keith Busch 0 siblings, 0 replies; 8+ messages in thread From: Keith Busch @ 2019-04-16 15:01 UTC (permalink / raw) To: Brice Goglin Cc: linux-kernel, linux-acpi, linux-mm, Rafael Wysocki, Dave Hansen, Dan Williams On Tue, Apr 16, 2019 at 04:55:21PM +0200, Brice Goglin wrote: > Hello Keith > > Several issues: > > * We always get a memory_side_cache, even if nothing was found in ACPI. > You should at least ignore the cache if size==0? > > * Your code seems to only work with a single level of cache, since > there's a single cache_attrs entry in each target structure. > > * I was getting a section mismatch warning and a crash on PMEM node > hotplug until I applied the patch below. > > WARNING: vmlinux.o(.text+0x47d3f7): Section mismatch in reference from the function hmat_callback() to the function .init.text:hmat_register_target() > The function hmat_callback() references > the function __init hmat_register_target(). > This is often because hmat_callback lacks a __init > annotation or the annotation of hmat_register_target is wrong. > > Thanks > > Brice Oh, thanks for the notice. I'll add multi-level and no-caches into my test, as I had it fixed to one. Will need to respin this series. ^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2019-04-16 15:01 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-04-15 15:16 [PATCHv2 0/2] HMAT memroy hotplug support Keith Busch
2019-04-15 15:16 ` Keith Busch
2019-04-15 15:16 ` [PATCHv2 1/2] hmat: Register memory-side cache after parsing Keith Busch
2019-04-15 15:16 ` Keith Busch
2019-04-15 15:16 ` [PATCHv2 2/2] hmat: Register attributes for memory hot add Keith Busch
2019-04-15 15:16 ` Keith Busch
[not found] ` <9f130b73-e5ae-0529-69a1-28bd2ca29581@inria.fr>
2019-04-16 15:01 ` Keith Busch
2019-04-16 15:01 ` Keith Busch
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).