Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752050AbbLJSv5 (ORCPT ); Thu, 10 Dec 2015 13:51:57 -0500 Received: from mga11.intel.com ([192.55.52.93]:18039 "EHLO mga11.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751415AbbLJSvb (ORCPT ); Thu, 10 Dec 2015 13:51:31 -0500 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.20,409,1444719600"; d="scan'208";a="869036632" From: Tom Zanussi To: rostedt@goodmis.org Cc: masami.hiramatsu.pt@hitachi.com, namhyung@kernel.org, josh@joshtriplett.org, andi@firstfloor.org, mathieu.desnoyers@efficios.com, peterz@infradead.org, linux-kernel@vger.kernel.org, Tom Zanussi Subject: [PATCH v13 08/29] tracing: Add lock-free tracing_map Date: Thu, 10 Dec 2015 12:50:50 -0600 Message-Id: X-Mailer: git-send-email 1.9.3 In-Reply-To: References: In-Reply-To: References: Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 44025 Lines: 1425 Add tracing_map, a special-purpose lock-free map for tracing. tracing_map is designed to aggregate or 'sum' one or more values associated with a specific object of type tracing_map_elt, which is associated by the map to a given key. It provides various hooks allowing per-tracer customization and is separated out into a separate file in order to allow it to be shared between multiple tracers, but isn't meant to be generally used outside of that context. The tracing_map implementation was inspired by lock-free map algorithms originated by Dr. Cliff Click: http://www.azulsystems.com/blog/cliff/2007-03-26-non-blocking-hashtable http://www.azulsystems.com/events/javaone_2007/2007_LockFreeHash.pdf Signed-off-by: Tom Zanussi Tested-by: Masami Hiramatsu Reviewed-by: Namhyung Kim --- kernel/trace/Kconfig | 13 + kernel/trace/Makefile | 1 + kernel/trace/tracing_map.c | 1058 ++++++++++++++++++++++++++++++++++++++++++++ kernel/trace/tracing_map.h | 282 ++++++++++++ 4 files changed, 1354 insertions(+) create mode 100644 kernel/trace/tracing_map.c create mode 100644 kernel/trace/tracing_map.h diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig index e45db6b..48b7329 100644 --- a/kernel/trace/Kconfig +++ b/kernel/trace/Kconfig @@ -528,6 +528,19 @@ config MMIOTRACE See Documentation/trace/mmiotrace.txt. If you are not helping to develop drivers, say N. +config TRACING_MAP + bool + depends on ARCH_HAVE_NMI_SAFE_CMPXCHG + default n + help + tracing_map is a special-purpose lock-free map for tracing, + separated out as a stand-alone facility in order to allow it + to be shared between multiple tracers. It isn't meant to be + generally used outside of that context, and is normally + selected by tracers that use it. + + If in doubt, say N. + config MMIOTRACE_TEST tristate "Test module for mmiotrace" depends on MMIOTRACE && m diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile index 9b1044e..4255c40 100644 --- a/kernel/trace/Makefile +++ b/kernel/trace/Makefile @@ -31,6 +31,7 @@ obj-$(CONFIG_TRACING) += trace_output.o obj-$(CONFIG_TRACING) += trace_seq.o obj-$(CONFIG_TRACING) += trace_stat.o obj-$(CONFIG_TRACING) += trace_printk.o +obj-$(CONFIG_TRACING_MAP) += tracing_map.o obj-$(CONFIG_CONTEXT_SWITCH_TRACER) += trace_sched_switch.o obj-$(CONFIG_FUNCTION_TRACER) += trace_functions.o obj-$(CONFIG_IRQSOFF_TRACER) += trace_irqsoff.o diff --git a/kernel/trace/tracing_map.c b/kernel/trace/tracing_map.c new file mode 100644 index 0000000..dd6401d --- /dev/null +++ b/kernel/trace/tracing_map.c @@ -0,0 +1,1058 @@ +/* + * tracing_map - lock-free map for tracing + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * Copyright (C) 2015 Tom Zanussi + * + * tracing_map implementation inspired by lock-free map algorithms + * originated by Dr. Cliff Click: + * + * http://www.azulsystems.com/blog/cliff/2007-03-26-non-blocking-hashtable + * http://www.azulsystems.com/events/javaone_2007/2007_LockFreeHash.pdf + */ + +#include +#include +#include +#include + +#include "tracing_map.h" +#include "trace.h" + +/* + * NOTE: For a detailed description of the data structures used by + * these functions (such as tracing_map_elt) please see the overview + * of tracing_map data structures at the beginning of tracing_map.h. + */ + +/** + * tracing_map_update_sum - Add a value to a tracing_map_elt's sum field + * @elt: The tracing_map_elt + * @i: The index of the given sum associated with the tracing_map_elt + * @n: The value to add to the sum + * + * Add n to sum i associated with the specified tracing_map_elt + * instance. The index i is the index returned by the call to + * tracing_map_add_sum_field() when the tracing map was set up. + */ +void tracing_map_update_sum(struct tracing_map_elt *elt, unsigned int i, u64 n) +{ + atomic64_add(n, &elt->fields[i].sum); +} + +/** + * tracing_map_read_sum - Return the value of a tracing_map_elt's sum field + * @elt: The tracing_map_elt + * @i: The index of the given sum associated with the tracing_map_elt + * + * Retrieve the value of the sum i associated with the specified + * tracing_map_elt instance. The index i is the index returned by the + * call to tracing_map_add_sum_field() when the tracing map was set + * up. + * + * Return: The sum associated with field i for elt. + */ +u64 tracing_map_read_sum(struct tracing_map_elt *elt, unsigned int i) +{ + return (u64)atomic64_read(&elt->fields[i].sum); +} + +int tracing_map_cmp_string(void *val_a, void *val_b) +{ + char *a = val_a; + char *b = val_b; + + return strcmp(a, b); +} + +int tracing_map_cmp_none(void *val_a, void *val_b) +{ + return 0; +} + +static int tracing_map_cmp_atomic64(void *val_a, void *val_b) +{ + u64 a = atomic64_read((atomic64_t *)val_a); + u64 b = atomic64_read((atomic64_t *)val_b); + + return (a > b) ? 1 : ((a < b) ? -1 : 0); +} + +#define DEFINE_TRACING_MAP_CMP_FN(type) \ +static int tracing_map_cmp_##type(void *val_a, void *val_b) \ +{ \ + type a = *(type *)val_a; \ + type b = *(type *)val_b; \ + \ + return (a > b) ? 1 : ((a < b) ? -1 : 0); \ +} + +DEFINE_TRACING_MAP_CMP_FN(s64); +DEFINE_TRACING_MAP_CMP_FN(u64); +DEFINE_TRACING_MAP_CMP_FN(s32); +DEFINE_TRACING_MAP_CMP_FN(u32); +DEFINE_TRACING_MAP_CMP_FN(s16); +DEFINE_TRACING_MAP_CMP_FN(u16); +DEFINE_TRACING_MAP_CMP_FN(s8); +DEFINE_TRACING_MAP_CMP_FN(u8); + +tracing_map_cmp_fn_t tracing_map_cmp_num(int field_size, + int field_is_signed) +{ + tracing_map_cmp_fn_t fn = tracing_map_cmp_none; + + switch (field_size) { + case 8: + if (field_is_signed) + fn = tracing_map_cmp_s64; + else + fn = tracing_map_cmp_u64; + break; + case 4: + if (field_is_signed) + fn = tracing_map_cmp_s32; + else + fn = tracing_map_cmp_u32; + break; + case 2: + if (field_is_signed) + fn = tracing_map_cmp_s16; + else + fn = tracing_map_cmp_u16; + break; + case 1: + if (field_is_signed) + fn = tracing_map_cmp_s8; + else + fn = tracing_map_cmp_u8; + break; + } + + return fn; +} + +static int tracing_map_add_field(struct tracing_map *map, + tracing_map_cmp_fn_t cmp_fn) +{ + int ret = -EINVAL; + + if (map->n_fields < TRACING_MAP_FIELDS_MAX) { + ret = map->n_fields; + map->fields[map->n_fields++].cmp_fn = cmp_fn; + } + + return ret; +} + +/** + * tracing_map_add_sum_field - Add a field describing a tracing_map sum + * @map: The tracing_map + * + * Add a sum field to the key and return the index identifying it in + * the map and associated tracing_map_elts. This is the index used + * for instance to update a sum for a particular tracing_map_elt using + * tracing_map_update_sum() or reading it via tracing_map_read_sum(). + * + * Return: The index identifying the field in the map and associated + * tracing_map_elts. + */ +int tracing_map_add_sum_field(struct tracing_map *map) +{ + return tracing_map_add_field(map, tracing_map_cmp_atomic64); +} + +/** + * tracing_map_add_key_field - Add a field describing a tracing_map key + * @map: The tracing_map + * @offset: The offset within the key + * @cmp_fn: The comparison function that will be used to sort on the key + * + * Let the map know there is a key and that if it's used as a sort key + * to use cmp_fn. + * + * A key can be a subset of a compound key; for that purpose, the + * offset param is used to describe where within the the compound key + * the key referenced by this key field resides. + * + * Return: The index identifying the field in the map and associated + * tracing_map_elts. + */ +int tracing_map_add_key_field(struct tracing_map *map, + unsigned int offset, + tracing_map_cmp_fn_t cmp_fn) + +{ + int idx = tracing_map_add_field(map, cmp_fn); + + if (idx < 0) + return idx; + + map->fields[idx].offset = offset; + + map->key_idx[map->n_keys++] = idx; + + return idx; +} + +void tracing_map_array_clear(struct tracing_map_array *a) +{ + unsigned int i; + + if (!a->pages) + return; + + for (i = 0; i < a->n_pages; i++) + memset(a->pages[i], 0, PAGE_SIZE); +} + +void tracing_map_array_free(struct tracing_map_array *a) +{ + unsigned int i; + + if (!a) + return; + + if (!a->pages) { + kfree(a); + return; + } + + for (i = 0; i < a->n_pages; i++) { + if (!a->pages[i]) + break; + free_page((unsigned long)a->pages[i]); + } +} + +struct tracing_map_array *tracing_map_array_alloc(unsigned int n_elts, + unsigned int entry_size) +{ + struct tracing_map_array *a; + unsigned int i; + + a = kzalloc(sizeof(*a), GFP_KERNEL); + if (!a) + return NULL; + + a->entry_size_shift = fls(roundup_pow_of_two(entry_size) - 1); + a->entries_per_page = PAGE_SIZE / (1 << a->entry_size_shift); + a->n_pages = n_elts / a->entries_per_page; + if (!a->n_pages) + a->n_pages = 1; + a->entry_shift = fls(a->entries_per_page) - 1; + a->entry_mask = (1 << a->entry_shift) - 1; + + a->pages = kcalloc(a->n_pages, sizeof(void *), GFP_KERNEL); + if (!a->pages) + goto free; + + for (i = 0; i < a->n_pages; i++) { + a->pages[i] = (void *)get_zeroed_page(GFP_KERNEL); + if (!a->pages[i]) + goto free; + } + out: + return a; + free: + tracing_map_array_free(a); + a = NULL; + + goto out; +} + +static void tracing_map_elt_clear(struct tracing_map_elt *elt) +{ + unsigned i; + + for (i = 0; i < elt->map->n_fields; i++) + if (elt->fields[i].cmp_fn == tracing_map_cmp_atomic64) + atomic64_set(&elt->fields[i].sum, 0); + + if (elt->map->ops && elt->map->ops->elt_clear) + elt->map->ops->elt_clear(elt); +} + +static void tracing_map_elt_init_fields(struct tracing_map_elt *elt) +{ + unsigned int i; + + tracing_map_elt_clear(elt); + + for (i = 0; i < elt->map->n_fields; i++) { + elt->fields[i].cmp_fn = elt->map->fields[i].cmp_fn; + + if (elt->fields[i].cmp_fn != tracing_map_cmp_atomic64) + elt->fields[i].offset = elt->map->fields[i].offset; + } +} + +static void tracing_map_elt_free(struct tracing_map_elt *elt) +{ + if (!elt) + return; + + if (elt->map->ops && elt->map->ops->elt_free) + elt->map->ops->elt_free(elt); + kfree(elt->fields); + kfree(elt->key); + kfree(elt); +} + +static struct tracing_map_elt *tracing_map_elt_alloc(struct tracing_map *map) +{ + struct tracing_map_elt *elt; + int err = 0; + + elt = kzalloc(sizeof(*elt), GFP_KERNEL); + if (!elt) + return ERR_PTR(-ENOMEM); + + elt->map = map; + + elt->key = kzalloc(map->key_size, GFP_KERNEL); + if (!elt->key) { + err = -ENOMEM; + goto free; + } + + elt->fields = kcalloc(map->n_fields, sizeof(*elt->fields), GFP_KERNEL); + if (!elt->fields) { + err = -ENOMEM; + goto free; + } + + tracing_map_elt_init_fields(elt); + + if (map->ops && map->ops->elt_alloc) { + err = map->ops->elt_alloc(elt); + if (err) + goto free; + } + return elt; + free: + tracing_map_elt_free(elt); + + return ERR_PTR(err); +} + +static struct tracing_map_elt *get_free_elt(struct tracing_map *map) +{ + struct tracing_map_elt *elt = NULL; + int idx; + + idx = atomic_inc_return(&map->next_elt); + if (idx < map->max_elts) { + elt = *(TRACING_MAP_ELT(map->elts, idx)); + if (map->ops && map->ops->elt_init) + map->ops->elt_init(elt); + } + + return elt; +} + +static void tracing_map_free_elts(struct tracing_map *map) +{ + unsigned int i; + + if (!map->elts) + return; + + for (i = 0; i < map->max_elts; i++) + tracing_map_elt_free(*(TRACING_MAP_ELT(map->elts, i))); + + tracing_map_array_free(map->elts); +} + +static int tracing_map_alloc_elts(struct tracing_map *map) +{ + unsigned int i; + + map->elts = tracing_map_array_alloc(map->max_elts, + sizeof(struct tracing_map_elt *)); + if (!map->elts) + return -ENOMEM; + + for (i = 0; i < map->max_elts; i++) { + *(TRACING_MAP_ELT(map->elts, i)) = tracing_map_elt_alloc(map); + if (!(*(TRACING_MAP_ELT(map->elts, i)))) { + tracing_map_free_elts(map); + + return -ENOMEM; + } + } + + return 0; +} + +static inline bool keys_match(void *key, void *test_key, unsigned key_size) +{ + bool match = true; + + if (memcmp(key, test_key, key_size)) + match = false; + + return match; +} + +static inline struct tracing_map_elt * +__tracing_map_insert(struct tracing_map *map, void *key, bool lookup_only) +{ + u32 idx, key_hash, test_key; + struct tracing_map_entry *entry; + + key_hash = jhash(key, map->key_size, 0); + if (key_hash == 0) + key_hash = 1; + idx = key_hash >> (32 - (map->map_bits + 1)); + + while (1) { + idx &= (map->map_size - 1); + entry = TRACING_MAP_ENTRY(map->map, idx); + test_key = entry->key; + + if (test_key && test_key == key_hash && entry->val && + keys_match(key, entry->val->key, map->key_size)) { + atomic64_inc(&map->hits); + return entry->val; + } + + if (!test_key) { + if (lookup_only) + break; + + if (!cmpxchg(&entry->key, 0, key_hash)) { + struct tracing_map_elt *elt; + + elt = get_free_elt(map); + if (!elt) { + atomic64_inc(&map->drops); + entry->key = 0; + break; + } + + memcpy(elt->key, key, map->key_size); + entry->val = elt; + atomic64_inc(&map->hits); + + return entry->val; + } + } + + idx++; + } + + return NULL; +} + +/** + * tracing_map_insert - Insert key and/or retrieve val from a tracing_map + * @map: The tracing_map to insert into + * @key: The key to insert + * + * Inserts a key into a tracing_map and creates and returns a new + * tracing_map_elt for it, or if the key has already been inserted by + * a previous call, returns the tracing_map_elt already associated + * with it. When the map was created, the number of elements to be + * allocated for the map was specified (internally maintained as + * 'max_elts' in struct tracing_map), and that number of + * tracing_map_elts was created by tracing_map_init(). This is the + * pre-allocated pool of tracing_map_elts that tracing_map_insert() + * will allocate from when adding new keys. Once that pool is + * exhausted, tracing_map_insert() is useless and will return NULL to + * signal that state. There are two user-visible tracing_map + * variables, 'hits' and 'drops', which are updated by this function. + * Every time an element is either successfully inserted or retrieved, + * the 'hits' value is incrememented. Every time an element insertion + * fails, the 'drops' value is incremented. + * + * This is a lock-free tracing map insertion function implementing a + * modified form of Cliff Click's basic insertion algorithm. It + * requires the table size be a power of two. To prevent any + * possibility of an infinite loop we always make the internal table + * size double the size of the requested table size (max_elts * 2). + * Likewise, we never reuse a slot or resize or delete elements - when + * we've reached max_elts entries, we simply return NULL once we've + * run out of entries. Readers can at any point in time traverse the + * tracing map and safely access the key/val pairs. + * + * Return: the tracing_map_elt pointer val associated with the key. + * If this was a newly inserted key, the val will be a newly allocated + * and associated tracing_map_elt pointer val. If the key wasn't + * found and the pool of tracing_map_elts has been exhausted, NULL is + * returned and no further insertions will succeed. + */ +struct tracing_map_elt *tracing_map_insert(struct tracing_map *map, void *key) +{ + return __tracing_map_insert(map, key, false); +} + +/** + * tracing_map_lookup - Retrieve val from a tracing_map + * @map: The tracing_map to perform the lookup on + * @key: The key to look up + * + * Looks up key in tracing_map and if found returns the matching + * tracing_map_elt. This is a lock-free lookup; see + * tracing_map_insert() for details on tracing_map and how it works. + * Every time an element is retrieved, the 'hits' value is + * incrememented. There is one user-visible tracing_map variable, + * 'hits', which is updated by this function. Every time an element + * is successfully retrieved, the 'hits' value is incrememented. The + * 'drops' value is never updated by this function. + * + * Return: the tracing_map_elt pointer val associated with the key. + * If the key wasn't found, NULL is returned. + */ +struct tracing_map_elt *tracing_map_lookup(struct tracing_map *map, void *key) +{ + return __tracing_map_insert(map, key, true); +} + +/** + * tracing_map_destroy - Destroy a tracing_map + * @map: The tracing_map to destroy + * + * Frees a tracing_map along with its associated array of + * tracing_map_elts. + * + * Callers should make sure there are no readers or writers actively + * reading or inserting into the map before calling this. + */ +void tracing_map_destroy(struct tracing_map *map) +{ + if (!map) + return; + + tracing_map_free_elts(map); + + tracing_map_array_free(map->map); + kfree(map); +} + +/** + * tracing_map_clear - Clear a tracing_map + * @map: The tracing_map to clear + * + * Resets the tracing map to a cleared or initial state. The + * tracing_map_elts are all cleared, and the array of struct + * tracing_map_entry is reset to an initialized state. + * + * Callers should make sure there are no writers actively inserting + * into the map before calling this. + */ +void tracing_map_clear(struct tracing_map *map) +{ + unsigned int i; + + atomic_set(&map->next_elt, -1); + atomic64_set(&map->hits, 0); + atomic64_set(&map->drops, 0); + + tracing_map_array_clear(map->map); + + for (i = 0; i < map->max_elts; i++) + tracing_map_elt_clear(*(TRACING_MAP_ELT(map->elts, i))); +} + +static void set_sort_key(struct tracing_map *map, + struct tracing_map_sort_key *sort_key) +{ + map->sort_key = *sort_key; +} + +/** + * tracing_map_create - Create a lock-free map and element pool + * @map_bits: The size of the map (2 ** map_bits) + * @key_size: The size of the key for the map in bytes + * @ops: Optional client-defined tracing_map_ops instance + * @private_data: Client data associated with the map + * + * Creates and sets up a map to contain 2 ** map_bits number of + * elements (internally maintained as 'max_elts' in struct + * tracing_map). Before using, map fields should be added to the map + * with tracing_map_add_sum_field() and tracing_map_add_key_field(). + * tracing_map_init() should then be called to allocate the array of + * tracing_map_elts, in order to avoid allocating anything in the map + * insertion path. The user-specified map size reflects the maximum + * number of elements that can be contained in the table requested by + * the user - internally we double that in order to keep the table + * sparse and keep collisions manageable. + * + * A tracing_map is a special-purpose map designed to aggregate or + * 'sum' one or more values associated with a specific object of type + * tracing_map_elt, which is attached by the map to a given key. + * + * tracing_map_create() sets up the map itself, and provides + * operations for inserting tracing_map_elts, but doesn't allocate the + * tracing_map_elts themselves, or provide a means for describing the + * keys or sums associated with the tracing_map_elts. All + * tracing_map_elts for a given map have the same set of sums and + * keys, which are defined by the client using the functions + * tracing_map_add_key_field() and tracing_map_add_sum_field(). Once + * the fields are defined, the pool of elements allocated for the map + * can be created, which occurs when the client code calls + * tracing_map_init(). + * + * When tracing_map_init() returns, tracing_map_elt elements can be + * inserted into the map using tracing_map_insert(). When called, + * tracing_map_insert() grabs a free tracing_map_elt from the pool, or + * finds an existing match in the map and in either case returns it. + * The client can then use tracing_map_update_sum() and + * tracing_map_read_sum() to update or read a given sum field for the + * tracing_map_elt. + * + * The client can at any point retrieve and traverse the current set + * of inserted tracing_map_elts in a tracing_map, via + * tracing_map_sort_entries(). Sorting can be done on any field, + * including keys. + * + * See tracing_map.h for a description of tracing_map_ops. + * + * Return: the tracing_map pointer if successful, ERR_PTR if not. + */ +struct tracing_map *tracing_map_create(unsigned int map_bits, + unsigned int key_size, + const struct tracing_map_ops *ops, + void *private_data) +{ + struct tracing_map *map; + unsigned int i; + + if (map_bits < TRACING_MAP_BITS_MIN || + map_bits > TRACING_MAP_BITS_MAX) + return ERR_PTR(-EINVAL); + + map = kzalloc(sizeof(*map), GFP_KERNEL); + if (!map) + return ERR_PTR(-ENOMEM); + + map->map_bits = map_bits; + map->max_elts = (1 << map_bits); + atomic_set(&map->next_elt, -1); + + map->map_size = (1 << (map_bits + 1)); + map->ops = ops; + + map->private_data = private_data; + + map->map = tracing_map_array_alloc(map->map_size, + sizeof(struct tracing_map_entry)); + if (!map->map) + goto free; + + map->key_size = key_size; + for (i = 0; i < TRACING_MAP_KEYS_MAX; i++) + map->key_idx[i] = -1; + out: + return map; + free: + tracing_map_destroy(map); + map = ERR_PTR(-ENOMEM); + + goto out; +} + +/** + * tracing_map_init - Allocate and clear a map's tracing_map_elts + * @map: The tracing_map to initialize + * + * Allocates a clears a pool of tracing_map_elts equal to the + * user-specified size of 2 ** map_bits (internally maintained as + * 'max_elts' in struct tracing_map). Before using, the map fields + * should be added to the map with tracing_map_add_sum_field() and + * tracing_map_add_key_field(). tracing_map_init() should then be + * called to allocate the array of tracing_map_elts, in order to avoid + * allocating anything in the map insertion path. The user-specified + * map size reflects the max number of elements requested by the user + * - internally we double that in order to keep the table sparse and + * keep collisions manageable. + * + * See tracing_map.h for a description of tracing_map_ops. + * + * Return: the tracing_map pointer if successful, ERR_PTR if not. + */ +int tracing_map_init(struct tracing_map *map) +{ + int err; + + if (map->n_fields < 2) + return -EINVAL; /* need at least 1 key and 1 val */ + + err = tracing_map_alloc_elts(map); + if (err) + return err; + + tracing_map_clear(map); + + return err; +} + +static int cmp_entries_dup(const struct tracing_map_sort_entry **a, + const struct tracing_map_sort_entry **b) +{ + int ret = 0; + + if (memcmp((*a)->key, (*b)->key, (*a)->elt->map->key_size)) + ret = 1; + + return ret; +} + +static int cmp_entries_sum(const struct tracing_map_sort_entry **a, + const struct tracing_map_sort_entry **b) +{ + const struct tracing_map_elt *elt_a, *elt_b; + struct tracing_map_sort_key *sort_key; + struct tracing_map_field *field; + tracing_map_cmp_fn_t cmp_fn; + void *val_a, *val_b; + int ret = 0; + + elt_a = (*a)->elt; + elt_b = (*b)->elt; + + sort_key = &elt_a->map->sort_key; + + field = &elt_a->fields[sort_key->field_idx]; + cmp_fn = field->cmp_fn; + + val_a = &elt_a->fields[sort_key->field_idx].sum; + val_b = &elt_b->fields[sort_key->field_idx].sum; + + ret = cmp_fn(val_a, val_b); + if (sort_key->descending) + ret = -ret; + + return ret; +} + +static int cmp_entries_key(const struct tracing_map_sort_entry **a, + const struct tracing_map_sort_entry **b) +{ + const struct tracing_map_elt *elt_a, *elt_b; + struct tracing_map_sort_key *sort_key; + struct tracing_map_field *field; + tracing_map_cmp_fn_t cmp_fn; + void *val_a, *val_b; + int ret = 0; + + elt_a = (*a)->elt; + elt_b = (*b)->elt; + + sort_key = &elt_a->map->sort_key; + + field = &elt_a->fields[sort_key->field_idx]; + + cmp_fn = field->cmp_fn; + + val_a = elt_a->key + field->offset; + val_b = elt_b->key + field->offset; + + ret = cmp_fn(val_a, val_b); + if (sort_key->descending) + ret = -ret; + + return ret; +} + +static void destroy_sort_entry(struct tracing_map_sort_entry *entry) +{ + if (!entry) + return; + + if (entry->elt_copied) + tracing_map_elt_free(entry->elt); + + kfree(entry); +} + +/** + * tracing_map_destroy_sort_entries - Destroy an array of sort entries + * @entries: The entries to destroy + * @n_entries: The number of entries in the array + * + * Destroy the elements returned by a tracing_map_sort_entries() call. + */ +void tracing_map_destroy_sort_entries(struct tracing_map_sort_entry **entries, + unsigned int n_entries) +{ + unsigned int i; + + for (i = 0; i < n_entries; i++) + destroy_sort_entry(entries[i]); + + vfree(entries); +} + +static struct tracing_map_sort_entry * +create_sort_entry(void *key, struct tracing_map_elt *elt) +{ + struct tracing_map_sort_entry *sort_entry; + + sort_entry = kzalloc(sizeof(*sort_entry), GFP_KERNEL); + if (!sort_entry) + return NULL; + + sort_entry->key = key; + sort_entry->elt = elt; + + return sort_entry; +} + +static struct tracing_map_elt *copy_elt(struct tracing_map_elt *elt) +{ + struct tracing_map_elt *dup_elt; + unsigned int i; + + dup_elt = tracing_map_elt_alloc(elt->map); + if (!dup_elt) + return NULL; + + if (elt->map->ops && elt->map->ops->elt_copy) + elt->map->ops->elt_copy(dup_elt, elt); + + dup_elt->private_data = elt->private_data; + memcpy(dup_elt->key, elt->key, elt->map->key_size); + + for (i = 0; i < elt->map->n_fields; i++) { + atomic64_set(&dup_elt->fields[i].sum, + atomic64_read(&elt->fields[i].sum)); + dup_elt->fields[i].cmp_fn = elt->fields[i].cmp_fn; + } + + return dup_elt; +} + +static int merge_dup(struct tracing_map_sort_entry **sort_entries, + unsigned int target, unsigned int dup) +{ + struct tracing_map_elt *target_elt, *elt; + bool first_dup = (target - dup) == 1; + int i; + + if (first_dup) { + elt = sort_entries[target]->elt; + target_elt = copy_elt(elt); + if (!target_elt) + return -ENOMEM; + sort_entries[target]->elt = target_elt; + sort_entries[target]->elt_copied = true; + } else + target_elt = sort_entries[target]->elt; + + elt = sort_entries[dup]->elt; + + for (i = 0; i < elt->map->n_fields; i++) + atomic64_add(atomic64_read(&elt->fields[i].sum), + &target_elt->fields[i].sum); + + sort_entries[dup]->dup = true; + + return 0; +} + +static int merge_dups(struct tracing_map_sort_entry **sort_entries, + int n_entries, unsigned int key_size) +{ + unsigned int dups = 0, total_dups = 0; + int err, i, j; + void *key; + + if (n_entries < 2) + return total_dups; + + sort(sort_entries, n_entries, sizeof(struct tracing_map_sort_entry *), + (int (*)(const void *, const void *))cmp_entries_dup, NULL); + + key = sort_entries[0]->key; + for (i = 1; i < n_entries; i++) { + if (!memcmp(sort_entries[i]->key, key, key_size)) { + dups++; total_dups++; + err = merge_dup(sort_entries, i - dups, i); + if (err) + return err; + continue; + } + key = sort_entries[i]->key; + dups = 0; + } + + if (!total_dups) + return total_dups; + + for (i = 0, j = 0; i < n_entries; i++) { + if (!sort_entries[i]->dup) { + sort_entries[j] = sort_entries[i]; + if (j++ != i) + sort_entries[i] = NULL; + } else { + destroy_sort_entry(sort_entries[i]); + sort_entries[i] = NULL; + } + } + + return total_dups; +} + +static bool is_key(struct tracing_map *map, unsigned int field_idx) +{ + unsigned int i; + + for (i = 0; i < map->n_keys; i++) + if (map->key_idx[i] == field_idx) + return true; + return false; +} + +static void sort_secondary(struct tracing_map *map, + const struct tracing_map_sort_entry **entries, + unsigned int n_entries, + struct tracing_map_sort_key *primary_key, + struct tracing_map_sort_key *secondary_key) +{ + int (*primary_fn)(const struct tracing_map_sort_entry **, + const struct tracing_map_sort_entry **); + int (*secondary_fn)(const struct tracing_map_sort_entry **, + const struct tracing_map_sort_entry **); + unsigned i, start = 0, n_sub = 1; + + if (is_key(map, primary_key->field_idx)) + primary_fn = cmp_entries_key; + else + primary_fn = cmp_entries_sum; + + if (is_key(map, secondary_key->field_idx)) + secondary_fn = cmp_entries_key; + else + secondary_fn = cmp_entries_sum; + + for (i = 0; i < n_entries - 1; i++) { + const struct tracing_map_sort_entry **a = &entries[i]; + const struct tracing_map_sort_entry **b = &entries[i + 1]; + + if (primary_fn(a, b) == 0) { + n_sub++; + if (i < n_entries - 2) + continue; + } + + if (n_sub < 2) { + start = i + 1; + n_sub = 1; + continue; + } + + set_sort_key(map, secondary_key); + sort(&entries[start], n_sub, + sizeof(struct tracing_map_sort_entry *), + (int (*)(const void *, const void *))secondary_fn, NULL); + set_sort_key(map, primary_key); + + start = i + 1; + n_sub = 1; + } +} + +/** + * tracing_map_sort_entries - Sort the current set of tracing_map_elts in a map + * @map: The tracing_map + * @sort_key: The sort key to use for sorting + * @sort_entries: outval: pointer to allocated and sorted array of entries + * + * tracing_map_sort_entries() sorts the current set of entries in the + * map and returns the list of tracing_map_sort_entries containing + * them to the client in the sort_entries param. The client can + * access the struct tracing_map_elt element of interest directly as + * the 'elt' field of a returned struct tracing_map_sort_entry object. + * + * The sort_key has only two fields: idx and descending. 'idx' refers + * to the index of the field added via tracing_map_add_sum_field() or + * tracing_map_add_key_field() when the tracing_map was initialized. + * 'descending' is a flag that if set reverses the sort order, which + * by default is ascending. + * + * The client should not hold on to the returned array but should use + * it and call tracing_map_destroy_sort_entries() when done. + * + * Return: the number of sort_entries in the struct tracing_map_sort_entry + * array, negative on error + */ +int tracing_map_sort_entries(struct tracing_map *map, + struct tracing_map_sort_key *sort_keys, + unsigned int n_sort_keys, + struct tracing_map_sort_entry ***sort_entries) +{ + int (*cmp_entries_fn)(const struct tracing_map_sort_entry **, + const struct tracing_map_sort_entry **); + struct tracing_map_sort_entry *sort_entry, **entries; + int i, n_entries, ret; + + entries = vmalloc(map->max_elts * sizeof(sort_entry)); + if (!entries) + return -ENOMEM; + + for (i = 0, n_entries = 0; i < map->map_size; i++) { + struct tracing_map_entry *entry; + + entry = TRACING_MAP_ENTRY(map->map, i); + + if (!entry->key || !entry->val) + continue; + + entries[n_entries] = create_sort_entry(entry->val->key, + entry->val); + if (!entries[n_entries++]) { + ret = -ENOMEM; + goto free; + } + } + + if (n_entries == 0) { + ret = 0; + goto free; + } + + if (n_entries == 1) { + *sort_entries = entries; + return 1; + } + + ret = merge_dups(entries, n_entries, map->key_size); + if (ret < 0) + goto free; + n_entries -= ret; + + if (is_key(map, sort_keys[0].field_idx)) + cmp_entries_fn = cmp_entries_key; + else + cmp_entries_fn = cmp_entries_sum; + + set_sort_key(map, &sort_keys[0]); + + sort(entries, n_entries, sizeof(struct tracing_map_sort_entry *), + (int (*)(const void *, const void *))cmp_entries_fn, NULL); + + if (n_sort_keys > 1) + sort_secondary(map, + (const struct tracing_map_sort_entry **)entries, + n_entries, + &sort_keys[0], + &sort_keys[1]); + + *sort_entries = entries; + + return n_entries; + free: + tracing_map_destroy_sort_entries(entries, n_entries); + + return ret; +} diff --git a/kernel/trace/tracing_map.h b/kernel/trace/tracing_map.h new file mode 100644 index 0000000..7571825 --- /dev/null +++ b/kernel/trace/tracing_map.h @@ -0,0 +1,282 @@ +#ifndef __TRACING_MAP_H +#define __TRACING_MAP_H + +#define TRACING_MAP_BITS_DEFAULT 11 +#define TRACING_MAP_BITS_MAX 17 +#define TRACING_MAP_BITS_MIN 7 + +#define TRACING_MAP_FIELDS_MAX 4 +#define TRACING_MAP_KEYS_MAX 2 + +#define TRACING_MAP_SORT_KEYS_MAX 2 + +typedef int (*tracing_map_cmp_fn_t) (void *val_a, void *val_b); + +/* + * This is an overview of the tracing_map data structures and how they + * relate to the tracing_map API. The details of the algorithms + * aren't discussed here - this is just a general overview of the data + * structures and how they interact with the API. + * + * The central data structure of the tracing_map is an initially + * zeroed array of struct tracing_map_entry (stored in the map field + * of struct tracing_map). tracing_map_entry is a very simple data + * structure containing only two fields: a 32-bit unsigned 'key' + * variable and a pointer named 'val'. This array of struct + * tracing_map_entry is essentially a hash table which will be + * modified by a single function, tracing_map_insert(), but which can + * be traversed and read by a user at any time (though the user does + * this indirectly via an array of tracing_map_sort_entry - see the + * explanation of that data structure in the discussion of the + * sorting-related data structures below). + * + * The central function of the tracing_map API is + * tracing_map_insert(). tracing_map_insert() hashes the + * arbitrarily-sized key passed into it into a 32-bit unsigned key. + * It then uses this key, truncated to the array size, as an index + * into the array of tracing_map_entries. If the value of the 'key' + * field of the tracing_map_entry found at that location is 0, then + * that entry is considered to be free and can be claimed, by + * replacing the 0 in the 'key' field of the tracing_map_entry with + * the new 32-bit hashed key. Once claimed, that tracing_map_entry's + * 'val' field is then used to store a unique element which will be + * forever associated with that 32-bit hashed key in the + * tracing_map_entry. + * + * That unique element now in the tracing_map_entry's 'val' field is + * an instance of tracing_map_elt, where 'elt' in the latter part of + * that variable name is short for 'element'. The purpose of a + * tracing_map_elt is to hold values specific to the the particular + * 32-bit hashed key it's assocated with. Things such as the unique + * set of aggregated sums associated with the 32-bit hashed key, along + * with a copy of the full key associated with the entry, and which + * was used to produce the 32-bit hashed key. + * + * When tracing_map_create() is called to create the tracing map, the + * user specifies (indirectly via the map_bits param, the details are + * unimportant for this discussion) the maximum number of elements + * that the map can hold (stored in the max_elts field of struct + * tracing_map). This is the maximum possible number of + * tracing_map_entries in the tracing_map_entry array which can be + * 'claimed' as described in the above discussion, and therefore is + * also the maximum number of tracing_map_elts that can be associated + * with the tracing_map_entry array in the tracing_map. Because of + * the way the insertion algorithm works, the size of the allocated + * tracing_map_entry array is always twice the maximum number of + * elements (2 * max_elts). This value is stored in the map_size + * field of struct tracing_map. + * + * Because tracing_map_insert() needs to work from any context, + * including from within the memory allocation functions themselves, + * both the tracing_map_entry array and a pool of max_elts + * tracing_map_elts are pre-allocated before any call is made to + * tracing_map_insert(). + * + * The tracing_map_entry array is allocated as a single block by + * tracing_map_create(). + * + * Because the tracing_map_elts are much larger objects and can't + * generally be allocated together as a single large array without + * failure, they're allocated individually, by tracing_map_init(). + * + * The pool of tracing_map_elts are allocated by tracing_map_init() + * rather than by tracing_map_create() because at the time + * tracing_map_create() is called, there isn't enough information to + * create the tracing_map_elts. Specifically,the user first needs to + * tell the tracing_map implementation how many fields the + * tracing_map_elts contain, and which types of fields they are (key + * or sum). The user does this via the tracing_map_add_sum_field() + * and tracing_map_add_key_field() functions, following which the user + * calls tracing_map_init() to finish up the tracing map setup. The + * array holding the pointers which make up the pre-allocated pool of + * tracing_map_elts is allocated as a single block and is stored in + * the elts field of struct tracing_map. + * + * There is also a set of structures used for sorting that might + * benefit from some minimal explanation. + * + * struct tracing_map_sort_key is used to drive the sort at any given + * time. By 'any given time' we mean that a different + * tracing_map_sort_key will be used at different times depending on + * whether the sort currently being performed is a primary or a + * secondary sort. + * + * The sort key is very simple, consisting of the field index of the + * tracing_map_elt field to sort on (which the user saved when adding + * the field), and whether the sort should be done in an ascending or + * descending order. + * + * For the convenience of the sorting code, a tracing_map_sort_entry + * is created for each tracing_map_elt, again individually allocated + * to avoid failures that might be expected if allocated as a single + * large array of struct tracing_map_sort_entry. + * tracing_map_sort_entry instances are the objects expected by the + * various internal sorting functions, and are also what the user + * ultimately receives after calling tracing_map_sort_entries(). + * Because it doesn't make sense for users to access an unordered and + * sparsely populated tracing_map directly, the + * tracing_map_sort_entries() function is provided so that users can + * retrieve a sorted list of all existing elements. In addition to + * the associated tracing_map_elt 'elt' field contained within the + * tracing_map_sort_entry, which is the object of interest to the + * user, tracing_map_sort_entry objects contain a number of additional + * fields which are used for caching and internal purposes and can + * safely be ignored. +*/ + +struct tracing_map_field { + tracing_map_cmp_fn_t cmp_fn; + union { + atomic64_t sum; + unsigned int offset; + }; +}; + +struct tracing_map_elt { + struct tracing_map *map; + struct tracing_map_field *fields; + void *key; + void *private_data; +}; + +struct tracing_map_entry { + u32 key; + struct tracing_map_elt *val; +}; + +struct tracing_map_sort_key { + unsigned int field_idx; + bool descending; +}; + +struct tracing_map_sort_entry { + void *key; + struct tracing_map_elt *elt; + bool elt_copied; + bool dup; +}; + +struct tracing_map_array { + unsigned int entries_per_page; + unsigned int entry_size_shift; + unsigned int entry_shift; + unsigned int entry_mask; + unsigned int n_pages; + void **pages; +}; + +#define TRACING_MAP_ARRAY_ELT(array, idx) \ + (array->pages[idx >> array->entry_shift] + \ + ((idx & array->entry_mask) << array->entry_size_shift)) + +#define TRACING_MAP_ENTRY(array, idx) \ + ((struct tracing_map_entry *)TRACING_MAP_ARRAY_ELT(array, idx)) + +#define TRACING_MAP_ELT(array, idx) \ + ((struct tracing_map_elt **)TRACING_MAP_ARRAY_ELT(array, idx)) + +struct tracing_map { + unsigned int key_size; + unsigned int map_bits; + unsigned int map_size; + unsigned int max_elts; + atomic_t next_elt; + struct tracing_map_array *elts; + struct tracing_map_array *map; + const struct tracing_map_ops *ops; + void *private_data; + struct tracing_map_field fields[TRACING_MAP_FIELDS_MAX]; + unsigned int n_fields; + int key_idx[TRACING_MAP_KEYS_MAX]; + unsigned int n_keys; + struct tracing_map_sort_key sort_key; + atomic64_t hits; + atomic64_t drops; +}; + +/** + * struct tracing_map_ops - callbacks for tracing_map + * + * The methods in this structure define callback functions for various + * operations on a tracing_map or objects related to a tracing_map. + * + * For a detailed description of tracing_map_elt objects please see + * the overview of tracing_map data structures at the beginning of + * this file. + * + * All the methods below are optional. + * + * @elt_alloc: When a tracing_map_elt is allocated, this function, if + * defined, will be called and gives clients the opportunity to + * allocate additional data and attach it to the element + * (tracing_map_elt->private_data is meant for that purpose). + * Element allocation occurs before tracing begins, when the + * tracing_map_init() call is made by client code. + * + * @elt_copy: At certain points in the lifetime of an element, it may + * need to be copied. The copy should include a copy of the + * client-allocated data, which can be copied into the 'to' + * element from the 'from' element. + * + * @elt_free: When a tracing_map_elt is freed, this function is called + * and allows client-allocated per-element data to be freed. + * + * @elt_clear: This callback allows per-element client-defined data to + * be cleared, if applicable. + * + * @elt_init: This callback allows per-element client-defined data to + * be initialized when used i.e. when the element is actually + * claimed by tracing_map_insert() in the context of the map + * insertion. + */ +struct tracing_map_ops { + int (*elt_alloc)(struct tracing_map_elt *elt); + void (*elt_copy)(struct tracing_map_elt *to, + struct tracing_map_elt *from); + void (*elt_free)(struct tracing_map_elt *elt); + void (*elt_clear)(struct tracing_map_elt *elt); + void (*elt_init)(struct tracing_map_elt *elt); +}; + +extern struct tracing_map * +tracing_map_create(unsigned int map_bits, + unsigned int key_size, + const struct tracing_map_ops *ops, + void *private_data); +extern int tracing_map_init(struct tracing_map *map); + +extern int tracing_map_add_sum_field(struct tracing_map *map); +extern int tracing_map_add_key_field(struct tracing_map *map, + unsigned int offset, + tracing_map_cmp_fn_t cmp_fn); + +extern void tracing_map_destroy(struct tracing_map *map); +extern void tracing_map_clear(struct tracing_map *map); + +extern struct tracing_map_elt * +tracing_map_insert(struct tracing_map *map, void *key); +extern struct tracing_map_elt * +tracing_map_lookup(struct tracing_map *map, void *key); + +extern tracing_map_cmp_fn_t tracing_map_cmp_num(int field_size, + int field_is_signed); +extern int tracing_map_cmp_string(void *val_a, void *val_b); +extern int tracing_map_cmp_none(void *val_a, void *val_b); + +extern void tracing_map_update_sum(struct tracing_map_elt *elt, + unsigned int i, u64 n); +extern u64 tracing_map_read_sum(struct tracing_map_elt *elt, unsigned int i); +extern void tracing_map_set_field_descr(struct tracing_map *map, + unsigned int i, + unsigned int key_offset, + tracing_map_cmp_fn_t cmp_fn); +extern int +tracing_map_sort_entries(struct tracing_map *map, + struct tracing_map_sort_key *sort_keys, + unsigned int n_sort_keys, + struct tracing_map_sort_entry ***sort_entries); + +extern void +tracing_map_destroy_sort_entries(struct tracing_map_sort_entry **entries, + unsigned int n_entries); +#endif /* __TRACING_MAP_H */ -- 1.9.3 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/