From patchwork Tue Nov 16 00:53:10 2010 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Huang, Ying" X-Patchwork-Id: 326742 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by demeter1.kernel.org (8.14.4/8.14.3) with ESMTP id oAG0sgwn015265 for ; Tue, 16 Nov 2010 00:54:42 GMT Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758974Ab0KPAxy (ORCPT ); Mon, 15 Nov 2010 19:53:54 -0500 Received: from mga14.intel.com ([143.182.124.37]:40415 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752826Ab0KPAxx (ORCPT ); Mon, 15 Nov 2010 19:53:53 -0500 Received: from azsmga001.ch.intel.com ([10.2.17.19]) by azsmga102.ch.intel.com with ESMTP; 15 Nov 2010 16:53:52 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.59,203,1288594800"; d="scan'208";a="349121662" Received: from yhuang-dev.sh.intel.com ([10.239.13.2]) by azsmga001.ch.intel.com with ESMTP; 15 Nov 2010 16:53:50 -0800 From: Huang Ying To: Len Brown Cc: linux-kernel@vger.kernel.org, Andi Kleen , ying.huang@intel.com, linux-acpi@vger.kernel.org, Linus Torvalds , Andrew Morton , Thomas Gleixner , Ingo Molnar , Mauro Carvalho Chehab , Peter Zijlstra Subject: [PATCH -v4 1/2] lib, Make gen_pool memory allocator lockless Date: Tue, 16 Nov 2010 08:53:10 +0800 Message-Id: <1289868791-16658-2-git-send-email-ying.huang@intel.com> X-Mailer: git-send-email 1.7.2.3 In-Reply-To: <1289868791-16658-1-git-send-email-ying.huang@intel.com> References: <1289868791-16658-1-git-send-email-ying.huang@intel.com> Sender: linux-acpi-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-acpi@vger.kernel.org X-Greylist: IP, sender and recipient auto-whitelisted, not delayed by milter-greylist-4.2.3 (demeter1.kernel.org [140.211.167.41]); Tue, 16 Nov 2010 00:54:42 +0000 (UTC) --- a/include/linux/bitmap.h +++ b/include/linux/bitmap.h @@ -142,6 +142,7 @@ extern void bitmap_release_region(unsign extern int bitmap_allocate_region(unsigned long *bitmap, int pos, int order); extern void bitmap_copy_le(void *dst, const unsigned long *src, int nbits); +#define BITMAP_FIRST_WORD_MASK(start) (~0UL << ((start) % BITS_PER_LONG)) #define BITMAP_LAST_WORD_MASK(nbits) \ ( \ ((nbits) % BITS_PER_LONG) ? \ --- a/include/linux/genalloc.h +++ b/include/linux/genalloc.h @@ -1,19 +1,26 @@ +#ifndef GENALLOC_H +#define GENALLOC_H /* - * Basic general purpose allocator for managing special purpose memory - * not managed by the regular kmalloc/kfree interface. - * Uses for this includes on-device special memory, uncached memory - * etc. + * Basic general purpose allocator for managing special purpose + * memory, for example, memory not managed by the regular + * kmalloc/kfree interface. Uses for this includes on-device special + * memory, uncached memory etc. + * + * The gen_pool_alloc, gen_pool_free, gen_pool_avail and gen_pool_size + * implementation is lockless, that is, multiple users can + * allocate/free memory in the pool simultaneously without lock. This + * also makes the gen_pool memory allocator can be used to + * allocate/free memory in IRQ/NMI handler. * * This source code is licensed under the GNU General Public License, * Version 2. See the file COPYING for more details. */ - /* * General purpose special memory pool descriptor. */ struct gen_pool { - rwlock_t lock; + spinlock_t lock; struct list_head chunks; /* list of chunks in this pool */ int min_alloc_order; /* minimum allocation order */ }; @@ -22,15 +29,29 @@ struct gen_pool { * General purpose special memory pool chunk descriptor. */ struct gen_pool_chunk { - spinlock_t lock; struct list_head next_chunk; /* next chunk in pool */ + atomic_t avail; unsigned long start_addr; /* starting address of memory chunk */ unsigned long end_addr; /* ending address of memory chunk */ unsigned long bits[0]; /* bitmap for allocating memory chunk */ }; +/** + * gen_pool_for_each_chunk - iterate over chunks of generic memory pool + * @chunk: the struct gen_pool_chunk * to use as a loop cursor + * @pool: the generic memory pool + * + * Not lockless, proper mutual exclusion is needed to use this macro + * with other gen_pool function simultaneously. + */ +#define gen_pool_for_each_chunk(chunk, pool) \ + list_for_each_entry(chunk, &pool->chunks, next_chunk) + extern struct gen_pool *gen_pool_create(int, int); extern int gen_pool_add(struct gen_pool *, unsigned long, size_t, int); extern void gen_pool_destroy(struct gen_pool *); extern unsigned long gen_pool_alloc(struct gen_pool *, size_t); extern void gen_pool_free(struct gen_pool *, unsigned long, size_t); +extern size_t gen_pool_avail(struct gen_pool *); +extern size_t gen_pool_size(struct gen_pool *); +#endif /* GENALLOC_H */ --- a/lib/bitmap.c +++ b/lib/bitmap.c @@ -271,8 +271,6 @@ int __bitmap_weight(const unsigned long } EXPORT_SYMBOL(__bitmap_weight); -#define BITMAP_FIRST_WORD_MASK(start) (~0UL << ((start) % BITS_PER_LONG)) - void bitmap_set(unsigned long *map, int start, int nr) { unsigned long *p = map + BIT_WORD(start); --- a/lib/genalloc.c +++ b/lib/genalloc.c @@ -1,8 +1,25 @@ /* - * Basic general purpose allocator for managing special purpose memory - * not managed by the regular kmalloc/kfree interface. - * Uses for this includes on-device special memory, uncached memory - * etc. + * Basic general purpose allocator for managing special purpose + * memory, for example, memory that is not managed by the regular + * kmalloc/kfree interface. Uses for this includes on-device special + * memory, uncached memory etc. + * + * It is safe to use the allocator in NMI handlers and other special + * unblockable contexts that could otherwise deadlock on locks. This + * is implemented by using atomic operations and retries on any + * conflicts. The disadvantage is that there may be livelocks in + * extreme cases. For better scalability, one allocator can be used + * for each CPU. + * + * The lockless operation only works if there is enough memory + * available. If new memory is added to the pool a lock has to be + * still taken. So any user relying on locklessness has to ensure + * that sufficient memory is preallocated. + * + * The basic atomic operation of this allocator is cmpxchg on long. + * On architectures that don't support cmpxchg natively a fallback is + * used. If the fallback uses locks it may not be safe to use it in + * NMI contexts on these architectures. * * Copyright 2005 (C) Jes Sorensen * @@ -13,8 +30,107 @@ #include #include #include +#include #include +static inline int set_bits_ll(unsigned long *addr, unsigned long mask_to_set) +{ + unsigned long val, nval; + + nval = *addr; + do { + val = nval; + if (val & mask_to_set) + return -EBUSY; + } while ((nval = cmpxchg(addr, val, val | mask_to_set)) != val); + + return 0; +} + +static inline int clear_bits_ll(unsigned long *addr, + unsigned long mask_to_clear) +{ + unsigned long val, nval; + + nval = *addr; + do { + val = nval; + if ((val & mask_to_clear) != mask_to_clear) + return -EBUSY; + } while ((nval = cmpxchg(addr, val, val & ~mask_to_clear)) != val); + + return 0; +} + +/* + * bitmap_set_ll - set the specified number of bits at the specified position + * @map: pointer to a bitmap + * @start: a bit position in @map + * @nr: number of bits to set + * + * Set @nr bits start from @start in @map lock-lessly. Several users + * can set/clear the same bitmap simultaneously without lock. If two + * users set the same bit, one user will return remain bits, otherwise + * return 0. + */ +static int bitmap_set_ll(unsigned long *map, int start, int nr) +{ + unsigned long *p = map + BIT_WORD(start); + const int size = start + nr; + int bits_to_set = BITS_PER_LONG - (start % BITS_PER_LONG); + unsigned long mask_to_set = BITMAP_FIRST_WORD_MASK(start); + + while (nr - bits_to_set >= 0) { + if (set_bits_ll(p, mask_to_set)) + return nr; + nr -= bits_to_set; + bits_to_set = BITS_PER_LONG; + mask_to_set = ~0UL; + p++; + } + if (nr) { + mask_to_set &= BITMAP_LAST_WORD_MASK(size); + if (set_bits_ll(p, mask_to_set)) + return nr; + } + + return 0; +} + +/* + * bitmap_clear_ll - clear the specified number of bits at the specified position + * @map: pointer to a bitmap + * @start: a bit position in @map + * @nr: number of bits to set + * + * Clear @nr bits start from @start in @map lock-lessly. Several users + * can set/clear the same bitmap simultaneously without lock. If two + * users clear the same bit, one user will return remain bits, + * otherwise return 0. + */ +static int bitmap_clear_ll(unsigned long *map, int start, int nr) +{ + unsigned long *p = map + BIT_WORD(start); + const int size = start + nr; + int bits_to_clear = BITS_PER_LONG - (start % BITS_PER_LONG); + unsigned long mask_to_clear = BITMAP_FIRST_WORD_MASK(start); + + while (nr - bits_to_clear >= 0) { + if (clear_bits_ll(p, mask_to_clear)) + return nr; + nr -= bits_to_clear; + bits_to_clear = BITS_PER_LONG; + mask_to_clear = ~0UL; + p++; + } + if (nr) { + mask_to_clear &= BITMAP_LAST_WORD_MASK(size); + if (clear_bits_ll(p, mask_to_clear)) + return nr; + } + + return 0; +} /** * gen_pool_create - create a new special memory pool @@ -30,7 +146,7 @@ struct gen_pool *gen_pool_create(int min pool = kmalloc_node(sizeof(struct gen_pool), GFP_KERNEL, nid); if (pool != NULL) { - rwlock_init(&pool->lock); + spin_lock_init(&pool->lock); INIT_LIST_HEAD(&pool->chunks); pool->min_alloc_order = min_alloc_order; } @@ -58,15 +174,15 @@ int gen_pool_add(struct gen_pool *pool, chunk = kmalloc_node(nbytes, GFP_KERNEL | __GFP_ZERO, nid); if (unlikely(chunk == NULL)) - return -1; + return -ENOMEM; - spin_lock_init(&chunk->lock); chunk->start_addr = addr; chunk->end_addr = addr + size; + atomic_set(&chunk->avail, size); - write_lock(&pool->lock); - list_add(&chunk->next_chunk, &pool->chunks); - write_unlock(&pool->lock); + spin_lock(&pool->lock); + list_add_rcu(&chunk->next_chunk, &pool->chunks); + spin_unlock(&pool->lock); return 0; } @@ -112,39 +228,39 @@ EXPORT_SYMBOL(gen_pool_destroy); */ unsigned long gen_pool_alloc(struct gen_pool *pool, size_t size) { - struct list_head *_chunk; struct gen_pool_chunk *chunk; - unsigned long addr, flags; + unsigned long addr; int order = pool->min_alloc_order; - int nbits, start_bit, end_bit; + int nbits, start_bit = 0, end_bit, remain; if (size == 0) return 0; nbits = (size + (1UL << order) - 1) >> order; - read_lock(&pool->lock); - list_for_each(_chunk, &pool->chunks) { - chunk = list_entry(_chunk, struct gen_pool_chunk, next_chunk); + list_for_each_entry_rcu(chunk, &pool->chunks, next_chunk) { + if (size > atomic_read(&chunk->avail)) + continue; end_bit = (chunk->end_addr - chunk->start_addr) >> order; - - spin_lock_irqsave(&chunk->lock, flags); - start_bit = bitmap_find_next_zero_area(chunk->bits, end_bit, 0, - nbits, 0); - if (start_bit >= end_bit) { - spin_unlock_irqrestore(&chunk->lock, flags); +retry: + start_bit = bitmap_find_next_zero_area(chunk->bits, end_bit, + start_bit, nbits, 0); + if (start_bit >= end_bit) continue; + remain = bitmap_set_ll(chunk->bits, start_bit, nbits); + if (remain) { + remain = bitmap_clear_ll(chunk->bits, start_bit, + nbits - remain); + BUG_ON(remain); + goto retry; } addr = chunk->start_addr + ((unsigned long)start_bit << order); - - bitmap_set(chunk->bits, start_bit, nbits); - spin_unlock_irqrestore(&chunk->lock, flags); - read_unlock(&pool->lock); + size = nbits << order; + atomic_sub(size, &chunk->avail); return addr; } - read_unlock(&pool->lock); return 0; } EXPORT_SYMBOL(gen_pool_alloc); @@ -159,29 +275,57 @@ EXPORT_SYMBOL(gen_pool_alloc); */ void gen_pool_free(struct gen_pool *pool, unsigned long addr, size_t size) { - struct list_head *_chunk; struct gen_pool_chunk *chunk; - unsigned long flags; int order = pool->min_alloc_order; - int bit, nbits; + int start_bit, nbits, remain; nbits = (size + (1UL << order) - 1) >> order; - read_lock(&pool->lock); - list_for_each(_chunk, &pool->chunks) { - chunk = list_entry(_chunk, struct gen_pool_chunk, next_chunk); - + list_for_each_entry_rcu(chunk, &pool->chunks, next_chunk) { if (addr >= chunk->start_addr && addr < chunk->end_addr) { BUG_ON(addr + size > chunk->end_addr); - spin_lock_irqsave(&chunk->lock, flags); - bit = (addr - chunk->start_addr) >> order; - while (nbits--) - __clear_bit(bit++, chunk->bits); - spin_unlock_irqrestore(&chunk->lock, flags); - break; + start_bit = (addr - chunk->start_addr) >> order; + remain = bitmap_clear_ll(chunk->bits, start_bit, nbits); + BUG_ON(remain); + size = nbits << order; + atomic_add(size, &chunk->avail); + return; } } - BUG_ON(nbits > 0); - read_unlock(&pool->lock); + BUG(); } EXPORT_SYMBOL(gen_pool_free); + +/** + * gen_pool_avail - get available free space of the pool + * @pool: pool to get available free space + * + * Return available free space of the specified pool. + */ +size_t gen_pool_avail(struct gen_pool *pool) +{ + struct gen_pool_chunk *chunk; + size_t avail = 0; + + list_for_each_entry_rcu(chunk, &pool->chunks, next_chunk) + avail += atomic_read(&chunk->avail); + return avail; +} +EXPORT_SYMBOL_GPL(gen_pool_avail); + +/** + * gen_pool_size - get size in bytes of memory managed by the pool + * @pool: pool to get size + * + * Return size in bytes of memory managed by the pool. + */ +size_t gen_pool_size(struct gen_pool *pool) +{ + struct gen_pool_chunk *chunk; + size_t size = 0; + + list_for_each_entry_rcu(chunk, &pool->chunks, next_chunk) + size += chunk->end_addr - chunk->start_addr; + return size; +} +EXPORT_SYMBOL_GPL(gen_pool_size);