kernel-fxtec-pro1x/lib/genalloc.c
Benjamin Gaignard ca279cf106 genalloc: make it possible to use a custom allocation algorithm
Premit use of another algorithm than the default first-fit one.  For
example a custom algorithm could be used to manage alignment requirements.

As I can't predict all the possible requirements/needs for all allocation
uses cases, I add a "free" field 'void *data' to pass any needed
information to the allocation function.  For example 'data' could be used
to handle a structure where you store the alignment, the expected memory
bank, the requester device, or any information that could influence the
allocation algorithm.

An usage example may look like this:
struct my_pool_constraints {
	int align;
	int bank;
	...
};

unsigned long my_custom_algo(unsigned long *map, unsigned long size,
		unsigned long start, unsigned int nr, void *data)
{
	struct my_pool_constraints *constraints = data;
	...
	deal with allocation contraints
	...
	return the index in bitmap where perform the allocation
}

void create_my_pool()
{
	struct my_pool_constraints c;
	struct gen_pool *pool = gen_pool_create(...);
	gen_pool_add(pool, ...);
	gen_pool_set_algo(pool, my_custom_algo, &c);
}

Add of best-fit algorithm function:
most of the time best-fit is slower then first-fit but memory fragmentation
is lower. The random buffer allocation/free tests don't show any arithmetic
relation between the allocation time and fragmentation but the
best-fit algorithm
is sometime able to perform the allocation when the first-fit can't.

This new algorithm help to remove static allocations on ESRAM, a small but
fast on-chip RAM of few KB, used for high-performance uses cases like DMA
linked lists, graphic accelerators, encoders/decoders. On the Ux500
(in the ARM tree) we have define 5 ESRAM banks of 128 KB each and use of
static allocations becomes unmaintainable:
cd arch/arm/mach-ux500 && grep -r ESRAM .
./include/mach/db8500-regs.h:/* Base address and bank offsets for ESRAM */
./include/mach/db8500-regs.h:#define U8500_ESRAM_BASE   0x40000000
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK_SIZE      0x00020000
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK0  U8500_ESRAM_BASE
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK1       (U8500_ESRAM_BASE + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK2       (U8500_ESRAM_BANK1 + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK3       (U8500_ESRAM_BANK2 + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK4       (U8500_ESRAM_BANK3 + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_DMA_LCPA_OFFSET     0x10000
./include/mach/db8500-regs.h:#define U8500_DMA_LCPA_BASE
(U8500_ESRAM_BANK0 + U8500_ESRAM_DMA_LCPA_OFFSET)
./include/mach/db8500-regs.h:#define U8500_DMA_LCLA_BASE U8500_ESRAM_BANK4

I want to use genalloc to do dynamic allocations but I need to be able to
fine tune the allocation algorithm. I my case best-fit algorithm give
better results than first-fit, but it will not be true for every use case.

Signed-off-by: Benjamin Gaignard <benjamin.gaignard@stericsson.com>
Cc: Huang Ying <ying.huang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-10-06 03:04:57 +09:00

482 lines
13 KiB
C

/*
* 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 have NMI-safe cmpxchg implementation,
* the allocator can NOT be used in NMI handler. So code uses the
* allocator in NMI handler should depend on
* CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG.
*
* Copyright 2005 (C) Jes Sorensen <jes@trained-monkey.org>
*
* This source code is licensed under the GNU General Public License,
* Version 2. See the file COPYING for more details.
*/
#include <linux/slab.h>
#include <linux/export.h>
#include <linux/bitmap.h>
#include <linux/rculist.h>
#include <linux/interrupt.h>
#include <linux/genalloc.h>
static 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;
cpu_relax();
} while ((nval = cmpxchg(addr, val, val | mask_to_set)) != val);
return 0;
}
static 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;
cpu_relax();
} 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
* @min_alloc_order: log base 2 of number of bytes each bitmap bit represents
* @nid: node id of the node the pool structure should be allocated on, or -1
*
* Create a new special memory pool that can be used to manage special purpose
* memory not managed by the regular kmalloc/kfree interface.
*/
struct gen_pool *gen_pool_create(int min_alloc_order, int nid)
{
struct gen_pool *pool;
pool = kmalloc_node(sizeof(struct gen_pool), GFP_KERNEL, nid);
if (pool != NULL) {
spin_lock_init(&pool->lock);
INIT_LIST_HEAD(&pool->chunks);
pool->min_alloc_order = min_alloc_order;
pool->algo = gen_pool_first_fit;
pool->data = NULL;
}
return pool;
}
EXPORT_SYMBOL(gen_pool_create);
/**
* gen_pool_add_virt - add a new chunk of special memory to the pool
* @pool: pool to add new memory chunk to
* @virt: virtual starting address of memory chunk to add to pool
* @phys: physical starting address of memory chunk to add to pool
* @size: size in bytes of the memory chunk to add to pool
* @nid: node id of the node the chunk structure and bitmap should be
* allocated on, or -1
*
* Add a new chunk of special memory to the specified pool.
*
* Returns 0 on success or a -ve errno on failure.
*/
int gen_pool_add_virt(struct gen_pool *pool, unsigned long virt, phys_addr_t phys,
size_t size, int nid)
{
struct gen_pool_chunk *chunk;
int nbits = size >> pool->min_alloc_order;
int nbytes = sizeof(struct gen_pool_chunk) +
(nbits + BITS_PER_BYTE - 1) / BITS_PER_BYTE;
chunk = kmalloc_node(nbytes, GFP_KERNEL | __GFP_ZERO, nid);
if (unlikely(chunk == NULL))
return -ENOMEM;
chunk->phys_addr = phys;
chunk->start_addr = virt;
chunk->end_addr = virt + size;
atomic_set(&chunk->avail, size);
spin_lock(&pool->lock);
list_add_rcu(&chunk->next_chunk, &pool->chunks);
spin_unlock(&pool->lock);
return 0;
}
EXPORT_SYMBOL(gen_pool_add_virt);
/**
* gen_pool_virt_to_phys - return the physical address of memory
* @pool: pool to allocate from
* @addr: starting address of memory
*
* Returns the physical address on success, or -1 on error.
*/
phys_addr_t gen_pool_virt_to_phys(struct gen_pool *pool, unsigned long addr)
{
struct gen_pool_chunk *chunk;
phys_addr_t paddr = -1;
rcu_read_lock();
list_for_each_entry_rcu(chunk, &pool->chunks, next_chunk) {
if (addr >= chunk->start_addr && addr < chunk->end_addr) {
paddr = chunk->phys_addr + (addr - chunk->start_addr);
break;
}
}
rcu_read_unlock();
return paddr;
}
EXPORT_SYMBOL(gen_pool_virt_to_phys);
/**
* gen_pool_destroy - destroy a special memory pool
* @pool: pool to destroy
*
* Destroy the specified special memory pool. Verifies that there are no
* outstanding allocations.
*/
void gen_pool_destroy(struct gen_pool *pool)
{
struct list_head *_chunk, *_next_chunk;
struct gen_pool_chunk *chunk;
int order = pool->min_alloc_order;
int bit, end_bit;
list_for_each_safe(_chunk, _next_chunk, &pool->chunks) {
chunk = list_entry(_chunk, struct gen_pool_chunk, next_chunk);
list_del(&chunk->next_chunk);
end_bit = (chunk->end_addr - chunk->start_addr) >> order;
bit = find_next_bit(chunk->bits, end_bit, 0);
BUG_ON(bit < end_bit);
kfree(chunk);
}
kfree(pool);
return;
}
EXPORT_SYMBOL(gen_pool_destroy);
/**
* gen_pool_alloc - allocate special memory from the pool
* @pool: pool to allocate from
* @size: number of bytes to allocate from the pool
*
* Allocate the requested number of bytes from the specified pool.
* Uses the pool allocation function (with first-fit algorithm by default).
* Can not be used in NMI handler on architectures without
* NMI-safe cmpxchg implementation.
*/
unsigned long gen_pool_alloc(struct gen_pool *pool, size_t size)
{
struct gen_pool_chunk *chunk;
unsigned long addr = 0;
int order = pool->min_alloc_order;
int nbits, start_bit = 0, end_bit, remain;
#ifndef CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG
BUG_ON(in_nmi());
#endif
if (size == 0)
return 0;
nbits = (size + (1UL << order) - 1) >> order;
rcu_read_lock();
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;
retry:
start_bit = pool->algo(chunk->bits, end_bit, start_bit, nbits,
pool->data);
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);
size = nbits << order;
atomic_sub(size, &chunk->avail);
break;
}
rcu_read_unlock();
return addr;
}
EXPORT_SYMBOL(gen_pool_alloc);
/**
* gen_pool_free - free allocated special memory back to the pool
* @pool: pool to free to
* @addr: starting address of memory to free back to pool
* @size: size in bytes of memory to free
*
* Free previously allocated special memory back to the specified
* pool. Can not be used in NMI handler on architectures without
* NMI-safe cmpxchg implementation.
*/
void gen_pool_free(struct gen_pool *pool, unsigned long addr, size_t size)
{
struct gen_pool_chunk *chunk;
int order = pool->min_alloc_order;
int start_bit, nbits, remain;
#ifndef CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG
BUG_ON(in_nmi());
#endif
nbits = (size + (1UL << order) - 1) >> order;
rcu_read_lock();
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);
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);
rcu_read_unlock();
return;
}
}
rcu_read_unlock();
BUG();
}
EXPORT_SYMBOL(gen_pool_free);
/**
* gen_pool_for_each_chunk - call func for every chunk of generic memory pool
* @pool: the generic memory pool
* @func: func to call
* @data: additional data used by @func
*
* Call @func for every chunk of generic memory pool. The @func is
* called with rcu_read_lock held.
*/
void gen_pool_for_each_chunk(struct gen_pool *pool,
void (*func)(struct gen_pool *pool, struct gen_pool_chunk *chunk, void *data),
void *data)
{
struct gen_pool_chunk *chunk;
rcu_read_lock();
list_for_each_entry_rcu(chunk, &(pool)->chunks, next_chunk)
func(pool, chunk, data);
rcu_read_unlock();
}
EXPORT_SYMBOL(gen_pool_for_each_chunk);
/**
* 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;
rcu_read_lock();
list_for_each_entry_rcu(chunk, &pool->chunks, next_chunk)
avail += atomic_read(&chunk->avail);
rcu_read_unlock();
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;
rcu_read_lock();
list_for_each_entry_rcu(chunk, &pool->chunks, next_chunk)
size += chunk->end_addr - chunk->start_addr;
rcu_read_unlock();
return size;
}
EXPORT_SYMBOL_GPL(gen_pool_size);
/**
* gen_pool_set_algo - set the allocation algorithm
* @pool: pool to change allocation algorithm
* @algo: custom algorithm function
* @data: additional data used by @algo
*
* Call @algo for each memory allocation in the pool.
* If @algo is NULL use gen_pool_first_fit as default
* memory allocation function.
*/
void gen_pool_set_algo(struct gen_pool *pool, genpool_algo_t algo, void *data)
{
rcu_read_lock();
pool->algo = algo;
if (!pool->algo)
pool->algo = gen_pool_first_fit;
pool->data = data;
rcu_read_unlock();
}
EXPORT_SYMBOL(gen_pool_set_algo);
/**
* gen_pool_first_fit - find the first available region
* of memory matching the size requirement (no alignment constraint)
* @map: The address to base the search on
* @size: The bitmap size in bits
* @start: The bitnumber to start searching at
* @nr: The number of zeroed bits we're looking for
* @data: additional data - unused
*/
unsigned long gen_pool_first_fit(unsigned long *map, unsigned long size,
unsigned long start, unsigned int nr, void *data)
{
return bitmap_find_next_zero_area(map, size, start, nr, 0);
}
EXPORT_SYMBOL(gen_pool_first_fit);
/**
* gen_pool_best_fit - find the best fitting region of memory
* macthing the size requirement (no alignment constraint)
* @map: The address to base the search on
* @size: The bitmap size in bits
* @start: The bitnumber to start searching at
* @nr: The number of zeroed bits we're looking for
* @data: additional data - unused
*
* Iterate over the bitmap to find the smallest free region
* which we can allocate the memory.
*/
unsigned long gen_pool_best_fit(unsigned long *map, unsigned long size,
unsigned long start, unsigned int nr, void *data)
{
unsigned long start_bit = size;
unsigned long len = size + 1;
unsigned long index;
index = bitmap_find_next_zero_area(map, size, start, nr, 0);
while (index < size) {
int next_bit = find_next_bit(map, size, index + nr);
if ((next_bit - index) < len) {
len = next_bit - index;
start_bit = index;
if (len == nr)
return start_bit;
}
index = bitmap_find_next_zero_area(map, size,
next_bit + 1, nr, 0);
}
return start_bit;
}
EXPORT_SYMBOL(gen_pool_best_fit);