提交 c0d25450 编写于 作者: L Linus Torvalds

Merge branch 'percpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'percpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  percpu: remove rbtree and use page->index instead
  percpu: don't put the first chunk in reverse-map rbtree
...@@ -23,7 +23,7 @@ ...@@ -23,7 +23,7 @@
* Allocation is done in offset-size areas of single unit space. Ie, * Allocation is done in offset-size areas of single unit space. Ie,
* an area of 512 bytes at 6k in c1 occupies 512 bytes at 6k of c1:u0, * an area of 512 bytes at 6k in c1 occupies 512 bytes at 6k of c1:u0,
* c1:u1, c1:u2 and c1:u3. Percpu access can be done by configuring * c1:u1, c1:u2 and c1:u3. Percpu access can be done by configuring
* percpu base registers UNIT_SIZE apart. * percpu base registers pcpu_unit_size apart.
* *
* There are usually many small percpu allocations many of them as * There are usually many small percpu allocations many of them as
* small as 4 bytes. The allocator organizes chunks into lists * small as 4 bytes. The allocator organizes chunks into lists
...@@ -38,8 +38,8 @@ ...@@ -38,8 +38,8 @@
* region and negative allocated. Allocation inside a chunk is done * region and negative allocated. Allocation inside a chunk is done
* by scanning this map sequentially and serving the first matching * by scanning this map sequentially and serving the first matching
* entry. This is mostly copied from the percpu_modalloc() allocator. * entry. This is mostly copied from the percpu_modalloc() allocator.
* Chunks are also linked into a rb tree to ease address to chunk * Chunks can be determined from the address using the index field
* mapping during free. * in the page struct. The index field contains a pointer to the chunk.
* *
* To use this allocator, arch code should do the followings. * To use this allocator, arch code should do the followings.
* *
...@@ -61,7 +61,6 @@ ...@@ -61,7 +61,6 @@
#include <linux/mutex.h> #include <linux/mutex.h>
#include <linux/percpu.h> #include <linux/percpu.h>
#include <linux/pfn.h> #include <linux/pfn.h>
#include <linux/rbtree.h>
#include <linux/slab.h> #include <linux/slab.h>
#include <linux/spinlock.h> #include <linux/spinlock.h>
#include <linux/vmalloc.h> #include <linux/vmalloc.h>
...@@ -88,7 +87,6 @@ ...@@ -88,7 +87,6 @@
struct pcpu_chunk { struct pcpu_chunk {
struct list_head list; /* linked to pcpu_slot lists */ struct list_head list; /* linked to pcpu_slot lists */
struct rb_node rb_node; /* key is chunk->vm->addr */
int free_size; /* free bytes in the chunk */ int free_size; /* free bytes in the chunk */
int contig_hint; /* max contiguous size hint */ int contig_hint; /* max contiguous size hint */
struct vm_struct *vm; /* mapped vmalloc region */ struct vm_struct *vm; /* mapped vmalloc region */
...@@ -110,9 +108,21 @@ static size_t pcpu_chunk_struct_size __read_mostly; ...@@ -110,9 +108,21 @@ static size_t pcpu_chunk_struct_size __read_mostly;
void *pcpu_base_addr __read_mostly; void *pcpu_base_addr __read_mostly;
EXPORT_SYMBOL_GPL(pcpu_base_addr); EXPORT_SYMBOL_GPL(pcpu_base_addr);
/* optional reserved chunk, only accessible for reserved allocations */ /*
* The first chunk which always exists. Note that unlike other
* chunks, this one can be allocated and mapped in several different
* ways and thus often doesn't live in the vmalloc area.
*/
static struct pcpu_chunk *pcpu_first_chunk;
/*
* Optional reserved chunk. This chunk reserves part of the first
* chunk and serves it for reserved allocations. The amount of
* reserved offset is in pcpu_reserved_chunk_limit. When reserved
* area doesn't exist, the following variables contain NULL and 0
* respectively.
*/
static struct pcpu_chunk *pcpu_reserved_chunk; static struct pcpu_chunk *pcpu_reserved_chunk;
/* offset limit of the reserved chunk */
static int pcpu_reserved_chunk_limit; static int pcpu_reserved_chunk_limit;
/* /*
...@@ -121,7 +131,7 @@ static int pcpu_reserved_chunk_limit; ...@@ -121,7 +131,7 @@ static int pcpu_reserved_chunk_limit;
* There are two locks - pcpu_alloc_mutex and pcpu_lock. The former * There are two locks - pcpu_alloc_mutex and pcpu_lock. The former
* protects allocation/reclaim paths, chunks and chunk->page arrays. * protects allocation/reclaim paths, chunks and chunk->page arrays.
* The latter is a spinlock and protects the index data structures - * The latter is a spinlock and protects the index data structures -
* chunk slots, rbtree, chunks and area maps in chunks. * chunk slots, chunks and area maps in chunks.
* *
* During allocation, pcpu_alloc_mutex is kept locked all the time and * During allocation, pcpu_alloc_mutex is kept locked all the time and
* pcpu_lock is grabbed and released as necessary. All actual memory * pcpu_lock is grabbed and released as necessary. All actual memory
...@@ -140,7 +150,6 @@ static DEFINE_MUTEX(pcpu_alloc_mutex); /* protects whole alloc and reclaim */ ...@@ -140,7 +150,6 @@ static DEFINE_MUTEX(pcpu_alloc_mutex); /* protects whole alloc and reclaim */
static DEFINE_SPINLOCK(pcpu_lock); /* protects index data structures */ static DEFINE_SPINLOCK(pcpu_lock); /* protects index data structures */
static struct list_head *pcpu_slot __read_mostly; /* chunk list slots */ static struct list_head *pcpu_slot __read_mostly; /* chunk list slots */
static struct rb_root pcpu_addr_root = RB_ROOT; /* chunks by address */
/* reclaim work to release fully free chunks, scheduled from free path */ /* reclaim work to release fully free chunks, scheduled from free path */
static void pcpu_reclaim(struct work_struct *work); static void pcpu_reclaim(struct work_struct *work);
...@@ -191,6 +200,18 @@ static bool pcpu_chunk_page_occupied(struct pcpu_chunk *chunk, ...@@ -191,6 +200,18 @@ static bool pcpu_chunk_page_occupied(struct pcpu_chunk *chunk,
return *pcpu_chunk_pagep(chunk, 0, page_idx) != NULL; return *pcpu_chunk_pagep(chunk, 0, page_idx) != NULL;
} }
/* set the pointer to a chunk in a page struct */
static void pcpu_set_page_chunk(struct page *page, struct pcpu_chunk *pcpu)
{
page->index = (unsigned long)pcpu;
}
/* obtain pointer to a chunk from a page struct */
static struct pcpu_chunk *pcpu_get_page_chunk(struct page *page)
{
return (struct pcpu_chunk *)page->index;
}
/** /**
* pcpu_mem_alloc - allocate memory * pcpu_mem_alloc - allocate memory
* @size: bytes to allocate * @size: bytes to allocate
...@@ -257,93 +278,26 @@ static void pcpu_chunk_relocate(struct pcpu_chunk *chunk, int oslot) ...@@ -257,93 +278,26 @@ static void pcpu_chunk_relocate(struct pcpu_chunk *chunk, int oslot)
} }
} }
static struct rb_node **pcpu_chunk_rb_search(void *addr,
struct rb_node **parentp)
{
struct rb_node **p = &pcpu_addr_root.rb_node;
struct rb_node *parent = NULL;
struct pcpu_chunk *chunk;
while (*p) {
parent = *p;
chunk = rb_entry(parent, struct pcpu_chunk, rb_node);
if (addr < chunk->vm->addr)
p = &(*p)->rb_left;
else if (addr > chunk->vm->addr)
p = &(*p)->rb_right;
else
break;
}
if (parentp)
*parentp = parent;
return p;
}
/** /**
* pcpu_chunk_addr_search - search for chunk containing specified address * pcpu_chunk_addr_search - determine chunk containing specified address
* @addr: address to search for * @addr: address for which the chunk needs to be determined.
*
* Look for chunk which might contain @addr. More specifically, it
* searchs for the chunk with the highest start address which isn't
* beyond @addr.
*
* CONTEXT:
* pcpu_lock.
* *
* RETURNS: * RETURNS:
* The address of the found chunk. * The address of the found chunk.
*/ */
static struct pcpu_chunk *pcpu_chunk_addr_search(void *addr) static struct pcpu_chunk *pcpu_chunk_addr_search(void *addr)
{ {
struct rb_node *n, *parent; void *first_start = pcpu_first_chunk->vm->addr;
struct pcpu_chunk *chunk;
/* is it in the reserved chunk? */ /* is it in the first chunk? */
if (pcpu_reserved_chunk) { if (addr >= first_start && addr < first_start + pcpu_chunk_size) {
void *start = pcpu_reserved_chunk->vm->addr; /* is it in the reserved area? */
if (addr < first_start + pcpu_reserved_chunk_limit)
if (addr >= start && addr < start + pcpu_reserved_chunk_limit)
return pcpu_reserved_chunk; return pcpu_reserved_chunk;
return pcpu_first_chunk;
} }
/* nah... search the regular ones */ return pcpu_get_page_chunk(vmalloc_to_page(addr));
n = *pcpu_chunk_rb_search(addr, &parent);
if (!n) {
/* no exactly matching chunk, the parent is the closest */
n = parent;
BUG_ON(!n);
}
chunk = rb_entry(n, struct pcpu_chunk, rb_node);
if (addr < chunk->vm->addr) {
/* the parent was the next one, look for the previous one */
n = rb_prev(n);
BUG_ON(!n);
chunk = rb_entry(n, struct pcpu_chunk, rb_node);
}
return chunk;
}
/**
* pcpu_chunk_addr_insert - insert chunk into address rb tree
* @new: chunk to insert
*
* Insert @new into address rb tree.
*
* CONTEXT:
* pcpu_lock.
*/
static void pcpu_chunk_addr_insert(struct pcpu_chunk *new)
{
struct rb_node **p, *parent;
p = pcpu_chunk_rb_search(new->vm->addr, &parent);
BUG_ON(*p);
rb_link_node(&new->rb_node, parent, p);
rb_insert_color(&new->rb_node, &pcpu_addr_root);
} }
/** /**
...@@ -755,6 +709,7 @@ static int pcpu_populate_chunk(struct pcpu_chunk *chunk, int off, int size) ...@@ -755,6 +709,7 @@ static int pcpu_populate_chunk(struct pcpu_chunk *chunk, int off, int size)
alloc_mask, 0); alloc_mask, 0);
if (!*pagep) if (!*pagep)
goto err; goto err;
pcpu_set_page_chunk(*pagep, chunk);
} }
} }
...@@ -879,7 +834,6 @@ static void *pcpu_alloc(size_t size, size_t align, bool reserved) ...@@ -879,7 +834,6 @@ static void *pcpu_alloc(size_t size, size_t align, bool reserved)
spin_lock_irq(&pcpu_lock); spin_lock_irq(&pcpu_lock);
pcpu_chunk_relocate(chunk, -1); pcpu_chunk_relocate(chunk, -1);
pcpu_chunk_addr_insert(chunk);
goto restart; goto restart;
area_found: area_found:
...@@ -968,7 +922,6 @@ static void pcpu_reclaim(struct work_struct *work) ...@@ -968,7 +922,6 @@ static void pcpu_reclaim(struct work_struct *work)
if (chunk == list_first_entry(head, struct pcpu_chunk, list)) if (chunk == list_first_entry(head, struct pcpu_chunk, list))
continue; continue;
rb_erase(&chunk->rb_node, &pcpu_addr_root);
list_move(&chunk->list, &todo); list_move(&chunk->list, &todo);
} }
...@@ -1147,7 +1100,8 @@ size_t __init pcpu_setup_first_chunk(pcpu_get_page_fn_t get_page_fn, ...@@ -1147,7 +1100,8 @@ size_t __init pcpu_setup_first_chunk(pcpu_get_page_fn_t get_page_fn,
if (reserved_size) { if (reserved_size) {
schunk->free_size = reserved_size; schunk->free_size = reserved_size;
pcpu_reserved_chunk = schunk; /* not for dynamic alloc */ pcpu_reserved_chunk = schunk;
pcpu_reserved_chunk_limit = static_size + reserved_size;
} else { } else {
schunk->free_size = dyn_size; schunk->free_size = dyn_size;
dyn_size = 0; /* dynamic area covered */ dyn_size = 0; /* dynamic area covered */
...@@ -1158,8 +1112,6 @@ size_t __init pcpu_setup_first_chunk(pcpu_get_page_fn_t get_page_fn, ...@@ -1158,8 +1112,6 @@ size_t __init pcpu_setup_first_chunk(pcpu_get_page_fn_t get_page_fn,
if (schunk->free_size) if (schunk->free_size)
schunk->map[schunk->map_used++] = schunk->free_size; schunk->map[schunk->map_used++] = schunk->free_size;
pcpu_reserved_chunk_limit = static_size + schunk->free_size;
/* init dynamic chunk if necessary */ /* init dynamic chunk if necessary */
if (dyn_size) { if (dyn_size) {
dchunk = alloc_bootmem(sizeof(struct pcpu_chunk)); dchunk = alloc_bootmem(sizeof(struct pcpu_chunk));
...@@ -1226,13 +1178,8 @@ size_t __init pcpu_setup_first_chunk(pcpu_get_page_fn_t get_page_fn, ...@@ -1226,13 +1178,8 @@ size_t __init pcpu_setup_first_chunk(pcpu_get_page_fn_t get_page_fn,
} }
/* link the first chunk in */ /* link the first chunk in */
if (!dchunk) { pcpu_first_chunk = dchunk ?: schunk;
pcpu_chunk_relocate(schunk, -1); pcpu_chunk_relocate(pcpu_first_chunk, -1);
pcpu_chunk_addr_insert(schunk);
} else {
pcpu_chunk_relocate(dchunk, -1);
pcpu_chunk_addr_insert(dchunk);
}
/* we're done */ /* we're done */
pcpu_base_addr = (void *)pcpu_chunk_addr(schunk, 0, 0); pcpu_base_addr = (void *)pcpu_chunk_addr(schunk, 0, 0);
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册