linux_dsm_epyc7002/include/linux/idr.h
Matthew Wilcox 424251a4a9 idr: reduce the number of bits per level from 8 to 6
In preparation for merging the IDR and radix tree, reduce the fanout at
each level from 256 to 64.  If this causes a performance problem then a
bisect will point to this commit, and we'll have a better idea about
what we might do to fix it.

Link: http://lkml.kernel.org/r/1480369871-5271-66-git-send-email-mawilcox@linuxonhyperv.com
Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
Tested-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Konstantin Khlebnikov <koct9i@gmail.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-12-14 16:04:10 -08:00

231 lines
6.7 KiB
C

/*
* include/linux/idr.h
*
* 2002-10-18 written by Jim Houston jim.houston@ccur.com
* Copyright (C) 2002 by Concurrent Computer Corporation
* Distributed under the GNU GPL license version 2.
*
* Small id to pointer translation service avoiding fixed sized
* tables.
*/
#ifndef __IDR_H__
#define __IDR_H__
#include <linux/types.h>
#include <linux/bitops.h>
#include <linux/init.h>
#include <linux/rcupdate.h>
/*
* Using 6 bits at each layer allows us to allocate 7 layers out of each page.
* 8 bits only gave us 3 layers out of every pair of pages, which is less
* efficient except for trees with a largest element between 192-255 inclusive.
*/
#define IDR_BITS 6
#define IDR_SIZE (1 << IDR_BITS)
#define IDR_MASK ((1 << IDR_BITS)-1)
struct idr_layer {
int prefix; /* the ID prefix of this idr_layer */
int layer; /* distance from leaf */
struct idr_layer __rcu *ary[1<<IDR_BITS];
int count; /* When zero, we can release it */
union {
/* A zero bit means "space here" */
DECLARE_BITMAP(bitmap, IDR_SIZE);
struct rcu_head rcu_head;
};
};
struct idr {
struct idr_layer __rcu *hint; /* the last layer allocated from */
struct idr_layer __rcu *top;
int layers; /* only valid w/o concurrent changes */
int cur; /* current pos for cyclic allocation */
spinlock_t lock;
int id_free_cnt;
struct idr_layer *id_free;
};
#define IDR_INIT(name) \
{ \
.lock = __SPIN_LOCK_UNLOCKED(name.lock), \
}
#define DEFINE_IDR(name) struct idr name = IDR_INIT(name)
/**
* idr_get_cursor - Return the current position of the cyclic allocator
* @idr: idr handle
*
* The value returned is the value that will be next returned from
* idr_alloc_cyclic() if it is free (otherwise the search will start from
* this position).
*/
static inline unsigned int idr_get_cursor(struct idr *idr)
{
return READ_ONCE(idr->cur);
}
/**
* idr_set_cursor - Set the current position of the cyclic allocator
* @idr: idr handle
* @val: new position
*
* The next call to idr_alloc_cyclic() will return @val if it is free
* (otherwise the search will start from this position).
*/
static inline void idr_set_cursor(struct idr *idr, unsigned int val)
{
WRITE_ONCE(idr->cur, val);
}
/**
* DOC: idr sync
* idr synchronization (stolen from radix-tree.h)
*
* idr_find() is able to be called locklessly, using RCU. The caller must
* ensure calls to this function are made within rcu_read_lock() regions.
* Other readers (lock-free or otherwise) and modifications may be running
* concurrently.
*
* It is still required that the caller manage the synchronization and
* lifetimes of the items. So if RCU lock-free lookups are used, typically
* this would mean that the items have their own locks, or are amenable to
* lock-free access; and that the items are freed by RCU (or only freed after
* having been deleted from the idr tree *and* a synchronize_rcu() grace
* period).
*/
/*
* This is what we export.
*/
void *idr_find_slowpath(struct idr *idp, int id);
void idr_preload(gfp_t gfp_mask);
int idr_alloc(struct idr *idp, void *ptr, int start, int end, gfp_t gfp_mask);
int idr_alloc_cyclic(struct idr *idr, void *ptr, int start, int end, gfp_t gfp_mask);
int idr_for_each(struct idr *idp,
int (*fn)(int id, void *p, void *data), void *data);
void *idr_get_next(struct idr *idp, int *nextid);
void *idr_replace(struct idr *idp, void *ptr, int id);
void idr_remove(struct idr *idp, int id);
void idr_destroy(struct idr *idp);
void idr_init(struct idr *idp);
bool idr_is_empty(struct idr *idp);
/**
* idr_preload_end - end preload section started with idr_preload()
*
* Each idr_preload() should be matched with an invocation of this
* function. See idr_preload() for details.
*/
static inline void idr_preload_end(void)
{
preempt_enable();
}
/**
* idr_find - return pointer for given id
* @idr: idr handle
* @id: lookup key
*
* Return the pointer given the id it has been registered with. A %NULL
* return indicates that @id is not valid or you passed %NULL in
* idr_get_new().
*
* This function can be called under rcu_read_lock(), given that the leaf
* pointers lifetimes are correctly managed.
*/
static inline void *idr_find(struct idr *idr, int id)
{
struct idr_layer *hint = rcu_dereference_raw(idr->hint);
if (hint && (id & ~IDR_MASK) == hint->prefix)
return rcu_dereference_raw(hint->ary[id & IDR_MASK]);
return idr_find_slowpath(idr, id);
}
/**
* idr_for_each_entry - iterate over an idr's elements of a given type
* @idp: idr handle
* @entry: the type * to use as cursor
* @id: id entry's key
*
* @entry and @id do not need to be initialized before the loop, and
* after normal terminatinon @entry is left with the value NULL. This
* is convenient for a "not found" value.
*/
#define idr_for_each_entry(idp, entry, id) \
for (id = 0; ((entry) = idr_get_next(idp, &(id))) != NULL; ++id)
/**
* idr_for_each_entry - continue iteration over an idr's elements of a given type
* @idp: idr handle
* @entry: the type * to use as cursor
* @id: id entry's key
*
* Continue to iterate over list of given type, continuing after
* the current position.
*/
#define idr_for_each_entry_continue(idp, entry, id) \
for ((entry) = idr_get_next((idp), &(id)); \
entry; \
++id, (entry) = idr_get_next((idp), &(id)))
/*
* IDA - IDR based id allocator, use when translation from id to
* pointer isn't necessary.
*
* IDA_BITMAP_LONGS is calculated to be one less to accommodate
* ida_bitmap->nr_busy so that the whole struct fits in 128 bytes.
*/
#define IDA_CHUNK_SIZE 128 /* 128 bytes per chunk */
#define IDA_BITMAP_LONGS (IDA_CHUNK_SIZE / sizeof(long) - 1)
#define IDA_BITMAP_BITS (IDA_BITMAP_LONGS * sizeof(long) * 8)
struct ida_bitmap {
long nr_busy;
unsigned long bitmap[IDA_BITMAP_LONGS];
};
struct ida {
struct idr idr;
struct ida_bitmap *free_bitmap;
};
#define IDA_INIT(name) { .idr = IDR_INIT((name).idr), .free_bitmap = NULL, }
#define DEFINE_IDA(name) struct ida name = IDA_INIT(name)
int ida_pre_get(struct ida *ida, gfp_t gfp_mask);
int ida_get_new_above(struct ida *ida, int starting_id, int *p_id);
void ida_remove(struct ida *ida, int id);
void ida_destroy(struct ida *ida);
void ida_init(struct ida *ida);
int ida_simple_get(struct ida *ida, unsigned int start, unsigned int end,
gfp_t gfp_mask);
void ida_simple_remove(struct ida *ida, unsigned int id);
/**
* ida_get_new - allocate new ID
* @ida: idr handle
* @p_id: pointer to the allocated handle
*
* Simple wrapper around ida_get_new_above() w/ @starting_id of zero.
*/
static inline int ida_get_new(struct ida *ida, int *p_id)
{
return ida_get_new_above(ida, 0, p_id);
}
static inline bool ida_is_empty(struct ida *ida)
{
return idr_is_empty(&ida->idr);
}
void __init idr_init_cache(void);
#endif /* __IDR_H__ */