linux_dsm_epyc7002/mm/percpu-internal.h

225 lines
5.9 KiB
C
Raw Normal View History

#ifndef _MM_PERCPU_INTERNAL_H
#define _MM_PERCPU_INTERNAL_H
#include <linux/types.h>
#include <linux/percpu.h>
/*
* pcpu_block_md is the metadata block struct.
* Each chunk's bitmap is split into a number of full blocks.
* All units are in terms of bits.
*/
struct pcpu_block_md {
int contig_hint; /* contig hint for block */
int contig_hint_start; /* block relative starting
position of the contig hint */
int left_free; /* size of free space along
the left side of the block */
int right_free; /* size of free space along
the right side of the block */
int first_free; /* block position of first free */
};
struct pcpu_chunk {
#ifdef CONFIG_PERCPU_STATS
int nr_alloc; /* # of allocations */
size_t max_alloc_size; /* largest allocation size */
#endif
struct list_head list; /* linked to pcpu_slot lists */
percpu: replace area map allocator with bitmap The percpu memory allocator is experiencing scalability issues when allocating and freeing large numbers of counters as in BPF. Additionally, there is a corner case where iteration is triggered over all chunks if the contig_hint is the right size, but wrong alignment. This patch replaces the area map allocator with a basic bitmap allocator implementation. Each subsequent patch will introduce new features and replace full scanning functions with faster non-scanning options when possible. Implementation: This patchset removes the area map allocator in favor of a bitmap allocator backed by metadata blocks. The primary goal is to provide consistency in performance and memory footprint with a focus on small allocations (< 64 bytes). The bitmap removes the heavy memmove from the freeing critical path and provides a consistent memory footprint. The metadata blocks provide a bound on the amount of scanning required by maintaining a set of hints. In an effort to make freeing fast, the metadata is updated on the free path if the new free area makes a page free, a block free, or spans across blocks. This causes the chunk's contig hint to potentially be smaller than what it could allocate by up to the smaller of a page or a block. If the chunk's contig hint is contained within a block, a check occurs and the hint is kept accurate. Metadata is always kept accurate on allocation, so there will not be a situation where a chunk has a later contig hint than available. Evaluation: I have primarily done testing against a simple workload of allocation of 1 million objects (2^20) of varying size. Deallocation was done by in order, alternating, and in reverse. These numbers were collected after rebasing ontop of a80099a152. I present the worst-case numbers here: Area Map Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 310 | 4770 16B | 557 | 1325 64B | 436 | 273 256B | 776 | 131 1024B | 3280 | 122 Bitmap Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 490 | 70 16B | 515 | 75 64B | 610 | 80 256B | 950 | 100 1024B | 3520 | 200 This data demonstrates the inability for the area map allocator to handle less than ideal situations. In the best case of reverse deallocation, the area map allocator was able to perform within range of the bitmap allocator. In the worst case situation, freeing took nearly 5 seconds for 1 million 4-byte objects. The bitmap allocator dramatically improves the consistency of the free path. The small allocations performed nearly identical regardless of the freeing pattern. While it does add to the allocation latency, the allocation scenario here is optimal for the area map allocator. The area map allocator runs into trouble when it is allocating in chunks where the latter half is full. It is difficult to replicate this, so I present a variant where the pages are second half filled. Freeing was done sequentially. Below are the numbers for this scenario: Area Map Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 4118 | 4892 16B | 1651 | 1163 64B | 598 | 285 256B | 771 | 158 1024B | 3034 | 160 Bitmap Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 481 | 67 16B | 506 | 69 64B | 636 | 75 256B | 892 | 90 1024B | 3262 | 147 The data shows a parabolic curve of performance for the area map allocator. This is due to the memmove operation being the dominant cost with the lower object sizes as more objects are packed in a chunk and at higher object sizes, the traversal of the chunk slots is the dominating cost. The bitmap allocator suffers this problem as well. The above data shows the inability to scale for the allocation path with the area map allocator and that the bitmap allocator demonstrates consistent performance in general. The second problem of additional scanning can result in the area map allocator completing in 52 minutes when trying to allocate 1 million 4-byte objects with 8-byte alignment. The same workload takes approximately 16 seconds to complete for the bitmap allocator. V2: Fixed a bug in pcpu_alloc_first_chunk end_offset was setting the bitmap using bytes instead of bits. Added a comment to pcpu_cnt_pop_pages to explain bitmap_weight. Signed-off-by: Dennis Zhou <dennisszhou@gmail.com> Reviewed-by: Josef Bacik <jbacik@fb.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2017-07-13 01:27:32 +07:00
int free_bytes; /* free bytes in the chunk */
int contig_bits; /* max contiguous size hint */
int contig_bits_start; /* contig_bits starting
offset */
void *base_addr; /* base address of this chunk */
percpu: replace area map allocator with bitmap The percpu memory allocator is experiencing scalability issues when allocating and freeing large numbers of counters as in BPF. Additionally, there is a corner case where iteration is triggered over all chunks if the contig_hint is the right size, but wrong alignment. This patch replaces the area map allocator with a basic bitmap allocator implementation. Each subsequent patch will introduce new features and replace full scanning functions with faster non-scanning options when possible. Implementation: This patchset removes the area map allocator in favor of a bitmap allocator backed by metadata blocks. The primary goal is to provide consistency in performance and memory footprint with a focus on small allocations (< 64 bytes). The bitmap removes the heavy memmove from the freeing critical path and provides a consistent memory footprint. The metadata blocks provide a bound on the amount of scanning required by maintaining a set of hints. In an effort to make freeing fast, the metadata is updated on the free path if the new free area makes a page free, a block free, or spans across blocks. This causes the chunk's contig hint to potentially be smaller than what it could allocate by up to the smaller of a page or a block. If the chunk's contig hint is contained within a block, a check occurs and the hint is kept accurate. Metadata is always kept accurate on allocation, so there will not be a situation where a chunk has a later contig hint than available. Evaluation: I have primarily done testing against a simple workload of allocation of 1 million objects (2^20) of varying size. Deallocation was done by in order, alternating, and in reverse. These numbers were collected after rebasing ontop of a80099a152. I present the worst-case numbers here: Area Map Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 310 | 4770 16B | 557 | 1325 64B | 436 | 273 256B | 776 | 131 1024B | 3280 | 122 Bitmap Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 490 | 70 16B | 515 | 75 64B | 610 | 80 256B | 950 | 100 1024B | 3520 | 200 This data demonstrates the inability for the area map allocator to handle less than ideal situations. In the best case of reverse deallocation, the area map allocator was able to perform within range of the bitmap allocator. In the worst case situation, freeing took nearly 5 seconds for 1 million 4-byte objects. The bitmap allocator dramatically improves the consistency of the free path. The small allocations performed nearly identical regardless of the freeing pattern. While it does add to the allocation latency, the allocation scenario here is optimal for the area map allocator. The area map allocator runs into trouble when it is allocating in chunks where the latter half is full. It is difficult to replicate this, so I present a variant where the pages are second half filled. Freeing was done sequentially. Below are the numbers for this scenario: Area Map Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 4118 | 4892 16B | 1651 | 1163 64B | 598 | 285 256B | 771 | 158 1024B | 3034 | 160 Bitmap Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 481 | 67 16B | 506 | 69 64B | 636 | 75 256B | 892 | 90 1024B | 3262 | 147 The data shows a parabolic curve of performance for the area map allocator. This is due to the memmove operation being the dominant cost with the lower object sizes as more objects are packed in a chunk and at higher object sizes, the traversal of the chunk slots is the dominating cost. The bitmap allocator suffers this problem as well. The above data shows the inability to scale for the allocation path with the area map allocator and that the bitmap allocator demonstrates consistent performance in general. The second problem of additional scanning can result in the area map allocator completing in 52 minutes when trying to allocate 1 million 4-byte objects with 8-byte alignment. The same workload takes approximately 16 seconds to complete for the bitmap allocator. V2: Fixed a bug in pcpu_alloc_first_chunk end_offset was setting the bitmap using bytes instead of bits. Added a comment to pcpu_cnt_pop_pages to explain bitmap_weight. Signed-off-by: Dennis Zhou <dennisszhou@gmail.com> Reviewed-by: Josef Bacik <jbacik@fb.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2017-07-13 01:27:32 +07:00
unsigned long *alloc_map; /* allocation map */
unsigned long *bound_map; /* boundary map */
struct pcpu_block_md *md_blocks; /* metadata blocks */
void *data; /* chunk data */
int first_bit; /* no free below this */
bool immutable; /* no [de]population allowed */
int start_offset; /* the overlap with the previous
region to have a page aligned
base_addr */
int end_offset; /* additional area required to
have the region end page
aligned */
int nr_pages; /* # of pages served by this chunk */
int nr_populated; /* # of populated pages */
int nr_empty_pop_pages; /* # of empty populated pages */
unsigned long populated[]; /* populated bitmap */
};
extern spinlock_t pcpu_lock;
extern struct list_head *pcpu_slot;
extern int pcpu_nr_slots;
extern int pcpu_nr_empty_pop_pages;
extern struct pcpu_chunk *pcpu_first_chunk;
extern struct pcpu_chunk *pcpu_reserved_chunk;
/**
* pcpu_chunk_nr_blocks - converts nr_pages to # of md_blocks
* @chunk: chunk of interest
*
* This conversion is from the number of physical pages that the chunk
* serves to the number of bitmap blocks used.
*/
static inline int pcpu_chunk_nr_blocks(struct pcpu_chunk *chunk)
{
return chunk->nr_pages * PAGE_SIZE / PCPU_BITMAP_BLOCK_SIZE;
}
percpu: replace area map allocator with bitmap The percpu memory allocator is experiencing scalability issues when allocating and freeing large numbers of counters as in BPF. Additionally, there is a corner case where iteration is triggered over all chunks if the contig_hint is the right size, but wrong alignment. This patch replaces the area map allocator with a basic bitmap allocator implementation. Each subsequent patch will introduce new features and replace full scanning functions with faster non-scanning options when possible. Implementation: This patchset removes the area map allocator in favor of a bitmap allocator backed by metadata blocks. The primary goal is to provide consistency in performance and memory footprint with a focus on small allocations (< 64 bytes). The bitmap removes the heavy memmove from the freeing critical path and provides a consistent memory footprint. The metadata blocks provide a bound on the amount of scanning required by maintaining a set of hints. In an effort to make freeing fast, the metadata is updated on the free path if the new free area makes a page free, a block free, or spans across blocks. This causes the chunk's contig hint to potentially be smaller than what it could allocate by up to the smaller of a page or a block. If the chunk's contig hint is contained within a block, a check occurs and the hint is kept accurate. Metadata is always kept accurate on allocation, so there will not be a situation where a chunk has a later contig hint than available. Evaluation: I have primarily done testing against a simple workload of allocation of 1 million objects (2^20) of varying size. Deallocation was done by in order, alternating, and in reverse. These numbers were collected after rebasing ontop of a80099a152. I present the worst-case numbers here: Area Map Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 310 | 4770 16B | 557 | 1325 64B | 436 | 273 256B | 776 | 131 1024B | 3280 | 122 Bitmap Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 490 | 70 16B | 515 | 75 64B | 610 | 80 256B | 950 | 100 1024B | 3520 | 200 This data demonstrates the inability for the area map allocator to handle less than ideal situations. In the best case of reverse deallocation, the area map allocator was able to perform within range of the bitmap allocator. In the worst case situation, freeing took nearly 5 seconds for 1 million 4-byte objects. The bitmap allocator dramatically improves the consistency of the free path. The small allocations performed nearly identical regardless of the freeing pattern. While it does add to the allocation latency, the allocation scenario here is optimal for the area map allocator. The area map allocator runs into trouble when it is allocating in chunks where the latter half is full. It is difficult to replicate this, so I present a variant where the pages are second half filled. Freeing was done sequentially. Below are the numbers for this scenario: Area Map Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 4118 | 4892 16B | 1651 | 1163 64B | 598 | 285 256B | 771 | 158 1024B | 3034 | 160 Bitmap Allocator: Object Size | Alloc Time (ms) | Free Time (ms) ---------------------------------------------- 4B | 481 | 67 16B | 506 | 69 64B | 636 | 75 256B | 892 | 90 1024B | 3262 | 147 The data shows a parabolic curve of performance for the area map allocator. This is due to the memmove operation being the dominant cost with the lower object sizes as more objects are packed in a chunk and at higher object sizes, the traversal of the chunk slots is the dominating cost. The bitmap allocator suffers this problem as well. The above data shows the inability to scale for the allocation path with the area map allocator and that the bitmap allocator demonstrates consistent performance in general. The second problem of additional scanning can result in the area map allocator completing in 52 minutes when trying to allocate 1 million 4-byte objects with 8-byte alignment. The same workload takes approximately 16 seconds to complete for the bitmap allocator. V2: Fixed a bug in pcpu_alloc_first_chunk end_offset was setting the bitmap using bytes instead of bits. Added a comment to pcpu_cnt_pop_pages to explain bitmap_weight. Signed-off-by: Dennis Zhou <dennisszhou@gmail.com> Reviewed-by: Josef Bacik <jbacik@fb.com> Signed-off-by: Tejun Heo <tj@kernel.org>
2017-07-13 01:27:32 +07:00
/**
* pcpu_nr_pages_to_map_bits - converts the pages to size of bitmap
* @pages: number of physical pages
*
* This conversion is from physical pages to the number of bits
* required in the bitmap.
*/
static inline int pcpu_nr_pages_to_map_bits(int pages)
{
return pages * PAGE_SIZE / PCPU_MIN_ALLOC_SIZE;
}
/**
* pcpu_chunk_map_bits - helper to convert nr_pages to size of bitmap
* @chunk: chunk of interest
*
* This conversion is from the number of physical pages that the chunk
* serves to the number of bits in the bitmap.
*/
static inline int pcpu_chunk_map_bits(struct pcpu_chunk *chunk)
{
return pcpu_nr_pages_to_map_bits(chunk->nr_pages);
}
#ifdef CONFIG_PERCPU_STATS
#include <linux/spinlock.h>
struct percpu_stats {
u64 nr_alloc; /* lifetime # of allocations */
u64 nr_dealloc; /* lifetime # of deallocations */
u64 nr_cur_alloc; /* current # of allocations */
u64 nr_max_alloc; /* max # of live allocations */
u32 nr_chunks; /* current # of live chunks */
u32 nr_max_chunks; /* max # of live chunks */
size_t min_alloc_size; /* min allocaiton size */
size_t max_alloc_size; /* max allocation size */
};
extern struct percpu_stats pcpu_stats;
extern struct pcpu_alloc_info pcpu_stats_ai;
/*
* For debug purposes. We don't care about the flexible array.
*/
static inline void pcpu_stats_save_ai(const struct pcpu_alloc_info *ai)
{
memcpy(&pcpu_stats_ai, ai, sizeof(struct pcpu_alloc_info));
/* initialize min_alloc_size to unit_size */
pcpu_stats.min_alloc_size = pcpu_stats_ai.unit_size;
}
/*
* pcpu_stats_area_alloc - increment area allocation stats
* @chunk: the location of the area being allocated
* @size: size of area to allocate in bytes
*
* CONTEXT:
* pcpu_lock.
*/
static inline void pcpu_stats_area_alloc(struct pcpu_chunk *chunk, size_t size)
{
lockdep_assert_held(&pcpu_lock);
pcpu_stats.nr_alloc++;
pcpu_stats.nr_cur_alloc++;
pcpu_stats.nr_max_alloc =
max(pcpu_stats.nr_max_alloc, pcpu_stats.nr_cur_alloc);
pcpu_stats.min_alloc_size =
min(pcpu_stats.min_alloc_size, size);
pcpu_stats.max_alloc_size =
max(pcpu_stats.max_alloc_size, size);
chunk->nr_alloc++;
chunk->max_alloc_size = max(chunk->max_alloc_size, size);
}
/*
* pcpu_stats_area_dealloc - decrement allocation stats
* @chunk: the location of the area being deallocated
*
* CONTEXT:
* pcpu_lock.
*/
static inline void pcpu_stats_area_dealloc(struct pcpu_chunk *chunk)
{
lockdep_assert_held(&pcpu_lock);
pcpu_stats.nr_dealloc++;
pcpu_stats.nr_cur_alloc--;
chunk->nr_alloc--;
}
/*
* pcpu_stats_chunk_alloc - increment chunk stats
*/
static inline void pcpu_stats_chunk_alloc(void)
{
unsigned long flags;
spin_lock_irqsave(&pcpu_lock, flags);
pcpu_stats.nr_chunks++;
pcpu_stats.nr_max_chunks =
max(pcpu_stats.nr_max_chunks, pcpu_stats.nr_chunks);
spin_unlock_irqrestore(&pcpu_lock, flags);
}
/*
* pcpu_stats_chunk_dealloc - decrement chunk stats
*/
static inline void pcpu_stats_chunk_dealloc(void)
{
unsigned long flags;
spin_lock_irqsave(&pcpu_lock, flags);
pcpu_stats.nr_chunks--;
spin_unlock_irqrestore(&pcpu_lock, flags);
}
#else
static inline void pcpu_stats_save_ai(const struct pcpu_alloc_info *ai)
{
}
static inline void pcpu_stats_area_alloc(struct pcpu_chunk *chunk, size_t size)
{
}
static inline void pcpu_stats_area_dealloc(struct pcpu_chunk *chunk)
{
}
static inline void pcpu_stats_chunk_alloc(void)
{
}
static inline void pcpu_stats_chunk_dealloc(void)
{
}
#endif /* !CONFIG_PERCPU_STATS */
#endif