Permalink
Fetching contributors…
Cannot retrieve contributors at this time
363 lines (320 sloc) 13 KB
/** @file
A brief file description
@section license License
Licensed to the Apache Software Foundation (ASF) under one
or more contributor license agreements. See the NOTICE file
distributed with this work for additional information
regarding copyright ownership. The ASF licenses this file
to you under the Apache License, Version 2.0 (the
"License"); you may not use this file except in compliance
with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
#pragma once
#include "P_CacheHttp.h"
struct Vol;
struct InterimCacheVol;
struct CacheVC;
/*
Directory layout
*/
// Constants
#define DIR_TAG_WIDTH 12
#define DIR_MASK_TAG(_t) ((_t) & ((1 << DIR_TAG_WIDTH) - 1))
#define SIZEOF_DIR 10
#define ESTIMATED_OBJECT_SIZE 8000
#define MAX_DIR_SEGMENTS (32 * (1 << 16))
#define DIR_DEPTH 4
#define MAX_ENTRIES_PER_SEGMENT (1 << 16)
#define MAX_BUCKETS_PER_SEGMENT (MAX_ENTRIES_PER_SEGMENT / DIR_DEPTH)
#define DIR_SIZE_WIDTH 6
#define DIR_BLOCK_SIZES 4
#define DIR_BLOCK_SHIFT(_i) (3 * (_i))
#define DIR_BLOCK_SIZE(_i) (CACHE_BLOCK_SIZE << DIR_BLOCK_SHIFT(_i))
#define DIR_SIZE_WITH_BLOCK(_i) ((1 << DIR_SIZE_WIDTH) * DIR_BLOCK_SIZE(_i))
#define DIR_OFFSET_BITS 40
#define DIR_OFFSET_MAX ((((off_t)1) << DIR_OFFSET_BITS) - 1)
#define SYNC_MAX_WRITE (2 * 1024 * 1024)
#define SYNC_DELAY HRTIME_MSECONDS(500)
#define DO_NOT_REMOVE_THIS 0
// Debugging Options
//#define DO_CHECK_DIR_FAST
//#define DO_CHECK_DIR
// Macros
#ifdef DO_CHECK_DIR
#define CHECK_DIR(_d) ink_assert(check_dir(_d))
#else
#define CHECK_DIR(_d) ((void)0)
#endif
#define dir_index(_e, _i) ((Dir *)((char *)(_e)->dir + (SIZEOF_DIR * (_i))))
#define dir_assign(_e, _x) \
do { \
(_e)->w[0] = (_x)->w[0]; \
(_e)->w[1] = (_x)->w[1]; \
(_e)->w[2] = (_x)->w[2]; \
(_e)->w[3] = (_x)->w[3]; \
(_e)->w[4] = (_x)->w[4]; \
} while (0)
#define dir_assign_data(_e, _x) \
do { \
unsigned short next = dir_next(_e); \
dir_assign(_e, _x); \
dir_set_next(_e, next); \
} while (0)
// entry is valid
#define dir_valid(_d, _e) (_d->header->phase == dir_phase(_e) ? _d->vol_in_phase_valid(_e) : _d->vol_out_of_phase_valid(_e))
// entry is valid and outside of write aggregation region
#define dir_agg_valid(_d, _e) (_d->header->phase == dir_phase(_e) ? _d->vol_in_phase_valid(_e) : _d->vol_out_of_phase_agg_valid(_e))
// entry may be valid or overwritten in the last aggregated write
#define dir_write_valid(_d, _e) \
(_d->header->phase == dir_phase(_e) ? vol_in_phase_valid(_d, _e) : vol_out_of_phase_write_valid(_d, _e))
#define dir_agg_buf_valid(_d, _e) (_d->header->phase == dir_phase(_e) && _d->vol_in_phase_agg_buf_valid(_e))
#define dir_is_empty(_e) (!dir_offset(_e))
#define dir_clear(_e) \
do { \
(_e)->w[0] = 0; \
(_e)->w[1] = 0; \
(_e)->w[2] = 0; \
(_e)->w[3] = 0; \
(_e)->w[4] = 0; \
} while (0)
#define dir_clean(_e) dir_set_offset(_e, 0)
// OpenDir
#define OPEN_DIR_BUCKETS 256
struct EvacuationBlock;
typedef uint32_t DirInfo;
// Cache Directory
// INTERNAL: do not access these members directly, use the
// accessors below (e.g. dir_offset, dir_set_offset).
// These structures are stored in memory 2 byte aligned.
// The accessors prevent unaligned memory access which
// is often either less efficient or unsupported depending
// on the processor.
struct Dir {
#if DO_NOT_REMOVE_THIS
// THE BIT-FIELD INTERPRETATION OF THIS STRUCT WHICH HAS TO
// USE MACROS TO PREVENT UNALIGNED LOADS
// bits are numbered from lowest in u16 to highest
// always index as u16 to avoid byte order issues
unsigned int offset : 24; // (0,1:0-7) 16M * 512 = 8GB
unsigned int big : 2; // (1:8-9) 512 << (3 * big)
unsigned int size : 6; // (1:10-15) 6**2 = 64, 64*512 = 32768 .. 64*256=16MB
unsigned int tag : 12; // (2:0-11) 2048 / 8 entries/bucket = .4%
unsigned int phase : 1; // (2:12)
unsigned int head : 1; // (2:13) first segment in a document
unsigned int pinned : 1; // (2:14)
unsigned int token : 1; // (2:15)
unsigned int next : 16; // (3)
unsigned int offset_high : 16; // 8GB * 65k = 0.5PB (4)
#else
uint16_t w[5];
Dir() { dir_clear(this); }
#endif
};
// INTERNAL: do not access these members directly, use the
// accessors below (e.g. dir_offset, dir_set_offset)
struct FreeDir {
#if DO_NOT_REMOVE_THIS
// THE BIT-FIELD INTERPRETATION OF THIS STRUCT WHICH HAS TO
// USE MACROS TO PREVENT UNALIGNED LOADS
unsigned int offset : 24; // 0: empty
unsigned int reserved : 8;
unsigned int prev : 16; // (2)
unsigned int next : 16; // (3)
unsigned int offset_high : 16; // 0: empty
#else
uint16_t w[5];
FreeDir() { dir_clear(this); }
#endif
};
#define dir_offset(_e) \
((int64_t)(((uint64_t)(_e)->w[0]) | (((uint64_t)((_e)->w[1] & 0xFF)) << 16) | (((uint64_t)(_e)->w[4]) << 24)))
#define dir_set_offset(_e, _o) \
do { \
(_e)->w[0] = (uint16_t)_o; \
(_e)->w[1] = (uint16_t)((((_o) >> 16) & 0xFF) | ((_e)->w[1] & 0xFF00)); \
(_e)->w[4] = (uint16_t)((_o) >> 24); \
} while (0)
#define dir_bit(_e, _w, _b) ((uint32_t)(((_e)->w[_w] >> (_b)) & 1))
#define dir_set_bit(_e, _w, _b, _v) (_e)->w[_w] = (uint16_t)(((_e)->w[_w] & ~(1 << (_b))) | (((_v) ? 1 : 0) << (_b)))
#define dir_big(_e) ((uint32_t)((((_e)->w[1]) >> 8) & 0x3))
#define dir_set_big(_e, _v) (_e)->w[1] = (uint16_t)(((_e)->w[1] & 0xFCFF) | (((uint16_t)(_v)) & 0x3) << 8)
#define dir_size(_e) ((uint32_t)(((_e)->w[1]) >> 10))
#define dir_set_size(_e, _v) (_e)->w[1] = (uint16_t)(((_e)->w[1] & ((1 << 10) - 1)) | ((_v) << 10))
#define dir_set_approx_size(_e, _s) \
do { \
if ((_s) <= DIR_SIZE_WITH_BLOCK(0)) { \
dir_set_big(_e, 0); \
dir_set_size(_e, ((_s)-1) / DIR_BLOCK_SIZE(0)); \
} else if ((_s) <= DIR_SIZE_WITH_BLOCK(1)) { \
dir_set_big(_e, 1); \
dir_set_size(_e, ((_s)-1) / DIR_BLOCK_SIZE(1)); \
} else if ((_s) <= DIR_SIZE_WITH_BLOCK(2)) { \
dir_set_big(_e, 2); \
dir_set_size(_e, ((_s)-1) / DIR_BLOCK_SIZE(2)); \
} else { \
dir_set_big(_e, 3); \
dir_set_size(_e, ((_s)-1) / DIR_BLOCK_SIZE(3)); \
} \
} while (0)
#define dir_approx_size(_e) ((dir_size(_e) + 1) * DIR_BLOCK_SIZE(dir_big(_e)))
#define round_to_approx_dir_size(_s) \
(_s <= DIR_SIZE_WITH_BLOCK(0) ? \
ROUND_TO(_s, DIR_BLOCK_SIZE(0)) : \
(_s <= DIR_SIZE_WITH_BLOCK(1) ? \
ROUND_TO(_s, DIR_BLOCK_SIZE(1)) : \
(_s <= DIR_SIZE_WITH_BLOCK(2) ? ROUND_TO(_s, DIR_BLOCK_SIZE(2)) : ROUND_TO(_s, DIR_BLOCK_SIZE(3)))))
#define dir_tag(_e) ((uint32_t)((_e)->w[2] & ((1 << DIR_TAG_WIDTH) - 1)))
#define dir_set_tag(_e, _t) \
(_e)->w[2] = (uint16_t)(((_e)->w[2] & ~((1 << DIR_TAG_WIDTH) - 1)) | ((_t) & ((1 << DIR_TAG_WIDTH) - 1)))
#define dir_phase(_e) dir_bit(_e, 2, 12)
#define dir_set_phase(_e, _v) dir_set_bit(_e, 2, 12, _v)
#define dir_head(_e) dir_bit(_e, 2, 13)
#define dir_set_head(_e, _v) dir_set_bit(_e, 2, 13, _v)
#define dir_pinned(_e) dir_bit(_e, 2, 14)
#define dir_set_pinned(_e, _v) dir_set_bit(_e, 2, 14, _v)
#define dir_token(_e) dir_bit(_e, 2, 15)
#define dir_set_token(_e, _v) dir_set_bit(_e, 2, 15, _v)
#define dir_next(_e) (_e)->w[3]
#define dir_set_next(_e, _o) (_e)->w[3] = (uint16_t)(_o)
#define dir_prev(_e) (_e)->w[2]
#define dir_set_prev(_e, _o) (_e)->w[2] = (uint16_t)(_o)
// INKqa11166 - Cache can not store 2 HTTP alternates simultaneously.
// To allow this, move the vector from the CacheVC to the OpenDirEntry.
// Each CacheVC now maintains a pointer to this vector. Adding/Deleting
// alternates from this vector is done under the Vol::lock. The alternate
// is deleted/inserted into the vector just before writing the vector disk
// (CacheVC::updateVector).
LINK_FORWARD_DECLARATION(CacheVC, opendir_link) // forward declaration
struct OpenDirEntry {
DLL<CacheVC, Link_CacheVC_opendir_link> writers; // list of all the current writers
DLL<CacheVC, Link_CacheVC_opendir_link> readers; // list of all the current readers - not used
CacheHTTPInfoVector vector; // Vector for the http document. Each writer
// maintains a pointer to this vector and
// writes it down to disk.
CacheKey single_doc_key; // Key for the resident alternate.
Dir single_doc_dir; // Directory for the resident alternate
Dir first_dir; // Dir for the vector. If empty, a new dir is
// inserted, otherwise this dir is overwritten
uint16_t num_writers; // num of current writers
uint16_t max_writers; // max number of simultaneous writers allowed
bool dont_update_directory; // if set, the first_dir is not updated.
bool move_resident_alt; // if set, single_doc_dir is inserted.
bool reading_vec; // somebody is currently reading the vector
bool writing_vec; // somebody is currently writing the vector
LINK(OpenDirEntry, link);
int wait(CacheVC *c, int msec);
bool
has_multiple_writers()
{
return num_writers > 1;
}
};
struct OpenDir : public Continuation {
Queue<CacheVC, Link_CacheVC_opendir_link> delayed_readers;
DLL<OpenDirEntry> bucket[OPEN_DIR_BUCKETS];
int open_write(CacheVC *c, int allow_if_writers, int max_writers);
int close_write(CacheVC *c);
OpenDirEntry *open_read(const CryptoHash *key);
int signal_readers(int event, Event *e);
OpenDir();
};
struct CacheSync : public Continuation {
int vol_idx;
char *buf;
size_t buflen;
bool buf_huge;
off_t writepos;
AIOCallbackInternal io;
Event *trigger;
ink_hrtime start_time;
int mainEvent(int event, Event *e);
void aio_write(int fd, char *b, int n, off_t o);
CacheSync()
: Continuation(new_ProxyMutex()),
vol_idx(0),
buf(nullptr),
buflen(0),
buf_huge(false),
writepos(0),
trigger(nullptr),
start_time(0)
{
SET_HANDLER(&CacheSync::mainEvent);
}
};
// Global Functions
void vol_init_dir(Vol *d);
int dir_token_probe(const CacheKey *, Vol *, Dir *);
int dir_probe(const CacheKey *, Vol *, Dir *, Dir **);
int dir_insert(const CacheKey *key, Vol *d, Dir *to_part);
int dir_overwrite(const CacheKey *key, Vol *d, Dir *to_part, Dir *overwrite, bool must_overwrite = true);
int dir_delete(const CacheKey *key, Vol *d, Dir *del);
int dir_lookaside_probe(const CacheKey *key, Vol *d, Dir *result, EvacuationBlock **eblock);
int dir_lookaside_insert(EvacuationBlock *b, Vol *d, Dir *to);
int dir_lookaside_fixup(const CacheKey *key, Vol *d);
void dir_lookaside_cleanup(Vol *d);
void dir_lookaside_remove(const CacheKey *key, Vol *d);
void dir_free_entry(Dir *e, int s, Vol *d);
void dir_sync_init();
int check_dir(Vol *d);
void dir_clean_vol(Vol *d);
void dir_clear_range(off_t start, off_t end, Vol *d);
int dir_segment_accounted(int s, Vol *d, int offby = 0, int *free = nullptr, int *used = nullptr, int *empty = nullptr,
int *valid = nullptr, int *agg_valid = nullptr, int *avg_size = nullptr);
uint64_t dir_entries_used(Vol *d);
void sync_cache_dir_on_shutdown();
// Global Data
extern Dir empty_dir;
// Inline Funtions
#define dir_in_seg(_s, _i) ((Dir *)(((char *)(_s)) + (SIZEOF_DIR * (_i))))
TS_INLINE bool
dir_compare_tag(const Dir *e, const CacheKey *key)
{
return (dir_tag(e) == DIR_MASK_TAG(key->slice32(2)));
}
TS_INLINE Dir *
dir_from_offset(int64_t i, Dir *seg)
{
#if DIR_DEPTH < 5
if (!i) {
return nullptr;
}
return dir_in_seg(seg, i);
#else
i = i + ((i - 1) / (DIR_DEPTH - 1));
return dir_in_seg(seg, i);
#endif
}
TS_INLINE Dir *
next_dir(Dir *d, Dir *seg)
{
int i = dir_next(d);
return dir_from_offset(i, seg);
}
TS_INLINE int64_t
dir_to_offset(const Dir *d, const Dir *seg)
{
#if DIR_DEPTH < 5
return (((char *)d) - ((char *)seg)) / SIZEOF_DIR;
#else
int64_t i = (int64_t)((((char *)d) - ((char *)seg)) / SIZEOF_DIR);
i = i - (i / DIR_DEPTH);
return i;
#endif
}
TS_INLINE Dir *
dir_bucket(int64_t b, Dir *seg)
{
return dir_in_seg(seg, b * DIR_DEPTH);
}
TS_INLINE Dir *
dir_bucket_row(Dir *b, int64_t i)
{
return dir_in_seg(b, i);
}