-
Notifications
You must be signed in to change notification settings - Fork 31
/
tuple.h
191 lines (169 loc) · 5.85 KB
/
tuple.h
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
#pragma once
#include "macros.h"
#include "varstr.h"
#include "dbcore/serial.h"
#include "dbcore/sm-object.h"
#include "dbcore/xid.h"
#ifdef SSN
// Indicate somebody has read this tuple and thought it was an old one
#define PERSISTENT_READER_MARK 0x1
#endif
namespace ermia {
/**
* A dbtuple is the type of value which we stick
* into underlying (non-transactional) data structures- it
* also contains the memory of the value
*/
struct dbtuple {
public:
#if defined(SSN) || defined(SSI)
TXN::readers_list::bitmap_t readers_bitmap; // bitmap of in-flight readers
fat_ptr sstamp; // successor (overwriter) stamp (\pi in ssn), set to writer
// XID during
// normal write to indicate its existence; become writer cstamp at commit
uint64_t xstamp; // access (reader) stamp (\eta), updated when reader commits
uint64_t preader; // did I have some reader thinking I'm old?
#endif
#ifdef MVOCC
fat_ptr sstamp;
#endif
#ifdef SSI
uint64_t s2; // smallest successor stamp of all reads performed by the tx
// that clobbered this version
// Consider a transaction T which clobbers this version, upon
// commit, T writes its cstamp in sstamp, and the smallest
// sstamp among all its reads in s2 of this version. This
// basically means T has clobbered this version, and meantime,
// some other transaction C clobbered T's read.
// So [X] r:w T r:w C. If anyone reads this version again,
// it will become the X in the dangerous structure above
// and must abort.
#endif
uint32_t size; // actual size of record
varstr *pvalue; // points to the value that will be put into value_start if
// committed
// so that read-my-own-update can copy from here.
uint8_t value_start[0]; // must be last field
dbtuple(uint32_t size)
:
#if defined(SSN) || defined(SSI)
sstamp(NULL_PTR),
xstamp(0),
preader(0),
#endif
#ifdef SSI
s2(0),
#endif
size(CheckBounds(size)),
pvalue(NULL) {
}
~dbtuple() {}
#if defined(SSN)
/* return the tuple's age based on a safe_lsn provided by the calling tx.
* safe_lsn usually = the calling tx's begin offset.
*
* and we need to check the type of clsn because a "committed" tx might
* change the clsn to ASI_LOG type after changing state.
*/
ALWAYS_INLINE uint64_t age(TXN::xid_context *visitor) {
uint64_t end = 0;
XID owner = volatile_read(visitor->owner);
retry:
fat_ptr cstamp = GetObject()->GetClsn();
if (cstamp.asi_type() == fat_ptr::ASI_XID) {
XID xid = XID::from_ptr(cstamp);
if (xid == owner) // my own update
return 0;
TXN::xid_context *xc = TXN::xid_get_context(xid);
end = volatile_read(xc->end);
if (not xc or xc->owner != xid) goto retry;
} else {
ASSERT(cstamp.asi_type() == fat_ptr::ASI_LOG);
end = cstamp.offset();
}
// the caller must be alive...
return volatile_read(visitor->begin) - end;
}
bool is_old(TXN::xid_context *visitor); // FOR READERS ONLY!
ALWAYS_INLINE bool set_persistent_reader() {
uint64_t pr = 0;
do {
pr = volatile_read(preader);
if (pr >> 7) // some updater already locked it
return false;
} while (
volatile_read(preader) != PERSISTENT_READER_MARK and
not __sync_bool_compare_and_swap(&preader, pr, PERSISTENT_READER_MARK));
return true;
}
#endif
#if defined(SSN)
ALWAYS_INLINE bool has_persistent_reader() {
return volatile_read(preader) & PERSISTENT_READER_MARK;
}
// XXX: for the writer who's updating this tuple only
ALWAYS_INLINE void lockout_read_mostly_tx() {
if (config::ssn_read_opt_enabled()) {
if (not(volatile_read(preader) >> 7))
__sync_fetch_and_xor(&preader, uint64_t{1} << 7);
ASSERT(volatile_read(preader) >> 7);
}
}
// XXX: for the writer who's updating this tuple only
ALWAYS_INLINE void welcome_read_mostly_tx() {
if (config::ssn_read_opt_enabled()) {
if (volatile_read(preader) >> 7)
__sync_fetch_and_xor(&preader, uint64_t{1} << 7);
ASSERT(not(volatile_read(preader) >> 7));
}
}
#endif
ALWAYS_INLINE uint8_t *get_value_start() { return &value_start[0]; }
ALWAYS_INLINE const uint8_t *get_value_start() const {
return &value_start[0];
}
inline Object *GetObject() {
Object *obj = (Object *)((char *)this - sizeof(Object));
ASSERT(obj->GetPayload() == (char *)this);
return obj;
}
inline dbtuple *NextVolatile() {
// So far this is only used by the primary
ALWAYS_ASSERT(!config::is_backup_srv());
Object *myobj = GetObject();
ASSERT(myobj->GetPayload() == (char *)this);
Object *next_obj = (Object*)myobj->GetNextVolatile().offset();
return next_obj ? next_obj->GetPinnedTuple() : nullptr;
}
private:
static ALWAYS_INLINE uint32_t CheckBounds(uint32_t s) {
ASSERT(s <= std::numeric_limits<uint32_t>::max());
return s;
}
public:
// Note: the stable=false option will try to read from pvalue,
// instead of the real data area; so giving stable=false is only
// safe for the updating transaction itself to read its own write.
inline rc_t DoRead(varstr *out_v, bool stable) const {
if (stable) {
out_v->p = get_value_start();
} else {
if (!pvalue) { // so I just deleted this tuple... return empty?
ASSERT(size == 0);
return rc_t{RC_FALSE};
}
out_v->p = pvalue->data();
ASSERT(pvalue->size() == size);
}
out_v->l = size;
return size > 0 ? rc_t{RC_TRUE} : rc_t{RC_FALSE};
}
// move data from the user's varstr pvalue to this tuple
inline void DoWrite() const {
if (pvalue) {
ASSERT(pvalue->size() == size);
memcpy((void *)get_value_start(), pvalue->data(), pvalue->size());
}
}
};
} // namespace ermia