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
|
/*-
* See the file LICENSE for redistribution information.
*
* Copyright (c) 1997, 2015 Oracle and/or its affiliates. All rights reserved.
*
* $Id$
*/
#include "db_config.h"
#include "db_int.h"
/*
* __db_zero_fill --
* Zero out bytes in the file.
*
* Pages allocated by writing pages past end-of-file are not zeroed,
* on some systems. Recovery could theoretically be fooled by a page
* showing up that contained garbage. In order to avoid this, we
* have to write the pages out to disk, and flush them. The reason
* for the flush is because if we don't sync, the allocation of another
* page subsequent to this one might reach the disk first, and if we
* crashed at the right moment, leave us with this page as the one
* allocated by writing a page past it in the file.
*
* PUBLIC: int __db_zero_fill __P((ENV *, DB_FH *));
*/
int
__db_zero_fill(env, fhp)
ENV *env;
DB_FH *fhp;
{
#ifdef HAVE_FILESYSTEM_NOTZERO
off_t stat_offset, write_offset;
size_t blen, nw;
u_int32_t bytes, mbytes;
int group_sync, ret;
u_int8_t *bp;
/* Calculate the byte offset of the next write. */
write_offset = (off_t)fhp->pgno * fhp->pgsize + fhp->offset;
/* Stat the file. */
if ((ret = __os_ioinfo(env, NULL, fhp, &mbytes, &bytes, NULL)) != 0)
return (ret);
stat_offset = (off_t)mbytes * MEGABYTE + bytes;
/* Check if the file is large enough. */
if (stat_offset >= write_offset)
return (0);
/* Get a large buffer if we're writing lots of data. */
#undef ZF_LARGE_WRITE
#define ZF_LARGE_WRITE (64 * 1024)
if ((ret = __os_calloc(env, 1, ZF_LARGE_WRITE, &bp)) != 0)
return (ret);
blen = ZF_LARGE_WRITE;
/* Seek to the current end of the file. */
if ((ret = __os_seek(env, fhp, mbytes, MEGABYTE, bytes)) != 0)
goto err;
/*
* Hash is the only access method that allocates groups of pages. Hash
* uses the existence of the last page in a group to signify the entire
* group is OK; so, write all the pages but the last one in the group,
* flush them to disk, then write the last one to disk and flush it.
*/
for (group_sync = 0; stat_offset < write_offset; group_sync = 1) {
if (write_offset - stat_offset <= (off_t)blen) {
blen = (size_t)(write_offset - stat_offset);
if (group_sync && (ret = __os_fsync(env, fhp)) != 0)
goto err;
}
if ((ret = __os_physwrite(env, fhp, bp, blen, &nw)) != 0)
goto err;
stat_offset += blen;
}
if ((ret = __os_fsync(env, fhp)) != 0)
goto err;
/* Seek back to where we started. */
mbytes = (u_int32_t)(write_offset / MEGABYTE);
bytes = (u_int32_t)(write_offset % MEGABYTE);
ret = __os_seek(env, fhp, mbytes, MEGABYTE, bytes);
err: __os_free(env, bp);
return (ret);
#else
COMPQUIET(env, NULL);
COMPQUIET(fhp, NULL);
return (0);
#endif /* HAVE_FILESYSTEM_NOTZERO */
}
/*
* __db_zero --
* Zero to the end of the file.
*
* PUBLIC: int __db_zero_extend __P((ENV *,
* PUBLIC: DB_FH *, db_pgno_t, db_pgno_t, u_int32_t));
*/
int
__db_zero_extend(env, fhp, pgno, last_pgno, pgsize)
ENV *env;
DB_FH *fhp;
db_pgno_t pgno, last_pgno;
u_int32_t pgsize;
{
int ret;
size_t nwrote;
u_int8_t *buf;
if ((ret = __os_calloc(env, 1, pgsize, &buf)) != 0)
return (ret);
memset(buf, 0, pgsize);
for (; pgno <= last_pgno; pgno++)
if ((ret = __os_io(env, DB_IO_WRITE,
fhp, pgno, pgsize, 0, pgsize, buf, &nwrote)) != 0) {
if (ret == 0) {
ret = EIO;
goto err;
}
goto err;
}
err: __os_free(env, buf);
return (ret);
}
|