Skip to content

Commit 4b71d28

Browse files
robertmhaasrobertmhaas
committed
Recursively fsync() the data directory after a crash.
Otherwise, if there's another crash, some writes from after the first crash might make it to disk while writes from before the crash fail to make it to disk. This could lead to data corruption. Back-patch to all supported versions. Abhijit Menon-Sen, reviewed by Andres Freund and slightly revised by me.
1 parent 1dadd36 commit 4b71d28

File tree

3 files changed

+158
-0
lines changed

3 files changed

+158
-0
lines changed

‎src/backend/access/transam/xlog.c

Lines changed: 41 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -678,6 +678,7 @@ static bool read_backup_label(XLogRecPtr *checkPointLoc,
678678
static void rm_redo_error_callback(void *arg);
679679
static int get_sync_bit(int method);
680680

681+
static void fsync_pgdata(char *datadir);
681682

682683
/*
683684
* Insert an XLOG record having the specified RMID and info bytes,
@@ -6229,6 +6230,18 @@ StartupXLOG(void)
62296230
(errmsg("database system was interrupted; last known up at %s",
62306231
str_time(ControlFile->time))));
62316232

6233+
/*
6234+
* If we previously crashed, there might be data which we had written,
6235+
* intending to fsync it, but which we had not actually fsync'd yet.
6236+
* Therefore, a power failure in the near future might cause earlier
6237+
* unflushed writes to be lost, even though more recent data written to
6238+
* disk from here on would be persisted. To avoid that, fsync the entire
6239+
* data directory.
6240+
*/
6241+
if (ControlFile->state != DB_SHUTDOWNED &&
6242+
ControlFile->state != DB_SHUTDOWNED_IN_RECOVERY)
6243+
fsync_pgdata(data_directory);
6244+
62326245
/* This is just to allow attaching to startup process with a debugger */
62336246
#ifdef XLOG_REPLAY_DELAY
62346247
if (ControlFile->state != DB_SHUTDOWNED)
@@ -10825,3 +10838,31 @@ WakeupRecovery(void)
1082510838
{
1082610839
SetLatch(&XLogCtl->recoveryWakeupLatch);
1082710840
}
10841+
10842+
/*
10843+
* Issue fsync recursively on PGDATA and all its contents.
10844+
*/
10845+
static void
10846+
fsync_pgdata(char *datadir)
10847+
{
10848+
if (!enableFsync)
10849+
return;
10850+
10851+
/*
10852+
* If possible, hint to the kernel that we're soon going to fsync
10853+
* the data directory and its contents.
10854+
*/
10855+
#if defined(HAVE_SYNC_FILE_RANGE) || \
10856+
(defined(USE_POSIX_FADVISE) && defined(POSIX_FADV_DONTNEED))
10857+
walkdir(datadir, pre_sync_fname);
10858+
#endif
10859+
10860+
/*
10861+
* Now we do the fsync()s in the same order.
10862+
*
10863+
* It's important to fsync the destination directory itself as individual
10864+
* file fsyncs don't guarantee that the directory entry for the file is
10865+
* synced.
10866+
*/
10867+
walkdir(datadir, fsync_fname);
10868+
}

‎src/backend/storage/file/fd.c

Lines changed: 115 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -2206,3 +2206,118 @@ looks_like_temp_rel_name(const char *name)
22062206
return false;
22072207
return true;
22082208
}
2209+
2210+
/*
2211+
* Hint to the OS that it should get ready to fsync() this file.
2212+
*
2213+
* Adapted from pre_sync_fname in initdb.c
2214+
*/
2215+
void
2216+
pre_sync_fname(char *fname, bool isdir)
2217+
{
2218+
int fd;
2219+
2220+
fd = open(fname, O_RDONLY | PG_BINARY);
2221+
2222+
/*
2223+
* Some OSs don't allow us to open directories at all (Windows returns
2224+
* EACCES)
2225+
*/
2226+
if (fd < 0 && isdir && (errno == EISDIR || errno == EACCES))
2227+
return;
2228+
2229+
if (fd < 0)
2230+
ereport(FATAL,
2231+
(errmsg("could not open file \"%s\" before fsync",
2232+
fname)));
2233+
2234+
pg_flush_data(fd, 0, 0);
2235+
2236+
close(fd);
2237+
}
2238+
2239+
/*
2240+
* walkdir: recursively walk a directory, applying the action to each
2241+
* regular file and directory (including the named directory itself)
2242+
* and following symbolic links.
2243+
*
2244+
* NB: There is another version of walkdir in initdb.c, but that version
2245+
* behaves differently with respect to symbolic links. Caveat emptor!
2246+
*/
2247+
void
2248+
walkdir(char *path, void (*action) (char *fname, bool isdir))
2249+
{
2250+
DIR *dir;
2251+
struct dirent *de;
2252+
2253+
dir = AllocateDir(path);
2254+
while ((de = ReadDir(dir, path)) != NULL)
2255+
{
2256+
char subpath[MAXPGPATH];
2257+
struct stat fst;
2258+
2259+
CHECK_FOR_INTERRUPTS();
2260+
2261+
if (strcmp(de->d_name, ".") == 0 ||
2262+
strcmp(de->d_name, "..") == 0)
2263+
continue;
2264+
2265+
snprintf(subpath, MAXPGPATH, "%s/%s", path, de->d_name);
2266+
2267+
if (lstat(subpath, &fst) < 0)
2268+
ereport(ERROR,
2269+
(errcode_for_file_access(),
2270+
errmsg("could not stat file \"%s\": %m", subpath)));
2271+
2272+
if (S_ISREG(fst.st_mode))
2273+
(*action) (subpath, false);
2274+
else if (S_ISDIR(fst.st_mode))
2275+
walkdir(subpath, action);
2276+
#ifndef WIN32
2277+
else if (S_ISLNK(fst.st_mode))
2278+
#else
2279+
else if (pg_win32_is_junction(subpath))
2280+
#endif
2281+
{
2282+
#if defined(HAVE_READLINK) || defined(WIN32)
2283+
char linkpath[MAXPGPATH];
2284+
int len;
2285+
struct stat lst;
2286+
2287+
len = readlink(subpath, linkpath, sizeof(linkpath)-1);
2288+
if (len < 0)
2289+
ereport(ERROR,
2290+
(errcode_for_file_access(),
2291+
errmsg("could not read symbolic link \"%s\": %m",
2292+
subpath)));
2293+
2294+
if (len >= sizeof(linkpath)-1)
2295+
ereport(ERROR,
2296+
(errmsg("symbolic link \"%s\" target is too long",
2297+
subpath)));
2298+
2299+
linkpath[len] = '\0';
2300+
2301+
if (lstat(linkpath, &lst) == 0)
2302+
{
2303+
if (S_ISREG(lst.st_mode))
2304+
(*action) (linkpath, false);
2305+
else if (S_ISDIR(lst.st_mode))
2306+
walkdir(subpath, action);
2307+
}
2308+
else if (errno != ENOENT)
2309+
ereport(ERROR,
2310+
(errcode_for_file_access(),
2311+
errmsg("could not stat file \"%s\": %m", linkpath)));
2312+
#else
2313+
ereport(WARNING,
2314+
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
2315+
errmsg("this platform does not support symbolic links; ignoring \"%s\"",
2316+
subpath)));
2317+
#endif
2318+
}
2319+
}
2320+
FreeDir(dir);
2321+
2322+
(*action) (path, true);
2323+
}

‎src/include/storage/fd.h

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -100,6 +100,8 @@ extern int pg_fsync_writethrough(int fd);
100100
extern int pg_fdatasync(int fd);
101101
extern int pg_flush_data(int fd, off_t offset, off_t amount);
102102
extern void fsync_fname(char *fname, bool isdir);
103+
extern void pre_sync_fname(char *fname, bool isdir);
104+
extern void walkdir(char *path, void (*action) (char *fname, bool isdir));
103105

104106
/* Filename components for OpenTemporaryFile */
105107
#define PG_TEMP_FILES_DIR "pgsql_tmp"

0 commit comments

Comments
 (0)
pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy