/*
* Using the Hadoop connector requires proper setup of GUCs
* This procedure does sanity check on gp_hadoop_connector_jardir,
* gp_hadoop_target_version and gp_hadoop_home.
* It also update GUC gp_hadoop_connector_version for the current gp_hadoop_target_version.
*
* It checks the following:
* 1. $GPHOME/<gp_hadoop_jardir>/$GP_HADOOP_CONN_VERSION.jar must exists.
* 2. if gp_hadoop_home is set, then gp_hadoop_home must exists.
*/
static void
checkHadoopGUCs()
{
char gphome[MAXPGPATH];
StringInfoData path;
int jarFD;
/* Check the existence of $GPHOME/<gp_hadoop_jardir>/$GP_HADOOP_CONN_VERSION.jar
*
* To get $GPHOME, we go from my_exec_path, which is $GPHOME/bin/postgres, and
* go up 2 levels.
*
* Currently, gp_hadoop_connector_jardir is fixed. We look up $GP_HADOOP_CONN_VERSION
* using gp_hadoop_target_version.
*/
snprintf(gphome, sizeof(gphome), "%s", my_exec_path);
get_parent_directory(gphome);
get_parent_directory(gphome);
initStringInfoOfSize(&path, MAXPGPATH);
gp_hadoop_connector_version = (char*)getConnectorVersion();
appendStringInfo(&path, "%s/%s/%s.jar",
gphome, gp_hadoop_connector_jardir, gp_hadoop_connector_version);
jarFD = BasicOpenFile(path.data, O_RDONLY | PG_BINARY, 0);
if (jarFD == -1)
{
ereport(ERROR,
(errcode(ERRCODE_EXTERNAL_ROUTINE_EXCEPTION),
errmsg("cannot open Hadoop Cross Connect in %s: %m", path.data)));
}
close(jarFD);
/* Check the existence of gp_hadoop_home, if specified.
*
* If user has already specified $HADOOP_HOME in the env, then
* there's no need to setup this GUC.
*/
if (strlen(gp_hadoop_home)> 0)
{
int hdHomeFD = BasicOpenFile(gp_hadoop_home, O_RDONLY, 0);
if (hdHomeFD == -1)
{
ereport(ERROR,
(errcode(ERRCODE_EXTERNAL_ROUTINE_EXCEPTION),
errmsg("cannot open gp_hadoop_home in %s: %m", gp_hadoop_home)));
}
close(hdHomeFD);
}
}
/*
* createtbspc_abort_callback: Error cleanup callback for create-tablespace.
* This function should be executed only on successful creation of tablespace
* directory structure. This way we are sure that the directory and the symlink
* that we are removing are created by the same transaction, and are not
* pre-existing. Otherwise, we might delete any pre-existing directories.
*/
static void
createtbspc_abort_callback(bool isCommit, void *arg)
{
Oid tablespaceoid = *(Oid *) arg;
char *linkloc_with_version_dir;
char *linkloc;
struct stat st;
if (isCommit)
return;
linkloc_with_version_dir = palloc(9 + 1 + OIDCHARS + 1 +
strlen(PGXCNodeName) + 1 +
strlen(TABLESPACE_VERSION_DIRECTORY));
sprintf(linkloc_with_version_dir, "pg_tblspc/%u/%s_%s", tablespaceoid,
TABLESPACE_VERSION_DIRECTORY,
PGXCNodeName);
/* First, remove version directory */
if (rmdir(linkloc_with_version_dir) < 0)
{
ereport(WARNING,
(errcode_for_file_access(),
errmsg("could not remove directory \"%s\": %m",
linkloc_with_version_dir)));
pfree(linkloc_with_version_dir);
return;
}
/*
* Now remove the symlink.
* This has been borrowed from destroy_tablespace_directories().
*/
linkloc = pstrdup(linkloc_with_version_dir);
get_parent_directory(linkloc);
if (lstat(linkloc, &st) == 0 && S_ISDIR(st.st_mode))
{
/*
* We are here possibly because this is Windows, and lstat has identified
* the junction point as a directory.
*/
if (rmdir(linkloc) < 0)
ereport(WARNING,
(errcode_for_file_access(),
errmsg("could not remove directory \"%s\": %m",
linkloc)));
}
else
{
if (unlink(linkloc) < 0)
ereport(WARNING,
(errcode_for_file_access(),
errmsg("could not remove symbolic link \"%s\": %m",
linkloc)));
}
pfree(linkloc_with_version_dir);
pfree(linkloc);
}
int fopen(char *path, int flags)
{
if (!fs_initialized())
{
printf("fopen: file system is not initialized. use mkfs().\n");
return SYSERR;
}
int fd;
for (fd = 0; fd < NUM_FD; fd++)
if (oft[fd].state == O_CLOSED)
break;
if (fd >= NUM_FD)
{
printf("fopen: Can't open file, reached file descriptor limit (%d). "
"Close some files first.\n", NUM_FD);
return SYSERR;
}
directory parent_dir;
if (get_parent_directory(&fsd.root_dir, path, &parent_dir) == SYSERR)
{
printf("fopen: Can't get parent directory of %s\n.", path);
return SYSERR;
}
// now that we have parent directory, drop slashes and search the file in
// directory
char *filename = strrchr(path, '/');
if (filename == NULL)
filename = path;
else
filename++; // skip '/'
inode file_inode;
if (get_file_inode(&parent_dir, filename, &file_inode, INODE_TYPE_FILE) == SYSERR)
{
printf("fopen: Can't get file inode: %s\n", filename);
return SYSERR;
}
// Make sure the file is not already open
int i;
for (i = 0; i < NUM_FD; i++)
if (oft[i].state != O_CLOSED && oft[i].in.inode_idx == file_inode.inode_idx)
{
printf("fopen: Can't open the file, it is already open.\n");
return SYSERR;
}
oft[fd].in = file_inode;
oft[fd].state = flags;
oft[fd].cursor = 0;
return fd;
}
/*
* fsync_parent_path -- fsync the parent path of a file or directory
*
* This is aimed at making file operations persistent on disk in case of
* an OS crash or power failure.
*/
int
fsync_parent_path(const char *fname, const char *progname)
{
char parentpath[MAXPGPATH];
strlcpy(parentpath, fname, MAXPGPATH);
get_parent_directory(parentpath);
/*
* get_parent_directory() returns an empty string if the input argument is
* just a file name (see comments in path.c), so handle that as being the
* current directory.
*/
if (strlen(parentpath) == 0)
strlcpy(parentpath, ".", MAXPGPATH);
if (fsync_fname(parentpath, true, progname) != 0)
return -1;
return 0;
}
/*
* Try open file. Also create parent directries if open for writes.
*
* mode can contain 'R', that is same as 'r' but missing ok.
*/
FILE *
pgut_fopen(const char *path, const char *omode)
{
FILE *fp;
bool missing_ok = false;
char mode[16];
strlcpy(mode, omode, lengthof(mode));
if (mode[0] == 'R')
{
mode[0] = 'r';
missing_ok = true;
}
retry:
if ((fp = fopen(path, mode)) == NULL)
{
if (errno == ENOENT)
{
if (missing_ok)
return NULL;
if (mode[0] == 'w' || mode[0] == 'a')
{
char dir[MAXPGPATH];
strlcpy(dir, path, MAXPGPATH);
get_parent_directory(dir);
pgut_mkdir(dir);
goto retry;
}
}
ereport(ERROR,
(errcode_errno(),
errmsg("could not open file \"%s\": ", path)));
}
return fp;
}
void
nautilus_directory_notify_files_moved (GList *file_pairs)
{
GList *p, *affected_files, *node;
GFilePair *pair;
NautilusFile *file;
NautilusDirectory *old_directory, *new_directory;
GHashTable *parent_directories;
GList *new_files_list, *unref_list;
GHashTable *added_lists, *changed_lists;
char *name;
NautilusFileAttributes cancel_attributes;
GFile *to_location, *from_location;
/* Make a list of added and changed files in each directory. */
new_files_list = NULL;
added_lists = g_hash_table_new (NULL, NULL);
changed_lists = g_hash_table_new (NULL, NULL);
unref_list = NULL;
/* Make a list of parent directories that will need their counts updated. */
parent_directories = g_hash_table_new (NULL, NULL);
cancel_attributes = nautilus_file_get_all_attributes ();
for (p = file_pairs; p != NULL; p = p->next) {
pair = p->data;
from_location = pair->from;
to_location = pair->to;
/* Handle overwriting a file. */
file = nautilus_file_get_existing (to_location);
if (file != NULL) {
/* Mark it gone and prepare to send the changed signal. */
nautilus_file_mark_gone (file);
new_directory = file->details->directory;
hash_table_list_prepend (changed_lists,
new_directory,
file);
collect_parent_directories (parent_directories,
new_directory);
}
/* Update any directory objects that are affected. */
affected_files = nautilus_directory_moved_internal (from_location,
to_location);
for (node = affected_files; node != NULL; node = node->next) {
file = NAUTILUS_FILE (node->data);
hash_table_list_prepend (changed_lists,
file->details->directory,
file);
}
unref_list = g_list_concat (unref_list, affected_files);
/* Move an existing file. */
file = nautilus_file_get_existing (from_location);
if (file == NULL) {
/* Handle this as if it was a new file. */
new_files_list = g_list_prepend (new_files_list,
to_location);
} else {
/* Handle notification in the old directory. */
old_directory = file->details->directory;
collect_parent_directories (parent_directories, old_directory);
/* Cancel loading of attributes in the old directory */
nautilus_directory_cancel_loading_file_attributes
(old_directory, file, cancel_attributes);
/* Locate the new directory. */
new_directory = get_parent_directory (to_location);
collect_parent_directories (parent_directories, new_directory);
/* We can unref now -- new_directory is in the
* parent directories list so it will be
* around until the end of this function
* anyway.
*/
nautilus_directory_unref (new_directory);
/* Update the file's name and directory. */
name = g_file_get_basename (to_location);
nautilus_file_update_name_and_directory
(file, name, new_directory);
g_free (name);
/* Update file attributes */
nautilus_file_invalidate_attributes (file, NAUTILUS_FILE_ATTRIBUTE_INFO);
hash_table_list_prepend (changed_lists,
old_directory,
file);
if (old_directory != new_directory) {
hash_table_list_prepend (added_lists,
new_directory,
file);
}
/* Unref each file once to balance out nautilus_file_get_by_uri. */
unref_list = g_list_prepend (unref_list, file);
}
}
/* Now send out the changed and added signals for existing file objects. */
g_hash_table_foreach (changed_lists, call_files_changed_free_list, NULL);
g_hash_table_destroy (changed_lists);
g_hash_table_foreach (added_lists, call_files_added_free_list, NULL);
g_hash_table_destroy (added_lists);
/* Let the file objects go. */
nautilus_file_list_free (unref_list);
/* Invalidate count for each parent directory. */
g_hash_table_foreach (parent_directories, invalidate_count_and_unref, NULL);
g_hash_table_destroy (parent_directories);
/* Separate handling for brand new file objects. */
nautilus_directory_notify_files_added (new_files_list);
g_list_free (new_files_list);
}
/*
* Each database using a table space is isolated into its own name space
* by a subdirectory named for the database OID. On first creation of an
* object in the tablespace, create the subdirectory. If the subdirectory
* already exists, fall through quietly.
*
* isRedo indicates that we are creating an object during WAL replay.
* In this case we will cope with the possibility of the tablespace
* directory not being there either --- this could happen if we are
* replaying an operation on a table in a subsequently-dropped tablespace.
* We handle this by making a directory in the place where the tablespace
* symlink would normally be. This isn't an exact replay of course, but
* it's the best we can do given the available information.
*
* If tablespaces are not supported, we still need it in case we have to
* re-create a database subdirectory (of $PGDATA/base) during WAL replay.
*/
void
TablespaceCreateDbspace(Oid spcNode, Oid dbNode, bool isRedo)
{
struct stat st;
char *dir;
/*
* The global tablespace doesn't have per-database subdirectories, so
* nothing to do for it.
*/
if (spcNode == GLOBALTABLESPACE_OID)
return;
Assert(OidIsValid(spcNode));
Assert(OidIsValid(dbNode));
dir = GetDatabasePath(dbNode, spcNode);
if (stat(dir, &st) < 0)
{
/* Directory does not exist? */
if (errno == ENOENT)
{
/*
* Acquire TablespaceCreateLock to ensure that no DROP TABLESPACE
* or TablespaceCreateDbspace is running concurrently.
*/
LWLockAcquire(TablespaceCreateLock, LW_EXCLUSIVE);
/*
* Recheck to see if someone created the directory while we were
* waiting for lock.
*/
if (stat(dir, &st) == 0 && S_ISDIR(st.st_mode))
{
/* Directory was created */
}
else
{
/* Directory creation failed? */
if (mkdir(dir, S_IRWXU) < 0)
{
char *parentdir;
/* Failure other than not exists or not in WAL replay? */
if (errno != ENOENT || !isRedo)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not create directory \"%s\": %m",
dir)));
/*
* Parent directories are missing during WAL replay, so
* continue by creating simple parent directories rather
* than a symlink.
*/
/* create two parents up if not exist */
parentdir = pstrdup(dir);
get_parent_directory(parentdir);
get_parent_directory(parentdir);
/* Can't create parent and it doesn't already exist? */
if (mkdir(parentdir, S_IRWXU) < 0 && errno != EEXIST)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not create directory \"%s\": %m",
parentdir)));
pfree(parentdir);
/* create one parent up if not exist */
parentdir = pstrdup(dir);
get_parent_directory(parentdir);
/* Can't create parent and it doesn't already exist? */
if (mkdir(parentdir, S_IRWXU) < 0 && errno != EEXIST)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not create directory \"%s\": %m",
parentdir)));
pfree(parentdir);
/* Create database directory */
if (mkdir(dir, S_IRWXU) < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not create directory \"%s\": %m",
dir)));
}
}
LWLockRelease(TablespaceCreateLock);
}
else
{
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not stat directory \"%s\": %m", dir)));
}
}
else
{
/* Is it not a directory? */
if (!S_ISDIR(st.st_mode))
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("\"%s\" exists but is not a directory",
dir)));
}
pfree(dir);
}
/*
* destroy_tablespace_directories
*
* Attempt to remove filesystem infrastructure for the tablespace.
*
* 'redo' indicates we are redoing a drop from XLOG; in that case we should
* not throw an ERROR for problems, just LOG them. The worst consequence of
* not removing files here would be failure to release some disk space, which
* does not justify throwing an error that would require manual intervention
* to get the database running again.
*
* Returns TRUE if successful, FALSE if some subdirectory is not empty
*/
static bool
destroy_tablespace_directories(Oid tablespaceoid, bool redo)
{
char *linkloc;
char *linkloc_with_version_dir;
DIR *dirdesc;
struct dirent *de;
char *subfile;
struct stat st;
linkloc_with_version_dir = psprintf("pg_tblspc/%u/%s", tablespaceoid,
TABLESPACE_VERSION_DIRECTORY);
/*
* Check if the tablespace still contains any files. We try to rmdir each
* per-database directory we find in it. rmdir failure implies there are
* still files in that subdirectory, so give up. (We do not have to worry
* about undoing any already completed rmdirs, since the next attempt to
* use the tablespace from that database will simply recreate the
* subdirectory via TablespaceCreateDbspace.)
*
* Since we hold TablespaceCreateLock, no one else should be creating any
* fresh subdirectories in parallel. It is possible that new files are
* being created within subdirectories, though, so the rmdir call could
* fail. Worst consequence is a less friendly error message.
*
* If redo is true then ENOENT is a likely outcome here, and we allow it
* to pass without comment. In normal operation we still allow it, but
* with a warning. This is because even though ProcessUtility disallows
* DROP TABLESPACE in a transaction block, it's possible that a previous
* DROP failed and rolled back after removing the tablespace directories
* and/or symlink. We want to allow a new DROP attempt to succeed at
* removing the catalog entries (and symlink if still present), so we
* should not give a hard error here.
*/
dirdesc = AllocateDir(linkloc_with_version_dir);
if (dirdesc == NULL)
{
if (errno == ENOENT)
{
if (!redo)
ereport(WARNING,
(errcode_for_file_access(),
errmsg("could not open directory \"%s\": %m",
linkloc_with_version_dir)));
/* The symlink might still exist, so go try to remove it */
goto remove_symlink;
}
else if (redo)
{
/* in redo, just log other types of error */
ereport(LOG,
(errcode_for_file_access(),
errmsg("could not open directory \"%s\": %m",
linkloc_with_version_dir)));
pfree(linkloc_with_version_dir);
return false;
}
/* else let ReadDir report the error */
}
while ((de = ReadDir(dirdesc, linkloc_with_version_dir)) != NULL)
{
if (strcmp(de->d_name, ".") == 0 ||
strcmp(de->d_name, "..") == 0)
continue;
subfile = psprintf("%s/%s", linkloc_with_version_dir, de->d_name);
/* This check is just to deliver a friendlier error message */
if (!redo && !directory_is_empty(subfile))
{
FreeDir(dirdesc);
pfree(subfile);
pfree(linkloc_with_version_dir);
return false;
}
/* remove empty directory */
if (rmdir(subfile) < 0)
ereport(redo ? LOG : ERROR,
(errcode_for_file_access(),
errmsg("could not remove directory \"%s\": %m",
subfile)));
pfree(subfile);
}
FreeDir(dirdesc);
/* remove version directory */
if (rmdir(linkloc_with_version_dir) < 0)
{
ereport(redo ? LOG : ERROR,
(errcode_for_file_access(),
errmsg("could not remove directory \"%s\": %m",
linkloc_with_version_dir)));
pfree(linkloc_with_version_dir);
return false;
}
/*
* Try to remove the symlink. We must however deal with the possibility
* that it's a directory instead of a symlink --- this could happen during
* WAL replay (see TablespaceCreateDbspace), and it is also the case on
* Windows where junction points lstat() as directories.
*
* Note: in the redo case, we'll return true if this final step fails;
* there's no point in retrying it. Also, ENOENT should provoke no more
* than a warning.
*/
remove_symlink:
linkloc = pstrdup(linkloc_with_version_dir);
get_parent_directory(linkloc);
if (lstat(linkloc, &st) < 0)
{
int saved_errno = errno;
ereport(redo ? LOG : (saved_errno == ENOENT ? WARNING : ERROR),
(errcode_for_file_access(),
errmsg("could not stat file \"%s\": %m",
linkloc)));
}
else if (S_ISDIR(st.st_mode))
{
if (rmdir(linkloc) < 0)
{
int saved_errno = errno;
ereport(redo ? LOG : (saved_errno == ENOENT ? WARNING : ERROR),
(errcode_for_file_access(),
errmsg("could not remove directory \"%s\": %m",
linkloc)));
}
}
#ifdef S_ISLNK
else if (S_ISLNK(st.st_mode))
{
if (unlink(linkloc) < 0)
{
int saved_errno = errno;
ereport(redo ? LOG : (saved_errno == ENOENT ? WARNING : ERROR),
(errcode_for_file_access(),
errmsg("could not remove symbolic link \"%s\": %m",
linkloc)));
}
}
#endif
else
{
/* Refuse to remove anything that's not a directory or symlink */
ereport(redo ? LOG : ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("\"%s\" is not a directory or symbolic link",
linkloc)));
}
pfree(linkloc_with_version_dir);
pfree(linkloc);
return true;
}
/**
* @brief Entry point for pg_bulkload command.
*
* Flow:
* <ol>
* <li> Parses command arguments. </li>
* <li> Without -r option: Starts the loading. </li>
* <li> With -r option: Starts the recovery. </li>
* </ol>
*
* @param argc [in] Number of arguments.
* @param argv [in] Argument list.
* @return Returns zero if successful, 1 otherwise.
*/
int
main(int argc, char *argv[])
{
char cwd[MAXPGPATH];
char control_file[MAXPGPATH] = "";
int i;
pgut_init(argc, argv);
if (argc < 2)
{
help(false);
return E_PG_OTHER;
}
if (getcwd(cwd, MAXPGPATH) == NULL)
ereport(ERROR,
(errcode(EXIT_FAILURE),
errmsg("cannot read current directory: ")));
i = pgut_getopt(argc, argv, options);
for (; i < argc; i++)
{
if (control_file[0])
ereport(ERROR,
(errcode(EXIT_FAILURE),
errmsg("too many arguments")));
/* make absolute control file path */
if (is_absolute_path(argv[i]))
strlcpy(control_file, argv[i], MAXPGPATH);
else
join_path_components(control_file, cwd, argv[i]);
canonicalize_path(control_file);
}
/*
* Determines data loading or recovery.
*/
if (recovery)
{
/* verify arguments */
if (!DataDir && (DataDir = getenv("PGDATA")) == NULL)
elog(ERROR, "no $PGDATA specified");
if (strlen(DataDir) + MAX_LOADSTATUS_NAME >= MAXPGPATH)
elog(ERROR, "too long $PGDATA path length");
if (control_file[0] != '\0')
elog(ERROR, "invalid argument 'control file' for recovery");
return LoaderRecoveryMain();
}
else
{
/* verify arguments */
if (DataDir)
elog(ERROR, "invalid option '-D' for data load");
if (control_file[0])
bulkload_options = list_concat(
ParseControlFile(control_file), bulkload_options);
/* chdir control_file to the parent directory */
get_parent_directory(control_file);
/* add path options */
for (i = 0; i < NUM_PATH_OPTIONS; i++)
{
const pgut_option *opt = &options[i];
const char *path = *(const char **) opt->var;
char abspath[MAXPGPATH];
char item[MAXPGPATH + 32];
if (path == NULL)
continue;
if ((i == 0 || i == 1) &&
(pg_strcasecmp(path, "stdin") == 0 || type_function))
{
/* special case for stdin and input from function */
strlcpy(abspath, path, lengthof(abspath));
}
else if (is_absolute_path(path) || (i == 2 && !writer_binary))
{
/* absolute path */
strlcpy(abspath, path, lengthof(abspath));
}
else if (opt->source == SOURCE_FILE)
{
/* control file relative path */
join_path_components(abspath, control_file, path);
}
else
{
/* current working directory relative path */
join_path_components(abspath, cwd, path);
}
canonicalize_path(abspath);
snprintf(item, lengthof(item), "%s=%s", opt->lname, abspath);
bulkload_options = lappend(bulkload_options, pgut_strdup(item));
}
return LoaderLoadMain(bulkload_options);
}
}
/*
* Each database using a table space is isolated into its own name space
* by a subdirectory named for the database OID. On first creation of an
* object in the tablespace, create the subdirectory. If the subdirectory
* already exists, just fall through quietly.
*
* If tablespaces are not supported, this is just a no-op; CREATE DATABASE
* is expected to create the default subdirectory for the database.
*
* isRedo indicates that we are creating an object during WAL replay;
* we can skip doing locking in that case (and should do so to avoid
* any possible problems with pg_tablespace not being valid).
*
* Also, when isRedo is true, we will cope with the possibility of the
* tablespace not being there either --- this could happen if we are
* replaying an operation on a table in a subsequently-dropped tablespace.
* We handle this by making a directory in the place where the tablespace
* symlink would normally be. This isn't an exact replay of course, but
* it's the best we can do given the available information.
*/
void
TablespaceCreateDbspace(Oid spcNode, Oid dbNode, bool isRedo)
{
#ifdef HAVE_SYMLINK
struct stat st;
char *dir;
/*
* The global tablespace doesn't have per-database subdirectories, so
* nothing to do for it.
*/
if (spcNode == GLOBALTABLESPACE_OID)
return;
Assert(OidIsValid(spcNode));
Assert(OidIsValid(dbNode));
dir = GetDatabasePath(dbNode, spcNode);
if (stat(dir, &st) < 0)
{
if (errno == ENOENT)
{
/*
* Acquire ExclusiveLock on pg_tablespace to ensure that no DROP
* TABLESPACE or TablespaceCreateDbspace is running concurrently.
* Simple reads from pg_tablespace are OK.
*/
Relation rel;
if (!isRedo)
rel = heap_open(TableSpaceRelationId, ExclusiveLock);
else
rel = NULL;
/*
* Recheck to see if someone created the directory while we were
* waiting for lock.
*/
if (stat(dir, &st) == 0 && S_ISDIR(st.st_mode))
{
/* need not do anything */
}
else
{
/* OK, go for it */
if (mkdir(dir, S_IRWXU) < 0)
{
char *parentdir;
if (errno != ENOENT || !isRedo)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not create directory \"%s\": %m",
dir)));
/* Try to make parent directory too */
parentdir = pstrdup(dir);
get_parent_directory(parentdir);
if (mkdir(parentdir, S_IRWXU) < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not create directory \"%s\": %m",
parentdir)));
pfree(parentdir);
if (mkdir(dir, S_IRWXU) < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not create directory \"%s\": %m",
dir)));
}
}
/* OK to drop the exclusive lock */
if (!isRedo)
heap_close(rel, ExclusiveLock);
}
else
{
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not stat directory \"%s\": %m", dir)));
}
}
else
{
/* be paranoid */
if (!S_ISDIR(st.st_mode))
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("\"%s\" exists but is not a directory",
dir)));
}
pfree(dir);
#endif /* HAVE_SYMLINK */
}
/*
* destroy_tablespace_directories
*
* Attempt to remove filesystem infrastructure
*
* 'redo' indicates we are redoing a drop from XLOG; okay if nothing there
*
* Returns TRUE if successful, FALSE if some subdirectory is not empty
*/
static bool
destroy_tablespace_directories(Oid tablespaceoid, bool redo)
{
char *linkloc;
char *linkloc_with_version_dir;
DIR *dirdesc;
struct dirent *de;
char *subfile;
struct stat st;
linkloc_with_version_dir = palloc(9 + 1 + OIDCHARS + 1 +
strlen(TABLESPACE_VERSION_DIRECTORY));
sprintf(linkloc_with_version_dir, "pg_tblspc/%u/%s", tablespaceoid,
TABLESPACE_VERSION_DIRECTORY);
/*
* Check if the tablespace still contains any files. We try to rmdir each
* per-database directory we find in it. rmdir failure implies there are
* still files in that subdirectory, so give up. (We do not have to worry
* about undoing any already completed rmdirs, since the next attempt to
* use the tablespace from that database will simply recreate the
* subdirectory via TablespaceCreateDbspace.)
*
* Since we hold TablespaceCreateLock, no one else should be creating any
* fresh subdirectories in parallel. It is possible that new files are
* being created within subdirectories, though, so the rmdir call could
* fail. Worst consequence is a less friendly error message.
*
* If redo is true then ENOENT is a likely outcome here, and we allow it
* to pass without comment. In normal operation we still allow it, but
* with a warning. This is because even though ProcessUtility disallows
* DROP TABLESPACE in a transaction block, it's possible that a previous
* DROP failed and rolled back after removing the tablespace directories
* and symlink. We want to allow a new DROP attempt to succeed at
* removing the catalog entries, so we should not give a hard error here.
*/
dirdesc = AllocateDir(linkloc_with_version_dir);
if (dirdesc == NULL)
{
if (errno == ENOENT)
{
if (!redo)
ereport(WARNING,
(errcode_for_file_access(),
errmsg("could not open directory \"%s\": %m",
linkloc_with_version_dir)));
pfree(linkloc_with_version_dir);
return true;
}
/* else let ReadDir report the error */
}
while ((de = ReadDir(dirdesc, linkloc_with_version_dir)) != NULL)
{
if (strcmp(de->d_name, ".") == 0 ||
strcmp(de->d_name, "..") == 0)
continue;
subfile = palloc(strlen(linkloc_with_version_dir) + 1 + strlen(de->d_name) + 1);
sprintf(subfile, "%s/%s", linkloc_with_version_dir, de->d_name);
/* This check is just to deliver a friendlier error message */
if (!directory_is_empty(subfile))
{
FreeDir(dirdesc);
pfree(subfile);
pfree(linkloc_with_version_dir);
return false;
}
/* remove empty directory */
if (rmdir(subfile) < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not remove directory \"%s\": %m",
subfile)));
pfree(subfile);
}
FreeDir(dirdesc);
/* remove version directory */
if (rmdir(linkloc_with_version_dir) < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not remove directory \"%s\": %m",
linkloc_with_version_dir)));
/*
* Try to remove the symlink. We must however deal with the possibility
* that it's a directory instead of a symlink --- this could happen during
* WAL replay (see TablespaceCreateDbspace), and it is also the case on
* Windows where junction points lstat() as directories.
*/
linkloc = pstrdup(linkloc_with_version_dir);
get_parent_directory(linkloc);
if (lstat(linkloc, &st) == 0 && S_ISDIR(st.st_mode))
{
if (rmdir(linkloc) < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not remove directory \"%s\": %m",
linkloc)));
}
else
{
if (unlink(linkloc) < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not remove symbolic link \"%s\": %m",
linkloc)));
}
pfree(linkloc_with_version_dir);
pfree(linkloc);
return true;
}
static int
ReadStringFromToast(const char *buffer,
unsigned int buff_size,
unsigned int* out_size)
{
int result = 0;
/* If toasted value is on disk, we'll try to restore it. */
if (VARATT_IS_EXTERNAL_ONDISK(buffer))
{
varatt_external toast_ptr;
char *toast_data = NULL;
/* Number of chunks the TOAST data is divided into */
int32 num_chunks;
/* Actual size of external TOASTed value */
int32 toast_ext_size;
/* Path to directory with TOAST realtion file */
char *toast_relation_path;
/* Filename of TOAST relation file */
char toast_relation_filename[MAXPGPATH];
FILE *toast_rel_fp;
unsigned int block_options = 0;
unsigned int control_options = 0;
VARATT_EXTERNAL_GET_POINTER(toast_ptr, buffer);
printf(" TOAST value. Raw size: %8d, external size: %8d, "
"value id: %6d, toast relation id: %6d\n",
toast_ptr.va_rawsize,
toast_ptr.va_extsize,
toast_ptr.va_valueid,
toast_ptr.va_toastrelid);
/* Extract TOASTed value */
toast_ext_size = toast_ptr.va_extsize;
num_chunks = (toast_ext_size - 1) / TOAST_MAX_CHUNK_SIZE + 1;
printf(" Number of chunks: %d\n", num_chunks);
/* Open TOAST relation file */
toast_relation_path = strdup(fileName);
get_parent_directory(toast_relation_path);
sprintf(toast_relation_filename, "%s/%d", toast_relation_path,
toast_ptr.va_toastrelid);
printf(" Read TOAST relation %s\n", toast_relation_filename);
toast_rel_fp = fopen(toast_relation_filename, "rb");
if (!toast_rel_fp) {
printf("Cannot open TOAST relation %s\n",
toast_relation_filename);
result = -1;
}
if (result == 0)
{
unsigned int toast_relation_block_size = GetBlockSize(toast_rel_fp);
fseek(toast_rel_fp, 0, SEEK_SET);
toast_data = malloc(toast_ptr.va_rawsize);
result = DumpFileContents(block_options,
control_options,
toast_rel_fp,
toast_relation_block_size,
-1, /* no start block */
-1, /* no end block */
true, /* is toast relation */
toast_ptr.va_valueid,
toast_ptr.va_extsize,
toast_data);
if (result == 0)
{
if (VARATT_EXTERNAL_IS_COMPRESSED(toast_ptr))
result = DumpCompressedString(toast_data, toast_ext_size);
else
CopyAppendEncode(toast_data, toast_ext_size);
}
else
{
printf("Error in TOAST file.\n");
}
free(toast_data);
}
fclose(toast_rel_fp);
free(toast_relation_path);
}
/* If tag is indirect or expanded, it was stored in memory. */
else
{
CopyAppend("(TOASTED IN MEMORY)");
}
return result;
}
int
main(int argc, char *argv[])
{
/*
* options with no short version return a low integer, the rest return
* their short version value
*/
static struct option long_options[] = {
{"pgdata", required_argument, NULL, 'D'},
{"help", no_argument, NULL, '?'},
{"version", no_argument, NULL, 'V'},
{"debug", no_argument, NULL, 'd'},
{"show", no_argument, NULL, 's'},
{"noclean", no_argument, NULL, 'n'},
{NULL, 0, NULL, 0}
};
int c, ret;
int option_index;
char *effective_user;
char bin_dir[MAXPGPATH];
char *pg_data_native;
bool node_type_specified = false;
progname = get_progname(argv[0]);
set_pglocale_pgservice(argv[0], PG_TEXTDOMAIN("initgtm"));
if (argc > 1)
{
if (strcmp(argv[1], "--help") == 0 || strcmp(argv[1], "-?") == 0)
{
usage(progname);
exit(0);
}
if (strcmp(argv[1], "--version") == 0 || strcmp(argv[1], "-V") == 0)
{
puts("initgtm (Postgres-XL) " PGXC_VERSION);
exit(0);
}
}
/* process command-line options */
while ((c = getopt_long(argc, argv, "dD:nsZ:", long_options, &option_index)) != -1)
{
switch (c)
{
case 'D':
pg_data = xstrdup(optarg);
break;
case 'd':
debug = true;
printf(_("Running in debug mode.\n"));
break;
case 'n':
noclean = true;
printf(_("Running in noclean mode. Mistakes will not be cleaned up.\n"));
break;
case 's':
show_setting = true;
break;
case 'Z':
if (strcmp(xstrdup(optarg), "gtm") == 0)
is_gtm = true;
else if (strcmp(xstrdup(optarg), "gtm_proxy") == 0)
is_gtm = false;
else
{
fprintf(stderr, _("Try \"%s --help\" for more information.\n"),
progname);
exit(1);
}
node_type_specified = true;
break;
default:
/* getopt_long already emitted a complaint */
fprintf(stderr, _("Try \"%s --help\" for more information.\n"),
progname);
exit(1);
}
}
/* Non-option argument specifies data directory */
if (optind < argc)
{
pg_data = xstrdup(argv[optind]);
optind++;
}
if (optind < argc)
{
fprintf(stderr, _("%s: too many command-line arguments (first is \"%s\")\n"),
progname, argv[optind + 1]);
fprintf(stderr, _("Try \"%s --help\" for more information.\n"),
progname);
exit(1);
}
/* Check on definition of GTM data folder */
if (strlen(pg_data) == 0)
{
fprintf(stderr,
_("%s: no data directory specified\n"
"You must identify the directory where the data for this GTM system\n"
"will reside. Do this with either the invocation option -D or the\n"
"environment variable PGDATA.\n"),
progname);
exit(1);
}
if (!node_type_specified)
{
fprintf(stderr,
_("%s: no node type specified\n"
"You must identify the node type chosen for initialization.\n"
"Do this with the invocation option -Z by choosing \"gtm\" or"
"\"gtm_proxy\"\n"),
progname);
exit(1);
}
pg_data_native = pg_data;
canonicalize_path(pg_data);
#ifdef WIN32
/*
* Before we execute another program, make sure that we are running with a
* restricted token. If not, re-execute ourselves with one.
*/
if ((restrict_env = getenv("PG_RESTRICT_EXEC")) == NULL
|| strcmp(restrict_env, "1") != 0)
{
PROCESS_INFORMATION pi;
char *cmdline;
ZeroMemory(&pi, sizeof(pi));
cmdline = xstrdup(GetCommandLine());
putenv("PG_RESTRICT_EXEC=1");
if (!CreateRestrictedProcess(cmdline, &pi))
{
fprintf(stderr, "Failed to re-exec with restricted token: %lu.\n", GetLastError());
}
else
{
/*
* Successfully re-execed. Now wait for child process to capture
* exitcode.
*/
DWORD x;
CloseHandle(pi.hThread);
WaitForSingleObject(pi.hProcess, INFINITE);
if (!GetExitCodeProcess(pi.hProcess, &x))
{
fprintf(stderr, "Failed to get exit code from subprocess: %lu\n", GetLastError());
exit(1);
}
exit(x);
}
}
#endif
/* Like for initdb, check if a valid version of Postgres is running */
if ((ret = find_other_exec(argv[0], "postgres", PG_BACKEND_VERSIONSTR,
backend_exec)) < 0)
{
char full_path[MAXPGPATH];
if (find_my_exec(argv[0], full_path) < 0)
strlcpy(full_path, progname, sizeof(full_path));
if (ret == -1)
fprintf(stderr,
_("The program \"postgres\" is needed by %s "
"but was not found in the\n"
"same directory as \"%s\".\n"
"Check your installation.\n"),
progname, full_path);
else
fprintf(stderr,
_("The program \"postgres\" was found by \"%s\"\n"
"but was not the same version as %s.\n"
"Check your installation.\n"),
full_path, progname);
exit(1);
}
/* store binary directory */
strcpy(bin_path, backend_exec);
*last_dir_separator(bin_path) = '\0';
canonicalize_path(bin_path);
if (!share_path)
{
share_path = pg_malloc(MAXPGPATH);
get_share_path(backend_exec, share_path);
}
else if (!is_absolute_path(share_path))
{
fprintf(stderr, _("%s: input file location must be an absolute path\n"), progname);
exit(1);
}
canonicalize_path(share_path);
effective_user = get_id();
/* Take into account GTM and GTM-proxy cases */
if (is_gtm)
set_input(&conf_file, "gtm.conf.sample");
else
set_input(&conf_file, "gtm_proxy.conf.sample");
if (show_setting || debug)
{
fprintf(stderr,
"VERSION=%s\n"
"GTMDATA=%s\nshare_path=%s\nGTMPATH=%s\n"
"GTM_CONF_SAMPLE=%s\n",
PGXC_VERSION,
pg_data, share_path, bin_path,
conf_file);
if (show_setting)
exit(0);
}
check_input(conf_file);
printf(_("The files belonging to this GTM system will be owned "
"by user \"%s\".\n"
"This user must also own the server process.\n\n"),
effective_user);
printf("\n");
umask(S_IRWXG | S_IRWXO);
/*
* now we are starting to do real work, trap signals so we can clean up
*/
/* some of these are not valid on Windows */
#ifdef SIGHUP
pqsignal(SIGHUP, trapsig);
#endif
#ifdef SIGINT
pqsignal(SIGINT, trapsig);
#endif
#ifdef SIGQUIT
pqsignal(SIGQUIT, trapsig);
#endif
#ifdef SIGTERM
pqsignal(SIGTERM, trapsig);
#endif
/* Ignore SIGPIPE when writing to backend, so we can clean up */
#ifdef SIGPIPE
pqsignal(SIGPIPE, SIG_IGN);
#endif
switch (pg_check_dir(pg_data))
{
case 0:
/* PGDATA not there, must create it */
printf(_("creating directory %s ... "),
pg_data);
fflush(stdout);
if (!mkdatadir(NULL))
exit_nicely();
else
check_ok();
made_new_pgdata = true;
break;
case 1:
/* Present but empty, fix permissions and use it */
printf(_("fixing permissions on existing directory %s ... "),
pg_data);
fflush(stdout);
if (chmod(pg_data, S_IRWXU) != 0)
{
fprintf(stderr, _("%s: could not change permissions of directory \"%s\": %s\n"),
progname, pg_data, strerror(errno));
exit_nicely();
}
else
check_ok();
found_existing_pgdata = true;
break;
case 2:
/* Present and not empty */
fprintf(stderr,
_("%s: directory \"%s\" exists but is not empty\n"),
progname, pg_data);
fprintf(stderr,
_("If you want to create a new GTM system, either remove or empty\n"
"the directory \"%s\" or run %s\n"
"with an argument other than \"%s\".\n"),
pg_data, progname, pg_data);
exit(1); /* no further message needed */
default:
/* Trouble accessing directory */
fprintf(stderr, _("%s: could not access directory \"%s\": %s\n"),
progname, pg_data, strerror(errno));
exit_nicely();
}
/* Select suitable configuration settings */
set_null_conf();
/* Now create all the text config files */
setup_config();
/* Get directory specification used to start this executable */
strcpy(bin_dir, argv[0]);
get_parent_directory(bin_dir);
if (is_gtm)
printf(_("\nSuccess. You can now start the GTM server using:\n\n"
" %s%s%sgtm%s -D %s%s%s\n"
"or\n"
" %s%s%sgtm_ctl%s -Z gtm -D %s%s%s -l logfile start\n\n"),
QUOTE_PATH, bin_dir, (strlen(bin_dir) > 0) ? DIR_SEP : "", QUOTE_PATH,
QUOTE_PATH, pg_data_native, QUOTE_PATH,
QUOTE_PATH, bin_dir, (strlen(bin_dir) > 0) ? DIR_SEP : "", QUOTE_PATH,
QUOTE_PATH, pg_data_native, QUOTE_PATH);
else
printf(_("\nSuccess. You can now start the GTM proxy server using:\n\n"
" %s%s%sgtm_proxy%s -D %s%s%s\n"
"or\n"
" %s%s%sgtm_ctl%s -Z gtm_proxy -D %s%s%s -l logfile start\n\n"),
QUOTE_PATH, bin_dir, (strlen(bin_dir) > 0) ? DIR_SEP : "", QUOTE_PATH,
QUOTE_PATH, pg_data_native, QUOTE_PATH,
QUOTE_PATH, bin_dir, (strlen(bin_dir) > 0) ? DIR_SEP : "", QUOTE_PATH,
QUOTE_PATH, pg_data_native, QUOTE_PATH);
return 0;
}
CIRCA_EXPORT void circa_get_parent_directory(Value* filename, Value* result)
{
get_parent_directory(filename, result);
}
/*
* Each database using a table space is isolated into its own name space
* by a subdirectory named for the database OID. On first creation of an
* object in the tablespace, create the subdirectory. If the subdirectory
* already exists, just fall through quietly.
*
* isRedo indicates that we are creating an object during WAL replay.
* In this case we will cope with the possibility of the tablespace
* directory not being there either --- this could happen if we are
* replaying an operation on a table in a subsequently-dropped tablespace.
* We handle this by making a directory in the place where the tablespace
* symlink would normally be. This isn't an exact replay of course, but
* it's the best we can do given the available information.
*
* If tablespaces are not supported, you might think this could be a no-op,
* but you'd be wrong: we still need it in case we have to re-create a
* database subdirectory (of $PGDATA/base) during WAL replay.
*/
void
TablespaceCreateDbspace(Oid spcNode, Oid dbNode, bool isRedo)
{
struct stat st;
char *dir;
/*
* The global tablespace doesn't have per-database subdirectories, so
* nothing to do for it.
*/
if (spcNode == GLOBALTABLESPACE_OID)
return;
Assert(OidIsValid(spcNode));
Assert(OidIsValid(dbNode));
dir = GetDatabasePath(dbNode, spcNode);
if (stat(dir, &st) < 0)
{
if (errno == ENOENT)
{
/*
* Acquire TablespaceCreateLock to ensure that no DROP TABLESPACE
* or TablespaceCreateDbspace is running concurrently.
*/
LWLockAcquire(TablespaceCreateLock, LW_EXCLUSIVE);
/*
* Recheck to see if someone created the directory while we were
* waiting for lock.
*/
if (stat(dir, &st) == 0 && S_ISDIR(st.st_mode))
{
/* need not do anything */
}
else
{
/* OK, go for it */
if (mkdir(dir, S_IRWXU) < 0)
{
char *parentdir;
if (errno != ENOENT || !isRedo)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not create directory \"%s\": %m",
dir)));
/* Try to make parent directory too */
parentdir = pstrdup(dir);
get_parent_directory(parentdir);
if (mkdir(parentdir, S_IRWXU) < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not create directory \"%s\": %m",
parentdir)));
pfree(parentdir);
if (mkdir(dir, S_IRWXU) < 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not create directory \"%s\": %m",
dir)));
}
}
LWLockRelease(TablespaceCreateLock);
}
else
{
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not stat directory \"%s\": %m", dir)));
}
}
else
{
/* be paranoid */
if (!S_ISDIR(st.st_mode))
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("\"%s\" exists but is not a directory",
dir)));
}
pfree(dir);
}
