/*
* Validity test for an old-format hstore.
* 0 = not valid
* 1 = valid but with "slop" in the length
* 2 = exactly valid
*/
static int
hstoreValidOldFormat(HStore *hs)
{
int count = hs->size_;
HOldEntry *entries = (HOldEntry *) ARRPTR(hs);
int vsize;
int lastpos = 0;
int i;
if (hs->size_ & HS_FLAG_NEWVERSION)
return 0;
/* New format uses an HEntry for key and another for value */
StaticAssertStmt(sizeof(HOldEntry) == 2 * sizeof(HEntry),
"old hstore format is not upward-compatible");
if (count == 0)
return 2;
if (count > 0xFFFFFFF)
return 0;
if (CALCDATASIZE(count, 0) > VARSIZE(hs))
return 0;
if (entries[0].pos != 0)
return 0;
/* key length must be nondecreasing */
for (i = 1; i < count; ++i)
{
if (entries[i].keylen < entries[i - 1].keylen)
return 0;
}
/*
* entry position must be strictly increasing, except for the first entry
* (which can be ""=>"" and thus zero-length); and all entries must be
* properly contiguous
*/
for (i = 0; i < count; ++i)
{
if (entries[i].pos != lastpos)
return 0;
lastpos += (entries[i].keylen
+ ((entries[i].valisnull) ? 0 : entries[i].vallen));
}
vsize = CALCDATASIZE(count, lastpos);
if (vsize > VARSIZE(hs))
return 0;
if (vsize != VARSIZE(hs))
return 1;
return 2;
}
/*
* InitCatalogCache - initialize the caches
*
* Note that no database access is done here; we only allocate memory
* and initialize the cache structure. Interrogation of the database
* to complete initialization of a cache happens upon first use
* of that cache.
*/
void
InitCatalogCache(void)
{
int cacheId;
int i,
j;
StaticAssertStmt(SysCacheSize == (int) lengthof(cacheinfo),
"SysCacheSize does not match syscache.c's array");
Assert(!CacheInitialized);
SysCacheRelationOidSize = SysCacheSupportingRelOidSize = 0;
for (cacheId = 0; cacheId < SysCacheSize; cacheId++)
{
SysCache[cacheId] = InitCatCache(cacheId,
cacheinfo[cacheId].reloid,
cacheinfo[cacheId].indoid,
cacheinfo[cacheId].nkeys,
cacheinfo[cacheId].key,
cacheinfo[cacheId].nbuckets);
if (!PointerIsValid(SysCache[cacheId]))
elog(ERROR, "could not initialize cache %u (%d)",
cacheinfo[cacheId].reloid, cacheId);
/* Accumulate data for OID lists, too */
SysCacheRelationOid[SysCacheRelationOidSize++] =
cacheinfo[cacheId].reloid;
SysCacheSupportingRelOid[SysCacheSupportingRelOidSize++] =
cacheinfo[cacheId].reloid;
SysCacheSupportingRelOid[SysCacheSupportingRelOidSize++] =
cacheinfo[cacheId].indoid;
/* see comments for RelationInvalidatesSnapshotsOnly */
Assert(!RelationInvalidatesSnapshotsOnly(cacheinfo[cacheId].reloid));
}
Assert(SysCacheRelationOidSize <= lengthof(SysCacheRelationOid));
Assert(SysCacheSupportingRelOidSize <= lengthof(SysCacheSupportingRelOid));
/* Sort and de-dup OID arrays, so we can use binary search. */
pg_qsort(SysCacheRelationOid, SysCacheRelationOidSize,
sizeof(Oid), oid_compare);
for (i = 1, j = 0; i < SysCacheRelationOidSize; i++)
{
if (SysCacheRelationOid[i] != SysCacheRelationOid[j])
SysCacheRelationOid[++j] = SysCacheRelationOid[i];
}
SysCacheRelationOidSize = j + 1;
pg_qsort(SysCacheSupportingRelOid, SysCacheSupportingRelOidSize,
sizeof(Oid), oid_compare);
for (i = 1, j = 0; i < SysCacheSupportingRelOidSize; i++)
{
if (SysCacheSupportingRelOid[i] != SysCacheSupportingRelOid[j])
SysCacheSupportingRelOid[++j] = SysCacheSupportingRelOid[i];
}
SysCacheSupportingRelOidSize = j + 1;
CacheInitialized = true;
}
/*
* txid_current_snapshot() returns txid_snapshot
*
* Return current snapshot in TXID format
*
* Note that only top-transaction XIDs are included in the snapshot.
*/
Datum
txid_current_snapshot(PG_FUNCTION_ARGS)
{
TxidSnapshot *snap;
uint32 nxip,
i;
TxidEpoch state;
Snapshot cur;
cur = GetActiveSnapshot();
if (cur == NULL)
elog(ERROR, "no active snapshot set");
load_xid_epoch(&state);
/*
* Compile-time limits on the procarray (MAX_BACKENDS processes plus
* MAX_BACKENDS prepared transactions) guarantee nxip won't be too large.
*/
StaticAssertStmt(MAX_BACKENDS * 2 <= TXID_SNAPSHOT_MAX_NXIP,
"possible overflow in txid_current_snapshot()");
/* allocate */
nxip = cur->xcnt;
snap = palloc(TXID_SNAPSHOT_SIZE(nxip));
/* fill */
snap->xmin = convert_xid(cur->xmin, &state);
snap->xmax = convert_xid(cur->xmax, &state);
snap->nxip = nxip;
for (i = 0; i < nxip; i++)
snap->xip[i] = convert_xid(cur->xip[i], &state);
/*
* We want them guaranteed to be in ascending order. This also removes
* any duplicate xids. Normally, an XID can only be assigned to one
* backend, but when preparing a transaction for two-phase commit, there
* is a transient state when both the original backend and the dummy
* PGPROC entry reserved for the prepared transaction hold the same XID.
*/
sort_snapshot(snap);
/* set size after sorting, because it may have removed duplicate xips */
SET_VARSIZE(snap, TXID_SNAPSHOT_SIZE(snap->nxip));
PG_RETURN_POINTER(snap);
}
void
pg_atomic_init_u32_impl(volatile pg_atomic_uint32 *ptr, uint32 val_)
{
StaticAssertStmt(sizeof(ptr->sema) >= sizeof(slock_t),
"size mismatch of atomic_flag vs slock_t");
/*
* If we're using semaphore based atomic flags, be careful about nested
* usage of atomics while a spinlock is held.
*/
#ifndef HAVE_SPINLOCKS
s_init_lock_sema((slock_t *) &ptr->sema, true);
#else
SpinLockInit((slock_t *) &ptr->sema);
#endif
ptr->value = val_;
}
void
pg_atomic_init_flag_impl(volatile pg_atomic_flag *ptr)
{
StaticAssertStmt(sizeof(ptr->sema) >= sizeof(slock_t),
"size mismatch of atomic_flag vs slock_t");
#ifndef HAVE_SPINLOCKS
/*
* NB: If we're using semaphore based TAS emulation, be careful to use a
* separate set of semaphores. Otherwise we'd get in trouble if a atomic
* var would be manipulated while spinlock is held.
*/
s_init_lock_sema((slock_t *) &ptr->sema, true);
#else
SpinLockInit((slock_t *) &ptr->sema);
#endif
}
/*
* ItemPointerEquals
* Returns true if both item pointers point to the same item,
* otherwise returns false.
*
* Note:
* Asserts that the disk item pointers are both valid!
*/
bool
ItemPointerEquals(ItemPointer pointer1, ItemPointer pointer2)
{
/*
* We really want ItemPointerData to be exactly 6 bytes. This is rather a
* random place to check, but there is no better place.
*/
StaticAssertStmt(sizeof(ItemPointerData) == 3 * sizeof(uint16),
"ItemPointerData struct is improperly padded");
if (ItemPointerGetBlockNumber(pointer1) ==
ItemPointerGetBlockNumber(pointer2) &&
ItemPointerGetOffsetNumber(pointer1) ==
ItemPointerGetOffsetNumber(pointer2))
return true;
else
return false;
}
/*
* Determinisitcally generate salt for mock authentication, using a SHA256
* hash based on the username and a cluster-level secret key. Returns a
* pointer to a static buffer of size SCRAM_DEFAULT_SALT_LEN.
*/
static char *
scram_mock_salt(const char *username)
{
pg_sha256_ctx ctx;
static uint8 sha_digest[PG_SHA256_DIGEST_LENGTH];
char *mock_auth_nonce = GetMockAuthenticationNonce();
/*
* Generate salt using a SHA256 hash of the username and the cluster's
* mock authentication nonce. (This works as long as the salt length is
* not larger the SHA256 digest length. If the salt is smaller, the caller
* will just ignore the extra data.)
*/
StaticAssertStmt(PG_SHA256_DIGEST_LENGTH >= SCRAM_DEFAULT_SALT_LEN,
"salt length greater than SHA256 digest length");
pg_sha256_init(&ctx);
pg_sha256_update(&ctx, (uint8 *) username, strlen(username));
pg_sha256_update(&ctx, (uint8 *) mock_auth_nonce, MOCK_AUTH_NONCE_LEN);
pg_sha256_final(&ctx, sha_digest);
return (char *) sha_digest;
}
/*
* PageIsVerified
* Check that the page header and checksum (if any) appear valid.
*
* This is called when a page has just been read in from disk. The idea is
* to cheaply detect trashed pages before we go nuts following bogus item
* pointers, testing invalid transaction identifiers, etc.
*
* It turns out to be necessary to allow zeroed pages here too. Even though
* this routine is *not* called when deliberately adding a page to a relation,
* there are scenarios in which a zeroed page might be found in a table.
* (Example: a backend extends a relation, then crashes before it can write
* any WAL entry about the new page. The kernel will already have the
* zeroed page in the file, and it will stay that way after restart.) So we
* allow zeroed pages here, and are careful that the page access macros
* treat such a page as empty and without free space. Eventually, VACUUM
* will clean up such a page and make it usable.
*/
bool
PageIsVerified(Page page, BlockNumber blkno)
{
PageHeader p = (PageHeader) page;
size_t *pagebytes;
int i;
bool checksum_failure = false;
bool header_sane = false;
bool all_zeroes = false;
uint16 checksum = 0;
/*
* Don't verify page data unless the page passes basic non-zero test
*/
if (!PageIsNew(page))
{
if (DataChecksumsEnabled())
{
checksum = pg_checksum_page((char *) page, blkno);
if (checksum != p->pd_checksum)
checksum_failure = true;
}
/*
* The following checks don't prove the header is correct, only that
* it looks sane enough to allow into the buffer pool. Later usage of
* the block can still reveal problems, which is why we offer the
* checksum option.
*/
if ((p->pd_flags & ~PD_VALID_FLAG_BITS) == 0 &&
p->pd_lower <= p->pd_upper &&
p->pd_upper <= p->pd_special &&
p->pd_special <= BLCKSZ &&
p->pd_special == MAXALIGN(p->pd_special))
header_sane = true;
if (header_sane && !checksum_failure)
return true;
}
/*
* Check all-zeroes case. Luckily BLCKSZ is guaranteed to always be a
* multiple of size_t - and it's much faster to compare memory using the
* native word size.
*/
StaticAssertStmt(BLCKSZ == (BLCKSZ / sizeof(size_t)) * sizeof(size_t),
"BLCKSZ has to be a multiple of sizeof(size_t)");
all_zeroes = true;
pagebytes = (size_t *) page;
for (i = 0; i < (BLCKSZ / sizeof(size_t)); i++)
{
if (pagebytes[i] != 0)
{
all_zeroes = false;
break;
}
}
if (all_zeroes)
return true;
/*
* Throw a WARNING if the checksum fails, but only after we've checked for
* the all-zeroes case.
*/
if (checksum_failure)
{
ereport(WARNING,
(ERRCODE_DATA_CORRUPTED,
errmsg("page verification failed, calculated checksum %u but expected %u",
checksum, p->pd_checksum)));
if (header_sane && ignore_checksum_failure)
return true;
}
return false;
}
/*
* There are a few ways to arrive in the initsequencer.
* 1. From _PG_init (called exactly once when the library is loaded for ANY
* reason).
* 1a. Because of the command LOAD 'libraryname';
* This case can be distinguished because _PG_init will have found the
* LOAD command and saved the 'libraryname' in pljavaLoadPath.
* 1b. Because of a CREATE FUNCTION naming this library. pljavaLoadPath will
* be NULL.
* 1c. By the first actual use of a PL/Java function, causing this library
* to be loaded. pljavaLoadPath will be NULL. The called function's Oid
* will be available to the call handler once we return from _PG_init,
* but it isn't (easily) available here.
* 2. From the call handler, if initialization isn't complete yet. That can only
* mean something failed in the earlier call to _PG_init, and whatever it was
* is highly likely to fail again. That may lead to the untidyness of
* duplicated diagnostic messages, but for now I like the belt-and-suspenders
* approach of making sure the init sequence gets as many chances as possible
* to succeed.
* 3. From a GUC assign hook, if the user has updated a setting that might allow
* initialization to succeed. It resumes from where it left off.
*
* In all cases, the sequence must progress as far as starting the VM and
* initializing the PL/Java classes. In all cases except 1a, that's enough,
* assuming the language handlers and schema have all been set up already (or,
* in case 1b, the user is intent on setting them up explicitly).
*
* In case 1a, we can go ahead and test for, and create, the schema, functions,
* and language entries as needed, using pljavaLoadPath as the library path
* if creating the language handler functions. One-stop shopping. (The presence
* of pljavaLoadPath in any of the other cases, such as resumption by an assign
* hook, indicates it is really a continuation of case 1a.)
*/
static void initsequencer(enum initstage is, bool tolerant)
{
JVMOptList optList;
Invocation ctx;
jint JNIresult;
char *greeting;
switch (is)
{
case IS_FORMLESS_VOID:
initstage = IS_GUCS_REGISTERED;
case IS_GUCS_REGISTERED:
libjvmlocation = strdup("libjvm.so");
initstage = IS_PLJAVA_ENABLED;
case IS_PLJAVA_ENABLED:
libjvm_handle = pg_dlopen(libjvmlocation);
if ( NULL == libjvm_handle )
{
ereport(ERROR, (
errmsg("Cannot load libjvm.so library, check that it is available in LD_LIBRARY_PATH"),
errdetail("%s", (char *)pg_dlerror())));
goto check_tolerant;
}
initstage = IS_CAND_JVMOPENED;
case IS_CAND_JVMOPENED:
pljava_createvm =
(jint (JNICALL *)(JavaVM **, void **, void *))
pg_dlsym(libjvm_handle, "JNI_CreateJavaVM");
if ( NULL == pljava_createvm )
{
/*
* If it hasn't got the symbol, it can't be the right
* library, so close/unload it so another can be tried.
* Format the dlerror string first: dlclose may clobber it.
*/
char *dle = MemoryContextStrdup(ErrorContext, pg_dlerror());
pg_dlclose(libjvm_handle);
initstage = IS_CAND_JVMLOCATION;
ereport(ERROR, (
errmsg("Cannot start Java VM"),
errdetail("%s", dle),
errhint("Check that libjvm.so is available in LD_LIBRARY_PATH")));
goto check_tolerant;
}
initstage = IS_CREATEVM_SYM_FOUND;
case IS_CREATEVM_SYM_FOUND:
s_javaLogLevel = INFO;
checkIntTimeType();
HashMap_initialize(); /* creates things in TopMemoryContext */
#ifdef PLJAVA_DEBUG
/* Hard setting for debug. Don't forget to recompile...
*/
pljava_debug = 1;
#endif
initstage = IS_MISC_ONCE_DONE;
case IS_MISC_ONCE_DONE:
JVMOptList_init(&optList); /* uses CurrentMemoryContext */
seenVisualVMName = false;
addUserJVMOptions(&optList);
if ( ! seenVisualVMName )
JVMOptList_addVisualVMName(&optList);
JVMOptList_add(&optList, "vfprintf", (void*)my_vfprintf, true);
#ifndef GCJ
JVMOptList_add(&optList, "-Xrs", 0, true);
#endif
effectiveClassPath = getClassPath("-Djava.class.path=");
if(effectiveClassPath != 0)
{
JVMOptList_add(&optList, effectiveClassPath, 0, true);
}
initstage = IS_JAVAVM_OPTLIST;
case IS_JAVAVM_OPTLIST:
JNIresult = initializeJavaVM(&optList); /* frees the optList */
if( JNI_OK != JNIresult )
{
initstage = IS_MISC_ONCE_DONE; /* optList has been freed */
StaticAssertStmt(sizeof(jint) <= sizeof(long int),
"jint wider than long int?!");
ereport(WARNING,
(errmsg("failed to create Java virtual machine"),
errdetail("JNI_CreateJavaVM returned an error code: %ld",
(long int)JNIresult),
jvmStartedAtLeastOnce ?
errhint("Because an earlier attempt during this session "
"did start a VM before failing, this probably means your "
"Java runtime environment does not support more than one "
"VM creation per session. You may need to exit this "
"session and start a new one.") : 0));
goto check_tolerant;
}
jvmStartedAtLeastOnce = true;
elog(DEBUG2, "successfully created Java virtual machine");
initstage = IS_JAVAVM_STARTED;
case IS_JAVAVM_STARTED:
#ifdef USE_PLJAVA_SIGHANDLERS
pqsignal(SIGINT, pljavaStatementCancelHandler);
pqsignal(SIGTERM, pljavaDieHandler);
#endif
/* Register an on_proc_exit handler that destroys the VM
*/
on_proc_exit(_destroyJavaVM, 0);
initstage = IS_SIGHANDLERS;
case IS_SIGHANDLERS:
Invocation_pushBootContext(&ctx);
PG_TRY();
{
initPLJavaClasses();
initJavaSession();
Invocation_popBootContext();
initstage = IS_PLJAVA_FOUND;
}
PG_CATCH();
{
MemoryContextSwitchTo(ctx.upperContext); /* leave ErrorContext */
Invocation_popBootContext();
initstage = IS_MISC_ONCE_DONE;
/* We can't stay here...
*/
if ( tolerant )
reLogWithChangedLevel(WARNING); /* so xact is not aborted */
else
{
EmitErrorReport(); /* no more unwinding, just log it */
/* Seeing an ERROR emitted to the log, without leaving the
* transaction aborted, would violate the principle of least
* astonishment. But at check_tolerant below, another ERROR will
* be thrown immediately, so the transaction effect will be as
* expected and this ERROR will contribute information beyond
* what is in the generic one thrown down there.
*/
FlushErrorState();
}
}
PG_END_TRY();
if ( IS_PLJAVA_FOUND != initstage )
{
/* JVM initialization failed for some reason. Destroy
* the VM if it exists. Perhaps the user will try
* fixing the pljava.classpath and make a new attempt.
*/
ereport(WARNING, (
errmsg("failed to load initial PL/Java classes"),
errhint("The most common reason is that \"pljava_classpath\" "
"needs to be set, naming the proper \"pljava.jar\" file.")
));
_destroyJavaVM(0, 0);
goto check_tolerant;
}
case IS_PLJAVA_FOUND:
greeting = InstallHelper_hello();
ereport(NULL != pljavaLoadPath ? NOTICE : DEBUG1, (
errmsg("PL/Java loaded"),
errdetail("versions:\n%s", greeting)));
pfree(greeting);
if ( NULL != pljavaLoadPath )
InstallHelper_groundwork(); /* sqlj schema, language handlers, ...*/
initstage = IS_COMPLETE;
case IS_COMPLETE:
pljavaLoadingAsExtension = false;
if ( alteredSettingsWereNeeded )
{
/* Use this StringInfoData to conditionally construct part of the
* hint string suggesting ALTER DATABASE ... SET ... FROM CURRENT
* provided the server is >= 9.2 where that will actually work.
* In 9.3, psprintf appeared, which would make this all simpler,
* but if 9.3+ were all that had to be supported, this would all
* be moot anyway. Doing the initStringInfo inside the ereport
* ensures the string is allocated in ErrorContext and won't leak.
* Don't remove the extra parens grouping
* (initStringInfo, appendStringInfo, errhint) ... with the parens,
* that's a comma expression, which is sequenced; without them, they
* are just function parameters with evaluation order unknown.
*/
StringInfoData buf;
#if PG_VERSION_NUM >= 90200
#define MOREHINT \
appendStringInfo(&buf, \
"using ALTER DATABASE %s SET ... FROM CURRENT or ", \
pljavaDbName()),
#else
#define MOREHINT
#endif
ereport(NOTICE, (
errmsg("PL/Java successfully started after adjusting settings"),
(initStringInfo(&buf),
MOREHINT
errhint("The settings that worked should be saved (%s"
"in the \"%s\" file). For a reminder of what has been set, "
"try: SELECT name, setting FROM pg_settings WHERE name LIKE"
" 'pljava.%%' AND source = 'session'",
buf.data,
superuser()
? PG_GETCONFIGOPTION("config_file")
: "postgresql.conf"))));
#undef MOREHINT
if ( loadAsExtensionFailed )
{
ereport(NOTICE, (errmsg(
"PL/Java load successful after failed CREATE EXTENSION"),
errdetail(
"PL/Java is now installed, but not as an extension."),
errhint(
"To correct that, either COMMIT or ROLLBACK, make sure "
"the working settings are saved, exit this session, and "
"in a new session, either: "
"1. if committed, run "
"\"CREATE EXTENSION pljava FROM unpackaged\", or 2. "
"if rolled back, simply \"CREATE EXTENSION pljava\" again."
)));
}
}
return;
default:
ereport(ERROR, (
errmsg("cannot set up PL/Java"),
errdetail(
"An unexpected stage was reached in the startup sequence."),
errhint(
"Please report the circumstances to the PL/Java maintainers.")
));
}
check_tolerant:
if ( pljavaLoadingAsExtension )
{
tolerant = false;
loadAsExtensionFailed = true;
pljavaLoadingAsExtension = false;
}
if ( !tolerant )
{
ereport(ERROR, (
errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg(
"cannot use PL/Java before successfully completing its setup"),
errhint(
"Check the log for messages closely preceding this one, "
"detailing what step of setup failed and what will be needed, "
"probably setting one of the \"pljava.\" configuration "
"variables, to complete the setup. If there is not enough "
"help in the log, try again with different settings for "
"\"log_min_messages\" or \"log_error_verbosity\".")));
}
}
/*
* Create a WaitEventSet with space for nevents different events to wait for.
*
* These events can then efficiently waited upon together, using
* WaitEventSetWait().
*/
WaitEventSet *
CreateWaitEventSet(MemoryContext context, int nevents)
{
WaitEventSet *set;
char *data;
Size sz = 0;
sz += sizeof(WaitEventSet);
sz += sizeof(WaitEvent) * nevents;
#if defined(WAIT_USE_EPOLL)
sz += sizeof(struct epoll_event) * nevents;
#elif defined(WAIT_USE_POLL)
sz += sizeof(struct pollfd) * nevents;
#elif defined(WAIT_USE_WIN32)
/* need space for the pgwin32_signal_event */
sz += sizeof(HANDLE) * (nevents + 1);
#endif
data = (char *) MemoryContextAllocZero(context, sz);
set = (WaitEventSet *) data;
data += sizeof(WaitEventSet);
set->events = (WaitEvent *) data;
data += sizeof(WaitEvent) * nevents;
#if defined(WAIT_USE_EPOLL)
set->epoll_ret_events = (struct epoll_event *) data;
data += sizeof(struct epoll_event) * nevents;
#elif defined(WAIT_USE_POLL)
set->pollfds = (struct pollfd *) data;
data += sizeof(struct pollfd) * nevents;
#elif defined(WAIT_USE_WIN32)
set->handles = (HANDLE) data;
data += sizeof(HANDLE) * nevents;
#endif
set->latch = NULL;
set->nevents_space = nevents;
#if defined(WAIT_USE_EPOLL)
set->epoll_fd = epoll_create(nevents);
if (set->epoll_fd < 0)
elog(ERROR, "epoll_create failed: %m");
#elif defined(WAIT_USE_WIN32)
/*
* To handle signals while waiting, we need to add a win32 specific event.
* We accounted for the additional event at the top of this routine. See
* port/win32/signal.c for more details.
*
* Note: pgwin32_signal_event should be first to ensure that it will be
* reported when multiple events are set. We want to guarantee that
* pending signals are serviced.
*/
set->handles[0] = pgwin32_signal_event;
StaticAssertStmt(WSA_INVALID_EVENT == NULL, "");
#endif
return set;
}
static void
ReadBlocks(int filenum)
{
FILE *file;
char record_type;
char *dbname;
Oid record_filenode;
ForkNumber record_forknum;
BlockNumber record_blocknum;
BlockNumber record_range;
int log_level = DEBUG3;
Oid relOid = InvalidOid;
Relation rel = NULL;
bool skip_relation = false;
bool skip_fork = false;
bool skip_block = false;
BlockNumber nblocks = 0;
BlockNumber blocks_restored = 0;
const char *filepath;
/*
* If this condition changes, then this code, and the code in the writer
* will need to be changed; especially the format specifiers in log and
* error messages.
*/
StaticAssertStmt(MaxBlockNumber == 0xFFFFFFFE, "Code may need review.");
filepath = getSavefileName(filenum);
file = fileOpen(filepath, PG_BINARY_R);
dbname = readDBName(file, filepath);
/*
* When restoring global objects, the dbname is zero-length string, and non-
* zero length otherwise. And filenum is never expected to be smaller than 1.
*/
Assert(filenum >= 1);
Assert(filenum == 1 ? strlen(dbname) == 0 : strlen(dbname) > 0);
/* To restore the global objects, use default database */
BackgroundWorkerInitializeConnection(filenum == 1 ? guc_default_database : dbname, NULL);
SetCurrentStatementStartTimestamp();
StartTransactionCommand();
SPI_connect();
PushActiveSnapshot(GetTransactionSnapshot());
pgstat_report_activity(STATE_RUNNING, "restoring buffers");
/*
* Note that in case of a read error, we will leak relcache entry that we may
* currently have open. In case of EOF, we close the relation after the loop.
*/
while (fileRead(&record_type, 1, file, true, filepath))
{
/*
* If we want to process the signals, this seems to be the best place
* to do it. Generally the backends refrain from processing config file
* while in transaction, but that's more for the fear of allowing GUC
* changes to affect expression evaluation, causing different results
* for the same expression in a transaction. Since this worker is not
* processing any queries, it is okay to process the config file here.
*
* Even though it's okay to process SIGHUP here, doing so doesn't add
* any value. The only reason we might want to process config file here
* would be to allow the user to interrupt the BlockReader's operation
* by changing this extenstion's GUC parameter. But the user can do that
* anyway, using SIGTERM or pg_terminate_backend().
*/
/* Stop processing the save-file if the Postmaster wants us to die. */
if (got_sigterm)
break;
ereport(log_level,
(errmsg("record type %x - %c", record_type, record_type)));
switch (record_type)
{
case 'r':
{
/* Close the previous relation, if any. */
if (rel)
{
relation_close(rel, AccessShareLock);
rel = NULL;
}
record_forknum = InvalidForkNumber;
record_blocknum = InvalidBlockNumber;
nblocks = 0;
fileRead(&record_filenode, sizeof(Oid), file, false, filepath);
relOid = GetRelOid(record_filenode);
ereport(log_level, (errmsg("processing filenode %u, relation %u",
record_filenode, relOid)));
/*
* If the relation has been rewritten/dropped since we saved it,
* just skip it and process the next relation.
*/
if (relOid == InvalidOid)
skip_relation = true;
else
{
skip_relation = false;
/* Open the relation */
rel = relation_open(relOid, AccessShareLock);
RelationOpenSmgr(rel);
}
}
break;
case 'f':
{
record_blocknum = InvalidBlockNumber;
nblocks = 0;
fileRead(&record_forknum, sizeof(ForkNumber), file, false, filepath);
if (skip_relation)
continue;
if (rel == NULL)
ereport(ERROR,
(errmsg("found a fork record without a preceeding relation record")));
ereport(log_level, (errmsg("processing fork %d", record_forknum)));
if (!smgrexists(rel->rd_smgr, record_forknum))
skip_fork = true;
else
{
skip_fork = false;
nblocks = RelationGetNumberOfBlocksInFork(rel, record_forknum);
}
}
break;
case 'b':
{
if (record_forknum == InvalidForkNumber)
ereport(ERROR,
(errmsg("found a block record without a preceeding fork record")));
fileRead(&record_blocknum, sizeof(BlockNumber), file, false, filepath);
if (skip_relation || skip_fork)
continue;
/*
* Don't try to read past the file; the file may have been shrunk
* by a vaccum/truncate operation.
*/
if (record_blocknum >= nblocks)
{
ereport(log_level,
(errmsg("reader %d skipping block filenode %u forknum %d blocknum %u",
filenum, record_filenode, record_forknum, record_blocknum)));
skip_block = true;
continue;
}
else
{
Buffer buf;
skip_block = false;
ereport(log_level,
(errmsg("reader %d reading block filenode %u forknum %d blocknum %u",
filenum, record_filenode, record_forknum, record_blocknum)));
buf = ReadBufferExtended(rel, record_forknum, record_blocknum, RBM_NORMAL, NULL);
ReleaseBuffer(buf);
++blocks_restored;
}
}
break;
case 'N':
{
BlockNumber block;
Assert(record_blocknum != InvalidBlockNumber);
if (record_blocknum == InvalidBlockNumber)
ereport(ERROR,
(errmsg("found a block range record without a preceeding block record")));
fileRead(&record_range, sizeof(int), file, false, filepath);
if (skip_relation || skip_fork || skip_block)
continue;
ereport(log_level,
(errmsg("reader %d reading range filenode %u forknum %d blocknum %u range %u",
filenum, record_filenode, record_forknum, record_blocknum, record_range)));
for (block = record_blocknum + 1; block <= (record_blocknum + record_range); ++block)
{
Buffer buf;
/*
* Don't try to read past the file; the file may have been
* shrunk by a vaccum operation.
*/
if (block >= nblocks)
{
ereport(log_level,
(errmsg("reader %d skipping block range filenode %u forknum %d start %u end %u",
filenum, record_filenode, record_forknum,
block, record_blocknum + record_range)));
break;
}
buf = ReadBufferExtended(rel, record_forknum, block, RBM_NORMAL, NULL);
ReleaseBuffer(buf);
++blocks_restored;
}
}
break;
default:
{
ereport(ERROR,
(errmsg("found unexpected save-file marker %x - %c)", record_type, record_type)));
Assert(false);
}
break;
}
}
if (rel)
relation_close(rel, AccessShareLock);
ereport(LOG,
(errmsg("Block Reader %d: restored %u blocks",
filenum, blocks_restored)));
SPI_finish();
PopActiveSnapshot();
CommitTransactionCommand();
pgstat_report_activity(STATE_IDLE, NULL);
fileClose(file, filepath);
/* Remove the save-file */
if (remove(filepath) != 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("error removing file \"%s\" : %m", filepath)));
}