postgresql/src/backend/storage/ipc/dsm.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1305 lines
40 KiB
C
Raw Normal View History

/*-------------------------------------------------------------------------
*
* dsm.c
* manage dynamic shared memory segments
*
* This file provides a set of services to make programming with dynamic
* shared memory segments more convenient. Unlike the low-level
* facilities provided by dsm_impl.h and dsm_impl.c, mappings and segments
* created using this module will be cleaned up automatically. Mappings
* will be removed when the resource owner under which they were created
* is cleaned up, unless dsm_pin_mapping() is used, in which case they
* have session lifespan. Segments will be removed when there are no
* remaining mappings, or at postmaster shutdown in any case. After a
* hard postmaster crash, remaining segments will be removed, if they
* still exist, at the next postmaster startup.
*
* Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
*
* IDENTIFICATION
* src/backend/storage/ipc/dsm.c
*
*-------------------------------------------------------------------------
*/
#include "postgres.h"
#include <fcntl.h>
#include <unistd.h>
#ifndef WIN32
#include <sys/mman.h>
#endif
#include <sys/stat.h>
#include "common/pg_prng.h"
#include "lib/ilist.h"
#include "miscadmin.h"
#include "port/pg_bitutils.h"
#include "storage/dsm.h"
Make ResourceOwners more easily extensible. Instead of having a separate array/hash for each resource kind, use a single array and hash to hold all kinds of resources. This makes it possible to introduce new resource "kinds" without having to modify the ResourceOwnerData struct. In particular, this makes it possible for extensions to register custom resource kinds. The old approach was to have a small array of resources of each kind, and if it fills up, switch to a hash table. The new approach also uses an array and a hash, but now the array and the hash are used at the same time. The array is used to hold the recently added resources, and when it fills up, they are moved to the hash. This keeps the access to recent entries fast, even when there are a lot of long-held resources. All the resource-specific ResourceOwnerEnlarge*(), ResourceOwnerRemember*(), and ResourceOwnerForget*() functions have been replaced with three generic functions that take resource kind as argument. For convenience, we still define resource-specific wrapper macros around the generic functions with the old names, but they are now defined in the source files that use those resource kinds. The release callback no longer needs to call ResourceOwnerForget on the resource being released. ResourceOwnerRelease unregisters the resource from the owner before calling the callback. That needed some changes in bufmgr.c and some other files, where releasing the resources previously always called ResourceOwnerForget. Each resource kind specifies a release priority, and ResourceOwnerReleaseAll releases the resources in priority order. To make that possible, we have to restrict what you can do between phases. After calling ResourceOwnerRelease(), you are no longer allowed to remember any more resources in it or to forget any previously remembered resources by calling ResourceOwnerForget. There was one case where that was done previously. At subtransaction commit, AtEOSubXact_Inval() would handle the invalidation messages and call RelationFlushRelation(), which temporarily increased the reference count on the relation being flushed. We now switch to the parent subtransaction's resource owner before calling AtEOSubXact_Inval(), so that there is a valid ResourceOwner to temporarily hold that relcache reference. Other end-of-xact routines make similar calls to AtEOXact_Inval() between release phases, but I didn't see any regression test failures from those, so I'm not sure if they could reach a codepath that needs remembering extra resources. There were two exceptions to how the resource leak WARNINGs on commit were printed previously: llvmjit silently released the context without printing the warning, and a leaked buffer io triggered a PANIC. Now everything prints a WARNING, including those cases. Add tests in src/test/modules/test_resowner. Reviewed-by: Aleksander Alekseev, Michael Paquier, Julien Rouhaud Reviewed-by: Kyotaro Horiguchi, Hayato Kuroda, Álvaro Herrera, Zhihong Yu Reviewed-by: Peter Eisentraut, Andres Freund Discussion: https://www.postgresql.org/message-id/cbfabeb0-cd3c-e951-a572-19b365ed314d%40iki.fi
2023-11-08 12:30:50 +01:00
#include "storage/fd.h"
#include "storage/ipc.h"
#include "storage/lwlock.h"
#include "storage/pg_shmem.h"
Make ResourceOwners more easily extensible. Instead of having a separate array/hash for each resource kind, use a single array and hash to hold all kinds of resources. This makes it possible to introduce new resource "kinds" without having to modify the ResourceOwnerData struct. In particular, this makes it possible for extensions to register custom resource kinds. The old approach was to have a small array of resources of each kind, and if it fills up, switch to a hash table. The new approach also uses an array and a hash, but now the array and the hash are used at the same time. The array is used to hold the recently added resources, and when it fills up, they are moved to the hash. This keeps the access to recent entries fast, even when there are a lot of long-held resources. All the resource-specific ResourceOwnerEnlarge*(), ResourceOwnerRemember*(), and ResourceOwnerForget*() functions have been replaced with three generic functions that take resource kind as argument. For convenience, we still define resource-specific wrapper macros around the generic functions with the old names, but they are now defined in the source files that use those resource kinds. The release callback no longer needs to call ResourceOwnerForget on the resource being released. ResourceOwnerRelease unregisters the resource from the owner before calling the callback. That needed some changes in bufmgr.c and some other files, where releasing the resources previously always called ResourceOwnerForget. Each resource kind specifies a release priority, and ResourceOwnerReleaseAll releases the resources in priority order. To make that possible, we have to restrict what you can do between phases. After calling ResourceOwnerRelease(), you are no longer allowed to remember any more resources in it or to forget any previously remembered resources by calling ResourceOwnerForget. There was one case where that was done previously. At subtransaction commit, AtEOSubXact_Inval() would handle the invalidation messages and call RelationFlushRelation(), which temporarily increased the reference count on the relation being flushed. We now switch to the parent subtransaction's resource owner before calling AtEOSubXact_Inval(), so that there is a valid ResourceOwner to temporarily hold that relcache reference. Other end-of-xact routines make similar calls to AtEOXact_Inval() between release phases, but I didn't see any regression test failures from those, so I'm not sure if they could reach a codepath that needs remembering extra resources. There were two exceptions to how the resource leak WARNINGs on commit were printed previously: llvmjit silently released the context without printing the warning, and a leaked buffer io triggered a PANIC. Now everything prints a WARNING, including those cases. Add tests in src/test/modules/test_resowner. Reviewed-by: Aleksander Alekseev, Michael Paquier, Julien Rouhaud Reviewed-by: Kyotaro Horiguchi, Hayato Kuroda, Álvaro Herrera, Zhihong Yu Reviewed-by: Peter Eisentraut, Andres Freund Discussion: https://www.postgresql.org/message-id/cbfabeb0-cd3c-e951-a572-19b365ed314d%40iki.fi
2023-11-08 12:30:50 +01:00
#include "storage/shmem.h"
#include "utils/freepage.h"
#include "utils/guc.h"
#include "utils/memutils.h"
Make ResourceOwners more easily extensible. Instead of having a separate array/hash for each resource kind, use a single array and hash to hold all kinds of resources. This makes it possible to introduce new resource "kinds" without having to modify the ResourceOwnerData struct. In particular, this makes it possible for extensions to register custom resource kinds. The old approach was to have a small array of resources of each kind, and if it fills up, switch to a hash table. The new approach also uses an array and a hash, but now the array and the hash are used at the same time. The array is used to hold the recently added resources, and when it fills up, they are moved to the hash. This keeps the access to recent entries fast, even when there are a lot of long-held resources. All the resource-specific ResourceOwnerEnlarge*(), ResourceOwnerRemember*(), and ResourceOwnerForget*() functions have been replaced with three generic functions that take resource kind as argument. For convenience, we still define resource-specific wrapper macros around the generic functions with the old names, but they are now defined in the source files that use those resource kinds. The release callback no longer needs to call ResourceOwnerForget on the resource being released. ResourceOwnerRelease unregisters the resource from the owner before calling the callback. That needed some changes in bufmgr.c and some other files, where releasing the resources previously always called ResourceOwnerForget. Each resource kind specifies a release priority, and ResourceOwnerReleaseAll releases the resources in priority order. To make that possible, we have to restrict what you can do between phases. After calling ResourceOwnerRelease(), you are no longer allowed to remember any more resources in it or to forget any previously remembered resources by calling ResourceOwnerForget. There was one case where that was done previously. At subtransaction commit, AtEOSubXact_Inval() would handle the invalidation messages and call RelationFlushRelation(), which temporarily increased the reference count on the relation being flushed. We now switch to the parent subtransaction's resource owner before calling AtEOSubXact_Inval(), so that there is a valid ResourceOwner to temporarily hold that relcache reference. Other end-of-xact routines make similar calls to AtEOXact_Inval() between release phases, but I didn't see any regression test failures from those, so I'm not sure if they could reach a codepath that needs remembering extra resources. There were two exceptions to how the resource leak WARNINGs on commit were printed previously: llvmjit silently released the context without printing the warning, and a leaked buffer io triggered a PANIC. Now everything prints a WARNING, including those cases. Add tests in src/test/modules/test_resowner. Reviewed-by: Aleksander Alekseev, Michael Paquier, Julien Rouhaud Reviewed-by: Kyotaro Horiguchi, Hayato Kuroda, Álvaro Herrera, Zhihong Yu Reviewed-by: Peter Eisentraut, Andres Freund Discussion: https://www.postgresql.org/message-id/cbfabeb0-cd3c-e951-a572-19b365ed314d%40iki.fi
2023-11-08 12:30:50 +01:00
#include "utils/resowner.h"
#define PG_DYNSHMEM_CONTROL_MAGIC 0x9a503d32
#define PG_DYNSHMEM_FIXED_SLOTS 64
#define PG_DYNSHMEM_SLOTS_PER_BACKEND 5
#define INVALID_CONTROL_SLOT ((uint32) -1)
/* Backend-local tracking for on-detach callbacks. */
typedef struct dsm_segment_detach_callback
{
on_dsm_detach_callback function;
Datum arg;
slist_node node;
} dsm_segment_detach_callback;
/* Backend-local state for a dynamic shared memory segment. */
struct dsm_segment
{
dlist_node node; /* List link in dsm_segment_list. */
ResourceOwner resowner; /* Resource owner. */
dsm_handle handle; /* Segment name. */
uint32 control_slot; /* Slot in control segment. */
void *impl_private; /* Implementation-specific private data. */
void *mapped_address; /* Mapping address, or NULL if unmapped. */
Size mapped_size; /* Size of our mapping. */
slist_head on_detach; /* On-detach callbacks. */
};
/* Shared-memory state for a dynamic shared memory segment. */
typedef struct dsm_control_item
{
dsm_handle handle;
uint32 refcnt; /* 2+ = active, 1 = moribund, 0 = gone */
size_t first_page;
size_t npages;
void *impl_private_pm_handle; /* only needed on Windows */
bool pinned;
} dsm_control_item;
/* Layout of the dynamic shared memory control segment. */
typedef struct dsm_control_header
{
uint32 magic;
uint32 nitems;
uint32 maxitems;
dsm_control_item item[FLEXIBLE_ARRAY_MEMBER];
} dsm_control_header;
static void dsm_cleanup_for_mmap(void);
static void dsm_postmaster_shutdown(int code, Datum arg);
static dsm_segment *dsm_create_descriptor(void);
static bool dsm_control_segment_sane(dsm_control_header *control,
Size mapped_size);
static uint64 dsm_control_bytes_needed(uint32 nitems);
static inline dsm_handle make_main_region_dsm_handle(int slot);
static inline bool is_main_region_dsm_handle(dsm_handle handle);
/* Has this backend initialized the dynamic shared memory system yet? */
static bool dsm_init_done = false;
/* Preallocated DSM space in the main shared memory region. */
static void *dsm_main_space_begin = NULL;
/*
* List of dynamic shared memory segments used by this backend.
*
* At process exit time, we must decrement the reference count of each
* segment we have attached; this list makes it possible to find all such
* segments.
*
* This list should always be empty in the postmaster. We could probably
* allow the postmaster to map dynamic shared memory segments before it
* begins to start child processes, provided that each process adjusted
* the reference counts for those segments in the control segment at
* startup time, but there's no obvious need for such a facility, which
* would also be complex to handle in the EXEC_BACKEND case. Once the
* postmaster has begun spawning children, there's an additional problem:
* each new mapping would require an update to the control segment,
* which requires locking, in which the postmaster must not be involved.
*/
static dlist_head dsm_segment_list = DLIST_STATIC_INIT(dsm_segment_list);
/*
* Control segment information.
*
* Unlike ordinary shared memory segments, the control segment is not
* reference counted; instead, it lasts for the postmaster's entire
* life cycle. For simplicity, it doesn't have a dsm_segment object either.
*/
static dsm_handle dsm_control_handle;
static dsm_control_header *dsm_control;
static Size dsm_control_mapped_size = 0;
static void *dsm_control_impl_private = NULL;
Make ResourceOwners more easily extensible. Instead of having a separate array/hash for each resource kind, use a single array and hash to hold all kinds of resources. This makes it possible to introduce new resource "kinds" without having to modify the ResourceOwnerData struct. In particular, this makes it possible for extensions to register custom resource kinds. The old approach was to have a small array of resources of each kind, and if it fills up, switch to a hash table. The new approach also uses an array and a hash, but now the array and the hash are used at the same time. The array is used to hold the recently added resources, and when it fills up, they are moved to the hash. This keeps the access to recent entries fast, even when there are a lot of long-held resources. All the resource-specific ResourceOwnerEnlarge*(), ResourceOwnerRemember*(), and ResourceOwnerForget*() functions have been replaced with three generic functions that take resource kind as argument. For convenience, we still define resource-specific wrapper macros around the generic functions with the old names, but they are now defined in the source files that use those resource kinds. The release callback no longer needs to call ResourceOwnerForget on the resource being released. ResourceOwnerRelease unregisters the resource from the owner before calling the callback. That needed some changes in bufmgr.c and some other files, where releasing the resources previously always called ResourceOwnerForget. Each resource kind specifies a release priority, and ResourceOwnerReleaseAll releases the resources in priority order. To make that possible, we have to restrict what you can do between phases. After calling ResourceOwnerRelease(), you are no longer allowed to remember any more resources in it or to forget any previously remembered resources by calling ResourceOwnerForget. There was one case where that was done previously. At subtransaction commit, AtEOSubXact_Inval() would handle the invalidation messages and call RelationFlushRelation(), which temporarily increased the reference count on the relation being flushed. We now switch to the parent subtransaction's resource owner before calling AtEOSubXact_Inval(), so that there is a valid ResourceOwner to temporarily hold that relcache reference. Other end-of-xact routines make similar calls to AtEOXact_Inval() between release phases, but I didn't see any regression test failures from those, so I'm not sure if they could reach a codepath that needs remembering extra resources. There were two exceptions to how the resource leak WARNINGs on commit were printed previously: llvmjit silently released the context without printing the warning, and a leaked buffer io triggered a PANIC. Now everything prints a WARNING, including those cases. Add tests in src/test/modules/test_resowner. Reviewed-by: Aleksander Alekseev, Michael Paquier, Julien Rouhaud Reviewed-by: Kyotaro Horiguchi, Hayato Kuroda, Álvaro Herrera, Zhihong Yu Reviewed-by: Peter Eisentraut, Andres Freund Discussion: https://www.postgresql.org/message-id/cbfabeb0-cd3c-e951-a572-19b365ed314d%40iki.fi
2023-11-08 12:30:50 +01:00
/* ResourceOwner callbacks to hold DSM segments */
static void ResOwnerReleaseDSM(Datum res);
static char *ResOwnerPrintDSM(Datum res);
static const ResourceOwnerDesc dsm_resowner_desc =
{
.name = "dynamic shared memory segment",
.release_phase = RESOURCE_RELEASE_BEFORE_LOCKS,
.release_priority = RELEASE_PRIO_DSMS,
.ReleaseResource = ResOwnerReleaseDSM,
.DebugPrint = ResOwnerPrintDSM
};
/* Convenience wrappers over ResourceOwnerRemember/Forget */
static inline void
ResourceOwnerRememberDSM(ResourceOwner owner, dsm_segment *seg)
{
ResourceOwnerRemember(owner, PointerGetDatum(seg), &dsm_resowner_desc);
}
static inline void
ResourceOwnerForgetDSM(ResourceOwner owner, dsm_segment *seg)
{
ResourceOwnerForget(owner, PointerGetDatum(seg), &dsm_resowner_desc);
}
/*
* Start up the dynamic shared memory system.
*
* This is called just once during each cluster lifetime, at postmaster
* startup time.
*/
void
dsm_postmaster_startup(PGShmemHeader *shim)
{
void *dsm_control_address = NULL;
uint32 maxitems;
Size segsize;
Assert(!IsUnderPostmaster);
/*
* If we're using the mmap implementations, clean up any leftovers.
* Cleanup isn't needed on Windows, and happens earlier in startup for
* POSIX and System V shared memory, via a direct call to
* dsm_cleanup_using_control_segment.
*/
if (dynamic_shared_memory_type == DSM_IMPL_MMAP)
dsm_cleanup_for_mmap();
/* Determine size for new control segment. */
maxitems = PG_DYNSHMEM_FIXED_SLOTS
+ PG_DYNSHMEM_SLOTS_PER_BACKEND * MaxBackends;
elog(DEBUG2, "dynamic shared memory system will support %u segments",
maxitems);
segsize = dsm_control_bytes_needed(maxitems);
/*
* Loop until we find an unused identifier for the new control segment. We
* sometimes use DSM_HANDLE_INVALID as a sentinel value indicating "no
* control segment", so avoid generating that value for a real handle.
*/
for (;;)
{
Assert(dsm_control_address == NULL);
Assert(dsm_control_mapped_size == 0);
/* Use even numbers only */
dsm_control_handle = pg_prng_uint32(&pg_global_prng_state) << 1;
if (dsm_control_handle == DSM_HANDLE_INVALID)
continue;
if (dsm_impl_op(DSM_OP_CREATE, dsm_control_handle, segsize,
&dsm_control_impl_private, &dsm_control_address,
&dsm_control_mapped_size, ERROR))
break;
}
dsm_control = dsm_control_address;
on_shmem_exit(dsm_postmaster_shutdown, PointerGetDatum(shim));
elog(DEBUG2,
"created dynamic shared memory control segment %u (%zu bytes)",
dsm_control_handle, segsize);
shim->dsm_control = dsm_control_handle;
/* Initialize control segment. */
dsm_control->magic = PG_DYNSHMEM_CONTROL_MAGIC;
dsm_control->nitems = 0;
dsm_control->maxitems = maxitems;
}
/*
* Determine whether the control segment from the previous postmaster
* invocation still exists. If so, remove the dynamic shared memory
* segments to which it refers, and then the control segment itself.
*/
void
dsm_cleanup_using_control_segment(dsm_handle old_control_handle)
{
void *mapped_address = NULL;
void *junk_mapped_address = NULL;
void *impl_private = NULL;
void *junk_impl_private = NULL;
Size mapped_size = 0;
Size junk_mapped_size = 0;
uint32 nitems;
uint32 i;
dsm_control_header *old_control;
/*
* Try to attach the segment. If this fails, it probably just means that
* the operating system has been rebooted and the segment no longer
* exists, or an unrelated process has used the same shm ID. So just fall
* out quietly.
*/
if (!dsm_impl_op(DSM_OP_ATTACH, old_control_handle, 0, &impl_private,
&mapped_address, &mapped_size, DEBUG1))
return;
/*
* We've managed to reattach it, but the contents might not be sane. If
* they aren't, we disregard the segment after all.
*/
old_control = (dsm_control_header *) mapped_address;
if (!dsm_control_segment_sane(old_control, mapped_size))
{
dsm_impl_op(DSM_OP_DETACH, old_control_handle, 0, &impl_private,
&mapped_address, &mapped_size, LOG);
return;
}
/*
2016-06-07 18:34:33 +02:00
* OK, the control segment looks basically valid, so we can use it to get
* a list of segments that need to be removed.
*/
nitems = old_control->nitems;
for (i = 0; i < nitems; ++i)
{
dsm_handle handle;
uint32 refcnt;
/* If the reference count is 0, the slot is actually unused. */
refcnt = old_control->item[i].refcnt;
if (refcnt == 0)
continue;
/* If it was using the main shmem area, there is nothing to do. */
handle = old_control->item[i].handle;
if (is_main_region_dsm_handle(handle))
continue;
/* Log debugging information. */
elog(DEBUG2, "cleaning up orphaned dynamic shared memory with ID %u (reference count %u)",
handle, refcnt);
/* Destroy the referenced segment. */
dsm_impl_op(DSM_OP_DESTROY, handle, 0, &junk_impl_private,
&junk_mapped_address, &junk_mapped_size, LOG);
}
/* Destroy the old control segment, too. */
elog(DEBUG2,
"cleaning up dynamic shared memory control segment with ID %u",
old_control_handle);
dsm_impl_op(DSM_OP_DESTROY, old_control_handle, 0, &impl_private,
&mapped_address, &mapped_size, LOG);
}
/*
* When we're using the mmap shared memory implementation, "shared memory"
* segments might even manage to survive an operating system reboot.
* But there's no guarantee as to exactly what will survive: some segments
* may survive, and others may not, and the contents of some may be out
* of date. In particular, the control segment may be out of date, so we
* can't rely on it to figure out what to remove. However, since we know
* what directory contains the files we used as shared memory, we can simply
* scan the directory and blow everything away that shouldn't be there.
*/
static void
dsm_cleanup_for_mmap(void)
{
DIR *dir;
struct dirent *dent;
Clean up assorted messiness around AllocateDir() usage. This patch fixes a couple of low-probability bugs that could lead to reporting an irrelevant errno value (and hence possibly a wrong SQLSTATE) concerning directory-open or file-open failures. It also fixes places where we took shortcuts in reporting such errors, either by using elog instead of ereport or by using ereport but forgetting to specify an errcode. And it eliminates a lot of just plain redundant error-handling code. In service of all this, export fd.c's formerly-static function ReadDirExtended, so that external callers can make use of the coding pattern dir = AllocateDir(path); while ((de = ReadDirExtended(dir, path, LOG)) != NULL) if they'd like to treat directory-open failures as mere LOG conditions rather than errors. Also fix FreeDir to be a no-op if we reach it with dir == NULL, as such a coding pattern would cause. Then, remove code at many call sites that was throwing an error or log message for AllocateDir failure, as ReadDir or ReadDirExtended can handle that job just fine. Aside from being a net code savings, this gets rid of a lot of not-quite-up-to-snuff reports, as mentioned above. (In some places these changes result in replacing a custom error message such as "could not open tablespace directory" with more generic wording "could not open directory", but it was agreed that the custom wording buys little as long as we report the directory name.) In some other call sites where we can't just remove code, change the error reports to be fully project-style-compliant. Also reorder code in restoreTwoPhaseData that was acquiring a lock between AllocateDir and ReadDir; in the unlikely but surely not impossible case that LWLockAcquire changes errno, AllocateDir failures would be misreported. There is no great value in opening the directory before acquiring TwoPhaseStateLock, so just do it in the other order. Also fix CheckXLogRemoved to guarantee that it preserves errno, as quite a number of call sites are implicitly assuming. (Again, it's unlikely but I think not impossible that errno could change during a SpinLockAcquire. If so, this function was broken for its own purposes as well as breaking callers.) And change a few places that were using not-per-project-style messages, such as "could not read directory" when "could not open directory" is more correct. Back-patch the exporting of ReadDirExtended, in case we have occasion to back-patch some fix that makes use of it; it's not needed right now but surely making it global is pretty harmless. Also back-patch the restoreTwoPhaseData and CheckXLogRemoved fixes. The rest of this is essentially cosmetic and need not get back-patched. Michael Paquier, with a bit of additional work by me Discussion: https://postgr.es/m/CAB7nPqRpOCxjiirHmebEFhXVTK7V5Jvw4bz82p7Oimtsm3TyZA@mail.gmail.com
2017-12-04 23:02:52 +01:00
/* Scan the directory for something with a name of the correct format. */
dir = AllocateDir(PG_DYNSHMEM_DIR);
while ((dent = ReadDir(dir, PG_DYNSHMEM_DIR)) != NULL)
{
if (strncmp(dent->d_name, PG_DYNSHMEM_MMAP_FILE_PREFIX,
strlen(PG_DYNSHMEM_MMAP_FILE_PREFIX)) == 0)
{
char buf[MAXPGPATH + sizeof(PG_DYNSHMEM_DIR)];
snprintf(buf, sizeof(buf), PG_DYNSHMEM_DIR "/%s", dent->d_name);
elog(DEBUG2, "removing file \"%s\"", buf);
/* We found a matching file; so remove it. */
if (unlink(buf) != 0)
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not remove file \"%s\": %m", buf)));
}
}
/* Cleanup complete. */
FreeDir(dir);
}
/*
* At shutdown time, we iterate over the control segment and remove all
* remaining dynamic shared memory segments. We avoid throwing errors here;
* the postmaster is shutting down either way, and this is just non-critical
* resource cleanup.
*/
static void
dsm_postmaster_shutdown(int code, Datum arg)
{
uint32 nitems;
uint32 i;
void *dsm_control_address;
void *junk_mapped_address = NULL;
void *junk_impl_private = NULL;
Size junk_mapped_size = 0;
PGShmemHeader *shim = (PGShmemHeader *) DatumGetPointer(arg);
/*
* If some other backend exited uncleanly, it might have corrupted the
* control segment while it was dying. In that case, we warn and ignore
* the contents of the control segment. This may end up leaving behind
* stray shared memory segments, but there's not much we can do about that
* if the metadata is gone.
*/
nitems = dsm_control->nitems;
if (!dsm_control_segment_sane(dsm_control, dsm_control_mapped_size))
{
ereport(LOG,
(errmsg("dynamic shared memory control segment is corrupt")));
return;
}
/* Remove any remaining segments. */
for (i = 0; i < nitems; ++i)
{
dsm_handle handle;
/* If the reference count is 0, the slot is actually unused. */
if (dsm_control->item[i].refcnt == 0)
continue;
handle = dsm_control->item[i].handle;
if (is_main_region_dsm_handle(handle))
continue;
/* Log debugging information. */
elog(DEBUG2, "cleaning up orphaned dynamic shared memory with ID %u",
handle);
/* Destroy the segment. */
dsm_impl_op(DSM_OP_DESTROY, handle, 0, &junk_impl_private,
&junk_mapped_address, &junk_mapped_size, LOG);
}
/* Remove the control segment itself. */
elog(DEBUG2,
"cleaning up dynamic shared memory control segment with ID %u",
dsm_control_handle);
dsm_control_address = dsm_control;
dsm_impl_op(DSM_OP_DESTROY, dsm_control_handle, 0,
&dsm_control_impl_private, &dsm_control_address,
&dsm_control_mapped_size, LOG);
dsm_control = dsm_control_address;
shim->dsm_control = 0;
}
/*
* Prepare this backend for dynamic shared memory usage. Under EXEC_BACKEND,
* we must reread the state file and map the control segment; in other cases,
* we'll have inherited the postmaster's mapping and global variables.
*/
static void
dsm_backend_startup(void)
{
#ifdef EXEC_BACKEND
if (IsUnderPostmaster)
{
void *control_address = NULL;
/* Attach control segment. */
Assert(dsm_control_handle != 0);
dsm_impl_op(DSM_OP_ATTACH, dsm_control_handle, 0,
&dsm_control_impl_private, &control_address,
&dsm_control_mapped_size, ERROR);
dsm_control = control_address;
/* If control segment doesn't look sane, something is badly wrong. */
if (!dsm_control_segment_sane(dsm_control, dsm_control_mapped_size))
{
dsm_impl_op(DSM_OP_DETACH, dsm_control_handle, 0,
&dsm_control_impl_private, &control_address,
&dsm_control_mapped_size, WARNING);
ereport(FATAL,
(errcode(ERRCODE_INTERNAL_ERROR),
errmsg("dynamic shared memory control segment is not valid")));
}
}
#endif
dsm_init_done = true;
}
#ifdef EXEC_BACKEND
/*
* When running under EXEC_BACKEND, we get a callback here when the main
* shared memory segment is re-attached, so that we can record the control
* handle retrieved from it.
*/
void
dsm_set_control_handle(dsm_handle h)
{
Assert(dsm_control_handle == 0 && h != 0);
dsm_control_handle = h;
}
#endif
/*
* Reserve some space in the main shared memory segment for DSM segments.
*/
size_t
dsm_estimate_size(void)
{
return 1024 * 1024 * (size_t) min_dynamic_shared_memory;
}
/*
* Initialize space in the main shared memory segment for DSM segments.
*/
void
dsm_shmem_init(void)
{
size_t size = dsm_estimate_size();
bool found;
if (size == 0)
return;
dsm_main_space_begin = ShmemInitStruct("Preallocated DSM", size, &found);
if (!found)
{
FreePageManager *fpm = (FreePageManager *) dsm_main_space_begin;
size_t first_page = 0;
size_t pages;
/* Reserve space for the FreePageManager. */
while (first_page * FPM_PAGE_SIZE < sizeof(FreePageManager))
++first_page;
/* Initialize it and give it all the rest of the space. */
FreePageManagerInitialize(fpm, dsm_main_space_begin);
pages = (size / FPM_PAGE_SIZE) - first_page;
FreePageManagerPut(fpm, first_page, pages);
}
}
/*
* Create a new dynamic shared memory segment.
*
* If there is a non-NULL CurrentResourceOwner, the new segment is associated
* with it and must be detached before the resource owner releases, or a
* warning will be logged. If CurrentResourceOwner is NULL, the segment
* remains attached until explicitly detached or the session ends.
* Creating with a NULL CurrentResourceOwner is equivalent to creating
* with a non-NULL CurrentResourceOwner and then calling dsm_pin_mapping.
*/
dsm_segment *
dsm_create(Size size, int flags)
{
dsm_segment *seg;
uint32 i;
uint32 nitems;
size_t npages = 0;
size_t first_page = 0;
FreePageManager *dsm_main_space_fpm = dsm_main_space_begin;
bool using_main_dsm_region = false;
/*
* Unsafe in postmaster. It might seem pointless to allow use of dsm in
* single user mode, but otherwise some subsystems will need dedicated
* single user mode code paths.
*/
Assert(IsUnderPostmaster || !IsPostmasterEnvironment);
if (!dsm_init_done)
dsm_backend_startup();
/* Create a new segment descriptor. */
seg = dsm_create_descriptor();
/*
* Lock the control segment while we try to allocate from the main shared
* memory area, if configured.
*/
if (dsm_main_space_fpm)
{
npages = size / FPM_PAGE_SIZE;
if (size % FPM_PAGE_SIZE > 0)
++npages;
LWLockAcquire(DynamicSharedMemoryControlLock, LW_EXCLUSIVE);
if (FreePageManagerGet(dsm_main_space_fpm, npages, &first_page))
{
/* We can carve out a piece of the main shared memory segment. */
seg->mapped_address = (char *) dsm_main_space_begin +
first_page * FPM_PAGE_SIZE;
seg->mapped_size = npages * FPM_PAGE_SIZE;
using_main_dsm_region = true;
/* We'll choose a handle below. */
}
}
if (!using_main_dsm_region)
{
/*
* We need to create a new memory segment. Loop until we find an
* unused segment identifier.
*/
if (dsm_main_space_fpm)
LWLockRelease(DynamicSharedMemoryControlLock);
for (;;)
{
Assert(seg->mapped_address == NULL && seg->mapped_size == 0);
/* Use even numbers only */
seg->handle = pg_prng_uint32(&pg_global_prng_state) << 1;
if (seg->handle == DSM_HANDLE_INVALID) /* Reserve sentinel */
continue;
if (dsm_impl_op(DSM_OP_CREATE, seg->handle, size, &seg->impl_private,
&seg->mapped_address, &seg->mapped_size, ERROR))
break;
}
LWLockAcquire(DynamicSharedMemoryControlLock, LW_EXCLUSIVE);
}
/* Search the control segment for an unused slot. */
nitems = dsm_control->nitems;
for (i = 0; i < nitems; ++i)
{
if (dsm_control->item[i].refcnt == 0)
{
if (using_main_dsm_region)
{
seg->handle = make_main_region_dsm_handle(i);
dsm_control->item[i].first_page = first_page;
dsm_control->item[i].npages = npages;
}
else
Assert(!is_main_region_dsm_handle(seg->handle));
dsm_control->item[i].handle = seg->handle;
/* refcnt of 1 triggers destruction, so start at 2 */
dsm_control->item[i].refcnt = 2;
dsm_control->item[i].impl_private_pm_handle = NULL;
dsm_control->item[i].pinned = false;
seg->control_slot = i;
LWLockRelease(DynamicSharedMemoryControlLock);
return seg;
}
}
/* Verify that we can support an additional mapping. */
if (nitems >= dsm_control->maxitems)
{
if (using_main_dsm_region)
FreePageManagerPut(dsm_main_space_fpm, first_page, npages);
LWLockRelease(DynamicSharedMemoryControlLock);
if (!using_main_dsm_region)
dsm_impl_op(DSM_OP_DESTROY, seg->handle, 0, &seg->impl_private,
&seg->mapped_address, &seg->mapped_size, WARNING);
if (seg->resowner != NULL)
ResourceOwnerForgetDSM(seg->resowner, seg);
dlist_delete(&seg->node);
pfree(seg);
if ((flags & DSM_CREATE_NULL_IF_MAXSEGMENTS) != 0)
return NULL;
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_RESOURCES),
errmsg("too many dynamic shared memory segments")));
}
/* Enter the handle into a new array slot. */
if (using_main_dsm_region)
{
seg->handle = make_main_region_dsm_handle(nitems);
dsm_control->item[i].first_page = first_page;
dsm_control->item[i].npages = npages;
}
dsm_control->item[nitems].handle = seg->handle;
/* refcnt of 1 triggers destruction, so start at 2 */
dsm_control->item[nitems].refcnt = 2;
dsm_control->item[nitems].impl_private_pm_handle = NULL;
dsm_control->item[nitems].pinned = false;
seg->control_slot = nitems;
dsm_control->nitems++;
LWLockRelease(DynamicSharedMemoryControlLock);
return seg;
}
/*
* Attach a dynamic shared memory segment.
*
* See comments for dsm_segment_handle() for an explanation of how this
* is intended to be used.
*
* This function will return NULL if the segment isn't known to the system.
* This can happen if we're asked to attach the segment, but then everyone
* else detaches it (causing it to be destroyed) before we get around to
* attaching it.
*
* If there is a non-NULL CurrentResourceOwner, the attached segment is
* associated with it and must be detached before the resource owner releases,
* or a warning will be logged. Otherwise the segment remains attached until
* explicitly detached or the session ends. See the note atop dsm_create().
*/
dsm_segment *
dsm_attach(dsm_handle h)
{
dsm_segment *seg;
dlist_iter iter;
uint32 i;
uint32 nitems;
/* Unsafe in postmaster (and pointless in a stand-alone backend). */
Assert(IsUnderPostmaster);
if (!dsm_init_done)
dsm_backend_startup();
/*
* Since this is just a debugging cross-check, we could leave it out
* altogether, or include it only in assert-enabled builds. But since the
* list of attached segments should normally be very short, let's include
* it always for right now.
*
* If you're hitting this error, you probably want to attempt to find an
* existing mapping via dsm_find_mapping() before calling dsm_attach() to
* create a new one.
*/
dlist_foreach(iter, &dsm_segment_list)
{
seg = dlist_container(dsm_segment, node, iter.cur);
if (seg->handle == h)
elog(ERROR, "can't attach the same segment more than once");
}
/* Create a new segment descriptor. */
seg = dsm_create_descriptor();
seg->handle = h;
/* Bump reference count for this segment in shared memory. */
LWLockAcquire(DynamicSharedMemoryControlLock, LW_EXCLUSIVE);
nitems = dsm_control->nitems;
for (i = 0; i < nitems; ++i)
{
/*
* If the reference count is 0, the slot is actually unused. If the
* reference count is 1, the slot is still in use, but the segment is
* in the process of going away; even if the handle matches, another
* slot may already have started using the same handle value by
* coincidence so we have to keep searching.
*/
if (dsm_control->item[i].refcnt <= 1)
continue;
/* If the handle doesn't match, it's not the slot we want. */
if (dsm_control->item[i].handle != seg->handle)
continue;
/* Otherwise we've found a match. */
dsm_control->item[i].refcnt++;
seg->control_slot = i;
if (is_main_region_dsm_handle(seg->handle))
{
seg->mapped_address = (char *) dsm_main_space_begin +
dsm_control->item[i].first_page * FPM_PAGE_SIZE;
seg->mapped_size = dsm_control->item[i].npages * FPM_PAGE_SIZE;
}
break;
}
LWLockRelease(DynamicSharedMemoryControlLock);
/*
* If we didn't find the handle we're looking for in the control segment,
* it probably means that everyone else who had it mapped, including the
* original creator, died before we got to this point. It's up to the
* caller to decide what to do about that.
*/
if (seg->control_slot == INVALID_CONTROL_SLOT)
{
dsm_detach(seg);
return NULL;
}
/* Here's where we actually try to map the segment. */
if (!is_main_region_dsm_handle(seg->handle))
dsm_impl_op(DSM_OP_ATTACH, seg->handle, 0, &seg->impl_private,
&seg->mapped_address, &seg->mapped_size, ERROR);
return seg;
}
/*
* At backend shutdown time, detach any segments that are still attached.
* (This is similar to dsm_detach_all, except that there's no reason to
* unmap the control segment before exiting, so we don't bother.)
*/
void
dsm_backend_shutdown(void)
{
while (!dlist_is_empty(&dsm_segment_list))
{
dsm_segment *seg;
seg = dlist_head_element(dsm_segment, node, &dsm_segment_list);
dsm_detach(seg);
}
}
/*
* Detach all shared memory segments, including the control segments. This
* should be called, along with PGSharedMemoryDetach, in processes that
* might inherit mappings but are not intended to be connected to dynamic
* shared memory.
*/
void
dsm_detach_all(void)
{
void *control_address = dsm_control;
while (!dlist_is_empty(&dsm_segment_list))
{
dsm_segment *seg;
seg = dlist_head_element(dsm_segment, node, &dsm_segment_list);
dsm_detach(seg);
}
if (control_address != NULL)
dsm_impl_op(DSM_OP_DETACH, dsm_control_handle, 0,
&dsm_control_impl_private, &control_address,
&dsm_control_mapped_size, ERROR);
}
/*
* Detach from a shared memory segment, destroying the segment if we
* remove the last reference.
*
* This function should never fail. It will often be invoked when aborting
* a transaction, and a further error won't serve any purpose. It's not a
* complete disaster if we fail to unmap or destroy the segment; it means a
* resource leak, but that doesn't necessarily preclude further operations.
*/
void
dsm_detach(dsm_segment *seg)
{
/*
* Invoke registered callbacks. Just in case one of those callbacks
* throws a further error that brings us back here, pop the callback
* before invoking it, to avoid infinite error recursion. Don't allow
* interrupts while running the individual callbacks in non-error code
* paths, to avoid leaving cleanup work unfinished if we're interrupted by
* a statement timeout or similar.
*/
HOLD_INTERRUPTS();
while (!slist_is_empty(&seg->on_detach))
{
slist_node *node;
dsm_segment_detach_callback *cb;
on_dsm_detach_callback function;
Datum arg;
node = slist_pop_head_node(&seg->on_detach);
cb = slist_container(dsm_segment_detach_callback, node, node);
function = cb->function;
arg = cb->arg;
pfree(cb);
function(seg, arg);
}
RESUME_INTERRUPTS();
/*
* Try to remove the mapping, if one exists. Normally, there will be, but
* maybe not, if we failed partway through a create or attach operation.
* We remove the mapping before decrementing the reference count so that
* the process that sees a zero reference count can be certain that no
* remaining mappings exist. Even if this fails, we pretend that it
* works, because retrying is likely to fail in the same way.
*/
if (seg->mapped_address != NULL)
{
if (!is_main_region_dsm_handle(seg->handle))
dsm_impl_op(DSM_OP_DETACH, seg->handle, 0, &seg->impl_private,
&seg->mapped_address, &seg->mapped_size, WARNING);
seg->impl_private = NULL;
seg->mapped_address = NULL;
seg->mapped_size = 0;
}
/* Reduce reference count, if we previously increased it. */
if (seg->control_slot != INVALID_CONTROL_SLOT)
{
uint32 refcnt;
uint32 control_slot = seg->control_slot;
LWLockAcquire(DynamicSharedMemoryControlLock, LW_EXCLUSIVE);
Assert(dsm_control->item[control_slot].handle == seg->handle);
Assert(dsm_control->item[control_slot].refcnt > 1);
refcnt = --dsm_control->item[control_slot].refcnt;
seg->control_slot = INVALID_CONTROL_SLOT;
LWLockRelease(DynamicSharedMemoryControlLock);
/* If new reference count is 1, try to destroy the segment. */
if (refcnt == 1)
{
/* A pinned segment should never reach 1. */
Assert(!dsm_control->item[control_slot].pinned);
/*
* If we fail to destroy the segment here, or are killed before we
* finish doing so, the reference count will remain at 1, which
* will mean that nobody else can attach to the segment. At
* postmaster shutdown time, or when a new postmaster is started
* after a hard kill, another attempt will be made to remove the
* segment.
*
* The main case we're worried about here is being killed by a
* signal before we can finish removing the segment. In that
* case, it's important to be sure that the segment still gets
* removed. If we actually fail to remove the segment for some
* other reason, the postmaster may not have any better luck than
* we did. There's not much we can do about that, though.
*/
if (is_main_region_dsm_handle(seg->handle) ||
dsm_impl_op(DSM_OP_DESTROY, seg->handle, 0, &seg->impl_private,
&seg->mapped_address, &seg->mapped_size, WARNING))
{
LWLockAcquire(DynamicSharedMemoryControlLock, LW_EXCLUSIVE);
if (is_main_region_dsm_handle(seg->handle))
FreePageManagerPut((FreePageManager *) dsm_main_space_begin,
dsm_control->item[control_slot].first_page,
dsm_control->item[control_slot].npages);
Assert(dsm_control->item[control_slot].handle == seg->handle);
Assert(dsm_control->item[control_slot].refcnt == 1);
dsm_control->item[control_slot].refcnt = 0;
LWLockRelease(DynamicSharedMemoryControlLock);
}
}
}
/* Clean up our remaining backend-private data structures. */
if (seg->resowner != NULL)
ResourceOwnerForgetDSM(seg->resowner, seg);
dlist_delete(&seg->node);
pfree(seg);
}
/*
* Keep a dynamic shared memory mapping until end of session.
*
* By default, mappings are owned by the current resource owner, which
* typically means they stick around for the duration of the current query
* only.
*/
void
dsm_pin_mapping(dsm_segment *seg)
{
if (seg->resowner != NULL)
{
ResourceOwnerForgetDSM(seg->resowner, seg);
seg->resowner = NULL;
}
}
/*
* Arrange to remove a dynamic shared memory mapping at cleanup time.
*
* dsm_pin_mapping() can be used to preserve a mapping for the entire
* lifetime of a process; this function reverses that decision, making
* the segment owned by the current resource owner. This may be useful
* just before performing some operation that will invalidate the segment
* for future use by this backend.
*/
void
dsm_unpin_mapping(dsm_segment *seg)
{
Assert(seg->resowner == NULL);
Make ResourceOwners more easily extensible. Instead of having a separate array/hash for each resource kind, use a single array and hash to hold all kinds of resources. This makes it possible to introduce new resource "kinds" without having to modify the ResourceOwnerData struct. In particular, this makes it possible for extensions to register custom resource kinds. The old approach was to have a small array of resources of each kind, and if it fills up, switch to a hash table. The new approach also uses an array and a hash, but now the array and the hash are used at the same time. The array is used to hold the recently added resources, and when it fills up, they are moved to the hash. This keeps the access to recent entries fast, even when there are a lot of long-held resources. All the resource-specific ResourceOwnerEnlarge*(), ResourceOwnerRemember*(), and ResourceOwnerForget*() functions have been replaced with three generic functions that take resource kind as argument. For convenience, we still define resource-specific wrapper macros around the generic functions with the old names, but they are now defined in the source files that use those resource kinds. The release callback no longer needs to call ResourceOwnerForget on the resource being released. ResourceOwnerRelease unregisters the resource from the owner before calling the callback. That needed some changes in bufmgr.c and some other files, where releasing the resources previously always called ResourceOwnerForget. Each resource kind specifies a release priority, and ResourceOwnerReleaseAll releases the resources in priority order. To make that possible, we have to restrict what you can do between phases. After calling ResourceOwnerRelease(), you are no longer allowed to remember any more resources in it or to forget any previously remembered resources by calling ResourceOwnerForget. There was one case where that was done previously. At subtransaction commit, AtEOSubXact_Inval() would handle the invalidation messages and call RelationFlushRelation(), which temporarily increased the reference count on the relation being flushed. We now switch to the parent subtransaction's resource owner before calling AtEOSubXact_Inval(), so that there is a valid ResourceOwner to temporarily hold that relcache reference. Other end-of-xact routines make similar calls to AtEOXact_Inval() between release phases, but I didn't see any regression test failures from those, so I'm not sure if they could reach a codepath that needs remembering extra resources. There were two exceptions to how the resource leak WARNINGs on commit were printed previously: llvmjit silently released the context without printing the warning, and a leaked buffer io triggered a PANIC. Now everything prints a WARNING, including those cases. Add tests in src/test/modules/test_resowner. Reviewed-by: Aleksander Alekseev, Michael Paquier, Julien Rouhaud Reviewed-by: Kyotaro Horiguchi, Hayato Kuroda, Álvaro Herrera, Zhihong Yu Reviewed-by: Peter Eisentraut, Andres Freund Discussion: https://www.postgresql.org/message-id/cbfabeb0-cd3c-e951-a572-19b365ed314d%40iki.fi
2023-11-08 12:30:50 +01:00
ResourceOwnerEnlarge(CurrentResourceOwner);
seg->resowner = CurrentResourceOwner;
ResourceOwnerRememberDSM(seg->resowner, seg);
}
/*
* Keep a dynamic shared memory segment until postmaster shutdown, or until
* dsm_unpin_segment is called.
*
* This function should not be called more than once per segment, unless the
* segment is explicitly unpinned with dsm_unpin_segment in between calls.
*
* Note that this function does not arrange for the current process to
* keep the segment mapped indefinitely; if that behavior is desired,
* dsm_pin_mapping() should be used from each process that needs to
* retain the mapping.
*/
void
dsm_pin_segment(dsm_segment *seg)
{
void *handle = NULL;
/*
* Bump reference count for this segment in shared memory. This will
* ensure that even if there is no session which is attached to this
* segment, it will remain until postmaster shutdown or an explicit call
* to unpin.
*/
LWLockAcquire(DynamicSharedMemoryControlLock, LW_EXCLUSIVE);
if (dsm_control->item[seg->control_slot].pinned)
elog(ERROR, "cannot pin a segment that is already pinned");
if (!is_main_region_dsm_handle(seg->handle))
dsm_impl_pin_segment(seg->handle, seg->impl_private, &handle);
dsm_control->item[seg->control_slot].pinned = true;
dsm_control->item[seg->control_slot].refcnt++;
dsm_control->item[seg->control_slot].impl_private_pm_handle = handle;
LWLockRelease(DynamicSharedMemoryControlLock);
}
/*
* Unpin a dynamic shared memory segment that was previously pinned with
* dsm_pin_segment. This function should not be called unless dsm_pin_segment
* was previously called for this segment.
*
* The argument is a dsm_handle rather than a dsm_segment in case you want
* to unpin a segment to which you haven't attached. This turns out to be
* useful if, for example, a reference to one shared memory segment is stored
* within another shared memory segment. You might want to unpin the
* referenced segment before destroying the referencing segment.
*/
void
dsm_unpin_segment(dsm_handle handle)
{
uint32 control_slot = INVALID_CONTROL_SLOT;
bool destroy = false;
uint32 i;
/* Find the control slot for the given handle. */
LWLockAcquire(DynamicSharedMemoryControlLock, LW_EXCLUSIVE);
for (i = 0; i < dsm_control->nitems; ++i)
{
/* Skip unused slots and segments that are concurrently going away. */
if (dsm_control->item[i].refcnt <= 1)
continue;
/* If we've found our handle, we can stop searching. */
if (dsm_control->item[i].handle == handle)
{
control_slot = i;
break;
}
}
/*
* We should definitely have found the slot, and it should not already be
* in the process of going away, because this function should only be
* called on a segment which is pinned.
*/
if (control_slot == INVALID_CONTROL_SLOT)
elog(ERROR, "cannot unpin unknown segment handle");
if (!dsm_control->item[control_slot].pinned)
elog(ERROR, "cannot unpin a segment that is not pinned");
Assert(dsm_control->item[control_slot].refcnt > 1);
/*
* Allow implementation-specific code to run. We have to do this before
* releasing the lock, because impl_private_pm_handle may get modified by
* dsm_impl_unpin_segment.
*/
if (!is_main_region_dsm_handle(handle))
dsm_impl_unpin_segment(handle,
&dsm_control->item[control_slot].impl_private_pm_handle);
/* Note that 1 means no references (0 means unused slot). */
if (--dsm_control->item[control_slot].refcnt == 1)
destroy = true;
dsm_control->item[control_slot].pinned = false;
/* Now we can release the lock. */
LWLockRelease(DynamicSharedMemoryControlLock);
/* Clean up resources if that was the last reference. */
if (destroy)
{
void *junk_impl_private = NULL;
void *junk_mapped_address = NULL;
Size junk_mapped_size = 0;
/*
* For an explanation of how error handling works in this case, see
* comments in dsm_detach. Note that if we reach this point, the
* current process certainly does not have the segment mapped, because
* if it did, the reference count would have still been greater than 1
* even after releasing the reference count held by the pin. The fact
* that there can't be a dsm_segment for this handle makes it OK to
* pass the mapped size, mapped address, and private data as NULL
* here.
*/
if (is_main_region_dsm_handle(handle) ||
dsm_impl_op(DSM_OP_DESTROY, handle, 0, &junk_impl_private,
&junk_mapped_address, &junk_mapped_size, WARNING))
{
LWLockAcquire(DynamicSharedMemoryControlLock, LW_EXCLUSIVE);
if (is_main_region_dsm_handle(handle))
FreePageManagerPut((FreePageManager *) dsm_main_space_begin,
dsm_control->item[control_slot].first_page,
dsm_control->item[control_slot].npages);
Assert(dsm_control->item[control_slot].handle == handle);
Assert(dsm_control->item[control_slot].refcnt == 1);
dsm_control->item[control_slot].refcnt = 0;
LWLockRelease(DynamicSharedMemoryControlLock);
}
}
}
/*
* Find an existing mapping for a shared memory segment, if there is one.
*/
dsm_segment *
dsm_find_mapping(dsm_handle handle)
{
dlist_iter iter;
dsm_segment *seg;
dlist_foreach(iter, &dsm_segment_list)
{
seg = dlist_container(dsm_segment, node, iter.cur);
if (seg->handle == handle)
return seg;
}
return NULL;
}
/*
* Get the address at which a dynamic shared memory segment is mapped.
*/
void *
dsm_segment_address(dsm_segment *seg)
{
Assert(seg->mapped_address != NULL);
return seg->mapped_address;
}
/*
* Get the size of a mapping.
*/
Size
dsm_segment_map_length(dsm_segment *seg)
{
Assert(seg->mapped_address != NULL);
return seg->mapped_size;
}
/*
* Get a handle for a mapping.
*
* To establish communication via dynamic shared memory between two backends,
* one of them should first call dsm_create() to establish a new shared
* memory mapping. That process should then call dsm_segment_handle() to
* obtain a handle for the mapping, and pass that handle to the
* coordinating backend via some means (e.g. bgw_main_arg, or via the
2016-06-07 18:34:33 +02:00
* main shared memory segment). The recipient, once in possession of the
* handle, should call dsm_attach().
*/
dsm_handle
dsm_segment_handle(dsm_segment *seg)
{
return seg->handle;
}
/*
* Register an on-detach callback for a dynamic shared memory segment.
*/
void
on_dsm_detach(dsm_segment *seg, on_dsm_detach_callback function, Datum arg)
{
dsm_segment_detach_callback *cb;
cb = MemoryContextAlloc(TopMemoryContext,
sizeof(dsm_segment_detach_callback));
cb->function = function;
cb->arg = arg;
slist_push_head(&seg->on_detach, &cb->node);
}
/*
* Unregister an on-detach callback for a dynamic shared memory segment.
*/
void
cancel_on_dsm_detach(dsm_segment *seg, on_dsm_detach_callback function,
Datum arg)
{
slist_mutable_iter iter;
slist_foreach_modify(iter, &seg->on_detach)
{
dsm_segment_detach_callback *cb;
cb = slist_container(dsm_segment_detach_callback, node, iter.cur);
if (cb->function == function && cb->arg == arg)
{
slist_delete_current(&iter);
pfree(cb);
break;
}
}
}
/*
* Discard all registered on-detach callbacks without executing them.
*/
void
reset_on_dsm_detach(void)
{
dlist_iter iter;
dlist_foreach(iter, &dsm_segment_list)
{
dsm_segment *seg = dlist_container(dsm_segment, node, iter.cur);
/* Throw away explicit on-detach actions one by one. */
while (!slist_is_empty(&seg->on_detach))
{
slist_node *node;
dsm_segment_detach_callback *cb;
node = slist_pop_head_node(&seg->on_detach);
cb = slist_container(dsm_segment_detach_callback, node, node);
pfree(cb);
}
/*
* Decrementing the reference count is a sort of implicit on-detach
* action; make sure we don't do that, either.
*/
seg->control_slot = INVALID_CONTROL_SLOT;
}
}
/*
* Create a segment descriptor.
*/
static dsm_segment *
dsm_create_descriptor(void)
{
dsm_segment *seg;
if (CurrentResourceOwner)
Make ResourceOwners more easily extensible. Instead of having a separate array/hash for each resource kind, use a single array and hash to hold all kinds of resources. This makes it possible to introduce new resource "kinds" without having to modify the ResourceOwnerData struct. In particular, this makes it possible for extensions to register custom resource kinds. The old approach was to have a small array of resources of each kind, and if it fills up, switch to a hash table. The new approach also uses an array and a hash, but now the array and the hash are used at the same time. The array is used to hold the recently added resources, and when it fills up, they are moved to the hash. This keeps the access to recent entries fast, even when there are a lot of long-held resources. All the resource-specific ResourceOwnerEnlarge*(), ResourceOwnerRemember*(), and ResourceOwnerForget*() functions have been replaced with three generic functions that take resource kind as argument. For convenience, we still define resource-specific wrapper macros around the generic functions with the old names, but they are now defined in the source files that use those resource kinds. The release callback no longer needs to call ResourceOwnerForget on the resource being released. ResourceOwnerRelease unregisters the resource from the owner before calling the callback. That needed some changes in bufmgr.c and some other files, where releasing the resources previously always called ResourceOwnerForget. Each resource kind specifies a release priority, and ResourceOwnerReleaseAll releases the resources in priority order. To make that possible, we have to restrict what you can do between phases. After calling ResourceOwnerRelease(), you are no longer allowed to remember any more resources in it or to forget any previously remembered resources by calling ResourceOwnerForget. There was one case where that was done previously. At subtransaction commit, AtEOSubXact_Inval() would handle the invalidation messages and call RelationFlushRelation(), which temporarily increased the reference count on the relation being flushed. We now switch to the parent subtransaction's resource owner before calling AtEOSubXact_Inval(), so that there is a valid ResourceOwner to temporarily hold that relcache reference. Other end-of-xact routines make similar calls to AtEOXact_Inval() between release phases, but I didn't see any regression test failures from those, so I'm not sure if they could reach a codepath that needs remembering extra resources. There were two exceptions to how the resource leak WARNINGs on commit were printed previously: llvmjit silently released the context without printing the warning, and a leaked buffer io triggered a PANIC. Now everything prints a WARNING, including those cases. Add tests in src/test/modules/test_resowner. Reviewed-by: Aleksander Alekseev, Michael Paquier, Julien Rouhaud Reviewed-by: Kyotaro Horiguchi, Hayato Kuroda, Álvaro Herrera, Zhihong Yu Reviewed-by: Peter Eisentraut, Andres Freund Discussion: https://www.postgresql.org/message-id/cbfabeb0-cd3c-e951-a572-19b365ed314d%40iki.fi
2023-11-08 12:30:50 +01:00
ResourceOwnerEnlarge(CurrentResourceOwner);
seg = MemoryContextAlloc(TopMemoryContext, sizeof(dsm_segment));
dlist_push_head(&dsm_segment_list, &seg->node);
/* seg->handle must be initialized by the caller */
seg->control_slot = INVALID_CONTROL_SLOT;
seg->impl_private = NULL;
seg->mapped_address = NULL;
seg->mapped_size = 0;
seg->resowner = CurrentResourceOwner;
if (CurrentResourceOwner)
ResourceOwnerRememberDSM(CurrentResourceOwner, seg);
slist_init(&seg->on_detach);
return seg;
}
/*
* Sanity check a control segment.
*
* The goal here isn't to detect everything that could possibly be wrong with
* the control segment; there's not enough information for that. Rather, the
* goal is to make sure that someone can iterate over the items in the segment
* without overrunning the end of the mapping and crashing. We also check
* the magic number since, if that's messed up, this may not even be one of
* our segments at all.
*/
static bool
dsm_control_segment_sane(dsm_control_header *control, Size mapped_size)
{
if (mapped_size < offsetof(dsm_control_header, item))
return false; /* Mapped size too short to read header. */
if (control->magic != PG_DYNSHMEM_CONTROL_MAGIC)
return false; /* Magic number doesn't match. */
if (dsm_control_bytes_needed(control->maxitems) > mapped_size)
return false; /* Max item count won't fit in map. */
if (control->nitems > control->maxitems)
return false; /* Overfull. */
return true;
}
/*
* Compute the number of control-segment bytes needed to store a given
* number of items.
*/
static uint64
dsm_control_bytes_needed(uint32 nitems)
{
return offsetof(dsm_control_header, item)
+ sizeof(dsm_control_item) * (uint64) nitems;
}
static inline dsm_handle
make_main_region_dsm_handle(int slot)
{
dsm_handle handle;
/*
* We need to create a handle that doesn't collide with any existing extra
* segment created by dsm_impl_op(), so we'll make it odd. It also
* mustn't collide with any other main area pseudo-segment, so we'll
* include the slot number in some of the bits. We also want to make an
* effort to avoid newly created and recently destroyed handles from being
* confused, so we'll make the rest of the bits random.
*/
handle = 1;
handle |= slot << 1;
handle |= pg_prng_uint32(&pg_global_prng_state) << (pg_leftmost_one_pos32(dsm_control->maxitems) + 1);
return handle;
}
static inline bool
is_main_region_dsm_handle(dsm_handle handle)
{
return handle & 1;
}
Make ResourceOwners more easily extensible. Instead of having a separate array/hash for each resource kind, use a single array and hash to hold all kinds of resources. This makes it possible to introduce new resource "kinds" without having to modify the ResourceOwnerData struct. In particular, this makes it possible for extensions to register custom resource kinds. The old approach was to have a small array of resources of each kind, and if it fills up, switch to a hash table. The new approach also uses an array and a hash, but now the array and the hash are used at the same time. The array is used to hold the recently added resources, and when it fills up, they are moved to the hash. This keeps the access to recent entries fast, even when there are a lot of long-held resources. All the resource-specific ResourceOwnerEnlarge*(), ResourceOwnerRemember*(), and ResourceOwnerForget*() functions have been replaced with three generic functions that take resource kind as argument. For convenience, we still define resource-specific wrapper macros around the generic functions with the old names, but they are now defined in the source files that use those resource kinds. The release callback no longer needs to call ResourceOwnerForget on the resource being released. ResourceOwnerRelease unregisters the resource from the owner before calling the callback. That needed some changes in bufmgr.c and some other files, where releasing the resources previously always called ResourceOwnerForget. Each resource kind specifies a release priority, and ResourceOwnerReleaseAll releases the resources in priority order. To make that possible, we have to restrict what you can do between phases. After calling ResourceOwnerRelease(), you are no longer allowed to remember any more resources in it or to forget any previously remembered resources by calling ResourceOwnerForget. There was one case where that was done previously. At subtransaction commit, AtEOSubXact_Inval() would handle the invalidation messages and call RelationFlushRelation(), which temporarily increased the reference count on the relation being flushed. We now switch to the parent subtransaction's resource owner before calling AtEOSubXact_Inval(), so that there is a valid ResourceOwner to temporarily hold that relcache reference. Other end-of-xact routines make similar calls to AtEOXact_Inval() between release phases, but I didn't see any regression test failures from those, so I'm not sure if they could reach a codepath that needs remembering extra resources. There were two exceptions to how the resource leak WARNINGs on commit were printed previously: llvmjit silently released the context without printing the warning, and a leaked buffer io triggered a PANIC. Now everything prints a WARNING, including those cases. Add tests in src/test/modules/test_resowner. Reviewed-by: Aleksander Alekseev, Michael Paquier, Julien Rouhaud Reviewed-by: Kyotaro Horiguchi, Hayato Kuroda, Álvaro Herrera, Zhihong Yu Reviewed-by: Peter Eisentraut, Andres Freund Discussion: https://www.postgresql.org/message-id/cbfabeb0-cd3c-e951-a572-19b365ed314d%40iki.fi
2023-11-08 12:30:50 +01:00
/* ResourceOwner callbacks */
static void
ResOwnerReleaseDSM(Datum res)
{
dsm_segment *seg = (dsm_segment *) DatumGetPointer(res);
seg->resowner = NULL;
dsm_detach(seg);
}
static char *
ResOwnerPrintDSM(Datum res)
{
dsm_segment *seg = (dsm_segment *) DatumGetPointer(res);
return psprintf("dynamic shared memory segment %u",
dsm_segment_handle(seg));
}