/* ----------------------------------------------------------------------------- * (c) The GHC Team 1998-2005 * * STM implementation. * * Overview * -------- * * See the PPoPP 2005 paper "Composable memory transactions". In summary, each * transaction has a TRec (transaction record) holding entries for each of the * TVars (transactional variables) that it has accessed. Each entry records (a) * the TVar, (b) the expected value seen in the TVar, (c) the new value that the * transaction wants to write to the TVar, (d) during commit, the identity of * the TRec that wrote the expected value. * * Separate TRecs are used for each level in a nest of transactions. This * allows a nested transaction to be aborted without condemning its enclosing * transactions. This is needed in the implementation of catchRetry. Note that * the "expected value" in a nested transaction's TRec is the value expected to * be *held in memory* if the transaction commits -- not the "new value" stored * in one of the enclosing transactions. This means that validation can be done * without searching through a nest of TRecs. * * Concurrency control * ------------------- * * Three different concurrency control schemes can be built according to the * settings in STM.h: * * STM_UNIPROC assumes that the caller serialises invocations on the STM * interface. In the Haskell RTS this means it is suitable only for * non-THREADED_RTS builds. * * STM_CG_LOCK uses coarse-grained locking -- a single 'stm lock' is acquired * during an invocation on the STM interface. Note that this does not mean that * transactions are simply serialized -- the lock is only held *within* the * implementation of stmCommitTransaction, stmWait etc. * * STM_FG_LOCKS uses fine-grained locking -- locking is done on a per-TVar basis * and, when committing a transaction, no locks are acquired for TVars that have * been read but not updated. * * Concurrency control is implemented in the functions: * * lock_stm * unlock_stm * lock_tvar / cond_lock_tvar * unlock_tvar * * The choice between STM_UNIPROC / STM_CG_LOCK / STM_FG_LOCKS affects the * implementation of these functions. * * lock_stm & unlock_stm are straightforward : they acquire a simple spin-lock * using STM_CG_LOCK, and otherwise they are no-ops. * * lock_tvar / cond_lock_tvar and unlock_tvar are more complex because they have * other effects (present in STM_UNIPROC and STM_CG_LOCK builds) as well as the * actual business of manipulating a lock (present only in STM_FG_LOCKS builds). * This is because locking a TVar is implemented by writing the lock holder's * TRec into the TVar's current_value field: * * lock_tvar - lock a specified TVar (STM_FG_LOCKS only), returning the value * it contained. * * cond_lock_tvar - lock a specified TVar (STM_FG_LOCKS only) if it * contains a specified value. Return true if this succeeds, * false otherwise. * * unlock_tvar - release the lock on a specified TVar (STM_FG_LOCKS only), * storing a specified value in place of the lock entry. * * Using these operations, the typical pattern of a commit/validate/wait * operation is to (a) lock the STM, (b) lock all the TVars being updated, (c) * check that the TVars that were only read from still contain their expected * values, (d) release the locks on the TVars, writing updates to them in the * case of a commit, (e) unlock the STM. * * Queues of waiting threads hang off the first_watch_queue_entry field of each * TVar. This may only be manipulated when holding that TVar's lock. In * particular, when a thread is putting itself to sleep, it mustn't release the * TVar's lock until it has added itself to the wait queue and marked its TSO as * BlockedOnSTM -- this makes sure that other threads will know to wake it. * * ---------------------------------------------------------------------------*/ #include "rts/PosixSource.h" #include "Rts.h" #include "RtsUtils.h" #include "Schedule.h" #include "STM.h" #include "Trace.h" #include "Threads.h" #include "sm/Storage.h" #include "SMPClosureOps.h" #include // ACQ_ASSERT is used for assertions which are only required for // THREADED_RTS builds with fine-grained locking. #if defined(STM_FG_LOCKS) #define ACQ_ASSERT(_X) ASSERT(_X) #define NACQ_ASSERT(_X) /*Nothing*/ #else #define ACQ_ASSERT(_X) /*Nothing*/ #define NACQ_ASSERT(_X) ASSERT(_X) #endif /*......................................................................*/ #define TRACE(_x...) debugTrace(DEBUG_stm, "STM: " _x) // If SHAKE is defined then validation will sometimes spuriously fail. They help test // unusual code paths if genuine contention is rare #if defined(SHAKE) static int shake_ctr = 0; static int shake_lim = 1; static int shake(void) { if (((shake_ctr++) % shake_lim) == 0) { shake_ctr = 1; shake_lim ++; return true; } return false; } #else static int shake(void) { return false; } #endif /*......................................................................*/ // Helper macros for iterating over entries within a transaction // record #define FOR_EACH_ENTRY(_t,_x,CODE) do { \ StgTRecHeader *__t = (_t); \ StgTRecChunk *__c = __t -> current_chunk; \ StgWord __limit = __c -> next_entry_idx; \ TRACE("%p : FOR_EACH_ENTRY, current_chunk=%p limit=%ld", __t, __c, __limit); \ while (__c != END_STM_CHUNK_LIST) { \ StgWord __i; \ for (__i = 0; __i < __limit; __i ++) { \ TRecEntry *_x = &(__c -> entries[__i]); \ do { CODE } while (0); \ } \ __c = __c -> prev_chunk; \ __limit = TREC_CHUNK_NUM_ENTRIES; \ } \ exit_for_each: \ if (false) goto exit_for_each; \ } while (0) #define BREAK_FOR_EACH goto exit_for_each /*......................................................................*/ // if REUSE_MEMORY is defined then attempt to re-use descriptors, log chunks, // and wait queue entries without GC #define REUSE_MEMORY /*......................................................................*/ #define IF_STM_UNIPROC(__X) do { } while (0) #define IF_STM_CG_LOCK(__X) do { } while (0) #define IF_STM_FG_LOCKS(__X) do { } while (0) #if defined(STM_UNIPROC) #undef IF_STM_UNIPROC #define IF_STM_UNIPROC(__X) do { __X } while (0) static const StgBool config_use_read_phase = false; static void lock_stm(StgTRecHeader *trec STG_UNUSED) { TRACE("%p : lock_stm()", trec); } static void unlock_stm(StgTRecHeader *trec STG_UNUSED) { TRACE("%p : unlock_stm()", trec); } static StgClosure *lock_tvar(Capability *cap STG_UNUSED, StgTRecHeader *trec STG_UNUSED, StgTVar *s STG_UNUSED) { StgClosure *result; TRACE("%p : lock_tvar(%p)", trec, s); result = SEQ_CST_LOAD(&s->current_value); return result; } static void unlock_tvar(Capability *cap, StgTRecHeader *trec STG_UNUSED, StgTVar *s, StgClosure *c, StgBool force_update) { TRACE("%p : unlock_tvar(%p)", trec, s); if (force_update) { StgClosure *old_value = SEQ_CST_LOAD(&s->current_value); RELEASE_STORE(&s->current_value, c); dirty_TVAR(cap, s, old_value); } } static StgBool cond_lock_tvar(Capability *cap STG_UNUSED, StgTRecHeader *trec STG_UNUSED, StgTVar *s STG_UNUSED, StgClosure *expected) { StgClosure *result; TRACE("%p : cond_lock_tvar(%p, %p)", trec, s, expected); result = SEQ_CST_LOAD(&s->current_value); TRACE("%p : %s", trec, (result == expected) ? "success" : "failure"); return (result == expected); } #endif #if defined(STM_CG_LOCK) /*........................................*/ #undef IF_STM_CG_LOCK #define IF_STM_CG_LOCK(__X) do { __X } while (0) static const StgBool config_use_read_phase = false; static volatile StgTRecHeader *smp_locked = NULL; static void lock_stm(StgTRecHeader *trec) { while (cas(&smp_locked, NULL, trec) != NULL) { } TRACE("%p : lock_stm()", trec); } static void unlock_stm(StgTRecHeader *trec STG_UNUSED) { TRACE("%p : unlock_stm()", trec); ASSERT(smp_locked == trec); SEQ_CST_STORE(&smp_locked, 0); } static StgClosure *lock_tvar(Capability *cap STG_UNUSED, StgTRecHeader *trec STG_UNUSED, StgTVar *s STG_UNUSED) { StgClosure *result; TRACE("%p : lock_tvar(%p)", trec, s); ASSERT(smp_locked == trec); result = SEQ_CST_LOAD(&s->current_value); return result; } static void *unlock_tvar(Capability *cap, StgTRecHeader *trec STG_UNUSED, StgTVar *s, StgClosure *c, StgBool force_update) { TRACE("%p : unlock_tvar(%p, %p)", trec, s, c); ASSERT(smp_locked == trec); if (force_update) { StgClosure *old_value = SEQ_CST_LOAD(&s->current_value); RELEASE_STORE(&s->current_value, c); dirty_TVAR(cap, s, old_value); } } static StgBool cond_lock_tvar(Capability *cap STG_UNUSED, StgTRecHeader *trec STG_UNUSED, StgTVar *s STG_UNUSED, StgClosure *expected) { StgClosure *result; TRACE("%p : cond_lock_tvar(%p, %p)", trec, s, expected); ASSERT(smp_locked == trec); result = SEQ_CST_LOAD(&s->current_value); TRACE("%p : %d", result ? "success" : "failure"); return (result == expected); } #endif #if defined(STM_FG_LOCKS) /*...................................*/ #undef IF_STM_FG_LOCKS #define IF_STM_FG_LOCKS(__X) do { __X } while (0) static const StgBool config_use_read_phase = true; static void lock_stm(StgTRecHeader *trec STG_UNUSED) { TRACE("%p : lock_stm()", trec); } static void unlock_stm(StgTRecHeader *trec STG_UNUSED) { TRACE("%p : unlock_stm()", trec); } static StgClosure *lock_tvar(Capability *cap, StgTRecHeader *trec, StgTVar *s STG_UNUSED) { StgClosure *result; TRACE("%p : lock_tvar(%p)", trec, s); do { do { result = SEQ_CST_LOAD(&s->current_value); } while (GET_INFO(UNTAG_CLOSURE(result)) == &stg_TREC_HEADER_info); } while (cas((void *) &s->current_value, (StgWord)result, (StgWord)trec) != (StgWord)result); IF_NONMOVING_WRITE_BARRIER_ENABLED { if (result) updateRemembSetPushClosure(cap, result); } return result; } static void unlock_tvar(Capability *cap, StgTRecHeader *trec STG_UNUSED, StgTVar *s, StgClosure *c, StgBool force_update STG_UNUSED) { TRACE("%p : unlock_tvar(%p, %p)", trec, s, c); ASSERT(SEQ_CST_LOAD(&s->current_value) == (StgClosure *)trec); RELEASE_STORE(&s->current_value, c); dirty_TVAR(cap, s, (StgClosure *) trec); } static StgBool cond_lock_tvar(Capability *cap, StgTRecHeader *trec, StgTVar *s, StgClosure *expected) { StgClosure *result; StgWord w; TRACE("%p : cond_lock_tvar(%p, %p)", trec, s, expected); w = cas((void *)&(s -> current_value), (StgWord)expected, (StgWord)trec); result = (StgClosure *)w; IF_NONMOVING_WRITE_BARRIER_ENABLED { if (result) updateRemembSetPushClosure(cap, expected); } TRACE("%p : %s", trec, result ? "success" : "failure"); return (result == expected); } #endif /*......................................................................*/ // Helper functions for thread blocking and unblocking static void park_tso(StgTSO *tso) { ASSERT(tso -> why_blocked == NotBlocked); tso -> why_blocked = BlockedOnSTM; tso -> block_info.closure = (StgClosure *) END_TSO_QUEUE; TRACE("park_tso on tso=%p", tso); } static void unpark_tso(Capability *cap, StgTSO *tso) { // We will continue unparking threads while they remain on one of the wait // queues: it's up to the thread itself to remove it from the wait queues // if it decides to do so when it is scheduled. // Only the capability that owns this TSO may unblock it. We can // call tryWakeupThread() which will either unblock it directly if // it belongs to this cap, or send a message to the owning cap // otherwise. // TODO: This sends multiple messages if we write to the same TVar multiple // times and the owning cap hasn't yet woken up the thread and removed it // from the TVar's watch list. We tried to optimise this in D4961, but that // patch was incorrect and broke other things, see #15544 comment:17. See // #15626 for the tracking ticket. // Safety Note: we hold the TVar lock at this point, so we know // that this thread is definitely still blocked, since the first // thing a thread will do when it runs is remove itself from the // TVar watch queues, and to do that it would need to lock the // TVar. tryWakeupThread(cap,tso); } static void unpark_waiters_on(Capability *cap, StgTVar *s) { StgTVarWatchQueue *q; StgTVarWatchQueue *trail; TRACE("unpark_waiters_on tvar=%p", s); // unblock TSOs in reverse order, to be a bit fairer (#2319) for (q = SEQ_CST_LOAD(&s->first_watch_queue_entry), trail = q; q != END_STM_WATCH_QUEUE; q = q -> next_queue_entry) { trail = q; } q = trail; for (; q != END_STM_WATCH_QUEUE; q = q -> prev_queue_entry) { unpark_tso(cap, (StgTSO *)(q -> closure)); } } /*......................................................................*/ // Helper functions for downstream allocation and initialization static StgTVarWatchQueue *new_stg_tvar_watch_queue(Capability *cap, StgClosure *closure) { StgTVarWatchQueue *result; result = (StgTVarWatchQueue *)allocate(cap, sizeofW(StgTVarWatchQueue)); SET_HDR (result, &stg_TVAR_WATCH_QUEUE_info, CCS_SYSTEM); result -> closure = closure; return result; } static StgTRecChunk *new_stg_trec_chunk(Capability *cap) { StgTRecChunk *result; result = (StgTRecChunk *)allocate(cap, sizeofW(StgTRecChunk)); SET_HDR (result, &stg_TREC_CHUNK_info, CCS_SYSTEM); result -> prev_chunk = END_STM_CHUNK_LIST; result -> next_entry_idx = 0; return result; } static StgTRecHeader *new_stg_trec_header(Capability *cap, StgTRecHeader *enclosing_trec) { StgTRecHeader *result; result = (StgTRecHeader *) allocate(cap, sizeofW(StgTRecHeader)); SET_HDR (result, &stg_TREC_HEADER_info, CCS_SYSTEM); result -> enclosing_trec = enclosing_trec; result -> current_chunk = new_stg_trec_chunk(cap); if (enclosing_trec == NO_TREC) { result -> state = TREC_ACTIVE; } else { ASSERT(enclosing_trec -> state == TREC_ACTIVE || enclosing_trec -> state == TREC_CONDEMNED); result -> state = enclosing_trec -> state; } return result; } /*......................................................................*/ // Allocation / deallocation functions that retain per-capability lists // of closures that can be re-used static StgTVarWatchQueue *alloc_stg_tvar_watch_queue(Capability *cap, StgClosure *closure) { StgTVarWatchQueue *result = NULL; if (cap -> free_tvar_watch_queues == END_STM_WATCH_QUEUE) { result = new_stg_tvar_watch_queue(cap, closure); } else { result = cap -> free_tvar_watch_queues; result -> closure = closure; cap -> free_tvar_watch_queues = result -> next_queue_entry; } return result; } static void free_stg_tvar_watch_queue(Capability *cap, StgTVarWatchQueue *wq) { #if defined(REUSE_MEMORY) wq -> next_queue_entry = cap -> free_tvar_watch_queues; cap -> free_tvar_watch_queues = wq; #endif } static StgTRecChunk *alloc_stg_trec_chunk(Capability *cap) { StgTRecChunk *result = NULL; if (cap -> free_trec_chunks == END_STM_CHUNK_LIST) { result = new_stg_trec_chunk(cap); } else { result = cap -> free_trec_chunks; cap -> free_trec_chunks = result -> prev_chunk; result -> prev_chunk = END_STM_CHUNK_LIST; result -> next_entry_idx = 0; } return result; } static void free_stg_trec_chunk(Capability *cap, StgTRecChunk *c) { #if defined(REUSE_MEMORY) c -> prev_chunk = cap -> free_trec_chunks; cap -> free_trec_chunks = c; #endif } static StgTRecHeader *alloc_stg_trec_header(Capability *cap, StgTRecHeader *enclosing_trec) { StgTRecHeader *result = NULL; if (cap -> free_trec_headers == NO_TREC) { result = new_stg_trec_header(cap, enclosing_trec); } else { result = cap -> free_trec_headers; cap -> free_trec_headers = result -> enclosing_trec; result -> enclosing_trec = enclosing_trec; result -> current_chunk -> next_entry_idx = 0; if (enclosing_trec == NO_TREC) { result -> state = TREC_ACTIVE; } else { ASSERT(enclosing_trec -> state == TREC_ACTIVE || enclosing_trec -> state == TREC_CONDEMNED); result -> state = enclosing_trec -> state; } } return result; } static void free_stg_trec_header(Capability *cap, StgTRecHeader *trec) { #if defined(REUSE_MEMORY) StgTRecChunk *chunk = trec -> current_chunk -> prev_chunk; while (chunk != END_STM_CHUNK_LIST) { StgTRecChunk *prev_chunk = chunk -> prev_chunk; free_stg_trec_chunk(cap, chunk); chunk = prev_chunk; } trec -> current_chunk -> prev_chunk = END_STM_CHUNK_LIST; trec -> enclosing_trec = cap -> free_trec_headers; cap -> free_trec_headers = trec; #endif } /*......................................................................*/ // Helper functions for managing waiting lists static void build_watch_queue_entries_for_trec(Capability *cap, StgTSO *tso, StgTRecHeader *trec) { ASSERT(trec != NO_TREC); ASSERT(trec -> enclosing_trec == NO_TREC); ASSERT(trec -> state == TREC_ACTIVE); TRACE("%p : build_watch_queue_entries_for_trec()", trec); FOR_EACH_ENTRY(trec, e, { StgTVar *s; StgTVarWatchQueue *q; StgTVarWatchQueue *fq; s = e -> tvar; TRACE("%p : adding tso=%p to watch queue for tvar=%p", trec, tso, s); ACQ_ASSERT(SEQ_CST_LOAD(&s->current_value) == (StgClosure *)trec); NACQ_ASSERT(SEQ_CST_LOAD(&s->current_value) == e -> expected_value); fq = SEQ_CST_LOAD(&s->first_watch_queue_entry); q = alloc_stg_tvar_watch_queue(cap, (StgClosure*) tso); q -> next_queue_entry = fq; q -> prev_queue_entry = END_STM_WATCH_QUEUE; if (fq != END_STM_WATCH_QUEUE) { fq -> prev_queue_entry = q; } SEQ_CST_STORE(&s->first_watch_queue_entry, q); e -> new_value = (StgClosure *) q; dirty_TVAR(cap, s, (StgClosure *) fq); // we modified first_watch_queue_entry }); } static void remove_watch_queue_entries_for_trec(Capability *cap, StgTRecHeader *trec) { ASSERT(trec != NO_TREC); ASSERT(trec -> enclosing_trec == NO_TREC); ASSERT(trec -> state == TREC_WAITING || trec -> state == TREC_CONDEMNED); TRACE("%p : remove_watch_queue_entries_for_trec()", trec); FOR_EACH_ENTRY(trec, e, { StgTVar *s; StgTVarWatchQueue *pq; StgTVarWatchQueue *nq; StgTVarWatchQueue *q; StgClosure *saw; s = e -> tvar; saw = lock_tvar(cap, trec, s); q = (StgTVarWatchQueue *) (e -> new_value); TRACE("%p : removing tso=%p from watch queue for tvar=%p", trec, q -> closure, s); ACQ_ASSERT(SEQ_CST_LOAD(&s->current_value) == (StgClosure *)trec); nq = q -> next_queue_entry; pq = q -> prev_queue_entry; if (nq != END_STM_WATCH_QUEUE) { nq -> prev_queue_entry = pq; } if (pq != END_STM_WATCH_QUEUE) { pq -> next_queue_entry = nq; } else { ASSERT(SEQ_CST_LOAD(&s->first_watch_queue_entry) == q); SEQ_CST_STORE(&s->first_watch_queue_entry, nq); dirty_TVAR(cap, s, (StgClosure *) q); // we modified first_watch_queue_entry } free_stg_tvar_watch_queue(cap, q); unlock_tvar(cap, trec, s, saw, false); }); } /*......................................................................*/ static TRecEntry *get_new_entry(Capability *cap, StgTRecHeader *t) { TRecEntry *result; StgTRecChunk *c; int i; c = t -> current_chunk; i = c -> next_entry_idx; ASSERT(c != END_STM_CHUNK_LIST); if (i < TREC_CHUNK_NUM_ENTRIES) { // Continue to use current chunk result = &(c -> entries[i]); c -> next_entry_idx ++; } else { // Current chunk is full: allocate a fresh one StgTRecChunk *nc; nc = alloc_stg_trec_chunk(cap); nc -> prev_chunk = c; nc -> next_entry_idx = 1; t -> current_chunk = nc; result = &(nc -> entries[0]); } return result; } /*......................................................................*/ static void merge_update_into(Capability *cap, StgTRecHeader *t, StgTVar *tvar, StgClosure *expected_value, StgClosure *new_value) { // Look for an entry in this trec bool found = false; FOR_EACH_ENTRY(t, e, { StgTVar *s; s = e -> tvar; if (s == tvar) { found = true; if (e -> expected_value != expected_value) { // Must abort if the two entries start from different values TRACE("%p : update entries inconsistent at %p (%p vs %p)", t, tvar, e -> expected_value, expected_value); t -> state = TREC_CONDEMNED; } e -> new_value = new_value; BREAK_FOR_EACH; } }); if (!found) { // No entry so far in this trec TRecEntry *ne; ne = get_new_entry(cap, t); ne -> tvar = tvar; ne -> expected_value = expected_value; ne -> new_value = new_value; } } /*......................................................................*/ static void merge_read_into(Capability *cap, StgTRecHeader *trec, StgTVar *tvar, StgClosure *expected_value) { StgTRecHeader *t; bool found = false; // // See #7493 // // We need to look for an existing entry *anywhere* in the stack of // nested transactions. Otherwise, in stmCommitNestedTransaction() // we can't tell the difference between // // (1) a read-only entry // (2) an entry that writes back the original value // // Since in both cases e->new_value == e->expected_value. But in (1) // we want to do nothing, and in (2) we want to update e->new_value // in the outer transaction. // // Here we deal with the first possibility: we never create a // read-only entry in an inner transaction if there is an existing // outer entry; so we never have an inner read and an outer update. // So then in stmCommitNestedTransaction() we know we can always // write e->new_value over the outer entry, because the inner entry // is the most up to date. // for (t = trec; !found && t != NO_TREC; t = t -> enclosing_trec) { FOR_EACH_ENTRY(t, e, { if (e -> tvar == tvar) { found = true; if (e -> expected_value != expected_value) { // Must abort if the two entries start from different values TRACE("%p : read entries inconsistent at %p (%p vs %p)", t, tvar, e -> expected_value, expected_value); t -> state = TREC_CONDEMNED; } BREAK_FOR_EACH; } }); } if (!found) { // No entry found TRecEntry *ne; ne = get_new_entry(cap, trec); ne -> tvar = tvar; ne -> expected_value = expected_value; ne -> new_value = expected_value; } } /*......................................................................*/ static StgBool entry_is_update(TRecEntry *e) { StgBool result; result = (e -> expected_value != e -> new_value); return result; } #if defined(STM_FG_LOCKS) static StgBool entry_is_read_only(TRecEntry *e) { StgBool result; result = (e -> expected_value == e -> new_value); return result; } static StgBool tvar_is_locked(StgTVar *s, StgTRecHeader *h) { StgClosure *c; StgBool result; c = SEQ_CST_LOAD(&s->current_value); result = (c == (StgClosure *) h); return result; } #endif // revert_ownership : release a lock on a TVar, storing back // the value that it held when the lock was acquired. "revert_all" // is set in stmWait and stmReWait when we acquired locks on all of // the TVars involved. "revert_all" is not set in commit operations // where we don't lock TVars that have been read from but not updated. static void revert_ownership(Capability *cap STG_UNUSED, StgTRecHeader *trec STG_UNUSED, StgBool revert_all STG_UNUSED) { #if defined(STM_FG_LOCKS) FOR_EACH_ENTRY(trec, e, { if (revert_all || entry_is_update(e)) { StgTVar *s; s = e -> tvar; if (tvar_is_locked(s, trec)) { unlock_tvar(cap, trec, s, e -> expected_value, true); } } }); #endif } /*......................................................................*/ // validate_and_acquire_ownership : this performs the twin functions // of checking that the TVars referred to by entries in trec hold the // expected values and: // // - locking the TVar (on updated TVars during commit, or all TVars // during wait) // // - recording the identity of the TRec who wrote the value seen in the // TVar (on non-updated TVars during commit). These values are // stashed in the TRec entries and are then checked in check_read_only // to ensure that an atomic snapshot of all of these locations has been // seen. static StgBool validate_and_acquire_ownership (Capability *cap, StgTRecHeader *trec, int acquire_all, int retain_ownership) { StgBool result; if (shake()) { TRACE("%p : shake, pretending trec is invalid when it may not be", trec); return false; } ASSERT((trec -> state == TREC_ACTIVE) || (trec -> state == TREC_WAITING) || (trec -> state == TREC_CONDEMNED)); result = !((trec -> state) == TREC_CONDEMNED); if (result) { FOR_EACH_ENTRY(trec, e, { StgTVar *s; s = e -> tvar; if (acquire_all || entry_is_update(e)) { TRACE("%p : trying to acquire %p", trec, s); if (!cond_lock_tvar(cap, trec, s, e -> expected_value)) { TRACE("%p : failed to acquire %p", trec, s); result = false; BREAK_FOR_EACH; } } else { ASSERT(config_use_read_phase); IF_STM_FG_LOCKS({ TRACE("%p : will need to check %p", trec, s); // The memory ordering here must ensure that we have two distinct // reads to current_value, with the read from num_updates between // them. if (SEQ_CST_LOAD(&s->current_value) != e -> expected_value) { TRACE("%p : doesn't match", trec); result = false; BREAK_FOR_EACH; } e->num_updates = SEQ_CST_LOAD(&s->num_updates); if (SEQ_CST_LOAD(&s->current_value) != e -> expected_value) { TRACE("%p : doesn't match (race)", trec); result = false; BREAK_FOR_EACH; } else { TRACE("%p : need to check version %ld", trec, e -> num_updates); } }); } }); } if ((!result) || (!retain_ownership)) { revert_ownership(cap, trec, acquire_all); } return result; } // check_read_only : check that we've seen an atomic snapshot of the // non-updated TVars accessed by a trec. This checks that the last TRec to // commit an update to the TVar is unchanged since the value was stashed in // validate_and_acquire_ownership. If no update is seen to any TVar then // all of them contained their expected values at the start of the call to // check_read_only. // // The paper "Concurrent programming without locks" (under submission), or // Keir Fraser's PhD dissertation "Practical lock-free programming" discuss // this kind of algorithm. static StgBool check_read_only(StgTRecHeader *trec STG_UNUSED) { StgBool result = true; ASSERT(config_use_read_phase); IF_STM_FG_LOCKS({ FOR_EACH_ENTRY(trec, e, { StgTVar *s; s = e -> tvar; if (entry_is_read_only(e)) { TRACE("%p : check_read_only for TVar %p, saw %ld", trec, s, e -> num_updates); // We must first load current_value then num_updates; this is inverse of // the order of the stores in stmCommitTransaction. StgClosure *current_value = SEQ_CST_LOAD(&s->current_value); StgInt num_updates = SEQ_CST_LOAD(&s->num_updates); // Note we need both checks and in this order as the TVar could be // locked by another transaction that is committing but has not yet // incremented `num_updates` (See #7815). if (current_value != e->expected_value || num_updates != e->num_updates) { TRACE("%p : mismatch", trec); result = false; BREAK_FOR_EACH; } } }); }); return result; } /************************************************************************/ void stmPreGCHook (Capability *cap) { lock_stm(NO_TREC); TRACE("stmPreGCHook"); cap->free_tvar_watch_queues = END_STM_WATCH_QUEUE; cap->free_trec_chunks = END_STM_CHUNK_LIST; cap->free_trec_headers = NO_TREC; unlock_stm(NO_TREC); } /************************************************************************/ // check_read_only relies on version numbers held in TVars' "num_updates" // fields not wrapping around while a transaction is committed. The version // number is incremented each time an update is committed to the TVar // This is unlikely to wrap around when 32-bit integers are used for the counts, // but to ensure correctness we maintain a shared count on the maximum // number of commit operations that may occur and check that this has // not increased by more than 2^32 during a commit. #define TOKEN_BATCH_SIZE 1024 #if defined(THREADED_RTS) static volatile StgInt64 max_commits = 0; static volatile StgWord token_locked = false; static StgInt64 getMaxCommits(void) { return RELAXED_LOAD(&max_commits); } static void getTokenBatch(Capability *cap) { while (cas((void *)&token_locked, false, true) == true) { /* nothing */ } NONATOMIC_ADD(&max_commits, TOKEN_BATCH_SIZE); TRACE("%p : cap got token batch, max_commits=%" FMT_Int64, cap, RELAXED_LOAD(&max_commits)); cap -> transaction_tokens = TOKEN_BATCH_SIZE; RELEASE_STORE(&token_locked, false); } static void getToken(Capability *cap) { if (cap -> transaction_tokens == 0) { getTokenBatch(cap); } cap -> transaction_tokens --; } #else static StgInt64 getMaxCommits(void) { return 0; } static void getToken(Capability *cap STG_UNUSED) { // Nothing } #endif /*......................................................................*/ StgTRecHeader *stmStartTransaction(Capability *cap, StgTRecHeader *outer) { StgTRecHeader *t; TRACE("%p : stmStartTransaction with %d tokens", outer, cap -> transaction_tokens); getToken(cap); t = alloc_stg_trec_header(cap, outer); TRACE("%p : stmStartTransaction()=%p", outer, t); return t; } /*......................................................................*/ void stmAbortTransaction(Capability *cap, StgTRecHeader *trec) { StgTRecHeader *et; TRACE("%p : stmAbortTransaction", trec); ASSERT(trec != NO_TREC); ASSERT((trec -> state == TREC_ACTIVE) || (trec -> state == TREC_WAITING) || (trec -> state == TREC_CONDEMNED)); lock_stm(trec); et = trec -> enclosing_trec; if (et == NO_TREC) { // We're a top-level transaction: remove any watch queue entries that // we may have. TRACE("%p : aborting top-level transaction", trec); if (trec -> state == TREC_WAITING) { ASSERT(trec -> enclosing_trec == NO_TREC); TRACE("%p : stmAbortTransaction aborting waiting transaction", trec); remove_watch_queue_entries_for_trec(cap, trec); } } else { // We're a nested transaction: merge our read set into our parent's TRACE("%p : retaining read-set into parent %p", trec, et); FOR_EACH_ENTRY(trec, e, { StgTVar *s = e -> tvar; merge_read_into(cap, et, s, e -> expected_value); }); } trec -> state = TREC_ABORTED; unlock_stm(trec); TRACE("%p : stmAbortTransaction done", trec); } /*......................................................................*/ void stmFreeAbortedTRec(Capability *cap, StgTRecHeader *trec) { TRACE("%p : stmFreeAbortedTRec", trec); ASSERT(trec != NO_TREC); ASSERT((trec -> state == TREC_CONDEMNED) || (trec -> state == TREC_ABORTED)); free_stg_trec_header(cap, trec); TRACE("%p : stmFreeAbortedTRec done", trec); } /*......................................................................*/ void stmCondemnTransaction(Capability *cap, StgTRecHeader *trec) { TRACE("%p : stmCondemnTransaction", trec); ASSERT(trec != NO_TREC); ASSERT((trec -> state == TREC_ACTIVE) || (trec -> state == TREC_WAITING) || (trec -> state == TREC_CONDEMNED)); lock_stm(trec); if (trec -> state == TREC_WAITING) { ASSERT(trec -> enclosing_trec == NO_TREC); TRACE("%p : stmCondemnTransaction condemning waiting transaction", trec); remove_watch_queue_entries_for_trec(cap, trec); } trec -> state = TREC_CONDEMNED; unlock_stm(trec); TRACE("%p : stmCondemnTransaction done", trec); } /*......................................................................*/ StgBool stmValidateNestOfTransactions(Capability *cap, StgTRecHeader *trec) { StgTRecHeader *t; TRACE("%p : stmValidateNestOfTransactions", trec); ASSERT(trec != NO_TREC); ASSERT((trec -> state == TREC_ACTIVE) || (trec -> state == TREC_WAITING) || (trec -> state == TREC_CONDEMNED)); lock_stm(trec); t = trec; StgBool result = true; while (t != NO_TREC) { result &= validate_and_acquire_ownership(cap, t, true, false); t = t -> enclosing_trec; } if (!result && trec -> state != TREC_WAITING) { trec -> state = TREC_CONDEMNED; } unlock_stm(trec); TRACE("%p : stmValidateNestOfTransactions()=%d", trec, result); return result; } /*......................................................................*/ static TRecEntry *get_entry_for(StgTRecHeader *trec, StgTVar *tvar, StgTRecHeader **in) { TRecEntry *result = NULL; TRACE("%p : get_entry_for TVar %p", trec, tvar); ASSERT(trec != NO_TREC); do { FOR_EACH_ENTRY(trec, e, { if (e -> tvar == tvar) { result = e; if (in != NULL) { *in = trec; } BREAK_FOR_EACH; } }); trec = trec -> enclosing_trec; } while (result == NULL && trec != NO_TREC); return result; } /*......................................................................*/ StgBool stmCommitTransaction(Capability *cap, StgTRecHeader *trec) { StgInt64 max_commits_at_start = getMaxCommits(); TRACE("%p : stmCommitTransaction()", trec); ASSERT(trec != NO_TREC); lock_stm(trec); ASSERT(trec -> enclosing_trec == NO_TREC); ASSERT((trec -> state == TREC_ACTIVE) || (trec -> state == TREC_CONDEMNED)); // Use a read-phase (i.e. don't lock TVars we've read but not updated) if // the configuration lets us use a read phase. bool result = validate_and_acquire_ownership(cap, trec, (!config_use_read_phase), true); if (result) { // We now know that all the updated locations hold their expected values. ASSERT(trec -> state == TREC_ACTIVE); if (config_use_read_phase) { StgInt64 max_commits_at_end; StgInt64 max_concurrent_commits; TRACE("%p : doing read check", trec); result = check_read_only(trec); TRACE("%p : read-check %s", trec, result ? "succeeded" : "failed"); max_commits_at_end = getMaxCommits(); max_concurrent_commits = ((max_commits_at_end - max_commits_at_start) + (getNumCapabilities() * TOKEN_BATCH_SIZE)); if (((max_concurrent_commits >> 32) > 0) || shake()) { result = false; } } if (result) { // We now know that all of the read-only locations held their expected values // at the end of the call to validate_and_acquire_ownership. This forms the // linearization point of the commit. // Make the updates required by the transaction. FOR_EACH_ENTRY(trec, e, { StgTVar *s; s = e -> tvar; if ((!config_use_read_phase) || (e -> new_value != e -> expected_value)) { // Either the entry is an update or we're not using a read phase: // write the value back to the TVar, unlocking it if necessary. ACQ_ASSERT(tvar_is_locked(s, trec)); TRACE("%p : writing %p to %p, waking waiters", trec, e -> new_value, s); unpark_waiters_on(cap,s); IF_STM_FG_LOCKS({ // We have locked the TVar therefore nonatomic addition is sufficient NONATOMIC_ADD(&s->num_updates, 1); }); unlock_tvar(cap, trec, s, e -> new_value, true); } ACQ_ASSERT(!tvar_is_locked(s, trec)); }); } else { revert_ownership(cap, trec, false); } } unlock_stm(trec); free_stg_trec_header(cap, trec); TRACE("%p : stmCommitTransaction()=%d", trec, result); return result; } /*......................................................................*/ StgBool stmCommitNestedTransaction(Capability *cap, StgTRecHeader *trec) { StgTRecHeader *et; ASSERT(trec != NO_TREC && trec -> enclosing_trec != NO_TREC); TRACE("%p : stmCommitNestedTransaction() into %p", trec, trec -> enclosing_trec); ASSERT((trec -> state == TREC_ACTIVE) || (trec -> state == TREC_CONDEMNED)); lock_stm(trec); et = trec -> enclosing_trec; bool result = validate_and_acquire_ownership(cap, trec, (!config_use_read_phase), true); if (result) { // We now know that all the updated locations hold their expected values. if (config_use_read_phase) { TRACE("%p : doing read check", trec); result = check_read_only(trec); } if (result) { // We now know that all of the read-only locations held their expected values // at the end of the call to validate_and_acquire_ownership. This forms the // linearization point of the commit. TRACE("%p : read-check succeeded", trec); FOR_EACH_ENTRY(trec, e, { // Merge each entry into the enclosing transaction record, release all // locks. StgTVar *s; s = e -> tvar; if (entry_is_update(e)) { unlock_tvar(cap, trec, s, e -> expected_value, false); } merge_update_into(cap, et, s, e -> expected_value, e -> new_value); ACQ_ASSERT(s -> current_value != (StgClosure *)trec); }); } else { revert_ownership(cap, trec, false); } } unlock_stm(trec); free_stg_trec_header(cap, trec); TRACE("%p : stmCommitNestedTransaction()=%d", trec, result); return result; } /*......................................................................*/ StgBool stmWait(Capability *cap, StgTSO *tso, StgTRecHeader *trec) { TRACE("%p : stmWait(%p)", trec, tso); ASSERT(trec != NO_TREC); ASSERT(trec -> enclosing_trec == NO_TREC); ASSERT((trec -> state == TREC_ACTIVE) || (trec -> state == TREC_CONDEMNED)); lock_stm(trec); bool result = validate_and_acquire_ownership(cap, trec, true, true); if (result) { // The transaction is valid so far so we can actually start waiting. // (Otherwise the transaction was not valid and the thread will have to // retry it). // Put ourselves to sleep. We retain locks on all the TVars involved // until we are sound asleep : (a) on the wait queues, (b) BlockedOnSTM // in the TSO, (c) TREC_WAITING in the Trec. build_watch_queue_entries_for_trec(cap, tso, trec); park_tso(tso); trec -> state = TREC_WAITING; // We haven't released ownership of the transaction yet. The TSO // has been put on the wait queue for the TVars it is waiting for, // but we haven't yet tidied up the TSO's stack and made it safe // to wake up the TSO. Therefore, we must wait until the TSO is // safe to wake up before we release ownership - when all is well, // the runtime will call stmWaitUnlock() below, with the same // TRec. } else { unlock_stm(trec); free_stg_trec_header(cap, trec); } TRACE("%p : stmWait(%p)=%d", trec, tso, result); return result; } void stmWaitUnlock(Capability *cap, StgTRecHeader *trec) { revert_ownership(cap, trec, true); unlock_stm(trec); } /*......................................................................*/ StgBool stmReWait(Capability *cap, StgTSO *tso) { StgTRecHeader *trec = tso->trec; TRACE("%p : stmReWait", trec); ASSERT(trec != NO_TREC); ASSERT(trec -> enclosing_trec == NO_TREC); ASSERT((trec -> state == TREC_WAITING) || (trec -> state == TREC_CONDEMNED)); lock_stm(trec); bool result = validate_and_acquire_ownership(cap, trec, true, true); TRACE("%p : validation %s", trec, result ? "succeeded" : "failed"); if (result) { // The transaction remains valid -- do nothing because it is already on // the wait queues ASSERT(trec -> state == TREC_WAITING); park_tso(tso); revert_ownership(cap, trec, true); } else { // The transaction has become invalid. We can now remove it from the wait // queues. if (trec -> state != TREC_CONDEMNED) { remove_watch_queue_entries_for_trec (cap, trec); } free_stg_trec_header(cap, trec); } unlock_stm(trec); TRACE("%p : stmReWait()=%d", trec, result); return result; } /*......................................................................*/ static StgClosure *read_current_value(StgTRecHeader *trec STG_UNUSED, StgTVar *tvar) { StgClosure *result; result = ACQUIRE_LOAD(&tvar->current_value); #if defined(STM_FG_LOCKS) while (GET_INFO(UNTAG_CLOSURE(result)) == &stg_TREC_HEADER_info) { TRACE("%p : read_current_value(%p) saw %p", trec, tvar, result); result = ACQUIRE_LOAD(&tvar->current_value); } #endif TRACE("%p : read_current_value(%p)=%p", trec, tvar, result); return result; } /*......................................................................*/ StgClosure *stmReadTVar(Capability *cap, StgTRecHeader *trec, StgTVar *tvar) { StgTRecHeader *entry_in = NULL; StgClosure *result = NULL; TRecEntry *entry = NULL; TRACE("%p : stmReadTVar(%p)", trec, tvar); ASSERT(trec != NO_TREC); ASSERT(trec -> state == TREC_ACTIVE || trec -> state == TREC_CONDEMNED); entry = get_entry_for(trec, tvar, &entry_in); if (entry != NULL) { if (entry_in == trec) { // Entry found in our trec result = entry -> new_value; } else { // Entry found in another trec TRecEntry *new_entry = get_new_entry(cap, trec); new_entry -> tvar = tvar; new_entry -> expected_value = entry -> expected_value; new_entry -> new_value = entry -> new_value; result = new_entry -> new_value; } } else { // No entry found StgClosure *current_value = read_current_value(trec, tvar); TRecEntry *new_entry = get_new_entry(cap, trec); new_entry -> tvar = tvar; new_entry -> expected_value = current_value; new_entry -> new_value = current_value; result = current_value; } TRACE("%p : stmReadTVar(%p)=%p", trec, tvar, result); return result; } /*......................................................................*/ void stmWriteTVar(Capability *cap, StgTRecHeader *trec, StgTVar *tvar, StgClosure *new_value) { StgTRecHeader *entry_in = NULL; TRecEntry *entry = NULL; TRACE("%p : stmWriteTVar(%p, %p)", trec, tvar, new_value); ASSERT(trec != NO_TREC); ASSERT(trec -> state == TREC_ACTIVE || trec -> state == TREC_CONDEMNED); entry = get_entry_for(trec, tvar, &entry_in); if (entry != NULL) { if (entry_in == trec) { // Entry found in our trec IF_NONMOVING_WRITE_BARRIER_ENABLED { updateRemembSetPushClosure(cap, (StgClosure *) entry->new_value); } entry -> new_value = new_value; } else { // Entry found in another trec TRecEntry *new_entry = get_new_entry(cap, trec); new_entry -> tvar = tvar; new_entry -> expected_value = entry -> expected_value; new_entry -> new_value = new_value; } } else { // No entry found StgClosure *current_value = read_current_value(trec, tvar); TRecEntry *new_entry = get_new_entry(cap, trec); new_entry -> tvar = tvar; new_entry -> expected_value = current_value; new_entry -> new_value = new_value; } TRACE("%p : stmWriteTVar done", trec); } /*......................................................................*/