asgardius/virtualx-engine
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge pull request #10484 from Sipaha/pr_mt_queue_deadlock_fix

Browse source 
Fix deadlock in CommandQueueMT
This commit is contained in:
Rémi Verschelde 2017-08-21 22:55:07 +02:00 committed by GitHub
commit de261ef380
2 changed files with 47 additions and 18 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
core/command_queue_mt.cpp
View file

@ -55,6 +55,7 @@ CommandQueueMT::SyncSemaphore *CommandQueueMT::_alloc_sync_sem() {
while (true) { while (true) {
lock();
for (int i = 0; i < SYNC_SEMAPHORES; i++) { for (int i = 0; i < SYNC_SEMAPHORES; i++) {
if (!sync_sems[i].in_use) { if (!sync_sems[i].in_use) {
@ -63,6 +64,7 @@ CommandQueueMT::SyncSemaphore *CommandQueueMT::_alloc_sync_sem() {
break; break;
} }
} }
unlock();
if (idx == -1) { if (idx == -1) {
wait_for_flush(); wait_for_flush();

63
core/command_queue_mt.h
View file

@ -911,12 +911,14 @@ public:
template <class T, class M, class R> template <class T, class M, class R>
void push_and_ret(T *p_instance, M p_method, R *r_ret) { void push_and_ret(T *p_instance, M p_method, R *r_ret) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandRet0<T, M, R> *cmd = allocate_and_lock<CommandRet0<T, M, R> >(); CommandRet0<T, M, R> *cmd = allocate_and_lock<CommandRet0<T, M, R> >();
cmd->instance = p_instance; cmd->instance = p_instance;
cmd->method = p_method; cmd->method = p_method;
cmd->ret = r_ret; cmd->ret = r_ret;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -928,13 +930,15 @@ public:
template <class T, class M, class P1, class R> template <class T, class M, class P1, class R>
void push_and_ret(T *p_instance, M p_method, P1 p1, R *r_ret) { void push_and_ret(T *p_instance, M p_method, P1 p1, R *r_ret) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandRet1<T, M, P1, R> *cmd = allocate_and_lock<CommandRet1<T, M, P1, R> >(); CommandRet1<T, M, P1, R> *cmd = allocate_and_lock<CommandRet1<T, M, P1, R> >();
cmd->instance = p_instance; cmd->instance = p_instance;
cmd->method = p_method; cmd->method = p_method;
cmd->p1 = p1; cmd->p1 = p1;
cmd->ret = r_ret; cmd->ret = r_ret;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -946,6 +950,8 @@ public:
template <class T, class M, class P1, class P2, class R> template <class T, class M, class P1, class P2, class R>
void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, R *r_ret) { void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, R *r_ret) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandRet2<T, M, P1, P2, R> *cmd = allocate_and_lock<CommandRet2<T, M, P1, P2, R> >(); CommandRet2<T, M, P1, P2, R> *cmd = allocate_and_lock<CommandRet2<T, M, P1, P2, R> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -953,7 +959,7 @@ public:
cmd->p1 = p1; cmd->p1 = p1;
cmd->p2 = p2; cmd->p2 = p2;
cmd->ret = r_ret; cmd->ret = r_ret;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -965,6 +971,8 @@ public:
template <class T, class M, class P1, class P2, class P3, class R> template <class T, class M, class P1, class P2, class P3, class R>
void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, R *r_ret) { void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, R *r_ret) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandRet3<T, M, P1, P2, P3, R> *cmd = allocate_and_lock<CommandRet3<T, M, P1, P2, P3, R> >(); CommandRet3<T, M, P1, P2, P3, R> *cmd = allocate_and_lock<CommandRet3<T, M, P1, P2, P3, R> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -973,7 +981,7 @@ public:
cmd->p2 = p2; cmd->p2 = p2;
cmd->p3 = p3; cmd->p3 = p3;
cmd->ret = r_ret; cmd->ret = r_ret;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -985,6 +993,8 @@ public:
template <class T, class M, class P1, class P2, class P3, class P4, class R> template <class T, class M, class P1, class P2, class P3, class P4, class R>
void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, R *r_ret) { void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, R *r_ret) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandRet4<T, M, P1, P2, P3, P4, R> *cmd = allocate_and_lock<CommandRet4<T, M, P1, P2, P3, P4, R> >(); CommandRet4<T, M, P1, P2, P3, P4, R> *cmd = allocate_and_lock<CommandRet4<T, M, P1, P2, P3, P4, R> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -994,7 +1004,7 @@ public:
cmd->p3 = p3; cmd->p3 = p3;
cmd->p4 = p4; cmd->p4 = p4;
cmd->ret = r_ret; cmd->ret = r_ret;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1006,6 +1016,8 @@ public:
template <class T, class M, class P1, class P2, class P3, class P4, class P5, class R> template <class T, class M, class P1, class P2, class P3, class P4, class P5, class R>
void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, R *r_ret) { void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, R *r_ret) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandRet5<T, M, P1, P2, P3, P4, P5, R> *cmd = allocate_and_lock<CommandRet5<T, M, P1, P2, P3, P4, P5, R> >(); CommandRet5<T, M, P1, P2, P3, P4, P5, R> *cmd = allocate_and_lock<CommandRet5<T, M, P1, P2, P3, P4, P5, R> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -1016,7 +1028,7 @@ public:
cmd->p4 = p4; cmd->p4 = p4;
cmd->p5 = p5; cmd->p5 = p5;
cmd->ret = r_ret; cmd->ret = r_ret;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1028,6 +1040,8 @@ public:
template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6, class R> template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6, class R>
void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6, R *r_ret) { void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6, R *r_ret) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandRet6<T, M, P1, P2, P3, P4, P5, P6, R> *cmd = allocate_and_lock<CommandRet6<T, M, P1, P2, P3, P4, P5, P6, R> >(); CommandRet6<T, M, P1, P2, P3, P4, P5, P6, R> *cmd = allocate_and_lock<CommandRet6<T, M, P1, P2, P3, P4, P5, P6, R> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -1039,7 +1053,7 @@ public:
cmd->p5 = p5; cmd->p5 = p5;
cmd->p6 = p6; cmd->p6 = p6;
cmd->ret = r_ret; cmd->ret = r_ret;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1051,6 +1065,8 @@ public:
template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6, class P7, class R> template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6, class P7, class R>
void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6, P7 p7, R *r_ret) { void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6, P7 p7, R *r_ret) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandRet7<T, M, P1, P2, P3, P4, P5, P6, P7, R> *cmd = allocate_and_lock<CommandRet7<T, M, P1, P2, P3, P4, P5, P6, P7, R> >(); CommandRet7<T, M, P1, P2, P3, P4, P5, P6, P7, R> *cmd = allocate_and_lock<CommandRet7<T, M, P1, P2, P3, P4, P5, P6, P7, R> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -1063,7 +1079,7 @@ public:
cmd->p6 = p6; cmd->p6 = p6;
cmd->p7 = p7; cmd->p7 = p7;
cmd->ret = r_ret; cmd->ret = r_ret;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1075,6 +1091,8 @@ public:
template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6, class P7, class P8, class R> template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6, class P7, class P8, class R>
void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6, P7 p7, P8 p8, R *r_ret) { void push_and_ret(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6, P7 p7, P8 p8, R *r_ret) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandRet8<T, M, P1, P2, P3, P4, P5, P6, P7, P8, R> *cmd = allocate_and_lock<CommandRet8<T, M, P1, P2, P3, P4, P5, P6, P7, P8, R> >(); CommandRet8<T, M, P1, P2, P3, P4, P5, P6, P7, P8, R> *cmd = allocate_and_lock<CommandRet8<T, M, P1, P2, P3, P4, P5, P6, P7, P8, R> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -1088,7 +1106,7 @@ public:
cmd->p7 = p7; cmd->p7 = p7;
cmd->p8 = p8; cmd->p8 = p8;
cmd->ret = r_ret; cmd->ret = r_ret;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1100,12 +1118,13 @@ public:
template <class T, class M> template <class T, class M>
void push_and_sync(T *p_instance, M p_method) { void push_and_sync(T *p_instance, M p_method) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandSync0<T, M> *cmd = allocate_and_lock<CommandSync0<T, M> >(); CommandSync0<T, M> *cmd = allocate_and_lock<CommandSync0<T, M> >();
cmd->instance = p_instance; cmd->instance = p_instance;
cmd->method = p_method; cmd->method = p_method;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1117,13 +1136,14 @@ public:
template <class T, class M, class P1> template <class T, class M, class P1>
void push_and_sync(T *p_instance, M p_method, P1 p1) { void push_and_sync(T *p_instance, M p_method, P1 p1) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandSync1<T, M, P1> *cmd = allocate_and_lock<CommandSync1<T, M, P1> >(); CommandSync1<T, M, P1> *cmd = allocate_and_lock<CommandSync1<T, M, P1> >();
cmd->instance = p_instance; cmd->instance = p_instance;
cmd->method = p_method; cmd->method = p_method;
cmd->p1 = p1; cmd->p1 = p1;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1135,6 +1155,8 @@ public:
template <class T, class M, class P1, class P2> template <class T, class M, class P1, class P2>
void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2) { void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandSync2<T, M, P1, P2> *cmd = allocate_and_lock<CommandSync2<T, M, P1, P2> >(); CommandSync2<T, M, P1, P2> *cmd = allocate_and_lock<CommandSync2<T, M, P1, P2> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -1142,7 +1164,6 @@ public:
cmd->p1 = p1; cmd->p1 = p1;
cmd->p2 = p2; cmd->p2 = p2;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1154,6 +1175,8 @@ public:
template <class T, class M, class P1, class P2, class P3> template <class T, class M, class P1, class P2, class P3>
void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3) { void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandSync3<T, M, P1, P2, P3> *cmd = allocate_and_lock<CommandSync3<T, M, P1, P2, P3> >(); CommandSync3<T, M, P1, P2, P3> *cmd = allocate_and_lock<CommandSync3<T, M, P1, P2, P3> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -1162,7 +1185,6 @@ public:
cmd->p2 = p2; cmd->p2 = p2;
cmd->p3 = p3; cmd->p3 = p3;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1174,6 +1196,8 @@ public:
template <class T, class M, class P1, class P2, class P3, class P4> template <class T, class M, class P1, class P2, class P3, class P4>
void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4) { void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandSync4<T, M, P1, P2, P3, P4> *cmd = allocate_and_lock<CommandSync4<T, M, P1, P2, P3, P4> >(); CommandSync4<T, M, P1, P2, P3, P4> *cmd = allocate_and_lock<CommandSync4<T, M, P1, P2, P3, P4> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -1183,7 +1207,6 @@ public:
cmd->p3 = p3; cmd->p3 = p3;
cmd->p4 = p4; cmd->p4 = p4;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1195,6 +1218,8 @@ public:
template <class T, class M, class P1, class P2, class P3, class P4, class P5> template <class T, class M, class P1, class P2, class P3, class P4, class P5>
void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5) { void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandSync5<T, M, P1, P2, P3, P4, P5> *cmd = allocate_and_lock<CommandSync5<T, M, P1, P2, P3, P4, P5> >(); CommandSync5<T, M, P1, P2, P3, P4, P5> *cmd = allocate_and_lock<CommandSync5<T, M, P1, P2, P3, P4, P5> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -1205,7 +1230,6 @@ public:
cmd->p4 = p4; cmd->p4 = p4;
cmd->p5 = p5; cmd->p5 = p5;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1217,6 +1241,8 @@ public:
template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6> template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6>
void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6) { void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandSync6<T, M, P1, P2, P3, P4, P5, P6> *cmd = allocate_and_lock<CommandSync6<T, M, P1, P2, P3, P4, P5, P6> >(); CommandSync6<T, M, P1, P2, P3, P4, P5, P6> *cmd = allocate_and_lock<CommandSync6<T, M, P1, P2, P3, P4, P5, P6> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -1228,7 +1254,6 @@ public:
cmd->p5 = p5; cmd->p5 = p5;
cmd->p6 = p6; cmd->p6 = p6;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1240,6 +1265,8 @@ public:
template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6, class P7> template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6, class P7>
void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6, P7 p7) { void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6, P7 p7) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandSync7<T, M, P1, P2, P3, P4, P5, P6, P7> *cmd = allocate_and_lock<CommandSync7<T, M, P1, P2, P3, P4, P5, P6, P7> >(); CommandSync7<T, M, P1, P2, P3, P4, P5, P6, P7> *cmd = allocate_and_lock<CommandSync7<T, M, P1, P2, P3, P4, P5, P6, P7> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -1252,7 +1279,6 @@ public:
cmd->p6 = p6; cmd->p6 = p6;
cmd->p7 = p7; cmd->p7 = p7;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();
@ -1264,6 +1290,8 @@ public:
template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6, class P7, class P8> template <class T, class M, class P1, class P2, class P3, class P4, class P5, class P6, class P7, class P8>
void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6, P7 p7, P8 p8) { void push_and_sync(T *p_instance, M p_method, P1 p1, P2 p2, P3 p3, P4 p4, P5 p5, P6 p6, P7 p7, P8 p8) {
SyncSemaphore *ss = _alloc_sync_sem();
CommandSync8<T, M, P1, P2, P3, P4, P5, P6, P7, P8> *cmd = allocate_and_lock<CommandSync8<T, M, P1, P2, P3, P4, P5, P6, P7, P8> >(); CommandSync8<T, M, P1, P2, P3, P4, P5, P6, P7, P8> *cmd = allocate_and_lock<CommandSync8<T, M, P1, P2, P3, P4, P5, P6, P7, P8> >();
cmd->instance = p_instance; cmd->instance = p_instance;
@ -1277,7 +1305,6 @@ public:
cmd->p7 = p7; cmd->p7 = p7;
cmd->p8 = p8; cmd->p8 = p8;
SyncSemaphore *ss = _alloc_sync_sem();
cmd->sync = ss; cmd->sync = ss;
unlock(); unlock();