Implement thread cancellation for aarch64

libc/thread/posixthread.internal.h

@@ -97,8 +97,8 @@ void _pthread_free(struct PosixThread *);
 void _pthread_onfork_prepare(void);
 void _pthread_onfork_parent(void);
 void _pthread_onfork_child(void);
+long _pthread_cancel_sys(void);
 void _pthread_ungarbage(void);
-int _pthread_cancel_sys(void);
 
 COSMOPOLITAN_C_END_
 #endif /* !(__ASSEMBLER__ + __LINKER__ + 0) */

libc/thread/pthread_cancel.c

@@ -21,11 +21,13 @@
 #include "libc/calls/struct/sigaction.h"
 #include "libc/calls/struct/siginfo.h"
 #include "libc/calls/struct/sigset.h"
+#include "libc/calls/struct/ucontext.internal.h"
 #include "libc/calls/syscall_support-sysv.internal.h"
 #include "libc/calls/ucontext.h"
 #include "libc/dce.h"
 #include "libc/errno.h"
 #include "libc/intrin/atomic.h"
+#include "libc/intrin/kprintf.h"
 #include "libc/runtime/runtime.h"
 #include "libc/str/str.h"
 #include "libc/sysv/consts/sa.h"
@@ -34,14 +36,13 @@
 #include "libc/thread/posixthread.internal.h"
 #include "libc/thread/thread.h"
 #include "libc/thread/tls.h"
-#ifdef __x86_64__
 
 int systemfive_cancel(void);
 
 extern const char systemfive_cancellable[];
 extern const char systemfive_cancellable_end[];
 
-int _pthread_cancel_sys(void) {
+long _pthread_cancel_sys(void) {
   struct PosixThread *pt;
   pt = (struct PosixThread *)__get_tls()->tib_pthread;
   if (!(pt->flags & (PT_NOCANCEL | PT_MASKED)) || (pt->flags & PT_ASYNC)) {
@@ -60,9 +61,9 @@ static void OnSigThr(int sig, siginfo_t *si, void *ctx) {
       !(pt->flags & PT_NOCANCEL) &&
       atomic_load_explicit(&pt->cancelled, memory_order_acquire)) {
     sigaddset(&uc->uc_sigmask, sig);
-    if (systemfive_cancellable <= (char *)uc->uc_mcontext.rip &&
-        (char *)uc->uc_mcontext.rip < systemfive_cancellable_end) {
-      uc->uc_mcontext.rip = (intptr_t)systemfive_cancel;
+    if (systemfive_cancellable <= (char *)uc->uc_mcontext.PC &&
+        (char *)uc->uc_mcontext.PC < systemfive_cancellable_end) {
+      uc->uc_mcontext.PC = (intptr_t)systemfive_cancel;
     } else if (pt->flags & PT_ASYNC) {
       pthread_exit(PTHREAD_CANCELED);
     } else {
@@ -302,8 +303,6 @@ errno_t pthread_cancel(pthread_t thread) {
   return rc;
 }
 
-#endif /* __x86_64__ */
-
 /**
  * Creates cancellation point in calling thread.
  *

libc/thread/pthread_create.c

@@ -17,6 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/assert.h"
+#include "libc/atomic.h"
 #include "libc/calls/blocksigs.internal.h"
 #include "libc/calls/calls.h"
 #include "libc/calls/struct/sigaltstack.h"
@@ -28,6 +29,7 @@
 #include "libc/intrin/bits.h"
 #include "libc/intrin/bsr.h"
 #include "libc/intrin/dll.h"
+#include "libc/intrin/popcnt.h"
 #include "libc/log/internal.h"
 #include "libc/macros.internal.h"
 #include "libc/mem/mem.h"
@@ -58,6 +60,7 @@ static unsigned long roundup2pow(unsigned long x) {
 }
 
 void _pthread_free(struct PosixThread *pt) {
+  static atomic_uint freed;
   if (pt->flags & PT_STATIC) return;
   free(pt->tls);
   if ((pt->flags & PT_OWNSTACK) &&  //
@@ -69,6 +72,9 @@ void _pthread_free(struct PosixThread *pt) {
     free(pt->altstack);
   }
   free(pt);
+  if (popcnt(atomic_fetch_add_explicit(&freed, 1, memory_order_acq_rel)) == 1) {
+    malloc_trim(0);
+  }
 }
 
 void pthread_kill_siblings_np(void) {

libc/thread/pthread_decimate_np.c

@@ -16,19 +16,18 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
+#include "libc/atomic.h"
 #include "libc/intrin/atomic.h"
 #include "libc/intrin/dll.h"
 #include "libc/runtime/runtime.h"
 #include "libc/thread/posixthread.internal.h"
 #include "libc/thread/thread.h"
 #include "libc/thread/tls.h"
-#include "third_party/dlmalloc/dlmalloc.h"
 
 /**
  * Releases memory of detached threads that have terminated.
  */
 void pthread_decimate_np(void) {
-  bool empty;
   struct Dll *e;
   struct PosixThread *pt;
   enum PosixThreadStatus status;
@@ -46,9 +45,5 @@ StartOver:
       goto StartOver;
     }
   }
-  empty = dll_is_empty(_pthread_list);
   pthread_spin_unlock(&_pthread_lock);
-  if (empty) {
-    dlmalloc_trim(0);
-  }
 }

libc/thread/thread.mk

@@ -34,7 +34,6 @@ LIBC_THREAD_A_DIRECTDEPS =				\
 	LIBC_STR					\
 	LIBC_SYSV					\
 	LIBC_SYSV_CALLS					\
-	THIRD_PARTY_DLMALLOC				\
 	THIRD_PARTY_NSYNC				\
 	THIRD_PARTY_NSYNC_MEM
 

libc/thread/tls.h

@@ -44,9 +44,15 @@ extern unsigned __tls_index;
 #ifdef __x86_64__
 extern bool __tls_enabled;
 #define __tls_enabled_set(x) __tls_enabled = x
-#else
-#define __tls_enabled        true
+#elif defined(__aarch64__)
+#define __tls_enabled                        \
+  ({                                         \
+    register struct CosmoTib *_t asm("x28"); \
+    !!_t;                                    \
+  })
 #define __tls_enabled_set(x) (void)0
+#else
+#error "unsupported architecture"
 #endif
 
 void __require_tls(void);