Date: Mon, 24 Jul 2017 22:26:03 +0000 From: bugzilla-noreply@freebsd.org To: freebsd-bugs@FreeBSD.org Subject: [Bug 219399] System panics after several hours of 14-threads-compilation orgies using poudriere on AMD Ryzen... Message-ID: <bug-219399-8-s0Dh4tw28A@https.bugs.freebsd.org/bugzilla/> In-Reply-To: <bug-219399-8@https.bugs.freebsd.org/bugzilla/> References: <bug-219399-8@https.bugs.freebsd.org/bugzilla/>
next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=219399 --- Comment #108 from Don Lewis <truckman@FreeBSD.org> --- (In reply to Nils Beyer from comment #91) I'm pretty sure that ryzen_segv_test is actually broken. The first iteration of the loop in the t2 threadx() is unlocked and there is no guarantee that it will have initialized things before thread1() tries to use them. Try this patch: --- ryzen_segv_test.c.orig 2017-07-24 14:26:23.851846000 -0700 +++ ryzen_segv_test.c 2017-07-24 15:02:33.998102000 -0700 @@ -291,29 +291,32 @@ atomic_store(&flg, 0); } +void threadx_core() +{ + uint8_t offset; + uint32_t randval; + + offset = random() % 256; + randval = random(); + memset(func_set, 0, sizeof(func_set_t)); + memcpy(&func_set->func[offset], func_base, FUNC_BYTES); + func_set->offset = offset; + func_set->ret = randval; +} + void threadx(void *p) { uint8_t offset; uint32_t randval; int init = 0; - if(p != NULL) { - init = 1; - } //usleep(1000); while(atomic_load(&flg)) { offset = random() % 256; randval = random(); - if(!init) { - lock_enter(); - } else { - if(func_set == MAP_FAILED) { - fprintf(stderr, "mmap returns MAP_FAILED!\n"); - return; - } - init = 0; - } + lock_enter(); + // threadx_core(); memset(func_set, 0, sizeof(func_set_t)); memcpy(&func_set->func[offset], func_base, FUNC_BYTES); func_set->offset = offset; @@ -330,8 +333,7 @@ { int64_t loops; pthread_t t1, t2, t3; -#ifdef _MSC_VER -#else +#if !defined(_MSC_VER) && !defined(__FreeBSD__) cpu_set_t cpuset; int cpu; #endif @@ -349,19 +351,23 @@ n_cpus = sysconf(_SC_NPROCESSORS_ONLN); func_set = mmap (NULL, sizeof(func_set_t), PROT_READ | PROT_WRITE | PROT_EXEC, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); #endif + if(func_set == MAP_FAILED) { + fprintf(stderr, "mmap returns MAP_FAILED!\n"); + exit (1); + } atomic_store(&flg, 1); atomic_store(&locked, 1); srandom(time(NULL) + pid); // You should confirm assembly of generated code, just in case the compiler reorders mfence instruction + threadx_core(); mfence(); // Assure that flags are stored properly pthread_create(&t1, NULL, (void*)thread1, &loops); - pthread_create(&t2, NULL, (void*)threadx, (void*)1); + pthread_create(&t2, NULL, (void*)threadx, NULL); pthread_create(&t3, NULL, (void*)threadx, NULL); -#ifdef _MSC_VER -#else +#if !defined(_MSC_VER) && !defined(__FreeBSD__) cpu = random() % n_cpus; CPU_ZERO(&cpuset); CPU_SET(cpu, &cpuset); -- You are receiving this mail because: You are the assignee for the bug.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-219399-8-s0Dh4tw28A>