aboutsummaryrefslogtreecommitdiff
path: root/core
diff options
context:
space:
mode:
authorBenjamin Herrenschmidt <benh@kernel.crashing.org>2017-12-20 13:16:23 +1100
committerStewart Smith <stewart@linux.vnet.ibm.com>2017-12-20 22:15:36 -0600
commit76d9bcdca58936d761458f8f05960239c4dd8dec (patch)
treec8377b11be33e62d312810645b8044d3a6f427aa /core
parentca612b802adac0c72cd0f10c51a51275e5914101 (diff)
downloadskiboot-76d9bcdca58936d761458f8f05960239c4dd8dec.zip
skiboot-76d9bcdca58936d761458f8f05960239c4dd8dec.tar.gz
skiboot-76d9bcdca58936d761458f8f05960239c4dd8dec.tar.bz2
lock: Add additional lock auditing code
Keep track of lock owner name and replace lock_depth counter with a per-cpu list of locks held by the cpu. This allows us to print the actual locks held in case we hit the (in)famous message about opal_pollers being run with a lock held. It also allows us to warn (and drop them) if locks are still held when returning to the OS or completing a scheduled job. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Reviewed-by: Nicholas Piggin <npiggin@gmail.com> [stewart: fix unit tests] Signed-off-by: Stewart Smith <stewart@linux.vnet.ibm.com>
Diffstat (limited to 'core')
-rw-r--r--core/cpu.c6
-rw-r--r--core/lock.c41
-rw-r--r--core/opal.c8
-rw-r--r--core/stack.c1
-rw-r--r--core/test/run-malloc-speed.c3
-rw-r--r--core/test/run-malloc.c3
-rw-r--r--core/test/run-mem_range_is_reserved.c3
-rw-r--r--core/test/run-mem_region.c3
-rw-r--r--core/test/run-mem_region_init.c3
-rw-r--r--core/test/run-mem_region_next.c3
-rw-r--r--core/test/run-mem_region_release_unused.c3
-rw-r--r--core/test/run-mem_region_release_unused_noalloc.c3
-rw-r--r--core/test/run-mem_region_reservations.c3
-rw-r--r--core/test/run-msg.c3
-rw-r--r--core/test/run-timer.c6
-rw-r--r--core/test/run-trace.c3
-rw-r--r--core/timebase.c2
17 files changed, 74 insertions, 23 deletions
diff --git a/core/cpu.c b/core/cpu.c
index 7dd7c86..3e110c3 100644
--- a/core/cpu.c
+++ b/core/cpu.c
@@ -285,6 +285,11 @@ void cpu_process_jobs(void)
if (no_return)
free(job);
func(data);
+ if (!list_empty(&cpu->locks_held)) {
+ prlog(PR_ERR, "OPAL job %s returning with locks held\n",
+ job->name);
+ drop_my_locks(true);
+ }
lock(&cpu->job_lock);
if (!no_return) {
cpu->job_count--;
@@ -822,6 +827,7 @@ static void init_cpu_thread(struct cpu_thread *t,
init_lock(&t->dctl_lock);
init_lock(&t->job_lock);
list_head_init(&t->job_queue);
+ list_head_init(&t->locks_held);
t->stack_guard = STACK_CHECK_GUARD_BASE ^ pir;
t->state = state;
t->pir = pir;
diff --git a/core/lock.c b/core/lock.c
index b5e3323..edfe1c7 100644
--- a/core/lock.c
+++ b/core/lock.c
@@ -57,8 +57,8 @@ static void unlock_check(struct lock *l)
if (l->in_con_path && this_cpu()->con_suspend == 0)
lock_error(l, "Unlock con lock with console not suspended", 3);
- if (this_cpu()->lock_depth == 0)
- lock_error(l, "Releasing lock with 0 depth", 4);
+ if (list_empty(&this_cpu()->locks_held))
+ lock_error(l, "Releasing lock we don't hold depth", 4);
}
#else
@@ -89,7 +89,7 @@ static inline bool __try_lock(struct cpu_thread *cpu, struct lock *l)
return false;
}
-bool try_lock(struct lock *l)
+bool try_lock_caller(struct lock *l, const char *owner)
{
struct cpu_thread *cpu = this_cpu();
@@ -97,22 +97,23 @@ bool try_lock(struct lock *l)
return true;
if (__try_lock(cpu, l)) {
+ l->owner = owner;
if (l->in_con_path)
cpu->con_suspend++;
- cpu->lock_depth++;
+ list_add(&cpu->locks_held, &l->list);
return true;
}
return false;
}
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *owner)
{
if (bust_locks)
return;
lock_check(l);
for (;;) {
- if (try_lock(l))
+ if (try_lock_caller(l, owner))
break;
smt_lowest();
while (l->lock_val)
@@ -130,8 +131,9 @@ void unlock(struct lock *l)
unlock_check(l);
+ l->owner = NULL;
+ list_del(&l->list);
lwsync();
- this_cpu()->lock_depth--;
l->lock_val = 0;
/* WARNING: On fast reboot, we can be reset right at that
@@ -144,7 +146,7 @@ void unlock(struct lock *l)
}
}
-bool lock_recursive(struct lock *l)
+bool lock_recursive_caller(struct lock *l, const char *caller)
{
if (bust_locks)
return false;
@@ -152,7 +154,7 @@ bool lock_recursive(struct lock *l)
if (lock_held_by_me(l))
return false;
- lock(l);
+ lock_caller(l, caller);
return true;
}
@@ -160,3 +162,24 @@ void init_locks(void)
{
bust_locks = false;
}
+
+void dump_locks_list(void)
+{
+ struct lock *l;
+
+ prlog(PR_ERR, "Locks held:\n");
+ list_for_each(&this_cpu()->locks_held, l, list)
+ prlog(PR_ERR, " %s\n", l->owner);
+}
+
+void drop_my_locks(bool warn)
+{
+ struct lock *l;
+
+ while((l = list_pop(&this_cpu()->locks_held, struct lock, list)) != NULL) {
+ if (warn)
+ prlog(PR_ERR, " %s\n", l->owner);
+ unlock(l);
+ }
+}
+
diff --git a/core/opal.c b/core/opal.c
index d33d527..1bca774 100644
--- a/core/opal.c
+++ b/core/opal.c
@@ -182,6 +182,11 @@ int64_t opal_exit_check(int64_t retval, struct stack_frame *eframe)
printf("CPU UN-ACCOUNTED FIRMWARE ENTRY! PIR=%04lx cpu @%p -> pir=%04x token=%llu retval=%lld\n",
mfspr(SPR_PIR), cpu, cpu->pir, token, retval);
} else {
+ if (!list_empty(&cpu->locks_held)) {
+ prlog(PR_ERR, "OPAL exiting with locks held, token=%llu retval=%lld\n",
+ token, retval);
+ drop_my_locks(true);
+ }
sync(); /* release barrier vs quiescing */
cpu->in_opal_call--;
}
@@ -557,7 +562,7 @@ void opal_run_pollers(void)
}
this_cpu()->in_poller = true;
- if (this_cpu()->lock_depth && pollers_with_lock_warnings < 64) {
+ if (!list_empty(&this_cpu()->locks_held) && pollers_with_lock_warnings < 64) {
/**
* @fwts-label OPALPollerWithLock
* @fwts-advice opal_run_pollers() was called with a lock
@@ -565,6 +570,7 @@ void opal_run_pollers(void)
* lucky/careful.
*/
prlog(PR_ERR, "Running pollers with lock held !\n");
+ dump_locks_list();
backtrace();
pollers_with_lock_warnings++;
if (pollers_with_lock_warnings == 64) {
diff --git a/core/stack.c b/core/stack.c
index bb7fa62..af4d37d 100644
--- a/core/stack.c
+++ b/core/stack.c
@@ -21,6 +21,7 @@
#include <stack.h>
#include <mem_region.h>
#include <unistd.h>
+#include <lock.h>
#define STACK_BUF_ENTRIES 60
static struct bt_entry bt_buf[STACK_BUF_ENTRIES];
diff --git a/core/test/run-malloc-speed.c b/core/test/run-malloc-speed.c
index 279216e..d842bd6 100644
--- a/core/test/run-malloc-speed.c
+++ b/core/test/run-malloc-speed.c
@@ -55,8 +55,9 @@ static inline void real_free(void *p)
char __rodata_start[1], __rodata_end[1];
struct dt_node *dt_root;
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *caller)
{
+ (void)caller;
assert(!l->lock_val);
l->lock_val = 1;
}
diff --git a/core/test/run-malloc.c b/core/test/run-malloc.c
index d79e6f9..2feaacb 100644
--- a/core/test/run-malloc.c
+++ b/core/test/run-malloc.c
@@ -57,8 +57,9 @@ static inline void real_free(void *p)
struct dt_node *dt_root;
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *caller)
{
+ (void)caller;
assert(!l->lock_val);
l->lock_val = 1;
}
diff --git a/core/test/run-mem_range_is_reserved.c b/core/test/run-mem_range_is_reserved.c
index 95c790d..37f7db3 100644
--- a/core/test/run-mem_range_is_reserved.c
+++ b/core/test/run-mem_range_is_reserved.c
@@ -57,8 +57,9 @@ static void real_free(void *p)
#include <assert.h>
#include <stdio.h>
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *caller)
{
+ (void)caller;
assert(!l->lock_val);
l->lock_val++;
}
diff --git a/core/test/run-mem_region.c b/core/test/run-mem_region.c
index 6b7f6fb..f2506d6 100644
--- a/core/test/run-mem_region.c
+++ b/core/test/run-mem_region.c
@@ -55,8 +55,9 @@ static inline void real_free(void *p)
struct dt_node *dt_root;
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *caller)
{
+ (void)caller;
assert(!l->lock_val);
l->lock_val++;
}
diff --git a/core/test/run-mem_region_init.c b/core/test/run-mem_region_init.c
index ee7e189..d4265af 100644
--- a/core/test/run-mem_region_init.c
+++ b/core/test/run-mem_region_init.c
@@ -63,8 +63,9 @@ static inline char *skiboot_strdup(const char *str)
#include <assert.h>
#include <stdio.h>
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *caller)
{
+ (void)caller;
assert(!l->lock_val);
l->lock_val = 1;
}
diff --git a/core/test/run-mem_region_next.c b/core/test/run-mem_region_next.c
index 7daa269..72d02a9 100644
--- a/core/test/run-mem_region_next.c
+++ b/core/test/run-mem_region_next.c
@@ -52,8 +52,9 @@ static void real_free(void *p)
#include <assert.h>
#include <stdio.h>
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *caller)
{
+ (void)caller;
assert(!l->lock_val);
l->lock_val++;
}
diff --git a/core/test/run-mem_region_release_unused.c b/core/test/run-mem_region_release_unused.c
index 712f98a..4941453 100644
--- a/core/test/run-mem_region_release_unused.c
+++ b/core/test/run-mem_region_release_unused.c
@@ -60,8 +60,9 @@ static inline void __free(void *p, const char *location __attribute__((unused)))
#include <assert.h>
#include <stdio.h>
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *caller)
{
+ (void)caller;
l->lock_val++;
}
diff --git a/core/test/run-mem_region_release_unused_noalloc.c b/core/test/run-mem_region_release_unused_noalloc.c
index a79485b..1ad4abb 100644
--- a/core/test/run-mem_region_release_unused_noalloc.c
+++ b/core/test/run-mem_region_release_unused_noalloc.c
@@ -60,8 +60,9 @@ static inline void __free(void *p, const char *location __attribute__((unused)))
#include <assert.h>
#include <stdio.h>
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *caller)
{
+ (void)caller;
l->lock_val++;
}
diff --git a/core/test/run-mem_region_reservations.c b/core/test/run-mem_region_reservations.c
index 584b7c3..f593d9a 100644
--- a/core/test/run-mem_region_reservations.c
+++ b/core/test/run-mem_region_reservations.c
@@ -57,8 +57,9 @@ static void real_free(void *p)
#include <assert.h>
#include <stdio.h>
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *caller)
{
+ (void)caller;
assert(!l->lock_val);
l->lock_val++;
}
diff --git a/core/test/run-msg.c b/core/test/run-msg.c
index 2cee515..67418a9 100644
--- a/core/test/run-msg.c
+++ b/core/test/run-msg.c
@@ -41,8 +41,9 @@ static void *zalloc(size_t size)
#include "../opal-msg.c"
#include <skiboot.h>
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *caller)
{
+ (void)caller;
assert(!l->lock_val);
l->lock_val = 1;
}
diff --git a/core/test/run-timer.c b/core/test/run-timer.c
index 0270cfe..e45cf63 100644
--- a/core/test/run-timer.c
+++ b/core/test/run-timer.c
@@ -12,7 +12,11 @@
static uint64_t stamp, last;
struct lock;
-static inline void lock(struct lock *l) { (void)l; }
+static inline void lock_caller(struct lock *l, const char *caller)
+{
+ (void)caller;
+ (void)l;
+}
static inline void unlock(struct lock *l) { (void)l; }
unsigned long tb_hz = 512000000;
diff --git a/core/test/run-trace.c b/core/test/run-trace.c
index c319c05..dd4cd45 100644
--- a/core/test/run-trace.c
+++ b/core/test/run-trace.c
@@ -113,8 +113,9 @@ struct debug_descriptor debug_descriptor = {
.trace_mask = -1
};
-void lock(struct lock *l)
+void lock_caller(struct lock *l, const char *caller)
{
+ (void)caller;
assert(!l->lock_val);
l->lock_val = 1;
}
diff --git a/core/timebase.c b/core/timebase.c
index f2dff44..777e4ba 100644
--- a/core/timebase.c
+++ b/core/timebase.c
@@ -53,7 +53,7 @@ void time_wait(unsigned long duration)
{
struct cpu_thread *c = this_cpu();
- if (this_cpu()->lock_depth) {
+ if (!list_empty(&this_cpu()->locks_held)) {
time_wait_nopoll(duration);
return;
}