block/graph-lock: Make WITH_GRAPH_RDLOCK_GUARD() fully checked
Subsystems (2)
cover block/graph-lock: Make WITH_GRAPH_RDLOCK_GUARD() fully checked
Newer clang versions allow us to check scoped guards more thoroughly. Surprisingly, we only seem to have missed one instance with the old incomplete checks.
Kevin Wolf (2):
block-copy: Fix missing graph lock
block/graph-lock: Make WITH_GRAPH_RDLOCK_GUARD() fully checked
include/block/graph-lock.h | 21 ++++++++++++++------- block/block-copy.c | 4 +++- meson.build | 14 +++++++++++++- 3 files changed, 30 insertions(+), 9 deletions(-)
Signature · 1 line
2.45.2
PATCH 1/2 block-copy: Fix missing graph lock
The graph lock needs to be held when calling bdrv_co_pdiscard(). Fix block_copy_task_entry() to take it for the call.
WITH_GRAPH_RDLOCK_GUARD() was implemented in a weak way because of limitations in clang's Thread Safety Analysis at the time, so that it only asserts that the lock is held (which allows calling functions that require the lock), but we never deal with the unlocking (so even after the scope of the guard, the compiler assumes that the lock is still held). This is why the compiler didn't catch this locking error.
--- block/block-copy.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
block/block-copy.c +3 -1
@@ -595,7 +595,9 @@ static coroutine_fn int block_copy_task_entry(AioTask *task)if (s->discard_source && ret ==0 ) {int64_t nbytes =MIN (t->req.offset + t->req.bytes, s->len) - t->req.offset; -bdrv_co_pdiscard (s->source, t->req.offset, nbytes); + WITH_GRAPH_RDLOCK_GUARD() { +bdrv_co_pdiscard (s->source, t->req.offset, nbytes); + } }return ret;
Signature · 1 line
2.45.2
PATCH 2/2 block/graph-lock: Make WITH_GRAPH_RDLOCK_GUARD() fully checked
In reply to Kevin WolfUpstream clang 18 (and backports to clang 17 in Fedora and RHEL) implemented support for __attribute__((cleanup())) in its Thread Safety Analysis, so we can now actually have a proper implementation of WITH_GRAPH_RDLOCK_GUARD() that understands when we acquire and when we release the lock.
-Wthread-safety is now only enabled if the compiler is new enough to understand this pattern. In theory, we could have used some #ifdefs to keep the existing basic checks on old compilers, but as long as someone runs a newer compiler (and our CI does), we will catch locking problems, so it's probably not worth keeping multiple implementations for this.
The implementation can't use g_autoptr any more because the glib macros define wrapper functions that don't have the right TSA attributes, so the compiler would complain about them. Just use the cleanup attribute directly instead.
--- include/block/graph-lock.h | 21 ++++++++++++++------- meson.build | 14 +++++++++++++- 2 files changed, 27 insertions(+), 8 deletions(-)
include/block/graph-lock.h +14 -7
@@ -209,31 +209,38 @@ typedef struct GraphLockable { } GraphLockable;* unlocked. TSA_ASSERT_SHARED() makes sure that the following calls know that * we hold the lock while unlocking is left unchecked. */ -static inline GraphLockable *TSA_ASSERT_SHARED (graph_lock )TSA_NO_TSA coroutine_fn +static inline GraphLockable *TSA_ACQUIRE_SHARED (graph_lock ) coroutine_fngraph_lockable_auto_lock (GraphLockable *x) {bdrv_graph_co_rdlock ();return x; } -static inline void TSA_NO_TSA coroutine_fn -graph_lockable_auto_unlock (GraphLockable *x) +static inline void TSA_RELEASE_SHARED(graph_lock) coroutine_fn +graph_lockable_auto_unlock(GraphLockable **x) { bdrv_graph_co_rdunlock(); } -G_DEFINE_AUTOPTR_CLEANUP_FUNC(GraphLockable, graph_lockable_auto_unlock) +#define GRAPH_AUTO_UNLOCK __attribute__((cleanup(graph_lockable_auto_unlock))) +/* + * @var is only used to break the loop after the first iteration. + * @unlock_var can't be unlocked and then set to NULL because TSA wants the lock + * to be held at the start of every iteration of the loop. + */ #define WITH_GRAPH_RDLOCK_GUARD_(var) \ - for (g_autoptr(GraphLockable) var = graph_lockable_auto_lock(GML_OBJ_()); \ + for (GraphLockable *unlock_var GRAPH_AUTO_UNLOCK = \ + graph_lockable_auto_lock(GML_OBJ_()), \ + *var = unlock_var; \ var; \ - graph_lockable_auto_unlock(var), var = NULL) + var = NULL) #define WITH_GRAPH_RDLOCK_GUARD() \ WITH_GRAPH_RDLOCK_GUARD_(glue(graph_lockable_auto, __COUNTER__)) #define GRAPH_RDLOCK_GUARD(x) \ - g_autoptr(GraphLockable) \ + GraphLockable * GRAPH_AUTO_UNLOCK \ glue(graph_lockable_auto, __COUNTER__) G_GNUC_UNUSED = \ graph_lockable_auto_lock(GML_OBJ_())
meson.build +13 -1
@@ -624,7 +624,19 @@ warn_flags = [ ] if host_os != 'darwin' - warn_flags += ['-Wthread-safety'] + tsa_has_cleanup = cc.compiles(''' + struct __attribute__((capability("mutex"))) mutex {}; + void lock(struct mutex *m) __attribute__((acquire_capability(m))); + void unlock(struct mutex *m) __attribute__((release_capability(m))); + + void test(void) { + struct mutex __attribute__((cleanup(unlock))) m; + lock(&m); + } + ''', args: ['-Wthread-safety', '-Werror']) + if tsa_has_cleanup + warn_flags += ['-Wthread-safety'] + endif endif # Set up C++ compiler flags
Signature · 1 line
2.45.2
Quoted 100 lines · Kevin Wolf
Upstream clang 18 (and backports to clang 17 in Fedora and RHEL) implemented support for __attribute__((cleanup())) in its Thread Safety Analysis, so we can now actually have a proper implementation of
WITH_GRAPH_RDLOCK_GUARD()that understands when we acquire and when we release the lock.-Wthread-safety is now only enabled if the compiler is new enough to understand this pattern. In theory, we could have used some #ifdefs to keep the existing basic checks on old compilers, but as long as someone runs a newer compiler (and our CI does), we will catch locking problems, so it's probably not worth keeping multiple implementations for this.
The implementation can't use
g_autoptrany more because the glib macros define wrapper functions that don't have the right TSA attributes, so the compiler would complain about them. Just use the cleanup attribute directly instead.Signed-off-by Kevin Wolf <[email protected]>--- include/block/graph-lock.h | 21 ++++++++++++++------- meson.build | 14 +++++++++++++- 2 files changed, 27 insertions(+), 8 deletions(-)include/block/graph-lock.h +14
-7@@ -209,31 +209,38 @@ typedef struct GraphLockable { } GraphLockable;* unlocked. TSA_ASSERT_SHARED() makes sure that the following calls know that * we hold the lock while unlocking is left unchecked. */ -static inline GraphLockable *TSA_ASSERT_SHARED (graph_lock )TSA_NO_TSA coroutine_fn +static inline GraphLockable *TSA_ACQUIRE_SHARED (graph_lock ) coroutine_fngraph_lockable_auto_lock (GraphLockable *x) {bdrv_graph_co_rdlock ();return x; } -static inline void TSA_NO_TSA coroutine_fn -graph_lockable_auto_unlock (GraphLockable *x) +static inline void TSA_RELEASE_SHARED(graph_lock) coroutine_fn +graph_lockable_auto_unlock(GraphLockable **x) { bdrv_graph_co_rdunlock(); } -G_DEFINE_AUTOPTR_CLEANUP_FUNC(GraphLockable, graph_lockable_auto_unlock) +#define GRAPH_AUTO_UNLOCK __attribute__((cleanup(graph_lockable_auto_unlock))) +/* + * @var is only used to break the loop after the first iteration. + * @unlock_var can't be unlocked and then set to NULL because TSA wants the lock + * to be held at the start of every iteration of the loop. + */ #define WITH_GRAPH_RDLOCK_GUARD_(var) \ - for (g_autoptr(GraphLockable) var = graph_lockable_auto_lock(GML_OBJ_()); \ + for (GraphLockable *unlock_var GRAPH_AUTO_UNLOCK = \ + graph_lockable_auto_lock(GML_OBJ_()), \ + *var = unlock_var; \ var; \ - graph_lockable_auto_unlock(var), var = NULL) + var = NULL) #define WITH_GRAPH_RDLOCK_GUARD() \ WITH_GRAPH_RDLOCK_GUARD_(glue(graph_lockable_auto, __COUNTER__)) #define GRAPH_RDLOCK_GUARD(x) \ - g_autoptr(GraphLockable) \ + GraphLockable * GRAPH_AUTO_UNLOCK \ glue(graph_lockable_auto, __COUNTER__) G_GNUC_UNUSED = \ graph_lockable_auto_lock(GML_OBJ_())meson.build +13
-2@@ -624,7 +624,19 @@ warn_flags = [ ] if host_os != 'darwin' - warn_flags += ['-Wthread-safety'] + tsa_has_cleanup = cc.compiles(''' + struct __attribute__((capability("mutex"))) mutex {}; + void lock(struct mutex *m) __attribute__((acquire_capability(m))); + void unlock(struct mutex *m) __attribute__((release_capability(m))); + + void test(void) { + struct mutex __attribute__((cleanup(unlock))) m; + lock(&m); + } + ''', args: ['-Wthread-safety', '-Werror']) + if tsa_has_cleanup + warn_flags += ['-Wthread-safety'] + endif endif # Set up C++ compiler flags --
Quoted 2 lines
2.45.2
Signature · 2 lines
Manos Pitsidianakis Emulation and Virtualization Engineer at Linaro Ltd
Quoted 4 lines · Kevin Wolf
Newer clang versions allow us to check scoped guards more thoroughly. Surprisingly, we only seem to have missed one instance with the old incomplete checks.
Quoted 12 lines
Kevin Wolf (2):
block-copy: Fix missing graph lock
block/graph-lock: MakeWITH_GRAPH_RDLOCK_GUARD()fully checkedinclude/block/graph-lock.h | 21 ++++++++++++++------- block/block-copy.c | 4 +++- meson.build | 14 +++++++++++++- 3 files changed, 30 insertions(+), 9 deletions(-)Signature · 1 line
2.45.2