Switch from _Noreturn to __attribute__((noreturn))

Parts of the Linux kernel build system struggle with _Noreturn.  This
results in the following warnings when building on RHEL 8.5, and likely
other environments.  Switch to using the __attribute__((noreturn)).

  warning: objtool: dbuf_free_range()+0x2b8:
    return with modified stack frame
  warning: objtool: dbuf_free_range()+0x0:
    stack state mismatch: cfa1=7+40 cfa2=7+8
  ...
  WARNING: EXPORT symbol "arc_buf_size" [zfs.ko] version generation
    failed, symbol will not be versioned.
  WARNING: EXPORT symbol "spa_open" [zfs.ko] version generation
    failed, symbol will not be versioned.
  ...

Additionally, __thread_exit() has been renamed spl_thread_exit() and
made a static inline function.  This was needed because the kernel
will generate a warning for symbols which are __attribute__((noreturn))
and then exported with EXPORT_SYMBOL.

While we could continue to use _Noreturn in user space I've also
switched it to __attribute__((noreturn)) purely for consistency
throughout the code base.

Reviewed-by: Ryan Moeller <freqlabs@FreeBSD.org>
Reviewed-by: Brian Atkinson <batkinson@lanl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #13238

module/zfs/arc.c

@@ -928,7 +928,7 @@ static unsigned long l2arc_rebuild_blocks_min_l2size = 1024 * 1024 * 1024;
 
 /* L2ARC persistence rebuild control routines. */
 void l2arc_rebuild_vdev(vdev_t *vd, boolean_t reopen);
-static _Noreturn void l2arc_dev_rebuild_thread(void *arg);
+static __attribute__((noreturn)) void l2arc_dev_rebuild_thread(void *arg);
 static int l2arc_rebuild(l2arc_dev_t *dev);
 
 /* L2ARC persistence read I/O routines. */
@@ -9707,7 +9707,7 @@ l2arc_hdr_limit_reached(void)
  * This thread feeds the L2ARC at regular intervals.  This is the beating
  * heart of the L2ARC.
  */
-static _Noreturn void
+static  __attribute__((noreturn)) void
 l2arc_feed_thread(void *unused)
 {
 	(void) unused;
@@ -10146,7 +10146,7 @@ l2arc_spa_rebuild_start(spa_t *spa)
 /*
  * Main entry point for L2ARC rebuilding.
  */
-static _Noreturn void
+static __attribute__((noreturn)) void
 l2arc_dev_rebuild_thread(void *arg)
 {
 	l2arc_dev_t *dev = arg;

module/zfs/dbuf.c

@@ -778,7 +778,7 @@ dbuf_evict_one(void)
  * of the dbuf cache is at or below the maximum size. Once the dbuf is aged
  * out of the cache it is destroyed and becomes eligible for arc eviction.
  */
-static _Noreturn void
+static __attribute__((noreturn)) void
 dbuf_evict_thread(void *unused)
 {
 	(void) unused;

module/zfs/dmu_recv.c

@@ -2801,7 +2801,7 @@ receive_process_record(struct receive_writer_arg *rwa,
  * dmu_recv_stream's worker thread; pull records off the queue, and then call
  * receive_process_record  When we're done, signal the main thread and exit.
  */
-static _Noreturn void
+static __attribute__((noreturn)) void
 receive_writer_thread(void *arg)
 {
 	struct receive_writer_arg *rwa = arg;

module/zfs/dmu_redact.c

@@ -351,7 +351,7 @@ redact_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
 	return (0);
 }
 
-static _Noreturn void
+static __attribute__((noreturn)) void
 redact_traverse_thread(void *arg)
 {
 	struct redact_thread_arg *rt_arg = arg;
@@ -837,7 +837,7 @@ struct redact_merge_thread_arg {
 	int error_code;
 };
 
-static _Noreturn void
+static __attribute__((noreturn)) void
 redact_merge_thread(void *arg)
 {
 	struct redact_merge_thread_arg *rmta = arg;

module/zfs/dmu_send.c

@@ -1234,7 +1234,7 @@ redact_list_cb(redact_block_phys_t *rb, void *arg)
  * error code of the thread in case something goes wrong, and pushes the End of
  * Stream record when the traverse_dataset call has finished.
  */
-static _Noreturn void
+static __attribute__((noreturn)) void
 send_traverse_thread(void *arg)
 {
 	struct send_thread_arg *st_arg = arg;
@@ -1324,7 +1324,7 @@ get_next_range(bqueue_t *bq, struct send_range *prev)
 	return (next);
 }
 
-static _Noreturn void
+static __attribute__((noreturn)) void
 redact_list_thread(void *arg)
 {
 	struct redact_list_thread_arg *rlt_arg = arg;
@@ -1519,7 +1519,7 @@ find_next_range(struct send_range **ranges, bqueue_t **qs, uint64_t *out_mask)
  * data from the redact_list_thread and use that to determine which blocks
  * should be redacted.
  */
-static _Noreturn void
+static __attribute__((noreturn)) void
 send_merge_thread(void *arg)
 {
 	struct send_merge_thread_arg *smt_arg = arg;
@@ -1744,7 +1744,7 @@ enqueue_range(struct send_reader_thread_arg *srta, bqueue_t *q, dnode_t *dn,
  * some indirect blocks can be discarded because they're not holes. Second,
  * it issues prefetches for the data we need to send.
  */
-static _Noreturn void
+static __attribute__((noreturn)) void
 send_reader_thread(void *arg)
 {
 	struct send_reader_thread_arg *srta = arg;

module/zfs/mmp.c

@@ -187,7 +187,7 @@ uint_t zfs_multihost_import_intervals = MMP_DEFAULT_IMPORT_INTERVALS;
 uint_t zfs_multihost_fail_intervals = MMP_DEFAULT_FAIL_INTERVALS;
 
 static void *const mmp_tag = "mmp_write_uberblock";
-static _Noreturn void mmp_thread(void *arg);
+static __attribute__((noreturn)) void mmp_thread(void *arg);
 
 void
 mmp_init(spa_t *spa)
@@ -217,14 +217,13 @@ mmp_thread_enter(mmp_thread_t *mmp, callb_cpr_t *cpr)
 	mutex_enter(&mmp->mmp_thread_lock);
 }
 
-static _Noreturn void
+static void
 mmp_thread_exit(mmp_thread_t *mmp, kthread_t **mpp, callb_cpr_t *cpr)
 {
 	ASSERT(*mpp != NULL);
 	*mpp = NULL;
 	cv_broadcast(&mmp->mmp_thread_cv);
 	CALLB_CPR_EXIT(cpr);		/* drops &mmp->mmp_thread_lock */
-	thread_exit();
 }
 
 void
@@ -537,7 +536,7 @@ mmp_write_uberblock(spa_t *spa)
 	zio_nowait(zio);
 }
 
-static _Noreturn void
+static __attribute__((noreturn)) void
 mmp_thread(void *arg)
 {
 	spa_t *spa = (spa_t *)arg;
@@ -698,6 +697,8 @@ mmp_thread(void *arg)
 
 	mmp->mmp_zio_root = NULL;
 	mmp_thread_exit(mmp, &mmp->mmp_thread, &cpr);
+
+	thread_exit();
 }
 
 /*

module/zfs/spa.c

@@ -8139,7 +8139,7 @@ spa_async_autoexpand(spa_t *spa, vdev_t *vd)
 	spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_AUTOEXPAND);
 }
 
-static _Noreturn void
+static __attribute__((noreturn)) void
 spa_async_thread(void *arg)
 {
 	spa_t *spa = (spa_t *)arg;

module/zfs/txg.c

@@ -108,8 +108,8 @@
  * now transition to the syncing state.
  */
 
-static _Noreturn void txg_sync_thread(void *arg);
-static _Noreturn void txg_quiesce_thread(void *arg);
+static __attribute__((noreturn)) void txg_sync_thread(void *arg);
+static __attribute__((noreturn)) void txg_quiesce_thread(void *arg);
 
 int zfs_txg_timeout = 5;	/* max seconds worth of delta per txg */
 
@@ -514,7 +514,7 @@ txg_has_quiesced_to_sync(dsl_pool_t *dp)
 	return (tx->tx_quiesced_txg != 0);
 }
 
-static _Noreturn void
+static __attribute__((noreturn)) void
 txg_sync_thread(void *arg)
 {
 	dsl_pool_t *dp = arg;
@@ -605,7 +605,7 @@ txg_sync_thread(void *arg)
 	}
 }
 
-static _Noreturn void
+static __attribute__((noreturn)) void
 txg_quiesce_thread(void *arg)
 {
 	dsl_pool_t *dp = arg;

module/zfs/vdev_initialize.c

@@ -488,7 +488,7 @@ vdev_initialize_range_add(void *arg, uint64_t start, uint64_t size)
 	vdev_xlate_walk(vd, &logical_rs, vdev_initialize_xlate_range_add, arg);
 }
 
-static _Noreturn void
+static __attribute__((noreturn)) void
 vdev_initialize_thread(void *arg)
 {
 	vdev_t *vd = arg;

module/zfs/vdev_rebuild.c

@@ -133,7 +133,7 @@ static int zfs_rebuild_scrub_enabled = 1;
 /*
  * For vdev_rebuild_initiate_sync() and vdev_rebuild_reset_sync().
  */
-static _Noreturn void vdev_rebuild_thread(void *arg);
+static __attribute__((noreturn)) void vdev_rebuild_thread(void *arg);
 
 /*
  * Clear the per-vdev rebuild bytes value for a vdev tree.
@@ -736,7 +736,7 @@ vdev_rebuild_load(vdev_t *vd)
  * Each scan thread is responsible for rebuilding a top-level vdev.  The
  * rebuild progress in tracked on-disk in VDEV_TOP_ZAP_VDEV_REBUILD_PHYS.
  */
-static _Noreturn void
+static __attribute__((noreturn)) void
 vdev_rebuild_thread(void *arg)
 {
 	vdev_t *vd = arg;

module/zfs/vdev_removal.c

@@ -140,7 +140,7 @@ int zfs_removal_suspend_progress = 0;
 
 #define	VDEV_REMOVAL_ZAP_OBJS	"lzap"
 
-static _Noreturn void spa_vdev_remove_thread(void *arg);
+static __attribute__((noreturn)) void spa_vdev_remove_thread(void *arg);
 static int spa_vdev_remove_cancel_impl(spa_t *spa);
 
 static void
@@ -1589,7 +1589,7 @@ spa_remove_max_segment(spa_t *spa)
  * TXG have completed (see spa_txg_zio) and writes the new mappings to disk
  * (see vdev_mapping_sync()).
  */
-static _Noreturn void
+static __attribute__((noreturn)) void
 spa_vdev_remove_thread(void *arg)
 {
 	spa_t *spa = arg;

module/zfs/vdev_trim.c

@@ -834,7 +834,7 @@ vdev_trim_range_add(void *arg, uint64_t start, uint64_t size)
  * by its ms_allocatable.  While a metaslab is undergoing trimming it is
  * not eligible for new allocations.
  */
-static _Noreturn void
+static __attribute__((noreturn)) void
 vdev_trim_thread(void *arg)
 {
 	vdev_t *vd = arg;
@@ -1175,7 +1175,7 @@ vdev_trim_range_verify(void *arg, uint64_t start, uint64_t size)
  * N.B. This behavior is different from a manual TRIM where a thread
  * is created for each leaf vdev, instead of each top-level vdev.
  */
-static _Noreturn void
+static __attribute__((noreturn)) void
 vdev_autotrim_thread(void *arg)
 {
 	vdev_t *vd = arg;
@@ -1514,7 +1514,7 @@ vdev_autotrim_restart(spa_t *spa)
 		vdev_autotrim(spa);
 }
 
-static _Noreturn void
+static __attribute__((noreturn)) void
 vdev_trim_l2arc_thread(void *arg)
 {
 	vdev_t		*vd = arg;

module/zfs/zthr.c

@@ -231,7 +231,7 @@ struct zthr {
 	const char	*zthr_name;
 };
 
-static _Noreturn void
+static __attribute__((noreturn)) void
 zthr_procedure(void *arg)
 {
 	zthr_t *t = arg;