bcache: Debug code improvements
authorKent Overstreet <kmo@daterainc.com>
Thu, 24 Oct 2013 23:36:03 +0000 (16:36 -0700)
committerKent Overstreet <kmo@daterainc.com>
Mon, 11 Nov 2013 05:56:34 +0000 (21:56 -0800)
Couple changes:
 * Consolidate bch_check_keys() and bch_check_key_order(), and move the
   checks that only check_key_order() could do to bch_btree_iter_next().

 * Get rid of CONFIG_BCACHE_EDEBUG - now, all that code is compiled in
   when CONFIG_BCACHE_DEBUG is enabled, and there's now a sysfs file to
   flip on the EDEBUG checks at runtime.

 * Dropped an old not terribly useful check in rw_unlock(), and
   refactored/improved a some of the other debug code.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
drivers/md/bcache/Kconfig
drivers/md/bcache/alloc.c
drivers/md/bcache/bcache.h
drivers/md/bcache/bset.c
drivers/md/bcache/bset.h
drivers/md/bcache/btree.c
drivers/md/bcache/btree.h
drivers/md/bcache/debug.c
drivers/md/bcache/debug.h
drivers/md/bcache/sysfs.c
drivers/md/bcache/util.h

index f950c9d29f3e5400a97b7eb43400d6e2a79b9e19..2638417b19aa74d1207de209b93bbb1dfc2689e8 100644 (file)
@@ -13,15 +13,8 @@ config BCACHE_DEBUG
        ---help---
        Don't select this option unless you're a developer
 
-       Enables extra debugging tools (primarily a fuzz tester)
-
-config BCACHE_EDEBUG
-       bool "Extended runtime checks"
-       depends on BCACHE
-       ---help---
-       Don't select this option unless you're a developer
-
-       Enables extra runtime checks which significantly affect performance
+       Enables extra debugging tools, allows expensive runtime checks to be
+       turned on.
 
 config BCACHE_CLOSURES_DEBUG
        bool "Debug closures"
index 4970ddc6a7f69bcc0087a22134ed09eb51c18bf2..ed5920b20c614f783b03b944525a79b9d90e702e 100644 (file)
@@ -398,8 +398,7 @@ long bch_bucket_alloc(struct cache *ca, unsigned watermark, bool wait)
 out:
        wake_up_process(ca->alloc_thread);
 
-#ifdef CONFIG_BCACHE_EDEBUG
-       {
+       if (expensive_debug_checks(ca->set)) {
                size_t iter;
                long i;
 
@@ -413,7 +412,7 @@ out:
                fifo_for_each(i, &ca->unused, iter)
                        BUG_ON(i == r);
        }
-#endif
+
        b = ca->buckets + r;
 
        BUG_ON(atomic_read(&b->pin) != 1);
index 045cb99f1ca69dd2ddc3ffe5edfda0a2df54297e..d03bc6f66493791e90f6f1fc80828cf9757650ff 100644 (file)
@@ -690,6 +690,7 @@ struct cache_set {
        unsigned short          journal_delay_ms;
        unsigned                verify:1;
        unsigned                key_merging_disabled:1;
+       unsigned                expensive_debug_checks:1;
        unsigned                gc_always_rewrite:1;
        unsigned                shrinker_disabled:1;
        unsigned                copy_gc_enabled:1;
@@ -698,15 +699,6 @@ struct cache_set {
        struct hlist_head       bucket_hash[1 << BUCKET_HASH_BITS];
 };
 
-static inline bool key_merging_disabled(struct cache_set *c)
-{
-#ifdef CONFIG_BCACHE_DEBUG
-       return c->key_merging_disabled;
-#else
-       return 0;
-#endif
-}
-
 struct bbio {
        unsigned                submit_time_us;
        union {
index f32216c759480a670e675dcf877d27627f8d39c4..6bffde478926abad17a1d8dce6d1eb151f107f6c 100644 (file)
@@ -106,6 +106,43 @@ bad:
        return true;
 }
 
+static bool ptr_bad_expensive_checks(struct btree *b, const struct bkey *k,
+                                    unsigned ptr)
+{
+       struct bucket *g = PTR_BUCKET(b->c, k, ptr);
+       char buf[80];
+
+       if (mutex_trylock(&b->c->bucket_lock)) {
+               if (b->level) {
+                       if (KEY_DIRTY(k) ||
+                           g->prio != BTREE_PRIO ||
+                           (b->c->gc_mark_valid &&
+                            GC_MARK(g) != GC_MARK_METADATA))
+                               goto err;
+
+               } else {
+                       if (g->prio == BTREE_PRIO)
+                               goto err;
+
+                       if (KEY_DIRTY(k) &&
+                           b->c->gc_mark_valid &&
+                           GC_MARK(g) != GC_MARK_DIRTY)
+                               goto err;
+               }
+               mutex_unlock(&b->c->bucket_lock);
+       }
+
+       return false;
+err:
+       mutex_unlock(&b->c->bucket_lock);
+       bch_bkey_to_text(buf, sizeof(buf), k);
+       btree_bug(b,
+"inconsistent pointer %s: bucket %zu pin %i prio %i gen %i last_gc %i mark %llu gc_gen %i",
+                 buf, PTR_BUCKET_NR(b->c, k, ptr), atomic_read(&g->pin),
+                 g->prio, g->gen, g->last_gc, GC_MARK(g), g->gc_gen);
+       return true;
+}
+
 bool bch_ptr_bad(struct btree *b, const struct bkey *k)
 {
        struct bucket *g;
@@ -133,46 +170,12 @@ bool bch_ptr_bad(struct btree *b, const struct bkey *k)
                if (stale)
                        return true;
 
-#ifdef CONFIG_BCACHE_EDEBUG
-               if (!mutex_trylock(&b->c->bucket_lock))
-                       continue;
-
-               if (b->level) {
-                       if (KEY_DIRTY(k) ||
-                           g->prio != BTREE_PRIO ||
-                           (b->c->gc_mark_valid &&
-                            GC_MARK(g) != GC_MARK_METADATA))
-                               goto bug;
-
-               } else {
-                       if (g->prio == BTREE_PRIO)
-                               goto bug;
-
-                       if (KEY_DIRTY(k) &&
-                           b->c->gc_mark_valid &&
-                           GC_MARK(g) != GC_MARK_DIRTY)
-                               goto bug;
-               }
-               mutex_unlock(&b->c->bucket_lock);
-#endif
+               if (expensive_debug_checks(b->c) &&
+                   ptr_bad_expensive_checks(b, k, i))
+                       return true;
        }
 
        return false;
-#ifdef CONFIG_BCACHE_EDEBUG
-bug:
-       mutex_unlock(&b->c->bucket_lock);
-
-       {
-               char buf[80];
-
-               bch_bkey_to_text(buf, sizeof(buf), k);
-               btree_bug(b,
-"inconsistent pointer %s: bucket %zu pin %i prio %i gen %i last_gc %i mark %llu gc_gen %i",
-                         buf, PTR_BUCKET_NR(b->c, k, i), atomic_read(&g->pin),
-                         g->prio, g->gen, g->last_gc, GC_MARK(g), g->gc_gen);
-       }
-       return true;
-#endif
 }
 
 /* Key/pointer manipulation */
@@ -821,16 +824,16 @@ struct bkey *__bch_bset_search(struct btree *b, struct bset_tree *t,
        } else
                i = bset_search_write_set(b, t, search);
 
-#ifdef CONFIG_BCACHE_EDEBUG
-       BUG_ON(bset_written(b, t) &&
-              i.l != t->data->start &&
-              bkey_cmp(tree_to_prev_bkey(t,
-                 inorder_to_tree(bkey_to_cacheline(t, i.l), t)),
-                       search) > 0);
+       if (expensive_debug_checks(b->c)) {
+               BUG_ON(bset_written(b, t) &&
+                      i.l != t->data->start &&
+                      bkey_cmp(tree_to_prev_bkey(t,
+                         inorder_to_tree(bkey_to_cacheline(t, i.l), t)),
+                               search) > 0);
 
-       BUG_ON(i.r != end(t->data) &&
-              bkey_cmp(i.r, search) <= 0);
-#endif
+               BUG_ON(i.r != end(t->data) &&
+                      bkey_cmp(i.r, search) <= 0);
+       }
 
        while (likely(i.l != i.r) &&
               bkey_cmp(i.l, search) <= 0)
@@ -871,12 +874,16 @@ void bch_btree_iter_push(struct btree_iter *iter, struct bkey *k,
 }
 
 struct bkey *__bch_btree_iter_init(struct btree *b, struct btree_iter *iter,
-                              struct bkey *search, struct bset_tree *start)
+                                  struct bkey *search, struct bset_tree *start)
 {
        struct bkey *ret = NULL;
        iter->size = ARRAY_SIZE(iter->data);
        iter->used = 0;
 
+#ifdef CONFIG_BCACHE_DEBUG
+       iter->b = b;
+#endif
+
        for (; start <= &b->sets[b->nsets]; start++) {
                ret = bch_bset_search(b, start, search);
                bch_btree_iter_push(iter, ret, end(start->data));
@@ -891,6 +898,8 @@ struct bkey *bch_btree_iter_next(struct btree_iter *iter)
        struct bkey *ret = NULL;
 
        if (!btree_iter_end(iter)) {
+               bch_btree_iter_next_check(iter);
+
                ret = iter->data->k;
                iter->data->k = bkey_next(iter->data->k);
 
@@ -1002,7 +1011,6 @@ static void btree_mergesort(struct btree *b, struct bset *out,
        out->keys = last ? (uint64_t *) bkey_next(last) - out->d : 0;
 
        pr_debug("sorted %i keys", out->keys);
-       bch_check_key_order(b, out);
 }
 
 static void __btree_sort(struct btree *b, struct btree_iter *iter,
@@ -1063,15 +1071,15 @@ static void __btree_sort(struct btree *b, struct btree_iter *iter,
 
 void bch_btree_sort_partial(struct btree *b, unsigned start)
 {
-       size_t oldsize = 0, order = b->page_order, keys = 0;
+       size_t order = b->page_order, keys = 0;
        struct btree_iter iter;
+       int oldsize = bch_count_data(b);
+
        __bch_btree_iter_init(b, &iter, NULL, &b->sets[start]);
 
        BUG_ON(b->sets[b->nsets].data == write_block(b) &&
               (b->sets[b->nsets].size || b->nsets));
 
-       if (b->written)
-               oldsize = bch_count_data(b);
 
        if (start) {
                unsigned i;
@@ -1087,7 +1095,7 @@ void bch_btree_sort_partial(struct btree *b, unsigned start)
 
        __btree_sort(b, &iter, start, order, false);
 
-       EBUG_ON(b->written && bch_count_data(b) != oldsize);
+       EBUG_ON(b->written && oldsize >= 0 && bch_count_data(b) != oldsize);
 }
 
 void bch_btree_sort_and_fix_extents(struct btree *b, struct btree_iter *iter)
index 5cd90565dfe2f738388c836d9dc1def40e650fe1..a043a92d4dc9fb1b6597212ac227895d658a9d70 100644 (file)
 
 struct btree_iter {
        size_t size, used;
+#ifdef CONFIG_BCACHE_DEBUG
+       struct btree *b;
+#endif
        struct btree_iter_set {
                struct bkey *k, *end;
        } data[MAX_BSETS];
index aba787d954e5739e82e1c2b5c9794d4023f2917f..fa4d0b1f6d758f3792e46d3fc4a2be59b05aa222 100644 (file)
@@ -216,6 +216,10 @@ static void bch_btree_node_read_done(struct btree *b)
        iter->size = b->c->sb.bucket_size / b->c->sb.block_size;
        iter->used = 0;
 
+#ifdef CONFIG_BCACHE_DEBUG
+       iter->b = b;
+#endif
+
        if (!i->seq)
                goto err;
 
@@ -454,7 +458,7 @@ void bch_btree_node_write(struct btree *b, struct closure *parent)
        BUG_ON(b->written >= btree_blocks(b));
        BUG_ON(b->written && !i->keys);
        BUG_ON(b->sets->data->seq != i->seq);
-       bch_check_key_order(b, i);
+       bch_check_keys(b, "writing");
 
        cancel_delayed_work(&b->work);
 
@@ -1917,7 +1921,7 @@ static bool bch_btree_insert_keys(struct btree *b, struct btree_op *op,
                                  struct bkey *replace_key)
 {
        bool ret = false;
-       unsigned oldsize = bch_count_data(b);
+       int oldsize = bch_count_data(b);
 
        while (!bch_keylist_empty(insert_keys)) {
                struct bset *i = write_block(b);
index 8fc1e8925399a5760b6ceab27b182ec27963ebe4..27e90b18911202f9045672bfde816cca892e9a5d 100644 (file)
@@ -259,14 +259,6 @@ static inline void rw_lock(bool w, struct btree *b, int level)
 
 static inline void rw_unlock(bool w, struct btree *b)
 {
-#ifdef CONFIG_BCACHE_EDEBUG
-       unsigned i;
-
-       if (w && b->key.ptr[0])
-               for (i = 0; i <= b->nsets; i++)
-                       bch_check_key_order(b, b->sets[i].data);
-#endif
-
        if (w)
                b->seq++;
        (w ? up_write : up_read)(&b->lock);
index d9ccb3169aa28660270738671be7d5067063df30..e99e6b8852b228294856dc9b23ab77f494fa1588 100644 (file)
@@ -76,29 +76,17 @@ int bch_bkey_to_text(char *buf, size_t size, const struct bkey *k)
        return out - buf;
 }
 
-int bch_btree_to_text(char *buf, size_t size, const struct btree *b)
-{
-       return scnprintf(buf, size, "%zu level %i/%i",
-                        PTR_BUCKET_NR(b->c, &b->key, 0),
-                        b->level, b->c->root ? b->c->root->level : -1);
-}
-
-#if defined(CONFIG_BCACHE_DEBUG) || defined(CONFIG_BCACHE_EDEBUG)
-
-static bool skipped_backwards(struct btree *b, struct bkey *k)
-{
-       return bkey_cmp(k, (!b->level)
-                       ? &START_KEY(bkey_next(k))
-                       : bkey_next(k)) > 0;
-}
+#ifdef CONFIG_BCACHE_DEBUG
 
 static void dump_bset(struct btree *b, struct bset *i)
 {
-       struct bkey *k;
+       struct bkey *k, *next;
        unsigned j;
        char buf[80];
 
-       for (k = i->start; k < end(i); k = bkey_next(k)) {
+       for (k = i->start; k < end(i); k = next) {
+               next = bkey_next(k);
+
                bch_bkey_to_text(buf, sizeof(buf), k);
                printk(KERN_ERR "block %zu key %zi/%u: %s", index(i, b),
                       (uint64_t *) k - i->d, i->keys, buf);
@@ -114,15 +102,21 @@ static void dump_bset(struct btree *b, struct bset *i)
 
                printk(" %s\n", bch_ptr_status(b->c, k));
 
-               if (bkey_next(k) < end(i) &&
-                   skipped_backwards(b, k))
+               if (next < end(i) &&
+                   bkey_cmp(k, !b->level ? &START_KEY(next) : next) > 0)
                        printk(KERN_ERR "Key skipped backwards\n");
        }
 }
 
-#endif
+static void bch_dump_bucket(struct btree *b)
+{
+       unsigned i;
 
-#ifdef CONFIG_BCACHE_DEBUG
+       console_lock();
+       for (i = 0; i <= b->nsets; i++)
+               dump_bset(b, b->sets[i].data);
+       console_unlock();
+}
 
 void bch_btree_verify(struct btree *b, struct bset *new)
 {
@@ -211,11 +205,7 @@ out_put:
        bio_put(check);
 }
 
-#endif
-
-#ifdef CONFIG_BCACHE_EDEBUG
-
-unsigned bch_count_data(struct btree *b)
+int __bch_count_data(struct btree *b)
 {
        unsigned ret = 0;
        struct btree_iter iter;
@@ -227,72 +217,60 @@ unsigned bch_count_data(struct btree *b)
        return ret;
 }
 
-static void vdump_bucket_and_panic(struct btree *b, const char *fmt,
-                                  va_list args)
-{
-       unsigned i;
-       char buf[80];
-
-       console_lock();
-
-       for (i = 0; i <= b->nsets; i++)
-               dump_bset(b, b->sets[i].data);
-
-       vprintk(fmt, args);
-
-       console_unlock();
-
-       bch_btree_to_text(buf, sizeof(buf), b);
-       panic("at %s\n", buf);
-}
-
-void bch_check_key_order_msg(struct btree *b, struct bset *i,
-                            const char *fmt, ...)
-{
-       struct bkey *k;
-
-       if (!i->keys)
-               return;
-
-       for (k = i->start; bkey_next(k) < end(i); k = bkey_next(k))
-               if (skipped_backwards(b, k)) {
-                       va_list args;
-                       va_start(args, fmt);
-
-                       vdump_bucket_and_panic(b, fmt, args);
-                       va_end(args);
-               }
-}
-
-void bch_check_keys(struct btree *b, const char *fmt, ...)
+void __bch_check_keys(struct btree *b, const char *fmt, ...)
 {
        va_list args;
        struct bkey *k, *p = NULL;
        struct btree_iter iter;
-
-       if (b->level)
-               return;
+       const char *err;
 
        for_each_key(b, k, &iter) {
-               if (p && bkey_cmp(&START_KEY(p), &START_KEY(k)) > 0) {
-                       printk(KERN_ERR "Keys out of order:\n");
-                       goto bug;
-               }
-
-               if (bch_ptr_invalid(b, k))
-                       continue;
-
-               if (p && bkey_cmp(p, &START_KEY(k)) > 0) {
-                       printk(KERN_ERR "Overlapping keys:\n");
-                       goto bug;
+               if (!b->level) {
+                       err = "Keys out of order";
+                       if (p && bkey_cmp(&START_KEY(p), &START_KEY(k)) > 0)
+                               goto bug;
+
+                       if (bch_ptr_invalid(b, k))
+                               continue;
+
+                       err =  "Overlapping keys";
+                       if (p && bkey_cmp(p, &START_KEY(k)) > 0)
+                               goto bug;
+               } else {
+                       if (bch_ptr_bad(b, k))
+                               continue;
+
+                       err = "Duplicate keys";
+                       if (p && !bkey_cmp(p, k))
+                               goto bug;
                }
                p = k;
        }
+
+       err = "Key larger than btree node key";
+       if (p && bkey_cmp(p, &b->key) > 0)
+               goto bug;
+
        return;
 bug:
+       bch_dump_bucket(b);
+
        va_start(args, fmt);
-       vdump_bucket_and_panic(b, fmt, args);
+       vprintk(fmt, args);
        va_end(args);
+
+       panic("bcache error: %s:\n", err);
+}
+
+void bch_btree_iter_next_check(struct btree_iter *iter)
+{
+       struct bkey *k = iter->data->k, *next = bkey_next(k);
+
+       if (next < iter->data->end &&
+           bkey_cmp(k, iter->b->level ? next : &START_KEY(next)) > 0) {
+               bch_dump_bucket(iter->b);
+               panic("Key skipped backwards\n");
+       }
 }
 
 #endif
index 0f4b3440512c83dcfade48c81f8be60b14624b70..7914ba0ff316d70676ecdb5f98780b838b4c7844 100644 (file)
@@ -4,40 +4,42 @@
 /* Btree/bkey debug printing */
 
 int bch_bkey_to_text(char *buf, size_t size, const struct bkey *k);
-int bch_btree_to_text(char *buf, size_t size, const struct btree *b);
-
-#ifdef CONFIG_BCACHE_EDEBUG
-
-unsigned bch_count_data(struct btree *);
-void bch_check_key_order_msg(struct btree *, struct bset *, const char *, ...);
-void bch_check_keys(struct btree *, const char *, ...);
-
-#define bch_check_key_order(b, i)                      \
-       bch_check_key_order_msg(b, i, "keys out of order")
-#define EBUG_ON(cond)          BUG_ON(cond)
-
-#else /* EDEBUG */
-
-#define bch_count_data(b)                              0
-#define bch_check_key_order(b, i)                      do {} while (0)
-#define bch_check_key_order_msg(b, i, ...)             do {} while (0)
-#define bch_check_keys(b, ...)                         do {} while (0)
-#define EBUG_ON(cond)                                  do {} while (0)
-
-#endif
 
 #ifdef CONFIG_BCACHE_DEBUG
 
 void bch_btree_verify(struct btree *, struct bset *);
 void bch_data_verify(struct cached_dev *, struct bio *);
+int __bch_count_data(struct btree *);
+void __bch_check_keys(struct btree *, const char *, ...);
+void bch_btree_iter_next_check(struct btree_iter *);
+
+#define EBUG_ON(cond)                  BUG_ON(cond)
+#define expensive_debug_checks(c)      ((c)->expensive_debug_checks)
+#define key_merging_disabled(c)                ((c)->key_merging_disabled)
 
 #else /* DEBUG */
 
 static inline void bch_btree_verify(struct btree *b, struct bset *i) {}
-static inline void bch_data_verify(struct cached_dev *dc, struct bio *bio) {};
+static inline void bch_data_verify(struct cached_dev *dc, struct bio *bio) {}
+static inline int __bch_count_data(struct btree *b) { return -1; }
+static inline void __bch_check_keys(struct btree *b, const char *fmt, ...) {}
+static inline void bch_btree_iter_next_check(struct btree_iter *iter) {}
+
+#define EBUG_ON(cond)                  do { if (cond); } while (0)
+#define expensive_debug_checks(c)      0
+#define key_merging_disabled(c)                0
 
 #endif
 
+#define bch_count_data(b)                                              \
+       (expensive_debug_checks((b)->c) ? __bch_count_data(b) : -1)
+
+#define bch_check_keys(b, ...)                                         \
+do {                                                                   \
+       if (expensive_debug_checks((b)->c))                             \
+               __bch_check_keys(b, __VA_ARGS__);                       \
+} while (0)
+
 #ifdef CONFIG_DEBUG_FS
 void bch_debug_init_cache_set(struct cache_set *);
 #else
index ab286b9b5e40a1eb9e11c36d767258b0071f0969..9687771ec6f3286400d99566cdd98678aafbb018 100644 (file)
@@ -102,6 +102,7 @@ rw_attribute(io_error_halflife);
 rw_attribute(verify);
 rw_attribute(key_merging_disabled);
 rw_attribute(gc_always_rewrite);
+rw_attribute(expensive_debug_checks);
 rw_attribute(freelist_percent);
 rw_attribute(cache_replacement_policy);
 rw_attribute(btree_shrinker_disabled);
@@ -517,6 +518,8 @@ lock_root:
        sysfs_print(active_journal_entries,     fifo_used(&c->journal.pin));
        sysfs_printf(verify,                    "%i", c->verify);
        sysfs_printf(key_merging_disabled,      "%i", c->key_merging_disabled);
+       sysfs_printf(expensive_debug_checks,
+                    "%i", c->expensive_debug_checks);
        sysfs_printf(gc_always_rewrite,         "%i", c->gc_always_rewrite);
        sysfs_printf(btree_shrinker_disabled,   "%i", c->shrinker_disabled);
        sysfs_printf(copy_gc_enabled,           "%i", c->copy_gc_enabled);
@@ -599,6 +602,7 @@ STORE(__bch_cache_set)
        sysfs_strtoul(journal_delay_ms,         c->journal_delay_ms);
        sysfs_strtoul(verify,                   c->verify);
        sysfs_strtoul(key_merging_disabled,     c->key_merging_disabled);
+       sysfs_strtoul(expensive_debug_checks,   c->expensive_debug_checks);
        sysfs_strtoul(gc_always_rewrite,        c->gc_always_rewrite);
        sysfs_strtoul(btree_shrinker_disabled,  c->shrinker_disabled);
        sysfs_strtoul(copy_gc_enabled,          c->copy_gc_enabled);
@@ -674,6 +678,7 @@ static struct attribute *bch_cache_set_internal_files[] = {
 #ifdef CONFIG_BCACHE_DEBUG
        &sysfs_verify,
        &sysfs_key_merging_disabled,
+       &sysfs_expensive_debug_checks,
 #endif
        &sysfs_gc_always_rewrite,
        &sysfs_btree_shrinker_disabled,
index 38ae7a4ce928740cf7efeb702363dfe5bdbe36a1..8ce5aab55962b8bfc16c2e0429151273c5f3b737 100644 (file)
 
 struct closure;
 
-#ifdef CONFIG_BCACHE_EDEBUG
+#ifdef CONFIG_BCACHE_DEBUG
 
 #define atomic_dec_bug(v)      BUG_ON(atomic_dec_return(v) < 0)
 #define atomic_inc_bug(v, i)   BUG_ON(atomic_inc_return(v) <= i)
 
-#else /* EDEBUG */
+#else /* DEBUG */
 
 #define atomic_dec_bug(v)      atomic_dec(v)
 #define atomic_inc_bug(v, i)   atomic_inc(v)