diff --git a/include/jemalloc/internal/arena_inlines_b.h b/include/jemalloc/internal/arena_inlines_b.h
index 005f422e..de1c8149 100644
--- a/include/jemalloc/internal/arena_inlines_b.h
+++ b/include/jemalloc/internal/arena_inlines_b.h
@@ -9,7 +9,7 @@
 #include "jemalloc/internal/emap.h"
 #include "jemalloc/internal/jemalloc_internal_inlines_b.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
-#include "jemalloc/internal/large_externs.h"
+#include "jemalloc/internal/large.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/prof_externs.h"
 #include "jemalloc/internal/prof_structs.h"
diff --git a/include/jemalloc/internal/background_thread_structs.h b/include/jemalloc/internal/background_thread.h
similarity index 58%
rename from include/jemalloc/internal/background_thread_structs.h
rename to include/jemalloc/internal/background_thread.h
index d56673da..db26f1b8 100644
--- a/include/jemalloc/internal/background_thread_structs.h
+++ b/include/jemalloc/internal/background_thread.h
@@ -1,11 +1,10 @@
-#ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_STRUCTS_H
-#define JEMALLOC_INTERNAL_BACKGROUND_THREAD_STRUCTS_H
+#ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_H
+#define JEMALLOC_INTERNAL_BACKGROUND_THREAD_H
 
 #include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/base.h"
 #include "jemalloc/internal/mutex.h"
 
-/* This file really combines "structs" and "types", but only transitionally. */
-
 #if defined(JEMALLOC_BACKGROUND_THREAD) || defined(JEMALLOC_LAZY_LOCK)
 #	define JEMALLOC_PTHREAD_CREATE_WRAPPER
 #endif
@@ -66,4 +65,33 @@ struct background_thread_stats_s {
 };
 typedef struct background_thread_stats_s background_thread_stats_t;
 
-#endif /* JEMALLOC_INTERNAL_BACKGROUND_THREAD_STRUCTS_H */
+extern bool                      opt_background_thread;
+extern size_t                    opt_max_background_threads;
+extern malloc_mutex_t            background_thread_lock;
+extern atomic_b_t                background_thread_enabled_state;
+extern size_t                    n_background_threads;
+extern size_t                    max_background_threads;
+extern background_thread_info_t *background_thread_info;
+
+bool background_thread_create(tsd_t *tsd, unsigned arena_ind);
+bool background_threads_enable(tsd_t *tsd);
+bool background_threads_disable(tsd_t *tsd);
+bool background_thread_is_started(background_thread_info_t *info);
+void background_thread_wakeup_early(
+    background_thread_info_t *info, nstime_t *remaining_sleep);
+void background_thread_prefork0(tsdn_t *tsdn);
+void background_thread_prefork1(tsdn_t *tsdn);
+void background_thread_postfork_parent(tsdn_t *tsdn);
+void background_thread_postfork_child(tsdn_t *tsdn);
+bool background_thread_stats_read(
+    tsdn_t *tsdn, background_thread_stats_t *stats);
+void background_thread_ctl_init(tsdn_t *tsdn);
+
+#ifdef JEMALLOC_PTHREAD_CREATE_WRAPPER
+extern int pthread_create_wrapper(pthread_t *__restrict, const pthread_attr_t *,
+    void *(*)(void *), void *__restrict);
+#endif
+bool background_thread_boot0(void);
+bool background_thread_boot1(tsdn_t *tsdn, base_t *base);
+
+#endif /* JEMALLOC_INTERNAL_BACKGROUND_THREAD_H */
diff --git a/include/jemalloc/internal/background_thread_externs.h b/include/jemalloc/internal/background_thread_externs.h
deleted file mode 100644
index efc0aaa4..00000000
--- a/include/jemalloc/internal/background_thread_externs.h
+++ /dev/null
@@ -1,38 +0,0 @@
-#ifndef JEMALLOC_INTERNAL_BACKGROUND_THREAD_EXTERNS_H
-#define JEMALLOC_INTERNAL_BACKGROUND_THREAD_EXTERNS_H
-
-#include "jemalloc/internal/jemalloc_preamble.h"
-#include "jemalloc/internal/background_thread_structs.h"
-#include "jemalloc/internal/base.h"
-#include "jemalloc/internal/mutex.h"
-
-extern bool                      opt_background_thread;
-extern size_t                    opt_max_background_threads;
-extern malloc_mutex_t            background_thread_lock;
-extern atomic_b_t                background_thread_enabled_state;
-extern size_t                    n_background_threads;
-extern size_t                    max_background_threads;
-extern background_thread_info_t *background_thread_info;
-
-bool background_thread_create(tsd_t *tsd, unsigned arena_ind);
-bool background_threads_enable(tsd_t *tsd);
-bool background_threads_disable(tsd_t *tsd);
-bool background_thread_is_started(background_thread_info_t *info);
-void background_thread_wakeup_early(
-    background_thread_info_t *info, nstime_t *remaining_sleep);
-void background_thread_prefork0(tsdn_t *tsdn);
-void background_thread_prefork1(tsdn_t *tsdn);
-void background_thread_postfork_parent(tsdn_t *tsdn);
-void background_thread_postfork_child(tsdn_t *tsdn);
-bool background_thread_stats_read(
-    tsdn_t *tsdn, background_thread_stats_t *stats);
-void background_thread_ctl_init(tsdn_t *tsdn);
-
-#ifdef JEMALLOC_PTHREAD_CREATE_WRAPPER
-extern int pthread_create_wrapper(pthread_t *__restrict, const pthread_attr_t *,
-    void *(*)(void *), void *__restrict);
-#endif
-bool background_thread_boot0(void);
-bool background_thread_boot1(tsdn_t *tsdn, base_t *base);
-
-#endif /* JEMALLOC_INTERNAL_BACKGROUND_THREAD_EXTERNS_H */
diff --git a/include/jemalloc/internal/background_thread_inlines.h b/include/jemalloc/internal/background_thread_inlines.h
index e822a3f7..67517626 100644
--- a/include/jemalloc/internal/background_thread_inlines.h
+++ b/include/jemalloc/internal/background_thread_inlines.h
@@ -4,7 +4,7 @@
 #include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/arena_inlines_a.h"
 #include "jemalloc/internal/atomic.h"
-#include "jemalloc/internal/background_thread_externs.h"
+#include "jemalloc/internal/background_thread.h"
 
 JEMALLOC_ALWAYS_INLINE bool
 background_thread_enabled(void) {
diff --git a/include/jemalloc/internal/bin.h b/include/jemalloc/internal/bin.h
index 45c6c5dd..90fc3d31 100644
--- a/include/jemalloc/internal/bin.h
+++ b/include/jemalloc/internal/bin.h
@@ -4,11 +4,17 @@
 #include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/bin_info.h"
 #include "jemalloc/internal/bin_stats.h"
-#include "jemalloc/internal/bin_types.h"
+#include "jemalloc/internal/div.h"
 #include "jemalloc/internal/edata.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/sc.h"
 
+/* Forward decl; only used as a pointer type in bin_choose. */
+typedef struct arena_s arena_t;
+
+#define BIN_SHARDS_MAX (1 << EDATA_BITS_BINSHARD_WIDTH)
+#define N_BIN_SHARDS_DEFAULT 1
+
 /*
  * A bin contains a set of extents that are currently being used for slab
  * allocations.
@@ -102,30 +108,11 @@ void *bin_current_slab_addr(tsdn_t *tsdn, bin_t *bin);
 bin_t *bin_choose(tsdn_t *tsdn, arena_t *arena, szind_t binind,
     unsigned *binshard_p);
 
-/* Stats. */
-static inline void
-bin_stats_nrequests_add(tsdn_t *tsdn, bin_t *bin, uint64_t n) {
-	malloc_mutex_lock(tsdn, &bin->lock);
-	bin->stats.nrequests += n;
-	malloc_mutex_unlock(tsdn, &bin->lock);
-}
-
-static inline void
-bin_stats_merge(tsdn_t *tsdn, bin_stats_data_t *dst_bin_stats, bin_t *bin) {
-	malloc_mutex_lock(tsdn, &bin->lock);
-	malloc_mutex_prof_accum(tsdn, &dst_bin_stats->mutex_data, &bin->lock);
-	bin_stats_t *stats = &dst_bin_stats->stats_data;
-	stats->nmalloc += bin->stats.nmalloc;
-	stats->ndalloc += bin->stats.ndalloc;
-	stats->nrequests += bin->stats.nrequests;
-	stats->curregs += bin->stats.curregs;
-	stats->nfills += bin->stats.nfills;
-	stats->nflushes += bin->stats.nflushes;
-	stats->nslabs += bin->stats.nslabs;
-	stats->reslabs += bin->stats.reslabs;
-	stats->curslabs += bin->stats.curslabs;
-	stats->nonfull_slabs += bin->stats.nonfull_slabs;
-	malloc_mutex_unlock(tsdn, &bin->lock);
-}
+typedef struct bin_dalloc_locked_info_s bin_dalloc_locked_info_t;
+struct bin_dalloc_locked_info_s {
+	div_info_t div_info;
+	uint32_t   nregs;
+	uint64_t   ndalloc;
+};
 
 #endif /* JEMALLOC_INTERNAL_BIN_H */
diff --git a/include/jemalloc/internal/bin_inlines.h b/include/jemalloc/internal/bin_inlines.h
index 31fe4818..5d7a9d04 100644
--- a/include/jemalloc/internal/bin_inlines.h
+++ b/include/jemalloc/internal/bin_inlines.h
@@ -7,20 +7,33 @@
 #include "jemalloc/internal/bitmap.h"
 #include "jemalloc/internal/div.h"
 #include "jemalloc/internal/edata.h"
-#include "jemalloc/internal/sc.h"
+#include "jemalloc/internal/mutex.h"
 
-/*
- * The dalloc bin info contains just the information that the common paths need
- * during tcache flushes.  By force-inlining these paths, and using local copies
- * of data (so that the compiler knows it's constant), we avoid a whole bunch of
- * redundant loads and stores by leaving this information in registers.
- */
-typedef struct bin_dalloc_locked_info_s bin_dalloc_locked_info_t;
-struct bin_dalloc_locked_info_s {
-	div_info_t div_info;
-	uint32_t   nregs;
-	uint64_t   ndalloc;
-};
+/* Stats. */
+static inline void
+bin_stats_nrequests_add(tsdn_t *tsdn, bin_t *bin, uint64_t n) {
+	malloc_mutex_lock(tsdn, &bin->lock);
+	bin->stats.nrequests += n;
+	malloc_mutex_unlock(tsdn, &bin->lock);
+}
+
+static inline void
+bin_stats_merge(tsdn_t *tsdn, bin_stats_data_t *dst_bin_stats, bin_t *bin) {
+	malloc_mutex_lock(tsdn, &bin->lock);
+	malloc_mutex_prof_accum(tsdn, &dst_bin_stats->mutex_data, &bin->lock);
+	bin_stats_t *stats = &dst_bin_stats->stats_data;
+	stats->nmalloc += bin->stats.nmalloc;
+	stats->ndalloc += bin->stats.ndalloc;
+	stats->nrequests += bin->stats.nrequests;
+	stats->curregs += bin->stats.curregs;
+	stats->nfills += bin->stats.nfills;
+	stats->nflushes += bin->stats.nflushes;
+	stats->nslabs += bin->stats.nslabs;
+	stats->reslabs += bin->stats.reslabs;
+	stats->curslabs += bin->stats.curslabs;
+	stats->nonfull_slabs += bin->stats.nonfull_slabs;
+	malloc_mutex_unlock(tsdn, &bin->lock);
+}
 
 /* Find the region index of a pointer within a slab. */
 JEMALLOC_ALWAYS_INLINE size_t
@@ -52,14 +65,6 @@ bin_slab_regind(const bin_dalloc_locked_info_t *info, szind_t binind,
 	return regind;
 }
 
-JEMALLOC_ALWAYS_INLINE void
-bin_dalloc_locked_begin(
-    bin_dalloc_locked_info_t *info, szind_t binind) {
-	info->div_info = arena_binind_div_info[binind];
-	info->nregs = bin_infos[binind].nregs;
-	info->ndalloc = 0;
-}
-
 /*
  * Does the deallocation work associated with freeing a single pointer (a
  * "step") in between a bin_dalloc_locked begin and end call.
diff --git a/include/jemalloc/internal/bin_types.h b/include/jemalloc/internal/bin_types.h
deleted file mode 100644
index b6bad37e..00000000
--- a/include/jemalloc/internal/bin_types.h
+++ /dev/null
@@ -1,21 +0,0 @@
-#ifndef JEMALLOC_INTERNAL_BIN_TYPES_H
-#define JEMALLOC_INTERNAL_BIN_TYPES_H
-
-#include "jemalloc/internal/jemalloc_preamble.h"
-#include "jemalloc/internal/sc.h"
-
-#define BIN_SHARDS_MAX (1 << EDATA_BITS_BINSHARD_WIDTH)
-#define N_BIN_SHARDS_DEFAULT 1
-
-/* Used in TSD static initializer only. Real init in arena_bind(). */
-#define TSD_BINSHARDS_ZERO_INITIALIZER                                         \
-	{                                                                      \
-		{ UINT8_MAX }                                                  \
-	}
-
-typedef struct tsd_binshards_s tsd_binshards_t;
-struct tsd_binshards_s {
-	uint8_t binshard[SC_NBINS];
-};
-
-#endif /* JEMALLOC_INTERNAL_BIN_TYPES_H */
diff --git a/include/jemalloc/internal/ctl.h b/include/jemalloc/internal/ctl.h
index e7a8221c..09f1825c 100644
--- a/include/jemalloc/internal/ctl.h
+++ b/include/jemalloc/internal/ctl.h
@@ -3,7 +3,7 @@
 
 #include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/arena_stats.h"
-#include "jemalloc/internal/background_thread_structs.h"
+#include "jemalloc/internal/background_thread.h"
 #include "jemalloc/internal/bin_stats.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
 #include "jemalloc/internal/malloc_io.h"
diff --git a/include/jemalloc/internal/jemalloc_internal_includes.h b/include/jemalloc/internal/jemalloc_internal_includes.h
index d198d09c..63b65599 100644
--- a/include/jemalloc/internal/jemalloc_internal_includes.h
+++ b/include/jemalloc/internal/jemalloc_internal_includes.h
@@ -51,7 +51,6 @@
 #include "jemalloc/internal/prof_structs.h"
 #include "jemalloc/internal/arena_structs.h"
 #include "jemalloc/internal/tcache_structs.h"
-#include "jemalloc/internal/background_thread_structs.h"
 
 /******************************************************************************/
 /* EXTERNS */
@@ -59,11 +58,11 @@
 
 #include "jemalloc/internal/jemalloc_internal_externs.h"
 #include "jemalloc/internal/arena_externs.h"
-#include "jemalloc/internal/large_externs.h"
+#include "jemalloc/internal/large.h"
 #include "jemalloc/internal/tcache_externs.h"
 #include "jemalloc/internal/malloc_dispatch.h"
 #include "jemalloc/internal/prof_externs.h"
-#include "jemalloc/internal/background_thread_externs.h"
+#include "jemalloc/internal/background_thread.h"
 
 /******************************************************************************/
 /* INLINES */
diff --git a/include/jemalloc/internal/large_externs.h b/include/jemalloc/internal/large.h
similarity index 87%
rename from include/jemalloc/internal/large_externs.h
rename to include/jemalloc/internal/large.h
index 82abd7de..e28841c1 100644
--- a/include/jemalloc/internal/large_externs.h
+++ b/include/jemalloc/internal/large.h
@@ -1,5 +1,5 @@
-#ifndef JEMALLOC_INTERNAL_LARGE_EXTERNS_H
-#define JEMALLOC_INTERNAL_LARGE_EXTERNS_H
+#ifndef JEMALLOC_INTERNAL_LARGE_H
+#define JEMALLOC_INTERNAL_LARGE_H
 
 #include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/edata.h"
@@ -20,4 +20,4 @@ void   large_prof_info_get(
 void large_prof_tctx_reset(edata_t *edata);
 void large_prof_info_set(edata_t *edata, prof_tctx_t *tctx, size_t size);
 
-#endif /* JEMALLOC_INTERNAL_LARGE_EXTERNS_H */
+#endif /* JEMALLOC_INTERNAL_LARGE_H */
diff --git a/include/jemalloc/internal/malloc_dispatch_inlines.h b/include/jemalloc/internal/malloc_dispatch_inlines.h
index e18192bf..5632e27a 100644
--- a/include/jemalloc/internal/malloc_dispatch_inlines.h
+++ b/include/jemalloc/internal/malloc_dispatch_inlines.h
@@ -9,7 +9,7 @@
 #include "jemalloc/internal/emap.h"
 #include "jemalloc/internal/jemalloc_internal_inlines_b.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
-#include "jemalloc/internal/large_externs.h"
+#include "jemalloc/internal/large.h"
 #include "jemalloc/internal/malloc_dispatch.h"
 #include "jemalloc/internal/safety_check.h"
 #include "jemalloc/internal/sc.h"
diff --git a/include/jemalloc/internal/tcache_inlines.h b/include/jemalloc/internal/tcache_inlines.h
index b3ce81bd..181db1b3 100644
--- a/include/jemalloc/internal/tcache_inlines.h
+++ b/include/jemalloc/internal/tcache_inlines.h
@@ -6,7 +6,7 @@
 #include "jemalloc/internal/bin.h"
 #include "jemalloc/internal/jemalloc_internal_inlines_b.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
-#include "jemalloc/internal/large_externs.h"
+#include "jemalloc/internal/large.h"
 #include "jemalloc/internal/san.h"
 #include "jemalloc/internal/sc.h"
 #include "jemalloc/internal/sz.h"
diff --git a/include/jemalloc/internal/tsd_binshards.h b/include/jemalloc/internal/tsd_binshards.h
new file mode 100644
index 00000000..a4afc496
--- /dev/null
+++ b/include/jemalloc/internal/tsd_binshards.h
@@ -0,0 +1,24 @@
+#ifndef JEMALLOC_INTERNAL_TSD_BINSHARDS_H
+#define JEMALLOC_INTERNAL_TSD_BINSHARDS_H
+
+#include "jemalloc/internal/jemalloc_preamble.h"
+#include "jemalloc/internal/sc.h"
+
+/*
+ * Per-thread cache of bin-shard assignments.  This lives in its own header
+ * (separate from bin.h) so that tsd_internals.h can pull it in for X-macro
+ * expansion without dragging in mutex.h, which itself depends on TSD machinery
+ * and would form an include-order dependency cycle.
+ */
+
+#define TSD_BINSHARDS_ZERO_INITIALIZER                                         \
+	{                                                                      \
+		{ UINT8_MAX }                                                  \
+	}
+
+typedef struct tsd_binshards_s tsd_binshards_t;
+struct tsd_binshards_s {
+	uint8_t binshard[SC_NBINS];
+};
+
+#endif /* JEMALLOC_INTERNAL_TSD_BINSHARDS_H */
diff --git a/include/jemalloc/internal/tsd_internals.h b/include/jemalloc/internal/tsd_internals.h
index 46b4930f..a7b6fa5e 100644
--- a/include/jemalloc/internal/tsd_internals.h
+++ b/include/jemalloc/internal/tsd_internals.h
@@ -6,7 +6,7 @@
 #include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/arena_types.h"
 #include "jemalloc/internal/assert.h"
-#include "jemalloc/internal/bin_types.h"
+#include "jemalloc/internal/tsd_binshards.h"
 #include "jemalloc/internal/jemalloc_internal_externs.h"
 #include "jemalloc/internal/peak.h"
 #include "jemalloc/internal/prof_types.h"
diff --git a/src/arena.c b/src/arena.c
index 3f4e7b65..e8d90458 100644
--- a/src/arena.c
+++ b/src/arena.c
@@ -36,6 +36,14 @@ static pa_central_t arena_pa_central_global;
 
 div_info_t arena_binind_div_info[SC_NBINS];
 
+JET_EXTERN void
+bin_dalloc_locked_begin(
+    bin_dalloc_locked_info_t *info, szind_t binind) {
+	info->div_info = arena_binind_div_info[binind];
+	info->nregs = bin_infos[binind].nregs;
+	info->ndalloc = 0;
+}
+
 size_t opt_oversize_threshold = OVERSIZE_THRESHOLD_DEFAULT;
 size_t oversize_threshold = OVERSIZE_THRESHOLD_DEFAULT;
 
diff --git a/test/unit/bin.c b/test/unit/bin.c
index 08dd4665..a7091544 100644
--- a/test/unit/bin.c
+++ b/test/unit/bin.c
@@ -1,5 +1,8 @@
 #include "test/jemalloc_test.h"
 
+extern void bin_dalloc_locked_begin(
+    bin_dalloc_locked_info_t *info, szind_t binind);
+
 #define INVALID_ARENA_IND ((1U << MALLOCX_ARENA_BITS) - 1)
 
 /* Create a page-aligned mock slab with all regions free. */
diff --git a/test/unit/slab.c b/test/unit/slab.c
index d98663e8..e7adafea 100644
--- a/test/unit/slab.c
+++ b/test/unit/slab.c
@@ -1,5 +1,8 @@
 #include "test/jemalloc_test.h"
 
+extern void bin_dalloc_locked_begin(
+    bin_dalloc_locked_info_t *info, szind_t binind);
+
 #define INVALID_ARENA_IND ((1U << MALLOCX_ARENA_BITS) - 1)
 
 TEST_BEGIN(test_bin_slab_regind) {