Implement the arena.<i>.reset mallctl.

This makes it possible to discard all of an arena's allocations in a
single operation.

This resolves #146.

8 files changed, 411 insertions, 39 deletions

diff --git a/Makefile.in b/Makefile.in
index a872eb5..ddc8915 100644
--- a/Makefile.in
+++ b/Makefile.in
@@ -135,7 +135,9 @@ C_TESTLIB_SRCS := $(srcroot)test/src/btalloc.c $(srcroot)test/src/btalloc_0.c \
 	$(srcroot)test/src/SFMT.c $(srcroot)test/src/test.c \
 	$(srcroot)test/src/thd.c $(srcroot)test/src/timer.c
 C_UTIL_INTEGRATION_SRCS := $(srcroot)src/nstime.c $(srcroot)src/util.c
-TESTS_UNIT := $(srcroot)test/unit/atomic.c \
+TESTS_UNIT := \
+	$(srcroot)test/unit/arena_reset.c \
+	$(srcroot)test/unit/atomic.c \
 	$(srcroot)test/unit/bitmap.c \
 	$(srcroot)test/unit/ckh.c \
 	$(srcroot)test/unit/decay.c \
diff --git a/doc/jemalloc.xml.in b/doc/jemalloc.xml.in
index 9814c22..7b602a5 100644
--- a/doc/jemalloc.xml.in
+++ b/doc/jemalloc.xml.in
@@ -1558,6 +1558,23 @@ malloc_conf = "xmalloc:true";]]></programlisting>
         details.</para></listitem>
       </varlistentry>
 
+      <varlistentry id="arena.i.reset">
+        <term>
+          <mallctl>arena.&lt;i&gt;.reset</mallctl>
+          (<type>void</type>)
+          <literal>--</literal>
+        </term>
+        <listitem><para>Discard all of the arena's extant allocations.  This
+        interface can only be used with arenas created via <link
+        linkend="arenas.extend"><mallctl>arenas.extend</mallctl></link>.  None
+        of the arena's discarded/cached allocations may accessed afterward.  As
+        part of this requirement, all thread caches which were used to
+        allocate/deallocate in conjunction with the arena must be flushed
+        beforehand.  This interface cannot be used if running inside Valgrind,
+        nor if the <link linkend="opt.quarantine">quarantine</link> size is
+        non-zero.</para></listitem>
+      </varlistentry>
+
       <varlistentry id="arena.i.dss">
         <term>
           <mallctl>arena.&lt;i&gt;.dss</mallctl>
diff --git a/include/jemalloc/internal/arena.h b/include/jemalloc/internal/arena.h
index 103a4c9..f2685f6 100644
--- a/include/jemalloc/internal/arena.h
+++ b/include/jemalloc/internal/arena.h
@@ -330,6 +330,10 @@ struct arena_s {
 
 	dss_prec_t		dss_prec;
 
+
+	/* Extant arena chunks. */
+	ql_head(extent_node_t)	achunks;
+
 	/*
 	 * In order to avoid rapid chunk allocation/deallocation when an arena
 	 * oscillates right on the cusp of needing a new chunk, cache the most
@@ -533,6 +537,7 @@ ssize_t	arena_decay_time_get(tsd_t *tsd, arena_t *arena);
 bool	arena_decay_time_set(tsd_t *tsd, arena_t *arena, ssize_t decay_time);
 void	arena_purge(tsd_t *tsd, arena_t *arena, bool all);
 void	arena_maybe_purge(tsd_t *tsd, arena_t *arena);
+void	arena_reset(tsd_t *tsd, arena_t *arena);
 void	arena_tcache_fill_small(tsd_t *tsd, arena_t *arena, tcache_bin_t *tbin,
     szind_t binind, uint64_t prof_accumbytes);
 void	arena_alloc_junk_small(void *ptr, arena_bin_info_t *bin_info,
diff --git a/include/jemalloc/internal/extent.h b/include/jemalloc/internal/extent.h
index 386d50e..49d76a5 100644
--- a/include/jemalloc/internal/extent.h
+++ b/include/jemalloc/internal/extent.h
@@ -48,7 +48,7 @@ struct extent_node_s {
 		/* Linkage for the size/address-ordered tree. */
 		rb_node(extent_node_t)	szad_link;
 
-		/* Linkage for arena's huge and node_cache lists. */
+		/* Linkage for arena's achunks, huge, and node_cache lists. */
 		ql_elm(extent_node_t)	ql_link;
 	};
 
diff --git a/include/jemalloc/internal/private_symbols.txt b/include/jemalloc/internal/private_symbols.txt
index eacc7c6..c7ff852 100644
--- a/include/jemalloc/internal/private_symbols.txt
+++ b/include/jemalloc/internal/private_symbols.txt
@@ -100,6 +100,7 @@ arena_ralloc_junk_large
 arena_ralloc_no_move
 arena_rd_to_miscelm
 arena_redzone_corruption
+arena_reset
 arena_run_regind
 arena_run_to_miscelm
 arena_salloc
diff --git a/src/arena.c b/src/arena.c
index 0da832e..f752aca 100644
--- a/src/arena.c
+++ b/src/arena.c
@@ -738,14 +738,61 @@ arena_chunk_alloc(tsd_t *tsd, arena_t *arena)
 			return (NULL);
 	}
 
+	ql_elm_new(&chunk->node, ql_link);
+	ql_tail_insert(&arena->achunks, &chunk->node, ql_link);
 	arena_avail_insert(arena, chunk, map_bias, chunk_npages-map_bias);
 
 	return (chunk);
 }
 
 static void
+arena_chunk_discard(tsd_t *tsd, arena_t *arena, arena_chunk_t *chunk)
+{
+	bool committed;
+	chunk_hooks_t chunk_hooks = CHUNK_HOOKS_INITIALIZER;
+
+	chunk_deregister(chunk, &chunk->node);
+
+	committed = (arena_mapbits_decommitted_get(chunk, map_bias) == 0);
+	if (!committed) {
+		/*
+		 * Decommit the header.  Mark the chunk as decommitted even if
+		 * header decommit fails, since treating a partially committed
+		 * chunk as committed has a high potential for causing later
+		 * access of decommitted memory.
+		 */
+		chunk_hooks = chunk_hooks_get(tsd, arena);
+		chunk_hooks.decommit(chunk, chunksize, 0, map_bias << LG_PAGE,
+		    arena->ind);
+	}
+
+	chunk_dalloc_cache(tsd, arena, &chunk_hooks, (void *)chunk, chunksize,
+	    committed);
+
+	if (config_stats) {
+		arena->stats.mapped -= chunksize;
+		arena->stats.metadata_mapped -= (map_bias << LG_PAGE);
+	}
+}
+
+static void
+arena_spare_discard(tsd_t *tsd, arena_t *arena, arena_chunk_t *spare)
+{
+
+	assert(arena->spare != spare);
+
+	if (arena_mapbits_dirty_get(spare, map_bias) != 0) {
+		arena_run_dirty_remove(arena, spare, map_bias,
+		    chunk_npages-map_bias);
+	}
+
+	arena_chunk_discard(tsd, arena, spare);
+}
+
+static void
 arena_chunk_dalloc(tsd_t *tsd, arena_t *arena, arena_chunk_t *chunk)
 {
+	arena_chunk_t *spare;
 
 	assert(arena_mapbits_allocated_get(chunk, map_bias) == 0);
 	assert(arena_mapbits_allocated_get(chunk, chunk_npages-1) == 0);
@@ -761,43 +808,11 @@ arena_chunk_dalloc(tsd_t *tsd, arena_t *arena, arena_chunk_t *chunk)
 	/* Remove run from runs_avail, so that the arena does not use it. */
 	arena_avail_remove(arena, chunk, map_bias, chunk_npages-map_bias);
 
-	if (arena->spare != NULL) {
-		arena_chunk_t *spare = arena->spare;
-		chunk_hooks_t chunk_hooks = CHUNK_HOOKS_INITIALIZER;
-		bool committed;
-
-		arena->spare = chunk;
-		if (arena_mapbits_dirty_get(spare, map_bias) != 0) {
-			arena_run_dirty_remove(arena, spare, map_bias,
-			    chunk_npages-map_bias);
-		}
-
-		chunk_deregister(spare, &spare->node);
-
-		committed = (arena_mapbits_decommitted_get(spare, map_bias) ==
-		    0);
-		if (!committed) {
-			/*
-			 * Decommit the header.  Mark the chunk as decommitted
-			 * even if header decommit fails, since treating a
-			 * partially committed chunk as committed has a high
-			 * potential for causing later access of decommitted
-			 * memory.
-			 */
-			chunk_hooks = chunk_hooks_get(tsd, arena);
-			chunk_hooks.decommit(spare, chunksize, 0, map_bias <<
-			    LG_PAGE, arena->ind);
-		}
-
-		chunk_dalloc_cache(tsd, arena, &chunk_hooks, (void *)spare,
-		    chunksize, committed);
-
-		if (config_stats) {
-			arena->stats.mapped -= chunksize;
-			arena->stats.metadata_mapped -= (map_bias << LG_PAGE);
-		}
-	} else
-		arena->spare = chunk;
+	ql_remove(&arena->achunks, &chunk->node, ql_link);
+	spare = arena->spare;
+	arena->spare = chunk;
+	if (spare != NULL)
+		arena_spare_discard(tsd, arena, spare);
 }
 
 static void
@@ -1803,6 +1818,140 @@ arena_purge(tsd_t *tsd, arena_t *arena, bool all)
 }
 
 static void
+arena_achunk_prof_reset(tsd_t *tsd, arena_t *arena, arena_chunk_t *chunk)
+{
+	size_t pageind, npages;
+
+	cassert(config_prof);
+	assert(opt_prof);
+
+	/*
+	 * Iterate over the allocated runs and remove profiled allocations from
+	 * the sample set.
+	 */
+	for (pageind = map_bias; pageind < chunk_npages; pageind += npages) {
+		if (arena_mapbits_allocated_get(chunk, pageind) != 0) {
+			if (arena_mapbits_large_get(chunk, pageind) != 0) {
+				void *ptr = (void *)((uintptr_t)chunk + (pageind
+				    << LG_PAGE));
+				size_t usize = isalloc(tsd, ptr, config_prof);
+
+				prof_free(tsd, ptr, usize);
+				npages = arena_mapbits_large_size_get(chunk,
+				    pageind) >> LG_PAGE;
+			} else {
+				/* Skip small run. */
+				size_t binind = arena_mapbits_binind_get(chunk,
+				    pageind);
+				arena_bin_info_t *bin_info =
+				    &arena_bin_info[binind];
+				npages = bin_info->run_size >> LG_PAGE;
+			}
+		} else {
+			/* Skip unallocated run. */
+			npages = arena_mapbits_unallocated_size_get(chunk,
+			    pageind) >> LG_PAGE;
+		}
+		assert(pageind + npages <= chunk_npages);
+	}
+}
+
+void
+arena_reset(tsd_t *tsd, arena_t *arena)
+{
+	unsigned i;
+	extent_node_t *node;
+
+	/*
+	 * Locking in this function is unintuitive.  The caller guarantees that
+	 * no concurrent operations are happening in this arena, but there are
+	 * still reasons that some locking is necessary:
+	 *
+	 * - Some of the functions in the transitive closure of calls assume
+	 *   appropriate locks are held, and in some cases these locks are
+	 *   temporarily dropped to avoid lock order reversal or deadlock due to
+	 *   reentry.
+	 * - mallctl("epoch", ...) may concurrently refresh stats.  While
+	 *   strictly speaking this is a "concurrent operation", disallowing
+	 *   stats refreshes would impose an inconvenient burden.
+	 */
+
+	/* Remove large allocations from prof sample set. */
+	if (config_prof && opt_prof) {
+		ql_foreach(node, &arena->achunks, ql_link) {
+			arena_achunk_prof_reset(tsd, arena,
+			    extent_node_addr_get(node));
+		}
+	}
+
+	/* Huge allocations. */
+	malloc_mutex_lock(tsd, &arena->huge_mtx);
+	for (node = ql_last(&arena->huge, ql_link); node != NULL; node =
+	    ql_last(&arena->huge, ql_link)) {
+		void *ptr = extent_node_addr_get(node);
+
+		malloc_mutex_unlock(tsd, &arena->huge_mtx);
+		/* Remove huge allocation from prof sample set. */
+		if (config_prof && opt_prof) {
+			size_t usize;
+
+			usize = isalloc(tsd, ptr, config_prof);
+			prof_free(tsd, ptr, usize);
+		}
+		huge_dalloc(tsd, ptr);
+		malloc_mutex_lock(tsd, &arena->huge_mtx);
+	}
+	malloc_mutex_unlock(tsd, &arena->huge_mtx);
+
+	malloc_mutex_lock(tsd, &arena->lock);
+
+	/* Bins. */
+	for (i = 0; i < NBINS; i++) {
+		arena_bin_t *bin = &arena->bins[i];
+		malloc_mutex_lock(tsd, &bin->lock);
+		bin->runcur = NULL;
+		arena_run_heap_new(&bin->runs);
+		if (config_stats) {
+			bin->stats.curregs = 0;
+			bin->stats.curruns = 0;
+		}
+		malloc_mutex_unlock(tsd, &bin->lock);
+	}
+
+	/*
+	 * Re-initialize runs_dirty such that the chunks_cache and runs_dirty
+	 * chains directly correspond.
+	 */
+	qr_new(&arena->runs_dirty, rd_link);
+	for (node = qr_next(&arena->chunks_cache, cc_link);
+	    node != &arena->chunks_cache; node = qr_next(node, cc_link)) {
+		qr_new(&node->rd, rd_link);
+		qr_meld(&arena->runs_dirty, &node->rd, rd_link);
+	}
+
+	/* Arena chunks. */
+	for (node = ql_last(&arena->achunks, ql_link); node != NULL; node =
+	    ql_last(&arena->achunks, ql_link)) {
+		ql_remove(&arena->achunks, node, ql_link);
+		arena_chunk_discard(tsd, arena, extent_node_addr_get(node));
+	}
+
+	/* Spare. */
+	if (arena->spare != NULL) {
+		arena_chunk_discard(tsd, arena, arena->spare);
+		arena->spare = NULL;
+	}
+
+	assert(!arena->purging);
+	arena->nactive = 0;
+
+	for(i = 0; i < runs_avail_nclasses; i++)
+		arena_run_heap_new(&arena->runs_avail[i]);
+
+	malloc_mutex_unlock(tsd, &arena->lock);
+}
+
+static void
 arena_run_coalesce(arena_t *arena, arena_chunk_t *chunk, size_t *p_size,
     size_t *p_run_ind, size_t *p_run_pages, size_t flag_dirty,
     size_t flag_decommitted)
@@ -3373,6 +3522,8 @@ arena_new(tsd_t *tsd, unsigned ind)
 
 	arena->dss_prec = chunk_dss_prec_get(tsd);
 
+	ql_new(&arena->achunks);
+
 	arena->spare = NULL;
 
 	arena->lg_dirty_mult = arena_lg_dirty_mult_default_get();
diff --git a/src/ctl.c b/src/ctl.c
index fad2fdd..2e81143 100644
--- a/src/ctl.c
+++ b/src/ctl.c
@@ -120,6 +120,7 @@ CTL_PROTO(tcache_destroy)
 static void	arena_i_purge(tsd_t *tsd, unsigned arena_ind, bool all);
 CTL_PROTO(arena_i_purge)
 CTL_PROTO(arena_i_decay)
+CTL_PROTO(arena_i_reset)
 CTL_PROTO(arena_i_dss)
 CTL_PROTO(arena_i_lg_dirty_mult)
 CTL_PROTO(arena_i_decay_time)
@@ -299,6 +300,7 @@ static const ctl_named_node_t	tcache_node[] = {
 static const ctl_named_node_t arena_i_node[] = {
 	{NAME("purge"),		CTL(arena_i_purge)},
 	{NAME("decay"),		CTL(arena_i_decay)},
+	{NAME("reset"),		CTL(arena_i_reset)},
 	{NAME("dss"),		CTL(arena_i_dss)},
 	{NAME("lg_dirty_mult"),	CTL(arena_i_lg_dirty_mult)},
 	{NAME("decay_time"),	CTL(arena_i_decay_time)},
@@ -1603,6 +1605,40 @@ label_return:
 }
 
 static int
+arena_i_reset_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
+    size_t *oldlenp, void *newp, size_t newlen)
+{
+	int ret;
+	unsigned arena_ind;
+	arena_t *arena;
+
+	READONLY();
+	WRITEONLY();
+
+	if ((config_valgrind && unlikely(in_valgrind)) || (config_fill &&
+	    unlikely(opt_quarantine))) {
+		ret = EFAULT;
+		goto label_return;
+	}
+
+	arena_ind = (unsigned)mib[1];
+	if (config_debug) {
+		malloc_mutex_lock(tsd, &ctl_mtx);
+		assert(arena_ind < ctl_stats.narenas);
+		malloc_mutex_unlock(tsd, &ctl_mtx);
+	}
+	assert(arena_ind >= opt_narenas);
+
+	arena = arena_get(tsd, arena_ind, false);
+
+	arena_reset(tsd, arena);
+
+	ret = 0;
+label_return:
+	return (ret);
+}
+
+static int
 arena_i_dss_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
     size_t *oldlenp, void *newp, size_t newlen)
 {
diff --git a/test/unit/arena_reset.c b/test/unit/arena_reset.c
new file mode 100644
index 0000000..52170cc
--- /dev/null
+++ b/test/unit/arena_reset.c
@@ -0,0 +1,160 @@
+#include "test/jemalloc_test.h"
+
+#ifdef JEMALLOC_PROF
+const char *malloc_conf = "prof:true,lg_prof_sample:0";
+#endif
+
+static unsigned
+get_nsizes_impl(const char *cmd)
+{
+	unsigned ret;
+	size_t z;
+
+	z = sizeof(unsigned);
+	assert_d_eq(mallctl(cmd, &ret, &z, NULL, 0), 0,
+	    "Unexpected mallctl(\"%s\", ...) failure", cmd);
+
+	return (ret);
+}
+
+static unsigned
+get_nsmall(void)
+{
+
+	return (get_nsizes_impl("arenas.nbins"));
+}
+
+static unsigned
+get_nlarge(void)
+{
+
+	return (get_nsizes_impl("arenas.nlruns"));
+}
+
+static unsigned
+get_nhuge(void)
+{
+
+	return (get_nsizes_impl("arenas.nhchunks"));
+}
+
+static size_t
+get_size_impl(const char *cmd, size_t ind)
+{
+	size_t ret;
+	size_t z;
+	size_t mib[4];
+	size_t miblen = 4;
+
+	z = sizeof(size_t);
+	assert_d_eq(mallctlnametomib(cmd, mib, &miblen),
+	    0, "Unexpected mallctlnametomib(\"%s\", ...) failure", cmd);
+	mib[2] = ind;
+	z = sizeof(size_t);
+	assert_d_eq(mallctlbymib(mib, miblen, &ret, &z, NULL, 0),
+	    0, "Unexpected mallctlbymib([\"%s\", %zu], ...) failure", cmd, ind);
+
+	return (ret);
+}
+
+static size_t
+get_small_size(size_t ind)
+{
+
+	return (get_size_impl("arenas.bin.0.size", ind));
+}
+
+static size_t
+get_large_size(size_t ind)
+{
+
+	return (get_size_impl("arenas.lrun.0.size", ind));
+}
+
+static size_t
+get_huge_size(size_t ind)
+{
+
+	return (get_size_impl("arenas.hchunk.0.size", ind));
+}
+
+TEST_BEGIN(test_arena_reset)
+{
+#define	NHUGE	4
+	unsigned arena_ind, nsmall, nlarge, nhuge, nptrs, i;
+	size_t sz, miblen;
+	void **ptrs;
+	size_t mib[3];
+	tsd_t *tsd;
+
+	test_skip_if((config_valgrind && unlikely(in_valgrind)) || (config_fill
+	    && unlikely(opt_quarantine)));
+
+	sz = sizeof(unsigned);
+	assert_d_eq(mallctl("arenas.extend", &arena_ind, &sz, NULL, 0), 0,
+	    "Unexpected mallctl() failure");
+
+	nsmall = get_nsmall();
+	nlarge = get_nlarge();
+	nhuge = get_nhuge() > NHUGE ? NHUGE : get_nhuge();
+	nptrs = nsmall + nlarge + nhuge;
+	ptrs = (void **)malloc(nptrs * sizeof(void *));
+	assert_ptr_not_null(ptrs, "Unexpected malloc() failure");
+
+	/* Allocate objects with a wide range of sizes. */
+	for (i = 0; i < nsmall; i++) {
+		sz = get_small_size(i);
+		ptrs[i] = mallocx(sz, MALLOCX_ARENA(arena_ind));
+		assert_ptr_not_null(ptrs[i],
+		    "Unexpected mallocx(%zu, MALLOCX_ARENA(%u)) failure", sz,
+		    arena_ind);
+	}
+	for (i = 0; i < nlarge; i++) {
+		sz = get_large_size(i);
+		ptrs[nsmall + i] = mallocx(sz, MALLOCX_ARENA(arena_ind));
+		assert_ptr_not_null(ptrs[i],
+		    "Unexpected mallocx(%zu, MALLOCX_ARENA(%u)) failure", sz,
+		    arena_ind);
+	}
+	for (i = 0; i < nhuge; i++) {
+		sz = get_huge_size(i);
+		ptrs[nsmall + nlarge + i] = mallocx(sz,
+		    MALLOCX_ARENA(arena_ind));
+		assert_ptr_not_null(ptrs[i],
+		    "Unexpected mallocx(%zu, MALLOCX_ARENA(%u)) failure", sz,
+		    arena_ind);
+	}
+
+	tsd = tsd_fetch();
+
+	/* Verify allocations. */
+	for (i = 0; i < nptrs; i++) {
+		assert_zu_gt(ivsalloc(tsd, ptrs[i], false), 0,
+		    "Allocation should have queryable size");
+	}
+
+	/* Reset. */
+	miblen = sizeof(mib)/sizeof(size_t);
+	assert_d_eq(mallctlnametomib("arena.0.reset", mib, &miblen), 0,
+	    "Unexpected mallctlnametomib() failure");
+	mib[1] = (size_t)arena_ind;
+	assert_d_eq(mallctlbymib(mib, miblen, NULL, NULL, NULL, 0), 0,
+	    "Unexpected mallctlbymib() failure");
+
+	/* Verify allocations no longer exist. */
+	for (i = 0; i < nptrs; i++) {
+		assert_zu_eq(ivsalloc(tsd, ptrs[i], false), 0,
+		    "Allocation should no longer exist");
+	}
+
+	free(ptrs);
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test(
+	    test_arena_reset));
+}