1 files changed, 151 insertions, 86 deletions

diff --git a/sysdep/unix/alloc.c b/sysdep/unix/alloc.c
index c8f1c83f..6c68a865 100644
--- a/sysdep/unix/alloc.c
+++ b/sysdep/unix/alloc.c
@@ -10,6 +10,7 @@
 #include "lib/resource.h"
 #include "lib/lists.h"
 #include "lib/event.h"
+#include "lib/rcu.h"
 
 #include <errno.h>
 #include <stdlib.h>
@@ -29,56 +30,55 @@
 long page_size = 0;
 
 #ifdef HAVE_MMAP
-#define KEEP_PAGES_MAIN_MAX	256
-#define KEEP_PAGES_MAIN_MIN	8
-#define CLEANUP_PAGES_BULK	256
+#define KEEP_PAGES_MAX	512
+#define KEEP_PAGES_MIN	32
+#define KEEP_PAGES_MAX_LOCAL	16
+#define ALLOC_PAGES_AT_ONCE	8
 
-STATIC_ASSERT(KEEP_PAGES_MAIN_MIN * 4 < KEEP_PAGES_MAIN_MAX);
+STATIC_ASSERT(KEEP_PAGES_MIN * 4 < KEEP_PAGES_MAX);
+STATIC_ASSERT(ALLOC_PAGES_AT_ONCE < KEEP_PAGES_MAX_LOCAL);
 
 static _Bool use_fake = 0;
+static _Bool initialized = 0;
 
 #if DEBUGGING
 struct free_page {
   node unused[42];
-  node n;
+  struct free_page * _Atomic next;
 };
 #else
 struct free_page {
-  node n;
+  struct free_page * _Atomic next;
 };
 #endif
 
 #define EP_POS_MAX	((page_size - OFFSETOF(struct empty_pages, pages)) / sizeof (void *))
 
 struct empty_pages {
-  node n;
+  struct empty_pages *next;
   uint pos;
   void *pages[0];
 };
 
-struct free_pages {
-  list pages;		/* List of (struct free_page) keeping free pages without releasing them (hot) */
-  list empty;		/* List of (struct empty_pages) keeping invalidated pages mapped for us (cold) */
-  u16 min, max;		/* Minimal and maximal number of free pages kept */
-  uint cnt;		/* Number of free pages in list */
-  event cleanup;
-};
+DEFINE_DOMAIN(resource);
+static DOMAIN(resource) empty_pages_domain;
+static struct empty_pages *empty_pages = NULL;
 
-static void global_free_pages_cleanup_event(void *);
-static void *alloc_cold_page(void);
+static struct free_page * _Atomic page_stack = NULL;
+static _Thread_local struct free_page * local_page_stack = NULL;
 
-static struct free_pages global_free_pages = {
-  .min = KEEP_PAGES_MAIN_MIN,
-  .max = KEEP_PAGES_MAIN_MAX,
-  .cleanup = { .hook = global_free_pages_cleanup_event },
-};
+static void page_cleanup(void *);
+static event page_cleanup_event = { .hook = page_cleanup, };
+#define SCHEDULE_CLEANUP  do if (initialized && !shutting_down) ev_send(&global_event_list, &page_cleanup_event); while (0)
 
-uint *pages_kept = &global_free_pages.cnt;
+_Atomic int pages_kept = 0;
+_Atomic int pages_kept_locally = 0;
+static int pages_kept_here = 0;
 
 static void *
 alloc_sys_page(void)
 {
-  void *ptr = mmap(NULL, page_size, PROT_WRITE | PROT_READ, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+  void *ptr = mmap(NULL, page_size * ALLOC_PAGES_AT_ONCE, PROT_WRITE | PROT_READ, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
 
   if (ptr == MAP_FAILED)
     die("mmap(%ld) failed: %m", (s64) page_size);
@@ -108,45 +108,55 @@ alloc_page(void)
   }
 
 #ifdef HAVE_MMAP
-  struct free_pages *fps = &global_free_pages;
-
-  /* If there is any free page kept hot, we use it. */
-  if (fps->cnt)
+  /* If there is any free page kept hot in this thread, we use it. */
+  struct free_page *fp = local_page_stack;
+  if (fp)
   {
-    struct free_page *fp = SKIP_BACK(struct free_page, n, HEAD(fps->pages));
-    rem_node(&fp->n);
+    local_page_stack = atomic_load_explicit(&fp->next, memory_order_acquire);
+    atomic_fetch_sub_explicit(&pages_kept_locally, 1, memory_order_relaxed);
+    pages_kept_here--;
+    return fp;
+  }
 
-    /* If the hot-free-page cache is getting short, request the cleanup routine to replenish the cache */
-    if ((--fps->cnt < fps->min) && !shutting_down)
-      ev_schedule(&fps->cleanup);
+  /* If there is any free page kept hot in global storage, we use it. */
+  rcu_read_lock();
+  fp = atomic_load_explicit(&page_stack, memory_order_acquire);
+  while (fp && !atomic_compare_exchange_strong_explicit(
+	&page_stack, &fp, atomic_load_explicit(&fp->next, memory_order_acquire),
+	memory_order_acq_rel, memory_order_acquire))
+    ;
+  rcu_read_unlock();
 
+  if (fp)
+  {
+    atomic_fetch_sub_explicit(&pages_kept, 1, memory_order_relaxed);
     return fp;
   }
-  else
-    return alloc_cold_page();
-}
-
-static void *
-alloc_cold_page(void)
-{
-  struct free_pages *fps = &global_free_pages;
 
   /* If there is any free page kept cold, we use that. */
-  if (!EMPTY_LIST(fps->empty))
-  {
-    struct empty_pages *ep = HEAD(fps->empty);
+  LOCK_DOMAIN(resource, empty_pages_domain);
+  if (empty_pages) {
+    if (empty_pages->pos)
+      /* Either the keeper page contains at least one cold page pointer, return that */
+      fp = empty_pages->pages[--empty_pages->pos];
+    else
+    {
+      /* Or the keeper page has no more cold page pointer, return the keeper page */
+      fp = (struct free_page *) empty_pages;
+      empty_pages = empty_pages->next;
+    }
+  }
+  UNLOCK_DOMAIN(resource, empty_pages_domain);
 
-    /* Either the keeper page contains at least one cold page pointer, return that */
-    if (ep->pos)
-      return ep->pages[--ep->pos];
+  if (fp)
+    return fp;
 
-    /* Or the keeper page has no more cold page pointer, return the keeper page */
-    rem_node(&ep->n);
-    return ep;
-  }
+  /* And in the worst case, allocate some new pages by mmap() */
+  void *ptr = alloc_sys_page();
+  for (int i=1; i<ALLOC_PAGES_AT_ONCE; i++)
+    free_page(ptr + page_size * i);
 
-  /* And in the worst case, allocate a new page by mmap() */
-  return alloc_sys_page();
+  return ptr;
 #endif
 }
 
@@ -161,53 +171,106 @@ free_page(void *ptr)
   }
 
 #ifdef HAVE_MMAP
-  struct free_pages *fps = &global_free_pages;
+  /* We primarily try to keep the pages locally. */
   struct free_page *fp = ptr;
+  if (shutting_down || (pages_kept_here < KEEP_PAGES_MAX_LOCAL))
+  {
+    atomic_store_explicit(&fp->next, local_page_stack, memory_order_relaxed);
+    atomic_fetch_add_explicit(&pages_kept_locally, 1, memory_order_relaxed);
+    pages_kept_here++;
+    return;
+  }
+
+  /* If there are too many local pages, we add the free page to the global hot-free-page list */
+  rcu_read_lock();
+  struct free_page *next = atomic_load_explicit(&page_stack, memory_order_acquire);
 
-  /* Otherwise, we add the free page to the hot-free-page list */
-  fp->n = (node) {};
-  add_tail(&fps->pages, &fp->n);
+  do atomic_store_explicit(&fp->next, next, memory_order_release);
+  while (!atomic_compare_exchange_strong_explicit(
+	&page_stack, &next, fp,
+	memory_order_acq_rel, memory_order_acquire));
+  rcu_read_unlock();
 
-  /* And if there are too many hot free pages, we ask for page cleanup */
-  if ((++fps->cnt > fps->max) && !shutting_down)
-    ev_schedule(&fps->cleanup);
+  /* And if there are too many global hot free pages, we ask for page cleanup */
+  if (atomic_fetch_add_explicit(&pages_kept, 1, memory_order_relaxed) >= KEEP_PAGES_MAX)
+    SCHEDULE_CLEANUP;
 #endif
 }
 
+/* When the routine is going to sleep for a long time, we flush the local
+ * hot page cache to not keep dirty pages for nothing. */
+void
+flush_local_pages(void)
+{
+  if (use_fake || !local_page_stack || shutting_down)
+    return;
+
+  /* We first count the pages to enable consistency checking.
+   * Also, we need to know the last page. */
+  struct free_page *last = local_page_stack, *next;
+  int check_count = 1;
+  while (next = atomic_load_explicit(&last->next, memory_order_acquire))
+  {
+    check_count++;
+    last = next;
+  }
+
+  /* The actual number of pages must be equal to the counter value. */
+  ASSERT_DIE(check_count == pages_kept_here);
+
+  /* Repeatedly trying to insert the whole page list into global page stack at once. */
+  rcu_read_lock();
+  next = atomic_load_explicit(&page_stack, memory_order_acquire);
+
+  /* First we set the outwards pointer (from our last),
+   * then we try to set the inwards pointer to our first page. */
+  do atomic_store_explicit(&last->next, next, memory_order_release);
+  while (!atomic_compare_exchange_strong_explicit(
+	&page_stack, &next, local_page_stack,
+	memory_order_acq_rel, memory_order_acquire));
+  rcu_read_unlock();
+
+  /* Finished. Now the local stack is empty. */
+  local_page_stack = NULL;
+  pages_kept_here = 0;
+
+  /* Check the state of global page cache and maybe schedule its cleanup. */
+  atomic_fetch_sub_explicit(&pages_kept_locally, check_count, memory_order_relaxed);
+  if (atomic_fetch_add_explicit(&pages_kept, check_count, memory_order_relaxed) >= KEEP_PAGES_MAX)
+    SCHEDULE_CLEANUP;
+}
+
 #ifdef HAVE_MMAP
 static void
-global_free_pages_cleanup_event(void *data UNUSED)
+page_cleanup(void *_ UNUSED)
 {
   /* Cleanup on shutdown is ignored. All pages may be kept hot, OS will take care. */
   if (shutting_down)
     return;
 
-  struct free_pages *fps = &global_free_pages;
+  struct free_page *stack = atomic_exchange_explicit(&page_stack, NULL, memory_order_acq_rel);
+  if (!stack)
+    return;
 
-  /* Cleanup may get called when hot free page cache is short of pages. Replenishing. */
-  while (fps->cnt / 2 < fps->min)
-    free_page(alloc_cold_page());
 
-  /* Or the hot free page cache is too big. Moving some pages to the cold free page cache. */
-  for (int limit = CLEANUP_PAGES_BULK; limit && (fps->cnt > fps->max / 2); fps->cnt--, limit--)
-  {
-    struct free_page *fp = SKIP_BACK(struct free_page, n, TAIL(fps->pages));
-    rem_node(&fp->n);
+  do {
+    synchronize_rcu();
+    struct free_page *fp = stack;
+    stack = atomic_load_explicit(&fp->next, memory_order_acquire);
 
+    LOCK_DOMAIN(resource, empty_pages_domain);
     /* Empty pages are stored as pointers. To store them, we need a pointer block. */
-    struct empty_pages *ep;
-    if (EMPTY_LIST(fps->empty) || ((ep = HEAD(fps->empty))->pos == EP_POS_MAX))
+    if (!empty_pages || (empty_pages->pos == EP_POS_MAX))
     {
       /* There is either no pointer block or the last block is full. We use this block as a pointer block. */
-      ep = (struct empty_pages *) fp;
-      *ep = (struct empty_pages) {};
-      add_head(&fps->empty, &ep->n);
+      empty_pages = (struct empty_pages *) fp;
+      *empty_pages = (struct empty_pages) {};
     }
     else
     {
       /* We store this block as a pointer into the first free place
        * and tell the OS that the underlying memory is trash. */
-      ep->pages[ep->pos++] = fp;
+      empty_pages->pages[empty_pages->pos++] = fp;
       if (madvise(fp, page_size,
 #ifdef CONFIG_MADV_DONTNEED_TO_FREE
 	    MADV_DONTNEED
@@ -217,12 +280,18 @@ global_free_pages_cleanup_event(void *data UNUSED)
 	    ) < 0)
 	bug("madvise(%p) failed: %m", fp);
     }
+    UNLOCK_DOMAIN(resource, empty_pages_domain);
   }
+  while ((atomic_fetch_sub_explicit(&pages_kept, 1, memory_order_relaxed) >= KEEP_PAGES_MAX / 2) && stack);
 
-  /* If the hot free page cleanup hit the limit, re-schedule this routine
-   * to allow for other routines to run. */
-  if (fps->cnt > fps->max)
-    ev_schedule(&fps->cleanup);
+  while (stack)
+  {
+    struct free_page *f = stack;
+    stack = atomic_load_explicit(&f->next, memory_order_acquire);
+    free_page(f);
+
+    atomic_fetch_sub_explicit(&pages_kept, 1, memory_order_relaxed);
+  }
 }
 #endif
 
@@ -236,8 +305,6 @@ resource_sys_init(void)
 #endif
 
 #ifdef HAVE_MMAP
-  ASSERT_DIE(global_free_pages.cnt == 0);
-
   /* Check what page size the system supports */
   if (!(page_size = sysconf(_SC_PAGESIZE)))
     die("System page size must be non-zero");
@@ -247,11 +314,8 @@ resource_sys_init(void)
     /* We assume that page size has only one bit and is between 1K and 256K (incl.).
      * Otherwise, the assumptions in lib/slab.c (sl_head's num_full range) aren't met. */
 
-    struct free_pages *fps = &global_free_pages;
-
-    init_list(&fps->pages);
-    init_list(&fps->empty);
-    global_free_pages_cleanup_event(NULL);
+    empty_pages_domain = DOMAIN_NEW(resource, "Empty Pages");
+    initialized = 1;
     return;
   }
 
@@ -261,4 +325,5 @@ resource_sys_init(void)
 #endif
 
   page_size = 4096;
+  initialized = 1;
 }