From 6cd37713781a3092f8166b2178fae35cbfec1e28 Mon Sep 17 00:00:00 2001
From: Maria Matejka <mq@ucw.cz>
Date: Wed, 8 Sep 2021 11:29:49 +0200
Subject: [PATCH] Multipage allocation

We can also quite simply allocate bigger blocks. Anyway, we need these
blocks to be aligned to their size which needs one mmap() two times
bigger and then two munmap()s returning the unaligned parts.

The user can specify -B <N> on startup when <N> is the exponent of 2,
setting the block size to 2^N. On most systems, N is 12, anyway if you
know that your configuration is going to eat gigabytes of RAM, you are
almost forced to raise your block size as you may easily get into memory
fragmentation issues or you have to raise your maximum mapping count,
e.g. "sysctl vm.max_map_count=(number)".
---
 doc/bird.sgml       |  7 +++++++
 lib/resource.c      |  2 --
 lib/slab.c          |  2 +-
 sysdep/unix/alloc.c | 23 +++++++++++++++++++++++
 sysdep/unix/main.c  | 32 +++++++++++++++++++++++++++++++-
 5 files changed, 62 insertions(+), 4 deletions(-)

diff --git a/doc/bird.sgml b/doc/bird.sgml
index 39dadaf2..ddad4d98 100644
--- a/doc/bird.sgml
+++ b/doc/bird.sgml
@@ -145,6 +145,13 @@ BIRD executable by configuring out routing protocols you don't use, and
 <p>You can pass several command-line options to bird:
 
 <descrip>
+	<tag><label id="argv-block">-B <m/exp/</tag>
+	allocate memory using 2^<cf/exp/ byte sized blocks;
+	if you're expecting high memory load, raise this to
+        reduce number of allocated memory pages. For a million routes
+	in one table, the recommended setting is 18.
+	Default is your system page size, typically 12 for 4096 bytes.
+
 	<tag><label id="argv-config">-c <m/config name/</tag>
 	use given configuration file instead of <it/prefix/<file>/etc/bird.conf</file>.
 
diff --git a/lib/resource.c b/lib/resource.c
index 0ad886d9..e80b315b 100644
--- a/lib/resource.c
+++ b/lib/resource.c
@@ -61,7 +61,6 @@ pool root_pool;
 
 void *alloc_sys_page(void);
 void free_sys_page(void *);
-void resource_sys_init(void);
 
 static int indent;
 
@@ -283,7 +282,6 @@ rlookup(unsigned long a)
 void
 resource_init(void)
 {
-  resource_sys_init();
   root_pool.r.class = &pool_class;
   root_pool.name = "Root";
   init_list(&root_pool.inside);
diff --git a/lib/slab.c b/lib/slab.c
index 2fc88ee0..6348e29b 100644
--- a/lib/slab.c
+++ b/lib/slab.c
@@ -178,7 +178,7 @@ struct sl_alignment {			/* Magic structure for testing of alignment */
   int x[0];
 };
 
-#define SL_GET_HEAD(x)	((struct sl_head *) (((uintptr_t) (x)) & ~(page_size-1)))
+#define SL_GET_HEAD(x)	((struct sl_head *) PAGE_HEAD(x))
 
 /**
  * sl_new - create a new Slab
diff --git a/sysdep/unix/alloc.c b/sysdep/unix/alloc.c
index f6296afe..4c9d5eb5 100644
--- a/sysdep/unix/alloc.c
+++ b/sysdep/unix/alloc.c
@@ -17,6 +17,7 @@
 #endif
 
 long page_size = 0;
+_Bool alloc_multipage = 0;
 
 #ifdef HAVE_MMAP
 static _Bool use_fake = 0;
@@ -45,9 +46,31 @@ alloc_sys_page(void)
 #ifdef HAVE_MMAP
   if (!use_fake)
   {
+    if (alloc_multipage)
+    {
+      void *big = mmap(NULL, page_size * 2, PROT_WRITE | PROT_READ, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+      if (big == MAP_FAILED)
+	bug("mmap(%lu) failed: %m", page_size);
+
+      uintptr_t offset = ((uintptr_t) big) % page_size;
+      if (offset)
+      {
+	void *ret = big + page_size - offset;
+	munmap(big, page_size - offset);
+	munmap(ret + page_size, offset);
+	return ret;
+      }
+      else
+      {
+	munmap(big + page_size, page_size);
+	return big;
+      }
+    }
+
     void *ret = mmap(NULL, page_size, PROT_WRITE | PROT_READ, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
     if (ret == MAP_FAILED)
       bug("mmap(%lu) failed: %m", page_size);
+
     return ret;
   }
   else
diff --git a/sysdep/unix/main.c b/sysdep/unix/main.c
index cdf0a310..7e8ea0dc 100644
--- a/sysdep/unix/main.c
+++ b/sysdep/unix/main.c
@@ -682,7 +682,7 @@ signal_init(void)
  *	Parsing of command-line arguments
  */
 
-static char *opt_list = "bc:dD:ps:P:u:g:flRh";
+static char *opt_list = "B:c:dD:ps:P:u:g:flRh";
 int parse_and_exit;
 char *bird_name;
 static char *use_user;
@@ -703,6 +703,7 @@ display_help(void)
   fprintf(stderr,
     "\n"
     "Options: \n"
+    "  -B <block-size>	    Use 2^this number as memory allocation block size (default: 12)\n"
     "  -c <config-file>     Use given configuration file instead of\n"
     "                       "  PATH_CONFIG_FILE "\n"
     "  -d                   Enable debug messages and run bird in foreground\n"
@@ -789,12 +790,15 @@ get_gid(const char *s)
   return gr->gr_gid;
 }
 
+extern _Bool alloc_multipage;
+
 static void
 parse_args(int argc, char **argv)
 {
   int config_changed = 0;
   int socket_changed = 0;
   int c;
+  int bp;
 
   bird_name = get_bird_name(argv[0], "bird");
   if (argc == 2)
@@ -807,6 +811,29 @@ parse_args(int argc, char **argv)
   while ((c = getopt(argc, argv, opt_list)) >= 0)
     switch (c)
       {
+      case 'B':
+	bp = atoi(optarg);
+	if (bp < 1)
+	{
+	  fprintf(stderr, "Strange block size power %d\n\n", bp);
+	  display_usage();
+	  exit(1);
+	}
+
+	if ((1 << bp) < page_size)
+	{
+	  fprintf(stderr, "Requested block size %ld is lesser than page size %ld\n\n", (1L<<bp), page_size);
+	  display_usage();
+	  exit(1);
+	}
+
+	if ((1L << bp) > page_size)
+	{
+	  alloc_multipage = 1;
+	  page_size = (1L << bp);
+	}
+
+	break;
       case 'c':
 	config_name = optarg;
 	config_changed = 1;
@@ -861,6 +888,8 @@ parse_args(int argc, char **argv)
    }
 }
 
+void resource_sys_init(void);
+
 /*
  *	Hic Est main()
  */
@@ -873,6 +902,7 @@ main(int argc, char **argv)
     dmalloc_debug(0x2f03d00);
 #endif
 
+  resource_sys_init();
   parse_args(argc, argv);
   log_switch(1, NULL, NULL);