summaryrefslogtreecommitdiffstats
path: root/src/vnet/ip/ip4.h
blob: e969594ec00f520a0546ba25ba54214080b8edff (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
/*
 * Copyright (c) 2015 Cisco and/or its affiliates.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at:
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
/*
 * ip/ip4.h: ip4 main include file
 *
 * Copyright (c) 2008 Eliot Dresselhaus
 *
 * Permission is hereby granted, free of charge, to any person obtaining
 * a copy of this software and associated documentation files (the
 * "Software"), to deal in the Software without restriction, including
 * without limitation the rights to use, copy, modify, merge, publish,
 * distribute, sublicense, and/or sell copies of the Software, and to
 * permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 *
 * The above copyright notice and this permission notice shall be
 * included in all copies or substantial portions of the Software.
 *
 *  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 *  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 *  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 *  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
 *  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
 *  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
 *  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 */

#ifndef included_ip_ip4_h
#define included_ip_ip4_h

#include <vnet/ip/ip4_packet.h>
#include <vnet/ip/ip_flow_hash.h>

#include <vnet/ip/lookup.h>
#include <vnet/ip/ip_interface.h>
#include <vnet/buffer.h>
#include <vnet/feature/feature.h>
#include <vnet/ip/icmp46_packet.h>

typedef struct ip4_mfib_t
{
  /* Hash table for each prefix length mapping. */
  uword *fib_entry_by_dst_address[65];

  /* Table ID (hash key) for this FIB. */
  u32 table_id;

  /* Index into FIB vector. */
  u32 index;
} ip4_mfib_t;

struct ip4_main_t;

typedef void (ip4_add_del_interface_address_function_t)
  (struct ip4_main_t * im,
   uword opaque,
   u32 sw_if_index,
   ip4_address_t * address,
   u32 address_length, u32 if_address_index, u32 is_del);

typedef struct
{
  ip4_add_del_interface_address_function_t *function;
  uword function_opaque;
} ip4_add_del_interface_address_callback_t;

typedef void (ip4_enable_disable_interface_function_t)
  (struct ip4_main_t * im, uword opaque, u32 sw_if_index, u32 is_enable);

typedef struct
{
  ip4_enable_disable_interface_function_t *function;
  uword function_opaque;
} ip4_enable_disable_interface_callback_t;

typedef void (ip4_table_bind_function_t)
  (struct ip4_main_t * im,
   uword opaque, u32 sw_if_index, u32 new_fib_index, u32 old_fib_index);

typedef struct
{
  ip4_table_bind_function_t *function;
  uword function_opaque;
} ip4_table_bind_callback_t;

/**
 * @brief IPv4 main type.
 *
 * State of IPv4 VPP processing including:
 * - FIBs
 * - Feature indices used in feature topological sort
 * - Feature node run time references
 */

typedef struct ip4_main_t
{
  ip_lookup_main_t lookup_main;

  /** Vector of FIBs. */
  struct fib_table_t_ *fibs;

  /** Vector of MFIBs. */
  struct mfib_table_t_ *mfibs;

  u32 fib_masks[33];

  /** Table index indexed by software interface. */
  u32 *fib_index_by_sw_if_index;

  /** Table index indexed by software interface. */
  u32 *mfib_index_by_sw_if_index;

  /* IP4 enabled count by software interface */
  u8 *ip_enabled_by_sw_if_index;

  /** Hash table mapping table id to fib index.
     ID space is not necessarily dense; index space is dense. */
  uword *fib_index_by_table_id;

  /** Hash table mapping table id to multicast fib index.
     ID space is not necessarily dense; index space is dense. */
  uword *mfib_index_by_table_id;

  /** Functions to call when interface address changes. */
    ip4_add_del_interface_address_callback_t
    * add_del_interface_address_callbacks;

  /** Functions to call when interface becomes IPv4 enabled/disable. */
    ip4_enable_disable_interface_callback_t
    * enable_disable_interface_callbacks;

  /** Functions to call when interface to table biding changes. */
  ip4_table_bind_callback_t *table_bind_callbacks;

  /** Template used to generate IP4 ARP packets. */
  vlib_packet_template_t ip4_arp_request_packet_template;

  /** Seed for Jenkins hash used to compute ip4 flow hash. */
  u32 flow_hash_seed;

  /** @brief Template information for VPP generated packets */
  struct
  {
    /** TTL to use for host generated packets. */
    u8 ttl;

    /** TOS byte to use for host generated packets. */
    u8 tos;

    u8 pad[2];
  } host_config;

  u16 msg_id_base;
} ip4_main_t;

#define ARP_THROTTLE_BITS	(512)

/** Global ip4 main structure. */
extern ip4_main_t ip4_main;

/** Global ip4 input node.  Errors get attached to ip4 input node. */
extern vlib_node_registration_t ip4_input_node;
extern vlib_node_registration_t ip4_lookup_node;
extern vlib_node_registration_t ip4_local_node;
extern vlib_node_registration_t ip4_rewrite_node;
extern vlib_node_registration_t ip4_rewrite_mcast_node;
extern vlib_node_registration_t ip4_rewrite_local_node;
extern vlib_node_registration_t ip4_arp_node;
extern vlib_node_registration_t ip4_glean_node;
extern vlib_node_registration_t ip4_midchain_node;
extern vlib_node_registration_t ip4_punt_node;

always_inline uword
ip4_destination_matches_route (const ip4_main_t * im,
			       const ip4_address_t * key,
			       const ip4_address_t * dest, uword dest_length)
{
  return 0 == ((key->data_u32 ^ dest->data_u32) & im->fib_masks[dest_length]);
}

always_inline uword
ip4_destination_matches_interface (ip4_main_t * im,
				   ip4_address_t * key,
				   ip_interface_address_t * ia)
{
  ip4_address_t *a = ip_interface_address_get_address (&im->lookup_main, ia);
  return ip4_destination_matches_route (im, key, a, ia->address_length);
}

/* Find interface address which matches destination. */
always_inline ip4_address_t *
ip4_interface_address_matching_destination (ip4_main_t * im,
					    const ip4_address_t * dst,
					    u32 sw_if_index,
					    ip_interface_address_t **
					    result_ia)
{
  ip_lookup_main_t *lm = &im->lookup_main;
  ip_interface_address_t *ia;
  ip4_address_t *result = 0;

  /* *INDENT-OFF* */
  foreach_ip_interface_address (lm, ia, sw_if_index,
                                1 /* honor unnumbered */,
  ({
    ip4_address_t * a = ip_interface_address_get_address (lm, ia);
    if (ip4_destination_matches_route (im, dst, a, ia->address_length))
      {
	result = a;
	break;
      }
  }));
  /* *INDENT-ON* */
  if (result_ia)
    *result_ia = result ? ia : 0;
  return result;
}

ip4_address_t *ip4_interface_first_address (ip4_main_t * im, u32 sw_if_index,
					    ip_interface_address_t **
					    result_ia);

clib_error_t *ip4_add_del_interface_address (vlib_main_t * vm,
					     u32 sw_if_index,
					     ip4_address_t * address,
					     u32 address_length, u32 is_del);

void ip4_directed_broadcast (u32 sw_if_index, u8 enable);

void ip4_sw_interface_enable_disable (u32 sw_if_index, u32 is_enable);

int ip4_address_compare (ip4_address_t * a1, ip4_address_t * a2);

uword
ip4_udp_register_listener (vlib_main_t * vm,
			   u16 dst_port, u32 next_node_index);

u16 ip4_tcp_udp_compute_checksum (vlib_main_t * vm, vlib_buffer_t * p0,
				  ip4_header_t * ip0);

void ip4_register_protocol (u32 protocol, u32 node_index);
void ip4_unregister_protocol (u32 protocolx);

serialize_function_t serialize_vnet_ip4_main, unserialize_vnet_ip4_main;

int vnet_set_ip4_classify_intfc (vlib_main_t * vm, u32 sw_if_index,
				 u32 table_index);

void ip4_punt_policer_add_del (u8 is_add, u32 policer_index);

void ip4_punt_redirect_add_paths (u32 rx_sw_if_index,
				  const fib_route_path_t *paths);

void ip4_punt_redirect_del (u32 rx_sw_if_index);


void
ip4_forward_next_trace (vlib_main_t * vm,
			vlib_node_runtime_t * node,
			vlib_frame_t * frame,
			vlib_rx_or_tx_t which_adj_index);

u8 *format_ip4_forward_next_trace (u8 * s, va_list * args);

u32 ip4_tcp_udp_validate_checksum (vlib_main_t * vm, vlib_buffer_t * p0);

always_inline u32
vlib_buffer_get_ip4_fib_index (vlib_buffer_t * b)
{
  u32 fib_index, sw_if_index;
  sw_if_index = vnet_buffer (b)->sw_if_index[VLIB_RX];
  fib_index = vnet_buffer (b)->sw_if_index[VLIB_TX];
  return (fib_index == (u32) ~ 0) ?
    vec_elt (ip4_main.fib_index_by_sw_if_index, sw_if_index) : fib_index;
}

#endif /* included_ip_ip4_h */

/*
 * fd.io coding-style-patch-verification: ON
 *
 * Local Variables:
 * eval: (c-set-style "gnu")
 * End:
 */
light .kt { color: #888888; font-weight: bold } /* Keyword.Type */ .highlight .m { color: #0000DD; font-weight: bold } /* Literal.Number */ .highlight .s { color: #dd2200; background-color: #fff0f0 } /* Literal.String */ .highlight .na { color: #336699 } /* Name.Attribute */ .highlight .nb { color: #003388 } /* Name.Builtin */ .highlight .nc { color: #bb0066; font-weight: bold } /* Name.Class */ .highlight .no { color: #003366; font-weight: bold } /* Name.Constant */ .highlight .nd { color: #555555 } /* Name.Decorator */ .highlight .ne { color: #bb0066; font-weight: bold } /* Name.Exception */ .highlight .nf { color: #0066bb; font-weight: bold } /* Name.Function */ .highlight .nl { color: #336699; font-style: italic } /* Name.Label */ .highlight .nn { color: #bb0066; font-weight: bold } /* Name.Namespace */ .highlight .py { color: #336699; font-weight: bold } /* Name.Property */ .highlight .nt { color: #bb0066; font-weight: bold } /* Name.Tag */ .highlight .nv { color: #336699 } /* Name.Variable */ .highlight .ow { color: #008800 } /* Operator.Word */ .highlight .w { color: #bbbbbb } /* Text.Whitespace */ .highlight .mb { color: #0000DD; font-weight: bold } /* Literal.Number.Bin */ .highlight .mf { color: #0000DD; font-weight: bold } /* Literal.Number.Float */ .highlight .mh { color: #0000DD; font-weight: bold } /* Literal.Number.Hex */ .highlight .mi { color: #0000DD; font-weight: bold } /* Literal.Number.Integer */ .highlight .mo { color: #0000DD; font-weight: bold } /* Literal.Number.Oct */ .highlight .sa { color: #dd2200; background-color: #fff0f0 } /* Literal.String.Affix */ .highlight .sb { color: #dd2200; background-color: #fff0f0 } /* Literal.String.Backtick */ .highlight .sc { color: #dd2200; background-color: #fff0f0 } /* Literal.String.Char */ .highlight .dl { color: #dd2200; background-color: #fff0f0 } /* Literal.String.Delimiter */ .highlight .sd { color: #dd2200; background-color: #fff0f0 } /* Literal.String.Doc */ .highlight .s2 { color: #dd2200; background-color: #fff0f0 } /* Literal.String.Double */ .highlight .se { color: #0044dd; background-color: #fff0f0 } /* Literal.String.Escape */ .highlight .sh { color: #dd2200; background-color: #fff0f0 } /* Literal.String.Heredoc */ .highlight .si { color: #3333bb; background-color: #fff0f0 } /* Literal.String.Interpol */ .highlight .sx { color: #22bb22; background-color: #f0fff0 } /* Literal.String.Other */ .highlight .sr { color: #008800; background-color: #fff0ff } /* Literal.String.Regex */ .highlight .s1 { color: #dd2200; background-color: #fff0f0 } /* Literal.String.Single */ .highlight .ss { color: #aa6600; background-color: #fff0f0 } /* Literal.String.Symbol */ .highlight .bp { color: #003388 } /* Name.Builtin.Pseudo */ .highlight .fm { color: #0066bb; font-weight: bold } /* Name.Function.Magic */ .highlight .vc { color: #336699 } /* Name.Variable.Class */ .highlight .vg { color: #dd7700 } /* Name.Variable.Global */ .highlight .vi { color: #3333bb } /* Name.Variable.Instance */ .highlight .vm { color: #336699 } /* Name.Variable.Magic */ .highlight .il { color: #0000DD; font-weight: bold } /* Literal.Number.Integer.Long */ }
/*
 *------------------------------------------------------------------
 * svm.c - shared VM allocation, mmap(...MAP_FIXED...)
 * library
 *
 * Copyright (c) 2009 Cisco and/or its affiliates.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at:
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 *------------------------------------------------------------------
 */

#include <stdio.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <netinet/in.h>
#include <signal.h>
#include <pthread.h>
#include <unistd.h>
#include <time.h>
#include <fcntl.h>
#include <string.h>
#include <vppinfra/clib.h>
#include <vppinfra/vec.h>
#include <vppinfra/hash.h>
#include <vppinfra/bitmap.h>
#include <vppinfra/fifo.h>
#include <vppinfra/time.h>
#include <vppinfra/mheap.h>
#include <vppinfra/heap.h>
#include <vppinfra/pool.h>
#include <vppinfra/format.h>

#include "svm.h"

static svm_region_t *root_rp;
static int root_rp_refcount;

#define MAXLOCK 2
static pthread_mutex_t *mutexes_held[MAXLOCK];
static int nheld;

svm_region_t *
svm_get_root_rp (void)
{
  return root_rp;
}

#define MUTEX_DEBUG

static void
region_lock (svm_region_t * rp, int tag)
{
  pthread_mutex_lock (&rp->mutex);
#ifdef MUTEX_DEBUG
  rp->mutex_owner_pid = getpid ();
  rp->mutex_owner_tag = tag;
#endif
  ASSERT (nheld < MAXLOCK);
  /*
   * Keep score of held mutexes so we can try to exit
   * cleanly if the world comes to an end at the worst possible
   * moment
   */
  mutexes_held[nheld++] = &rp->mutex;
}

static void
region_unlock (svm_region_t * rp)
{
  int i, j;
#ifdef MUTEX_DEBUG
  rp->mutex_owner_pid = 0;
  rp->mutex_owner_tag = 0;
#endif

  for (i = nheld - 1; i >= 0; i--)
    {
      if (mutexes_held[i] == &rp->mutex)
	{
	  for (j = i; j < MAXLOCK - 1; j++)
	    mutexes_held[j] = mutexes_held[j + 1];
	  nheld--;
	  goto found;
	}
    }
  ASSERT (0);

found:
  CLIB_MEMORY_BARRIER ();
  pthread_mutex_unlock (&rp->mutex);
}


static u8 *
format_svm_flags (u8 * s, va_list * args)
{
  uword f = va_arg (*args, uword);

  if (f & SVM_FLAGS_MHEAP)
    s = format (s, "MHEAP ");
  if (f & SVM_FLAGS_FILE)
    s = format (s, "FILE ");
  if (f & SVM_FLAGS_NODATA)
    s = format (s, "NODATA ");
  if (f & SVM_FLAGS_NEED_DATA_INIT)
    s = format (s, "INIT ");

  return (s);
}

static u8 *
format_svm_size (u8 * s, va_list * args)
{
  uword size = va_arg (*args, uword);

  if (size >= (1 << 20))
    {
      s = format (s, "(%d mb)", size >> 20);
    }
  else if (size >= (1 << 10))
    {
      s = format (s, "(%d kb)", size >> 10);
    }
  else
    {
      s = format (s, "(%d bytes)", size);
    }
  return (s);
}

u8 *
format_svm_region (u8 * s, va_list * args)
{
  svm_region_t *rp = va_arg (*args, svm_region_t *);
  int verbose = va_arg (*args, int);
  int i;
  uword lo, hi;

  s = format (s, "%s: base va 0x%x size 0x%x %U\n",
	      rp->region_name, rp->virtual_base,
	      rp->virtual_size, format_svm_size, rp->virtual_size);
  s = format (s, "  user_ctx 0x%x, bitmap_size %d\n",
	      rp->user_ctx, rp->bitmap_size);

  if (verbose)
    {
      s = format (s, "  flags: 0x%x %U\n", rp->flags,
		  format_svm_flags, rp->flags);
      s = format (s,
		  "  region_heap 0x%x data_base 0x%x data_heap 0x%x\n",
		  rp->region_heap, rp->data_base, rp->data_heap);
    }

  s = format (s, "  %d clients, pids: ", vec_len (rp->client_pids));

  for (i = 0; i < vec_len (rp->client_pids); i++)
    s = format (s, "%d ", rp->client_pids[i]);

  s = format (s, "\n");

  if (verbose)
    {
      lo = hi = ~0;

      s = format (s, "  VM in use: ");

      for (i = 0; i < rp->bitmap_size; i++)
	{
	  if (clib_bitmap_get_no_check (rp->bitmap, i) != 0)
	    {
	      if (lo == ~0)
		{
		  hi = lo = rp->virtual_base + i * MMAP_PAGESIZE;
		}
	      else
		{
		  hi = rp->virtual_base + i * MMAP_PAGESIZE;
		}
	    }
	  else
	    {
	      if (lo != ~0)
		{
		  hi = rp->virtual_base + i * MMAP_PAGESIZE - 1;
		  s = format (s, "   0x%x - 0x%x (%dk)\n", lo, hi,
			      (hi - lo) >> 10);
		  lo = hi = ~0;
		}
	    }
	}
      s = format (s, "  rgn heap stats: %U", format_mheap,
		  rp->region_heap, 0);
      if ((rp->flags & SVM_FLAGS_MHEAP) && rp->data_heap)
	{
	  s = format (s, "\n  data heap stats: %U", format_mheap,
		      rp->data_heap, 1);
	}
      s = format (s, "\n");
    }

  return (s);
}

/*
 * rnd_pagesize
 * Round to a pagesize multiple, presumably 4k works
 */
static u64
rnd_pagesize (u64 size)
{
  u64 rv;

  rv = (size + (MMAP_PAGESIZE - 1)) & ~(MMAP_PAGESIZE - 1);
  return (rv);
}

/*
 * svm_data_region_setup
 */
static int
svm_data_region_create (svm_map_region_args_t * a, svm_region_t * rp)
{
  int fd;
  u8 junk = 0;
  uword map_size;

  map_size = rp->virtual_size - (MMAP_PAGESIZE +
				 (a->pvt_heap_size ? a->pvt_heap_size :
				  SVM_PVT_MHEAP_SIZE));

  if (a->flags & SVM_FLAGS_FILE)
    {
      struct stat statb;

      fd = open (a->backing_file, O_RDWR | O_CREAT, 0777);

      if (fd < 0)
	{
	  clib_unix_warning ("open");
	  return -1;
	}

      if (fstat (fd, &statb) < 0)
	{
	  clib_unix_warning ("fstat");
	  close (fd);
	  return -2;
	}

      if (statb.st_mode & S_IFREG)
	{
	  if (statb.st_size == 0)
	    {
	      if (lseek (fd, map_size, SEEK_SET) == (off_t) - 1)
		{
		  clib_unix_warning ("seek region size");
		  close (fd);
		  return -3;
		}
	      if (write (fd, &junk, 1) != 1)
		{
		  clib_unix_warning ("set region size");
		  close (fd);
		  return -3;
		}
	    }
	  else
	    {
	      map_size = rnd_pagesize (statb.st_size);
	    }
	}
      else
	{
	  map_size = a->backing_mmap_size;
	}

      ASSERT (map_size <= rp->virtual_size -
	      (MMAP_PAGESIZE + SVM_PVT_MHEAP_SIZE));

      if (mmap (rp->data_base, map_size, PROT_READ | PROT_WRITE,
		MAP_SHARED | MAP_FIXED, fd, 0) == MAP_FAILED)
	{
	  clib_unix_warning ("mmap");
	  close (fd);
	  return -3;
	}
      close (fd);
      rp->backing_file = (char *) format (0, "%s\0", a->backing_file);
      rp->flags |= SVM_FLAGS_FILE;
    }

  if (a->flags & SVM_FLAGS_MHEAP)
    {
      rp->data_heap =
	mheap_alloc_with_flags ((void *) (rp->data_base), map_size,
				MHEAP_FLAG_DISABLE_VM);
      rp->flags |= SVM_FLAGS_MHEAP;
    }
  return 0;
}

static int
svm_data_region_map (svm_map_region_args_t * a, svm_region_t * rp)
{
  int fd;
  u8 junk = 0;
  uword map_size;
  struct stat statb;

  map_size = rp->virtual_size -
    (MMAP_PAGESIZE
     + (a->pvt_heap_size ? a->pvt_heap_size : SVM_PVT_MHEAP_SIZE));

  if (a->flags & SVM_FLAGS_FILE)
    {

      fd = open (a->backing_file, O_RDWR, 0777);

      if (fd < 0)
	{
	  clib_unix_warning ("open");
	  return -1;
	}

      if (fstat (fd, &statb) < 0)
	{
	  clib_unix_warning ("fstat");
	  close (fd);
	  return -2;
	}

      if (statb.st_mode & S_IFREG)
	{
	  if (statb.st_size == 0)
	    {
	      if (lseek (fd, map_size, SEEK_SET) == (off_t) - 1)
		{
		  clib_unix_warning ("seek region size");
		  close (fd);
		  return -3;
		}
	      if (write (fd, &junk, 1) != 1)
		{
		  clib_unix_warning ("set region size");
		  close (fd);
		  return -3;
		}
	    }
	  else
	    {
	      map_size = rnd_pagesize (statb.st_size);
	    }
	}
      else
	{
	  map_size = a->backing_mmap_size;
	}

      ASSERT (map_size <= rp->virtual_size
	      - (MMAP_PAGESIZE
		 +
		 (a->pvt_heap_size ? a->pvt_heap_size : SVM_PVT_MHEAP_SIZE)));

      if (mmap (rp->data_base, map_size, PROT_READ | PROT_WRITE,
		MAP_SHARED | MAP_FIXED, fd, 0) == MAP_FAILED)
	{
	  clib_unix_warning ("mmap");
	  close (fd);
	  return -3;
	}
      close (fd);
    }
  return 0;
}

u8 *
shm_name_from_svm_map_region_args (svm_map_region_args_t * a)
{
  u8 *path;
  u8 *shm_name;
  u8 *split_point;
  u8 *mkdir_arg = 0;
  int root_path_offset = 0;
  int name_offset = 0;

  if (a->root_path)
    {
      /* Tolerate present or absent slashes */
      if (a->root_path[0] == '/')
	root_path_offset++;

      /* create the root_path under /dev/shm
         iterate through path creating directories */

      path = format (0, "/dev/shm/%s%c", &a->root_path[root_path_offset], 0);
      split_point = path + 1;
      vec_add1 (mkdir_arg, '-');

      while (*split_point)
	{
	  while (*split_point && *split_point != '/')
	    {
	      vec_add1 (mkdir_arg, *split_point);
	      split_point++;
	    }
	  vec_add1 (mkdir_arg, 0);

	  /* ready to descend another level */
	  mkdir_arg[vec_len (mkdir_arg) - 1] = '-';
	  split_point++;
	}
      vec_free (mkdir_arg);
      vec_free (path);

      if (a->name[0] == '/')
	name_offset = 1;

      shm_name = format (0, "/%s-%s%c", a->root_path,
			 &a->name[name_offset], 0);
    }
  else
    shm_name = format (0, "%s%c", a->name, 0);
  return (shm_name);
}

/*
 * svm_map_region
 */
void *
svm_map_region (svm_map_region_args_t * a)
{
  int svm_fd;
  svm_region_t *rp;
  pthread_mutexattr_t attr;
  pthread_condattr_t cattr;
  int deadman = 0;
  u8 junk = 0;
  void *oldheap;
  int overhead_space;
  int rv;
  uword data_base;
  int nbits, words, bit;
  int pid_holding_region_lock;
  u8 *shm_name;
  int dead_region_recovery = 0;
  int time_left;
  struct stat stat;
  struct timespec ts, tsrem;

  if (CLIB_DEBUG > 1)
    clib_warning ("[%d] map region %s", getpid (), a->name);

  ASSERT ((a->size & ~(MMAP_PAGESIZE - 1)) == a->size);
  ASSERT (a->name);

  shm_name = shm_name_from_svm_map_region_args (a);

  svm_fd = shm_open ((char *) shm_name, O_RDWR | O_CREAT | O_EXCL, 0777);

  if (svm_fd >= 0)
    {
      if (fchmod (svm_fd, 0770) < 0)
	clib_unix_warning ("segment chmod");
      /* This turns out to fail harmlessly if the client starts first */
      if (fchown (svm_fd, a->uid, a->gid) < 0)
	clib_unix_warning ("segment chown [ok if client starts first]");

      vec_free (shm_name);

      if (lseek (svm_fd, a->size, SEEK_SET) == (off_t) - 1)
	{
	  clib_warning ("seek region size");
	  close (svm_fd);
	  return (0);
	}
      if (write (svm_fd, &junk, 1) != 1)
	{
	  clib_warning ("set region size");
	  close (svm_fd);
	  return (0);
	}

      rp = mmap ((void *) a->baseva, a->size,
		 PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, svm_fd, 0);

      if (rp == (svm_region_t *) MAP_FAILED)
	{
	  clib_unix_warning ("mmap create");
	  close (svm_fd);
	  return (0);
	}
      close (svm_fd);
      memset (rp, 0, sizeof (*rp));

      if (pthread_mutexattr_init (&attr))
	clib_unix_warning ("mutexattr_init");

      if (pthread_mutexattr_setpshared (&attr, PTHREAD_PROCESS_SHARED))
	clib_unix_warning ("mutexattr_setpshared");

      if (pthread_mutex_init (&rp->mutex, &attr))
	clib_unix_warning ("mutex_init");

      if (pthread_mutexattr_destroy (&attr))
	clib_unix_warning ("mutexattr_destroy");

      if (pthread_condattr_init (&cattr))
	clib_unix_warning ("condattr_init");

      if (pthread_condattr_setpshared (&cattr, PTHREAD_PROCESS_SHARED))
	clib_unix_warning ("condattr_setpshared");

      if (pthread_cond_init (&rp->condvar, &cattr))
	clib_unix_warning ("cond_init");

      if (pthread_condattr_destroy (&cattr))
	clib_unix_warning ("condattr_destroy");

      region_lock (rp, 1);

      rp->virtual_base = a->baseva;
      rp->virtual_size = a->size;

      rp->region_heap =
	mheap_alloc_with_flags ((void *) (a->baseva + MMAP_PAGESIZE),
				(a->pvt_heap_size != 0) ?
				a->pvt_heap_size : SVM_PVT_MHEAP_SIZE,
				MHEAP_FLAG_DISABLE_VM);
      oldheap = svm_push_pvt_heap (rp);

      rp->region_name = (char *) format (0, "%s%c", a->name, 0);
      vec_add1 (rp->client_pids, getpid ());

      nbits = rp->virtual_size / MMAP_PAGESIZE;

      ASSERT (nbits > 0);
      rp->bitmap_size = nbits;
      words = (nbits + BITS (uword) - 1) / BITS (uword);
      vec_validate (rp->bitmap, words - 1);

      overhead_space = MMAP_PAGESIZE /* header */  +
	((a->pvt_heap_size != 0) ? a->pvt_heap_size : SVM_PVT_MHEAP_SIZE);

      bit = 0;
      data_base = (uword) rp->virtual_base;

      if (a->flags & SVM_FLAGS_NODATA)
	rp->flags |= SVM_FLAGS_NEED_DATA_INIT;

      do
	{
	  clib_bitmap_set_no_check (rp->bitmap, bit, 1);
	  bit++;
	  overhead_space -= MMAP_PAGESIZE;
	  data_base += MMAP_PAGESIZE;
	}
      while (overhead_space > 0);

      rp->data_base = (void *) data_base;

      /*
       * Note: although the POSIX spec guarantees that only one
       * process enters this block, we have to play games
       * to hold off clients until e.g. the mutex is ready
       */
      rp->version = SVM_VERSION;

      /* setup the data portion of the region */

      rv = svm_data_region_create (a, rp);
      if (rv)
	{
	  clib_warning ("data_region_create: %d", rv);
	}

      region_unlock (rp);

      svm_pop_heap (oldheap);

      return ((void *) rp);
    }
  else
    {
      svm_fd = shm_open ((char *) shm_name, O_RDWR, 0777);

      vec_free (shm_name);

      if (svm_fd < 0)
	{
	  perror ("svm_region_map(mmap open)");
	  return (0);
	}

      time_left = 20;
      while (1)
	{
	  if (0 != fstat (svm_fd, &stat))
	    {
	      clib_warning ("fstat failed: %d", errno);
	      close (svm_fd);
	      return (0);
	    }
	  if (stat.st_size > 0)
	    {
	      break;
	    }
	  if (0 == time_left)
	    {
	      clib_warning ("waiting for resize of shm file timed out");
	      close (svm_fd);
	      return (0);
	    }
	  ts.tv_sec = 0;
	  ts.tv_nsec = 100000000;
	  while (nanosleep (&ts, &tsrem) < 0)
	    ts = tsrem;
	  time_left--;
	}

      rp = mmap (0, MMAP_PAGESIZE,
		 PROT_READ | PROT_WRITE, MAP_SHARED, svm_fd, 0);

      if (rp == (svm_region_t *) MAP_FAILED)
	{
	  close (svm_fd);
	  clib_warning ("mmap");
	  return (0);
	}
      /*
       * We lost the footrace to create this region; make sure
       * the winner has crossed the finish line.
       */
      while (rp->version == 0 && deadman++ < 5)
	{
	  sleep (1);
	}

      /*
       * <bleep>-ed?
       */
      if (rp->version == 0)
	{
	  clib_warning ("rp->version %d not %d", rp->version, SVM_VERSION);
	  close (svm_fd);
	  munmap (rp, a->size);
	  return (0);
	}
      /* Remap now that the region has been placed */
      a->baseva = rp->virtual_base;
      a->size = rp->virtual_size;
      munmap (rp, MMAP_PAGESIZE);

      rp = (void *) mmap ((void *) a->baseva, a->size,
			  PROT_READ | PROT_WRITE,
			  MAP_SHARED | MAP_FIXED, svm_fd, 0);
      if ((uword) rp == (uword) MAP_FAILED)
	{
	  clib_unix_warning ("mmap");
	  close (svm_fd);
	  return (0);
	}

      if ((uword) rp != rp->virtual_base)
	{
	  clib_warning ("mmap botch");
	}

      /*
       * Try to fix the region mutex if it is held by
       * a dead process
       */
      pid_holding_region_lock = rp->mutex_owner_pid;
      if (pid_holding_region_lock && kill (pid_holding_region_lock, 0) < 0)
	{
	  clib_warning
	    ("region %s mutex held by dead pid %d, tag %d, force unlock",
	     rp->region_name, pid_holding_region_lock, rp->mutex_owner_tag);
	  /* owner pid is nonexistent */
	  rp->mutex.__data.__owner = 0;
	  rp->mutex.__data.__lock = 0;
	  dead_region_recovery = 1;
	}

      if (dead_region_recovery)
	clib_warning ("recovery: attempt to re-lock region");

      region_lock (rp, 2);
      oldheap = svm_push_pvt_heap (rp);
      vec_add1 (rp->client_pids, getpid ());

      if (dead_region_recovery)
	clib_warning ("recovery: attempt svm_data_region_map");

      rv = svm_data_region_map (a, rp);
      if (rv)
	{
	  clib_warning ("data_region_map: %d", rv);
	}

      if (dead_region_recovery)
	clib_warning ("unlock and continue");

      region_unlock (rp);

      svm_pop_heap (oldheap);

      return ((void *) rp);

    }
  return 0;			/* NOTREACHED */
}

static void
svm_mutex_cleanup (void)
{
  int i;
  for (i = 0; i < nheld; i++)
    {
      pthread_mutex_unlock (mutexes_held[i]);
    }
}

static void
svm_region_init_internal (svm_map_region_args_t * a)
{
  svm_region_t *rp;
  u64 ticks = clib_cpu_time_now ();
  uword randomize_baseva;

  /* guard against klutz calls */
  if (root_rp)
    return;

  root_rp_refcount++;

  atexit (svm_mutex_cleanup);

  /* Randomize the shared-VM base at init time */
  if (MMAP_PAGESIZE <= (4 << 10))
    randomize_baseva = (ticks & 15) * MMAP_PAGESIZE;
  else
    randomize_baseva = (ticks & 3) * MMAP_PAGESIZE;

  a->baseva += randomize_baseva;

  rp = svm_map_region (a);
  ASSERT (rp);

  region_lock (rp, 3);

  /* Set up the main region data structures */
  if (rp->flags & SVM_FLAGS_NEED_DATA_INIT)
    {
      svm_main_region_t *mp = 0;
      void *oldheap;

      rp->flags &= ~(SVM_FLAGS_NEED_DATA_INIT);

      oldheap = svm_push_pvt_heap (rp);
      vec_validate (mp, 0);
      mp->name_hash = hash_create_string (0, sizeof (uword));
      mp->root_path = a->root_path ? format (0, "%s%c", a->root_path, 0) : 0;
      rp->data_base = mp;
      svm_pop_heap (oldheap);
    }
  region_unlock (rp);
  root_rp = rp;
}

void
svm_region_init (void)
{
  svm_map_region_args_t _a, *a = &_a;

  memset (a, 0, sizeof (*a));
  a->root_path = 0;
  a->name = SVM_GLOBAL_REGION_NAME;
  a->baseva = SVM_GLOBAL_REGION_BASEVA;
  a->size = SVM_GLOBAL_REGION_SIZE;
  a->flags = SVM_FLAGS_NODATA;
  a->uid = 0;
  a->gid = 0;

  svm_region_init_internal (a);
}

void
svm_region_init_chroot (char *root_path)
{
  svm_map_region_args_t _a, *a = &_a;

  memset (a, 0, sizeof (*a));
  a->root_path = root_path;
  a->name = SVM_GLOBAL_REGION_NAME;
  a->baseva = SVM_GLOBAL_REGION_BASEVA;
  a->size = SVM_GLOBAL_REGION_SIZE;
  a->flags = SVM_FLAGS_NODATA;
  a->uid = 0;
  a->gid = 0;

  svm_region_init_internal (a);
}

void
svm_region_init_chroot_uid_gid (char *root_path, int uid, int gid)
{
  svm_map_region_args_t _a, *a = &_a;

  memset (a, 0, sizeof (*a));
  a->root_path = root_path;
  a->name = SVM_GLOBAL_REGION_NAME;
  a->baseva = SVM_GLOBAL_REGION_BASEVA;
  a->size = SVM_GLOBAL_REGION_SIZE;
  a->flags = SVM_FLAGS_NODATA;
  a->uid = uid;
  a->gid = gid;

  svm_region_init_internal (a);
}

void
svm_region_init_args (svm_map_region_args_t * a)
{
  svm_region_init_internal (a);
}

void *
svm_region_find_or_create (svm_map_region_args_t * a)
{
  svm_main_region_t *mp;
  svm_region_t *rp;
  uword need_nbits;
  int index, i;
  void *oldheap;
  uword *p;
  u8 *name;
  svm_subregion_t *subp;

  ASSERT (root_rp);

  a->size += MMAP_PAGESIZE +
    ((a->pvt_heap_size != 0) ? a->pvt_heap_size : SVM_PVT_MHEAP_SIZE);
  a->size = rnd_pagesize (a->size);

  region_lock (root_rp, 4);
  oldheap = svm_push_pvt_heap (root_rp);
  mp = root_rp->data_base;

  ASSERT (mp);

  /* Map the named region from the correct chroot environment */
  a->root_path = (char *) mp->root_path;

  /*
   * See if this region is already known. If it is, we're
   * almost done...
   */
  p = hash_get_mem (mp->name_hash, a->name);

  if (p)
    {
      rp = svm_map_region (a);
      region_unlock (root_rp);
      svm_pop_heap (oldheap);
      return rp;
    }

  /* Create the region. */
  ASSERT ((a->size & ~(MMAP_PAGESIZE - 1)) == a->size);

  need_nbits = a->size / MMAP_PAGESIZE;

  index = 1;			/* $$$ fixme, figure out how many bit to really skip */

  /*
   * Scan the virtual space allocation bitmap, looking for a large
   * enough chunk
   */
  do
    {
      if (clib_bitmap_get_no_check (root_rp->bitmap, index) == 0)
	{
	  for (i = 0; i < (need_nbits - 1); i++)
	    {
	      if (clib_bitmap_get_no_check (root_rp->bitmap, index + i) == 1)
		{
		  index = index + i;
		  goto next;
		}
	    }
	  break;
	}
      index++;
    next:;
    }
  while (index < root_rp->bitmap_size);

  /* Completely out of VM? */
  if (index >= root_rp->bitmap_size)
    {
      clib_warning ("region %s: not enough VM to allocate 0x%llx (%lld)",
		    root_rp->region_name, a->size, a->size);
      svm_pop_heap (oldheap);
      region_unlock (root_rp);
      return 0;
    }

  /*
   * Mark virtual space allocated
   */
#if CLIB_DEBUG > 1
  clib_warning ("set %d bits at index %d", need_nbits, index);
#endif

  for (i = 0; i < need_nbits; i++)
    {
      clib_bitmap_set_no_check (root_rp->bitmap, index + i, 1);
    }

  /* Place this region where it goes... */
  a->baseva = root_rp->virtual_base + index * MMAP_PAGESIZE;

  rp = svm_map_region (a);

  pool_get (mp->subregions, subp);
  name = format (0, "%s%c", a->name, 0);
  subp->subregion_name = name;

  hash_set_mem (mp->name_hash, name, subp - mp->subregions);

  svm_pop_heap (oldheap);

  region_unlock (root_rp);

  return (rp);
}

/*
 * svm_region_unmap
 *
 * Let go of the indicated region. If the calling process
 * is the last customer, throw it away completely.
 * The root region mutex guarantees atomicity with respect to
 * a new region client showing up at the wrong moment.
 */
void
svm_region_unmap (void *rp_arg)
{
  int i, mypid = getpid ();
  int nclients_left;
  void *oldheap;
  uword virtual_base, virtual_size;
  svm_region_t *rp = rp_arg;
  char *name;

  /*
   * If we take a signal while holding one or more shared-memory
   * mutexes, we may end up back here from an otherwise
   * benign exit handler. Bail out to avoid a recursive
   * mutex screw-up.
   */
  if (nheld)
    return;

  ASSERT (rp);
  ASSERT (root_rp);

  if (CLIB_DEBUG > 1)
    clib_warning ("[%d] unmap region %s", getpid (), rp->region_name);

  region_lock (root_rp, 5);
  region_lock (rp, 6);

  oldheap = svm_push_pvt_heap (rp);	/* nb vec_delete() in the loop */

  /* Remove the caller from the list of mappers */
  for (i = 0; i < vec_len (rp->client_pids); i++)
    {
      if (rp->client_pids[i] == mypid)
	{
	  vec_delete (rp->client_pids, 1, i);
	  goto found;
	}
    }
  clib_warning ("pid %d AWOL", mypid);

found:

  svm_pop_heap (oldheap);

  nclients_left = vec_len (rp->client_pids);
  virtual_base = rp->virtual_base;
  virtual_size = rp->virtual_size;

  if (nclients_left == 0)
    {
      int index, nbits, i;
      svm_main_region_t *mp;
      uword *p;
      svm_subregion_t *subp;

      /* Kill the region, last guy on his way out */

      oldheap = svm_push_pvt_heap (root_rp);
      name = vec_dup (rp->region_name);

      virtual_base = rp->virtual_base;
      virtual_size = rp->virtual_size;

      /* Figure out which bits to clear in the root region bitmap */
      index = (virtual_base - root_rp->virtual_base) / MMAP_PAGESIZE;

      nbits = (virtual_size + MMAP_PAGESIZE - 1) / MMAP_PAGESIZE;

#if CLIB_DEBUG > 1
      clib_warning ("clear %d bits at index %d", nbits, index);
#endif
      /* Give back the allocated VM */
      for (i = 0; i < nbits; i++)
	{
	  clib_bitmap_set_no_check (root_rp->bitmap, index + i, 0);
	}

      mp = root_rp->data_base;

      p = hash_get_mem (mp->name_hash, name);

      /* Better never happen ... */
      if (p == NULL)
	{
	  region_unlock (rp);
	  region_unlock (root_rp);
	  svm_pop_heap (oldheap);
	  clib_warning ("Region name '%s' not found?", name);
	  return;
	}

      /* Remove from the root region subregion pool */
      subp = mp->subregions + p[0];
      pool_put (mp->subregions, subp);

      hash_unset_mem (mp->name_hash, name);

      vec_free (name);

      region_unlock (rp);
      shm_unlink (rp->region_name);
      munmap ((void *) virtual_base, virtual_size);
      region_unlock (root_rp);
      svm_pop_heap (oldheap);
      return;
    }

  region_unlock (rp);
  region_unlock (root_rp);

  munmap ((void *) virtual_base, virtual_size);
}

/*
 * svm_region_exit
 * There is no clean way to unlink the
 * root region when all clients go away,
 * so remove the pid entry and call it a day.
 */
void
svm_region_exit ()
{
  void *oldheap;
  int i, mypid = getpid ();
  uword virtual_base, virtual_size;

  /* It felt so nice we did it twice... */
  if (root_rp == 0)
    return;

  if (--root_rp_refcount > 0)
    return;

  /*
   * If we take a signal while holding one or more shared-memory
   * mutexes, we may end up back here from an otherwise
   * benign exit handler. Bail out to avoid a recursive
   * mutex screw-up.
   */
  if (nheld)
    return;

  region_lock (root_rp, 7);
  oldheap = svm_push_pvt_heap (root_rp);

  virtual_base = root_rp->virtual_base;
  virtual_size = root_rp->virtual_size;

  for (i = 0; i < vec_len (root_rp->client_pids); i++)
    {
      if (root_rp->client_pids[i] == mypid)
	{
	  vec_delete (root_rp->client_pids, 1, i);
	  goto found;
	}
    }
  clib_warning ("pid %d AWOL", mypid);

found:

  region_unlock (root_rp);
  svm_pop_heap (oldheap);

  root_rp = 0;
  munmap ((void *) virtual_base, virtual_size);
}

void
svm_client_scan_this_region_nolock (svm_region_t * rp)
{
  int j;
  int mypid = getpid ();
  void *oldheap;

  for (j = 0; j < vec_len (rp->client_pids); j++)
    {
      if (mypid == rp->client_pids[j])
	continue;
      if (rp->client_pids[j] && (kill (rp->client_pids[j], 0) < 0))
	{
	  clib_warning ("%s: cleanup ghost pid %d",
			rp->region_name, rp->client_pids[j]);
	  /* nb: client vec in rp->region_heap */
	  oldheap = svm_push_pvt_heap (rp);
	  vec_delete (rp->client_pids, 1, j);
	  j--;
	  svm_pop_heap (oldheap);
	}
    }
}


/*
 * Scan svm regions for dead clients
 */
void
svm_client_scan (char *root_path)
{
  int i, j;
  svm_main_region_t *mp;
  svm_map_region_args_t *a = 0;
  svm_region_t *root_rp;
  svm_region_t *rp;
  svm_subregion_t *subp;
  u8 *name = 0;
  u8 **svm_names = 0;
  void *oldheap;
  int mypid = getpid ();

  vec_validate (a, 0);

  svm_region_init_chroot (root_path);

  root_rp = svm_get_root_rp ();

  pthread_mutex_lock (&root_rp->mutex);

  mp = root_rp->data_base;

  for (j = 0; j < vec_len (root_rp->client_pids); j++)
    {
      if (mypid == root_rp->client_pids[j])
	continue;
      if (root_rp->client_pids[j] && (kill (root_rp->client_pids[j], 0) < 0))
	{
	  clib_warning ("%s: cleanup ghost pid %d",
			root_rp->region_name, root_rp->client_pids[j]);
	  /* nb: client vec in root_rp->region_heap */
	  oldheap = svm_push_pvt_heap (root_rp);
	  vec_delete (root_rp->client_pids, 1, j);
	  j--;
	  svm_pop_heap (oldheap);
	}
    }

  /*
   * Snapshoot names, can't hold root rp mutex across
   * find_or_create.
   */
  /* *INDENT-OFF* */
  pool_foreach (subp, mp->subregions, ({
        name = vec_dup (subp->subregion_name);
        vec_add1(svm_names, name);
      }));
  /* *INDENT-ON* */

  pthread_mutex_unlock (&root_rp->mutex);

  for (i = 0; i < vec_len (svm_names); i++)
    {
      vec_validate (a, 0);
      a->root_path = root_path;
      a->name = (char *) svm_names[i];
      rp = svm_region_find_or_create (a);
      if (rp)
	{
	  pthread_mutex_lock (&rp->mutex);

	  svm_client_scan_this_region_nolock (rp);

	  pthread_mutex_unlock (&rp->mutex);
	  svm_region_unmap (rp);
	  vec_free (svm_names[i]);
	}
      vec_free (a);
    }
  vec_free (svm_names);

  svm_region_exit ();

  vec_free (a);
}

/*
 * fd.io coding-style-patch-verification: ON
 *
 * Local Variables:
 * eval: (c-set-style "gnu")
 * End:
 */