1 /* -----------------------------------------------------------------------------
3 * (c) The GHC Team 1998-1999
5 * MegaBlock Allocator Interface. This file contains all the dirty
6 * architecture-dependent hackery required to get a chunk of aligned
7 * memory from the operating system.
9 * ---------------------------------------------------------------------------*/
11 /* This is non-posix compliant. */
12 /* #include "PosixSource.h" */
18 #include "BlockAlloc.h"
29 #ifdef HAVE_SYS_TYPES_H
30 #include <sys/types.h>
32 #ifndef mingw32_HOST_OS
33 # ifdef HAVE_SYS_MMAN_H
34 # include <sys/mman.h>
44 #include <mach/vm_map.h>
49 lnat mblocks_allocated = 0;
51 /* -----------------------------------------------------------------------------
52 The MBlock Map: provides our implementation of HEAP_ALLOCED()
53 -------------------------------------------------------------------------- */
55 #if SIZEOF_VOID_P == 4
56 StgWord8 mblock_map[MBLOCK_MAP_SIZE]; // initially all zeros
57 #elif SIZEOF_VOID_P == 8
58 static MBlockMap dummy_mblock_map;
59 MBlockMap *mblock_cache = &dummy_mblock_map;
60 int mblock_map_count = 0;
61 MBlockMap **mblock_maps = NULL;
64 findMBlockMap(void *p)
67 StgWord32 hi = (StgWord32) (((StgWord)p) >> 32);
68 for( i = 0; i < mblock_map_count; i++ )
70 if(mblock_maps[i]->addrHigh32 == hi)
72 return mblock_maps[i];
79 slowIsHeapAlloced(void *p)
81 MBlockMap *map = findMBlockMap(p);
85 return map->mblocks[MBLOCK_MAP_ENTRY(p)];
93 markHeapAlloced(void *p)
95 #if SIZEOF_VOID_P == 4
96 mblock_map[MBLOCK_MAP_ENTRY(p)] = 1;
97 #elif SIZEOF_VOID_P == 8
98 MBlockMap *map = findMBlockMap(p);
102 mblock_maps = realloc(mblock_maps,
103 sizeof(MBlockMap*) * mblock_map_count);
104 map = mblock_maps[mblock_map_count-1] = calloc(1,sizeof(MBlockMap));
105 map->addrHigh32 = (StgWord32) (((StgWord)p) >> 32);
107 map->mblocks[MBLOCK_MAP_ENTRY(p)] = 1;
112 /* -----------------------------------------------------------------------------
113 Allocate new mblock(s)
114 -------------------------------------------------------------------------- */
119 return getMBlocks(1);
122 /* -----------------------------------------------------------------------------
125 On Unix-like systems, we use mmap() to allocate our memory. We
126 want memory in chunks of MBLOCK_SIZE, and aligned on an MBLOCK_SIZE
127 boundary. The mmap() interface doesn't give us this level of
128 control, so we have to use some heuristics.
130 In the general case, if we want a block of n megablocks, then we
131 allocate n+1 and trim off the slop from either side (using
132 munmap()) to get an aligned chunk of size n. However, the next
133 time we'll try to allocate directly after the previously allocated
134 chunk, on the grounds that this is aligned and likely to be free.
135 If it turns out that we were wrong, we have to munmap() and try
136 again using the general method.
138 Note on posix_memalign(): this interface is available on recent
139 systems and appears to provide exactly what we want. However, it
140 turns out not to be as good as our mmap() implementation, because
141 it wastes extra space (using double the address space, in a test on
142 x86_64/Linux). The problem seems to be that posix_memalign()
143 returns memory that can be free()'d, so the library must store
144 extra information along with the allocated block, thus messing up
145 the alignment. Hence, we don't use posix_memalign() for now.
147 -------------------------------------------------------------------------- */
149 #if !defined(mingw32_HOST_OS) && !defined(cygwin32_HOST_OS)
151 // A wrapper around mmap(), to abstract away from OS differences in
152 // the mmap() interface.
155 my_mmap (void *addr, lnat size)
159 #if defined(solaris2_HOST_OS) || defined(irix_HOST_OS)
161 int fd = open("/dev/zero",O_RDONLY);
162 ret = mmap(addr, size, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0);
166 ret = mmap(addr, size, PROT_READ | PROT_WRITE,
167 MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
169 // Without MAP_FIXED, Apple's mmap ignores addr.
170 // With MAP_FIXED, it overwrites already mapped regions, whic
171 // mmap(0, ... MAP_FIXED ...) is worst of all: It unmaps the program text
172 // and replaces it with zeroes, causing instant death.
173 // This behaviour seems to be conformant with IEEE Std 1003.1-2001.
174 // Let's just use the underlying Mach Microkernel calls directly,
175 // they're much nicer.
179 if(addr) // try to allocate at adress
180 err = vm_allocate(mach_task_self(),(vm_address_t*) &ret, size, FALSE);
181 if(!addr || err) // try to allocate anywhere
182 err = vm_allocate(mach_task_self(),(vm_address_t*) &ret, size, TRUE);
184 if(err) // don't know what the error codes mean exactly
185 barf("memory allocation failed (requested %lu bytes)", size);
187 vm_protect(mach_task_self(),ret,size,FALSE,VM_PROT_READ|VM_PROT_WRITE);
189 ret = mmap(addr, size, PROT_READ | PROT_WRITE | PROT_EXEC,
190 MAP_ANON | MAP_PRIVATE, -1, 0);
193 if (ret == (void *)-1) {
194 if (errno == ENOMEM ||
195 (errno == EINVAL && sizeof(void*)==4 && size >= 0xc0000000)) {
196 // If we request more than 3Gig, then we get EINVAL
197 // instead of ENOMEM (at least on Linux).
198 errorBelch("out of memory (requested %lu bytes)", size);
199 stg_exit(EXIT_FAILURE);
201 barf("getMBlock: mmap: %s", strerror(errno));
208 // Implements the general case: allocate a chunk of memory of 'size'
212 gen_map_mblocks (lnat size)
217 // Try to map a larger block, and take the aligned portion from
218 // it (unmap the rest).
220 ret = my_mmap(0, size);
222 // unmap the slop bits around the chunk we allocated
223 slop = (W_)ret & MBLOCK_MASK;
225 if (munmap(ret, MBLOCK_SIZE - slop) == -1) {
226 barf("gen_map_mblocks: munmap failed");
228 if (slop > 0 && munmap(ret+size-slop, slop) == -1) {
229 barf("gen_map_mblocks: munmap failed");
232 // ToDo: if we happened to get an aligned block, then don't
233 // unmap the excess, just use it. For this to work, you
234 // need to keep in mind the following:
235 // * Calling my_mmap() with an 'addr' arg pointing to
236 // already my_mmap()ed space is OK and won't fail.
237 // * If my_mmap() can't satisfy the request at the
238 // given 'next_request' address in getMBlocks(), that
239 // you unmap the extra mblock mmap()ed here (or simply
240 // satisfy yourself that the slop introduced isn't worth
244 // next time, try after the block we just got.
245 ret += MBLOCK_SIZE - slop;
250 // The external interface: allocate 'n' mblocks, and return the
256 static caddr_t next_request = (caddr_t)HEAP_BASE;
258 lnat size = MBLOCK_SIZE * n;
261 if (next_request == 0) {
262 // use gen_map_mblocks the first time.
263 ret = gen_map_mblocks(size);
265 ret = my_mmap(next_request, size);
267 if (((W_)ret & MBLOCK_MASK) != 0) {
269 #if 0 // defined(DEBUG)
270 errorBelch("warning: getMBlock: misaligned block %p returned when allocating %d megablock(s) at %p", ret, n, next_request);
273 // unmap this block...
274 if (munmap(ret, size) == -1) {
275 barf("getMBlock: munmap failed");
277 // and do it the hard way
278 ret = gen_map_mblocks(size);
282 // Next time, we'll try to allocate right after the block we just got.
283 // ToDo: check that we haven't already grabbed the memory at next_request
284 next_request = ret + size;
286 IF_DEBUG(gc,debugBelch("Allocated %d megablock(s) at %p\n",n,ret));
289 for (i = 0; i < n; i++) {
290 markHeapAlloced( ret + i * MBLOCK_SIZE );
293 mblocks_allocated += n;
298 #else /* defined(mingw32_HOST_OS) || defined(cygwin32_HOST_OS) */
301 On Win32 platforms we make use of the two-phased virtual memory API
302 to allocate mega blocks. We proceed as follows:
304 Reserve a large chunk of VM (256M at the time, or what the user asked
305 for via the -M option), but don't supply a base address that's aligned on
306 a MB boundary. Instead we round up to the nearest mblock from the chunk of
307 VM we're handed back from the OS (at the moment we just leave the 'slop' at
308 the beginning of the reserved chunk unused - ToDo: reuse it .)
310 Reserving memory doesn't allocate physical storage (not even in the
311 page file), this is done later on by committing pages (or mega-blocks in
315 char* base_non_committed = (char*)0;
316 char* end_non_committed = (char*)0;
318 /* Default is to reserve 256M of VM to minimise the slop cost. */
319 #define SIZE_RESERVED_POOL ( 256 * 1024 * 1024 )
321 /* Number of bytes reserved */
322 static unsigned long size_reserved_pool = SIZE_RESERVED_POOL;
327 static char* base_mblocks = (char*)0;
328 static char* next_request = (char*)0;
329 void* ret = (void*)0;
332 lnat size = MBLOCK_SIZE * n;
334 if ( (base_non_committed == 0) || (next_request + size > end_non_committed) ) {
335 if (base_non_committed) {
336 /* Tacky, but if no user-provided -M option is in effect,
337 * set it to the default (==256M) in time for the heap overflow PSA.
339 if (RtsFlags.GcFlags.maxHeapSize == 0) {
340 RtsFlags.GcFlags.maxHeapSize = size_reserved_pool / BLOCK_SIZE;
344 if (RtsFlags.GcFlags.maxHeapSize != 0) {
345 size_reserved_pool = BLOCK_SIZE * RtsFlags.GcFlags.maxHeapSize;
346 if (size_reserved_pool < MBLOCK_SIZE) {
347 size_reserved_pool = 2*MBLOCK_SIZE;
350 base_non_committed = VirtualAlloc ( NULL
355 if ( base_non_committed == 0 ) {
356 errorBelch("getMBlocks: VirtualAlloc failed with: %ld\n", GetLastError());
359 end_non_committed = (char*)base_non_committed + (unsigned long)size_reserved_pool;
360 /* The returned pointer is not aligned on a mega-block boundary. Make it. */
361 base_mblocks = (char*)((unsigned long)base_non_committed & (unsigned long)~MBLOCK_MASK) + MBLOCK_SIZE;
363 debugBelch("getMBlocks: Dropping %d bytes off of 256M chunk\n",
364 (unsigned)base_mblocks - (unsigned)base_non_committed);
367 if ( ((char*)base_mblocks + size) > end_non_committed ) {
368 debugBelch("getMBlocks: oops, committed too small a region to start with.");
371 next_request = base_mblocks;
375 /* Commit the mega block(s) to phys mem */
376 if ( ret != (void*)-1 ) {
377 ret = VirtualAlloc(next_request, size, MEM_COMMIT, PAGE_READWRITE);
379 debugBelch("getMBlocks: VirtualAlloc failed with: %ld\n", GetLastError());
384 if (((W_)ret & MBLOCK_MASK) != 0) {
385 barf("getMBlocks: misaligned block returned");
388 if (ret == (void*)-1) {
389 barf("getMBlocks: unknown memory allocation failure on Win32.");
392 IF_DEBUG(gc,debugBelch("Allocated %d megablock(s) at 0x%x\n",n,(nat)ret));
393 next_request = (char*)next_request + size;
395 mblocks_allocated += n;
398 for (i = 0; i < n; i++) {
399 markHeapAlloced( ret + i * MBLOCK_SIZE );
405 /* Hand back the physical memory that is allocated to a mega-block.
406 ToDo: chain the released mega block onto some list so that
407 getMBlocks() can get at it.
413 freeMBlock(void* p, nat n)
417 rc = VirtualFree(p, n * MBLOCK_SIZE , MEM_DECOMMIT );
421 debugBelch("freeMBlocks: VirtualFree failed with: %d\n", GetLastError());