diff options
Diffstat (limited to 'rts/MBlock.c')
-rw-r--r-- | rts/MBlock.c | 453 |
1 files changed, 453 insertions, 0 deletions
diff --git a/rts/MBlock.c b/rts/MBlock.c new file mode 100644 index 0000000000..fa8fd49d88 --- /dev/null +++ b/rts/MBlock.c @@ -0,0 +1,453 @@ +/* ----------------------------------------------------------------------------- + * + * (c) The GHC Team 1998-1999 + * + * MegaBlock Allocator Interface. This file contains all the dirty + * architecture-dependent hackery required to get a chunk of aligned + * memory from the operating system. + * + * ---------------------------------------------------------------------------*/ + +/* This is non-posix compliant. */ +/* #include "PosixSource.h" */ + +#include "Rts.h" +#include "RtsUtils.h" +#include "RtsFlags.h" +#include "MBlock.h" +#include "BlockAlloc.h" + +#ifdef HAVE_STDLIB_H +#include <stdlib.h> +#endif +#ifdef HAVE_STRING_H +#include <string.h> +#endif +#ifdef HAVE_UNISTD_H +#include <unistd.h> +#endif +#ifdef HAVE_SYS_TYPES_H +#include <sys/types.h> +#endif +#ifndef mingw32_HOST_OS +# ifdef HAVE_SYS_MMAN_H +# include <sys/mman.h> +# endif +#endif +#ifdef HAVE_FCNTL_H +#include <fcntl.h> +#endif +#if HAVE_WINDOWS_H +#include <windows.h> +#endif +#if darwin_HOST_OS +#include <mach/vm_map.h> +#endif + +#include <errno.h> + +lnat mblocks_allocated = 0; + +/* ----------------------------------------------------------------------------- + The MBlock Map: provides our implementation of HEAP_ALLOCED() + -------------------------------------------------------------------------- */ + +#if SIZEOF_VOID_P == 4 +StgWord8 mblock_map[MBLOCK_MAP_SIZE]; // initially all zeros +#elif SIZEOF_VOID_P == 8 +static MBlockMap dummy_mblock_map; +MBlockMap *mblock_cache = &dummy_mblock_map; +int mblock_map_count = 0; +MBlockMap **mblock_maps = NULL; + +static MBlockMap * +findMBlockMap(void *p) +{ + int i; + StgWord32 hi = (StgWord32) (((StgWord)p) >> 32); + for( i = 0; i < mblock_map_count; i++ ) + { + if(mblock_maps[i]->addrHigh32 == hi) + { + return mblock_maps[i]; + } + } + return NULL; +} + +StgBool +slowIsHeapAlloced(void *p) +{ + MBlockMap *map = findMBlockMap(p); + if(map) + { + mblock_cache = map; + return map->mblocks[MBLOCK_MAP_ENTRY(p)]; + } + else + return 0; +} +#endif + +static void +markHeapAlloced(void *p) +{ +#if SIZEOF_VOID_P == 4 + mblock_map[MBLOCK_MAP_ENTRY(p)] = 1; +#elif SIZEOF_VOID_P == 8 + MBlockMap *map = findMBlockMap(p); + if(map == NULL) + { + mblock_map_count++; + mblock_maps = realloc(mblock_maps, + sizeof(MBlockMap*) * mblock_map_count); + map = mblock_maps[mblock_map_count-1] = calloc(1,sizeof(MBlockMap)); + map->addrHigh32 = (StgWord32) (((StgWord)p) >> 32); + } + map->mblocks[MBLOCK_MAP_ENTRY(p)] = 1; + mblock_cache = map; +#endif +} + +/* ----------------------------------------------------------------------------- + Allocate new mblock(s) + -------------------------------------------------------------------------- */ + +void * +getMBlock(void) +{ + return getMBlocks(1); +} + +/* ----------------------------------------------------------------------------- + The mmap() method + + On Unix-like systems, we use mmap() to allocate our memory. We + want memory in chunks of MBLOCK_SIZE, and aligned on an MBLOCK_SIZE + boundary. The mmap() interface doesn't give us this level of + control, so we have to use some heuristics. + + In the general case, if we want a block of n megablocks, then we + allocate n+1 and trim off the slop from either side (using + munmap()) to get an aligned chunk of size n. However, the next + time we'll try to allocate directly after the previously allocated + chunk, on the grounds that this is aligned and likely to be free. + If it turns out that we were wrong, we have to munmap() and try + again using the general method. + + Note on posix_memalign(): this interface is available on recent + systems and appears to provide exactly what we want. However, it + turns out not to be as good as our mmap() implementation, because + it wastes extra space (using double the address space, in a test on + x86_64/Linux). The problem seems to be that posix_memalign() + returns memory that can be free()'d, so the library must store + extra information along with the allocated block, thus messing up + the alignment. Hence, we don't use posix_memalign() for now. + + -------------------------------------------------------------------------- */ + +#if !defined(mingw32_HOST_OS) && !defined(cygwin32_HOST_OS) + +// A wrapper around mmap(), to abstract away from OS differences in +// the mmap() interface. + +static void * +my_mmap (void *addr, lnat size) +{ + void *ret; + +#if defined(solaris2_HOST_OS) || defined(irix_HOST_OS) + { + int fd = open("/dev/zero",O_RDONLY); + ret = mmap(addr, size, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0); + close(fd); + } +#elif hpux_HOST_OS + ret = mmap(addr, size, PROT_READ | PROT_WRITE, + MAP_ANONYMOUS | MAP_PRIVATE, -1, 0); +#elif darwin_HOST_OS + // Without MAP_FIXED, Apple's mmap ignores addr. + // With MAP_FIXED, it overwrites already mapped regions, whic + // mmap(0, ... MAP_FIXED ...) is worst of all: It unmaps the program text + // and replaces it with zeroes, causing instant death. + // This behaviour seems to be conformant with IEEE Std 1003.1-2001. + // Let's just use the underlying Mach Microkernel calls directly, + // they're much nicer. + + kern_return_t err; + ret = addr; + if(addr) // try to allocate at adress + err = vm_allocate(mach_task_self(),(vm_address_t*) &ret, size, FALSE); + if(!addr || err) // try to allocate anywhere + err = vm_allocate(mach_task_self(),(vm_address_t*) &ret, size, TRUE); + + if(err) { + // don't know what the error codes mean exactly, assume it's + // not our problem though. + errorBelch("memory allocation failed (requested %lu bytes)", size); + stg_exit(EXIT_FAILURE); + } else { + vm_protect(mach_task_self(),ret,size,FALSE,VM_PROT_READ|VM_PROT_WRITE); + } +#else + ret = mmap(addr, size, PROT_READ | PROT_WRITE | PROT_EXEC, + MAP_ANON | MAP_PRIVATE, -1, 0); +#endif + + if (ret == (void *)-1) { + if (errno == ENOMEM || + (errno == EINVAL && sizeof(void*)==4 && size >= 0xc0000000)) { + // If we request more than 3Gig, then we get EINVAL + // instead of ENOMEM (at least on Linux). + errorBelch("out of memory (requested %lu bytes)", size); + stg_exit(EXIT_FAILURE); + } else { + barf("getMBlock: mmap: %s", strerror(errno)); + } + } + + return ret; +} + +// Implements the general case: allocate a chunk of memory of 'size' +// mblocks. + +static void * +gen_map_mblocks (lnat size) +{ + int slop; + void *ret; + + // Try to map a larger block, and take the aligned portion from + // it (unmap the rest). + size += MBLOCK_SIZE; + ret = my_mmap(0, size); + + // unmap the slop bits around the chunk we allocated + slop = (W_)ret & MBLOCK_MASK; + + if (munmap(ret, MBLOCK_SIZE - slop) == -1) { + barf("gen_map_mblocks: munmap failed"); + } + if (slop > 0 && munmap(ret+size-slop, slop) == -1) { + barf("gen_map_mblocks: munmap failed"); + } + + // ToDo: if we happened to get an aligned block, then don't + // unmap the excess, just use it. For this to work, you + // need to keep in mind the following: + // * Calling my_mmap() with an 'addr' arg pointing to + // already my_mmap()ed space is OK and won't fail. + // * If my_mmap() can't satisfy the request at the + // given 'next_request' address in getMBlocks(), that + // you unmap the extra mblock mmap()ed here (or simply + // satisfy yourself that the slop introduced isn't worth + // salvaging.) + // + + // next time, try after the block we just got. + ret += MBLOCK_SIZE - slop; + return ret; +} + + +// The external interface: allocate 'n' mblocks, and return the +// address. + +void * +getMBlocks(nat n) +{ + static caddr_t next_request = (caddr_t)HEAP_BASE; + caddr_t ret; + lnat size = MBLOCK_SIZE * n; + nat i; + + if (next_request == 0) { + // use gen_map_mblocks the first time. + ret = gen_map_mblocks(size); + } else { + ret = my_mmap(next_request, size); + + if (((W_)ret & MBLOCK_MASK) != 0) { + // misaligned block! +#if 0 // defined(DEBUG) + errorBelch("warning: getMBlock: misaligned block %p returned when allocating %d megablock(s) at %p", ret, n, next_request); +#endif + + // unmap this block... + if (munmap(ret, size) == -1) { + barf("getMBlock: munmap failed"); + } + // and do it the hard way + ret = gen_map_mblocks(size); + } + } + + // Next time, we'll try to allocate right after the block we just got. + // ToDo: check that we haven't already grabbed the memory at next_request + next_request = ret + size; + + IF_DEBUG(gc,debugBelch("Allocated %d megablock(s) at %p\n",n,ret)); + + // fill in the table + for (i = 0; i < n; i++) { + markHeapAlloced( ret + i * MBLOCK_SIZE ); + } + + mblocks_allocated += n; + + return ret; +} + +void +freeAllMBlocks(void) +{ + /* XXX Do something here */ +} + +#else /* defined(mingw32_HOST_OS) || defined(cygwin32_HOST_OS) */ + +/* + On Win32 platforms we make use of the two-phased virtual memory API + to allocate mega blocks. We proceed as follows: + + Reserve a large chunk of VM (256M at the time, or what the user asked + for via the -M option), but don't supply a base address that's aligned on + a MB boundary. Instead we round up to the nearest mblock from the chunk of + VM we're handed back from the OS (at the moment we just leave the 'slop' at + the beginning of the reserved chunk unused - ToDo: reuse it .) + + Reserving memory doesn't allocate physical storage (not even in the + page file), this is done later on by committing pages (or mega-blocks in + our case). +*/ + +static char* base_non_committed = (char*)0; +static char* end_non_committed = (char*)0; + +static void *membase; + +/* Default is to reserve 256M of VM to minimise the slop cost. */ +#define SIZE_RESERVED_POOL ( 256 * 1024 * 1024 ) + +/* Number of bytes reserved */ +static unsigned long size_reserved_pool = SIZE_RESERVED_POOL; + +void * +getMBlocks(nat n) +{ + static char* base_mblocks = (char*)0; + static char* next_request = (char*)0; + void* ret = (void*)0; + nat i; + + lnat size = MBLOCK_SIZE * n; + + if ( (base_non_committed == 0) || (next_request + size > end_non_committed) ) { + if (base_non_committed) { + /* Tacky, but if no user-provided -M option is in effect, + * set it to the default (==256M) in time for the heap overflow PSA. + */ + if (RtsFlags.GcFlags.maxHeapSize == 0) { + RtsFlags.GcFlags.maxHeapSize = size_reserved_pool / BLOCK_SIZE; + } + heapOverflow(); + } + if (RtsFlags.GcFlags.maxHeapSize != 0) { + size_reserved_pool = BLOCK_SIZE * RtsFlags.GcFlags.maxHeapSize; + if (size_reserved_pool < MBLOCK_SIZE) { + size_reserved_pool = 2*MBLOCK_SIZE; + } + } + base_non_committed = VirtualAlloc ( NULL + , size_reserved_pool + , MEM_RESERVE + , PAGE_READWRITE + ); + membase = base_non_committed; + if ( base_non_committed == 0 ) { + errorBelch("getMBlocks: VirtualAlloc MEM_RESERVE %lu failed with: %ld\n", size_reserved_pool, GetLastError()); + ret=(void*)-1; + } else { + end_non_committed = (char*)base_non_committed + (unsigned long)size_reserved_pool; + /* The returned pointer is not aligned on a mega-block boundary. Make it. */ + base_mblocks = (char*)((unsigned long)base_non_committed & (unsigned long)~MBLOCK_MASK) + MBLOCK_SIZE; +# if 0 + debugBelch("getMBlocks: Dropping %d bytes off of 256M chunk\n", + (unsigned)base_mblocks - (unsigned)base_non_committed); +# endif + + if ( ((char*)base_mblocks + size) > end_non_committed ) { + debugBelch("getMBlocks: oops, committed too small a region to start with."); + ret=(void*)-1; + } else { + next_request = base_mblocks; + } + } + } + /* Commit the mega block(s) to phys mem */ + if ( ret != (void*)-1 ) { + ret = VirtualAlloc(next_request, size, MEM_COMMIT, PAGE_READWRITE); + if (ret == NULL) { + debugBelch("getMBlocks: VirtualAlloc MEM_COMMIT %lu failed with: %ld\n", size, GetLastError()); + ret=(void*)-1; + } + } + + if (((W_)ret & MBLOCK_MASK) != 0) { + barf("getMBlocks: misaligned block returned"); + } + + if (ret == (void*)-1) { + barf("getMBlocks: unknown memory allocation failure on Win32."); + } + + IF_DEBUG(gc,debugBelch("Allocated %d megablock(s) at 0x%x\n",n,(nat)ret)); + next_request = (char*)next_request + size; + + mblocks_allocated += n; + + // fill in the table + for (i = 0; i < n; i++) { + markHeapAlloced( ret + i * MBLOCK_SIZE ); + } + + return ret; +} + +void +freeAllMBlocks(void) +{ + BOOL rc; + + rc = VirtualFree(membase, 0, MEM_RELEASE); + + if (rc == FALSE) { + debugBelch("freeAllMBlocks: VirtualFree failed with: %ld\n", GetLastError()); + } +} + +/* Hand back the physical memory that is allocated to a mega-block. + ToDo: chain the released mega block onto some list so that + getMBlocks() can get at it. + + Currently unused. +*/ +#if 0 +void +freeMBlock(void* p, nat n) +{ + BOOL rc; + + rc = VirtualFree(p, n * MBLOCK_SIZE , MEM_DECOMMIT ); + + if (rc == FALSE) { +# ifdef DEBUG + debugBelch("freeMBlocks: VirtualFree failed with: %d\n", GetLastError()); +# endif + } + +} +#endif + +#endif |