///////////////////////////////////////////////////////////////////////// // $Id$ ///////////////////////////////////////////////////////////////////////// // // Copyright (C) 2002-2012 The Bochs Project // // This library is free software; you can redistribute it and/or // modify it under the terms of the GNU Lesser General Public // License as published by the Free Software Foundation; either // version 2 of the License, or (at your option) any later version. // // This library is distributed in the hope that it will be useful, // but WITHOUT ANY WARRANTY; without even the implied warranty of // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU // Lesser General Public License for more details. // // You should have received a copy of the GNU Lesser General Public // License along with this library; if not, write to the Free Software // Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA ///////////////////////////////////////////////////////////////////////// // Define BX_PLUGGABLE in files that can be compiled into plugins. For // platforms that require a special tag on exported symbols, BX_PLUGGABLE // is used to know when we are exporting symbols and when we are importing. #define BX_PLUGGABLE #include "iodev.h" #include "cdrom.h" #include "hdimage.h" #include "vmware3.h" #include "vmware4.h" #include "vvfat.h" #include "vpc-img.h" #if BX_HAVE_SYS_MMAN_H #include #endif #ifdef linux #include #include #endif #define LOG_THIS theHDImageCtl-> bx_hdimage_ctl_c* theHDImageCtl = NULL; int libhdimage_LTX_plugin_init(plugin_t *plugin, plugintype_t type, int argc, char *argv[]) { if (type == PLUGTYPE_CORE) { theHDImageCtl = new bx_hdimage_ctl_c; bx_devices.pluginHDImageCtl = theHDImageCtl; return 0; // Success } else { return -1; } } void libhdimage_LTX_plugin_fini(void) { delete theHDImageCtl; } bx_hdimage_ctl_c::bx_hdimage_ctl_c() { put("hdimage", "IMG"); } device_image_t* bx_hdimage_ctl_c::init_image(Bit8u image_mode, Bit64u disk_size, const char *journal) { device_image_t *hdimage = NULL; // instantiate the right class switch (image_mode) { case BX_HDIMAGE_MODE_FLAT: hdimage = new default_image_t(); break; case BX_HDIMAGE_MODE_CONCAT: hdimage = new concat_image_t(); break; #if EXTERNAL_DISK_SIMULATOR case BX_HDIMAGE_MODE_EXTDISKSIM: hdimage = new EXTERNAL_DISK_SIMULATOR_CLASS(); break; #endif //EXTERNAL_DISK_SIMULATOR #if DLL_HD_SUPPORT case BX_HDIMAGE_MODE_DLL_HD: hdimage = new dll_image_t(); break; #endif //DLL_HD_SUPPORT case BX_HDIMAGE_MODE_SPARSE: hdimage = new sparse_image_t(); break; case BX_HDIMAGE_MODE_VMWARE3: hdimage = new vmware3_image_t(); break; case BX_HDIMAGE_MODE_VMWARE4: hdimage = new vmware4_image_t(); break; case BX_HDIMAGE_MODE_UNDOABLE: hdimage = new undoable_image_t(journal); break; case BX_HDIMAGE_MODE_GROWING: hdimage = new growing_image_t(); break; case BX_HDIMAGE_MODE_VOLATILE: hdimage = new volatile_image_t(journal); break; case BX_HDIMAGE_MODE_VVFAT: hdimage = new vvfat_image_t(disk_size, journal); break; case BX_HDIMAGE_MODE_VPC: hdimage = new vpc_image_t(); break; default: BX_PANIC(("unsupported HD mode : '%s'", hdimage_mode_names[image_mode])); break; } return hdimage; } #ifdef LOWLEVEL_CDROM LOWLEVEL_CDROM* bx_hdimage_ctl_c::init_cdrom(const char *dev) { return new LOWLEVEL_CDROM(dev); } #endif // helper functions int bx_read_image(int fd, Bit64s offset, void *buf, int count) { if (lseek(fd, offset, SEEK_SET) == -1) { return -1; } return read(fd, buf, count); } int bx_write_image(int fd, Bit64s offset, void *buf, int count) { if (lseek(fd, offset, SEEK_SET) == -1) { return -1; } return write(fd, buf, count); } Bit64s hdimage_save_handler(void *class_ptr, bx_param_c *param) { char imgname[BX_PATHNAME_LEN]; char path[BX_PATHNAME_LEN]; param->get_param_path(imgname, BX_PATHNAME_LEN); if (!strncmp(imgname, "bochs.", 6)) { strcpy(imgname, imgname+6); } sprintf(path, "%s/%s", SIM->get_param_string(BXPN_RESTORE_PATH)->getptr(), imgname); return ((device_image_t*)class_ptr)->save_state(path); } void hdimage_restore_handler(void *class_ptr, bx_param_c *param, Bit64s value) { char imgname[BX_PATHNAME_LEN]; char path[BX_PATHNAME_LEN]; if (value != 0) { param->get_param_path(imgname, BX_PATHNAME_LEN); if (!strncmp(imgname, "bochs.", 6)) { strcpy(imgname, imgname+6); } sprintf(path, "%s/%s", SIM->get_param_string(BXPN_RESTORE_PATH)->getptr(), imgname); ((device_image_t*)class_ptr)->restore_state(path); } } bx_bool hdimage_backup_file(int fd, const char *backup_fname) { char *buf; off_t offset; int nread, size; bx_bool ret = 1; int backup_fd = ::open(backup_fname, O_RDWR | O_CREAT | O_TRUNC #ifdef O_BINARY | O_BINARY #endif , S_IWUSR | S_IRUSR | S_IRGRP | S_IWGRP); if (backup_fd >= 0) { offset = 0; size = 0x20000; buf = (char*)malloc(size); if (buf == NULL) { ::close(backup_fd); return 0; } while ((nread = bx_read_image(fd, offset, buf, size)) > 0) { if (bx_write_image(backup_fd, offset, buf, nread) < 0) { ret = 0; break; } if (nread < size) { break; } offset += size; }; if (nread < 0) { ret = 0; } free(buf); ::close(backup_fd); return ret; } return 0; } bx_bool hdimage_copy_file(const char *src, const char *dst) { #ifdef WIN32 return (bx_bool)CopyFile(src, dst, FALSE); #else int fd1, fd2; char *buf; off_t offset; int nread, size; bx_bool ret = 1; fd1 = ::open(src, O_RDONLY #ifdef O_BINARY | O_BINARY #endif ); if (fd1 < 0) return 0; fd2 = ::open(dst, O_RDWR | O_CREAT | O_TRUNC #ifdef O_BINARY | O_BINARY #endif , S_IWUSR | S_IRUSR | S_IRGRP | S_IWGRP); if (fd2 < 0) return 0; offset = 0; size = 0x20000; buf = (char*)malloc(size); if (buf == NULL) { ::close(fd1); ::close(fd2); return 0; } while ((nread = bx_read_image(fd1, offset, buf, size)) > 0) { if (bx_write_image(fd2, offset, buf, nread) < 0) { ret = 0; break; } if (nread < size) { break; } offset += size; }; if (nread < 0) { ret = 0; } free(buf); ::close(fd1); ::close(fd2); return ret; #endif } /*** base class device_image_t ***/ device_image_t::device_image_t() { hd_size = 0; } Bit32u device_image_t::get_capabilities() { return (cylinders == 0) ? HDIMAGE_AUTO_GEOMETRY : 0; } void device_image_t::register_state(bx_list_c *parent) { bx_param_bool_c *image = new bx_param_bool_c(parent, "image", NULL, NULL, 0); // TODO: restore image image->set_sr_handlers(this, hdimage_save_handler, hdimage_restore_handler); } /*** default_image_t function definitions ***/ int default_image_t::open(const char* pathname) { return open(pathname, O_RDWR); } int default_image_t::open(const char* pathname, int flags) { #ifdef WIN32 HANDLE hFile = CreateFile(pathname, GENERIC_READ, FILE_SHARE_READ, NULL, OPEN_EXISTING, FILE_FLAG_RANDOM_ACCESS, NULL); if (hFile != INVALID_HANDLE_VALUE) { ULARGE_INTEGER FileSize; FileSize.LowPart = GetFileSize(hFile, &FileSize.HighPart); GetFileTime(hFile, NULL, NULL, &mtime); CloseHandle(hFile); if ((FileSize.LowPart != INVALID_FILE_SIZE) || (GetLastError() == NO_ERROR)) { hd_size = FileSize.QuadPart; } else { return -1; } } else { return -1; } #endif fd = ::open(pathname, flags #ifdef O_BINARY | O_BINARY #endif ); if (fd < 0) { return fd; } #ifndef WIN32 /* look at size of image file to calculate disk geometry */ struct stat stat_buf; if (fstat(fd, &stat_buf)) { BX_PANIC(("fstat() returns error!")); } #ifdef linux if (stat_buf.st_rdev) { // Is this a special device file (e.g. /dev/sde) ? ioctl(fd, BLKGETSIZE64, &hd_size); // yes it's! } else #endif { hd_size = (Bit64u)stat_buf.st_size; // standard unix procedure to get size of regular files } mtime = stat_buf.st_mtime; #endif BX_INFO(("hd_size: "FMT_LL"u", hd_size)); if (hd_size <= 0) BX_PANIC(("size of disk image not detected / invalid")); if ((hd_size % 512) != 0) BX_PANIC(("size of disk image must be multiple of 512 bytes")); imgpath = pathname; return fd; } void default_image_t::close() { if (fd > -1) { ::close(fd); } } Bit64s default_image_t::lseek(Bit64s offset, int whence) { return (Bit64s)::lseek(fd, (off_t)offset, whence); } ssize_t default_image_t::read(void* buf, size_t count) { return ::read(fd, (char*) buf, count); } ssize_t default_image_t::write(const void* buf, size_t count) { return ::write(fd, (char*) buf, count); } Bit32u default_image_t::get_timestamp() { return (fat_datetime(mtime, 1) | (fat_datetime(mtime, 0) << 16)); } bx_bool default_image_t::save_state(const char *backup_fname) { return hdimage_backup_file(fd, backup_fname); } void default_image_t::restore_state(const char *backup_fname) { close(); if (hdimage_copy_file(backup_fname, imgpath)) { BX_PANIC(("Failed to restore image '%s'", imgpath)); } if (open(imgpath) < 0) { BX_PANIC(("Failed to oprn restored image '%s'", imgpath)); } } // helper function for concat and sparse mode images char increment_string(char *str, int diff) { // find the last character of the string, and increment it. char *p = str; while (*p != 0) p++; BX_ASSERT(p>str); // choke on zero length strings p--; // point to last character of the string (*p) += diff; // increment to next/previous ascii code. BX_DEBUG(("increment string returning '%s'", str)); return (*p); } /*** concat_image_t function definitions ***/ concat_image_t::concat_image_t() { fd = -1; } void concat_image_t::increment_string(char *str) { ::increment_string(str, +1); } int concat_image_t::open(const char* pathname0) { char *pathname = strdup(pathname0); BX_DEBUG(("concat_image_t.open")); Bit64s start_offset = 0; for (int i=0; i -1) { ::close(fd_table[index]); } } } Bit64s concat_image_t::lseek(Bit64s offset, int whence) { if ((offset % 512) != 0) BX_PANIC(("lseek HD with offset not multiple of 512")); BX_DEBUG(("concat_image_t.lseek(%d)", whence)); // is this offset in this disk image? if (offset < thismin) { // no, look at previous images for (int i=index-1; i>=0; i--) { if (offset >= start_offset_table[i]) { index = i; fd = fd_table[i]; thismin = start_offset_table[i]; thismax = thismin + length_table[i] - 1; BX_DEBUG(("concat_image_t.lseek to earlier image, index=%d", index)); break; } } } else if (offset > thismax) { // no, look at later images for (int i=index+1; i= length_table[index]) { BX_PANIC(("concat_image_t.lseek to byte %ld failed", (long)offset)); return -1; } seek_was_last_op = 1; return (Bit64s)::lseek(fd, (off_t)offset, whence); } ssize_t concat_image_t::read(void* buf, size_t count) { BX_DEBUG(("concat_image_t.read %ld bytes", (long)count)); // notice if anyone does sequential read or write without seek in between. // This can be supported pretty easily, but needs additional checks for // end of a partial image. if (!seek_was_last_op) BX_PANIC(("no seek before read")); return ::read(fd, (char*) buf, count); } ssize_t concat_image_t::write(const void* buf, size_t count) { BX_DEBUG(("concat_image_t.write %ld bytes", (long)count)); // notice if anyone does sequential read or write without seek in between. // This can be supported pretty easily, but needs additional checks for // end of a partial image. if (!seek_was_last_op) BX_PANIC(("no seek before write")); return ::write(fd, (char*) buf, count); } bx_bool concat_image_t::save_state(const char *backup_fname) { bx_bool ret = 1; char tempfn[BX_PATHNAME_LEN]; for (int index = 0; index < maxfd; index++) { sprintf(tempfn, "%s%d", backup_fname, index); ret &= hdimage_backup_file(fd_table[index], tempfn); if (ret == 0) break; } return ret; } /*** sparse_image_t function definitions ***/ sparse_image_t::sparse_image_t() { fd = -1; pathname = NULL; #ifdef _POSIX_MAPPED_FILES mmap_header = NULL; #endif pagetable = NULL; parent_image = NULL; } /* void showpagetable(Bit32u * pagetable, size_t numpages) { printf("Non null pages: "); for (int i = 0; i < numpages; i++) { if (pagetable[i] != 0xffffffff) { printf("%d ", i); } } printf("\n"); } */ void sparse_image_t::read_header() { BX_ASSERT(sizeof(header) == SPARSE_HEADER_SIZE); int ret = ::read(fd, &header, sizeof(header)); if (-1 == ret) { panic(strerror(errno)); } if (sizeof(header) != ret) { panic("could not read entire header"); } if (dtoh32(header.magic) != SPARSE_HEADER_MAGIC) { panic("failed header magic check"); } if ((dtoh32(header.version) != SPARSE_HEADER_VERSION) && (dtoh32(header.version) != SPARSE_HEADER_V1)) { panic("unknown version in header"); } pagesize = dtoh32(header.pagesize); Bit32u numpages = dtoh32(header.numpages); total_size = pagesize; total_size *= numpages; pagesize_shift = 0; while ((pagesize >> pagesize_shift) > 1) pagesize_shift++; if ((Bit32u)(1 << pagesize_shift) != pagesize) { panic("failed block size header check"); } pagesize_mask = pagesize - 1; size_t preamble_size = (sizeof(Bit32u) * numpages) + sizeof(header); data_start = 0; while ((size_t)data_start < preamble_size) data_start += pagesize; bx_bool did_mmap = 0; #ifdef _POSIX_MAPPED_FILES // Try to memory map from the beginning of the file (0 is trivially a page multiple) void *mmap_header = mmap(NULL, preamble_size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0); if (mmap_header == MAP_FAILED) { BX_INFO(("failed to mmap sparse disk file - using conventional file access")); mmap_header = NULL; } else { mmap_length = preamble_size; did_mmap = 1; pagetable = ((Bit32u *) (((Bit8u *) mmap_header) + sizeof(header))); system_pagesize_mask = getpagesize() - 1; } #endif if (!did_mmap) { pagetable = new Bit32u[numpages]; if (pagetable == NULL) { panic("could not allocate memory for sparse disk block table"); } ret = ::read(fd, pagetable, sizeof(Bit32u) * numpages); if (-1 == ret) { panic(strerror(errno)); } if ((int)(sizeof(Bit32u) * numpages) != ret) { panic("could not read entire block table"); } } } int sparse_image_t::open(const char* pathname0) { pathname = strdup(pathname0); BX_DEBUG(("sparse_image_t.open")); fd = ::open(pathname, O_RDWR #ifdef O_BINARY | O_BINARY #endif ); if (fd < 0) { return -1; // open failed } BX_DEBUG(("sparse_image: open image %s", pathname)); read_header(); struct stat stat_buf; if (fstat(fd, &stat_buf) != 0) panic(("fstat() returns error!")); underlying_filesize = stat_buf.st_size; if ((underlying_filesize % pagesize) != 0) panic("size of sparse disk image is not multiple of page size"); underlying_current_filepos = 0; if (-1 == ::lseek(fd, 0, SEEK_SET)) panic("error while seeking to start of file"); lseek(0, SEEK_SET); //showpagetable(pagetable, header.numpages); char * parentpathname = strdup(pathname); char lastchar = ::increment_string(parentpathname, -1); if ((lastchar >= '0') && (lastchar <= '9')) { struct stat stat_buf; if (0 == stat(parentpathname, &stat_buf)) { parent_image = new sparse_image_t(); int ret = parent_image->open(parentpathname); if (ret != 0) return ret; if ( (parent_image->pagesize != pagesize) || (parent_image->total_size != total_size)) { panic("child drive image does not have same page count/page size configuration"); } } } if (parentpathname != NULL) free(parentpathname); if (dtoh32(header.version) == SPARSE_HEADER_VERSION) { hd_size = dtoh64(header.disk); BX_INFO(("sparse: pagesize = 0x%x, data_start = 0x" FMT_LL "x", pagesize, data_start)); } return 0; // success } void sparse_image_t::close() { BX_DEBUG(("concat_image_t.close")); if (pathname != NULL) { free(pathname); } #ifdef _POSIX_MAPPED_FILES if (mmap_header != NULL) { int ret = munmap(mmap_header, mmap_length); if (ret != 0) BX_INFO(("failed to un-memory map sparse disk file")); } pagetable = NULL; // We didn't malloc it #endif if (fd > -1) { ::close(fd); } if (pagetable != NULL) { delete [] pagetable; } if (parent_image != NULL) { delete parent_image; } } Bit64s sparse_image_t::lseek(Bit64s offset, int whence) { //showpagetable(pagetable, header.numpages); if ((offset % 512) != 0) BX_PANIC(("lseek HD with offset not multiple of 512")); if (whence != SEEK_SET) BX_PANIC(("lseek HD with whence not SEEK_SET")); BX_DEBUG(("sparse_image_t.lseek(%d)", whence)); if (offset > total_size) { BX_PANIC(("sparse_image_t.lseek to byte %ld failed", (long)offset)); return -1; } //printf("Seeking to position %ld\n", (long) offset); set_virtual_page((Bit32u)(offset >> pagesize_shift)); position_page_offset = (Bit32u)(offset & pagesize_mask); return 0; } inline Bit64s sparse_image_t::get_physical_offset() { Bit64s physical_offset = data_start; physical_offset += ((Bit64s)position_physical_page << pagesize_shift); physical_offset += position_page_offset; return physical_offset; } void sparse_image_t::set_virtual_page(Bit32u new_virtual_page) { position_virtual_page = new_virtual_page; position_physical_page = dtoh32(pagetable[position_virtual_page]); } ssize_t sparse_image_t::read_page_fragment(Bit32u read_virtual_page, Bit32u read_page_offset, size_t read_size, void * buf) { if (read_virtual_page != position_virtual_page) { set_virtual_page(read_virtual_page); } position_page_offset = read_page_offset; if (position_physical_page == SPARSE_PAGE_NOT_ALLOCATED) { if (parent_image != NULL) { return parent_image->read_page_fragment(read_virtual_page, read_page_offset, read_size, buf); } else { memset(buf, 0, read_size); } } else { Bit64s physical_offset = get_physical_offset(); if (physical_offset != underlying_current_filepos) { off_t ret = ::lseek(fd, (off_t)physical_offset, SEEK_SET); // underlying_current_filepos update deferred if (ret == -1) panic(strerror(errno)); } //printf("Reading %s at position %ld size %d\n", pathname, (long) physical_offset, (long) read_size); ssize_t readret = ::read(fd, buf, read_size); if (readret == -1) { panic(strerror(errno)); } if ((size_t)readret != read_size) { panic("could not read block contents from file"); } underlying_current_filepos = physical_offset + read_size; } return read_size; } ssize_t sparse_image_t::read(void* buf, size_t count) { //showpagetable(pagetable, header.numpages); ssize_t total_read = 0; BX_DEBUG(("sparse_image_t.read %ld bytes", (long)count)); while (count != 0) { size_t can_read = pagesize - position_page_offset; if (count < can_read) can_read = count; BX_ASSERT (can_read != 0); size_t was_read = read_page_fragment(position_virtual_page, position_page_offset, can_read, buf); if (was_read != can_read) { BX_PANIC(("could not read from sparse disk")); } total_read += can_read; position_page_offset += can_read; if (position_page_offset == pagesize) { position_page_offset = 0; set_virtual_page(position_virtual_page + 1); } BX_ASSERT(position_page_offset < pagesize); buf = (((Bit8u *) buf) + can_read); count -= can_read; } return total_read; } void sparse_image_t::panic(const char * message) { char buffer[1024]; if (message == NULL) { snprintf(buffer, sizeof(buffer), "error with sparse disk image %s", pathname); } else { snprintf(buffer, sizeof(buffer), "error with sparse disk image %s - %s", pathname, message); } BX_PANIC(("%s", buffer)); } ssize_t sparse_image_t::write(const void* buf, size_t count) { //showpagetable(pagetable, header.numpages); ssize_t total_written = 0; Bit32u update_pagetable_start = position_virtual_page; Bit32u update_pagetable_count = 0; BX_DEBUG(("sparse_image_t.write %ld bytes", (long)count)); while (count != 0) { size_t can_write = pagesize - position_page_offset; if (count < can_write) can_write = count; BX_ASSERT (can_write != 0); if (position_physical_page == SPARSE_PAGE_NOT_ALLOCATED) { // We just add on another page at the end of the file // Reclamation, compaction etc should currently be done off-line Bit64s data_size = underlying_filesize - data_start; BX_ASSERT((data_size % pagesize) == 0); Bit32u data_size_pages = (Bit32u)(data_size / pagesize); Bit32u next_data_page = data_size_pages; pagetable[position_virtual_page] = htod32(next_data_page); position_physical_page = next_data_page; Bit64s page_file_start = data_start + ((Bit64s)position_physical_page << pagesize_shift); if (parent_image != NULL) { // If we have a parent, we must merge our portion with the parent void *writebuffer = NULL; if (can_write == pagesize) { writebuffer = (void *) buf; } else { writebuffer = malloc(pagesize); if (writebuffer == NULL) panic("Cannot allocate sufficient memory for page-merge in write"); // Read entire page - could optimize, but simple for now parent_image->read_page_fragment(position_virtual_page, 0, pagesize, writebuffer); void *dest_start = ((Bit8u *) writebuffer) + position_page_offset; memcpy(dest_start, buf, can_write); } int ret = (int)::lseek(fd, page_file_start, SEEK_SET); // underlying_current_filepos update deferred if (ret == -1) panic(strerror(errno)); ret = ::write(fd, writebuffer, pagesize); if (ret == -1) panic(strerror(errno)); if (pagesize != (Bit32u)ret) panic("failed to write entire merged page to disk"); if (can_write != pagesize) { free(writebuffer); } } else { // We need to write a zero page because read has been returning zeroes // We seek as close to the page end as possible, and then write a little // This produces a sparse file which has blanks // Also very quick, even when pagesize is massive int ret = (int)::lseek(fd, page_file_start + pagesize - 4, SEEK_SET); // underlying_current_filepos update deferred if (ret == -1) panic(strerror(errno)); Bit32u zero = 0; ret = ::write(fd, &zero, 4); if (ret == -1) panic(strerror(errno)); if (ret != 4) panic("failed to write entire blank page to disk"); } update_pagetable_count = (position_virtual_page - update_pagetable_start) + 1; underlying_filesize = underlying_current_filepos = page_file_start + pagesize; } BX_ASSERT(position_physical_page != SPARSE_PAGE_NOT_ALLOCATED); Bit64s physical_offset = get_physical_offset(); if (physical_offset != underlying_current_filepos) { off_t ret = ::lseek(fd, (off_t)physical_offset, SEEK_SET); // underlying_current_filepos update deferred if (ret == -1) panic(strerror(errno)); } //printf("Writing at position %ld size %d\n", (long) physical_offset, can_write); ssize_t writeret = ::write(fd, buf, can_write); if (writeret == -1) { panic(strerror(errno)); } if ((size_t)writeret != can_write) { panic("could not write block contents to file"); } underlying_current_filepos = physical_offset + can_write; total_written += can_write; position_page_offset += can_write; if (position_page_offset == pagesize) { position_page_offset = 0; set_virtual_page(position_virtual_page + 1); } BX_ASSERT(position_page_offset < pagesize); buf = (((Bit8u *) buf) + can_write); count -= can_write; } if (update_pagetable_count != 0) { bx_bool done = 0; off_t pagetable_write_from = sizeof(header) + (sizeof(Bit32u) * update_pagetable_start); size_t write_bytecount = update_pagetable_count * sizeof(Bit32u); #ifdef _POSIX_MAPPED_FILES if (mmap_header != NULL) { // Sync from the beginning of the page size_t system_page_offset = pagetable_write_from & system_pagesize_mask; void *start = ((Bit8u *) mmap_header + pagetable_write_from - system_page_offset); int ret = msync(start, system_page_offset + write_bytecount, MS_ASYNC); if (ret != 0) panic(strerror(errno)); done = 1; } #endif if (!done) { int ret = (int)::lseek(fd, pagetable_write_from, SEEK_SET); // underlying_current_filepos update deferred if (ret == -1) panic(strerror(errno)); //printf("Writing header at position %ld size %ld\n", (long) pagetable_write_from, (long) write_bytecount); ret = ::write(fd, &pagetable[update_pagetable_start], write_bytecount); if (ret == -1) panic(strerror(errno)); if ((size_t)ret != write_bytecount) panic("could not write entire updated block header"); underlying_current_filepos = pagetable_write_from + write_bytecount; } } return total_written; } bx_bool sparse_image_t::save_state(const char *backup_fname) { return hdimage_backup_file(fd, backup_fname); } #if DLL_HD_SUPPORT /*** dll_image_t function definitions ***/ /* function vdisk_open(path:PChar;numclusters,clustersize:integer):integer; procedure vdisk_read(vunit:integer;blk:integer;var buf:TBlock); procedure vdisk_write(vunit:integer;blk:integer;var buf:TBlock); procedure vdisk_close(vunit:integer); */ HINSTANCE hlib_vdisk = 0; int (*vdisk_open) (const char *path,int numclusters,int clustersize); void (*vdisk_read) (int vunit,int blk,void *buf); void (*vdisk_write) (int vunit,int blk,const void *buf); void (*vdisk_close) (int vunit); int dll_image_t::open (const char* pathname) { if (hlib_vdisk == 0) { hlib_vdisk = LoadLibrary("vdisk.dll"); if (hlib_vdisk != 0) { vdisk_read = (void (*)(int,int,void*)) GetProcAddress(hlib_vdisk,"vdisk_read"); vdisk_write = (void (*)(int,int,const void*)) GetProcAddress(hlib_vdisk,"vdisk_write"); vdisk_open = (int (*)(const char *,int,int)) GetProcAddress(hlib_vdisk,"vdisk_open"); vdisk_close = (void (*)(int)) GetProcAddress(hlib_vdisk,"vdisk_close"); } } if (hlib_vdisk != 0) { vunit = vdisk_open(pathname,0x10000,64); vblk = 0; } else { vunit = -2; } return vunit; } void dll_image_t::close () { if (vunit >= 0 && hlib_vdisk != 0) { vdisk_close(vunit); } } Bit64s dll_image_t::lseek(Bit64s offset, int whence) { vblk = (int)(offset >> 9); return 0; } ssize_t dll_image_t::read (void* buf, size_t count) { if (vunit >= 0 && hlib_vdisk != 0) { vdisk_read(vunit,vblk,buf); return count; } else { return -1; } } ssize_t dll_image_t::write (const void* buf, size_t count) { if (vunit >= 0 && hlib_vdisk != 0) { vdisk_write(vunit,vblk,buf); return count; } else { return -1; } } #endif // DLL_HD_SUPPORT // redolog implementation redolog_t::redolog_t() { fd = -1; catalog = NULL; bitmap = NULL; extent_index = (Bit32u)0; extent_offset = (Bit32u)0; extent_next = (Bit32u)0; } void redolog_t::print_header() { BX_INFO(("redolog : Standard Header : magic='%s', type='%s', subtype='%s', version = %d.%d", header.standard.magic, header.standard.type, header.standard.subtype, dtoh32(header.standard.version)/0x10000, dtoh32(header.standard.version)%0x10000)); if (dtoh32(header.standard.version) == STANDARD_HEADER_VERSION) { BX_INFO(("redolog : Specific Header : #entries=%d, bitmap size=%d, exent size = %d disk size = " FMT_LL "d", dtoh32(header.specific.catalog), dtoh32(header.specific.bitmap), dtoh32(header.specific.extent), dtoh64(header.specific.disk))); } else if (dtoh32(header.standard.version) == STANDARD_HEADER_V1) { redolog_header_v1_t header_v1; memcpy(&header_v1, &header, STANDARD_HEADER_SIZE); BX_INFO(("redolog : Specific Header : #entries=%d, bitmap size=%d, exent size = %d disk size = " FMT_LL "d", dtoh32(header_v1.specific.catalog), dtoh32(header_v1.specific.bitmap), dtoh32(header_v1.specific.extent), dtoh64(header_v1.specific.disk))); } } int redolog_t::make_header(const char* type, Bit64u size) { Bit32u entries, extent_size, bitmap_size; Bit64u maxsize; Bit32u flip=0; // Set standard header values memset(&header, 0, sizeof(redolog_header_t)); strcpy((char*)header.standard.magic, STANDARD_HEADER_MAGIC); strcpy((char*)header.standard.type, REDOLOG_TYPE); strcpy((char*)header.standard.subtype, type); header.standard.version = htod32(STANDARD_HEADER_VERSION); header.standard.header = htod32(STANDARD_HEADER_SIZE); entries = 512; bitmap_size = 1; // Compute #entries and extent size values do { extent_size = 8 * bitmap_size * 512; header.specific.catalog = htod32(entries); header.specific.bitmap = htod32(bitmap_size); header.specific.extent = htod32(extent_size); maxsize = (Bit64u)entries * (Bit64u)extent_size; flip++; if(flip&0x01) bitmap_size *= 2; else entries *= 2; } while (maxsize < size); header.specific.timestamp = 0; header.specific.disk = htod64(size); print_header(); catalog = (Bit32u*)malloc(dtoh32(header.specific.catalog) * sizeof(Bit32u)); bitmap = (Bit8u*)malloc(dtoh32(header.specific.bitmap)); if ((catalog == NULL) || (bitmap==NULL)) BX_PANIC(("redolog : could not malloc catalog or bitmap")); for (Bit32u i=0; i= extent_next) extent_next = dtoh32(catalog[i]) + 1; } } BX_INFO(("redolog : next extent will be at index %d",extent_next)); // memory used for storing bitmaps bitmap = (Bit8u *)malloc(dtoh32(header.specific.bitmap)); bitmap_blocks = 1 + (dtoh32(header.specific.bitmap) - 1) / 512; extent_blocks = 1 + (dtoh32(header.specific.extent) - 1) / 512; BX_DEBUG(("redolog : each bitmap is %d blocks", bitmap_blocks)); BX_DEBUG(("redolog : each extent is %d blocks", extent_blocks)); imagepos = 0; bitmap_update = 1; return 0; } void redolog_t::close() { if (fd >= 0) ::close(fd); if (catalog != NULL) free(catalog); if (bitmap != NULL) free(bitmap); } Bit64u redolog_t::get_size() { return dtoh64(header.specific.disk); } Bit32u redolog_t::get_timestamp() { return dtoh32(header.specific.timestamp); } bx_bool redolog_t::set_timestamp(Bit32u timestamp) { header.specific.timestamp = htod32(timestamp); // Update header bx_write_image(fd, 0, &header, dtoh32(header.standard.header)); return 1; } Bit64s redolog_t::lseek(Bit64s offset, int whence) { if ((offset % 512) != 0) { BX_PANIC(("redolog : lseek() offset not multiple of 512")); return -1; } if (whence == SEEK_SET) { imagepos = offset; } else if (whence == SEEK_CUR) { imagepos += offset; } else { BX_PANIC(("redolog: lseek() mode not supported yet")); return -1; } if (imagepos > (Bit64s)dtoh64(header.specific.disk)) { BX_PANIC(("redolog : lseek() to byte %ld failed", (long)offset)); return -1; } Bit32u old_extent_index = extent_index; extent_index = (Bit32u)(imagepos / dtoh32(header.specific.extent)); if (extent_index != old_extent_index) { bitmap_update = 1; } extent_offset = (Bit32u)((imagepos % dtoh32(header.specific.extent)) / 512); BX_DEBUG(("redolog : lseeking extent index %d, offset %d",extent_index, extent_offset)); return imagepos; } ssize_t redolog_t::read(void* buf, size_t count) { Bit64s block_offset, bitmap_offset; ssize_t ret; if (count != 512) { BX_PANIC(("redolog : read() with count not 512")); return -1; } BX_DEBUG(("redolog : reading index %d, mapping to %d", extent_index, dtoh32(catalog[extent_index]))); if (dtoh32(catalog[extent_index]) == REDOLOG_PAGE_NOT_ALLOCATED) { // page not allocated return 0; } bitmap_offset = (Bit64s)STANDARD_HEADER_SIZE + (dtoh32(header.specific.catalog) * sizeof(Bit32u)); bitmap_offset += (Bit64s)512 * dtoh32(catalog[extent_index]) * (extent_blocks + bitmap_blocks); block_offset = bitmap_offset + ((Bit64s)512 * (bitmap_blocks + extent_offset)); BX_DEBUG(("redolog : bitmap offset is %x", (Bit32u)bitmap_offset)); BX_DEBUG(("redolog : block offset is %x", (Bit32u)block_offset)); if (bitmap_update) { if (bx_read_image(fd, (off_t)bitmap_offset, bitmap, dtoh32(header.specific.bitmap)) != (ssize_t)dtoh32(header.specific.bitmap)) { BX_PANIC(("redolog : failed to read bitmap for extent %d", extent_index)); return -1; } bitmap_update = 0; } if (((bitmap[extent_offset/8] >> (extent_offset%8)) & 0x01) == 0x00) { BX_DEBUG(("read not in redolog")); // bitmap says block not in redolog return 0; } ret = bx_read_image(fd, (off_t)block_offset, buf, count); if (ret >= 0) lseek(512, SEEK_CUR); return ret; } ssize_t redolog_t::write(const void* buf, size_t count) { Bit32u i; Bit64s block_offset, bitmap_offset, catalog_offset; ssize_t written; bx_bool update_catalog = 0; if (count != 512) { BX_PANIC(("redolog : write() with count not 512")); return -1; } BX_DEBUG(("redolog : writing index %d, mapping to %d", extent_index, dtoh32(catalog[extent_index]))); if (dtoh32(catalog[extent_index]) == REDOLOG_PAGE_NOT_ALLOCATED) { if (extent_next >= dtoh32(header.specific.catalog)) { BX_PANIC(("redolog : can't allocate new extent... catalog is full")); return -1; } BX_DEBUG(("redolog : allocating new extent at %d", extent_next)); // Extent not allocated, allocate new catalog[extent_index] = htod32(extent_next); extent_next += 1; char *zerobuffer = (char*)malloc(512); memset(zerobuffer, 0, 512); // Write bitmap bitmap_offset = (Bit64s)STANDARD_HEADER_SIZE + (dtoh32(header.specific.catalog) * sizeof(Bit32u)); bitmap_offset += (Bit64s)512 * dtoh32(catalog[extent_index]) * (extent_blocks + bitmap_blocks); ::lseek(fd, (off_t)bitmap_offset, SEEK_SET); for (i=0; i> (extent_offset%8)) & 0x01) == 0x00) { bitmap[extent_offset/8] |= 1 << (extent_offset%8); bx_write_image(fd, (off_t)bitmap_offset, bitmap, dtoh32(header.specific.bitmap)); } // Write catalog if (update_catalog) { // FIXME if mmap catalog_offset = (Bit64s)STANDARD_HEADER_SIZE + (extent_index * sizeof(Bit32u)); BX_DEBUG(("redolog : writing catalog at offset %x", (Bit32u)catalog_offset)); bx_write_image(fd, (off_t)catalog_offset, &catalog[extent_index], sizeof(Bit32u)); } if (written >= 0) lseek(512, SEEK_CUR); return written; } bx_bool redolog_t::save_state(const char *backup_fname) { return hdimage_backup_file(fd, backup_fname); } /*** growing_image_t function definitions ***/ growing_image_t::growing_image_t() { redolog = new redolog_t(); } growing_image_t::~growing_image_t() { delete redolog; } int growing_image_t::open(const char* pathname) { int filedes = redolog->open(pathname, REDOLOG_SUBTYPE_GROWING); hd_size = redolog->get_size(); BX_INFO(("'growing' disk opened, growing file is '%s'", pathname)); return filedes; } void growing_image_t::close() { redolog->close(); } Bit64s growing_image_t::lseek(Bit64s offset, int whence) { return redolog->lseek(offset, whence); } ssize_t growing_image_t::read(void* buf, size_t count) { size_t n = 0; ssize_t ret = 0; memset(buf, 0, count); while (n < count) { ret = redolog->read((char*) buf, 512); if (ret < 0) break; n += 512; } return (ret < 0) ? ret : count; } ssize_t growing_image_t::write(const void* buf, size_t count) { size_t n = 0; ssize_t ret = 0; while (n < count) { ret = redolog->write((char*) buf, 512); if (ret < 0) break; n += 512; } return (ret < 0) ? ret : count; } bx_bool growing_image_t::save_state(const char *backup_fname) { return redolog->save_state(backup_fname); } // compare hd_size and modification time of r/o disk and journal bx_bool coherency_check(default_image_t *ro_disk, redolog_t *redolog) { Bit32u timestamp1, timestamp2; if (ro_disk->hd_size != redolog->get_size()) { BX_PANIC(("size reported by redolog doesn't match r/o disk size")); return 0; } timestamp1 = ro_disk->get_timestamp(); timestamp2 = redolog->get_timestamp(); if (timestamp2 != 0) { if (timestamp1 != timestamp2) { BX_PANIC(("unexpected modification time of the r/o disk")); return 0; } } else if (timestamp1 != 0) { redolog->set_timestamp(timestamp1); } return 1; } /*** undoable_image_t function definitions ***/ undoable_image_t::undoable_image_t(const char* _redolog_name) { redolog = new redolog_t(); ro_disk = new default_image_t(); redolog_name = NULL; if (_redolog_name != NULL) { if (strcmp(_redolog_name,"") != 0) { redolog_name = strdup(_redolog_name); } } } undoable_image_t::~undoable_image_t() { delete redolog; delete ro_disk; } int undoable_image_t::open(const char* pathname) { if (ro_disk->open(pathname, O_RDONLY) < 0) return -1; hd_size = ro_disk->hd_size; // If not set, we make up the redolog filename from the pathname if (redolog_name == NULL) { redolog_name = (char*)malloc(strlen(pathname) + UNDOABLE_REDOLOG_EXTENSION_LENGTH + 1); sprintf(redolog_name, "%s%s", pathname, UNDOABLE_REDOLOG_EXTENSION); } if (redolog->open(redolog_name, REDOLOG_SUBTYPE_UNDOABLE) < 0) { if (redolog->create(redolog_name, REDOLOG_SUBTYPE_UNDOABLE, hd_size) < 0) { BX_PANIC(("Can't open or create redolog '%s'",redolog_name)); return -1; } } if (!coherency_check(ro_disk, redolog)) { return -1; } BX_INFO(("'undoable' disk opened: ro-file is '%s', redolog is '%s'", pathname, redolog_name)); return 0; } void undoable_image_t::close() { redolog->close(); ro_disk->close(); if (redolog_name != NULL) free(redolog_name); } Bit64s undoable_image_t::lseek(Bit64s offset, int whence) { redolog->lseek(offset, whence); return ro_disk->lseek(offset, whence); } ssize_t undoable_image_t::read(void* buf, size_t count) { size_t n = 0; ssize_t ret = 0; while (n < count) { if ((size_t)redolog->read((char*) buf, 512) != 512) { ret = ro_disk->read((char*) buf, 512); if (ret < 0) break; } n += 512; } return (ret < 0) ? ret : count; } ssize_t undoable_image_t::write(const void* buf, size_t count) { size_t n = 0; ssize_t ret = 0; while (n < count) { ret = redolog->write((char*) buf, 512); if (ret < 0) break; n += 512; } return (ret < 0) ? ret : count; } bx_bool undoable_image_t::save_state(const char *backup_fname) { return redolog->save_state(backup_fname); } void undoable_image_t::restore_state(const char *backup_fname) { redolog_t *temp_redolog = new redolog_t(); if (temp_redolog->open(backup_fname, REDOLOG_SUBTYPE_UNDOABLE) < 0) { delete temp_redolog; BX_PANIC(("Can't open redolog backup '%s'", backup_fname)); return; } else { if (!coherency_check(ro_disk, temp_redolog)) { delete temp_redolog; return; } delete temp_redolog; } redolog->close(); if (!hdimage_copy_file(backup_fname, redolog_name)) { BX_PANIC(("Failed to restore redolog '%s'", redolog_name)); } else { if (redolog->open(redolog_name, REDOLOG_SUBTYPE_UNDOABLE) < 0) { BX_PANIC(("Can't open restored redolog '%s'", redolog_name)); } } } /*** volatile_image_t function definitions ***/ volatile_image_t::volatile_image_t(const char* _redolog_name) { redolog = new redolog_t(); ro_disk = new default_image_t(); redolog_temp = NULL; redolog_name = NULL; if (_redolog_name != NULL) { if (strcmp(_redolog_name,"") != 0) { redolog_name = strdup(_redolog_name); } } } volatile_image_t::~volatile_image_t() { delete redolog; delete ro_disk; } int volatile_image_t::open(const char* pathname) { int filedes; const char *logname=NULL; Bit32u timestamp; if (ro_disk->open(pathname, O_RDONLY)<0) return -1; hd_size = ro_disk->hd_size; // if redolog name was set if (redolog_name != NULL) { if (strcmp(redolog_name, "") != 0) { logname = redolog_name; } } // otherwise use pathname as template if (logname == NULL) { logname = pathname; } redolog_temp = (char*)malloc(strlen(logname) + VOLATILE_REDOLOG_EXTENSION_LENGTH + 1); sprintf (redolog_temp, "%s%s", logname, VOLATILE_REDOLOG_EXTENSION); filedes = mkstemp (redolog_temp); if (filedes < 0) { BX_PANIC(("Can't create volatile redolog '%s'", redolog_temp)); return -1; } if (redolog->create(filedes, REDOLOG_SUBTYPE_VOLATILE, hd_size) < 0) { BX_PANIC(("Can't create volatile redolog '%s'", redolog_temp)); return -1; } #if (!defined(WIN32)) && !BX_WITH_MACOS // on unix it is legal to delete an open file unlink(redolog_temp); #endif // timestamp required for save/restore support timestamp = ro_disk->get_timestamp(); redolog->set_timestamp(timestamp); BX_INFO(("'volatile' disk opened: ro-file is '%s', redolog is '%s'", pathname, redolog_temp)); return 0; } void volatile_image_t::close() { redolog->close(); ro_disk->close(); #if defined(WIN32) || BX_WITH_MACOS // on non-unix we have to wait till the file is closed to delete it unlink(redolog_temp); #endif if (redolog_temp!=NULL) free(redolog_temp); if (redolog_name!=NULL) free(redolog_name); } Bit64s volatile_image_t::lseek(Bit64s offset, int whence) { redolog->lseek(offset, whence); return ro_disk->lseek(offset, whence); } ssize_t volatile_image_t::read(void* buf, size_t count) { size_t n = 0; ssize_t ret = 0; while (n < count) { if ((size_t)redolog->read((char*) buf, 512) != 512) { ret = ro_disk->read((char*) buf, 512); if (ret < 0) break; } n += 512; } return (ret < 0) ? ret : count; } ssize_t volatile_image_t::write(const void* buf, size_t count) { size_t n = 0; ssize_t ret = 0; while (n < count) { ret = redolog->write((char*) buf, 512); if (ret < 0) break; n += 512; } return (ret < 0) ? ret : count; } bx_bool volatile_image_t::save_state(const char *backup_fname) { return redolog->save_state(backup_fname); } void volatile_image_t::restore_state(const char *backup_fname) { redolog_t *temp_redolog = new redolog_t(); if (temp_redolog->open(backup_fname, REDOLOG_SUBTYPE_VOLATILE) < 0) { delete temp_redolog; BX_PANIC(("Can't open redolog backup '%s'", backup_fname)); return; } else { if (!coherency_check(ro_disk, temp_redolog)) { delete temp_redolog; return; } delete temp_redolog; } BX_ERROR(("volatile_image_t::restore_state(): UNIMPLEMENTED")); }