Index: head/stand/common/bcache.c =================================================================== --- head/stand/common/bcache.c (revision 345065) +++ head/stand/common/bcache.c (revision 345066) @@ -1,488 +1,488 @@ /*- * Copyright (c) 1998 Michael Smith * Copyright 2015 Toomas Soome * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include #include __FBSDID("$FreeBSD$"); /* * Simple hashed block cache */ #include #include #include #include #include "bootstrap.h" /* #define BCACHE_DEBUG */ #ifdef BCACHE_DEBUG -# define DEBUG(fmt, args...) printf("%s: " fmt "\n" , __func__ , ## args) +# define DPRINTF(fmt, args...) printf("%s: " fmt "\n" , __func__ , ## args) #else -# define DEBUG(fmt, args...) +# define DPRINTF(fmt, args...) #endif struct bcachectl { daddr_t bc_blkno; int bc_count; }; /* * bcache per device node. cache is allocated on device first open and freed * on last close, to save memory. The issue there is the size; biosdisk * supports up to 31 (0x1f) devices. Classic setup would use single disk * to boot from, but this has changed with zfs. */ struct bcache { struct bcachectl *bcache_ctl; caddr_t bcache_data; size_t bcache_nblks; size_t ra; }; static u_int bcache_total_nblks; /* set by bcache_init */ static u_int bcache_blksize; /* set by bcache_init */ static u_int bcache_numdev; /* set by bcache_add_dev */ /* statistics */ static u_int bcache_units; /* number of devices with cache */ static u_int bcache_unit_nblks; /* nblocks per unit */ static u_int bcache_hits; static u_int bcache_misses; static u_int bcache_ops; static u_int bcache_bypasses; static u_int bcache_bcount; static u_int bcache_rablks; #define BHASH(bc, blkno) ((blkno) & ((bc)->bcache_nblks - 1)) #define BCACHE_LOOKUP(bc, blkno) \ ((bc)->bcache_ctl[BHASH((bc), (blkno))].bc_blkno != (blkno)) #define BCACHE_READAHEAD 256 #define BCACHE_MINREADAHEAD 32 static void bcache_invalidate(struct bcache *bc, daddr_t blkno); static void bcache_insert(struct bcache *bc, daddr_t blkno); static void bcache_free_instance(struct bcache *bc); /* * Initialise the cache for (nblks) of (bsize). */ void bcache_init(size_t nblks, size_t bsize) { /* set up control data */ bcache_total_nblks = nblks; bcache_blksize = bsize; } /* * add number of devices to bcache. we have to divide cache space * between the devices, so bcache_add_dev() can be used to set up the * number. The issue is, we need to get the number before actual allocations. * bcache_add_dev() is supposed to be called from device init() call, so the * assumption is, devsw dv_init is called for plain devices first, and * for zfs, last. */ void bcache_add_dev(int devices) { bcache_numdev += devices; } void * bcache_allocate(void) { u_int i; struct bcache *bc = malloc(sizeof (struct bcache)); int disks = bcache_numdev; if (disks == 0) disks = 1; /* safe guard */ if (bc == NULL) { errno = ENOMEM; return (bc); } /* * the bcache block count must be power of 2 for hash function */ i = fls(disks) - 1; /* highbit - 1 */ if (disks > (1 << i)) /* next power of 2 */ i++; bc->bcache_nblks = bcache_total_nblks >> i; bcache_unit_nblks = bc->bcache_nblks; bc->bcache_data = malloc(bc->bcache_nblks * bcache_blksize); if (bc->bcache_data == NULL) { /* dont error out yet. fall back to 32 blocks and try again */ bc->bcache_nblks = 32; bc->bcache_data = malloc(bc->bcache_nblks * bcache_blksize + sizeof(uint32_t)); } bc->bcache_ctl = malloc(bc->bcache_nblks * sizeof(struct bcachectl)); if ((bc->bcache_data == NULL) || (bc->bcache_ctl == NULL)) { bcache_free_instance(bc); errno = ENOMEM; return (NULL); } /* Flush the cache */ for (i = 0; i < bc->bcache_nblks; i++) { bc->bcache_ctl[i].bc_count = -1; bc->bcache_ctl[i].bc_blkno = -1; } bcache_units++; bc->ra = BCACHE_READAHEAD; /* optimistic read ahead */ return (bc); } void bcache_free(void *cache) { struct bcache *bc = cache; if (bc == NULL) return; bcache_free_instance(bc); bcache_units--; } /* * Handle a write request; write directly to the disk, and populate the * cache with the new values. */ static int write_strategy(void *devdata, int rw, daddr_t blk, size_t size, char *buf, size_t *rsize) { struct bcache_devdata *dd = (struct bcache_devdata *)devdata; struct bcache *bc = dd->dv_cache; daddr_t i, nblk; nblk = size / bcache_blksize; /* Invalidate the blocks being written */ for (i = 0; i < nblk; i++) { bcache_invalidate(bc, blk + i); } /* Write the blocks */ return (dd->dv_strategy(dd->dv_devdata, rw, blk, size, buf, rsize)); } /* * Handle a read request; fill in parts of the request that can * be satisfied by the cache, use the supplied strategy routine to do * device I/O and then use the I/O results to populate the cache. */ static int read_strategy(void *devdata, int rw, daddr_t blk, size_t size, char *buf, size_t *rsize) { struct bcache_devdata *dd = (struct bcache_devdata *)devdata; struct bcache *bc = dd->dv_cache; size_t i, nblk, p_size, r_size, complete, ra; int result; daddr_t p_blk; caddr_t p_buf; if (bc == NULL) { errno = ENODEV; return (-1); } if (rsize != NULL) *rsize = 0; nblk = size / bcache_blksize; if (nblk == 0 && size != 0) nblk++; result = 0; complete = 1; /* Satisfy any cache hits up front, break on first miss */ for (i = 0; i < nblk; i++) { if (BCACHE_LOOKUP(bc, (daddr_t)(blk + i))) { bcache_misses += (nblk - i); complete = 0; if (nblk - i > BCACHE_MINREADAHEAD && bc->ra > BCACHE_MINREADAHEAD) bc->ra >>= 1; /* reduce read ahead */ break; } else { bcache_hits++; } } if (complete) { /* whole set was in cache, return it */ if (bc->ra < BCACHE_READAHEAD) bc->ra <<= 1; /* increase read ahead */ bcopy(bc->bcache_data + (bcache_blksize * BHASH(bc, blk)), buf, size); goto done; } /* * Fill in any misses. From check we have i pointing to first missing * block, read in all remaining blocks + readahead. * We have space at least for nblk - i before bcache wraps. */ p_blk = blk + i; p_buf = bc->bcache_data + (bcache_blksize * BHASH(bc, p_blk)); r_size = bc->bcache_nblks - BHASH(bc, p_blk); /* remaining blocks */ p_size = MIN(r_size, nblk - i); /* read at least those blocks */ /* * The read ahead size setup. * While the read ahead can save us IO, it also can complicate things: * 1. We do not want to read ahead by wrapping around the * bcache end - this would complicate the cache management. * 2. We are using bc->ra as dynamic hint for read ahead size, * detected cache hits will increase the read-ahead block count, and * misses will decrease, see the code above. * 3. The bcache is sized by 512B blocks, however, the underlying device * may have a larger sector size, and we should perform the IO by * taking into account these larger sector sizes. We could solve this by * passing the sector size to bcache_allocate(), or by using ioctl(), but * in this version we are using the constant, 16 blocks, and are rounding * read ahead block count down to multiple of 16. * Using the constant has two reasons, we are not entirely sure if the * BIOS disk interface is providing the correct value for sector size. * And secondly, this way we get the most conservative setup for the ra. * * The selection of multiple of 16 blocks (8KB) is quite arbitrary, however, * we want to cover CDs (2K) and 4K disks. * bcache_allocate() will always fall back to a minimum of 32 blocks. * Our choice of 16 read ahead blocks will always fit inside the bcache. */ if ((rw & F_NORA) == F_NORA) ra = 0; else ra = bc->bcache_nblks - BHASH(bc, p_blk + p_size); if (ra != 0 && ra != bc->bcache_nblks) { /* do we have RA space? */ ra = MIN(bc->ra, ra - 1); ra = rounddown(ra, 16); /* multiple of 16 blocks */ p_size += ra; } /* invalidate bcache */ for (i = 0; i < p_size; i++) { bcache_invalidate(bc, p_blk + i); } r_size = 0; /* * with read-ahead, it may happen we are attempting to read past * disk end, as bcache has no information about disk size. * in such case we should get partial read if some blocks can be * read or error, if no blocks can be read. * in either case we should return the data in bcache and only * return error if there is no data. */ rw &= F_MASK; result = dd->dv_strategy(dd->dv_devdata, rw, p_blk, p_size * bcache_blksize, p_buf, &r_size); r_size /= bcache_blksize; for (i = 0; i < r_size; i++) bcache_insert(bc, p_blk + i); /* update ra statistics */ if (r_size != 0) { if (r_size < p_size) bcache_rablks += (p_size - r_size); else bcache_rablks += ra; } /* check how much data can we copy */ for (i = 0; i < nblk; i++) { if (BCACHE_LOOKUP(bc, (daddr_t)(blk + i))) break; } if (size > i * bcache_blksize) size = i * bcache_blksize; if (size != 0) { bcopy(bc->bcache_data + (bcache_blksize * BHASH(bc, blk)), buf, size); result = 0; } done: if ((result == 0) && (rsize != NULL)) *rsize = size; return(result); } /* * Requests larger than 1/2 cache size will be bypassed and go * directly to the disk. XXX tune this. */ int bcache_strategy(void *devdata, int rw, daddr_t blk, size_t size, char *buf, size_t *rsize) { struct bcache_devdata *dd = (struct bcache_devdata *)devdata; struct bcache *bc = dd->dv_cache; u_int bcache_nblks = 0; int nblk, cblk, ret; size_t csize, isize, total; bcache_ops++; if (bc != NULL) bcache_nblks = bc->bcache_nblks; /* bypass large requests, or when the cache is inactive */ if (bc == NULL || ((size * 2 / bcache_blksize) > bcache_nblks)) { - DEBUG("bypass %zu from %qu", size / bcache_blksize, blk); + DPRINTF("bypass %zu from %qu", size / bcache_blksize, blk); bcache_bypasses++; rw &= F_MASK; return (dd->dv_strategy(dd->dv_devdata, rw, blk, size, buf, rsize)); } switch (rw & F_MASK) { case F_READ: nblk = size / bcache_blksize; if (size != 0 && nblk == 0) nblk++; /* read at least one block */ ret = 0; total = 0; while(size) { cblk = bcache_nblks - BHASH(bc, blk); /* # of blocks left */ cblk = MIN(cblk, nblk); if (size <= bcache_blksize) csize = size; else csize = cblk * bcache_blksize; ret = read_strategy(devdata, rw, blk, csize, buf+total, &isize); /* * we may have error from read ahead, if we have read some data * return partial read. */ if (ret != 0 || isize == 0) { if (total != 0) ret = 0; break; } blk += isize / bcache_blksize; total += isize; size -= isize; nblk = size / bcache_blksize; } if (rsize) *rsize = total; return (ret); case F_WRITE: return write_strategy(devdata, F_WRITE, blk, size, buf, rsize); } return -1; } /* * Free allocated bcache instance */ static void bcache_free_instance(struct bcache *bc) { if (bc != NULL) { if (bc->bcache_ctl) free(bc->bcache_ctl); if (bc->bcache_data) free(bc->bcache_data); free(bc); } } /* * Insert a block into the cache. */ static void bcache_insert(struct bcache *bc, daddr_t blkno) { u_int cand; cand = BHASH(bc, blkno); - DEBUG("insert blk %llu -> %u # %d", blkno, cand, bcache_bcount); + DPRINTF("insert blk %llu -> %u # %d", blkno, cand, bcache_bcount); bc->bcache_ctl[cand].bc_blkno = blkno; bc->bcache_ctl[cand].bc_count = bcache_bcount++; } /* * Invalidate a block from the cache. */ static void bcache_invalidate(struct bcache *bc, daddr_t blkno) { u_int i; i = BHASH(bc, blkno); if (bc->bcache_ctl[i].bc_blkno == blkno) { bc->bcache_ctl[i].bc_count = -1; bc->bcache_ctl[i].bc_blkno = -1; - DEBUG("invalidate blk %llu", blkno); + DPRINTF("invalidate blk %llu", blkno); } } #ifndef BOOT2 COMMAND_SET(bcachestat, "bcachestat", "get disk block cache stats", command_bcache); static int command_bcache(int argc, char *argv[]) { if (argc != 1) { command_errmsg = "wrong number of arguments"; return(CMD_ERROR); } printf("\ncache blocks: %u\n", bcache_total_nblks); printf("cache blocksz: %u\n", bcache_blksize); printf("cache readahead: %u\n", bcache_rablks); printf("unit cache blocks: %u\n", bcache_unit_nblks); printf("cached units: %u\n", bcache_units); printf("%u ops %d bypasses %u hits %u misses\n", bcache_ops, bcache_bypasses, bcache_hits, bcache_misses); return(CMD_OK); } #endif Index: head/stand/common/disk.c =================================================================== --- head/stand/common/disk.c (revision 345065) +++ head/stand/common/disk.c (revision 345066) @@ -1,443 +1,443 @@ /*- * Copyright (c) 1998 Michael Smith * Copyright (c) 2012 Andrey V. Elsukov * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include "disk.h" #ifdef DISK_DEBUG -# define DEBUG(fmt, args...) printf("%s: " fmt "\n" , __func__ , ## args) +# define DPRINTF(fmt, args...) printf("%s: " fmt "\n" , __func__ , ## args) #else -# define DEBUG(fmt, args...) +# define DPRINTF(fmt, args...) #endif struct open_disk { struct ptable *table; uint64_t mediasize; uint64_t entrysize; u_int sectorsize; }; struct print_args { struct disk_devdesc *dev; const char *prefix; int verbose; }; /* Convert size to a human-readable number. */ static char * display_size(uint64_t size, u_int sectorsize) { static char buf[80]; char unit; size = size * sectorsize / 1024; unit = 'K'; if (size >= 10485760000LL) { size /= 1073741824; unit = 'T'; } else if (size >= 10240000) { size /= 1048576; unit = 'G'; } else if (size >= 10000) { size /= 1024; unit = 'M'; } sprintf(buf, "%4ld%cB", (long)size, unit); return (buf); } int ptblread(void *d, void *buf, size_t blocks, uint64_t offset) { struct disk_devdesc *dev; struct open_disk *od; dev = (struct disk_devdesc *)d; od = (struct open_disk *)dev->dd.d_opendata; /* * The strategy function assumes the offset is in units of 512 byte * sectors. For larger sector sizes, we need to adjust the offset to * match the actual sector size. */ offset *= (od->sectorsize / 512); /* * As the GPT backup partition is located at the end of the disk, * to avoid reading past disk end, flag bcache not to use RA. */ return (dev->dd.d_dev->dv_strategy(dev, F_READ | F_NORA, offset, blocks * od->sectorsize, (char *)buf, NULL)); } static int ptable_print(void *arg, const char *pname, const struct ptable_entry *part) { struct disk_devdesc dev; struct print_args *pa, bsd; struct open_disk *od; struct ptable *table; char line[80]; int res; u_int sectsize; uint64_t partsize; pa = (struct print_args *)arg; od = (struct open_disk *)pa->dev->dd.d_opendata; sectsize = od->sectorsize; partsize = part->end - part->start + 1; sprintf(line, " %s%s: %s\t%s\n", pa->prefix, pname, parttype2str(part->type), pa->verbose ? display_size(partsize, sectsize) : ""); if (pager_output(line)) return 1; res = 0; if (part->type == PART_FREEBSD) { /* Open slice with BSD label */ dev.dd.d_dev = pa->dev->dd.d_dev; dev.dd.d_unit = pa->dev->dd.d_unit; dev.d_slice = part->index; dev.d_partition = -1; if (disk_open(&dev, partsize, sectsize) == 0) { /* * disk_open() for partition -1 on a bsd slice assumes * you want the first bsd partition. Reset things so * that we're looking at the start of the raw slice. */ dev.d_partition = -1; dev.d_offset = part->start; table = ptable_open(&dev, partsize, sectsize, ptblread); if (table != NULL) { sprintf(line, " %s%s", pa->prefix, pname); bsd.dev = pa->dev; bsd.prefix = line; bsd.verbose = pa->verbose; res = ptable_iterate(table, &bsd, ptable_print); ptable_close(table); } disk_close(&dev); } } return (res); } int disk_print(struct disk_devdesc *dev, char *prefix, int verbose) { struct open_disk *od; struct print_args pa; /* Disk should be opened */ od = (struct open_disk *)dev->dd.d_opendata; pa.dev = dev; pa.prefix = prefix; pa.verbose = verbose; return (ptable_iterate(od->table, &pa, ptable_print)); } int disk_read(struct disk_devdesc *dev, void *buf, uint64_t offset, u_int blocks) { struct open_disk *od; int ret; od = (struct open_disk *)dev->dd.d_opendata; ret = dev->dd.d_dev->dv_strategy(dev, F_READ, dev->d_offset + offset, blocks * od->sectorsize, buf, NULL); return (ret); } int disk_write(struct disk_devdesc *dev, void *buf, uint64_t offset, u_int blocks) { struct open_disk *od; int ret; od = (struct open_disk *)dev->dd.d_opendata; ret = dev->dd.d_dev->dv_strategy(dev, F_WRITE, dev->d_offset + offset, blocks * od->sectorsize, buf, NULL); return (ret); } int disk_ioctl(struct disk_devdesc *dev, u_long cmd, void *data) { struct open_disk *od = dev->dd.d_opendata; if (od == NULL) return (ENOTTY); switch (cmd) { case DIOCGSECTORSIZE: *(u_int *)data = od->sectorsize; break; case DIOCGMEDIASIZE: if (dev->d_offset == 0) *(uint64_t *)data = od->mediasize; else *(uint64_t *)data = od->entrysize * od->sectorsize; break; default: return (ENOTTY); } return (0); } int disk_open(struct disk_devdesc *dev, uint64_t mediasize, u_int sectorsize) { struct disk_devdesc partdev; struct open_disk *od; struct ptable *table; struct ptable_entry part; int rc, slice, partition; rc = 0; od = (struct open_disk *)malloc(sizeof(struct open_disk)); if (od == NULL) { - DEBUG("no memory"); + DPRINTF("no memory"); return (ENOMEM); } dev->dd.d_opendata = od; od->entrysize = 0; od->mediasize = mediasize; od->sectorsize = sectorsize; /* * While we are reading disk metadata, make sure we do it relative * to the start of the disk */ memcpy(&partdev, dev, sizeof(partdev)); partdev.d_offset = 0; partdev.d_slice = -1; partdev.d_partition = -1; dev->d_offset = 0; table = NULL; slice = dev->d_slice; partition = dev->d_partition; - DEBUG("%s unit %d, slice %d, partition %d => %p", + DPRINTF("%s unit %d, slice %d, partition %d => %p", disk_fmtdev(dev), dev->dd.d_unit, dev->d_slice, dev->d_partition, od); /* Determine disk layout. */ od->table = ptable_open(&partdev, mediasize / sectorsize, sectorsize, ptblread); if (od->table == NULL) { - DEBUG("Can't read partition table"); + DPRINTF("Can't read partition table"); rc = ENXIO; goto out; } if (ptable_getsize(od->table, &mediasize) != 0) { rc = ENXIO; goto out; } od->mediasize = mediasize; if (ptable_gettype(od->table) == PTABLE_BSD && partition >= 0) { /* It doesn't matter what value has d_slice */ rc = ptable_getpart(od->table, &part, partition); if (rc == 0) { dev->d_offset = part.start; od->entrysize = part.end - part.start + 1; } } else if (ptable_gettype(od->table) == PTABLE_ISO9660) { dev->d_offset = 0; od->entrysize = mediasize; } else if (slice >= 0) { /* Try to get information about partition */ if (slice == 0) rc = ptable_getbestpart(od->table, &part); else rc = ptable_getpart(od->table, &part, slice); if (rc != 0) /* Partition doesn't exist */ goto out; dev->d_offset = part.start; od->entrysize = part.end - part.start + 1; slice = part.index; if (ptable_gettype(od->table) == PTABLE_GPT) { partition = 255; goto out; /* Nothing more to do */ } else if (partition == 255) { /* * When we try to open GPT partition, but partition * table isn't GPT, reset d_partition value to -1 * and try to autodetect appropriate value. */ partition = -1; } /* * If d_partition < 0 and we are looking at a BSD slice, * then try to read BSD label, otherwise return the * whole MBR slice. */ if (partition == -1 && part.type != PART_FREEBSD) goto out; /* Try to read BSD label */ table = ptable_open(dev, part.end - part.start + 1, od->sectorsize, ptblread); if (table == NULL) { - DEBUG("Can't read BSD label"); + DPRINTF("Can't read BSD label"); rc = ENXIO; goto out; } /* * If slice contains BSD label and d_partition < 0, then * assume the 'a' partition. Otherwise just return the * whole MBR slice, because it can contain ZFS. */ if (partition < 0) { if (ptable_gettype(table) != PTABLE_BSD) goto out; partition = 0; } rc = ptable_getpart(table, &part, partition); if (rc != 0) goto out; dev->d_offset += part.start; od->entrysize = part.end - part.start + 1; } out: if (table != NULL) ptable_close(table); if (rc != 0) { if (od->table != NULL) ptable_close(od->table); free(od); - DEBUG("%s could not open", disk_fmtdev(dev)); + DPRINTF("%s could not open", disk_fmtdev(dev)); } else { /* Save the slice and partition number to the dev */ dev->d_slice = slice; dev->d_partition = partition; - DEBUG("%s offset %lld => %p", disk_fmtdev(dev), + DPRINTF("%s offset %lld => %p", disk_fmtdev(dev), (long long)dev->d_offset, od); } return (rc); } int disk_close(struct disk_devdesc *dev) { struct open_disk *od; od = (struct open_disk *)dev->dd.d_opendata; - DEBUG("%s closed => %p", disk_fmtdev(dev), od); + DPRINTF("%s closed => %p", disk_fmtdev(dev), od); ptable_close(od->table); free(od); return (0); } char* disk_fmtdev(struct disk_devdesc *dev) { static char buf[128]; char *cp; cp = buf + sprintf(buf, "%s%d", dev->dd.d_dev->dv_name, dev->dd.d_unit); if (dev->d_slice >= 0) { #ifdef LOADER_GPT_SUPPORT if (dev->d_partition == 255) { sprintf(cp, "p%d:", dev->d_slice); return (buf); } else #endif #ifdef LOADER_MBR_SUPPORT cp += sprintf(cp, "s%d", dev->d_slice); #endif } if (dev->d_partition >= 0) cp += sprintf(cp, "%c", dev->d_partition + 'a'); strcat(cp, ":"); return (buf); } int disk_parsedev(struct disk_devdesc *dev, const char *devspec, const char **path) { int unit, slice, partition; const char *np; char *cp; np = devspec; unit = slice = partition = -1; if (*np != '\0' && *np != ':') { unit = strtol(np, &cp, 10); if (cp == np) return (EUNIT); #ifdef LOADER_GPT_SUPPORT if (*cp == 'p') { np = cp + 1; slice = strtol(np, &cp, 10); if (np == cp) return (ESLICE); /* we don't support nested partitions on GPT */ if (*cp != '\0' && *cp != ':') return (EINVAL); partition = 255; } else #endif #ifdef LOADER_MBR_SUPPORT if (*cp == 's') { np = cp + 1; slice = strtol(np, &cp, 10); if (np == cp) return (ESLICE); } #endif if (*cp != '\0' && *cp != ':') { partition = *cp - 'a'; if (partition < 0) return (EPART); cp++; } } else return (EINVAL); if (*cp != '\0' && *cp != ':') return (EINVAL); dev->dd.d_unit = unit; dev->d_slice = slice; dev->d_partition = partition; if (path != NULL) *path = (*cp == '\0') ? cp: cp + 1; return (0); } Index: head/stand/common/interp_forth.c =================================================================== --- head/stand/common/interp_forth.c (revision 345065) +++ head/stand/common/interp_forth.c (revision 345066) @@ -1,452 +1,452 @@ /*- * Copyright (c) 1998 Michael Smith * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include /* to pick up __FreeBSD_version */ #include #include #include "bootstrap.h" #include "ficl.h" extern unsigned bootprog_rev; INTERP_DEFINE("4th"); /* #define BFORTH_DEBUG */ #ifdef BFORTH_DEBUG -#define DEBUG(fmt, args...) printf("%s: " fmt "\n" , __func__ , ## args) +#define DPRINTF(fmt, args...) printf("%s: " fmt "\n" , __func__ , ## args) #else -#define DEBUG(fmt, args...) +#define DPRINTF(fmt, args...) #endif /* * Eventually, all builtin commands throw codes must be defined * elsewhere, possibly bootstrap.h. For now, just this code, used * just in this file, it is getting defined. */ #define BF_PARSE 100 /* * FreeBSD loader default dictionary cells */ #ifndef BF_DICTSIZE #define BF_DICTSIZE 10000 #endif /* * BootForth Interface to Ficl Forth interpreter. */ FICL_SYSTEM *bf_sys; FICL_VM *bf_vm; /* * Shim for taking commands from BF and passing them out to 'standard' * argv/argc command functions. */ static void bf_command(FICL_VM *vm) { char *name, *line, *tail, *cp; size_t len; struct bootblk_command **cmdp; bootblk_cmd_t *cmd; int nstrings, i; int argc, result; char **argv; /* Get the name of the current word */ name = vm->runningWord->name; /* Find our command structure */ cmd = NULL; SET_FOREACH(cmdp, Xcommand_set) { if (((*cmdp)->c_name != NULL) && !strcmp(name, (*cmdp)->c_name)) cmd = (*cmdp)->c_fn; } if (cmd == NULL) panic("callout for unknown command '%s'", name); /* Check whether we have been compiled or are being interpreted */ if (stackPopINT(vm->pStack)) { /* * Get parameters from stack, in the format: * an un ... a2 u2 a1 u1 n -- * Where n is the number of strings, a/u are pairs of * address/size for strings, and they will be concatenated * in LIFO order. */ nstrings = stackPopINT(vm->pStack); for (i = 0, len = 0; i < nstrings; i++) len += stackFetch(vm->pStack, i * 2).i + 1; line = malloc(strlen(name) + len + 1); strcpy(line, name); if (nstrings) for (i = 0; i < nstrings; i++) { len = stackPopINT(vm->pStack); cp = stackPopPtr(vm->pStack); strcat(line, " "); strncat(line, cp, len); } } else { /* Get remainder of invocation */ tail = vmGetInBuf(vm); for (cp = tail, len = 0; cp != vm->tib.end && *cp != 0 && *cp != '\n'; cp++, len++) ; line = malloc(strlen(name) + len + 2); strcpy(line, name); if (len > 0) { strcat(line, " "); strncat(line, tail, len); vmUpdateTib(vm, tail + len); } } - DEBUG("cmd '%s'", line); + DPRINTF("cmd '%s'", line); command_errmsg = command_errbuf; command_errbuf[0] = 0; if (!parse(&argc, &argv, line)) { result = (cmd)(argc, argv); free(argv); } else { result=BF_PARSE; } switch (result) { case CMD_CRIT: printf("%s\n", command_errmsg); command_errmsg = NULL; break; case CMD_FATAL: panic("%s", command_errmsg); } free(line); /* * If there was error during nested ficlExec(), we may no longer have * valid environment to return. Throw all exceptions from here. */ if (result != CMD_OK) vmThrow(vm, result); /* This is going to be thrown!!! */ stackPushINT(vm->pStack,result); } /* * Replace a word definition (a builtin command) with another * one that: * * - Throw error results instead of returning them on the stack * - Pass a flag indicating whether the word was compiled or is * being interpreted. * * There is one major problem with builtins that cannot be overcome * in anyway, except by outlawing it. We want builtins to behave * differently depending on whether they have been compiled or they * are being interpreted. Notice that this is *not* the interpreter's * current state. For example: * * : example ls ; immediate * : problem example ; \ "ls" gets executed while compiling * example \ "ls" gets executed while interpreting * * Notice that, though the current state is different in the two * invocations of "example", in both cases "ls" has been * *compiled in*, which is what we really want. * * The problem arises when you tick the builtin. For example: * * : example-1 ['] ls postpone literal ; immediate * : example-2 example-1 execute ; immediate * : problem example-2 ; * example-2 * * We have no way, when we get EXECUTEd, of knowing what our behavior * should be. Thus, our only alternative is to "outlaw" this. See RFI * 0007, and ANS Forth Standard's appendix D, item 6.7 for a related * problem, concerning compile semantics. * * The problem is compounded by the fact that "' builtin CATCH" is valid * and desirable. The only solution is to create an intermediary word. * For example: * * : my-ls ls ; * : example ['] my-ls catch ; * * So, with the below implementation, here is a summary of the behavior * of builtins: * * ls -l \ "interpret" behavior, ie, * \ takes parameters from TIB * : ex-1 s" -l" 1 ls ; \ "compile" behavior, ie, * \ takes parameters from the stack * : ex-2 ['] ls catch ; immediate \ undefined behavior * : ex-3 ['] ls catch ; \ undefined behavior * ex-2 ex-3 \ "interpret" behavior, * \ catch works * : ex-4 ex-2 ; \ "compile" behavior, * \ catch does not work * : ex-5 ex-3 ; immediate \ same as ex-2 * : ex-6 ex-3 ; \ same as ex-3 * : ex-7 ['] ex-1 catch ; \ "compile" behavior, * \ catch works * : ex-8 postpone ls ; immediate \ same as ex-2 * : ex-9 postpone ls ; \ same as ex-3 * * As the definition below is particularly tricky, and it's side effects * must be well understood by those playing with it, I'll be heavy on * the comments. * * (if you edit this definition, pay attention to trailing spaces after * each word -- I warned you! :-) ) */ #define BUILTIN_CONSTRUCTOR \ ": builtin: " \ ">in @ " /* save the tib index pointer */ \ "' " /* get next word's xt */ \ "swap >in ! " /* point again to next word */ \ "create " /* create a new definition of the next word */ \ ", " /* save previous definition's xt */ \ "immediate " /* make the new definition an immediate word */ \ \ "does> " /* Now, the *new* definition will: */ \ "state @ if " /* if in compiling state: */ \ "1 postpone literal " /* pass 1 flag to indicate compile */ \ "@ compile, " /* compile in previous definition */ \ "postpone throw " /* throw stack-returned result */ \ "else " /* if in interpreting state: */ \ "0 swap " /* pass 0 flag to indicate interpret */ \ "@ execute " /* call previous definition */ \ "throw " /* throw stack-returned result */ \ "then ; " /* * Initialise the Forth interpreter, create all our commands as words. */ void bf_init(void) { struct bootblk_command **cmdp; char create_buf[41]; /* 31 characters-long builtins */ int fd; bf_sys = ficlInitSystem(BF_DICTSIZE); bf_vm = ficlNewVM(bf_sys); /* Put all private definitions in a "builtins" vocabulary */ ficlExec(bf_vm, "vocabulary builtins also builtins definitions"); /* Builtin constructor word */ ficlExec(bf_vm, BUILTIN_CONSTRUCTOR); /* make all commands appear as Forth words */ SET_FOREACH(cmdp, Xcommand_set) { ficlBuild(bf_sys, (char *)(*cmdp)->c_name, bf_command, FW_DEFAULT); ficlExec(bf_vm, "forth definitions builtins"); sprintf(create_buf, "builtin: %s", (*cmdp)->c_name); ficlExec(bf_vm, create_buf); ficlExec(bf_vm, "builtins definitions"); } ficlExec(bf_vm, "only forth definitions"); /* Export some version numbers so that code can detect the loader/host version */ ficlSetEnv(bf_sys, "FreeBSD_version", __FreeBSD_version); ficlSetEnv(bf_sys, "loader_version", bootprog_rev); /* try to load and run init file if present */ if ((fd = open("/boot/boot.4th", O_RDONLY)) != -1) { (void)ficlExecFD(bf_vm, fd); close(fd); } } /* * Feed a line of user input to the Forth interpreter */ static int bf_run(const char *line) { int result; /* * ficl would require extensive changes to accept a const char * * interface. Instead, cast it away here and hope for the best. * We know at the present time the caller for us in the boot * forth loader can tolerate the string being modified because * the string is passed in here and then not touched again. */ result = ficlExec(bf_vm, __DECONST(char *, line)); - DEBUG("ficlExec '%s' = %d", line, result); + DPRINTF("ficlExec '%s' = %d", line, result); switch (result) { case VM_OUTOFTEXT: case VM_ABORTQ: case VM_QUIT: case VM_ERREXIT: break; case VM_USEREXIT: printf("No where to leave to!\n"); break; case VM_ABORT: printf("Aborted!\n"); break; case BF_PARSE: printf("Parse error!\n"); break; default: if (command_errmsg != NULL) { printf("%s\n", command_errmsg); command_errmsg = NULL; } } if (result == VM_USEREXIT) panic("interpreter exit"); setenv("interpret", bf_vm->state ? "" : "OK", 1); return (result); } void interp_init(void) { setenv("script.lang", "forth", 1); bf_init(); /* Read our default configuration. */ interp_include("/boot/loader.rc"); } int interp_run(const char *input) { bf_vm->sourceID.i = 0; return bf_run(input); } /* * Header prepended to each line. The text immediately follows the header. * We try to make this short in order to save memory -- the loader has * limited memory available, and some of the forth files are very long. */ struct includeline { struct includeline *next; char text[0]; }; int interp_include(const char *filename) { struct includeline *script, *se, *sp; char input[256]; /* big enough? */ int res; char *cp; int prevsrcid, fd, line; if (((fd = open(filename, O_RDONLY)) == -1)) { snprintf(command_errbuf, sizeof(command_errbuf), "can't open '%s': %s", filename, strerror(errno)); return(CMD_ERROR); } #ifdef LOADER_VERIEXEC if (verify_file(fd, filename, 0, VE_GUESS) < 0) { close(fd); sprintf(command_errbuf,"can't verify '%s'", filename); return(CMD_ERROR); } #endif /* * Read the script into memory. */ script = se = NULL; line = 0; while (fgetstr(input, sizeof(input), fd) >= 0) { line++; cp = input; /* Allocate script line structure and copy line, flags */ if (*cp == '\0') continue; /* ignore empty line, save memory */ sp = malloc(sizeof(struct includeline) + strlen(cp) + 1); /* On malloc failure (it happens!), free as much as possible and exit */ if (sp == NULL) { while (script != NULL) { se = script; script = script->next; free(se); } snprintf(command_errbuf, sizeof(command_errbuf), "file '%s' line %d: memory allocation failure - aborting", filename, line); close(fd); return (CMD_ERROR); } strcpy(sp->text, cp); sp->next = NULL; if (script == NULL) { script = sp; } else { se->next = sp; } se = sp; } close(fd); /* * Execute the script */ prevsrcid = bf_vm->sourceID.i; bf_vm->sourceID.i = fd; res = CMD_OK; for (sp = script; sp != NULL; sp = sp->next) { res = bf_run(sp->text); if (res != VM_OUTOFTEXT) { snprintf(command_errbuf, sizeof(command_errbuf), "Error while including %s, in the line:\n%s", filename, sp->text); res = CMD_ERROR; break; } else res = CMD_OK; } bf_vm->sourceID.i = prevsrcid; while (script != NULL) { se = script; script = script->next; free(se); } return(res); } Index: head/stand/common/part.c =================================================================== --- head/stand/common/part.c (revision 345065) +++ head/stand/common/part.c (revision 345066) @@ -1,949 +1,949 @@ /*- * Copyright (c) 2012 Andrey V. Elsukov * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef PART_DEBUG -#define DEBUG(fmt, args...) printf("%s: " fmt "\n", __func__, ## args) +#define DPRINTF(fmt, args...) printf("%s: " fmt "\n", __func__, ## args) #else -#define DEBUG(fmt, args...) +#define DPRINTF(fmt, args...) #endif #ifdef LOADER_GPT_SUPPORT #define MAXTBLSZ 64 static const uuid_t gpt_uuid_unused = GPT_ENT_TYPE_UNUSED; static const uuid_t gpt_uuid_ms_basic_data = GPT_ENT_TYPE_MS_BASIC_DATA; static const uuid_t gpt_uuid_freebsd_ufs = GPT_ENT_TYPE_FREEBSD_UFS; static const uuid_t gpt_uuid_efi = GPT_ENT_TYPE_EFI; static const uuid_t gpt_uuid_freebsd = GPT_ENT_TYPE_FREEBSD; static const uuid_t gpt_uuid_freebsd_boot = GPT_ENT_TYPE_FREEBSD_BOOT; static const uuid_t gpt_uuid_freebsd_nandfs = GPT_ENT_TYPE_FREEBSD_NANDFS; static const uuid_t gpt_uuid_freebsd_swap = GPT_ENT_TYPE_FREEBSD_SWAP; static const uuid_t gpt_uuid_freebsd_zfs = GPT_ENT_TYPE_FREEBSD_ZFS; static const uuid_t gpt_uuid_freebsd_vinum = GPT_ENT_TYPE_FREEBSD_VINUM; #endif struct pentry { struct ptable_entry part; uint64_t flags; union { uint8_t bsd; uint8_t mbr; uuid_t gpt; uint16_t vtoc8; } type; STAILQ_ENTRY(pentry) entry; }; struct ptable { enum ptable_type type; uint16_t sectorsize; uint64_t sectors; STAILQ_HEAD(, pentry) entries; }; static struct parttypes { enum partition_type type; const char *desc; } ptypes[] = { { PART_UNKNOWN, "Unknown" }, { PART_EFI, "EFI" }, { PART_FREEBSD, "FreeBSD" }, { PART_FREEBSD_BOOT, "FreeBSD boot" }, { PART_FREEBSD_NANDFS, "FreeBSD nandfs" }, { PART_FREEBSD_UFS, "FreeBSD UFS" }, { PART_FREEBSD_ZFS, "FreeBSD ZFS" }, { PART_FREEBSD_SWAP, "FreeBSD swap" }, { PART_FREEBSD_VINUM, "FreeBSD vinum" }, { PART_LINUX, "Linux" }, { PART_LINUX_SWAP, "Linux swap" }, { PART_DOS, "DOS/Windows" }, { PART_ISO9660, "ISO9660" }, }; const char * parttype2str(enum partition_type type) { size_t i; for (i = 0; i < nitems(ptypes); i++) if (ptypes[i].type == type) return (ptypes[i].desc); return (ptypes[0].desc); } #ifdef LOADER_GPT_SUPPORT static void uuid_letoh(uuid_t *uuid) { uuid->time_low = le32toh(uuid->time_low); uuid->time_mid = le16toh(uuid->time_mid); uuid->time_hi_and_version = le16toh(uuid->time_hi_and_version); } static enum partition_type gpt_parttype(uuid_t type) { if (uuid_equal(&type, &gpt_uuid_efi, NULL)) return (PART_EFI); else if (uuid_equal(&type, &gpt_uuid_ms_basic_data, NULL)) return (PART_DOS); else if (uuid_equal(&type, &gpt_uuid_freebsd_boot, NULL)) return (PART_FREEBSD_BOOT); else if (uuid_equal(&type, &gpt_uuid_freebsd_ufs, NULL)) return (PART_FREEBSD_UFS); else if (uuid_equal(&type, &gpt_uuid_freebsd_zfs, NULL)) return (PART_FREEBSD_ZFS); else if (uuid_equal(&type, &gpt_uuid_freebsd_swap, NULL)) return (PART_FREEBSD_SWAP); else if (uuid_equal(&type, &gpt_uuid_freebsd_vinum, NULL)) return (PART_FREEBSD_VINUM); else if (uuid_equal(&type, &gpt_uuid_freebsd_nandfs, NULL)) return (PART_FREEBSD_NANDFS); else if (uuid_equal(&type, &gpt_uuid_freebsd, NULL)) return (PART_FREEBSD); return (PART_UNKNOWN); } static struct gpt_hdr * gpt_checkhdr(struct gpt_hdr *hdr, uint64_t lba_self, uint64_t lba_last, uint16_t sectorsize) { uint32_t sz, crc; if (memcmp(hdr->hdr_sig, GPT_HDR_SIG, sizeof(hdr->hdr_sig)) != 0) { - DEBUG("no GPT signature"); + DPRINTF("no GPT signature"); return (NULL); } sz = le32toh(hdr->hdr_size); if (sz < 92 || sz > sectorsize) { - DEBUG("invalid GPT header size: %d", sz); + DPRINTF("invalid GPT header size: %d", sz); return (NULL); } crc = le32toh(hdr->hdr_crc_self); hdr->hdr_crc_self = 0; if (crc32(hdr, sz) != crc) { - DEBUG("GPT header's CRC doesn't match"); + DPRINTF("GPT header's CRC doesn't match"); return (NULL); } hdr->hdr_crc_self = crc; hdr->hdr_revision = le32toh(hdr->hdr_revision); if (hdr->hdr_revision < GPT_HDR_REVISION) { - DEBUG("unsupported GPT revision %d", hdr->hdr_revision); + DPRINTF("unsupported GPT revision %d", hdr->hdr_revision); return (NULL); } hdr->hdr_lba_self = le64toh(hdr->hdr_lba_self); if (hdr->hdr_lba_self != lba_self) { - DEBUG("self LBA doesn't match"); + DPRINTF("self LBA doesn't match"); return (NULL); } hdr->hdr_lba_alt = le64toh(hdr->hdr_lba_alt); if (hdr->hdr_lba_alt == hdr->hdr_lba_self) { - DEBUG("invalid alternate LBA"); + DPRINTF("invalid alternate LBA"); return (NULL); } hdr->hdr_entries = le32toh(hdr->hdr_entries); hdr->hdr_entsz = le32toh(hdr->hdr_entsz); if (hdr->hdr_entries == 0 || hdr->hdr_entsz < sizeof(struct gpt_ent) || sectorsize % hdr->hdr_entsz != 0) { - DEBUG("invalid entry size or number of entries"); + DPRINTF("invalid entry size or number of entries"); return (NULL); } hdr->hdr_lba_start = le64toh(hdr->hdr_lba_start); hdr->hdr_lba_end = le64toh(hdr->hdr_lba_end); hdr->hdr_lba_table = le64toh(hdr->hdr_lba_table); hdr->hdr_crc_table = le32toh(hdr->hdr_crc_table); uuid_letoh(&hdr->hdr_uuid); return (hdr); } static int gpt_checktbl(const struct gpt_hdr *hdr, uint8_t *tbl, size_t size, uint64_t lba_last) { struct gpt_ent *ent; uint32_t i, cnt; cnt = size / hdr->hdr_entsz; if (hdr->hdr_entries <= cnt) { cnt = hdr->hdr_entries; /* Check CRC only when buffer size is enough for table. */ if (hdr->hdr_crc_table != crc32(tbl, hdr->hdr_entries * hdr->hdr_entsz)) { - DEBUG("GPT table's CRC doesn't match"); + DPRINTF("GPT table's CRC doesn't match"); return (-1); } } for (i = 0; i < cnt; i++) { ent = (struct gpt_ent *)(tbl + i * hdr->hdr_entsz); uuid_letoh(&ent->ent_type); if (uuid_equal(&ent->ent_type, &gpt_uuid_unused, NULL)) continue; ent->ent_lba_start = le64toh(ent->ent_lba_start); ent->ent_lba_end = le64toh(ent->ent_lba_end); } return (0); } static struct ptable * ptable_gptread(struct ptable *table, void *dev, diskread_t dread) { struct pentry *entry; struct gpt_hdr *phdr, hdr; struct gpt_ent *ent; uint8_t *buf, *tbl; uint64_t offset; int pri, sec; size_t size, i; buf = malloc(table->sectorsize); if (buf == NULL) return (NULL); tbl = malloc(table->sectorsize * MAXTBLSZ); if (tbl == NULL) { free(buf); return (NULL); } /* Read the primary GPT header. */ if (dread(dev, buf, 1, 1) != 0) { ptable_close(table); table = NULL; goto out; } pri = sec = 0; /* Check the primary GPT header. */ phdr = gpt_checkhdr((struct gpt_hdr *)buf, 1, table->sectors - 1, table->sectorsize); if (phdr != NULL) { /* Read the primary GPT table. */ size = MIN(MAXTBLSZ, howmany(phdr->hdr_entries * phdr->hdr_entsz, table->sectorsize)); if (dread(dev, tbl, size, phdr->hdr_lba_table) == 0 && gpt_checktbl(phdr, tbl, size * table->sectorsize, table->sectors - 1) == 0) { memcpy(&hdr, phdr, sizeof(hdr)); pri = 1; } } offset = pri ? hdr.hdr_lba_alt: table->sectors - 1; /* Read the backup GPT header. */ if (dread(dev, buf, 1, offset) != 0) phdr = NULL; else phdr = gpt_checkhdr((struct gpt_hdr *)buf, offset, table->sectors - 1, table->sectorsize); if (phdr != NULL) { /* * Compare primary and backup headers. * If they are equal, then we do not need to read backup * table. If they are different, then prefer backup header * and try to read backup table. */ if (pri == 0 || uuid_equal(&hdr.hdr_uuid, &phdr->hdr_uuid, NULL) == 0 || hdr.hdr_revision != phdr->hdr_revision || hdr.hdr_size != phdr->hdr_size || hdr.hdr_lba_start != phdr->hdr_lba_start || hdr.hdr_lba_end != phdr->hdr_lba_end || hdr.hdr_entries != phdr->hdr_entries || hdr.hdr_entsz != phdr->hdr_entsz || hdr.hdr_crc_table != phdr->hdr_crc_table) { /* Read the backup GPT table. */ size = MIN(MAXTBLSZ, howmany(phdr->hdr_entries * phdr->hdr_entsz, table->sectorsize)); if (dread(dev, tbl, size, phdr->hdr_lba_table) == 0 && gpt_checktbl(phdr, tbl, size * table->sectorsize, table->sectors - 1) == 0) { memcpy(&hdr, phdr, sizeof(hdr)); sec = 1; } } } if (pri == 0 && sec == 0) { /* Both primary and backup tables are invalid. */ table->type = PTABLE_NONE; goto out; } - DEBUG("GPT detected"); + DPRINTF("GPT detected"); size = MIN(hdr.hdr_entries * hdr.hdr_entsz, MAXTBLSZ * table->sectorsize); /* * If the disk's sector count is smaller than the sector count recorded * in the disk's GPT table header, set the table->sectors to the value * recorded in GPT tables. This is done to work around buggy firmware * that returns truncated disk sizes. * * Note, this is still not a foolproof way to get disk's size. For * example, an image file can be truncated when copied to smaller media. */ table->sectors = hdr.hdr_lba_alt + 1; for (i = 0; i < size / hdr.hdr_entsz; i++) { ent = (struct gpt_ent *)(tbl + i * hdr.hdr_entsz); if (uuid_equal(&ent->ent_type, &gpt_uuid_unused, NULL)) continue; /* Simple sanity checks. */ if (ent->ent_lba_start < hdr.hdr_lba_start || ent->ent_lba_end > hdr.hdr_lba_end || ent->ent_lba_start > ent->ent_lba_end) continue; entry = malloc(sizeof(*entry)); if (entry == NULL) break; entry->part.start = ent->ent_lba_start; entry->part.end = ent->ent_lba_end; entry->part.index = i + 1; entry->part.type = gpt_parttype(ent->ent_type); entry->flags = le64toh(ent->ent_attr); memcpy(&entry->type.gpt, &ent->ent_type, sizeof(uuid_t)); STAILQ_INSERT_TAIL(&table->entries, entry, entry); - DEBUG("new GPT partition added"); + DPRINTF("new GPT partition added"); } out: free(buf); free(tbl); return (table); } #endif /* LOADER_GPT_SUPPORT */ #ifdef LOADER_MBR_SUPPORT /* We do not need to support too many EBR partitions in the loader */ #define MAXEBRENTRIES 8 static enum partition_type mbr_parttype(uint8_t type) { switch (type) { case DOSPTYP_386BSD: return (PART_FREEBSD); case DOSPTYP_LINSWP: return (PART_LINUX_SWAP); case DOSPTYP_LINUX: return (PART_LINUX); case 0x01: case 0x04: case 0x06: case 0x07: case 0x0b: case 0x0c: case 0x0e: return (PART_DOS); } return (PART_UNKNOWN); } static struct ptable * ptable_ebrread(struct ptable *table, void *dev, diskread_t dread) { struct dos_partition *dp; struct pentry *e1, *entry; uint32_t start, end, offset; u_char *buf; int i, index; STAILQ_FOREACH(e1, &table->entries, entry) { if (e1->type.mbr == DOSPTYP_EXT || e1->type.mbr == DOSPTYP_EXTLBA) break; } if (e1 == NULL) return (table); index = 5; offset = e1->part.start; buf = malloc(table->sectorsize); if (buf == NULL) return (table); - DEBUG("EBR detected"); + DPRINTF("EBR detected"); for (i = 0; i < MAXEBRENTRIES; i++) { #if 0 /* Some BIOSes return an incorrect number of sectors */ if (offset >= table->sectors) break; #endif if (dread(dev, buf, 1, offset) != 0) break; dp = (struct dos_partition *)(buf + DOSPARTOFF); if (dp[0].dp_typ == 0) break; start = le32toh(dp[0].dp_start); if (dp[0].dp_typ == DOSPTYP_EXT && dp[1].dp_typ == 0) { offset = e1->part.start + start; continue; } end = le32toh(dp[0].dp_size); entry = malloc(sizeof(*entry)); if (entry == NULL) break; entry->part.start = offset + start; entry->part.end = entry->part.start + end - 1; entry->part.index = index++; entry->part.type = mbr_parttype(dp[0].dp_typ); entry->flags = dp[0].dp_flag; entry->type.mbr = dp[0].dp_typ; STAILQ_INSERT_TAIL(&table->entries, entry, entry); - DEBUG("new EBR partition added"); + DPRINTF("new EBR partition added"); if (dp[1].dp_typ == 0) break; offset = e1->part.start + le32toh(dp[1].dp_start); } free(buf); return (table); } #endif /* LOADER_MBR_SUPPORT */ static enum partition_type bsd_parttype(uint8_t type) { switch (type) { case FS_NANDFS: return (PART_FREEBSD_NANDFS); case FS_SWAP: return (PART_FREEBSD_SWAP); case FS_BSDFFS: return (PART_FREEBSD_UFS); case FS_VINUM: return (PART_FREEBSD_VINUM); case FS_ZFS: return (PART_FREEBSD_ZFS); } return (PART_UNKNOWN); } static struct ptable * ptable_bsdread(struct ptable *table, void *dev, diskread_t dread) { struct disklabel *dl; struct partition *part; struct pentry *entry; uint8_t *buf; uint32_t raw_offset; int i; if (table->sectorsize < sizeof(struct disklabel)) { - DEBUG("Too small sectorsize"); + DPRINTF("Too small sectorsize"); return (table); } buf = malloc(table->sectorsize); if (buf == NULL) return (table); if (dread(dev, buf, 1, 1) != 0) { - DEBUG("read failed"); + DPRINTF("read failed"); ptable_close(table); table = NULL; goto out; } dl = (struct disklabel *)buf; if (le32toh(dl->d_magic) != DISKMAGIC && le32toh(dl->d_magic2) != DISKMAGIC) goto out; if (le32toh(dl->d_secsize) != table->sectorsize) { - DEBUG("unsupported sector size"); + DPRINTF("unsupported sector size"); goto out; } dl->d_npartitions = le16toh(dl->d_npartitions); if (dl->d_npartitions > 20 || dl->d_npartitions < 8) { - DEBUG("invalid number of partitions"); + DPRINTF("invalid number of partitions"); goto out; } - DEBUG("BSD detected"); + DPRINTF("BSD detected"); part = &dl->d_partitions[0]; raw_offset = le32toh(part[RAW_PART].p_offset); for (i = 0; i < dl->d_npartitions; i++, part++) { if (i == RAW_PART) continue; if (part->p_size == 0) continue; entry = malloc(sizeof(*entry)); if (entry == NULL) break; entry->part.start = le32toh(part->p_offset) - raw_offset; entry->part.end = entry->part.start + le32toh(part->p_size) - 1; entry->part.type = bsd_parttype(part->p_fstype); entry->part.index = i; /* starts from zero */ entry->type.bsd = part->p_fstype; STAILQ_INSERT_TAIL(&table->entries, entry, entry); - DEBUG("new BSD partition added"); + DPRINTF("new BSD partition added"); } table->type = PTABLE_BSD; out: free(buf); return (table); } #ifdef LOADER_VTOC8_SUPPORT static enum partition_type vtoc8_parttype(uint16_t type) { switch (type) { case VTOC_TAG_FREEBSD_NANDFS: return (PART_FREEBSD_NANDFS); case VTOC_TAG_FREEBSD_SWAP: return (PART_FREEBSD_SWAP); case VTOC_TAG_FREEBSD_UFS: return (PART_FREEBSD_UFS); case VTOC_TAG_FREEBSD_VINUM: return (PART_FREEBSD_VINUM); case VTOC_TAG_FREEBSD_ZFS: return (PART_FREEBSD_ZFS); } return (PART_UNKNOWN); } static struct ptable * ptable_vtoc8read(struct ptable *table, void *dev, diskread_t dread) { struct pentry *entry; struct vtoc8 *dl; uint8_t *buf; uint16_t sum, heads, sectors; int i; if (table->sectorsize != sizeof(struct vtoc8)) return (table); buf = malloc(table->sectorsize); if (buf == NULL) return (table); if (dread(dev, buf, 1, 0) != 0) { - DEBUG("read failed"); + DPRINTF("read failed"); ptable_close(table); table = NULL; goto out; } dl = (struct vtoc8 *)buf; /* Check the sum */ for (i = sum = 0; i < sizeof(struct vtoc8); i += sizeof(sum)) sum ^= be16dec(buf + i); if (sum != 0) { - DEBUG("incorrect checksum"); + DPRINTF("incorrect checksum"); goto out; } if (be16toh(dl->nparts) != VTOC8_NPARTS) { - DEBUG("invalid number of entries"); + DPRINTF("invalid number of entries"); goto out; } sectors = be16toh(dl->nsecs); heads = be16toh(dl->nheads); if (sectors * heads == 0) { - DEBUG("invalid geometry"); + DPRINTF("invalid geometry"); goto out; } - DEBUG("VTOC8 detected"); + DPRINTF("VTOC8 detected"); for (i = 0; i < VTOC8_NPARTS; i++) { dl->part[i].tag = be16toh(dl->part[i].tag); if (i == VTOC_RAW_PART || dl->part[i].tag == VTOC_TAG_UNASSIGNED) continue; entry = malloc(sizeof(*entry)); if (entry == NULL) break; entry->part.start = be32toh(dl->map[i].cyl) * heads * sectors; entry->part.end = be32toh(dl->map[i].nblks) + entry->part.start - 1; entry->part.type = vtoc8_parttype(dl->part[i].tag); entry->part.index = i; /* starts from zero */ entry->type.vtoc8 = dl->part[i].tag; STAILQ_INSERT_TAIL(&table->entries, entry, entry); - DEBUG("new VTOC8 partition added"); + DPRINTF("new VTOC8 partition added"); } table->type = PTABLE_VTOC8; out: free(buf); return (table); } #endif /* LOADER_VTOC8_SUPPORT */ #define cdb2devb(bno) ((bno) * ISO_DEFAULT_BLOCK_SIZE / table->sectorsize) static struct ptable * ptable_iso9660read(struct ptable *table, void *dev, diskread_t dread) { uint8_t *buf; struct iso_primary_descriptor *vd; struct pentry *entry; buf = malloc(table->sectorsize); if (buf == NULL) return (table); if (dread(dev, buf, 1, cdb2devb(16)) != 0) { - DEBUG("read failed"); + DPRINTF("read failed"); ptable_close(table); table = NULL; goto out; } vd = (struct iso_primary_descriptor *)buf; if (bcmp(vd->id, ISO_STANDARD_ID, sizeof vd->id) != 0) goto out; entry = malloc(sizeof(*entry)); if (entry == NULL) goto out; entry->part.start = 0; entry->part.end = table->sectors; entry->part.type = PART_ISO9660; entry->part.index = 0; STAILQ_INSERT_TAIL(&table->entries, entry, entry); table->type = PTABLE_ISO9660; out: free(buf); return (table); } struct ptable * ptable_open(void *dev, uint64_t sectors, uint16_t sectorsize, diskread_t *dread) { struct dos_partition *dp; struct ptable *table; uint8_t *buf; int i, count; #ifdef LOADER_MBR_SUPPORT struct pentry *entry; uint32_t start, end; int has_ext; #endif table = NULL; buf = malloc(sectorsize); if (buf == NULL) return (NULL); /* First, read the MBR. */ if (dread(dev, buf, 1, DOSBBSECTOR) != 0) { - DEBUG("read failed"); + DPRINTF("read failed"); goto out; } table = malloc(sizeof(*table)); if (table == NULL) goto out; table->sectors = sectors; table->sectorsize = sectorsize; table->type = PTABLE_NONE; STAILQ_INIT(&table->entries); if (ptable_iso9660read(table, dev, dread) == NULL) { /* Read error. */ table = NULL; goto out; } else if (table->type == PTABLE_ISO9660) goto out; #ifdef LOADER_VTOC8_SUPPORT if (be16dec(buf + offsetof(struct vtoc8, magic)) == VTOC_MAGIC) { if (ptable_vtoc8read(table, dev, dread) == NULL) { /* Read error. */ table = NULL; goto out; } else if (table->type == PTABLE_VTOC8) goto out; } #endif /* Check the BSD label. */ if (ptable_bsdread(table, dev, dread) == NULL) { /* Read error. */ table = NULL; goto out; } else if (table->type == PTABLE_BSD) goto out; #if defined(LOADER_GPT_SUPPORT) || defined(LOADER_MBR_SUPPORT) /* Check the MBR magic. */ if (buf[DOSMAGICOFFSET] != 0x55 || buf[DOSMAGICOFFSET + 1] != 0xaa) { - DEBUG("magic sequence not found"); + DPRINTF("magic sequence not found"); #if defined(LOADER_GPT_SUPPORT) /* There is no PMBR, check that we have backup GPT */ table->type = PTABLE_GPT; table = ptable_gptread(table, dev, dread); #endif goto out; } /* Check that we have PMBR. Also do some validation. */ dp = (struct dos_partition *)(buf + DOSPARTOFF); for (i = 0, count = 0; i < NDOSPART; i++) { if (dp[i].dp_flag != 0 && dp[i].dp_flag != 0x80) { - DEBUG("invalid partition flag %x", dp[i].dp_flag); + DPRINTF("invalid partition flag %x", dp[i].dp_flag); goto out; } #ifdef LOADER_GPT_SUPPORT if (dp[i].dp_typ == DOSPTYP_PMBR) { table->type = PTABLE_GPT; - DEBUG("PMBR detected"); + DPRINTF("PMBR detected"); } #endif if (dp[i].dp_typ != 0) count++; } /* Do we have some invalid values? */ if (table->type == PTABLE_GPT && count > 1) { if (dp[1].dp_typ != DOSPTYP_HFS) { table->type = PTABLE_NONE; - DEBUG("Incorrect PMBR, ignore it"); + DPRINTF("Incorrect PMBR, ignore it"); } else { - DEBUG("Bootcamp detected"); + DPRINTF("Bootcamp detected"); } } #ifdef LOADER_GPT_SUPPORT if (table->type == PTABLE_GPT) { table = ptable_gptread(table, dev, dread); goto out; } #endif #ifdef LOADER_MBR_SUPPORT /* Read MBR. */ - DEBUG("MBR detected"); + DPRINTF("MBR detected"); table->type = PTABLE_MBR; for (i = has_ext = 0; i < NDOSPART; i++) { if (dp[i].dp_typ == 0) continue; start = le32dec(&(dp[i].dp_start)); end = le32dec(&(dp[i].dp_size)); if (start == 0 || end == 0) continue; #if 0 /* Some BIOSes return an incorrect number of sectors */ if (start + end - 1 >= sectors) continue; /* XXX: ignore */ #endif if (dp[i].dp_typ == DOSPTYP_EXT || dp[i].dp_typ == DOSPTYP_EXTLBA) has_ext = 1; entry = malloc(sizeof(*entry)); if (entry == NULL) break; entry->part.start = start; entry->part.end = start + end - 1; entry->part.index = i + 1; entry->part.type = mbr_parttype(dp[i].dp_typ); entry->flags = dp[i].dp_flag; entry->type.mbr = dp[i].dp_typ; STAILQ_INSERT_TAIL(&table->entries, entry, entry); - DEBUG("new MBR partition added"); + DPRINTF("new MBR partition added"); } if (has_ext) { table = ptable_ebrread(table, dev, dread); /* FALLTHROUGH */ } #endif /* LOADER_MBR_SUPPORT */ #endif /* LOADER_MBR_SUPPORT || LOADER_GPT_SUPPORT */ out: free(buf); return (table); } void ptable_close(struct ptable *table) { struct pentry *entry; if (table == NULL) return; while (!STAILQ_EMPTY(&table->entries)) { entry = STAILQ_FIRST(&table->entries); STAILQ_REMOVE_HEAD(&table->entries, entry); free(entry); } free(table); } enum ptable_type ptable_gettype(const struct ptable *table) { return (table->type); } int ptable_getsize(const struct ptable *table, uint64_t *sizep) { uint64_t tmp = table->sectors * table->sectorsize; if (tmp < table->sectors) return (EOVERFLOW); if (sizep != NULL) *sizep = tmp; return (0); } int ptable_getpart(const struct ptable *table, struct ptable_entry *part, int index) { struct pentry *entry; if (part == NULL || table == NULL) return (EINVAL); STAILQ_FOREACH(entry, &table->entries, entry) { if (entry->part.index != index) continue; memcpy(part, &entry->part, sizeof(*part)); return (0); } return (ENOENT); } /* * Search for a slice with the following preferences: * * 1: Active FreeBSD slice * 2: Non-active FreeBSD slice * 3: Active Linux slice * 4: non-active Linux slice * 5: Active FAT/FAT32 slice * 6: non-active FAT/FAT32 slice */ #define PREF_RAWDISK 0 #define PREF_FBSD_ACT 1 #define PREF_FBSD 2 #define PREF_LINUX_ACT 3 #define PREF_LINUX 4 #define PREF_DOS_ACT 5 #define PREF_DOS 6 #define PREF_NONE 7 int ptable_getbestpart(const struct ptable *table, struct ptable_entry *part) { struct pentry *entry, *best; int pref, preflevel; if (part == NULL || table == NULL) return (EINVAL); best = NULL; preflevel = pref = PREF_NONE; STAILQ_FOREACH(entry, &table->entries, entry) { #ifdef LOADER_MBR_SUPPORT if (table->type == PTABLE_MBR) { switch (entry->type.mbr) { case DOSPTYP_386BSD: pref = entry->flags & 0x80 ? PREF_FBSD_ACT: PREF_FBSD; break; case DOSPTYP_LINUX: pref = entry->flags & 0x80 ? PREF_LINUX_ACT: PREF_LINUX; break; case 0x01: /* DOS/Windows */ case 0x04: case 0x06: case 0x0c: case 0x0e: case DOSPTYP_FAT32: pref = entry->flags & 0x80 ? PREF_DOS_ACT: PREF_DOS; break; default: pref = PREF_NONE; } } #endif /* LOADER_MBR_SUPPORT */ #ifdef LOADER_GPT_SUPPORT if (table->type == PTABLE_GPT) { if (entry->part.type == PART_DOS) pref = PREF_DOS; else if (entry->part.type == PART_FREEBSD_UFS || entry->part.type == PART_FREEBSD_ZFS) pref = PREF_FBSD; else pref = PREF_NONE; } #endif /* LOADER_GPT_SUPPORT */ if (pref < preflevel) { preflevel = pref; best = entry; } } if (best != NULL) { memcpy(part, &best->part, sizeof(*part)); return (0); } return (ENOENT); } int ptable_iterate(const struct ptable *table, void *arg, ptable_iterate_t *iter) { struct pentry *entry; char name[32]; int ret = 0; name[0] = '\0'; STAILQ_FOREACH(entry, &table->entries, entry) { #ifdef LOADER_MBR_SUPPORT if (table->type == PTABLE_MBR) sprintf(name, "s%d", entry->part.index); else #endif #ifdef LOADER_GPT_SUPPORT if (table->type == PTABLE_GPT) sprintf(name, "p%d", entry->part.index); else #endif #ifdef LOADER_VTOC8_SUPPORT if (table->type == PTABLE_VTOC8) sprintf(name, "%c", (uint8_t) 'a' + entry->part.index); else #endif if (table->type == PTABLE_BSD) sprintf(name, "%c", (uint8_t) 'a' + entry->part.index); if ((ret = iter(arg, name, &entry->part)) != 0) return (ret); } return (ret); } Index: head/stand/userboot/userboot/elf64_freebsd.c =================================================================== --- head/stand/userboot/userboot/elf64_freebsd.c (revision 345065) +++ head/stand/userboot/userboot/elf64_freebsd.c (revision 345066) @@ -1,172 +1,175 @@ /*- * Copyright (c) 1998 Michael Smith * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #define __ELF_WORD_SIZE 64 #include #include #include +#ifdef DEBUG +#include +#endif #include #include #include #include #include "bootstrap.h" #include "libuserboot.h" static int elf64_exec(struct preloaded_file *amp); static int elf64_obj_exec(struct preloaded_file *amp); struct file_format amd64_elf = { elf64_loadfile, elf64_exec }; struct file_format amd64_elf_obj = { elf64_obj_loadfile, elf64_obj_exec }; #define MSR_EFER 0xc0000080 #define EFER_LME 0x00000100 #define EFER_LMA 0x00000400 /* Long mode active (R) */ #define CR4_PAE 0x00000020 #define CR4_VMXE (1UL << 13) #define CR4_PSE 0x00000010 #define CR0_PG 0x80000000 #define CR0_PE 0x00000001 /* Protected mode Enable */ #define CR0_NE 0x00000020 /* Numeric Error enable (EX16 vs IRQ13) */ #define PG_V 0x001 #define PG_RW 0x002 #define PG_U 0x004 #define PG_PS 0x080 typedef uint64_t p4_entry_t; typedef uint64_t p3_entry_t; typedef uint64_t p2_entry_t; #define GUEST_NULL_SEL 0 #define GUEST_CODE_SEL 1 #define GUEST_DATA_SEL 2 #define GUEST_GDTR_LIMIT (3 * 8 - 1) static void setup_freebsd_gdt(uint64_t *gdtr) { gdtr[GUEST_NULL_SEL] = 0; gdtr[GUEST_CODE_SEL] = 0x0020980000000000; gdtr[GUEST_DATA_SEL] = 0x0000900000000000; } /* * There is an ELF kernel and one or more ELF modules loaded. * We wish to start executing the kernel image, so make such * preparations as are required, and do so. */ static int elf64_exec(struct preloaded_file *fp) { struct file_metadata *md; Elf_Ehdr *ehdr; vm_offset_t modulep, kernend; int err; int i; uint32_t stack[1024]; p4_entry_t PT4[512]; p3_entry_t PT3[512]; p2_entry_t PT2[512]; uint64_t gdtr[3]; if ((md = file_findmetadata(fp, MODINFOMD_ELFHDR)) == NULL) return(EFTYPE); ehdr = (Elf_Ehdr *)&(md->md_data); err = bi_load64(fp->f_args, &modulep, &kernend); if (err != 0) return(err); bzero(PT4, PAGE_SIZE); bzero(PT3, PAGE_SIZE); bzero(PT2, PAGE_SIZE); /* * Build a scratch stack at physical 0x1000, page tables: * PT4 at 0x2000, * PT3 at 0x3000, * PT2 at 0x4000, * gdtr at 0x5000, */ /* * This is kinda brutal, but every single 1GB VM memory segment * points to the same first 1GB of physical memory. But it is * more than adequate. */ for (i = 0; i < 512; i++) { /* Each slot of the level 4 pages points to the same level 3 page */ PT4[i] = (p4_entry_t) 0x3000; PT4[i] |= PG_V | PG_RW | PG_U; /* Each slot of the level 3 pages points to the same level 2 page */ PT3[i] = (p3_entry_t) 0x4000; PT3[i] |= PG_V | PG_RW | PG_U; /* The level 2 page slots are mapped with 2MB pages for 1GB. */ PT2[i] = i * (2 * 1024 * 1024); PT2[i] |= PG_V | PG_RW | PG_PS | PG_U; } #ifdef DEBUG - printf("Start @ %#llx ...\n", ehdr->e_entry); + printf("Start @ %#"PRIx64" ...\n", ehdr->e_entry); #endif dev_cleanup(); stack[0] = 0; /* return address */ stack[1] = modulep; stack[2] = kernend; CALLBACK(copyin, stack, 0x1000, sizeof(stack)); CALLBACK(copyin, PT4, 0x2000, sizeof(PT4)); CALLBACK(copyin, PT3, 0x3000, sizeof(PT3)); CALLBACK(copyin, PT2, 0x4000, sizeof(PT2)); CALLBACK(setreg, 4, 0x1000); CALLBACK(setmsr, MSR_EFER, EFER_LMA | EFER_LME); CALLBACK(setcr, 4, CR4_PAE | CR4_VMXE); CALLBACK(setcr, 3, 0x2000); CALLBACK(setcr, 0, CR0_PG | CR0_PE | CR0_NE); setup_freebsd_gdt(gdtr); CALLBACK(copyin, gdtr, 0x5000, sizeof(gdtr)); CALLBACK(setgdt, 0x5000, sizeof(gdtr)); CALLBACK(exec, ehdr->e_entry); panic("exec returned"); } static int elf64_obj_exec(struct preloaded_file *fp) { return (EFTYPE); }