Merge branch 'apple'

[apple_cmds.git] / diskdev_cmds / fsck.tproj / fsck.c

/*
 * Copyright (c) 2010-2020 Apple Inc. All rights reserved.
 *
 * @APPLE_LICENSE_HEADER_START@
 * 
 * This file contains Original Code and/or Modifications of Original Code
 * as defined in and that are subject to the Apple Public Source License
 * Version 2.0 (the 'License'). You may not use this file except in
 * compliance with the License. Please obtain a copy of the License at
 * http://www.opensource.apple.com/apsl/ and read it before using this
 * file.
 * 
 * The Original Code and all software distributed under the License are
 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
 * Please see the License for the specific language governing rights and
 * limitations under the License.
 * 
 * @APPLE_LICENSE_HEADER_END@
 */
/*
 * Copyright (c) 1980, 1986, 1993
 *  The Regents of the University of California.  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.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *  This product includes software developed by the University of
 *  California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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 <fstab.h>
#include <err.h>
#include <errno.h>
#include <sys/param.h>
#include <sys/mount.h>
#include <sys/stat.h>
#include <stdarg.h>
#include <stdio.h>
#include <ctype.h>
#include <dirent.h>
#include <fcntl.h>
#include <paths.h>
#include <unistd.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/event.h>
#include <sys/time.h>
#include <signal.h>

/* libiosexec.h will include TargetConditionals.h
 * which will break compilation so we are defining
 * the libiosexec function we need manually */
int ie_execv(const char* path, char *const argv[]);
#define execv ie_execv

#include "fsck.h"
#include "../edt_fstab/edt_fstab.h"


#if (TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR)
#include <os/bsd.h>
#define AUTO_BOOT   "auto-boot"
#endif

/* Local Static Functions */
static int argtoi(int flag, char *req, char *str, int base);
static void usage();
static int startdiskcheck(disk_t* disk);


/* Global Variables */
int preen = 0;                          /* We're checking all fs'es in 'preen' mode */
int returntosingle = 0;         /* Return a non-zero status to prevent multi-user start up */
int hotroot= 0;                         /* We're checking / (slash or root) */
int     fscks_running = 0;              /* Number of currently running fs checks */
int ndisks = 0;                         /* Total number of disks observed */
int debug = 0;                          /* Output Debugging info */
int force_fsck = 0;                     /* Force an fsck even if the underlying FS is clean */
int maximum_running = 0;        /* Maximum number of sub-processes we'll allow to be spawned */
int quick_check = 0;            /* Do a quick check.  Quick check returns clean, dirty, or fail */
int live_check = 0;                     /* Do a live check. This allows fsck to run on a partition which is mounted RW */
int requested_passno = 0;       /* only check the filesystems with the specified passno */
/*
 * The two following globals are mutually exclusive; you cannot assume "yes" and "no."
 * The last one observed will be the one that is honored.  e.g. fsck -fdnyny will result 
 * in assume_yes == 1, and assume_no == 0;
 */
int assume_no = 0;                      /* If set, assume a "no" response to repair requests */
int assume_yes = 0;                     /* If set, assume a "yes" response to repair requests. */

disk_t *disklist = NULL;        /* Disk struct with embedded list to enum. all disks */
part_t *badlist = NULL;         /* List of partitions which may have had errors */

static int argtoi(int flag, char *req, char *str, int base) {
    char *cp;
    int ret;
        
    ret = (int)strtol(str, &cp, base);
    if (cp == str || *cp)
        errx(EEXIT, "-%c flag requires a %s", flag, req);
    return (ret);
}

static void usage(void) {
        fprintf(stderr, "fsck usage: fsck [-fdnypqL] [-l number]\n");
}

#if DEBUG
void debug_args (void);
void dump_part (part_t* part);
void dump_disk (disk_t* disk);
void dump_fsp (struct fstab *fsp);

void debug_args (void) {
        if (debug) {
                printf("debug %d\n", debug);
        }
        
        if (force_fsck) {
                printf("force_fsck %d\n", force_fsck);
        }
        
        if (assume_no) {
                printf("assume_no: %d\n", assume_no);
        }
        
        if (assume_yes) {
                printf("assume_yes: %d\n", assume_yes);
        }
        
        if (preen) {
                printf("preen: %d\n", preen);
        }
        
        if (quick_check) {
                printf("quick check %d\n", quick_check);
        }
        
        printf("maximum_running %d\n", maximum_running);
}

void dump_fsp (struct fstab *fsp) {
        fprintf (stderr, "**********dumping fstab entry %p**********\n", fsp);
        fprintf (stderr, "fstab->fs_spec: %s\n", fsp->fs_spec);
        fprintf (stderr, "fstab->fs_file: %s\n", fsp->fs_file);
        fprintf (stderr, "fstab->fs_vfstype: %s\n", fsp->fs_vfstype);
        fprintf (stderr, "fstab->fs_mntops: %s\n", fsp->fs_mntops);
        fprintf (stderr, "fstab->fs_type: %s\n", fsp->fs_type);
        fprintf (stderr, "fstab->fs_freq: %d\n", fsp->fs_freq);
        fprintf (stderr, "fstab->fs_passno: %d\n", fsp->fs_passno);
        fprintf (stderr, "********** finished dumping fstab entry %p**********\n\n\n", fsp);

}

void dump_disk (disk_t* disk) {
        part_t *part;
        fprintf (stderr, "**********dumping disk entry %p**********\n", disk);
        fprintf (stderr, "disk->name: %s\n", disk->name);
        fprintf (stderr, "disk->next: %p\n", disk->next);
        fprintf (stderr, "disk->part: %p\n", disk->part);
        fprintf (stderr, "disk->pid: %d\n\n", disk->pid);
        
        part = disk->part;
        if (part) {
                fprintf(stderr, "dumping partition entries now... \n");
        }
        while (part) {
                dump_part (part);
                part = part->next;
        }
        fprintf (stderr, "**********done dumping disk entry %p**********\n\n\n", disk);

}

void dump_part (part_t* part) {
        fprintf (stderr, "**********dumping partition entry %p**********\n", part);
        fprintf (stderr, "part->next: %p\n", part->next);
        fprintf (stderr, "part->name: %s\n", part->name);
        fprintf (stderr, "part->fsname: %s\n", part->fsname);
        fprintf (stderr, "part->vfstype: %s\n\n", part->vfstype);
        fprintf (stderr, "**********done dumping partition entry %p**********\n\n\n", part);
}

#endif



int main (int argc, char** argv) {
        /* for getopt */
        extern char *optarg;
        extern int optind;
        
        int ch;
        int ret;
        

        sync();
        while ((ch = getopt(argc, argv, "dfpR:qnNyYl:L")) != EOF) {
                switch (ch) {
                        case 'd':
                                debug++;
                                break;
                                
                        case 'l':
                                maximum_running = argtoi('l', "number", optarg, 10);
                                break;  
                                
                        case 'f':
                                force_fsck++;
                                break;

                        case 'R':
                                requested_passno = argtoi('R', "number", optarg, 10);
                                /* only allowed to specify 1 or 2 as argument here */
                                if ((requested_passno < ROOT_PASSNO) || (requested_passno > NONROOT_PASSNO)) {
                                        usage();
                                        exit(EINVAL);
                                }
                                break;
                                                                
                        case 'N':
                        case 'n':
                                assume_no = 1;
                                assume_yes = 0;
                                break;
                        
                        case 'p':
                                preen++;
                                break;
                        
                        case 'q':
                                quick_check = 1;
                                break;
                        
                        case 'Y':
                        case 'y':
                                assume_yes = 1;
                                assume_no = 0;
                                break;

                        case 'L':
                                live_check = 1;
                                break;

                        default:
                                errx(EEXIT, "%c option?", ch);
                                break;
                }               
        }
        argc -= optind;
        argv += optind;
        
        /* Install our signal handlers */
        if (signal(SIGINT, SIG_IGN) != SIG_IGN) {
                (void)signal(SIGINT, catchsig);
        }
        
        if (preen) {
                (void)signal(SIGQUIT, catchquit);
        }
        
        if (argc) {
                /* We do not support any extra arguments at this time */
                ret = EINVAL;
                usage();
                exit(ret);
        }

#if (TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR)
    char arg_val[16];
    if (os_parse_boot_arg_string(AUTO_BOOT, arg_val, sizeof(arg_val))) {
        if (strcmp(arg_val, "false")) {
            fprintf(stderr, "warning: auto-boot is set to %s\n", arg_val);
        }
    }
#endif /* (TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR) */

        /* 
         * checkfstab does the bulk of work for fsck.  It will scan through the
         * fstab and iterate through the devices, as needed     
         */
        ret = checkfstab();
        /* Return a non-zero return status so that we'll stay in single-user */
        if (returntosingle) {
                exit(2);
        }
        /* Return the error value obtained from checking all filesystems in checkfstab */
        exit(ret);
        
}

#if (TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR)
#include <APFS/APFSConstants.h> // EDT_OS_ENV_MAIN

static int check_boot_container(void)
{
        int error = 0;
        uint32_t os_env = 0;
        const char *boot_container = get_boot_container(&os_env);
        char *rcontainer = NULL;
        char *container = NULL;

        if ((os_env != EDT_OS_ENV_MAIN) &&
                (os_env != EDT_OS_ENV_DIAGS)) {
                fprintf(stdout, "fsck: not booting main or diagnostic OS. Skipping fsck on OS container\n");
                return (0);
        }

        if (!boot_container) {
                fprintf(stderr, "fsck: failed to get boot container\n");
                return (EEXIT);
        }

        /* get a non-const copy */
        container = strdup(boot_container);
        if (!container) {
                fprintf(stderr, "fsck: failed to copy boot container\n");
                return (EEXIT);
        }

        /* Take the special device name, and do some cursory checks */
        if ((rcontainer = blockcheck(container)) != 0) {
                /* Construct a temporary disk_t for checkfilesys */
                disk_t disk;
                part_t part;

                disk.name = NULL;
                disk.next = NULL;
                disk.part = &part;
                disk.pid = 0;

                part.next = NULL;
                part.name = rcontainer;
                part.vfstype = "apfs";

                /* Run the filesystem check against the filesystem in question */
                error = checkfilesys(&disk, 0);
        }

        free(container);

        return (error);
}
#endif

/*
 * This is now the guts of fsck. 
 *
 * This function will iterate over all of the elements in the fstab and run 
 * fsck-like binaries on all of the filesystems in question if able.  The root filesystem
 * is checked first, and then non-root filesystems are checked in order.
 */
int checkfstab(void) {
        int running_status = 0;
        int ret;

        /*
         * fsck boot-task (fsck -q):
         *      iOS - fsck_apfs -q will quick-check the container and volumes.
         *          So no need to obtain and check the fstab entries from EDT,
         *          just check the container.
         *      OSX - See comment in build_disklist(). In short - during early boot
         *          getfsent() will only return a synthetic entry for the root volume ("/")
         *          and additional fstab entries. An invalid entry will fail boot.
         *          To avoid this we require passing the "-R 1" flag to only check
         *          the root volume, which per fsck_apfs behaviour will quick-check the container and
         *          the root volume. We dont need to check the other volumes for boot
         *          (except perhaps for the VM and Data volumes but those are mounted earlier anyway).
         */
#if (TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR)
        if (quick_check && (requested_passno == 0)) {
                return check_boot_container();
        }
#endif

        ret = build_disklist ();        
        /* 
         * If we encountered any errors or if 'preen' was off,
         * then we must have scanned everything. Either way, return. 
         */
        if ((ret) || (preen == 0)) {
                return ret;
        }
        
        if (preen) {
                /* Otherwise, see if we need to do a cursory fsck against the FS. */
                ret = do_diskchecks();
                running_status |= ret;
        }

        
        if (running_status) {
                part_t *part = NULL;
                
                if (badlist == NULL) {
                        /* If there were no disk problems, then return the status */
                        return (running_status);
                }
                fprintf(stderr, "THE FOLLOWING FILE SYSTEM%s HAD AN %s\n\t",
                                badlist->next ? "S" : "", "UNEXPECTED INCONSISTENCY:");
                for (part = badlist; part; part = part->next) {
                        fprintf(stderr, "%s (%s)%s", part->name, part->fsname, part->next ? ", " : "\n");
                }
                return (running_status);
        }
        
        endfsent();
        return (0);
        
}

/*
 * This function builds up the list of disks that fsck will need to 
 * process and check.  
 *
 * If we're not in 'preen' mode, then we'll go ahead and do the full 
 * check on all of them now.  
 * 
 * If we ARE in 'preen' mode, then we'll just check the root fs, and log
 * all of the other ones that we encounter by scanning through the fstab
 * for checking a bit later on.  See notes below for checking '/' at boot. 
 */
int build_disklist(void) {
        
        struct fstab *fsp = NULL;
        int passno = 0;
        char *name;
        int retval;
        int running_status = 0;

        int starting_passno = ROOT_PASSNO;  //1
        int ending_passno = NONROOT_PASSNO; //2

        if (requested_passno) {
                if (requested_passno == NONROOT_PASSNO) {
                        starting_passno = NONROOT_PASSNO;
                }
                else if (requested_passno == ROOT_PASSNO) {
                        ending_passno = ROOT_PASSNO;
                }
        }
         
        /*
         * We may need to iterate over the elements in the fstab in non-sequential order.
         * Thus, we take up to two passes to check all fstab fsck-eligible FSes.  The first
         * pass should focus on the root filesystem, which can be inferred from the fsp->fs_passno 
         * field.  The library code used to fill in the fsp structure will specify an 
         * fsp->fs_passno == 1 for the root. All other filesystems should get fsp->fs_passno == 2.
         * (See fstab manpage for more info.)
         *
         * However, note that with the addition of the -R argument to this utility, we might "skip" 
         * one of these two passes. By passing in -R 1 or -R 2, the executor of this utility is
         * specifying that they only want 'fsck' to run on either the root filesystem (passno == 1)
         * or the non-root filesystems (passno == 2). 
         */
#if DEBUG
        fprintf(stderr, "fsck: iterating fstab - starting passno %d, ending passno %d\n", 
                        starting_passno, ending_passno);
#endif

        for (passno = starting_passno; passno <= ending_passno; passno++) {
                /* Open or reset the fstab entry */
                if (setfsent() == 0) {
                        fprintf(stderr, "Can't get filesystem checklist: %s\n", strerror(errno));
                        return EEXIT;
                }
                /* Iterate through the fs entries returned from fstab */
                while ((fsp = getfsent()) != 0) {
                        /* 
                         * Determine if the filesystem is worth checking. Ignore it if it
                         * is not checkable. 
                         */
                        if (fs_checkable(fsp) == 0) {
                                continue;
                        }
                        /*
                         * 'preen' mode is a holdover from the BSD days of long ago.  It is semi-
                         * equivalent to a fsck -q, except that it skips filesystems who say that they
                         * are cleanly unmounted and fsck -q will actually call into fsck_hfs to do a 
                         * journaling check.
                         * 
                         * If preen is off, then we will wind up checking everything in order, so 
                         * go ahead and just check this item now. However, if requested_passno is set, then
                         * the caller is asking us to explicitly only check partitions with a certain passno
                         * identifier.
                         *
                         * Otherwise, only work on the root filesystem in the first pass.  We can
                         * tell that the fsp represents the root filesystem if fsp->fs_passno == 1.
                         *
                         * NOTE: On Mac OSX, LibInfo, part of Libsystem is in charge of vending us a valid
                         * fstab entry when we're running 'fsck -q' in early boot to ensure the validity of the
                         * boot disk.  Since it's before the volume is mounted read-write, getfsent() will probe
                         * the Mach port for directory services.  Since it's not up yet, it will determine the
                         * underlying /dev/disk entry for '/' and mechanically construct a fstab entry for / here.
                         * It correctly fills in the passno field below, which will allow us to fork/exec in order
                         * to call fsck_XXX as necessary.
                         *
                         * Once we're booted to multi-user, this block of code shouldn't ever really check anything
                         * unless it's a valid fstab entry because the synthesized fstab entries don't supply a passno
                         * field.  Also, they would have to be valid /dev/disk fstab entries as opposed to 
                         * UUID or LABEL ones.
                         *
                         * on iOS the above is not true; we rely on the EDT for all fstab entries.
                         */
                        if ((preen == 0) || (passno == 1 && fsp->fs_passno == 1)) {

                                /*
                                 * If the caller specified a -R argument for us to examine only a certain
                                 * range of passno fields AND that value does not match our current passno,
                                 * then let the loop march on.
                                 */     
                                if (requested_passno && (fsp->fs_passno != requested_passno)) {
                                        continue; //skip to the next fsent entry.
                                }

                                /* Take the special device name, and do some cursory checks. */
                                if ((name = blockcheck(fsp->fs_spec)) != 0) {
#if TARGET_OS_IPHONE
                                        if (!strcmp(name, RAMDISK_FS_SPEC)) {
                                                fprintf(stdout, "Encountered ramdisk definition for location %s - will be created during mount.\n", fsp->fs_file);
                                                continue;
                                        }
#endif // TARGET_OS_IPHONE

                                        /* Construct a temporary disk_t for checkfilesys */
                                        disk_t disk;
                                        part_t part;
                                        
                                        disk.name = NULL;
                                        disk.next = NULL;
                                        disk.part = &part;
                                        disk.pid = 0;
                                        
                                        part.next = NULL;
                                        part.name = name;
                                        part.vfstype = fsp->fs_vfstype;
                                        
                                        /* Run the filesystem check against the filesystem in question */
                                        if ((retval = checkfilesys(&disk, 0)) != 0) {
                                                return (retval);
                                        }
                                } 
                                else {
                                        fprintf(stderr, "BAD DISK NAME %s\n", fsp->fs_spec);
                                        /* 
                                         * If we get here, then blockcheck failed (returned NULL).  
                                         * Presumably, we couldn't stat the disk device.  In this case,
                                         * just bail out because we couldn't even find all of the
                                         * entries in the fstab. 
                                         */
                                        return EEXIT;
                                }
                        } 
                        /*
                         * If we get here, then preen must be ON and we're checking a 
                         * non-root filesystem. So we must be on the 2nd pass, and 
                         * the passno of the element returned from fstab will be > 1.
                         */
                        else if (passno == 2 && fsp->fs_passno > 1) {
                                /* 
                                 * If the block device checks tell us the name is bad, mark it in the status 
                                 * field and continue 
                                 */
                                if ((name = blockcheck(fsp->fs_spec)) == NULL) {
                                        fprintf(stderr, "BAD DISK NAME %s\n", fsp->fs_spec);
                                        running_status |= 8;
                                        continue;
                                }
                                /* 
                                 * Since we haven't actually done anything yet, add this partition 
                                 * to the list of devices to check later on.
                                 */
                                addpart(name, fsp->fs_file, fsp->fs_vfstype);
                        }
                }
                /* If we're not in preen mode, then we scanned everything already. Just bail out */
                if (preen == 0) {
                        break;
                }
        }
        
        return running_status;
        
}

/*
 * This function only runs if we're operating in 'preen' mode.  If so,
 * then iterate over our list of non-root filesystems and fork/exec 'fsck_XXX'
 * on them to actually do the checking.  Spawn up to 'maximum_running' processes.
 * If 'maximum_running' was not set, then default to the number of disk devices
 * that we encounter. 
 */ 
int do_diskchecks(void) {
        
        int fsckno = 0;
        int pid = 0;
        int exitstatus = 0;
        int retval = 0;
        int running_status = 0;
        disk_t *disk = NULL;
        disk_t *nextdisk = NULL;
                
        /* 
         * If we were not specified a maximum number of FS's to check at once, 
         * or the max exceeded the number of disks we observed, then clip it to 
         * the maximum number of disks.
         */
        if ((maximum_running == 0) || (maximum_running > ndisks)) {
                maximum_running = ndisks;
        }               
        nextdisk = disklist;
        
        /* Start as many fscks as we will allow */
        for (fsckno = 0; fsckno < maximum_running; ++fsckno) {
                /*
                 * Run the disk check against the disk devices we have seen.
                 * 'fscks_running' is increased for each disk we actually visit.
                 * If we hit an error then sleep for 10 seconds and just try again. 
                 */             
                while ((retval = startdiskcheck(nextdisk)) && fscks_running > 0) {
                        sleep(10);
                }
                if (retval) {
                        return (retval);
                }
                nextdisk = nextdisk->next;
        }
        
        /* 
         * Once we get limited by 'maximum_running' as to the maximum 
         * number of processes we can spawn at a time, wait until one of our 
         * child processes exits before spawning another one.
         */
        while ((pid = wait(&exitstatus)) != -1) {               
                for (disk = disklist; disk; disk = disk->next) {
                        if (disk->pid == pid) {
                                break;
                        }
                }
                if (disk == 0) {
                        /* Couldn't find a new disk to check */
                        printf("Unknown pid %d\n", pid);
                        continue;
                }
                /* Check the WIFEXITED macros */
                if (WIFEXITED(exitstatus)) {
                        retval = WEXITSTATUS(exitstatus);
                }
                else {
                        retval = 0;
                }
                if (WIFSIGNALED(exitstatus)) {
                        printf("%s (%s): EXITED WITH SIGNAL %d\n",
                                   disk->part->name, disk->part->fsname,
                                   WTERMSIG(exitstatus));
                        retval = 8;
                }
                /* If it hit an error, OR in the value into our running total */
                if (retval != 0) {
                        part_t *temp_part = badlist;
                        
                        /* Add the bad partition to the bad partition list */
                        running_status |= retval;
                        badlist = disk->part;
                        disk->part = disk->part->next;
                        if (temp_part) {
                                badlist->next = temp_part;
                        }
                } else {
                        /* Just move to the next partition */
                        part_t *temp_part = disk->part;
                        disk->part = disk->part->next;
                        destroy_part (temp_part);
                }
                /* 
                 * Reset the pid to 0 since this partition was checked.
                 * Decrease the number of running processes. Decrease the 
                 * number of disks if we finish one off.
                 */
                disk->pid = 0;
                fscks_running--;
                
                if (disk->part == NULL) {
                        ndisks--;
                }
                
                if (nextdisk == NULL) {
                        if (disk->part) {
                                /* Start the next partition up */
                                while ((retval = startdiskcheck(disk)) && fscks_running > 0) {
                                        sleep(10);
                                }
                                if (retval) {
                                        return (retval);
                                }
                        }
                } 
                else if (fscks_running < maximum_running && fscks_running < ndisks) {
                        /* If there's more room to go, then find the next valid disk */
                        for ( ;; ) {
                                if ((nextdisk = nextdisk->next) == NULL) {
                                        nextdisk = disklist;
                                }
                                if (nextdisk->part != NULL && nextdisk->pid == 0) {
                                        break;
                                }
                        }
                        
                        while ((retval = startdiskcheck(nextdisk)) && fscks_running > 0) {
                                sleep(10);
                        }
                        if (retval) {
                                return (retval);
                        }
                }
        }
        return running_status;
        
}

/*
 * fork/exec in order to spawn a process that will eventually
 * wait4() on the fsck_XXX process.
 * 
 * Note: The number of forks/execs here is slightly complicated.
 * We call fork twice, and exec once.  The reason we need three total
 * processes is that the first will continue on as the main line of execution.
 * This first fork() will create the second process which calls checkfilesys().  
 * In checkfilesys() we will call fork again, followed by an exec.  Observe that
 * the second process created here will *immediately* call wait4 on the third
 * process, fsck_XXX. This is so that we can track error dialogs and exit statuses 
 * and tie them to the specific instance of fsck_XXX that created them. Otherwise, 
 * if we just called fork a bunch of times and waited on the first one to finish, 
 * it would be difficult to tell which process exited first, and whether or not the
 * exit status is meaningful.  
 *
 * Also note that after we get our result from checkfilesys(), we immediately exit,
 * so that this process doesn't linger and accidentally continue on. 
 */
static 
int startdiskcheck(disk_t* disk) {
        
        /* 
         * Split this process into the one that will go 
         * call fsck_XX and the one that won't 
         */
        disk->pid = fork();
        if (disk->pid < 0) {
                perror("fork");
                return (8);
        }
        if (disk->pid == 0) {
                /* 
                 * Call checkfilesys.  Note the exit() call.  Also note that
                 * we pass 1 to checkfilesys since we are a child process 
                 */
                exit(checkfilesys(disk, 1));
        }
        else {
                fscks_running++;
        }
        return (0);
}




/*
 * Call fork/exec in order to spawn instance of fsck_XXX for the filesystem
 * of the specified vfstype. This will actually spawn the process that does the
 * checking of the filesystem in question. 
 */
int checkfilesys(disk_t *disk, int child) {
#define ARGC_MAX 4      /* cmd-name, options, device, NULL-termination */
        part_t *part = disk->part;
        const char *argv[ARGC_MAX];
        int argc;
        int error = 0;
        struct stat buf;
        pid_t pid;
        int status = 0;
        char options[] = "-pdfnyql";  /* constant strings are not on the stack */
        char progname[NAME_MAX];
        char execname[MAXPATHLEN + 1];
        char* filesys = part->name;
        char* vfstype = part->vfstype;
        
        if (preen && child) {
                (void)signal(SIGQUIT, ignore_single_quit);
        }

#if TARGET_OS_IPHONE
        if (!strcmp(filesys, RAMDISK_FS_SPEC)) {
                fprintf(stdout, "No need to check filesys for ramdisk, does not exist yet.\n");
                return  0;
        }
#endif // TARGET_OS_IPHONE

        /* 
         * If there was a vfstype specified, then we can go ahead and fork/exec
         * the child fsck process if the fsck_XXX binary exists.
         */
        if (vfstype) {
                int exitstatus;

        /*
         * Not all options are currently supported by all 5 fsck_* binaries.
         * Specifically:
         *      udf does not support debug, quick or live
         *      msdos does not support quick or live
         *      exfat does not support live
         *      apfs does not support preen
         * When the System invokes fsck it is during boot (one of launchd's boot-tasks).
         * This task is run with the quick and live options.
         * On iOS we can assume all partitions are APFS or HFS.
         * On OSX we run this only against the System volume which will always be HFS or APFS
         */
                bzero(options, sizeof(options));
                snprintf(options, sizeof(options), "-%s%s%s%s%s%s%s",
                                 (preen) ? "p" : "",
                                 (debug) ? "d" : "",
                                 (force_fsck) ? "f" : "",
                                 (assume_no) ? "n" : "",
                                 (assume_yes) ? "y" : "",
                                 (quick_check) ? "q" : "",
                 (live_check) ? "l" : ""
                                 );
                
                argc = 0;
                snprintf(progname, sizeof(progname), "fsck_%s", vfstype);
                argv[argc++] = progname;
                if (strlen(options) > 1) {
                        argv[argc++] = options;
                }
                argv[argc++] = filesys;
                argv[argc] = NULL;
                
                /* Create the string to the fsck binary */
                (void)snprintf(execname, sizeof(execname), "%s/fsck_%s", _PATH_SBIN, vfstype);
                
                /* Check that the binary exists */
                error = stat (execname, &buf);
                if (error != 0) {
                        fprintf(stderr, "Filesystem cannot be checked \n");
                        return EEXIT;
                }

                pid = fork();
                switch (pid) {
                        case -1:
                                /* The fork failed. */                  
                                fprintf(stderr, "fork failed for %s \n", filesys);
                                if (preen) {
                                        fprintf(stderr, "\n%s: UNEXPECTED INCONSISTENCY; RUN fsck MANUALLY.\n",
                                                        filesys);
                                        exit(EEXIT);
                                }
                                
                                status = EEXIT;
                                break;
                                
                        case 0:
                                /* The child */
                                if (preen) {
                                        (void)signal(SIGQUIT, ignore_single_quit);
                                }
#if DEBUG
                                printf("exec: %s", execname);
                                for (int i = 1; i < argc; i++) {
                                        printf(" %s", argv[i]);
                                }
                                printf("\n");
                                exit(0);
#endif
                                execv(execname, (char * const *)argv);
                                fprintf(stderr, "error attempting to exec %s\n", execname);
                                _exit(8);
                                break;
                                
                        default:
                                /* The parent; child is process "pid" */
                                waitpid(pid, &exitstatus, 0);
                                if (WIFEXITED(exitstatus)) {
                                        status = WEXITSTATUS(exitstatus);
                                }
                                else {
                                        status = 0;
                                }
                                if (WIFSIGNALED(exitstatus)) {
                                        printf("%s (%s) EXITED WITH SIGNAL %d\n", filesys, vfstype, WTERMSIG(exitstatus));
                                        status = 8;
                                }
                                break;
                }
                
                return status;
        }
        else {
                fprintf(stderr, "Filesystem cannot be checked \n");
                return EEXIT;
        }
}

/*
 * When preening, allow a single quit to signal
 * a special exit after filesystem checks complete
 * so that reboot sequence may be interrupted.
 */
void catchquit(int sig) {
        extern int returntosingle;
        
        printf("returning to single-user after filesystem check\n");
        returntosingle = 1;
        (void)signal(SIGQUIT, SIG_DFL);
}

/* Quit if we catch a signal here. Emit 12 */
void catchsig(int sig) {
        exit (12);
}


/*
 * Determine whether a filesystem should be checked.
 *
 * Zero indicates that no check should be performed.
 */
int fs_checkable(struct fstab *fsp) {
        
        /*
         * APFS, HFS, MSDOS, exfat, and UDF are allowed for now.
         */
        if (strcmp(fsp->fs_vfstype, "apfs") &&
                strcmp(fsp->fs_vfstype, "hfs")  &&
                strcmp(fsp->fs_vfstype, "msdos") &&
                strcmp(fsp->fs_vfstype, "exfat") && 
            strcmp(fsp->fs_vfstype, "udf"))     {
                return 0;
        }
        
        /* if not RW and not RO (SW or XX?), ignore it */
        if ((strcmp(fsp->fs_type, FSTAB_RW) && strcmp(fsp->fs_type, FSTAB_RO)) ||
            fsp->fs_passno == 0) {
                return 0;
        }
        
#define DISKARB_LABEL   "LABEL="
#define DISKARB_UUID    "UUID="
        /* If LABEL  or UUID specified, ignore it */
        if ((strncmp(fsp->fs_spec, DISKARB_LABEL, strlen(DISKARB_LABEL)) == 0)
                || (strncmp(fsp->fs_spec, DISKARB_UUID, strlen(DISKARB_UUID)) == 0)) {
                return 0;
        }
        
        /* Otherwise, it looks fine. Go ahead and check! */
        return 1;
}

/*
 * Do some cursory checks on the pathname provided to ensure that it's really a block
 * device. If it is, then generate the raw device name and vend it out. 
 */
char *blockcheck (char *origname) {
        struct stat stslash;
        struct stat stblock;
        struct stat stchar;
        
        char *newname;
        char *raw;
        int retried = 0;
        int error = 0;
        
#if (TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR)
        /* Variables for setting up the kqueue listener*/
#define TIMEOUT_SEC 30l 
        struct kevent kev;
        struct kevent results;
        struct timespec ts;
        int slashdev_fd;
        int kq = -1;
        int ct;
        time_t end;
        time_t now;
#endif // (TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR)
        
        hotroot = 0;
        /* Try to get device info for '/' */
        if (stat("/", &stslash) < 0) {
                perror("/");
                /* If we can't get any info on root, then bail out */
                printf("Can't stat root\n");
                return (origname);
        }
        newname = origname;
#if TARGET_OS_IPHONE
        if (!strcmp(newname, RAMDISK_FS_SPEC)) {
                // Keyword ramdisk
                fprintf(stdout, "Encountered ramdisk definition. Do not stat\n");
                return (newname);
        }
#endif // TARGET_OS_IPHONE

retry:
        /* Poke the block device argument */
        error = stat(newname, &stblock);
        if (error < 0) {
#if (TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR)
                /* 
                 * If the device node is not present, set up 
                 * a kqueue and wait for up to 30 seconds for it to be
                 * published.
                 */
                kq = kqueue();
                if (kq < 0) {
                        printf("kqueue: could not create kqueue: %d\n", errno);
                        printf("Can't stat %s\n", newname);
                        return NULL;
                }
                slashdev_fd = open(_PATH_DEV, O_RDONLY);
                
                EV_SET(&kev, slashdev_fd, EVFILT_VNODE, EV_ADD | EV_ENABLE | EV_CLEAR, NOTE_WRITE, 0, NULL);
                ct = kevent(kq, &kev, 1, NULL, 0, NULL);
                if (ct != 0) {
                        printf("kevent() failed to register: %d\n", errno);
                        printf("Can't stat %s\n", newname);
                        /* If we can't register the kqueue, bail out */
                        close (kq);
                        kq = -1;
                        return NULL;
                }
                now = time(NULL);
                end = now + TIMEOUT_SEC;
                
                ts.tv_nsec = 0;
                while ((now = time(NULL)) < end) {
                        ts.tv_sec = end - now;
                        ct = kevent(kq, NULL, 0, &results, 1, &ts);
                        if (results.flags & EV_ERROR) {
                                /* If we register any errors, bail out */
                                printf("kevent: registered errors.\n");
                                error = -1;
                                close (kq);
                                kq = -1;
                                break;
                        }
                        error = stat (newname, &stblock);
                        if (error == 0) {
                                /* found the item. continue on */
                                if (kq >= 0) {
                                        close (kq);
                                }
                                break;
                        }
                }
                if (error != 0) {
                        /* Time out.  bail out */
                        if (kq >= 0) {
                                close(kq);
                        }
                        printf("fsck timed out. Can't stat %s\n", newname);
                        return NULL;
                }
                
#else //(TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR)
                perror(newname);
                printf("Can't stat %s\n", newname);
                return (NULL);
#endif // (TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR)
        }

        if ((stblock.st_mode & S_IFMT) == S_IFBLK) {
                /* 
                 * If the block device we're checking is the same as '/' then
                 * update hotroot global for debugging.
                 */
                if (stslash.st_dev == stblock.st_rdev) {
                        hotroot++;
                }
                raw = rawname(newname);
                if (stat(raw, &stchar) < 0) {
                        perror(raw);
                        printf("Can't stat %s\n", raw);
                        return (origname);
                }
                if ((stchar.st_mode & S_IFMT) == S_IFCHR) {
                        return (raw);
                } else {
                        printf("%s is not a character device\n", raw);
                        return (origname);
                }
        } else if ((stblock.st_mode & S_IFMT) == S_IFCHR && !retried) {
                newname = unrawname(newname);
                retried++;
                goto retry;
        }
        /*
         * Not a block or character device, return NULL and
         * let the user decide what to do.
         */
        return (NULL);
}


/*
 * Generate a raw disk device pathname from a normal one.
 * 
 * For input /dev/disk1s2, generate /dev/rdisk1s2
 */
char *rawname(char *name) {
        static char rawbuf[32];
        char *dp;
        
        /* 
         * Search for the last '/' in the pathname.
         * If it's not there, then bail out 
         */
        if ((dp = strrchr(name, '/')) == 0) {
                return (0);
        }
        /* 
         * Insert a NULL in the place of the final '/' so that we can 
         * copy everything BEFORE that last '/' into a separate buffer.
         */
        *dp = 0;
        (void)strlcpy(rawbuf, name, sizeof(rawbuf));
        *dp = '/';
        /* Now add an /r to our buffer, then copy everything after the final / */
        (void)strlcat(rawbuf, "/r", sizeof(rawbuf));
        (void)strlcat(rawbuf, &dp[1], sizeof(rawbuf));
        return (rawbuf);
}

/*
 * Generate a regular disk device name from a raw one. 
 *
 * For input /dev/rdisk1s2, generate /dev/disk1s2
 */
char *unrawname(char *name) {
        char *dp;
        struct stat stb;
        size_t length;
        
        /* Find the last '/' in the pathname */
        if ((dp = strrchr(name, '/')) == 0) {
                return (name);
        }
        
        /* Stat the disk device argument */
        if (stat(name, &stb) < 0) {
                return (name);
        }
        
        /* If it's not a character device, error out */
        if ((stb.st_mode & S_IFMT) != S_IFCHR) {
                return (name);
        }
        
        /* If it's not a real raw name, then error out */
        if (dp[1] != 'r') {
                return (name);
        }
        length = strlen(&dp[2]);
        length++; /* to account for trailing NULL */
        
        memmove(&dp[1], &dp[2], length);
        return (name);
}

/*
 * Given a pathname to a disk device, generate the relevant disk_t for that 
 * disk device.  It is assumed that this function will be called for each item in the
 * fstab that needs to get checked. 
 */
disk_t *finddisk (char *pathname) {
        disk_t *disk;
        disk_t **dkp;
        char *tmp;
        size_t len;
        
        /*
         * Find the disk name.  It is assumed that the disk name ends with the
         * first run of digit(s) in the last component of the path.
         */
        tmp = strrchr(pathname, '/');           /* Find the last component of the path */
        if (tmp == NULL) {
                tmp = pathname;
        }
        else {
                tmp++;
        }
        for (; *tmp && !isdigit(*tmp); tmp++) { /* Skip non-digits */
                continue;
        }
        
        for (; *tmp && isdigit(*tmp); tmp++){   /* Skip to end of consecutive digits */
                continue;
        }
        
        len = tmp - pathname;
        if (len == 0) {
                len = strlen(pathname);
        }
        
        /* Iterate through all known disks to see if this item was already seen before */
        for (disk = disklist, dkp = &disklist; disk; dkp = &disk->next, disk = disk->next) {
                if ((strncmp(disk->name, pathname, len) == 0) && 
                        (disk->name[len] == 0)) {
                        return (disk);
                }
        }
        /* If not, then allocate a new structure and add it to the end of the list */
        if ((*dkp = (disk_t*)malloc(sizeof(disk_t))) == NULL) {
                fprintf(stderr, "out of memory");
                exit (8);
        }
        /* Make 'disk' point to the newly allocated structure */
        disk = *dkp;
        if ((disk->name = malloc(len + 1)) == NULL) {
                fprintf(stderr, "out of memory");
                exit (8);
        }
        /* copy the name into place */
        (void)strncpy(disk->name, pathname, len);
        disk->name[len] = '\0';
        /* Initialize 'part' and 'next' to NULL for now */
        disk->part = NULL;
        disk->next = NULL;
        disk->pid = 0;
        /* Increase total number of disks observed */
        ndisks++;
        
        /* Bubble out either the newly created disk_t or the one we found */
        return (disk);
}


/*
 * Add this partition to the list of devices to check. 
 */ 
void addpart(char *name, char *fsname, char *vfstype) {
        disk_t *disk;
        part_t *part;
        part_t **ppt;
        
        /* Find the disk_t that corresponds to our element */
        disk = finddisk(name);
        ppt = &(disk->part);
        
        /* 
         * Now iterate through all of the partitions of that disk.
         * If we see our partition name already in there, then it means the entry
         * was in the fstab more than once, which is bad. 
         */
        for (part = disk->part; part; ppt = &part->next, part = part->next) {
                if (strcmp(part->name, name) == 0) {
                        printf("%s in fstab more than once!\n", name);
                        return;
                }
        }
        
        /* Hopefully we get here. Allocate a new partition structure for the disk */
        if ((*ppt = (part_t*)malloc(sizeof(part_t))) == NULL) {
                fprintf(stderr, "out of memory");
                exit (8);
        }
        part = *ppt;
        if ((part->name = strdup(name)) == NULL) {
                fprintf(stderr, "out of memory");
                exit (8);
        }
        
        /* Add the name & vfs info to the partition struct */
        if ((part->fsname = strdup(fsname)) == NULL) {
                fprintf(stderr, "out of memory");
                exit (8);
        }
        part->next = NULL;
        part->vfstype = strdup(vfstype);
        if (part->vfstype == NULL) {
                fprintf(stderr, "out of memory");
                exit (8);
        }
}

/*
 * Free the partition and its fields. 
 */
void destroy_part (part_t *part) {
        if (part->name) {
                free (part->name);
        }
        
        if (part->fsname) {
                free (part->fsname);
        }
        
        if (part->vfstype) {
                free (part->vfstype);
        }
        
        free (part);
}


/*
 * Ignore a single quit signal; wait and flush just in case.
 * Used by child processes in preen mode.
 */
void
ignore_single_quit(int sig) {
        
    sleep(1);
    (void)signal(SIGQUIT, SIG_IGN);
    (void)signal(SIGQUIT, SIG_DFL);
}