Whamcloud - gitweb
LU-16750 tune2fs: add "-E iops" to set/clear IOPS groups
[tools/e2fsprogs.git] / misc / mke2fs.c
index 27e7d17..b62252a 100644 (file)
 #include <strings.h>
 #include <ctype.h>
 #include <time.h>
-#ifdef __linux__
-#include <sys/utsname.h>
-#define KERNEL_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c))
-#endif
 #ifdef HAVE_GETOPT_H
 #include <getopt.h>
 #else
@@ -71,8 +67,6 @@ extern int optind;
 #define ZAP_BOOTBLOCK
 #endif
 
-#define DISCARD_STEP_MB                (2048)
-
 extern int isatty(int);
 extern FILE *fpopen(const char *cmd, const char *mode);
 
@@ -93,7 +87,10 @@ static uid_t root_uid;
 static gid_t   root_gid;
 int    journal_size;
 int    journal_flags;
+int    journal_fc_size;
+static e2_blkcnt_t     orphan_file_blocks;
 static int     lazy_itable_init;
+static int     assume_storage_prezeroed;
 static int     packed_meta_blocks;
 int            no_copy_xattrs;
 static char    *bad_blocks_filename = NULL;
@@ -104,6 +101,10 @@ static __u64       offset;
 static blk64_t journal_location = ~0LL;
 static int     proceed_delay = -1;
 static blk64_t dev_size;
+blk64_t                iops_array[64];
+unsigned int   iops_size = sizeof(iops_array);
+unsigned int   iops_count = 0;
+blk64_t                *iops_range = iops_array;
 
 static struct ext2_super_block fs_param;
 static __u32 zero_buf[4];
@@ -164,54 +165,6 @@ int int_log10(unsigned long long arg)
        return l;
 }
 
-#ifdef __linux__
-static int parse_version_number(const char *s)
-{
-       int     major, minor, rev;
-       char    *endptr;
-       const char *cp = s;
-
-       if (!s)
-               return 0;
-       major = strtol(cp, &endptr, 10);
-       if (cp == endptr || *endptr != '.')
-               return 0;
-       cp = endptr + 1;
-       minor = strtol(cp, &endptr, 10);
-       if (cp == endptr || *endptr != '.')
-               return 0;
-       cp = endptr + 1;
-       rev = strtol(cp, &endptr, 10);
-       if (cp == endptr)
-               return 0;
-       return KERNEL_VERSION(major, minor, rev);
-}
-
-static int is_before_linux_ver(unsigned int major, unsigned int minor,
-                              unsigned int rev)
-{
-       struct          utsname ut;
-       static int      linux_version_code = -1;
-
-       if (uname(&ut)) {
-               perror("uname");
-               exit(1);
-       }
-       if (linux_version_code < 0)
-               linux_version_code = parse_version_number(ut.release);
-       if (linux_version_code == 0)
-               return 0;
-
-       return linux_version_code < (int) KERNEL_VERSION(major, minor, rev);
-}
-#else
-static int is_before_linux_ver(unsigned int major, unsigned int minor,
-                              unsigned int rev)
-{
-       return 0;
-}
-#endif
-
 /*
  * Helper function for read_bb_file and test_disk
  */
@@ -256,7 +209,8 @@ static void test_disk(ext2_filsys fs, badblocks_list *bb_list)
 
        sprintf(buf, "badblocks -b %d -X %s%s%s %llu", fs->blocksize,
                quiet ? "" : "-s ", (cflag > 1) ? "-w " : "",
-               fs->device_name, ext2fs_blocks_count(fs->super)-1);
+               fs->device_name,
+               (unsigned long long) ext2fs_blocks_count(fs->super)-1);
        if (verbose)
                printf(_("Running command: %s\n"), buf);
        f = popen(buf, "r");
@@ -441,7 +395,8 @@ static void write_inode_tables(ext2_filsys fs, int lazy_flag, int itable_zeroed)
                        if (retval) {
                                fprintf(stderr, _("\nCould not write %d "
                                          "blocks in inode table starting at %llu: %s\n"),
-                                       num, blk, error_message(retval));
+                                       num, (unsigned long long) blk,
+                                       error_message(retval));
                                exit(1);
                        }
                }
@@ -582,8 +537,10 @@ static void zap_sector(ext2_filsys fs, int sect, int nsect)
                else {
                        magic = (unsigned int *) (buf + BSD_LABEL_OFFSET);
                        if ((*magic == BSD_DISKMAGIC) ||
-                           (*magic == BSD_MAGICDISK))
+                           (*magic == BSD_MAGICDISK)) {
+                               free(buf);
                                return;
+                       }
                }
        }
 
@@ -604,9 +561,16 @@ static void create_journal_dev(ext2_filsys fs)
        char                    *buf;
        blk64_t                 blk, err_blk;
        int                     c, count, err_count;
+       struct ext2fs_journal_params    jparams;
 
-       retval = ext2fs_create_journal_superblock(fs,
-                                 ext2fs_blocks_count(fs->super), 0, &buf);
+       retval = ext2fs_get_journal_params(&jparams, fs);
+       if (retval) {
+               com_err("create_journal_dev", retval, "%s",
+                       _("while splitting the journal size"));
+               exit(1);
+       }
+
+       retval = ext2fs_create_journal_superblock2(fs, &jparams, 0, &buf);
        if (retval) {
                com_err("create_journal_dev", retval, "%s",
                        _("while initializing journal superblock"));
@@ -631,7 +595,7 @@ static void create_journal_dev(ext2_filsys fs)
                        com_err("create_journal_dev", retval,
                                _("while zeroing journal device "
                                  "(block %llu, count %d)"),
-                               err_blk, err_count);
+                               (unsigned long long) err_blk, err_count);
                        exit(1);
                }
                blk += c;
@@ -663,14 +627,15 @@ static void show_stats(ext2_filsys fs)
        if (!verbose) {
                printf(_("Creating filesystem with %llu %dk blocks and "
                         "%u inodes\n"),
-                      ext2fs_blocks_count(s), fs->blocksize >> 10,
-                      s->s_inodes_count);
+                      (unsigned long long) ext2fs_blocks_count(s),
+                      fs->blocksize >> 10, s->s_inodes_count);
                goto skip_details;
        }
 
        if (ext2fs_blocks_count(&fs_param) != ext2fs_blocks_count(s))
                fprintf(stderr, _("warning: %llu blocks unused.\n\n"),
-                      ext2fs_blocks_count(&fs_param) - ext2fs_blocks_count(s));
+                       (unsigned long long) (ext2fs_blocks_count(&fs_param) -
+                                             ext2fs_blocks_count(s)));
 
        printf(_("Filesystem label=%.*s\n"), EXT2_LEN_STR(s->s_volume_name));
 
@@ -690,9 +655,9 @@ static void show_stats(ext2_filsys fs)
        printf(_("Stride=%u blocks, Stripe width=%u blocks\n"),
               s->s_raid_stride, s->s_raid_stripe_width);
        printf(_("%u inodes, %llu blocks\n"), s->s_inodes_count,
-              ext2fs_blocks_count(s));
+              (unsigned long long) ext2fs_blocks_count(s));
        printf(_("%llu blocks (%2.2f%%) reserved for the super user\n"),
-               ext2fs_r_blocks_count(s),
+              (unsigned long long) ext2fs_r_blocks_count(s),
               100.0 *  ext2fs_r_blocks_count(s) / ext2fs_blocks_count(s));
        printf(_("First data block=%u\n"), s->s_first_data_block);
        if (root_uid != 0 || root_gid != 0)
@@ -736,7 +701,7 @@ skip_details:
                        col_left = 72;
                }
                col_left -= need;
-               printf("%llu", group_block);
+               printf("%llu", (unsigned long long) group_block);
        }
        printf("\n\n");
 }
@@ -779,6 +744,54 @@ static int set_os(struct ext2_super_block *sb, char *os)
        return 1;
 }
 
+static int parse_range(char *p_start, char *p_end, char *p_hyphen)
+{
+       blk64_t start, end;
+       blk64_t *new_array;
+
+       /**
+        * e.g  0-1024G
+        *      ^      ^
+        *      |      |
+        *   p_start  p_end
+        */
+       end = parse_num_blocks(p_hyphen + 1, -1);
+
+       if (!isdigit(*(p_end - 1)) && isdigit(*(p_hyphen -1))) {
+               /* copy G/M/K unit to start value */
+               *p_hyphen = *(p_end - 1);
+               p_hyphen++;
+       }
+       *p_hyphen = 0;
+
+       start = parse_num_blocks(p_start, -1);
+
+       /* add to iops_range */
+       if (iops_count == iops_size) {
+               iops_size <<= 1;
+               if (iops_size == 0) {
+                       iops_size = iops_count;
+                       return -E2BIG;
+               }
+               if (iops_range == iops_array)
+                       new_array = malloc(iops_size * sizeof(blk64_t));
+               else
+                       new_array = realloc(iops_range,
+                                           iops_size * sizeof(blk64_t));
+               if (!new_array) {
+                       iops_size >>= 1;
+                       return -ENOMEM;
+               } else {
+                       iops_range = new_array;
+               }
+       }
+
+       iops_range[iops_count++] = start;
+       iops_range[iops_count++] = end;
+
+       return 0;
+}
+
 #define PATH_SET "PATH=/sbin"
 
 static void parse_extended_opts(struct ext2_super_block *param,
@@ -999,6 +1012,11 @@ static void parse_extended_opts(struct ext2_super_block *param,
                                lazy_itable_init = strtoul(arg, &p, 0);
                        else
                                lazy_itable_init = 1;
+               } else if (!strcmp(token, "assume_storage_prezeroed")) {
+                       if (arg)
+                               assume_storage_prezeroed = strtoul(arg, &p, 0);
+                       else
+                               assume_storage_prezeroed = 1;
                } else if (!strcmp(token, "lazy_journal_init")) {
                        if (arg)
                                journal_flags |= strtoul(arg, &p, 0) ?
@@ -1076,6 +1094,77 @@ static void parse_extended_opts(struct ext2_super_block *param,
                                continue;
                        }
                        encoding_flags = arg;
+               } else if (!strcmp(token, "orphan_file_size")) {
+                       if (!arg) {
+                               r_usage++;
+                               badopt = token;
+                               continue;
+                       }
+                       orphan_file_blocks = parse_num_blocks2(arg,
+                                               fs_param.s_log_block_size);
+                       if (orphan_file_blocks == 0) {
+                               fprintf(stderr,
+                                       _("Invalid size of orphan file %s\n"),
+                                       arg);
+                               r_usage++;
+                               continue;
+                       }
+               } else if (!strcmp(token, "iops")) {
+                       char *p_colon, *p_hyphen;
+                       blk64_t start, end;
+
+                       /* example: iops=0-1024G:4096-8192G */
+
+                       if (!arg) {
+                               r_usage++;
+                               badopt = token;
+                               continue;
+                       }
+                       p_colon = strchr(arg, ':');
+                       while (p_colon != NULL) {
+                               *p_colon = 0;
+
+                               p_hyphen = strchr(arg, '-');
+                               if (p_hyphen == NULL) {
+                                       fprintf(stderr,
+                                               _("error: parse iops %s\n"),
+                                               arg);
+                                       r_usage++;
+                                       badopt = token;
+                                       break;
+                               }
+
+                               ret = parse_range(arg, p_colon, p_hyphen);
+                               if (ret < 0) {
+                                       fprintf(stderr,
+                                               _("error: parse iops %s:%d\n"),
+                                               arg, ret);
+                                       r_usage++;
+                                       badopt = token;
+                                       break;
+                               }
+
+                               arg = p_colon + 1;
+                               p_colon = strchr(arg, ':');
+                       }
+                       p_hyphen = strchr(arg, '-');
+                       if (p_hyphen == NULL) {
+                               fprintf(stderr,
+                                       _("error: parse iops %s\n"), arg);
+                               r_usage++;
+                               badopt = token;
+                               continue;
+                       }
+
+                       ret = parse_range(arg, arg + strlen(arg), p_hyphen);
+                       if (ret < 0) {
+                               fprintf(stderr,
+                                       _("error: parse iops %s:%d\n"),
+                                       arg, ret);
+                               r_usage++;
+                               badopt = token;
+                               continue;
+                       }
                } else {
                        r_usage++;
                        badopt = token;
@@ -1102,9 +1191,13 @@ static void parse_extended_opts(struct ext2_super_block *param,
                        "\tnodiscard\n"
                        "\tencoding=<encoding>\n"
                        "\tencoding_flags=<flags>\n"
-                       "\tquotatype=<quota type(s) to be enabled>\n\n"),
+                       "\tiops=<iops storage size range>\n"
+                       "\tquotatype=<quota type(s) to be enabled>\n"
+                       "\tassume_storage_prezeroed=<0 to disable, 1 to enable>\n\n"),
                        badopt ? badopt : "");
                free(buf);
+               if (iops_range != iops_array)
+                       free(iops_range);
                exit(1);
        }
        if (param->s_raid_stride &&
@@ -1141,7 +1234,10 @@ static __u32 ok_features[3] = {
                EXT2_FEATURE_COMPAT_RESIZE_INODE |
                EXT2_FEATURE_COMPAT_DIR_INDEX |
                EXT2_FEATURE_COMPAT_EXT_ATTR |
-               EXT4_FEATURE_COMPAT_SPARSE_SUPER2,
+               EXT4_FEATURE_COMPAT_SPARSE_SUPER2 |
+               EXT4_FEATURE_COMPAT_FAST_COMMIT |
+               EXT4_FEATURE_COMPAT_STABLE_INODES |
+               EXT4_FEATURE_COMPAT_ORPHAN_FILE,
        /* Incompat */
        EXT2_FEATURE_INCOMPAT_FILETYPE|
                EXT3_FEATURE_INCOMPAT_EXTENTS|
@@ -1150,6 +1246,7 @@ static __u32 ok_features[3] = {
                EXT4_FEATURE_INCOMPAT_FLEX_BG|
                EXT4_FEATURE_INCOMPAT_EA_INODE|
                EXT4_FEATURE_INCOMPAT_MMP |
+               EXT4_FEATURE_INCOMPAT_DIRDATA|
                EXT4_FEATURE_INCOMPAT_64BIT|
                EXT4_FEATURE_INCOMPAT_INLINE_DATA|
                EXT4_FEATURE_INCOMPAT_ENCRYPT |
@@ -1466,23 +1563,33 @@ int get_bool_from_profile(char **types, const char *opt, int def_val)
 extern const char *mke2fs_default_profile;
 static const char *default_files[] = { "<default>", 0 };
 
+struct device_param {
+       unsigned long min_io;           /* preferred minimum IO size */
+       unsigned long opt_io;           /* optimal IO size */
+       unsigned long alignment_offset; /* alignment offset wrt physical block size */
+       unsigned int dax:1;             /* supports dax? */
+};
+
+#define        OPTIMIZED_STRIPE_WIDTH  512
+#define        OPTIMIZED_STRIDE        512
+
 #ifdef HAVE_BLKID_PROBE_GET_TOPOLOGY
 /*
  * Sets the geometry of a device (stripe/stride), and returns the
  * device's alignment offset, if any, or a negative error.
  */
 static int get_device_geometry(const char *file,
-                              struct ext2_super_block *param,
-                              unsigned int psector_size)
+                              unsigned int blocksize,
+                              unsigned int psector_size,
+                              struct device_param *dev_param)
 {
        int rc = -1;
-       unsigned int blocksize;
        blkid_probe pr;
        blkid_topology tp;
-       unsigned long min_io;
-       unsigned long opt_io;
        struct stat statbuf;
 
+       memset(dev_param, 0, sizeof(*dev_param));
+
        /* Nothing to do for a regular file */
        if (!stat(file, &statbuf) && S_ISREG(statbuf.st_mode))
                return 0;
@@ -1495,23 +1602,32 @@ static int get_device_geometry(const char *file,
        if (!tp)
                goto out;
 
-       min_io = blkid_topology_get_minimum_io_size(tp);
-       opt_io = blkid_topology_get_optimal_io_size(tp);
-       blocksize = EXT2_BLOCK_SIZE(param);
-       if ((min_io == 0) && (psector_size > blocksize))
-               min_io = psector_size;
-       if ((opt_io == 0) && min_io)
-               opt_io = min_io;
-       if ((opt_io == 0) && (psector_size > blocksize))
-               opt_io = psector_size;
-
-       /* setting stripe/stride to blocksize is pointless */
-       if (min_io > blocksize)
-               param->s_raid_stride = min_io / blocksize;
-       if (opt_io > blocksize)
-               param->s_raid_stripe_width = opt_io / blocksize;
-
-       rc = blkid_topology_get_alignment_offset(tp);
+       dev_param->min_io = blkid_topology_get_minimum_io_size(tp);
+       if (dev_param->min_io > OPTIMIZED_STRIDE) {
+               fprintf(stdout,
+                       "detected raid stride %lu too large, use optimum %lu\n",
+                       dev_param->min_io, OPTIMIZED_STRIDE);
+               dev_param->min_io = OPTIMIZED_STRIDE;
+       }
+       dev_param->opt_io = blkid_topology_get_optimal_io_size(tp);
+       if (dev_param->opt_io > OPTIMIZED_STRIPE_WIDTH) {
+               fprintf(stdout,
+                       "detected raid stripe width %lu too large, use optimum %lu\n",
+                       dev_param->opt_io, OPTIMIZED_STRIPE_WIDTH);
+               dev_param->opt_io = OPTIMIZED_STRIPE_WIDTH;
+       }
+       if ((dev_param->min_io == 0) && (psector_size > blocksize))
+               dev_param->min_io = psector_size;
+       if ((dev_param->opt_io == 0) && dev_param->min_io > 0)
+               dev_param->opt_io = dev_param->min_io;
+       if ((dev_param->opt_io == 0) && (psector_size > blocksize))
+               dev_param->opt_io = psector_size;
+
+       dev_param->alignment_offset = blkid_topology_get_alignment_offset(tp);
+#ifdef HAVE_BLKID_TOPOLOGY_GET_DAX
+       dev_param->dax = blkid_topology_get_dax(tp);
+#endif
+       rc = 0;
 out:
        blkid_free_probe(pr);
        return rc;
@@ -1532,6 +1648,8 @@ static void PRS(int argc, char *argv[])
        int             lsector_size = 0, psector_size = 0;
        int             show_version_only = 0, is_device = 0;
        unsigned long long num_inodes = 0; /* unsigned long long to catch too-large input */
+       int             default_orphan_file = 0;
+       int             default_csum_seed = 0;
        errcode_t       retval;
        char *          oldpath = getenv("PATH");
        char *          extended_opts = 0;
@@ -1553,13 +1671,15 @@ static void PRS(int argc, char *argv[])
         * Finally, we complain about fs_blocks_count > 2^32 on a non-64bit fs.
         */
        blk64_t         fs_blocks_count = 0;
-       long            sysval;
        int             s_opt = -1, r_opt = -1;
        char            *fs_features = 0;
        int             fs_features_size = 0;
        int             use_bsize;
        char            *newpath;
        int             pathlen = sizeof(PATH_SET) + 1;
+#ifdef HAVE_BLKID_PROBE_GET_TOPOLOGY
+       struct device_param dev_param;
+#endif
 
        if (oldpath)
                pathlen += strlen(oldpath);
@@ -1584,9 +1704,12 @@ static void PRS(int argc, char *argv[])
 #define _SC_PAGESIZE _SC_PAGE_SIZE
 #endif
 #ifdef _SC_PAGESIZE
-       sysval = sysconf(_SC_PAGESIZE);
-       if (sysval > 0)
-               sys_page_size = sysval;
+       {
+               long sysval = sysconf(_SC_PAGESIZE);
+
+               if (sysval > 0)
+                       sys_page_size = sysval;
+       }
 #endif /* _SC_PAGESIZE */
 #endif /* HAVE_SYSCONF */
 
@@ -1615,7 +1738,7 @@ profile_error:
        memset(&fs_param, 0, sizeof(struct ext2_super_block));
        fs_param.s_rev_level = 1;  /* Create revision 1 filesystems now */
 
-       if (is_before_linux_ver(2, 2, 0))
+       if (ext2fs_is_before_linux_ver(2, 2, 0))
                fs_param.s_rev_level = 0;
 
        if (argc && *argv) {
@@ -1746,6 +1869,8 @@ profile_error:
                case 'j':
                        if (!journal_size)
                                journal_size = -1;
+                       if (!journal_fc_size)
+                               journal_fc_size = -1;
                        break;
                case 'J':
                        parse_journal_opts(optarg);
@@ -1900,10 +2025,10 @@ profile_error:
 #ifdef CONFIG_TESTIO_DEBUG
                if (getenv("TEST_IO_FLAGS") || getenv("TEST_IO_BLOCK")) {
                        io_ptr = test_io_manager;
-                       test_io_backing_manager = unix_io_manager;
+                       test_io_backing_manager = default_io_manager;
                } else
 #endif
-                       io_ptr = unix_io_manager;
+                       io_ptr = default_io_manager;
                retval = ext2fs_open(journal_device,
                                     EXT2_FLAG_JOURNAL_DEV_OK, 0,
                                     0, io_ptr, &jfs);
@@ -1949,18 +2074,8 @@ profile_error:
        profile_get_integer(profile, "options", "proceed_delay", 0, 0,
                            &proceed_delay);
 
-       /* The isatty() test is so we don't break existing scripts */
-       flags = CREATE_FILE;
-       if (isatty(0) && isatty(1) && !offset)
-               flags |= CHECK_FS_EXIST;
-       if (!quiet)
-               flags |= VERBOSE_CREATE;
-       if (fs_blocks_count == 0)
-               flags |= NO_SIZE;
-       else
+       if (fs_blocks_count)
                explicit_fssize = 1;
-       if (!check_plausibility(device_name, flags, &is_device) && !force)
-               proceed_question(proceed_delay);
 
        check_mount(device_name, force, _("filesystem"));
 
@@ -1969,15 +2084,29 @@ profile_error:
                dev_size = fs_blocks_count;
                retval = 0;
        } else
-#ifndef _WIN32
                retval = ext2fs_get_device_size2(device_name,
                                                 EXT2_BLOCK_SIZE(&fs_param),
                                                 &dev_size);
-#else
-               retval = ext2fs_get_device_size(device_name,
-                                               EXT2_BLOCK_SIZE(&fs_param),
-                                               &dev_size);
-#endif
+       if (retval == ENOENT) {
+               int fd;
+
+               if (!explicit_fssize) {
+                       fprintf(stderr,
+                               _("The file %s does not exist and no "
+                                 "size was specified.\n"), device_name);
+                       exit(1);
+               }
+               fd = ext2fs_open_file(device_name,
+                                     O_CREAT | O_WRONLY, 0666);
+               if (fd < 0) {
+                       retval = errno;
+               } else {
+                       dev_size = 0;
+                       retval = 0;
+                       close(fd);
+                       printf(_("Creating regular file %s\n"), device_name);
+               }
+       }
        if (retval && (retval != EXT2_ET_UNIMPLEMENTED)) {
                com_err(program_name, retval, "%s",
                        _("while trying to determine filesystem size"));
@@ -2020,6 +2149,9 @@ profile_error:
        if (!usage_types)
                profile_get_string(profile, "devices", device_name,
                                   "usage_types", 0, &usage_types);
+       if (!creator_os)
+               profile_get_string(profile, "defaults", "creator_os", 0,
+                                  0, &creator_os);
 
        /*
         * We have the file system (or device) size, so we can now
@@ -2027,8 +2159,7 @@ profile_error:
         * be appropriately configured.
         */
        fs_types = parse_fs_type(fs_type, usage_types, &fs_param,
-                                fs_blocks_count ? fs_blocks_count : dev_size,
-                                argv[0]);
+                                fs_blocks_count, argv[0]);
        if (!fs_types) {
                fprintf(stderr, "%s", _("Failed to parse fs types list\n"));
                exit(1);
@@ -2070,8 +2201,24 @@ profile_error:
                ext2fs_clear_feature_ea_inode(&fs_param);
                ext2fs_clear_feature_casefold(&fs_param);
        }
-       edit_feature(fs_features ? fs_features : tmp,
-                    &fs_param.s_feature_compat);
+       if (!fs_features && tmp)
+               edit_feature(tmp, &fs_param.s_feature_compat);
+       /*
+        * Now all the defaults are incorporated in fs_param. Check the state
+        * of orphan_file feature so that we know whether we should silently
+        * disabled in case journal gets disabled.
+        */
+       if (ext2fs_has_feature_orphan_file(&fs_param))
+               default_orphan_file = 1;
+       if (ext2fs_has_feature_csum_seed(&fs_param))
+               default_csum_seed = 1;
+       if (fs_features)
+               edit_feature(fs_features, &fs_param.s_feature_compat);
+       /* Silently disable orphan_file if user chose fs without journal */
+       if (default_orphan_file && !ext2fs_has_feature_journal(&fs_param))
+               ext2fs_clear_feature_orphan_file(&fs_param);
+       if (default_csum_seed && !ext2fs_has_feature_metadata_csum(&fs_param))
+               ext2fs_clear_feature_csum_seed(&fs_param);
        if (tmp)
                free(tmp);
        (void) ext2fs_free_mem(&fs_features);
@@ -2131,7 +2278,8 @@ profile_error:
 
                if (use_bsize == -1) {
                        use_bsize = sys_page_size;
-                       if (is_before_linux_ver(2, 6, 0) && use_bsize > 4096)
+                       if (ext2fs_is_before_linux_ver(2, 6, 0) &&
+                           use_bsize > 4096)
                                use_bsize = 4096;
                }
                if (lsector_size && use_bsize < lsector_size)
@@ -2161,24 +2309,25 @@ profile_error:
         * We now need to do a sanity check of fs_blocks_count for
         * 32-bit vs 64-bit block number support.
         */
-       if ((fs_blocks_count > MAX_32_NUM) &&
-           ext2fs_has_feature_64bit(&fs_param))
-               ext2fs_clear_feature_resize_inode(&fs_param);
-       if ((fs_blocks_count > MAX_32_NUM) &&
-           !ext2fs_has_feature_64bit(&fs_param) &&
-           get_bool_from_profile(fs_types, "auto_64-bit_support", 0)) {
-               ext2fs_set_feature_64bit(&fs_param);
-               ext2fs_clear_feature_resize_inode(&fs_param);
-       }
-       if ((fs_blocks_count > MAX_32_NUM) &&
-           !ext2fs_has_feature_64bit(&fs_param)) {
-               fprintf(stderr, _("%s: Size of device (0x%llx blocks) %s "
+       if (fs_blocks_count > MAX_32_NUM) {
+               if (!ext2fs_has_feature_64bit(&fs_param) &&
+                   get_bool_from_profile(fs_types, "auto_64-bit_support", 0))
+                       ext2fs_set_feature_64bit(&fs_param);
+
+               if (ext2fs_has_feature_64bit(&fs_param)) {
+                       ext2fs_clear_feature_resize_inode(&fs_param);
+               } else {
+                       fprintf(stderr,
+                               _("%s: Size of device (0x%llx blocks) %s "
                                  "too big to be expressed\n\t"
                                  "in 32 bits using a blocksize of %d.\n"),
-                       program_name, fs_blocks_count, device_name,
-                       EXT2_BLOCK_SIZE(&fs_param));
-               exit(1);
+                               program_name,
+                               (unsigned long long) fs_blocks_count,
+                               device_name, EXT2_BLOCK_SIZE(&fs_param));
+                       exit(1);
+               }
        }
+
        /*
         * Guard against group descriptor count overflowing... Mostly to avoid
         * strange results for absurdly large devices.  This is in log2:
@@ -2189,8 +2338,8 @@ profile_error:
                fprintf(stderr, _("%s: Size of device (0x%llx blocks) %s "
                                  "too big to create\n\t"
                                  "a filesystem using a blocksize of %d.\n"),
-                       program_name, fs_blocks_count, device_name,
-                       EXT2_BLOCK_SIZE(&fs_param));
+                       program_name, (unsigned long long) fs_blocks_count,
+                       device_name, EXT2_BLOCK_SIZE(&fs_param));
                exit(1);
        }
 
@@ -2258,13 +2407,13 @@ profile_error:
                fs_param.s_feature_compat = 0;
                fs_param.s_feature_ro_compat &=
                                        EXT4_FEATURE_RO_COMPAT_METADATA_CSUM;
-       }
+       }
 
        /* Check the user's mkfs options for 64bit */
-       if (ext2fs_has_feature_64bit(&fs_param) &&
+       if (fs_blocks_count > MAX_32_NUM &&
            !ext2fs_has_feature_extents(&fs_param)) {
-               printf("%s", _("Extents MUST be enabled for a 64-bit "
-                              "filesystem.  Pass -O extents to rectify.\n"));
+               printf("%s", _("Extents MUST be enabled for filesystems with "
+                              "over 2^32 blocks. Use '-O extents' to fix.\n"));
                exit(1);
        }
 
@@ -2305,17 +2454,35 @@ profile_error:
        }
 
 #ifdef HAVE_BLKID_PROBE_GET_TOPOLOGY
-       retval = get_device_geometry(device_name, &fs_param,
-                                    (unsigned int) psector_size);
+       retval = get_device_geometry(device_name, blocksize,
+                                    psector_size, &dev_param);
        if (retval < 0) {
                fprintf(stderr,
                        _("warning: Unable to get device geometry for %s\n"),
                        device_name);
-       } else if (retval) {
-               printf(_("%s alignment is offset by %lu bytes.\n"),
-                      device_name, retval);
-               printf(_("This may result in very poor performance, "
-                         "(re)-partitioning suggested.\n"));
+       } else {
+               /* setting stripe/stride to blocksize is pointless */
+               if (dev_param.min_io > (unsigned) blocksize)
+                       fs_param.s_raid_stride = dev_param.min_io / blocksize;
+               if (dev_param.opt_io > (unsigned) blocksize) {
+                       fs_param.s_raid_stripe_width =
+                                               dev_param.opt_io / blocksize;
+               }
+
+               if (dev_param.alignment_offset) {
+                       printf(_("%s alignment is offset by %lu bytes.\n"),
+                              device_name, dev_param.alignment_offset);
+                       printf(_("This may result in very poor performance, "
+                                 "(re)-partitioning suggested.\n"));
+               }
+
+               if (dev_param.dax && blocksize != sys_page_size) {
+                       fprintf(stderr,
+                               _("%s is capable of DAX but current block size "
+                                 "%u is different from system page size %u so "
+                                 "filesystem will not support DAX.\n"),
+                               device_name, blocksize, sys_page_size);
+               }
        }
 #endif
 
@@ -2345,7 +2512,7 @@ profile_error:
        }
 
        /* Metadata checksumming wasn't totally stable before 3.18. */
-       if (is_before_linux_ver(3, 18, 0) &&
+       if (ext2fs_is_before_linux_ver(3, 18, 0) &&
            ext2fs_has_feature_metadata_csum(&fs_param))
                fprintf(stderr, _("Suggestion: Use Linux kernel >= 3.18 for "
                        "improved stability of the metadata and journal "
@@ -2355,7 +2522,7 @@ profile_error:
         * On newer kernels we do have lazy_itable_init support. So pick the
         * right default in case ext4 module is not loaded.
         */
-       if (is_before_linux_ver(2, 6, 37))
+       if (ext2fs_is_before_linux_ver(2, 6, 37))
                lazy_itable_init = 0;
        else
                lazy_itable_init = 1;
@@ -2457,15 +2624,6 @@ profile_error:
                }
        }
 
-       if (ext2fs_has_feature_casefold(&fs_param) &&
-           ext2fs_has_feature_encrypt(&fs_param)) {
-               com_err(program_name, 0, "%s",
-                       _("The encrypt and casefold features are not "
-                         "compatible.\nThey can not be both enabled "
-                         "simultaneously.\n"));
-                     exit (1);
-       }
-
        /* Don't allow user to set both metadata_csum and uninit_bg bits. */
        if (ext2fs_has_feature_metadata_csum(&fs_param) &&
            ext2fs_has_feature_gdt_csum(&fs_param))
@@ -2489,11 +2647,12 @@ profile_error:
                exit(1);
        }
 
-       if (!quiet && ext2fs_has_feature_bigalloc(&fs_param))
-               fprintf(stderr, "%s", _("\nWarning: the bigalloc feature is "
-                                 "still under development\n"
-                                 "See https://ext4.wiki.kernel.org/"
-                                 "index.php/Bigalloc for more information\n\n"));
+       if (!quiet && ext2fs_has_feature_bigalloc(&fs_param) &&
+           EXT2_CLUSTER_SIZE(&fs_param) > 16 * EXT2_BLOCK_SIZE(&fs_param))
+               fprintf(stderr, "%s", _("\nWarning: bigalloc file systems "
+                               "with a cluster size greater than\n"
+                               "16 times the block size is considered "
+                               "experimental\n"));
 
        /*
         * Since sparse_super is the default, we would only have a problem
@@ -2566,24 +2725,32 @@ profile_error:
                exit(1);
        }
 
+       /*
+        * Warn the user that filesystems with 128-byte inodes will
+        * not work properly beyond 2038.  This can be suppressed via
+        * a boolean in the mke2fs.conf file, and we will disable this
+        * warning for file systems created for the GNU Hurd.
+        */
+       if (inode_size == EXT2_GOOD_OLD_INODE_SIZE &&
+           get_bool_from_profile(fs_types, "warn_y2038_dates", 1))
+               printf(
+_("128-byte inodes cannot handle dates beyond 2038 and are deprecated\n"));
+
        /* Make sure number of inodes specified will fit in 32 bits */
        if (num_inodes == 0) {
                unsigned long long n;
                n = ext2fs_blocks_count(&fs_param) * blocksize / inode_ratio;
                if (n > MAX_32_NUM) {
-                       if (ext2fs_has_feature_64bit(&fs_param))
-                               num_inodes = MAX_32_NUM;
-                       else {
+                       num_inodes = MAX_32_NUM;
+                       if (!ext2fs_has_feature_64bit(&fs_param))
                                com_err(program_name, 0,
-                                       _("too many inodes (%llu), raise "
-                                         "inode ratio?"), n);
-                               exit(1);
-                       }
+                                       _("too many inodes (%llu), reduced to "
+                                         "%llu"), n, MAX_32_NUM);
                }
        } else if (num_inodes > MAX_32_NUM) {
                com_err(program_name, 0,
                        _("too many inodes (%llu), specify < 2^32 inodes"),
-                         num_inodes);
+                       (unsigned long long) num_inodes);
                exit(1);
        }
        /*
@@ -2622,6 +2789,17 @@ profile_error:
 
        free(fs_type);
        free(usage_types);
+
+       /* The isatty() test is so we don't break existing scripts */
+       flags = CREATE_FILE;
+       if (isatty(0) && isatty(1) && !offset)
+               flags |= CHECK_FS_EXIST;
+       if (!quiet)
+               flags |= VERBOSE_CREATE;
+       if (!explicit_fssize)
+               flags |= NO_SIZE;
+       if (!check_plausibility(device_name, flags, &is_device) && !force)
+               proceed_question(proceed_delay);
 }
 
 static int should_do_undo(const char *name)
@@ -2630,7 +2808,7 @@ static int should_do_undo(const char *name)
        io_channel channel;
        __u16   s_magic;
        struct ext2_super_block super;
-       io_manager manager = unix_io_manager;
+       io_manager manager = default_io_manager;
        int csum_flag, force_undo;
 
        csum_flag = ext2fs_has_feature_metadata_csum(&fs_param) ||
@@ -2764,9 +2942,7 @@ err:
 static int mke2fs_discard_device(ext2_filsys fs)
 {
        struct ext2fs_numeric_progress_struct progress;
-       blk64_t blocks = ext2fs_blocks_count(fs->super);
-       blk64_t count = DISCARD_STEP_MB;
-       blk64_t cur;
+       dgrp_t group;
        int retval = 0;
 
        /*
@@ -2774,27 +2950,30 @@ static int mke2fs_discard_device(ext2_filsys fs)
         * we do not print numeric progress resulting in failure
         * afterwards.
         */
-       retval = io_channel_discard(fs->io, 0, fs->blocksize);
+       retval = io_channel_discard(fs->io, 0, 1);
        if (retval)
                return retval;
-       cur = fs->blocksize;
-
-       count *= (1024 * 1024);
-       count /= fs->blocksize;
 
        ext2fs_numeric_progress_init(fs, &progress,
                                     _("Discarding device blocks: "),
-                                    blocks);
-       while (cur < blocks) {
-               ext2fs_numeric_progress_update(fs, &progress, cur);
+                                    ext2fs_blocks_count(fs->super));
 
-               if (cur + count > blocks)
-                       count = blocks - cur;
+       for (group = 0; group < fs->group_desc_count; group++) {
+               blk64_t start = ext2fs_group_first_block2(fs, group);
+               blk64_t count = ext2fs_group_blocks_count(fs, group);
+               int retval_discard = 0;
 
-               retval = io_channel_discard(fs->io, cur, count);
-               if (retval)
-                       break;
-               cur += count;
+               retval_discard = io_channel_discard(fs->io, start, count);
+               if (!retval_discard) {
+                       ext2fs_bg_flags_set(fs, group, EXT2_BG_TRIMMED);
+                       ext2fs_group_desc_csum_set(fs, group);
+                       fs->super->s_flags |= EXT2_FLAGS_TRACK_TRIM;
+               } else if (!retval) {
+                       retval = retval_discard;
+               }
+
+               ext2fs_numeric_progress_update(fs, &progress,
+                                       ext2fs_group_last_block2(fs, group));
        }
 
        if (retval) {
@@ -2910,7 +3089,9 @@ int main (int argc, char *argv[])
        errcode_t       retval = 0;
        ext2_filsys     fs;
        badblocks_list  bb_list = 0;
-       unsigned int    journal_blocks = 0;
+       badblocks_iterate       bb_iter;
+       blk_t           blk;
+       struct ext2fs_journal_params    jparams = {0};
        unsigned int    i, checkinterval;
        int             max_mnt_count;
        int             val, hash_alg;
@@ -2920,6 +3101,7 @@ int main (int argc, char *argv[])
        char            opt_string[40];
        char            *hash_alg_str;
        int             itable_zeroed = 0;
+       blk64_t         overhead;
 
 #ifdef ENABLE_NLS
        setlocale(LC_MESSAGES, "");
@@ -2933,10 +3115,10 @@ int main (int argc, char *argv[])
 #ifdef CONFIG_TESTIO_DEBUG
        if (getenv("TEST_IO_FLAGS") || getenv("TEST_IO_BLOCK")) {
                io_ptr = test_io_manager;
-               test_io_backing_manager = unix_io_manager;
+               test_io_backing_manager = default_io_manager;
        } else
 #endif
-               io_ptr = unix_io_manager;
+               io_ptr = default_io_manager;
 
        if (undo_file != NULL || should_do_undo(device_name)) {
                retval = mke2fs_setup_tdb(device_name, &io_ptr);
@@ -2983,6 +3165,16 @@ int main (int argc, char *argv[])
                        _("while setting up superblock"));
                exit(1);
        }
+
+       if (iops_range && iops_count) {
+               ext2fs_set_iops_group(fs, iops_range, iops_count);
+               fs->super->s_flags |= EXT2_FLAGS_HAS_IOPS;
+               ext2fs_mark_super_dirty(fs);
+
+               if (iops_range != iops_array)
+                       free(iops_range);
+       }
+
        fs->progress_ops = &ext2fs_numeric_progress_ops;
 
        /* Set the error behavior */
@@ -3016,19 +3208,44 @@ int main (int argc, char *argv[])
                exit(1);
        }
 
+       if (ext2fs_has_feature_dirdata(fs->super)) {
+               if (ext2fs_has_feature_inline_data(fs->super)) {
+                       printf("%s", _("The dirdata feature can not enabled "
+                                      "with inline data feature.\n"));
+                       exit(1);
+               }
+               if (ext2fs_has_feature_casefold(fs->super)) {
+                       printf("%s", _("The dirdata feature can not enabled "
+                                      "with casefold feature.\n"));
+                       exit(1);
+               }
+       }
+
        /* Calculate journal blocks */
        if (!journal_device && ((journal_size) ||
            ext2fs_has_feature_journal(&fs_param)))
-               journal_blocks = figure_journal_size(journal_size, fs);
+               figure_journal_size(&jparams, journal_size, journal_fc_size, fs);
 
        sprintf(opt_string, "tdb_data_size=%d", fs->blocksize <= 4096 ?
                32768 : fs->blocksize * 8);
        io_channel_set_options(fs->io, opt_string);
        if (offset) {
-               sprintf(opt_string, "offset=%llu", offset);
+               sprintf(opt_string, "offset=%llu", (unsigned long long) offset);
                io_channel_set_options(fs->io, opt_string);
        }
 
+       if (assume_storage_prezeroed) {
+               if (verbose)
+                       printf("%s",
+                              _("Assuming the storage device is prezeroed "
+                              "- skipping inode table and journal wipe\n"));
+
+               lazy_itable_init = 1;
+               itable_zeroed = 1;
+               zero_hugefile = 0;
+               journal_flags |= EXT2_MKJOURNAL_LAZYINIT;
+       }
+
        /* Can't undo discard ... */
        if (!noaction && discard && dev_size && (io_ptr != undo_io_manager)) {
                retval = mke2fs_discard_device(fs);
@@ -3151,7 +3368,7 @@ int main (int argc, char *argv[])
        if (volume_label) {
                memset(fs->super->s_volume_name, 0,
                       sizeof(fs->super->s_volume_name));
-               strncpy(fs->super->s_volume_name, volume_label,
+               strncpy((char *) fs->super->s_volume_name, volume_label,
                        sizeof(fs->super->s_volume_name));
        }
 
@@ -3161,7 +3378,7 @@ int main (int argc, char *argv[])
        if (mount_dir) {
                memset(fs->super->s_last_mounted, 0,
                       sizeof(fs->super->s_last_mounted));
-               strncpy(fs->super->s_last_mounted, mount_dir,
+               strncpy((char *) fs->super->s_last_mounted, mount_dir,
                        sizeof(fs->super->s_last_mounted));
        }
 
@@ -3211,6 +3428,23 @@ int main (int argc, char *argv[])
        if (!quiet)
                printf("%s", _("done                            \n"));
 
+       /*
+        * Unmark bad blocks to calculate overhead, because metadata
+        * blocks and bad blocks can land on the same allocation cluster.
+        */
+       if (bb_list) {
+               retval = ext2fs_badblocks_list_iterate_begin(bb_list,
+                                                            &bb_iter);
+               if (retval) {
+                       com_err("ext2fs_badblocks_list_iterate_begin", retval,
+                               "%s", _("while unmarking bad blocks"));
+                       exit(1);
+               }
+               while (ext2fs_badblocks_list_iterate(bb_iter, &blk))
+                       ext2fs_unmark_block_bitmap2(fs->block_map, blk);
+               ext2fs_badblocks_list_iterate_end(bb_iter);
+       }
+
        retval = ext2fs_convert_subcluster_bitmap(fs, &fs->block_map);
        if (retval) {
                com_err(program_name, retval, "%s",
@@ -3218,6 +3452,28 @@ int main (int argc, char *argv[])
                exit(1);
        }
 
+       retval = ext2fs_count_used_clusters(fs, fs->super->s_first_data_block,
+                                       ext2fs_blocks_count(fs->super) - 1,
+                                       &overhead);
+       if (retval) {
+               com_err(program_name, retval, "%s",
+                       _("while calculating overhead"));
+               exit(1);
+       }
+
+       if (bb_list) {
+               retval = ext2fs_badblocks_list_iterate_begin(bb_list,
+                                                            &bb_iter);
+               if (retval) {
+                       com_err("ext2fs_badblocks_list_iterate_begin", retval,
+                               "%s", _("while marking bad blocks as used"));
+                       exit(1);
+               }
+               while (ext2fs_badblocks_list_iterate(bb_iter, &blk))
+                       ext2fs_mark_block_bitmap2(fs->block_map, blk);
+               ext2fs_badblocks_list_iterate_end(bb_iter);
+       }
+
        if (super_only) {
                check_plausibility(device_name, CHECK_FS_EXIST, NULL);
                printf(_("%s may be further corrupted by superblock rewrite\n"),
@@ -3262,7 +3518,7 @@ int main (int argc, char *argv[])
                if (retval) {
                        com_err(program_name, retval,
                                _("while zeroing block %llu at end of filesystem"),
-                               ret_blk);
+                               (unsigned long long) ret_blk);
                }
                write_inode_tables(fs, lazy_itable_init, itable_zeroed);
                create_root_dir(fs);
@@ -3290,7 +3546,7 @@ int main (int argc, char *argv[])
 
                retval = ext2fs_open(journal_device, EXT2_FLAG_RW|
                                     EXT2_FLAG_JOURNAL_DEV_OK, 0,
-                                    fs->blocksize, unix_io_manager, &jfs);
+                                    fs->blocksize, default_io_manager, &jfs);
                if (retval) {
                        com_err(program_name, retval,
                                _("while trying to open journal device %s\n"),
@@ -3315,22 +3571,26 @@ int main (int argc, char *argv[])
                free(journal_device);
        } else if ((journal_size) ||
                   ext2fs_has_feature_journal(&fs_param)) {
+               overhead += EXT2FS_NUM_B2C(fs, jparams.num_journal_blocks + jparams.num_fc_blocks);
                if (super_only) {
                        printf("%s", _("Skipping journal creation in super-only mode\n"));
                        fs->super->s_journal_inum = EXT2_JOURNAL_INO;
                        goto no_journal;
                }
 
-               if (!journal_blocks) {
+               if (!jparams.num_journal_blocks) {
                        ext2fs_clear_feature_journal(fs->super);
+                       ext2fs_clear_feature_orphan_file(fs->super);
+                       ext2fs_clear_feature_journal(&fs_param);
+                       ext2fs_clear_feature_orphan_file(&fs_param);
                        goto no_journal;
                }
                if (!quiet) {
                        printf(_("Creating journal (%u blocks): "),
-                              journal_blocks);
+                              jparams.num_journal_blocks + jparams.num_fc_blocks);
                        fflush(stdout);
                }
-               retval = ext2fs_add_journal_inode2(fs, journal_blocks,
+               retval = ext2fs_add_journal_inode3(fs, &jparams,
                                                   journal_location,
                                                   journal_flags);
                if (retval) {
@@ -3357,10 +3617,31 @@ no_journal:
                               fs->super->s_mmp_update_interval);
        }
 
+       overhead += fs->super->s_first_data_block;
+       if (!super_only)
+               fs->super->s_overhead_clusters = overhead;
+
        if (ext2fs_has_feature_bigalloc(&fs_param))
                fix_cluster_bg_counts(fs);
        if (ext2fs_has_feature_quota(&fs_param))
                create_quota_inodes(fs);
+       if (ext2fs_has_feature_orphan_file(&fs_param)) {
+               if (!ext2fs_has_feature_journal(&fs_param)) {
+                       com_err(program_name, 0, _("cannot set orphan_file "
+                               "feature without a journal."));
+                       exit(1);
+               }
+               if (!orphan_file_blocks) {
+                       orphan_file_blocks =
+                               ext2fs_default_orphan_file_blocks(fs);
+               }
+               retval = ext2fs_create_orphan_file(fs, orphan_file_blocks);
+               if (retval) {
+                       com_err(program_name, retval,
+                               _("while creating orphan file"));
+                       exit(1);
+               }
+       }
 
        retval = mk_hugefiles(fs, device_name);
        if (retval)