static int zfs_refresh_properties(vfs_t *vfsp) { zfsvfs_t *zfsvfs = vfsp->vfs_data; /* * Remount operations default to "rw" unless "ro" is explicitly * specified. */ if (vfs_optionisset(vfsp, MNTOPT_RO, NULL)) { readonly_changed_cb(zfsvfs, B_TRUE); } else { if (!dmu_objset_is_snapshot(zfsvfs->z_os)) readonly_changed_cb(zfsvfs, B_FALSE); else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) return (EROFS); } if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) { devices_changed_cb(zfsvfs, B_FALSE); setuid_changed_cb(zfsvfs, B_FALSE); } else { if (vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL)) devices_changed_cb(zfsvfs, B_FALSE); else if (vfs_optionisset(vfsp, MNTOPT_DEVICES, NULL)) devices_changed_cb(zfsvfs, B_TRUE); if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) setuid_changed_cb(zfsvfs, B_FALSE); else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) setuid_changed_cb(zfsvfs, B_TRUE); } if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) exec_changed_cb(zfsvfs, B_FALSE); else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) exec_changed_cb(zfsvfs, B_TRUE); if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) atime_changed_cb(zfsvfs, B_TRUE); else if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) atime_changed_cb(zfsvfs, B_FALSE); return (0); }
int zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting) { int error; error = zfs_register_callbacks(zsb); if (error) return (error); /* * Set the objset user_ptr to track its zsb. */ mutex_enter(&zsb->z_os->os_user_ptr_lock); dmu_objset_set_user(zsb->z_os, zsb); mutex_exit(&zsb->z_os->os_user_ptr_lock); zsb->z_log = zil_open(zsb->z_os, zfs_get_data); /* * If we are not mounting (ie: online recv), then we don't * have to worry about replaying the log as we blocked all * operations out since we closed the ZIL. */ if (mounting) { boolean_t readonly; /* * During replay we remove the read only flag to * allow replays to succeed. */ readonly = zfs_is_readonly(zsb); if (readonly != 0) readonly_changed_cb(zsb, B_FALSE); else zfs_unlinked_drain(zsb); /* * Parse and replay the intent log. * * Because of ziltest, this must be done after * zfs_unlinked_drain(). (Further note: ziltest * doesn't use readonly mounts, where * zfs_unlinked_drain() isn't called.) This is because * ziltest causes spa_sync() to think it's committed, * but actually it is not, so the intent log contains * many txg's worth of changes. * * In particular, if object N is in the unlinked set in * the last txg to actually sync, then it could be * actually freed in a later txg and then reallocated * in a yet later txg. This would write a "create * object N" record to the intent log. Normally, this * would be fine because the spa_sync() would have * written out the fact that object N is free, before * we could write the "create object N" intent log * record. * * But when we are in ziltest mode, we advance the "open * txg" without actually spa_sync()-ing the changes to * disk. So we would see that object N is still * allocated and in the unlinked set, and there is an * intent log record saying to allocate it. */ if (spa_writeable(dmu_objset_spa(zsb->z_os))) { if (zil_replay_disable) { zil_destroy(zsb->z_log, B_FALSE); } else { zsb->z_replay = B_TRUE; zil_replay(zsb->z_os, zsb, zfs_replay_vector); zsb->z_replay = B_FALSE; } } /* restore readonly bit */ if (readonly != 0) readonly_changed_cb(zsb, B_TRUE); } return (0); }
int zfs_register_callbacks(zfs_sb_t *zsb) { struct dsl_dataset *ds = NULL; objset_t *os = zsb->z_os; zfs_mntopts_t *zmo = zsb->z_mntopts; int error = 0; ASSERT(zsb); ASSERT(zmo); /* * The act of registering our callbacks will destroy any mount * options we may have. In order to enable temporary overrides * of mount options, we stash away the current values and * restore them after we register the callbacks. */ if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) { zmo->z_do_readonly = B_TRUE; zmo->z_readonly = B_TRUE; } /* * Register property callbacks. * * It would probably be fine to just check for i/o error from * the first prop_register(), but I guess I like to go * overboard... */ ds = dmu_objset_ds(os); dsl_pool_config_enter(dmu_objset_pool(os), FTAG); error = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb); dsl_pool_config_exit(dmu_objset_pool(os), FTAG); if (error) goto unregister; /* * Invoke our callbacks to restore temporary mount options. */ if (zmo->z_do_readonly) readonly_changed_cb(zsb, zmo->z_readonly); if (zmo->z_do_setuid) setuid_changed_cb(zsb, zmo->z_setuid); if (zmo->z_do_exec) exec_changed_cb(zsb, zmo->z_exec); if (zmo->z_do_devices) devices_changed_cb(zsb, zmo->z_devices); if (zmo->z_do_xattr) xattr_changed_cb(zsb, zmo->z_xattr); if (zmo->z_do_atime) atime_changed_cb(zsb, zmo->z_atime); if (zmo->z_do_relatime) relatime_changed_cb(zsb, zmo->z_relatime); if (zmo->z_do_nbmand) nbmand_changed_cb(zsb, zmo->z_nbmand); return (0); unregister: /* * We may attempt to unregister some callbacks that are not * registered, but this is OK; it will simply return ENOMSG, * which we will ignore. */ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb); return (error); }
int zfs_domount(struct super_block *sb, zfs_mntopts_t *zmo, int silent) { const char *osname = zmo->z_osname; zfs_sb_t *zsb; struct inode *root_inode; uint64_t recordsize; int error; error = zfs_sb_create(osname, zmo, &zsb); if (error) return (error); if ((error = dsl_prop_get_integer(osname, "recordsize", &recordsize, NULL))) goto out; zsb->z_sb = sb; sb->s_fs_info = zsb; sb->s_magic = ZFS_SUPER_MAGIC; sb->s_maxbytes = MAX_LFS_FILESIZE; sb->s_time_gran = 1; sb->s_blocksize = recordsize; sb->s_blocksize_bits = ilog2(recordsize); zsb->z_bdi.ra_pages = 0; sb->s_bdi = &zsb->z_bdi; error = -zpl_bdi_setup_and_register(&zsb->z_bdi, "zfs"); if (error) goto out; /* Set callback operations for the file system. */ sb->s_op = &zpl_super_operations; sb->s_xattr = zpl_xattr_handlers; sb->s_export_op = &zpl_export_operations; #ifdef HAVE_S_D_OP sb->s_d_op = &zpl_dentry_operations; #endif /* HAVE_S_D_OP */ /* Set features for file system. */ zfs_set_fuid_feature(zsb); if (dmu_objset_is_snapshot(zsb->z_os)) { uint64_t pval; atime_changed_cb(zsb, B_FALSE); readonly_changed_cb(zsb, B_TRUE); if ((error = dsl_prop_get_integer(osname, "xattr", &pval, NULL))) goto out; xattr_changed_cb(zsb, pval); if ((error = dsl_prop_get_integer(osname, "acltype", &pval, NULL))) goto out; acltype_changed_cb(zsb, pval); zsb->z_issnap = B_TRUE; zsb->z_os->os_sync = ZFS_SYNC_DISABLED; zsb->z_snap_defer_time = jiffies; mutex_enter(&zsb->z_os->os_user_ptr_lock); dmu_objset_set_user(zsb->z_os, zsb); mutex_exit(&zsb->z_os->os_user_ptr_lock); } else { error = zfs_sb_setup(zsb, B_TRUE); } /* Allocate a root inode for the filesystem. */ error = zfs_root(zsb, &root_inode); if (error) { (void) zfs_umount(sb); goto out; } /* Allocate a root dentry for the filesystem */ sb->s_root = d_make_root(root_inode); if (sb->s_root == NULL) { (void) zfs_umount(sb); error = SET_ERROR(ENOMEM); goto out; } if (!zsb->z_issnap) zfsctl_create(zsb); zsb->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb); out: if (error) { dmu_objset_disown(zsb->z_os, zsb); zfs_sb_free(zsb); } return (error); }
static int zfs_domount(struct mount *mp, dev_t mount_dev, char *osname, vfs_context_t ctx) { uint64_t readonly; int error = 0; int mode; zfsvfs_t *zfsvfs; znode_t *zp = NULL; struct timeval tv; ASSERT(mp); ASSERT(osname); /* * Initialize the zfs-specific filesystem structure. * Should probably make this a kmem cache, shuffle fields, * and just bzero up to z_hold_mtx[]. */ zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); zfsvfs->z_vfs = mp; zfsvfs->z_parent = zfsvfs; zfsvfs->z_assign = TXG_NOWAIT; zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE; zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), offsetof(znode_t, z_link_node)); rw_init(&zfsvfs->z_unmount_lock, NULL, RW_DEFAULT, NULL); rw_init(&zfsvfs->z_unmount_inactive_lock, NULL, RW_DEFAULT, NULL); #ifndef __APPLE__ /* Initialize the generic filesystem structure. */ vfsp->vfs_bcount = 0; vfsp->vfs_data = NULL; if (zfs_create_unique_device(&mount_dev) == -1) { error = ENODEV; goto out; } ASSERT(vfs_devismounted(mount_dev) == 0); #endif vfs_setfsprivate(mp, zfsvfs); if (error = dsl_prop_get_integer(osname, "readonly", &readonly, NULL)) goto out; if (readonly) { mode = DS_MODE_PRIMARY | DS_MODE_READONLY; vfs_setflags(mp, (u_int64_t)((unsigned int)MNT_RDONLY)); } else { mode = DS_MODE_PRIMARY; } error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os); if (error == EROFS) { mode = DS_MODE_PRIMARY | DS_MODE_READONLY; error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os); } if (error) goto out; if (error = zfs_init_fs(zfsvfs, &zp, (cred_t *) vfs_context_ucred(ctx))) goto out; /* The call to zfs_init_fs leaves the vnode held, release it here. */ vnode_put(ZTOV(zp)); if (dmu_objset_is_snapshot(zfsvfs->z_os)) { uint64_t xattr; ASSERT(mode & DS_MODE_READONLY); #if 0 atime_changed_cb(zfsvfs, B_FALSE); readonly_changed_cb(zfsvfs, B_TRUE); if (error = dsl_prop_get_integer(osname, "xattr", &xattr, NULL)) goto out; xattr_changed_cb(zfsvfs, xattr); #endif zfsvfs->z_issnap = B_TRUE; } else { if (!vfs_isrdonly(mp)) zfs_unlinked_drain(zfsvfs); #ifndef __APPLE__ /* * Parse and replay the intent log. * * Because of ziltest, this must be done after * zfs_unlinked_drain(). (Further note: ziltest doesn't * use readonly mounts, where zfs_unlinked_drain() isn't * called.) This is because ziltest causes spa_sync() * to think it's committed, but actually it is not, so * the intent log contains many txg's worth of changes. * * In particular, if object N is in the unlinked set in * the last txg to actually sync, then it could be * actually freed in a later txg and then reallocated in * a yet later txg. This would write a "create object * N" record to the intent log. Normally, this would be * fine because the spa_sync() would have written out * the fact that object N is free, before we could write * the "create object N" intent log record. * * But when we are in ziltest mode, we advance the "open * txg" without actually spa_sync()-ing the changes to * disk. So we would see that object N is still * allocated and in the unlinked set, and there is an * intent log record saying to allocate it. */ zil_replay(zfsvfs->z_os, zfsvfs, &zfsvfs->z_assign, zfs_replay_vector); if (!zil_disable) zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data); #endif } #if 0 if (!zfsvfs->z_issnap) zfsctl_create(zfsvfs); #endif /* * Record the mount time (for Spotlight) */ microtime(&tv); zfsvfs->z_mount_time = tv.tv_sec; out: if (error) { if (zfsvfs->z_os) dmu_objset_close(zfsvfs->z_os); mutex_destroy(&zfsvfs->z_znodes_lock); list_destroy(&zfsvfs->z_all_znodes); rw_destroy(&zfsvfs->z_unmount_lock); rw_destroy(&zfsvfs->z_unmount_inactive_lock); kmem_free(zfsvfs, sizeof (zfsvfs_t)); } else { OSIncrementAtomic(&zfs_active_fs_count); (void) copystr(osname, vfs_statfs(mp)->f_mntfromname, MNAMELEN - 1, 0); vfs_getnewfsid(mp); } return (error); }
int zfs_register_callbacks(zfs_sb_t *zsb) { struct dsl_dataset *ds = NULL; objset_t *os = zsb->z_os; boolean_t do_readonly = B_FALSE; int error = 0; if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) do_readonly = B_TRUE; /* * Register property callbacks. * * It would probably be fine to just check for i/o error from * the first prop_register(), but I guess I like to go * overboard... */ ds = dmu_objset_ds(os); dsl_pool_config_enter(dmu_objset_pool(os), FTAG); error = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb); dsl_pool_config_exit(dmu_objset_pool(os), FTAG); if (error) goto unregister; if (do_readonly) readonly_changed_cb(zsb, B_TRUE); return (0); unregister: /* * We may attempt to unregister some callbacks that are not * registered, but this is OK; it will simply return ENOMSG, * which we will ignore. */ (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb); (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb); return (error); }
int zfs_domount(struct super_block *sb, void *data, int silent) { zpl_mount_data_t *zmd = data; const char *osname = zmd->z_osname; zfs_sb_t *zsb; struct inode *root_inode; uint64_t recordsize; int error; error = zfs_sb_create(osname, &zsb); if (error) return (error); if ((error = dsl_prop_get_integer(osname, "recordsize", &recordsize, NULL))) goto out; zsb->z_sb = sb; sb->s_fs_info = zsb; sb->s_magic = ZFS_SUPER_MAGIC; sb->s_maxbytes = MAX_LFS_FILESIZE; sb->s_time_gran = 1; sb->s_blocksize = recordsize; sb->s_blocksize_bits = ilog2(recordsize); #ifdef HAVE_BDI /* * 2.6.32 API change, * Added backing_device_info (BDI) per super block interfaces. A BDI * must be configured when using a non-device backed filesystem for * proper writeback. This is not required for older pdflush kernels. * * NOTE: Linux read-ahead is disabled in favor of zfs read-ahead. */ zsb->z_bdi.ra_pages = 0; sb->s_bdi = &zsb->z_bdi; error = -bdi_setup_and_register(&zsb->z_bdi, "zfs", BDI_CAP_MAP_COPY); if (error) goto out; #endif /* HAVE_BDI */ /* Set callback operations for the file system. */ sb->s_op = &zpl_super_operations; sb->s_xattr = zpl_xattr_handlers; sb->s_export_op = &zpl_export_operations; #ifdef HAVE_S_D_OP sb->s_d_op = &zpl_dentry_operations; #endif /* HAVE_S_D_OP */ /* Set features for file system. */ zfs_set_fuid_feature(zsb); if (dmu_objset_is_snapshot(zsb->z_os)) { uint64_t pval; atime_changed_cb(zsb, B_FALSE); readonly_changed_cb(zsb, B_TRUE); if ((error = dsl_prop_get_integer(osname,"xattr",&pval,NULL))) goto out; xattr_changed_cb(zsb, pval); zsb->z_issnap = B_TRUE; zsb->z_os->os_sync = ZFS_SYNC_DISABLED; mutex_enter(&zsb->z_os->os_user_ptr_lock); dmu_objset_set_user(zsb->z_os, zsb); mutex_exit(&zsb->z_os->os_user_ptr_lock); } else { error = zfs_sb_setup(zsb, B_TRUE); } /* Allocate a root inode for the filesystem. */ error = zfs_root(zsb, &root_inode); if (error) { (void) zfs_umount(sb); goto out; } /* Allocate a root dentry for the filesystem */ sb->s_root = d_make_root(root_inode); if (sb->s_root == NULL) { (void) zfs_umount(sb); error = ENOMEM; goto out; } if (!zsb->z_issnap) zfsctl_create(zsb); out: if (error) { dmu_objset_disown(zsb->z_os, zsb); zfs_sb_free(zsb); } return (error); }
static int zfs_mountroot(vfs_t *vfsp, enum whymountroot why) { int error = 0; int ret = 0; static int zfsrootdone = 0; zfsvfs_t *zfsvfs = NULL; znode_t *zp = NULL; vnode_t *vp = NULL; ASSERT(vfsp); /* * The filesystem that we mount as root is defined in * /etc/system using the zfsroot variable. The value defined * there is copied early in startup code to zfs_bootpath * (defined in modsysfile.c). */ if (why == ROOT_INIT) { if (zfsrootdone++) return (EBUSY); /* * This needs to be done here, so that when we return from * mountroot, the vfs resource name will be set correctly. */ if (snprintf(rootfs.bo_name, BO_MAXOBJNAME, "%s", zfs_bootpath) >= BO_MAXOBJNAME) return (ENAMETOOLONG); if (error = vfs_lock(vfsp)) return (error); if (error = zfs_domount(vfsp, zfs_bootpath, CRED())) goto out; zfsvfs = (zfsvfs_t *)vfsp->vfs_data; ASSERT(zfsvfs); if (error = zfs_zget(zfsvfs, zfsvfs->z_root, &zp)) goto out; vp = ZTOV(zp); mutex_enter(&vp->v_lock); vp->v_flag |= VROOT; mutex_exit(&vp->v_lock); rootvp = vp; /* * The zfs_zget call above returns with a hold on vp, we release * it here. */ VN_RELE(vp); /* * Mount root as readonly initially, it will be remouted * read/write by /lib/svc/method/fs-usr. */ readonly_changed_cb(vfsp->vfs_data, B_TRUE); vfs_add((struct vnode *)0, vfsp, (vfsp->vfs_flag & VFS_RDONLY) ? MS_RDONLY : 0); out: vfs_unlock(vfsp); ret = (error) ? error : 0; return (ret); } else if (why == ROOT_REMOUNT) { readonly_changed_cb(vfsp->vfs_data, B_FALSE); vfsp->vfs_flag |= VFS_REMOUNT; return (zfs_refresh_properties(vfsp)); } else if (why == ROOT_UNMOUNT) { zfs_unregister_callbacks((zfsvfs_t *)vfsp->vfs_data); (void) zfs_sync(vfsp, 0, 0); return (0); } /* * if "why" is equal to anything else other than ROOT_INIT, * ROOT_REMOUNT, or ROOT_UNMOUNT, we do not support it. */ return (ENOTSUP); }
static int zfs_domount(vfs_t *vfsp, char *osname, cred_t *cr) { dev_t mount_dev; uint64_t recordsize, readonly; int error = 0; int mode; zfsvfs_t *zfsvfs; znode_t *zp = NULL; ASSERT(vfsp); ASSERT(osname); /* * Initialize the zfs-specific filesystem structure. * Should probably make this a kmem cache, shuffle fields, * and just bzero up to z_hold_mtx[]. */ zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); zfsvfs->z_vfs = vfsp; zfsvfs->z_parent = zfsvfs; zfsvfs->z_assign = TXG_NOWAIT; zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE; zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), offsetof(znode_t, z_link_node)); rw_init(&zfsvfs->z_um_lock, NULL, RW_DEFAULT, NULL); /* Initialize the generic filesystem structure. */ vfsp->vfs_bcount = 0; vfsp->vfs_data = NULL; if (zfs_create_unique_device(&mount_dev) == -1) { error = ENODEV; goto out; } ASSERT(vfs_devismounted(mount_dev) == 0); if (error = dsl_prop_get_integer(osname, "recordsize", &recordsize, NULL)) goto out; vfsp->vfs_dev = mount_dev; vfsp->vfs_fstype = zfsfstype; vfsp->vfs_bsize = recordsize; vfsp->vfs_flag |= VFS_NOTRUNC; vfsp->vfs_data = zfsvfs; if (error = dsl_prop_get_integer(osname, "readonly", &readonly, NULL)) goto out; if (readonly) mode = DS_MODE_PRIMARY | DS_MODE_READONLY; else mode = DS_MODE_PRIMARY; error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os); if (error == EROFS) { mode = DS_MODE_PRIMARY | DS_MODE_READONLY; error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os); } if (error) goto out; if (error = zfs_init_fs(zfsvfs, &zp, cr)) goto out; /* The call to zfs_init_fs leaves the vnode held, release it here. */ VN_RELE(ZTOV(zp)); if (dmu_objset_is_snapshot(zfsvfs->z_os)) { ASSERT(mode & DS_MODE_READONLY); atime_changed_cb(zfsvfs, B_FALSE); readonly_changed_cb(zfsvfs, B_TRUE); zfsvfs->z_issnap = B_TRUE; } else { error = zfs_register_callbacks(vfsp); if (error) goto out; /* * Start a delete thread running. */ (void) zfs_delete_thread_target(zfsvfs, 1); /* * Parse and replay the intent log. */ zil_replay(zfsvfs->z_os, zfsvfs, &zfsvfs->z_assign, zfs_replay_vector, (void (*)(void *))zfs_delete_wait_empty); if (!zil_disable) zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data); } if (!zfsvfs->z_issnap) zfsctl_create(zfsvfs); out: if (error) { if (zfsvfs->z_os) dmu_objset_close(zfsvfs->z_os); kmem_free(zfsvfs, sizeof (zfsvfs_t)); } else { atomic_add_32(&zfs_active_fs_count, 1); } return (error); }
static int zfs_register_callbacks(vfs_t *vfsp) { struct dsl_dataset *ds = NULL; objset_t *os = NULL; zfsvfs_t *zfsvfs = NULL; int do_readonly = FALSE, readonly; int do_setuid = FALSE, setuid; int do_exec = FALSE, exec; int do_devices = FALSE, devices; int error = 0; ASSERT(vfsp); zfsvfs = vfsp->vfs_data; ASSERT(zfsvfs); os = zfsvfs->z_os; /* * The act of registering our callbacks will destroy any mount * options we may have. In order to enable temporary overrides * of mount options, we stash away the current values and restore * restore them after we register the callbacks. */ if (vfs_optionisset(vfsp, MNTOPT_RO, NULL)) { readonly = B_TRUE; do_readonly = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) { readonly = B_FALSE; do_readonly = B_TRUE; } if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) { devices = B_FALSE; setuid = B_FALSE; do_devices = B_TRUE; do_setuid = B_TRUE; } else { if (vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL)) { devices = B_FALSE; do_devices = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_DEVICES, NULL)) { devices = B_TRUE; do_devices = B_TRUE; } if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) { setuid = B_FALSE; do_setuid = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) { setuid = B_TRUE; do_setuid = B_TRUE; } } if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) { exec = B_FALSE; do_exec = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) { exec = B_TRUE; do_exec = B_TRUE; } /* * Register property callbacks. * * It would probably be fine to just check for i/o error from * the first prop_register(), but I guess I like to go * overboard... */ ds = dmu_objset_ds(os); error = dsl_prop_register(ds, "atime", atime_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "recordsize", blksz_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "readonly", readonly_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "devices", devices_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "setuid", setuid_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "exec", exec_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "snapdir", snapdir_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "aclmode", acl_mode_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "aclinherit", acl_inherit_changed_cb, zfsvfs); if (error) goto unregister; /* * Invoke our callbacks to restore temporary mount options. */ if (do_readonly) readonly_changed_cb(zfsvfs, readonly); if (do_setuid) setuid_changed_cb(zfsvfs, setuid); if (do_exec) exec_changed_cb(zfsvfs, exec); if (do_devices) devices_changed_cb(zfsvfs, devices); return (0); unregister: /* * We may attempt to unregister some callbacks that are not * registered, but this is OK; it will simply return ENOMSG, * which we will ignore. */ (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, zfsvfs); return (error); }
static int zfs_domount(vfs_t *vfsp, char *osname) { uint64_t recordsize, readonly; int error = 0; int mode; zfsvfs_t *zfsvfs; znode_t *zp = NULL; ASSERT(vfsp); ASSERT(osname); /* * Initialize the zfs-specific filesystem structure. * Should probably make this a kmem cache, shuffle fields, * and just bzero up to z_hold_mtx[]. */ zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); zfsvfs->z_vfs = vfsp; zfsvfs->z_parent = zfsvfs; zfsvfs->z_assign = TXG_NOWAIT; zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE; zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&zfsvfs->z_online_recv_lock, NULL, MUTEX_DEFAULT, NULL); list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), offsetof(znode_t, z_link_node)); rrw_init(&zfsvfs->z_teardown_lock); rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL); if (error = dsl_prop_get_integer(osname, "recordsize", &recordsize, NULL)) goto out; zfsvfs->z_vfs->vfs_bsize = recordsize; vfsp->vfs_data = zfsvfs; vfsp->mnt_flag |= MNT_LOCAL; vfsp->mnt_kern_flag |= MNTK_MPSAFE; vfsp->mnt_kern_flag |= MNTK_LOOKUP_SHARED; vfsp->mnt_kern_flag |= MNTK_SHARED_WRITES; if (error = dsl_prop_get_integer(osname, "readonly", &readonly, NULL)) goto out; mode = DS_MODE_OWNER; if (readonly) mode |= DS_MODE_READONLY; error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os); if (error == EROFS) { mode = DS_MODE_OWNER | DS_MODE_READONLY; error = dmu_objset_open(osname, DMU_OST_ZFS, mode, &zfsvfs->z_os); } if (error) goto out; if (error = zfs_init_fs(zfsvfs, &zp)) goto out; /* * Set features for file system. */ zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); if (zfsvfs->z_use_fuids) { vfs_set_feature(vfsp, VFSFT_XVATTR); vfs_set_feature(vfsp, VFSFT_SYSATTR_VIEWS); vfs_set_feature(vfsp, VFSFT_ACEMASKONACCESS); vfs_set_feature(vfsp, VFSFT_ACLONCREATE); } if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) { vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS); vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE); vfs_set_feature(vfsp, VFSFT_NOCASESENSITIVE); } else if (zfsvfs->z_case == ZFS_CASE_MIXED) { vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS); vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE); } if (dmu_objset_is_snapshot(zfsvfs->z_os)) { uint64_t pval; ASSERT(mode & DS_MODE_READONLY); atime_changed_cb(zfsvfs, B_FALSE); readonly_changed_cb(zfsvfs, B_TRUE); if (error = dsl_prop_get_integer(osname, "xattr", &pval, NULL)) goto out; xattr_changed_cb(zfsvfs, pval); zfsvfs->z_issnap = B_TRUE; } else { error = zfsvfs_setup(zfsvfs, B_TRUE); } vfs_mountedfrom(vfsp, osname); if (!zfsvfs->z_issnap) zfsctl_create(zfsvfs); out: if (error) { if (zfsvfs->z_os) dmu_objset_close(zfsvfs->z_os); zfs_freezfsvfs(zfsvfs); } else { atomic_add_32(&zfs_active_fs_count, 1); } return (error); }
static int zfs_register_callbacks(vfs_t *vfsp) { struct dsl_dataset *ds = NULL; objset_t *os = NULL; zfsvfs_t *zfsvfs = NULL; uint64_t nbmand; int readonly, do_readonly = FALSE; int setuid, do_setuid = FALSE; int exec, do_exec = FALSE; int xattr, do_xattr = FALSE; int atime, do_atime = FALSE; int error = 0; ASSERT(vfsp); zfsvfs = vfsp->vfs_data; ASSERT(zfsvfs); os = zfsvfs->z_os; /* * This function can be called for a snapshot when we update snapshot's * mount point, which isn't really supported. */ if (dmu_objset_is_snapshot(os)) return (EOPNOTSUPP); /* * The act of registering our callbacks will destroy any mount * options we may have. In order to enable temporary overrides * of mount options, we stash away the current values and * restore them after we register the callbacks. */ if (vfs_optionisset(vfsp, MNTOPT_RO, NULL)) { readonly = B_TRUE; do_readonly = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) { readonly = B_FALSE; do_readonly = B_TRUE; } if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) { setuid = B_FALSE; do_setuid = B_TRUE; } else { if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) { setuid = B_FALSE; do_setuid = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) { setuid = B_TRUE; do_setuid = B_TRUE; } } if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) { exec = B_FALSE; do_exec = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) { exec = B_TRUE; do_exec = B_TRUE; } if (vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL)) { xattr = B_FALSE; do_xattr = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_XATTR, NULL)) { xattr = B_TRUE; do_xattr = B_TRUE; } if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) { atime = B_FALSE; do_atime = B_TRUE; } else if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) { atime = B_TRUE; do_atime = B_TRUE; } /* * nbmand is a special property. It can only be changed at * mount time. * * This is weird, but it is documented to only be changeable * at mount time. */ if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) { nbmand = B_FALSE; } else if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL)) { nbmand = B_TRUE; } else { char osname[MAXNAMELEN]; dmu_objset_name(os, osname); if (error = dsl_prop_get_integer(osname, "nbmand", &nbmand, NULL)) { return (error); } } /* * Register property callbacks. * * It would probably be fine to just check for i/o error from * the first prop_register(), but I guess I like to go * overboard... */ ds = dmu_objset_ds(os); error = dsl_prop_register(ds, "atime", atime_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "xattr", xattr_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "recordsize", blksz_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "readonly", readonly_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "setuid", setuid_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "exec", exec_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "snapdir", snapdir_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "aclmode", acl_mode_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "aclinherit", acl_inherit_changed_cb, zfsvfs); error = error ? error : dsl_prop_register(ds, "vscan", vscan_changed_cb, zfsvfs); if (error) goto unregister; /* * Invoke our callbacks to restore temporary mount options. */ if (do_readonly) readonly_changed_cb(zfsvfs, readonly); if (do_setuid) setuid_changed_cb(zfsvfs, setuid); if (do_exec) exec_changed_cb(zfsvfs, exec); if (do_xattr) xattr_changed_cb(zfsvfs, xattr); if (do_atime) atime_changed_cb(zfsvfs, atime); nbmand_changed_cb(zfsvfs, nbmand); return (0); unregister: /* * We may attempt to unregister some callbacks that are not * registered, but this is OK; it will simply return ENOMSG, * which we will ignore. */ (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, zfsvfs); (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zfsvfs); return (error); }
int zfs_register_callbacks(zfs_sb_t *zsb) { struct dsl_dataset *ds = NULL; objset_t *os = zsb->z_os; int error = 0; if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) readonly_changed_cb(zsb, B_TRUE); /* * Register property callbacks. * * It would probably be fine to just check for i/o error from * the first prop_register(), but I guess I like to go * overboard... */ ds = dmu_objset_ds(os); error = dsl_prop_register(ds, "atime", atime_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "xattr", xattr_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "recordsize", blksz_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "readonly", readonly_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "devices", devices_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "setuid", setuid_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "exec", exec_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "snapdir", snapdir_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "aclinherit", acl_inherit_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "vscan", vscan_changed_cb, zsb); error = error ? error : dsl_prop_register(ds, "nbmand", nbmand_changed_cb, zsb); if (error) goto unregister; return (0); unregister: /* * We may attempt to unregister some callbacks that are not * registered, but this is OK; it will simply return ENOMSG, * which we will ignore. */ (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zsb); (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zsb); (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zsb); (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zsb); (void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zsb); (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zsb); (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zsb); (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zsb); (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, zsb); (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zsb); (void) dsl_prop_unregister(ds, "nbmand", nbmand_changed_cb, zsb); return (error); }