summaryrefslogtreecommitdiffstats
path: root/libblkid/libfdisk/src/gpt.c
diff options
context:
space:
mode:
Diffstat (limited to 'libblkid/libfdisk/src/gpt.c')
-rw-r--r--libblkid/libfdisk/src/gpt.c2565
1 files changed, 2565 insertions, 0 deletions
diff --git a/libblkid/libfdisk/src/gpt.c b/libblkid/libfdisk/src/gpt.c
new file mode 100644
index 000000000..8c1c96c37
--- /dev/null
+++ b/libblkid/libfdisk/src/gpt.c
@@ -0,0 +1,2565 @@
+/*
+ * Copyright (C) 2007 Karel Zak <kzak@redhat.com>
+ * Copyright (C) 2012 Davidlohr Bueso <dave@gnu.org>
+ *
+ * GUID Partition Table (GPT) support. Based on UEFI Specs 2.3.1
+ * Chapter 5: GUID Partition Table (GPT) Disk Layout (Jun 27th, 2012).
+ * Some ideas and inspiration from GNU parted and gptfdisk.
+ */
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+#include <inttypes.h>
+#include <sys/stat.h>
+#include <sys/utsname.h>
+#include <sys/types.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <errno.h>
+#include <ctype.h>
+#include <uuid.h>
+
+#include "fdiskP.h"
+
+#include "nls.h"
+#include "crc32.h"
+#include "blkdev.h"
+#include "bitops.h"
+#include "strutils.h"
+#include "all-io.h"
+
+/**
+ * SECTION: gpt
+ * @title: UEFI GPT
+ * @short_description: specific functionality
+ */
+
+#define GPT_HEADER_SIGNATURE 0x5452415020494645LL /* EFI PART */
+#define GPT_HEADER_REVISION_V1_02 0x00010200
+#define GPT_HEADER_REVISION_V1_00 0x00010000
+#define GPT_HEADER_REVISION_V0_99 0x00009900
+#define GPT_HEADER_MINSZ 92 /* bytes */
+
+#define GPT_PMBR_LBA 0
+#define GPT_MBR_PROTECTIVE 1
+#define GPT_MBR_HYBRID 2
+
+#define GPT_PRIMARY_PARTITION_TABLE_LBA 0x00000001
+
+#define EFI_PMBR_OSTYPE 0xEE
+#define MSDOS_MBR_SIGNATURE 0xAA55
+#define GPT_PART_NAME_LEN (72 / sizeof(uint16_t))
+#define GPT_NPARTITIONS 128
+
+/* Globally unique identifier */
+struct gpt_guid {
+ uint32_t time_low;
+ uint16_t time_mid;
+ uint16_t time_hi_and_version;
+ uint8_t clock_seq_hi;
+ uint8_t clock_seq_low;
+ uint8_t node[6];
+};
+
+
+/* only checking that the GUID is 0 is enough to verify an empty partition. */
+#define GPT_UNUSED_ENTRY_GUID \
+ ((struct gpt_guid) { 0x00000000, 0x0000, 0x0000, 0x00, 0x00, \
+ { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 }})
+
+/* Linux native partition type */
+#define GPT_DEFAULT_ENTRY_TYPE "0FC63DAF-8483-4772-8E79-3D69D8477DE4"
+
+/*
+ * Attribute bits
+ */
+enum {
+ /* UEFI specific */
+ GPT_ATTRBIT_REQ = 0,
+ GPT_ATTRBIT_NOBLOCK = 1,
+ GPT_ATTRBIT_LEGACY = 2,
+
+ /* GUID specific (range 48..64)*/
+ GPT_ATTRBIT_GUID_FIRST = 48,
+ GPT_ATTRBIT_GUID_COUNT = 16
+};
+
+#define GPT_ATTRSTR_REQ "RequiredPartiton"
+#define GPT_ATTRSTR_NOBLOCK "NoBlockIOProtocol"
+#define GPT_ATTRSTR_LEGACY "LegacyBIOSBootable"
+
+/* The GPT Partition entry array contains an array of GPT entries. */
+struct gpt_entry {
+ struct gpt_guid type; /* purpose and type of the partition */
+ struct gpt_guid partition_guid;
+ uint64_t lba_start;
+ uint64_t lba_end;
+ uint64_t attrs;
+ uint16_t name[GPT_PART_NAME_LEN];
+} __attribute__ ((packed));
+
+/* GPT header */
+struct gpt_header {
+ uint64_t signature; /* header identification */
+ uint32_t revision; /* header version */
+ uint32_t size; /* in bytes */
+ uint32_t crc32; /* header CRC checksum */
+ uint32_t reserved1; /* must be 0 */
+ uint64_t my_lba; /* LBA of block that contains this struct (LBA 1) */
+ uint64_t alternative_lba; /* backup GPT header */
+ uint64_t first_usable_lba; /* first usable logical block for partitions */
+ uint64_t last_usable_lba; /* last usable logical block for partitions */
+ struct gpt_guid disk_guid; /* unique disk identifier */
+ uint64_t partition_entry_lba; /* LBA of start of partition entries array */
+ uint32_t npartition_entries; /* total partition entries - normally 128 */
+ uint32_t sizeof_partition_entry; /* bytes for each GUID pt */
+ uint32_t partition_entry_array_crc32; /* partition CRC checksum */
+ uint8_t reserved2[512 - 92]; /* must all be 0 */
+} __attribute__ ((packed));
+
+struct gpt_record {
+ uint8_t boot_indicator; /* unused by EFI, set to 0x80 for bootable */
+ uint8_t start_head; /* unused by EFI, pt start in CHS */
+ uint8_t start_sector; /* unused by EFI, pt start in CHS */
+ uint8_t start_track;
+ uint8_t os_type; /* EFI and legacy non-EFI OS types */
+ uint8_t end_head; /* unused by EFI, pt end in CHS */
+ uint8_t end_sector; /* unused by EFI, pt end in CHS */
+ uint8_t end_track; /* unused by EFI, pt end in CHS */
+ uint32_t starting_lba; /* used by EFI - start addr of the on disk pt */
+ uint32_t size_in_lba; /* used by EFI - size of pt in LBA */
+} __attribute__ ((packed));
+
+/* Protected MBR and legacy MBR share same structure */
+struct gpt_legacy_mbr {
+ uint8_t boot_code[440];
+ uint32_t unique_mbr_signature;
+ uint16_t unknown;
+ struct gpt_record partition_record[4];
+ uint16_t signature;
+} __attribute__ ((packed));
+
+/*
+ * Here be dragons!
+ * See: http://en.wikipedia.org/wiki/GUID_Partition_Table#Partition_type_GUIDs
+ */
+#define DEF_GUID(_u, _n) \
+ { \
+ .typestr = (_u), \
+ .name = (_n), \
+ }
+
+static struct fdisk_parttype gpt_parttypes[] =
+{
+ /* Generic OS */
+ DEF_GUID("C12A7328-F81F-11D2-BA4B-00A0C93EC93B", N_("EFI System")),
+
+ DEF_GUID("024DEE41-33E7-11D3-9D69-0008C781F39F", N_("MBR partition scheme")),
+ DEF_GUID("D3BFE2DE-3DAF-11DF-BA40-E3A556D89593", N_("Intel Fast Flash")),
+
+ /* Hah!IdontneedEFI */
+ DEF_GUID("21686148-6449-6E6F-744E-656564454649", N_("BIOS boot")),
+
+ /* Windows */
+ DEF_GUID("E3C9E316-0B5C-4DB8-817D-F92DF00215AE", N_("Microsoft reserved")),
+ DEF_GUID("EBD0A0A2-B9E5-4433-87C0-68B6B72699C7", N_("Microsoft basic data")),
+ DEF_GUID("5808C8AA-7E8F-42E0-85D2-E1E90434CFB3", N_("Microsoft LDM metadata")),
+ DEF_GUID("AF9B60A0-1431-4F62-BC68-3311714A69AD", N_("Microsoft LDM data")),
+ DEF_GUID("DE94BBA4-06D1-4D40-A16A-BFD50179D6AC", N_("Windows recovery environment")),
+ DEF_GUID("37AFFC90-EF7D-4E96-91C3-2D7AE055B174", N_("IBM General Parallel Fs")),
+ DEF_GUID("E75CAF8F-F680-4CEE-AFA3-B001E56EFC2D", N_("Microsoft Storage Spaces")),
+
+ /* HP-UX */
+ DEF_GUID("75894C1E-3AEB-11D3-B7C1-7B03A0000000", N_("HP-UX data")),
+ DEF_GUID("E2A1E728-32E3-11D6-A682-7B03A0000000", N_("HP-UX service")),
+
+ /* Linux (http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec) */
+ DEF_GUID("0657FD6D-A4AB-43C4-84E5-0933C84B4F4F", N_("Linux swap")),
+ DEF_GUID("0FC63DAF-8483-4772-8E79-3D69D8477DE4", N_("Linux filesystem")),
+ DEF_GUID("3B8F8425-20E0-4F3B-907F-1A25A76F98E8", N_("Linux server data")),
+ DEF_GUID("44479540-F297-41B2-9AF7-D131D5F0458A", N_("Linux root (x86)")),
+ DEF_GUID("4F68BCE3-E8CD-4DB1-96E7-FBCAF984B709", N_("Linux root (x86-64)")),
+ DEF_GUID("8DA63339-0007-60C0-C436-083AC8230908", N_("Linux reserved")),
+ DEF_GUID("933AC7E1-2EB4-4F13-B844-0E14E2AEF915", N_("Linux home")),
+ DEF_GUID("A19D880F-05FC-4D3B-A006-743F0F84911E", N_("Linux RAID")),
+ DEF_GUID("BC13C2FF-59E6-4262-A352-B275FD6F7172", N_("Linux extended boot")),
+ DEF_GUID("E6D6D379-F507-44C2-A23C-238F2A3DF928", N_("Linux LVM")),
+
+ /* FreeBSD */
+ DEF_GUID("516E7CB4-6ECF-11D6-8FF8-00022D09712B", N_("FreeBSD data")),
+ DEF_GUID("83BD6B9D-7F41-11DC-BE0B-001560B84F0F", N_("FreeBSD boot")),
+ DEF_GUID("516E7CB5-6ECF-11D6-8FF8-00022D09712B", N_("FreeBSD swap")),
+ DEF_GUID("516E7CB6-6ECF-11D6-8FF8-00022D09712B", N_("FreeBSD UFS")),
+ DEF_GUID("516E7CBA-6ECF-11D6-8FF8-00022D09712B", N_("FreeBSD ZFS")),
+ DEF_GUID("516E7CB8-6ECF-11D6-8FF8-00022D09712B", N_("FreeBSD Vinum")),
+
+ /* Apple OSX */
+ DEF_GUID("48465300-0000-11AA-AA11-00306543ECAC", N_("Apple HFS/HFS+")),
+ DEF_GUID("55465300-0000-11AA-AA11-00306543ECAC", N_("Apple UFS")),
+ DEF_GUID("52414944-0000-11AA-AA11-00306543ECAC", N_("Apple RAID")),
+ DEF_GUID("52414944-5F4F-11AA-AA11-00306543ECAC", N_("Apple RAID offline")),
+ DEF_GUID("426F6F74-0000-11AA-AA11-00306543ECAC", N_("Apple boot")),
+ DEF_GUID("4C616265-6C00-11AA-AA11-00306543ECAC", N_("Apple label")),
+ DEF_GUID("5265636F-7665-11AA-AA11-00306543ECAC", N_("Apple TV recovery")),
+ DEF_GUID("53746F72-6167-11AA-AA11-00306543ECAC", N_("Apple Core storage")),
+
+ /* Solaris */
+ DEF_GUID("6A82CB45-1DD2-11B2-99A6-080020736631", N_("Solaris boot")),
+ DEF_GUID("6A85CF4D-1DD2-11B2-99A6-080020736631", N_("Solaris root")),
+ /* same as Apple ZFS */
+ DEF_GUID("6A898CC3-1DD2-11B2-99A6-080020736631", N_("Solaris /usr & Apple ZFS")),
+ DEF_GUID("6A87C46F-1DD2-11B2-99A6-080020736631", N_("Solaris swap")),
+ DEF_GUID("6A8B642B-1DD2-11B2-99A6-080020736631", N_("Solaris backup")),
+ DEF_GUID("6A8EF2E9-1DD2-11B2-99A6-080020736631", N_("Solaris /var")),
+ DEF_GUID("6A90BA39-1DD2-11B2-99A6-080020736631", N_("Solaris /home")),
+ DEF_GUID("6A9283A5-1DD2-11B2-99A6-080020736631", N_("Solaris alternate sector")),
+ DEF_GUID("6A945A3B-1DD2-11B2-99A6-080020736631", N_("Solaris reserved 1")),
+ DEF_GUID("6A9630D1-1DD2-11B2-99A6-080020736631", N_("Solaris reserved 2")),
+ DEF_GUID("6A980767-1DD2-11B2-99A6-080020736631", N_("Solaris reserved 3")),
+ DEF_GUID("6A96237F-1DD2-11B2-99A6-080020736631", N_("Solaris reserved 4")),
+ DEF_GUID("6A8D2AC7-1DD2-11B2-99A6-080020736631", N_("Solaris reserved 5")),
+
+ /* NetBSD */
+ DEF_GUID("49F48D32-B10E-11DC-B99B-0019D1879648", N_("NetBSD swap")),
+ DEF_GUID("49F48D5A-B10E-11DC-B99B-0019D1879648", N_("NetBSD FFS")),
+ DEF_GUID("49F48D82-B10E-11DC-B99B-0019D1879648", N_("NetBSD LFS")),
+ DEF_GUID("2DB519C4-B10E-11DC-B99B-0019D1879648", N_("NetBSD concatenated")),
+ DEF_GUID("2DB519EC-B10E-11DC-B99B-0019D1879648", N_("NetBSD encrypted")),
+ DEF_GUID("49F48DAA-B10E-11DC-B99B-0019D1879648", N_("NetBSD RAID")),
+
+ /* ChromeOS */
+ DEF_GUID("FE3A2A5D-4F32-41A7-B725-ACCC3285A309", N_("ChromeOS kernel")),
+ DEF_GUID("3CB8E202-3B7E-47DD-8A3C-7FF2A13CFCEC", N_("ChromeOS root fs")),
+ DEF_GUID("2E0A753D-9E48-43B0-8337-B15192CB1B5E", N_("ChromeOS reserved")),
+
+ /* MidnightBSD */
+ DEF_GUID("85D5E45A-237C-11E1-B4B3-E89A8F7FC3A7", N_("MidnightBSD data")),
+ DEF_GUID("85D5E45E-237C-11E1-B4B3-E89A8F7FC3A7", N_("MidnightBSD boot")),
+ DEF_GUID("85D5E45B-237C-11E1-B4B3-E89A8F7FC3A7", N_("MidnightBSD swap")),
+ DEF_GUID("0394Ef8B-237C-11E1-B4B3-E89A8F7FC3A7", N_("MidnightBSD UFS")),
+ DEF_GUID("85D5E45D-237C-11E1-B4B3-E89A8F7FC3A7", N_("MidnightBSD ZFS")),
+ DEF_GUID("85D5E45C-237C-11E1-B4B3-E89A8F7FC3A7", N_("MidnightBSD Vinum")),
+};
+
+/* gpt_entry macros */
+#define gpt_partition_start(_e) le64_to_cpu((_e)->lba_start)
+#define gpt_partition_end(_e) le64_to_cpu((_e)->lba_end)
+
+/*
+ * in-memory fdisk GPT stuff
+ */
+struct fdisk_gpt_label {
+ struct fdisk_label head; /* generic part */
+
+ /* gpt specific part */
+ struct gpt_header *pheader; /* primary header */
+ struct gpt_header *bheader; /* backup header */
+ struct gpt_entry *ents; /* entries (partitions) */
+};
+
+static void gpt_deinit(struct fdisk_label *lb);
+
+static inline struct fdisk_gpt_label *self_label(struct fdisk_context *cxt)
+{
+ return (struct fdisk_gpt_label *) cxt->label;
+}
+
+/*
+ * Returns the partition length, or 0 if end is before beginning.
+ */
+static uint64_t gpt_partition_size(const struct gpt_entry *e)
+{
+ uint64_t start = gpt_partition_start(e);
+ uint64_t end = gpt_partition_end(e);
+
+ return start > end ? 0 : end - start + 1ULL;
+}
+
+/* prints UUID in the real byte order! */
+static void gpt_debug_uuid(const char *mesg, struct gpt_guid *guid)
+{
+ const unsigned char *uuid = (unsigned char *) guid;
+
+ fprintf(stderr, "%s: "
+ "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x\n",
+ mesg,
+ uuid[0], uuid[1], uuid[2], uuid[3],
+ uuid[4], uuid[5],
+ uuid[6], uuid[7],
+ uuid[8], uuid[9],
+ uuid[10], uuid[11], uuid[12], uuid[13], uuid[14],uuid[15]);
+}
+
+/*
+ * UUID is traditionally 16 byte big-endian array, except Intel EFI
+ * specification where the UUID is a structure of little-endian fields.
+ */
+static void swap_efi_guid(struct gpt_guid *uid)
+{
+ uid->time_low = swab32(uid->time_low);
+ uid->time_mid = swab16(uid->time_mid);
+ uid->time_hi_and_version = swab16(uid->time_hi_and_version);
+}
+
+static int string_to_guid(const char *in, struct gpt_guid *guid)
+{
+ if (uuid_parse(in, (unsigned char *) guid)) /* BE */
+ return -1;
+ swap_efi_guid(guid); /* LE */
+ return 0;
+}
+
+static char *guid_to_string(const struct gpt_guid *guid, char *out)
+{
+ struct gpt_guid u = *guid; /* LE */
+
+ swap_efi_guid(&u); /* BE */
+ uuid_unparse_upper((unsigned char *) &u, out);
+
+ return out;
+}
+
+static struct fdisk_parttype *gpt_partition_parttype(
+ struct fdisk_context *cxt,
+ const struct gpt_entry *e)
+{
+ struct fdisk_parttype *t;
+ char str[37];
+
+ guid_to_string(&e->type, str);
+ t = fdisk_label_get_parttype_from_string(cxt->label, str);
+ return t ? : fdisk_new_unknown_parttype(0, str);
+}
+
+static void gpt_entry_set_type(struct gpt_entry *e, struct gpt_guid *uuid)
+{
+ e->type = *uuid;
+ DBG(LABEL, gpt_debug_uuid("new type", &(e->type)));
+}
+
+static void gpt_entry_set_name(struct gpt_entry *e, char *str)
+{
+ char name[GPT_PART_NAME_LEN] = { 0 };
+ size_t i, sz = strlen(str);
+
+ if (sz) {
+ if (sz > GPT_PART_NAME_LEN)
+ sz = GPT_PART_NAME_LEN;
+ memcpy(name, str, sz);
+ }
+
+ for (i = 0; i < GPT_PART_NAME_LEN; i++)
+ e->name[i] = cpu_to_le16((uint16_t) name[i]);
+}
+
+static int gpt_entry_set_uuid(struct gpt_entry *e, char *str)
+{
+ struct gpt_guid uuid;
+ int rc;
+
+ rc = string_to_guid(str, &uuid);
+ if (rc)
+ return rc;
+
+ e->partition_guid = uuid;
+ return 0;
+}
+
+
+static const char *gpt_get_header_revstr(struct gpt_header *header)
+{
+ if (!header)
+ goto unknown;
+
+ switch (header->revision) {
+ case GPT_HEADER_REVISION_V1_02:
+ return "1.2";
+ case GPT_HEADER_REVISION_V1_00:
+ return "1.0";
+ case GPT_HEADER_REVISION_V0_99:
+ return "0.99";
+ default:
+ goto unknown;
+ }
+
+unknown:
+ return "unknown";
+}
+
+static inline int partition_unused(const struct gpt_entry *e)
+{
+ return !memcmp(&e->type, &GPT_UNUSED_ENTRY_GUID,
+ sizeof(struct gpt_guid));
+}
+
+/*
+ * Builds a clean new valid protective MBR - will wipe out any existing data.
+ * Returns 0 on success, otherwise < 0 on error.
+ */
+static int gpt_mknew_pmbr(struct fdisk_context *cxt)
+{
+ struct gpt_legacy_mbr *pmbr = NULL;
+ int rc;
+
+ if (!cxt || !cxt->firstsector)
+ return -ENOSYS;
+
+ rc = fdisk_init_firstsector_buffer(cxt);
+ if (rc)
+ return rc;
+
+ pmbr = (struct gpt_legacy_mbr *) cxt->firstsector;
+
+ pmbr->signature = cpu_to_le16(MSDOS_MBR_SIGNATURE);
+ pmbr->partition_record[0].os_type = EFI_PMBR_OSTYPE;
+ pmbr->partition_record[0].start_sector = 1;
+ pmbr->partition_record[0].end_head = 0xFE;
+ pmbr->partition_record[0].end_sector = 0xFF;
+ pmbr->partition_record[0].end_track = 0xFF;
+ pmbr->partition_record[0].starting_lba = cpu_to_le32(1);
+ pmbr->partition_record[0].size_in_lba =
+ cpu_to_le32(min((uint32_t) cxt->total_sectors - 1, 0xFFFFFFFF));
+
+ return 0;
+}
+
+/* some universal differences between the headers */
+static void gpt_mknew_header_common(struct fdisk_context *cxt,
+ struct gpt_header *header, uint64_t lba)
+{
+ if (!cxt || !header)
+ return;
+
+ header->my_lba = cpu_to_le64(lba);
+
+ if (lba == GPT_PRIMARY_PARTITION_TABLE_LBA) { /* primary */
+ header->alternative_lba = cpu_to_le64(cxt->total_sectors - 1);
+ header->partition_entry_lba = cpu_to_le64(2);
+ } else { /* backup */
+ uint64_t esz = le32_to_cpu(header->npartition_entries) * sizeof(struct gpt_entry);
+ uint64_t esects = (esz + cxt->sector_size - 1) / cxt->sector_size;
+
+ header->alternative_lba = cpu_to_le64(GPT_PRIMARY_PARTITION_TABLE_LBA);
+ header->partition_entry_lba = cpu_to_le64(cxt->total_sectors - 1 - esects);
+ }
+}
+
+/*
+ * Builds a new GPT header (at sector lba) from a backup header2.
+ * If building a primary header, then backup is the secondary, and vice versa.
+ *
+ * Always pass a new (zeroized) header to build upon as we don't
+ * explicitly zero-set some values such as CRCs and reserved.
+ *
+ * Returns 0 on success, otherwise < 0 on error.
+ */
+static int gpt_mknew_header_from_bkp(struct fdisk_context *cxt,
+ struct gpt_header *header,
+ uint64_t lba,
+ struct gpt_header *header2)
+{
+ if (!cxt || !header || !header2)
+ return -ENOSYS;
+
+ header->signature = header2->signature;
+ header->revision = header2->revision;
+ header->size = header2->size;
+ header->npartition_entries = header2->npartition_entries;
+ header->sizeof_partition_entry = header2->sizeof_partition_entry;
+ header->first_usable_lba = header2->first_usable_lba;
+ header->last_usable_lba = header2->last_usable_lba;
+
+ memcpy(&header->disk_guid,
+ &header2->disk_guid, sizeof(header2->disk_guid));
+ gpt_mknew_header_common(cxt, header, lba);
+
+ return 0;
+}
+
+static struct gpt_header *gpt_copy_header(struct fdisk_context *cxt,
+ struct gpt_header *src)
+{
+ struct gpt_header *res;
+
+ if (!cxt || !src)
+ return NULL;
+
+ res = calloc(1, sizeof(*res));
+ if (!res) {
+ fdisk_warn(cxt, _("failed to allocate GPT header"));
+ return NULL;
+ }
+
+ res->my_lba = src->alternative_lba;
+ res->alternative_lba = src->my_lba;
+
+ res->signature = src->signature;
+ res->revision = src->revision;
+ res->size = src->size;
+ res->npartition_entries = src->npartition_entries;
+ res->sizeof_partition_entry = src->sizeof_partition_entry;
+ res->first_usable_lba = src->first_usable_lba;
+ res->last_usable_lba = src->last_usable_lba;
+
+ memcpy(&res->disk_guid, &src->disk_guid, sizeof(src->disk_guid));
+
+
+ if (res->my_lba == GPT_PRIMARY_PARTITION_TABLE_LBA)
+ res->partition_entry_lba = cpu_to_le64(2);
+ else {
+ uint64_t esz = le32_to_cpu(src->npartition_entries) * sizeof(struct gpt_entry);
+ uint64_t esects = (esz + cxt->sector_size - 1) / cxt->sector_size;
+
+ res->partition_entry_lba = cpu_to_le64(cxt->total_sectors - 1 - esects);
+ }
+
+ return res;
+}
+
+static void count_first_last_lba(struct fdisk_context *cxt,
+ uint64_t *first, uint64_t *last)
+{
+ uint64_t esz = 0;
+
+ assert(cxt);
+
+ esz = sizeof(struct gpt_entry) * GPT_NPARTITIONS / cxt->sector_size;
+ *last = cxt->total_sectors - 2 - esz;
+ *first = esz + 2;
+
+ if (*first < cxt->first_lba && cxt->first_lba < *last)
+ /* Align according to topology */
+ *first = cxt->first_lba;
+}
+
+/*
+ * Builds a clean new GPT header (currently under revision 1.0).
+ *
+ * Always pass a new (zeroized) header to build upon as we don't
+ * explicitly zero-set some values such as CRCs and reserved.
+ *
+ * Returns 0 on success, otherwise < 0 on error.
+ */
+static int gpt_mknew_header(struct fdisk_context *cxt,
+ struct gpt_header *header, uint64_t lba)
+{
+ uint64_t first, last;
+ int has_id = 0;
+
+ if (!cxt || !header)
+ return -ENOSYS;
+
+ header->signature = cpu_to_le64(GPT_HEADER_SIGNATURE);
+ header->revision = cpu_to_le32(GPT_HEADER_REVISION_V1_00);
+ header->size = cpu_to_le32(sizeof(struct gpt_header));
+
+ /*
+ * 128 partitions are the default. It can go beyond that, but
+ * we're creating a de facto header here, so no funny business.
+ */
+ header->npartition_entries = cpu_to_le32(GPT_NPARTITIONS);
+ header->sizeof_partition_entry = cpu_to_le32(sizeof(struct gpt_entry));
+
+ count_first_last_lba(cxt, &first, &last);
+ header->first_usable_lba = cpu_to_le64(first);
+ header->last_usable_lba = cpu_to_le64(last);
+
+ gpt_mknew_header_common(cxt, header, lba);
+
+ if (cxt->script) {
+ const char *id = fdisk_script_get_header(cxt->script, "label-id");
+ if (id && string_to_guid(id, &header->disk_guid) == 0)
+ has_id = 1;
+ }
+
+ if (!has_id) {
+ uuid_generate_random((unsigned char *) &header->disk_guid);
+ swap_efi_guid(&header->disk_guid);
+ }
+ return 0;
+}
+
+/*
+ * Checks if there is a valid protective MBR partition table.
+ * Returns 0 if it is invalid or failure. Otherwise, return
+ * GPT_MBR_PROTECTIVE or GPT_MBR_HYBRID, depeding on the detection.
+ */
+static int valid_pmbr(struct fdisk_context *cxt)
+{
+ int i, part = 0, ret = 0; /* invalid by default */
+ struct gpt_legacy_mbr *pmbr = NULL;
+ uint32_t sz_lba = 0;
+
+ if (!cxt->firstsector)
+ goto done;
+
+ pmbr = (struct gpt_legacy_mbr *) cxt->firstsector;
+
+ if (le16_to_cpu(pmbr->signature) != MSDOS_MBR_SIGNATURE)
+ goto done;
+
+ /* LBA of the GPT partition header */
+ if (pmbr->partition_record[0].starting_lba !=
+ cpu_to_le32(GPT_PRIMARY_PARTITION_TABLE_LBA))
+ goto done;
+
+ /* seems like a valid MBR was found, check DOS primary partitions */
+ for (i = 0; i < 4; i++) {
+ if (pmbr->partition_record[i].os_type == EFI_PMBR_OSTYPE) {
+ /*
+ * Ok, we at least know that there's a protective MBR,
+ * now check if there are other partition types for
+ * hybrid MBR.
+ */
+ part = i;
+ ret = GPT_MBR_PROTECTIVE;
+ goto check_hybrid;
+ }
+ }
+
+ if (ret != GPT_MBR_PROTECTIVE)
+ goto done;
+check_hybrid:
+ for (i = 0 ; i < 4; i++) {
+ if ((pmbr->partition_record[i].os_type != EFI_PMBR_OSTYPE) &&
+ (pmbr->partition_record[i].os_type != 0x00))
+ ret = GPT_MBR_HYBRID;
+ }
+
+ /*
+ * Protective MBRs take up the lesser of the whole disk
+ * or 2 TiB (32bit LBA), ignoring the rest of the disk.
+ * Some partitioning programs, nonetheless, choose to set
+ * the size to the maximum 32-bit limitation, disregarding
+ * the disk size.
+ *
+ * Hybrid MBRs do not necessarily comply with this.
+ *
+ * Consider a bad value here to be a warning to support dd-ing
+ * an image from a smaller disk to a bigger disk.
+ */
+ if (ret == GPT_MBR_PROTECTIVE) {
+ sz_lba = le32_to_cpu(pmbr->partition_record[part].size_in_lba);
+ if (sz_lba != (uint32_t) cxt->total_sectors - 1 && sz_lba != 0xFFFFFFFF) {
+ fdisk_warnx(cxt, _("GPT PMBR size mismatch (%u != %u) "
+ "will be corrected by w(rite)."),
+ sz_lba,
+ (uint32_t) cxt->total_sectors - 1);
+ fdisk_label_set_changed(cxt->label, 1);
+ }
+ }
+done:
+ return ret;
+}
+
+static uint64_t last_lba(struct fdisk_context *cxt)
+{
+ struct stat s;
+ uint64_t sectors = 0;
+
+ memset(&s, 0, sizeof(s));
+ if (fstat(cxt->dev_fd, &s) == -1) {
+ fdisk_warn(cxt, _("gpt: stat() failed"));
+ return 0;
+ }
+
+ if (S_ISBLK(s.st_mode))
+ sectors = cxt->total_sectors - 1;
+ else if (S_ISREG(s.st_mode))
+ sectors = ((uint64_t) s.st_size /
+ (uint64_t) cxt->sector_size) - 1ULL;
+ else
+ fdisk_warnx(cxt, _("gpt: cannot handle files with mode %o"), s.st_mode);
+
+ DBG(LABEL, ul_debug("GPT last LBA: %ju", sectors));
+ return sectors;
+}
+
+static ssize_t read_lba(struct fdisk_context *cxt, uint64_t lba,
+ void *buffer, const size_t bytes)
+{
+ off_t offset = lba * cxt->sector_size;
+
+ if (lseek(cxt->dev_fd, offset, SEEK_SET) == (off_t) -1)
+ return -1;
+ return read(cxt->dev_fd, buffer, bytes) != bytes;
+}
+
+
+/* Returns the GPT entry array */
+static struct gpt_entry *gpt_read_entries(struct fdisk_context *cxt,
+ struct gpt_header *header)
+{
+ ssize_t sz;
+ struct gpt_entry *ret = NULL;
+ off_t offset;
+
+ assert(cxt);
+ assert(header);
+
+ sz = le32_to_cpu(header->npartition_entries) *
+ le32_to_cpu(header->sizeof_partition_entry);
+
+ ret = calloc(1, sz);
+ if (!ret)
+ return NULL;
+ offset = le64_to_cpu(header->partition_entry_lba) *
+ cxt->sector_size;
+
+ if (offset != lseek(cxt->dev_fd, offset, SEEK_SET))
+ goto fail;
+ if (sz != read(cxt->dev_fd, ret, sz))
+ goto fail;
+
+ return ret;
+
+fail:
+ free(ret);
+ return NULL;
+}
+
+static inline uint32_t count_crc32(const unsigned char *buf, size_t len)
+{
+ return (crc32(~0L, buf, len) ^ ~0L);
+}
+
+/*
+ * Recompute header and partition array 32bit CRC checksums.
+ * This function does not fail - if there's corruption, then it
+ * will be reported when checksuming it again (ie: probing or verify).
+ */
+static void gpt_recompute_crc(struct gpt_header *header, struct gpt_entry *ents)
+{
+ uint32_t crc = 0;
+ size_t entry_sz = 0;
+
+ if (!header)
+ return;
+
+ /* header CRC */
+ header->crc32 = 0;
+ crc = count_crc32((unsigned char *) header, le32_to_cpu(header->size));
+ header->crc32 = cpu_to_le32(crc);
+
+ /* partition entry array CRC */
+ header->partition_entry_array_crc32 = 0;
+ entry_sz = le32_to_cpu(header->npartition_entries) *
+ le32_to_cpu(header->sizeof_partition_entry);
+
+ crc = count_crc32((unsigned char *) ents, entry_sz);
+ header->partition_entry_array_crc32 = cpu_to_le32(crc);
+}
+
+/*
+ * Compute the 32bit CRC checksum of the partition table header.
+ * Returns 1 if it is valid, otherwise 0.
+ */
+static int gpt_check_header_crc(struct gpt_header *header, struct gpt_entry *ents)
+{
+ uint32_t crc, orgcrc = le32_to_cpu(header->crc32);
+
+ header->crc32 = 0;
+ crc = count_crc32((unsigned char *) header, le32_to_cpu(header->size));
+ header->crc32 = cpu_to_le32(orgcrc);
+
+ if (crc == le32_to_cpu(header->crc32))
+ return 1;
+
+ /*
+ * If we have checksum mismatch it may be due to stale data,
+ * like a partition being added or deleted. Recompute the CRC again
+ * and make sure this is not the case.
+ */
+ if (ents) {
+ gpt_recompute_crc(header, ents);
+ orgcrc = le32_to_cpu(header->crc32);
+ header->crc32 = 0;
+ crc = count_crc32((unsigned char *) header, le32_to_cpu(header->size));
+ header->crc32 = cpu_to_le32(orgcrc);
+
+ return crc == le32_to_cpu(header->crc32);
+ }
+
+ return 0;
+}
+
+/*
+ * It initializes the partition entry array.
+ * Returns 1 if the checksum is valid, otherwise 0.
+ */
+static int gpt_check_entryarr_crc(struct gpt_header *header,
+ struct gpt_entry *ents)
+{
+ int ret = 0;
+ ssize_t entry_sz;
+ uint32_t crc;
+
+ if (!header || !ents)
+ goto done;
+
+ entry_sz = le32_to_cpu(header->npartition_entries) *
+ le32_to_cpu(header->sizeof_partition_entry);
+
+ if (!entry_sz)
+ goto done;
+
+ crc = count_crc32((unsigned char *) ents, entry_sz);
+ ret = (crc == le32_to_cpu(header->partition_entry_array_crc32));
+done:
+ return ret;
+}
+
+static int gpt_check_lba_sanity(struct fdisk_context *cxt, struct gpt_header *header)
+{
+ int ret = 0;
+ uint64_t lu, fu, lastlba = last_lba(cxt);
+
+ fu = le64_to_cpu(header->first_usable_lba);
+ lu = le64_to_cpu(header->last_usable_lba);
+
+ /* check if first and last usable LBA make sense */
+ if (lu < fu) {
+ DBG(LABEL, ul_debug("error: header last LBA is before first LBA"));
+ goto done;
+ }
+
+ /* check if first and last usable LBAs with the disk's last LBA */
+ if (fu > lastlba || lu > lastlba) {
+ DBG(LABEL, ul_debug("error: header LBAs are after the disk's last LBA"));
+ goto done;
+ }
+
+ /* the header has to be outside usable range */
+ if (fu < GPT_PRIMARY_PARTITION_TABLE_LBA &&
+ GPT_PRIMARY_PARTITION_TABLE_LBA < lu) {
+ DBG(LABEL, ul_debug("error: header outside of usable range"));
+ goto done;
+ }
+
+ ret = 1; /* sane */
+done:
+ return ret;
+}
+
+/* Check if there is a valid header signature */
+static int gpt_check_signature(struct gpt_header *header)
+{
+ return header->signature == cpu_to_le64(GPT_HEADER_SIGNATURE);
+}
+
+/*
+ * Return the specified GPT Header, or NULL upon failure/invalid.
+ * Note that all tests must pass to ensure a valid header,
+ * we do not rely on only testing the signature for a valid probe.
+ */
+static struct gpt_header *gpt_read_header(struct fdisk_context *cxt,
+ uint64_t lba,
+ struct gpt_entry **_ents)
+{
+ struct gpt_header *header = NULL;
+ struct gpt_entry *ents = NULL;
+ uint32_t hsz;
+
+ if (!cxt)
+ return NULL;
+
+ header = calloc(1, sizeof(*header));
+ if (!header)
+ return NULL;
+
+ /* read and verify header */
+ if (read_lba(cxt, lba, header, sizeof(struct gpt_header)) != 0)
+ goto invalid;
+
+ if (!gpt_check_signature(header))
+ goto invalid;
+
+ if (!gpt_check_header_crc(header, NULL))
+ goto invalid;
+
+ /* read and verify entries */
+ ents = gpt_read_entries(cxt, header);
+ if (!ents)
+ goto invalid;
+
+ if (!gpt_check_entryarr_crc(header, ents))
+ goto invalid;
+
+ if (!gpt_check_lba_sanity(cxt, header))
+ goto invalid;
+
+ /* valid header must be at MyLBA */
+ if (le64_to_cpu(header->my_lba) != lba)
+ goto invalid;
+
+ /* make sure header size is between 92 and sector size bytes */
+ hsz = le32_to_cpu(header->size);
+ if (hsz < GPT_HEADER_MINSZ || hsz > cxt->sector_size)
+ goto invalid;
+
+ if (_ents)
+ *_ents = ents;
+ else
+ free(ents);
+
+ DBG(LABEL, ul_debug("found valid GPT Header on LBA %ju", lba));
+ return header;
+invalid:
+ free(header);
+ free(ents);
+
+ DBG(LABEL, ul_debug("read GPT Header on LBA %ju failed", lba));
+ return NULL;
+}
+
+
+static int gpt_locate_disklabel(struct fdisk_context *cxt, int n,
+ const char **name, off_t *offset, size_t *size)
+{
+ struct fdisk_gpt_label *gpt;
+
+ assert(cxt);
+
+ *name = NULL;
+ *offset = 0;
+ *size = 0;
+
+ switch (n) {
+ case 0:
+ *name = "PMBR";
+ *offset = 0;
+ *size = 512;
+ break;
+ case 1:
+ *name = _("GPT Header");
+ *offset = GPT_PRIMARY_PARTITION_TABLE_LBA * cxt->sector_size;
+ *size = sizeof(struct gpt_header);
+ break;
+ case 2:
+ *name = _("GPT Entries");
+ gpt = self_label(cxt);
+ *offset = le64_to_cpu(gpt->pheader->partition_entry_lba) * cxt->sector_size;
+ *size = le32_to_cpu(gpt->pheader->npartition_entries) *
+ le32_to_cpu(gpt->pheader->sizeof_partition_entry);
+ break;
+ default:
+ return 1; /* no more chunks */
+ }
+
+ return 0;
+}
+
+
+
+/*
+ * Returns the number of partitions that are in use.
+ */
+static unsigned partitions_in_use(struct gpt_header *header,
+ struct gpt_entry *ents)
+{
+ uint32_t i, used = 0;
+
+ if (!header || ! ents)
+ return 0;
+
+ for (i = 0; i < le32_to_cpu(header->npartition_entries); i++)
+ if (!partition_unused(&ents[i]))
+ used++;
+ return used;
+}
+
+
+/*
+ * Check if a partition is too big for the disk (sectors).
+ * Returns the faulting partition number, otherwise 0.
+ */
+static uint32_t check_too_big_partitions(struct gpt_header *header,
+ struct gpt_entry *ents, uint64_t sectors)
+{
+ uint32_t i;
+
+ for (i = 0; i < le32_to_cpu(header->npartition_entries); i++) {
+ if (partition_unused(&ents[i]))
+ continue;
+ if (gpt_partition_end(&ents[i]) >= sectors)
+ return i + 1;
+ }
+
+ return 0;
+}
+
+/*
+ * Check if a partition ends before it begins
+ * Returns the faulting partition number, otherwise 0.
+ */
+static uint32_t check_start_after_end_paritions(struct gpt_header *header,
+ struct gpt_entry *ents)
+{
+ uint32_t i;
+
+ for (i = 0; i < le32_to_cpu(header->npartition_entries); i++) {
+ if (partition_unused(&ents[i]))
+ continue;
+ if (gpt_partition_start(&ents[i]) > gpt_partition_end(&ents[i]))
+ return i + 1;
+ }
+
+ return 0;
+}
+
+/*
+ * Check if partition e1 overlaps with partition e2.
+ */
+static inline int partition_overlap(struct gpt_entry *e1, struct gpt_entry *e2)
+{
+ uint64_t start1 = gpt_partition_start(e1);
+ uint64_t end1 = gpt_partition_end(e1);
+ uint64_t start2 = gpt_partition_start(e2);
+ uint64_t end2 = gpt_partition_end(e2);
+
+ return (start1 && start2 && (start1 <= end2) != (end1 < start2));
+}
+
+/*
+ * Find any partitions that overlap.
+ */
+static uint32_t check_overlap_partitions(struct gpt_header *header,
+ struct gpt_entry *ents)
+{
+ uint32_t i, j;
+
+ for (i = 0; i < le32_to_cpu(header->npartition_entries); i++)
+ for (j = 0; j < i; j++) {
+ if (partition_unused(&ents[i]) ||
+ partition_unused(&ents[j]))
+ continue;
+ if (partition_overlap(&ents[i], &ents[j])) {
+ DBG(LABEL, ul_debug("GPT partitions overlap detected [%u vs. %u]", i, j));
+ return i + 1;
+ }
+ }
+
+ return 0;
+}
+
+/*
+ * Find the first available block after the starting point; returns 0 if
+ * there are no available blocks left, or error. From gdisk.
+ */
+static uint64_t find_first_available(struct gpt_header *header,
+ struct gpt_entry *ents, uint64_t start)
+{
+ uint64_t first;
+ uint32_t i, first_moved = 0;
+
+ uint64_t fu, lu;
+
+ if (!header || !ents)
+ return 0;
+
+ fu = le64_to_cpu(header->first_usable_lba);
+ lu = le64_to_cpu(header->last_usable_lba);
+
+ /*
+ * Begin from the specified starting point or from the first usable
+ * LBA, whichever is greater...
+ */
+ first = start < fu ? fu : start;
+
+ /*
+ * Now search through all partitions; if first is within an
+ * existing partition, move it to the next sector after that
+ * partition and repeat. If first was moved, set firstMoved
+ * flag; repeat until firstMoved is not set, so as to catch
+ * cases where partitions are out of sequential order....
+ */
+ do {
+ first_moved = 0;
+ for (i = 0; i < le32_to_cpu(header->npartition_entries); i++) {
+ if (partition_unused(&ents[i]))
+ continue;
+ if (first < gpt_partition_start(&ents[i]))
+ continue;
+ if (first <= gpt_partition_end(&ents[i])) {
+ first = gpt_partition_end(&ents[i]) + 1;
+ first_moved = 1;
+ }
+ }
+ } while (first_moved == 1);
+
+ if (first > lu)
+ first = 0;
+
+ return first;
+}
+
+
+/* Returns last available sector in the free space pointed to by start. From gdisk. */
+static uint64_t find_last_free(struct gpt_header *header,
+ struct gpt_entry *ents, uint64_t start)
+{
+ uint32_t i;
+ uint64_t nearest_start;
+
+ if (!header || !ents)
+ return 0;
+
+ nearest_start = le64_to_cpu(header->last_usable_lba);
+
+ for (i = 0; i < le32_to_cpu(header->npartition_entries); i++) {
+ uint64_t ps = gpt_partition_start(&ents[i]);
+
+ if (nearest_start > ps && ps > start)
+ nearest_start = ps - 1;
+ }
+
+ return nearest_start;
+}
+
+/* Returns the last free sector on the disk. From gdisk. */
+static uint64_t find_last_free_sector(struct gpt_header *header,
+ struct gpt_entry *ents)
+{
+ uint32_t i, last_moved;
+ uint64_t last = 0;
+
+ if (!header || !ents)
+ goto done;
+
+ /* start by assuming the last usable LBA is available */
+ last = le64_to_cpu(header->last_usable_lba);
+ do {
+ last_moved = 0;
+ for (i = 0; i < le32_to_cpu(header->npartition_entries); i++) {
+ if ((last >= gpt_partition_start(&ents[i])) &&
+ (last <= gpt_partition_end(&ents[i]))) {
+ last = gpt_partition_start(&ents[i]) - 1;
+ last_moved = 1;
+ }
+ }
+ } while (last_moved == 1);
+done:
+ return last;
+}
+
+/*
+ * Finds the first available sector in the largest block of unallocated
+ * space on the disk. Returns 0 if there are no available blocks left.
+ * From gdisk.
+ */
+static uint64_t find_first_in_largest(struct gpt_header *header,
+ struct gpt_entry *ents)
+{
+ uint64_t start = 0, first_sect, last_sect;
+ uint64_t segment_size, selected_size = 0, selected_segment = 0;
+
+ if (!header || !ents)
+ goto done;
+
+ do {
+ first_sect = find_first_available(header, ents, start);
+ if (first_sect != 0) {
+ last_sect = find_last_free(header, ents, first_sect);
+ segment_size = last_sect - first_sect + 1;
+
+ if (segment_size > selected_size) {
+ selected_size = segment_size;
+ selected_segment = first_sect;
+ }
+ start = last_sect + 1;
+ }
+ } while (first_sect != 0);
+
+done:
+ return selected_segment;
+}
+
+/*
+ * Find the total number of free sectors, the number of segments in which
+ * they reside, and the size of the largest of those segments. From gdisk.
+ */
+static uint64_t get_free_sectors(struct fdisk_context *cxt, struct gpt_header *header,
+ struct gpt_entry *ents, uint32_t *nsegments,
+ uint64_t *largest_segment)
+{
+ uint32_t num = 0;
+ uint64_t first_sect, last_sect;
+ uint64_t largest_seg = 0, segment_sz;
+ uint64_t totfound = 0, start = 0; /* starting point for each search */
+
+ if (!cxt->total_sectors)
+ goto done;
+
+ do {
+ first_sect = find_first_available(header, ents, start);
+ if (first_sect) {
+ last_sect = find_last_free(header, ents, first_sect);
+ segment_sz = last_sect - first_sect + 1;
+
+ if (segment_sz > largest_seg)
+ largest_seg = segment_sz;
+ totfound += segment_sz;
+ num++;
+ start = last_sect + 1;
+ }
+ } while (first_sect);
+
+done:
+ if (nsegments)
+ *nsegments = num;
+ if (largest_segment)
+ *largest_segment = largest_seg;
+
+ return totfound;
+}
+
+static int gpt_probe_label(struct fdisk_context *cxt)
+{
+ int mbr_type;
+ struct fdisk_gpt_label *gpt;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+
+ /* TODO: it would be nice to support scenario when GPT headers are OK,
+ * but PMBR is corrupt */
+ mbr_type = valid_pmbr(cxt);
+ if (!mbr_type)
+ goto failed;
+
+ DBG(LABEL, ul_debug("found a %s MBR", mbr_type == GPT_MBR_PROTECTIVE ?
+ "protective" : "hybrid"));
+
+ /* primary header */
+ gpt->pheader = gpt_read_header(cxt, GPT_PRIMARY_PARTITION_TABLE_LBA,
+ &gpt->ents);
+
+ if (gpt->pheader)
+ /* primary OK, try backup from alternative LBA */
+ gpt->bheader = gpt_read_header(cxt,
+ le64_to_cpu(gpt->pheader->alternative_lba),
+ NULL);
+ else
+ /* primary corrupted -- try last LBA */
+ gpt->bheader = gpt_read_header(cxt, last_lba(cxt), &gpt->ents);
+
+ if (!gpt->pheader && !gpt->bheader)
+ goto failed;
+
+ /* primary OK, backup corrupted -- recovery */
+ if (gpt->pheader && !gpt->bheader) {
+ fdisk_warnx(cxt, _("The backup GPT table is corrupt, but the "
+ "primary appears OK, so that will be used."));
+ gpt->bheader = gpt_copy_header(cxt, gpt->pheader);
+ if (!gpt->bheader)
+ goto failed;
+ gpt_recompute_crc(gpt->bheader, gpt->ents);
+
+ /* primary corrupted, backup OK -- recovery */
+ } else if (!gpt->pheader && gpt->bheader) {
+ fdisk_warnx(cxt, _("The primary GPT table is corrupt, but the "
+ "backup appears OK, so that will be used."));
+ gpt->pheader = gpt_copy_header(cxt, gpt->bheader);
+ if (!gpt->pheader)
+ goto failed;
+ gpt_recompute_crc(gpt->pheader, gpt->ents);
+ }
+
+ cxt->label->nparts_max = le32_to_cpu(gpt->pheader->npartition_entries);
+ cxt->label->nparts_cur = partitions_in_use(gpt->pheader, gpt->ents);
+ return 1;
+failed:
+ DBG(LABEL, ul_debug("GPT probe failed"));
+ gpt_deinit(cxt->label);
+ return 0;
+}
+
+/*
+ * Stolen from libblkid - can be removed once partition semantics
+ * are added to the fdisk API.
+ */
+static char *encode_to_utf8(unsigned char *src, size_t count)
+{
+ uint16_t c;
+ char *dest;
+ size_t i, j, len = count;
+
+ dest = calloc(1, count);
+ if (!dest)
+ return NULL;
+
+ for (j = i = 0; i + 2 <= count; i += 2) {
+ /* always little endian */
+ c = (src[i+1] << 8) | src[i];
+ if (c == 0) {
+ dest[j] = '\0';
+ break;
+ } else if (c < 0x80) {
+ if (j+1 >= len)
+ break;
+ dest[j++] = (uint8_t) c;
+ } else if (c < 0x800) {
+ if (j+2 >= len)
+ break;
+ dest[j++] = (uint8_t) (0xc0 | (c >> 6));
+ dest[j++] = (uint8_t) (0x80 | (c & 0x3f));
+ } else {
+ if (j+3 >= len)
+ break;
+ dest[j++] = (uint8_t) (0xe0 | (c >> 12));
+ dest[j++] = (uint8_t) (0x80 | ((c >> 6) & 0x3f));
+ dest[j++] = (uint8_t) (0x80 | (c & 0x3f));
+ }
+ }
+ dest[j] = '\0';
+
+ return dest;
+}
+
+static int gpt_entry_attrs_to_string(struct gpt_entry *e, char **res)
+{
+ unsigned int n, count = 0;
+ size_t l;
+ char *bits, *p;
+ uint64_t attrs;
+
+ assert(e);
+ assert(res);
+
+ *res = NULL;
+ attrs = le64_to_cpu(e->attrs);
+ if (!attrs)
+ return 0; /* no attributes at all */
+
+ bits = (char *) &attrs;
+
+ /* Note that sizeof() is correct here, we need separators between
+ * the strings so also count \0 is correct */
+ *res = calloc(1, sizeof(GPT_ATTRSTR_NOBLOCK) +
+ sizeof(GPT_ATTRSTR_REQ) +
+ sizeof(GPT_ATTRSTR_LEGACY) +
+ sizeof("GUID:") + (GPT_ATTRBIT_GUID_COUNT * 3));
+ if (!*res)
+ return -errno;
+
+ p = *res;
+ if (isset(bits, GPT_ATTRBIT_REQ)) {
+ memcpy(p, GPT_ATTRSTR_REQ, (l = sizeof(GPT_ATTRSTR_REQ)));
+ p += l - 1;
+ }
+ if (isset(bits, GPT_ATTRBIT_NOBLOCK)) {
+ if (p > *res)
+ *p++ = ' ';
+ memcpy(p, GPT_ATTRSTR_NOBLOCK, (l = sizeof(GPT_ATTRSTR_NOBLOCK)));
+ p += l - 1;
+ }
+ if (isset(bits, GPT_ATTRBIT_LEGACY)) {
+ if (p > *res)
+ *p++ = ' ';
+ memcpy(p, GPT_ATTRSTR_LEGACY, (l = sizeof(GPT_ATTRSTR_LEGACY)));
+ p += l - 1;
+ }
+
+ for (n = GPT_ATTRBIT_GUID_FIRST;
+ n < GPT_ATTRBIT_GUID_FIRST + GPT_ATTRBIT_GUID_COUNT; n++) {
+
+ if (!isset(bits, n))
+ continue;
+ if (!count) {
+ if (p > *res)
+ *p++ = ' ';
+ p += sprintf(p, "GUID:%u", n);
+ } else
+ p += sprintf(p, ",%u", n);
+ count++;
+ }
+
+ return 0;
+}
+
+static int gpt_entry_attrs_from_string(
+ struct fdisk_context *cxt,
+ struct gpt_entry *e,
+ const char *str)
+{
+ const char *p = str;
+ uint64_t attrs = 0;
+ char *bits;
+
+ assert(e);
+ assert(p);
+
+ DBG(LABEL, ul_debug("GPT: parsing string attributes '%s'", p));
+
+ bits = (char *) &attrs;
+
+ while (p && *p) {
+ int bit = -1;
+
+ while (isblank(*p)) p++;
+ if (!*p)
+ break;
+
+ DBG(LABEL, ul_debug(" parsing item '%s'", p));
+
+ if (strncmp(p, "GUID:", 5) == 0) {
+ p += 5;
+ continue;
+ } else if (strncmp(p, GPT_ATTRSTR_REQ,
+ sizeof(GPT_ATTRSTR_REQ) - 1) == 0) {
+ bit = GPT_ATTRBIT_REQ;
+ p += sizeof(GPT_ATTRSTR_REQ) - 1;
+ } else if (strncmp(p, GPT_ATTRSTR_LEGACY,
+ sizeof(GPT_ATTRSTR_LEGACY) - 1) == 0) {
+ bit = GPT_ATTRBIT_LEGACY;
+ p += sizeof(GPT_ATTRSTR_LEGACY) - 1;
+ } else if (strncmp(p, GPT_ATTRSTR_NOBLOCK,
+ sizeof(GPT_ATTRSTR_NOBLOCK) - 1) == 0) {
+ bit = GPT_ATTRBIT_NOBLOCK;
+ p += sizeof(GPT_ATTRSTR_NOBLOCK) - 1;
+ } else if (isdigit((unsigned int) *p)) {
+ char *end = NULL;
+
+ errno = 0;
+ bit = strtol(p, &end, 0);
+ if (errno || !end || end == str
+ || bit < GPT_ATTRBIT_GUID_FIRST
+ || bit >= GPT_ATTRBIT_GUID_FIRST + GPT_ATTRBIT_GUID_COUNT)
+ bit = -1;
+ else
+ p = end;
+ }
+
+ if (bit < 0) {
+ fdisk_warnx(cxt, _("unssuported GPT attribute bit '%s'"), p);
+ return -EINVAL;
+ }
+
+ setbit(bits, bit);
+
+ while (isblank(*p)) p++;
+ if (*p == ',')
+ p++;
+ }
+
+ e->attrs = cpu_to_le64(attrs);
+ return 0;
+}
+
+static int gpt_get_partition(struct fdisk_context *cxt, size_t n,
+ struct fdisk_partition *pa)
+{
+ struct fdisk_gpt_label *gpt;
+ struct gpt_entry *e;
+ char u_str[37];
+ int rc = 0;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+
+ if ((uint32_t) n >= le32_to_cpu(gpt->pheader->npartition_entries))
+ return -EINVAL;
+
+ gpt = self_label(cxt);
+ e = &gpt->ents[n];
+
+ pa->used = !partition_unused(e) || gpt_partition_start(e);
+ if (!pa->used)
+ return 0;
+
+ pa->start = gpt_partition_start(e);
+ pa->size = gpt_partition_size(e);
+ pa->type = gpt_partition_parttype(cxt, e);
+
+ if (guid_to_string(&e->partition_guid, u_str)) {
+ pa->uuid = strdup(u_str);
+ if (!pa->uuid) {
+ rc = -errno;
+ goto done;
+ }
+ } else
+ pa->uuid = NULL;
+
+ rc = gpt_entry_attrs_to_string(e, &pa->attrs);
+ if (rc)
+ goto done;
+
+ pa->name = encode_to_utf8((unsigned char *)e->name, sizeof(e->name));
+ return 0;
+done:
+ fdisk_reset_partition(pa);
+ return rc;
+}
+
+
+static int gpt_set_partition(struct fdisk_context *cxt, size_t n,
+ struct fdisk_partition *pa)
+{
+ struct fdisk_gpt_label *gpt;
+ struct gpt_entry *e;
+ int rc = 0;
+ uint64_t start, end;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+
+ if ((uint32_t) n >= le32_to_cpu(gpt->pheader->npartition_entries))
+ return -EINVAL;
+
+ FDISK_INIT_UNDEF(start);
+ FDISK_INIT_UNDEF(end);
+
+ gpt = self_label(cxt);
+ e = &gpt->ents[n];
+
+ if (pa->uuid) {
+ char new_u[37], old_u[37];
+
+ guid_to_string(&e->partition_guid, old_u);
+ rc = gpt_entry_set_uuid(e, pa->uuid);
+ if (rc)
+ return rc;
+ guid_to_string(&e->partition_guid, new_u);
+ fdisk_info(cxt, _("Partition UUID changed from %s to %s."),
+ old_u, new_u);
+ }
+
+ if (pa->name) {
+ char *old = encode_to_utf8((unsigned char *)e->name, sizeof(e->name));
+ gpt_entry_set_name(e, pa->name);
+
+ fdisk_info(cxt, _("Partition name changed from '%s' to '%.*s'."),
+ old, (int) GPT_PART_NAME_LEN, pa->name);
+ free(old);
+ }
+
+ if (pa->type && pa->type->typestr) {
+ struct gpt_guid typeid;
+
+ rc = string_to_guid(pa->type->typestr, &typeid);
+ if (rc)
+ return rc;
+ gpt_entry_set_type(e, &typeid);
+ }
+ if (pa->attrs) {
+ rc = gpt_entry_attrs_from_string(cxt, e, pa->attrs);
+ if (rc)
+ return rc;
+ }
+
+ if (fdisk_partition_has_start(pa))
+ start = pa->start;
+ if (fdisk_partition_has_size(pa))
+ end = gpt_partition_start(e) + pa->size - 1ULL;
+
+ if (pa->end_follow_default) {
+ /* enlarge */
+ if (!FDISK_IS_UNDEF(start))
+ start = gpt_partition_start(e);
+ end = find_last_free(gpt->bheader, gpt->ents, start);
+ if (!end)
+ FDISK_INIT_UNDEF(end);
+ }
+
+ if (!FDISK_IS_UNDEF(start))
+ e->lba_start = cpu_to_le64(start);
+ if (!FDISK_IS_UNDEF(end))
+ e->lba_end = cpu_to_le64(end);
+
+ gpt_recompute_crc(gpt->pheader, gpt->ents);
+ gpt_recompute_crc(gpt->bheader, gpt->ents);
+
+ fdisk_label_set_changed(cxt->label, 1);
+ return rc;
+}
+
+
+/*
+ * List label partitions.
+ */
+static int gpt_list_disklabel(struct fdisk_context *cxt)
+{
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ if (fdisk_is_details(cxt)) {
+ struct gpt_header *h = self_label(cxt)->pheader;
+
+ fdisk_info(cxt, _("First LBA: %ju"), h->first_usable_lba);
+ fdisk_info(cxt, _("Last LBA: %ju"), h->last_usable_lba);
+ /* TRANSLATORS: The LBA (Logical Block Address) of the backup GPT header. */
+ fdisk_info(cxt, _("Alternative LBA: %ju"), h->alternative_lba);
+ /* TRANSLATORS: The start of the array of partition entries. */
+ fdisk_info(cxt, _("Partition entries LBA: %ju"), h->partition_entry_lba);
+ fdisk_info(cxt, _("Allocated partition entries: %u"), h->npartition_entries);
+ }
+
+ return 0;
+}
+
+/*
+ * Write partitions.
+ * Returns 0 on success, or corresponding error otherwise.
+ */
+static int gpt_write_partitions(struct fdisk_context *cxt,
+ struct gpt_header *header, struct gpt_entry *ents)
+{
+ off_t offset = le64_to_cpu(header->partition_entry_lba) * cxt->sector_size;
+ uint32_t nparts = le32_to_cpu(header->npartition_entries);
+ uint32_t totwrite = nparts * le32_to_cpu(header->sizeof_partition_entry);
+ ssize_t rc;
+
+ if (offset != lseek(cxt->dev_fd, offset, SEEK_SET))
+ goto fail;
+
+ rc = write(cxt->dev_fd, ents, totwrite);
+ if (rc > 0 && totwrite == (uint32_t) rc)
+ return 0;
+fail:
+ return -errno;
+}
+
+/*
+ * Write a GPT header to a specified LBA
+ * Returns 0 on success, or corresponding error otherwise.
+ */
+static int gpt_write_header(struct fdisk_context *cxt,
+ struct gpt_header *header, uint64_t lba)
+{
+ off_t offset = lba * cxt->sector_size;
+
+ if (offset != lseek(cxt->dev_fd, offset, SEEK_SET))
+ goto fail;
+ if (cxt->sector_size ==
+ (size_t) write(cxt->dev_fd, header, cxt->sector_size))
+ return 0;
+fail:
+ return -errno;
+}
+
+/*
+ * Write the protective MBR.
+ * Returns 0 on success, or corresponding error otherwise.
+ */
+static int gpt_write_pmbr(struct fdisk_context *cxt)
+{
+ off_t offset;
+ struct gpt_legacy_mbr *pmbr = NULL;
+
+ assert(cxt);
+ assert(cxt->firstsector);
+
+ pmbr = (struct gpt_legacy_mbr *) cxt->firstsector;
+
+ /* zero out the legacy partitions */
+ memset(pmbr->partition_record, 0, sizeof(pmbr->partition_record));
+
+ pmbr->signature = cpu_to_le16(MSDOS_MBR_SIGNATURE);
+ pmbr->partition_record[0].os_type = EFI_PMBR_OSTYPE;
+ pmbr->partition_record[0].start_sector = 1;
+ pmbr->partition_record[0].end_head = 0xFE;
+ pmbr->partition_record[0].end_sector = 0xFF;
+ pmbr->partition_record[0].end_track = 0xFF;
+ pmbr->partition_record[0].starting_lba = cpu_to_le32(1);
+
+ /*
+ * Set size_in_lba to the size of the disk minus one. If the size of the disk
+ * is too large to be represented by a 32bit LBA (2Tb), set it to 0xFFFFFFFF.
+ */
+ if (cxt->total_sectors - 1 > 0xFFFFFFFFULL)
+ pmbr->partition_record[0].size_in_lba = cpu_to_le32(0xFFFFFFFF);
+ else
+ pmbr->partition_record[0].size_in_lba =
+ cpu_to_le32(cxt->total_sectors - 1UL);
+
+ offset = GPT_PMBR_LBA * cxt->sector_size;
+ if (offset != lseek(cxt->dev_fd, offset, SEEK_SET))
+ goto fail;
+
+ /* pMBR covers the first sector (LBA) of the disk */
+ if (write_all(cxt->dev_fd, pmbr, cxt->sector_size))
+ goto fail;
+ return 0;
+fail:
+ return -errno;
+}
+
+/*
+ * Writes in-memory GPT and pMBR data to disk.
+ * Returns 0 if successful write, otherwise, a corresponding error.
+ * Any indication of error will abort the operation.
+ */
+static int gpt_write_disklabel(struct fdisk_context *cxt)
+{
+ struct fdisk_gpt_label *gpt;
+ int mbr_type;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+ mbr_type = valid_pmbr(cxt);
+
+ /* check that disk is big enough to handle the backup header */
+ if (le64_to_cpu(gpt->pheader->alternative_lba) > cxt->total_sectors)
+ goto err0;
+
+ /* check that the backup header is properly placed */
+ if (le64_to_cpu(gpt->pheader->alternative_lba) < cxt->total_sectors - 1)
+ /* TODO: correct this (with user authorization) and write */
+ goto err0;
+
+ if (check_overlap_partitions(gpt->pheader, gpt->ents))
+ goto err0;
+
+ /* recompute CRCs for both headers */
+ gpt_recompute_crc(gpt->pheader, gpt->ents);
+ gpt_recompute_crc(gpt->bheader, gpt->ents);
+
+ /*
+ * UEFI requires writing in this specific order:
+ * 1) backup partition tables
+ * 2) backup GPT header
+ * 3) primary partition tables
+ * 4) primary GPT header
+ * 5) protective MBR
+ *
+ * If any write fails, we abort the rest.
+ */
+ if (gpt_write_partitions(cxt, gpt->bheader, gpt->ents) != 0)
+ goto err1;
+ if (gpt_write_header(cxt, gpt->bheader,
+ le64_to_cpu(gpt->pheader->alternative_lba)) != 0)
+ goto err1;
+ if (gpt_write_partitions(cxt, gpt->pheader, gpt->ents) != 0)
+ goto err1;
+ if (gpt_write_header(cxt, gpt->pheader, GPT_PRIMARY_PARTITION_TABLE_LBA) != 0)
+ goto err1;
+
+ if (mbr_type == GPT_MBR_HYBRID)
+ fdisk_warnx(cxt, _("The device contains hybrid MBR -- writing GPT only. "
+ "You have to sync the MBR manually."));
+ else if (gpt_write_pmbr(cxt) != 0)
+ goto err1;
+
+ DBG(LABEL, ul_debug("GPT write success"));
+ return 0;
+err0:
+ DBG(LABEL, ul_debug("GPT write failed: incorrect input"));
+ errno = EINVAL;
+ return -EINVAL;
+err1:
+ DBG(LABEL, ul_debug("GPT write failed: %m"));
+ return -errno;
+}
+
+/*
+ * Verify data integrity and report any found problems for:
+ * - primary and backup header validations
+ * - paritition validations
+ */
+static int gpt_verify_disklabel(struct fdisk_context *cxt)
+{
+ int nerror = 0;
+ unsigned int ptnum;
+ struct fdisk_gpt_label *gpt;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+
+ if (!gpt || !gpt->bheader) {
+ nerror++;
+ fdisk_warnx(cxt, _("Disk does not contain a valid backup header."));
+ }
+
+ if (!gpt_check_header_crc(gpt->pheader, gpt->ents)) {
+ nerror++;
+ fdisk_warnx(cxt, _("Invalid primary header CRC checksum."));
+ }
+ if (gpt->bheader && !gpt_check_header_crc(gpt->bheader, gpt->ents)) {
+ nerror++;
+ fdisk_warnx(cxt, _("Invalid backup header CRC checksum."));
+ }
+
+ if (!gpt_check_entryarr_crc(gpt->pheader, gpt->ents)) {
+ nerror++;
+ fdisk_warnx(cxt, _("Invalid partition entry checksum."));
+ }
+
+ if (!gpt_check_lba_sanity(cxt, gpt->pheader)) {
+ nerror++;
+ fdisk_warnx(cxt, _("Invalid primary header LBA sanity checks."));
+ }
+ if (gpt->bheader && !gpt_check_lba_sanity(cxt, gpt->bheader)) {
+ nerror++;
+ fdisk_warnx(cxt, _("Invalid backup header LBA sanity checks."));
+ }
+
+ if (le64_to_cpu(gpt->pheader->my_lba) != GPT_PRIMARY_PARTITION_TABLE_LBA) {
+ nerror++;
+ fdisk_warnx(cxt, _("MyLBA mismatch with real position at primary header."));
+ }
+ if (gpt->bheader && le64_to_cpu(gpt->bheader->my_lba) != last_lba(cxt)) {
+ nerror++;
+ fdisk_warnx(cxt, _("MyLBA mismatch with real position at backup header."));
+
+ }
+ if (le64_to_cpu(gpt->pheader->alternative_lba) >= cxt->total_sectors) {
+ nerror++;
+ fdisk_warnx(cxt, _("Disk is too small to hold all data."));
+ }
+
+ /*
+ * if the GPT is the primary table, check the alternateLBA
+ * to see if it is a valid GPT
+ */
+ if (gpt->bheader && (le64_to_cpu(gpt->pheader->my_lba) !=
+ le64_to_cpu(gpt->bheader->alternative_lba))) {
+ nerror++;
+ fdisk_warnx(cxt, _("Primary and backup header mismatch."));
+ }
+
+ ptnum = check_overlap_partitions(gpt->pheader, gpt->ents);
+ if (ptnum) {
+ nerror++;
+ fdisk_warnx(cxt, _("Partition %u overlaps with partition %u."),
+ ptnum, ptnum+1);
+ }
+
+ ptnum = check_too_big_partitions(gpt->pheader, gpt->ents, cxt->total_sectors);
+ if (ptnum) {
+ nerror++;
+ fdisk_warnx(cxt, _("Partition %u is too big for the disk."),
+ ptnum);
+ }
+
+ ptnum = check_start_after_end_paritions(gpt->pheader, gpt->ents);
+ if (ptnum) {
+ nerror++;
+ fdisk_warnx(cxt, _("Partition %u ends before it starts."),
+ ptnum);
+ }
+
+ if (!nerror) { /* yay :-) */
+ uint32_t nsegments = 0;
+ uint64_t free_sectors = 0, largest_segment = 0;
+ char *strsz = NULL;
+
+ fdisk_info(cxt, _("No errors detected."));
+ fdisk_info(cxt, _("Header version: %s"), gpt_get_header_revstr(gpt->pheader));
+ fdisk_info(cxt, _("Using %u out of %d partitions."),
+ partitions_in_use(gpt->pheader, gpt->ents),
+ le32_to_cpu(gpt->pheader->npartition_entries));
+
+ free_sectors = get_free_sectors(cxt, gpt->pheader, gpt->ents,
+ &nsegments, &largest_segment);
+ if (largest_segment)
+ strsz = size_to_human_string(SIZE_SUFFIX_SPACE | SIZE_SUFFIX_3LETTER,
+ largest_segment * cxt->sector_size);
+
+ fdisk_info(cxt,
+ P_("A total of %ju free sectors is available in %u segment.",
+ "A total of %ju free sectors is available in %u segments "
+ "(the largest is %s).", nsegments),
+ free_sectors, nsegments, strsz);
+ free(strsz);
+
+ } else
+ fdisk_warnx(cxt,
+ P_("%d error detected.", "%d errors detected.", nerror),
+ nerror);
+
+ return 0;
+}
+
+/* Delete a single GPT partition, specified by partnum. */
+static int gpt_delete_partition(struct fdisk_context *cxt,
+ size_t partnum)
+{
+ struct fdisk_gpt_label *gpt;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+
+ if (partnum >= cxt->label->nparts_max
+ || partition_unused(&gpt->ents[partnum]))
+ return -EINVAL;
+
+ /* hasta la vista, baby! */
+ memset(&gpt->ents[partnum], 0, sizeof(struct gpt_entry));
+ if (!partition_unused(&gpt->ents[partnum]))
+ return -EINVAL;
+ else {
+ gpt_recompute_crc(gpt->pheader, gpt->ents);
+ gpt_recompute_crc(gpt->bheader, gpt->ents);
+ cxt->label->nparts_cur--;
+ fdisk_label_set_changed(cxt->label, 1);
+ }
+
+ return 0;
+}
+
+
+/* Performs logical checks to add a new partition entry */
+static int gpt_add_partition(
+ struct fdisk_context *cxt,
+ struct fdisk_partition *pa,
+ size_t *partno)
+{
+ uint64_t user_f, user_l; /* user input ranges for first and last sectors */
+ uint64_t disk_f, disk_l; /* first and last available sector ranges on device*/
+ uint64_t dflt_f, dflt_l; /* largest segment (default) */
+ struct gpt_guid typeid;
+ struct fdisk_gpt_label *gpt;
+ struct gpt_header *pheader;
+ struct gpt_entry *e, *ents;
+ struct fdisk_ask *ask = NULL;
+ size_t partnum;
+ int rc;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+ pheader = gpt->pheader;
+ ents = gpt->ents;
+
+ rc = fdisk_partition_next_partno(pa, cxt, &partnum);
+ if (rc) {
+ DBG(LABEL, ul_debug("GPT failed to get next partno"));
+ return rc;
+ }
+ if (!partition_unused(&ents[partnum])) {
+ fdisk_warnx(cxt, _("Partition %zu is already defined. "
+ "Delete it before re-adding it."), partnum +1);
+ return -ERANGE;
+ }
+ if (le32_to_cpu(pheader->npartition_entries) ==
+ partitions_in_use(pheader, ents)) {
+ fdisk_warnx(cxt, _("All partitions are already in use."));
+ return -ENOSPC;
+ }
+ if (!get_free_sectors(cxt, pheader, ents, NULL, NULL)) {
+ fdisk_warnx(cxt, _("No free sectors available."));
+ return -ENOSPC;
+ }
+
+ string_to_guid(pa && pa->type && pa->type->typestr ?
+ pa->type->typestr:
+ GPT_DEFAULT_ENTRY_TYPE, &typeid);
+
+ disk_f = find_first_available(pheader, ents, pheader->first_usable_lba);
+
+ /* if first sector no explicitly defined then ignore small gaps before
+ * the first partition */
+ if ((!pa || !fdisk_partition_has_start(pa))
+ && !partition_unused(&ents[0])
+ && disk_f < gpt_partition_start(&ents[0])) {
+
+ do {
+ uint64_t x;
+ DBG(LABEL, ul_debug("testing first sector %ju", disk_f));
+ disk_f = find_first_available(pheader, ents, disk_f);
+ if (!disk_f)
+ break;
+ x = find_last_free(pheader, ents, disk_f);
+ if (x - disk_f >= cxt->grain / cxt->sector_size)
+ break;
+ DBG(LABEL, ul_debug("first sector %ju addresses to small space, continue...", disk_f));
+ disk_f = x + 1;
+ } while(1);
+
+ if (disk_f == 0)
+ disk_f = find_first_available(pheader, ents, pheader->first_usable_lba);
+ }
+
+ disk_l = find_last_free_sector(pheader, ents);
+
+ /* the default is the largest free space */
+ dflt_f = find_first_in_largest(pheader, ents);
+ dflt_l = find_last_free(pheader, ents, dflt_f);
+
+ /* align the default in range <dflt_f,dflt_l>*/
+ dflt_f = fdisk_align_lba_in_range(cxt, dflt_f, dflt_f, dflt_l);
+
+ /* first sector */
+ if (pa && pa->start_follow_default) {
+ user_f = dflt_f;
+
+ } else if (pa && fdisk_partition_has_start(pa)) {
+ DBG(LABEL, ul_debug("first sector defined: %ju", pa->start));
+ if (pa->start != find_first_available(pheader, ents, pa->start)) {
+ fdisk_warnx(cxt, _("Sector %ju already used."), pa->start);
+ return -ERANGE;
+ }
+ user_f = pa->start;
+ } else {
+ /* ask by dialog */
+ for (;;) {
+ if (!ask)
+ ask = fdisk_new_ask();
+ else
+ fdisk_reset_ask(ask);
+
+ /* First sector */
+ fdisk_ask_set_query(ask, _("First sector"));
+ fdisk_ask_set_type(ask, FDISK_ASKTYPE_NUMBER);
+ fdisk_ask_number_set_low(ask, disk_f); /* minimal */
+ fdisk_ask_number_set_default(ask, dflt_f); /* default */
+ fdisk_ask_number_set_high(ask, disk_l); /* maximal */
+
+ rc = fdisk_do_ask(cxt, ask);
+ if (rc)
+ goto done;
+
+ user_f = fdisk_ask_number_get_result(ask);
+ if (user_f != find_first_available(pheader, ents, user_f)) {
+ fdisk_warnx(cxt, _("Sector %ju already used."), user_f);
+ continue;
+ }
+ break;
+ }
+ }
+
+
+ /* Last sector */
+ dflt_l = find_last_free(pheader, ents, user_f);
+
+ if (pa && pa->end_follow_default) {
+ user_l = dflt_l;
+
+ } else if (pa && fdisk_partition_has_size(pa)) {
+ user_l = user_f + pa->size - 1;
+ DBG(LABEL, ul_debug("size defined: %ju, end: %ju (last possible: %ju)",
+ pa->size, user_l, dflt_l));
+ if (user_l != dflt_l && !pa->size_explicit)
+ user_l = fdisk_align_lba_in_range(cxt, user_l, user_f, dflt_l) - 1;
+
+ } else {
+ for (;;) {
+ if (!ask)
+ ask = fdisk_new_ask();
+ else
+ fdisk_reset_ask(ask);
+
+ fdisk_ask_set_query(ask, _("Last sector, +sectors or +size{K,M,G,T,P}"));
+ fdisk_ask_set_type(ask, FDISK_ASKTYPE_OFFSET);
+ fdisk_ask_number_set_low(ask, user_f); /* minimal */
+ fdisk_ask_number_set_default(ask, dflt_l); /* default */
+ fdisk_ask_number_set_high(ask, dflt_l); /* maximal */
+ fdisk_ask_number_set_base(ask, user_f); /* base for relative input */
+ fdisk_ask_number_set_unit(ask, cxt->sector_size);
+
+ rc = fdisk_do_ask(cxt, ask);
+ if (rc)
+ goto done;
+
+ user_l = fdisk_ask_number_get_result(ask);
+ if (fdisk_ask_number_is_relative(ask)) {
+ user_l = fdisk_align_lba_in_range(cxt, user_l, user_f, dflt_l) - 1;
+
+ /* no space for anything useful, use all space
+ if (user_l + (cxt->grain / cxt->sector_size) > dflt_l)
+ user_l = dflt_l;
+ */
+ }
+
+ if (user_l > user_f && user_l <= disk_l)
+ break;
+ }
+ }
+
+
+ if (user_f > user_l || partnum >= cxt->label->nparts_max) {
+ fdisk_warnx(cxt, _("Could not create partition %zu"), partnum + 1);
+ rc = -EINVAL;
+ goto done;
+ }
+
+ assert(!FDISK_IS_UNDEF(user_l));
+ assert(!FDISK_IS_UNDEF(user_f));
+
+ e = &ents[partnum];
+ e->lba_end = cpu_to_le64(user_l);
+ e->lba_start = cpu_to_le64(user_f);
+
+ gpt_entry_set_type(e, &typeid);
+
+ if (pa && pa->uuid) {
+ /* Sometimes it's necessary to create a copy of the PT and
+ * reuse already defined UUID
+ */
+ rc = gpt_entry_set_uuid(e, pa->uuid);
+ if (rc)
+ goto done;
+ } else {
+ /* Any time a new partition entry is created a new GUID must be
+ * generated for that partition, and every partition is guaranteed
+ * to have a unique GUID.
+ */
+ uuid_generate_random((unsigned char *) &e->partition_guid);
+ swap_efi_guid(&e->partition_guid);
+ }
+
+ if (pa && pa->name && *pa->name)
+ gpt_entry_set_name(e, pa->name);
+ if (pa && pa->attrs)
+ gpt_entry_attrs_from_string(cxt, e, pa->attrs);
+
+ DBG(LABEL, ul_debug("GPT new partition: partno=%zu, start=%ju, end=%ju, size=%ju",
+ partnum,
+ gpt_partition_start(e),
+ gpt_partition_end(e),
+ gpt_partition_size(e)));
+
+ gpt_recompute_crc(gpt->pheader, ents);
+ gpt_recompute_crc(gpt->bheader, ents);
+
+ /* report result */
+ {
+ struct fdisk_parttype *t;
+
+ cxt->label->nparts_cur++;
+ fdisk_label_set_changed(cxt->label, 1);
+
+ t = gpt_partition_parttype(cxt, &ents[partnum]);
+ fdisk_info_new_partition(cxt, partnum + 1, user_f, user_l, t);
+ fdisk_unref_parttype(t);
+ }
+
+ rc = 0;
+ if (partno)
+ *partno = partnum;
+done:
+ fdisk_unref_ask(ask);
+ return rc;
+}
+
+/*
+ * Create a new GPT disklabel - destroys any previous data.
+ */
+static int gpt_create_disklabel(struct fdisk_context *cxt)
+{
+ int rc = 0;
+ ssize_t esz = 0;
+ char str[37];
+ struct fdisk_gpt_label *gpt;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+
+ /* label private stuff has to be empty, see gpt_deinit() */
+ assert(gpt->pheader == NULL);
+ assert(gpt->bheader == NULL);
+
+ /*
+ * When no header, entries or pmbr is set, we're probably
+ * dealing with a new, empty disk - so always allocate memory
+ * to deal with the data structures whatever the case is.
+ */
+ rc = gpt_mknew_pmbr(cxt);
+ if (rc < 0)
+ goto done;
+
+ /* primary */
+ gpt->pheader = calloc(1, sizeof(*gpt->pheader));
+ if (!gpt->pheader) {
+ rc = -ENOMEM;
+ goto done;
+ }
+ rc = gpt_mknew_header(cxt, gpt->pheader, GPT_PRIMARY_PARTITION_TABLE_LBA);
+ if (rc < 0)
+ goto done;
+
+ /* backup ("copy" primary) */
+ gpt->bheader = calloc(1, sizeof(*gpt->bheader));
+ if (!gpt->bheader) {
+ rc = -ENOMEM;
+ goto done;
+ }
+ rc = gpt_mknew_header_from_bkp(cxt, gpt->bheader,
+ last_lba(cxt), gpt->pheader);
+ if (rc < 0)
+ goto done;
+
+ esz = le32_to_cpu(gpt->pheader->npartition_entries) *
+ le32_to_cpu(gpt->pheader->sizeof_partition_entry);
+ gpt->ents = calloc(1, esz);
+ if (!gpt->ents) {
+ rc = -ENOMEM;
+ goto done;
+ }
+ gpt_recompute_crc(gpt->pheader, gpt->ents);
+ gpt_recompute_crc(gpt->bheader, gpt->ents);
+
+ cxt->label->nparts_max = le32_to_cpu(gpt->pheader->npartition_entries);
+ cxt->label->nparts_cur = 0;
+
+ guid_to_string(&gpt->pheader->disk_guid, str);
+ fdisk_label_set_changed(cxt->label, 1);
+ fdisk_info(cxt, _("Created a new GPT disklabel (GUID: %s)."), str);
+done:
+ return rc;
+}
+
+static int gpt_get_disklabel_id(struct fdisk_context *cxt, char **id)
+{
+ struct fdisk_gpt_label *gpt;
+ char str[37];
+
+ assert(cxt);
+ assert(id);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+ guid_to_string(&gpt->pheader->disk_guid, str);
+
+ *id = strdup(str);
+ if (!*id)
+ return -ENOMEM;
+ return 0;
+}
+
+static int gpt_set_disklabel_id(struct fdisk_context *cxt)
+{
+ struct fdisk_gpt_label *gpt;
+ struct gpt_guid uuid;
+ char *str, *old, *new;
+ int rc;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+ if (fdisk_ask_string(cxt,
+ _("Enter new disk UUID (in 8-4-4-4-12 format)"), &str))
+ return -EINVAL;
+
+ rc = string_to_guid(str, &uuid);
+ free(str);
+
+ if (rc) {
+ fdisk_warnx(cxt, _("Failed to parse your UUID."));
+ return rc;
+ }
+
+ gpt_get_disklabel_id(cxt, &old);
+
+ gpt->pheader->disk_guid = uuid;
+ gpt->bheader->disk_guid = uuid;
+
+ gpt_recompute_crc(gpt->pheader, gpt->ents);
+ gpt_recompute_crc(gpt->bheader, gpt->ents);
+
+ gpt_get_disklabel_id(cxt, &new);
+
+ fdisk_info(cxt, _("Disk identifier changed from %s to %s."), old, new);
+
+ free(old);
+ free(new);
+ fdisk_label_set_changed(cxt->label, 1);
+ return 0;
+}
+
+static int gpt_part_is_used(struct fdisk_context *cxt, size_t i)
+{
+ struct fdisk_gpt_label *gpt;
+ struct gpt_entry *e;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+
+ if ((uint32_t) i >= le32_to_cpu(gpt->pheader->npartition_entries))
+ return 0;
+ e = &gpt->ents[i];
+
+ return !partition_unused(e) || gpt_partition_start(e);
+}
+
+/**
+ * fdisk_gpt_is_hybrid:
+ * @cxt: context
+ *
+ * The regular GPT contains PMBR (dummy protective MBR) where the protective
+ * MBR does not address any partitions.
+ *
+ * Hybrid GPT contains regular MBR where this partition table addresses the
+ * same partitions as GPT. It's recommended to not use hybrid GPT due to MBR
+ * limits.
+ *
+ * The libfdisk does not provide functionality to sync GPT and MBR, you have to
+ * directly access and modify (P)MBR (see fdisk_new_nested_context()).
+ *
+ * Returns: 1 if partition table detected as hybrid otherwise return 0
+ */
+int fdisk_gpt_is_hybrid(struct fdisk_context *cxt)
+{
+ assert(cxt);
+ return valid_pmbr(cxt) == GPT_MBR_HYBRID;
+}
+
+static int gpt_toggle_partition_flag(
+ struct fdisk_context *cxt,
+ size_t i,
+ unsigned long flag)
+{
+ struct fdisk_gpt_label *gpt;
+ uint64_t attrs, tmp;
+ char *bits;
+ const char *name = NULL;
+ int bit = -1, rc;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ DBG(LABEL, ul_debug("GPT entry attribute change requested partno=%zu", i));
+ gpt = self_label(cxt);
+
+ if ((uint32_t) i >= le32_to_cpu(gpt->pheader->npartition_entries))
+ return -EINVAL;
+
+ attrs = le64_to_cpu(gpt->ents[i].attrs);
+ bits = (char *) &attrs;
+
+ switch (flag) {
+ case GPT_FLAG_REQUIRED:
+ bit = GPT_ATTRBIT_REQ;
+ name = GPT_ATTRSTR_REQ;
+ break;
+ case GPT_FLAG_NOBLOCK:
+ bit = GPT_ATTRBIT_NOBLOCK;
+ name = GPT_ATTRSTR_NOBLOCK;
+ break;
+ case GPT_FLAG_LEGACYBOOT:
+ bit = GPT_ATTRBIT_LEGACY;
+ name = GPT_ATTRSTR_LEGACY;
+ break;
+ case GPT_FLAG_GUIDSPECIFIC:
+ rc = fdisk_ask_number(cxt, 48, 48, 63, _("Enter GUID specific bit"), &tmp);
+ if (rc)
+ return rc;
+ bit = tmp;
+ break;
+ default:
+ /* already specified PT_FLAG_GUIDSPECIFIC bit */
+ if (flag >= 48 && flag <= 63) {
+ bit = flag;
+ flag = GPT_FLAG_GUIDSPECIFIC;
+ }
+ break;
+ }
+
+ if (bit < 0) {
+ fdisk_warnx(cxt, _("failed to toggle unsupported bit %lu"), flag);
+ return -EINVAL;
+ }
+
+ if (!isset(bits, bit))
+ setbit(bits, bit);
+ else
+ clrbit(bits, bit);
+
+ gpt->ents[i].attrs = cpu_to_le64(attrs);
+
+ if (flag == GPT_FLAG_GUIDSPECIFIC)
+ fdisk_info(cxt, isset(bits, bit) ?
+ _("The GUID specific bit %d on partition %zu is enabled now.") :
+ _("The GUID specific bit %d on partition %zu is disabled now."),
+ bit, i + 1);
+ else
+ fdisk_info(cxt, isset(bits, bit) ?
+ _("The %s flag on partition %zu is enabled now.") :
+ _("The %s flag on partition %zu is disabled now."),
+ name, i + 1);
+
+ gpt_recompute_crc(gpt->pheader, gpt->ents);
+ gpt_recompute_crc(gpt->bheader, gpt->ents);
+ fdisk_label_set_changed(cxt->label, 1);
+ return 0;
+}
+
+static int gpt_entry_cmp_start(const void *a, const void *b)
+{
+ struct gpt_entry *ae = (struct gpt_entry *) a,
+ *be = (struct gpt_entry *) b;
+ int au = partition_unused(ae),
+ bu = partition_unused(be);
+
+ if (au && bu)
+ return 0;
+ if (au)
+ return 1;
+ if (bu)
+ return -1;
+
+ return cmp_numbers(gpt_partition_start(ae), gpt_partition_start(be));
+}
+
+/* sort partition by start sector */
+static int gpt_reorder(struct fdisk_context *cxt)
+{
+ struct fdisk_gpt_label *gpt;
+ size_t nparts;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+ nparts = le32_to_cpu(gpt->pheader->npartition_entries);
+
+ qsort(gpt->ents, nparts, sizeof(struct gpt_entry),
+ gpt_entry_cmp_start);
+
+ gpt_recompute_crc(gpt->pheader, gpt->ents);
+ gpt_recompute_crc(gpt->bheader, gpt->ents);
+ fdisk_label_set_changed(cxt->label, 1);
+
+ fdisk_info(cxt, _("Done."));
+ return 0;
+}
+
+static int gpt_reset_alignment(struct fdisk_context *cxt)
+{
+ struct fdisk_gpt_label *gpt;
+ struct gpt_header *h;
+
+ assert(cxt);
+ assert(cxt->label);
+ assert(fdisk_is_label(cxt, GPT));
+
+ gpt = self_label(cxt);
+ h = gpt ? gpt->pheader : NULL;
+
+ if (h) {
+ /* always follow existing table */
+ cxt->first_lba = h->first_usable_lba;
+ cxt->last_lba = h->last_usable_lba;
+ } else {
+ /* estimate ranges for GPT */
+ uint64_t first, last;
+
+ count_first_last_lba(cxt, &first, &last);
+
+ if (cxt->first_lba < first)
+ cxt->first_lba = first;
+ if (cxt->last_lba > last)
+ cxt->last_lba = last;
+ }
+
+ return 0;
+}
+/*
+ * Deinitialize fdisk-specific variables
+ */
+static void gpt_deinit(struct fdisk_label *lb)
+{
+ struct fdisk_gpt_label *gpt = (struct fdisk_gpt_label *) lb;
+
+ if (!gpt)
+ return;
+
+ free(gpt->ents);
+ free(gpt->pheader);
+ free(gpt->bheader);
+
+ gpt->ents = NULL;
+ gpt->pheader = NULL;
+ gpt->bheader = NULL;
+}
+
+static const struct fdisk_label_operations gpt_operations =
+{
+ .probe = gpt_probe_label,
+ .write = gpt_write_disklabel,
+ .verify = gpt_verify_disklabel,
+ .create = gpt_create_disklabel,
+ .list = gpt_list_disklabel,
+ .locate = gpt_locate_disklabel,
+ .reorder = gpt_reorder,
+ .get_id = gpt_get_disklabel_id,
+ .set_id = gpt_set_disklabel_id,
+
+ .get_part = gpt_get_partition,
+ .set_part = gpt_set_partition,
+ .add_part = gpt_add_partition,
+ .del_part = gpt_delete_partition,
+
+ .part_is_used = gpt_part_is_used,
+ .part_toggle_flag = gpt_toggle_partition_flag,
+
+ .deinit = gpt_deinit,
+
+ .reset_alignment = gpt_reset_alignment
+};
+
+static const struct fdisk_field gpt_fields[] =
+{
+ /* basic */
+ { FDISK_FIELD_DEVICE, N_("Device"), 10, 0 },
+ { FDISK_FIELD_START, N_("Start"), 5, FDISK_FIELDFL_NUMBER },
+ { FDISK_FIELD_END, N_("End"), 5, FDISK_FIELDFL_NUMBER },
+ { FDISK_FIELD_SECTORS, N_("Sectors"), 5, FDISK_FIELDFL_NUMBER },
+ { FDISK_FIELD_SIZE, N_("Size"), 5, FDISK_FIELDFL_NUMBER | FDISK_FIELDFL_EYECANDY },
+ { FDISK_FIELD_TYPE, N_("Type"), 0.1, FDISK_FIELDFL_EYECANDY },
+ /* expert */
+ { FDISK_FIELD_TYPEID, N_("Type-UUID"), 36, FDISK_FIELDFL_DETAIL },
+ { FDISK_FIELD_UUID, N_("UUID"), 36, FDISK_FIELDFL_DETAIL },
+ { FDISK_FIELD_NAME, N_("Name"), 0.2, FDISK_FIELDFL_DETAIL },
+ { FDISK_FIELD_ATTR, N_("Attrs"), 0, FDISK_FIELDFL_DETAIL }
+};
+
+/*
+ * allocates GPT in-memory stuff
+ */
+struct fdisk_label *fdisk_new_gpt_label(struct fdisk_context *cxt)
+{
+ struct fdisk_label *lb;
+ struct fdisk_gpt_label *gpt;
+
+ assert(cxt);
+
+ gpt = calloc(1, sizeof(*gpt));
+ if (!gpt)
+ return NULL;
+
+ /* initialize generic part of the driver */
+ lb = (struct fdisk_label *) gpt;
+ lb->name = "gpt";
+ lb->id = FDISK_DISKLABEL_GPT;
+ lb->op = &gpt_operations;
+ lb->parttypes = gpt_parttypes;
+ lb->nparttypes = ARRAY_SIZE(gpt_parttypes);
+
+ lb->fields = gpt_fields;
+ lb->nfields = ARRAY_SIZE(gpt_fields);
+
+ return lb;
+}