| 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
| 2 | /* |
| 3 | * Simple zone file system for zoned block devices. |
| 4 | * |
| 5 | * Copyright (C) 2019 Western Digital Corporation or its affiliates. |
| 6 | */ |
| 7 | #ifndef __ZONEFS_H__ |
| 8 | #define __ZONEFS_H__ |
| 9 | |
| 10 | #include <linux/fs.h> |
| 11 | #include <linux/magic.h> |
| 12 | #include <linux/uuid.h> |
| 13 | #include <linux/mutex.h> |
| 14 | #include <linux/rwsem.h> |
| 15 | #include <linux/kobject.h> |
| 16 | |
| 17 | /* |
| 18 | * Maximum length of file names: this only needs to be large enough to fit |
| 19 | * the zone group directory names and a decimal zone number for file names. |
| 20 | * 16 characters is plenty. |
| 21 | */ |
| 22 | #define ZONEFS_NAME_MAX 16 |
| 23 | |
| 24 | /* |
| 25 | * Zone types: ZONEFS_ZTYPE_SEQ is used for all sequential zone types |
| 26 | * defined in linux/blkzoned.h, that is, BLK_ZONE_TYPE_SEQWRITE_REQ and |
| 27 | * BLK_ZONE_TYPE_SEQWRITE_PREF. |
| 28 | */ |
| 29 | enum zonefs_ztype { |
| 30 | ZONEFS_ZTYPE_CNV, |
| 31 | ZONEFS_ZTYPE_SEQ, |
| 32 | ZONEFS_ZTYPE_MAX, |
| 33 | }; |
| 34 | |
| 35 | static inline enum zonefs_ztype zonefs_zone_type(struct blk_zone *zone) |
| 36 | { |
| 37 | if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) |
| 38 | return ZONEFS_ZTYPE_CNV; |
| 39 | return ZONEFS_ZTYPE_SEQ; |
| 40 | } |
| 41 | |
| 42 | #define ZONEFS_ZONE_INIT_MODE (1U << 0) |
| 43 | #define ZONEFS_ZONE_OPEN (1U << 1) |
| 44 | #define ZONEFS_ZONE_ACTIVE (1U << 2) |
| 45 | #define ZONEFS_ZONE_OFFLINE (1U << 3) |
| 46 | #define ZONEFS_ZONE_READONLY (1U << 4) |
| 47 | #define ZONEFS_ZONE_CNV (1U << 31) |
| 48 | |
| 49 | /* |
| 50 | * In-memory per-file inode zone data. |
| 51 | */ |
| 52 | struct zonefs_zone { |
| 53 | /* Zone state flags */ |
| 54 | unsigned int z_flags; |
| 55 | |
| 56 | /* Zone start sector (512B unit) */ |
| 57 | sector_t z_sector; |
| 58 | |
| 59 | /* Zone size (bytes) */ |
| 60 | loff_t z_size; |
| 61 | |
| 62 | /* Zone capacity (file maximum size, bytes) */ |
| 63 | loff_t z_capacity; |
| 64 | |
| 65 | /* Write pointer offset in the zone (sequential zones only, bytes) */ |
| 66 | loff_t z_wpoffset; |
| 67 | |
| 68 | /* Saved inode uid, gid and access rights */ |
| 69 | umode_t z_mode; |
| 70 | kuid_t z_uid; |
| 71 | kgid_t z_gid; |
| 72 | }; |
| 73 | |
| 74 | /* |
| 75 | * In memory zone group information: all zones of a group are exposed |
| 76 | * as files, one file per zone. |
| 77 | */ |
| 78 | struct zonefs_zone_group { |
| 79 | struct inode *g_inode; |
| 80 | unsigned int g_nr_zones; |
| 81 | struct zonefs_zone *g_zones; |
| 82 | }; |
| 83 | |
| 84 | /* |
| 85 | * In-memory inode data. |
| 86 | */ |
| 87 | struct zonefs_inode_info { |
| 88 | struct inode i_vnode; |
| 89 | |
| 90 | /* |
| 91 | * To serialise fully against both syscall and mmap based IO and |
| 92 | * sequential file truncation, two locks are used. For serializing |
| 93 | * zonefs_seq_file_truncate() against zonefs_iomap_begin(), that is, |
| 94 | * file truncate operations against block mapping, i_truncate_mutex is |
| 95 | * used. i_truncate_mutex also protects against concurrent accesses |
| 96 | * and changes to the inode private data, and in particular changes to |
| 97 | * a sequential file size on completion of direct IO writes. |
| 98 | * Serialization of mmap read IOs with truncate and syscall IO |
| 99 | * operations is done with invalidate_lock in addition to |
| 100 | * i_truncate_mutex. Only zonefs_seq_file_truncate() takes both lock |
| 101 | * (invalidate_lock first, i_truncate_mutex second). |
| 102 | */ |
| 103 | struct mutex i_truncate_mutex; |
| 104 | |
| 105 | /* guarded by i_truncate_mutex */ |
| 106 | unsigned int i_wr_refcnt; |
| 107 | }; |
| 108 | |
| 109 | static inline struct zonefs_inode_info *ZONEFS_I(struct inode *inode) |
| 110 | { |
| 111 | return container_of(inode, struct zonefs_inode_info, i_vnode); |
| 112 | } |
| 113 | |
| 114 | static inline bool zonefs_zone_is_cnv(struct zonefs_zone *z) |
| 115 | { |
| 116 | return z->z_flags & ZONEFS_ZONE_CNV; |
| 117 | } |
| 118 | |
| 119 | static inline bool zonefs_zone_is_seq(struct zonefs_zone *z) |
| 120 | { |
| 121 | return !zonefs_zone_is_cnv(z); |
| 122 | } |
| 123 | |
| 124 | static inline struct zonefs_zone *zonefs_inode_zone(struct inode *inode) |
| 125 | { |
| 126 | return inode->i_private; |
| 127 | } |
| 128 | |
| 129 | static inline bool zonefs_inode_is_cnv(struct inode *inode) |
| 130 | { |
| 131 | return zonefs_zone_is_cnv(z: zonefs_inode_zone(inode)); |
| 132 | } |
| 133 | |
| 134 | static inline bool zonefs_inode_is_seq(struct inode *inode) |
| 135 | { |
| 136 | return zonefs_zone_is_seq(z: zonefs_inode_zone(inode)); |
| 137 | } |
| 138 | |
| 139 | /* |
| 140 | * On-disk super block (block 0). |
| 141 | */ |
| 142 | #define ZONEFS_LABEL_LEN 64 |
| 143 | #define ZONEFS_UUID_SIZE 16 |
| 144 | #define ZONEFS_SUPER_SIZE 4096 |
| 145 | |
| 146 | struct zonefs_super { |
| 147 | |
| 148 | /* Magic number */ |
| 149 | __le32 s_magic; |
| 150 | |
| 151 | /* Checksum */ |
| 152 | __le32 s_crc; |
| 153 | |
| 154 | /* Volume label */ |
| 155 | char s_label[ZONEFS_LABEL_LEN]; |
| 156 | |
| 157 | /* 128-bit uuid */ |
| 158 | __u8 s_uuid[ZONEFS_UUID_SIZE]; |
| 159 | |
| 160 | /* Features */ |
| 161 | __le64 s_features; |
| 162 | |
| 163 | /* UID/GID to use for files */ |
| 164 | __le32 s_uid; |
| 165 | __le32 s_gid; |
| 166 | |
| 167 | /* File permissions */ |
| 168 | __le32 s_perm; |
| 169 | |
| 170 | /* Padding to ZONEFS_SUPER_SIZE bytes */ |
| 171 | __u8 s_reserved[3988]; |
| 172 | |
| 173 | } __packed; |
| 174 | |
| 175 | /* |
| 176 | * Feature flags: specified in the s_features field of the on-disk super |
| 177 | * block struct zonefs_super and in-memory in the s_feartures field of |
| 178 | * struct zonefs_sb_info. |
| 179 | */ |
| 180 | enum zonefs_features { |
| 181 | /* |
| 182 | * Aggregate contiguous conventional zones into a single file. |
| 183 | */ |
| 184 | ZONEFS_F_AGGRCNV = 1ULL << 0, |
| 185 | /* |
| 186 | * Use super block specified UID for files instead of default 0. |
| 187 | */ |
| 188 | ZONEFS_F_UID = 1ULL << 1, |
| 189 | /* |
| 190 | * Use super block specified GID for files instead of default 0. |
| 191 | */ |
| 192 | ZONEFS_F_GID = 1ULL << 2, |
| 193 | /* |
| 194 | * Use super block specified file permissions instead of default 640. |
| 195 | */ |
| 196 | ZONEFS_F_PERM = 1ULL << 3, |
| 197 | }; |
| 198 | |
| 199 | #define ZONEFS_F_DEFINED_FEATURES \ |
| 200 | (ZONEFS_F_AGGRCNV | ZONEFS_F_UID | ZONEFS_F_GID | ZONEFS_F_PERM) |
| 201 | |
| 202 | /* |
| 203 | * Mount options for zone write pointer error handling. |
| 204 | */ |
| 205 | #define ZONEFS_MNTOPT_ERRORS_RO (1 << 0) /* Make zone file readonly */ |
| 206 | #define ZONEFS_MNTOPT_ERRORS_ZRO (1 << 1) /* Make zone file offline */ |
| 207 | #define ZONEFS_MNTOPT_ERRORS_ZOL (1 << 2) /* Make zone file offline */ |
| 208 | #define ZONEFS_MNTOPT_ERRORS_REPAIR (1 << 3) /* Remount read-only */ |
| 209 | #define ZONEFS_MNTOPT_ERRORS_MASK \ |
| 210 | (ZONEFS_MNTOPT_ERRORS_RO | ZONEFS_MNTOPT_ERRORS_ZRO | \ |
| 211 | ZONEFS_MNTOPT_ERRORS_ZOL | ZONEFS_MNTOPT_ERRORS_REPAIR) |
| 212 | #define ZONEFS_MNTOPT_EXPLICIT_OPEN (1 << 4) /* Explicit open/close of zones on open/close */ |
| 213 | |
| 214 | /* |
| 215 | * In-memory Super block information. |
| 216 | */ |
| 217 | struct zonefs_sb_info { |
| 218 | |
| 219 | unsigned long s_mount_opts; |
| 220 | |
| 221 | spinlock_t s_lock; |
| 222 | |
| 223 | unsigned long long s_features; |
| 224 | kuid_t s_uid; |
| 225 | kgid_t s_gid; |
| 226 | umode_t s_perm; |
| 227 | uuid_t s_uuid; |
| 228 | unsigned int s_zone_sectors_shift; |
| 229 | |
| 230 | struct zonefs_zone_group s_zgroup[ZONEFS_ZTYPE_MAX]; |
| 231 | |
| 232 | loff_t s_blocks; |
| 233 | loff_t s_used_blocks; |
| 234 | |
| 235 | unsigned int s_max_wro_seq_files; |
| 236 | atomic_t s_wro_seq_files; |
| 237 | |
| 238 | unsigned int s_max_active_seq_files; |
| 239 | atomic_t s_active_seq_files; |
| 240 | |
| 241 | bool s_sysfs_registered; |
| 242 | struct kobject s_kobj; |
| 243 | struct completion s_kobj_unregister; |
| 244 | }; |
| 245 | |
| 246 | static inline struct zonefs_sb_info *ZONEFS_SB(struct super_block *sb) |
| 247 | { |
| 248 | return sb->s_fs_info; |
| 249 | } |
| 250 | |
| 251 | #define zonefs_info(sb, format, args...) \ |
| 252 | pr_info("zonefs (%s): " format, sb->s_id, ## args) |
| 253 | #define zonefs_err(sb, format, args...) \ |
| 254 | pr_err("zonefs (%s) ERROR: " format, sb->s_id, ## args) |
| 255 | #define zonefs_warn(sb, format, args...) \ |
| 256 | pr_warn("zonefs (%s) WARNING: " format, sb->s_id, ## args) |
| 257 | |
| 258 | /* In super.c */ |
| 259 | void zonefs_inode_account_active(struct inode *inode); |
| 260 | int zonefs_inode_zone_mgmt(struct inode *inode, enum req_op op); |
| 261 | void zonefs_i_size_write(struct inode *inode, loff_t isize); |
| 262 | void zonefs_update_stats(struct inode *inode, loff_t new_isize); |
| 263 | void __zonefs_io_error(struct inode *inode, bool write); |
| 264 | |
| 265 | static inline void zonefs_io_error(struct inode *inode, bool write) |
| 266 | { |
| 267 | struct zonefs_inode_info *zi = ZONEFS_I(inode); |
| 268 | |
| 269 | mutex_lock(&zi->i_truncate_mutex); |
| 270 | __zonefs_io_error(inode, write); |
| 271 | mutex_unlock(lock: &zi->i_truncate_mutex); |
| 272 | } |
| 273 | |
| 274 | /* In super.c */ |
| 275 | extern const struct inode_operations zonefs_dir_inode_operations; |
| 276 | extern const struct file_operations zonefs_dir_operations; |
| 277 | |
| 278 | /* In file.c */ |
| 279 | extern const struct address_space_operations zonefs_file_aops; |
| 280 | extern const struct file_operations zonefs_file_operations; |
| 281 | int zonefs_file_truncate(struct inode *inode, loff_t isize); |
| 282 | |
| 283 | /* In sysfs.c */ |
| 284 | int zonefs_sysfs_register(struct super_block *sb); |
| 285 | void zonefs_sysfs_unregister(struct super_block *sb); |
| 286 | int zonefs_sysfs_init(void); |
| 287 | void zonefs_sysfs_exit(void); |
| 288 | |
| 289 | #endif |
| 290 | |