1 /* 2 * linux/fs/nfs/blocklayout/blocklayout.h 3 * 4 * Module for the NFSv4.1 pNFS block layout driver. 5 * 6 * Copyright (c) 2006 The Regents of the University of Michigan. 7 * All rights reserved. 8 * 9 * Andy Adamson <andros@citi.umich.edu> 10 * Fred Isaman <iisaman@umich.edu> 11 * 12 * permission is granted to use, copy, create derivative works and 13 * redistribute this software and such derivative works for any purpose, 14 * so long as the name of the university of michigan is not used in 15 * any advertising or publicity pertaining to the use or distribution 16 * of this software without specific, written prior authorization. if 17 * the above copyright notice or any other identification of the 18 * university of michigan is included in any copy of any portion of 19 * this software, then the disclaimer below must also be included. 20 * 21 * this software is provided as is, without representation from the 22 * university of michigan as to its fitness for any purpose, and without 23 * warranty by the university of michigan of any kind, either express 24 * or implied, including without limitation the implied warranties of 25 * merchantability and fitness for a particular purpose. the regents 26 * of the university of michigan shall not be liable for any damages, 27 * including special, indirect, incidental, or consequential damages, 28 * with respect to any claim arising out or in connection with the use 29 * of the software, even if it has been or is hereafter advised of the 30 * possibility of such damages. 31 */ 32 #ifndef FS_NFS_NFS4BLOCKLAYOUT_H 33 #define FS_NFS_NFS4BLOCKLAYOUT_H 34 35 #include <linux/device-mapper.h> 36 #include <linux/nfs_fs.h> 37 #include <linux/sunrpc/rpc_pipe_fs.h> 38 39 #include "../nfs4_fs.h" 40 #include "../pnfs.h" 41 #include "../netns.h" 42 43 #define PAGE_CACHE_SECTORS (PAGE_CACHE_SIZE >> SECTOR_SHIFT) 44 #define PAGE_CACHE_SECTOR_SHIFT (PAGE_CACHE_SHIFT - SECTOR_SHIFT) 45 #define SECTOR_SIZE (1 << SECTOR_SHIFT) 46 47 struct block_mount_id { 48 spinlock_t bm_lock; /* protects list */ 49 struct list_head bm_devlist; /* holds pnfs_block_dev */ 50 }; 51 52 struct pnfs_block_dev { 53 struct list_head bm_node; 54 struct nfs4_deviceid bm_mdevid; /* associated devid */ 55 struct block_device *bm_mdev; /* meta device itself */ 56 struct net *net; 57 }; 58 59 enum exstate4 { 60 PNFS_BLOCK_READWRITE_DATA = 0, 61 PNFS_BLOCK_READ_DATA = 1, 62 PNFS_BLOCK_INVALID_DATA = 2, /* mapped, but data is invalid */ 63 PNFS_BLOCK_NONE_DATA = 3 /* unmapped, it's a hole */ 64 }; 65 66 #define MY_MAX_TAGS (15) /* tag bitnums used must be less than this */ 67 68 struct my_tree { 69 sector_t mtt_step_size; /* Internal sector alignment */ 70 struct list_head mtt_stub; /* Should be a radix tree */ 71 }; 72 73 struct pnfs_inval_markings { 74 spinlock_t im_lock; 75 struct my_tree im_tree; /* Sectors that need LAYOUTCOMMIT */ 76 sector_t im_block_size; /* Server blocksize in sectors */ 77 struct list_head im_extents; /* Short extents for INVAL->RW conversion */ 78 }; 79 80 struct pnfs_inval_tracking { 81 struct list_head it_link; 82 int it_sector; 83 int it_tags; 84 }; 85 86 /* sector_t fields are all in 512-byte sectors */ 87 struct pnfs_block_extent { 88 struct kref be_refcnt; 89 struct list_head be_node; /* link into lseg list */ 90 struct nfs4_deviceid be_devid; /* FIXME: could use device cache instead */ 91 struct block_device *be_mdev; 92 sector_t be_f_offset; /* the starting offset in the file */ 93 sector_t be_length; /* the size of the extent */ 94 sector_t be_v_offset; /* the starting offset in the volume */ 95 enum exstate4 be_state; /* the state of this extent */ 96 struct pnfs_inval_markings *be_inval; /* tracks INVAL->RW transition */ 97 }; 98 99 /* Shortened extent used by LAYOUTCOMMIT */ 100 struct pnfs_block_short_extent { 101 struct list_head bse_node; 102 struct nfs4_deviceid bse_devid; 103 struct block_device *bse_mdev; 104 sector_t bse_f_offset; /* the starting offset in the file */ 105 sector_t bse_length; /* the size of the extent */ 106 }; 107 108 static inline void 109 BL_INIT_INVAL_MARKS(struct pnfs_inval_markings *marks, sector_t blocksize) 110 { 111 spin_lock_init(&marks->im_lock); 112 INIT_LIST_HEAD(&marks->im_tree.mtt_stub); 113 INIT_LIST_HEAD(&marks->im_extents); 114 marks->im_block_size = blocksize; 115 marks->im_tree.mtt_step_size = min((sector_t)PAGE_CACHE_SECTORS, 116 blocksize); 117 } 118 119 enum extentclass4 { 120 RW_EXTENT = 0, /* READWRTE and INVAL */ 121 RO_EXTENT = 1, /* READ and NONE */ 122 EXTENT_LISTS = 2, 123 }; 124 125 static inline int bl_choose_list(enum exstate4 state) 126 { 127 if (state == PNFS_BLOCK_READ_DATA || state == PNFS_BLOCK_NONE_DATA) 128 return RO_EXTENT; 129 else 130 return RW_EXTENT; 131 } 132 133 struct pnfs_block_layout { 134 struct pnfs_layout_hdr bl_layout; 135 struct pnfs_inval_markings bl_inval; /* tracks INVAL->RW transition */ 136 spinlock_t bl_ext_lock; /* Protects list manipulation */ 137 struct list_head bl_extents[EXTENT_LISTS]; /* R and RW extents */ 138 struct list_head bl_commit; /* Needs layout commit */ 139 struct list_head bl_committing; /* Layout committing */ 140 unsigned int bl_count; /* entries in bl_commit */ 141 sector_t bl_blocksize; /* Server blocksize in sectors */ 142 }; 143 144 #define BLK_ID(lo) ((struct block_mount_id *)(NFS_SERVER(lo->plh_inode)->pnfs_ld_data)) 145 146 static inline struct pnfs_block_layout * 147 BLK_LO2EXT(struct pnfs_layout_hdr *lo) 148 { 149 return container_of(lo, struct pnfs_block_layout, bl_layout); 150 } 151 152 static inline struct pnfs_block_layout * 153 BLK_LSEG2EXT(struct pnfs_layout_segment *lseg) 154 { 155 return BLK_LO2EXT(lseg->pls_layout); 156 } 157 158 struct bl_pipe_msg { 159 struct rpc_pipe_msg msg; 160 wait_queue_head_t *bl_wq; 161 }; 162 163 struct bl_msg_hdr { 164 u8 type; 165 u16 totallen; /* length of entire message, including hdr itself */ 166 }; 167 168 #define BL_DEVICE_UMOUNT 0x0 /* Umount--delete devices */ 169 #define BL_DEVICE_MOUNT 0x1 /* Mount--create devices*/ 170 #define BL_DEVICE_REQUEST_INIT 0x0 /* Start request */ 171 #define BL_DEVICE_REQUEST_PROC 0x1 /* User level process succeeds */ 172 #define BL_DEVICE_REQUEST_ERR 0x2 /* User level process fails */ 173 174 /* blocklayoutdev.c */ 175 ssize_t bl_pipe_downcall(struct file *, const char __user *, size_t); 176 void bl_pipe_destroy_msg(struct rpc_pipe_msg *); 177 void nfs4_blkdev_put(struct block_device *bdev); 178 struct pnfs_block_dev *nfs4_blk_decode_device(struct nfs_server *server, 179 struct pnfs_device *dev); 180 int nfs4_blk_process_layoutget(struct pnfs_layout_hdr *lo, 181 struct nfs4_layoutget_res *lgr, gfp_t gfp_flags); 182 183 /* blocklayoutdm.c */ 184 void bl_free_block_dev(struct pnfs_block_dev *bdev); 185 186 /* extents.c */ 187 struct pnfs_block_extent * 188 bl_find_get_extent(struct pnfs_block_layout *bl, sector_t isect, 189 struct pnfs_block_extent **cow_read); 190 int bl_mark_sectors_init(struct pnfs_inval_markings *marks, 191 sector_t offset, sector_t length); 192 void bl_put_extent(struct pnfs_block_extent *be); 193 struct pnfs_block_extent *bl_alloc_extent(void); 194 int bl_is_sector_init(struct pnfs_inval_markings *marks, sector_t isect); 195 int encode_pnfs_block_layoutupdate(struct pnfs_block_layout *bl, 196 struct xdr_stream *xdr, 197 const struct nfs4_layoutcommit_args *arg); 198 void clean_pnfs_block_layoutupdate(struct pnfs_block_layout *bl, 199 const struct nfs4_layoutcommit_args *arg, 200 int status); 201 int bl_add_merge_extent(struct pnfs_block_layout *bl, 202 struct pnfs_block_extent *new); 203 int bl_mark_for_commit(struct pnfs_block_extent *be, 204 sector_t offset, sector_t length, 205 struct pnfs_block_short_extent *new); 206 int bl_push_one_short_extent(struct pnfs_inval_markings *marks); 207 struct pnfs_block_short_extent * 208 bl_pop_one_short_extent(struct pnfs_inval_markings *marks); 209 void bl_free_short_extents(struct pnfs_inval_markings *marks, int num_to_free); 210 211 #endif /* FS_NFS_NFS4BLOCKLAYOUT_H */ 212