vfs_wapbl.c revision 1.3.2.2
1 1.3.2.2 wrstuden /* $NetBSD: vfs_wapbl.c,v 1.3.2.2 2008/09/18 04:31:45 wrstuden Exp $ */ 2 1.3.2.2 wrstuden 3 1.3.2.2 wrstuden /*- 4 1.3.2.2 wrstuden * Copyright (c) 2003,2008 The NetBSD Foundation, Inc. 5 1.3.2.2 wrstuden * All rights reserved. 6 1.3.2.2 wrstuden * 7 1.3.2.2 wrstuden * This code is derived from software contributed to The NetBSD Foundation 8 1.3.2.2 wrstuden * by Wasabi Systems, Inc. 9 1.3.2.2 wrstuden * 10 1.3.2.2 wrstuden * Redistribution and use in source and binary forms, with or without 11 1.3.2.2 wrstuden * modification, are permitted provided that the following conditions 12 1.3.2.2 wrstuden * are met: 13 1.3.2.2 wrstuden * 1. Redistributions of source code must retain the above copyright 14 1.3.2.2 wrstuden * notice, this list of conditions and the following disclaimer. 15 1.3.2.2 wrstuden * 2. Redistributions in binary form must reproduce the above copyright 16 1.3.2.2 wrstuden * notice, this list of conditions and the following disclaimer in the 17 1.3.2.2 wrstuden * documentation and/or other materials provided with the distribution. 18 1.3.2.2 wrstuden * 19 1.3.2.2 wrstuden * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 1.3.2.2 wrstuden * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 1.3.2.2 wrstuden * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 1.3.2.2 wrstuden * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 1.3.2.2 wrstuden * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 1.3.2.2 wrstuden * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 1.3.2.2 wrstuden * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 1.3.2.2 wrstuden * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 1.3.2.2 wrstuden * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 1.3.2.2 wrstuden * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 1.3.2.2 wrstuden * POSSIBILITY OF SUCH DAMAGE. 30 1.3.2.2 wrstuden */ 31 1.3.2.2 wrstuden 32 1.3.2.2 wrstuden /* 33 1.3.2.2 wrstuden * This implements file system independent write ahead filesystem logging. 34 1.3.2.2 wrstuden */ 35 1.3.2.2 wrstuden #include <sys/cdefs.h> 36 1.3.2.2 wrstuden __KERNEL_RCSID(0, "$NetBSD: vfs_wapbl.c,v 1.3.2.2 2008/09/18 04:31:45 wrstuden Exp $"); 37 1.3.2.2 wrstuden 38 1.3.2.2 wrstuden #include <sys/param.h> 39 1.3.2.2 wrstuden 40 1.3.2.2 wrstuden #ifdef _KERNEL 41 1.3.2.2 wrstuden #include <sys/param.h> 42 1.3.2.2 wrstuden #include <sys/namei.h> 43 1.3.2.2 wrstuden #include <sys/proc.h> 44 1.3.2.2 wrstuden #include <sys/uio.h> 45 1.3.2.2 wrstuden #include <sys/vnode.h> 46 1.3.2.2 wrstuden #include <sys/file.h> 47 1.3.2.2 wrstuden #include <sys/malloc.h> 48 1.3.2.2 wrstuden #include <sys/resourcevar.h> 49 1.3.2.2 wrstuden #include <sys/conf.h> 50 1.3.2.2 wrstuden #include <sys/mount.h> 51 1.3.2.2 wrstuden #include <sys/kernel.h> 52 1.3.2.2 wrstuden #include <sys/kauth.h> 53 1.3.2.2 wrstuden #include <sys/mutex.h> 54 1.3.2.2 wrstuden #include <sys/atomic.h> 55 1.3.2.2 wrstuden #include <sys/wapbl.h> 56 1.3.2.2 wrstuden 57 1.3.2.2 wrstuden #if WAPBL_UVM_ALLOC 58 1.3.2.2 wrstuden #include <uvm/uvm.h> 59 1.3.2.2 wrstuden #endif 60 1.3.2.2 wrstuden 61 1.3.2.2 wrstuden #include <miscfs/specfs/specdev.h> 62 1.3.2.2 wrstuden 63 1.3.2.2 wrstuden MALLOC_JUSTDEFINE(M_WAPBL, "wapbl", "write-ahead physical block logging"); 64 1.3.2.2 wrstuden #define wapbl_malloc(s) malloc((s), M_WAPBL, M_WAITOK) 65 1.3.2.2 wrstuden #define wapbl_free(a) free((a), M_WAPBL) 66 1.3.2.2 wrstuden #define wapbl_calloc(n, s) malloc((n)*(s), M_WAPBL, M_WAITOK | M_ZERO) 67 1.3.2.2 wrstuden 68 1.3.2.2 wrstuden #else /* !_KERNEL */ 69 1.3.2.2 wrstuden #include <assert.h> 70 1.3.2.2 wrstuden #include <errno.h> 71 1.3.2.2 wrstuden #include <stdio.h> 72 1.3.2.2 wrstuden #include <stdbool.h> 73 1.3.2.2 wrstuden #include <stdlib.h> 74 1.3.2.2 wrstuden #include <string.h> 75 1.3.2.2 wrstuden 76 1.3.2.2 wrstuden #include <sys/time.h> 77 1.3.2.2 wrstuden #include <sys/wapbl.h> 78 1.3.2.2 wrstuden 79 1.3.2.2 wrstuden #define KDASSERT(x) assert(x) 80 1.3.2.2 wrstuden #define KASSERT(x) assert(x) 81 1.3.2.2 wrstuden #define wapbl_malloc(s) malloc(s) 82 1.3.2.2 wrstuden #define wapbl_free(a) free(a) 83 1.3.2.2 wrstuden #define wapbl_calloc(n, s) calloc((n), (s)) 84 1.3.2.2 wrstuden 85 1.3.2.2 wrstuden #endif /* !_KERNEL */ 86 1.3.2.2 wrstuden 87 1.3.2.2 wrstuden /* 88 1.3.2.2 wrstuden * INTERNAL DATA STRUCTURES 89 1.3.2.2 wrstuden */ 90 1.3.2.2 wrstuden 91 1.3.2.2 wrstuden /* 92 1.3.2.2 wrstuden * This structure holds per-mount log information. 93 1.3.2.2 wrstuden * 94 1.3.2.2 wrstuden * Legend: a = atomic access only 95 1.3.2.2 wrstuden * r = read-only after init 96 1.3.2.2 wrstuden * l = rwlock held 97 1.3.2.2 wrstuden * m = mutex held 98 1.3.2.2 wrstuden * u = unlocked access ok 99 1.3.2.2 wrstuden * b = bufcache_lock held 100 1.3.2.2 wrstuden */ 101 1.3.2.2 wrstuden struct wapbl { 102 1.3.2.2 wrstuden struct vnode *wl_logvp; /* r: log here */ 103 1.3.2.2 wrstuden struct vnode *wl_devvp; /* r: log on this device */ 104 1.3.2.2 wrstuden struct mount *wl_mount; /* r: mountpoint wl is associated with */ 105 1.3.2.2 wrstuden daddr_t wl_logpbn; /* r: Physical block number of start of log */ 106 1.3.2.2 wrstuden int wl_log_dev_bshift; /* r: logarithm of device block size of log 107 1.3.2.2 wrstuden device */ 108 1.3.2.2 wrstuden int wl_fs_dev_bshift; /* r: logarithm of device block size of 109 1.3.2.2 wrstuden filesystem device */ 110 1.3.2.2 wrstuden 111 1.3.2.2 wrstuden unsigned wl_lock_count; /* m: Count of transactions in progress */ 112 1.3.2.2 wrstuden 113 1.3.2.2 wrstuden size_t wl_circ_size; /* r: Number of bytes in buffer of log */ 114 1.3.2.2 wrstuden size_t wl_circ_off; /* r: Number of bytes reserved at start */ 115 1.3.2.2 wrstuden 116 1.3.2.2 wrstuden size_t wl_bufcount_max; /* r: Number of buffers reserved for log */ 117 1.3.2.2 wrstuden size_t wl_bufbytes_max; /* r: Number of buf bytes reserved for log */ 118 1.3.2.2 wrstuden 119 1.3.2.2 wrstuden off_t wl_head; /* l: Byte offset of log head */ 120 1.3.2.2 wrstuden off_t wl_tail; /* l: Byte offset of log tail */ 121 1.3.2.2 wrstuden /* 122 1.3.2.2 wrstuden * head == tail == 0 means log is empty 123 1.3.2.2 wrstuden * head == tail != 0 means log is full 124 1.3.2.2 wrstuden * see assertions in wapbl_advance() for other boundary conditions. 125 1.3.2.2 wrstuden * only truncate moves the tail, except when flush sets it to 126 1.3.2.2 wrstuden * wl_header_size only flush moves the head, except when truncate 127 1.3.2.2 wrstuden * sets it to 0. 128 1.3.2.2 wrstuden */ 129 1.3.2.2 wrstuden 130 1.3.2.2 wrstuden struct wapbl_wc_header *wl_wc_header; /* l */ 131 1.3.2.2 wrstuden void *wl_wc_scratch; /* l: scratch space (XXX: por que?!?) */ 132 1.3.2.2 wrstuden 133 1.3.2.2 wrstuden kmutex_t wl_mtx; /* u: short-term lock */ 134 1.3.2.2 wrstuden krwlock_t wl_rwlock; /* u: File system transaction lock */ 135 1.3.2.2 wrstuden 136 1.3.2.2 wrstuden /* 137 1.3.2.2 wrstuden * Must be held while accessing 138 1.3.2.2 wrstuden * wl_count or wl_bufs or head or tail 139 1.3.2.2 wrstuden */ 140 1.3.2.2 wrstuden 141 1.3.2.2 wrstuden /* 142 1.3.2.2 wrstuden * Callback called from within the flush routine to flush any extra 143 1.3.2.2 wrstuden * bits. Note that flush may be skipped without calling this if 144 1.3.2.2 wrstuden * there are no outstanding buffers in the transaction. 145 1.3.2.2 wrstuden */ 146 1.3.2.2 wrstuden wapbl_flush_fn_t wl_flush; /* r */ 147 1.3.2.2 wrstuden wapbl_flush_fn_t wl_flush_abort;/* r */ 148 1.3.2.2 wrstuden 149 1.3.2.2 wrstuden size_t wl_bufbytes; /* m: Byte count of pages in wl_bufs */ 150 1.3.2.2 wrstuden size_t wl_bufcount; /* m: Count of buffers in wl_bufs */ 151 1.3.2.2 wrstuden size_t wl_bcount; /* m: Total bcount of wl_bufs */ 152 1.3.2.2 wrstuden 153 1.3.2.2 wrstuden LIST_HEAD(, buf) wl_bufs; /* m: Buffers in current transaction */ 154 1.3.2.2 wrstuden 155 1.3.2.2 wrstuden kcondvar_t wl_reclaimable_cv; /* m (obviously) */ 156 1.3.2.2 wrstuden size_t wl_reclaimable_bytes; /* m: Amount of space available for 157 1.3.2.2 wrstuden reclamation by truncate */ 158 1.3.2.2 wrstuden int wl_error_count; /* m: # of wl_entries with errors */ 159 1.3.2.2 wrstuden size_t wl_reserved_bytes; /* never truncate log smaller than this */ 160 1.3.2.2 wrstuden 161 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 162 1.3.2.2 wrstuden size_t wl_unsynced_bufbytes; /* Byte count of unsynced buffers */ 163 1.3.2.2 wrstuden #endif 164 1.3.2.2 wrstuden 165 1.3.2.2 wrstuden daddr_t *wl_deallocblks;/* l: address of block */ 166 1.3.2.2 wrstuden int *wl_dealloclens; /* l: size of block (fragments, kom ihg) */ 167 1.3.2.2 wrstuden int wl_dealloccnt; /* l: total count */ 168 1.3.2.2 wrstuden int wl_dealloclim; /* l: max count */ 169 1.3.2.2 wrstuden 170 1.3.2.2 wrstuden /* hashtable of inode numbers for allocated but unlinked inodes */ 171 1.3.2.2 wrstuden /* synch ??? */ 172 1.3.2.2 wrstuden LIST_HEAD(wapbl_ino_head, wapbl_ino) *wl_inohash; 173 1.3.2.2 wrstuden u_long wl_inohashmask; 174 1.3.2.2 wrstuden int wl_inohashcnt; 175 1.3.2.2 wrstuden 176 1.3.2.2 wrstuden SIMPLEQ_HEAD(, wapbl_entry) wl_entries; /* On disk transaction 177 1.3.2.2 wrstuden accounting */ 178 1.3.2.2 wrstuden }; 179 1.3.2.2 wrstuden 180 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_PRINT 181 1.3.2.2 wrstuden int wapbl_debug_print = WAPBL_DEBUG_PRINT; 182 1.3.2.2 wrstuden #endif 183 1.3.2.2 wrstuden 184 1.3.2.2 wrstuden /****************************************************************/ 185 1.3.2.2 wrstuden #ifdef _KERNEL 186 1.3.2.2 wrstuden 187 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG 188 1.3.2.2 wrstuden struct wapbl *wapbl_debug_wl; 189 1.3.2.2 wrstuden #endif 190 1.3.2.2 wrstuden 191 1.3.2.2 wrstuden static int wapbl_write_commit(struct wapbl *wl, off_t head, off_t tail); 192 1.3.2.2 wrstuden static int wapbl_write_blocks(struct wapbl *wl, off_t *offp); 193 1.3.2.2 wrstuden static int wapbl_write_revocations(struct wapbl *wl, off_t *offp); 194 1.3.2.2 wrstuden static int wapbl_write_inodes(struct wapbl *wl, off_t *offp); 195 1.3.2.2 wrstuden #endif /* _KERNEL */ 196 1.3.2.2 wrstuden 197 1.3.2.2 wrstuden static int wapbl_replay_prescan(struct wapbl_replay *wr); 198 1.3.2.2 wrstuden static int wapbl_replay_get_inodes(struct wapbl_replay *wr); 199 1.3.2.2 wrstuden 200 1.3.2.2 wrstuden static __inline size_t wapbl_space_free(size_t avail, off_t head, 201 1.3.2.2 wrstuden off_t tail); 202 1.3.2.2 wrstuden static __inline size_t wapbl_space_used(size_t avail, off_t head, 203 1.3.2.2 wrstuden off_t tail); 204 1.3.2.2 wrstuden 205 1.3.2.2 wrstuden #ifdef _KERNEL 206 1.3.2.2 wrstuden 207 1.3.2.2 wrstuden #define WAPBL_INODETRK_SIZE 83 208 1.3.2.2 wrstuden static int wapbl_ino_pool_refcount; 209 1.3.2.2 wrstuden static struct pool wapbl_ino_pool; 210 1.3.2.2 wrstuden struct wapbl_ino { 211 1.3.2.2 wrstuden LIST_ENTRY(wapbl_ino) wi_hash; 212 1.3.2.2 wrstuden ino_t wi_ino; 213 1.3.2.2 wrstuden mode_t wi_mode; 214 1.3.2.2 wrstuden }; 215 1.3.2.2 wrstuden 216 1.3.2.2 wrstuden static void wapbl_inodetrk_init(struct wapbl *wl, u_int size); 217 1.3.2.2 wrstuden static void wapbl_inodetrk_free(struct wapbl *wl); 218 1.3.2.2 wrstuden static struct wapbl_ino *wapbl_inodetrk_get(struct wapbl *wl, ino_t ino); 219 1.3.2.2 wrstuden 220 1.3.2.2 wrstuden static size_t wapbl_transaction_len(struct wapbl *wl); 221 1.3.2.2 wrstuden static __inline size_t wapbl_transaction_inodes_len(struct wapbl *wl); 222 1.3.2.2 wrstuden 223 1.3.2.2 wrstuden /* 224 1.3.2.2 wrstuden * This is useful for debugging. If set, the log will 225 1.3.2.2 wrstuden * only be truncated when necessary. 226 1.3.2.2 wrstuden */ 227 1.3.2.2 wrstuden int wapbl_lazy_truncate = 0; 228 1.3.2.2 wrstuden 229 1.3.2.2 wrstuden struct wapbl_ops wapbl_ops = { 230 1.3.2.2 wrstuden .wo_wapbl_discard = wapbl_discard, 231 1.3.2.2 wrstuden .wo_wapbl_replay_isopen = wapbl_replay_isopen1, 232 1.3.2.2 wrstuden .wo_wapbl_replay_read = wapbl_replay_read, 233 1.3.2.2 wrstuden .wo_wapbl_add_buf = wapbl_add_buf, 234 1.3.2.2 wrstuden .wo_wapbl_remove_buf = wapbl_remove_buf, 235 1.3.2.2 wrstuden .wo_wapbl_resize_buf = wapbl_resize_buf, 236 1.3.2.2 wrstuden .wo_wapbl_begin = wapbl_begin, 237 1.3.2.2 wrstuden .wo_wapbl_end = wapbl_end, 238 1.3.2.2 wrstuden .wo_wapbl_junlock_assert= wapbl_junlock_assert, 239 1.3.2.2 wrstuden 240 1.3.2.2 wrstuden /* XXX: the following is only used to say "this is a wapbl buf" */ 241 1.3.2.2 wrstuden .wo_wapbl_biodone = wapbl_biodone, 242 1.3.2.2 wrstuden }; 243 1.3.2.2 wrstuden 244 1.3.2.2 wrstuden void 245 1.3.2.2 wrstuden wapbl_init() 246 1.3.2.2 wrstuden { 247 1.3.2.2 wrstuden 248 1.3.2.2 wrstuden malloc_type_attach(M_WAPBL); 249 1.3.2.2 wrstuden } 250 1.3.2.2 wrstuden 251 1.3.2.2 wrstuden int 252 1.3.2.2 wrstuden wapbl_start(struct wapbl ** wlp, struct mount *mp, struct vnode *vp, 253 1.3.2.2 wrstuden daddr_t off, size_t count, size_t blksize, struct wapbl_replay *wr, 254 1.3.2.2 wrstuden wapbl_flush_fn_t flushfn, wapbl_flush_fn_t flushabortfn) 255 1.3.2.2 wrstuden { 256 1.3.2.2 wrstuden struct wapbl *wl; 257 1.3.2.2 wrstuden struct vnode *devvp; 258 1.3.2.2 wrstuden daddr_t logpbn; 259 1.3.2.2 wrstuden int error; 260 1.3.2.2 wrstuden int log_dev_bshift = DEV_BSHIFT; 261 1.3.2.2 wrstuden int fs_dev_bshift = DEV_BSHIFT; 262 1.3.2.2 wrstuden int run; 263 1.3.2.2 wrstuden 264 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_OPEN, ("wapbl_start: vp=%p off=%" PRId64 265 1.3.2.2 wrstuden " count=%zu blksize=%zu\n", vp, off, count, blksize)); 266 1.3.2.2 wrstuden 267 1.3.2.2 wrstuden if (log_dev_bshift > fs_dev_bshift) { 268 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_OPEN, 269 1.3.2.2 wrstuden ("wapbl: log device's block size cannot be larger " 270 1.3.2.2 wrstuden "than filesystem's\n")); 271 1.3.2.2 wrstuden /* 272 1.3.2.2 wrstuden * Not currently implemented, although it could be if 273 1.3.2.2 wrstuden * needed someday. 274 1.3.2.2 wrstuden */ 275 1.3.2.2 wrstuden return ENOSYS; 276 1.3.2.2 wrstuden } 277 1.3.2.2 wrstuden 278 1.3.2.2 wrstuden if (off < 0) 279 1.3.2.2 wrstuden return EINVAL; 280 1.3.2.2 wrstuden 281 1.3.2.2 wrstuden if (blksize < DEV_BSIZE) 282 1.3.2.2 wrstuden return EINVAL; 283 1.3.2.2 wrstuden if (blksize % DEV_BSIZE) 284 1.3.2.2 wrstuden return EINVAL; 285 1.3.2.2 wrstuden 286 1.3.2.2 wrstuden /* XXXTODO: verify that the full load is writable */ 287 1.3.2.2 wrstuden 288 1.3.2.2 wrstuden /* 289 1.3.2.2 wrstuden * XXX check for minimum log size 290 1.3.2.2 wrstuden * minimum is governed by minimum amount of space 291 1.3.2.2 wrstuden * to complete a transaction. (probably truncate) 292 1.3.2.2 wrstuden */ 293 1.3.2.2 wrstuden /* XXX for now pick something minimal */ 294 1.3.2.2 wrstuden if ((count * blksize) < MAXPHYS) { 295 1.3.2.2 wrstuden return ENOSPC; 296 1.3.2.2 wrstuden } 297 1.3.2.2 wrstuden 298 1.3.2.2 wrstuden if ((error = VOP_BMAP(vp, off, &devvp, &logpbn, &run)) != 0) { 299 1.3.2.2 wrstuden return error; 300 1.3.2.2 wrstuden } 301 1.3.2.2 wrstuden 302 1.3.2.2 wrstuden wl = wapbl_calloc(1, sizeof(*wl)); 303 1.3.2.2 wrstuden rw_init(&wl->wl_rwlock); 304 1.3.2.2 wrstuden mutex_init(&wl->wl_mtx, MUTEX_DEFAULT, IPL_NONE); 305 1.3.2.2 wrstuden cv_init(&wl->wl_reclaimable_cv, "wapblrec"); 306 1.3.2.2 wrstuden LIST_INIT(&wl->wl_bufs); 307 1.3.2.2 wrstuden SIMPLEQ_INIT(&wl->wl_entries); 308 1.3.2.2 wrstuden 309 1.3.2.2 wrstuden wl->wl_logvp = vp; 310 1.3.2.2 wrstuden wl->wl_devvp = devvp; 311 1.3.2.2 wrstuden wl->wl_mount = mp; 312 1.3.2.2 wrstuden wl->wl_logpbn = logpbn; 313 1.3.2.2 wrstuden wl->wl_log_dev_bshift = log_dev_bshift; 314 1.3.2.2 wrstuden wl->wl_fs_dev_bshift = fs_dev_bshift; 315 1.3.2.2 wrstuden 316 1.3.2.2 wrstuden wl->wl_flush = flushfn; 317 1.3.2.2 wrstuden wl->wl_flush_abort = flushabortfn; 318 1.3.2.2 wrstuden 319 1.3.2.2 wrstuden /* Reserve two log device blocks for the commit headers */ 320 1.3.2.2 wrstuden wl->wl_circ_off = 2<<wl->wl_log_dev_bshift; 321 1.3.2.2 wrstuden wl->wl_circ_size = ((count * blksize) - wl->wl_circ_off); 322 1.3.2.2 wrstuden /* truncate the log usage to a multiple of log_dev_bshift */ 323 1.3.2.2 wrstuden wl->wl_circ_size >>= wl->wl_log_dev_bshift; 324 1.3.2.2 wrstuden wl->wl_circ_size <<= wl->wl_log_dev_bshift; 325 1.3.2.2 wrstuden 326 1.3.2.2 wrstuden /* 327 1.3.2.2 wrstuden * wl_bufbytes_max limits the size of the in memory transaction space. 328 1.3.2.2 wrstuden * - Since buffers are allocated and accounted for in units of 329 1.3.2.2 wrstuden * PAGE_SIZE it is required to be a multiple of PAGE_SIZE 330 1.3.2.2 wrstuden * (i.e. 1<<PAGE_SHIFT) 331 1.3.2.2 wrstuden * - Since the log device has to be written in units of 332 1.3.2.2 wrstuden * 1<<wl_log_dev_bshift it is required to be a mulitple of 333 1.3.2.2 wrstuden * 1<<wl_log_dev_bshift. 334 1.3.2.2 wrstuden * - Since filesystem will provide data in units of 1<<wl_fs_dev_bshift, 335 1.3.2.2 wrstuden * it is convenient to be a multiple of 1<<wl_fs_dev_bshift. 336 1.3.2.2 wrstuden * Therefore it must be multiple of the least common multiple of those 337 1.3.2.2 wrstuden * three quantities. Fortunately, all of those quantities are 338 1.3.2.2 wrstuden * guaranteed to be a power of two, and the least common multiple of 339 1.3.2.2 wrstuden * a set of numbers which are all powers of two is simply the maximum 340 1.3.2.2 wrstuden * of those numbers. Finally, the maximum logarithm of a power of two 341 1.3.2.2 wrstuden * is the same as the log of the maximum power of two. So we can do 342 1.3.2.2 wrstuden * the following operations to size wl_bufbytes_max: 343 1.3.2.2 wrstuden */ 344 1.3.2.2 wrstuden 345 1.3.2.2 wrstuden /* XXX fix actual number of pages reserved per filesystem. */ 346 1.3.2.2 wrstuden wl->wl_bufbytes_max = MIN(wl->wl_circ_size, buf_memcalc() / 2); 347 1.3.2.2 wrstuden 348 1.3.2.2 wrstuden /* Round wl_bufbytes_max to the largest power of two constraint */ 349 1.3.2.2 wrstuden wl->wl_bufbytes_max >>= PAGE_SHIFT; 350 1.3.2.2 wrstuden wl->wl_bufbytes_max <<= PAGE_SHIFT; 351 1.3.2.2 wrstuden wl->wl_bufbytes_max >>= wl->wl_log_dev_bshift; 352 1.3.2.2 wrstuden wl->wl_bufbytes_max <<= wl->wl_log_dev_bshift; 353 1.3.2.2 wrstuden wl->wl_bufbytes_max >>= wl->wl_fs_dev_bshift; 354 1.3.2.2 wrstuden wl->wl_bufbytes_max <<= wl->wl_fs_dev_bshift; 355 1.3.2.2 wrstuden 356 1.3.2.2 wrstuden /* XXX maybe use filesystem fragment size instead of 1024 */ 357 1.3.2.2 wrstuden /* XXX fix actual number of buffers reserved per filesystem. */ 358 1.3.2.2 wrstuden wl->wl_bufcount_max = (nbuf / 2) * 1024; 359 1.3.2.2 wrstuden 360 1.3.2.2 wrstuden /* XXX tie this into resource estimation */ 361 1.3.2.2 wrstuden wl->wl_dealloclim = 2 * btodb(wl->wl_bufbytes_max); 362 1.3.2.2 wrstuden 363 1.3.2.2 wrstuden #if WAPBL_UVM_ALLOC 364 1.3.2.2 wrstuden wl->wl_deallocblks = (void *) uvm_km_zalloc(kernel_map, 365 1.3.2.2 wrstuden round_page(sizeof(*wl->wl_deallocblks) * wl->wl_dealloclim)); 366 1.3.2.2 wrstuden KASSERT(wl->wl_deallocblks != NULL); 367 1.3.2.2 wrstuden wl->wl_dealloclens = (void *) uvm_km_zalloc(kernel_map, 368 1.3.2.2 wrstuden round_page(sizeof(*wl->wl_dealloclens) * wl->wl_dealloclim)); 369 1.3.2.2 wrstuden KASSERT(wl->wl_dealloclens != NULL); 370 1.3.2.2 wrstuden #else 371 1.3.2.2 wrstuden wl->wl_deallocblks = wapbl_malloc(sizeof(*wl->wl_deallocblks) * 372 1.3.2.2 wrstuden wl->wl_dealloclim); 373 1.3.2.2 wrstuden wl->wl_dealloclens = wapbl_malloc(sizeof(*wl->wl_dealloclens) * 374 1.3.2.2 wrstuden wl->wl_dealloclim); 375 1.3.2.2 wrstuden #endif 376 1.3.2.2 wrstuden 377 1.3.2.2 wrstuden wapbl_inodetrk_init(wl, WAPBL_INODETRK_SIZE); 378 1.3.2.2 wrstuden 379 1.3.2.2 wrstuden /* Initialize the commit header */ 380 1.3.2.2 wrstuden { 381 1.3.2.2 wrstuden struct wapbl_wc_header *wc; 382 1.3.2.2 wrstuden size_t len = 1<<wl->wl_log_dev_bshift; 383 1.3.2.2 wrstuden wc = wapbl_calloc(1, len); 384 1.3.2.2 wrstuden wc->wc_type = WAPBL_WC_HEADER; 385 1.3.2.2 wrstuden wc->wc_len = len; 386 1.3.2.2 wrstuden wc->wc_circ_off = wl->wl_circ_off; 387 1.3.2.2 wrstuden wc->wc_circ_size = wl->wl_circ_size; 388 1.3.2.2 wrstuden /* XXX wc->wc_fsid */ 389 1.3.2.2 wrstuden wc->wc_log_dev_bshift = wl->wl_log_dev_bshift; 390 1.3.2.2 wrstuden wc->wc_fs_dev_bshift = wl->wl_fs_dev_bshift; 391 1.3.2.2 wrstuden wl->wl_wc_header = wc; 392 1.3.2.2 wrstuden wl->wl_wc_scratch = wapbl_malloc(len); 393 1.3.2.2 wrstuden } 394 1.3.2.2 wrstuden 395 1.3.2.2 wrstuden /* 396 1.3.2.2 wrstuden * if there was an existing set of unlinked but 397 1.3.2.2 wrstuden * allocated inodes, preserve it in the new 398 1.3.2.2 wrstuden * log. 399 1.3.2.2 wrstuden */ 400 1.3.2.2 wrstuden if (wr && wr->wr_inodescnt) { 401 1.3.2.2 wrstuden int i; 402 1.3.2.2 wrstuden 403 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_REPLAY, 404 1.3.2.2 wrstuden ("wapbl_start: reusing log with %d inodes\n", 405 1.3.2.2 wrstuden wr->wr_inodescnt)); 406 1.3.2.2 wrstuden 407 1.3.2.2 wrstuden /* 408 1.3.2.2 wrstuden * Its only valid to reuse the replay log if its 409 1.3.2.2 wrstuden * the same as the new log we just opened. 410 1.3.2.2 wrstuden */ 411 1.3.2.2 wrstuden KDASSERT(!wapbl_replay_isopen(wr)); 412 1.3.2.2 wrstuden KASSERT(devvp->v_rdev == wr->wr_devvp->v_rdev); 413 1.3.2.2 wrstuden KASSERT(logpbn == wr->wr_logpbn); 414 1.3.2.2 wrstuden KASSERT(wl->wl_circ_size == wr->wr_wc_header.wc_circ_size); 415 1.3.2.2 wrstuden KASSERT(wl->wl_circ_off == wr->wr_wc_header.wc_circ_off); 416 1.3.2.2 wrstuden KASSERT(wl->wl_log_dev_bshift == 417 1.3.2.2 wrstuden wr->wr_wc_header.wc_log_dev_bshift); 418 1.3.2.2 wrstuden KASSERT(wl->wl_fs_dev_bshift == 419 1.3.2.2 wrstuden wr->wr_wc_header.wc_fs_dev_bshift); 420 1.3.2.2 wrstuden 421 1.3.2.2 wrstuden wl->wl_wc_header->wc_generation = 422 1.3.2.2 wrstuden wr->wr_wc_header.wc_generation + 1; 423 1.3.2.2 wrstuden 424 1.3.2.2 wrstuden for (i = 0; i < wr->wr_inodescnt; i++) 425 1.3.2.2 wrstuden wapbl_register_inode(wl, wr->wr_inodes[i].wr_inumber, 426 1.3.2.2 wrstuden wr->wr_inodes[i].wr_imode); 427 1.3.2.2 wrstuden 428 1.3.2.2 wrstuden /* Make sure new transaction won't overwrite old inodes list */ 429 1.3.2.2 wrstuden KDASSERT(wapbl_transaction_len(wl) <= 430 1.3.2.2 wrstuden wapbl_space_free(wl->wl_circ_size, wr->wr_inodeshead, 431 1.3.2.2 wrstuden wr->wr_inodestail)); 432 1.3.2.2 wrstuden 433 1.3.2.2 wrstuden wl->wl_head = wl->wl_tail = wr->wr_inodeshead; 434 1.3.2.2 wrstuden wl->wl_reclaimable_bytes = wl->wl_reserved_bytes = 435 1.3.2.2 wrstuden wapbl_transaction_len(wl); 436 1.3.2.2 wrstuden 437 1.3.2.2 wrstuden error = wapbl_write_inodes(wl, &wl->wl_head); 438 1.3.2.2 wrstuden if (error) 439 1.3.2.2 wrstuden goto errout; 440 1.3.2.2 wrstuden 441 1.3.2.2 wrstuden KASSERT(wl->wl_head != wl->wl_tail); 442 1.3.2.2 wrstuden KASSERT(wl->wl_head != 0); 443 1.3.2.2 wrstuden } 444 1.3.2.2 wrstuden 445 1.3.2.2 wrstuden error = wapbl_write_commit(wl, wl->wl_head, wl->wl_tail); 446 1.3.2.2 wrstuden if (error) { 447 1.3.2.2 wrstuden goto errout; 448 1.3.2.2 wrstuden } 449 1.3.2.2 wrstuden 450 1.3.2.2 wrstuden *wlp = wl; 451 1.3.2.2 wrstuden #if defined(WAPBL_DEBUG) 452 1.3.2.2 wrstuden wapbl_debug_wl = wl; 453 1.3.2.2 wrstuden #endif 454 1.3.2.2 wrstuden 455 1.3.2.2 wrstuden return 0; 456 1.3.2.2 wrstuden errout: 457 1.3.2.2 wrstuden wapbl_discard(wl); 458 1.3.2.2 wrstuden wapbl_free(wl->wl_wc_scratch); 459 1.3.2.2 wrstuden wapbl_free(wl->wl_wc_header); 460 1.3.2.2 wrstuden #if WAPBL_UVM_ALLOC 461 1.3.2.2 wrstuden uvm_km_free_wakeup(kernel_map, (vaddr_t) wl->wl_deallocblks, 462 1.3.2.2 wrstuden round_page(sizeof(*wl->wl_deallocblks * 463 1.3.2.2 wrstuden wl->wl_dealloclim))); 464 1.3.2.2 wrstuden uvm_km_free_wakeup(kernel_map, (vaddr_t) wl->wl_dealloclens, 465 1.3.2.2 wrstuden round_page(sizeof(*wl->wl_dealloclens * 466 1.3.2.2 wrstuden wl->wl_dealloclim))); 467 1.3.2.2 wrstuden #else 468 1.3.2.2 wrstuden wapbl_free(wl->wl_deallocblks); 469 1.3.2.2 wrstuden wapbl_free(wl->wl_dealloclens); 470 1.3.2.2 wrstuden #endif 471 1.3.2.2 wrstuden wapbl_inodetrk_free(wl); 472 1.3.2.2 wrstuden wapbl_free(wl); 473 1.3.2.2 wrstuden 474 1.3.2.2 wrstuden return error; 475 1.3.2.2 wrstuden } 476 1.3.2.2 wrstuden 477 1.3.2.2 wrstuden /* 478 1.3.2.2 wrstuden * Like wapbl_flush, only discards the transaction 479 1.3.2.2 wrstuden * completely 480 1.3.2.2 wrstuden */ 481 1.3.2.2 wrstuden 482 1.3.2.2 wrstuden void 483 1.3.2.2 wrstuden wapbl_discard(struct wapbl *wl) 484 1.3.2.2 wrstuden { 485 1.3.2.2 wrstuden struct wapbl_entry *we; 486 1.3.2.2 wrstuden struct buf *bp; 487 1.3.2.2 wrstuden int i; 488 1.3.2.2 wrstuden 489 1.3.2.2 wrstuden /* 490 1.3.2.2 wrstuden * XXX we may consider using upgrade here 491 1.3.2.2 wrstuden * if we want to call flush from inside a transaction 492 1.3.2.2 wrstuden */ 493 1.3.2.2 wrstuden rw_enter(&wl->wl_rwlock, RW_WRITER); 494 1.3.2.2 wrstuden wl->wl_flush(wl->wl_mount, wl->wl_deallocblks, wl->wl_dealloclens, 495 1.3.2.2 wrstuden wl->wl_dealloccnt); 496 1.3.2.2 wrstuden 497 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_PRINT 498 1.3.2.2 wrstuden { 499 1.3.2.2 wrstuden struct wapbl_entry *we; 500 1.3.2.2 wrstuden pid_t pid = -1; 501 1.3.2.2 wrstuden lwpid_t lid = -1; 502 1.3.2.2 wrstuden if (curproc) 503 1.3.2.2 wrstuden pid = curproc->p_pid; 504 1.3.2.2 wrstuden if (curlwp) 505 1.3.2.2 wrstuden lid = curlwp->l_lid; 506 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 507 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_DISCARD, 508 1.3.2.2 wrstuden ("wapbl_discard: thread %d.%d discarding " 509 1.3.2.2 wrstuden "transaction\n" 510 1.3.2.2 wrstuden "\tbufcount=%zu bufbytes=%zu bcount=%zu " 511 1.3.2.2 wrstuden "deallocs=%d inodes=%d\n" 512 1.3.2.2 wrstuden "\terrcnt = %u, reclaimable=%zu reserved=%zu " 513 1.3.2.2 wrstuden "unsynced=%zu\n", 514 1.3.2.2 wrstuden pid, lid, wl->wl_bufcount, wl->wl_bufbytes, 515 1.3.2.2 wrstuden wl->wl_bcount, wl->wl_dealloccnt, 516 1.3.2.2 wrstuden wl->wl_inohashcnt, wl->wl_error_count, 517 1.3.2.2 wrstuden wl->wl_reclaimable_bytes, wl->wl_reserved_bytes, 518 1.3.2.2 wrstuden wl->wl_unsynced_bufbytes)); 519 1.3.2.2 wrstuden SIMPLEQ_FOREACH(we, &wl->wl_entries, we_entries) { 520 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_DISCARD, 521 1.3.2.2 wrstuden ("\tentry: bufcount = %zu, reclaimable = %zu, " 522 1.3.2.2 wrstuden "error = %d, unsynced = %zu\n", 523 1.3.2.2 wrstuden we->we_bufcount, we->we_reclaimable_bytes, 524 1.3.2.2 wrstuden we->we_error, we->we_unsynced_bufbytes)); 525 1.3.2.2 wrstuden } 526 1.3.2.2 wrstuden #else /* !WAPBL_DEBUG_BUFBYTES */ 527 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_DISCARD, 528 1.3.2.2 wrstuden ("wapbl_discard: thread %d.%d discarding transaction\n" 529 1.3.2.2 wrstuden "\tbufcount=%zu bufbytes=%zu bcount=%zu " 530 1.3.2.2 wrstuden "deallocs=%d inodes=%d\n" 531 1.3.2.2 wrstuden "\terrcnt = %u, reclaimable=%zu reserved=%zu\n", 532 1.3.2.2 wrstuden pid, lid, wl->wl_bufcount, wl->wl_bufbytes, 533 1.3.2.2 wrstuden wl->wl_bcount, wl->wl_dealloccnt, 534 1.3.2.2 wrstuden wl->wl_inohashcnt, wl->wl_error_count, 535 1.3.2.2 wrstuden wl->wl_reclaimable_bytes, wl->wl_reserved_bytes)); 536 1.3.2.2 wrstuden SIMPLEQ_FOREACH(we, &wl->wl_entries, we_entries) { 537 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_DISCARD, 538 1.3.2.2 wrstuden ("\tentry: bufcount = %zu, reclaimable = %zu, " 539 1.3.2.2 wrstuden "error = %d\n", 540 1.3.2.2 wrstuden we->we_bufcount, we->we_reclaimable_bytes, 541 1.3.2.2 wrstuden we->we_error)); 542 1.3.2.2 wrstuden } 543 1.3.2.2 wrstuden #endif /* !WAPBL_DEBUG_BUFBYTES */ 544 1.3.2.2 wrstuden } 545 1.3.2.2 wrstuden #endif /* WAPBL_DEBUG_PRINT */ 546 1.3.2.2 wrstuden 547 1.3.2.2 wrstuden for (i = 0; i <= wl->wl_inohashmask; i++) { 548 1.3.2.2 wrstuden struct wapbl_ino_head *wih; 549 1.3.2.2 wrstuden struct wapbl_ino *wi; 550 1.3.2.2 wrstuden 551 1.3.2.2 wrstuden wih = &wl->wl_inohash[i]; 552 1.3.2.2 wrstuden while ((wi = LIST_FIRST(wih)) != NULL) { 553 1.3.2.2 wrstuden LIST_REMOVE(wi, wi_hash); 554 1.3.2.2 wrstuden pool_put(&wapbl_ino_pool, wi); 555 1.3.2.2 wrstuden KASSERT(wl->wl_inohashcnt > 0); 556 1.3.2.2 wrstuden wl->wl_inohashcnt--; 557 1.3.2.2 wrstuden } 558 1.3.2.2 wrstuden } 559 1.3.2.2 wrstuden 560 1.3.2.2 wrstuden /* 561 1.3.2.2 wrstuden * clean buffer list 562 1.3.2.2 wrstuden */ 563 1.3.2.2 wrstuden mutex_enter(&bufcache_lock); 564 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 565 1.3.2.2 wrstuden while ((bp = LIST_FIRST(&wl->wl_bufs)) != NULL) { 566 1.3.2.2 wrstuden if (bbusy(bp, 0, 0, &wl->wl_mtx) == 0) { 567 1.3.2.2 wrstuden /* 568 1.3.2.2 wrstuden * The buffer will be unlocked and 569 1.3.2.2 wrstuden * removed from the transaction in brelse 570 1.3.2.2 wrstuden */ 571 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 572 1.3.2.2 wrstuden brelsel(bp, 0); 573 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 574 1.3.2.2 wrstuden } 575 1.3.2.2 wrstuden } 576 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 577 1.3.2.2 wrstuden mutex_exit(&bufcache_lock); 578 1.3.2.2 wrstuden 579 1.3.2.2 wrstuden /* 580 1.3.2.2 wrstuden * Remove references to this wl from wl_entries, free any which 581 1.3.2.2 wrstuden * no longer have buffers, others will be freed in wapbl_biodone 582 1.3.2.2 wrstuden * when they no longer have any buffers. 583 1.3.2.2 wrstuden */ 584 1.3.2.2 wrstuden while ((we = SIMPLEQ_FIRST(&wl->wl_entries)) != NULL) { 585 1.3.2.2 wrstuden SIMPLEQ_REMOVE_HEAD(&wl->wl_entries, we_entries); 586 1.3.2.2 wrstuden /* XXX should we be accumulating wl_error_count 587 1.3.2.2 wrstuden * and increasing reclaimable bytes ? */ 588 1.3.2.2 wrstuden we->we_wapbl = NULL; 589 1.3.2.2 wrstuden if (we->we_bufcount == 0) { 590 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 591 1.3.2.2 wrstuden KASSERT(we->we_unsynced_bufbytes == 0); 592 1.3.2.2 wrstuden #endif 593 1.3.2.2 wrstuden wapbl_free(we); 594 1.3.2.2 wrstuden } 595 1.3.2.2 wrstuden } 596 1.3.2.2 wrstuden 597 1.3.2.2 wrstuden /* Discard list of deallocs */ 598 1.3.2.2 wrstuden wl->wl_dealloccnt = 0; 599 1.3.2.2 wrstuden /* XXX should we clear wl_reserved_bytes? */ 600 1.3.2.2 wrstuden 601 1.3.2.2 wrstuden KASSERT(wl->wl_bufbytes == 0); 602 1.3.2.2 wrstuden KASSERT(wl->wl_bcount == 0); 603 1.3.2.2 wrstuden KASSERT(wl->wl_bufcount == 0); 604 1.3.2.2 wrstuden KASSERT(LIST_EMPTY(&wl->wl_bufs)); 605 1.3.2.2 wrstuden KASSERT(SIMPLEQ_EMPTY(&wl->wl_entries)); 606 1.3.2.2 wrstuden KASSERT(wl->wl_inohashcnt == 0); 607 1.3.2.2 wrstuden 608 1.3.2.2 wrstuden rw_exit(&wl->wl_rwlock); 609 1.3.2.2 wrstuden } 610 1.3.2.2 wrstuden 611 1.3.2.2 wrstuden int 612 1.3.2.2 wrstuden wapbl_stop(struct wapbl *wl, int force) 613 1.3.2.2 wrstuden { 614 1.3.2.2 wrstuden struct vnode *vp; 615 1.3.2.2 wrstuden int error; 616 1.3.2.2 wrstuden 617 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_OPEN, ("wapbl_stop called\n")); 618 1.3.2.2 wrstuden error = wapbl_flush(wl, 1); 619 1.3.2.2 wrstuden if (error) { 620 1.3.2.2 wrstuden if (force) 621 1.3.2.2 wrstuden wapbl_discard(wl); 622 1.3.2.2 wrstuden else 623 1.3.2.2 wrstuden return error; 624 1.3.2.2 wrstuden } 625 1.3.2.2 wrstuden 626 1.3.2.2 wrstuden /* Unlinked inodes persist after a flush */ 627 1.3.2.2 wrstuden if (wl->wl_inohashcnt) { 628 1.3.2.2 wrstuden if (force) { 629 1.3.2.2 wrstuden wapbl_discard(wl); 630 1.3.2.2 wrstuden } else { 631 1.3.2.2 wrstuden return EBUSY; 632 1.3.2.2 wrstuden } 633 1.3.2.2 wrstuden } 634 1.3.2.2 wrstuden 635 1.3.2.2 wrstuden KASSERT(wl->wl_bufbytes == 0); 636 1.3.2.2 wrstuden KASSERT(wl->wl_bcount == 0); 637 1.3.2.2 wrstuden KASSERT(wl->wl_bufcount == 0); 638 1.3.2.2 wrstuden KASSERT(LIST_EMPTY(&wl->wl_bufs)); 639 1.3.2.2 wrstuden KASSERT(wl->wl_dealloccnt == 0); 640 1.3.2.2 wrstuden KASSERT(SIMPLEQ_EMPTY(&wl->wl_entries)); 641 1.3.2.2 wrstuden KASSERT(wl->wl_inohashcnt == 0); 642 1.3.2.2 wrstuden 643 1.3.2.2 wrstuden vp = wl->wl_logvp; 644 1.3.2.2 wrstuden 645 1.3.2.2 wrstuden wapbl_free(wl->wl_wc_scratch); 646 1.3.2.2 wrstuden wapbl_free(wl->wl_wc_header); 647 1.3.2.2 wrstuden #if WAPBL_UVM_ALLOC 648 1.3.2.2 wrstuden uvm_km_free_wakeup(kernel_map, (vaddr_t) wl->wl_deallocblks, 649 1.3.2.2 wrstuden round_page(sizeof(*wl->wl_deallocblks * 650 1.3.2.2 wrstuden wl->wl_dealloclim))); 651 1.3.2.2 wrstuden uvm_km_free_wakeup(kernel_map, (vaddr_t) wl->wl_dealloclens, 652 1.3.2.2 wrstuden round_page(sizeof(*wl->wl_dealloclens * 653 1.3.2.2 wrstuden wl->wl_dealloclim))); 654 1.3.2.2 wrstuden #else 655 1.3.2.2 wrstuden wapbl_free(wl->wl_deallocblks); 656 1.3.2.2 wrstuden wapbl_free(wl->wl_dealloclens); 657 1.3.2.2 wrstuden #endif 658 1.3.2.2 wrstuden wapbl_inodetrk_free(wl); 659 1.3.2.2 wrstuden 660 1.3.2.2 wrstuden cv_destroy(&wl->wl_reclaimable_cv); 661 1.3.2.2 wrstuden mutex_destroy(&wl->wl_mtx); 662 1.3.2.2 wrstuden rw_destroy(&wl->wl_rwlock); 663 1.3.2.2 wrstuden wapbl_free(wl); 664 1.3.2.2 wrstuden 665 1.3.2.2 wrstuden return 0; 666 1.3.2.2 wrstuden } 667 1.3.2.2 wrstuden 668 1.3.2.2 wrstuden static int 669 1.3.2.2 wrstuden wapbl_doio(void *data, size_t len, struct vnode *devvp, daddr_t pbn, int flags) 670 1.3.2.2 wrstuden { 671 1.3.2.2 wrstuden struct pstats *pstats = curlwp->l_proc->p_stats; 672 1.3.2.2 wrstuden struct buf *bp; 673 1.3.2.2 wrstuden int error; 674 1.3.2.2 wrstuden 675 1.3.2.2 wrstuden KASSERT((flags & ~(B_WRITE | B_READ)) == 0); 676 1.3.2.2 wrstuden KASSERT(devvp->v_type == VBLK); 677 1.3.2.2 wrstuden 678 1.3.2.2 wrstuden if ((flags & (B_WRITE | B_READ)) == B_WRITE) { 679 1.3.2.2 wrstuden mutex_enter(&devvp->v_interlock); 680 1.3.2.2 wrstuden devvp->v_numoutput++; 681 1.3.2.2 wrstuden mutex_exit(&devvp->v_interlock); 682 1.3.2.2 wrstuden pstats->p_ru.ru_oublock++; 683 1.3.2.2 wrstuden } else { 684 1.3.2.2 wrstuden pstats->p_ru.ru_inblock++; 685 1.3.2.2 wrstuden } 686 1.3.2.2 wrstuden 687 1.3.2.2 wrstuden bp = getiobuf(devvp, true); 688 1.3.2.2 wrstuden bp->b_flags = flags; 689 1.3.2.2 wrstuden bp->b_cflags = BC_BUSY; /* silly & dubious */ 690 1.3.2.2 wrstuden bp->b_dev = devvp->v_rdev; 691 1.3.2.2 wrstuden bp->b_data = data; 692 1.3.2.2 wrstuden bp->b_bufsize = bp->b_resid = bp->b_bcount = len; 693 1.3.2.2 wrstuden bp->b_blkno = pbn; 694 1.3.2.2 wrstuden 695 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_IO, 696 1.3.2.2 wrstuden ("wapbl_doio: %s %d bytes at block %"PRId64" on dev 0x%x\n", 697 1.3.2.2 wrstuden BUF_ISWRITE(bp) ? "write" : "read", bp->b_bcount, 698 1.3.2.2 wrstuden bp->b_blkno, bp->b_dev)); 699 1.3.2.2 wrstuden 700 1.3.2.2 wrstuden VOP_STRATEGY(devvp, bp); 701 1.3.2.2 wrstuden 702 1.3.2.2 wrstuden error = biowait(bp); 703 1.3.2.2 wrstuden putiobuf(bp); 704 1.3.2.2 wrstuden 705 1.3.2.2 wrstuden if (error) { 706 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_ERROR, 707 1.3.2.2 wrstuden ("wapbl_doio: %s %zu bytes at block %" PRId64 708 1.3.2.2 wrstuden " on dev 0x%x failed with error %d\n", 709 1.3.2.2 wrstuden (((flags & (B_WRITE | B_READ)) == B_WRITE) ? 710 1.3.2.2 wrstuden "write" : "read"), 711 1.3.2.2 wrstuden len, pbn, devvp->v_rdev, error)); 712 1.3.2.2 wrstuden } 713 1.3.2.2 wrstuden 714 1.3.2.2 wrstuden return error; 715 1.3.2.2 wrstuden } 716 1.3.2.2 wrstuden 717 1.3.2.2 wrstuden int 718 1.3.2.2 wrstuden wapbl_write(void *data, size_t len, struct vnode *devvp, daddr_t pbn) 719 1.3.2.2 wrstuden { 720 1.3.2.2 wrstuden 721 1.3.2.2 wrstuden return wapbl_doio(data, len, devvp, pbn, B_WRITE); 722 1.3.2.2 wrstuden } 723 1.3.2.2 wrstuden 724 1.3.2.2 wrstuden int 725 1.3.2.2 wrstuden wapbl_read(void *data, size_t len, struct vnode *devvp, daddr_t pbn) 726 1.3.2.2 wrstuden { 727 1.3.2.2 wrstuden 728 1.3.2.2 wrstuden return wapbl_doio(data, len, devvp, pbn, B_READ); 729 1.3.2.2 wrstuden } 730 1.3.2.2 wrstuden 731 1.3.2.2 wrstuden /* 732 1.3.2.2 wrstuden * Off is byte offset returns new offset for next write 733 1.3.2.2 wrstuden * handles log wraparound 734 1.3.2.2 wrstuden */ 735 1.3.2.2 wrstuden static int 736 1.3.2.2 wrstuden wapbl_circ_write(struct wapbl *wl, void *data, size_t len, off_t *offp) 737 1.3.2.2 wrstuden { 738 1.3.2.2 wrstuden size_t slen; 739 1.3.2.2 wrstuden off_t off = *offp; 740 1.3.2.2 wrstuden int error; 741 1.3.2.2 wrstuden 742 1.3.2.2 wrstuden KDASSERT(((len >> wl->wl_log_dev_bshift) << 743 1.3.2.2 wrstuden wl->wl_log_dev_bshift) == len); 744 1.3.2.2 wrstuden 745 1.3.2.2 wrstuden if (off < wl->wl_circ_off) 746 1.3.2.2 wrstuden off = wl->wl_circ_off; 747 1.3.2.2 wrstuden slen = wl->wl_circ_off + wl->wl_circ_size - off; 748 1.3.2.2 wrstuden if (slen < len) { 749 1.3.2.2 wrstuden error = wapbl_write(data, slen, wl->wl_devvp, 750 1.3.2.2 wrstuden wl->wl_logpbn + (off >> wl->wl_log_dev_bshift)); 751 1.3.2.2 wrstuden if (error) 752 1.3.2.2 wrstuden return error; 753 1.3.2.2 wrstuden data = (uint8_t *)data + slen; 754 1.3.2.2 wrstuden len -= slen; 755 1.3.2.2 wrstuden off = wl->wl_circ_off; 756 1.3.2.2 wrstuden } 757 1.3.2.2 wrstuden error = wapbl_write(data, len, wl->wl_devvp, 758 1.3.2.2 wrstuden wl->wl_logpbn + (off >> wl->wl_log_dev_bshift)); 759 1.3.2.2 wrstuden if (error) 760 1.3.2.2 wrstuden return error; 761 1.3.2.2 wrstuden off += len; 762 1.3.2.2 wrstuden if (off >= wl->wl_circ_off + wl->wl_circ_size) 763 1.3.2.2 wrstuden off = wl->wl_circ_off; 764 1.3.2.2 wrstuden *offp = off; 765 1.3.2.2 wrstuden return 0; 766 1.3.2.2 wrstuden } 767 1.3.2.2 wrstuden 768 1.3.2.2 wrstuden /****************************************************************/ 769 1.3.2.2 wrstuden 770 1.3.2.2 wrstuden int 771 1.3.2.2 wrstuden wapbl_begin(struct wapbl *wl, const char *file, int line) 772 1.3.2.2 wrstuden { 773 1.3.2.2 wrstuden int doflush; 774 1.3.2.2 wrstuden unsigned lockcount; 775 1.3.2.2 wrstuden krw_t op; 776 1.3.2.2 wrstuden 777 1.3.2.2 wrstuden KDASSERT(wl); 778 1.3.2.2 wrstuden 779 1.3.2.2 wrstuden /* 780 1.3.2.2 wrstuden * XXX: The original code calls for the use of a RW_READER lock 781 1.3.2.2 wrstuden * here, but it turns out there are performance issues with high 782 1.3.2.2 wrstuden * metadata-rate workloads (e.g. multiple simultaneous tar 783 1.3.2.2 wrstuden * extractions). For now, we force the lock to be RW_WRITER, 784 1.3.2.2 wrstuden * since that currently has the best performance characteristics 785 1.3.2.2 wrstuden * (even for a single tar-file extraction). 786 1.3.2.2 wrstuden * 787 1.3.2.2 wrstuden */ 788 1.3.2.2 wrstuden #define WAPBL_DEBUG_SERIALIZE 1 789 1.3.2.2 wrstuden 790 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_SERIALIZE 791 1.3.2.2 wrstuden op = RW_WRITER; 792 1.3.2.2 wrstuden #else 793 1.3.2.2 wrstuden op = RW_READER; 794 1.3.2.2 wrstuden #endif 795 1.3.2.2 wrstuden 796 1.3.2.2 wrstuden /* 797 1.3.2.2 wrstuden * XXX this needs to be made much more sophisticated. 798 1.3.2.2 wrstuden * perhaps each wapbl_begin could reserve a specified 799 1.3.2.2 wrstuden * number of buffers and bytes. 800 1.3.2.2 wrstuden */ 801 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 802 1.3.2.2 wrstuden lockcount = wl->wl_lock_count; 803 1.3.2.2 wrstuden doflush = ((wl->wl_bufbytes + (lockcount * MAXPHYS)) > 804 1.3.2.2 wrstuden wl->wl_bufbytes_max / 2) || 805 1.3.2.2 wrstuden ((wl->wl_bufcount + (lockcount * 10)) > 806 1.3.2.2 wrstuden wl->wl_bufcount_max / 2) || 807 1.3.2.2 wrstuden (wapbl_transaction_len(wl) > wl->wl_circ_size / 2); 808 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 809 1.3.2.2 wrstuden 810 1.3.2.2 wrstuden if (doflush) { 811 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_FLUSH, 812 1.3.2.2 wrstuden ("force flush lockcnt=%d bufbytes=%zu " 813 1.3.2.2 wrstuden "(max=%zu) bufcount=%zu (max=%zu)\n", 814 1.3.2.2 wrstuden lockcount, wl->wl_bufbytes, 815 1.3.2.2 wrstuden wl->wl_bufbytes_max, wl->wl_bufcount, 816 1.3.2.2 wrstuden wl->wl_bufcount_max)); 817 1.3.2.2 wrstuden } 818 1.3.2.2 wrstuden 819 1.3.2.2 wrstuden if (doflush) { 820 1.3.2.2 wrstuden int error = wapbl_flush(wl, 0); 821 1.3.2.2 wrstuden if (error) 822 1.3.2.2 wrstuden return error; 823 1.3.2.2 wrstuden } 824 1.3.2.2 wrstuden 825 1.3.2.2 wrstuden rw_enter(&wl->wl_rwlock, op); 826 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 827 1.3.2.2 wrstuden wl->wl_lock_count++; 828 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 829 1.3.2.2 wrstuden 830 1.3.2.2 wrstuden #if defined(WAPBL_DEBUG_PRINT) && defined(WAPBL_DEBUG_SERIALIZE) 831 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_TRANSACTION, 832 1.3.2.2 wrstuden ("wapbl_begin thread %d.%d with bufcount=%zu " 833 1.3.2.2 wrstuden "bufbytes=%zu bcount=%zu at %s:%d\n", 834 1.3.2.2 wrstuden curproc->p_pid, curlwp->l_lid, wl->wl_bufcount, 835 1.3.2.2 wrstuden wl->wl_bufbytes, wl->wl_bcount, file, line)); 836 1.3.2.2 wrstuden #endif 837 1.3.2.2 wrstuden 838 1.3.2.2 wrstuden return 0; 839 1.3.2.2 wrstuden } 840 1.3.2.2 wrstuden 841 1.3.2.2 wrstuden void 842 1.3.2.2 wrstuden wapbl_end(struct wapbl *wl) 843 1.3.2.2 wrstuden { 844 1.3.2.2 wrstuden 845 1.3.2.2 wrstuden #if defined(WAPBL_DEBUG_PRINT) && defined(WAPBL_DEBUG_SERIALIZE) 846 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_TRANSACTION, 847 1.3.2.2 wrstuden ("wapbl_end thread %d.%d with bufcount=%zu " 848 1.3.2.2 wrstuden "bufbytes=%zu bcount=%zu\n", 849 1.3.2.2 wrstuden curproc->p_pid, curlwp->l_lid, wl->wl_bufcount, 850 1.3.2.2 wrstuden wl->wl_bufbytes, wl->wl_bcount)); 851 1.3.2.2 wrstuden #endif 852 1.3.2.2 wrstuden 853 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 854 1.3.2.2 wrstuden KASSERT(wl->wl_lock_count > 0); 855 1.3.2.2 wrstuden wl->wl_lock_count--; 856 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 857 1.3.2.2 wrstuden 858 1.3.2.2 wrstuden rw_exit(&wl->wl_rwlock); 859 1.3.2.2 wrstuden } 860 1.3.2.2 wrstuden 861 1.3.2.2 wrstuden void 862 1.3.2.2 wrstuden wapbl_add_buf(struct wapbl *wl, struct buf * bp) 863 1.3.2.2 wrstuden { 864 1.3.2.2 wrstuden 865 1.3.2.2 wrstuden KASSERT(bp->b_cflags & BC_BUSY); 866 1.3.2.2 wrstuden KASSERT(bp->b_vp); 867 1.3.2.2 wrstuden 868 1.3.2.2 wrstuden wapbl_jlock_assert(wl); 869 1.3.2.2 wrstuden 870 1.3.2.2 wrstuden #if 0 871 1.3.2.2 wrstuden /* 872 1.3.2.2 wrstuden * XXX this might be an issue for swapfiles. 873 1.3.2.2 wrstuden * see uvm_swap.c:1702 874 1.3.2.2 wrstuden * 875 1.3.2.2 wrstuden * XXX2 why require it then? leap of semantics? 876 1.3.2.2 wrstuden */ 877 1.3.2.2 wrstuden KASSERT((bp->b_cflags & BC_NOCACHE) == 0); 878 1.3.2.2 wrstuden #endif 879 1.3.2.2 wrstuden 880 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 881 1.3.2.2 wrstuden if (bp->b_flags & B_LOCKED) { 882 1.3.2.2 wrstuden LIST_REMOVE(bp, b_wapbllist); 883 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_BUFFER2, 884 1.3.2.2 wrstuden ("wapbl_add_buf thread %d.%d re-adding buf %p " 885 1.3.2.2 wrstuden "with %d bytes %d bcount\n", 886 1.3.2.2 wrstuden curproc->p_pid, curlwp->l_lid, bp, bp->b_bufsize, 887 1.3.2.2 wrstuden bp->b_bcount)); 888 1.3.2.2 wrstuden } else { 889 1.3.2.2 wrstuden /* unlocked by dirty buffers shouldn't exist */ 890 1.3.2.2 wrstuden KASSERT(!(bp->b_oflags & BO_DELWRI)); 891 1.3.2.2 wrstuden wl->wl_bufbytes += bp->b_bufsize; 892 1.3.2.2 wrstuden wl->wl_bcount += bp->b_bcount; 893 1.3.2.2 wrstuden wl->wl_bufcount++; 894 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_BUFFER, 895 1.3.2.2 wrstuden ("wapbl_add_buf thread %d.%d adding buf %p " 896 1.3.2.2 wrstuden "with %d bytes %d bcount\n", 897 1.3.2.2 wrstuden curproc->p_pid, curlwp->l_lid, bp, bp->b_bufsize, 898 1.3.2.2 wrstuden bp->b_bcount)); 899 1.3.2.2 wrstuden } 900 1.3.2.2 wrstuden LIST_INSERT_HEAD(&wl->wl_bufs, bp, b_wapbllist); 901 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 902 1.3.2.2 wrstuden 903 1.3.2.2 wrstuden bp->b_flags |= B_LOCKED; 904 1.3.2.2 wrstuden } 905 1.3.2.2 wrstuden 906 1.3.2.2 wrstuden static void 907 1.3.2.2 wrstuden wapbl_remove_buf_locked(struct wapbl * wl, struct buf *bp) 908 1.3.2.2 wrstuden { 909 1.3.2.2 wrstuden 910 1.3.2.2 wrstuden KASSERT(mutex_owned(&wl->wl_mtx)); 911 1.3.2.2 wrstuden KASSERT(bp->b_cflags & BC_BUSY); 912 1.3.2.2 wrstuden wapbl_jlock_assert(wl); 913 1.3.2.2 wrstuden 914 1.3.2.2 wrstuden #if 0 915 1.3.2.2 wrstuden /* 916 1.3.2.2 wrstuden * XXX this might be an issue for swapfiles. 917 1.3.2.2 wrstuden * see uvm_swap.c:1725 918 1.3.2.2 wrstuden * 919 1.3.2.2 wrstuden * XXXdeux: see above 920 1.3.2.2 wrstuden */ 921 1.3.2.2 wrstuden KASSERT((bp->b_flags & BC_NOCACHE) == 0); 922 1.3.2.2 wrstuden #endif 923 1.3.2.2 wrstuden KASSERT(bp->b_flags & B_LOCKED); 924 1.3.2.2 wrstuden 925 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_BUFFER, 926 1.3.2.2 wrstuden ("wapbl_remove_buf thread %d.%d removing buf %p with " 927 1.3.2.2 wrstuden "%d bytes %d bcount\n", 928 1.3.2.2 wrstuden curproc->p_pid, curlwp->l_lid, bp, bp->b_bufsize, bp->b_bcount)); 929 1.3.2.2 wrstuden 930 1.3.2.2 wrstuden KASSERT(wl->wl_bufbytes >= bp->b_bufsize); 931 1.3.2.2 wrstuden wl->wl_bufbytes -= bp->b_bufsize; 932 1.3.2.2 wrstuden KASSERT(wl->wl_bcount >= bp->b_bcount); 933 1.3.2.2 wrstuden wl->wl_bcount -= bp->b_bcount; 934 1.3.2.2 wrstuden KASSERT(wl->wl_bufcount > 0); 935 1.3.2.2 wrstuden wl->wl_bufcount--; 936 1.3.2.2 wrstuden KASSERT((wl->wl_bufcount == 0) == (wl->wl_bufbytes == 0)); 937 1.3.2.2 wrstuden KASSERT((wl->wl_bufcount == 0) == (wl->wl_bcount == 0)); 938 1.3.2.2 wrstuden LIST_REMOVE(bp, b_wapbllist); 939 1.3.2.2 wrstuden 940 1.3.2.2 wrstuden bp->b_flags &= ~B_LOCKED; 941 1.3.2.2 wrstuden } 942 1.3.2.2 wrstuden 943 1.3.2.2 wrstuden /* called from brelsel() in vfs_bio among other places */ 944 1.3.2.2 wrstuden void 945 1.3.2.2 wrstuden wapbl_remove_buf(struct wapbl * wl, struct buf *bp) 946 1.3.2.2 wrstuden { 947 1.3.2.2 wrstuden 948 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 949 1.3.2.2 wrstuden wapbl_remove_buf_locked(wl, bp); 950 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 951 1.3.2.2 wrstuden } 952 1.3.2.2 wrstuden 953 1.3.2.2 wrstuden void 954 1.3.2.2 wrstuden wapbl_resize_buf(struct wapbl *wl, struct buf *bp, long oldsz, long oldcnt) 955 1.3.2.2 wrstuden { 956 1.3.2.2 wrstuden 957 1.3.2.2 wrstuden KASSERT(bp->b_cflags & BC_BUSY); 958 1.3.2.2 wrstuden 959 1.3.2.2 wrstuden /* 960 1.3.2.2 wrstuden * XXX: why does this depend on B_LOCKED? otherwise the buf 961 1.3.2.2 wrstuden * is not for a transaction? if so, why is this called in the 962 1.3.2.2 wrstuden * first place? 963 1.3.2.2 wrstuden */ 964 1.3.2.2 wrstuden if (bp->b_flags & B_LOCKED) { 965 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 966 1.3.2.2 wrstuden wl->wl_bufbytes += bp->b_bufsize - oldsz; 967 1.3.2.2 wrstuden wl->wl_bcount += bp->b_bcount - oldcnt; 968 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 969 1.3.2.2 wrstuden } 970 1.3.2.2 wrstuden } 971 1.3.2.2 wrstuden 972 1.3.2.2 wrstuden #endif /* _KERNEL */ 973 1.3.2.2 wrstuden 974 1.3.2.2 wrstuden /****************************************************************/ 975 1.3.2.2 wrstuden /* Some utility inlines */ 976 1.3.2.2 wrstuden 977 1.3.2.2 wrstuden /* This is used to advance the pointer at old to new value at old+delta */ 978 1.3.2.2 wrstuden static __inline off_t 979 1.3.2.2 wrstuden wapbl_advance(size_t size, size_t off, off_t old, size_t delta) 980 1.3.2.2 wrstuden { 981 1.3.2.2 wrstuden off_t new; 982 1.3.2.2 wrstuden 983 1.3.2.2 wrstuden /* Define acceptable ranges for inputs. */ 984 1.3.2.2 wrstuden KASSERT(delta <= size); 985 1.3.2.2 wrstuden KASSERT((old == 0) || (old >= off)); 986 1.3.2.2 wrstuden KASSERT(old < (size + off)); 987 1.3.2.2 wrstuden 988 1.3.2.2 wrstuden if ((old == 0) && (delta != 0)) 989 1.3.2.2 wrstuden new = off + delta; 990 1.3.2.2 wrstuden else if ((old + delta) < (size + off)) 991 1.3.2.2 wrstuden new = old + delta; 992 1.3.2.2 wrstuden else 993 1.3.2.2 wrstuden new = (old + delta) - size; 994 1.3.2.2 wrstuden 995 1.3.2.2 wrstuden /* Note some interesting axioms */ 996 1.3.2.2 wrstuden KASSERT((delta != 0) || (new == old)); 997 1.3.2.2 wrstuden KASSERT((delta == 0) || (new != 0)); 998 1.3.2.2 wrstuden KASSERT((delta != (size)) || (new == old)); 999 1.3.2.2 wrstuden 1000 1.3.2.2 wrstuden /* Define acceptable ranges for output. */ 1001 1.3.2.2 wrstuden KASSERT((new == 0) || (new >= off)); 1002 1.3.2.2 wrstuden KASSERT(new < (size + off)); 1003 1.3.2.2 wrstuden return new; 1004 1.3.2.2 wrstuden } 1005 1.3.2.2 wrstuden 1006 1.3.2.2 wrstuden static __inline size_t 1007 1.3.2.2 wrstuden wapbl_space_used(size_t avail, off_t head, off_t tail) 1008 1.3.2.2 wrstuden { 1009 1.3.2.2 wrstuden 1010 1.3.2.2 wrstuden if (tail == 0) { 1011 1.3.2.2 wrstuden KASSERT(head == 0); 1012 1.3.2.2 wrstuden return 0; 1013 1.3.2.2 wrstuden } 1014 1.3.2.2 wrstuden return ((head + (avail - 1) - tail) % avail) + 1; 1015 1.3.2.2 wrstuden } 1016 1.3.2.2 wrstuden 1017 1.3.2.2 wrstuden static __inline size_t 1018 1.3.2.2 wrstuden wapbl_space_free(size_t avail, off_t head, off_t tail) 1019 1.3.2.2 wrstuden { 1020 1.3.2.2 wrstuden 1021 1.3.2.2 wrstuden return avail - wapbl_space_used(avail, head, tail); 1022 1.3.2.2 wrstuden } 1023 1.3.2.2 wrstuden 1024 1.3.2.2 wrstuden static __inline void 1025 1.3.2.2 wrstuden wapbl_advance_head(size_t size, size_t off, size_t delta, off_t *headp, 1026 1.3.2.2 wrstuden off_t *tailp) 1027 1.3.2.2 wrstuden { 1028 1.3.2.2 wrstuden off_t head = *headp; 1029 1.3.2.2 wrstuden off_t tail = *tailp; 1030 1.3.2.2 wrstuden 1031 1.3.2.2 wrstuden KASSERT(delta <= wapbl_space_free(size, head, tail)); 1032 1.3.2.2 wrstuden head = wapbl_advance(size, off, head, delta); 1033 1.3.2.2 wrstuden if ((tail == 0) && (head != 0)) 1034 1.3.2.2 wrstuden tail = off; 1035 1.3.2.2 wrstuden *headp = head; 1036 1.3.2.2 wrstuden *tailp = tail; 1037 1.3.2.2 wrstuden } 1038 1.3.2.2 wrstuden 1039 1.3.2.2 wrstuden static __inline void 1040 1.3.2.2 wrstuden wapbl_advance_tail(size_t size, size_t off, size_t delta, off_t *headp, 1041 1.3.2.2 wrstuden off_t *tailp) 1042 1.3.2.2 wrstuden { 1043 1.3.2.2 wrstuden off_t head = *headp; 1044 1.3.2.2 wrstuden off_t tail = *tailp; 1045 1.3.2.2 wrstuden 1046 1.3.2.2 wrstuden KASSERT(delta <= wapbl_space_used(size, head, tail)); 1047 1.3.2.2 wrstuden tail = wapbl_advance(size, off, tail, delta); 1048 1.3.2.2 wrstuden if (head == tail) { 1049 1.3.2.2 wrstuden head = tail = 0; 1050 1.3.2.2 wrstuden } 1051 1.3.2.2 wrstuden *headp = head; 1052 1.3.2.2 wrstuden *tailp = tail; 1053 1.3.2.2 wrstuden } 1054 1.3.2.2 wrstuden 1055 1.3.2.2 wrstuden #ifdef _KERNEL 1056 1.3.2.2 wrstuden 1057 1.3.2.2 wrstuden /****************************************************************/ 1058 1.3.2.2 wrstuden 1059 1.3.2.2 wrstuden /* 1060 1.3.2.2 wrstuden * Remove transactions whose buffers are completely flushed to disk. 1061 1.3.2.2 wrstuden * Will block until at least minfree space is available. 1062 1.3.2.2 wrstuden * only intended to be called from inside wapbl_flush and therefore 1063 1.3.2.2 wrstuden * does not protect against commit races with itself or with flush. 1064 1.3.2.2 wrstuden */ 1065 1.3.2.2 wrstuden static int 1066 1.3.2.2 wrstuden wapbl_truncate(struct wapbl *wl, size_t minfree, int waitonly) 1067 1.3.2.2 wrstuden { 1068 1.3.2.2 wrstuden size_t delta; 1069 1.3.2.2 wrstuden size_t avail; 1070 1.3.2.2 wrstuden off_t head; 1071 1.3.2.2 wrstuden off_t tail; 1072 1.3.2.2 wrstuden int error = 0; 1073 1.3.2.2 wrstuden 1074 1.3.2.2 wrstuden KASSERT(minfree <= (wl->wl_circ_size - wl->wl_reserved_bytes)); 1075 1.3.2.2 wrstuden KASSERT(rw_write_held(&wl->wl_rwlock)); 1076 1.3.2.2 wrstuden 1077 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 1078 1.3.2.2 wrstuden 1079 1.3.2.2 wrstuden /* 1080 1.3.2.2 wrstuden * First check to see if we have to do a commit 1081 1.3.2.2 wrstuden * at all. 1082 1.3.2.2 wrstuden */ 1083 1.3.2.2 wrstuden avail = wapbl_space_free(wl->wl_circ_size, wl->wl_head, wl->wl_tail); 1084 1.3.2.2 wrstuden if (minfree < avail) { 1085 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1086 1.3.2.2 wrstuden return 0; 1087 1.3.2.2 wrstuden } 1088 1.3.2.2 wrstuden minfree -= avail; 1089 1.3.2.2 wrstuden while ((wl->wl_error_count == 0) && 1090 1.3.2.2 wrstuden (wl->wl_reclaimable_bytes < minfree)) { 1091 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_TRUNCATE, 1092 1.3.2.2 wrstuden ("wapbl_truncate: sleeping on %p wl=%p bytes=%zd " 1093 1.3.2.2 wrstuden "minfree=%zd\n", 1094 1.3.2.2 wrstuden &wl->wl_reclaimable_bytes, wl, wl->wl_reclaimable_bytes, 1095 1.3.2.2 wrstuden minfree)); 1096 1.3.2.2 wrstuden 1097 1.3.2.2 wrstuden cv_wait(&wl->wl_reclaimable_cv, &wl->wl_mtx); 1098 1.3.2.2 wrstuden } 1099 1.3.2.2 wrstuden if (wl->wl_reclaimable_bytes < minfree) { 1100 1.3.2.2 wrstuden KASSERT(wl->wl_error_count); 1101 1.3.2.2 wrstuden /* XXX maybe get actual error from buffer instead someday? */ 1102 1.3.2.2 wrstuden error = EIO; 1103 1.3.2.2 wrstuden } 1104 1.3.2.2 wrstuden head = wl->wl_head; 1105 1.3.2.2 wrstuden tail = wl->wl_tail; 1106 1.3.2.2 wrstuden delta = wl->wl_reclaimable_bytes; 1107 1.3.2.2 wrstuden 1108 1.3.2.2 wrstuden /* If all of of the entries are flushed, then be sure to keep 1109 1.3.2.2 wrstuden * the reserved bytes reserved. Watch out for discarded transactions, 1110 1.3.2.2 wrstuden * which could leave more bytes reserved than are reclaimable. 1111 1.3.2.2 wrstuden */ 1112 1.3.2.2 wrstuden if (SIMPLEQ_EMPTY(&wl->wl_entries) && 1113 1.3.2.2 wrstuden (delta >= wl->wl_reserved_bytes)) { 1114 1.3.2.2 wrstuden delta -= wl->wl_reserved_bytes; 1115 1.3.2.2 wrstuden } 1116 1.3.2.2 wrstuden wapbl_advance_tail(wl->wl_circ_size, wl->wl_circ_off, delta, &head, 1117 1.3.2.2 wrstuden &tail); 1118 1.3.2.2 wrstuden KDASSERT(wl->wl_reserved_bytes <= 1119 1.3.2.2 wrstuden wapbl_space_used(wl->wl_circ_size, head, tail)); 1120 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1121 1.3.2.2 wrstuden 1122 1.3.2.2 wrstuden if (error) 1123 1.3.2.2 wrstuden return error; 1124 1.3.2.2 wrstuden 1125 1.3.2.2 wrstuden if (waitonly) 1126 1.3.2.2 wrstuden return 0; 1127 1.3.2.2 wrstuden 1128 1.3.2.2 wrstuden /* 1129 1.3.2.2 wrstuden * This is where head, tail and delta are unprotected 1130 1.3.2.2 wrstuden * from races against itself or flush. This is ok since 1131 1.3.2.2 wrstuden * we only call this routine from inside flush itself. 1132 1.3.2.2 wrstuden * 1133 1.3.2.2 wrstuden * XXX: how can it race against itself when accessed only 1134 1.3.2.2 wrstuden * from behind the write-locked rwlock? 1135 1.3.2.2 wrstuden */ 1136 1.3.2.2 wrstuden error = wapbl_write_commit(wl, head, tail); 1137 1.3.2.2 wrstuden if (error) 1138 1.3.2.2 wrstuden return error; 1139 1.3.2.2 wrstuden 1140 1.3.2.2 wrstuden wl->wl_head = head; 1141 1.3.2.2 wrstuden wl->wl_tail = tail; 1142 1.3.2.2 wrstuden 1143 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 1144 1.3.2.2 wrstuden KASSERT(wl->wl_reclaimable_bytes >= delta); 1145 1.3.2.2 wrstuden wl->wl_reclaimable_bytes -= delta; 1146 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1147 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_TRUNCATE, 1148 1.3.2.2 wrstuden ("wapbl_truncate thread %d.%d truncating %zu bytes\n", 1149 1.3.2.2 wrstuden curproc->p_pid, curlwp->l_lid, delta)); 1150 1.3.2.2 wrstuden 1151 1.3.2.2 wrstuden return 0; 1152 1.3.2.2 wrstuden } 1153 1.3.2.2 wrstuden 1154 1.3.2.2 wrstuden /****************************************************************/ 1155 1.3.2.2 wrstuden 1156 1.3.2.2 wrstuden void 1157 1.3.2.2 wrstuden wapbl_biodone(struct buf *bp) 1158 1.3.2.2 wrstuden { 1159 1.3.2.2 wrstuden struct wapbl_entry *we = bp->b_private; 1160 1.3.2.2 wrstuden struct wapbl *wl = we->we_wapbl; 1161 1.3.2.2 wrstuden 1162 1.3.2.2 wrstuden /* 1163 1.3.2.2 wrstuden * Handle possible flushing of buffers after log has been 1164 1.3.2.2 wrstuden * decomissioned. 1165 1.3.2.2 wrstuden */ 1166 1.3.2.2 wrstuden if (!wl) { 1167 1.3.2.2 wrstuden KASSERT(we->we_bufcount > 0); 1168 1.3.2.2 wrstuden we->we_bufcount--; 1169 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 1170 1.3.2.2 wrstuden KASSERT(we->we_unsynced_bufbytes >= bp->b_bufsize); 1171 1.3.2.2 wrstuden we->we_unsynced_bufbytes -= bp->b_bufsize; 1172 1.3.2.2 wrstuden #endif 1173 1.3.2.2 wrstuden 1174 1.3.2.2 wrstuden if (we->we_bufcount == 0) { 1175 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 1176 1.3.2.2 wrstuden KASSERT(we->we_unsynced_bufbytes == 0); 1177 1.3.2.2 wrstuden #endif 1178 1.3.2.2 wrstuden wapbl_free(we); 1179 1.3.2.2 wrstuden } 1180 1.3.2.2 wrstuden 1181 1.3.2.2 wrstuden brelse(bp, 0); 1182 1.3.2.2 wrstuden return; 1183 1.3.2.2 wrstuden } 1184 1.3.2.2 wrstuden 1185 1.3.2.2 wrstuden #ifdef ohbother 1186 1.3.2.2 wrstuden KDASSERT(bp->b_flags & B_DONE); 1187 1.3.2.2 wrstuden KDASSERT(!(bp->b_flags & B_DELWRI)); 1188 1.3.2.2 wrstuden KDASSERT(bp->b_flags & B_ASYNC); 1189 1.3.2.2 wrstuden KDASSERT(bp->b_flags & B_BUSY); 1190 1.3.2.2 wrstuden KDASSERT(!(bp->b_flags & B_LOCKED)); 1191 1.3.2.2 wrstuden KDASSERT(!(bp->b_flags & B_READ)); 1192 1.3.2.2 wrstuden KDASSERT(!(bp->b_flags & B_INVAL)); 1193 1.3.2.2 wrstuden KDASSERT(!(bp->b_flags & B_NOCACHE)); 1194 1.3.2.2 wrstuden #endif 1195 1.3.2.2 wrstuden 1196 1.3.2.2 wrstuden if (bp->b_error) { 1197 1.3.2.2 wrstuden #ifdef notyet /* Can't currently handle possible dirty buffer reuse */ 1198 1.3.2.2 wrstuden XXXpooka: interfaces not fully updated 1199 1.3.2.2 wrstuden Note: this was not enabled in the original patch 1200 1.3.2.2 wrstuden against netbsd4 either. I don't know if comment 1201 1.3.2.2 wrstuden above is true or not. 1202 1.3.2.2 wrstuden 1203 1.3.2.2 wrstuden /* 1204 1.3.2.2 wrstuden * If an error occurs, report the error and leave the 1205 1.3.2.2 wrstuden * buffer as a delayed write on the LRU queue. 1206 1.3.2.2 wrstuden * restarting the write would likely result in 1207 1.3.2.2 wrstuden * an error spinloop, so let it be done harmlessly 1208 1.3.2.2 wrstuden * by the syncer. 1209 1.3.2.2 wrstuden */ 1210 1.3.2.2 wrstuden bp->b_flags &= ~(B_DONE); 1211 1.3.2.2 wrstuden simple_unlock(&bp->b_interlock); 1212 1.3.2.2 wrstuden 1213 1.3.2.2 wrstuden if (we->we_error == 0) { 1214 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 1215 1.3.2.2 wrstuden wl->wl_error_count++; 1216 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1217 1.3.2.2 wrstuden cv_broadcast(&wl->wl_reclaimable_cv); 1218 1.3.2.2 wrstuden } 1219 1.3.2.2 wrstuden we->we_error = bp->b_error; 1220 1.3.2.2 wrstuden bp->b_error = 0; 1221 1.3.2.2 wrstuden brelse(bp); 1222 1.3.2.2 wrstuden return; 1223 1.3.2.2 wrstuden #else 1224 1.3.2.2 wrstuden /* For now, just mark the log permanently errored out */ 1225 1.3.2.2 wrstuden 1226 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 1227 1.3.2.2 wrstuden if (wl->wl_error_count == 0) { 1228 1.3.2.2 wrstuden wl->wl_error_count++; 1229 1.3.2.2 wrstuden cv_broadcast(&wl->wl_reclaimable_cv); 1230 1.3.2.2 wrstuden } 1231 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1232 1.3.2.2 wrstuden #endif 1233 1.3.2.2 wrstuden } 1234 1.3.2.2 wrstuden 1235 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 1236 1.3.2.2 wrstuden 1237 1.3.2.2 wrstuden KASSERT(we->we_bufcount > 0); 1238 1.3.2.2 wrstuden we->we_bufcount--; 1239 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 1240 1.3.2.2 wrstuden KASSERT(we->we_unsynced_bufbytes >= bp->b_bufsize); 1241 1.3.2.2 wrstuden we->we_unsynced_bufbytes -= bp->b_bufsize; 1242 1.3.2.2 wrstuden KASSERT(wl->wl_unsynced_bufbytes >= bp->b_bufsize); 1243 1.3.2.2 wrstuden wl->wl_unsynced_bufbytes -= bp->b_bufsize; 1244 1.3.2.2 wrstuden #endif 1245 1.3.2.2 wrstuden 1246 1.3.2.2 wrstuden /* 1247 1.3.2.2 wrstuden * If the current transaction can be reclaimed, start 1248 1.3.2.2 wrstuden * at the beginning and reclaim any consecutive reclaimable 1249 1.3.2.2 wrstuden * transactions. If we successfully reclaim anything, 1250 1.3.2.2 wrstuden * then wakeup anyone waiting for the reclaim. 1251 1.3.2.2 wrstuden */ 1252 1.3.2.2 wrstuden if (we->we_bufcount == 0) { 1253 1.3.2.2 wrstuden size_t delta = 0; 1254 1.3.2.2 wrstuden int errcnt = 0; 1255 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 1256 1.3.2.2 wrstuden KDASSERT(we->we_unsynced_bufbytes == 0); 1257 1.3.2.2 wrstuden #endif 1258 1.3.2.2 wrstuden /* 1259 1.3.2.2 wrstuden * clear any posted error, since the buffer it came from 1260 1.3.2.2 wrstuden * has successfully flushed by now 1261 1.3.2.2 wrstuden */ 1262 1.3.2.2 wrstuden while ((we = SIMPLEQ_FIRST(&wl->wl_entries)) && 1263 1.3.2.2 wrstuden (we->we_bufcount == 0)) { 1264 1.3.2.2 wrstuden delta += we->we_reclaimable_bytes; 1265 1.3.2.2 wrstuden if (we->we_error) 1266 1.3.2.2 wrstuden errcnt++; 1267 1.3.2.2 wrstuden SIMPLEQ_REMOVE_HEAD(&wl->wl_entries, we_entries); 1268 1.3.2.2 wrstuden wapbl_free(we); 1269 1.3.2.2 wrstuden } 1270 1.3.2.2 wrstuden 1271 1.3.2.2 wrstuden if (delta) { 1272 1.3.2.2 wrstuden wl->wl_reclaimable_bytes += delta; 1273 1.3.2.2 wrstuden KASSERT(wl->wl_error_count >= errcnt); 1274 1.3.2.2 wrstuden wl->wl_error_count -= errcnt; 1275 1.3.2.2 wrstuden cv_broadcast(&wl->wl_reclaimable_cv); 1276 1.3.2.2 wrstuden } 1277 1.3.2.2 wrstuden } 1278 1.3.2.2 wrstuden 1279 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1280 1.3.2.2 wrstuden brelse(bp, 0); 1281 1.3.2.2 wrstuden } 1282 1.3.2.2 wrstuden 1283 1.3.2.2 wrstuden /* 1284 1.3.2.2 wrstuden * Write transactions to disk + start I/O for contents 1285 1.3.2.2 wrstuden */ 1286 1.3.2.2 wrstuden int 1287 1.3.2.2 wrstuden wapbl_flush(struct wapbl *wl, int waitfor) 1288 1.3.2.2 wrstuden { 1289 1.3.2.2 wrstuden struct buf *bp; 1290 1.3.2.2 wrstuden struct wapbl_entry *we; 1291 1.3.2.2 wrstuden off_t off; 1292 1.3.2.2 wrstuden off_t head; 1293 1.3.2.2 wrstuden off_t tail; 1294 1.3.2.2 wrstuden size_t delta = 0; 1295 1.3.2.2 wrstuden size_t flushsize; 1296 1.3.2.2 wrstuden size_t reserved; 1297 1.3.2.2 wrstuden int error = 0; 1298 1.3.2.2 wrstuden 1299 1.3.2.2 wrstuden /* 1300 1.3.2.2 wrstuden * Do a quick check to see if a full flush can be skipped 1301 1.3.2.2 wrstuden * This assumes that the flush callback does not need to be called 1302 1.3.2.2 wrstuden * unless there are other outstanding bufs. 1303 1.3.2.2 wrstuden */ 1304 1.3.2.2 wrstuden if (!waitfor) { 1305 1.3.2.2 wrstuden size_t nbufs; 1306 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); /* XXX need mutex here to 1307 1.3.2.2 wrstuden protect the KASSERTS */ 1308 1.3.2.2 wrstuden nbufs = wl->wl_bufcount; 1309 1.3.2.2 wrstuden KASSERT((wl->wl_bufcount == 0) == (wl->wl_bufbytes == 0)); 1310 1.3.2.2 wrstuden KASSERT((wl->wl_bufcount == 0) == (wl->wl_bcount == 0)); 1311 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1312 1.3.2.2 wrstuden if (nbufs == 0) 1313 1.3.2.2 wrstuden return 0; 1314 1.3.2.2 wrstuden } 1315 1.3.2.2 wrstuden 1316 1.3.2.2 wrstuden /* 1317 1.3.2.2 wrstuden * XXX we may consider using LK_UPGRADE here 1318 1.3.2.2 wrstuden * if we want to call flush from inside a transaction 1319 1.3.2.2 wrstuden */ 1320 1.3.2.2 wrstuden rw_enter(&wl->wl_rwlock, RW_WRITER); 1321 1.3.2.2 wrstuden wl->wl_flush(wl->wl_mount, wl->wl_deallocblks, wl->wl_dealloclens, 1322 1.3.2.2 wrstuden wl->wl_dealloccnt); 1323 1.3.2.2 wrstuden 1324 1.3.2.2 wrstuden /* 1325 1.3.2.2 wrstuden * Now that we are fully locked and flushed, 1326 1.3.2.2 wrstuden * do another check for nothing to do. 1327 1.3.2.2 wrstuden */ 1328 1.3.2.2 wrstuden if (wl->wl_bufcount == 0) { 1329 1.3.2.2 wrstuden goto out; 1330 1.3.2.2 wrstuden } 1331 1.3.2.2 wrstuden 1332 1.3.2.2 wrstuden #if 0 1333 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_FLUSH, 1334 1.3.2.2 wrstuden ("wapbl_flush thread %d.%d flushing entries with " 1335 1.3.2.2 wrstuden "bufcount=%zu bufbytes=%zu\n", 1336 1.3.2.2 wrstuden curproc->p_pid, curlwp->l_lid, wl->wl_bufcount, 1337 1.3.2.2 wrstuden wl->wl_bufbytes)); 1338 1.3.2.2 wrstuden #endif 1339 1.3.2.2 wrstuden 1340 1.3.2.2 wrstuden /* Calculate amount of space needed to flush */ 1341 1.3.2.2 wrstuden flushsize = wapbl_transaction_len(wl); 1342 1.3.2.2 wrstuden 1343 1.3.2.2 wrstuden if (flushsize > (wl->wl_circ_size - wl->wl_reserved_bytes)) { 1344 1.3.2.2 wrstuden /* 1345 1.3.2.2 wrstuden * XXX this could be handled more gracefully, perhaps place 1346 1.3.2.2 wrstuden * only a partial transaction in the log and allow the 1347 1.3.2.2 wrstuden * remaining to flush without the protection of the journal. 1348 1.3.2.2 wrstuden */ 1349 1.3.2.2 wrstuden panic("wapbl_flush: current transaction too big to flush\n"); 1350 1.3.2.2 wrstuden } 1351 1.3.2.2 wrstuden 1352 1.3.2.2 wrstuden error = wapbl_truncate(wl, flushsize, 0); 1353 1.3.2.2 wrstuden if (error) 1354 1.3.2.2 wrstuden goto out2; 1355 1.3.2.2 wrstuden 1356 1.3.2.2 wrstuden off = wl->wl_head; 1357 1.3.2.2 wrstuden KASSERT((off == 0) || ((off >= wl->wl_circ_off) && 1358 1.3.2.2 wrstuden (off < wl->wl_circ_off + wl->wl_circ_size))); 1359 1.3.2.2 wrstuden error = wapbl_write_blocks(wl, &off); 1360 1.3.2.2 wrstuden if (error) 1361 1.3.2.2 wrstuden goto out2; 1362 1.3.2.2 wrstuden error = wapbl_write_revocations(wl, &off); 1363 1.3.2.2 wrstuden if (error) 1364 1.3.2.2 wrstuden goto out2; 1365 1.3.2.2 wrstuden error = wapbl_write_inodes(wl, &off); 1366 1.3.2.2 wrstuden if (error) 1367 1.3.2.2 wrstuden goto out2; 1368 1.3.2.2 wrstuden 1369 1.3.2.2 wrstuden reserved = 0; 1370 1.3.2.2 wrstuden if (wl->wl_inohashcnt) 1371 1.3.2.2 wrstuden reserved = wapbl_transaction_inodes_len(wl); 1372 1.3.2.2 wrstuden 1373 1.3.2.2 wrstuden head = wl->wl_head; 1374 1.3.2.2 wrstuden tail = wl->wl_tail; 1375 1.3.2.2 wrstuden 1376 1.3.2.2 wrstuden wapbl_advance_head(wl->wl_circ_size, wl->wl_circ_off, flushsize, 1377 1.3.2.2 wrstuden &head, &tail); 1378 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG 1379 1.3.2.2 wrstuden if (head != off) { 1380 1.3.2.2 wrstuden panic("lost head! head=%"PRIdMAX" tail=%" PRIdMAX 1381 1.3.2.2 wrstuden " off=%"PRIdMAX" flush=%zu\n", 1382 1.3.2.2 wrstuden (intmax_t)head, (intmax_t)tail, (intmax_t)off, 1383 1.3.2.2 wrstuden flushsize); 1384 1.3.2.2 wrstuden } 1385 1.3.2.2 wrstuden #else 1386 1.3.2.2 wrstuden KASSERT(head == off); 1387 1.3.2.2 wrstuden #endif 1388 1.3.2.2 wrstuden 1389 1.3.2.2 wrstuden /* Opportunistically move the tail forward if we can */ 1390 1.3.2.2 wrstuden if (!wapbl_lazy_truncate) { 1391 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 1392 1.3.2.2 wrstuden delta = wl->wl_reclaimable_bytes; 1393 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1394 1.3.2.2 wrstuden wapbl_advance_tail(wl->wl_circ_size, wl->wl_circ_off, delta, 1395 1.3.2.2 wrstuden &head, &tail); 1396 1.3.2.2 wrstuden } 1397 1.3.2.2 wrstuden 1398 1.3.2.2 wrstuden error = wapbl_write_commit(wl, head, tail); 1399 1.3.2.2 wrstuden if (error) 1400 1.3.2.2 wrstuden goto out2; 1401 1.3.2.2 wrstuden 1402 1.3.2.2 wrstuden /* poolme? or kmemme? */ 1403 1.3.2.2 wrstuden we = wapbl_calloc(1, sizeof(*we)); 1404 1.3.2.2 wrstuden 1405 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 1406 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_FLUSH, 1407 1.3.2.2 wrstuden ("wapbl_flush: thread %d.%d head+=%zu tail+=%zu used=%zu" 1408 1.3.2.2 wrstuden " unsynced=%zu" 1409 1.3.2.2 wrstuden "\n\tbufcount=%zu bufbytes=%zu bcount=%zu deallocs=%d " 1410 1.3.2.2 wrstuden "inodes=%d\n", 1411 1.3.2.2 wrstuden curproc->p_pid, curlwp->l_lid, flushsize, delta, 1412 1.3.2.2 wrstuden wapbl_space_used(wl->wl_circ_size, head, tail), 1413 1.3.2.2 wrstuden wl->wl_unsynced_bufbytes, wl->wl_bufcount, 1414 1.3.2.2 wrstuden wl->wl_bufbytes, wl->wl_bcount, wl->wl_dealloccnt, 1415 1.3.2.2 wrstuden wl->wl_inohashcnt)); 1416 1.3.2.2 wrstuden #else 1417 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_FLUSH, 1418 1.3.2.2 wrstuden ("wapbl_flush: thread %d.%d head+=%zu tail+=%zu used=%zu" 1419 1.3.2.2 wrstuden "\n\tbufcount=%zu bufbytes=%zu bcount=%zu deallocs=%d " 1420 1.3.2.2 wrstuden "inodes=%d\n", 1421 1.3.2.2 wrstuden curproc->p_pid, curlwp->l_lid, flushsize, delta, 1422 1.3.2.2 wrstuden wapbl_space_used(wl->wl_circ_size, head, tail), 1423 1.3.2.2 wrstuden wl->wl_bufcount, wl->wl_bufbytes, wl->wl_bcount, 1424 1.3.2.2 wrstuden wl->wl_dealloccnt, wl->wl_inohashcnt)); 1425 1.3.2.2 wrstuden #endif 1426 1.3.2.2 wrstuden 1427 1.3.2.2 wrstuden 1428 1.3.2.2 wrstuden mutex_enter(&bufcache_lock); 1429 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 1430 1.3.2.2 wrstuden 1431 1.3.2.2 wrstuden wl->wl_reserved_bytes = reserved; 1432 1.3.2.2 wrstuden wl->wl_head = head; 1433 1.3.2.2 wrstuden wl->wl_tail = tail; 1434 1.3.2.2 wrstuden KASSERT(wl->wl_reclaimable_bytes >= delta); 1435 1.3.2.2 wrstuden wl->wl_reclaimable_bytes -= delta; 1436 1.3.2.2 wrstuden wl->wl_dealloccnt = 0; 1437 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 1438 1.3.2.2 wrstuden wl->wl_unsynced_bufbytes += wl->wl_bufbytes; 1439 1.3.2.2 wrstuden #endif 1440 1.3.2.2 wrstuden 1441 1.3.2.2 wrstuden we->we_wapbl = wl; 1442 1.3.2.2 wrstuden we->we_bufcount = wl->wl_bufcount; 1443 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 1444 1.3.2.2 wrstuden we->we_unsynced_bufbytes = wl->wl_bufbytes; 1445 1.3.2.2 wrstuden #endif 1446 1.3.2.2 wrstuden we->we_reclaimable_bytes = flushsize; 1447 1.3.2.2 wrstuden we->we_error = 0; 1448 1.3.2.2 wrstuden SIMPLEQ_INSERT_TAIL(&wl->wl_entries, we, we_entries); 1449 1.3.2.2 wrstuden 1450 1.3.2.2 wrstuden /* 1451 1.3.2.2 wrstuden * this flushes bufs in reverse order than they were queued 1452 1.3.2.2 wrstuden * it shouldn't matter, but if we care we could use TAILQ instead. 1453 1.3.2.2 wrstuden * XXX Note they will get put on the lru queue when they flush 1454 1.3.2.2 wrstuden * so we might actually want to change this to preserve order. 1455 1.3.2.2 wrstuden */ 1456 1.3.2.2 wrstuden while ((bp = LIST_FIRST(&wl->wl_bufs)) != NULL) { 1457 1.3.2.2 wrstuden if (bbusy(bp, 0, 0, &wl->wl_mtx)) { 1458 1.3.2.2 wrstuden continue; 1459 1.3.2.2 wrstuden } 1460 1.3.2.2 wrstuden bp->b_iodone = wapbl_biodone; 1461 1.3.2.2 wrstuden bp->b_private = we; 1462 1.3.2.2 wrstuden bremfree(bp); 1463 1.3.2.2 wrstuden wapbl_remove_buf_locked(wl, bp); 1464 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1465 1.3.2.2 wrstuden mutex_exit(&bufcache_lock); 1466 1.3.2.2 wrstuden bawrite(bp); 1467 1.3.2.2 wrstuden mutex_enter(&bufcache_lock); 1468 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 1469 1.3.2.2 wrstuden } 1470 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1471 1.3.2.2 wrstuden mutex_exit(&bufcache_lock); 1472 1.3.2.2 wrstuden 1473 1.3.2.2 wrstuden #if 0 1474 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_FLUSH, 1475 1.3.2.2 wrstuden ("wapbl_flush thread %d.%d done flushing entries...\n", 1476 1.3.2.2 wrstuden curproc->p_pid, curlwp->l_lid)); 1477 1.3.2.2 wrstuden #endif 1478 1.3.2.2 wrstuden 1479 1.3.2.2 wrstuden out: 1480 1.3.2.2 wrstuden 1481 1.3.2.2 wrstuden /* 1482 1.3.2.2 wrstuden * If the waitfor flag is set, don't return until everything is 1483 1.3.2.2 wrstuden * fully flushed and the on disk log is empty. 1484 1.3.2.2 wrstuden */ 1485 1.3.2.2 wrstuden if (waitfor) { 1486 1.3.2.2 wrstuden error = wapbl_truncate(wl, wl->wl_circ_size - 1487 1.3.2.2 wrstuden wl->wl_reserved_bytes, wapbl_lazy_truncate); 1488 1.3.2.2 wrstuden } 1489 1.3.2.2 wrstuden 1490 1.3.2.2 wrstuden out2: 1491 1.3.2.2 wrstuden if (error) { 1492 1.3.2.2 wrstuden wl->wl_flush_abort(wl->wl_mount, wl->wl_deallocblks, 1493 1.3.2.2 wrstuden wl->wl_dealloclens, wl->wl_dealloccnt); 1494 1.3.2.2 wrstuden } 1495 1.3.2.2 wrstuden 1496 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_PRINT 1497 1.3.2.2 wrstuden if (error) { 1498 1.3.2.2 wrstuden pid_t pid = -1; 1499 1.3.2.2 wrstuden lwpid_t lid = -1; 1500 1.3.2.2 wrstuden if (curproc) 1501 1.3.2.2 wrstuden pid = curproc->p_pid; 1502 1.3.2.2 wrstuden if (curlwp) 1503 1.3.2.2 wrstuden lid = curlwp->l_lid; 1504 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 1505 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 1506 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_ERROR, 1507 1.3.2.2 wrstuden ("wapbl_flush: thread %d.%d aborted flush: " 1508 1.3.2.2 wrstuden "error = %d\n" 1509 1.3.2.2 wrstuden "\tbufcount=%zu bufbytes=%zu bcount=%zu " 1510 1.3.2.2 wrstuden "deallocs=%d inodes=%d\n" 1511 1.3.2.2 wrstuden "\terrcnt = %d, reclaimable=%zu reserved=%zu " 1512 1.3.2.2 wrstuden "unsynced=%zu\n", 1513 1.3.2.2 wrstuden pid, lid, error, wl->wl_bufcount, 1514 1.3.2.2 wrstuden wl->wl_bufbytes, wl->wl_bcount, 1515 1.3.2.2 wrstuden wl->wl_dealloccnt, wl->wl_inohashcnt, 1516 1.3.2.2 wrstuden wl->wl_error_count, wl->wl_reclaimable_bytes, 1517 1.3.2.2 wrstuden wl->wl_reserved_bytes, wl->wl_unsynced_bufbytes)); 1518 1.3.2.2 wrstuden SIMPLEQ_FOREACH(we, &wl->wl_entries, we_entries) { 1519 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_ERROR, 1520 1.3.2.2 wrstuden ("\tentry: bufcount = %zu, reclaimable = %zu, " 1521 1.3.2.2 wrstuden "error = %d, unsynced = %zu\n", 1522 1.3.2.2 wrstuden we->we_bufcount, we->we_reclaimable_bytes, 1523 1.3.2.2 wrstuden we->we_error, we->we_unsynced_bufbytes)); 1524 1.3.2.2 wrstuden } 1525 1.3.2.2 wrstuden #else 1526 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_ERROR, 1527 1.3.2.2 wrstuden ("wapbl_flush: thread %d.%d aborted flush: " 1528 1.3.2.2 wrstuden "error = %d\n" 1529 1.3.2.2 wrstuden "\tbufcount=%zu bufbytes=%zu bcount=%zu " 1530 1.3.2.2 wrstuden "deallocs=%d inodes=%d\n" 1531 1.3.2.2 wrstuden "\terrcnt = %d, reclaimable=%zu reserved=%zu\n", 1532 1.3.2.2 wrstuden pid, lid, error, wl->wl_bufcount, 1533 1.3.2.2 wrstuden wl->wl_bufbytes, wl->wl_bcount, 1534 1.3.2.2 wrstuden wl->wl_dealloccnt, wl->wl_inohashcnt, 1535 1.3.2.2 wrstuden wl->wl_error_count, wl->wl_reclaimable_bytes, 1536 1.3.2.2 wrstuden wl->wl_reserved_bytes)); 1537 1.3.2.2 wrstuden SIMPLEQ_FOREACH(we, &wl->wl_entries, we_entries) { 1538 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_ERROR, 1539 1.3.2.2 wrstuden ("\tentry: bufcount = %zu, reclaimable = %zu, " 1540 1.3.2.2 wrstuden "error = %d\n", we->we_bufcount, 1541 1.3.2.2 wrstuden we->we_reclaimable_bytes, we->we_error)); 1542 1.3.2.2 wrstuden } 1543 1.3.2.2 wrstuden #endif 1544 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1545 1.3.2.2 wrstuden } 1546 1.3.2.2 wrstuden #endif 1547 1.3.2.2 wrstuden 1548 1.3.2.2 wrstuden rw_exit(&wl->wl_rwlock); 1549 1.3.2.2 wrstuden return error; 1550 1.3.2.2 wrstuden } 1551 1.3.2.2 wrstuden 1552 1.3.2.2 wrstuden /****************************************************************/ 1553 1.3.2.2 wrstuden 1554 1.3.2.2 wrstuden void 1555 1.3.2.2 wrstuden wapbl_jlock_assert(struct wapbl *wl) 1556 1.3.2.2 wrstuden { 1557 1.3.2.2 wrstuden 1558 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_SERIALIZE 1559 1.3.2.2 wrstuden KASSERT(rw_write_held(&wl->wl_rwlock)); 1560 1.3.2.2 wrstuden #else 1561 1.3.2.2 wrstuden KASSERT(rw_read_held(&wl->wl_rwlock) || rw_write_held(&wl->wl_rwlock)); 1562 1.3.2.2 wrstuden #endif 1563 1.3.2.2 wrstuden } 1564 1.3.2.2 wrstuden 1565 1.3.2.2 wrstuden void 1566 1.3.2.2 wrstuden wapbl_junlock_assert(struct wapbl *wl) 1567 1.3.2.2 wrstuden { 1568 1.3.2.2 wrstuden 1569 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_SERIALIZE 1570 1.3.2.2 wrstuden KASSERT(!rw_write_held(&wl->wl_rwlock)); 1571 1.3.2.2 wrstuden #endif 1572 1.3.2.2 wrstuden } 1573 1.3.2.2 wrstuden 1574 1.3.2.2 wrstuden /****************************************************************/ 1575 1.3.2.2 wrstuden 1576 1.3.2.2 wrstuden /* locks missing */ 1577 1.3.2.2 wrstuden void 1578 1.3.2.2 wrstuden wapbl_print(struct wapbl *wl, 1579 1.3.2.2 wrstuden int full, 1580 1.3.2.2 wrstuden void (*pr)(const char *, ...)) 1581 1.3.2.2 wrstuden { 1582 1.3.2.2 wrstuden struct buf *bp; 1583 1.3.2.2 wrstuden struct wapbl_entry *we; 1584 1.3.2.2 wrstuden (*pr)("wapbl %p", wl); 1585 1.3.2.2 wrstuden (*pr)("\nlogvp = %p, devvp = %p, logpbn = %"PRId64"\n", 1586 1.3.2.2 wrstuden wl->wl_logvp, wl->wl_devvp, wl->wl_logpbn); 1587 1.3.2.2 wrstuden (*pr)("circ = %zu, header = %zu, head = %"PRIdMAX" tail = %"PRIdMAX"\n", 1588 1.3.2.2 wrstuden wl->wl_circ_size, wl->wl_circ_off, 1589 1.3.2.2 wrstuden (intmax_t)wl->wl_head, (intmax_t)wl->wl_tail); 1590 1.3.2.2 wrstuden (*pr)("fs_dev_bshift = %d, log_dev_bshift = %d\n", 1591 1.3.2.2 wrstuden wl->wl_log_dev_bshift, wl->wl_fs_dev_bshift); 1592 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 1593 1.3.2.2 wrstuden (*pr)("bufcount = %zu, bufbytes = %zu bcount = %zu reclaimable = %zu " 1594 1.3.2.2 wrstuden "reserved = %zu errcnt = %d unsynced = %zu\n", 1595 1.3.2.2 wrstuden wl->wl_bufcount, wl->wl_bufbytes, wl->wl_bcount, 1596 1.3.2.2 wrstuden wl->wl_reclaimable_bytes, wl->wl_reserved_bytes, 1597 1.3.2.2 wrstuden wl->wl_error_count, wl->wl_unsynced_bufbytes); 1598 1.3.2.2 wrstuden #else 1599 1.3.2.2 wrstuden (*pr)("bufcount = %zu, bufbytes = %zu bcount = %zu reclaimable = %zu " 1600 1.3.2.2 wrstuden "reserved = %zu errcnt = %d\n", wl->wl_bufcount, wl->wl_bufbytes, 1601 1.3.2.2 wrstuden wl->wl_bcount, wl->wl_reclaimable_bytes, wl->wl_reserved_bytes, 1602 1.3.2.2 wrstuden wl->wl_error_count); 1603 1.3.2.2 wrstuden #endif 1604 1.3.2.2 wrstuden (*pr)("\tdealloccnt = %d, dealloclim = %d\n", 1605 1.3.2.2 wrstuden wl->wl_dealloccnt, wl->wl_dealloclim); 1606 1.3.2.2 wrstuden (*pr)("\tinohashcnt = %d, inohashmask = 0x%08x\n", 1607 1.3.2.2 wrstuden wl->wl_inohashcnt, wl->wl_inohashmask); 1608 1.3.2.2 wrstuden (*pr)("entries:\n"); 1609 1.3.2.2 wrstuden SIMPLEQ_FOREACH(we, &wl->wl_entries, we_entries) { 1610 1.3.2.2 wrstuden #ifdef WAPBL_DEBUG_BUFBYTES 1611 1.3.2.2 wrstuden (*pr)("\tbufcount = %zu, reclaimable = %zu, error = %d, " 1612 1.3.2.2 wrstuden "unsynced = %zu\n", 1613 1.3.2.2 wrstuden we->we_bufcount, we->we_reclaimable_bytes, 1614 1.3.2.2 wrstuden we->we_error, we->we_unsynced_bufbytes); 1615 1.3.2.2 wrstuden #else 1616 1.3.2.2 wrstuden (*pr)("\tbufcount = %zu, reclaimable = %zu, error = %d\n", 1617 1.3.2.2 wrstuden we->we_bufcount, we->we_reclaimable_bytes, we->we_error); 1618 1.3.2.2 wrstuden #endif 1619 1.3.2.2 wrstuden } 1620 1.3.2.2 wrstuden if (full) { 1621 1.3.2.2 wrstuden int cnt = 0; 1622 1.3.2.2 wrstuden (*pr)("bufs ="); 1623 1.3.2.2 wrstuden LIST_FOREACH(bp, &wl->wl_bufs, b_wapbllist) { 1624 1.3.2.2 wrstuden if (!LIST_NEXT(bp, b_wapbllist)) { 1625 1.3.2.2 wrstuden (*pr)(" %p", bp); 1626 1.3.2.2 wrstuden } else if ((++cnt % 6) == 0) { 1627 1.3.2.2 wrstuden (*pr)(" %p,\n\t", bp); 1628 1.3.2.2 wrstuden } else { 1629 1.3.2.2 wrstuden (*pr)(" %p,", bp); 1630 1.3.2.2 wrstuden } 1631 1.3.2.2 wrstuden } 1632 1.3.2.2 wrstuden (*pr)("\n"); 1633 1.3.2.2 wrstuden 1634 1.3.2.2 wrstuden (*pr)("dealloced blks = "); 1635 1.3.2.2 wrstuden { 1636 1.3.2.2 wrstuden int i; 1637 1.3.2.2 wrstuden cnt = 0; 1638 1.3.2.2 wrstuden for (i = 0; i < wl->wl_dealloccnt; i++) { 1639 1.3.2.2 wrstuden (*pr)(" %"PRId64":%d,", 1640 1.3.2.2 wrstuden wl->wl_deallocblks[i], 1641 1.3.2.2 wrstuden wl->wl_dealloclens[i]); 1642 1.3.2.2 wrstuden if ((++cnt % 4) == 0) { 1643 1.3.2.2 wrstuden (*pr)("\n\t"); 1644 1.3.2.2 wrstuden } 1645 1.3.2.2 wrstuden } 1646 1.3.2.2 wrstuden } 1647 1.3.2.2 wrstuden (*pr)("\n"); 1648 1.3.2.2 wrstuden 1649 1.3.2.2 wrstuden (*pr)("registered inodes = "); 1650 1.3.2.2 wrstuden { 1651 1.3.2.2 wrstuden int i; 1652 1.3.2.2 wrstuden cnt = 0; 1653 1.3.2.2 wrstuden for (i = 0; i <= wl->wl_inohashmask; i++) { 1654 1.3.2.2 wrstuden struct wapbl_ino_head *wih; 1655 1.3.2.2 wrstuden struct wapbl_ino *wi; 1656 1.3.2.2 wrstuden 1657 1.3.2.2 wrstuden wih = &wl->wl_inohash[i]; 1658 1.3.2.2 wrstuden LIST_FOREACH(wi, wih, wi_hash) { 1659 1.3.2.2 wrstuden if (wi->wi_ino == 0) 1660 1.3.2.2 wrstuden continue; 1661 1.3.2.2 wrstuden (*pr)(" %"PRId32"/0%06"PRIo32",", 1662 1.3.2.2 wrstuden wi->wi_ino, wi->wi_mode); 1663 1.3.2.2 wrstuden if ((++cnt % 4) == 0) { 1664 1.3.2.2 wrstuden (*pr)("\n\t"); 1665 1.3.2.2 wrstuden } 1666 1.3.2.2 wrstuden } 1667 1.3.2.2 wrstuden } 1668 1.3.2.2 wrstuden (*pr)("\n"); 1669 1.3.2.2 wrstuden } 1670 1.3.2.2 wrstuden } 1671 1.3.2.2 wrstuden } 1672 1.3.2.2 wrstuden 1673 1.3.2.2 wrstuden #if defined(WAPBL_DEBUG) || defined(DDB) 1674 1.3.2.2 wrstuden void 1675 1.3.2.2 wrstuden wapbl_dump(struct wapbl *wl) 1676 1.3.2.2 wrstuden { 1677 1.3.2.2 wrstuden #if defined(WAPBL_DEBUG) 1678 1.3.2.2 wrstuden if (!wl) 1679 1.3.2.2 wrstuden wl = wapbl_debug_wl; 1680 1.3.2.2 wrstuden #endif 1681 1.3.2.2 wrstuden if (!wl) 1682 1.3.2.2 wrstuden return; 1683 1.3.2.2 wrstuden wapbl_print(wl, 1, printf); 1684 1.3.2.2 wrstuden } 1685 1.3.2.2 wrstuden #endif 1686 1.3.2.2 wrstuden 1687 1.3.2.2 wrstuden /****************************************************************/ 1688 1.3.2.2 wrstuden 1689 1.3.2.2 wrstuden void 1690 1.3.2.2 wrstuden wapbl_register_deallocation(struct wapbl *wl, daddr_t blk, int len) 1691 1.3.2.2 wrstuden { 1692 1.3.2.2 wrstuden 1693 1.3.2.2 wrstuden wapbl_jlock_assert(wl); 1694 1.3.2.2 wrstuden 1695 1.3.2.2 wrstuden /* XXX should eventually instead tie this into resource estimation */ 1696 1.3.2.2 wrstuden /* XXX this KASSERT needs locking/mutex analysis */ 1697 1.3.2.2 wrstuden KASSERT(wl->wl_dealloccnt < wl->wl_dealloclim); 1698 1.3.2.2 wrstuden wl->wl_deallocblks[wl->wl_dealloccnt] = blk; 1699 1.3.2.2 wrstuden wl->wl_dealloclens[wl->wl_dealloccnt] = len; 1700 1.3.2.2 wrstuden wl->wl_dealloccnt++; 1701 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_ALLOC, 1702 1.3.2.2 wrstuden ("wapbl_register_deallocation: blk=%"PRId64" len=%d\n", blk, len)); 1703 1.3.2.2 wrstuden } 1704 1.3.2.2 wrstuden 1705 1.3.2.2 wrstuden /****************************************************************/ 1706 1.3.2.2 wrstuden 1707 1.3.2.2 wrstuden static void 1708 1.3.2.2 wrstuden wapbl_inodetrk_init(struct wapbl *wl, u_int size) 1709 1.3.2.2 wrstuden { 1710 1.3.2.2 wrstuden 1711 1.3.2.2 wrstuden wl->wl_inohash = hashinit(size, HASH_LIST, true, &wl->wl_inohashmask); 1712 1.3.2.2 wrstuden if (atomic_inc_uint_nv(&wapbl_ino_pool_refcount) == 1) { 1713 1.3.2.2 wrstuden pool_init(&wapbl_ino_pool, sizeof(struct wapbl_ino), 0, 0, 0, 1714 1.3.2.2 wrstuden "wapblinopl", &pool_allocator_nointr, IPL_NONE); 1715 1.3.2.2 wrstuden } 1716 1.3.2.2 wrstuden } 1717 1.3.2.2 wrstuden 1718 1.3.2.2 wrstuden static void 1719 1.3.2.2 wrstuden wapbl_inodetrk_free(struct wapbl *wl) 1720 1.3.2.2 wrstuden { 1721 1.3.2.2 wrstuden 1722 1.3.2.2 wrstuden /* XXX this KASSERT needs locking/mutex analysis */ 1723 1.3.2.2 wrstuden KASSERT(wl->wl_inohashcnt == 0); 1724 1.3.2.2 wrstuden hashdone(wl->wl_inohash, HASH_LIST, wl->wl_inohashmask); 1725 1.3.2.2 wrstuden if (atomic_dec_uint_nv(&wapbl_ino_pool_refcount) == 0) { 1726 1.3.2.2 wrstuden pool_destroy(&wapbl_ino_pool); 1727 1.3.2.2 wrstuden } 1728 1.3.2.2 wrstuden } 1729 1.3.2.2 wrstuden 1730 1.3.2.2 wrstuden static struct wapbl_ino * 1731 1.3.2.2 wrstuden wapbl_inodetrk_get(struct wapbl *wl, ino_t ino) 1732 1.3.2.2 wrstuden { 1733 1.3.2.2 wrstuden struct wapbl_ino_head *wih; 1734 1.3.2.2 wrstuden struct wapbl_ino *wi; 1735 1.3.2.2 wrstuden 1736 1.3.2.2 wrstuden KASSERT(mutex_owned(&wl->wl_mtx)); 1737 1.3.2.2 wrstuden 1738 1.3.2.2 wrstuden wih = &wl->wl_inohash[ino & wl->wl_inohashmask]; 1739 1.3.2.2 wrstuden LIST_FOREACH(wi, wih, wi_hash) { 1740 1.3.2.2 wrstuden if (ino == wi->wi_ino) 1741 1.3.2.2 wrstuden return wi; 1742 1.3.2.2 wrstuden } 1743 1.3.2.2 wrstuden return 0; 1744 1.3.2.2 wrstuden } 1745 1.3.2.2 wrstuden 1746 1.3.2.2 wrstuden void 1747 1.3.2.2 wrstuden wapbl_register_inode(struct wapbl *wl, ino_t ino, mode_t mode) 1748 1.3.2.2 wrstuden { 1749 1.3.2.2 wrstuden struct wapbl_ino_head *wih; 1750 1.3.2.2 wrstuden struct wapbl_ino *wi; 1751 1.3.2.2 wrstuden 1752 1.3.2.2 wrstuden wi = pool_get(&wapbl_ino_pool, PR_WAITOK); 1753 1.3.2.2 wrstuden 1754 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 1755 1.3.2.2 wrstuden if (wapbl_inodetrk_get(wl, ino) == NULL) { 1756 1.3.2.2 wrstuden wi->wi_ino = ino; 1757 1.3.2.2 wrstuden wi->wi_mode = mode; 1758 1.3.2.2 wrstuden wih = &wl->wl_inohash[ino & wl->wl_inohashmask]; 1759 1.3.2.2 wrstuden LIST_INSERT_HEAD(wih, wi, wi_hash); 1760 1.3.2.2 wrstuden wl->wl_inohashcnt++; 1761 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_INODE, 1762 1.3.2.2 wrstuden ("wapbl_register_inode: ino=%"PRId64"\n", ino)); 1763 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1764 1.3.2.2 wrstuden } else { 1765 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1766 1.3.2.2 wrstuden pool_put(&wapbl_ino_pool, wi); 1767 1.3.2.2 wrstuden } 1768 1.3.2.2 wrstuden } 1769 1.3.2.2 wrstuden 1770 1.3.2.2 wrstuden void 1771 1.3.2.2 wrstuden wapbl_unregister_inode(struct wapbl *wl, ino_t ino, mode_t mode) 1772 1.3.2.2 wrstuden { 1773 1.3.2.2 wrstuden struct wapbl_ino *wi; 1774 1.3.2.2 wrstuden 1775 1.3.2.2 wrstuden mutex_enter(&wl->wl_mtx); 1776 1.3.2.2 wrstuden wi = wapbl_inodetrk_get(wl, ino); 1777 1.3.2.2 wrstuden if (wi) { 1778 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_INODE, 1779 1.3.2.2 wrstuden ("wapbl_unregister_inode: ino=%"PRId64"\n", ino)); 1780 1.3.2.2 wrstuden KASSERT(wl->wl_inohashcnt > 0); 1781 1.3.2.2 wrstuden wl->wl_inohashcnt--; 1782 1.3.2.2 wrstuden LIST_REMOVE(wi, wi_hash); 1783 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1784 1.3.2.2 wrstuden 1785 1.3.2.2 wrstuden pool_put(&wapbl_ino_pool, wi); 1786 1.3.2.2 wrstuden } else { 1787 1.3.2.2 wrstuden mutex_exit(&wl->wl_mtx); 1788 1.3.2.2 wrstuden } 1789 1.3.2.2 wrstuden } 1790 1.3.2.2 wrstuden 1791 1.3.2.2 wrstuden /****************************************************************/ 1792 1.3.2.2 wrstuden 1793 1.3.2.2 wrstuden static __inline size_t 1794 1.3.2.2 wrstuden wapbl_transaction_inodes_len(struct wapbl *wl) 1795 1.3.2.2 wrstuden { 1796 1.3.2.2 wrstuden int blocklen = 1<<wl->wl_log_dev_bshift; 1797 1.3.2.2 wrstuden int iph; 1798 1.3.2.2 wrstuden 1799 1.3.2.2 wrstuden /* Calculate number of inodes described in a inodelist header */ 1800 1.3.2.2 wrstuden iph = (blocklen - offsetof(struct wapbl_wc_inodelist, wc_inodes)) / 1801 1.3.2.2 wrstuden sizeof(((struct wapbl_wc_inodelist *)0)->wc_inodes[0]); 1802 1.3.2.2 wrstuden 1803 1.3.2.2 wrstuden KASSERT(iph > 0); 1804 1.3.2.2 wrstuden 1805 1.3.2.2 wrstuden return MAX(1, howmany(wl->wl_inohashcnt, iph))*blocklen; 1806 1.3.2.2 wrstuden } 1807 1.3.2.2 wrstuden 1808 1.3.2.2 wrstuden 1809 1.3.2.2 wrstuden /* Calculate amount of space a transaction will take on disk */ 1810 1.3.2.2 wrstuden static size_t 1811 1.3.2.2 wrstuden wapbl_transaction_len(struct wapbl *wl) 1812 1.3.2.2 wrstuden { 1813 1.3.2.2 wrstuden int blocklen = 1<<wl->wl_log_dev_bshift; 1814 1.3.2.2 wrstuden size_t len; 1815 1.3.2.2 wrstuden int bph; 1816 1.3.2.2 wrstuden 1817 1.3.2.2 wrstuden /* Calculate number of blocks described in a blocklist header */ 1818 1.3.2.2 wrstuden bph = (blocklen - offsetof(struct wapbl_wc_blocklist, wc_blocks)) / 1819 1.3.2.2 wrstuden sizeof(((struct wapbl_wc_blocklist *)0)->wc_blocks[0]); 1820 1.3.2.2 wrstuden 1821 1.3.2.2 wrstuden KASSERT(bph > 0); 1822 1.3.2.2 wrstuden 1823 1.3.2.2 wrstuden len = wl->wl_bcount; 1824 1.3.2.2 wrstuden len += howmany(wl->wl_bufcount, bph)*blocklen; 1825 1.3.2.2 wrstuden len += howmany(wl->wl_dealloccnt, bph)*blocklen; 1826 1.3.2.2 wrstuden len += wapbl_transaction_inodes_len(wl); 1827 1.3.2.2 wrstuden 1828 1.3.2.2 wrstuden return len; 1829 1.3.2.2 wrstuden } 1830 1.3.2.2 wrstuden 1831 1.3.2.2 wrstuden /* 1832 1.3.2.2 wrstuden * Perform commit operation 1833 1.3.2.2 wrstuden * 1834 1.3.2.2 wrstuden * Note that generation number incrementation needs to 1835 1.3.2.2 wrstuden * be protected against racing with other invocations 1836 1.3.2.2 wrstuden * of wapbl_commit. This is ok since this routine 1837 1.3.2.2 wrstuden * is only invoked from wapbl_flush 1838 1.3.2.2 wrstuden */ 1839 1.3.2.2 wrstuden static int 1840 1.3.2.2 wrstuden wapbl_write_commit(struct wapbl *wl, off_t head, off_t tail) 1841 1.3.2.2 wrstuden { 1842 1.3.2.2 wrstuden struct wapbl_wc_header *wc = wl->wl_wc_header; 1843 1.3.2.2 wrstuden struct timespec ts; 1844 1.3.2.2 wrstuden int error; 1845 1.3.2.2 wrstuden int force = 1; 1846 1.3.2.2 wrstuden 1847 1.3.2.2 wrstuden /* XXX Calc checksum here, instead we do this for now */ 1848 1.3.2.2 wrstuden error = VOP_IOCTL(wl->wl_devvp, DIOCCACHESYNC, &force, FWRITE, FSCRED); 1849 1.3.2.2 wrstuden if (error) { 1850 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_ERROR, 1851 1.3.2.2 wrstuden ("wapbl_write_commit: DIOCCACHESYNC on dev 0x%x " 1852 1.3.2.2 wrstuden "returned %d\n", wl->wl_devvp->v_rdev, error)); 1853 1.3.2.2 wrstuden } 1854 1.3.2.2 wrstuden 1855 1.3.2.2 wrstuden wc->wc_head = head; 1856 1.3.2.2 wrstuden wc->wc_tail = tail; 1857 1.3.2.2 wrstuden wc->wc_checksum = 0; 1858 1.3.2.2 wrstuden wc->wc_version = 1; 1859 1.3.2.2 wrstuden getnanotime(&ts); 1860 1.3.2.2 wrstuden wc->wc_time = ts.tv_sec;; 1861 1.3.2.2 wrstuden wc->wc_timensec = ts.tv_nsec; 1862 1.3.2.2 wrstuden 1863 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_WRITE, 1864 1.3.2.2 wrstuden ("wapbl_write_commit: head = %"PRIdMAX "tail = %"PRIdMAX"\n", 1865 1.3.2.2 wrstuden (intmax_t)head, (intmax_t)tail)); 1866 1.3.2.2 wrstuden 1867 1.3.2.2 wrstuden /* 1868 1.3.2.2 wrstuden * XXX if generation will rollover, then first zero 1869 1.3.2.2 wrstuden * over second commit header before trying to write both headers. 1870 1.3.2.2 wrstuden */ 1871 1.3.2.2 wrstuden 1872 1.3.2.2 wrstuden error = wapbl_write(wc, wc->wc_len, wl->wl_devvp, 1873 1.3.2.2 wrstuden wl->wl_logpbn + wc->wc_generation % 2); 1874 1.3.2.2 wrstuden if (error) 1875 1.3.2.2 wrstuden return error; 1876 1.3.2.2 wrstuden 1877 1.3.2.2 wrstuden error = VOP_IOCTL(wl->wl_devvp, DIOCCACHESYNC, &force, FWRITE, FSCRED); 1878 1.3.2.2 wrstuden if (error) { 1879 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_ERROR, 1880 1.3.2.2 wrstuden ("wapbl_write_commit: DIOCCACHESYNC on dev 0x%x " 1881 1.3.2.2 wrstuden "returned %d\n", wl->wl_devvp->v_rdev, error)); 1882 1.3.2.2 wrstuden } 1883 1.3.2.2 wrstuden 1884 1.3.2.2 wrstuden /* 1885 1.3.2.2 wrstuden * If the generation number was zero, write it out a second time. 1886 1.3.2.2 wrstuden * This handles initialization and generation number rollover 1887 1.3.2.2 wrstuden */ 1888 1.3.2.2 wrstuden if (wc->wc_generation++ == 0) { 1889 1.3.2.2 wrstuden error = wapbl_write_commit(wl, head, tail); 1890 1.3.2.2 wrstuden /* 1891 1.3.2.2 wrstuden * This panic should be able to be removed if we do the 1892 1.3.2.2 wrstuden * zero'ing mentioned above, and we are certain to roll 1893 1.3.2.2 wrstuden * back generation number on failure. 1894 1.3.2.2 wrstuden */ 1895 1.3.2.2 wrstuden if (error) 1896 1.3.2.2 wrstuden panic("wapbl_write_commit: error writing duplicate " 1897 1.3.2.2 wrstuden "log header: %d\n", error); 1898 1.3.2.2 wrstuden } 1899 1.3.2.2 wrstuden return 0; 1900 1.3.2.2 wrstuden } 1901 1.3.2.2 wrstuden 1902 1.3.2.2 wrstuden /* Returns new offset value */ 1903 1.3.2.2 wrstuden static int 1904 1.3.2.2 wrstuden wapbl_write_blocks(struct wapbl *wl, off_t *offp) 1905 1.3.2.2 wrstuden { 1906 1.3.2.2 wrstuden struct wapbl_wc_blocklist *wc = 1907 1.3.2.2 wrstuden (struct wapbl_wc_blocklist *)wl->wl_wc_scratch; 1908 1.3.2.2 wrstuden int blocklen = 1<<wl->wl_log_dev_bshift; 1909 1.3.2.2 wrstuden int bph; 1910 1.3.2.2 wrstuden struct buf *bp; 1911 1.3.2.2 wrstuden off_t off = *offp; 1912 1.3.2.2 wrstuden int error; 1913 1.3.2.2 wrstuden 1914 1.3.2.2 wrstuden KASSERT(rw_write_held(&wl->wl_rwlock)); 1915 1.3.2.2 wrstuden 1916 1.3.2.2 wrstuden bph = (blocklen - offsetof(struct wapbl_wc_blocklist, wc_blocks)) / 1917 1.3.2.2 wrstuden sizeof(((struct wapbl_wc_blocklist *)0)->wc_blocks[0]); 1918 1.3.2.2 wrstuden 1919 1.3.2.2 wrstuden bp = LIST_FIRST(&wl->wl_bufs); 1920 1.3.2.2 wrstuden 1921 1.3.2.2 wrstuden while (bp) { 1922 1.3.2.2 wrstuden int cnt; 1923 1.3.2.2 wrstuden struct buf *obp = bp; 1924 1.3.2.2 wrstuden 1925 1.3.2.2 wrstuden KASSERT(bp->b_flags & B_LOCKED); 1926 1.3.2.2 wrstuden 1927 1.3.2.2 wrstuden wc->wc_type = WAPBL_WC_BLOCKS; 1928 1.3.2.2 wrstuden wc->wc_len = blocklen; 1929 1.3.2.2 wrstuden wc->wc_blkcount = 0; 1930 1.3.2.2 wrstuden while (bp && (wc->wc_blkcount < bph)) { 1931 1.3.2.2 wrstuden /* 1932 1.3.2.2 wrstuden * Make sure all the physical block numbers are up to 1933 1.3.2.2 wrstuden * date. If this is not always true on a given 1934 1.3.2.2 wrstuden * filesystem, then VOP_BMAP must be called. We 1935 1.3.2.2 wrstuden * could call VOP_BMAP here, or else in the filesystem 1936 1.3.2.2 wrstuden * specific flush callback, although neither of those 1937 1.3.2.2 wrstuden * solutions allow us to take the vnode lock. If a 1938 1.3.2.2 wrstuden * filesystem requires that we must take the vnode lock 1939 1.3.2.2 wrstuden * to call VOP_BMAP, then we can probably do it in 1940 1.3.2.2 wrstuden * bwrite when the vnode lock should already be held 1941 1.3.2.2 wrstuden * by the invoking code. 1942 1.3.2.2 wrstuden */ 1943 1.3.2.2 wrstuden KASSERT((bp->b_vp->v_type == VBLK) || 1944 1.3.2.2 wrstuden (bp->b_blkno != bp->b_lblkno)); 1945 1.3.2.2 wrstuden KASSERT(bp->b_blkno > 0); 1946 1.3.2.2 wrstuden 1947 1.3.2.2 wrstuden wc->wc_blocks[wc->wc_blkcount].wc_daddr = bp->b_blkno; 1948 1.3.2.2 wrstuden wc->wc_blocks[wc->wc_blkcount].wc_dlen = bp->b_bcount; 1949 1.3.2.2 wrstuden wc->wc_len += bp->b_bcount; 1950 1.3.2.2 wrstuden wc->wc_blkcount++; 1951 1.3.2.2 wrstuden bp = LIST_NEXT(bp, b_wapbllist); 1952 1.3.2.2 wrstuden } 1953 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_WRITE, 1954 1.3.2.2 wrstuden ("wapbl_write_blocks: len = %u off = %"PRIdMAX"\n", 1955 1.3.2.2 wrstuden wc->wc_len, (intmax_t)off)); 1956 1.3.2.2 wrstuden 1957 1.3.2.2 wrstuden error = wapbl_circ_write(wl, wc, blocklen, &off); 1958 1.3.2.2 wrstuden if (error) 1959 1.3.2.2 wrstuden return error; 1960 1.3.2.2 wrstuden bp = obp; 1961 1.3.2.2 wrstuden cnt = 0; 1962 1.3.2.2 wrstuden while (bp && (cnt++ < bph)) { 1963 1.3.2.2 wrstuden error = wapbl_circ_write(wl, bp->b_data, 1964 1.3.2.2 wrstuden bp->b_bcount, &off); 1965 1.3.2.2 wrstuden if (error) 1966 1.3.2.2 wrstuden return error; 1967 1.3.2.2 wrstuden bp = LIST_NEXT(bp, b_wapbllist); 1968 1.3.2.2 wrstuden } 1969 1.3.2.2 wrstuden } 1970 1.3.2.2 wrstuden *offp = off; 1971 1.3.2.2 wrstuden return 0; 1972 1.3.2.2 wrstuden } 1973 1.3.2.2 wrstuden 1974 1.3.2.2 wrstuden static int 1975 1.3.2.2 wrstuden wapbl_write_revocations(struct wapbl *wl, off_t *offp) 1976 1.3.2.2 wrstuden { 1977 1.3.2.2 wrstuden struct wapbl_wc_blocklist *wc = 1978 1.3.2.2 wrstuden (struct wapbl_wc_blocklist *)wl->wl_wc_scratch; 1979 1.3.2.2 wrstuden int i; 1980 1.3.2.2 wrstuden int blocklen = 1<<wl->wl_log_dev_bshift; 1981 1.3.2.2 wrstuden int bph; 1982 1.3.2.2 wrstuden off_t off = *offp; 1983 1.3.2.2 wrstuden int error; 1984 1.3.2.2 wrstuden 1985 1.3.2.2 wrstuden if (wl->wl_dealloccnt == 0) 1986 1.3.2.2 wrstuden return 0; 1987 1.3.2.2 wrstuden 1988 1.3.2.2 wrstuden bph = (blocklen - offsetof(struct wapbl_wc_blocklist, wc_blocks)) / 1989 1.3.2.2 wrstuden sizeof(((struct wapbl_wc_blocklist *)0)->wc_blocks[0]); 1990 1.3.2.2 wrstuden 1991 1.3.2.2 wrstuden i = 0; 1992 1.3.2.2 wrstuden while (i < wl->wl_dealloccnt) { 1993 1.3.2.2 wrstuden wc->wc_type = WAPBL_WC_REVOCATIONS; 1994 1.3.2.2 wrstuden wc->wc_len = blocklen; 1995 1.3.2.2 wrstuden wc->wc_blkcount = 0; 1996 1.3.2.2 wrstuden while ((i < wl->wl_dealloccnt) && (wc->wc_blkcount < bph)) { 1997 1.3.2.2 wrstuden wc->wc_blocks[wc->wc_blkcount].wc_daddr = 1998 1.3.2.2 wrstuden wl->wl_deallocblks[i]; 1999 1.3.2.2 wrstuden wc->wc_blocks[wc->wc_blkcount].wc_dlen = 2000 1.3.2.2 wrstuden wl->wl_dealloclens[i]; 2001 1.3.2.2 wrstuden wc->wc_blkcount++; 2002 1.3.2.2 wrstuden i++; 2003 1.3.2.2 wrstuden } 2004 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_WRITE, 2005 1.3.2.2 wrstuden ("wapbl_write_revocations: len = %u off = %"PRIdMAX"\n", 2006 1.3.2.2 wrstuden wc->wc_len, (intmax_t)off)); 2007 1.3.2.2 wrstuden error = wapbl_circ_write(wl, wc, blocklen, &off); 2008 1.3.2.2 wrstuden if (error) 2009 1.3.2.2 wrstuden return error; 2010 1.3.2.2 wrstuden } 2011 1.3.2.2 wrstuden *offp = off; 2012 1.3.2.2 wrstuden return 0; 2013 1.3.2.2 wrstuden } 2014 1.3.2.2 wrstuden 2015 1.3.2.2 wrstuden static int 2016 1.3.2.2 wrstuden wapbl_write_inodes(struct wapbl *wl, off_t *offp) 2017 1.3.2.2 wrstuden { 2018 1.3.2.2 wrstuden struct wapbl_wc_inodelist *wc = 2019 1.3.2.2 wrstuden (struct wapbl_wc_inodelist *)wl->wl_wc_scratch; 2020 1.3.2.2 wrstuden int i; 2021 1.3.2.2 wrstuden int blocklen = 1<<wl->wl_log_dev_bshift; 2022 1.3.2.2 wrstuden off_t off = *offp; 2023 1.3.2.2 wrstuden int error; 2024 1.3.2.2 wrstuden 2025 1.3.2.2 wrstuden struct wapbl_ino_head *wih; 2026 1.3.2.2 wrstuden struct wapbl_ino *wi; 2027 1.3.2.2 wrstuden int iph; 2028 1.3.2.2 wrstuden 2029 1.3.2.2 wrstuden iph = (blocklen - offsetof(struct wapbl_wc_inodelist, wc_inodes)) / 2030 1.3.2.2 wrstuden sizeof(((struct wapbl_wc_inodelist *)0)->wc_inodes[0]); 2031 1.3.2.2 wrstuden 2032 1.3.2.2 wrstuden i = 0; 2033 1.3.2.2 wrstuden wih = &wl->wl_inohash[0]; 2034 1.3.2.2 wrstuden wi = 0; 2035 1.3.2.2 wrstuden do { 2036 1.3.2.2 wrstuden wc->wc_type = WAPBL_WC_INODES; 2037 1.3.2.2 wrstuden wc->wc_len = blocklen; 2038 1.3.2.2 wrstuden wc->wc_inocnt = 0; 2039 1.3.2.2 wrstuden wc->wc_clear = (i == 0); 2040 1.3.2.2 wrstuden while ((i < wl->wl_inohashcnt) && (wc->wc_inocnt < iph)) { 2041 1.3.2.2 wrstuden while (!wi) { 2042 1.3.2.2 wrstuden KASSERT((wih - &wl->wl_inohash[0]) 2043 1.3.2.2 wrstuden <= wl->wl_inohashmask); 2044 1.3.2.2 wrstuden wi = LIST_FIRST(wih++); 2045 1.3.2.2 wrstuden } 2046 1.3.2.2 wrstuden wc->wc_inodes[wc->wc_inocnt].wc_inumber = wi->wi_ino; 2047 1.3.2.2 wrstuden wc->wc_inodes[wc->wc_inocnt].wc_imode = wi->wi_mode; 2048 1.3.2.2 wrstuden wc->wc_inocnt++; 2049 1.3.2.2 wrstuden i++; 2050 1.3.2.2 wrstuden wi = LIST_NEXT(wi, wi_hash); 2051 1.3.2.2 wrstuden } 2052 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_WRITE, 2053 1.3.2.2 wrstuden ("wapbl_write_inodes: len = %u off = %"PRIdMAX"\n", 2054 1.3.2.2 wrstuden wc->wc_len, (intmax_t)off)); 2055 1.3.2.2 wrstuden error = wapbl_circ_write(wl, wc, blocklen, &off); 2056 1.3.2.2 wrstuden if (error) 2057 1.3.2.2 wrstuden return error; 2058 1.3.2.2 wrstuden } while (i < wl->wl_inohashcnt); 2059 1.3.2.2 wrstuden 2060 1.3.2.2 wrstuden *offp = off; 2061 1.3.2.2 wrstuden return 0; 2062 1.3.2.2 wrstuden } 2063 1.3.2.2 wrstuden 2064 1.3.2.2 wrstuden #endif /* _KERNEL */ 2065 1.3.2.2 wrstuden 2066 1.3.2.2 wrstuden /****************************************************************/ 2067 1.3.2.2 wrstuden 2068 1.3.2.2 wrstuden #ifdef _KERNEL 2069 1.3.2.2 wrstuden static struct pool wapbl_blk_pool; 2070 1.3.2.2 wrstuden static int wapbl_blk_pool_refcount; 2071 1.3.2.2 wrstuden #endif 2072 1.3.2.2 wrstuden struct wapbl_blk { 2073 1.3.2.2 wrstuden LIST_ENTRY(wapbl_blk) wb_hash; 2074 1.3.2.2 wrstuden daddr_t wb_blk; 2075 1.3.2.2 wrstuden off_t wb_off; /* Offset of this block in the log */ 2076 1.3.2.2 wrstuden }; 2077 1.3.2.2 wrstuden #define WAPBL_BLKPOOL_MIN 83 2078 1.3.2.2 wrstuden 2079 1.3.2.2 wrstuden static void 2080 1.3.2.2 wrstuden wapbl_blkhash_init(struct wapbl_replay *wr, u_int size) 2081 1.3.2.2 wrstuden { 2082 1.3.2.2 wrstuden if (size < WAPBL_BLKPOOL_MIN) 2083 1.3.2.2 wrstuden size = WAPBL_BLKPOOL_MIN; 2084 1.3.2.2 wrstuden KASSERT(wr->wr_blkhash == 0); 2085 1.3.2.2 wrstuden #ifdef _KERNEL 2086 1.3.2.2 wrstuden wr->wr_blkhash = hashinit(size, HASH_LIST, true, &wr->wr_blkhashmask); 2087 1.3.2.2 wrstuden if (atomic_inc_uint_nv(&wapbl_blk_pool_refcount) == 1) { 2088 1.3.2.2 wrstuden pool_init(&wapbl_blk_pool, sizeof(struct wapbl_blk), 0, 0, 0, 2089 1.3.2.2 wrstuden "wapblblkpl", &pool_allocator_nointr, IPL_NONE); 2090 1.3.2.2 wrstuden } 2091 1.3.2.2 wrstuden #else /* ! _KERNEL */ 2092 1.3.2.2 wrstuden /* Manually implement hashinit */ 2093 1.3.2.2 wrstuden { 2094 1.3.2.2 wrstuden int i; 2095 1.3.2.2 wrstuden unsigned long hashsize; 2096 1.3.2.2 wrstuden for (hashsize = 1; hashsize < size; hashsize <<= 1) 2097 1.3.2.2 wrstuden continue; 2098 1.3.2.2 wrstuden wr->wr_blkhash = wapbl_malloc(hashsize * sizeof(*wr->wr_blkhash)); 2099 1.3.2.2 wrstuden for (i = 0; i < wr->wr_blkhashmask; i++) 2100 1.3.2.2 wrstuden LIST_INIT(&wr->wr_blkhash[i]); 2101 1.3.2.2 wrstuden wr->wr_blkhashmask = hashsize - 1; 2102 1.3.2.2 wrstuden } 2103 1.3.2.2 wrstuden #endif /* ! _KERNEL */ 2104 1.3.2.2 wrstuden } 2105 1.3.2.2 wrstuden 2106 1.3.2.2 wrstuden static void 2107 1.3.2.2 wrstuden wapbl_blkhash_free(struct wapbl_replay *wr) 2108 1.3.2.2 wrstuden { 2109 1.3.2.2 wrstuden KASSERT(wr->wr_blkhashcnt == 0); 2110 1.3.2.2 wrstuden #ifdef _KERNEL 2111 1.3.2.2 wrstuden hashdone(wr->wr_blkhash, HASH_LIST, wr->wr_blkhashmask); 2112 1.3.2.2 wrstuden if (atomic_dec_uint_nv(&wapbl_blk_pool_refcount) == 0) { 2113 1.3.2.2 wrstuden pool_destroy(&wapbl_blk_pool); 2114 1.3.2.2 wrstuden } 2115 1.3.2.2 wrstuden #else /* ! _KERNEL */ 2116 1.3.2.2 wrstuden wapbl_free(wr->wr_blkhash); 2117 1.3.2.2 wrstuden #endif /* ! _KERNEL */ 2118 1.3.2.2 wrstuden } 2119 1.3.2.2 wrstuden 2120 1.3.2.2 wrstuden static struct wapbl_blk * 2121 1.3.2.2 wrstuden wapbl_blkhash_get(struct wapbl_replay *wr, daddr_t blk) 2122 1.3.2.2 wrstuden { 2123 1.3.2.2 wrstuden struct wapbl_blk_head *wbh; 2124 1.3.2.2 wrstuden struct wapbl_blk *wb; 2125 1.3.2.2 wrstuden wbh = &wr->wr_blkhash[blk & wr->wr_blkhashmask]; 2126 1.3.2.2 wrstuden LIST_FOREACH(wb, wbh, wb_hash) { 2127 1.3.2.2 wrstuden if (blk == wb->wb_blk) 2128 1.3.2.2 wrstuden return wb; 2129 1.3.2.2 wrstuden } 2130 1.3.2.2 wrstuden return 0; 2131 1.3.2.2 wrstuden } 2132 1.3.2.2 wrstuden 2133 1.3.2.2 wrstuden static void 2134 1.3.2.2 wrstuden wapbl_blkhash_ins(struct wapbl_replay *wr, daddr_t blk, off_t off) 2135 1.3.2.2 wrstuden { 2136 1.3.2.2 wrstuden struct wapbl_blk_head *wbh; 2137 1.3.2.2 wrstuden struct wapbl_blk *wb; 2138 1.3.2.2 wrstuden wb = wapbl_blkhash_get(wr, blk); 2139 1.3.2.2 wrstuden if (wb) { 2140 1.3.2.2 wrstuden KASSERT(wb->wb_blk == blk); 2141 1.3.2.2 wrstuden wb->wb_off = off; 2142 1.3.2.2 wrstuden } else { 2143 1.3.2.2 wrstuden #ifdef _KERNEL 2144 1.3.2.2 wrstuden wb = pool_get(&wapbl_blk_pool, PR_WAITOK); 2145 1.3.2.2 wrstuden #else /* ! _KERNEL */ 2146 1.3.2.2 wrstuden wb = wapbl_malloc(sizeof(*wb)); 2147 1.3.2.2 wrstuden #endif /* ! _KERNEL */ 2148 1.3.2.2 wrstuden wb->wb_blk = blk; 2149 1.3.2.2 wrstuden wb->wb_off = off; 2150 1.3.2.2 wrstuden wbh = &wr->wr_blkhash[blk & wr->wr_blkhashmask]; 2151 1.3.2.2 wrstuden LIST_INSERT_HEAD(wbh, wb, wb_hash); 2152 1.3.2.2 wrstuden wr->wr_blkhashcnt++; 2153 1.3.2.2 wrstuden } 2154 1.3.2.2 wrstuden } 2155 1.3.2.2 wrstuden 2156 1.3.2.2 wrstuden static void 2157 1.3.2.2 wrstuden wapbl_blkhash_rem(struct wapbl_replay *wr, daddr_t blk) 2158 1.3.2.2 wrstuden { 2159 1.3.2.2 wrstuden struct wapbl_blk *wb = wapbl_blkhash_get(wr, blk); 2160 1.3.2.2 wrstuden if (wb) { 2161 1.3.2.2 wrstuden KASSERT(wr->wr_blkhashcnt > 0); 2162 1.3.2.2 wrstuden wr->wr_blkhashcnt--; 2163 1.3.2.2 wrstuden LIST_REMOVE(wb, wb_hash); 2164 1.3.2.2 wrstuden #ifdef _KERNEL 2165 1.3.2.2 wrstuden pool_put(&wapbl_blk_pool, wb); 2166 1.3.2.2 wrstuden #else /* ! _KERNEL */ 2167 1.3.2.2 wrstuden wapbl_free(wb); 2168 1.3.2.2 wrstuden #endif /* ! _KERNEL */ 2169 1.3.2.2 wrstuden } 2170 1.3.2.2 wrstuden } 2171 1.3.2.2 wrstuden 2172 1.3.2.2 wrstuden static void 2173 1.3.2.2 wrstuden wapbl_blkhash_clear(struct wapbl_replay *wr) 2174 1.3.2.2 wrstuden { 2175 1.3.2.2 wrstuden int i; 2176 1.3.2.2 wrstuden for (i = 0; i <= wr->wr_blkhashmask; i++) { 2177 1.3.2.2 wrstuden struct wapbl_blk *wb; 2178 1.3.2.2 wrstuden 2179 1.3.2.2 wrstuden while ((wb = LIST_FIRST(&wr->wr_blkhash[i]))) { 2180 1.3.2.2 wrstuden KASSERT(wr->wr_blkhashcnt > 0); 2181 1.3.2.2 wrstuden wr->wr_blkhashcnt--; 2182 1.3.2.2 wrstuden LIST_REMOVE(wb, wb_hash); 2183 1.3.2.2 wrstuden #ifdef _KERNEL 2184 1.3.2.2 wrstuden pool_put(&wapbl_blk_pool, wb); 2185 1.3.2.2 wrstuden #else /* ! _KERNEL */ 2186 1.3.2.2 wrstuden wapbl_free(wb); 2187 1.3.2.2 wrstuden #endif /* ! _KERNEL */ 2188 1.3.2.2 wrstuden } 2189 1.3.2.2 wrstuden } 2190 1.3.2.2 wrstuden KASSERT(wr->wr_blkhashcnt == 0); 2191 1.3.2.2 wrstuden } 2192 1.3.2.2 wrstuden 2193 1.3.2.2 wrstuden /****************************************************************/ 2194 1.3.2.2 wrstuden 2195 1.3.2.2 wrstuden static int 2196 1.3.2.2 wrstuden wapbl_circ_read(struct wapbl_replay *wr, void *data, size_t len, off_t *offp) 2197 1.3.2.2 wrstuden { 2198 1.3.2.2 wrstuden size_t slen; 2199 1.3.2.2 wrstuden struct wapbl_wc_header *wc = &wr->wr_wc_header; 2200 1.3.2.2 wrstuden off_t off = *offp; 2201 1.3.2.2 wrstuden int error; 2202 1.3.2.2 wrstuden 2203 1.3.2.2 wrstuden KASSERT(((len >> wc->wc_log_dev_bshift) << 2204 1.3.2.2 wrstuden wc->wc_log_dev_bshift) == len); 2205 1.3.2.2 wrstuden if (off < wc->wc_circ_off) 2206 1.3.2.2 wrstuden off = wc->wc_circ_off; 2207 1.3.2.2 wrstuden slen = wc->wc_circ_off + wc->wc_circ_size - off; 2208 1.3.2.2 wrstuden if (slen < len) { 2209 1.3.2.2 wrstuden error = wapbl_read(data, slen, wr->wr_devvp, 2210 1.3.2.2 wrstuden wr->wr_logpbn + (off >> wc->wc_log_dev_bshift)); 2211 1.3.2.2 wrstuden if (error) 2212 1.3.2.2 wrstuden return error; 2213 1.3.2.2 wrstuden data = (uint8_t *)data + slen; 2214 1.3.2.2 wrstuden len -= slen; 2215 1.3.2.2 wrstuden off = wc->wc_circ_off; 2216 1.3.2.2 wrstuden } 2217 1.3.2.2 wrstuden error = wapbl_read(data, len, wr->wr_devvp, 2218 1.3.2.2 wrstuden wr->wr_logpbn + (off >> wc->wc_log_dev_bshift)); 2219 1.3.2.2 wrstuden if (error) 2220 1.3.2.2 wrstuden return error; 2221 1.3.2.2 wrstuden off += len; 2222 1.3.2.2 wrstuden if (off >= wc->wc_circ_off + wc->wc_circ_size) 2223 1.3.2.2 wrstuden off = wc->wc_circ_off; 2224 1.3.2.2 wrstuden *offp = off; 2225 1.3.2.2 wrstuden return 0; 2226 1.3.2.2 wrstuden } 2227 1.3.2.2 wrstuden 2228 1.3.2.2 wrstuden static void 2229 1.3.2.2 wrstuden wapbl_circ_advance(struct wapbl_replay *wr, size_t len, off_t *offp) 2230 1.3.2.2 wrstuden { 2231 1.3.2.2 wrstuden size_t slen; 2232 1.3.2.2 wrstuden struct wapbl_wc_header *wc = &wr->wr_wc_header; 2233 1.3.2.2 wrstuden off_t off = *offp; 2234 1.3.2.2 wrstuden 2235 1.3.2.2 wrstuden KASSERT(((len >> wc->wc_log_dev_bshift) << 2236 1.3.2.2 wrstuden wc->wc_log_dev_bshift) == len); 2237 1.3.2.2 wrstuden 2238 1.3.2.2 wrstuden if (off < wc->wc_circ_off) 2239 1.3.2.2 wrstuden off = wc->wc_circ_off; 2240 1.3.2.2 wrstuden slen = wc->wc_circ_off + wc->wc_circ_size - off; 2241 1.3.2.2 wrstuden if (slen < len) { 2242 1.3.2.2 wrstuden len -= slen; 2243 1.3.2.2 wrstuden off = wc->wc_circ_off; 2244 1.3.2.2 wrstuden } 2245 1.3.2.2 wrstuden off += len; 2246 1.3.2.2 wrstuden if (off >= wc->wc_circ_off + wc->wc_circ_size) 2247 1.3.2.2 wrstuden off = wc->wc_circ_off; 2248 1.3.2.2 wrstuden *offp = off; 2249 1.3.2.2 wrstuden } 2250 1.3.2.2 wrstuden 2251 1.3.2.2 wrstuden /****************************************************************/ 2252 1.3.2.2 wrstuden 2253 1.3.2.2 wrstuden int 2254 1.3.2.2 wrstuden wapbl_replay_start(struct wapbl_replay **wrp, struct vnode *vp, 2255 1.3.2.2 wrstuden daddr_t off, size_t count, size_t blksize) 2256 1.3.2.2 wrstuden { 2257 1.3.2.2 wrstuden struct wapbl_replay *wr; 2258 1.3.2.2 wrstuden int error; 2259 1.3.2.2 wrstuden struct vnode *devvp; 2260 1.3.2.2 wrstuden daddr_t logpbn; 2261 1.3.2.2 wrstuden uint8_t *scratch; 2262 1.3.2.2 wrstuden struct wapbl_wc_header *wch; 2263 1.3.2.2 wrstuden struct wapbl_wc_header *wch2; 2264 1.3.2.2 wrstuden /* Use this until we read the actual log header */ 2265 1.3.2.2 wrstuden int log_dev_bshift = DEV_BSHIFT; 2266 1.3.2.2 wrstuden size_t used; 2267 1.3.2.2 wrstuden 2268 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_REPLAY, 2269 1.3.2.2 wrstuden ("wapbl_replay_start: vp=%p off=%"PRId64 " count=%zu blksize=%zu\n", 2270 1.3.2.2 wrstuden vp, off, count, blksize)); 2271 1.3.2.2 wrstuden 2272 1.3.2.2 wrstuden if (off < 0) 2273 1.3.2.2 wrstuden return EINVAL; 2274 1.3.2.2 wrstuden 2275 1.3.2.2 wrstuden if (blksize < DEV_BSIZE) 2276 1.3.2.2 wrstuden return EINVAL; 2277 1.3.2.2 wrstuden if (blksize % DEV_BSIZE) 2278 1.3.2.2 wrstuden return EINVAL; 2279 1.3.2.2 wrstuden 2280 1.3.2.2 wrstuden #ifdef _KERNEL 2281 1.3.2.2 wrstuden #if 0 2282 1.3.2.2 wrstuden /* XXX vp->v_size isn't reliably set for VBLK devices, 2283 1.3.2.2 wrstuden * especially root. However, we might still want to verify 2284 1.3.2.2 wrstuden * that the full load is readable */ 2285 1.3.2.2 wrstuden if ((off + count) * blksize > vp->v_size) 2286 1.3.2.2 wrstuden return EINVAL; 2287 1.3.2.2 wrstuden #endif 2288 1.3.2.2 wrstuden 2289 1.3.2.2 wrstuden if ((error = VOP_BMAP(vp, off, &devvp, &logpbn, 0)) != 0) { 2290 1.3.2.2 wrstuden return error; 2291 1.3.2.2 wrstuden } 2292 1.3.2.2 wrstuden #else /* ! _KERNEL */ 2293 1.3.2.2 wrstuden devvp = vp; 2294 1.3.2.2 wrstuden logpbn = off; 2295 1.3.2.2 wrstuden #endif /* ! _KERNEL */ 2296 1.3.2.2 wrstuden 2297 1.3.2.2 wrstuden scratch = wapbl_malloc(MAXBSIZE); 2298 1.3.2.2 wrstuden 2299 1.3.2.2 wrstuden error = wapbl_read(scratch, 2<<log_dev_bshift, devvp, logpbn); 2300 1.3.2.2 wrstuden if (error) 2301 1.3.2.2 wrstuden goto errout; 2302 1.3.2.2 wrstuden 2303 1.3.2.2 wrstuden wch = (struct wapbl_wc_header *)scratch; 2304 1.3.2.2 wrstuden wch2 = 2305 1.3.2.2 wrstuden (struct wapbl_wc_header *)(scratch + (1<<log_dev_bshift)); 2306 1.3.2.2 wrstuden /* XXX verify checksums and magic numbers */ 2307 1.3.2.2 wrstuden if (wch->wc_type != WAPBL_WC_HEADER) { 2308 1.3.2.2 wrstuden printf("Unrecognized wapbl magic: 0x%08x\n", wch->wc_type); 2309 1.3.2.2 wrstuden error = EFTYPE; 2310 1.3.2.2 wrstuden goto errout; 2311 1.3.2.2 wrstuden } 2312 1.3.2.2 wrstuden 2313 1.3.2.2 wrstuden if (wch2->wc_generation > wch->wc_generation) 2314 1.3.2.2 wrstuden wch = wch2; 2315 1.3.2.2 wrstuden 2316 1.3.2.2 wrstuden wr = wapbl_calloc(1, sizeof(*wr)); 2317 1.3.2.2 wrstuden 2318 1.3.2.2 wrstuden wr->wr_logvp = vp; 2319 1.3.2.2 wrstuden wr->wr_devvp = devvp; 2320 1.3.2.2 wrstuden wr->wr_logpbn = logpbn; 2321 1.3.2.2 wrstuden 2322 1.3.2.2 wrstuden wr->wr_scratch = scratch; 2323 1.3.2.2 wrstuden 2324 1.3.2.2 wrstuden memcpy(&wr->wr_wc_header, wch, sizeof(wr->wr_wc_header)); 2325 1.3.2.2 wrstuden 2326 1.3.2.2 wrstuden used = wapbl_space_used(wch->wc_circ_size, wch->wc_head, wch->wc_tail); 2327 1.3.2.2 wrstuden 2328 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_REPLAY, 2329 1.3.2.2 wrstuden ("wapbl_replay: head=%"PRId64" tail=%"PRId64" off=%"PRId64 2330 1.3.2.2 wrstuden " len=%"PRId64" used=%zu\n", 2331 1.3.2.2 wrstuden wch->wc_head, wch->wc_tail, wch->wc_circ_off, 2332 1.3.2.2 wrstuden wch->wc_circ_size, used)); 2333 1.3.2.2 wrstuden 2334 1.3.2.2 wrstuden wapbl_blkhash_init(wr, (used >> wch->wc_fs_dev_bshift)); 2335 1.3.2.2 wrstuden error = wapbl_replay_prescan(wr); 2336 1.3.2.2 wrstuden if (error) { 2337 1.3.2.2 wrstuden wapbl_replay_stop(wr); 2338 1.3.2.2 wrstuden wapbl_replay_free(wr); 2339 1.3.2.2 wrstuden return error; 2340 1.3.2.2 wrstuden } 2341 1.3.2.2 wrstuden 2342 1.3.2.2 wrstuden error = wapbl_replay_get_inodes(wr); 2343 1.3.2.2 wrstuden if (error) { 2344 1.3.2.2 wrstuden wapbl_replay_stop(wr); 2345 1.3.2.2 wrstuden wapbl_replay_free(wr); 2346 1.3.2.2 wrstuden return error; 2347 1.3.2.2 wrstuden } 2348 1.3.2.2 wrstuden 2349 1.3.2.2 wrstuden *wrp = wr; 2350 1.3.2.2 wrstuden return 0; 2351 1.3.2.2 wrstuden 2352 1.3.2.2 wrstuden errout: 2353 1.3.2.2 wrstuden wapbl_free(scratch); 2354 1.3.2.2 wrstuden return error; 2355 1.3.2.2 wrstuden } 2356 1.3.2.2 wrstuden 2357 1.3.2.2 wrstuden void 2358 1.3.2.2 wrstuden wapbl_replay_stop(struct wapbl_replay *wr) 2359 1.3.2.2 wrstuden { 2360 1.3.2.2 wrstuden 2361 1.3.2.2 wrstuden WAPBL_PRINTF(WAPBL_PRINT_REPLAY, ("wapbl_replay_stop called\n")); 2362 1.3.2.2 wrstuden 2363 1.3.2.2 wrstuden KDASSERT(wapbl_replay_isopen(wr)); 2364 1.3.2.2 wrstuden 2365 1.3.2.2 wrstuden wapbl_free(wr->wr_scratch); 2366 1.3.2.2 wrstuden wr->wr_scratch = 0; 2367 1.3.2.2 wrstuden 2368 1.3.2.2 wrstuden wr->wr_logvp = 0; 2369 1.3.2.2 wrstuden 2370 1.3.2.2 wrstuden wapbl_blkhash_clear(wr); 2371 1.3.2.2 wrstuden wapbl_blkhash_free(wr); 2372 1.3.2.2 wrstuden } 2373 1.3.2.2 wrstuden 2374 1.3.2.2 wrstuden void 2375 1.3.2.2 wrstuden wapbl_replay_free(struct wapbl_replay *wr) 2376 1.3.2.2 wrstuden { 2377 1.3.2.2 wrstuden 2378 1.3.2.2 wrstuden KDASSERT(!wapbl_replay_isopen(wr)); 2379 1.3.2.2 wrstuden 2380 1.3.2.2 wrstuden if (wr->wr_inodes) 2381 1.3.2.2 wrstuden wapbl_free(wr->wr_inodes); 2382 1.3.2.2 wrstuden wapbl_free(wr); 2383 1.3.2.2 wrstuden } 2384 1.3.2.2 wrstuden 2385 1.3.2.2 wrstuden int 2386 1.3.2.2 wrstuden wapbl_replay_isopen1(struct wapbl_replay *wr) 2387 1.3.2.2 wrstuden { 2388 1.3.2.2 wrstuden 2389 1.3.2.2 wrstuden return wapbl_replay_isopen(wr); 2390 1.3.2.2 wrstuden } 2391 1.3.2.2 wrstuden 2392 1.3.2.2 wrstuden static int 2393 1.3.2.2 wrstuden wapbl_replay_prescan(struct wapbl_replay *wr) 2394 1.3.2.2 wrstuden { 2395 1.3.2.2 wrstuden off_t off; 2396 1.3.2.2 wrstuden struct wapbl_wc_header *wch = &wr->wr_wc_header; 2397 1.3.2.2 wrstuden int error; 2398 1.3.2.2 wrstuden 2399 1.3.2.2 wrstuden int logblklen = 1<<wch->wc_log_dev_bshift; 2400 1.3.2.2 wrstuden int fsblklen = 1<<wch->wc_fs_dev_bshift; 2401 1.3.2.2 wrstuden 2402 1.3.2.2 wrstuden wapbl_blkhash_clear(wr); 2403 1.3.2.2 wrstuden 2404 1.3.2.2 wrstuden off = wch->wc_tail; 2405 1.3.2.2 wrstuden while (off != wch->wc_head) { 2406 1.3.2.2 wrstuden struct wapbl_wc_null *wcn; 2407 1.3.2.2 wrstuden off_t saveoff = off; 2408 1.3.2.2 wrstuden error = wapbl_circ_read(wr, wr->wr_scratch, logblklen, &off); 2409 1.3.2.2 wrstuden if (error) 2410 1.3.2.2 wrstuden goto errout; 2411 1.3.2.2 wrstuden wcn = (struct wapbl_wc_null *)wr->wr_scratch; 2412 1.3.2.2 wrstuden switch (wcn->wc_type) { 2413 1.3.2.2 wrstuden case WAPBL_WC_BLOCKS: 2414 1.3.2.2 wrstuden { 2415 1.3.2.2 wrstuden struct wapbl_wc_blocklist *wc = 2416 1.3.2.2 wrstuden (struct wapbl_wc_blocklist *)wr->wr_scratch; 2417 1.3.2.2 wrstuden int i; 2418 1.3.2.2 wrstuden for (i = 0; i < wc->wc_blkcount; i++) { 2419 1.3.2.2 wrstuden int j, n; 2420 1.3.2.2 wrstuden /* 2421 1.3.2.2 wrstuden * Enter each physical block into the 2422 1.3.2.2 wrstuden * hashtable independently 2423 1.3.2.2 wrstuden */ 2424 1.3.2.2 wrstuden n = wc->wc_blocks[i].wc_dlen >> 2425 1.3.2.2 wrstuden wch->wc_fs_dev_bshift; 2426 1.3.2.2 wrstuden for (j = 0; j < n; j++) { 2427 1.3.2.2 wrstuden wapbl_blkhash_ins(wr, 2428 1.3.2.2 wrstuden wc->wc_blocks[i].wc_daddr + j, 2429 1.3.2.2 wrstuden off); 2430 1.3.2.2 wrstuden wapbl_circ_advance(wr, 2431 1.3.2.2 wrstuden fsblklen, &off); 2432 1.3.2.2 wrstuden } 2433 1.3.2.2 wrstuden } 2434 1.3.2.2 wrstuden } 2435 1.3.2.2 wrstuden break; 2436 1.3.2.2 wrstuden 2437 1.3.2.2 wrstuden case WAPBL_WC_REVOCATIONS: 2438 1.3.2.2 wrstuden { 2439 1.3.2.2 wrstuden struct wapbl_wc_blocklist *wc = 2440 1.3.2.2 wrstuden (struct wapbl_wc_blocklist *)wr->wr_scratch; 2441 1.3.2.2 wrstuden int i; 2442 1.3.2.2 wrstuden for (i = 0; i < wc->wc_blkcount; i++) { 2443 1.3.2.2 wrstuden int j, n; 2444 1.3.2.2 wrstuden /* 2445 1.3.2.2 wrstuden * Remove any blocks found from the 2446 1.3.2.2 wrstuden * hashtable 2447 1.3.2.2 wrstuden */ 2448 1.3.2.2 wrstuden n = wc->wc_blocks[i].wc_dlen >> 2449 1.3.2.2 wrstuden wch->wc_fs_dev_bshift; 2450 1.3.2.2 wrstuden for (j = 0; j < n; j++) { 2451 1.3.2.2 wrstuden wapbl_blkhash_rem(wr, 2452 1.3.2.2 wrstuden wc->wc_blocks[i].wc_daddr + j); 2453 1.3.2.2 wrstuden } 2454 1.3.2.2 wrstuden } 2455 1.3.2.2 wrstuden } 2456 1.3.2.2 wrstuden break; 2457 1.3.2.2 wrstuden 2458 1.3.2.2 wrstuden case WAPBL_WC_INODES: 2459 1.3.2.2 wrstuden { 2460 1.3.2.2 wrstuden struct wapbl_wc_inodelist *wc = 2461 1.3.2.2 wrstuden (struct wapbl_wc_inodelist *)wr->wr_scratch; 2462 1.3.2.2 wrstuden /* 2463 1.3.2.2 wrstuden * Keep track of where we found this so we 2464 1.3.2.2 wrstuden * can use it later 2465 1.3.2.2 wrstuden */ 2466 1.3.2.2 wrstuden if (wc->wc_clear) { 2467 1.3.2.2 wrstuden wr->wr_inodestail = saveoff; 2468 1.3.2.2 wrstuden wr->wr_inodescnt = 0; 2469 1.3.2.2 wrstuden } 2470 1.3.2.2 wrstuden if (wr->wr_inodestail) 2471 1.3.2.2 wrstuden wr->wr_inodeshead = off; 2472 1.3.2.2 wrstuden wr->wr_inodescnt += wc->wc_inocnt; 2473 1.3.2.2 wrstuden } 2474 1.3.2.2 wrstuden break; 2475 1.3.2.2 wrstuden default: 2476 1.3.2.2 wrstuden printf("Unrecognized wapbl type: 0x%08x\n", 2477 1.3.2.2 wrstuden wcn->wc_type); 2478 1.3.2.2 wrstuden error = EFTYPE; 2479 1.3.2.2 wrstuden goto errout; 2480 1.3.2.2 wrstuden } 2481 1.3.2.2 wrstuden wapbl_circ_advance(wr, wcn->wc_len, &saveoff); 2482 1.3.2.2 wrstuden if (off != saveoff) { 2483 1.3.2.2 wrstuden printf("wapbl_replay: corrupted records\n"); 2484 1.3.2.2 wrstuden error = EFTYPE; 2485 1.3.2.2 wrstuden goto errout; 2486 1.3.2.2 wrstuden } 2487 1.3.2.2 wrstuden } 2488 1.3.2.2 wrstuden return 0; 2489 1.3.2.2 wrstuden 2490 1.3.2.2 wrstuden errout: 2491 1.3.2.2 wrstuden wapbl_blkhash_clear(wr); 2492 1.3.2.2 wrstuden return error; 2493 1.3.2.2 wrstuden } 2494 1.3.2.2 wrstuden 2495 1.3.2.2 wrstuden static int 2496 1.3.2.2 wrstuden wapbl_replay_get_inodes(struct wapbl_replay *wr) 2497 1.3.2.2 wrstuden { 2498 1.3.2.2 wrstuden off_t off; 2499 1.3.2.2 wrstuden struct wapbl_wc_header *wch = &wr->wr_wc_header; 2500 1.3.2.2 wrstuden int logblklen = 1<<wch->wc_log_dev_bshift; 2501 1.3.2.2 wrstuden int cnt= 0; 2502 1.3.2.2 wrstuden 2503 1.3.2.2 wrstuden KDASSERT(wapbl_replay_isopen(wr)); 2504 1.3.2.2 wrstuden 2505 1.3.2.2 wrstuden if (wr->wr_inodescnt == 0) 2506 1.3.2.2 wrstuden return 0; 2507 1.3.2.2 wrstuden 2508 1.3.2.2 wrstuden KASSERT(!wr->wr_inodes); 2509 1.3.2.2 wrstuden 2510 1.3.2.2 wrstuden wr->wr_inodes = wapbl_malloc(wr->wr_inodescnt*sizeof(wr->wr_inodes[0])); 2511 1.3.2.2 wrstuden 2512 1.3.2.2 wrstuden off = wr->wr_inodestail; 2513 1.3.2.2 wrstuden 2514 1.3.2.2 wrstuden while (off != wr->wr_inodeshead) { 2515 1.3.2.2 wrstuden struct wapbl_wc_null *wcn; 2516 1.3.2.2 wrstuden int error; 2517 1.3.2.2 wrstuden off_t saveoff = off; 2518 1.3.2.2 wrstuden error = wapbl_circ_read(wr, wr->wr_scratch, logblklen, &off); 2519 1.3.2.2 wrstuden if (error) { 2520 1.3.2.2 wrstuden wapbl_free(wr->wr_inodes); 2521 1.3.2.2 wrstuden wr->wr_inodes = 0; 2522 1.3.2.2 wrstuden return error; 2523 1.3.2.2 wrstuden } 2524 1.3.2.2 wrstuden wcn = (struct wapbl_wc_null *)wr->wr_scratch; 2525 1.3.2.2 wrstuden switch (wcn->wc_type) { 2526 1.3.2.2 wrstuden case WAPBL_WC_BLOCKS: 2527 1.3.2.2 wrstuden case WAPBL_WC_REVOCATIONS: 2528 1.3.2.2 wrstuden break; 2529 1.3.2.2 wrstuden case WAPBL_WC_INODES: 2530 1.3.2.2 wrstuden { 2531 1.3.2.2 wrstuden struct wapbl_wc_inodelist *wc = 2532 1.3.2.2 wrstuden (struct wapbl_wc_inodelist *)wr->wr_scratch; 2533 1.3.2.2 wrstuden /* 2534 1.3.2.2 wrstuden * Keep track of where we found this so we 2535 1.3.2.2 wrstuden * can use it later 2536 1.3.2.2 wrstuden */ 2537 1.3.2.2 wrstuden if (wc->wc_clear) { 2538 1.3.2.2 wrstuden cnt = 0; 2539 1.3.2.2 wrstuden } 2540 1.3.2.2 wrstuden /* This memcpy assumes that wr_inodes is 2541 1.3.2.2 wrstuden * laid out the same as wc_inodes. */ 2542 1.3.2.2 wrstuden memcpy(&wr->wr_inodes[cnt], wc->wc_inodes, 2543 1.3.2.2 wrstuden wc->wc_inocnt*sizeof(wc->wc_inodes[0])); 2544 1.3.2.2 wrstuden cnt += wc->wc_inocnt; 2545 1.3.2.2 wrstuden } 2546 1.3.2.2 wrstuden break; 2547 1.3.2.2 wrstuden default: 2548 1.3.2.2 wrstuden KASSERT(0); 2549 1.3.2.2 wrstuden } 2550 1.3.2.2 wrstuden off = saveoff; 2551 1.3.2.2 wrstuden wapbl_circ_advance(wr, wcn->wc_len, &off); 2552 1.3.2.2 wrstuden } 2553 1.3.2.2 wrstuden KASSERT(cnt == wr->wr_inodescnt); 2554 1.3.2.2 wrstuden return 0; 2555 1.3.2.2 wrstuden } 2556 1.3.2.2 wrstuden 2557 1.3.2.2 wrstuden #ifdef DEBUG 2558 1.3.2.2 wrstuden int 2559 1.3.2.2 wrstuden wapbl_replay_verify(struct wapbl_replay *wr, struct vnode *fsdevvp) 2560 1.3.2.2 wrstuden { 2561 1.3.2.2 wrstuden off_t off; 2562 1.3.2.2 wrstuden struct wapbl_wc_header *wch = &wr->wr_wc_header; 2563 1.3.2.2 wrstuden int mismatchcnt = 0; 2564 1.3.2.2 wrstuden int logblklen = 1<<wch->wc_log_dev_bshift; 2565 1.3.2.2 wrstuden int fsblklen = 1<<wch->wc_fs_dev_bshift; 2566 1.3.2.2 wrstuden void *scratch1 = wapbl_malloc(MAXBSIZE); 2567 1.3.2.2 wrstuden void *scratch2 = wapbl_malloc(MAXBSIZE); 2568 1.3.2.2 wrstuden int error = 0; 2569 1.3.2.2 wrstuden 2570 1.3.2.2 wrstuden KDASSERT(wapbl_replay_isopen(wr)); 2571 1.3.2.2 wrstuden 2572 1.3.2.2 wrstuden off = wch->wc_tail; 2573 1.3.2.2 wrstuden while (off != wch->wc_head) { 2574 1.3.2.2 wrstuden struct wapbl_wc_null *wcn; 2575 1.3.2.2 wrstuden #ifdef DEBUG 2576 1.3.2.2 wrstuden off_t saveoff = off; 2577 1.3.2.2 wrstuden #endif 2578 1.3.2.2 wrstuden error = wapbl_circ_read(wr, wr->wr_scratch, logblklen, &off); 2579 1.3.2.2 wrstuden if (error) 2580 1.3.2.2 wrstuden goto out; 2581 1.3.2.2 wrstuden wcn = (struct wapbl_wc_null *)wr->wr_scratch; 2582 1.3.2.2 wrstuden switch (wcn->wc_type) { 2583 1.3.2.2 wrstuden case WAPBL_WC_BLOCKS: 2584 1.3.2.2 wrstuden { 2585 1.3.2.2 wrstuden struct wapbl_wc_blocklist *wc = 2586 1.3.2.2 wrstuden (struct wapbl_wc_blocklist *)wr->wr_scratch; 2587 1.3.2.2 wrstuden int i; 2588 1.3.2.2 wrstuden for (i = 0; i < wc->wc_blkcount; i++) { 2589 1.3.2.2 wrstuden int foundcnt = 0; 2590 1.3.2.2 wrstuden int dirtycnt = 0; 2591 1.3.2.2 wrstuden int j, n; 2592 1.3.2.2 wrstuden /* 2593 1.3.2.2 wrstuden * Check each physical block into the 2594 1.3.2.2 wrstuden * hashtable independently 2595 1.3.2.2 wrstuden */ 2596 1.3.2.2 wrstuden n = wc->wc_blocks[i].wc_dlen >> 2597 1.3.2.2 wrstuden wch->wc_fs_dev_bshift; 2598 1.3.2.2 wrstuden for (j = 0; j < n; j++) { 2599 1.3.2.2 wrstuden struct wapbl_blk *wb = 2600 1.3.2.2 wrstuden wapbl_blkhash_get(wr, 2601 1.3.2.2 wrstuden wc->wc_blocks[i].wc_daddr + j); 2602 1.3.2.2 wrstuden if (wb && (wb->wb_off == off)) { 2603 1.3.2.2 wrstuden foundcnt++; 2604 1.3.2.2 wrstuden error = 2605 1.3.2.2 wrstuden wapbl_circ_read(wr, 2606 1.3.2.2 wrstuden scratch1, fsblklen, 2607 1.3.2.2 wrstuden &off); 2608 1.3.2.2 wrstuden if (error) 2609 1.3.2.2 wrstuden goto out; 2610 1.3.2.2 wrstuden error = 2611 1.3.2.2 wrstuden wapbl_read(scratch2, 2612 1.3.2.2 wrstuden fsblklen, fsdevvp, 2613 1.3.2.2 wrstuden wb->wb_blk); 2614 1.3.2.2 wrstuden if (error) 2615 1.3.2.2 wrstuden goto out; 2616 1.3.2.2 wrstuden if (memcmp(scratch1, 2617 1.3.2.2 wrstuden scratch2, 2618 1.3.2.2 wrstuden fsblklen)) { 2619 1.3.2.2 wrstuden printf( 2620 1.3.2.2 wrstuden "wapbl_verify: mismatch block %"PRId64" at off %"PRIdMAX"\n", 2621 1.3.2.2 wrstuden wb->wb_blk, (intmax_t)off); 2622 1.3.2.2 wrstuden dirtycnt++; 2623 1.3.2.2 wrstuden mismatchcnt++; 2624 1.3.2.2 wrstuden } 2625 1.3.2.2 wrstuden } else { 2626 1.3.2.2 wrstuden wapbl_circ_advance(wr, 2627 1.3.2.2 wrstuden fsblklen, &off); 2628 1.3.2.2 wrstuden } 2629 1.3.2.2 wrstuden } 2630 1.3.2.2 wrstuden #if 0 2631 1.3.2.2 wrstuden /* 2632 1.3.2.2 wrstuden * If all of the blocks in an entry 2633 1.3.2.2 wrstuden * are clean, then remove all of its 2634 1.3.2.2 wrstuden * blocks from the hashtable since they 2635 1.3.2.2 wrstuden * never will need replay. 2636 1.3.2.2 wrstuden */ 2637 1.3.2.2 wrstuden if ((foundcnt != 0) && 2638 1.3.2.2 wrstuden (dirtycnt == 0)) { 2639 1.3.2.2 wrstuden off = saveoff; 2640 1.3.2.2 wrstuden wapbl_circ_advance(wr, 2641 1.3.2.2 wrstuden logblklen, &off); 2642 1.3.2.2 wrstuden for (j = 0; j < n; j++) { 2643 1.3.2.2 wrstuden struct wapbl_blk *wb = 2644 1.3.2.2 wrstuden wapbl_blkhash_get(wr, 2645 1.3.2.2 wrstuden wc->wc_blocks[i].wc_daddr + j); 2646 1.3.2.2 wrstuden if (wb && 2647 1.3.2.2 wrstuden (wb->wb_off == off)) { 2648 1.3.2.2 wrstuden wapbl_blkhash_rem(wr, wb->wb_blk); 2649 1.3.2.2 wrstuden } 2650 1.3.2.2 wrstuden wapbl_circ_advance(wr, 2651 1.3.2.2 wrstuden fsblklen, &off); 2652 1.3.2.2 wrstuden } 2653 1.3.2.2 wrstuden } 2654 1.3.2.2 wrstuden #endif 2655 1.3.2.2 wrstuden } 2656 1.3.2.2 wrstuden } 2657 1.3.2.2 wrstuden break; 2658 1.3.2.2 wrstuden case WAPBL_WC_REVOCATIONS: 2659 1.3.2.2 wrstuden case WAPBL_WC_INODES: 2660 1.3.2.2 wrstuden break; 2661 1.3.2.2 wrstuden default: 2662 1.3.2.2 wrstuden KASSERT(0); 2663 1.3.2.2 wrstuden } 2664 1.3.2.2 wrstuden #ifdef DEBUG 2665 1.3.2.2 wrstuden wapbl_circ_advance(wr, wcn->wc_len, &saveoff); 2666 1.3.2.2 wrstuden KASSERT(off == saveoff); 2667 1.3.2.2 wrstuden #endif 2668 1.3.2.2 wrstuden } 2669 1.3.2.2 wrstuden out: 2670 1.3.2.2 wrstuden wapbl_free(scratch1); 2671 1.3.2.2 wrstuden wapbl_free(scratch2); 2672 1.3.2.2 wrstuden if (!error && mismatchcnt) 2673 1.3.2.2 wrstuden error = EFTYPE; 2674 1.3.2.2 wrstuden return error; 2675 1.3.2.2 wrstuden } 2676 1.3.2.2 wrstuden #endif 2677 1.3.2.2 wrstuden 2678 1.3.2.2 wrstuden int 2679 1.3.2.2 wrstuden wapbl_replay_write(struct wapbl_replay *wr, struct vnode *fsdevvp) 2680 1.3.2.2 wrstuden { 2681 1.3.2.2 wrstuden off_t off; 2682 1.3.2.2 wrstuden struct wapbl_wc_header *wch = &wr->wr_wc_header; 2683 1.3.2.2 wrstuden int logblklen = 1<<wch->wc_log_dev_bshift; 2684 1.3.2.2 wrstuden int fsblklen = 1<<wch->wc_fs_dev_bshift; 2685 1.3.2.2 wrstuden void *scratch1 = wapbl_malloc(MAXBSIZE); 2686 1.3.2.2 wrstuden int error = 0; 2687 1.3.2.2 wrstuden 2688 1.3.2.2 wrstuden KDASSERT(wapbl_replay_isopen(wr)); 2689 1.3.2.2 wrstuden 2690 1.3.2.2 wrstuden /* 2691 1.3.2.2 wrstuden * This parses the journal for replay, although it could 2692 1.3.2.2 wrstuden * just as easily walk the hashtable instead. 2693 1.3.2.2 wrstuden */ 2694 1.3.2.2 wrstuden 2695 1.3.2.2 wrstuden off = wch->wc_tail; 2696 1.3.2.2 wrstuden while (off != wch->wc_head) { 2697 1.3.2.2 wrstuden struct wapbl_wc_null *wcn; 2698 1.3.2.2 wrstuden #ifdef DEBUG 2699 1.3.2.2 wrstuden off_t saveoff = off; 2700 1.3.2.2 wrstuden #endif 2701 1.3.2.2 wrstuden error = wapbl_circ_read(wr, wr->wr_scratch, logblklen, &off); 2702 1.3.2.2 wrstuden if (error) 2703 1.3.2.2 wrstuden goto out; 2704 1.3.2.2 wrstuden wcn = (struct wapbl_wc_null *)wr->wr_scratch; 2705 1.3.2.2 wrstuden switch (wcn->wc_type) { 2706 1.3.2.2 wrstuden case WAPBL_WC_BLOCKS: 2707 1.3.2.2 wrstuden { 2708 1.3.2.2 wrstuden struct wapbl_wc_blocklist *wc = 2709 1.3.2.2 wrstuden (struct wapbl_wc_blocklist *)wr->wr_scratch; 2710 1.3.2.2 wrstuden int i; 2711 1.3.2.2 wrstuden for (i = 0; i < wc->wc_blkcount; i++) { 2712 1.3.2.2 wrstuden int j, n; 2713 1.3.2.2 wrstuden /* 2714 1.3.2.2 wrstuden * Check each physical block against 2715 1.3.2.2 wrstuden * the hashtable independently 2716 1.3.2.2 wrstuden */ 2717 1.3.2.2 wrstuden n = wc->wc_blocks[i].wc_dlen >> 2718 1.3.2.2 wrstuden wch->wc_fs_dev_bshift; 2719 1.3.2.2 wrstuden for (j = 0; j < n; j++) { 2720 1.3.2.2 wrstuden struct wapbl_blk *wb = 2721 1.3.2.2 wrstuden wapbl_blkhash_get(wr, 2722 1.3.2.2 wrstuden wc->wc_blocks[i].wc_daddr + j); 2723 1.3.2.2 wrstuden if (wb && (wb->wb_off == off)) { 2724 1.3.2.2 wrstuden error = wapbl_circ_read( 2725 1.3.2.2 wrstuden wr, scratch1, 2726 1.3.2.2 wrstuden fsblklen, &off); 2727 1.3.2.2 wrstuden if (error) 2728 1.3.2.2 wrstuden goto out; 2729 1.3.2.2 wrstuden error = 2730 1.3.2.2 wrstuden wapbl_write(scratch1, 2731 1.3.2.2 wrstuden fsblklen, fsdevvp, 2732 1.3.2.2 wrstuden wb->wb_blk); 2733 1.3.2.2 wrstuden if (error) 2734 1.3.2.2 wrstuden goto out; 2735 1.3.2.2 wrstuden } else { 2736 1.3.2.2 wrstuden wapbl_circ_advance(wr, 2737 1.3.2.2 wrstuden fsblklen, &off); 2738 1.3.2.2 wrstuden } 2739 1.3.2.2 wrstuden } 2740 1.3.2.2 wrstuden } 2741 1.3.2.2 wrstuden } 2742 1.3.2.2 wrstuden break; 2743 1.3.2.2 wrstuden case WAPBL_WC_REVOCATIONS: 2744 1.3.2.2 wrstuden case WAPBL_WC_INODES: 2745 1.3.2.2 wrstuden break; 2746 1.3.2.2 wrstuden default: 2747 1.3.2.2 wrstuden KASSERT(0); 2748 1.3.2.2 wrstuden } 2749 1.3.2.2 wrstuden #ifdef DEBUG 2750 1.3.2.2 wrstuden wapbl_circ_advance(wr, wcn->wc_len, &saveoff); 2751 1.3.2.2 wrstuden KASSERT(off == saveoff); 2752 1.3.2.2 wrstuden #endif 2753 1.3.2.2 wrstuden } 2754 1.3.2.2 wrstuden out: 2755 1.3.2.2 wrstuden wapbl_free(scratch1); 2756 1.3.2.2 wrstuden return error; 2757 1.3.2.2 wrstuden } 2758 1.3.2.2 wrstuden 2759 1.3.2.2 wrstuden int 2760 1.3.2.2 wrstuden wapbl_replay_read(struct wapbl_replay *wr, void *data, daddr_t blk, long len) 2761 1.3.2.2 wrstuden { 2762 1.3.2.2 wrstuden struct wapbl_wc_header *wch = &wr->wr_wc_header; 2763 1.3.2.2 wrstuden int fsblklen = 1<<wch->wc_fs_dev_bshift; 2764 1.3.2.2 wrstuden 2765 1.3.2.2 wrstuden KDASSERT(wapbl_replay_isopen(wr)); 2766 1.3.2.2 wrstuden 2767 1.3.2.2 wrstuden KASSERT((len % fsblklen) == 0); 2768 1.3.2.2 wrstuden 2769 1.3.2.2 wrstuden while (len != 0) { 2770 1.3.2.2 wrstuden struct wapbl_blk *wb = wapbl_blkhash_get(wr, blk); 2771 1.3.2.2 wrstuden if (wb) { 2772 1.3.2.2 wrstuden off_t off = wb->wb_off; 2773 1.3.2.2 wrstuden int error; 2774 1.3.2.2 wrstuden error = wapbl_circ_read(wr, data, fsblklen, &off); 2775 1.3.2.2 wrstuden if (error) 2776 1.3.2.2 wrstuden return error; 2777 1.3.2.2 wrstuden } 2778 1.3.2.2 wrstuden data = (uint8_t *)data + fsblklen; 2779 1.3.2.2 wrstuden len -= fsblklen; 2780 1.3.2.2 wrstuden blk++; 2781 1.3.2.2 wrstuden } 2782 1.3.2.2 wrstuden return 0; 2783 1.3.2.2 wrstuden } 2784
Indexes created Fri Oct 03 14:09:59 GMT 2025