git.sesse.net Git - bcachefs-tools-debian/blob - bcache.c

   1 #define _GNU_SOURCE
   2
   3 #include <ctype.h>
   4 #include <errno.h>
   5 #include <stdio.h>
   6 #include <stdlib.h>
   7 #include <stdint.h>
   8 #include <fcntl.h>
   9 #include <inttypes.h>
  10 #include <getopt.h>
  11 #include <limits.h>
  12 #include <linux/fs.h>
  13 #include <stdbool.h>
  14 #include <string.h>
  15 #include <sys/ioctl.h>
  16 #include <sys/types.h>
  17 #include <sys/stat.h>
  18 #include <dirent.h>
  19 #include <unistd.h>
  20 #include <uuid/uuid.h>
  21 #include <blkid.h>
  22
  23 #include "bcache.h"
  24
  25 #define __KERNEL__
  26 #include <linux/bcache-ioctl.h>
  27 #undef __KERNEL__
  28
  29 const char * const cache_state[] = {
  30         "active",
  31         "ro",
  32         "failed",
  33         "spare",
  34         NULL
  35 };
  36
  37 const char * const replacement_policies[] = {
  38         "lru",
  39         "fifo",
  40         "random",
  41         NULL
  42 };
  43
  44 const char * const csum_types[] = {
  45         "none",
  46         "crc32c",
  47         "crc64",
  48         NULL
  49 };
  50
  51 const char * const bdev_cache_mode[] = {
  52         "writethrough",
  53         "writeback",
  54         "writearound",
  55         "none",
  56         NULL
  57 };
  58
  59 const char * const bdev_state[] = {
  60         "detached",
  61         "clean",
  62         "dirty",
  63         "inconsistent",
  64         NULL
  65 };
  66
  67 const char * const set_attrs[] = {
  68         "btree_flush_delay",
  69         "btree_scan_ratelimit",
  70         "bucket_reserve_percent",
  71         "cache_reserve_percent",
  72         "checksum_type",
  73         "congested_read_threshold_us",
  74         "congested_write_threshold_us",
  75         "data_replicas",
  76         "errors",
  77         "foreground_target_percent",
  78         "gc_sector_percent",
  79         "journal_delay_ms",
  80         "meta_replicas",
  81         "sector_reserve_percent",
  82         "tiering_percent",
  83         NULL
  84 };
  85
  86 const char * const cache_attrs[] = {
  87         "cache_replacement_policy",
  88         "discard",
  89         "state",
  90         "tier",
  91         NULL
  92 };
  93
  94 const char * const internal_attrs[] = {
  95         "btree_shrinker_disabled",
  96         "copy_gc_enabled",
  97         "foreground_write_rate",
  98         "tiering_enabled",
  99         "tiering_rate",
 100         NULL
 101 };
 102
 103 char *skip_spaces(const char *str)
 104 {
 105         while (isspace(*str))
 106                 ++str;
 107         return (char *)str;
 108 }
 109
 110 char *strim(char *s)
 111 {
 112         size_t size;
 113         char *end;
 114
 115         s = skip_spaces(s);
 116         size = strlen(s);
 117         if (!size)
 118                 return s;
 119
 120         end = s + size - 1;
 121         while (end >= s && isspace(*end))
 122                 end--;
 123         *(end + 1) = '\0';
 124
 125         return s;
 126 }
 127
 128 ssize_t read_string_list(const char *buf, const char * const list[])
 129 {
 130         size_t i;
 131         char *s, *d = strdup(buf);
 132         if (!d)
 133                 return -ENOMEM;
 134
 135         s = strim(d);
 136
 137         for (i = 0; list[i]; i++)
 138                 if (!strcmp(list[i], s))
 139                         break;
 140
 141         free(d);
 142
 143         if (!list[i])
 144                 return -EINVAL;
 145
 146         return i;
 147 }
 148
 149 ssize_t read_string_list_or_die(const char *opt, const char * const list[],
 150                                 const char *msg)
 151 {
 152         ssize_t v = read_string_list(opt, list);
 153         if (v < 0) {
 154                 fprintf(stderr, "Bad %s %s\n", msg, opt);
 155                 exit(EXIT_FAILURE);
 156
 157         }
 158
 159         return v;
 160 }
 161
 162 void print_string_list(const char * const list[], size_t selected)
 163 {
 164         size_t i;
 165
 166         for (i = 0; list[i]; i++) {
 167                 if (i)
 168                         putchar(' ');
 169                 printf(i == selected ? "[%s] ": "%s", list[i]);
 170         }
 171 }
 172
 173 /*
 174  * This is the CRC-32C table
 175  * Generated with:
 176  * width = 32 bits
 177  * poly = 0x1EDC6F41
 178  * reflect input bytes = true
 179  * reflect output bytes = true
 180  */
 181
 182 static const u32 crc32c_table[256] = {
 183         0x00000000L, 0xF26B8303L, 0xE13B70F7L, 0x1350F3F4L,
 184         0xC79A971FL, 0x35F1141CL, 0x26A1E7E8L, 0xD4CA64EBL,
 185         0x8AD958CFL, 0x78B2DBCCL, 0x6BE22838L, 0x9989AB3BL,
 186         0x4D43CFD0L, 0xBF284CD3L, 0xAC78BF27L, 0x5E133C24L,
 187         0x105EC76FL, 0xE235446CL, 0xF165B798L, 0x030E349BL,
 188         0xD7C45070L, 0x25AFD373L, 0x36FF2087L, 0xC494A384L,
 189         0x9A879FA0L, 0x68EC1CA3L, 0x7BBCEF57L, 0x89D76C54L,
 190         0x5D1D08BFL, 0xAF768BBCL, 0xBC267848L, 0x4E4DFB4BL,
 191         0x20BD8EDEL, 0xD2D60DDDL, 0xC186FE29L, 0x33ED7D2AL,
 192         0xE72719C1L, 0x154C9AC2L, 0x061C6936L, 0xF477EA35L,
 193         0xAA64D611L, 0x580F5512L, 0x4B5FA6E6L, 0xB93425E5L,
 194         0x6DFE410EL, 0x9F95C20DL, 0x8CC531F9L, 0x7EAEB2FAL,
 195         0x30E349B1L, 0xC288CAB2L, 0xD1D83946L, 0x23B3BA45L,
 196         0xF779DEAEL, 0x05125DADL, 0x1642AE59L, 0xE4292D5AL,
 197         0xBA3A117EL, 0x4851927DL, 0x5B016189L, 0xA96AE28AL,
 198         0x7DA08661L, 0x8FCB0562L, 0x9C9BF696L, 0x6EF07595L,
 199         0x417B1DBCL, 0xB3109EBFL, 0xA0406D4BL, 0x522BEE48L,
 200         0x86E18AA3L, 0x748A09A0L, 0x67DAFA54L, 0x95B17957L,
 201         0xCBA24573L, 0x39C9C670L, 0x2A993584L, 0xD8F2B687L,
 202         0x0C38D26CL, 0xFE53516FL, 0xED03A29BL, 0x1F682198L,
 203         0x5125DAD3L, 0xA34E59D0L, 0xB01EAA24L, 0x42752927L,
 204         0x96BF4DCCL, 0x64D4CECFL, 0x77843D3BL, 0x85EFBE38L,
 205         0xDBFC821CL, 0x2997011FL, 0x3AC7F2EBL, 0xC8AC71E8L,
 206         0x1C661503L, 0xEE0D9600L, 0xFD5D65F4L, 0x0F36E6F7L,
 207         0x61C69362L, 0x93AD1061L, 0x80FDE395L, 0x72966096L,
 208         0xA65C047DL, 0x5437877EL, 0x4767748AL, 0xB50CF789L,
 209         0xEB1FCBADL, 0x197448AEL, 0x0A24BB5AL, 0xF84F3859L,
 210         0x2C855CB2L, 0xDEEEDFB1L, 0xCDBE2C45L, 0x3FD5AF46L,
 211         0x7198540DL, 0x83F3D70EL, 0x90A324FAL, 0x62C8A7F9L,
 212         0xB602C312L, 0x44694011L, 0x5739B3E5L, 0xA55230E6L,
 213         0xFB410CC2L, 0x092A8FC1L, 0x1A7A7C35L, 0xE811FF36L,
 214         0x3CDB9BDDL, 0xCEB018DEL, 0xDDE0EB2AL, 0x2F8B6829L,
 215         0x82F63B78L, 0x709DB87BL, 0x63CD4B8FL, 0x91A6C88CL,
 216         0x456CAC67L, 0xB7072F64L, 0xA457DC90L, 0x563C5F93L,
 217         0x082F63B7L, 0xFA44E0B4L, 0xE9141340L, 0x1B7F9043L,
 218         0xCFB5F4A8L, 0x3DDE77ABL, 0x2E8E845FL, 0xDCE5075CL,
 219         0x92A8FC17L, 0x60C37F14L, 0x73938CE0L, 0x81F80FE3L,
 220         0x55326B08L, 0xA759E80BL, 0xB4091BFFL, 0x466298FCL,
 221         0x1871A4D8L, 0xEA1A27DBL, 0xF94AD42FL, 0x0B21572CL,
 222         0xDFEB33C7L, 0x2D80B0C4L, 0x3ED04330L, 0xCCBBC033L,
 223         0xA24BB5A6L, 0x502036A5L, 0x4370C551L, 0xB11B4652L,
 224         0x65D122B9L, 0x97BAA1BAL, 0x84EA524EL, 0x7681D14DL,
 225         0x2892ED69L, 0xDAF96E6AL, 0xC9A99D9EL, 0x3BC21E9DL,
 226         0xEF087A76L, 0x1D63F975L, 0x0E330A81L, 0xFC588982L,
 227         0xB21572C9L, 0x407EF1CAL, 0x532E023EL, 0xA145813DL,
 228         0x758FE5D6L, 0x87E466D5L, 0x94B49521L, 0x66DF1622L,
 229         0x38CC2A06L, 0xCAA7A905L, 0xD9F75AF1L, 0x2B9CD9F2L,
 230         0xFF56BD19L, 0x0D3D3E1AL, 0x1E6DCDEEL, 0xEC064EEDL,
 231         0xC38D26C4L, 0x31E6A5C7L, 0x22B65633L, 0xD0DDD530L,
 232         0x0417B1DBL, 0xF67C32D8L, 0xE52CC12CL, 0x1747422FL,
 233         0x49547E0BL, 0xBB3FFD08L, 0xA86F0EFCL, 0x5A048DFFL,
 234         0x8ECEE914L, 0x7CA56A17L, 0x6FF599E3L, 0x9D9E1AE0L,
 235         0xD3D3E1ABL, 0x21B862A8L, 0x32E8915CL, 0xC083125FL,
 236         0x144976B4L, 0xE622F5B7L, 0xF5720643L, 0x07198540L,
 237         0x590AB964L, 0xAB613A67L, 0xB831C993L, 0x4A5A4A90L,
 238         0x9E902E7BL, 0x6CFBAD78L, 0x7FAB5E8CL, 0x8DC0DD8FL,
 239         0xE330A81AL, 0x115B2B19L, 0x020BD8EDL, 0xF0605BEEL,
 240         0x24AA3F05L, 0xD6C1BC06L, 0xC5914FF2L, 0x37FACCF1L,
 241         0x69E9F0D5L, 0x9B8273D6L, 0x88D28022L, 0x7AB90321L,
 242         0xAE7367CAL, 0x5C18E4C9L, 0x4F48173DL, 0xBD23943EL,
 243         0xF36E6F75L, 0x0105EC76L, 0x12551F82L, 0xE03E9C81L,
 244         0x34F4F86AL, 0xC69F7B69L, 0xD5CF889DL, 0x27A40B9EL,
 245         0x79B737BAL, 0x8BDCB4B9L, 0x988C474DL, 0x6AE7C44EL,
 246         0xBE2DA0A5L, 0x4C4623A6L, 0x5F16D052L, 0xAD7D5351L
 247 };
 248
 249 /*
 250  * Steps through buffer one byte at at time, calculates reflected
 251  * crc using table.
 252  */
 253
 254 static u32 crc32c(u32 crc, unsigned char const *data, size_t length)
 255 {
 256         while (length--)
 257                 crc =
 258                     crc32c_table[(crc ^ *data++) & 0xFFL] ^ (crc >> 8);
 259         return crc;
 260 }
 261
 262 /*
 263  * Portions Copyright (c) 1996-2001, PostgreSQL Global Development Group (Any
 264  * use permitted, subject to terms of PostgreSQL license; see.)
 265
 266  * If we have a 64-bit integer type, then a 64-bit CRC looks just like the
 267  * usual sort of implementation. (See Ross Williams' excellent introduction
 268  * A PAINLESS GUIDE TO CRC ERROR DETECTION ALGORITHMS, available from
 269  * ftp://ftp.rocksoft.com/papers/crc_v3.txt or several other net sites.)
 270  * If we have no working 64-bit type, then fake it with two 32-bit registers.
 271  *
 272  * The present implementation is a normal (not "reflected", in Williams'
 273  * terms) 64-bit CRC, using initial all-ones register contents and a final
 274  * bit inversion. The chosen polynomial is borrowed from the DLT1 spec
 275  * (ECMA-182, available from http://www.ecma.ch/ecma1/STAND/ECMA-182.HTM):
 276  *
 277  * x^64 + x^62 + x^57 + x^55 + x^54 + x^53 + x^52 + x^47 + x^46 + x^45 +
 278  * x^40 + x^39 + x^38 + x^37 + x^35 + x^33 + x^32 + x^31 + x^29 + x^27 +
 279  * x^24 + x^23 + x^22 + x^21 + x^19 + x^17 + x^13 + x^12 + x^10 + x^9 +
 280  * x^7 + x^4 + x + 1
 281 */
 282
 283 static const uint64_t crc_table[256] = {
 284         0x0000000000000000ULL, 0x42F0E1EBA9EA3693ULL, 0x85E1C3D753D46D26ULL,
 285         0xC711223CFA3E5BB5ULL, 0x493366450E42ECDFULL, 0x0BC387AEA7A8DA4CULL,
 286         0xCCD2A5925D9681F9ULL, 0x8E224479F47CB76AULL, 0x9266CC8A1C85D9BEULL,
 287         0xD0962D61B56FEF2DULL, 0x17870F5D4F51B498ULL, 0x5577EEB6E6BB820BULL,
 288         0xDB55AACF12C73561ULL, 0x99A54B24BB2D03F2ULL, 0x5EB4691841135847ULL,
 289         0x1C4488F3E8F96ED4ULL, 0x663D78FF90E185EFULL, 0x24CD9914390BB37CULL,
 290         0xE3DCBB28C335E8C9ULL, 0xA12C5AC36ADFDE5AULL, 0x2F0E1EBA9EA36930ULL,
 291         0x6DFEFF5137495FA3ULL, 0xAAEFDD6DCD770416ULL, 0xE81F3C86649D3285ULL,
 292         0xF45BB4758C645C51ULL, 0xB6AB559E258E6AC2ULL, 0x71BA77A2DFB03177ULL,
 293         0x334A9649765A07E4ULL, 0xBD68D2308226B08EULL, 0xFF9833DB2BCC861DULL,
 294         0x388911E7D1F2DDA8ULL, 0x7A79F00C7818EB3BULL, 0xCC7AF1FF21C30BDEULL,
 295         0x8E8A101488293D4DULL, 0x499B3228721766F8ULL, 0x0B6BD3C3DBFD506BULL,
 296         0x854997BA2F81E701ULL, 0xC7B97651866BD192ULL, 0x00A8546D7C558A27ULL,
 297         0x4258B586D5BFBCB4ULL, 0x5E1C3D753D46D260ULL, 0x1CECDC9E94ACE4F3ULL,
 298         0xDBFDFEA26E92BF46ULL, 0x990D1F49C77889D5ULL, 0x172F5B3033043EBFULL,
 299         0x55DFBADB9AEE082CULL, 0x92CE98E760D05399ULL, 0xD03E790CC93A650AULL,
 300         0xAA478900B1228E31ULL, 0xE8B768EB18C8B8A2ULL, 0x2FA64AD7E2F6E317ULL,
 301         0x6D56AB3C4B1CD584ULL, 0xE374EF45BF6062EEULL, 0xA1840EAE168A547DULL,
 302         0x66952C92ECB40FC8ULL, 0x2465CD79455E395BULL, 0x3821458AADA7578FULL,
 303         0x7AD1A461044D611CULL, 0xBDC0865DFE733AA9ULL, 0xFF3067B657990C3AULL,
 304         0x711223CFA3E5BB50ULL, 0x33E2C2240A0F8DC3ULL, 0xF4F3E018F031D676ULL,
 305         0xB60301F359DBE0E5ULL, 0xDA050215EA6C212FULL, 0x98F5E3FE438617BCULL,
 306         0x5FE4C1C2B9B84C09ULL, 0x1D14202910527A9AULL, 0x93366450E42ECDF0ULL,
 307         0xD1C685BB4DC4FB63ULL, 0x16D7A787B7FAA0D6ULL, 0x5427466C1E109645ULL,
 308         0x4863CE9FF6E9F891ULL, 0x0A932F745F03CE02ULL, 0xCD820D48A53D95B7ULL,
 309         0x8F72ECA30CD7A324ULL, 0x0150A8DAF8AB144EULL, 0x43A04931514122DDULL,
 310         0x84B16B0DAB7F7968ULL, 0xC6418AE602954FFBULL, 0xBC387AEA7A8DA4C0ULL,
 311         0xFEC89B01D3679253ULL, 0x39D9B93D2959C9E6ULL, 0x7B2958D680B3FF75ULL,
 312         0xF50B1CAF74CF481FULL, 0xB7FBFD44DD257E8CULL, 0x70EADF78271B2539ULL,
 313         0x321A3E938EF113AAULL, 0x2E5EB66066087D7EULL, 0x6CAE578BCFE24BEDULL,
 314         0xABBF75B735DC1058ULL, 0xE94F945C9C3626CBULL, 0x676DD025684A91A1ULL,
 315         0x259D31CEC1A0A732ULL, 0xE28C13F23B9EFC87ULL, 0xA07CF2199274CA14ULL,
 316         0x167FF3EACBAF2AF1ULL, 0x548F120162451C62ULL, 0x939E303D987B47D7ULL,
 317         0xD16ED1D631917144ULL, 0x5F4C95AFC5EDC62EULL, 0x1DBC74446C07F0BDULL,
 318         0xDAAD56789639AB08ULL, 0x985DB7933FD39D9BULL, 0x84193F60D72AF34FULL,
 319         0xC6E9DE8B7EC0C5DCULL, 0x01F8FCB784FE9E69ULL, 0x43081D5C2D14A8FAULL,
 320         0xCD2A5925D9681F90ULL, 0x8FDAB8CE70822903ULL, 0x48CB9AF28ABC72B6ULL,
 321         0x0A3B7B1923564425ULL, 0x70428B155B4EAF1EULL, 0x32B26AFEF2A4998DULL,
 322         0xF5A348C2089AC238ULL, 0xB753A929A170F4ABULL, 0x3971ED50550C43C1ULL,
 323         0x7B810CBBFCE67552ULL, 0xBC902E8706D82EE7ULL, 0xFE60CF6CAF321874ULL,
 324         0xE224479F47CB76A0ULL, 0xA0D4A674EE214033ULL, 0x67C58448141F1B86ULL,
 325         0x253565A3BDF52D15ULL, 0xAB1721DA49899A7FULL, 0xE9E7C031E063ACECULL,
 326         0x2EF6E20D1A5DF759ULL, 0x6C0603E6B3B7C1CAULL, 0xF6FAE5C07D3274CDULL,
 327         0xB40A042BD4D8425EULL, 0x731B26172EE619EBULL, 0x31EBC7FC870C2F78ULL,
 328         0xBFC9838573709812ULL, 0xFD39626EDA9AAE81ULL, 0x3A28405220A4F534ULL,
 329         0x78D8A1B9894EC3A7ULL, 0x649C294A61B7AD73ULL, 0x266CC8A1C85D9BE0ULL,
 330         0xE17DEA9D3263C055ULL, 0xA38D0B769B89F6C6ULL, 0x2DAF4F0F6FF541ACULL,
 331         0x6F5FAEE4C61F773FULL, 0xA84E8CD83C212C8AULL, 0xEABE6D3395CB1A19ULL,
 332         0x90C79D3FEDD3F122ULL, 0xD2377CD44439C7B1ULL, 0x15265EE8BE079C04ULL,
 333         0x57D6BF0317EDAA97ULL, 0xD9F4FB7AE3911DFDULL, 0x9B041A914A7B2B6EULL,
 334         0x5C1538ADB04570DBULL, 0x1EE5D94619AF4648ULL, 0x02A151B5F156289CULL,
 335         0x4051B05E58BC1E0FULL, 0x87409262A28245BAULL, 0xC5B073890B687329ULL,
 336         0x4B9237F0FF14C443ULL, 0x0962D61B56FEF2D0ULL, 0xCE73F427ACC0A965ULL,
 337         0x8C8315CC052A9FF6ULL, 0x3A80143F5CF17F13ULL, 0x7870F5D4F51B4980ULL,
 338         0xBF61D7E80F251235ULL, 0xFD913603A6CF24A6ULL, 0x73B3727A52B393CCULL,
 339         0x31439391FB59A55FULL, 0xF652B1AD0167FEEAULL, 0xB4A25046A88DC879ULL,
 340         0xA8E6D8B54074A6ADULL, 0xEA16395EE99E903EULL, 0x2D071B6213A0CB8BULL,
 341         0x6FF7FA89BA4AFD18ULL, 0xE1D5BEF04E364A72ULL, 0xA3255F1BE7DC7CE1ULL,
 342         0x64347D271DE22754ULL, 0x26C49CCCB40811C7ULL, 0x5CBD6CC0CC10FAFCULL,
 343         0x1E4D8D2B65FACC6FULL, 0xD95CAF179FC497DAULL, 0x9BAC4EFC362EA149ULL,
 344         0x158E0A85C2521623ULL, 0x577EEB6E6BB820B0ULL, 0x906FC95291867B05ULL,
 345         0xD29F28B9386C4D96ULL, 0xCEDBA04AD0952342ULL, 0x8C2B41A1797F15D1ULL,
 346         0x4B3A639D83414E64ULL, 0x09CA82762AAB78F7ULL, 0x87E8C60FDED7CF9DULL,
 347         0xC51827E4773DF90EULL, 0x020905D88D03A2BBULL, 0x40F9E43324E99428ULL,
 348         0x2CFFE7D5975E55E2ULL, 0x6E0F063E3EB46371ULL, 0xA91E2402C48A38C4ULL,
 349         0xEBEEC5E96D600E57ULL, 0x65CC8190991CB93DULL, 0x273C607B30F68FAEULL,
 350         0xE02D4247CAC8D41BULL, 0xA2DDA3AC6322E288ULL, 0xBE992B5F8BDB8C5CULL,
 351         0xFC69CAB42231BACFULL, 0x3B78E888D80FE17AULL, 0x7988096371E5D7E9ULL,
 352         0xF7AA4D1A85996083ULL, 0xB55AACF12C735610ULL, 0x724B8ECDD64D0DA5ULL,
 353         0x30BB6F267FA73B36ULL, 0x4AC29F2A07BFD00DULL, 0x08327EC1AE55E69EULL,
 354         0xCF235CFD546BBD2BULL, 0x8DD3BD16FD818BB8ULL, 0x03F1F96F09FD3CD2ULL,
 355         0x41011884A0170A41ULL, 0x86103AB85A2951F4ULL, 0xC4E0DB53F3C36767ULL,
 356         0xD8A453A01B3A09B3ULL, 0x9A54B24BB2D03F20ULL, 0x5D45907748EE6495ULL,
 357         0x1FB5719CE1045206ULL, 0x919735E51578E56CULL, 0xD367D40EBC92D3FFULL,
 358         0x1476F63246AC884AULL, 0x568617D9EF46BED9ULL, 0xE085162AB69D5E3CULL,
 359         0xA275F7C11F7768AFULL, 0x6564D5FDE549331AULL, 0x279434164CA30589ULL,
 360         0xA9B6706FB8DFB2E3ULL, 0xEB46918411358470ULL, 0x2C57B3B8EB0BDFC5ULL,
 361         0x6EA7525342E1E956ULL, 0x72E3DAA0AA188782ULL, 0x30133B4B03F2B111ULL,
 362         0xF7021977F9CCEAA4ULL, 0xB5F2F89C5026DC37ULL, 0x3BD0BCE5A45A6B5DULL,
 363         0x79205D0E0DB05DCEULL, 0xBE317F32F78E067BULL, 0xFCC19ED95E6430E8ULL,
 364         0x86B86ED5267CDBD3ULL, 0xC4488F3E8F96ED40ULL, 0x0359AD0275A8B6F5ULL,
 365         0x41A94CE9DC428066ULL, 0xCF8B0890283E370CULL, 0x8D7BE97B81D4019FULL,
 366         0x4A6ACB477BEA5A2AULL, 0x089A2AACD2006CB9ULL, 0x14DEA25F3AF9026DULL,
 367         0x562E43B4931334FEULL, 0x913F6188692D6F4BULL, 0xD3CF8063C0C759D8ULL,
 368         0x5DEDC41A34BBEEB2ULL, 0x1F1D25F19D51D821ULL, 0xD80C07CD676F8394ULL,
 369         0x9AFCE626CE85B507ULL
 370 };
 371
 372 static uint64_t bch_crc64_update(uint64_t crc, const void *_data, size_t len)
 373 {
 374         const unsigned char *data = _data;
 375
 376         while (len--) {
 377                 int i = ((int) (crc >> 56) ^ *data++) & 0xFF;
 378                 crc = crc_table[i] ^ (crc << 8);
 379         }
 380
 381         return crc;
 382 }
 383
 384 static uint64_t bch_checksum_update(unsigned type, uint64_t crc, const void *data, size_t len)
 385 {
 386         switch (type) {
 387         case BCH_CSUM_NONE:
 388                 return 0;
 389         case BCH_CSUM_CRC32C:
 390                 return crc32c(crc, data, len);
 391         case BCH_CSUM_CRC64:
 392                 return bch_crc64_update(crc, data, len);
 393         default:
 394                 fprintf(stderr, "Unknown checksum type %u\n", type);
 395                 exit(EXIT_FAILURE);
 396         }
 397 }
 398
 399 uint64_t bch_checksum(unsigned type, const void *data, size_t len)
 400 {
 401         uint64_t crc = 0xffffffffffffffffULL;
 402
 403         crc = bch_checksum_update(type, crc, data, len);
 404
 405         return crc ^ 0xffffffffffffffffULL;
 406 }
 407
 408 uint64_t getblocks(int fd)
 409 {
 410         uint64_t ret;
 411         struct stat statbuf;
 412         if (fstat(fd, &statbuf)) {
 413                 perror("getblocks: stat error\n");
 414                 exit(EXIT_FAILURE);
 415         }
 416         ret = statbuf.st_size / 512;
 417         if (S_ISBLK(statbuf.st_mode))
 418                 if (ioctl(fd, BLKGETSIZE, &ret)) {
 419                         perror("ioctl error getting blksize");
 420                         exit(EXIT_FAILURE);
 421                 }
 422         return ret;
 423 }
 424
 425 uint64_t hatoi(const char *s)
 426 {
 427         char *e;
 428         long long i = strtoll(s, &e, 10);
 429         switch (*e) {
 430                 case 't':
 431                 case 'T':
 432                         i *= 1024;
 433                 case 'g':
 434                 case 'G':
 435                         i *= 1024;
 436                 case 'm':
 437                 case 'M':
 438                         i *= 1024;
 439                 case 'k':
 440                 case 'K':
 441                         i *= 1024;
 442         }
 443         return i;
 444 }
 445
 446 unsigned hatoi_validate(const char *s, const char *msg)
 447 {
 448         uint64_t v = hatoi(s);
 449
 450         if (v & (v - 1)) {
 451                 fprintf(stderr, "%s must be a power of two\n", msg);
 452                 exit(EXIT_FAILURE);
 453         }
 454
 455         v /= 512;
 456
 457         if (v > USHRT_MAX) {
 458                 fprintf(stderr, "%s too large\n", msg);
 459                 exit(EXIT_FAILURE);
 460         }
 461
 462         if (!v) {
 463                 fprintf(stderr, "%s too small\n", msg);
 464                 exit(EXIT_FAILURE);
 465         }
 466
 467         return v;
 468 }
 469
 470 static void do_write_sb(int fd, struct cache_sb *sb)
 471 {
 472         char zeroes[SB_START] = {0};
 473         size_t bytes = ((void *) __bset_bkey_last(sb)) - (void *) sb;
 474
 475         /* Zero start of disk */
 476         if (pwrite(fd, zeroes, SB_START, 0) != SB_START) {
 477                 perror("write error trying to zero start of disk\n");
 478                 exit(EXIT_FAILURE);
 479         }
 480         /* Write superblock */
 481         if (pwrite(fd, sb, bytes, SB_START) != bytes) {
 482                 perror("write error trying to write superblock\n");
 483                 exit(EXIT_FAILURE);
 484         }
 485
 486         fsync(fd);
 487         close(fd);
 488 }
 489
 490 void write_backingdev_sb(int fd, unsigned block_size, unsigned *bucket_sizes,
 491                          unsigned mode, uint64_t data_offset, const char *label,
 492                          uuid_le user_uuid, uuid_le set_uuid)
 493 {
 494         char uuid_str[40], set_uuid_str[40];
 495         struct cache_sb sb;
 496
 497         memset(&sb, 0, sizeof(struct cache_sb));
 498
 499         sb.offset       = SB_SECTOR;
 500         sb.version      = BCACHE_SB_VERSION_BDEV;
 501         sb.magic        = BCACHE_MAGIC;
 502         uuid_generate(sb.disk_uuid.b);
 503         sb.user_uuid    = user_uuid;
 504         sb.set_uuid     = set_uuid;
 505         sb.block_size   = block_size;
 506
 507         uuid_unparse(sb.disk_uuid.b, uuid_str);
 508         uuid_unparse(sb.user_uuid.b, set_uuid_str);
 509         if (label)
 510                 memcpy(sb.label, label, SB_LABEL_SIZE);
 511
 512         SET_BDEV_CACHE_MODE(&sb, mode);
 513
 514         if (data_offset != BDEV_DATA_START_DEFAULT) {
 515                 sb.version = BCACHE_SB_VERSION_BDEV_WITH_OFFSET;
 516                 sb.bdev_data_offset = data_offset;
 517         }
 518
 519         sb.csum = csum_set(&sb, BCH_CSUM_CRC64);
 520
 521         printf("UUID:                   %s\n"
 522                "Set UUID:               %s\n"
 523                "version:                %u\n"
 524                "block_size:             %u\n"
 525                "data_offset:            %ju\n",
 526                uuid_str, set_uuid_str,
 527                (unsigned) sb.version,
 528                sb.block_size,
 529                data_offset);
 530
 531         do_write_sb(fd, &sb);
 532 }
 533
 534 int dev_open(const char *dev, bool wipe_bcache)
 535 {
 536         struct cache_sb sb;
 537         blkid_probe pr;
 538         int fd;
 539         char err[MAX_PATH];
 540
 541         if ((fd = open(dev, O_RDWR|O_EXCL)) == -1) {
 542                 sprintf(err, "Can't open dev %s: %s\n", dev, strerror(errno));
 543                 goto err;
 544         }
 545
 546         if (pread(fd, &sb, sizeof(sb), SB_START) != sizeof(sb)) {
 547                 sprintf(err, "Failed to read superblock");
 548                 goto err;
 549         }
 550
 551         if (!memcmp(&sb.magic, &BCACHE_MAGIC, 16) && !wipe_bcache) {
 552                 sprintf(err, "Already a bcache device on %s, "
 553                         "overwrite with --wipe-bcache\n", dev);
 554                 goto err;
 555         }
 556
 557         if (!(pr = blkid_new_probe())) {
 558                 sprintf(err, "Failed to create a new probe");
 559                 goto err;
 560         }
 561         if (blkid_probe_set_device(pr, fd, 0, 0)) {
 562                 sprintf(err, "failed to set probe to device");
 563                 goto err;
 564         }
 565         /* enable ptable probing; superblock probing is enabled by default */
 566         if (blkid_probe_enable_partitions(pr, true)) {
 567                 sprintf(err, "Failed to enable partitions on probe");
 568                 goto err;
 569         }
 570         if (!blkid_do_probe(pr)) {
 571                 /* XXX wipefs doesn't know how to remove partition tables */
 572                 sprintf(err, "Device %s already has a non-bcache superblock, "
 573                         "remove it using wipefs and wipefs -a\n", dev);
 574                 goto err;
 575         }
 576
 577         return fd;
 578
 579         err:
 580                 fprintf(stderr, "dev_open failed with: %s", err);
 581                 exit(EXIT_FAILURE);
 582 }
 583
 584 void write_cache_sbs(int *fds, struct cache_sb *sb,
 585                      unsigned block_size, unsigned *bucket_sizes,
 586                      int num_bucket_sizes, unsigned btree_node_size)
 587 {
 588         char uuid_str[40], set_uuid_str[40];
 589         size_t i;
 590
 591         if (!btree_node_size) {
 592                 btree_node_size = 512;
 593
 594                 for (i = 0; i < num_bucket_sizes; i++)
 595                         btree_node_size = min(btree_node_size,
 596                                               bucket_sizes[i]);
 597         }
 598
 599         sb->offset      = SB_SECTOR;
 600         sb->version     = BCACHE_SB_VERSION_CDEV_V3;
 601         sb->magic       = BCACHE_MAGIC;
 602         sb->block_size  = block_size;
 603         sb->u64s        = bch_journal_buckets_offset(sb);
 604
 605         /*
 606          * don't have a userspace crc32c implementation handy, just always use
 607          * crc64
 608          */
 609         SET_CACHE_SB_CSUM_TYPE(sb, BCH_CSUM_CRC64);
 610
 611         for (i = 0; i < sb->nr_in_set; i++) {
 612                 struct cache_member *m = sb->members + i;
 613
 614                 if (num_bucket_sizes <= 1) {
 615                         m->bucket_size = bucket_sizes[0];
 616                 } else {
 617                         if (!bucket_sizes[i]) {
 618                                 printf("No bucket size specified for cache %zu,"
 619                                        " using the default of %u",
 620                                        i, bucket_sizes[0]);
 621                                 m->bucket_size = bucket_sizes[0];
 622                         } else {
 623                                 m->bucket_size  = bucket_sizes[i];
 624                         }
 625                 }
 626
 627                 m->nbuckets             = getblocks(fds[i]) / m->bucket_size;
 628                 m->first_bucket         = (23 / m->bucket_size) + 1;
 629
 630                 if (m->nbuckets < 1 << 7) {
 631                         fprintf(stderr, "Not enough buckets: %llu, need %u\n",
 632                                 m->nbuckets, 1 << 7);
 633                         exit(EXIT_FAILURE);
 634                 }
 635         }
 636
 637         for (i = 0; i < sb->nr_in_set; i++) {
 638                 struct cache_member *m = sb->members + i;
 639
 640                 SET_CACHE_BTREE_NODE_SIZE(sb, btree_node_size);
 641
 642
 643                 sb->disk_uuid = m->uuid;
 644                 sb->nr_this_dev         = i;
 645
 646                 sb->csum = csum_set(sb, CACHE_SB_CSUM_TYPE(sb));
 647
 648                 uuid_unparse(sb->disk_uuid.b, uuid_str);
 649                 uuid_unparse(sb->user_uuid.b, set_uuid_str);
 650                 printf("UUID:                   %s\n"
 651                        "Set UUID:               %s\n"
 652                        "version:                %u\n"
 653                        "nbuckets:               %llu\n"
 654                        "block_size:             %u\n"
 655                        "bucket_size:            %u\n"
 656                        "nr_in_set:              %u\n"
 657                        "nr_this_dev:            %u\n"
 658                        "first_bucket:           %u\n",
 659                        uuid_str, set_uuid_str,
 660                        (unsigned) sb->version,
 661                        m->nbuckets,
 662                        sb->block_size,
 663                        m->bucket_size,
 664                        sb->nr_in_set,
 665                        sb->nr_this_dev,
 666                        m->first_bucket);
 667
 668                 do_write_sb(fds[i], sb);
 669         }
 670 }
 671
 672 void next_cache_device(struct cache_sb *sb,
 673                               unsigned replication_set,
 674                               int tier,
 675                               unsigned replacement_policy,
 676                               bool discard)
 677 {
 678         struct cache_member *m = sb->members + sb->nr_in_set;
 679
 680         SET_CACHE_REPLICATION_SET(m, replication_set);
 681         SET_CACHE_TIER(m, tier);
 682         SET_CACHE_REPLACEMENT(m, replacement_policy);
 683         SET_CACHE_DISCARD(m, discard);
 684         uuid_generate(m->uuid.b);
 685
 686         sb->nr_in_set++;
 687 }
 688
 689 unsigned get_blocksize(const char *path)
 690 {
 691         struct stat statbuf;
 692
 693         if (stat(path, &statbuf)) {
 694                 fprintf(stderr, "Error statting %s: %s\n",
 695                         path, strerror(errno));
 696                 exit(EXIT_FAILURE);
 697         }
 698
 699         if (S_ISBLK(statbuf.st_mode)) {
 700                 /* check IO limits:
 701                  * BLKALIGNOFF: alignment_offset
 702                  * BLKPBSZGET: physical_block_size
 703                  * BLKSSZGET: logical_block_size
 704                  * BLKIOMIN: minimum_io_size
 705                  * BLKIOOPT: optimal_io_size
 706                  *
 707                  * It may be tempting to use physical_block_size,
 708                  * or even minimum_io_size.
 709                  * But to be as transparent as possible,
 710                  * we want to use logical_block_size.
 711                  */
 712                 unsigned int logical_block_size;
 713                 int fd = open(path, O_RDONLY);
 714
 715                 if (fd < 0) {
 716                         fprintf(stderr, "open(%s) failed: %m\n", path);
 717                         exit(EXIT_FAILURE);
 718                 }
 719                 if (ioctl(fd, BLKSSZGET, &logical_block_size)) {
 720                         fprintf(stderr, "ioctl(%s, BLKSSZGET) failed: %m\n", path);
 721                         exit(EXIT_FAILURE);
 722                 }
 723                 close(fd);
 724                 return logical_block_size / 512;
 725
 726         }
 727         /* else: not a block device.
 728          * Why would we even want to write a bcache super block there? */
 729
 730         return statbuf.st_blksize / 512;
 731 }
 732
 733 long strtoul_or_die(const char *p, size_t max, const char *msg)
 734 {
 735         errno = 0;
 736         long v = strtol(p, NULL, 10);
 737         if (errno || v < 0 || v >= max) {
 738                 fprintf(stderr, "Invalid %s %zi\n", msg, v);
 739                 exit(EXIT_FAILURE);
 740         }
 741
 742         return v;
 743 }
 744
 745 static void print_encode(char *in)
 746 {
 747     char *pos;
 748         for (pos = in; *pos; pos++)
 749                 if (isalnum(*pos) || strchr(".-_", *pos))
 750                         putchar(*pos);
 751                 else
 752                         printf("%%%x", *pos);
 753 }
 754
 755 static void show_uuid_only(struct cache_sb *sb, char *dev_uuid) {
 756         uuid_unparse(sb->disk_uuid.b, dev_uuid);
 757 }
 758
 759 static void show_super_common(struct cache_sb *sb, bool force_csum)
 760 {
 761         char uuid[40];
 762         char label[SB_LABEL_SIZE + 1];
 763         uint64_t expected_csum;
 764
 765         printf("sb.magic\t\t");
 766         if (!memcmp(&sb->magic, &BCACHE_MAGIC, sizeof(sb->magic))) {
 767                 printf("ok\n");
 768         } else {
 769                 printf("bad magic\n");
 770                 fprintf(stderr, "Invalid superblock (bad magic)\n");
 771                 exit(2);
 772         }
 773
 774         printf("sb.first_sector\t\t%ju", (uint64_t) sb->offset);
 775         if (sb->offset == SB_SECTOR) {
 776                 printf(" [match]\n");
 777         } else {
 778                 printf(" [expected %ds]\n", SB_SECTOR);
 779                 fprintf(stderr, "Invalid superblock (bad sector)\n");
 780                 exit(2);
 781         }
 782
 783         printf("sb.csum\t\t\t%ju", (uint64_t) sb->csum);
 784         expected_csum = csum_set(sb,
 785                                  sb->version < BCACHE_SB_VERSION_CDEV_V3
 786                                  ? BCH_CSUM_CRC64
 787                                  : CACHE_SB_CSUM_TYPE(sb));
 788         if (sb->csum == expected_csum) {
 789                 printf(" [match]\n");
 790         } else {
 791                 printf(" [expected %" PRIX64 "]\n", expected_csum);
 792                 if (force_csum) {
 793                         fprintf(stderr, "Corrupt superblock (bad csum)\n");
 794                         exit(2);
 795                 }
 796         }
 797
 798         printf("sb.version\t\t%ju", (uint64_t) sb->version);
 799         switch (sb->version) {
 800                 // These are handled the same by the kernel
 801                 case BCACHE_SB_VERSION_CDEV:
 802                 case BCACHE_SB_VERSION_CDEV_WITH_UUID:
 803                         printf(" [cache device]\n");
 804                         break;
 805
 806                 // The second adds data offset support
 807                 case BCACHE_SB_VERSION_BDEV:
 808                 case BCACHE_SB_VERSION_BDEV_WITH_OFFSET:
 809                         printf(" [backing device]\n");
 810                         break;
 811
 812                 default:
 813                         printf(" [unknown]\n");
 814                         // exit code?
 815                         exit(EXIT_SUCCESS);
 816         }
 817
 818         putchar('\n');
 819
 820         strncpy(label, (char *) sb->label, SB_LABEL_SIZE);
 821         label[SB_LABEL_SIZE] = '\0';
 822         printf("dev.label\t\t");
 823         if (*label)
 824                 print_encode(label);
 825         else
 826                 printf("(empty)");
 827         putchar('\n');
 828
 829         uuid_unparse(sb->disk_uuid.b, uuid);
 830         printf("dev.uuid\t\t%s\n", uuid);
 831
 832         uuid_unparse(sb->user_uuid.b, uuid);
 833         printf("cset.uuid\t\t%s\n", uuid);
 834
 835         uuid_unparse(sb->set_uuid.b, uuid);
 836         printf("internal.uuid\t%s\n", uuid);
 837
 838 }
 839
 840 void show_super_backingdev(struct cache_sb *sb, bool force_csum)
 841 {
 842         uint64_t first_sector;
 843
 844         show_super_common(sb, force_csum);
 845
 846         if (sb->version == BCACHE_SB_VERSION_BDEV) {
 847                 first_sector = BDEV_DATA_START_DEFAULT;
 848         } else {
 849                 first_sector = sb->bdev_data_offset;
 850         }
 851
 852         printf("dev.data.first_sector\t%ju\n"
 853                "dev.data.cache_mode\t%s"
 854                "dev.data.cache_state\t%s\n",
 855                first_sector,
 856                bdev_cache_mode[BDEV_CACHE_MODE(sb)],
 857                bdev_state[BDEV_STATE(sb)]);
 858 }
 859
 860 void show_super_cache(struct cache_sb *sb, bool force_csum)
 861 {
 862         struct cache_member *m = sb->members + sb->nr_this_dev;
 863         char uuid[16];
 864
 865         show_super_common(sb, force_csum);
 866
 867         uuid_unparse(m->uuid.b, uuid);
 868         printf("dev.cache.uuid\t%s\n", uuid);
 869
 870         printf("dev.sectors_per_block\t%u\n"
 871                "dev.sectors_per_bucket\t%u\n",
 872                sb->block_size,
 873                m->bucket_size);
 874
 875         // total_sectors includes the superblock;
 876         printf("dev.cache.first_sector\t%u\n"
 877                "dev.cache.cache_sectors\t%llu\n"
 878                "dev.cache.total_sectors\t%llu\n"
 879                "dev.cache.ordered\t%s\n"
 880                "dev.cache.pos\t\t%u\n"
 881                "dev.cache.setsize\t\t%u\n",
 882                m->bucket_size * m->first_bucket,
 883                m->bucket_size * (m->nbuckets - m->first_bucket),
 884                m->bucket_size * m->nbuckets,
 885                CACHE_SYNC(sb) ? "yes" : "no",
 886                sb->nr_this_dev,
 887                sb->nr_in_set);
 888
 889         printf("cache.state\t%s\n",             cache_state[CACHE_STATE(m)]);
 890
 891         printf("cache.tier\t%llu\n",            CACHE_TIER(m));
 892
 893         printf("cache.replication_set\t%llu\n", CACHE_REPLICATION_SET(m));
 894         printf("cache.cur_meta_replicas\t%llu\n", REPLICATION_SET_CUR_META_REPLICAS(m));
 895         printf("cache.cur_data_replicas\t%llu\n", REPLICATION_SET_CUR_DATA_REPLICAS(m));
 896
 897         printf("cache.has_metadata\t%llu\n",    CACHE_HAS_METADATA(m));
 898         printf("cache.has_data\t%llu\n",        CACHE_HAS_DATA(m));
 899
 900         printf("cache.replacement\t%s\n",       replacement_policies[CACHE_REPLACEMENT(m)]);
 901         printf("cache.discard\t%llu\n",         CACHE_DISCARD(m));
 902 }
 903
 904 static int __sysfs_attr_type(const char *attr, const char * const *attr_arr)
 905 {
 906         for (unsigned i = 0; attr_arr[i] != NULL; i++)
 907                 if(!strcmp(attr, attr_arr[i]))
 908                         return 1;
 909         return 0;
 910 }
 911
 912 enum sysfs_attr sysfs_attr_type(const char *attr)
 913 {
 914         if(__sysfs_attr_type(attr, set_attrs))
 915                 return SET_ATTR;
 916         if(__sysfs_attr_type(attr, cache_attrs))
 917                 return CACHE_ATTR;
 918         if(__sysfs_attr_type(attr, internal_attrs))
 919                 return INTERNAL_ATTR;
 920
 921         printf("No attribute called %s, try --list to see options\n", attr);
 922
 923         return -1;
 924 }
 925
 926 static void __sysfs_attr_list(const char * const *attr_arr)
 927 {
 928         for (unsigned i = 0; attr_arr[i] != NULL; i++)
 929                 printf("%s\n", attr_arr[i]);
 930 }
 931
 932 void sysfs_attr_list()
 933 {
 934         __sysfs_attr_list(set_attrs);
 935         __sysfs_attr_list(cache_attrs);
 936         __sysfs_attr_list(internal_attrs);
 937 }
 938
 939 struct cache_sb *query_dev(char *dev, bool force_csum,
 940                 bool print_sb, bool uuid_only, char *dev_uuid)
 941 {
 942         size_t bytes = 4096;
 943         struct cache_sb *sb = aligned_alloc(bytes, bytes);
 944
 945         int fd = open(dev, O_RDONLY|O_DIRECT);
 946         if (fd < 0) {
 947                 printf("Can't open dev %s: %s\n", dev, strerror(errno));
 948                 return NULL;
 949         }
 950
 951         while (true) {
 952                 int ret = pread(fd, sb, bytes, SB_START);
 953                 if (ret < 0) {
 954                         fprintf(stderr, "Couldn't read superblock: %s\n",
 955                                         strerror(errno));
 956                         close(fd);
 957                         free(sb);
 958                         return NULL;
 959                 } else if (bytes > sizeof(sb) + sb->u64s * sizeof(u64)) {
 960                         /* We read the whole superblock */
 961                         break;
 962                 }
 963
 964                 /*
 965                  * otherwise double the size of our dest
 966                  * and read again
 967                  */
 968                 free(sb);
 969                 bytes *= 2;
 970                 sb = aligned_alloc(4096, bytes);
 971         }
 972
 973         close(fd);
 974
 975         if(uuid_only) {
 976                 show_uuid_only(sb, dev_uuid);
 977                 return sb;
 978         }
 979
 980         if(print_sb) {
 981                 if (!SB_IS_BDEV(sb))
 982                         show_super_cache(sb, force_csum);
 983                 else
 984                         show_super_backingdev(sb, force_csum);
 985         }
 986
 987         return sb;
 988 }
 989
 990 char *dev_name(const char *ugly_path) {
 991         char buf[32];
 992         int i, end = strlen(ugly_path);
 993
 994         //Chop off "/bcache", then look for the next '/' from the end
 995         for (i = end - 8; ; i--)
 996                 if(ugly_path[i] == '/')
 997                         break;
 998
 999         strcpy(buf, ugly_path + i);
1000         buf[end - i - 7] = 0;
1001
1002         // Is the dev guaranteed to be in /dev?
1003         // This is needed for finding the superblock with a query-dev
1004         return strdup(buf);
1005 }
1006
1007 static void list_cacheset_devs(char *cset_dir, char *cset_name, bool parse_dev_name)
1008 {
1009         DIR *cachedir, *dir;
1010         struct stat cache_stat;
1011         char entry[MAX_PATH];
1012         struct dirent *ent;
1013         snprintf(entry, MAX_PATH, "%s/%s", cset_dir, cset_name);
1014
1015         if((dir = opendir(entry)) != NULL) {
1016                 while((ent = readdir(dir)) != NULL) {
1017                         char buf[MAX_PATH];
1018                         int len;
1019                         char *tmp;
1020
1021                         /*
1022                          * We are looking for all cache# directories
1023                          * do a strlen < 9 to skip over other entries
1024                          * that also start with "cache"
1025                          */
1026                         if(strncmp(ent->d_name, "cache", 5) ||
1027                                         !(strlen(ent->d_name) < 9))
1028                                 continue;
1029
1030                         snprintf(entry, MAX_PATH, "%s/%s/%s",
1031                                         cset_dir,
1032                                         cset_name,
1033                                         ent->d_name);
1034
1035                         if((cachedir = opendir(entry)) == NULL)
1036                                 continue;
1037
1038                         if(stat(entry, &cache_stat))
1039                                 continue;
1040
1041                         if((len = readlink(entry, buf, sizeof(buf) - 1)) !=
1042                                         -1) {
1043                                 buf[len] = '\0';
1044                                 if(parse_dev_name) {
1045                                         tmp = dev_name(buf);
1046                                         printf("/dev%s\n", tmp);
1047                                         free(tmp);
1048                                 } else {
1049                                         printf("\t%s\n", buf);
1050                                 }
1051                         }
1052                 }
1053         }
1054 }
1055
1056 char *find_matching_uuid(char *stats_dir, char *subdir, const char *stats_dev_uuid)
1057 {
1058         /* Do a query-dev --uuid only to get the uuid
1059          * repeat on each dev until we find a matching one
1060          * append that cache# to subdir and return
1061          */
1062
1063         int i = 0;
1064         DIR *cachedir;
1065         struct stat cache_stat;
1066         char intbuf[4];
1067         char entry[MAX_PATH];
1068         char *err = NULL;
1069
1070         snprintf(entry, MAX_PATH, "%s%s", stats_dir, subdir);
1071         snprintf(intbuf, 4, "%d", i);
1072         strcat(entry, intbuf);
1073
1074         while(true) {
1075                 char buf[MAX_PATH];
1076                 int len;
1077
1078                 if((cachedir = opendir(entry)) == NULL)
1079                         break;
1080
1081                 if(stat(entry, &cache_stat))
1082                         break;
1083
1084                 if((len = readlink(entry, buf, sizeof(buf) - 1)) != -1) {
1085                         char dev_uuid[40];
1086                         buf[len] = '\0';
1087                         int i, end = strlen(buf);
1088                         char tmp[32], devname[32];
1089                         struct cache_sb *sb;
1090
1091                         /* Chop off "/bcache", then look for the
1092                          * next '/' from the end
1093                          */
1094                         for (i = end - 8; ; i--)
1095                                 if(buf[i] == '/')
1096                                         break;
1097
1098                         strcpy(tmp, buf + i);
1099                         tmp[end - i - 7] = 0;
1100                         strcpy(devname, "/dev");
1101                         strcat(devname, tmp);
1102
1103                         err = "Unable to open superblock";
1104                         sb = query_dev(devname, false, false, true, dev_uuid);
1105                         if(!sb)
1106                                 return err;
1107                         else
1108                                 free(sb);
1109
1110                         if(!strcmp(stats_dev_uuid, dev_uuid)) {
1111                                 strcat(subdir, intbuf);
1112                                 return NULL;
1113                         }
1114                 }
1115
1116                 /* remove i from end and append i++ */
1117                 entry[strlen(entry)-strlen(intbuf)] = 0;
1118                 i++;
1119                 snprintf(intbuf, 4, "%d", i);
1120                 strcat(entry, intbuf);
1121         }
1122
1123
1124         err = "dev uuid doesn't exist in cache_set";
1125         return err;
1126 }
1127
1128 char *list_cachesets(char *cset_dir, bool list_devs)
1129 {
1130         struct dirent *ent;
1131         DIR *dir;
1132         char *err = NULL;
1133
1134         dir = opendir(cset_dir);
1135         if (!dir) {
1136                 err = "Failed to open cacheset dir";
1137                 goto err;
1138         }
1139
1140         while ((ent = readdir(dir)) != NULL) {
1141                 struct stat statbuf;
1142                 char entry[MAX_PATH];
1143
1144                 if (!strcmp(ent->d_name, ".") || !strcmp(ent->d_name, ".."))
1145                         continue;
1146
1147                 snprintf(entry, MAX_PATH, "%s/%s", cset_dir, ent->d_name);
1148                 if(stat(entry, &statbuf) == -1) {
1149                         err = "Failed to stat cacheset subdir";
1150                         goto err;
1151                 }
1152
1153                 if (S_ISDIR(statbuf.st_mode)) {
1154                         printf("%s\n", ent->d_name);
1155
1156                         if(list_devs) {
1157                                 list_cacheset_devs(cset_dir, ent->d_name, true);
1158                         }
1159                 }
1160         }
1161
1162 err:
1163         closedir(dir);
1164         return err;
1165 }
1166
1167 char *register_bcache(char *const *devs)
1168 {
1169         int ret, bcachefd;
1170         char *err = NULL;
1171
1172         bcachefd = open("/dev/bcache", O_RDWR);
1173         if (bcachefd < 0) {
1174                 err = "Can't open bcache device";
1175                 goto err;
1176         }
1177
1178         ret = ioctl(bcachefd, BCH_IOCTL_REGISTER, devs);
1179         if (ret < 0) {
1180                 char tmp[64];
1181                 snprintf(tmp, 64, "ioctl register error: %s\n",
1182                                 strerror(ret));
1183                 err = strdup(tmp);
1184                 goto err;
1185         }
1186
1187 err:
1188         if (bcachefd)
1189                 close(bcachefd);
1190         return err;
1191 }
1192
1193 char *unregister_bcache(char *const *devs)
1194 {
1195         int ret, bcachefd;
1196         char *err = NULL;
1197
1198         bcachefd = open("/dev/bcache_extent0", O_RDWR);
1199         if (bcachefd < 0) {
1200                 err = "Can't open bcache device";
1201                 goto err;
1202         }
1203
1204         ret = ioctl(bcachefd, BCH_IOCTL_STOP);
1205         if (ret < 0) {
1206                 char tmp[64];
1207                 snprintf(tmp, 64, "ioctl unregister error: %s\n",
1208                                 strerror(ret));
1209                 err = strdup(tmp);
1210                 goto err;
1211         }
1212
1213 err:
1214         close(bcachefd);
1215         return err;
1216 }
1217
1218 char *add_devices(char *const *devs)
1219 {
1220         int ret, bcachefd;
1221         char *err = NULL;
1222
1223         bcachefd = open("/dev/bcache_extent0", O_RDWR);
1224         if (bcachefd < 0) {
1225                 err = "Can't open bcache device";
1226                 goto err;
1227         }
1228
1229         struct bch_ioctl_add_disks ia;
1230         ia.devs = devs;
1231
1232         ret = ioctl(bcachefd, BCH_IOCTL_ADD_DISKS, &ia);
1233         if (ret < 0) {
1234                 char tmp[128];
1235                 snprintf(tmp, 128, "ioctl add disk error: %s\n",
1236                                 strerror(ret));
1237                 err = strdup(tmp);
1238         }
1239
1240 err:
1241         close(bcachefd);
1242         return err;
1243 }
1244
1245 char *remove_device(const char *dev, bool force)
1246 {
1247         int ret, bcachefd;
1248         char *err = NULL;
1249
1250         bcachefd = open("/dev/bcache_extent0", O_RDWR);
1251         if (bcachefd < 0) {
1252                 err = "Can't open bcache device";
1253                 goto err;
1254         }
1255
1256         struct bch_ioctl_rm_disk ir;
1257         ir.dev = dev;
1258         ir.force = force ? 1 : 0;
1259
1260         ret = ioctl(bcachefd, BCH_IOCTL_RM_DISK, &ir);
1261         if (ret < 0) {
1262                 char tmp[128];
1263                 snprintf(tmp, 128, "ioctl add disk error: %s\n",
1264                                 strerror(ret));
1265                 err = strdup(tmp);
1266         }
1267
1268 err:
1269         close(bcachefd);
1270         return err;
1271 }
1272
1273 char *device_set_failed(const char *dev_uuid) {
1274         int ret, bcachefd;
1275         char *err = NULL;
1276         uuid_le dev;
1277         struct bch_ioctl_disk_failed df;
1278
1279         bcachefd = open("/dev/bcache_extent0", O_RDWR);
1280         if (bcachefd < 0) {
1281                 err = "Can't open bcache device";
1282                 goto err;
1283         }
1284
1285         uuid_parse(dev_uuid, dev.b);
1286         df.dev_uuid = dev;
1287
1288         ret = ioctl(bcachefd, BCH_IOCTL_SET_DISK_FAILED, &df);
1289         if (ret < 0) {
1290                 char tmp[128];
1291                 snprintf(tmp, 128, "ioctl set disk failed error %s\n",
1292                                 strerror(ret));
1293                 err = strdup(tmp);
1294         }
1295
1296 err:
1297         close(bcachefd);
1298         return err;
1299 }
1300
1301 char *probe(char *dev, int udev)
1302 {
1303         struct cache_sb sb;
1304         char uuid[40];
1305         blkid_probe pr;
1306         char *err = NULL;
1307
1308         int fd = open(dev, O_RDONLY);
1309         if (fd == -1) {
1310                 err = "Got file descriptor -1 trying to open dev";
1311                 goto err;
1312         }
1313
1314         if (!(pr = blkid_new_probe())) {
1315                 err = "Failed trying to get a blkid for new probe";
1316                 goto err;
1317         }
1318
1319         if (blkid_probe_set_device(pr, fd, 0, 0)) {
1320                 err = "Failed blkid probe set device";
1321                 goto err;
1322         }
1323
1324         /* probe partitions too */
1325         if (blkid_probe_enable_partitions(pr, true)) {
1326                 err = "Enable probe partitions";
1327                 goto err;
1328         }
1329
1330         /* bail if anything was found
1331          * probe-bcache isn't needed once blkid recognizes bcache */
1332         if (!blkid_do_probe(pr)) {
1333                 err = "blkid recognizes bcache";
1334                 goto err;
1335         }
1336
1337         if (pread(fd, &sb, sizeof(sb), SB_START) != sizeof(sb)) {
1338                 err = "Failed to read superblock";
1339                 goto err;
1340         }
1341
1342         if (memcmp(&sb.magic, &BCACHE_MAGIC, sizeof(sb.magic))) {
1343                 err = "Bcache magic incorrect";
1344                 goto err;
1345         }
1346
1347         uuid_unparse(sb.disk_uuid.b, uuid);
1348
1349         if (udev)
1350                 printf("ID_FS_UUID=%s\n"
1351                        "ID_FS_UUID_ENC=%s\n"
1352                        "ID_FS_TYPE=bcache\n",
1353                        uuid, uuid);
1354         else
1355                 printf("%s: UUID=\"\" TYPE=\"bcache\"\n", uuid);
1356
1357         return 0;
1358
1359 err:
1360         return err;
1361 }
1362
1363 char *read_stat_dir(DIR *dir, char *stats_dir, char *stat_name, char *ret)
1364 {
1365         struct stat statbuf;
1366         char entry[MAX_PATH];
1367         char *err = NULL;
1368
1369         snprintf(entry, MAX_PATH, "%s/%s", stats_dir, stat_name);
1370         if(stat(entry, &statbuf) == -1) {
1371                 char tmp[MAX_PATH];
1372                 snprintf(tmp, MAX_PATH, "Failed to stat %s\n", entry);
1373                 err = strdup(tmp);
1374                 goto err;
1375         }
1376
1377         if (S_ISREG(statbuf.st_mode)) {
1378                 FILE *fp = NULL;
1379
1380                 fp = fopen(entry, "r");
1381                 if(!fp) {
1382                         /* If we can't open the file, this is probably because
1383                          * of permissions, just move to the next file */
1384                         return NULL;
1385                 }
1386
1387                 while(fgets(ret, MAX_PATH, fp));
1388                 fclose(fp);
1389         }
1390 err:
1391         return err;
1392 }
1393
1394 char *bcache_get_capacity(const char *cset_dir, const char *capacity_uuid,
1395                           bool show_devs)
1396 {
1397         char *err = NULL;
1398         char bucket_size_path[MAX_PATH];
1399         char nbuckets_path[MAX_PATH];
1400         char avail_buckets_path[MAX_PATH];
1401         char cache_path[MAX_PATH];
1402
1403         double bucket_sizes[MAX_DEVS];
1404         double nbuckets[MAX_DEVS];
1405         double avail_buckets[MAX_DEVS];
1406         char *dev_names[MAX_DEVS];
1407         int dev_count = 0, i;
1408         double total_cap = 0, total_free = 0;
1409         int precision = 2;
1410
1411
1412         while (true) {
1413                 char buf[MAX_PATH];
1414                 int len;
1415                 DIR *cache_dir;
1416
1417                 snprintf(bucket_size_path, sizeof(bucket_size_path), "%s/%s/%s%d/%s", cset_dir,
1418                                 capacity_uuid, "cache", dev_count, "bucket_size_bytes");
1419                 snprintf(nbuckets_path, sizeof(nbuckets_path), "%s/%s/%s%d/%s", cset_dir,
1420                                 capacity_uuid, "cache", dev_count, "nbuckets");
1421                 snprintf(avail_buckets_path, sizeof(avail_buckets_path), "%s/%s/%s%d/%s", cset_dir,
1422                                 capacity_uuid, "cache", dev_count, "available_buckets");
1423                 snprintf(cache_path, sizeof(cache_path), "%s/%s/%s%d", cset_dir, capacity_uuid,
1424                                 "cache", dev_count);
1425
1426                 if((cache_dir = opendir(cache_path)) == NULL)
1427                         break;
1428
1429                 err = read_stat_dir(cache_dir, cache_path,
1430                                 "bucket_size_bytes", buf);
1431                 if (err)
1432                         goto err;
1433                 else
1434                         bucket_sizes[dev_count] = atof(buf);
1435
1436                 err = read_stat_dir(cache_dir, cache_path,
1437                                 "nbuckets", buf);
1438                 if (err)
1439                         goto err;
1440                 else
1441                         nbuckets[dev_count] = atof(buf);
1442
1443                 err = read_stat_dir(cache_dir, cache_path,
1444                                 "available_buckets", buf);
1445                 if (err)
1446                         goto err;
1447                 else
1448                         avail_buckets[dev_count] = atof(buf);
1449
1450                 if((len = readlink(cache_path, buf, sizeof(buf) - 1)) != -1) {
1451                         buf[len] = '\0';
1452                         dev_names[dev_count] = dev_name(buf);
1453                 }
1454
1455                 dev_count++;
1456         }
1457
1458         printf("%-15s%-25s%-25s\n", "Device Name", "Capacity (512 Blocks)", "Free (512 Blocks)");
1459
1460         if (show_devs) {
1461                 for (i = 0; i < dev_count; i++) {
1462                         printf("%s%-11s%-25.*f%-25.*f\n", "/dev", dev_names[i],
1463                                 precision,
1464                                 (bucket_sizes[i] * nbuckets[i]) / 512,
1465                                 precision,
1466                                 (bucket_sizes[i] * avail_buckets[i]) / 512);
1467                 }
1468         }
1469
1470         for (i = 0; i < dev_count; i++) {
1471                 total_cap += (bucket_sizes[i] * nbuckets[i]) / 512;
1472                 total_free += (bucket_sizes[i] * avail_buckets[i]) / 512;
1473
1474         }
1475
1476         printf("%-15s%-25.*f%-25.*f\n", "Total", precision, total_cap,
1477                         precision, total_free);
1478
1479 err:
1480         for (i = 0; i < dev_count; i++)
1481                 if (dev_names[i])
1482                         free(dev_names[i]);
1483         return err;
1484 }