Ext2FileSystem.cpp 70 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791
  1. /*
  2. * Copyright (c) 2018-2020, Andreas Kling <kling@serenityos.org>
  3. * Copyright (c) 2021, sin-ack <sin-ack@protonmail.com>
  4. *
  5. * SPDX-License-Identifier: BSD-2-Clause
  6. */
  7. #include <AK/HashMap.h>
  8. #include <AK/MemoryStream.h>
  9. #include <AK/StdLibExtras.h>
  10. #include <AK/StringView.h>
  11. #include <Kernel/Debug.h>
  12. #include <Kernel/Devices/BlockDevice.h>
  13. #include <Kernel/FileSystem/Ext2FileSystem.h>
  14. #include <Kernel/FileSystem/FileDescription.h>
  15. #include <Kernel/FileSystem/ext2_fs.h>
  16. #include <Kernel/Process.h>
  17. #include <Kernel/UnixTypes.h>
  18. #include <LibC/errno_numbers.h>
  19. namespace Kernel {
  20. static const size_t max_link_count = 65535;
  21. static const size_t max_block_size = 4096;
  22. static const ssize_t max_inline_symlink_length = 60;
  23. struct Ext2FSDirectoryEntry {
  24. String name;
  25. InodeIndex inode_index { 0 };
  26. u8 file_type { 0 };
  27. u16 record_length { 0 };
  28. };
  29. static u8 to_ext2_file_type(mode_t mode)
  30. {
  31. if (is_regular_file(mode))
  32. return EXT2_FT_REG_FILE;
  33. if (is_directory(mode))
  34. return EXT2_FT_DIR;
  35. if (is_character_device(mode))
  36. return EXT2_FT_CHRDEV;
  37. if (is_block_device(mode))
  38. return EXT2_FT_BLKDEV;
  39. if (is_fifo(mode))
  40. return EXT2_FT_FIFO;
  41. if (is_socket(mode))
  42. return EXT2_FT_SOCK;
  43. if (is_symlink(mode))
  44. return EXT2_FT_SYMLINK;
  45. return EXT2_FT_UNKNOWN;
  46. }
  47. static unsigned divide_rounded_up(unsigned a, unsigned b)
  48. {
  49. return (a / b) + (a % b != 0);
  50. }
  51. NonnullRefPtr<Ext2FS> Ext2FS::create(FileDescription& file_description)
  52. {
  53. return adopt_ref(*new Ext2FS(file_description));
  54. }
  55. Ext2FS::Ext2FS(FileDescription& file_description)
  56. : BlockBasedFS(file_description)
  57. {
  58. }
  59. Ext2FS::~Ext2FS()
  60. {
  61. }
  62. bool Ext2FS::flush_super_block()
  63. {
  64. Locker locker(m_lock);
  65. VERIFY((sizeof(ext2_super_block) % logical_block_size()) == 0);
  66. auto super_block_buffer = UserOrKernelBuffer::for_kernel_buffer((u8*)&m_super_block);
  67. bool success = raw_write_blocks(2, (sizeof(ext2_super_block) / logical_block_size()), super_block_buffer);
  68. VERIFY(success);
  69. return true;
  70. }
  71. const ext2_group_desc& Ext2FS::group_descriptor(GroupIndex group_index) const
  72. {
  73. // FIXME: Should this fail gracefully somehow?
  74. VERIFY(group_index <= m_block_group_count);
  75. VERIFY(group_index > 0);
  76. return block_group_descriptors()[group_index.value() - 1];
  77. }
  78. bool Ext2FS::initialize()
  79. {
  80. Locker locker(m_lock);
  81. VERIFY((sizeof(ext2_super_block) % logical_block_size()) == 0);
  82. auto super_block_buffer = UserOrKernelBuffer::for_kernel_buffer((u8*)&m_super_block);
  83. bool success = raw_read_blocks(2, (sizeof(ext2_super_block) / logical_block_size()), super_block_buffer);
  84. VERIFY(success);
  85. auto& super_block = this->super_block();
  86. if constexpr (EXT2_DEBUG) {
  87. dmesgln("Ext2FS: super block magic: {:04x} (super block size: {})", super_block.s_magic, sizeof(ext2_super_block));
  88. }
  89. if (super_block.s_magic != EXT2_SUPER_MAGIC)
  90. return false;
  91. if constexpr (EXT2_DEBUG) {
  92. dmesgln("Ext2FS: {} inodes, {} blocks", super_block.s_inodes_count, super_block.s_blocks_count);
  93. dmesgln("Ext2FS: Block size: {}", EXT2_BLOCK_SIZE(&super_block));
  94. dmesgln("Ext2FS: First data block: {}", super_block.s_first_data_block);
  95. dmesgln("Ext2FS: Inodes per block: {}", inodes_per_block());
  96. dmesgln("Ext2FS: Inodes per group: {}", inodes_per_group());
  97. dmesgln("Ext2FS: Free inodes: {}", super_block.s_free_inodes_count);
  98. dmesgln("Ext2FS: Descriptors per block: {}", EXT2_DESC_PER_BLOCK(&super_block));
  99. dmesgln("Ext2FS: Descriptor size: {}", EXT2_DESC_SIZE(&super_block));
  100. }
  101. set_block_size(EXT2_BLOCK_SIZE(&super_block));
  102. set_fragment_size(EXT2_FRAG_SIZE(&super_block));
  103. VERIFY(block_size() <= (int)max_block_size);
  104. m_block_group_count = ceil_div(super_block.s_blocks_count, super_block.s_blocks_per_group);
  105. if (m_block_group_count == 0) {
  106. dmesgln("Ext2FS: no block groups :(");
  107. return false;
  108. }
  109. unsigned blocks_to_read = ceil_div(m_block_group_count * sizeof(ext2_group_desc), block_size());
  110. BlockIndex first_block_of_bgdt = block_size() == 1024 ? 2 : 1;
  111. m_cached_group_descriptor_table = KBuffer::try_create_with_size(block_size() * blocks_to_read, Region::Access::Read | Region::Access::Write, "Ext2FS: Block group descriptors");
  112. if (!m_cached_group_descriptor_table) {
  113. dbgln("Ext2FS: Failed to allocate memory for group descriptor table");
  114. return false;
  115. }
  116. auto buffer = UserOrKernelBuffer::for_kernel_buffer(m_cached_group_descriptor_table->data());
  117. if (auto result = read_blocks(first_block_of_bgdt, blocks_to_read, buffer); result.is_error()) {
  118. // FIXME: Propagate the error
  119. dbgln("Ext2FS: initialize had error: {}", result.error());
  120. return false;
  121. }
  122. if constexpr (EXT2_DEBUG) {
  123. for (unsigned i = 1; i <= m_block_group_count; ++i) {
  124. auto& group = group_descriptor(i);
  125. dbgln("Ext2FS: group[{}] ( block_bitmap: {}, inode_bitmap: {}, inode_table: {} )", i, group.bg_block_bitmap, group.bg_inode_bitmap, group.bg_inode_table);
  126. }
  127. }
  128. return true;
  129. }
  130. const char* Ext2FS::class_name() const
  131. {
  132. return "Ext2FS";
  133. }
  134. NonnullRefPtr<Inode> Ext2FS::root_inode() const
  135. {
  136. return *get_inode({ fsid(), EXT2_ROOT_INO });
  137. }
  138. bool Ext2FS::find_block_containing_inode(InodeIndex inode, BlockIndex& block_index, unsigned& offset) const
  139. {
  140. auto& super_block = this->super_block();
  141. if (inode != EXT2_ROOT_INO && inode < EXT2_FIRST_INO(&super_block))
  142. return false;
  143. if (inode > super_block.s_inodes_count)
  144. return false;
  145. auto& bgd = group_descriptor(group_index_from_inode(inode));
  146. offset = ((inode.value() - 1) % inodes_per_group()) * inode_size();
  147. block_index = bgd.bg_inode_table + (offset >> EXT2_BLOCK_SIZE_BITS(&super_block));
  148. offset &= block_size() - 1;
  149. return true;
  150. }
  151. Ext2FS::BlockListShape Ext2FS::compute_block_list_shape(unsigned blocks) const
  152. {
  153. BlockListShape shape;
  154. const unsigned entries_per_block = EXT2_ADDR_PER_BLOCK(&super_block());
  155. unsigned blocks_remaining = blocks;
  156. shape.direct_blocks = min((unsigned)EXT2_NDIR_BLOCKS, blocks_remaining);
  157. blocks_remaining -= shape.direct_blocks;
  158. if (!blocks_remaining)
  159. return shape;
  160. shape.indirect_blocks = min(blocks_remaining, entries_per_block);
  161. shape.meta_blocks += 1;
  162. blocks_remaining -= shape.indirect_blocks;
  163. if (!blocks_remaining)
  164. return shape;
  165. shape.doubly_indirect_blocks = min(blocks_remaining, entries_per_block * entries_per_block);
  166. shape.meta_blocks += 1;
  167. shape.meta_blocks += divide_rounded_up(shape.doubly_indirect_blocks, entries_per_block);
  168. blocks_remaining -= shape.doubly_indirect_blocks;
  169. if (!blocks_remaining)
  170. return shape;
  171. shape.triply_indirect_blocks = min(blocks_remaining, entries_per_block * entries_per_block * entries_per_block);
  172. shape.meta_blocks += 1;
  173. shape.meta_blocks += divide_rounded_up(shape.triply_indirect_blocks, entries_per_block * entries_per_block);
  174. shape.meta_blocks += divide_rounded_up(shape.triply_indirect_blocks, entries_per_block);
  175. blocks_remaining -= shape.triply_indirect_blocks;
  176. VERIFY(blocks_remaining == 0);
  177. return shape;
  178. }
  179. KResult Ext2FSInode::write_indirect_block(BlockBasedFS::BlockIndex block, Span<BlockBasedFS::BlockIndex> blocks_indices)
  180. {
  181. const auto entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  182. VERIFY(blocks_indices.size() <= entries_per_block);
  183. auto block_contents = ByteBuffer::create_uninitialized(fs().block_size());
  184. OutputMemoryStream stream { block_contents };
  185. auto buffer = UserOrKernelBuffer::for_kernel_buffer(stream.data());
  186. VERIFY(blocks_indices.size() <= EXT2_ADDR_PER_BLOCK(&fs().super_block()));
  187. for (unsigned i = 0; i < blocks_indices.size(); ++i)
  188. stream << static_cast<u32>(blocks_indices[i].value());
  189. stream.fill_to_end(0);
  190. return fs().write_block(block, buffer, stream.size());
  191. }
  192. KResult Ext2FSInode::grow_doubly_indirect_block(BlockBasedFS::BlockIndex block, size_t old_blocks_length, Span<BlockBasedFS::BlockIndex> blocks_indices, Vector<Ext2FS::BlockIndex>& new_meta_blocks, unsigned& meta_blocks)
  193. {
  194. const auto entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  195. const auto entries_per_doubly_indirect_block = entries_per_block * entries_per_block;
  196. const auto old_indirect_blocks_length = divide_rounded_up(old_blocks_length, entries_per_block);
  197. const auto new_indirect_blocks_length = divide_rounded_up(blocks_indices.size(), entries_per_block);
  198. VERIFY(blocks_indices.size() > 0);
  199. VERIFY(blocks_indices.size() > old_blocks_length);
  200. VERIFY(blocks_indices.size() <= entries_per_doubly_indirect_block);
  201. auto block_contents = ByteBuffer::create_uninitialized(fs().block_size());
  202. auto* block_as_pointers = (unsigned*)block_contents.data();
  203. OutputMemoryStream stream { block_contents };
  204. auto buffer = UserOrKernelBuffer::for_kernel_buffer(stream.data());
  205. if (old_blocks_length > 0) {
  206. if (auto result = fs().read_block(block, &buffer, fs().block_size()); result.is_error())
  207. return result;
  208. }
  209. // Grow the doubly indirect block.
  210. for (unsigned i = 0; i < old_indirect_blocks_length; i++)
  211. stream << static_cast<u32>(block_as_pointers[i]);
  212. for (unsigned i = old_indirect_blocks_length; i < new_indirect_blocks_length; i++) {
  213. auto new_block = new_meta_blocks.take_last().value();
  214. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::grow_doubly_indirect_block(): Allocating indirect block {} at index {}", identifier(), new_block, i);
  215. stream << static_cast<u32>(new_block);
  216. meta_blocks++;
  217. }
  218. stream.fill_to_end(0);
  219. // Write out the indirect blocks.
  220. for (unsigned i = old_blocks_length / entries_per_block; i < new_indirect_blocks_length; i++) {
  221. const auto offset_block = i * entries_per_block;
  222. if (auto result = write_indirect_block(block_as_pointers[i], blocks_indices.slice(offset_block, min(blocks_indices.size() - offset_block, entries_per_block))); result.is_error())
  223. return result;
  224. }
  225. // Write out the doubly indirect block.
  226. return fs().write_block(block, buffer, stream.size());
  227. }
  228. KResult Ext2FSInode::shrink_doubly_indirect_block(BlockBasedFS::BlockIndex block, size_t old_blocks_length, size_t new_blocks_length, unsigned& meta_blocks)
  229. {
  230. const auto entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  231. const auto entries_per_doubly_indirect_block = entries_per_block * entries_per_block;
  232. const auto old_indirect_blocks_length = divide_rounded_up(old_blocks_length, entries_per_block);
  233. const auto new_indirect_blocks_length = divide_rounded_up(new_blocks_length, entries_per_block);
  234. VERIFY(old_blocks_length > 0);
  235. VERIFY(old_blocks_length >= new_blocks_length);
  236. VERIFY(new_blocks_length <= entries_per_doubly_indirect_block);
  237. auto block_contents = ByteBuffer::create_uninitialized(fs().block_size());
  238. auto* block_as_pointers = (unsigned*)block_contents.data();
  239. auto buffer = UserOrKernelBuffer::for_kernel_buffer(reinterpret_cast<u8*>(block_as_pointers));
  240. if (auto result = fs().read_block(block, &buffer, fs().block_size()); result.is_error())
  241. return result;
  242. // Free the unused indirect blocks.
  243. for (unsigned i = new_indirect_blocks_length; i < old_indirect_blocks_length; i++) {
  244. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::shrink_doubly_indirect_block(): Freeing indirect block {} at index {}", identifier(), block_as_pointers[i], i);
  245. if (auto result = fs().set_block_allocation_state(block_as_pointers[i], false); result.is_error())
  246. return result;
  247. meta_blocks--;
  248. }
  249. // Free the doubly indirect block if no longer needed.
  250. if (new_blocks_length == 0) {
  251. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::shrink_doubly_indirect_block(): Freeing doubly indirect block {}", identifier(), block);
  252. if (auto result = fs().set_block_allocation_state(block, false); result.is_error())
  253. return result;
  254. meta_blocks--;
  255. }
  256. return KSuccess;
  257. }
  258. KResult Ext2FSInode::grow_triply_indirect_block(BlockBasedFS::BlockIndex block, size_t old_blocks_length, Span<BlockBasedFS::BlockIndex> blocks_indices, Vector<Ext2FS::BlockIndex>& new_meta_blocks, unsigned& meta_blocks)
  259. {
  260. const auto entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  261. const auto entries_per_doubly_indirect_block = entries_per_block * entries_per_block;
  262. const auto entries_per_triply_indirect_block = entries_per_block * entries_per_block;
  263. const auto old_doubly_indirect_blocks_length = divide_rounded_up(old_blocks_length, entries_per_doubly_indirect_block);
  264. const auto new_doubly_indirect_blocks_length = divide_rounded_up(blocks_indices.size(), entries_per_doubly_indirect_block);
  265. VERIFY(blocks_indices.size() > 0);
  266. VERIFY(blocks_indices.size() > old_blocks_length);
  267. VERIFY(blocks_indices.size() <= entries_per_triply_indirect_block);
  268. auto block_contents = ByteBuffer::create_uninitialized(fs().block_size());
  269. auto* block_as_pointers = (unsigned*)block_contents.data();
  270. OutputMemoryStream stream { block_contents };
  271. auto buffer = UserOrKernelBuffer::for_kernel_buffer(stream.data());
  272. if (old_blocks_length > 0) {
  273. if (auto result = fs().read_block(block, &buffer, fs().block_size()); result.is_error())
  274. return result;
  275. }
  276. // Grow the triply indirect block.
  277. for (unsigned i = 0; i < old_doubly_indirect_blocks_length; i++)
  278. stream << static_cast<u32>(block_as_pointers[i]);
  279. for (unsigned i = old_doubly_indirect_blocks_length; i < new_doubly_indirect_blocks_length; i++) {
  280. auto new_block = new_meta_blocks.take_last().value();
  281. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::grow_triply_indirect_block(): Allocating doubly indirect block {} at index {}", identifier(), new_block, i);
  282. stream << static_cast<u32>(new_block);
  283. meta_blocks++;
  284. }
  285. stream.fill_to_end(0);
  286. // Write out the doubly indirect blocks.
  287. for (unsigned i = old_blocks_length / entries_per_doubly_indirect_block; i < new_doubly_indirect_blocks_length; i++) {
  288. const auto processed_blocks = i * entries_per_doubly_indirect_block;
  289. const auto old_doubly_indirect_blocks_length = min(old_blocks_length > processed_blocks ? old_blocks_length - processed_blocks : 0, entries_per_doubly_indirect_block);
  290. const auto new_doubly_indirect_blocks_length = min(blocks_indices.size() > processed_blocks ? blocks_indices.size() - processed_blocks : 0, entries_per_doubly_indirect_block);
  291. if (auto result = grow_doubly_indirect_block(block_as_pointers[i], old_doubly_indirect_blocks_length, blocks_indices.slice(processed_blocks, new_doubly_indirect_blocks_length), new_meta_blocks, meta_blocks); result.is_error())
  292. return result;
  293. }
  294. // Write out the triply indirect block.
  295. return fs().write_block(block, buffer, stream.size());
  296. }
  297. KResult Ext2FSInode::shrink_triply_indirect_block(BlockBasedFS::BlockIndex block, size_t old_blocks_length, size_t new_blocks_length, unsigned& meta_blocks)
  298. {
  299. const auto entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  300. const auto entries_per_doubly_indirect_block = entries_per_block * entries_per_block;
  301. const auto entries_per_triply_indirect_block = entries_per_doubly_indirect_block * entries_per_block;
  302. const auto old_triply_indirect_blocks_length = divide_rounded_up(old_blocks_length, entries_per_doubly_indirect_block);
  303. const auto new_triply_indirect_blocks_length = new_blocks_length / entries_per_doubly_indirect_block;
  304. VERIFY(old_blocks_length > 0);
  305. VERIFY(old_blocks_length >= new_blocks_length);
  306. VERIFY(new_blocks_length <= entries_per_triply_indirect_block);
  307. auto block_contents = ByteBuffer::create_uninitialized(fs().block_size());
  308. auto* block_as_pointers = (unsigned*)block_contents.data();
  309. auto buffer = UserOrKernelBuffer::for_kernel_buffer(reinterpret_cast<u8*>(block_as_pointers));
  310. if (auto result = fs().read_block(block, &buffer, fs().block_size()); result.is_error())
  311. return result;
  312. // Shrink the doubly indirect blocks.
  313. for (unsigned i = new_triply_indirect_blocks_length; i < old_triply_indirect_blocks_length; i++) {
  314. const auto processed_blocks = i * entries_per_doubly_indirect_block;
  315. const auto old_doubly_indirect_blocks_length = min(old_blocks_length > processed_blocks ? old_blocks_length - processed_blocks : 0, entries_per_doubly_indirect_block);
  316. const auto new_doubly_indirect_blocks_length = min(new_blocks_length > processed_blocks ? new_blocks_length - processed_blocks : 0, entries_per_doubly_indirect_block);
  317. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::shrink_triply_indirect_block(): Shrinking doubly indirect block {} at index {}", identifier(), block_as_pointers[i], i);
  318. if (auto result = shrink_doubly_indirect_block(block_as_pointers[i], old_doubly_indirect_blocks_length, new_doubly_indirect_blocks_length, meta_blocks); result.is_error())
  319. return result;
  320. }
  321. // Free the triply indirect block if no longer needed.
  322. if (new_blocks_length == 0) {
  323. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::shrink_triply_indirect_block(): Freeing triply indirect block {}", identifier(), block);
  324. if (auto result = fs().set_block_allocation_state(block, false); result.is_error())
  325. return result;
  326. meta_blocks--;
  327. }
  328. return KSuccess;
  329. }
  330. KResult Ext2FSInode::flush_block_list()
  331. {
  332. Locker locker(m_lock);
  333. if (m_block_list.is_empty()) {
  334. m_raw_inode.i_blocks = 0;
  335. memset(m_raw_inode.i_block, 0, sizeof(m_raw_inode.i_block));
  336. set_metadata_dirty(true);
  337. return KSuccess;
  338. }
  339. // NOTE: There is a mismatch between i_blocks and blocks.size() since i_blocks includes meta blocks and blocks.size() does not.
  340. const auto old_block_count = ceil_div(size(), static_cast<u64>(fs().block_size()));
  341. auto old_shape = fs().compute_block_list_shape(old_block_count);
  342. const auto new_shape = fs().compute_block_list_shape(m_block_list.size());
  343. Vector<Ext2FS::BlockIndex> new_meta_blocks;
  344. if (new_shape.meta_blocks > old_shape.meta_blocks) {
  345. auto blocks_or_error = fs().allocate_blocks(fs().group_index_from_inode(index()), new_shape.meta_blocks - old_shape.meta_blocks);
  346. if (blocks_or_error.is_error())
  347. return blocks_or_error.error();
  348. new_meta_blocks = blocks_or_error.release_value();
  349. }
  350. m_raw_inode.i_blocks = (m_block_list.size() + new_shape.meta_blocks) * (fs().block_size() / 512);
  351. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::flush_block_list(): Old shape=({};{};{};{}:{}), new shape=({};{};{};{}:{})", identifier(), old_shape.direct_blocks, old_shape.indirect_blocks, old_shape.doubly_indirect_blocks, old_shape.triply_indirect_blocks, old_shape.meta_blocks, new_shape.direct_blocks, new_shape.indirect_blocks, new_shape.doubly_indirect_blocks, new_shape.triply_indirect_blocks, new_shape.meta_blocks);
  352. unsigned output_block_index = 0;
  353. unsigned remaining_blocks = m_block_list.size();
  354. // Deal with direct blocks.
  355. bool inode_dirty = false;
  356. VERIFY(new_shape.direct_blocks <= EXT2_NDIR_BLOCKS);
  357. for (unsigned i = 0; i < new_shape.direct_blocks; ++i) {
  358. if (BlockBasedFS::BlockIndex(m_raw_inode.i_block[i]) != m_block_list[output_block_index])
  359. inode_dirty = true;
  360. m_raw_inode.i_block[i] = m_block_list[output_block_index].value();
  361. ++output_block_index;
  362. --remaining_blocks;
  363. }
  364. // e2fsck considers all blocks reachable through any of the pointers in
  365. // m_raw_inode.i_block as part of this inode regardless of the value in
  366. // m_raw_inode.i_size. When it finds more blocks than the amount that
  367. // is indicated by i_size or i_blocks it offers to repair the filesystem
  368. // by changing those values. That will actually cause further corruption.
  369. // So we must zero all pointers to blocks that are now unused.
  370. for (unsigned i = new_shape.direct_blocks; i < EXT2_NDIR_BLOCKS; ++i) {
  371. m_raw_inode.i_block[i] = 0;
  372. }
  373. if (inode_dirty) {
  374. if constexpr (EXT2_DEBUG) {
  375. dbgln("Ext2FSInode[{}]::flush_block_list(): Writing {} direct block(s) to i_block array of inode {}", identifier(), min((size_t)EXT2_NDIR_BLOCKS, m_block_list.size()), index());
  376. for (size_t i = 0; i < min((size_t)EXT2_NDIR_BLOCKS, m_block_list.size()); ++i)
  377. dbgln(" + {}", m_block_list[i]);
  378. }
  379. set_metadata_dirty(true);
  380. }
  381. // Deal with indirect blocks.
  382. if (old_shape.indirect_blocks != new_shape.indirect_blocks) {
  383. if (new_shape.indirect_blocks > old_shape.indirect_blocks) {
  384. // Write out the indirect block.
  385. if (old_shape.indirect_blocks == 0) {
  386. auto new_block = new_meta_blocks.take_last().value();
  387. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::flush_block_list(): Allocating indirect block: {}", identifier(), new_block);
  388. m_raw_inode.i_block[EXT2_IND_BLOCK] = new_block;
  389. set_metadata_dirty(true);
  390. old_shape.meta_blocks++;
  391. }
  392. if (auto result = write_indirect_block(m_raw_inode.i_block[EXT2_IND_BLOCK], m_block_list.span().slice(output_block_index, new_shape.indirect_blocks)); result.is_error())
  393. return result;
  394. } else if ((new_shape.indirect_blocks == 0) && (old_shape.indirect_blocks != 0)) {
  395. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::flush_block_list(): Freeing indirect block: {}", identifier(), m_raw_inode.i_block[EXT2_IND_BLOCK]);
  396. if (auto result = fs().set_block_allocation_state(m_raw_inode.i_block[EXT2_IND_BLOCK], false); result.is_error())
  397. return result;
  398. old_shape.meta_blocks--;
  399. m_raw_inode.i_block[EXT2_IND_BLOCK] = 0;
  400. }
  401. }
  402. remaining_blocks -= new_shape.indirect_blocks;
  403. output_block_index += new_shape.indirect_blocks;
  404. if (old_shape.doubly_indirect_blocks != new_shape.doubly_indirect_blocks) {
  405. // Write out the doubly indirect block.
  406. if (new_shape.doubly_indirect_blocks > old_shape.doubly_indirect_blocks) {
  407. if (old_shape.doubly_indirect_blocks == 0) {
  408. auto new_block = new_meta_blocks.take_last().value();
  409. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::flush_block_list(): Allocating doubly indirect block: {}", identifier(), new_block);
  410. m_raw_inode.i_block[EXT2_DIND_BLOCK] = new_block;
  411. set_metadata_dirty(true);
  412. old_shape.meta_blocks++;
  413. }
  414. if (auto result = grow_doubly_indirect_block(m_raw_inode.i_block[EXT2_DIND_BLOCK], old_shape.doubly_indirect_blocks, m_block_list.span().slice(output_block_index, new_shape.doubly_indirect_blocks), new_meta_blocks, old_shape.meta_blocks); result.is_error())
  415. return result;
  416. } else {
  417. if (auto result = shrink_doubly_indirect_block(m_raw_inode.i_block[EXT2_DIND_BLOCK], old_shape.doubly_indirect_blocks, new_shape.doubly_indirect_blocks, old_shape.meta_blocks); result.is_error())
  418. return result;
  419. if (new_shape.doubly_indirect_blocks == 0)
  420. m_raw_inode.i_block[EXT2_DIND_BLOCK] = 0;
  421. }
  422. }
  423. remaining_blocks -= new_shape.doubly_indirect_blocks;
  424. output_block_index += new_shape.doubly_indirect_blocks;
  425. if (old_shape.triply_indirect_blocks != new_shape.triply_indirect_blocks) {
  426. // Write out the triply indirect block.
  427. if (new_shape.triply_indirect_blocks > old_shape.triply_indirect_blocks) {
  428. if (old_shape.triply_indirect_blocks == 0) {
  429. auto new_block = new_meta_blocks.take_last().value();
  430. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::flush_block_list(): Allocating triply indirect block: {}", identifier(), new_block);
  431. m_raw_inode.i_block[EXT2_TIND_BLOCK] = new_block;
  432. set_metadata_dirty(true);
  433. old_shape.meta_blocks++;
  434. }
  435. if (auto result = grow_triply_indirect_block(m_raw_inode.i_block[EXT2_TIND_BLOCK], old_shape.triply_indirect_blocks, m_block_list.span().slice(output_block_index, new_shape.triply_indirect_blocks), new_meta_blocks, old_shape.meta_blocks); result.is_error())
  436. return result;
  437. } else {
  438. if (auto result = shrink_triply_indirect_block(m_raw_inode.i_block[EXT2_TIND_BLOCK], old_shape.triply_indirect_blocks, new_shape.triply_indirect_blocks, old_shape.meta_blocks); result.is_error())
  439. return result;
  440. if (new_shape.triply_indirect_blocks == 0)
  441. m_raw_inode.i_block[EXT2_TIND_BLOCK] = 0;
  442. }
  443. }
  444. remaining_blocks -= new_shape.triply_indirect_blocks;
  445. output_block_index += new_shape.triply_indirect_blocks;
  446. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::flush_block_list(): New meta blocks count at {}, expecting {}", identifier(), old_shape.meta_blocks, new_shape.meta_blocks);
  447. VERIFY(new_meta_blocks.size() == 0);
  448. VERIFY(old_shape.meta_blocks == new_shape.meta_blocks);
  449. if (!remaining_blocks)
  450. return KSuccess;
  451. dbgln("we don't know how to write qind ext2fs blocks, they don't exist anyway!");
  452. VERIFY_NOT_REACHED();
  453. }
  454. Vector<Ext2FS::BlockIndex> Ext2FSInode::compute_block_list() const
  455. {
  456. return compute_block_list_impl(false);
  457. }
  458. Vector<Ext2FS::BlockIndex> Ext2FSInode::compute_block_list_with_meta_blocks() const
  459. {
  460. return compute_block_list_impl(true);
  461. }
  462. Vector<Ext2FS::BlockIndex> Ext2FSInode::compute_block_list_impl(bool include_block_list_blocks) const
  463. {
  464. // FIXME: This is really awkwardly factored.. foo_impl_internal :|
  465. auto block_list = compute_block_list_impl_internal(m_raw_inode, include_block_list_blocks);
  466. while (!block_list.is_empty() && block_list.last() == 0)
  467. block_list.take_last();
  468. return block_list;
  469. }
  470. Vector<Ext2FS::BlockIndex> Ext2FSInode::compute_block_list_impl_internal(const ext2_inode& e2inode, bool include_block_list_blocks) const
  471. {
  472. unsigned entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  473. unsigned block_count = ceil_div(size(), static_cast<u64>(fs().block_size()));
  474. // If we are handling a symbolic link, the path is stored in the 60 bytes in
  475. // the inode that are used for the 12 direct and 3 indirect block pointers,
  476. // If the path is longer than 60 characters, a block is allocated, and the
  477. // block contains the destination path. The file size corresponds to the
  478. // path length of the destination.
  479. if (::is_symlink(e2inode.i_mode) && e2inode.i_blocks == 0)
  480. block_count = 0;
  481. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::block_list_for_inode(): i_size={}, i_blocks={}, block_count={}", identifier(), e2inode.i_size, e2inode.i_blocks, block_count);
  482. unsigned blocks_remaining = block_count;
  483. if (include_block_list_blocks) {
  484. auto shape = fs().compute_block_list_shape(block_count);
  485. blocks_remaining += shape.meta_blocks;
  486. }
  487. Vector<Ext2FS::BlockIndex> list;
  488. auto add_block = [&](auto bi) {
  489. if (blocks_remaining) {
  490. list.append(bi);
  491. --blocks_remaining;
  492. }
  493. };
  494. if (include_block_list_blocks) {
  495. // This seems like an excessive over-estimate but w/e.
  496. list.ensure_capacity(blocks_remaining * 2);
  497. } else {
  498. list.ensure_capacity(blocks_remaining);
  499. }
  500. unsigned direct_count = min(block_count, (unsigned)EXT2_NDIR_BLOCKS);
  501. for (unsigned i = 0; i < direct_count; ++i) {
  502. auto block_index = e2inode.i_block[i];
  503. add_block(block_index);
  504. }
  505. if (!blocks_remaining)
  506. return list;
  507. // Don't need to make copy of add_block, since this capture will only
  508. // be called before compute_block_list_impl_internal finishes.
  509. auto process_block_array = [&](auto array_block_index, auto&& callback) {
  510. if (include_block_list_blocks)
  511. add_block(array_block_index);
  512. auto count = min(blocks_remaining, entries_per_block);
  513. if (!count)
  514. return;
  515. size_t read_size = count * sizeof(u32);
  516. auto array_storage = ByteBuffer::create_uninitialized(read_size);
  517. auto* array = (u32*)array_storage.data();
  518. auto buffer = UserOrKernelBuffer::for_kernel_buffer((u8*)array);
  519. if (auto result = fs().read_block(array_block_index, &buffer, read_size, 0); result.is_error()) {
  520. // FIXME: Stop here and propagate this error.
  521. dbgln("Ext2FSInode[{}]::compute_block_list_impl_internal(): Error: {}", identifier(), result.error());
  522. }
  523. for (unsigned i = 0; i < count; ++i)
  524. callback(Ext2FS::BlockIndex(array[i]));
  525. };
  526. process_block_array(e2inode.i_block[EXT2_IND_BLOCK], [&](auto block_index) {
  527. add_block(block_index);
  528. });
  529. if (!blocks_remaining)
  530. return list;
  531. process_block_array(e2inode.i_block[EXT2_DIND_BLOCK], [&](auto block_index) {
  532. process_block_array(block_index, [&](auto block_index2) {
  533. add_block(block_index2);
  534. });
  535. });
  536. if (!blocks_remaining)
  537. return list;
  538. process_block_array(e2inode.i_block[EXT2_TIND_BLOCK], [&](auto block_index) {
  539. process_block_array(block_index, [&](auto block_index2) {
  540. process_block_array(block_index2, [&](auto block_index3) {
  541. add_block(block_index3);
  542. });
  543. });
  544. });
  545. return list;
  546. }
  547. void Ext2FS::free_inode(Ext2FSInode& inode)
  548. {
  549. Locker locker(m_lock);
  550. VERIFY(inode.m_raw_inode.i_links_count == 0);
  551. dbgln_if(EXT2_DEBUG, "Ext2FS[{}]::free_inode(): Inode {} has no more links, time to delete!", fsid(), inode.index());
  552. // Mark all blocks used by this inode as free.
  553. for (auto block_index : inode.compute_block_list_with_meta_blocks()) {
  554. VERIFY(block_index <= super_block().s_blocks_count);
  555. if (block_index.value()) {
  556. if (auto result = set_block_allocation_state(block_index, false); result.is_error()) {
  557. dbgln("Ext2FS[{}]::free_inode(): Failed to deallocate block {} for inode {}", fsid(), block_index, inode.index());
  558. }
  559. }
  560. }
  561. // If the inode being freed is a directory, update block group directory counter.
  562. if (inode.is_directory()) {
  563. auto& bgd = const_cast<ext2_group_desc&>(group_descriptor(group_index_from_inode(inode.index())));
  564. --bgd.bg_used_dirs_count;
  565. dbgln_if(EXT2_DEBUG, "Ext2FS[{}]::free_inode(): Decremented bg_used_dirs_count to {} for inode {}", fsid(), bgd.bg_used_dirs_count, inode.index());
  566. m_block_group_descriptors_dirty = true;
  567. }
  568. // NOTE: After this point, the inode metadata is wiped.
  569. memset(&inode.m_raw_inode, 0, sizeof(ext2_inode));
  570. inode.m_raw_inode.i_dtime = kgettimeofday().to_truncated_seconds();
  571. write_ext2_inode(inode.index(), inode.m_raw_inode);
  572. // Mark the inode as free.
  573. if (auto result = set_inode_allocation_state(inode.index(), false); result.is_error())
  574. dbgln("Ext2FS[{}]::free_inode(): Failed to free inode {}: {}", fsid(), inode.index(), result.error());
  575. }
  576. void Ext2FS::flush_block_group_descriptor_table()
  577. {
  578. Locker locker(m_lock);
  579. unsigned blocks_to_write = ceil_div(m_block_group_count * sizeof(ext2_group_desc), block_size());
  580. unsigned first_block_of_bgdt = block_size() == 1024 ? 2 : 1;
  581. auto buffer = UserOrKernelBuffer::for_kernel_buffer((u8*)block_group_descriptors());
  582. if (auto result = write_blocks(first_block_of_bgdt, blocks_to_write, buffer); result.is_error())
  583. dbgln("Ext2FS[{}]::flush_block_group_descriptor_table(): Failed to write blocks: {}", fsid(), result.error());
  584. }
  585. void Ext2FS::flush_writes()
  586. {
  587. Locker locker(m_lock);
  588. if (m_super_block_dirty) {
  589. flush_super_block();
  590. m_super_block_dirty = false;
  591. }
  592. if (m_block_group_descriptors_dirty) {
  593. flush_block_group_descriptor_table();
  594. m_block_group_descriptors_dirty = false;
  595. }
  596. for (auto& cached_bitmap : m_cached_bitmaps) {
  597. if (cached_bitmap->dirty) {
  598. auto buffer = UserOrKernelBuffer::for_kernel_buffer(cached_bitmap->buffer.data());
  599. if (auto result = write_block(cached_bitmap->bitmap_block_index, buffer, block_size()); result.is_error()) {
  600. dbgln("Ext2FS[{}]::flush_writes(): Failed to write blocks: {}", fsid(), result.error());
  601. }
  602. cached_bitmap->dirty = false;
  603. dbgln_if(EXT2_DEBUG, "Ext2FS[{}]::flush_writes(): Flushed bitmap block {}", fsid(), cached_bitmap->bitmap_block_index);
  604. }
  605. }
  606. BlockBasedFS::flush_writes();
  607. // Uncache Inodes that are only kept alive by the index-to-inode lookup cache.
  608. // We don't uncache Inodes that are being watched by at least one InodeWatcher.
  609. // FIXME: It would be better to keep a capped number of Inodes around.
  610. // The problem is that they are quite heavy objects, and use a lot of heap memory
  611. // for their (child name lookup) and (block list) caches.
  612. Vector<InodeIndex> unused_inodes;
  613. for (auto& it : m_inode_cache) {
  614. if (it.value->ref_count() != 1)
  615. continue;
  616. if (it.value->has_watchers())
  617. continue;
  618. unused_inodes.append(it.key);
  619. }
  620. for (auto index : unused_inodes)
  621. uncache_inode(index);
  622. }
  623. Ext2FSInode::Ext2FSInode(Ext2FS& fs, InodeIndex index)
  624. : Inode(fs, index)
  625. {
  626. }
  627. Ext2FSInode::~Ext2FSInode()
  628. {
  629. if (m_raw_inode.i_links_count == 0)
  630. fs().free_inode(*this);
  631. }
  632. u64 Ext2FSInode::size() const
  633. {
  634. if (Kernel::is_regular_file(m_raw_inode.i_mode) && ((u32)fs().get_features_readonly() & (u32)Ext2FS::FeaturesReadOnly::FileSize64bits))
  635. return static_cast<u64>(m_raw_inode.i_dir_acl) << 32 | m_raw_inode.i_size;
  636. return m_raw_inode.i_size;
  637. }
  638. InodeMetadata Ext2FSInode::metadata() const
  639. {
  640. Locker locker(m_lock);
  641. InodeMetadata metadata;
  642. metadata.inode = identifier();
  643. metadata.size = size();
  644. metadata.mode = m_raw_inode.i_mode;
  645. metadata.uid = m_raw_inode.i_uid;
  646. metadata.gid = m_raw_inode.i_gid;
  647. metadata.link_count = m_raw_inode.i_links_count;
  648. metadata.atime = m_raw_inode.i_atime;
  649. metadata.ctime = m_raw_inode.i_ctime;
  650. metadata.mtime = m_raw_inode.i_mtime;
  651. metadata.dtime = m_raw_inode.i_dtime;
  652. metadata.block_size = fs().block_size();
  653. metadata.block_count = m_raw_inode.i_blocks;
  654. if (Kernel::is_character_device(m_raw_inode.i_mode) || Kernel::is_block_device(m_raw_inode.i_mode)) {
  655. unsigned dev = m_raw_inode.i_block[0];
  656. if (!dev)
  657. dev = m_raw_inode.i_block[1];
  658. metadata.major_device = (dev & 0xfff00) >> 8;
  659. metadata.minor_device = (dev & 0xff) | ((dev >> 12) & 0xfff00);
  660. }
  661. return metadata;
  662. }
  663. void Ext2FSInode::flush_metadata()
  664. {
  665. Locker locker(m_lock);
  666. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::flush_metadata(): Flushing inode", identifier());
  667. fs().write_ext2_inode(index(), m_raw_inode);
  668. if (is_directory()) {
  669. // Unless we're about to go away permanently, invalidate the lookup cache.
  670. if (m_raw_inode.i_links_count != 0) {
  671. // FIXME: This invalidation is way too hardcore. It's sad to throw away the whole cache.
  672. m_lookup_cache.clear();
  673. }
  674. }
  675. set_metadata_dirty(false);
  676. }
  677. RefPtr<Inode> Ext2FS::get_inode(InodeIdentifier inode) const
  678. {
  679. Locker locker(m_lock);
  680. VERIFY(inode.fsid() == fsid());
  681. {
  682. auto it = m_inode_cache.find(inode.index());
  683. if (it != m_inode_cache.end())
  684. return (*it).value;
  685. }
  686. auto state_or_error = get_inode_allocation_state(inode.index());
  687. if (state_or_error.is_error())
  688. return {};
  689. if (!state_or_error.value()) {
  690. m_inode_cache.set(inode.index(), nullptr);
  691. return {};
  692. }
  693. BlockIndex block_index;
  694. unsigned offset;
  695. if (!find_block_containing_inode(inode.index(), block_index, offset))
  696. return {};
  697. auto new_inode = adopt_ref(*new Ext2FSInode(const_cast<Ext2FS&>(*this), inode.index()));
  698. auto buffer = UserOrKernelBuffer::for_kernel_buffer(reinterpret_cast<u8*>(&new_inode->m_raw_inode));
  699. if (auto result = read_block(block_index, &buffer, sizeof(ext2_inode), offset); result.is_error()) {
  700. // FIXME: Propagate the actual error.
  701. return nullptr;
  702. }
  703. m_inode_cache.set(inode.index(), new_inode);
  704. return new_inode;
  705. }
  706. KResultOr<ssize_t> Ext2FSInode::read_bytes(off_t offset, ssize_t count, UserOrKernelBuffer& buffer, FileDescription* description) const
  707. {
  708. Locker inode_locker(m_lock);
  709. VERIFY(offset >= 0);
  710. if (m_raw_inode.i_size == 0)
  711. return 0;
  712. if (static_cast<u64>(offset) >= size())
  713. return 0;
  714. // Symbolic links shorter than 60 characters are store inline inside the i_block array.
  715. // This avoids wasting an entire block on short links. (Most links are short.)
  716. if (is_symlink() && size() < max_inline_symlink_length) {
  717. VERIFY(offset == 0);
  718. ssize_t nread = min((off_t)size() - offset, static_cast<off_t>(count));
  719. if (!buffer.write(((const u8*)m_raw_inode.i_block) + offset, (size_t)nread))
  720. return EFAULT;
  721. return nread;
  722. }
  723. if (m_block_list.is_empty())
  724. m_block_list = compute_block_list();
  725. if (m_block_list.is_empty()) {
  726. dmesgln("Ext2FSInode[{}]::read_bytes(): Empty block list", identifier());
  727. return EIO;
  728. }
  729. bool allow_cache = !description || !description->is_direct();
  730. const int block_size = fs().block_size();
  731. BlockBasedFS::BlockIndex first_block_logical_index = offset / block_size;
  732. BlockBasedFS::BlockIndex last_block_logical_index = (offset + count) / block_size;
  733. if (last_block_logical_index >= m_block_list.size())
  734. last_block_logical_index = m_block_list.size() - 1;
  735. int offset_into_first_block = offset % block_size;
  736. ssize_t nread = 0;
  737. auto remaining_count = min((off_t)count, (off_t)size() - offset);
  738. dbgln_if(EXT2_VERY_DEBUG, "Ext2FSInode[{}]::read_bytes(): Reading up to {} bytes, {} bytes into inode to {}", identifier(), count, offset, buffer.user_or_kernel_ptr());
  739. for (auto bi = first_block_logical_index; remaining_count && bi <= last_block_logical_index; bi = bi.value() + 1) {
  740. auto block_index = m_block_list[bi.value()];
  741. size_t offset_into_block = (bi == first_block_logical_index) ? offset_into_first_block : 0;
  742. size_t num_bytes_to_copy = min((size_t)block_size - offset_into_block, (size_t)remaining_count);
  743. auto buffer_offset = buffer.offset(nread);
  744. if (block_index.value() == 0) {
  745. // This is a hole, act as if it's filled with zeroes.
  746. if (!buffer_offset.memset(0, num_bytes_to_copy))
  747. return EFAULT;
  748. } else {
  749. if (auto result = fs().read_block(block_index, &buffer_offset, num_bytes_to_copy, offset_into_block, allow_cache); result.is_error()) {
  750. dmesgln("Ext2FSInode[{}]::read_bytes(): Failed to read block {} (index {})", identifier(), block_index.value(), bi);
  751. return result.error();
  752. }
  753. }
  754. remaining_count -= num_bytes_to_copy;
  755. nread += num_bytes_to_copy;
  756. }
  757. return nread;
  758. }
  759. KResult Ext2FSInode::resize(u64 new_size)
  760. {
  761. auto old_size = size();
  762. if (old_size == new_size)
  763. return KSuccess;
  764. if (!((u32)fs().get_features_readonly() & (u32)Ext2FS::FeaturesReadOnly::FileSize64bits) && (new_size >= static_cast<u32>(-1)))
  765. return ENOSPC;
  766. u64 block_size = fs().block_size();
  767. auto blocks_needed_before = ceil_div(old_size, block_size);
  768. auto blocks_needed_after = ceil_div(new_size, block_size);
  769. if constexpr (EXT2_DEBUG) {
  770. dbgln("Ext2FSInode[{}]::resize(): Blocks needed before (size was {}): {}", identifier(), old_size, blocks_needed_before);
  771. dbgln("Ext2FSInode[{}]::resize(): Blocks needed after (size is {}): {}", identifier(), new_size, blocks_needed_after);
  772. }
  773. if (blocks_needed_after > blocks_needed_before) {
  774. auto additional_blocks_needed = blocks_needed_after - blocks_needed_before;
  775. if (additional_blocks_needed > fs().super_block().s_free_blocks_count)
  776. return ENOSPC;
  777. }
  778. if (m_block_list.is_empty())
  779. m_block_list = this->compute_block_list();
  780. if (blocks_needed_after > blocks_needed_before) {
  781. auto blocks_or_error = fs().allocate_blocks(fs().group_index_from_inode(index()), blocks_needed_after - blocks_needed_before);
  782. if (blocks_or_error.is_error())
  783. return blocks_or_error.error();
  784. if (!m_block_list.try_append(blocks_or_error.release_value()))
  785. return ENOMEM;
  786. } else if (blocks_needed_after < blocks_needed_before) {
  787. if constexpr (EXT2_VERY_DEBUG) {
  788. dbgln("Ext2FSInode[{}]::resize(): Shrinking inode, old block list is {} entries:", identifier(), m_block_list.size());
  789. for (auto block_index : m_block_list) {
  790. dbgln(" # {}", block_index);
  791. }
  792. }
  793. while (m_block_list.size() != blocks_needed_after) {
  794. auto block_index = m_block_list.take_last();
  795. if (block_index.value()) {
  796. if (auto result = fs().set_block_allocation_state(block_index, false); result.is_error()) {
  797. dbgln("Ext2FSInode[{}]::resize(): Failed to free block {}: {}", identifier(), block_index, result.error());
  798. return result;
  799. }
  800. }
  801. }
  802. }
  803. if (auto result = flush_block_list(); result.is_error())
  804. return result;
  805. m_raw_inode.i_size = new_size;
  806. if (Kernel::is_regular_file(m_raw_inode.i_mode))
  807. m_raw_inode.i_dir_acl = new_size >> 32;
  808. set_metadata_dirty(true);
  809. if (new_size > old_size) {
  810. // If we're growing the inode, make sure we zero out all the new space.
  811. // FIXME: There are definitely more efficient ways to achieve this.
  812. auto bytes_to_clear = new_size - old_size;
  813. auto clear_from = old_size;
  814. u8 zero_buffer[PAGE_SIZE] {};
  815. while (bytes_to_clear) {
  816. auto result = write_bytes(clear_from, min(static_cast<u64>(sizeof(zero_buffer)), bytes_to_clear), UserOrKernelBuffer::for_kernel_buffer(zero_buffer), nullptr);
  817. if (result.is_error())
  818. return result.error();
  819. VERIFY(result.value() != 0);
  820. bytes_to_clear -= result.value();
  821. clear_from += result.value();
  822. }
  823. }
  824. return KSuccess;
  825. }
  826. KResultOr<ssize_t> Ext2FSInode::write_bytes(off_t offset, ssize_t count, const UserOrKernelBuffer& data, FileDescription* description)
  827. {
  828. VERIFY(offset >= 0);
  829. VERIFY(count >= 0);
  830. if (count == 0)
  831. return 0;
  832. Locker inode_locker(m_lock);
  833. if (auto result = prepare_to_write_data(); result.is_error())
  834. return result;
  835. if (is_symlink()) {
  836. VERIFY(offset == 0);
  837. if (max((size_t)(offset + count), (size_t)m_raw_inode.i_size) < max_inline_symlink_length) {
  838. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::write_bytes(): Poking into i_block array for inline symlink '{}' ({} bytes)", identifier(), data.copy_into_string(count), count);
  839. if (!data.read(((u8*)m_raw_inode.i_block) + offset, (size_t)count))
  840. return EFAULT;
  841. if ((size_t)(offset + count) > (size_t)m_raw_inode.i_size)
  842. m_raw_inode.i_size = offset + count;
  843. set_metadata_dirty(true);
  844. return count;
  845. }
  846. }
  847. bool allow_cache = !description || !description->is_direct();
  848. const auto block_size = fs().block_size();
  849. auto new_size = max(static_cast<u64>(offset) + count, size());
  850. if (auto result = resize(new_size); result.is_error())
  851. return result;
  852. if (m_block_list.is_empty())
  853. m_block_list = compute_block_list();
  854. if (m_block_list.is_empty()) {
  855. dbgln("Ext2FSInode[{}]::write_bytes(): Empty block list", identifier());
  856. return EIO;
  857. }
  858. BlockBasedFS::BlockIndex first_block_logical_index = offset / block_size;
  859. BlockBasedFS::BlockIndex last_block_logical_index = (offset + count) / block_size;
  860. if (last_block_logical_index >= m_block_list.size())
  861. last_block_logical_index = m_block_list.size() - 1;
  862. size_t offset_into_first_block = offset % block_size;
  863. ssize_t nwritten = 0;
  864. auto remaining_count = min((off_t)count, (off_t)new_size - offset);
  865. dbgln_if(EXT2_VERY_DEBUG, "Ext2FSInode[{}]::write_bytes(): Writing {} bytes, {} bytes into inode from {}", identifier(), count, offset, data.user_or_kernel_ptr());
  866. for (auto bi = first_block_logical_index; remaining_count && bi <= last_block_logical_index; bi = bi.value() + 1) {
  867. size_t offset_into_block = (bi == first_block_logical_index) ? offset_into_first_block : 0;
  868. size_t num_bytes_to_copy = min((size_t)block_size - offset_into_block, (size_t)remaining_count);
  869. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::write_bytes(): Writing block {} (offset_into_block: {})", identifier(), m_block_list[bi.value()], offset_into_block);
  870. if (auto result = fs().write_block(m_block_list[bi.value()], data.offset(nwritten), num_bytes_to_copy, offset_into_block, allow_cache); result.is_error()) {
  871. dbgln("Ext2FSInode[{}]::write_bytes(): Failed to write block {} (index {})", identifier(), m_block_list[bi.value()], bi);
  872. return result;
  873. }
  874. remaining_count -= num_bytes_to_copy;
  875. nwritten += num_bytes_to_copy;
  876. }
  877. did_modify_contents();
  878. dbgln_if(EXT2_VERY_DEBUG, "Ext2FSInode[{}]::write_bytes(): After write, i_size={}, i_blocks={} ({} blocks in list)", identifier(), size(), m_raw_inode.i_blocks, m_block_list.size());
  879. return nwritten;
  880. }
  881. u8 Ext2FS::internal_file_type_to_directory_entry_type(const DirectoryEntryView& entry) const
  882. {
  883. switch (entry.file_type) {
  884. case EXT2_FT_REG_FILE:
  885. return DT_REG;
  886. case EXT2_FT_DIR:
  887. return DT_DIR;
  888. case EXT2_FT_CHRDEV:
  889. return DT_CHR;
  890. case EXT2_FT_BLKDEV:
  891. return DT_BLK;
  892. case EXT2_FT_FIFO:
  893. return DT_FIFO;
  894. case EXT2_FT_SOCK:
  895. return DT_SOCK;
  896. case EXT2_FT_SYMLINK:
  897. return DT_LNK;
  898. default:
  899. return DT_UNKNOWN;
  900. }
  901. }
  902. Ext2FS::FeaturesReadOnly Ext2FS::get_features_readonly() const
  903. {
  904. if (m_super_block.s_rev_level > 0)
  905. return static_cast<Ext2FS::FeaturesReadOnly>(m_super_block.s_feature_ro_compat);
  906. return Ext2FS::FeaturesReadOnly::None;
  907. }
  908. KResult Ext2FSInode::traverse_as_directory(Function<bool(const FS::DirectoryEntryView&)> callback) const
  909. {
  910. Locker locker(m_lock);
  911. VERIFY(is_directory());
  912. u8 buffer[max_block_size];
  913. auto buf = UserOrKernelBuffer::for_kernel_buffer(buffer);
  914. auto block_size = fs().block_size();
  915. bool allow_cache = true;
  916. if (m_block_list.is_empty())
  917. m_block_list = compute_block_list();
  918. // Directory entries are guaranteed not to span multiple blocks,
  919. // so we can iterate over blocks separately.
  920. for (auto& block_index : m_block_list) {
  921. VERIFY(block_index.value() != 0);
  922. if (auto result = fs().read_block(block_index, &buf, block_size, 0, allow_cache); result.is_error()) {
  923. return result;
  924. }
  925. auto* entry = reinterpret_cast<ext2_dir_entry_2*>(buffer);
  926. auto* entries_end = reinterpret_cast<ext2_dir_entry_2*>(buffer + block_size);
  927. while (entry < entries_end) {
  928. if (entry->inode != 0) {
  929. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::traverse_as_directory(): inode {}, name_len: {}, rec_len: {}, file_type: {}, name: {}", identifier(), entry->inode, entry->name_len, entry->rec_len, entry->file_type, StringView(entry->name, entry->name_len));
  930. if (!callback({ { entry->name, entry->name_len }, { fsid(), entry->inode }, entry->file_type }))
  931. return KSuccess;
  932. }
  933. entry = (ext2_dir_entry_2*)((char*)entry + entry->rec_len);
  934. }
  935. }
  936. return KSuccess;
  937. }
  938. KResult Ext2FSInode::write_directory(Vector<Ext2FSDirectoryEntry>& entries)
  939. {
  940. Locker locker(m_lock);
  941. auto block_size = fs().block_size();
  942. // Calculate directory size and record length of entries so that
  943. // the following constraints are met:
  944. // - All used blocks must be entirely filled.
  945. // - Entries are aligned on a 4-byte boundary.
  946. // - No entry may span multiple blocks.
  947. size_t directory_size = 0;
  948. size_t space_in_block = block_size;
  949. for (size_t i = 0; i < entries.size(); ++i) {
  950. auto& entry = entries[i];
  951. entry.record_length = EXT2_DIR_REC_LEN(entry.name.length());
  952. space_in_block -= entry.record_length;
  953. if (i + 1 < entries.size()) {
  954. if (EXT2_DIR_REC_LEN(entries[i + 1].name.length()) > space_in_block) {
  955. entry.record_length += space_in_block;
  956. space_in_block = block_size;
  957. }
  958. } else {
  959. entry.record_length += space_in_block;
  960. }
  961. directory_size += entry.record_length;
  962. }
  963. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::write_directory(): New directory contents to write (size {}):", identifier(), directory_size);
  964. auto directory_data = ByteBuffer::create_uninitialized(directory_size);
  965. OutputMemoryStream stream { directory_data };
  966. for (auto& entry : entries) {
  967. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::write_directory(): Writing inode: {}, name_len: {}, rec_len: {}, file_type: {}, name: {}", identifier(), entry.inode_index, u16(entry.name.length()), u16(entry.record_length), u8(entry.file_type), entry.name);
  968. stream << u32(entry.inode_index.value());
  969. stream << u16(entry.record_length);
  970. stream << u8(entry.name.length());
  971. stream << u8(entry.file_type);
  972. stream << entry.name.bytes();
  973. int padding = entry.record_length - entry.name.length() - 8;
  974. for (int j = 0; j < padding; ++j)
  975. stream << u8(0);
  976. }
  977. VERIFY(stream.is_end());
  978. if (auto result = resize(stream.size()); result.is_error())
  979. return result;
  980. auto buffer = UserOrKernelBuffer::for_kernel_buffer(stream.data());
  981. auto result = write_bytes(0, stream.size(), buffer, nullptr);
  982. if (result.is_error())
  983. return result.error();
  984. set_metadata_dirty(true);
  985. if (static_cast<size_t>(result.value()) != directory_data.size())
  986. return EIO;
  987. return KSuccess;
  988. }
  989. KResultOr<NonnullRefPtr<Inode>> Ext2FSInode::create_child(const String& name, mode_t mode, dev_t dev, uid_t uid, gid_t gid)
  990. {
  991. if (::is_directory(mode))
  992. return fs().create_directory(*this, name, mode, uid, gid);
  993. return fs().create_inode(*this, name, mode, dev, uid, gid);
  994. }
  995. KResult Ext2FSInode::add_child(Inode& child, const StringView& name, mode_t mode)
  996. {
  997. Locker locker(m_lock);
  998. VERIFY(is_directory());
  999. if (name.length() > EXT2_NAME_LEN)
  1000. return ENAMETOOLONG;
  1001. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::add_child(): Adding inode {} with name '{}' and mode {:o} to directory {}", identifier(), child.index(), name, mode, index());
  1002. Vector<Ext2FSDirectoryEntry> entries;
  1003. bool name_already_exists = false;
  1004. KResult result = traverse_as_directory([&](auto& entry) {
  1005. if (name == entry.name) {
  1006. name_already_exists = true;
  1007. return false;
  1008. }
  1009. entries.append({ entry.name, entry.inode.index(), entry.file_type });
  1010. return true;
  1011. });
  1012. if (result.is_error())
  1013. return result;
  1014. if (name_already_exists) {
  1015. dbgln("Ext2FSInode[{}]::add_child(): Name '{}' already exists", identifier(), name);
  1016. return EEXIST;
  1017. }
  1018. result = child.increment_link_count();
  1019. if (result.is_error())
  1020. return result;
  1021. entries.empend(name, child.index(), to_ext2_file_type(mode));
  1022. result = write_directory(entries);
  1023. if (result.is_error())
  1024. return result;
  1025. m_lookup_cache.set(name, child.index());
  1026. did_add_child(child.identifier(), name);
  1027. return KSuccess;
  1028. }
  1029. KResult Ext2FSInode::remove_child(const StringView& name)
  1030. {
  1031. Locker locker(m_lock);
  1032. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::remove_child(): Removing '{}'", identifier(), name);
  1033. VERIFY(is_directory());
  1034. auto it = m_lookup_cache.find(name);
  1035. if (it == m_lookup_cache.end())
  1036. return ENOENT;
  1037. auto child_inode_index = (*it).value;
  1038. InodeIdentifier child_id { fsid(), child_inode_index };
  1039. Vector<Ext2FSDirectoryEntry> entries;
  1040. KResult result = traverse_as_directory([&](auto& entry) {
  1041. if (name != entry.name)
  1042. entries.append({ entry.name, entry.inode.index(), entry.file_type });
  1043. return true;
  1044. });
  1045. if (result.is_error())
  1046. return result;
  1047. result = write_directory(entries);
  1048. if (result.is_error())
  1049. return result;
  1050. m_lookup_cache.remove(name);
  1051. auto child_inode = fs().get_inode(child_id);
  1052. result = child_inode->decrement_link_count();
  1053. if (result.is_error())
  1054. return result;
  1055. did_remove_child(child_id, name);
  1056. return KSuccess;
  1057. }
  1058. unsigned Ext2FS::inodes_per_block() const
  1059. {
  1060. return EXT2_INODES_PER_BLOCK(&super_block());
  1061. }
  1062. unsigned Ext2FS::inodes_per_group() const
  1063. {
  1064. return EXT2_INODES_PER_GROUP(&super_block());
  1065. }
  1066. unsigned Ext2FS::inode_size() const
  1067. {
  1068. return EXT2_INODE_SIZE(&super_block());
  1069. }
  1070. unsigned Ext2FS::blocks_per_group() const
  1071. {
  1072. return EXT2_BLOCKS_PER_GROUP(&super_block());
  1073. }
  1074. bool Ext2FS::write_ext2_inode(InodeIndex inode, const ext2_inode& e2inode)
  1075. {
  1076. BlockIndex block_index;
  1077. unsigned offset;
  1078. if (!find_block_containing_inode(inode, block_index, offset))
  1079. return false;
  1080. auto buffer = UserOrKernelBuffer::for_kernel_buffer(const_cast<u8*>((const u8*)&e2inode));
  1081. return write_block(block_index, buffer, inode_size(), offset) >= 0;
  1082. }
  1083. auto Ext2FS::allocate_blocks(GroupIndex preferred_group_index, size_t count) -> KResultOr<Vector<BlockIndex>>
  1084. {
  1085. Locker locker(m_lock);
  1086. dbgln_if(EXT2_DEBUG, "Ext2FS: allocate_blocks(preferred group: {}, count {})", preferred_group_index, count);
  1087. if (count == 0)
  1088. return Vector<BlockIndex> {};
  1089. Vector<BlockIndex> blocks;
  1090. dbgln_if(EXT2_DEBUG, "Ext2FS: allocate_blocks:");
  1091. blocks.ensure_capacity(count);
  1092. auto group_index = preferred_group_index;
  1093. if (!group_descriptor(preferred_group_index).bg_free_blocks_count) {
  1094. group_index = 1;
  1095. }
  1096. while (blocks.size() < count) {
  1097. bool found_a_group = false;
  1098. if (group_descriptor(group_index).bg_free_blocks_count) {
  1099. found_a_group = true;
  1100. } else {
  1101. if (group_index == preferred_group_index)
  1102. group_index = 1;
  1103. for (; group_index <= m_block_group_count; group_index = GroupIndex { group_index.value() + 1 }) {
  1104. if (group_descriptor(group_index).bg_free_blocks_count) {
  1105. found_a_group = true;
  1106. break;
  1107. }
  1108. }
  1109. }
  1110. VERIFY(found_a_group);
  1111. auto& bgd = group_descriptor(group_index);
  1112. auto cached_bitmap_or_error = get_bitmap_block(bgd.bg_block_bitmap);
  1113. if (cached_bitmap_or_error.is_error())
  1114. return cached_bitmap_or_error.error();
  1115. auto& cached_bitmap = *cached_bitmap_or_error.value();
  1116. int blocks_in_group = min(blocks_per_group(), super_block().s_blocks_count);
  1117. auto block_bitmap = cached_bitmap.bitmap(blocks_in_group);
  1118. BlockIndex first_block_in_group = (group_index.value() - 1) * blocks_per_group() + first_block_index().value();
  1119. size_t free_region_size = 0;
  1120. auto first_unset_bit_index = block_bitmap.find_longest_range_of_unset_bits(count - blocks.size(), free_region_size);
  1121. VERIFY(first_unset_bit_index.has_value());
  1122. dbgln_if(EXT2_DEBUG, "Ext2FS: allocating free region of size: {} [{}]", free_region_size, group_index);
  1123. for (size_t i = 0; i < free_region_size; ++i) {
  1124. BlockIndex block_index = (first_unset_bit_index.value() + i) + first_block_in_group.value();
  1125. if (auto result = set_block_allocation_state(block_index, true); result.is_error()) {
  1126. dbgln("Ext2FS: Failed to allocate block {} in allocate_blocks()", block_index);
  1127. return result;
  1128. }
  1129. blocks.unchecked_append(block_index);
  1130. dbgln_if(EXT2_DEBUG, " allocated > {}", block_index);
  1131. }
  1132. }
  1133. VERIFY(blocks.size() == count);
  1134. return blocks;
  1135. }
  1136. KResultOr<InodeIndex> Ext2FS::allocate_inode(GroupIndex preferred_group)
  1137. {
  1138. dbgln_if(EXT2_DEBUG, "Ext2FS: allocate_inode(preferred_group: {})", preferred_group);
  1139. Locker locker(m_lock);
  1140. // FIXME: We shouldn't refuse to allocate an inode if there is no group that can house the whole thing.
  1141. // In those cases we should just spread it across multiple groups.
  1142. auto is_suitable_group = [this](auto group_index) {
  1143. auto& bgd = group_descriptor(group_index);
  1144. return bgd.bg_free_inodes_count && bgd.bg_free_blocks_count >= 1;
  1145. };
  1146. GroupIndex group_index;
  1147. if (preferred_group.value() && is_suitable_group(preferred_group)) {
  1148. group_index = preferred_group;
  1149. } else {
  1150. for (unsigned i = 1; i <= m_block_group_count; ++i) {
  1151. if (is_suitable_group(i)) {
  1152. group_index = i;
  1153. break;
  1154. }
  1155. }
  1156. }
  1157. if (!group_index) {
  1158. dmesgln("Ext2FS: allocate_inode: no suitable group found for new inode");
  1159. return ENOSPC;
  1160. }
  1161. dbgln_if(EXT2_DEBUG, "Ext2FS: allocate_inode: found suitable group [{}] for new inode :^)", group_index);
  1162. auto& bgd = group_descriptor(group_index);
  1163. unsigned inodes_in_group = min(inodes_per_group(), super_block().s_inodes_count);
  1164. InodeIndex first_inode_in_group = (group_index.value() - 1) * inodes_per_group() + 1;
  1165. auto cached_bitmap_or_error = get_bitmap_block(bgd.bg_inode_bitmap);
  1166. if (cached_bitmap_or_error.is_error())
  1167. return cached_bitmap_or_error.error();
  1168. auto& cached_bitmap = *cached_bitmap_or_error.value();
  1169. auto inode_bitmap = cached_bitmap.bitmap(inodes_in_group);
  1170. for (size_t i = 0; i < inode_bitmap.size(); ++i) {
  1171. if (inode_bitmap.get(i))
  1172. continue;
  1173. inode_bitmap.set(i, true);
  1174. auto inode_index = InodeIndex(first_inode_in_group.value() + i);
  1175. cached_bitmap.dirty = true;
  1176. m_super_block.s_free_inodes_count--;
  1177. m_super_block_dirty = true;
  1178. const_cast<ext2_group_desc&>(bgd).bg_free_inodes_count--;
  1179. m_block_group_descriptors_dirty = true;
  1180. // In case the inode cache had this cached as "non-existent", uncache that info.
  1181. m_inode_cache.remove(inode_index.value());
  1182. return inode_index;
  1183. }
  1184. dmesgln("Ext2FS: allocate_inode found no available inode, despite bgd claiming there are inodes :(");
  1185. return EIO;
  1186. }
  1187. Ext2FS::GroupIndex Ext2FS::group_index_from_block_index(BlockIndex block_index) const
  1188. {
  1189. if (!block_index)
  1190. return 0;
  1191. return (block_index.value() - 1) / blocks_per_group() + 1;
  1192. }
  1193. auto Ext2FS::group_index_from_inode(InodeIndex inode) const -> GroupIndex
  1194. {
  1195. if (!inode)
  1196. return 0;
  1197. return (inode.value() - 1) / inodes_per_group() + 1;
  1198. }
  1199. KResultOr<bool> Ext2FS::get_inode_allocation_state(InodeIndex index) const
  1200. {
  1201. Locker locker(m_lock);
  1202. if (index == 0)
  1203. return EINVAL;
  1204. auto group_index = group_index_from_inode(index);
  1205. auto& bgd = group_descriptor(group_index);
  1206. unsigned index_in_group = index.value() - ((group_index.value() - 1) * inodes_per_group());
  1207. unsigned bit_index = (index_in_group - 1) % inodes_per_group();
  1208. auto cached_bitmap_or_error = const_cast<Ext2FS&>(*this).get_bitmap_block(bgd.bg_inode_bitmap);
  1209. if (cached_bitmap_or_error.is_error())
  1210. return cached_bitmap_or_error.error();
  1211. return cached_bitmap_or_error.value()->bitmap(inodes_per_group()).get(bit_index);
  1212. }
  1213. KResult Ext2FS::update_bitmap_block(BlockIndex bitmap_block, size_t bit_index, bool new_state, u32& super_block_counter, u16& group_descriptor_counter)
  1214. {
  1215. auto cached_bitmap_or_error = get_bitmap_block(bitmap_block);
  1216. if (cached_bitmap_or_error.is_error())
  1217. return cached_bitmap_or_error.error();
  1218. auto& cached_bitmap = *cached_bitmap_or_error.value();
  1219. bool current_state = cached_bitmap.bitmap(blocks_per_group()).get(bit_index);
  1220. if (current_state == new_state) {
  1221. dbgln("Ext2FS: Bit {} in bitmap block {} had unexpected state {}", bit_index, bitmap_block, current_state);
  1222. return EIO;
  1223. }
  1224. cached_bitmap.bitmap(blocks_per_group()).set(bit_index, new_state);
  1225. cached_bitmap.dirty = true;
  1226. if (new_state) {
  1227. --super_block_counter;
  1228. --group_descriptor_counter;
  1229. } else {
  1230. ++super_block_counter;
  1231. ++group_descriptor_counter;
  1232. }
  1233. m_super_block_dirty = true;
  1234. m_block_group_descriptors_dirty = true;
  1235. return KSuccess;
  1236. }
  1237. KResult Ext2FS::set_inode_allocation_state(InodeIndex inode_index, bool new_state)
  1238. {
  1239. Locker locker(m_lock);
  1240. auto group_index = group_index_from_inode(inode_index);
  1241. unsigned index_in_group = inode_index.value() - ((group_index.value() - 1) * inodes_per_group());
  1242. unsigned bit_index = (index_in_group - 1) % inodes_per_group();
  1243. dbgln_if(EXT2_DEBUG, "Ext2FS: set_inode_allocation_state: Inode {} -> {}", inode_index, new_state);
  1244. auto& bgd = const_cast<ext2_group_desc&>(group_descriptor(group_index));
  1245. return update_bitmap_block(bgd.bg_inode_bitmap, bit_index, new_state, m_super_block.s_free_inodes_count, bgd.bg_free_inodes_count);
  1246. }
  1247. Ext2FS::BlockIndex Ext2FS::first_block_index() const
  1248. {
  1249. return block_size() == 1024 ? 1 : 0;
  1250. }
  1251. KResultOr<Ext2FS::CachedBitmap*> Ext2FS::get_bitmap_block(BlockIndex bitmap_block_index)
  1252. {
  1253. for (auto& cached_bitmap : m_cached_bitmaps) {
  1254. if (cached_bitmap->bitmap_block_index == bitmap_block_index)
  1255. return cached_bitmap;
  1256. }
  1257. auto block = KBuffer::create_with_size(block_size(), Region::Access::Read | Region::Access::Write, "Ext2FS: Cached bitmap block");
  1258. auto buffer = UserOrKernelBuffer::for_kernel_buffer(block.data());
  1259. if (auto result = read_block(bitmap_block_index, &buffer, block_size()); result.is_error()) {
  1260. dbgln("Ext2FS: Failed to load bitmap block {}", bitmap_block_index);
  1261. return result;
  1262. }
  1263. auto new_bitmap = adopt_own_if_nonnull(new CachedBitmap(bitmap_block_index, move(block)));
  1264. if (!new_bitmap)
  1265. return ENOMEM;
  1266. if (!m_cached_bitmaps.try_append(move(new_bitmap)))
  1267. return ENOMEM;
  1268. return m_cached_bitmaps.last();
  1269. }
  1270. KResult Ext2FS::set_block_allocation_state(BlockIndex block_index, bool new_state)
  1271. {
  1272. VERIFY(block_index != 0);
  1273. Locker locker(m_lock);
  1274. auto group_index = group_index_from_block_index(block_index);
  1275. unsigned index_in_group = (block_index.value() - first_block_index().value()) - ((group_index.value() - 1) * blocks_per_group());
  1276. unsigned bit_index = index_in_group % blocks_per_group();
  1277. auto& bgd = const_cast<ext2_group_desc&>(group_descriptor(group_index));
  1278. dbgln_if(EXT2_DEBUG, "Ext2FS: Block {} state -> {} (in bitmap block {})", block_index, new_state, bgd.bg_block_bitmap);
  1279. return update_bitmap_block(bgd.bg_block_bitmap, bit_index, new_state, m_super_block.s_free_blocks_count, bgd.bg_free_blocks_count);
  1280. }
  1281. KResult Ext2FS::create_directory(Ext2FSInode& parent_inode, const String& name, mode_t mode, uid_t uid, gid_t gid)
  1282. {
  1283. Locker locker(m_lock);
  1284. VERIFY(is_directory(mode));
  1285. auto inode_or_error = create_inode(parent_inode, name, mode, 0, uid, gid);
  1286. if (inode_or_error.is_error())
  1287. return inode_or_error.error();
  1288. auto& inode = inode_or_error.value();
  1289. dbgln_if(EXT2_DEBUG, "Ext2FS: create_directory: created new directory named '{} with inode {}", name, inode->index());
  1290. Vector<Ext2FSDirectoryEntry> entries;
  1291. entries.empend(".", inode->index(), static_cast<u8>(EXT2_FT_DIR));
  1292. entries.empend("..", parent_inode.index(), static_cast<u8>(EXT2_FT_DIR));
  1293. if (auto result = static_cast<Ext2FSInode&>(*inode).write_directory(entries); result.is_error())
  1294. return result;
  1295. if (auto result = parent_inode.increment_link_count(); result.is_error())
  1296. return result;
  1297. auto& bgd = const_cast<ext2_group_desc&>(group_descriptor(group_index_from_inode(inode->identifier().index())));
  1298. ++bgd.bg_used_dirs_count;
  1299. m_block_group_descriptors_dirty = true;
  1300. return KSuccess;
  1301. }
  1302. KResultOr<NonnullRefPtr<Inode>> Ext2FS::create_inode(Ext2FSInode& parent_inode, const String& name, mode_t mode, dev_t dev, uid_t uid, gid_t gid)
  1303. {
  1304. if (name.length() > EXT2_NAME_LEN)
  1305. return ENAMETOOLONG;
  1306. if (parent_inode.m_raw_inode.i_links_count == 0)
  1307. return ENOENT;
  1308. ext2_inode e2inode {};
  1309. auto now = kgettimeofday().to_truncated_seconds();
  1310. e2inode.i_mode = mode;
  1311. e2inode.i_uid = uid;
  1312. e2inode.i_gid = gid;
  1313. e2inode.i_size = 0;
  1314. e2inode.i_atime = now;
  1315. e2inode.i_ctime = now;
  1316. e2inode.i_mtime = now;
  1317. e2inode.i_dtime = 0;
  1318. e2inode.i_flags = 0;
  1319. // For directories, add +1 link count for the "." entry in self.
  1320. e2inode.i_links_count = is_directory(mode);
  1321. if (is_character_device(mode))
  1322. e2inode.i_block[0] = dev;
  1323. else if (is_block_device(mode))
  1324. e2inode.i_block[1] = dev;
  1325. auto inode_id = allocate_inode();
  1326. if (inode_id.is_error())
  1327. return inode_id.error();
  1328. dbgln_if(EXT2_DEBUG, "Ext2FS: writing initial metadata for inode {}", inode_id.value());
  1329. auto success = write_ext2_inode(inode_id.value(), e2inode);
  1330. VERIFY(success);
  1331. auto new_inode = get_inode({ fsid(), inode_id.value() });
  1332. VERIFY(new_inode);
  1333. dbgln_if(EXT2_DEBUG, "Ext2FS: Adding inode '{}' (mode {:o}) to parent directory {}", name, mode, parent_inode.index());
  1334. if (auto result = parent_inode.add_child(*new_inode, name, mode); result.is_error())
  1335. return result;
  1336. return new_inode.release_nonnull();
  1337. }
  1338. bool Ext2FSInode::populate_lookup_cache() const
  1339. {
  1340. Locker locker(m_lock);
  1341. if (!m_lookup_cache.is_empty())
  1342. return true;
  1343. HashMap<String, InodeIndex> children;
  1344. KResult result = traverse_as_directory([&children](auto& entry) {
  1345. children.set(entry.name, entry.inode.index());
  1346. return true;
  1347. });
  1348. if (!result.is_success())
  1349. return false;
  1350. if (!m_lookup_cache.is_empty())
  1351. return false;
  1352. m_lookup_cache = move(children);
  1353. return true;
  1354. }
  1355. RefPtr<Inode> Ext2FSInode::lookup(StringView name)
  1356. {
  1357. VERIFY(is_directory());
  1358. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]:lookup(): Looking up '{}'", identifier(), name);
  1359. if (!populate_lookup_cache())
  1360. return {};
  1361. Locker locker(m_lock);
  1362. auto it = m_lookup_cache.find(name.hash(), [&](auto& entry) { return entry.key == name; });
  1363. if (it != m_lookup_cache.end())
  1364. return fs().get_inode({ fsid(), (*it).value });
  1365. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]:lookup(): '{}' not found", identifier(), name);
  1366. return {};
  1367. }
  1368. void Ext2FSInode::one_ref_left()
  1369. {
  1370. // FIXME: I would like to not live forever, but uncached Ext2FS is fucking painful right now.
  1371. }
  1372. KResult Ext2FSInode::set_atime(time_t t)
  1373. {
  1374. Locker locker(m_lock);
  1375. if (fs().is_readonly())
  1376. return EROFS;
  1377. m_raw_inode.i_atime = t;
  1378. set_metadata_dirty(true);
  1379. return KSuccess;
  1380. }
  1381. KResult Ext2FSInode::set_ctime(time_t t)
  1382. {
  1383. Locker locker(m_lock);
  1384. if (fs().is_readonly())
  1385. return EROFS;
  1386. m_raw_inode.i_ctime = t;
  1387. set_metadata_dirty(true);
  1388. return KSuccess;
  1389. }
  1390. KResult Ext2FSInode::set_mtime(time_t t)
  1391. {
  1392. Locker locker(m_lock);
  1393. if (fs().is_readonly())
  1394. return EROFS;
  1395. m_raw_inode.i_mtime = t;
  1396. set_metadata_dirty(true);
  1397. return KSuccess;
  1398. }
  1399. KResult Ext2FSInode::increment_link_count()
  1400. {
  1401. Locker locker(m_lock);
  1402. if (fs().is_readonly())
  1403. return EROFS;
  1404. if (m_raw_inode.i_links_count == max_link_count)
  1405. return EMLINK;
  1406. ++m_raw_inode.i_links_count;
  1407. set_metadata_dirty(true);
  1408. return KSuccess;
  1409. }
  1410. KResult Ext2FSInode::decrement_link_count()
  1411. {
  1412. Locker locker(m_lock);
  1413. if (fs().is_readonly())
  1414. return EROFS;
  1415. VERIFY(m_raw_inode.i_links_count);
  1416. --m_raw_inode.i_links_count;
  1417. set_metadata_dirty(true);
  1418. if (m_raw_inode.i_links_count == 0)
  1419. did_delete_self();
  1420. if (ref_count() == 1 && m_raw_inode.i_links_count == 0)
  1421. fs().uncache_inode(index());
  1422. return KSuccess;
  1423. }
  1424. void Ext2FS::uncache_inode(InodeIndex index)
  1425. {
  1426. Locker locker(m_lock);
  1427. m_inode_cache.remove(index);
  1428. }
  1429. KResultOr<size_t> Ext2FSInode::directory_entry_count() const
  1430. {
  1431. VERIFY(is_directory());
  1432. Locker locker(m_lock);
  1433. populate_lookup_cache();
  1434. return m_lookup_cache.size();
  1435. }
  1436. KResult Ext2FSInode::chmod(mode_t mode)
  1437. {
  1438. Locker locker(m_lock);
  1439. if (m_raw_inode.i_mode == mode)
  1440. return KSuccess;
  1441. m_raw_inode.i_mode = mode;
  1442. set_metadata_dirty(true);
  1443. return KSuccess;
  1444. }
  1445. KResult Ext2FSInode::chown(uid_t uid, gid_t gid)
  1446. {
  1447. Locker locker(m_lock);
  1448. if (m_raw_inode.i_uid == uid && m_raw_inode.i_gid == gid)
  1449. return KSuccess;
  1450. m_raw_inode.i_uid = uid;
  1451. m_raw_inode.i_gid = gid;
  1452. set_metadata_dirty(true);
  1453. return KSuccess;
  1454. }
  1455. KResult Ext2FSInode::truncate(u64 size)
  1456. {
  1457. Locker locker(m_lock);
  1458. if (static_cast<u64>(m_raw_inode.i_size) == size)
  1459. return KSuccess;
  1460. if (auto result = resize(size); result.is_error())
  1461. return result;
  1462. set_metadata_dirty(true);
  1463. return KSuccess;
  1464. }
  1465. KResultOr<int> Ext2FSInode::get_block_address(int index)
  1466. {
  1467. Locker locker(m_lock);
  1468. if (m_block_list.is_empty())
  1469. m_block_list = compute_block_list();
  1470. if (index < 0 || (size_t)index >= m_block_list.size())
  1471. return 0;
  1472. return m_block_list[index].value();
  1473. }
  1474. unsigned Ext2FS::total_block_count() const
  1475. {
  1476. Locker locker(m_lock);
  1477. return super_block().s_blocks_count;
  1478. }
  1479. unsigned Ext2FS::free_block_count() const
  1480. {
  1481. Locker locker(m_lock);
  1482. return super_block().s_free_blocks_count;
  1483. }
  1484. unsigned Ext2FS::total_inode_count() const
  1485. {
  1486. Locker locker(m_lock);
  1487. return super_block().s_inodes_count;
  1488. }
  1489. unsigned Ext2FS::free_inode_count() const
  1490. {
  1491. Locker locker(m_lock);
  1492. return super_block().s_free_inodes_count;
  1493. }
  1494. KResult Ext2FS::prepare_to_unmount() const
  1495. {
  1496. Locker locker(m_lock);
  1497. for (auto& it : m_inode_cache) {
  1498. if (it.value->ref_count() > 1)
  1499. return EBUSY;
  1500. }
  1501. m_inode_cache.clear();
  1502. return KSuccess;
  1503. }
  1504. }