Ext2FileSystem.cpp 69 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767
  1. /*
  2. * Copyright (c) 2018-2020, Andreas Kling <kling@serenityos.org>
  3. *
  4. * SPDX-License-Identifier: BSD-2-Clause
  5. */
  6. #include <AK/HashMap.h>
  7. #include <AK/MemoryStream.h>
  8. #include <AK/StdLibExtras.h>
  9. #include <AK/StringView.h>
  10. #include <Kernel/Debug.h>
  11. #include <Kernel/Devices/BlockDevice.h>
  12. #include <Kernel/FileSystem/Ext2FileSystem.h>
  13. #include <Kernel/FileSystem/FileDescription.h>
  14. #include <Kernel/FileSystem/ext2_fs.h>
  15. #include <Kernel/Process.h>
  16. #include <Kernel/UnixTypes.h>
  17. #include <LibC/errno_numbers.h>
  18. namespace Kernel {
  19. static const size_t max_link_count = 65535;
  20. static const size_t max_block_size = 4096;
  21. static const ssize_t max_inline_symlink_length = 60;
  22. struct Ext2FSDirectoryEntry {
  23. String name;
  24. InodeIndex inode_index { 0 };
  25. u8 file_type { 0 };
  26. u16 record_length { 0 };
  27. };
  28. static u8 to_ext2_file_type(mode_t mode)
  29. {
  30. if (is_regular_file(mode))
  31. return EXT2_FT_REG_FILE;
  32. if (is_directory(mode))
  33. return EXT2_FT_DIR;
  34. if (is_character_device(mode))
  35. return EXT2_FT_CHRDEV;
  36. if (is_block_device(mode))
  37. return EXT2_FT_BLKDEV;
  38. if (is_fifo(mode))
  39. return EXT2_FT_FIFO;
  40. if (is_socket(mode))
  41. return EXT2_FT_SOCK;
  42. if (is_symlink(mode))
  43. return EXT2_FT_SYMLINK;
  44. return EXT2_FT_UNKNOWN;
  45. }
  46. static unsigned divide_rounded_up(unsigned a, unsigned b)
  47. {
  48. return (a / b) + (a % b != 0);
  49. }
  50. NonnullRefPtr<Ext2FS> Ext2FS::create(FileDescription& file_description)
  51. {
  52. return adopt_ref(*new Ext2FS(file_description));
  53. }
  54. Ext2FS::Ext2FS(FileDescription& file_description)
  55. : BlockBasedFS(file_description)
  56. {
  57. }
  58. Ext2FS::~Ext2FS()
  59. {
  60. }
  61. bool Ext2FS::flush_super_block()
  62. {
  63. Locker locker(m_lock);
  64. VERIFY((sizeof(ext2_super_block) % logical_block_size()) == 0);
  65. auto super_block_buffer = UserOrKernelBuffer::for_kernel_buffer((u8*)&m_super_block);
  66. bool success = raw_write_blocks(2, (sizeof(ext2_super_block) / logical_block_size()), super_block_buffer);
  67. VERIFY(success);
  68. return true;
  69. }
  70. const ext2_group_desc& Ext2FS::group_descriptor(GroupIndex group_index) const
  71. {
  72. // FIXME: Should this fail gracefully somehow?
  73. VERIFY(group_index <= m_block_group_count);
  74. VERIFY(group_index > 0);
  75. return block_group_descriptors()[group_index.value() - 1];
  76. }
  77. bool Ext2FS::initialize()
  78. {
  79. Locker locker(m_lock);
  80. VERIFY((sizeof(ext2_super_block) % logical_block_size()) == 0);
  81. auto super_block_buffer = UserOrKernelBuffer::for_kernel_buffer((u8*)&m_super_block);
  82. bool success = raw_read_blocks(2, (sizeof(ext2_super_block) / logical_block_size()), super_block_buffer);
  83. VERIFY(success);
  84. auto& super_block = this->super_block();
  85. if constexpr (EXT2_DEBUG) {
  86. dmesgln("Ext2FS: super block magic: {:04x} (super block size: {})", super_block.s_magic, sizeof(ext2_super_block));
  87. }
  88. if (super_block.s_magic != EXT2_SUPER_MAGIC)
  89. return false;
  90. if constexpr (EXT2_DEBUG) {
  91. dmesgln("Ext2FS: {} inodes, {} blocks", super_block.s_inodes_count, super_block.s_blocks_count);
  92. dmesgln("Ext2FS: Block size: {}", EXT2_BLOCK_SIZE(&super_block));
  93. dmesgln("Ext2FS: First data block: {}", super_block.s_first_data_block);
  94. dmesgln("Ext2FS: Inodes per block: {}", inodes_per_block());
  95. dmesgln("Ext2FS: Inodes per group: {}", inodes_per_group());
  96. dmesgln("Ext2FS: Free inodes: {}", super_block.s_free_inodes_count);
  97. dmesgln("Ext2FS: Descriptors per block: {}", EXT2_DESC_PER_BLOCK(&super_block));
  98. dmesgln("Ext2FS: Descriptor size: {}", EXT2_DESC_SIZE(&super_block));
  99. }
  100. set_block_size(EXT2_BLOCK_SIZE(&super_block));
  101. VERIFY(block_size() <= (int)max_block_size);
  102. m_block_group_count = ceil_div(super_block.s_blocks_count, super_block.s_blocks_per_group);
  103. if (m_block_group_count == 0) {
  104. dmesgln("Ext2FS: no block groups :(");
  105. return false;
  106. }
  107. unsigned blocks_to_read = ceil_div(m_block_group_count * sizeof(ext2_group_desc), block_size());
  108. BlockIndex first_block_of_bgdt = block_size() == 1024 ? 2 : 1;
  109. 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");
  110. if (!m_cached_group_descriptor_table) {
  111. dbgln("Ext2FS: Failed to allocate memory for group descriptor table");
  112. return false;
  113. }
  114. auto buffer = UserOrKernelBuffer::for_kernel_buffer(m_cached_group_descriptor_table->data());
  115. if (auto result = read_blocks(first_block_of_bgdt, blocks_to_read, buffer); result.is_error()) {
  116. // FIXME: Propagate the error
  117. dbgln("Ext2FS: initialize had error: {}", result.error());
  118. return false;
  119. }
  120. if constexpr (EXT2_DEBUG) {
  121. for (unsigned i = 1; i <= m_block_group_count; ++i) {
  122. auto& group = group_descriptor(i);
  123. dbgln("Ext2FS: group[{}] ( block_bitmap: {}, inode_bitmap: {}, inode_table: {} )", i, group.bg_block_bitmap, group.bg_inode_bitmap, group.bg_inode_table);
  124. }
  125. }
  126. return true;
  127. }
  128. const char* Ext2FS::class_name() const
  129. {
  130. return "Ext2FS";
  131. }
  132. NonnullRefPtr<Inode> Ext2FS::root_inode() const
  133. {
  134. return *get_inode({ fsid(), EXT2_ROOT_INO });
  135. }
  136. bool Ext2FS::find_block_containing_inode(InodeIndex inode, BlockIndex& block_index, unsigned& offset) const
  137. {
  138. auto& super_block = this->super_block();
  139. if (inode != EXT2_ROOT_INO && inode < EXT2_FIRST_INO(&super_block))
  140. return false;
  141. if (inode > super_block.s_inodes_count)
  142. return false;
  143. auto& bgd = group_descriptor(group_index_from_inode(inode));
  144. offset = ((inode.value() - 1) % inodes_per_group()) * inode_size();
  145. block_index = bgd.bg_inode_table + (offset >> EXT2_BLOCK_SIZE_BITS(&super_block));
  146. offset &= block_size() - 1;
  147. return true;
  148. }
  149. Ext2FS::BlockListShape Ext2FS::compute_block_list_shape(unsigned blocks) const
  150. {
  151. BlockListShape shape;
  152. const unsigned entries_per_block = EXT2_ADDR_PER_BLOCK(&super_block());
  153. unsigned blocks_remaining = blocks;
  154. shape.direct_blocks = min((unsigned)EXT2_NDIR_BLOCKS, blocks_remaining);
  155. blocks_remaining -= shape.direct_blocks;
  156. if (!blocks_remaining)
  157. return shape;
  158. shape.indirect_blocks = min(blocks_remaining, entries_per_block);
  159. shape.meta_blocks += 1;
  160. blocks_remaining -= shape.indirect_blocks;
  161. if (!blocks_remaining)
  162. return shape;
  163. shape.doubly_indirect_blocks = min(blocks_remaining, entries_per_block * entries_per_block);
  164. shape.meta_blocks += 1;
  165. shape.meta_blocks += divide_rounded_up(shape.doubly_indirect_blocks, entries_per_block);
  166. blocks_remaining -= shape.doubly_indirect_blocks;
  167. if (!blocks_remaining)
  168. return shape;
  169. shape.triply_indirect_blocks = min(blocks_remaining, entries_per_block * entries_per_block * entries_per_block);
  170. shape.meta_blocks += 1;
  171. shape.meta_blocks += divide_rounded_up(shape.triply_indirect_blocks, entries_per_block * entries_per_block);
  172. shape.meta_blocks += divide_rounded_up(shape.triply_indirect_blocks, entries_per_block);
  173. blocks_remaining -= shape.triply_indirect_blocks;
  174. VERIFY(blocks_remaining == 0);
  175. return shape;
  176. }
  177. KResult Ext2FSInode::write_indirect_block(BlockBasedFS::BlockIndex block, Span<BlockBasedFS::BlockIndex> blocks_indices)
  178. {
  179. const auto entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  180. VERIFY(blocks_indices.size() <= entries_per_block);
  181. auto block_contents = ByteBuffer::create_uninitialized(fs().block_size());
  182. OutputMemoryStream stream { block_contents };
  183. auto buffer = UserOrKernelBuffer::for_kernel_buffer(stream.data());
  184. VERIFY(blocks_indices.size() <= EXT2_ADDR_PER_BLOCK(&fs().super_block()));
  185. for (unsigned i = 0; i < blocks_indices.size(); ++i)
  186. stream << static_cast<u32>(blocks_indices[i].value());
  187. stream.fill_to_end(0);
  188. return fs().write_block(block, buffer, stream.size());
  189. }
  190. 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)
  191. {
  192. const auto entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  193. const auto entries_per_doubly_indirect_block = entries_per_block * entries_per_block;
  194. const auto old_indirect_blocks_length = divide_rounded_up(old_blocks_length, entries_per_block);
  195. const auto new_indirect_blocks_length = divide_rounded_up(blocks_indices.size(), entries_per_block);
  196. VERIFY(blocks_indices.size() > 0);
  197. VERIFY(blocks_indices.size() > old_blocks_length);
  198. VERIFY(blocks_indices.size() <= entries_per_doubly_indirect_block);
  199. auto block_contents = ByteBuffer::create_uninitialized(fs().block_size());
  200. auto* block_as_pointers = (unsigned*)block_contents.data();
  201. OutputMemoryStream stream { block_contents };
  202. auto buffer = UserOrKernelBuffer::for_kernel_buffer(stream.data());
  203. if (old_blocks_length > 0) {
  204. if (auto result = fs().read_block(block, &buffer, fs().block_size()); result.is_error())
  205. return result;
  206. }
  207. // Grow the doubly indirect block.
  208. for (unsigned i = 0; i < old_indirect_blocks_length; i++)
  209. stream << static_cast<u32>(block_as_pointers[i]);
  210. for (unsigned i = old_indirect_blocks_length; i < new_indirect_blocks_length; i++) {
  211. auto new_block = new_meta_blocks.take_last().value();
  212. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::grow_doubly_indirect_block(): Allocating indirect block {} at index {}", identifier(), new_block, i);
  213. stream << static_cast<u32>(new_block);
  214. meta_blocks++;
  215. }
  216. stream.fill_to_end(0);
  217. // Write out the indirect blocks.
  218. for (unsigned i = old_blocks_length / entries_per_block; i < new_indirect_blocks_length; i++) {
  219. const auto offset_block = i * entries_per_block;
  220. 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())
  221. return result;
  222. }
  223. // Write out the doubly indirect block.
  224. return fs().write_block(block, buffer, stream.size());
  225. }
  226. KResult Ext2FSInode::shrink_doubly_indirect_block(BlockBasedFS::BlockIndex block, size_t old_blocks_length, size_t new_blocks_length, unsigned& meta_blocks)
  227. {
  228. const auto entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  229. const auto entries_per_doubly_indirect_block = entries_per_block * entries_per_block;
  230. const auto old_indirect_blocks_length = divide_rounded_up(old_blocks_length, entries_per_block);
  231. const auto new_indirect_blocks_length = divide_rounded_up(new_blocks_length, entries_per_block);
  232. VERIFY(old_blocks_length > 0);
  233. VERIFY(old_blocks_length >= new_blocks_length);
  234. VERIFY(new_blocks_length <= entries_per_doubly_indirect_block);
  235. auto block_contents = ByteBuffer::create_uninitialized(fs().block_size());
  236. auto* block_as_pointers = (unsigned*)block_contents.data();
  237. auto buffer = UserOrKernelBuffer::for_kernel_buffer(reinterpret_cast<u8*>(block_as_pointers));
  238. if (auto result = fs().read_block(block, &buffer, fs().block_size()); result.is_error())
  239. return result;
  240. // Free the unused indirect blocks.
  241. for (unsigned i = new_indirect_blocks_length; i < old_indirect_blocks_length; i++) {
  242. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::shrink_doubly_indirect_block(): Freeing indirect block {} at index {}", identifier(), block_as_pointers[i], i);
  243. if (auto result = fs().set_block_allocation_state(block_as_pointers[i], false); result.is_error())
  244. return result;
  245. meta_blocks--;
  246. }
  247. // Free the doubly indirect block if no longer needed.
  248. if (new_blocks_length == 0) {
  249. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::shrink_doubly_indirect_block(): Freeing doubly indirect block {}", identifier(), block);
  250. if (auto result = fs().set_block_allocation_state(block, false); result.is_error())
  251. return result;
  252. meta_blocks--;
  253. }
  254. return KSuccess;
  255. }
  256. 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)
  257. {
  258. const auto entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  259. const auto entries_per_doubly_indirect_block = entries_per_block * entries_per_block;
  260. const auto entries_per_triply_indirect_block = entries_per_block * entries_per_block;
  261. const auto old_doubly_indirect_blocks_length = divide_rounded_up(old_blocks_length, entries_per_doubly_indirect_block);
  262. const auto new_doubly_indirect_blocks_length = divide_rounded_up(blocks_indices.size(), entries_per_doubly_indirect_block);
  263. VERIFY(blocks_indices.size() > 0);
  264. VERIFY(blocks_indices.size() > old_blocks_length);
  265. VERIFY(blocks_indices.size() <= entries_per_triply_indirect_block);
  266. auto block_contents = ByteBuffer::create_uninitialized(fs().block_size());
  267. auto* block_as_pointers = (unsigned*)block_contents.data();
  268. OutputMemoryStream stream { block_contents };
  269. auto buffer = UserOrKernelBuffer::for_kernel_buffer(stream.data());
  270. if (old_blocks_length > 0) {
  271. if (auto result = fs().read_block(block, &buffer, fs().block_size()); result.is_error())
  272. return result;
  273. }
  274. // Grow the triply indirect block.
  275. for (unsigned i = 0; i < old_doubly_indirect_blocks_length; i++)
  276. stream << static_cast<u32>(block_as_pointers[i]);
  277. for (unsigned i = old_doubly_indirect_blocks_length; i < new_doubly_indirect_blocks_length; i++) {
  278. auto new_block = new_meta_blocks.take_last().value();
  279. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::grow_triply_indirect_block(): Allocating doubly indirect block {} at index {}", identifier(), new_block, i);
  280. stream << static_cast<u32>(new_block);
  281. meta_blocks++;
  282. }
  283. stream.fill_to_end(0);
  284. // Write out the doubly indirect blocks.
  285. for (unsigned i = old_blocks_length / entries_per_doubly_indirect_block; i < new_doubly_indirect_blocks_length; i++) {
  286. const auto processed_blocks = i * entries_per_doubly_indirect_block;
  287. const auto old_doubly_indirect_blocks_length = min(old_blocks_length > processed_blocks ? old_blocks_length - processed_blocks : 0, entries_per_doubly_indirect_block);
  288. const auto new_doubly_indirect_blocks_length = min(blocks_indices.size() > processed_blocks ? blocks_indices.size() - processed_blocks : 0, entries_per_doubly_indirect_block);
  289. 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())
  290. return result;
  291. }
  292. // Write out the triply indirect block.
  293. return fs().write_block(block, buffer, stream.size());
  294. }
  295. KResult Ext2FSInode::shrink_triply_indirect_block(BlockBasedFS::BlockIndex block, size_t old_blocks_length, size_t new_blocks_length, unsigned& meta_blocks)
  296. {
  297. const auto entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  298. const auto entries_per_doubly_indirect_block = entries_per_block * entries_per_block;
  299. const auto entries_per_triply_indirect_block = entries_per_doubly_indirect_block * entries_per_block;
  300. const auto old_triply_indirect_blocks_length = divide_rounded_up(old_blocks_length, entries_per_doubly_indirect_block);
  301. const auto new_triply_indirect_blocks_length = new_blocks_length / entries_per_doubly_indirect_block;
  302. VERIFY(old_blocks_length > 0);
  303. VERIFY(old_blocks_length >= new_blocks_length);
  304. VERIFY(new_blocks_length <= entries_per_triply_indirect_block);
  305. auto block_contents = ByteBuffer::create_uninitialized(fs().block_size());
  306. auto* block_as_pointers = (unsigned*)block_contents.data();
  307. auto buffer = UserOrKernelBuffer::for_kernel_buffer(reinterpret_cast<u8*>(block_as_pointers));
  308. if (auto result = fs().read_block(block, &buffer, fs().block_size()); result.is_error())
  309. return result;
  310. // Shrink the doubly indirect blocks.
  311. for (unsigned i = new_triply_indirect_blocks_length; i < old_triply_indirect_blocks_length; i++) {
  312. const auto processed_blocks = i * entries_per_doubly_indirect_block;
  313. const auto old_doubly_indirect_blocks_length = min(old_blocks_length > processed_blocks ? old_blocks_length - processed_blocks : 0, entries_per_doubly_indirect_block);
  314. const auto new_doubly_indirect_blocks_length = min(new_blocks_length > processed_blocks ? new_blocks_length - processed_blocks : 0, entries_per_doubly_indirect_block);
  315. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::shrink_triply_indirect_block(): Shrinking doubly indirect block {} at index {}", identifier(), block_as_pointers[i], i);
  316. 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())
  317. return result;
  318. }
  319. // Free the triply indirect block if no longer needed.
  320. if (new_blocks_length == 0) {
  321. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::shrink_triply_indirect_block(): Freeing triply indirect block {}", identifier(), block);
  322. if (auto result = fs().set_block_allocation_state(block, false); result.is_error())
  323. return result;
  324. meta_blocks--;
  325. }
  326. return KSuccess;
  327. }
  328. KResult Ext2FSInode::flush_block_list()
  329. {
  330. Locker locker(m_lock);
  331. if (m_block_list.is_empty()) {
  332. m_raw_inode.i_blocks = 0;
  333. memset(m_raw_inode.i_block, 0, sizeof(m_raw_inode.i_block));
  334. set_metadata_dirty(true);
  335. return KSuccess;
  336. }
  337. // NOTE: There is a mismatch between i_blocks and blocks.size() since i_blocks includes meta blocks and blocks.size() does not.
  338. const auto old_block_count = ceil_div(size(), static_cast<u64>(fs().block_size()));
  339. auto old_shape = fs().compute_block_list_shape(old_block_count);
  340. const auto new_shape = fs().compute_block_list_shape(m_block_list.size());
  341. Vector<Ext2FS::BlockIndex> new_meta_blocks;
  342. if (new_shape.meta_blocks > old_shape.meta_blocks) {
  343. auto blocks_or_error = fs().allocate_blocks(fs().group_index_from_inode(index()), new_shape.meta_blocks - old_shape.meta_blocks);
  344. if (blocks_or_error.is_error())
  345. return blocks_or_error.error();
  346. new_meta_blocks = blocks_or_error.release_value();
  347. }
  348. m_raw_inode.i_blocks = (m_block_list.size() + new_shape.meta_blocks) * (fs().block_size() / 512);
  349. 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);
  350. unsigned output_block_index = 0;
  351. unsigned remaining_blocks = m_block_list.size();
  352. // Deal with direct blocks.
  353. bool inode_dirty = false;
  354. VERIFY(new_shape.direct_blocks <= EXT2_NDIR_BLOCKS);
  355. for (unsigned i = 0; i < new_shape.direct_blocks; ++i) {
  356. if (BlockBasedFS::BlockIndex(m_raw_inode.i_block[i]) != m_block_list[output_block_index])
  357. inode_dirty = true;
  358. m_raw_inode.i_block[i] = m_block_list[output_block_index].value();
  359. ++output_block_index;
  360. --remaining_blocks;
  361. }
  362. // e2fsck considers all blocks reachable through any of the pointers in
  363. // m_raw_inode.i_block as part of this inode regardless of the value in
  364. // m_raw_inode.i_size. When it finds more blocks than the amount that
  365. // is indicated by i_size or i_blocks it offers to repair the filesystem
  366. // by changing those values. That will actually cause further corruption.
  367. // So we must zero all pointers to blocks that are now unused.
  368. for (unsigned i = new_shape.direct_blocks; i < EXT2_NDIR_BLOCKS; ++i) {
  369. m_raw_inode.i_block[i] = 0;
  370. }
  371. if (inode_dirty) {
  372. if constexpr (EXT2_DEBUG) {
  373. 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());
  374. for (size_t i = 0; i < min((size_t)EXT2_NDIR_BLOCKS, m_block_list.size()); ++i)
  375. dbgln(" + {}", m_block_list[i]);
  376. }
  377. set_metadata_dirty(true);
  378. }
  379. // Deal with indirect blocks.
  380. if (old_shape.indirect_blocks != new_shape.indirect_blocks) {
  381. if (new_shape.indirect_blocks > old_shape.indirect_blocks) {
  382. // Write out the indirect block.
  383. if (old_shape.indirect_blocks == 0) {
  384. auto new_block = new_meta_blocks.take_last().value();
  385. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::flush_block_list(): Allocating indirect block: {}", identifier(), new_block);
  386. m_raw_inode.i_block[EXT2_IND_BLOCK] = new_block;
  387. set_metadata_dirty(true);
  388. old_shape.meta_blocks++;
  389. }
  390. 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())
  391. return result;
  392. } else if ((new_shape.indirect_blocks == 0) && (old_shape.indirect_blocks != 0)) {
  393. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::flush_block_list(): Freeing indirect block: {}", identifier(), m_raw_inode.i_block[EXT2_IND_BLOCK]);
  394. if (auto result = fs().set_block_allocation_state(m_raw_inode.i_block[EXT2_IND_BLOCK], false); result.is_error())
  395. return result;
  396. old_shape.meta_blocks--;
  397. m_raw_inode.i_block[EXT2_IND_BLOCK] = 0;
  398. }
  399. }
  400. remaining_blocks -= new_shape.indirect_blocks;
  401. output_block_index += new_shape.indirect_blocks;
  402. if (old_shape.doubly_indirect_blocks != new_shape.doubly_indirect_blocks) {
  403. // Write out the doubly indirect block.
  404. if (new_shape.doubly_indirect_blocks > old_shape.doubly_indirect_blocks) {
  405. if (old_shape.doubly_indirect_blocks == 0) {
  406. auto new_block = new_meta_blocks.take_last().value();
  407. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::flush_block_list(): Allocating doubly indirect block: {}", identifier(), new_block);
  408. m_raw_inode.i_block[EXT2_DIND_BLOCK] = new_block;
  409. set_metadata_dirty(true);
  410. old_shape.meta_blocks++;
  411. }
  412. 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())
  413. return result;
  414. } else {
  415. 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())
  416. return result;
  417. if (new_shape.doubly_indirect_blocks == 0)
  418. m_raw_inode.i_block[EXT2_DIND_BLOCK] = 0;
  419. }
  420. }
  421. remaining_blocks -= new_shape.doubly_indirect_blocks;
  422. output_block_index += new_shape.doubly_indirect_blocks;
  423. if (old_shape.triply_indirect_blocks != new_shape.triply_indirect_blocks) {
  424. // Write out the triply indirect block.
  425. if (new_shape.triply_indirect_blocks > old_shape.triply_indirect_blocks) {
  426. if (old_shape.triply_indirect_blocks == 0) {
  427. auto new_block = new_meta_blocks.take_last().value();
  428. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::flush_block_list(): Allocating triply indirect block: {}", identifier(), new_block);
  429. m_raw_inode.i_block[EXT2_TIND_BLOCK] = new_block;
  430. set_metadata_dirty(true);
  431. old_shape.meta_blocks++;
  432. }
  433. 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())
  434. return result;
  435. } else {
  436. 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())
  437. return result;
  438. if (new_shape.triply_indirect_blocks == 0)
  439. m_raw_inode.i_block[EXT2_TIND_BLOCK] = 0;
  440. }
  441. }
  442. remaining_blocks -= new_shape.triply_indirect_blocks;
  443. output_block_index += new_shape.triply_indirect_blocks;
  444. dbgln_if(EXT2_BLOCKLIST_DEBUG, "Ext2FSInode[{}]::flush_block_list(): New meta blocks count at {}, expecting {}", identifier(), old_shape.meta_blocks, new_shape.meta_blocks);
  445. VERIFY(new_meta_blocks.size() == 0);
  446. VERIFY(old_shape.meta_blocks == new_shape.meta_blocks);
  447. if (!remaining_blocks)
  448. return KSuccess;
  449. dbgln("we don't know how to write qind ext2fs blocks, they don't exist anyway!");
  450. VERIFY_NOT_REACHED();
  451. }
  452. Vector<Ext2FS::BlockIndex> Ext2FSInode::compute_block_list() const
  453. {
  454. return compute_block_list_impl(false);
  455. }
  456. Vector<Ext2FS::BlockIndex> Ext2FSInode::compute_block_list_with_meta_blocks() const
  457. {
  458. return compute_block_list_impl(true);
  459. }
  460. Vector<Ext2FS::BlockIndex> Ext2FSInode::compute_block_list_impl(bool include_block_list_blocks) const
  461. {
  462. // FIXME: This is really awkwardly factored.. foo_impl_internal :|
  463. auto block_list = compute_block_list_impl_internal(m_raw_inode, include_block_list_blocks);
  464. while (!block_list.is_empty() && block_list.last() == 0)
  465. block_list.take_last();
  466. return block_list;
  467. }
  468. Vector<Ext2FS::BlockIndex> Ext2FSInode::compute_block_list_impl_internal(const ext2_inode& e2inode, bool include_block_list_blocks) const
  469. {
  470. unsigned entries_per_block = EXT2_ADDR_PER_BLOCK(&fs().super_block());
  471. unsigned block_count = ceil_div(size(), static_cast<u64>(fs().block_size()));
  472. // If we are handling a symbolic link, the path is stored in the 60 bytes in
  473. // the inode that are used for the 12 direct and 3 indirect block pointers,
  474. // If the path is longer than 60 characters, a block is allocated, and the
  475. // block contains the destination path. The file size corresponds to the
  476. // path length of the destination.
  477. if (::is_symlink(e2inode.i_mode) && e2inode.i_blocks == 0)
  478. block_count = 0;
  479. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::block_list_for_inode(): i_size={}, i_blocks={}, block_count={}", identifier(), e2inode.i_size, e2inode.i_blocks, block_count);
  480. unsigned blocks_remaining = block_count;
  481. if (include_block_list_blocks) {
  482. auto shape = fs().compute_block_list_shape(block_count);
  483. blocks_remaining += shape.meta_blocks;
  484. }
  485. Vector<Ext2FS::BlockIndex> list;
  486. auto add_block = [&](auto bi) {
  487. if (blocks_remaining) {
  488. list.append(bi);
  489. --blocks_remaining;
  490. }
  491. };
  492. if (include_block_list_blocks) {
  493. // This seems like an excessive over-estimate but w/e.
  494. list.ensure_capacity(blocks_remaining * 2);
  495. } else {
  496. list.ensure_capacity(blocks_remaining);
  497. }
  498. unsigned direct_count = min(block_count, (unsigned)EXT2_NDIR_BLOCKS);
  499. for (unsigned i = 0; i < direct_count; ++i) {
  500. auto block_index = e2inode.i_block[i];
  501. add_block(block_index);
  502. }
  503. if (!blocks_remaining)
  504. return list;
  505. // Don't need to make copy of add_block, since this capture will only
  506. // be called before compute_block_list_impl_internal finishes.
  507. auto process_block_array = [&](auto array_block_index, auto&& callback) {
  508. if (include_block_list_blocks)
  509. add_block(array_block_index);
  510. auto count = min(blocks_remaining, entries_per_block);
  511. if (!count)
  512. return;
  513. size_t read_size = count * sizeof(u32);
  514. auto array_storage = ByteBuffer::create_uninitialized(read_size);
  515. auto* array = (u32*)array_storage.data();
  516. auto buffer = UserOrKernelBuffer::for_kernel_buffer((u8*)array);
  517. if (auto result = fs().read_block(array_block_index, &buffer, read_size, 0); result.is_error()) {
  518. // FIXME: Stop here and propagate this error.
  519. dbgln("Ext2FSInode[{}]::compute_block_list_impl_internal(): Error: {}", identifier(), result.error());
  520. }
  521. for (unsigned i = 0; i < count; ++i)
  522. callback(Ext2FS::BlockIndex(array[i]));
  523. };
  524. process_block_array(e2inode.i_block[EXT2_IND_BLOCK], [&](auto block_index) {
  525. add_block(block_index);
  526. });
  527. if (!blocks_remaining)
  528. return list;
  529. process_block_array(e2inode.i_block[EXT2_DIND_BLOCK], [&](auto block_index) {
  530. process_block_array(block_index, [&](auto block_index2) {
  531. add_block(block_index2);
  532. });
  533. });
  534. if (!blocks_remaining)
  535. return list;
  536. process_block_array(e2inode.i_block[EXT2_TIND_BLOCK], [&](auto block_index) {
  537. process_block_array(block_index, [&](auto block_index2) {
  538. process_block_array(block_index2, [&](auto block_index3) {
  539. add_block(block_index3);
  540. });
  541. });
  542. });
  543. return list;
  544. }
  545. void Ext2FS::free_inode(Ext2FSInode& inode)
  546. {
  547. Locker locker(m_lock);
  548. VERIFY(inode.m_raw_inode.i_links_count == 0);
  549. dbgln_if(EXT2_DEBUG, "Ext2FS[{}]::free_inode(): Inode {} has no more links, time to delete!", fsid(), inode.index());
  550. // Mark all blocks used by this inode as free.
  551. for (auto block_index : inode.compute_block_list_with_meta_blocks()) {
  552. VERIFY(block_index <= super_block().s_blocks_count);
  553. if (block_index.value()) {
  554. if (auto result = set_block_allocation_state(block_index, false); result.is_error()) {
  555. dbgln("Ext2FS[{}]::free_inode(): Failed to deallocate block {} for inode {}", fsid(), block_index, inode.index());
  556. }
  557. }
  558. }
  559. // If the inode being freed is a directory, update block group directory counter.
  560. if (inode.is_directory()) {
  561. auto& bgd = const_cast<ext2_group_desc&>(group_descriptor(group_index_from_inode(inode.index())));
  562. --bgd.bg_used_dirs_count;
  563. dbgln_if(EXT2_DEBUG, "Ext2FS[{}]::free_inode(): Decremented bg_used_dirs_count to {} for inode {}", fsid(), bgd.bg_used_dirs_count, inode.index());
  564. m_block_group_descriptors_dirty = true;
  565. }
  566. // NOTE: After this point, the inode metadata is wiped.
  567. memset(&inode.m_raw_inode, 0, sizeof(ext2_inode));
  568. inode.m_raw_inode.i_dtime = kgettimeofday().to_truncated_seconds();
  569. write_ext2_inode(inode.index(), inode.m_raw_inode);
  570. // Mark the inode as free.
  571. if (auto result = set_inode_allocation_state(inode.index(), false); result.is_error())
  572. dbgln("Ext2FS[{}]::free_inode(): Failed to free inode {}: {}", fsid(), inode.index(), result.error());
  573. }
  574. void Ext2FS::flush_block_group_descriptor_table()
  575. {
  576. Locker locker(m_lock);
  577. unsigned blocks_to_write = ceil_div(m_block_group_count * sizeof(ext2_group_desc), block_size());
  578. unsigned first_block_of_bgdt = block_size() == 1024 ? 2 : 1;
  579. auto buffer = UserOrKernelBuffer::for_kernel_buffer((u8*)block_group_descriptors());
  580. if (auto result = write_blocks(first_block_of_bgdt, blocks_to_write, buffer); result.is_error())
  581. dbgln("Ext2FS[{}]::flush_block_group_descriptor_table(): Failed to write blocks: {}", fsid(), result.error());
  582. }
  583. void Ext2FS::flush_writes()
  584. {
  585. Locker locker(m_lock);
  586. if (m_super_block_dirty) {
  587. flush_super_block();
  588. m_super_block_dirty = false;
  589. }
  590. if (m_block_group_descriptors_dirty) {
  591. flush_block_group_descriptor_table();
  592. m_block_group_descriptors_dirty = false;
  593. }
  594. for (auto& cached_bitmap : m_cached_bitmaps) {
  595. if (cached_bitmap->dirty) {
  596. auto buffer = UserOrKernelBuffer::for_kernel_buffer(cached_bitmap->buffer.data());
  597. if (auto result = write_block(cached_bitmap->bitmap_block_index, buffer, block_size()); result.is_error()) {
  598. dbgln("Ext2FS[{}]::flush_writes(): Failed to write blocks: {}", fsid(), result.error());
  599. }
  600. cached_bitmap->dirty = false;
  601. dbgln_if(EXT2_DEBUG, "Ext2FS[{}]::flush_writes(): Flushed bitmap block {}", fsid(), cached_bitmap->bitmap_block_index);
  602. }
  603. }
  604. BlockBasedFS::flush_writes();
  605. // Uncache Inodes that are only kept alive by the index-to-inode lookup cache.
  606. // We don't uncache Inodes that are being watched by at least one InodeWatcher.
  607. // FIXME: It would be better to keep a capped number of Inodes around.
  608. // The problem is that they are quite heavy objects, and use a lot of heap memory
  609. // for their (child name lookup) and (block list) caches.
  610. Vector<InodeIndex> unused_inodes;
  611. for (auto& it : m_inode_cache) {
  612. if (it.value->ref_count() != 1)
  613. continue;
  614. if (it.value->has_watchers())
  615. continue;
  616. unused_inodes.append(it.key);
  617. }
  618. for (auto index : unused_inodes)
  619. uncache_inode(index);
  620. }
  621. Ext2FSInode::Ext2FSInode(Ext2FS& fs, InodeIndex index)
  622. : Inode(fs, index)
  623. {
  624. }
  625. Ext2FSInode::~Ext2FSInode()
  626. {
  627. if (m_raw_inode.i_links_count == 0)
  628. fs().free_inode(*this);
  629. }
  630. u64 Ext2FSInode::size() const
  631. {
  632. if (Kernel::is_regular_file(m_raw_inode.i_mode) && ((u32)fs().get_features_readonly() & (u32)Ext2FS::FeaturesReadOnly::FileSize64bits))
  633. return static_cast<u64>(m_raw_inode.i_dir_acl) << 32 | m_raw_inode.i_size;
  634. return m_raw_inode.i_size;
  635. }
  636. InodeMetadata Ext2FSInode::metadata() const
  637. {
  638. Locker locker(m_lock);
  639. InodeMetadata metadata;
  640. metadata.inode = identifier();
  641. metadata.size = size();
  642. metadata.mode = m_raw_inode.i_mode;
  643. metadata.uid = m_raw_inode.i_uid;
  644. metadata.gid = m_raw_inode.i_gid;
  645. metadata.link_count = m_raw_inode.i_links_count;
  646. metadata.atime = m_raw_inode.i_atime;
  647. metadata.ctime = m_raw_inode.i_ctime;
  648. metadata.mtime = m_raw_inode.i_mtime;
  649. metadata.dtime = m_raw_inode.i_dtime;
  650. metadata.block_size = fs().block_size();
  651. metadata.block_count = m_raw_inode.i_blocks;
  652. if (Kernel::is_character_device(m_raw_inode.i_mode) || Kernel::is_block_device(m_raw_inode.i_mode)) {
  653. unsigned dev = m_raw_inode.i_block[0];
  654. if (!dev)
  655. dev = m_raw_inode.i_block[1];
  656. metadata.major_device = (dev & 0xfff00) >> 8;
  657. metadata.minor_device = (dev & 0xff) | ((dev >> 12) & 0xfff00);
  658. }
  659. return metadata;
  660. }
  661. void Ext2FSInode::flush_metadata()
  662. {
  663. Locker locker(m_lock);
  664. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::flush_metadata(): Flushing inode", identifier());
  665. fs().write_ext2_inode(index(), m_raw_inode);
  666. if (is_directory()) {
  667. // Unless we're about to go away permanently, invalidate the lookup cache.
  668. if (m_raw_inode.i_links_count != 0) {
  669. // FIXME: This invalidation is way too hardcore. It's sad to throw away the whole cache.
  670. m_lookup_cache.clear();
  671. }
  672. }
  673. set_metadata_dirty(false);
  674. }
  675. RefPtr<Inode> Ext2FS::get_inode(InodeIdentifier inode) const
  676. {
  677. Locker locker(m_lock);
  678. VERIFY(inode.fsid() == fsid());
  679. {
  680. auto it = m_inode_cache.find(inode.index());
  681. if (it != m_inode_cache.end())
  682. return (*it).value;
  683. }
  684. auto state_or_error = get_inode_allocation_state(inode.index());
  685. if (state_or_error.is_error())
  686. return {};
  687. if (!state_or_error.value()) {
  688. m_inode_cache.set(inode.index(), nullptr);
  689. return {};
  690. }
  691. BlockIndex block_index;
  692. unsigned offset;
  693. if (!find_block_containing_inode(inode.index(), block_index, offset))
  694. return {};
  695. auto new_inode = adopt_ref(*new Ext2FSInode(const_cast<Ext2FS&>(*this), inode.index()));
  696. auto buffer = UserOrKernelBuffer::for_kernel_buffer(reinterpret_cast<u8*>(&new_inode->m_raw_inode));
  697. if (auto result = read_block(block_index, &buffer, sizeof(ext2_inode), offset); result.is_error()) {
  698. // FIXME: Propagate the actual error.
  699. return nullptr;
  700. }
  701. m_inode_cache.set(inode.index(), new_inode);
  702. return new_inode;
  703. }
  704. KResultOr<ssize_t> Ext2FSInode::read_bytes(off_t offset, ssize_t count, UserOrKernelBuffer& buffer, FileDescription* description) const
  705. {
  706. Locker inode_locker(m_lock);
  707. VERIFY(offset >= 0);
  708. if (m_raw_inode.i_size == 0)
  709. return 0;
  710. if (static_cast<u64>(offset) >= size())
  711. return 0;
  712. // Symbolic links shorter than 60 characters are store inline inside the i_block array.
  713. // This avoids wasting an entire block on short links. (Most links are short.)
  714. if (is_symlink() && size() < max_inline_symlink_length) {
  715. VERIFY(offset == 0);
  716. ssize_t nread = min((off_t)size() - offset, static_cast<off_t>(count));
  717. if (!buffer.write(((const u8*)m_raw_inode.i_block) + offset, (size_t)nread))
  718. return EFAULT;
  719. return nread;
  720. }
  721. if (m_block_list.is_empty())
  722. m_block_list = compute_block_list();
  723. if (m_block_list.is_empty()) {
  724. dmesgln("Ext2FSInode[{}]::read_bytes(): Empty block list", identifier());
  725. return EIO;
  726. }
  727. bool allow_cache = !description || !description->is_direct();
  728. const int block_size = fs().block_size();
  729. BlockBasedFS::BlockIndex first_block_logical_index = offset / block_size;
  730. BlockBasedFS::BlockIndex last_block_logical_index = (offset + count) / block_size;
  731. if (last_block_logical_index >= m_block_list.size())
  732. last_block_logical_index = m_block_list.size() - 1;
  733. int offset_into_first_block = offset % block_size;
  734. ssize_t nread = 0;
  735. auto remaining_count = min((off_t)count, (off_t)size() - offset);
  736. dbgln_if(EXT2_VERY_DEBUG, "Ext2FSInode[{}]::read_bytes(): Reading up to {} bytes, {} bytes into inode to {}", identifier(), count, offset, buffer.user_or_kernel_ptr());
  737. for (auto bi = first_block_logical_index; remaining_count && bi <= last_block_logical_index; bi = bi.value() + 1) {
  738. auto block_index = m_block_list[bi.value()];
  739. size_t offset_into_block = (bi == first_block_logical_index) ? offset_into_first_block : 0;
  740. size_t num_bytes_to_copy = min((size_t)block_size - offset_into_block, (size_t)remaining_count);
  741. auto buffer_offset = buffer.offset(nread);
  742. if (block_index.value() == 0) {
  743. // This is a hole, act as if it's filled with zeroes.
  744. if (!buffer_offset.memset(0, num_bytes_to_copy))
  745. return EFAULT;
  746. } else {
  747. if (auto result = fs().read_block(block_index, &buffer_offset, num_bytes_to_copy, offset_into_block, allow_cache); result.is_error()) {
  748. dmesgln("Ext2FSInode[{}]::read_bytes(): Failed to read block {} (index {})", identifier(), block_index.value(), bi);
  749. return result.error();
  750. }
  751. }
  752. remaining_count -= num_bytes_to_copy;
  753. nread += num_bytes_to_copy;
  754. }
  755. return nread;
  756. }
  757. KResult Ext2FSInode::resize(u64 new_size)
  758. {
  759. auto old_size = size();
  760. if (old_size == new_size)
  761. return KSuccess;
  762. if (!((u32)fs().get_features_readonly() & (u32)Ext2FS::FeaturesReadOnly::FileSize64bits) && (new_size >= static_cast<u32>(-1)))
  763. return ENOSPC;
  764. u64 block_size = fs().block_size();
  765. auto blocks_needed_before = ceil_div(old_size, block_size);
  766. auto blocks_needed_after = ceil_div(new_size, block_size);
  767. if constexpr (EXT2_DEBUG) {
  768. dbgln("Ext2FSInode[{}]::resize(): Blocks needed before (size was {}): {}", identifier(), old_size, blocks_needed_before);
  769. dbgln("Ext2FSInode[{}]::resize(): Blocks needed after (size is {}): {}", identifier(), new_size, blocks_needed_after);
  770. }
  771. if (blocks_needed_after > blocks_needed_before) {
  772. auto additional_blocks_needed = blocks_needed_after - blocks_needed_before;
  773. if (additional_blocks_needed > fs().super_block().s_free_blocks_count)
  774. return ENOSPC;
  775. }
  776. if (m_block_list.is_empty())
  777. m_block_list = this->compute_block_list();
  778. if (blocks_needed_after > blocks_needed_before) {
  779. auto blocks_or_error = fs().allocate_blocks(fs().group_index_from_inode(index()), blocks_needed_after - blocks_needed_before);
  780. if (blocks_or_error.is_error())
  781. return blocks_or_error.error();
  782. if (!m_block_list.try_append(blocks_or_error.release_value()))
  783. return ENOMEM;
  784. } else if (blocks_needed_after < blocks_needed_before) {
  785. if constexpr (EXT2_VERY_DEBUG) {
  786. dbgln("Ext2FSInode[{}]::resize(): Shrinking inode, old block list is {} entries:", identifier(), m_block_list.size());
  787. for (auto block_index : m_block_list) {
  788. dbgln(" # {}", block_index);
  789. }
  790. }
  791. while (m_block_list.size() != blocks_needed_after) {
  792. auto block_index = m_block_list.take_last();
  793. if (block_index.value()) {
  794. if (auto result = fs().set_block_allocation_state(block_index, false); result.is_error()) {
  795. dbgln("Ext2FSInode[{}]::resize(): Failed to free block {}: {}", identifier(), block_index, result.error());
  796. return result;
  797. }
  798. }
  799. }
  800. }
  801. if (auto result = flush_block_list(); result.is_error())
  802. return result;
  803. m_raw_inode.i_size = new_size;
  804. if (Kernel::is_regular_file(m_raw_inode.i_mode))
  805. m_raw_inode.i_dir_acl = new_size >> 32;
  806. set_metadata_dirty(true);
  807. if (new_size > old_size) {
  808. // If we're growing the inode, make sure we zero out all the new space.
  809. // FIXME: There are definitely more efficient ways to achieve this.
  810. auto bytes_to_clear = new_size - old_size;
  811. auto clear_from = old_size;
  812. u8 zero_buffer[PAGE_SIZE] {};
  813. while (bytes_to_clear) {
  814. auto result = write_bytes(clear_from, min(static_cast<u64>(sizeof(zero_buffer)), bytes_to_clear), UserOrKernelBuffer::for_kernel_buffer(zero_buffer), nullptr);
  815. if (result.is_error())
  816. return result.error();
  817. VERIFY(result.value() != 0);
  818. bytes_to_clear -= result.value();
  819. clear_from += result.value();
  820. }
  821. }
  822. return KSuccess;
  823. }
  824. KResultOr<ssize_t> Ext2FSInode::write_bytes(off_t offset, ssize_t count, const UserOrKernelBuffer& data, FileDescription* description)
  825. {
  826. VERIFY(offset >= 0);
  827. VERIFY(count >= 0);
  828. if (count == 0)
  829. return 0;
  830. Locker inode_locker(m_lock);
  831. if (auto result = prepare_to_write_data(); result.is_error())
  832. return result;
  833. if (is_symlink()) {
  834. VERIFY(offset == 0);
  835. if (max((size_t)(offset + count), (size_t)m_raw_inode.i_size) < max_inline_symlink_length) {
  836. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::write_bytes(): Poking into i_block array for inline symlink '{}' ({} bytes)", identifier(), data.copy_into_string(count), count);
  837. if (!data.read(((u8*)m_raw_inode.i_block) + offset, (size_t)count))
  838. return EFAULT;
  839. if ((size_t)(offset + count) > (size_t)m_raw_inode.i_size)
  840. m_raw_inode.i_size = offset + count;
  841. set_metadata_dirty(true);
  842. return count;
  843. }
  844. }
  845. bool allow_cache = !description || !description->is_direct();
  846. const auto block_size = fs().block_size();
  847. auto new_size = max(static_cast<u64>(offset) + count, size());
  848. if (auto result = resize(new_size); result.is_error())
  849. return result;
  850. if (m_block_list.is_empty())
  851. m_block_list = compute_block_list();
  852. if (m_block_list.is_empty()) {
  853. dbgln("Ext2FSInode[{}]::write_bytes(): Empty block list", identifier());
  854. return EIO;
  855. }
  856. BlockBasedFS::BlockIndex first_block_logical_index = offset / block_size;
  857. BlockBasedFS::BlockIndex last_block_logical_index = (offset + count) / block_size;
  858. if (last_block_logical_index >= m_block_list.size())
  859. last_block_logical_index = m_block_list.size() - 1;
  860. size_t offset_into_first_block = offset % block_size;
  861. ssize_t nwritten = 0;
  862. auto remaining_count = min((off_t)count, (off_t)new_size - offset);
  863. dbgln_if(EXT2_VERY_DEBUG, "Ext2FSInode[{}]::write_bytes(): Writing {} bytes, {} bytes into inode from {}", identifier(), count, offset, data.user_or_kernel_ptr());
  864. for (auto bi = first_block_logical_index; remaining_count && bi <= last_block_logical_index; bi = bi.value() + 1) {
  865. size_t offset_into_block = (bi == first_block_logical_index) ? offset_into_first_block : 0;
  866. size_t num_bytes_to_copy = min((size_t)block_size - offset_into_block, (size_t)remaining_count);
  867. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::write_bytes(): Writing block {} (offset_into_block: {})", identifier(), m_block_list[bi.value()], offset_into_block);
  868. 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()) {
  869. dbgln("Ext2FSInode[{}]::write_bytes(): Failed to write block {} (index {})", identifier(), m_block_list[bi.value()], bi);
  870. return result;
  871. }
  872. remaining_count -= num_bytes_to_copy;
  873. nwritten += num_bytes_to_copy;
  874. }
  875. 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());
  876. return nwritten;
  877. }
  878. u8 Ext2FS::internal_file_type_to_directory_entry_type(const DirectoryEntryView& entry) const
  879. {
  880. switch (entry.file_type) {
  881. case EXT2_FT_REG_FILE:
  882. return DT_REG;
  883. case EXT2_FT_DIR:
  884. return DT_DIR;
  885. case EXT2_FT_CHRDEV:
  886. return DT_CHR;
  887. case EXT2_FT_BLKDEV:
  888. return DT_BLK;
  889. case EXT2_FT_FIFO:
  890. return DT_FIFO;
  891. case EXT2_FT_SOCK:
  892. return DT_SOCK;
  893. case EXT2_FT_SYMLINK:
  894. return DT_LNK;
  895. default:
  896. return DT_UNKNOWN;
  897. }
  898. }
  899. Ext2FS::FeaturesReadOnly Ext2FS::get_features_readonly() const
  900. {
  901. if (m_super_block.s_rev_level > 0)
  902. return static_cast<Ext2FS::FeaturesReadOnly>(m_super_block.s_feature_ro_compat);
  903. return Ext2FS::FeaturesReadOnly::None;
  904. }
  905. KResult Ext2FSInode::traverse_as_directory(Function<bool(const FS::DirectoryEntryView&)> callback) const
  906. {
  907. Locker locker(m_lock);
  908. VERIFY(is_directory());
  909. auto buffer_or = read_entire();
  910. if (buffer_or.is_error())
  911. return buffer_or.error();
  912. auto& buffer = *buffer_or.value();
  913. auto* entry = reinterpret_cast<ext2_dir_entry_2*>(buffer.data());
  914. while (entry < buffer.end_pointer()) {
  915. if (entry->inode != 0) {
  916. 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));
  917. if (!callback({ { entry->name, entry->name_len }, { fsid(), entry->inode }, entry->file_type }))
  918. break;
  919. }
  920. entry = (ext2_dir_entry_2*)((char*)entry + entry->rec_len);
  921. }
  922. return KSuccess;
  923. }
  924. KResult Ext2FSInode::write_directory(Vector<Ext2FSDirectoryEntry>& entries)
  925. {
  926. Locker locker(m_lock);
  927. auto block_size = fs().block_size();
  928. // Calculate directory size and record length of entries so that
  929. // the following constraints are met:
  930. // - All used blocks must be entirely filled.
  931. // - Entries are aligned on a 4-byte boundary.
  932. // - No entry may span multiple blocks.
  933. size_t directory_size = 0;
  934. size_t space_in_block = block_size;
  935. for (size_t i = 0; i < entries.size(); ++i) {
  936. auto& entry = entries[i];
  937. entry.record_length = EXT2_DIR_REC_LEN(entry.name.length());
  938. space_in_block -= entry.record_length;
  939. if (i + 1 < entries.size()) {
  940. if (EXT2_DIR_REC_LEN(entries[i + 1].name.length()) > space_in_block) {
  941. entry.record_length += space_in_block;
  942. space_in_block = block_size;
  943. }
  944. } else {
  945. entry.record_length += space_in_block;
  946. }
  947. directory_size += entry.record_length;
  948. }
  949. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::write_directory(): New directory contents to write (size {}):", identifier(), directory_size);
  950. auto directory_data = ByteBuffer::create_uninitialized(directory_size);
  951. OutputMemoryStream stream { directory_data };
  952. for (auto& entry : entries) {
  953. 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);
  954. stream << u32(entry.inode_index.value());
  955. stream << u16(entry.record_length);
  956. stream << u8(entry.name.length());
  957. stream << u8(entry.file_type);
  958. stream << entry.name.bytes();
  959. int padding = entry.record_length - entry.name.length() - 8;
  960. for (int j = 0; j < padding; ++j)
  961. stream << u8(0);
  962. }
  963. VERIFY(stream.is_end());
  964. if (auto result = resize(stream.size()); result.is_error())
  965. return result;
  966. auto buffer = UserOrKernelBuffer::for_kernel_buffer(stream.data());
  967. auto result = write_bytes(0, stream.size(), buffer, nullptr);
  968. if (result.is_error())
  969. return result.error();
  970. set_metadata_dirty(true);
  971. if (static_cast<size_t>(result.value()) != directory_data.size())
  972. return EIO;
  973. return KSuccess;
  974. }
  975. KResultOr<NonnullRefPtr<Inode>> Ext2FSInode::create_child(const String& name, mode_t mode, dev_t dev, uid_t uid, gid_t gid)
  976. {
  977. if (::is_directory(mode))
  978. return fs().create_directory(*this, name, mode, uid, gid);
  979. return fs().create_inode(*this, name, mode, dev, uid, gid);
  980. }
  981. KResult Ext2FSInode::add_child(Inode& child, const StringView& name, mode_t mode)
  982. {
  983. Locker locker(m_lock);
  984. VERIFY(is_directory());
  985. if (name.length() > EXT2_NAME_LEN)
  986. return ENAMETOOLONG;
  987. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::add_child(): Adding inode {} with name '{}' and mode {:o} to directory {}", identifier(), child.index(), name, mode, index());
  988. Vector<Ext2FSDirectoryEntry> entries;
  989. bool name_already_exists = false;
  990. KResult result = traverse_as_directory([&](auto& entry) {
  991. if (name == entry.name) {
  992. name_already_exists = true;
  993. return false;
  994. }
  995. entries.append({ entry.name, entry.inode.index(), entry.file_type });
  996. return true;
  997. });
  998. if (result.is_error())
  999. return result;
  1000. if (name_already_exists) {
  1001. dbgln("Ext2FSInode[{}]::add_child(): Name '{}' already exists", identifier(), name);
  1002. return EEXIST;
  1003. }
  1004. result = child.increment_link_count();
  1005. if (result.is_error())
  1006. return result;
  1007. entries.empend(name, child.index(), to_ext2_file_type(mode));
  1008. result = write_directory(entries);
  1009. if (result.is_error())
  1010. return result;
  1011. m_lookup_cache.set(name, child.index());
  1012. did_add_child(child.identifier());
  1013. return KSuccess;
  1014. }
  1015. KResult Ext2FSInode::remove_child(const StringView& name)
  1016. {
  1017. Locker locker(m_lock);
  1018. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]::remove_child(): Removing '{}'", identifier(), name);
  1019. VERIFY(is_directory());
  1020. auto it = m_lookup_cache.find(name);
  1021. if (it == m_lookup_cache.end())
  1022. return ENOENT;
  1023. auto child_inode_index = (*it).value;
  1024. InodeIdentifier child_id { fsid(), child_inode_index };
  1025. Vector<Ext2FSDirectoryEntry> entries;
  1026. KResult result = traverse_as_directory([&](auto& entry) {
  1027. if (name != entry.name)
  1028. entries.append({ entry.name, entry.inode.index(), entry.file_type });
  1029. return true;
  1030. });
  1031. if (result.is_error())
  1032. return result;
  1033. result = write_directory(entries);
  1034. if (result.is_error())
  1035. return result;
  1036. m_lookup_cache.remove(name);
  1037. auto child_inode = fs().get_inode(child_id);
  1038. result = child_inode->decrement_link_count();
  1039. if (result.is_error())
  1040. return result;
  1041. did_remove_child(child_id);
  1042. return KSuccess;
  1043. }
  1044. unsigned Ext2FS::inodes_per_block() const
  1045. {
  1046. return EXT2_INODES_PER_BLOCK(&super_block());
  1047. }
  1048. unsigned Ext2FS::inodes_per_group() const
  1049. {
  1050. return EXT2_INODES_PER_GROUP(&super_block());
  1051. }
  1052. unsigned Ext2FS::inode_size() const
  1053. {
  1054. return EXT2_INODE_SIZE(&super_block());
  1055. }
  1056. unsigned Ext2FS::blocks_per_group() const
  1057. {
  1058. return EXT2_BLOCKS_PER_GROUP(&super_block());
  1059. }
  1060. bool Ext2FS::write_ext2_inode(InodeIndex inode, const ext2_inode& e2inode)
  1061. {
  1062. BlockIndex block_index;
  1063. unsigned offset;
  1064. if (!find_block_containing_inode(inode, block_index, offset))
  1065. return false;
  1066. auto buffer = UserOrKernelBuffer::for_kernel_buffer(const_cast<u8*>((const u8*)&e2inode));
  1067. return write_block(block_index, buffer, inode_size(), offset) >= 0;
  1068. }
  1069. auto Ext2FS::allocate_blocks(GroupIndex preferred_group_index, size_t count) -> KResultOr<Vector<BlockIndex>>
  1070. {
  1071. Locker locker(m_lock);
  1072. dbgln_if(EXT2_DEBUG, "Ext2FS: allocate_blocks(preferred group: {}, count {})", preferred_group_index, count);
  1073. if (count == 0)
  1074. return Vector<BlockIndex> {};
  1075. Vector<BlockIndex> blocks;
  1076. dbgln_if(EXT2_DEBUG, "Ext2FS: allocate_blocks:");
  1077. blocks.ensure_capacity(count);
  1078. auto group_index = preferred_group_index;
  1079. if (!group_descriptor(preferred_group_index).bg_free_blocks_count) {
  1080. group_index = 1;
  1081. }
  1082. while (blocks.size() < count) {
  1083. bool found_a_group = false;
  1084. if (group_descriptor(group_index).bg_free_blocks_count) {
  1085. found_a_group = true;
  1086. } else {
  1087. if (group_index == preferred_group_index)
  1088. group_index = 1;
  1089. for (; group_index <= m_block_group_count; group_index = GroupIndex { group_index.value() + 1 }) {
  1090. if (group_descriptor(group_index).bg_free_blocks_count) {
  1091. found_a_group = true;
  1092. break;
  1093. }
  1094. }
  1095. }
  1096. VERIFY(found_a_group);
  1097. auto& bgd = group_descriptor(group_index);
  1098. auto cached_bitmap_or_error = get_bitmap_block(bgd.bg_block_bitmap);
  1099. if (cached_bitmap_or_error.is_error())
  1100. return cached_bitmap_or_error.error();
  1101. auto& cached_bitmap = *cached_bitmap_or_error.value();
  1102. int blocks_in_group = min(blocks_per_group(), super_block().s_blocks_count);
  1103. auto block_bitmap = cached_bitmap.bitmap(blocks_in_group);
  1104. BlockIndex first_block_in_group = (group_index.value() - 1) * blocks_per_group() + first_block_index().value();
  1105. size_t free_region_size = 0;
  1106. auto first_unset_bit_index = block_bitmap.find_longest_range_of_unset_bits(count - blocks.size(), free_region_size);
  1107. VERIFY(first_unset_bit_index.has_value());
  1108. dbgln_if(EXT2_DEBUG, "Ext2FS: allocating free region of size: {} [{}]", free_region_size, group_index);
  1109. for (size_t i = 0; i < free_region_size; ++i) {
  1110. BlockIndex block_index = (first_unset_bit_index.value() + i) + first_block_in_group.value();
  1111. if (auto result = set_block_allocation_state(block_index, true); result.is_error()) {
  1112. dbgln("Ext2FS: Failed to allocate block {} in allocate_blocks()", block_index);
  1113. return result;
  1114. }
  1115. blocks.unchecked_append(block_index);
  1116. dbgln_if(EXT2_DEBUG, " allocated > {}", block_index);
  1117. }
  1118. }
  1119. VERIFY(blocks.size() == count);
  1120. return blocks;
  1121. }
  1122. KResultOr<InodeIndex> Ext2FS::allocate_inode(GroupIndex preferred_group)
  1123. {
  1124. dbgln_if(EXT2_DEBUG, "Ext2FS: allocate_inode(preferred_group: {})", preferred_group);
  1125. Locker locker(m_lock);
  1126. // FIXME: We shouldn't refuse to allocate an inode if there is no group that can house the whole thing.
  1127. // In those cases we should just spread it across multiple groups.
  1128. auto is_suitable_group = [this](auto group_index) {
  1129. auto& bgd = group_descriptor(group_index);
  1130. return bgd.bg_free_inodes_count && bgd.bg_free_blocks_count >= 1;
  1131. };
  1132. GroupIndex group_index;
  1133. if (preferred_group.value() && is_suitable_group(preferred_group)) {
  1134. group_index = preferred_group;
  1135. } else {
  1136. for (unsigned i = 1; i <= m_block_group_count; ++i) {
  1137. if (is_suitable_group(i)) {
  1138. group_index = i;
  1139. break;
  1140. }
  1141. }
  1142. }
  1143. if (!group_index) {
  1144. dmesgln("Ext2FS: allocate_inode: no suitable group found for new inode");
  1145. return ENOSPC;
  1146. }
  1147. dbgln_if(EXT2_DEBUG, "Ext2FS: allocate_inode: found suitable group [{}] for new inode :^)", group_index);
  1148. auto& bgd = group_descriptor(group_index);
  1149. unsigned inodes_in_group = min(inodes_per_group(), super_block().s_inodes_count);
  1150. InodeIndex first_inode_in_group = (group_index.value() - 1) * inodes_per_group() + 1;
  1151. auto cached_bitmap_or_error = get_bitmap_block(bgd.bg_inode_bitmap);
  1152. if (cached_bitmap_or_error.is_error())
  1153. return cached_bitmap_or_error.error();
  1154. auto& cached_bitmap = *cached_bitmap_or_error.value();
  1155. auto inode_bitmap = cached_bitmap.bitmap(inodes_in_group);
  1156. for (size_t i = 0; i < inode_bitmap.size(); ++i) {
  1157. if (inode_bitmap.get(i))
  1158. continue;
  1159. inode_bitmap.set(i, true);
  1160. auto inode_index = InodeIndex(first_inode_in_group.value() + i);
  1161. cached_bitmap.dirty = true;
  1162. m_super_block.s_free_inodes_count--;
  1163. m_super_block_dirty = true;
  1164. const_cast<ext2_group_desc&>(bgd).bg_free_inodes_count--;
  1165. m_block_group_descriptors_dirty = true;
  1166. // In case the inode cache had this cached as "non-existent", uncache that info.
  1167. m_inode_cache.remove(inode_index.value());
  1168. return inode_index;
  1169. }
  1170. dmesgln("Ext2FS: allocate_inode found no available inode, despite bgd claiming there are inodes :(");
  1171. return EIO;
  1172. }
  1173. Ext2FS::GroupIndex Ext2FS::group_index_from_block_index(BlockIndex block_index) const
  1174. {
  1175. if (!block_index)
  1176. return 0;
  1177. return (block_index.value() - 1) / blocks_per_group() + 1;
  1178. }
  1179. auto Ext2FS::group_index_from_inode(InodeIndex inode) const -> GroupIndex
  1180. {
  1181. if (!inode)
  1182. return 0;
  1183. return (inode.value() - 1) / inodes_per_group() + 1;
  1184. }
  1185. KResultOr<bool> Ext2FS::get_inode_allocation_state(InodeIndex index) const
  1186. {
  1187. Locker locker(m_lock);
  1188. if (index == 0)
  1189. return EINVAL;
  1190. auto group_index = group_index_from_inode(index);
  1191. auto& bgd = group_descriptor(group_index);
  1192. unsigned index_in_group = index.value() - ((group_index.value() - 1) * inodes_per_group());
  1193. unsigned bit_index = (index_in_group - 1) % inodes_per_group();
  1194. auto cached_bitmap_or_error = const_cast<Ext2FS&>(*this).get_bitmap_block(bgd.bg_inode_bitmap);
  1195. if (cached_bitmap_or_error.is_error())
  1196. return cached_bitmap_or_error.error();
  1197. return cached_bitmap_or_error.value()->bitmap(inodes_per_group()).get(bit_index);
  1198. }
  1199. KResult Ext2FS::update_bitmap_block(BlockIndex bitmap_block, size_t bit_index, bool new_state, u32& super_block_counter, u16& group_descriptor_counter)
  1200. {
  1201. auto cached_bitmap_or_error = get_bitmap_block(bitmap_block);
  1202. if (cached_bitmap_or_error.is_error())
  1203. return cached_bitmap_or_error.error();
  1204. auto& cached_bitmap = *cached_bitmap_or_error.value();
  1205. bool current_state = cached_bitmap.bitmap(blocks_per_group()).get(bit_index);
  1206. if (current_state == new_state) {
  1207. dbgln("Ext2FS: Bit {} in bitmap block {} had unexpected state {}", bit_index, bitmap_block, current_state);
  1208. return EIO;
  1209. }
  1210. cached_bitmap.bitmap(blocks_per_group()).set(bit_index, new_state);
  1211. cached_bitmap.dirty = true;
  1212. if (new_state) {
  1213. --super_block_counter;
  1214. --group_descriptor_counter;
  1215. } else {
  1216. ++super_block_counter;
  1217. ++group_descriptor_counter;
  1218. }
  1219. m_super_block_dirty = true;
  1220. m_block_group_descriptors_dirty = true;
  1221. return KSuccess;
  1222. }
  1223. KResult Ext2FS::set_inode_allocation_state(InodeIndex inode_index, bool new_state)
  1224. {
  1225. Locker locker(m_lock);
  1226. auto group_index = group_index_from_inode(inode_index);
  1227. unsigned index_in_group = inode_index.value() - ((group_index.value() - 1) * inodes_per_group());
  1228. unsigned bit_index = (index_in_group - 1) % inodes_per_group();
  1229. dbgln_if(EXT2_DEBUG, "Ext2FS: set_inode_allocation_state: Inode {} -> {}", inode_index, new_state);
  1230. auto& bgd = const_cast<ext2_group_desc&>(group_descriptor(group_index));
  1231. return update_bitmap_block(bgd.bg_inode_bitmap, bit_index, new_state, m_super_block.s_free_inodes_count, bgd.bg_free_inodes_count);
  1232. }
  1233. Ext2FS::BlockIndex Ext2FS::first_block_index() const
  1234. {
  1235. return block_size() == 1024 ? 1 : 0;
  1236. }
  1237. KResultOr<Ext2FS::CachedBitmap*> Ext2FS::get_bitmap_block(BlockIndex bitmap_block_index)
  1238. {
  1239. for (auto& cached_bitmap : m_cached_bitmaps) {
  1240. if (cached_bitmap->bitmap_block_index == bitmap_block_index)
  1241. return cached_bitmap;
  1242. }
  1243. auto block = KBuffer::create_with_size(block_size(), Region::Access::Read | Region::Access::Write, "Ext2FS: Cached bitmap block");
  1244. auto buffer = UserOrKernelBuffer::for_kernel_buffer(block.data());
  1245. if (auto result = read_block(bitmap_block_index, &buffer, block_size()); result.is_error()) {
  1246. dbgln("Ext2FS: Failed to load bitmap block {}", bitmap_block_index);
  1247. return result;
  1248. }
  1249. if (!m_cached_bitmaps.try_append(make<CachedBitmap>(bitmap_block_index, move(block))))
  1250. return ENOMEM;
  1251. return m_cached_bitmaps.last();
  1252. }
  1253. KResult Ext2FS::set_block_allocation_state(BlockIndex block_index, bool new_state)
  1254. {
  1255. VERIFY(block_index != 0);
  1256. Locker locker(m_lock);
  1257. auto group_index = group_index_from_block_index(block_index);
  1258. unsigned index_in_group = (block_index.value() - first_block_index().value()) - ((group_index.value() - 1) * blocks_per_group());
  1259. unsigned bit_index = index_in_group % blocks_per_group();
  1260. auto& bgd = const_cast<ext2_group_desc&>(group_descriptor(group_index));
  1261. dbgln_if(EXT2_DEBUG, "Ext2FS: Block {} state -> {} (in bitmap block {})", block_index, new_state, bgd.bg_block_bitmap);
  1262. return update_bitmap_block(bgd.bg_block_bitmap, bit_index, new_state, m_super_block.s_free_blocks_count, bgd.bg_free_blocks_count);
  1263. }
  1264. KResult Ext2FS::create_directory(Ext2FSInode& parent_inode, const String& name, mode_t mode, uid_t uid, gid_t gid)
  1265. {
  1266. Locker locker(m_lock);
  1267. VERIFY(is_directory(mode));
  1268. auto inode_or_error = create_inode(parent_inode, name, mode, 0, uid, gid);
  1269. if (inode_or_error.is_error())
  1270. return inode_or_error.error();
  1271. auto& inode = inode_or_error.value();
  1272. dbgln_if(EXT2_DEBUG, "Ext2FS: create_directory: created new directory named '{} with inode {}", name, inode->index());
  1273. Vector<Ext2FSDirectoryEntry> entries;
  1274. entries.empend(".", inode->index(), static_cast<u8>(EXT2_FT_DIR));
  1275. entries.empend("..", parent_inode.index(), static_cast<u8>(EXT2_FT_DIR));
  1276. if (auto result = static_cast<Ext2FSInode&>(*inode).write_directory(entries); result.is_error())
  1277. return result;
  1278. if (auto result = parent_inode.increment_link_count(); result.is_error())
  1279. return result;
  1280. auto& bgd = const_cast<ext2_group_desc&>(group_descriptor(group_index_from_inode(inode->identifier().index())));
  1281. ++bgd.bg_used_dirs_count;
  1282. m_block_group_descriptors_dirty = true;
  1283. return KSuccess;
  1284. }
  1285. KResultOr<NonnullRefPtr<Inode>> Ext2FS::create_inode(Ext2FSInode& parent_inode, const String& name, mode_t mode, dev_t dev, uid_t uid, gid_t gid)
  1286. {
  1287. if (name.length() > EXT2_NAME_LEN)
  1288. return ENAMETOOLONG;
  1289. if (parent_inode.m_raw_inode.i_links_count == 0)
  1290. return ENOENT;
  1291. ext2_inode e2inode {};
  1292. auto now = kgettimeofday().to_truncated_seconds();
  1293. e2inode.i_mode = mode;
  1294. e2inode.i_uid = uid;
  1295. e2inode.i_gid = gid;
  1296. e2inode.i_size = 0;
  1297. e2inode.i_atime = now;
  1298. e2inode.i_ctime = now;
  1299. e2inode.i_mtime = now;
  1300. e2inode.i_dtime = 0;
  1301. e2inode.i_flags = 0;
  1302. // For directories, add +1 link count for the "." entry in self.
  1303. e2inode.i_links_count = is_directory(mode);
  1304. if (is_character_device(mode))
  1305. e2inode.i_block[0] = dev;
  1306. else if (is_block_device(mode))
  1307. e2inode.i_block[1] = dev;
  1308. auto inode_id = allocate_inode();
  1309. if (inode_id.is_error())
  1310. return inode_id.error();
  1311. dbgln_if(EXT2_DEBUG, "Ext2FS: writing initial metadata for inode {}", inode_id.value());
  1312. auto success = write_ext2_inode(inode_id.value(), e2inode);
  1313. VERIFY(success);
  1314. auto new_inode = get_inode({ fsid(), inode_id.value() });
  1315. VERIFY(new_inode);
  1316. dbgln_if(EXT2_DEBUG, "Ext2FS: Adding inode '{}' (mode {:o}) to parent directory {}", name, mode, parent_inode.index());
  1317. if (auto result = parent_inode.add_child(*new_inode, name, mode); result.is_error())
  1318. return result;
  1319. return new_inode.release_nonnull();
  1320. }
  1321. bool Ext2FSInode::populate_lookup_cache() const
  1322. {
  1323. Locker locker(m_lock);
  1324. if (!m_lookup_cache.is_empty())
  1325. return true;
  1326. HashMap<String, InodeIndex> children;
  1327. KResult result = traverse_as_directory([&children](auto& entry) {
  1328. children.set(entry.name, entry.inode.index());
  1329. return true;
  1330. });
  1331. if (!result.is_success())
  1332. return false;
  1333. if (!m_lookup_cache.is_empty())
  1334. return false;
  1335. m_lookup_cache = move(children);
  1336. return true;
  1337. }
  1338. RefPtr<Inode> Ext2FSInode::lookup(StringView name)
  1339. {
  1340. VERIFY(is_directory());
  1341. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]:lookup(): Looking up '{}'", identifier(), name);
  1342. if (!populate_lookup_cache())
  1343. return {};
  1344. Locker locker(m_lock);
  1345. auto it = m_lookup_cache.find(name.hash(), [&](auto& entry) { return entry.key == name; });
  1346. if (it != m_lookup_cache.end())
  1347. return fs().get_inode({ fsid(), (*it).value });
  1348. dbgln_if(EXT2_DEBUG, "Ext2FSInode[{}]:lookup(): '{}' not found", identifier(), name);
  1349. return {};
  1350. }
  1351. void Ext2FSInode::one_ref_left()
  1352. {
  1353. // FIXME: I would like to not live forever, but uncached Ext2FS is fucking painful right now.
  1354. }
  1355. KResult Ext2FSInode::set_atime(time_t t)
  1356. {
  1357. Locker locker(m_lock);
  1358. if (fs().is_readonly())
  1359. return EROFS;
  1360. m_raw_inode.i_atime = t;
  1361. set_metadata_dirty(true);
  1362. return KSuccess;
  1363. }
  1364. KResult Ext2FSInode::set_ctime(time_t t)
  1365. {
  1366. Locker locker(m_lock);
  1367. if (fs().is_readonly())
  1368. return EROFS;
  1369. m_raw_inode.i_ctime = t;
  1370. set_metadata_dirty(true);
  1371. return KSuccess;
  1372. }
  1373. KResult Ext2FSInode::set_mtime(time_t t)
  1374. {
  1375. Locker locker(m_lock);
  1376. if (fs().is_readonly())
  1377. return EROFS;
  1378. m_raw_inode.i_mtime = t;
  1379. set_metadata_dirty(true);
  1380. return KSuccess;
  1381. }
  1382. KResult Ext2FSInode::increment_link_count()
  1383. {
  1384. Locker locker(m_lock);
  1385. if (fs().is_readonly())
  1386. return EROFS;
  1387. if (m_raw_inode.i_links_count == max_link_count)
  1388. return EMLINK;
  1389. ++m_raw_inode.i_links_count;
  1390. set_metadata_dirty(true);
  1391. return KSuccess;
  1392. }
  1393. KResult Ext2FSInode::decrement_link_count()
  1394. {
  1395. Locker locker(m_lock);
  1396. if (fs().is_readonly())
  1397. return EROFS;
  1398. VERIFY(m_raw_inode.i_links_count);
  1399. --m_raw_inode.i_links_count;
  1400. if (ref_count() == 1 && m_raw_inode.i_links_count == 0)
  1401. fs().uncache_inode(index());
  1402. set_metadata_dirty(true);
  1403. return KSuccess;
  1404. }
  1405. void Ext2FS::uncache_inode(InodeIndex index)
  1406. {
  1407. Locker locker(m_lock);
  1408. m_inode_cache.remove(index);
  1409. }
  1410. KResultOr<size_t> Ext2FSInode::directory_entry_count() const
  1411. {
  1412. VERIFY(is_directory());
  1413. Locker locker(m_lock);
  1414. populate_lookup_cache();
  1415. return m_lookup_cache.size();
  1416. }
  1417. KResult Ext2FSInode::chmod(mode_t mode)
  1418. {
  1419. Locker locker(m_lock);
  1420. if (m_raw_inode.i_mode == mode)
  1421. return KSuccess;
  1422. m_raw_inode.i_mode = mode;
  1423. set_metadata_dirty(true);
  1424. return KSuccess;
  1425. }
  1426. KResult Ext2FSInode::chown(uid_t uid, gid_t gid)
  1427. {
  1428. Locker locker(m_lock);
  1429. if (m_raw_inode.i_uid == uid && m_raw_inode.i_gid == gid)
  1430. return KSuccess;
  1431. m_raw_inode.i_uid = uid;
  1432. m_raw_inode.i_gid = gid;
  1433. set_metadata_dirty(true);
  1434. return KSuccess;
  1435. }
  1436. KResult Ext2FSInode::truncate(u64 size)
  1437. {
  1438. Locker locker(m_lock);
  1439. if (static_cast<u64>(m_raw_inode.i_size) == size)
  1440. return KSuccess;
  1441. if (auto result = resize(size); result.is_error())
  1442. return result;
  1443. set_metadata_dirty(true);
  1444. return KSuccess;
  1445. }
  1446. KResultOr<int> Ext2FSInode::get_block_address(int index)
  1447. {
  1448. Locker locker(m_lock);
  1449. if (m_block_list.is_empty())
  1450. m_block_list = compute_block_list();
  1451. if (index < 0 || (size_t)index >= m_block_list.size())
  1452. return 0;
  1453. return m_block_list[index].value();
  1454. }
  1455. unsigned Ext2FS::total_block_count() const
  1456. {
  1457. Locker locker(m_lock);
  1458. return super_block().s_blocks_count;
  1459. }
  1460. unsigned Ext2FS::free_block_count() const
  1461. {
  1462. Locker locker(m_lock);
  1463. return super_block().s_free_blocks_count;
  1464. }
  1465. unsigned Ext2FS::total_inode_count() const
  1466. {
  1467. Locker locker(m_lock);
  1468. return super_block().s_inodes_count;
  1469. }
  1470. unsigned Ext2FS::free_inode_count() const
  1471. {
  1472. Locker locker(m_lock);
  1473. return super_block().s_free_inodes_count;
  1474. }
  1475. KResult Ext2FS::prepare_to_unmount() const
  1476. {
  1477. Locker locker(m_lock);
  1478. for (auto& it : m_inode_cache) {
  1479. if (it.value->ref_count() > 1)
  1480. return EBUSY;
  1481. }
  1482. m_inode_cache.clear();
  1483. return KSuccess;
  1484. }
  1485. }