diff options
author | Nick Mathewson <nickm@torproject.org> | 2017-04-25 08:12:59 -0400 |
---|---|---|
committer | Nick Mathewson <nickm@torproject.org> | 2017-04-25 08:12:59 -0400 |
commit | ba405f86bf33ed0e6c9552b3d9aad87bb79c1237 (patch) | |
tree | 4f50dd06a62263594e69c701253d4a632d1c7574 | |
parent | db95a6e17129b51e0409dbcad1707367f9388173 (diff) | |
parent | 2210b330e757d234346191e5d05af761fe8561e6 (diff) | |
download | tor-ba405f86bf33ed0e6c9552b3d9aad87bb79c1237.tar.gz tor-ba405f86bf33ed0e6c9552b3d9aad87bb79c1237.zip |
Merge branch 'ahf_prop278_21662_squashed'
32 files changed, 2469 insertions, 416 deletions
diff --git a/Makefile.am b/Makefile.am index aeac0b85c4..be1dc7f7ef 100644 --- a/Makefile.am +++ b/Makefile.am @@ -16,7 +16,7 @@ noinst_PROGRAMS= DISTCLEANFILES= bin_SCRIPTS= AM_CPPFLAGS= -AM_CFLAGS=@TOR_SYSTEMD_CFLAGS@ @CFLAGS_BUGTRAP@ +AM_CFLAGS=@TOR_SYSTEMD_CFLAGS@ @CFLAGS_BUGTRAP@ @TOR_LZMA_CFLAGS@ @TOR_ZSTD_CFLAGS@ SHELL=@SHELL@ if COVERAGE_ENABLED diff --git a/configure.ac b/configure.ac index 0d4fa9838b..c7960fa4ed 100644 --- a/configure.ac +++ b/configure.ac @@ -732,6 +732,70 @@ else fi AC_SUBST(TOR_ZLIB_LIBS) +dnl ------------------------------------------------------ +dnl Where we do we find lzma? + +AC_ARG_ENABLE(lzma, + AS_HELP_STRING(--enable-lzma, [enable support for the Zstandard compression scheme.]), + [case "${enableval}" in + "yes") lzma=true ;; + "no") lzma=false ;; + * ) AC_MSG_ERROR(bad value for --enable-lzma) ;; + esac], [lzma=auto]) + +if test "x$enable_lzma" = "xno"; then + have_lzma=no; +else + PKG_CHECK_MODULES([LZMA], + [liblzma], + have_lzma=yes, + have_lzma=no) + + if test "x$have_lzma" = "xno" ; then + AC_MSG_WARN([Unable to find liblzma.]) + fi +fi + +if test "x$have_lzma" = "xyes"; then + AC_DEFINE(HAVE_LZMA,1,[Have LZMA]) + TOR_LZMA_CFLAGS="${LZMA_CFLAGS}" + TOR_LZMA_LIBS="${LZMA_LIBS}" +fi +AC_SUBST(TOR_LZMA_CFLAGS) +AC_SUBST(TOR_LZMA_LIBS) + +dnl ------------------------------------------------------ +dnl Where we do we find zstd? + +AC_ARG_ENABLE(zstd, + AS_HELP_STRING(--enable-zstd, [enable support for the Zstandard compression scheme.]), + [case "${enableval}" in + "yes") zstd=true ;; + "no") zstd=false ;; + * ) AC_MSG_ERROR(bad value for --enable-zstd) ;; + esac], [zstd=auto]) + +if test "x$enable_zstd" = "xno"; then + have_zstd=no; +else + PKG_CHECK_MODULES([ZSTD], + [libzstd], + have_zstd=yes, + have_zstd=no) + + if test "x$have_zstd" = "xno" ; then + AC_MSG_WARN([Unable to find libzstd.]) + fi +fi + +if test "x$have_zstd" = "xyes"; then + AC_DEFINE(HAVE_ZSTD,1,[Have Zstd]) + TOR_ZSTD_CFLAGS="${ZSTD_CFLAGS}" + TOR_ZSTD_LIBS="${ZSTD_LIBS}" +fi +AC_SUBST(TOR_ZSTD_CFLAGS) +AC_SUBST(TOR_ZSTD_LIBS) + dnl ---------------------------------------------------------------------- dnl Check if libcap is available for capabilities. diff --git a/src/common/Makefile.nmake b/src/common/Makefile.nmake index b8c5dd4fea..a1c819fffa 100644 --- a/src/common/Makefile.nmake +++ b/src/common/Makefile.nmake @@ -7,8 +7,8 @@ LIBOR_OBJECTS = address.obj backtrace.obj compat.obj container.obj di_ops.obj \ log.obj memarea.obj mempool.obj procmon.obj sandbox.obj util.obj \ util_codedigest.obj -LIBOR_CRYPTO_OBJECTS = aes.obj crypto.obj crypto_format.obj torgzip.obj tortls.obj \ - crypto_curve25519.obj curve25519-donna.obj +LIBOR_CRYPTO_OBJECTS = aes.obj crypto.obj crypto_format.obj compress.obj compress_zlib.obj \ + tortls.obj crypto_curve25519.obj curve25519-donna.obj LIBOR_EVENT_OBJECTS = compat_libevent.obj diff --git a/src/common/compress.c b/src/common/compress.c new file mode 100644 index 0000000000..38b8184573 --- /dev/null +++ b/src/common/compress.c @@ -0,0 +1,377 @@ +/* Copyright (c) 2004, Roger Dingledine. + * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson. + * Copyright (c) 2007-2017, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +/** + * \file compress.c + * \brief Common compression API. + **/ + +#include "orconfig.h" + +#include <stdlib.h> +#include <stdio.h> +#include <assert.h> +#include <string.h> +#include "torint.h" + +#ifdef HAVE_NETINET_IN_H +#include <netinet/in.h> +#endif + +#include "util.h" +#include "torlog.h" +#include "compress.h" +#include "compress_lzma.h" +#include "compress_zlib.h" +#include "compress_zstd.h" + +/** @{ */ +/* These macros define the maximum allowable compression factor. Anything of + * size greater than CHECK_FOR_COMPRESSION_BOMB_AFTER is not allowed to + * have an uncompression factor (uncompressed size:compressed size ratio) of + * any greater than MAX_UNCOMPRESSION_FACTOR. + * + * Picking a value for MAX_UNCOMPRESSION_FACTOR is a trade-off: we want it to + * be small to limit the attack multiplier, but we also want it to be large + * enough so that no legitimate document --even ones we might invent in the + * future -- ever compresses by a factor of greater than + * MAX_UNCOMPRESSION_FACTOR. Within those parameters, there's a reasonably + * large range of possible values. IMO, anything over 8 is probably safe; IMO + * anything under 50 is probably sufficient. + */ +#define MAX_UNCOMPRESSION_FACTOR 25 +#define CHECK_FOR_COMPRESSION_BOMB_AFTER (1024*64) +/** @} */ + +/** Return true if uncompressing an input of size <b>in_size</b> to an input of + * size at least <b>size_out</b> looks like a compression bomb. */ +int +tor_compress_is_compression_bomb(size_t size_in, size_t size_out) +{ + if (size_in == 0 || size_out < CHECK_FOR_COMPRESSION_BOMB_AFTER) + return 0; + + return (size_out / size_in > MAX_UNCOMPRESSION_FACTOR); +} + +/** Given <b>in_len</b> bytes at <b>in</b>, compress them into a newly + * allocated buffer, using the method described in <b>method</b>. Store the + * compressed string in *<b>out</b>, and its length in *<b>out_len</b>. + * Return 0 on success, -1 on failure. + */ +int +tor_compress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method) +{ + switch (method) { + case GZIP_METHOD: + case ZLIB_METHOD: + return tor_zlib_compress(out, out_len, in, in_len, method); + case LZMA_METHOD: + return tor_lzma_compress(out, out_len, in, in_len, method); + case ZSTD_METHOD: + return tor_zstd_compress(out, out_len, in, in_len, method); + case NO_METHOD: + case UNKNOWN_METHOD: + default: + return -1; + } +} + +/** Given zero or more zlib-compressed or gzip-compressed strings of + * total length + * <b>in_len</b> bytes at <b>in</b>, uncompress them into a newly allocated + * buffer, using the method described in <b>method</b>. Store the uncompressed + * string in *<b>out</b>, and its length in *<b>out_len</b>. Return 0 on + * success, -1 on failure. + * + * If <b>complete_only</b> is true, we consider a truncated input as a + * failure; otherwise we decompress as much as we can. Warn about truncated + * or corrupt inputs at <b>protocol_warn_level</b>. + */ +int +tor_uncompress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method, + int complete_only, + int protocol_warn_level) +{ + switch (method) { + case GZIP_METHOD: + case ZLIB_METHOD: + return tor_zlib_uncompress(out, out_len, in, in_len, + method, + complete_only, + protocol_warn_level); + case LZMA_METHOD: + return tor_lzma_uncompress(out, out_len, in, in_len, + method, + complete_only, + protocol_warn_level); + case ZSTD_METHOD: + return tor_zstd_uncompress(out, out_len, in, in_len, + method, + complete_only, + protocol_warn_level); + case NO_METHOD: + case UNKNOWN_METHOD: + default: + return -1; + } +} + +/** Try to tell whether the <b>in_len</b>-byte string in <b>in</b> is likely + * to be compressed or not. If it is, return the likeliest compression method. + * Otherwise, return UNKNOWN_METHOD. + */ +compress_method_t +detect_compression_method(const char *in, size_t in_len) +{ + if (in_len > 2 && fast_memeq(in, "\x1f\x8b", 2)) { + return GZIP_METHOD; + } else if (in_len > 2 && (in[0] & 0x0f) == 8 && + (ntohs(get_uint16(in)) % 31) == 0) { + return ZLIB_METHOD; + } else if (in_len > 3 && + fast_memeq(in, "\x5d\x00\x00\x00", 4)) { + return LZMA_METHOD; + } else if (in_len > 3 && + fast_memeq(in, "\x28\xb5\x2f\xfd", 4)) { + return ZSTD_METHOD; + } else { + return UNKNOWN_METHOD; + } +} + +/** Return 1 if a given <b>method</b> is supported; otherwise 0. */ +int +tor_compress_supports_method(compress_method_t method) +{ + switch (method) { + case GZIP_METHOD: + case ZLIB_METHOD: + return tor_zlib_method_supported(); + case LZMA_METHOD: + return tor_lzma_method_supported(); + case ZSTD_METHOD: + return tor_zstd_method_supported(); + case NO_METHOD: + case UNKNOWN_METHOD: + default: + return 0; + } +} + +/** Return a string representation of the version of the library providing the + * compression method given in <b>method</b>. Returns NULL if <b>method</b> is + * unknown or unsupported. */ +const char * +tor_compress_version_str(compress_method_t method) +{ + switch (method) { + case GZIP_METHOD: + case ZLIB_METHOD: + return tor_zlib_get_version_str(); + case LZMA_METHOD: + return tor_lzma_get_version_str(); + case ZSTD_METHOD: + return tor_zstd_get_version_str(); + case NO_METHOD: + case UNKNOWN_METHOD: + default: + return NULL; + } +} + +/** Return a string representation of the version of the library, found at + * compile time, providing the compression method given in <b>method</b>. + * Returns NULL if <b>method</b> is unknown or unsupported. */ +const char * +tor_compress_header_version_str(compress_method_t method) +{ + switch (method) { + case GZIP_METHOD: + case ZLIB_METHOD: + return tor_zlib_get_header_version_str(); + case LZMA_METHOD: + return tor_lzma_get_header_version_str(); + case ZSTD_METHOD: + return tor_zstd_get_header_version_str(); + case NO_METHOD: + case UNKNOWN_METHOD: + default: + return NULL; + } +} + +/** Return the approximate number of bytes allocated for all + * supported compression schemas. */ +size_t +tor_compress_get_total_allocation(void) +{ + return tor_zlib_get_total_allocation() + + tor_lzma_get_total_allocation() + + tor_zstd_get_total_allocation(); +} + +/** Internal state for an incremental compression/decompression. The body of + * this struct is not exposed. */ +struct tor_compress_state_t { + compress_method_t method; /**< The compression method. */ + + union { + tor_zlib_compress_state_t *zlib_state; + tor_lzma_compress_state_t *lzma_state; + tor_zstd_compress_state_t *zstd_state; + } u; /**< Compression backend state. */ +}; + +/** Construct and return a tor_compress_state_t object using <b>method</b>. If + * <b>compress</b>, it's for compression; otherwise it's for decompression. */ +tor_compress_state_t * +tor_compress_new(int compress, compress_method_t method, + compression_level_t compression_level) +{ + tor_compress_state_t *state; + + state = tor_malloc_zero(sizeof(tor_compress_state_t)); + state->method = method; + + switch (method) { + case GZIP_METHOD: + case ZLIB_METHOD: { + tor_zlib_compress_state_t *zlib_state = + tor_zlib_compress_new(compress, method, compression_level); + + if (zlib_state == NULL) + goto err; + + state->u.zlib_state = zlib_state; + break; + } + case LZMA_METHOD: { + tor_lzma_compress_state_t *lzma_state = + tor_lzma_compress_new(compress, method, compression_level); + + if (lzma_state == NULL) + goto err; + + state->u.lzma_state = lzma_state; + break; + } + case ZSTD_METHOD: { + tor_zstd_compress_state_t *zstd_state = + tor_zstd_compress_new(compress, method, compression_level); + + if (zstd_state == NULL) + goto err; + + state->u.zstd_state = zstd_state; + break; + } + case NO_METHOD: + case UNKNOWN_METHOD: + goto err; + } + + return state; + + err: + tor_free(state); + return NULL; +} + +/** Compress/decompress some bytes using <b>state</b>. Read up to + * *<b>in_len</b> bytes from *<b>in</b>, and write up to *<b>out_len</b> bytes + * to *<b>out</b>, adjusting the values as we go. If <b>finish</b> is true, + * we've reached the end of the input. + * + * Return TOR_COMPRESS_DONE if we've finished the entire + * compression/decompression. + * Return TOR_COMPRESS_OK if we're processed everything from the input. + * Return TOR_COMPRESS_BUFFER_FULL if we're out of space on <b>out</b>. + * Return TOR_COMPRESS_ERROR if the stream is corrupt. + */ +tor_compress_output_t +tor_compress_process(tor_compress_state_t *state, + char **out, size_t *out_len, + const char **in, size_t *in_len, + int finish) +{ + tor_assert(state != NULL); + + switch (state->method) { + case GZIP_METHOD: + case ZLIB_METHOD: + return tor_zlib_compress_process(state->u.zlib_state, + out, out_len, in, in_len, + finish); + case LZMA_METHOD: + return tor_lzma_compress_process(state->u.lzma_state, + out, out_len, in, in_len, + finish); + case ZSTD_METHOD: + return tor_zstd_compress_process(state->u.zstd_state, + out, out_len, in, in_len, + finish); + case NO_METHOD: + case UNKNOWN_METHOD: + goto err; + } + + err: + return TOR_COMPRESS_ERROR; +} + +/** Deallocate <b>state</b>. */ +void +tor_compress_free(tor_compress_state_t *state) +{ + if (state == NULL) + return; + + switch (state->method) { + case GZIP_METHOD: + case ZLIB_METHOD: + tor_zlib_compress_free(state->u.zlib_state); + break; + case LZMA_METHOD: + tor_lzma_compress_free(state->u.lzma_state); + break; + case ZSTD_METHOD: + tor_zstd_compress_free(state->u.zstd_state); + break; + case NO_METHOD: + case UNKNOWN_METHOD: + break; + } + + tor_free(state); +} + +/** Return the approximate number of bytes allocated for <b>state</b>. */ +size_t +tor_compress_state_size(const tor_compress_state_t *state) +{ + tor_assert(state != NULL); + + switch (state->method) { + case GZIP_METHOD: + case ZLIB_METHOD: + return tor_zlib_compress_state_size(state->u.zlib_state); + case LZMA_METHOD: + return tor_lzma_compress_state_size(state->u.lzma_state); + case ZSTD_METHOD: + return tor_zstd_compress_state_size(state->u.zstd_state); + case NO_METHOD: + case UNKNOWN_METHOD: + goto err; + } + + err: + return 0; +} + diff --git a/src/common/compress.h b/src/common/compress.h new file mode 100644 index 0000000000..2d812e4430 --- /dev/null +++ b/src/common/compress.h @@ -0,0 +1,83 @@ +/* Copyright (c) 2003, Roger Dingledine + * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson. + * Copyright (c) 2007-2017, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +/** + * \file compress.h + * \brief Headers for compress.c + **/ + +#ifndef TOR_COMPRESS_H +#define TOR_COMPRESS_H + +/** Enumeration of what kind of compression to use. Only ZLIB_METHOD and + * GZIP_METHOD is guaranteed to be supported by the compress/uncompress + * functions here. Call tor_compress_supports_method() to check if a given + * compression schema is supported by Tor. */ +typedef enum { + NO_METHOD=0, + GZIP_METHOD=1, + ZLIB_METHOD=2, + LZMA_METHOD=3, + ZSTD_METHOD=4, + UNKNOWN_METHOD=5 +} compress_method_t; + +/** + * Enumeration to define tradeoffs between memory usage and compression level. + * HIGH_COMPRESSION saves the most bandwidth; LOW_COMPRESSION saves the most + * memory. + **/ +typedef enum { + HIGH_COMPRESSION, MEDIUM_COMPRESSION, LOW_COMPRESSION +} compression_level_t; + +int tor_compress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method); + +int tor_uncompress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method, + int complete_only, + int protocol_warn_level); + +compress_method_t detect_compression_method(const char *in, size_t in_len); + +int tor_compress_is_compression_bomb(size_t size_in, size_t size_out); + +int tor_compress_supports_method(compress_method_t method); + +const char *tor_compress_version_str(compress_method_t method); + +const char *tor_compress_header_version_str(compress_method_t method); + +size_t tor_compress_get_total_allocation(void); + +/** Return values from tor_compress_process; see that function's documentation + * for details. */ +typedef enum { + TOR_COMPRESS_OK, + TOR_COMPRESS_DONE, + TOR_COMPRESS_BUFFER_FULL, + TOR_COMPRESS_ERROR +} tor_compress_output_t; + +/** Internal state for an incremental compression/decompression. */ +typedef struct tor_compress_state_t tor_compress_state_t; + +tor_compress_state_t *tor_compress_new(int compress, + compress_method_t method, + compression_level_t level); + +tor_compress_output_t tor_compress_process(tor_compress_state_t *state, + char **out, size_t *out_len, + const char **in, size_t *in_len, + int finish); +void tor_compress_free(tor_compress_state_t *state); + +size_t tor_compress_state_size(const tor_compress_state_t *state); + +#endif // TOR_COMPRESS_H. + diff --git a/src/common/compress_lzma.c b/src/common/compress_lzma.c new file mode 100644 index 0000000000..ae0327f581 --- /dev/null +++ b/src/common/compress_lzma.c @@ -0,0 +1,605 @@ +/* Copyright (c) 2004, Roger Dingledine. + * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson. + * Copyright (c) 2007-2017, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +/** + * \file compress_lzma.c + * \brief Compression backend for LZMA. + * + * This module should never be invoked directly. Use the compress module + * instead. + **/ + +#include "orconfig.h" + +#include "util.h" +#include "torlog.h" +#include "compress.h" +#include "compress_lzma.h" + +#ifdef HAVE_LZMA +#include <lzma.h> +#endif + +/** Total number of bytes allocated for LZMA state. */ +static size_t total_lzma_allocation = 0; + +#ifdef HAVE_LZMA +/** Given <b>level</b> return the memory level. */ +static int +memory_level(compression_level_t level) +{ + switch (level) { + default: + case HIGH_COMPRESSION: return 9; + case MEDIUM_COMPRESSION: return 6; + case LOW_COMPRESSION: return 3; + } +} + +/** Convert a given <b>error</b> to a human readable error string. */ +static const char * +lzma_error_str(lzma_ret error) +{ + switch (error) { + case LZMA_OK: + return "Operation completed successfully"; + case LZMA_STREAM_END: + return "End of stream"; + case LZMA_NO_CHECK: + return "Input stream lacks integrity check"; + case LZMA_UNSUPPORTED_CHECK: + return "Unable to calculate integrity check"; + case LZMA_GET_CHECK: + return "Integrity check available"; + case LZMA_MEM_ERROR: + return "Unable to allocate memory"; + case LZMA_MEMLIMIT_ERROR: + return "Memory limit reached"; + case LZMA_FORMAT_ERROR: + return "Unknown file format"; + case LZMA_OPTIONS_ERROR: + return "Unsupported options"; + case LZMA_DATA_ERROR: + return "Corrupt input data"; + case LZMA_BUF_ERROR: + return "Unable to progress"; + case LZMA_PROG_ERROR: + return "Programming error"; + default: + return "Unknown LZMA error"; + } +} +#endif // HAVE_LZMA. + +/** Return 1 if LZMA compression is supported; otherwise 0. */ +int +tor_lzma_method_supported(void) +{ +#ifdef HAVE_LZMA + return 1; +#else + return 0; +#endif +} + +/** Return a string representation of the version of the currently running + * version of liblzma. Returns NULL if LZMA is unsupported. */ +const char * +tor_lzma_get_version_str(void) +{ +#ifdef HAVE_LZMA + return lzma_version_string(); +#else + return NULL; +#endif +} + +/** Return a string representation of the version of liblzma used at + * compilation time. Returns NULL if LZMA is unsupported. */ +const char * +tor_lzma_get_header_version_str(void) +{ +#ifdef HAVE_LZMA + return LZMA_VERSION_STRING; +#else + return NULL; +#endif +} + +/** Given <b>in_len</b> bytes at <b>in</b>, compress them into a newly + * allocated buffer, using the LZMA method. Store the compressed string in + * *<b>out</b>, and its length in *<b>out_len</b>. Return 0 on success, -1 on + * failure. + */ +int +tor_lzma_compress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method) +{ +#ifdef HAVE_LZMA + lzma_stream stream = LZMA_STREAM_INIT; + lzma_options_lzma stream_options; + lzma_ret retval; + lzma_action action; + size_t out_size, old_size; + off_t offset; + + tor_assert(out); + tor_assert(out_len); + tor_assert(in); + tor_assert(in_len < UINT_MAX); + tor_assert(method == LZMA_METHOD); + + stream.next_in = (unsigned char *)in; + stream.avail_in = in_len; + + lzma_lzma_preset(&stream_options, + memory_level(HIGH_COMPRESSION)); + + retval = lzma_alone_encoder(&stream, &stream_options); + + if (retval != LZMA_OK) { + log_warn(LD_GENERAL, "Error from LZMA encoder: %s (%u).", + lzma_error_str(retval), retval); + goto err; + } + + out_size = in_len / 2; + if (out_size < 1024) + out_size = 1024; + + *out = tor_malloc(out_size); + + stream.next_out = (unsigned char *)*out; + stream.avail_out = out_size; + + action = LZMA_RUN; + + while (1) { + retval = lzma_code(&stream, action); + switch (retval) { + case LZMA_OK: + action = LZMA_FINISH; + break; + case LZMA_STREAM_END: + goto done; + case LZMA_BUF_ERROR: + offset = stream.next_out - ((unsigned char *)*out); + old_size = out_size; + out_size *= 2; + + if (out_size < old_size) { + log_warn(LD_GENERAL, "Size overflow in LZMA compression."); + goto err; + } + + *out = tor_realloc(*out, out_size); + stream.next_out = (unsigned char *)(*out + offset); + if (out_size - offset > UINT_MAX) { + log_warn(LD_BUG, "Ran over unsigned int limit of LZMA while " + "compressing."); + goto err; + } + stream.avail_out = (unsigned int)(out_size - offset); + break; + + // We list all the possible values of `lzma_ret` here to silence the + // `switch-enum` warning and to detect if a new member was added. + case LZMA_NO_CHECK: + case LZMA_UNSUPPORTED_CHECK: + case LZMA_GET_CHECK: + case LZMA_MEM_ERROR: + case LZMA_MEMLIMIT_ERROR: + case LZMA_FORMAT_ERROR: + case LZMA_OPTIONS_ERROR: + case LZMA_DATA_ERROR: + case LZMA_PROG_ERROR: + default: + log_warn(LD_GENERAL, "LZMA compression didn't finish: %s.", + lzma_error_str(retval)); + goto err; + } + } + + done: + *out_len = stream.total_out; + lzma_end(&stream); + + if (tor_compress_is_compression_bomb(*out_len, in_len)) { + log_warn(LD_BUG, "We compressed something and got an insanely high " + "compression factor; other Tor instances would think " + "this is a compression bomb."); + goto err; + } + + return 0; + + err: + lzma_end(&stream); + tor_free(*out); + return -1; +#else // HAVE_LZMA. + (void)out; + (void)out_len; + (void)in; + (void)in_len; + (void)method; + + return -1; +#endif // HAVE_LZMA. +} + +/** Given an LZMA compressed string of total length <b>in_len</b> bytes at + * <b>in</b>, uncompress them into a newly allocated buffer. Store the + * uncompressed string in *<b>out</b>, and its length in *<b>out_len</b>. + * Return 0 on success, -1 on failure. + * + * If <b>complete_only</b> is true, we consider a truncated input as a failure; + * otherwise we decompress as much as we can. Warn about truncated or corrupt + * inputs at <b>protocol_warn_level</b>. + */ +int +tor_lzma_uncompress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method, + int complete_only, + int protocol_warn_level) +{ +#ifdef HAVE_LZMA + lzma_stream stream = LZMA_STREAM_INIT; + lzma_ret retval; + lzma_action action; + size_t out_size, old_size; + off_t offset; + + tor_assert(out); + tor_assert(out_len); + tor_assert(in); + tor_assert(in_len < UINT_MAX); + tor_assert(method == LZMA_METHOD); + + stream.next_in = (unsigned char *)in; + stream.avail_in = in_len; + + // FIXME(ahf): This should be something more sensible than + // UINT64_MAX: See #21665. + retval = lzma_alone_decoder(&stream, UINT64_MAX); + + if (retval != LZMA_OK) { + log_warn(LD_GENERAL, "Error from LZMA decoder: %s (%u).", + lzma_error_str(retval), retval); + goto err; + } + + out_size = in_len * 2; + if (out_size < 1024) + out_size = 1024; + + if (out_size >= SIZE_T_CEILING || out_size > UINT_MAX) + goto err; + + *out = tor_malloc(out_size); + stream.next_out = (unsigned char *)*out; + stream.avail_out = out_size; + + // FIXME(ahf): We should figure out how to use LZMA_FULL_FLUSH to + // make the partial string read tests. + // action = complete_only ? LZMA_FINISH : LZMA_SYNC_FLUSH. // To do this, + // it seems like we have to use LZMA using their "xz" encoder instead of just + // regular LZMA. + (void)complete_only; + action = LZMA_FINISH; + + while (1) { + retval = lzma_code(&stream, action); + switch (retval) { + case LZMA_STREAM_END: + if (stream.avail_in == 0) + goto done; + + // We might have more data here. Reset our stream. + lzma_end(&stream); + + retval = lzma_alone_decoder(&stream, UINT64_MAX); + + if (retval != LZMA_OK) { + log_warn(LD_GENERAL, "Error from LZMA decoder: %s (%u).", + lzma_error_str(retval), retval); + goto err; + } + break; + case LZMA_OK: + break; + case LZMA_BUF_ERROR: + if (stream.avail_out > 0) { + log_fn(protocol_warn_level, LD_PROTOCOL, + "possible truncated or corrupt LZMA data."); + goto err; + } + + offset = stream.next_out - (unsigned char *)*out; + old_size = out_size; + out_size *= 2; + + if (out_size < old_size) { + log_warn(LD_GENERAL, "Size overflow in LZMA uncompression."); + goto err; + } + + if (tor_compress_is_compression_bomb(in_len, out_size)) { + log_warn(LD_GENERAL, "Input looks like a possible LZMA compression " + "bomb. Not proceeding."); + goto err; + } + + if (out_size >= SIZE_T_CEILING) { + log_warn(LD_BUG, "Hit SIZE_T_CEILING limit while uncompressing " + "LZMA data."); + goto err; + } + + *out = tor_realloc(*out, out_size); + stream.next_out = (unsigned char *)(*out + offset); + + if (out_size - offset > UINT_MAX) { + log_warn(LD_BUG, "Ran over unsigned int limit of LZMA while " + "uncompressing."); + goto err; + } + + stream.avail_out = (unsigned int)(out_size - offset); + break; + + // We list all the possible values of `lzma_ret` here to silence the + // `switch-enum` warning and to detect if a new member was added. + case LZMA_NO_CHECK: + case LZMA_UNSUPPORTED_CHECK: + case LZMA_GET_CHECK: + case LZMA_MEM_ERROR: + case LZMA_MEMLIMIT_ERROR: + case LZMA_FORMAT_ERROR: + case LZMA_OPTIONS_ERROR: + case LZMA_DATA_ERROR: + case LZMA_PROG_ERROR: + default: + log_warn(LD_GENERAL, "LZMA decompression didn't finish: %s.", + lzma_error_str(retval)); + goto err; + } + } + + done: + *out_len = stream.next_out - (unsigned char*)*out; + lzma_end(&stream); + + // NUL-terminate our output. + if (out_size == *out_len) + *out = tor_realloc(*out, out_size + 1); + (*out)[*out_len] = '\0'; + + return 0; + + err: + lzma_end(&stream); + tor_free(*out); + return -1; +#else // HAVE_LZMA. + (void)out; + (void)out_len; + (void)in; + (void)in_len; + (void)method; + (void)complete_only; + (void)protocol_warn_level; + + return -1; +#endif // HAVE_LZMA. +} + +/** Internal LZMA state for incremental compression/decompression. + * The body of this struct is not exposed. */ +struct tor_lzma_compress_state_t { +#ifdef HAVE_LZMA + lzma_stream stream; /**< The LZMA stream. */ +#endif + + int compress; /**< True if we are compressing; false if we are inflating */ + + /** Number of bytes read so far. Used to detect compression bombs. */ + size_t input_so_far; + /** Number of bytes written so far. Used to detect compression bombs. */ + size_t output_so_far; + + /** Approximate number of bytes allocated for this object. */ + size_t allocation; +}; + +/** Construct and return a tor_lzma_compress_state_t object using + * <b>method</b>. If <b>compress</b>, it's for compression; otherwise it's for + * decompression. */ +tor_lzma_compress_state_t * +tor_lzma_compress_new(int compress, + compress_method_t method, + compression_level_t compression_level) +{ + tor_assert(method == LZMA_METHOD); + +#ifdef HAVE_LZMA + tor_lzma_compress_state_t *result; + lzma_ret retval; + lzma_options_lzma stream_options; + + // Note that we do not explicitly initialize the lzma_stream object here, + // since the LZMA_STREAM_INIT "just" initializes all members to 0, which is + // also what `tor_malloc_zero()` does. + result = tor_malloc_zero(sizeof(tor_lzma_compress_state_t)); + result->compress = compress; + + // FIXME(ahf): We should either try to do the pre-calculation that is done + // with the zlib backend or use a custom allocator here where we pass our + // tor_lzma_compress_state_t as the opaque value. + result->allocation = 0; + + if (compress) { + lzma_lzma_preset(&stream_options, + memory_level(compression_level)); + + retval = lzma_alone_encoder(&result->stream, &stream_options); + + if (retval != LZMA_OK) { + log_warn(LD_GENERAL, "Error from LZMA encoder: %s (%u).", + lzma_error_str(retval), retval); + goto err; + } + } else { + // FIXME(ahf): This should be something more sensible than + // UINT64_MAX: See #21665. + retval = lzma_alone_decoder(&result->stream, UINT64_MAX); + + if (retval != LZMA_OK) { + log_warn(LD_GENERAL, "Error from LZMA decoder: %s (%u).", + lzma_error_str(retval), retval); + goto err; + } + } + + return result; + + err: + tor_free(result); + return NULL; +#else // HAVE_LZMA. + (void)compress; + (void)method; + (void)compression_level; + + return NULL; +#endif // HAVE_LZMA. +} + +/** Compress/decompress some bytes using <b>state</b>. Read up to + * *<b>in_len</b> bytes from *<b>in</b>, and write up to *<b>out_len</b> bytes + * to *<b>out</b>, adjusting the values as we go. If <b>finish</b> is true, + * we've reached the end of the input. + * + * Return TOR_COMPRESS_DONE if we've finished the entire + * compression/decompression. + * Return TOR_COMPRESS_OK if we're processed everything from the input. + * Return TOR_COMPRESS_BUFFER_FULL if we're out of space on <b>out</b>. + * Return TOR_COMPRESS_ERROR if the stream is corrupt. + */ +tor_compress_output_t +tor_lzma_compress_process(tor_lzma_compress_state_t *state, + char **out, size_t *out_len, + const char **in, size_t *in_len, + int finish) +{ +#ifdef HAVE_LZMA + lzma_ret retval; + lzma_action action; + + tor_assert(state != NULL); + tor_assert(*in_len <= UINT_MAX); + tor_assert(*out_len <= UINT_MAX); + + state->stream.next_in = (unsigned char *)*in; + state->stream.avail_in = *in_len; + state->stream.next_out = (unsigned char *)*out; + state->stream.avail_out = *out_len; + + action = finish ? LZMA_FINISH : LZMA_RUN; + + retval = lzma_code(&state->stream, action); + + state->input_so_far += state->stream.next_in - ((unsigned char *)*in); + state->output_so_far += state->stream.next_out - ((unsigned char *)*out); + + *out = (char *)state->stream.next_out; + *out_len = state->stream.avail_out; + *in = (const char *)state->stream.next_in; + *in_len = state->stream.avail_in; + + if (! state->compress && + tor_compress_is_compression_bomb(state->input_so_far, + state->output_so_far)) { + log_warn(LD_DIR, "Possible compression bomb; abandoning stream."); + return TOR_COMPRESS_ERROR; + } + + switch (retval) { + case LZMA_OK: + if (state->stream.avail_out == 0 || finish) + return TOR_COMPRESS_BUFFER_FULL; + + return TOR_COMPRESS_OK; + + case LZMA_BUF_ERROR: + if (state->stream.avail_in == 0 && !finish) + return TOR_COMPRESS_OK; + + return TOR_COMPRESS_BUFFER_FULL; + + case LZMA_STREAM_END: + return TOR_COMPRESS_DONE; + + // We list all the possible values of `lzma_ret` here to silence the + // `switch-enum` warning and to detect if a new member was added. + case LZMA_NO_CHECK: + case LZMA_UNSUPPORTED_CHECK: + case LZMA_GET_CHECK: + case LZMA_MEM_ERROR: + case LZMA_MEMLIMIT_ERROR: + case LZMA_FORMAT_ERROR: + case LZMA_OPTIONS_ERROR: + case LZMA_DATA_ERROR: + case LZMA_PROG_ERROR: + default: + log_warn(LD_GENERAL, "LZMA %s didn't finish: %s.", + state->compress ? "compression" : "decompression", + lzma_error_str(retval)); + return TOR_COMPRESS_ERROR; + } +#else // HAVE_LZMA. + (void)state; + (void)out; + (void)out_len; + (void)in; + (void)in_len; + (void)finish; + return TOR_COMPRESS_ERROR; +#endif // HAVE_LZMA. +} + +/** Deallocate <b>state</b>. */ +void +tor_lzma_compress_free(tor_lzma_compress_state_t *state) +{ + if (state == NULL) + return; + + total_lzma_allocation -= state->allocation; + +#ifdef HAVE_LZMA + lzma_end(&state->stream); +#endif + + tor_free(state); +} + +/** Return the approximate number of bytes allocated for <b>state</b>. */ +size_t +tor_lzma_compress_state_size(const tor_lzma_compress_state_t *state) +{ + tor_assert(state != NULL); + return state->allocation; +} + +/** Return the approximate number of bytes allocated for all LZMA states. */ +size_t +tor_lzma_get_total_allocation(void) +{ + return total_lzma_allocation; +} + diff --git a/src/common/compress_lzma.h b/src/common/compress_lzma.h new file mode 100644 index 0000000000..71de56a1c6 --- /dev/null +++ b/src/common/compress_lzma.h @@ -0,0 +1,51 @@ +/* Copyright (c) 2003, Roger Dingledine + * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson. + * Copyright (c) 2007-2017, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +/** + * \file compress_lzma.h + * \brief Header for compress_lzma.c + **/ + +#ifndef TOR_COMPRESS_LZMA_H +#define TOR_COMPRESS_LZMA_H + +int tor_lzma_method_supported(void); + +const char *tor_lzma_get_version_str(void); + +const char *tor_lzma_get_header_version_str(void); + +int tor_lzma_compress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method); + +int tor_lzma_uncompress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method, + int complete_only, + int protocol_warn_level); + +/** Internal state for an incremental LZMA compression/decompression. */ +typedef struct tor_lzma_compress_state_t tor_lzma_compress_state_t; + +tor_lzma_compress_state_t * +tor_lzma_compress_new(int compress, + compress_method_t method, + compression_level_t compression_level); + +tor_compress_output_t +tor_lzma_compress_process(tor_lzma_compress_state_t *state, + char **out, size_t *out_len, + const char **in, size_t *in_len, + int finish); + +void tor_lzma_compress_free(tor_lzma_compress_state_t *state); + +size_t tor_lzma_compress_state_size(const tor_lzma_compress_state_t *state); + +size_t tor_lzma_get_total_allocation(void); + +#endif // TOR_COMPRESS_LZMA_H. + diff --git a/src/common/torgzip.c b/src/common/compress_zlib.c index 223cd307a7..50bdf9ece1 100644 --- a/src/common/torgzip.c +++ b/src/common/compress_zlib.c @@ -4,25 +4,19 @@ /* See LICENSE for licensing information */ /** - * \file torgzip.c - * \brief A simple in-memory gzip implementation. + * \file compress_zlib.c + * \brief Compression backend for gzip and zlib. + * + * This module should never be invoked directly. Use the compress module + * instead. **/ #include "orconfig.h" -#include <stdlib.h> -#include <stdio.h> -#include <assert.h> -#include <string.h> -#include "torint.h" - -#ifdef HAVE_NETINET_IN_H -#include <netinet/in.h> -#endif - #include "util.h" #include "torlog.h" -#include "torgzip.h" +#include "compress.h" +#include "compress_zlib.h" /* zlib 1.2.4 and 1.2.5 do some "clever" things with macros. Instead of saying "(defined(FOO) ? FOO : 0)" they like to say "FOO-0", on the theory @@ -56,25 +50,21 @@ static size_t tor_zlib_state_size_precalc(int inflate, /** Total number of bytes allocated for zlib state */ static size_t total_zlib_allocation = 0; -/** Return a string representation of the version of the currently running - * version of zlib. */ -const char * -tor_zlib_get_version_str(void) -{ - return zlibVersion(); -} - -/** Return a string representation of the version of the version of zlib -* used at compilation. */ -const char * -tor_zlib_get_header_version_str(void) +/** Given <b>level</b> return the memory level. */ +static int +memory_level(compression_level_t level) { - return ZLIB_VERSION; + switch (level) { + default: + case HIGH_COMPRESSION: return 8; + case MEDIUM_COMPRESSION: return 7; + case LOW_COMPRESSION: return 6; + } } /** Return the 'bits' value to tell zlib to use <b>method</b>.*/ static inline int -method_bits(compress_method_t method, zlib_compression_level_t level) +method_bits(compress_method_t method, compression_level_t level) { /* Bits+16 means "use gzip" in zlib >= 1.2 */ const int flag = method == GZIP_METHOD ? 16 : 0; @@ -86,53 +76,39 @@ method_bits(compress_method_t method, zlib_compression_level_t level) } } -static inline int -get_memlevel(zlib_compression_level_t level) +/** Return 1 if zlib/gzip compression is supported; otherwise 0. */ +int +tor_zlib_method_supported(void) { - switch (level) { - default: - case HIGH_COMPRESSION: return 8; - case MEDIUM_COMPRESSION: return 7; - case LOW_COMPRESSION: return 6; - } + /* We currently always support zlib/gzip, but we keep this function around in + * case we some day decide to deprecate zlib/gzip support. + */ + return 1; } -/** @{ */ -/* These macros define the maximum allowable compression factor. Anything of - * size greater than CHECK_FOR_COMPRESSION_BOMB_AFTER is not allowed to - * have an uncompression factor (uncompressed size:compressed size ratio) of - * any greater than MAX_UNCOMPRESSION_FACTOR. - * - * Picking a value for MAX_UNCOMPRESSION_FACTOR is a trade-off: we want it to - * be small to limit the attack multiplier, but we also want it to be large - * enough so that no legitimate document --even ones we might invent in the - * future -- ever compresses by a factor of greater than - * MAX_UNCOMPRESSION_FACTOR. Within those parameters, there's a reasonably - * large range of possible values. IMO, anything over 8 is probably safe; IMO - * anything under 50 is probably sufficient. - */ -#define MAX_UNCOMPRESSION_FACTOR 25 -#define CHECK_FOR_COMPRESSION_BOMB_AFTER (1024*64) -/** @} */ - -/** Return true if uncompressing an input of size <b>in_size</b> to an input - * of size at least <b>size_out</b> looks like a compression bomb. */ -static int -is_compression_bomb(size_t size_in, size_t size_out) +/** Return a string representation of the version of the currently running + * version of zlib. */ +const char * +tor_zlib_get_version_str(void) { - if (size_in == 0 || size_out < CHECK_FOR_COMPRESSION_BOMB_AFTER) - return 0; + return zlibVersion(); +} - return (size_out / size_in > MAX_UNCOMPRESSION_FACTOR); +/** Return a string representation of the version of the version of zlib +* used at compilation. */ +const char * +tor_zlib_get_header_version_str(void) +{ + return ZLIB_VERSION; } /** Given <b>in_len</b> bytes at <b>in</b>, compress them into a newly * allocated buffer, using the method described in <b>method</b>. Store the - * compressed string in *<b>out</b>, and its length in *<b>out_len</b>. - * Return 0 on success, -1 on failure. + * compressed string in *<b>out</b>, and its length in *<b>out_len</b>. Return + * 0 on success, -1 on failure. */ int -tor_gzip_compress(char **out, size_t *out_len, +tor_zlib_compress(char **out, size_t *out_len, const char *in, size_t in_len, compress_method_t method) { @@ -156,7 +132,7 @@ tor_gzip_compress(char **out, size_t *out_len, if (deflateInit2(stream, Z_BEST_COMPRESSION, Z_DEFLATED, method_bits(method, HIGH_COMPRESSION), - get_memlevel(HIGH_COMPRESSION), + memory_level(HIGH_COMPRESSION), Z_DEFAULT_STRATEGY) != Z_OK) { //LCOV_EXCL_START -- we can only provoke failure by giving junk arguments. log_warn(LD_GENERAL, "Error from deflateInit2: %s", @@ -224,7 +200,7 @@ tor_gzip_compress(char **out, size_t *out_len, } tor_free(stream); - if (is_compression_bomb(*out_len, in_len)) { + if (tor_compress_is_compression_bomb(*out_len, in_len)) { log_warn(LD_BUG, "We compressed something and got an insanely high " "compression factor; other Tors would think this was a zlib bomb."); goto err; @@ -240,19 +216,17 @@ tor_gzip_compress(char **out, size_t *out_len, return -1; } -/** Given zero or more zlib-compressed or gzip-compressed strings of - * total length - * <b>in_len</b> bytes at <b>in</b>, uncompress them into a newly allocated - * buffer, using the method described in <b>method</b>. Store the uncompressed - * string in *<b>out</b>, and its length in *<b>out_len</b>. Return 0 on - * success, -1 on failure. +/** Given an Zlib/Gzip compressed string of total length <b>in_len</b> bytes + * at <b>in</b>, uncompress them into a newly allocated buffer. Store the + * uncompressed string in *<b>out</b>, and its length in *<b>out_len</b>. + * Return 0 on success, -1 on failure. * - * If <b>complete_only</b> is true, we consider a truncated input as a - * failure; otherwise we decompress as much as we can. Warn about truncated - * or corrupt inputs at <b>protocol_warn_level</b>. + * If <b>complete_only</b> is true, we consider a truncated input as a failure; + * otherwise we decompress as much as we can. Warn about truncated or corrupt + * inputs at <b>protocol_warn_level</b>. */ int -tor_gzip_uncompress(char **out, size_t *out_len, +tor_zlib_uncompress(char **out, size_t *out_len, const char *in, size_t in_len, compress_method_t method, int complete_only, @@ -332,7 +306,7 @@ tor_gzip_uncompress(char **out, size_t *out_len, log_warn(LD_GENERAL, "Size overflow in uncompression."); goto err; } - if (is_compression_bomb(in_len, out_size)) { + if (tor_compress_is_compression_bomb(in_len, out_size)) { log_warn(LD_GENERAL, "Input looks like a possible zlib bomb; " "not proceeding."); goto err; @@ -382,26 +356,9 @@ tor_gzip_uncompress(char **out, size_t *out_len, return -1; } -/** Try to tell whether the <b>in_len</b>-byte string in <b>in</b> is likely - * to be compressed or not. If it is, return the likeliest compression method. - * Otherwise, return UNKNOWN_METHOD. - */ -compress_method_t -detect_compression_method(const char *in, size_t in_len) -{ - if (in_len > 2 && fast_memeq(in, "\x1f\x8b", 2)) { - return GZIP_METHOD; - } else if (in_len > 2 && (in[0] & 0x0f) == 8 && - (ntohs(get_uint16(in)) % 31) == 0) { - return ZLIB_METHOD; - } else { - return UNKNOWN_METHOD; - } -} - -/** Internal state for an incremental zlib compression/decompression. The - * body of this struct is not exposed. */ -struct tor_zlib_state_t { +/** Internal zlib state for an incremental compression/decompression. + * The body of this struct is not exposed. */ +struct tor_zlib_compress_state_t { struct z_stream_s stream; /**< The zlib stream */ int compress; /**< True if we are compressing; false if we are inflating */ @@ -414,47 +371,79 @@ struct tor_zlib_state_t { size_t allocation; }; -/** Construct and return a tor_zlib_state_t object using <b>method</b>. If - * <b>compress</b>, it's for compression; otherwise it's for +/** Return an approximate number of bytes used in RAM to hold a state with + * window bits <b>windowBits</b> and compression level 'memlevel' */ +static size_t +tor_zlib_state_size_precalc(int inflate_, int windowbits, int memlevel) +{ + windowbits &= 15; + +#define A_FEW_KILOBYTES 2048 + + if (inflate_) { + /* From zconf.h: + + "The memory requirements for inflate are (in bytes) 1 << windowBits + that is, 32K for windowBits=15 (default value) plus a few kilobytes + for small objects." + */ + return sizeof(tor_zlib_compress_state_t) + sizeof(struct z_stream_s) + + (1 << 15) + A_FEW_KILOBYTES; + } else { + /* Also from zconf.h: + + "The memory requirements for deflate are (in bytes): + (1 << (windowBits+2)) + (1 << (memLevel+9)) + ... plus a few kilobytes for small objects." + */ + return sizeof(tor_zlib_compress_state_t) + sizeof(struct z_stream_s) + + (1 << (windowbits + 2)) + (1 << (memlevel + 9)) + A_FEW_KILOBYTES; + } +#undef A_FEW_KILOBYTES +} + +/** Construct and return a tor_zlib_compress_state_t object using + * <b>method</b>. If <b>compress</b>, it's for compression; otherwise it's for * decompression. */ -tor_zlib_state_t * -tor_zlib_new(int compress_, compress_method_t method, - zlib_compression_level_t compression_level) +tor_zlib_compress_state_t * +tor_zlib_compress_new(int compress, + compress_method_t method, + compression_level_t compression_level) { - tor_zlib_state_t *out; + tor_zlib_compress_state_t *out; int bits, memlevel; - if (! compress_) { - /* use this setting for decompression, since we might have the - * max number of window bits */ - compression_level = HIGH_COMPRESSION; - } - - out = tor_malloc_zero(sizeof(tor_zlib_state_t)); - out->stream.zalloc = Z_NULL; - out->stream.zfree = Z_NULL; - out->stream.opaque = NULL; - out->compress = compress_; - bits = method_bits(method, compression_level); - memlevel = get_memlevel(compression_level); - if (compress_) { - if (deflateInit2(&out->stream, Z_BEST_COMPRESSION, Z_DEFLATED, - bits, memlevel, - Z_DEFAULT_STRATEGY) != Z_OK) - goto err; // LCOV_EXCL_LINE - } else { - if (inflateInit2(&out->stream, bits) != Z_OK) - goto err; // LCOV_EXCL_LINE - } - out->allocation = tor_zlib_state_size_precalc(!compress_, bits, memlevel); - - total_zlib_allocation += out->allocation; - - return out; + if (! compress) { + /* use this setting for decompression, since we might have the + * max number of window bits */ + compression_level = HIGH_COMPRESSION; + } + + out = tor_malloc_zero(sizeof(tor_zlib_compress_state_t)); + out->stream.zalloc = Z_NULL; + out->stream.zfree = Z_NULL; + out->stream.opaque = NULL; + out->compress = compress; + bits = method_bits(method, compression_level); + memlevel = memory_level(compression_level); + if (compress) { + if (deflateInit2(&out->stream, Z_BEST_COMPRESSION, Z_DEFLATED, + bits, memlevel, + Z_DEFAULT_STRATEGY) != Z_OK) + goto err; // LCOV_EXCL_LINE + } else { + if (inflateInit2(&out->stream, bits) != Z_OK) + goto err; // LCOV_EXCL_LINE + } + out->allocation = tor_zlib_state_size_precalc(!compress, bits, memlevel); + + total_zlib_allocation += out->allocation; + + return out; err: - tor_free(out); - return NULL; + tor_free(out); + return NULL; } /** Compress/decompress some bytes using <b>state</b>. Read up to @@ -462,18 +451,20 @@ tor_zlib_new(int compress_, compress_method_t method, * to *<b>out</b>, adjusting the values as we go. If <b>finish</b> is true, * we've reached the end of the input. * - * Return TOR_ZLIB_DONE if we've finished the entire compression/decompression. - * Return TOR_ZLIB_OK if we're processed everything from the input. - * Return TOR_ZLIB_BUF_FULL if we're out of space on <b>out</b>. - * Return TOR_ZLIB_ERR if the stream is corrupt. + * Return TOR_COMPRESS_DONE if we've finished the entire + * compression/decompression. + * Return TOR_COMPRESS_OK if we're processed everything from the input. + * Return TOR_COMPRESS_BUFFER_FULL if we're out of space on <b>out</b>. + * Return TOR_COMPRESS_ERROR if the stream is corrupt. */ -tor_zlib_output_t -tor_zlib_process(tor_zlib_state_t *state, - char **out, size_t *out_len, - const char **in, size_t *in_len, - int finish) +tor_compress_output_t +tor_zlib_compress_process(tor_zlib_compress_state_t *state, + char **out, size_t *out_len, + const char **in, size_t *in_len, + int finish) { int err; + tor_assert(state != NULL); tor_assert(*in_len <= UINT_MAX); tor_assert(*out_len <= UINT_MAX); state->stream.next_in = (unsigned char*) *in; @@ -496,38 +487,39 @@ tor_zlib_process(tor_zlib_state_t *state, *in_len = state->stream.avail_in; if (! state->compress && - is_compression_bomb(state->input_so_far, state->output_so_far)) { + tor_compress_is_compression_bomb(state->input_so_far, + state->output_so_far)) { log_warn(LD_DIR, "Possible zlib bomb; abandoning stream."); - return TOR_ZLIB_ERR; + return TOR_COMPRESS_ERROR; } switch (err) { case Z_STREAM_END: - return TOR_ZLIB_DONE; + return TOR_COMPRESS_DONE; case Z_BUF_ERROR: if (state->stream.avail_in == 0 && !finish) - return TOR_ZLIB_OK; - return TOR_ZLIB_BUF_FULL; + return TOR_COMPRESS_OK; + return TOR_COMPRESS_BUFFER_FULL; case Z_OK: if (state->stream.avail_out == 0 || finish) - return TOR_ZLIB_BUF_FULL; - return TOR_ZLIB_OK; + return TOR_COMPRESS_BUFFER_FULL; + return TOR_COMPRESS_OK; default: log_warn(LD_GENERAL, "Gzip returned an error: %s", state->stream.msg ? state->stream.msg : "<no message>"); - return TOR_ZLIB_ERR; + return TOR_COMPRESS_ERROR; } } /** Deallocate <b>state</b>. */ void -tor_zlib_free(tor_zlib_state_t *state) +tor_zlib_compress_free(tor_zlib_compress_state_t *state) { - if (!state) + if (state == NULL) return; - total_zlib_allocation -= state->allocation; + total_zlib_allocation -= state->allocation; if (state->compress) deflateEnd(&state->stream); @@ -537,41 +529,11 @@ tor_zlib_free(tor_zlib_state_t *state) tor_free(state); } -/** Return an approximate number of bytes used in RAM to hold a state with - * window bits <b>windowBits</b> and compression level 'memlevel' */ -static size_t -tor_zlib_state_size_precalc(int inflate_, int windowbits, int memlevel) -{ - windowbits &= 15; - -#define A_FEW_KILOBYTES 2048 - - if (inflate_) { - /* From zconf.h: - - "The memory requirements for inflate are (in bytes) 1 << windowBits - that is, 32K for windowBits=15 (default value) plus a few kilobytes - for small objects." - */ - return sizeof(tor_zlib_state_t) + sizeof(struct z_stream_s) + - (1 << 15) + A_FEW_KILOBYTES; - } else { - /* Also from zconf.h: - - "The memory requirements for deflate are (in bytes): - (1 << (windowBits+2)) + (1 << (memLevel+9)) - ... plus a few kilobytes for small objects." - */ - return sizeof(tor_zlib_state_t) + sizeof(struct z_stream_s) + - (1 << (windowbits + 2)) + (1 << (memlevel + 9)) + A_FEW_KILOBYTES; - } -#undef A_FEW_KILOBYTES -} - /** Return the approximate number of bytes allocated for <b>state</b>. */ size_t -tor_zlib_state_size(const tor_zlib_state_t *state) +tor_zlib_compress_state_size(const tor_zlib_compress_state_t *state) { + tor_assert(state != NULL); return state->allocation; } diff --git a/src/common/compress_zlib.h b/src/common/compress_zlib.h new file mode 100644 index 0000000000..6e8e5c5136 --- /dev/null +++ b/src/common/compress_zlib.h @@ -0,0 +1,51 @@ +/* Copyright (c) 2003, Roger Dingledine + * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson. + * Copyright (c) 2007-2017, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +/** + * \file compress_zlib.h + * \brief Header for compress_zlib.c + **/ + +#ifndef TOR_COMPRESS_ZLIB_H +#define TOR_COMPRESS_ZLIB_H + +int tor_zlib_method_supported(void); + +const char *tor_zlib_get_version_str(void); + +const char *tor_zlib_get_header_version_str(void); + +int tor_zlib_compress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method); + +int tor_zlib_uncompress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method, + int complete_only, + int protocol_warn_level); + +/** Internal state for an incremental zlib/gzip compression/decompression. */ +typedef struct tor_zlib_compress_state_t tor_zlib_compress_state_t; + +tor_zlib_compress_state_t * +tor_zlib_compress_new(int compress, + compress_method_t method, + compression_level_t compression_level); + +tor_compress_output_t +tor_zlib_compress_process(tor_zlib_compress_state_t *state, + char **out, size_t *out_len, + const char **in, size_t *in_len, + int finish); + +void tor_zlib_compress_free(tor_zlib_compress_state_t *state); + +size_t tor_zlib_compress_state_size(const tor_zlib_compress_state_t *state); + +size_t tor_zlib_get_total_allocation(void); + +#endif // TOR_COMPRESS_ZLIB_H. + diff --git a/src/common/compress_zstd.c b/src/common/compress_zstd.c new file mode 100644 index 0000000000..664cce1700 --- /dev/null +++ b/src/common/compress_zstd.c @@ -0,0 +1,609 @@ +/* Copyright (c) 2004, Roger Dingledine. + * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson. + * Copyright (c) 2007-2017, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +/** + * \file compress_zstd.c + * \brief Compression backend for Zstandard. + * + * This module should never be invoked directly. Use the compress module + * instead. + **/ + +#include "orconfig.h" + +#include "util.h" +#include "torlog.h" +#include "compress.h" +#include "compress_zstd.h" + +#ifdef HAVE_ZSTD +#include <zstd.h> +#include <zstd_errors.h> +#endif + +/** Total number of bytes allocated for Zstandard state. */ +static size_t total_zstd_allocation = 0; + +#ifdef HAVE_ZSTD +/** Given <b>level</b> return the memory level. */ +static int +memory_level(compression_level_t level) +{ + switch (level) { + default: + case HIGH_COMPRESSION: return 9; + case MEDIUM_COMPRESSION: return 8; + case LOW_COMPRESSION: return 7; + } +} +#endif // HAVE_ZSTD. + +/** Return 1 if Zstandard compression is supported; otherwise 0. */ +int +tor_zstd_method_supported(void) +{ +#ifdef HAVE_ZSTD + return 1; +#else + return 0; +#endif +} + +/** Return a string representation of the version of the currently running + * version of libzstd. Returns NULL if Zstandard is unsupported. */ +const char * +tor_zstd_get_version_str(void) +{ +#ifdef HAVE_ZSTD + static char version_str[16]; + size_t version_number; + + version_number = ZSTD_versionNumber(); + tor_snprintf(version_str, sizeof(version_str), + "%lu.%lu.%lu", + version_number / 10000 % 100, + version_number / 100 % 100, + version_number % 100); + + return version_str; +#else + return NULL; +#endif +} + +/** Return a string representation of the version of the version of libzstd + * used at compilation time. Returns NULL if Zstandard is unsupported. */ +const char * +tor_zstd_get_header_version_str(void) +{ +#ifdef HAVE_ZSTD + return ZSTD_VERSION_STRING; +#else + return NULL; +#endif +} + +/** Given <b>in_len</b> bytes at <b>in</b>, compress them into a newly + * allocated buffer, using the Zstandard method. Store the compressed string + * in *<b>out</b>, and its length in *<b>out_len</b>. Return 0 on success, -1 + * on failure. + */ +int +tor_zstd_compress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method) +{ +#ifdef HAVE_ZSTD + ZSTD_CStream *stream = NULL; + size_t out_size, old_size; + size_t retval; + + tor_assert(out); + tor_assert(out_len); + tor_assert(in); + tor_assert(in_len < UINT_MAX); + tor_assert(method == ZSTD_METHOD); + + *out = NULL; + + stream = ZSTD_createCStream(); + + if (stream == NULL) { + // Zstandard does not give us any useful error message to why this + // happened. See https://github.com/facebook/zstd/issues/398 + log_warn(LD_GENERAL, "Error while creating Zstandard stream"); + goto err; + } + + retval = ZSTD_initCStream(stream, + memory_level(HIGH_COMPRESSION)); + + if (ZSTD_isError(retval)) { + log_warn(LD_GENERAL, "Zstandard stream initialization error: %s", + ZSTD_getErrorName(retval)); + goto err; + } + + // Assume 50% compression and update our buffer in case we need to. + out_size = in_len / 2; + if (out_size < 1024) + out_size = 1024; + + *out = tor_malloc(out_size); + *out_len = 0; + + ZSTD_inBuffer input = { in, in_len, 0 }; + ZSTD_outBuffer output = { *out, out_size, 0 }; + + while (input.pos < input.size) { + retval = ZSTD_compressStream(stream, &output, &input); + + if (ZSTD_isError(retval)) { + log_warn(LD_GENERAL, "Zstandard stream compression error: %s", + ZSTD_getErrorName(retval)); + goto err; + } + + if (input.pos < input.size && output.pos == output.size) { + old_size = out_size; + out_size *= 2; + + if (out_size < old_size) { + log_warn(LD_GENERAL, "Size overflow in Zstandard compression."); + goto err; + } + + if (out_size - output.pos > UINT_MAX) { + log_warn(LD_BUG, "Ran over unsigned int limit of Zstandard while " + "compressing."); + goto err; + } + + output.dst = *out = tor_realloc(*out, out_size); + output.size = out_size; + } + } + + while (1) { + retval = ZSTD_endStream(stream, &output); + + if (retval == 0) + break; + + if (ZSTD_isError(retval)) { + log_warn(LD_GENERAL, "Zstandard stream error: %s", + ZSTD_getErrorName(retval)); + goto err; + } + + if (output.pos == output.size) { + old_size = out_size; + out_size *= 2; + + if (out_size < old_size) { + log_warn(LD_GENERAL, "Size overflow in Zstandard compression."); + goto err; + } + + if (out_size - output.pos > UINT_MAX) { + log_warn(LD_BUG, "Ran over unsigned int limit of Zstandard while " + "compressing."); + goto err; + } + + output.dst = *out = tor_realloc(*out, out_size); + output.size = out_size; + } + } + + *out_len = output.pos; + + if (tor_compress_is_compression_bomb(*out_len, in_len)) { + log_warn(LD_BUG, "We compressed something and got an insanely high " + "compression factor; other Tor instances would think " + "this is a compression bomb."); + goto err; + } + + if (stream != NULL) { + ZSTD_freeCStream(stream); + } + + return 0; + + err: + if (stream != NULL) { + ZSTD_freeCStream(stream); + } + + tor_free(*out); + return -1; +#else // HAVE_ZSTD. + (void)out; + (void)out_len; + (void)in; + (void)in_len; + (void)method; + + return -1; +#endif // HAVE_ZSTD. +} + +/** Given a Zstandard compressed string of total length <b>in_len</b> bytes at + * <b>in</b>, uncompress them into a newly allocated buffer. Store the + * uncompressed string in *<b>out</b>, and its length in *<b>out_len</b>. + * Return 0 on success, -1 on failure. + * + * If <b>complete_only</b> is true, we consider a truncated input as a failure; + * otherwise we decompress as much as we can. Warn about truncated or corrupt + * inputs at <b>protocol_warn_level</b>. + */ +int +tor_zstd_uncompress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method, + int complete_only, + int protocol_warn_level) +{ +#ifdef HAVE_ZSTD + ZSTD_DStream *stream = NULL; + size_t retval; + size_t out_size, old_size; + + tor_assert(out); + tor_assert(out_len); + tor_assert(in); + tor_assert(in_len < UINT_MAX); + tor_assert(method == ZSTD_METHOD); + + // FIXME(ahf): Handle this? + (void)complete_only; + (void)protocol_warn_level; + + *out = NULL; + + stream = ZSTD_createDStream(); + + if (stream == NULL) { + // Zstandard does not give us any useful error message to why this + // happened. See https://github.com/facebook/zstd/issues/398 + log_warn(LD_GENERAL, "Error while creating Zstandard stream"); + goto err; + } + + retval = ZSTD_initDStream(stream); + + if (ZSTD_isError(retval)) { + log_warn(LD_GENERAL, "Zstandard stream initialization error: %s", + ZSTD_getErrorName(retval)); + goto err; + } + + out_size = in_len * 2; + if (out_size < 1024) + out_size = 1024; + + if (out_size >= SIZE_T_CEILING || out_size > UINT_MAX) + goto err; + + *out = tor_malloc(out_size); + *out_len = 0; + + ZSTD_inBuffer input = { in, in_len, 0 }; + ZSTD_outBuffer output = { *out, out_size, 0 }; + + while (input.pos < input.size) { + retval = ZSTD_decompressStream(stream, &output, &input); + + if (ZSTD_isError(retval)) { + log_warn(LD_GENERAL, "Zstandard stream decompression error: %s", + ZSTD_getErrorName(retval)); + goto err; + } + + if (input.pos < input.size && output.pos == output.size) { + old_size = out_size; + out_size *= 2; + + if (out_size < old_size) { + log_warn(LD_GENERAL, "Size overflow in Zstandard compression."); + goto err; + } + + if (tor_compress_is_compression_bomb(in_len, out_size)) { + log_warn(LD_GENERAL, "Input looks like a possible Zstandard " + "compression bomb. Not proceeding."); + goto err; + } + + if (out_size >= SIZE_T_CEILING) { + log_warn(LD_BUG, "Hit SIZE_T_CEILING limit while uncompressing " + "Zstandard data."); + goto err; + } + + if (out_size - output.pos > UINT_MAX) { + log_warn(LD_BUG, "Ran over unsigned int limit of Zstandard while " + "decompressing."); + goto err; + } + + output.dst = *out = tor_realloc(*out, out_size); + output.size = out_size; + } + } + + *out_len = output.pos; + + if (stream != NULL) { + ZSTD_freeDStream(stream); + } + + // NUL-terminate our output. + if (out_size == *out_len) + *out = tor_realloc(*out, out_size + 1); + (*out)[*out_len] = '\0'; + + return 0; + + err: + if (stream != NULL) { + ZSTD_freeDStream(stream); + } + + tor_free(*out); + return -1; +#else // HAVE_ZSTD. + (void)out; + (void)out_len; + (void)in; + (void)in_len; + (void)method; + (void)complete_only; + (void)protocol_warn_level; + + return -1; +#endif // HAVE_ZSTD. +} + +/** Internal Zstandard state for incremental compression/decompression. + * The body of this struct is not exposed. */ +struct tor_zstd_compress_state_t { +#ifdef HAVE_ZSTD + union { + /** Compression stream. Used when <b>compress</b> is true. */ + ZSTD_CStream *compress_stream; + /** Decompression stream. Used when <b>compress</b> is false. */ + ZSTD_DStream *decompress_stream; + } u; /**< Zstandard stream objects. */ +#endif // HAVE_ZSTD. + + int compress; /**< True if we are compressing; false if we are inflating */ + + /** Number of bytes read so far. Used to detect compression bombs. */ + size_t input_so_far; + /** Number of bytes written so far. Used to detect compression bombs. */ + size_t output_so_far; + + /** Approximate number of bytes allocated for this object. */ + size_t allocation; +}; + +/** Construct and return a tor_zstd_compress_state_t object using + * <b>method</b>. If <b>compress</b>, it's for compression; otherwise it's for + * decompression. */ +tor_zstd_compress_state_t * +tor_zstd_compress_new(int compress, + compress_method_t method, + compression_level_t compression_level) +{ + tor_assert(method == ZSTD_METHOD); + +#ifdef HAVE_ZSTD + tor_zstd_compress_state_t *result; + size_t retval; + + result = tor_malloc_zero(sizeof(tor_zstd_compress_state_t)); + result->compress = compress; + + // FIXME(ahf): We should either try to do the pre-calculation that is done + // with the zlib backend or use a custom allocator here where we pass our + // tor_zstd_compress_state_t as the opaque value. + result->allocation = 0; + + if (compress) { + result->u.compress_stream = ZSTD_createCStream(); + + if (result->u.compress_stream == NULL) { + log_warn(LD_GENERAL, "Error while creating Zstandard stream"); + goto err; + } + + retval = ZSTD_initCStream(result->u.compress_stream, + memory_level(compression_level)); + + if (ZSTD_isError(retval)) { + log_warn(LD_GENERAL, "Zstandard stream initialization error: %s", + ZSTD_getErrorName(retval)); + goto err; + } + } else { + result->u.decompress_stream = ZSTD_createDStream(); + + if (result->u.decompress_stream == NULL) { + log_warn(LD_GENERAL, "Error while creating Zstandard stream"); + goto err; + } + + retval = ZSTD_initDStream(result->u.decompress_stream); + + if (ZSTD_isError(retval)) { + log_warn(LD_GENERAL, "Zstandard stream initialization error: %s", + ZSTD_getErrorName(retval)); + goto err; + } + } + + return result; + + err: + if (compress) { + ZSTD_freeCStream(result->u.compress_stream); + } else { + ZSTD_freeDStream(result->u.decompress_stream); + } + + tor_free(result); + return NULL; +#else // HAVE_ZSTD. + (void)compress; + (void)method; + (void)compression_level; + + return NULL; +#endif // HAVE_ZSTD. +} + +/** Compress/decompress some bytes using <b>state</b>. Read up to + * *<b>in_len</b> bytes from *<b>in</b>, and write up to *<b>out_len</b> bytes + * to *<b>out</b>, adjusting the values as we go. If <b>finish</b> is true, + * we've reached the end of the input. + * + * Return TOR_COMPRESS_DONE if we've finished the entire + * compression/decompression. + * Return TOR_COMPRESS_OK if we're processed everything from the input. + * Return TOR_COMPRESS_BUFFER_FULL if we're out of space on <b>out</b>. + * Return TOR_COMPRESS_ERROR if the stream is corrupt. + */ +tor_compress_output_t +tor_zstd_compress_process(tor_zstd_compress_state_t *state, + char **out, size_t *out_len, + const char **in, size_t *in_len, + int finish) +{ +#ifdef HAVE_ZSTD + size_t retval; + + tor_assert(state != NULL); + tor_assert(*in_len <= UINT_MAX); + tor_assert(*out_len <= UINT_MAX); + + ZSTD_inBuffer input = { *in, *in_len, 0 }; + ZSTD_outBuffer output = { *out, *out_len, 0 }; + + if (state->compress) { + retval = ZSTD_compressStream(state->u.compress_stream, + &output, &input); + } else { + retval = ZSTD_decompressStream(state->u.decompress_stream, + &output, &input); + } + + state->input_so_far += input.pos; + state->output_so_far += output.pos; + + *out = (char *)output.dst + output.pos; + *out_len = output.size - output.pos; + *in = (char *)input.src + input.pos; + *in_len = input.size - input.pos; + + if (! state->compress && + tor_compress_is_compression_bomb(state->input_so_far, + state->output_so_far)) { + log_warn(LD_DIR, "Possible compression bomb; abandoning stream."); + return TOR_COMPRESS_ERROR; + } + + if (ZSTD_isError(retval)) { + log_warn(LD_GENERAL, "Zstandard %s didn't finish: %s.", + state->compress ? "compression" : "decompression", + ZSTD_getErrorName(retval)); + return TOR_COMPRESS_ERROR; + } + + if (state->compress && !finish) { + retval = ZSTD_flushStream(state->u.compress_stream, &output); + + *out = (char *)output.dst + output.pos; + *out_len = output.size - output.pos; + + if (ZSTD_isError(retval)) { + log_warn(LD_GENERAL, "Zstandard compression unable to flush: %s.", + ZSTD_getErrorName(retval)); + return TOR_COMPRESS_ERROR; + } + + if (retval > 0) + return TOR_COMPRESS_BUFFER_FULL; + } + + if (state->compress && finish) { + retval = ZSTD_endStream(state->u.compress_stream, &output); + + *out = (char *)output.dst + output.pos; + *out_len = output.size - output.pos; + + if (ZSTD_isError(retval)) { + log_warn(LD_GENERAL, "Zstandard compression unable to write " + "epilogue: %s.", + ZSTD_getErrorName(retval)); + return TOR_COMPRESS_ERROR; + } + + // endStream returns the number of bytes that is needed to write the + // epilogue. + if (retval > 0) + return TOR_COMPRESS_BUFFER_FULL; + } + + return finish ? TOR_COMPRESS_DONE : TOR_COMPRESS_OK; +#else // HAVE_ZSTD. + (void)state; + (void)out; + (void)out_len; + (void)in; + (void)in_len; + (void)finish; + + return TOR_COMPRESS_ERROR; +#endif // HAVE_ZSTD. +} + +/** Deallocate <b>state</b>. */ +void +tor_zstd_compress_free(tor_zstd_compress_state_t *state) +{ + if (state == NULL) + return; + + total_zstd_allocation -= state->allocation; + +#ifdef HAVE_ZSTD + if (state->compress) { + ZSTD_freeCStream(state->u.compress_stream); + } else { + ZSTD_freeDStream(state->u.decompress_stream); + } +#endif // HAVE_ZSTD. + + tor_free(state); +} + +/** Return the approximate number of bytes allocated for <b>state</b>. */ +size_t +tor_zstd_compress_state_size(const tor_zstd_compress_state_t *state) +{ + tor_assert(state != NULL); + return state->allocation; +} + +/** Return the approximate number of bytes allocated for all Zstandard + * states. */ +size_t +tor_zstd_get_total_allocation(void) +{ + return total_zstd_allocation; +} + diff --git a/src/common/compress_zstd.h b/src/common/compress_zstd.h new file mode 100644 index 0000000000..663cbdd2ff --- /dev/null +++ b/src/common/compress_zstd.h @@ -0,0 +1,51 @@ +/* Copyright (c) 2003, Roger Dingledine + * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson. + * Copyright (c) 2007-2017, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +/** + * \file compress_zstd.h + * \brief Header for compress_zstd.c + **/ + +#ifndef TOR_COMPRESS_ZSTD_H +#define TOR_COMPRESS_ZSTD_H + +int tor_zstd_method_supported(void); + +const char *tor_zstd_get_version_str(void); + +const char *tor_zstd_get_header_version_str(void); + +int tor_zstd_compress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method); + +int tor_zstd_uncompress(char **out, size_t *out_len, + const char *in, size_t in_len, + compress_method_t method, + int complete_only, + int protocol_warn_level); + +/** Internal state for an incremental Zstandard compression/decompression. */ +typedef struct tor_zstd_compress_state_t tor_zstd_compress_state_t; + +tor_zstd_compress_state_t * +tor_zstd_compress_new(int compress, + compress_method_t method, + compression_level_t compression_level); + +tor_compress_output_t +tor_zstd_compress_process(tor_zstd_compress_state_t *state, + char **out, size_t *out_len, + const char **in, size_t *in_len, + int finish); + +void tor_zstd_compress_free(tor_zstd_compress_state_t *state); + +size_t tor_zstd_compress_state_size(const tor_zstd_compress_state_t *state); + +size_t tor_zstd_get_total_allocation(void); + +#endif // TOR_COMPRESS_ZSTD_H. + diff --git a/src/common/include.am b/src/common/include.am index d6504c7466..e285ef5f86 100644 --- a/src/common/include.am +++ b/src/common/include.am @@ -105,11 +105,14 @@ src/common/src_common_libor_testing_a-log.$(OBJEXT) \ LIBOR_CRYPTO_A_SRC = \ src/common/aes.c \ + src/common/compress.c \ + src/common/compress_lzma.c \ + src/common/compress_zlib.c \ + src/common/compress_zstd.c \ src/common/crypto.c \ src/common/crypto_pwbox.c \ src/common/crypto_s2k.c \ src/common/crypto_format.c \ - src/common/torgzip.c \ src/common/tortls.c \ src/common/crypto_curve25519.c \ src/common/crypto_ed25519.c @@ -145,6 +148,10 @@ COMMONHEADERS = \ src/common/compat_openssl.h \ src/common/compat_threads.h \ src/common/compat_time.h \ + src/common/compress.h \ + src/common/compress_lzma.h \ + src/common/compress_zlib.h \ + src/common/compress_zstd.h \ src/common/confline.h \ src/common/container.h \ src/common/crypto.h \ @@ -163,7 +170,6 @@ COMMONHEADERS = \ src/common/storagedir.h \ src/common/testsupport.h \ src/common/timers.h \ - src/common/torgzip.h \ src/common/torint.h \ src/common/torlog.h \ src/common/tortls.h \ diff --git a/src/common/torgzip.h b/src/common/torgzip.h deleted file mode 100644 index 430b152fe4..0000000000 --- a/src/common/torgzip.h +++ /dev/null @@ -1,72 +0,0 @@ -/* Copyright (c) 2003, Roger Dingledine - * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson. - * Copyright (c) 2007-2017, The Tor Project, Inc. */ -/* See LICENSE for licensing information */ - -/** - * \file torgzip.h - * \brief Headers for torgzip.h - **/ - -#ifndef TOR_TORGZIP_H -#define TOR_TORGZIP_H - -/** Enumeration of what kind of compression to use. Only ZLIB_METHOD is - * guaranteed to be supported by the compress/uncompress functions here; - * GZIP_METHOD may be supported if we built against zlib version 1.2 or later - * and is_gzip_supported() returns true. */ -typedef enum { - NO_METHOD=0, GZIP_METHOD=1, ZLIB_METHOD=2, UNKNOWN_METHOD=3 -} compress_method_t; - -/** - * Enumeration to define tradeoffs between memory usage and compression level. - * HIGH_COMPRESSION saves the most bandwidth; LOW_COMPRESSION saves the most - * memory. - **/ -typedef enum { - HIGH_COMPRESSION, MEDIUM_COMPRESSION, LOW_COMPRESSION -} zlib_compression_level_t; - -int -tor_gzip_compress(char **out, size_t *out_len, - const char *in, size_t in_len, - compress_method_t method); -int -tor_gzip_uncompress(char **out, size_t *out_len, - const char *in, size_t in_len, - compress_method_t method, - int complete_only, - int protocol_warn_level); - -int is_gzip_supported(void); - -const char * -tor_zlib_get_version_str(void); - -const char * -tor_zlib_get_header_version_str(void); - -compress_method_t detect_compression_method(const char *in, size_t in_len); - -/** Return values from tor_zlib_process; see that function's documentation for - * details. */ -typedef enum { - TOR_ZLIB_OK, TOR_ZLIB_DONE, TOR_ZLIB_BUF_FULL, TOR_ZLIB_ERR -} tor_zlib_output_t; -/** Internal state for an incremental zlib compression/decompression. */ -typedef struct tor_zlib_state_t tor_zlib_state_t; -tor_zlib_state_t *tor_zlib_new(int compress, compress_method_t method, - zlib_compression_level_t level); - -tor_zlib_output_t tor_zlib_process(tor_zlib_state_t *state, - char **out, size_t *out_len, - const char **in, size_t *in_len, - int finish); -void tor_zlib_free(tor_zlib_state_t *state); - -size_t tor_zlib_state_size(const tor_zlib_state_t *state); -size_t tor_zlib_get_total_allocation(void); - -#endif - diff --git a/src/or/buffers.c b/src/or/buffers.c index e559f80a1e..4f22935d26 100644 --- a/src/or/buffers.c +++ b/src/or/buffers.c @@ -2088,13 +2088,13 @@ fetch_from_buf_line(buf_t *buf, char *data_out, size_t *data_len) } /** Compress on uncompress the <b>data_len</b> bytes in <b>data</b> using the - * zlib state <b>state</b>, appending the result to <b>buf</b>. If + * compression state <b>state</b>, appending the result to <b>buf</b>. If * <b>done</b> is true, flush the data in the state and finish the * compression/uncompression. Return -1 on failure, 0 on success. */ int -write_to_buf_zlib(buf_t *buf, tor_zlib_state_t *state, - const char *data, size_t data_len, - int done) +write_to_buf_compress(buf_t *buf, tor_compress_state_t *state, + const char *data, size_t data_len, + int done) { char *next; size_t old_avail, avail; @@ -2108,20 +2108,22 @@ write_to_buf_zlib(buf_t *buf, tor_zlib_state_t *state, } next = CHUNK_WRITE_PTR(buf->tail); avail = old_avail = CHUNK_REMAINING_CAPACITY(buf->tail); - switch (tor_zlib_process(state, &next, &avail, &data, &data_len, done)) { - case TOR_ZLIB_DONE: + switch (tor_compress_process(state, &next, &avail, + &data, &data_len, done)) { + case TOR_COMPRESS_DONE: over = 1; break; - case TOR_ZLIB_ERR: + case TOR_COMPRESS_ERROR: return -1; - case TOR_ZLIB_OK: + case TOR_COMPRESS_OK: if (data_len == 0) over = 1; break; - case TOR_ZLIB_BUF_FULL: + case TOR_COMPRESS_BUFFER_FULL: if (avail) { - /* Zlib says we need more room (ZLIB_BUF_FULL). Start a new chunk - * automatically, whether were going to or not. */ + /* The compression module says we need more room + * (TOR_COMPRESS_BUFFER_FULL). Start a new chunk automatically, + * whether were going to or not. */ need_new_chunk = 1; } break; diff --git a/src/or/buffers.h b/src/or/buffers.h index c6a5ffaad5..23b58a571a 100644 --- a/src/or/buffers.h +++ b/src/or/buffers.h @@ -36,8 +36,8 @@ int flush_buf(tor_socket_t s, buf_t *buf, size_t sz, size_t *buf_flushlen); int flush_buf_tls(tor_tls_t *tls, buf_t *buf, size_t sz, size_t *buf_flushlen); int write_to_buf(const char *string, size_t string_len, buf_t *buf); -int write_to_buf_zlib(buf_t *buf, tor_zlib_state_t *state, - const char *data, size_t data_len, int done); +int write_to_buf_compress(buf_t *buf, tor_compress_state_t *state, + const char *data, size_t data_len, int done); int move_buf_to_buf(buf_t *buf_out, buf_t *buf_in, size_t *buf_flushlen); int fetch_from_buf(char *string, size_t string_len, buf_t *buf); int fetch_var_cell_from_buf(buf_t *buf, var_cell_t **out, int linkproto); diff --git a/src/or/circuitlist.c b/src/or/circuitlist.c index 80bb7f69f3..365e5b2f89 100644 --- a/src/or/circuitlist.c +++ b/src/or/circuitlist.c @@ -1990,10 +1990,10 @@ single_conn_free_bytes(connection_t *conn) } if (conn->type == CONN_TYPE_DIR) { dir_connection_t *dir_conn = TO_DIR_CONN(conn); - if (dir_conn->zlib_state) { - result += tor_zlib_state_size(dir_conn->zlib_state); - tor_zlib_free(dir_conn->zlib_state); - dir_conn->zlib_state = NULL; + if (dir_conn->compress_state) { + result += tor_compress_state_size(dir_conn->compress_state); + tor_compress_free(dir_conn->compress_state); + dir_conn->compress_state = NULL; } } return result; diff --git a/src/or/config.c b/src/or/config.c index 809ff499fc..a73f397073 100644 --- a/src/or/config.c +++ b/src/or/config.c @@ -69,6 +69,7 @@ #include "circuitmux.h" #include "circuitmux_ewma.h" #include "circuitstats.h" +#include "compress.h" #include "config.h" #include "connection.h" #include "connection_edge.h" @@ -99,7 +100,6 @@ #include "statefile.h" #include "transports.h" #include "ext_orport.h" -#include "torgzip.h" #ifdef _WIN32 #include <shlobj.h> #endif @@ -4949,9 +4949,21 @@ options_init_from_torrc(int argc, char **argv) printf("OpenSSL \t\t%-15s\t\t%s\n", crypto_openssl_get_header_version_str(), crypto_openssl_get_version_str()); - printf("Zlib \t\t%-15s\t\t%s\n", - tor_zlib_get_header_version_str(), - tor_zlib_get_version_str()); + if (tor_compress_supports_method(ZLIB_METHOD)) { + printf("Zlib \t\t%-15s\t\t%s\n", + tor_compress_version_str(ZLIB_METHOD), + tor_compress_header_version_str(ZLIB_METHOD)); + } + if (tor_compress_supports_method(LZMA_METHOD)) { + printf("Liblzma \t\t%-15s\t\t%s\n", + tor_compress_version_str(LZMA_METHOD), + tor_compress_header_version_str(LZMA_METHOD)); + } + if (tor_compress_supports_method(ZSTD_METHOD)) { + printf("Libzstd \t\t%-15s\t\t%s\n", + tor_compress_version_str(ZSTD_METHOD), + tor_compress_header_version_str(ZSTD_METHOD)); + } //TODO: Hex versions? exit(0); } diff --git a/src/or/connection.c b/src/or/connection.c index 09e316d214..5fb2c53677 100644 --- a/src/or/connection.c +++ b/src/or/connection.c @@ -628,7 +628,7 @@ connection_free_(connection_t *conn) dir_connection_t *dir_conn = TO_DIR_CONN(conn); tor_free(dir_conn->requested_resource); - tor_zlib_free(dir_conn->zlib_state); + tor_compress_free(dir_conn->compress_state); if (dir_conn->spool) { SMARTLIST_FOREACH(dir_conn->spool, spooled_resource_t *, spooled, spooled_resource_free(spooled)); @@ -4060,9 +4060,9 @@ connection_write_to_buf_impl_,(const char *string, size_t len, if (zlib) { dir_connection_t *dir_conn = TO_DIR_CONN(conn); int done = zlib < 0; - CONN_LOG_PROTECT(conn, r = write_to_buf_zlib(conn->outbuf, - dir_conn->zlib_state, - string, len, done)); + CONN_LOG_PROTECT(conn, r = write_to_buf_compress(conn->outbuf, + dir_conn->compress_state, + string, len, done)); } else { CONN_LOG_PROTECT(conn, r = write_to_buf(string, len, conn->outbuf)); } diff --git a/src/or/connection.h b/src/or/connection.h index df6fc64709..36e45aef38 100644 --- a/src/or/connection.h +++ b/src/or/connection.h @@ -141,17 +141,17 @@ MOCK_DECL(void, connection_write_to_buf_impl_, /* DOCDOC connection_write_to_buf */ static void connection_write_to_buf(const char *string, size_t len, connection_t *conn); -/* DOCDOC connection_write_to_buf_zlib */ -static void connection_write_to_buf_zlib(const char *string, size_t len, - dir_connection_t *conn, int done); +/* DOCDOC connection_write_to_buf_compress */ +static void connection_write_to_buf_compress(const char *string, size_t len, + dir_connection_t *conn, int done); static inline void connection_write_to_buf(const char *string, size_t len, connection_t *conn) { connection_write_to_buf_impl_(string, len, conn, 0); } static inline void -connection_write_to_buf_zlib(const char *string, size_t len, - dir_connection_t *conn, int done) +connection_write_to_buf_compress(const char *string, size_t len, + dir_connection_t *conn, int done) { connection_write_to_buf_impl_(string, len, TO_CONN(conn), done ? -1 : 1); } diff --git a/src/or/directory.c b/src/or/directory.c index 1b999ee7c3..e0409e2021 100644 --- a/src/or/directory.c +++ b/src/or/directory.c @@ -2096,15 +2096,15 @@ connection_dir_client_reached_eof(dir_connection_t *conn) } /* Try declared compression first if we can. */ if (compression == GZIP_METHOD || compression == ZLIB_METHOD) - tor_gzip_uncompress(&new_body, &new_len, body, body_len, compression, - !allow_partial, LOG_PROTOCOL_WARN); + tor_uncompress(&new_body, &new_len, body, body_len, compression, + !allow_partial, LOG_PROTOCOL_WARN); /* Okay, if that didn't work, and we think that it was compressed * differently, try that. */ if (!new_body && (guessed == GZIP_METHOD || guessed == ZLIB_METHOD) && compression != guessed) - tor_gzip_uncompress(&new_body, &new_len, body, body_len, guessed, - !allow_partial, LOG_PROTOCOL_WARN); + tor_uncompress(&new_body, &new_len, body, body_len, guessed, + !allow_partial, LOG_PROTOCOL_WARN); /* If we're pretty sure that we have a compressed directory, and * we didn't manage to uncompress it, then warn and bail. */ if (!plausible && !new_body) { @@ -2845,7 +2845,7 @@ client_likes_consensus(networkstatus_t *v, const char *want_url) /** Return the compression level we should use for sending a compressed * response of size <b>n_bytes</b>. */ -STATIC zlib_compression_level_t +STATIC compression_level_t choose_compression_level(ssize_t n_bytes) { if (! have_been_under_memory_pressure()) { @@ -3178,7 +3178,8 @@ handle_get_current_consensus(dir_connection_t *conn, write_http_response_header(conn, -1, compressed, smartlist_len(conn->spool) == 1 ? lifetime : 0); if (! compressed) - conn->zlib_state = tor_zlib_new(0, ZLIB_METHOD, HIGH_COMPRESSION); + conn->compress_state = tor_compress_new(0, ZLIB_METHOD, + HIGH_COMPRESSION); /* Prime the connection with some data. */ const int initial_flush_result = connection_dirserv_flushed_some(conn); @@ -3276,11 +3277,11 @@ handle_get_status_vote(dir_connection_t *conn, const get_handler_args_t *args) if (smartlist_len(items)) { if (compressed) { - conn->zlib_state = tor_zlib_new(1, ZLIB_METHOD, - choose_compression_level(estimated_len)); + conn->compress_state = tor_compress_new(1, ZLIB_METHOD, + choose_compression_level(estimated_len)); SMARTLIST_FOREACH(items, const char *, c, - connection_write_to_buf_zlib(c, strlen(c), conn, 0)); - connection_write_to_buf_zlib("", 0, conn, 1); + connection_write_to_buf_compress(c, strlen(c), conn, 0)); + connection_write_to_buf_compress("", 0, conn, 1); } else { SMARTLIST_FOREACH(items, const char *, c, connection_write_to_buf(c, strlen(c), TO_CONN(conn))); @@ -3335,7 +3336,7 @@ handle_get_microdesc(dir_connection_t *conn, const get_handler_args_t *args) write_http_response_header(conn, -1, compressed, MICRODESC_CACHE_LIFETIME); if (compressed) - conn->zlib_state = tor_zlib_new(1, ZLIB_METHOD, + conn->compress_state = tor_compress_new(1, ZLIB_METHOD, choose_compression_level(size_guess)); const int initial_flush_result = connection_dirserv_flushed_some(conn); @@ -3428,7 +3429,7 @@ handle_get_descriptor(dir_connection_t *conn, const get_handler_args_t *args) } write_http_response_header(conn, -1, compressed, cache_lifetime); if (compressed) - conn->zlib_state = tor_zlib_new(1, ZLIB_METHOD, + conn->compress_state = tor_compress_new(1, ZLIB_METHOD, choose_compression_level(size_guess)); clear_spool = 0; /* Prime the connection with some data. */ @@ -3519,13 +3520,14 @@ handle_get_keys(dir_connection_t *conn, const get_handler_args_t *args) write_http_response_header(conn, compressed?-1:len, compressed, 60*60); if (compressed) { - conn->zlib_state = tor_zlib_new(1, ZLIB_METHOD, - choose_compression_level(len)); + conn->compress_state = tor_compress_new(1, ZLIB_METHOD, + choose_compression_level(len)); SMARTLIST_FOREACH(certs, authority_cert_t *, c, - connection_write_to_buf_zlib(c->cache_info.signed_descriptor_body, - c->cache_info.signed_descriptor_len, - conn, 0)); - connection_write_to_buf_zlib("", 0, conn, 1); + connection_write_to_buf_compress( + c->cache_info.signed_descriptor_body, + c->cache_info.signed_descriptor_len, + conn, 0)); + connection_write_to_buf_compress("", 0, conn, 1); } else { SMARTLIST_FOREACH(certs, authority_cert_t *, c, connection_write_to_buf(c->cache_info.signed_descriptor_body, diff --git a/src/or/directory.h b/src/or/directory.h index 0c5db3e070..4c52c24049 100644 --- a/src/or/directory.h +++ b/src/or/directory.h @@ -181,7 +181,7 @@ STATIC int handle_post_hs_descriptor(const char *url, const char *body); STATIC char* authdir_type_to_string(dirinfo_type_t auth); STATIC const char * dir_conn_purpose_to_string(int purpose); STATIC int should_use_directory_guards(const or_options_t *options); -STATIC zlib_compression_level_t choose_compression_level(ssize_t n_bytes); +STATIC compression_level_t choose_compression_level(ssize_t n_bytes); STATIC const smartlist_t *find_dl_schedule(download_status_t *dls, const or_options_t *options); STATIC void find_dl_min_and_max_delay(download_status_t *dls, diff --git a/src/or/dirserv.c b/src/or/dirserv.c index 70b0b22f25..e76fd932ca 100644 --- a/src/or/dirserv.c +++ b/src/or/dirserv.c @@ -1176,8 +1176,8 @@ new_cached_dir(char *s, time_t published) d->dir = s; d->dir_len = strlen(s); d->published = published; - if (tor_gzip_compress(&(d->dir_z), &(d->dir_z_len), d->dir, d->dir_len, - ZLIB_METHOD)) { + if (tor_compress(&(d->dir_z), &(d->dir_z_len), d->dir, d->dir_len, + ZLIB_METHOD)) { log_warn(LD_BUG, "Error compressing directory"); } return d; @@ -3497,8 +3497,8 @@ spooled_resource_flush_some(spooled_resource_t *spooled, /* Absent objects count as "done". */ return SRFS_DONE; } - if (conn->zlib_state) { - connection_write_to_buf_zlib((const char*)body, bodylen, conn, 0); + if (conn->compress_state) { + connection_write_to_buf_compress((const char*)body, bodylen, conn, 0); } else { connection_write_to_buf((const char*)body, bodylen, TO_CONN(conn)); } @@ -3523,9 +3523,10 @@ spooled_resource_flush_some(spooled_resource_t *spooled, if (BUG(remaining < 0)) return SRFS_ERR; ssize_t bytes = (ssize_t) MIN(DIRSERV_CACHED_DIR_CHUNK_SIZE, remaining); - if (conn->zlib_state) { - connection_write_to_buf_zlib(cached->dir_z + spooled->cached_dir_offset, - bytes, conn, 0); + if (conn->compress_state) { + connection_write_to_buf_compress( + cached->dir_z + spooled->cached_dir_offset, + bytes, conn, 0); } else { connection_write_to_buf(cached->dir_z + spooled->cached_dir_offset, bytes, TO_CONN(conn)); @@ -3788,12 +3789,12 @@ connection_dirserv_flushed_some(dir_connection_t *conn) /* If we get here, we're done. */ smartlist_free(conn->spool); conn->spool = NULL; - if (conn->zlib_state) { - /* Flush the zlib state: there could be more bytes pending in there, and - * we don't want to omit bytes. */ - connection_write_to_buf_zlib("", 0, conn, 1); - tor_zlib_free(conn->zlib_state); - conn->zlib_state = NULL; + if (conn->compress_state) { + /* Flush the compression state: there could be more bytes pending in there, + * and we don't want to omit bytes. */ + connection_write_to_buf_compress("", 0, conn, 1); + tor_compress_free(conn->compress_state); + conn->compress_state = NULL; } return 0; } diff --git a/src/or/include.am b/src/or/include.am index dd27bc8411..483ea2f2aa 100644 --- a/src/or/include.am +++ b/src/or/include.am @@ -121,7 +121,8 @@ src_or_tor_LDADD = src/or/libtor.a src/common/libor.a src/common/libor-ctime.a \ src/common/libor-crypto.a $(LIBKECCAK_TINY) $(LIBDONNA) \ src/common/libor-event.a src/trunnel/libor-trunnel.a \ @TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ @TOR_LIBEVENT_LIBS@ @TOR_OPENSSL_LIBS@ \ - @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ @TOR_SYSTEMD_LIBS@ + @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ @TOR_SYSTEMD_LIBS@ \ + @TOR_LZMA_LIBS@ @TOR_ZSTD_LIBS@ if COVERAGE_ENABLED src_or_tor_cov_SOURCES = src/or/tor_main.c @@ -133,7 +134,8 @@ src_or_tor_cov_LDADD = src/or/libtor-testing.a src/common/libor-testing.a \ src/common/libor-crypto-testing.a $(LIBKECCAK_TINY) $(LIBDONNA) \ src/common/libor-event-testing.a src/trunnel/libor-trunnel-testing.a \ @TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ @TOR_LIBEVENT_LIBS@ @TOR_OPENSSL_LIBS@ \ - @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ @TOR_SYSTEMD_LIBS@ + @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ @TOR_SYSTEMD_LIBS@ \ + @TOR_LZMA_LIBS@ @TOR_ZSTD_LIBS@ endif ORHEADERS = \ diff --git a/src/or/main.c b/src/or/main.c index 4505879adc..1ba6554d3e 100644 --- a/src/or/main.c +++ b/src/or/main.c @@ -58,6 +58,7 @@ #include "circuitlist.h" #include "circuituse.h" #include "command.h" +#include "compress.h" #include "config.h" #include "confparse.h" #include "connection.h" @@ -2998,11 +2999,16 @@ tor_init(int argc, char *argv[]) const char *version = get_version(); log_notice(LD_GENERAL, "Tor %s running on %s with Libevent %s, " - "OpenSSL %s and Zlib %s.", version, + "OpenSSL %s, Zlib %s, Liblzma %s, and Libzstd %s.", version, get_uname(), tor_libevent_get_version_str(), crypto_openssl_get_version_str(), - tor_zlib_get_version_str()); + tor_compress_supports_method(ZLIB_METHOD) ? + tor_compress_version_str(ZLIB_METHOD) : "N/A", + tor_compress_supports_method(LZMA_METHOD) ? + tor_compress_version_str(LZMA_METHOD) : "N/A", + tor_compress_supports_method(ZSTD_METHOD) ? + tor_compress_version_str(ZSTD_METHOD) : "N/A"); log_notice(LD_GENERAL, "Tor can't help you if you use it wrong! " "Learn how to be safe at " diff --git a/src/or/or.h b/src/or/or.h index a7b3a66561..9e3e409af9 100644 --- a/src/or/or.h +++ b/src/or/or.h @@ -71,7 +71,7 @@ #include "tortls.h" #include "torlog.h" #include "container.h" -#include "torgzip.h" +#include "compress.h" #include "address.h" #include "compat_libevent.h" #include "ht.h" @@ -1773,8 +1773,8 @@ typedef struct dir_connection_t { /** List of spooled_resource_t for objects that we're spooling. We use * it from back to front. */ smartlist_t *spool; - /** The zlib object doing on-the-fly compression for spooled data. */ - tor_zlib_state_t *zlib_state; + /** The compression object doing on-the-fly compression for spooled data. */ + tor_compress_state_t *compress_state; /** What rendezvous service are we querying for? */ rend_data_t *rend_data; diff --git a/src/or/relay.c b/src/or/relay.c index 5139036327..8524080939 100644 --- a/src/or/relay.c +++ b/src/or/relay.c @@ -54,6 +54,7 @@ #include "circuitbuild.h" #include "circuitlist.h" #include "circuituse.h" +#include "compress.h" #include "config.h" #include "connection.h" #include "connection_edge.h" @@ -2453,7 +2454,7 @@ cell_queues_check_size(void) { size_t alloc = cell_queues_get_total_allocation(); alloc += buf_get_total_allocation(); - alloc += tor_zlib_get_total_allocation(); + alloc += tor_compress_get_total_allocation(); const size_t rend_cache_total = rend_cache_get_total_allocation(); alloc += rend_cache_total; if (alloc >= get_options()->MaxMemInQueues_low_threshold) { diff --git a/src/test/fuzz/include.am b/src/test/fuzz/include.am index 1b608c6885..500377f6d7 100644 --- a/src/test/fuzz/include.am +++ b/src/test/fuzz/include.am @@ -18,7 +18,9 @@ FUZZING_LIBS = \ @TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ \ @TOR_LIBEVENT_LIBS@ \ @TOR_OPENSSL_LIBS@ @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ \ - @TOR_SYSTEMD_LIBS@ + @TOR_SYSTEMD_LIBS@ \ + @TOR_LZMA_LIBS@ \ + @TOR_ZSTD_LIBS@ oss-fuzz-prereqs: \ src/or/libtor-testing.a \ diff --git a/src/test/include.am b/src/test/include.am index 1b16d0f1a2..653adad300 100644 --- a/src/test/include.am +++ b/src/test/include.am @@ -180,7 +180,8 @@ src_test_test_switch_id_LDFLAGS = @TOR_LDFLAGS_zlib@ src_test_test_switch_id_LDADD = \ src/common/libor-testing.a \ src/common/libor-ctime-testing.a \ - @TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ + @TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ \ + @TOR_LZMA_LIBS@ @TOR_ZSTD_LIBS@ src_test_test_LDFLAGS = @TOR_LDFLAGS_zlib@ @TOR_LDFLAGS_openssl@ \ @TOR_LDFLAGS_libevent@ @@ -194,7 +195,7 @@ src_test_test_LDADD = src/or/libtor-testing.a \ src/trunnel/libor-trunnel-testing.a \ @TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ @TOR_LIBEVENT_LIBS@ \ @TOR_OPENSSL_LIBS@ @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ \ - @TOR_SYSTEMD_LIBS@ + @TOR_SYSTEMD_LIBS@ @TOR_LZMA_LIBS@ @TOR_ZSTD_LIBS@ src_test_test_slow_CPPFLAGS = $(src_test_test_CPPFLAGS) src_test_test_slow_CFLAGS = $(src_test_test_CFLAGS) @@ -217,7 +218,7 @@ src_test_bench_LDADD = src/or/libtor.a src/common/libor.a \ src/common/libor-event.a src/trunnel/libor-trunnel.a \ @TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ @TOR_LIBEVENT_LIBS@ \ @TOR_OPENSSL_LIBS@ @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ \ - @TOR_SYSTEMD_LIBS@ + @TOR_SYSTEMD_LIBS@ @TOR_LZMA_LIBS@ @TOR_ZSTD_LIBS@ src_test_test_workqueue_LDFLAGS = @TOR_LDFLAGS_zlib@ @TOR_LDFLAGS_openssl@ \ @TOR_LDFLAGS_libevent@ @@ -227,7 +228,8 @@ src_test_test_workqueue_LDADD = src/or/libtor-testing.a \ src/common/libor-crypto-testing.a $(LIBKECCAK_TINY) $(LIBDONNA) \ src/common/libor-event-testing.a \ @TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ @TOR_LIBEVENT_LIBS@ \ - @TOR_OPENSSL_LIBS@ @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ + @TOR_OPENSSL_LIBS@ @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ \ + @TOR_LZMA_LIBS@ @TOR_ZSTD_LIBS@ src_test_test_timers_CPPFLAGS = $(src_test_test_CPPFLAGS) src_test_test_timers_CFLAGS = $(src_test_test_CFLAGS) @@ -237,7 +239,8 @@ src_test_test_timers_LDADD = \ src/common/libor-event-testing.a \ src/common/libor-crypto-testing.a $(LIBKECCAK_TINY) $(LIBDONNA) \ @TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ @TOR_LIBEVENT_LIBS@ \ - @TOR_OPENSSL_LIBS@ @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ + @TOR_OPENSSL_LIBS@ @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ \ + @TOR_LZMA_LIBS@ src_test_test_timers_LDFLAGS = $(src_test_test_LDFLAGS) noinst_HEADERS+= \ @@ -262,7 +265,8 @@ src_test_test_ntor_cl_LDADD = src/or/libtor.a src/common/libor.a \ src/common/libor-ctime.a \ src/common/libor-crypto.a $(LIBKECCAK_TINY) $(LIBDONNA) \ @TOR_ZLIB_LIBS@ @TOR_LIB_MATH@ \ - @TOR_OPENSSL_LIBS@ @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ + @TOR_OPENSSL_LIBS@ @TOR_LIB_WS32@ @TOR_LIB_GDI@ @CURVE25519_LIBS@ \ + @TOR_LZMA_LIBS@ src_test_test_ntor_cl_AM_CPPFLAGS = \ -I"$(top_srcdir)/src/or" diff --git a/src/test/test.c b/src/test/test.c index 90e3d02f93..4d2cf1536b 100644 --- a/src/test/test.c +++ b/src/test/test.c @@ -44,13 +44,13 @@ double fabs(double x); #include "buffers.h" #include "circuitlist.h" #include "circuitstats.h" +#include "compress.h" #include "config.h" #include "connection_edge.h" #include "geoip.h" #include "rendcommon.h" #include "rendcache.h" #include "test.h" -#include "torgzip.h" #include "main.h" #include "memarea.h" #include "onion.h" diff --git a/src/test/test_buffers.c b/src/test/test_buffers.c index d14165c45c..43582d1b7a 100644 --- a/src/test/test_buffers.c +++ b/src/test/test_buffers.c @@ -584,22 +584,26 @@ test_buffers_zlib_impl(int finalize_with_nil) char *contents = NULL; char *expanded = NULL; buf_t *buf = NULL; - tor_zlib_state_t *zlib_state = NULL; + tor_compress_state_t *compress_state = NULL; size_t out_len, in_len; int done; buf = buf_new_with_capacity(128); /* will round up */ - zlib_state = tor_zlib_new(1, ZLIB_METHOD, HIGH_COMPRESSION); + compress_state = tor_compress_new(1, ZLIB_METHOD, HIGH_COMPRESSION); msg = tor_malloc(512); crypto_rand(msg, 512); - tt_int_op(write_to_buf_zlib(buf, zlib_state, msg, 128, 0), OP_EQ, 0); - tt_int_op(write_to_buf_zlib(buf, zlib_state, msg+128, 128, 0), OP_EQ, 0); - tt_int_op(write_to_buf_zlib(buf, zlib_state, msg+256, 256, 0), OP_EQ, 0); + tt_int_op(write_to_buf_compress(buf, compress_state, + msg, 128, 0), OP_EQ, 0); + tt_int_op(write_to_buf_compress(buf, compress_state, + msg+128, 128, 0), OP_EQ, 0); + tt_int_op(write_to_buf_compress(buf, compress_state, + msg+256, 256, 0), OP_EQ, 0); done = !finalize_with_nil; - tt_int_op(write_to_buf_zlib(buf, zlib_state, "all done", 9, done), OP_EQ, 0); + tt_int_op(write_to_buf_compress(buf, compress_state, + "all done", 9, done), OP_EQ, 0); if (finalize_with_nil) { - tt_int_op(write_to_buf_zlib(buf, zlib_state, "", 0, 1), OP_EQ, 0); + tt_int_op(write_to_buf_compress(buf, compress_state, "", 0, 1), OP_EQ, 0); } in_len = buf_datalen(buf); @@ -607,10 +611,10 @@ test_buffers_zlib_impl(int finalize_with_nil) tt_int_op(fetch_from_buf(contents, in_len, buf), OP_EQ, 0); - tt_int_op(0, OP_EQ, tor_gzip_uncompress(&expanded, &out_len, - contents, in_len, - ZLIB_METHOD, 1, - LOG_WARN)); + tt_int_op(0, OP_EQ, tor_uncompress(&expanded, &out_len, + contents, in_len, + ZLIB_METHOD, 1, + LOG_WARN)); tt_int_op(out_len, OP_GE, 128); tt_mem_op(msg, OP_EQ, expanded, 128); @@ -621,7 +625,7 @@ test_buffers_zlib_impl(int finalize_with_nil) done: buf_free(buf); - tor_zlib_free(zlib_state); + tor_compress_free(compress_state); tor_free(contents); tor_free(expanded); tor_free(msg); @@ -647,7 +651,7 @@ test_buffers_zlib_fin_at_chunk_end(void *arg) char *contents = NULL; char *expanded = NULL; buf_t *buf = NULL; - tor_zlib_state_t *zlib_state = NULL; + tor_compress_state_t *compress_state = NULL; size_t out_len, in_len; size_t sz, headerjunk; (void) arg; @@ -666,8 +670,8 @@ test_buffers_zlib_fin_at_chunk_end(void *arg) tt_uint_op(buf->head->datalen, OP_EQ, headerjunk); tt_uint_op(buf_datalen(buf), OP_EQ, headerjunk); /* Write an empty string, with finalization on. */ - zlib_state = tor_zlib_new(1, ZLIB_METHOD, HIGH_COMPRESSION); - tt_int_op(write_to_buf_zlib(buf, zlib_state, "", 0, 1), OP_EQ, 0); + compress_state = tor_compress_new(1, ZLIB_METHOD, HIGH_COMPRESSION); + tt_int_op(write_to_buf_compress(buf, compress_state, "", 0, 1), OP_EQ, 0); in_len = buf_datalen(buf); contents = tor_malloc(in_len); @@ -676,17 +680,18 @@ test_buffers_zlib_fin_at_chunk_end(void *arg) tt_uint_op(in_len, OP_GT, headerjunk); - tt_int_op(0, OP_EQ, tor_gzip_uncompress(&expanded, &out_len, - contents + headerjunk, in_len - headerjunk, - ZLIB_METHOD, 1, - LOG_WARN)); + tt_int_op(0, OP_EQ, tor_uncompress(&expanded, &out_len, + contents + headerjunk, + in_len - headerjunk, + ZLIB_METHOD, 1, + LOG_WARN)); tt_int_op(out_len, OP_EQ, 0); tt_assert(expanded); done: buf_free(buf); - tor_zlib_free(zlib_state); + tor_compress_free(compress_state); tor_free(contents); tor_free(expanded); tor_free(msg); diff --git a/src/test/test_dir_handle_get.c b/src/test/test_dir_handle_get.c index cfda314693..392fa4dde0 100644 --- a/src/test/test_dir_handle_get.c +++ b/src/test/test_dir_handle_get.c @@ -14,6 +14,7 @@ #include "connection.h" #include "directory.h" #include "test.h" +#include "compress.h" #include "connection.h" #include "rendcommon.h" #include "rendcache.h" @@ -28,7 +29,6 @@ #include "networkstatus.h" #include "geoip.h" #include "dirserv.h" -#include "torgzip.h" #include "dirvote.h" #include "log_test_helpers.h" @@ -1832,8 +1832,8 @@ test_dir_handle_get_status_vote_current_consensus_ns(void* data) comp_body_used); tt_int_op(ZLIB_METHOD, OP_EQ, compression); - tor_gzip_uncompress(&body, &body_used, comp_body, comp_body_used, - compression, 0, LOG_PROTOCOL_WARN); + tor_uncompress(&body, &body_used, comp_body, comp_body_used, + compression, 0, LOG_PROTOCOL_WARN); tt_str_op(NETWORK_STATUS, OP_EQ, body); tt_int_op(strlen(NETWORK_STATUS), OP_EQ, body_used); diff --git a/src/test/test_util.c b/src/test/test_util.c index 203f9dd1c4..1e33de82ae 100644 --- a/src/test/test_util.c +++ b/src/test/test_util.c @@ -2249,46 +2249,49 @@ test_util_gzip(void *arg) char *buf1=NULL, *buf2=NULL, *buf3=NULL, *cp1, *cp2; const char *ccp2; size_t len1, len2; - tor_zlib_state_t *state = NULL; + tor_compress_state_t *state = NULL; (void)arg; + tt_assert(tor_compress_supports_method(GZIP_METHOD)); + tt_assert(tor_compress_supports_method(ZLIB_METHOD)); + buf1 = tor_strdup("AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAZAAAAAAAAAAAAAAAAAAAZ"); tt_assert(detect_compression_method(buf1, strlen(buf1)) == UNKNOWN_METHOD); - tt_assert(!tor_gzip_compress(&buf2, &len1, buf1, strlen(buf1)+1, - GZIP_METHOD)); - tt_assert(buf2); - tt_assert(len1 < strlen(buf1)); - tt_assert(detect_compression_method(buf2, len1) == GZIP_METHOD); + tt_assert(!tor_compress(&buf2, &len1, buf1, strlen(buf1)+1, + GZIP_METHOD)); + tt_assert(buf2 != NULL); + tt_int_op(len1, OP_LT, strlen(buf1)); + tt_int_op(detect_compression_method(buf2, len1), OP_EQ, GZIP_METHOD); - tt_assert(!tor_gzip_uncompress(&buf3, &len2, buf2, len1, - GZIP_METHOD, 1, LOG_INFO)); - tt_assert(buf3); - tt_int_op(strlen(buf1) + 1,OP_EQ, len2); - tt_str_op(buf1,OP_EQ, buf3); + tt_assert(!tor_uncompress(&buf3, &len2, buf2, len1, + GZIP_METHOD, 1, LOG_INFO)); + tt_assert(buf3 != NULL); + tt_int_op(strlen(buf1) + 1, OP_EQ, len2); + tt_str_op(buf1, OP_EQ, buf3); tor_free(buf2); tor_free(buf3); - tt_assert(!tor_gzip_compress(&buf2, &len1, buf1, strlen(buf1)+1, - ZLIB_METHOD)); + tt_assert(!tor_compress(&buf2, &len1, buf1, strlen(buf1)+1, + ZLIB_METHOD)); tt_assert(buf2); - tt_assert(detect_compression_method(buf2, len1) == ZLIB_METHOD); + tt_int_op(detect_compression_method(buf2, len1), OP_EQ, ZLIB_METHOD); - tt_assert(!tor_gzip_uncompress(&buf3, &len2, buf2, len1, - ZLIB_METHOD, 1, LOG_INFO)); - tt_assert(buf3); - tt_int_op(strlen(buf1) + 1,OP_EQ, len2); - tt_str_op(buf1,OP_EQ, buf3); + tt_assert(!tor_uncompress(&buf3, &len2, buf2, len1, + ZLIB_METHOD, 1, LOG_INFO)); + tt_assert(buf3 != NULL); + tt_int_op(strlen(buf1) + 1, OP_EQ, len2); + tt_str_op(buf1, OP_EQ, buf3); /* Check whether we can uncompress concatenated, compressed strings. */ tor_free(buf3); buf2 = tor_reallocarray(buf2, len1, 2); memcpy(buf2+len1, buf2, len1); - tt_assert(!tor_gzip_uncompress(&buf3, &len2, buf2, len1*2, - ZLIB_METHOD, 1, LOG_INFO)); - tt_int_op((strlen(buf1)+1)*2,OP_EQ, len2); - tt_mem_op(buf3,OP_EQ, + tt_assert(!tor_uncompress(&buf3, &len2, buf2, len1*2, + ZLIB_METHOD, 1, LOG_INFO)); + tt_int_op((strlen(buf1)+1)*2, OP_EQ, len2); + tt_mem_op(buf3, OP_EQ, "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAZAAAAAAAAAAAAAAAAAAAZ\0" "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAZAAAAAAAAAAAAAAAAAAAZ\0", (strlen(buf1)+1)*2); @@ -2300,53 +2303,53 @@ test_util_gzip(void *arg) /* Check whether we can uncompress partial strings. */ buf1 = tor_strdup("String with low redundancy that won't be compressed much."); - tt_assert(!tor_gzip_compress(&buf2, &len1, buf1, strlen(buf1)+1, - ZLIB_METHOD)); - tt_assert(len1>16); + tt_assert(!tor_compress(&buf2, &len1, buf1, strlen(buf1)+1, + ZLIB_METHOD)); + tt_int_op(len1, OP_GT, 16); /* when we allow an incomplete string, we should succeed.*/ - tt_assert(!tor_gzip_uncompress(&buf3, &len2, buf2, len1-16, - ZLIB_METHOD, 0, LOG_INFO)); + tt_assert(!tor_uncompress(&buf3, &len2, buf2, len1-16, + ZLIB_METHOD, 0, LOG_INFO)); tt_assert(len2 > 5); buf3[len2]='\0'; tt_assert(!strcmpstart(buf1, buf3)); /* when we demand a complete string, this must fail. */ tor_free(buf3); - tt_assert(tor_gzip_uncompress(&buf3, &len2, buf2, len1-16, - ZLIB_METHOD, 1, LOG_INFO)); - tt_assert(!buf3); + tt_assert(tor_uncompress(&buf3, &len2, buf2, len1-16, + ZLIB_METHOD, 1, LOG_INFO)); + tt_assert(buf3 == NULL); /* Now, try streaming compression. */ tor_free(buf1); tor_free(buf2); tor_free(buf3); - state = tor_zlib_new(1, ZLIB_METHOD, HIGH_COMPRESSION); + state = tor_compress_new(1, ZLIB_METHOD, HIGH_COMPRESSION); tt_assert(state); cp1 = buf1 = tor_malloc(1024); len1 = 1024; ccp2 = "ABCDEFGHIJABCDEFGHIJ"; len2 = 21; - tt_assert(tor_zlib_process(state, &cp1, &len1, &ccp2, &len2, 0) - == TOR_ZLIB_OK); - tt_int_op(0,OP_EQ, len2); /* Make sure we compressed it all. */ + tt_int_op(tor_compress_process(state, &cp1, &len1, &ccp2, &len2, 0), + OP_EQ, TOR_COMPRESS_OK); + tt_int_op(0, OP_EQ, len2); /* Make sure we compressed it all. */ tt_assert(cp1 > buf1); len2 = 0; cp2 = cp1; - tt_assert(tor_zlib_process(state, &cp1, &len1, &ccp2, &len2, 1) - == TOR_ZLIB_DONE); - tt_int_op(0,OP_EQ, len2); + tt_int_op(tor_compress_process(state, &cp1, &len1, &ccp2, &len2, 1), + OP_EQ, TOR_COMPRESS_DONE); + tt_int_op(0, OP_EQ, len2); tt_assert(cp1 > cp2); /* Make sure we really added something. */ - tt_assert(!tor_gzip_uncompress(&buf3, &len2, buf1, 1024-len1, - ZLIB_METHOD, 1, LOG_WARN)); + tt_assert(!tor_uncompress(&buf3, &len2, buf1, 1024-len1, + ZLIB_METHOD, 1, LOG_WARN)); /* Make sure it compressed right. */ tt_str_op(buf3, OP_EQ, "ABCDEFGHIJABCDEFGHIJ"); - tt_int_op(21,OP_EQ, len2); + tt_int_op(21, OP_EQ, len2); done: if (state) - tor_zlib_free(state); + tor_compress_free(state); tor_free(buf2); tor_free(buf3); tor_free(buf1); @@ -2364,13 +2367,13 @@ test_util_gzip_compression_bomb(void *arg) char *one_mb = tor_malloc_zero(one_million); char *result = NULL; size_t result_len = 0; - tor_zlib_state_t *state = NULL; + tor_compress_state_t *state = NULL; /* Make sure we can't produce a compression bomb */ setup_full_capture_of_logs(LOG_WARN); - tt_int_op(-1, OP_EQ, tor_gzip_compress(&result, &result_len, - one_mb, one_million, - ZLIB_METHOD)); + tt_int_op(-1, OP_EQ, tor_compress(&result, &result_len, + one_mb, one_million, + ZLIB_METHOD)); expect_single_log_msg_containing( "We compressed something and got an insanely high " "compression factor; other Tors would think this " @@ -2381,27 +2384,251 @@ test_util_gzip_compression_bomb(void *arg) const char compression_bomb[1039] = { 0x78, 0xDA, 0xED, 0xC1, 0x31, 0x01, 0x00, 0x00, 0x00, 0xC2, 0xA0, 0xF5, 0x4F, 0x6D, 0x08, 0x5F, 0xA0 /* .... */ }; - tt_int_op(-1, OP_EQ, tor_gzip_uncompress(&result, &result_len, - compression_bomb, 1039, - ZLIB_METHOD, 0, LOG_WARN)); + tt_int_op(-1, OP_EQ, tor_uncompress(&result, &result_len, + compression_bomb, 1039, + ZLIB_METHOD, 0, LOG_WARN)); /* Now try streaming that. */ - state = tor_zlib_new(0, ZLIB_METHOD, HIGH_COMPRESSION); - tor_zlib_output_t r; + state = tor_compress_new(0, ZLIB_METHOD, HIGH_COMPRESSION); + tor_compress_output_t r; const char *inp = compression_bomb; size_t inlen = 1039; do { char *outp = one_mb; size_t outleft = 4096; /* small on purpose */ - r = tor_zlib_process(state, &outp, &outleft, &inp, &inlen, 0); + r = tor_compress_process(state, &outp, &outleft, &inp, &inlen, 0); tt_int_op(inlen, OP_NE, 0); - } while (r == TOR_ZLIB_BUF_FULL); + } while (r == TOR_COMPRESS_BUFFER_FULL); - tt_int_op(r, OP_EQ, TOR_ZLIB_ERR); + tt_int_op(r, OP_EQ, TOR_COMPRESS_ERROR); done: tor_free(one_mb); - tor_zlib_free(state); + tor_compress_free(state); +} + +static void +test_util_lzma(void *arg) +{ +#ifdef HAVE_LZMA + char *buf1=NULL, *buf2=NULL, *buf3=NULL, *cp1, *cp2; + const char *ccp2; + size_t len1, len2; + tor_compress_state_t *state = NULL; + + (void)arg; + tt_assert(tor_compress_supports_method(LZMA_METHOD)); + + buf1 = tor_strdup("AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAZAAAAAAAAAAAAAAAAAAAZ"); + tt_assert(detect_compression_method(buf1, strlen(buf1)) == UNKNOWN_METHOD); + + tt_assert(!tor_compress(&buf2, &len1, buf1, strlen(buf1)+1, + LZMA_METHOD)); + tt_assert(buf2 != NULL); + tt_int_op(len1, OP_LT, strlen(buf1)); + tt_int_op(detect_compression_method(buf2, len1), OP_EQ, LZMA_METHOD); + + tt_assert(!tor_uncompress(&buf3, &len2, buf2, len1, + LZMA_METHOD, 1, LOG_INFO)); + tt_assert(buf3 != NULL); + tt_int_op(strlen(buf1) + 1, OP_EQ, len2); + tt_str_op(buf1, OP_EQ, buf3); + + tor_free(buf1); + tor_free(buf2); + tor_free(buf3); + +#if 0 + /* Check whether we can uncompress concatenated, compressed strings. */ + tor_free(buf3); + buf2 = tor_reallocarray(buf2, len1, 2); + memcpy(buf2+len1, buf2, len1); + tt_assert(!tor_uncompress(&buf3, &len2, buf2, len1*2, + LZMA_METHOD, 1, LOG_INFO)); + tt_int_op((strlen(buf1)+1)*2, OP_EQ, len2); + tt_mem_op(buf3, OP_EQ, + "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAZAAAAAAAAAAAAAAAAAAAZ\0" + "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAZAAAAAAAAAAAAAAAAAAAZ\0", + (strlen(buf1)+1)*2); + + tor_free(buf1); + tor_free(buf2); + tor_free(buf3); + + /* Check whether we can uncompress partial strings. */ + buf1 = + tor_strdup("String with low redundancy that won't be compressed much."); + tt_assert(!tor_compress(&buf2, &len1, buf1, strlen(buf1)+1, + LZMA_METHOD)); + tt_int_op(len1, OP_GT, 16); + /* when we allow an incomplete string, we should succeed.*/ + tt_assert(!tor_uncompress(&buf3, &len2, buf2, len1-16, + LZMA_METHOD, 0, LOG_INFO)); + tt_assert(len2 > 5); + buf3[len2]='\0'; + tt_assert(!strcmpstart(buf1, buf3)); + + /* when we demand a complete string, this must fail. */ + tor_free(buf3); + tt_assert(tor_uncompress(&buf3, &len2, buf2, len1-16, + LZMA_METHOD, 1, LOG_INFO)); + tt_assert(buf3 == NULL); + + tor_free(buf1); + tor_free(buf2); + tor_free(buf3); +#endif + + /* Now, try streaming compression. */ + state = tor_compress_new(1, LZMA_METHOD, HIGH_COMPRESSION); + tt_assert(state); + cp1 = buf1 = tor_malloc(1024); + len1 = 1024; + ccp2 = "ABCDEFGHIJABCDEFGHIJ"; + len2 = 21; + tt_int_op(tor_compress_process(state, &cp1, &len1, &ccp2, &len2, 0), + OP_EQ, TOR_COMPRESS_OK); + tt_int_op(0, OP_EQ, len2); /* Make sure we compressed it all. */ + tt_assert(cp1 > buf1); + + len2 = 0; + cp2 = cp1; + tt_int_op(tor_compress_process(state, &cp1, &len1, &ccp2, &len2, 1), + OP_EQ, TOR_COMPRESS_DONE); + tt_int_op(0, OP_EQ, len2); + tt_assert(cp1 > cp2); /* Make sure we really added something. */ + + tt_assert(!tor_uncompress(&buf3, &len2, buf1, 1024-len1, + LZMA_METHOD, 1, LOG_WARN)); + /* Make sure it compressed right. */ + tt_str_op(buf3, OP_EQ, "ABCDEFGHIJABCDEFGHIJ"); + tt_int_op(21, OP_EQ, len2); + + done: + if (state) + tor_compress_free(state); + tor_free(buf2); + tor_free(buf3); + tor_free(buf1); +#else + (void)arg; + tt_assert(! tor_compress_supports_method(LZMA_METHOD)); + + done: + ; +#endif // HAVE_LZMA. +} + +static void +test_util_zstd(void *arg) +{ +#ifdef HAVE_ZSTD + char *buf1=NULL, *buf2=NULL, *buf3=NULL, *cp1, *cp2; + const char *ccp2; + size_t len1, len2; + tor_compress_state_t *state = NULL; + + (void)arg; + tt_assert(tor_compress_supports_method(ZSTD_METHOD)); + + buf1 = tor_strdup("AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAZAAAAAAAAAAAAAAAAAAAZ"); + tt_assert(detect_compression_method(buf1, strlen(buf1)) == UNKNOWN_METHOD); + + tt_assert(!tor_compress(&buf2, &len1, buf1, strlen(buf1)+1, + ZSTD_METHOD)); + tt_assert(buf2 != NULL); + tt_int_op(len1, OP_LT, strlen(buf1)); + tt_int_op(detect_compression_method(buf2, len1), OP_EQ, ZSTD_METHOD); + + tt_assert(!tor_uncompress(&buf3, &len2, buf2, len1, + ZSTD_METHOD, 1, LOG_INFO)); + tt_assert(buf3 != NULL); + tt_int_op(strlen(buf1) + 1, OP_EQ, len2); + tt_str_op(buf1, OP_EQ, buf3); + + tor_free(buf1); + tor_free(buf2); + tor_free(buf3); + +#if 0 + /* Check whether we can uncompress concatenated, compressed strings. */ + tor_free(buf3); + buf2 = tor_reallocarray(buf2, len1, 2); + memcpy(buf2+len1, buf2, len1); + tt_assert(!tor_uncompress(&buf3, &len2, buf2, len1*2, + ZSTD_METHOD, 1, LOG_INFO)); + tt_int_op((strlen(buf1)+1)*2, OP_EQ, len2); + tt_mem_op(buf3, OP_EQ, + "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAZAAAAAAAAAAAAAAAAAAAZ\0" + "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAZAAAAAAAAAAAAAAAAAAAZ\0", + (strlen(buf1)+1)*2); + + tor_free(buf1); + tor_free(buf2); + tor_free(buf3); + + /* Check whether we can uncompress partial strings. */ + buf1 = + tor_strdup("String with low redundancy that won't be compressed much."); + tt_assert(!tor_compress(&buf2, &len1, buf1, strlen(buf1)+1, + ZSTD_METHOD)); + tt_int_op(len1, OP_GT, 16); + /* when we allow an incomplete string, we should succeed.*/ + tt_assert(!tor_uncompress(&buf3, &len2, buf2, len1-16, + ZSTD_METHOD, 0, LOG_INFO)); + tt_assert(len2 > 5); + buf3[len2]='\0'; + tt_assert(!strcmpstart(buf1, buf3)); + + /* when we demand a complete string, this must fail. */ + tor_free(buf3); + tt_assert(tor_uncompress(&buf3, &len2, buf2, len1-16, + ZSTD_METHOD, 1, LOG_INFO)); + tt_assert(buf3 == NULL); + + tor_free(buf1); + tor_free(buf2); + tor_free(buf3); +#endif + + /* Now, try streaming compression. */ + state = tor_compress_new(1, ZSTD_METHOD, HIGH_COMPRESSION); + tt_assert(state); + cp1 = buf1 = tor_malloc(1024); + len1 = 1024; + ccp2 = "ABCDEFGHIJABCDEFGHIJ"; + len2 = 21; + tt_int_op(tor_compress_process(state, &cp1, &len1, &ccp2, &len2, 0), + OP_EQ, TOR_COMPRESS_OK); + tt_int_op(0, OP_EQ, len2); /* Make sure we compressed it all. */ +// tt_assert(cp1 > buf1); + + len2 = 0; + cp2 = cp1; + tt_int_op(tor_compress_process(state, &cp1, &len1, &ccp2, &len2, 1), + OP_EQ, TOR_COMPRESS_DONE); + tt_int_op(0, OP_EQ, len2); + tt_assert(cp1 > cp2); /* Make sure we really added something. */ + + tt_assert(!tor_uncompress(&buf3, &len2, buf1, 1024-len1, + ZSTD_METHOD, 1, LOG_WARN)); + /* Make sure it compressed right. */ + tt_str_op(buf3, OP_EQ, "ABCDEFGHIJABCDEFGHIJ"); + tt_int_op(21, OP_EQ, len2); + + done: + if (state) + tor_compress_free(state); + tor_free(buf2); + tor_free(buf3); + tor_free(buf1); +#else + (void)arg; + tt_assert(! tor_compress_supports_method(ZSTD_METHOD)); + + done: + ; +#endif // HAVE_ZSTD. } /** Run unit tests for mmap() wrapper functionality. */ @@ -5717,6 +5944,8 @@ struct testcase_t util_tests[] = { UTIL_LEGACY(pow2), UTIL_LEGACY(gzip), UTIL_TEST(gzip_compression_bomb, TT_FORK), + UTIL_LEGACY(lzma), + UTIL_LEGACY(zstd), UTIL_LEGACY(datadir), UTIL_LEGACY(memarea), UTIL_LEGACY(control_formats), |