aboutsummaryrefslogtreecommitdiff
path: root/src/liblzma
diff options
context:
space:
mode:
authorLasse Collin <lasse.collin@tukaani.org>2022-11-27 23:16:21 +0200
committerLasse Collin <lasse.collin@tukaani.org>2022-11-27 23:16:21 +0200
commit33b8a24b6646a9dbfd8358405aec466b13078559 (patch)
tree166b0977ca46fa664c0b7ceab6e64a59a7abe4d2 /src/liblzma
parentliblzma: Avoid unneeded use of void pointer in LZMA decoder. (diff)
downloadxz-33b8a24b6646a9dbfd8358405aec466b13078559.tar.xz
liblzma: Add LZMA_FILTER_LZMA1EXT to support LZMA1 without end marker.
Some file formats need support for LZMA1 streams that don't use the end of payload marker (EOPM) alias end of stream (EOS) marker. So far liblzma API has supported decompressing such streams via lzma_alone_decoder() when .lzma header specifies a known uncompressed size. Encoding support hasn't been available in the API. Instead of adding a new LZMA1-only API for this purpose, this commit adds a new filter ID for use with raw encoder and decoder. The main benefit of this approach is that then also filter chains are possible, for example, if someone wants to implement support for .7z files that use the x86 BCJ filter with LZMA1 (not BCJ2 as that isn't supported in liblzma).
Diffstat (limited to 'src/liblzma')
-rw-r--r--src/liblzma/api/lzma/lzma12.h123
-rw-r--r--src/liblzma/common/filter_common.c7
-rw-r--r--src/liblzma/common/filter_decoder.c6
-rw-r--r--src/liblzma/common/filter_encoder.c9
-rw-r--r--src/liblzma/lzma/lzma2_encoder.c2
-rw-r--r--src/liblzma/lzma/lzma_decoder.c26
-rw-r--r--src/liblzma/lzma/lzma_encoder.c40
-rw-r--r--src/liblzma/lzma/lzma_encoder.h3
-rw-r--r--src/liblzma/lzma/lzma_encoder_private.h3
9 files changed, 204 insertions, 15 deletions
diff --git a/src/liblzma/api/lzma/lzma12.h b/src/liblzma/api/lzma/lzma12.h
index df5f23b6..d34e7839 100644
--- a/src/liblzma/api/lzma/lzma12.h
+++ b/src/liblzma/api/lzma/lzma12.h
@@ -18,18 +18,41 @@
/**
- * \brief LZMA1 Filter ID
+ * \brief LZMA1 Filter ID (for raw encoder/decoder only, not in .xz)
*
* LZMA1 is the very same thing as what was called just LZMA in LZMA Utils,
* 7-Zip, and LZMA SDK. It's called LZMA1 here to prevent developers from
* accidentally using LZMA when they actually want LZMA2.
- *
- * LZMA1 shouldn't be used for new applications unless you _really_ know
- * what you are doing. LZMA2 is almost always a better choice.
*/
#define LZMA_FILTER_LZMA1 LZMA_VLI_C(0x4000000000000001)
/**
+ * \brief LZMA1 Filter ID with extended options (for raw encoder/decoder)
+ *
+ * This is like LZMA_FILTER_LZMA1 but with this ID a few extra options
+ * are supported in the lzma_options_lzma structure:
+ *
+ * - A flag to tell the encoder if the end of payload marker (EOPM) alias
+ * end of stream (EOS) marker must be written at the end of the stream.
+ * In contrast, LZMA_FILTER_LZMA1 always writes the end marker.
+ *
+ * - Decoder needs to be told the uncompressed size of the stream
+ * or that it is unknown (using the special value UINT64_MAX).
+ * If the size is known, a flag can be set to allow the presence of
+ * the end marker anyway. In contrast, LZMA_FILTER_LZMA1 always
+ * behaves as if the uncompressed size was unknown.
+ *
+ * This allows handling file formats where LZMA1 streams are used but where
+ * the end marker isn't allowed or where it might not (always) be present.
+ * This extended LZMA1 functionality is provided as a Filter ID for raw
+ * encoder and decoder instead of adding new encoder and decoder initialization
+ * functions because this way it is possible to also use extra filters,
+ * for example, LZMA_FILTER_X86 in a filter chain with LZMA_FILTER_LZMA1EXT,
+ * which might be needed to handle some file formats.
+ */
+#define LZMA_FILTER_LZMA1EXT LZMA_VLI_C(0x4000000000000002)
+
+/**
* \brief LZMA2 Filter ID
*
* Usually you want this instead of LZMA1. Compared to LZMA1, LZMA2 adds
@@ -374,6 +397,82 @@ typedef struct {
*/
uint32_t depth;
+ /**
+ * \brief For LZMA_FILTER_LZMA1EXT: Extended flags
+ *
+ * This is used only with LZMA_FILTER_LZMA1EXT.
+ *
+ * Currently only one flag is supported, LZMA_LZMA1EXT_ALLOW_EOPM:
+ *
+ * - Encoder: If the flag is set, then end marker is written just
+ * like it is with LZMA_FILTER_LZMA1. Without this flag the
+ * end marker isn't written and the application has to store
+ * the uncompressed size somewhere outside the compressed stream.
+ * To decompress streams without the end marker, the appliation
+ * has to set the correct uncompressed size in ext_size_low and
+ * ext_size_high.
+ *
+ * - Decoder: If the uncompressed size in ext_size_low and
+ * ext_size_high is set to the special value UINT64_MAX
+ * (indicating unknown uncompressed size) then this flag is
+ * ignored and the end marker must always be present, that is,
+ * the behavior is identical to LZMA_FILTER_LZMA1.
+ *
+ * Otherwise, if this flag isn't set, then the input stream
+ * must not have the end marker; if the end marker is detected
+ * then it will result in LZMA_DATA_ERROR. This is useful when
+ * it is known that the stream must not have the end marker and
+ * strict validation is wanted.
+ *
+ * If this flag is set, then it is autodetected if the end marker
+ * is present after the specified number of uncompressed bytes
+ * has been decompressed (ext_size_low and ext_size_high). The
+ * end marker isn't allowed in any other position. This behavior
+ * is useful when uncompressed size is known but the end marker
+ * may or may not be present. This is the case, for example,
+ * in .7z files (valid .7z files that have the end marker in
+ * LZMA1 streams are rare but they do exist).
+ */
+ uint32_t ext_flags;
+# define LZMA_LZMA1EXT_ALLOW_EOPM UINT32_C(0x01)
+
+ /**
+ * \brief For LZMA_FILTER_LZMA1EXT: Uncompressed size (low bits)
+ *
+ * The 64-bit uncompressed size is needed for decompression with
+ * LZMA_FILTER_LZMA1EXT. The size is ignored by the encoder.
+ *
+ * The special value UINT64_MAX indicates that the uncompressed size
+ * is unknown and that the end of payload marker (also known as
+ * end of stream marker) must be present to indicate the end of
+ * the LZMA1 stream. Any other value indicates the expected
+ * uncompressed size of the LZMA1 stream. (If LZMA1 was used together
+ * with filters that change the size of the data then the uncompressed
+ * size of the LZMA1 stream could be different than the final
+ * uncompressed size of the filtered stream.)
+ *
+ * ext_size_low holds the least significant 32 bits of the
+ * uncompressed size. The most significant 32 bits must be set
+ * in ext_size_high. The macro lzma_ext_size_set(opt_lzma, u64size)
+ * can be used to set these members.
+ *
+ * The 64-bit uncompressed size is split into two uint32_t variables
+ * because there were no reserved uint64_t members and using the
+ * same options structure for LZMA_FILTER_LZMA1, LZMA_FILTER_LZMA1EXT,
+ * and LZMA_FILTER_LZMA2 was otherwise more convenient than having
+ * a new options structure for LZMA_FILTER_LZMA1EXT. (Replacing two
+ * uint32_t members with one uint64_t changes the ABI on some systems
+ * as the alignment of this struct can increase from 4 bytes to 8.)
+ */
+ uint32_t ext_size_low;
+
+ /**
+ * \brief For LZMA_FILTER_LZMA1EXT: Uncompressed size (high bits)
+ *
+ * This holds the most significant 32 bits of the uncompressed size.
+ */
+ uint32_t ext_size_high;
+
/*
* Reserved space to allow possible future extensions without
* breaking the ABI. You should not touch these, because the names
@@ -381,9 +480,6 @@ typedef struct {
* with the currently supported options, so it is safe to leave these
* uninitialized.
*/
- uint32_t reserved_int1;
- uint32_t reserved_int2;
- uint32_t reserved_int3;
uint32_t reserved_int4;
uint32_t reserved_int5;
uint32_t reserved_int6;
@@ -400,6 +496,19 @@ typedef struct {
/**
+ * \brief Macro to set the 64-bit uncompressed size in ext_size_*
+ *
+ * This might be convenient when decoding using LZMA_FILTER_LZMA1EXT.
+ * This isn't used with LZMA_FILTER_LZMA1 or LZMA_FILTER_LZMA2.
+ */
+#define lzma_set_ext_size(opt_lzma2, u64size) \
+do { \
+ (opt_lzma2).ext_size_low = (uint32_t)(u64size); \
+ (opt_lzma2).ext_size_high = (uint32_t)((uint64_t)(u64size) >> 32); \
+} while (0)
+
+
+/**
* \brief Set a compression preset to lzma_options_lzma structure
*
* 0 is the fastest and 9 is the slowest. These match the switches -0 .. -9
diff --git a/src/liblzma/common/filter_common.c b/src/liblzma/common/filter_common.c
index a803b4c2..c8694e2c 100644
--- a/src/liblzma/common/filter_common.c
+++ b/src/liblzma/common/filter_common.c
@@ -42,6 +42,13 @@ static const struct {
.last_ok = true,
.changes_size = true,
},
+ {
+ .id = LZMA_FILTER_LZMA1EXT,
+ .options_size = sizeof(lzma_options_lzma),
+ .non_last_ok = false,
+ .last_ok = true,
+ .changes_size = true,
+ },
#endif
#if defined(HAVE_ENCODER_LZMA2) || defined(HAVE_DECODER_LZMA2)
{
diff --git a/src/liblzma/common/filter_decoder.c b/src/liblzma/common/filter_decoder.c
index b031ac62..fa53f5bd 100644
--- a/src/liblzma/common/filter_decoder.c
+++ b/src/liblzma/common/filter_decoder.c
@@ -50,6 +50,12 @@ static const lzma_filter_decoder decoders[] = {
.memusage = &lzma_lzma_decoder_memusage,
.props_decode = &lzma_lzma_props_decode,
},
+ {
+ .id = LZMA_FILTER_LZMA1EXT,
+ .init = &lzma_lzma_decoder_init,
+ .memusage = &lzma_lzma_decoder_memusage,
+ .props_decode = &lzma_lzma_props_decode,
+ },
#endif
#ifdef HAVE_DECODER_LZMA2
{
diff --git a/src/liblzma/common/filter_encoder.c b/src/liblzma/common/filter_encoder.c
index 5d6c1a7e..978b7a6b 100644
--- a/src/liblzma/common/filter_encoder.c
+++ b/src/liblzma/common/filter_encoder.c
@@ -64,6 +64,15 @@ static const lzma_filter_encoder encoders[] = {
.props_size_fixed = 5,
.props_encode = &lzma_lzma_props_encode,
},
+ {
+ .id = LZMA_FILTER_LZMA1EXT,
+ .init = &lzma_lzma_encoder_init,
+ .memusage = &lzma_lzma_encoder_memusage,
+ .block_size = NULL, // Not needed for LZMA1
+ .props_size_get = NULL,
+ .props_size_fixed = 5,
+ .props_encode = &lzma_lzma_props_encode,
+ },
#endif
#ifdef HAVE_ENCODER_LZMA2
{
diff --git a/src/liblzma/lzma/lzma2_encoder.c b/src/liblzma/lzma/lzma2_encoder.c
index f1252c57..4b6b2311 100644
--- a/src/liblzma/lzma/lzma2_encoder.c
+++ b/src/liblzma/lzma/lzma2_encoder.c
@@ -341,7 +341,7 @@ lzma2_encoder_init(lzma_lz_encoder *lz, const lzma_allocator *allocator,
// Initialize LZMA encoder
return_if_error(lzma_lzma_encoder_create(&coder->lzma, allocator,
- &coder->opt_cur, lz_options));
+ LZMA_FILTER_LZMA2, &coder->opt_cur, lz_options));
// Make sure that we will always have enough history available in
// case we need to use uncompressed chunks. They are used when the
diff --git a/src/liblzma/lzma/lzma_decoder.c b/src/liblzma/lzma/lzma_decoder.c
index 550963d1..26c148a9 100644
--- a/src/liblzma/lzma/lzma_decoder.c
+++ b/src/liblzma/lzma/lzma_decoder.c
@@ -1018,11 +1018,35 @@ lzma_decoder_init(lzma_lz_decoder *lz, const lzma_allocator *allocator,
if (!is_lclppb_valid(options))
return LZMA_PROG_ERROR;
+ lzma_vli uncomp_size = LZMA_VLI_UNKNOWN;
+ bool allow_eopm = true;
+
+ if (id == LZMA_FILTER_LZMA1EXT) {
+ const lzma_options_lzma *opt = options;
+
+ // Only one flag is supported.
+ if (opt->ext_flags & ~LZMA_LZMA1EXT_ALLOW_EOPM)
+ return LZMA_OPTIONS_ERROR;
+
+ // FIXME? Using lzma_vli instead of uint64_t is weird because
+ // this has nothing to do with .xz headers and variable-length
+ // integer encoding. On the other hand, using LZMA_VLI_UNKNOWN
+ // instead of UINT64_MAX is clearer when unknown size is
+ // meant. A problem with using lzma_vli is that now we
+ // allow > LZMA_VLI_MAX which is fine in this file but
+ // it's still confusing. Note that alone_decoder.c also
+ // allows > LZMA_VLI_MAX when setting uncompressed size.
+ uncomp_size = opt->ext_size_low
+ + ((uint64_t)(opt->ext_size_high) << 32);
+ allow_eopm = (opt->ext_flags & LZMA_LZMA1EXT_ALLOW_EOPM) != 0
+ || uncomp_size == LZMA_VLI_UNKNOWN;
+ }
+
return_if_error(lzma_lzma_decoder_create(
lz, allocator, options, lz_options));
lzma_decoder_reset(lz->coder, options);
- lzma_decoder_uncompressed(lz->coder, LZMA_VLI_UNKNOWN, true);
+ lzma_decoder_uncompressed(lz->coder, uncomp_size, allow_eopm);
return LZMA_OK;
}
diff --git a/src/liblzma/lzma/lzma_encoder.c b/src/liblzma/lzma/lzma_encoder.c
index e2dbbc03..dc62f44f 100644
--- a/src/liblzma/lzma/lzma_encoder.c
+++ b/src/liblzma/lzma/lzma_encoder.c
@@ -416,7 +416,7 @@ lzma_lzma_encode(lzma_lzma1_encoder *restrict coder, lzma_mf *restrict mf,
//
// Plain LZMA streams without EOPM aren't supported except when
// output size limiting is enabled.
- if (limit == UINT32_MAX && coder->out_limit == 0)
+ if (coder->use_eopm)
encode_eopm(coder, (uint32_t)(coder->uncomp_size));
// Flush the remaining bytes from the range encoder.
@@ -462,6 +462,7 @@ lzma_lzma_set_out_limit(
lzma_lzma1_encoder *coder = coder_ptr;
coder->out_limit = out_limit;
coder->uncomp_size_ptr = uncomp_size;
+ coder->use_eopm = false;
return LZMA_OK;
}
@@ -599,10 +600,13 @@ lzma_lzma_encoder_reset(lzma_lzma1_encoder *coder,
extern lzma_ret
-lzma_lzma_encoder_create(void **coder_ptr,
- const lzma_allocator *allocator,
- const lzma_options_lzma *options, lzma_lz_options *lz_options)
+lzma_lzma_encoder_create(void **coder_ptr, const lzma_allocator *allocator,
+ lzma_vli id, const lzma_options_lzma *options,
+ lzma_lz_options *lz_options)
{
+ assert(id == LZMA_FILTER_LZMA1 || id == LZMA_FILTER_LZMA1EXT
+ || id == LZMA_FILTER_LZMA2);
+
// Allocate lzma_lzma1_encoder if it wasn't already allocated.
if (*coder_ptr == NULL) {
*coder_ptr = lzma_alloc(sizeof(lzma_lzma1_encoder), allocator);
@@ -672,6 +676,32 @@ lzma_lzma_encoder_create(void **coder_ptr,
// Output size limitting is disabled by default.
coder->out_limit = 0;
+ // Determine if end marker is wanted:
+ // - It is never used with LZMA2.
+ // - It is always used with LZMA_FILTER_LZMA1 (unless
+ // lzma_lzma_set_out_limit() is called later).
+ // - LZMA_FILTER_LZMA1EXT has a flag for it in the options.
+ coder->use_eopm = (id == LZMA_FILTER_LZMA1);
+ if (id == LZMA_FILTER_LZMA1EXT) {
+ // Check if unsupported flags are present.
+ if (options->ext_flags & ~LZMA_LZMA1EXT_ALLOW_EOPM)
+ return LZMA_OPTIONS_ERROR;
+
+ coder->use_eopm = (options->ext_flags
+ & LZMA_LZMA1EXT_ALLOW_EOPM) != 0;
+
+ // TODO? As long as there are no filters that change the size
+ // of the data, it is enough to look at lzma_stream.total_in
+ // after encoding has been finished to know the uncompressed
+ // size of the LZMA1 stream. But in the future there could be
+ // filters that change the size of the data and then total_in
+ // doesn't work as the LZMA1 stream size might be different
+ // due to another filter in the chain. The problem is simple
+ // to solve: Add another flag to ext_flags and then set
+ // coder->uncomp_size_ptr to the address stored in
+ // lzma_options_lzma.reserved_ptr2 (or _ptr1).
+ }
+
set_lz_options(lz_options, options);
return lzma_lzma_encoder_reset(coder, options);
@@ -685,7 +715,7 @@ lzma_encoder_init(lzma_lz_encoder *lz, const lzma_allocator *allocator,
lz->code = &lzma_encode;
lz->set_out_limit = &lzma_lzma_set_out_limit;
return lzma_lzma_encoder_create(
- &lz->coder, allocator, options, lz_options);
+ &lz->coder, allocator, id, options, lz_options);
}
diff --git a/src/liblzma/lzma/lzma_encoder.h b/src/liblzma/lzma/lzma_encoder.h
index 6cfdf228..84d8c916 100644
--- a/src/liblzma/lzma/lzma_encoder.h
+++ b/src/liblzma/lzma/lzma_encoder.h
@@ -40,7 +40,8 @@ extern bool lzma_lzma_lclppb_encode(
/// Initializes raw LZMA encoder; this is used by LZMA2.
extern lzma_ret lzma_lzma_encoder_create(
void **coder_ptr, const lzma_allocator *allocator,
- const lzma_options_lzma *options, lzma_lz_options *lz_options);
+ lzma_vli id, const lzma_options_lzma *options,
+ lzma_lz_options *lz_options);
/// Resets an already initialized LZMA encoder; this is used by LZMA2.
diff --git a/src/liblzma/lzma/lzma_encoder_private.h b/src/liblzma/lzma/lzma_encoder_private.h
index 8960c52c..b228c577 100644
--- a/src/liblzma/lzma/lzma_encoder_private.h
+++ b/src/liblzma/lzma/lzma_encoder_private.h
@@ -111,6 +111,9 @@ struct lzma_lzma1_encoder_s {
/// have been written to the output buffer yet.
bool is_flushed;
+ /// True if end of payload marker will be written.
+ bool use_eopm;
+
uint32_t pos_mask; ///< (1 << pos_bits) - 1
uint32_t literal_context_bits;
uint32_t literal_pos_mask;