Accepting request 1119078 from home:dspinella:branches:devel:libraries:c_c++

- Update to 1.3:
  * Building using K&R (pre-ANSI) function definitions is no longer supported.
  * Fixed a bug in deflateBound() for level 0 and memLevel 9.
  * Fixed a bug when gzungetc() is used immediately after gzopen().
  * Fixed a bug when using gzflush() with a very small buffer.
  * Fixed a crash when gzsetparams() is attempted for a transparent write.
  * Fixed test/example.c to work with FORCE_STORED.
  * Fixed minizip to allow it to open an empty zip file.
  * Fixed reading disk number start on zip64 files in minizip.
  * Fixed a logic error in minizip argument processing. 
- Added patches:
  * zlib-1.3-IBM-Z-hw-accelerated-deflate-s390x.patch
- Refreshed patches:
  * zlib-1.2.12-add-optimized-slide_hash-for-power.patch
  * zlib-1.2.12-add-vectorized-longest_match-for-power.patch
  * zlib-1.2.12-adler32-vector-optimizations-for-power.patch
  * zlib-1.2.13-optimized-s390.patch
  * zlib-format.patch
  * zlib-no-version-check.patch
- Removed patches:
  * bsc1210593.patch
  * zlib-1.2.13-fix-bug-deflateBound.patch
  * zlib-1.2.12-s390-vectorize-crc32.patch
  * zlib-1.2.13-IBM-Z-hw-accelerated-deflate-s390x.patch
  * zlib-1.2.12-add-optimized-slide_hash-for-power.patch
  * zlib-1.2.12-fix-invalid-memory-access-on-ppc-and-ppc64.patch
  * zlib-1.2.12-add-vectorized-longest_match-for-power.patch
  * zlib-1.2.12-adler32-vector-optimizations-for-power.patch
- Fix CVE-2023-45853, integer overflow and resultant heap-based buffer
  overflow in zipOpenNewFileInZip4_6, bsc#1216378

OBS-URL: https://build.opensuse.org/request/show/1119078
OBS-URL: https://build.opensuse.org/package/show/devel:libraries:c_c++/zlib?expand=0&rev=95
This commit is contained in:
Danilo Spinella 2023-10-19 16:28:40 +00:00 committed by Git OBS Bridge
parent e6f111ebe2
commit 22d97a578b
18 changed files with 886 additions and 1909 deletions

38
CVE-2023-45853.patch Normal file
View File

@ -0,0 +1,38 @@
From 431e66398552effd82d5c0ea982a521821782ebd Mon Sep 17 00:00:00 2001
From: Hans Wennborg <hans@chromium.org>
Date: Fri, 18 Aug 2023 11:05:33 +0200
Subject: [PATCH] minizip: Check length of comment, filename, and extra field,
in zipOpenNewFileInZip4_64
These are stored in 16-bit fields in the zip file format. Passing longer
values would generate an invalid file.
Passing very long values could also cause the computation of
zi->ci.size_centralheader to overflow, which would cause heap buffer
overflow on subsequent writes to zi->ci.central_header.
---
contrib/minizip/zip.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/contrib/minizip/zip.c b/contrib/minizip/zip.c
index 3d3d4cadd..0446109b2 100644
--- a/contrib/minizip/zip.c
+++ b/contrib/minizip/zip.c
@@ -1043,6 +1043,17 @@ extern int ZEXPORT zipOpenNewFileInZip4_64(zipFile file, const char* filename, c
return ZIP_PARAMERROR;
#endif
+ // The filename and comment length must fit in 16 bits.
+ if ((filename!=NULL) && (strlen(filename)>0xffff))
+ return ZIP_PARAMERROR;
+ if ((comment!=NULL) && (strlen(comment)>0xffff))
+ return ZIP_PARAMERROR;
+ // The extra field length must fit in 16 bits. If the member also requires
+ // a Zip64 extra block, that will also need to fit within that 16-bit
+ // length, but that will be checked for later.
+ if ((size_extrafield_local>0xffff) || (size_extrafield_global>0xffff))
+ return ZIP_PARAMERROR;
+
zi = (zip64_internal*)file;
if (zi->in_opened_file_inzip == 1)

View File

@ -1,13 +0,0 @@
Index: zlib-1.2.11/contrib/s390/dfltcc_deflate.h
===================================================================
--- zlib-1.2.11.orig/contrib/s390/dfltcc_deflate.h
+++ zlib-1.2.11/contrib/s390/dfltcc_deflate.h
@@ -45,7 +45,7 @@ int ZLIB_INTERNAL dfltcc_deflate_get_dic
#define DEFLATE_DONE dfltcc_deflate_done
#define DEFLATE_BOUND_ADJUST_COMPLEN(strm, complen, source_len) \
do { \
- if (dfltcc_can_deflate((strm))) \
+ if (deflateStateCheck((strm)) || dfltcc_can_deflate((strm))) \
(complen) = DEFLATE_BOUND_COMPLEN(source_len); \
} while (0)
#define DEFLATE_NEED_CONSERVATIVE_BOUND(strm) (dfltcc_can_deflate((strm)))

View File

@ -1,217 +0,0 @@
From 4a8d89ae49aa17d1634a2816c8d159f533a07eae Mon Sep 17 00:00:00 2001
From: Matheus Castanho <msc@linux.ibm.com>
Date: Wed, 27 Nov 2019 10:18:10 -0300
Subject: [PATCH] Add optimized slide_hash for Power
Considerable time is spent on deflate.c:slide_hash() during
deflate. This commit introduces a new slide_hash function that
uses VSX vector instructions to slide 8 hash elements at a time,
instead of just one as the standard code does.
The choice between the optimized and default versions is made only
on the first call to the function, enabling a fallback to standard
behavior if the host processor does not support VSX instructions,
so the same binary can be used for multiple Power processor
versions.
Author: Matheus Castanho <msc@linux.ibm.com>
---
CMakeLists.txt | 3 +-
Makefile.in | 8 ++++
configure | 4 +-
contrib/power/power.h | 3 ++
contrib/power/slide_hash_power8.c | 63 +++++++++++++++++++++++++++++
contrib/power/slide_hash_resolver.c | 15 +++++++
deflate.c | 12 ++++++
7 files changed, 105 insertions(+), 3 deletions(-)
create mode 100644 contrib/power/slide_hash_power8.c
create mode 100644 contrib/power/slide_hash_resolver.c
Index: zlib-1.2.13/CMakeLists.txt
===================================================================
--- zlib-1.2.13.orig/CMakeLists.txt
+++ zlib-1.2.13/CMakeLists.txt
@@ -174,7 +174,8 @@ if(CMAKE_COMPILER_IS_GNUCC)
add_definitions(-DZ_POWER8)
set(ZLIB_POWER8
contrib/power/adler32_power8.c
- contrib/power/crc32_z_power8.c)
+ contrib/power/crc32_z_power8.c
+ contrib/power/slide_hash_power8.c)
set_source_files_properties(
${ZLIB_POWER8}
Index: zlib-1.2.13/Makefile.in
===================================================================
--- zlib-1.2.13.orig/Makefile.in
+++ zlib-1.2.13/Makefile.in
@@ -185,6 +185,9 @@ crc32-vx.o: $(SRCDIR)contrib/s390/crc32-
deflate.o: $(SRCDIR)deflate.c
$(CC) $(CFLAGS) $(ZINC) -c -o $@ $(SRCDIR)deflate.c
+slide_hash_power8.o: $(SRCDIR)contrib/power/slide_hash_power8.c
+ $(CC) $(CFLAGS) -mcpu=power8 $(ZINC) -c -o $@ $(SRCDIR)contrib/power/slide_hash_power8.c
+
infback.o: $(SRCDIR)infback.c
$(CC) $(CFLAGS) $(ZINC) -c -o $@ $(SRCDIR)infback.c
@@ -252,6 +255,11 @@ deflate.lo: $(SRCDIR)deflate.c
$(CC) $(SFLAGS) $(ZINC) -DPIC -c -o objs/deflate.o $(SRCDIR)deflate.c
-@mv objs/deflate.o $@
+slide_hash_power8.lo: $(SRCDIR)contrib/power/slide_hash_power8.c
+ -@mkdir objs 2>/dev/null || test -d objs
+ $(CC) $(SFLAGS) -mcpu=power8 $(ZINC) -DPIC -c -o objs/slide_hash_power8.o $(SRCDIR)contrib/power/slide_hash_power8.c
+ -@mv objs/slide_hash_power8.o $@
+
infback.lo: $(SRCDIR)infback.c
-@mkdir objs 2>/dev/null || test -d objs
$(CC) $(SFLAGS) $(ZINC) -DPIC -c -o objs/infback.o $(SRCDIR)infback.c
Index: zlib-1.2.13/configure
===================================================================
--- zlib-1.2.13.orig/configure
+++ zlib-1.2.13/configure
@@ -898,8 +898,8 @@ if tryboth $CC -c $CFLAGS $test.c; then
if tryboth $CC -c $CFLAGS -mcpu=power8 $test.c; then
POWER8="-DZ_POWER8"
- PIC_OBJC="${PIC_OBJC} adler32_power8.lo crc32_z_power8.lo"
- OBJC="${OBJC} adler32_power8.o crc32_z_power8.o"
+ PIC_OBJC="${PIC_OBJC} adler32_power8.lo crc32_z_power8.lo slide_hash_power8.lo"
+ OBJC="${OBJC} adler32_power8.o crc32_z_power8.o slide_hash_power8.o"
echo "Checking for -mcpu=power8 support... Yes." | tee -a configure.log
else
echo "Checking for -mcpu=power8 support... No." | tee -a configure.log
Index: zlib-1.2.13/contrib/power/power.h
===================================================================
--- zlib-1.2.13.orig/contrib/power/power.h
+++ zlib-1.2.13/contrib/power/power.h
@@ -4,7 +4,10 @@
*/
#include "../../zconf.h"
#include "../../zutil.h"
+#include "../../deflate.h"
uLong _adler32_power8(uLong adler, const Bytef* buf, uInt len);
unsigned long _crc32_z_power8(unsigned long, const Bytef *, z_size_t);
+
+void _slide_hash_power8(deflate_state *s);
Index: zlib-1.2.13/contrib/power/slide_hash_power8.c
===================================================================
--- /dev/null
+++ zlib-1.2.13/contrib/power/slide_hash_power8.c
@@ -0,0 +1,63 @@
+ /* Copyright (C) 2019 Matheus Castanho <msc@linux.ibm.com>, IBM
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include <altivec.h>
+#include "../../deflate.h"
+
+local inline void slide_hash_power8_loop OF((deflate_state *s,
+ unsigned n_elems, Posf *table_end)) __attribute__((always_inline));
+
+local void slide_hash_power8_loop(
+ deflate_state *s,
+ unsigned n_elems,
+ Posf *table_end)
+{
+ vector unsigned short vw, vm, *vp;
+ unsigned chunks;
+
+ /* Each vector register (chunk) corresponds to 128 bits == 8 Posf,
+ * so instead of processing each of the n_elems in the hash table
+ * individually, we can do it in chunks of 8 with vector instructions.
+ *
+ * This function is only called from slide_hash_power8(), and both calls
+ * pass n_elems as a power of 2 higher than 2^7, as defined by
+ * deflateInit2_(), so n_elems will always be a multiple of 8. */
+ chunks = n_elems >> 3;
+ Assert(n_elems % 8 == 0, "Weird hash table size!");
+
+ /* This type casting is safe since s->w_size is always <= 64KB
+ * as defined by deflateInit2_() and Posf == unsigned short */
+ vw[0] = (Posf) s->w_size;
+ vw = vec_splat(vw,0);
+
+ vp = (vector unsigned short *) table_end;
+
+ do {
+ /* Processing 8 elements at a time */
+ vp--;
+ vm = *vp;
+
+ /* This is equivalent to: m >= w_size ? m - w_size : 0
+ * Since we are using a saturated unsigned subtraction, any
+ * values that are > w_size will be set to 0, while the others
+ * will be subtracted by w_size. */
+ *vp = vec_subs(vm,vw);
+ } while (--chunks);
+};
+
+void ZLIB_INTERNAL _slide_hash_power8(deflate_state *s)
+{
+ unsigned n;
+ Posf *p;
+
+ n = s->hash_size;
+ p = &s->head[n];
+ slide_hash_power8_loop(s,n,p);
+
+#ifndef FASTEST
+ n = s->w_size;
+ p = &s->prev[n];
+ slide_hash_power8_loop(s,n,p);
+#endif
+}
Index: zlib-1.2.13/contrib/power/slide_hash_resolver.c
===================================================================
--- /dev/null
+++ zlib-1.2.13/contrib/power/slide_hash_resolver.c
@@ -0,0 +1,15 @@
+/* Copyright (C) 2019 Matheus Castanho <msc@linux.ibm.com>, IBM
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "../gcc/zifunc.h"
+#include "power.h"
+
+Z_IFUNC(slide_hash) {
+#ifdef Z_POWER8
+ if (__builtin_cpu_supports("arch_2_07"))
+ return _slide_hash_power8;
+#endif
+
+ return slide_hash_default;
+}
Index: zlib-1.2.13/deflate.c
===================================================================
--- zlib-1.2.13.orig/deflate.c
+++ zlib-1.2.13/deflate.c
@@ -204,6 +204,13 @@ local const config configuration_table[1
(unsigned)(s->hash_size - 1)*sizeof(*s->head)); \
} while (0)
+#ifdef Z_POWER_OPT
+/* Rename function so resolver can use its symbol. The default version will be
+ * returned by the resolver if the host has no support for an optimized version.
+ */
+#define slide_hash slide_hash_default
+#endif /* Z_POWER_OPT */
+
/* ===========================================================================
* Slide the hash table when sliding the window down (could be avoided with 32
* bit values at the expense of memory usage). We slide even when level == 0 to
@@ -235,6 +242,11 @@ local void slide_hash(s)
#endif
}
+#ifdef Z_POWER_OPT
+#undef slide_hash
+#include "contrib/power/slide_hash_resolver.c"
+#endif /* Z_POWER_OPT */
+
/* ========================================================================= */
int ZEXPORT deflateInit_(strm, level, version, stream_size)
z_streamp strm;

View File

@ -1,338 +0,0 @@
From aecdff0646c7e188b48f6db285d8d63a74f246c1 Mon Sep 17 00:00:00 2001
From: Matheus Castanho <msc@linux.ibm.com>
Date: Tue, 29 Oct 2019 18:04:11 -0300
Subject: [PATCH] Add vectorized longest_match for Power
This commit introduces an optimized version of the longest_match
function for Power processors. It uses VSX instructions to match
16 bytes at a time on each comparison, instead of one by one.
Author: Matheus Castanho <msc@linux.ibm.com>
---
CMakeLists.txt | 3 +-
Makefile.in | 8 +
configure | 4 +-
contrib/power/longest_match_power9.c | 194 +++++++++++++++++++++++++
contrib/power/longest_match_resolver.c | 15 ++
contrib/power/power.h | 2 +
deflate.c | 13 ++
7 files changed, 236 insertions(+), 3 deletions(-)
create mode 100644 contrib/power/longest_match_power9.c
create mode 100644 contrib/power/longest_match_resolver.c
Index: zlib-1.2.13/CMakeLists.txt
===================================================================
--- zlib-1.2.13.orig/CMakeLists.txt
+++ zlib-1.2.13/CMakeLists.txt
@@ -187,7 +187,8 @@ if(CMAKE_COMPILER_IS_GNUCC)
if(POWER9)
add_definitions(-DZ_POWER9)
- set(ZLIB_POWER9 )
+ set(ZLIB_POWER9
+ contrib/power/longest_match_power9.c)
set_source_files_properties(
${ZLIB_POWER9}
Index: zlib-1.2.13/Makefile.in
===================================================================
--- zlib-1.2.13.orig/Makefile.in
+++ zlib-1.2.13/Makefile.in
@@ -185,6 +185,9 @@ crc32-vx.o: $(SRCDIR)contrib/s390/crc32-
deflate.o: $(SRCDIR)deflate.c
$(CC) $(CFLAGS) $(ZINC) -c -o $@ $(SRCDIR)deflate.c
+longest_match_power9.o: $(SRCDIR)contrib/power/longest_match_power9.c
+ $(CC) $(CFLAGS) -mcpu=power9 $(ZINC) -c -o $@ $(SRCDIR)contrib/power/longest_match_power9.c
+
slide_hash_power8.o: $(SRCDIR)contrib/power/slide_hash_power8.c
$(CC) $(CFLAGS) -mcpu=power8 $(ZINC) -c -o $@ $(SRCDIR)contrib/power/slide_hash_power8.c
@@ -255,6 +258,11 @@ deflate.lo: $(SRCDIR)deflate.c
$(CC) $(SFLAGS) $(ZINC) -DPIC -c -o objs/deflate.o $(SRCDIR)deflate.c
-@mv objs/deflate.o $@
+longest_match_power9.lo: $(SRCDIR)contrib/power/longest_match_power9.c
+ -@mkdir objs 2>/dev/null || test -d objs
+ $(CC) $(SFLAGS) -mcpu=power9 $(ZINC) -DPIC -c -o objs/longest_match_power9.o $(SRCDIR)contrib/power/longest_match_power9.c
+ -@mv objs/longest_match_power9.o $@
+
slide_hash_power8.lo: $(SRCDIR)contrib/power/slide_hash_power8.c
-@mkdir objs 2>/dev/null || test -d objs
$(CC) $(SFLAGS) -mcpu=power8 $(ZINC) -DPIC -c -o objs/slide_hash_power8.o $(SRCDIR)contrib/power/slide_hash_power8.c
Index: zlib-1.2.13/configure
===================================================================
--- zlib-1.2.13.orig/configure
+++ zlib-1.2.13/configure
@@ -907,8 +907,8 @@ if tryboth $CC -c $CFLAGS $test.c; then
if tryboth $CC -c $CFLAGS -mcpu=power9 $test.c; then
POWER9="-DZ_POWER9"
- PIC_OBJC="${PIC_OBJC}"
- OBJC="${OBJC}"
+ PIC_OBJC="$PIC_OBJC longest_match_power9.lo"
+ OBJC="$OBJC longest_match_power9.o"
echo "Checking for -mcpu=power9 support... Yes." | tee -a configure.log
else
echo "Checking for -mcpu=power9 support... No." | tee -a configure.log
Index: zlib-1.2.13/contrib/power/longest_match_power9.c
===================================================================
--- /dev/null
+++ zlib-1.2.13/contrib/power/longest_match_power9.c
@@ -0,0 +1,194 @@
+/* Copyright (C) 2019 Matheus Castanho <msc@linux.ibm.com>, IBM
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include <altivec.h>
+#include "../../deflate.h"
+
+local inline int vec_match OF((Bytef* scan, Bytef* match))
+ __attribute__((always_inline));
+
+local inline int vec_match(Bytef* scan, Bytef* match)
+{
+ vector unsigned char vscan, vmatch, vc;
+ int len;
+
+ vscan = *((vector unsigned char *) scan);
+ vmatch = *((vector unsigned char *) match);
+
+ /* Compare 16 bytes at a time.
+ * Each byte of vc will be either all ones or all zeroes,
+ * depending on the result of the comparison
+ */
+ vc = (vector unsigned char) vec_cmpne(vscan,vmatch);
+
+ /* Since the index of matching bytes will contain only zeroes
+ * on vc (since we used cmpne), counting the number of consecutive
+ * bytes where LSB == 0 is the same as counting the length of the match.
+ *
+ * There was an issue in the way the vec_cnttz_lsbb builtin was implemented
+ * that got fixed on GCC 12, but now we have to use different builtins
+ * depending on the compiler version. To avoid that, let's use inline asm to
+ * generate the exact instruction we need.
+ */
+ #ifdef __LITTLE_ENDIAN__
+ asm volatile("vctzlsbb %0, %1\n\t" : "=r" (len) : "v" (vc));
+ #else
+ asm volatile("vclzlsbb %0, %1\n\t" : "=r" (len) : "v" (vc));
+ #endif
+
+ return len;
+}
+
+uInt ZLIB_INTERNAL _longest_match_power9(deflate_state *s, IPos cur_match)
+{
+ unsigned chain_length = s->max_chain_length;/* max hash chain length */
+ register Bytef *scan = s->window + s->strstart; /* current string */
+ register Bytef *match; /* matched string */
+ register int len; /* length of current match */
+ int best_len = (int)s->prev_length; /* best match length so far */
+ int nice_match = s->nice_match; /* stop if match long enough */
+ int mbytes; /* matched bytes inside loop */
+ IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
+ s->strstart - (IPos)MAX_DIST(s) : 0;
+ /* Stop when cur_match becomes <= limit. To simplify the code,
+ * we prevent matches with the string of window index 0.
+ */
+ Posf *prev = s->prev;
+ uInt wmask = s->w_mask;
+
+#if (MAX_MATCH == 258)
+ /* Compare the last two bytes at once. */
+ register Bytef *strend2 = s->window + s->strstart + MAX_MATCH - 2;
+ register ush scan_end = *(ushf*)(scan+best_len-1);
+#else
+ register Bytef *strend = s->window + s->strstart + MAX_MATCH;
+ register Byte scan_end1 = scan[best_len-1];
+ register Byte scan_end = scan[best_len];
+#endif
+
+ /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
+ * It is easy to get rid of this optimization if necessary.
+ */
+ Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
+
+ /* Do not waste too much time if we already have a good match: */
+ if (s->prev_length >= s->good_match) {
+ chain_length >>= 2;
+ }
+ /* Do not look for matches beyond the end of the input. This is necessary
+ * to make deflate deterministic.
+ */
+ if ((uInt)nice_match > s->lookahead) nice_match = (int)s->lookahead;
+
+ Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
+
+ do {
+ Assert(cur_match < s->strstart, "no future");
+ match = s->window + cur_match;
+
+ /* Skip to next match if the match length cannot increase
+ * or if the match length is less than 2. Note that the checks below
+ * for insufficient lookahead only occur occasionally for performance
+ * reasons. Therefore uninitialized memory will be accessed, and
+ * conditional jumps will be made that depend on those values.
+ * However the length of the match is limited to the lookahead, so
+ * the output of deflate is not affected by the uninitialized values.
+ */
+
+/* MAX_MATCH - 2 should be a multiple of 16 for this optimization to work. */
+#if (MAX_MATCH == 258)
+
+ /* Compare ending (2 bytes) and beginning of potential match.
+ *
+ * On Power processors, loading a 16-byte vector takes only 1 extra
+ * cycle compared to a regular byte load. So instead of comparing the
+ * first two bytes and then the rest later if they match, we can compare
+ * the first 16 at once, and when we have a match longer than 2, we will
+ * already have the result of comparing the first 16 bytes saved in mbytes.
+ */
+ if (*(ushf*)(match+best_len-1) != scan_end ||
+ (mbytes = vec_match(scan,match)) < 3) continue;
+
+ scan += mbytes;
+ match += mbytes;
+
+ /* In case when we may have a match longer than 16, we perform further
+ * comparisons in chunks of 16 and keep going while all bytes match.
+ */
+ while(mbytes == 16) {
+ mbytes = vec_match(scan,match);
+ scan += mbytes;
+ match += mbytes;
+
+ /* We also have to limit the maximum match based on MAX_MATCH.
+ * Since we are comparing 16 bytes at a time and MAX_MATCH == 258 (to
+ * comply with default implementation), we should stop comparing when
+ * we have matched 256 bytes, which happens when scan == strend2.
+ * In this ("rare") case, we have to check the remaining 2 bytes
+ * individually using common load and compare operations.
+ */
+ if(scan >= strend2) {
+ if(*scan == *match) {
+ if(*++scan == *++match)
+ scan++;
+ }
+ break;
+ }
+ }
+
+ Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
+
+ len = (MAX_MATCH - 2) - (int)(strend2 - scan);
+ scan = strend2 - (MAX_MATCH - 2);
+
+#else /* MAX_MATCH == 258 */
+
+ if (match[best_len] != scan_end ||
+ match[best_len-1] != scan_end1 ||
+ *match != *scan ||
+ *++match != scan[1]) continue;
+
+ /* The check at best_len-1 can be removed because it will be made
+ * again later. (This heuristic is not always a win.)
+ * It is not necessary to compare scan[2] and match[2] since they
+ * are always equal when the other bytes match, given that
+ * the hash keys are equal and that HASH_BITS >= 8.
+ */
+ scan += 2, match++;
+ Assert(*scan == *match, "match[2]?");
+
+ /* We check for insufficient lookahead only every 8th comparison;
+ * the 256th check will be made at strstart+258.
+ */
+ do {
+ } while (*++scan == *++match && *++scan == *++match &&
+ *++scan == *++match && *++scan == *++match &&
+ *++scan == *++match && *++scan == *++match &&
+ *++scan == *++match && *++scan == *++match &&
+ scan < strend);
+
+ Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
+
+ len = MAX_MATCH - (int)(strend - scan);
+ scan = strend - MAX_MATCH;
+
+#endif /* MAX_MATCH == 258 */
+
+ if (len > best_len) {
+ s->match_start = cur_match;
+ best_len = len;
+ if (len >= nice_match) break;
+#if (MAX_MATCH == 258)
+ scan_end = *(ushf*)(scan+best_len-1);
+#else
+ scan_end1 = scan[best_len-1];
+ scan_end = scan[best_len];
+#endif
+ }
+ } while ((cur_match = prev[cur_match & wmask]) > limit
+ && --chain_length != 0);
+
+ if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
+ return s->lookahead;
+}
Index: zlib-1.2.13/contrib/power/longest_match_resolver.c
===================================================================
--- /dev/null
+++ zlib-1.2.13/contrib/power/longest_match_resolver.c
@@ -0,0 +1,15 @@
+/* Copyright (C) 2019 Matheus Castanho <msc@linux.ibm.com>, IBM
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "../gcc/zifunc.h"
+#include "power.h"
+
+Z_IFUNC(longest_match) {
+#ifdef Z_POWER9
+ if (__builtin_cpu_supports("arch_3_00"))
+ return _longest_match_power9;
+#endif
+
+ return longest_match_default;
+}
Index: zlib-1.2.13/contrib/power/power.h
===================================================================
--- zlib-1.2.13.orig/contrib/power/power.h
+++ zlib-1.2.13/contrib/power/power.h
@@ -10,4 +10,6 @@ uLong _adler32_power8(uLong adler, const
unsigned long _crc32_z_power8(unsigned long, const Bytef *, z_size_t);
+uInt _longest_match_power9(deflate_state *s, IPos cur_match);
+
void _slide_hash_power8(deflate_state *s);
Index: zlib-1.2.13/deflate.c
===================================================================
--- zlib-1.2.13.orig/deflate.c
+++ zlib-1.2.13/deflate.c
@@ -1313,6 +1313,14 @@ local void lm_init(s)
* string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
* OUT assertion: the match length is not greater than s->lookahead.
*/
+
+#ifdef Z_POWER_OPT
+/* Rename function so resolver can use its symbol. The default version will be
+ * returned by the resolver if the host has no support for an optimized version.
+ */
+#define longest_match longest_match_default
+#endif /* Z_POWER_OPT */
+
local uInt longest_match(s, pcur_match)
deflate_state *s;
IPos pcur_match; /* current match */
@@ -1460,6 +1468,11 @@ local uInt longest_match(s, pcur_match)
return s->lookahead;
}
+#ifdef Z_POWER_OPT
+#undef longest_match
+#include "contrib/power/longest_match_resolver.c"
+#endif /* Z_POWER_OPT */
+
#else /* FASTEST */
/* ---------------------------------------------------------------------------

View File

@ -1,342 +0,0 @@
From 772f4bd0f880c4c193ab7da78728f38821572a02 Mon Sep 17 00:00:00 2001
From: Rogerio Alves <rcardoso@linux.ibm.com>
Date: Mon, 9 Dec 2019 14:40:53 -0300
Subject: [PATCH] Adler32 vector optimization for Power.
This commit implements a Power (POWER8+) vector optimization for Adler32
checksum using VSX (vector) instructions. The VSX adler32 checksum is up
to 10x fast than the adler32 baseline code.
Author: Rogerio Alves <rcardoso@linux.ibm.com>
---
CMakeLists.txt | 1 +
Makefile.in | 8 ++
adler32.c | 11 ++
configure | 4 +-
contrib/power/adler32_power8.c | 196 +++++++++++++++++++++++++++++++
contrib/power/adler32_resolver.c | 15 +++
contrib/power/power.h | 4 +-
7 files changed, 236 insertions(+), 3 deletions(-)
create mode 100644 contrib/power/adler32_power8.c
create mode 100644 contrib/power/adler32_resolver.c
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 581e1fa6d..c6296ee68 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -185,6 +185,7 @@ if(CMAKE_COMPILER_IS_GNUCC)
if(POWER8)
add_definitions(-DZ_POWER8)
set(ZLIB_POWER8
+ contrib/power/adler32_power8.c
contrib/power/crc32_z_power8.c)
set_source_files_properties(
diff --git a/Makefile.in b/Makefile.in
index 16943044e..a0ffac860 100644
--- a/Makefile.in
+++ b/Makefile.in
@@ -165,6 +165,9 @@ minigzip64.o: $(SRCDIR)test/minigzip.c $(SRCDIR)zlib.h zconf.h
adler32.o: $(SRCDIR)adler32.c
$(CC) $(CFLAGS) $(ZINC) -c -o $@ $(SRCDIR)adler32.c
+adler32_power8.o: $(SRCDIR)contrib/power/adler32_power8.c
+ $(CC) $(CFLAGS) -mcpu=power8 $(ZINC) -c -o $@ $(SRCDIR)contrib/power/adler32_power8.c
+
crc32.o: $(SRCDIR)crc32.c
$(CC) $(CFLAGS) $(ZINC) -c -o $@ $(SRCDIR)crc32.c
@@ -216,6 +219,11 @@ adler32.lo: $(SRCDIR)adler32.c
$(CC) $(SFLAGS) $(ZINC) -DPIC -c -o objs/adler32.o $(SRCDIR)adler32.c
-@mv objs/adler32.o $@
+adler32_power8.lo: $(SRCDIR)contrib/power/adler32_power8.c
+ -@mkdir objs 2>/dev/null || test -d objs
+ $(CC) $(SFLAGS) -mcpu=power8 $(ZINC) -DPIC -c -o objs/adler32_power8.o $(SRCDIR)contrib/power/adler32_power8.c
+ -@mv objs/adler32_power8.o $@
+
crc32.lo: $(SRCDIR)crc32.c
-@mkdir objs 2>/dev/null || test -d objs
$(CC) $(SFLAGS) $(ZINC) -DPIC -c -o objs/crc32.o $(SRCDIR)crc32.c
diff --git a/adler32.c b/adler32.c
index d0be4380a..4bde0fa18 100644
--- a/adler32.c
+++ b/adler32.c
@@ -131,6 +131,12 @@ uLong ZEXPORT adler32_z(adler, buf, len)
}
/* ========================================================================= */
+
+#ifdef Z_POWER_OPT
+/* Rename the default function to avoid naming conflicts */
+#define adler32 adler32_default
+#endif /* Z_POWER_OPT */
+
uLong ZEXPORT adler32(adler, buf, len)
uLong adler;
const Bytef *buf;
@@ -139,6 +145,11 @@ uLong ZEXPORT adler32(adler, buf, len)
return adler32_z(adler, buf, len);
}
+#ifdef Z_POWER_OPT
+#undef adler32
+#include "contrib/power/adler32_resolver.c"
+#endif /* Z_POWER_OPT */
+
/* ========================================================================= */
local uLong adler32_combine_(adler1, adler2, len2)
uLong adler1;
diff --git a/configure b/configure
index 914d9f4aa..810a7404d 100755
--- a/configure
+++ b/configure
@@ -879,8 +879,8 @@ if tryboth $CC -c $CFLAGS $test.c; then
if tryboth $CC -c $CFLAGS -mcpu=power8 $test.c; then
POWER8="-DZ_POWER8"
- PIC_OBJC="${PIC_OBJC} crc32_z_power8.lo"
- OBJC="${OBJC} crc32_z_power8.o"
+ PIC_OBJC="${PIC_OBJC} adler32_power8.lo crc32_z_power8.lo"
+ OBJC="${OBJC} adler32_power8.o crc32_z_power8.o"
echo "Checking for -mcpu=power8 support... Yes." | tee -a configure.log
else
echo "Checking for -mcpu=power8 support... No." | tee -a configure.log
diff --git a/contrib/power/adler32_power8.c b/contrib/power/adler32_power8.c
new file mode 100644
index 000000000..473c39457
--- /dev/null
+++ b/contrib/power/adler32_power8.c
@@ -0,0 +1,196 @@
+/*
+ * Adler32 for POWER 8+ using VSX instructions.
+ *
+ * Calculate adler32 checksum for 16 bytes at once using POWER8+ VSX (vector)
+ * instructions.
+ *
+ * If adler32 do 1 byte at time on the first iteration s1 is s1_0 (_n means
+ * iteration n) is the initial value of adler - at start _0 is 1 unless
+ * adler initial value is different than 1. So s1_1 = s1_0 + c[0] after
+ * the first calculation. For the iteration s1_2 = s1_1 + c[1] and so on.
+ * Hence, for iteration N, s1_N = s1_(N-1) + c[N] is the value of s1 on
+ * after iteration N.
+ *
+ * Therefore, for s2 and iteration N, s2_N = s2_0 + N*s1_N + N*c[0] +
+ * N-1*c[1] + ... + c[N]
+ *
+ * In a more general way:
+ *
+ * s1_N = s1_0 + sum(i=1 to N)c[i]
+ * s2_N = s2_0 + N*s1 + sum (i=1 to N)(N-i+1)*c[i]
+ *
+ * Where s1_N, s2_N are the values for s1, s2 after N iterations. So if we
+ * can process N-bit at time we can do this at once.
+ *
+ * Since VSX can support 16-bit vector instructions, we can process
+ * 16-bit at time using N = 16 we have:
+ *
+ * s1 = s1_16 = s1_(16-1) + c[16] = s1_0 + sum(i=1 to 16)c[i]
+ * s2 = s2_16 = s2_0 + 16*s1 + sum(i=1 to 16)(16-i+1)*c[i]
+ *
+ * After the first iteration we calculate the adler32 checksum for 16 bytes.
+ *
+ * For more background about adler32 please check the RFC:
+ * https://www.ietf.org/rfc/rfc1950.txt
+ *
+ * Copyright (C) 2019 Rogerio Alves <rcardoso@linux.ibm.com>, IBM
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ *
+ */
+
+#include "../../zutil.h"
+#include <altivec.h>
+
+/* Largest prime smaller than 65536. */
+#define BASE 65521U
+#define NMAX 5552
+/* NMAX is the largest n such that 255n(n+1)/2 + (n+1)(BASE-1) <= 2^32-1. */
+
+#define DO1(s1,s2,buf,i) {(s1) += buf[(i)]; (s2) += (s1);}
+#define DO2(s1,s2,buf,i) {DO1(s1,s2,buf,i); DO1(s1,s2,buf,i+1);}
+#define DO4(s1,s2,buf,i) {DO2(s1,s2,buf,i); DO2(s1,s2,buf,i+2);}
+#define DO8(s1,s2,buf,i) {DO4(s1,s2,buf,i); DO4(s1,s2,buf,i+4);}
+#define DO16(s1,s2,buf) {DO8(s1,s2,buf,0); DO8(s1,s2,buf,8);}
+
+/* Vector across sum unsigned int (saturate). */
+inline vector unsigned int vec_sumsu (vector unsigned int __a,
+ vector unsigned int __b)
+{
+ __b = vec_sld(__a, __a, 8);
+ __b = vec_add(__b, __a);
+ __a = vec_sld(__b, __b, 4);
+ __a = vec_add(__a, __b);
+
+ return __a;
+}
+
+uLong ZLIB_INTERNAL _adler32_power8 (uLong adler, const Bytef* buf, uInt len)
+{
+ /* If buffer is empty or len=0 we need to return adler initial value. */
+ if (buf == NULL)
+ return 1;
+
+ unsigned int s1 = adler & 0xffff;
+ unsigned int s2 = (adler >> 16) & 0xffff;
+
+ /* in case user likes doing a byte at a time, keep it fast */
+ if (len == 1) {
+ s1 += buf[0];
+ if (s1 >= BASE)
+ s1 -= BASE;
+ s2 += s1;
+ if (s2 >= BASE)
+ s2 -= BASE;
+ return (s2 << 16) | s1;
+ }
+
+ /* Keep it fast for short length buffers. */
+ if (len < 16) {
+ while (len--) {
+ s1 += *buf++;
+ s2 += s1;
+ }
+ if (s1 >= BASE)
+ s1 -= BASE;
+ s2 %= BASE;
+ return (s2 << 16) | s1;
+ }
+
+ /* This is faster than VSX code for len < 64. */
+ if (len < 64) {
+ while (len >= 16) {
+ len -= 16;
+ DO16(s1,s2,buf);
+ buf += 16;
+ }
+ } else {
+ /* Use POWER VSX instructions for len >= 64. */
+ const vector unsigned int v_zeros = { 0 };
+ const vector unsigned char v_mul = {16, 15, 14, 13, 12, 11, 10, 9, 8, 7,
+ 6, 5, 4, 3, 2, 1};
+ const vector unsigned char vsh = vec_splat_u8(4);
+ const vector unsigned int vmask = {0xffffffff, 0x0, 0x0, 0x0};
+ vector unsigned int vs1 = vec_xl(0, &s1);
+ vector unsigned int vs2 = vec_xl(0, &s2);
+ vector unsigned int vs1_save = { 0 };
+ vector unsigned int vsum1, vsum2;
+ vector unsigned char vbuf;
+ int n;
+
+ /* Zeros the undefined values of vectors vs1, vs2. */
+ vs1 = vec_and(vs1, vmask);
+ vs2 = vec_and(vs2, vmask);
+
+ /* Do length bigger than NMAX in blocks of NMAX size. */
+ while (len >= NMAX) {
+ len -= NMAX;
+ n = NMAX / 16;
+ do {
+ vbuf = vec_xl(0, (unsigned char *) buf);
+ vsum1 = vec_sum4s(vbuf, v_zeros); /* sum(i=1 to 16) buf[i]. */
+ /* sum(i=1 to 16) buf[i]*(16-i+1). */
+ vsum2 = vec_msum(vbuf, v_mul, v_zeros);
+ /* Save vs1. */
+ vs1_save = vec_add(vs1_save, vs1);
+ /* Accumulate the sums. */
+ vs1 = vec_add(vsum1, vs1);
+ vs2 = vec_add(vsum2, vs2);
+
+ buf += 16;
+ } while (--n);
+ /* Once each block of NMAX size. */
+ vs1 = vec_sumsu(vs1, vsum1);
+ vs1_save = vec_sll(vs1_save, vsh); /* 16*vs1_save. */
+ vs2 = vec_add(vs1_save, vs2);
+ vs2 = vec_sumsu(vs2, vsum2);
+
+ /* vs1[0] = (s1_i + sum(i=1 to 16)buf[i]) mod 65521. */
+ vs1[0] = vs1[0] % BASE;
+ /* vs2[0] = s2_i + 16*s1_save +
+ sum(i=1 to 16)(16-i+1)*buf[i] mod 65521. */
+ vs2[0] = vs2[0] % BASE;
+
+ vs1 = vec_and(vs1, vmask);
+ vs2 = vec_and(vs2, vmask);
+ vs1_save = v_zeros;
+ }
+
+ /* len is less than NMAX one modulo is needed. */
+ if (len >= 16) {
+ while (len >= 16) {
+ len -= 16;
+
+ vbuf = vec_xl(0, (unsigned char *) buf);
+
+ vsum1 = vec_sum4s(vbuf, v_zeros); /* sum(i=1 to 16) buf[i]. */
+ /* sum(i=1 to 16) buf[i]*(16-i+1). */
+ vsum2 = vec_msum(vbuf, v_mul, v_zeros);
+ /* Save vs1. */
+ vs1_save = vec_add(vs1_save, vs1);
+ /* Accumulate the sums. */
+ vs1 = vec_add(vsum1, vs1);
+ vs2 = vec_add(vsum2, vs2);
+
+ buf += 16;
+ }
+ /* Since the size will be always less than NMAX we do this once. */
+ vs1 = vec_sumsu(vs1, vsum1);
+ vs1_save = vec_sll(vs1_save, vsh); /* 16*vs1_save. */
+ vs2 = vec_add(vs1_save, vs2);
+ vs2 = vec_sumsu(vs2, vsum2);
+ }
+ /* Copy result back to s1, s2 (mod 65521). */
+ s1 = vs1[0] % BASE;
+ s2 = vs2[0] % BASE;
+ }
+
+ /* Process tail (len < 16). */
+ while (len--) {
+ s1 += *buf++;
+ s2 += s1;
+ }
+ s1 %= BASE;
+ s2 %= BASE;
+
+ return (s2 << 16) | s1;
+}
diff --git a/contrib/power/adler32_resolver.c b/contrib/power/adler32_resolver.c
new file mode 100644
index 000000000..07a1a2cb2
--- /dev/null
+++ b/contrib/power/adler32_resolver.c
@@ -0,0 +1,15 @@
+/* Copyright (C) 2019 Rogerio Alves <rcardoso@linux.ibm.com>, IBM
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "../gcc/zifunc.h"
+#include "power.h"
+
+Z_IFUNC(adler32) {
+#ifdef Z_POWER8
+ if (__builtin_cpu_supports("arch_2_07"))
+ return _adler32_power8;
+#endif
+
+ return adler32_default;
+}
diff --git a/contrib/power/power.h b/contrib/power/power.h
index 79123aa90..f57c76167 100644
--- a/contrib/power/power.h
+++ b/contrib/power/power.h
@@ -2,7 +2,9 @@
* 2019 Rogerio Alves <rogerio.alves@ibm.com>, IBM
* For conditions of distribution and use, see copyright notice in zlib.h
*/
-
#include "../../zconf.h"
+#include "../../zutil.h"
+
+uLong _adler32_power8(uLong adler, const Bytef* buf, uInt len);
unsigned long _crc32_z_power8(unsigned long, const Bytef *, z_size_t);

View File

@ -1,34 +0,0 @@
From 11b722e4ae91b611f605221587ec8e0829c27949 Mon Sep 17 00:00:00 2001
From: Matheus Castanho <msc@linux.ibm.com>
Date: Tue, 23 Jun 2020 10:26:19 -0300
Subject: [PATCH] Fix invalid memory access on ppc and ppc64
---
contrib/power/adler32_power8.c | 9 ++++-----
1 file changed, 4 insertions(+), 5 deletions(-)
diff --git a/contrib/power/adler32_power8.c b/contrib/power/adler32_power8.c
index 473c39457..fdd086453 100644
--- a/contrib/power/adler32_power8.c
+++ b/contrib/power/adler32_power8.c
@@ -110,16 +110,15 @@ uLong ZLIB_INTERNAL _adler32_power8 (uLong adler, const Bytef* buf, uInt len)
6, 5, 4, 3, 2, 1};
const vector unsigned char vsh = vec_splat_u8(4);
const vector unsigned int vmask = {0xffffffff, 0x0, 0x0, 0x0};
- vector unsigned int vs1 = vec_xl(0, &s1);
- vector unsigned int vs2 = vec_xl(0, &s2);
+ vector unsigned int vs1 = { 0 };
+ vector unsigned int vs2 = { 0 };
vector unsigned int vs1_save = { 0 };
vector unsigned int vsum1, vsum2;
vector unsigned char vbuf;
int n;
- /* Zeros the undefined values of vectors vs1, vs2. */
- vs1 = vec_and(vs1, vmask);
- vs2 = vec_and(vs2, vmask);
+ vs1[0] = s1;
+ vs2[0] = s2;
/* Do length bigger than NMAX in blocks of NMAX size. */
while (len >= NMAX) {

View File

@ -1,423 +0,0 @@
From 957bc67cfb4e01403c01fe6243850383183a7c19 Mon Sep 17 00:00:00 2001
From: Ilya Leoshkevich <iii@linux.ibm.com>
Date: Thu, 19 Mar 2020 11:52:03 +0100
Subject: [PATCH] s390x: vectorize crc32
Use vector extensions when compiling for s390x and binutils knows
about them. At runtime, check whether kernel supports vector
extensions (it has to be not just the CPU, but also the kernel) and
choose between the regular and the vectorized implementations.
---
Makefile.in | 9 ++
configure | 28 +++++
contrib/gcc/zifunc.h | 21 +++-
contrib/s390/crc32-vx.c | 195 ++++++++++++++++++++++++++++++++
contrib/s390/crc32_z_resolver.c | 41 +++++++
crc32.c | 11 +-
6 files changed, 301 insertions(+), 4 deletions(-)
create mode 100644 contrib/s390/crc32-vx.c
create mode 100644 contrib/s390/crc32_z_resolver.c
Index: zlib-1.2.13/Makefile.in
===================================================================
--- zlib-1.2.13.orig/Makefile.in
+++ zlib-1.2.13/Makefile.in
@@ -25,6 +25,7 @@ LDFLAGS=
TEST_LDFLAGS=$(LDFLAGS) -L. libz.a
LDSHARED=$(CC)
CPP=$(CC) -E
+VGFMAFLAG=
STATICLIB=libz.a
SHAREDLIB=libz.so
@@ -175,6 +176,9 @@ crc32.o: $(SRCDIR)crc32.c
crc32_z_power8.o: $(SRCDIR)contrib/power/crc32_z_power8.c
$(CC) $(CFLAGS) -mcpu=power8 $(ZINC) -c -o $@ $(SRCDIR)contrib/power/crc32_z_power8.c
+crc32-vx.o: $(SRCDIR)contrib/s390/crc32-vx.c
+ $(CC) $(CFLAGS) $(VGFMAFLAG) $(ZINC) -c -o $@ $(SRCDIR)contrib/s390/crc32-vx.c
+
deflate.o: $(SRCDIR)deflate.c
$(CC) $(CFLAGS) $(ZINC) -c -o $@ $(SRCDIR)deflate.c
@@ -225,6 +229,11 @@ crc32.lo: $(SRCDIR)crc32.c
$(CC) $(SFLAGS) $(ZINC) -DPIC -c -o objs/crc32.o $(SRCDIR)crc32.c
-@mv objs/crc32.o $@
+crc32-vx.lo: $(SRCDIR)contrib/s390/crc32-vx.c
+ -@mkdir objs 2>/dev/null || test -d objs
+ $(CC) $(SFLAGS) $(VGFMAFLAG) $(ZINC) -DPIC -c -o objs/crc32-vx.o $(SRCDIR)contrib/s390/crc32-vx.c
+ -@mv objs/crc32-vx.o $@
+
crc32_z_power8.lo: $(SRCDIR)contrib/power/crc32_z_power8.c
-@mkdir objs 2>/dev/null || test -d objs
$(CC) $(SFLAGS) -mcpu=power8 $(ZINC) -DPIC -c -o objs/crc32_z_power8.o $(SRCDIR)contrib/power/crc32_z_power8.c
Index: zlib-1.2.13/configure
===================================================================
--- zlib-1.2.13.orig/configure
+++ zlib-1.2.13/configure
@@ -921,6 +921,32 @@ else
echo "Checking for Power optimizations support... No." | tee -a configure.log
fi
+# check if we are compiling for s390 and binutils support vector extensions
+VGFMAFLAG=-march=z13
+cat > $test.c <<EOF
+#ifndef __s390__
+#error
+#endif
+EOF
+if try $CC -c $CFLAGS $VGFMAFLAG $test.c; then
+ CFLAGS="$CFLAGS -DHAVE_S390X_VX"
+ SFLAGS="$SFLAGS -DHAVE_S390X_VX"
+ OBJC="$OBJC crc32-vx.o"
+ PIC_OBJC="$PIC_OBJC crc32-vx.lo"
+ echo "Checking for s390 vector extensions... Yes." | tee -a configure.log
+
+ for flag in -mzarch -fzvector; do
+ if try $CC -c $CFLAGS $VGFMAFLAG $flag $test.c; then
+ VGFMAFLAG="$VGFMAFLAG $flag"
+ echo "Checking for $flag... Yes." | tee -a configure.log
+ else
+ echo "Checking for $flag... No." | tee -a configure.log
+ fi
+ done
+else
+ echo "Checking for s390 vector extensions... No." | tee -a configure.log
+fi
+
# show the results in the log
echo >> configure.log
echo ALL = $ALL >> configure.log
@@ -952,6 +978,7 @@ echo mandir = $mandir >> configure.log
echo prefix = $prefix >> configure.log
echo sharedlibdir = $sharedlibdir >> configure.log
echo uname = $uname >> configure.log
+echo VGFMAFLAG = $VGFMAFLAG >> configure.log
# udpate Makefile with the configure results
sed < ${SRCDIR}Makefile.in "
@@ -961,6 +988,7 @@ sed < ${SRCDIR}Makefile.in "
/^LDFLAGS *=/s#=.*#=$LDFLAGS#
/^LDSHARED *=/s#=.*#=$LDSHARED#
/^CPP *=/s#=.*#=$CPP#
+/^VGFMAFLAG *=/s#=.*#=$VGFMAFLAG#
/^STATICLIB *=/s#=.*#=$STATICLIB#
/^SHAREDLIB *=/s#=.*#=$SHAREDLIB#
/^SHAREDLIBV *=/s#=.*#=$SHAREDLIBV#
Index: zlib-1.2.13/contrib/gcc/zifunc.h
===================================================================
--- zlib-1.2.13.orig/contrib/gcc/zifunc.h
+++ zlib-1.2.13/contrib/gcc/zifunc.h
@@ -8,9 +8,28 @@
/* Helpers for arch optimizations */
+#if defined(__clang__)
+#if __has_feature(coverage_sanitizer)
+#define Z_IFUNC_NO_SANCOV __attribute__((no_sanitize("coverage")))
+#else /* __has_feature(coverage_sanitizer) */
+#define Z_IFUNC_NO_SANCOV
+#endif /* __has_feature(coverage_sanitizer) */
+#else /* __clang__ */
+#define Z_IFUNC_NO_SANCOV
+#endif /* __clang__ */
+
+#ifdef __s390__
+#define Z_IFUNC_PARAMS unsigned long hwcap
+#define Z_IFUNC_ATTRS Z_IFUNC_NO_SANCOV
+#else /* __s390__ */
+#define Z_IFUNC_PARAMS void
+#define Z_IFUNC_ATTRS
+#endif /* __s390__ */
+
#define Z_IFUNC(fname) \
typeof(fname) fname __attribute__ ((ifunc (#fname "_resolver"))); \
- local typeof(fname) *fname##_resolver(void)
+ Z_IFUNC_ATTRS \
+ local typeof(fname) *fname##_resolver(Z_IFUNC_PARAMS)
/* This is a helper macro to declare a resolver for an indirect function
* (ifunc). Let's say you have function
*
Index: zlib-1.2.13/contrib/s390/crc32-vx.c
===================================================================
--- /dev/null
+++ zlib-1.2.13/contrib/s390/crc32-vx.c
@@ -0,0 +1,195 @@
+/*
+ * Hardware-accelerated CRC-32 variants for Linux on z Systems
+ *
+ * Use the z/Architecture Vector Extension Facility to accelerate the
+ * computing of bitreflected CRC-32 checksums.
+ *
+ * This CRC-32 implementation algorithm is bitreflected and processes
+ * the least-significant bit first (Little-Endian).
+ *
+ * This code was originally written by Hendrik Brueckner
+ * <brueckner@linux.vnet.ibm.com> for use in the Linux kernel and has been
+ * relicensed under the zlib license.
+ */
+
+#include "../../zutil.h"
+
+#include <stdint.h>
+#include <vecintrin.h>
+
+typedef unsigned char uv16qi __attribute__((vector_size(16)));
+typedef unsigned int uv4si __attribute__((vector_size(16)));
+typedef unsigned long long uv2di __attribute__((vector_size(16)));
+
+uint32_t crc32_le_vgfm_16(uint32_t crc, const unsigned char *buf, size_t len) {
+ /*
+ * The CRC-32 constant block contains reduction constants to fold and
+ * process particular chunks of the input data stream in parallel.
+ *
+ * For the CRC-32 variants, the constants are precomputed according to
+ * these definitions:
+ *
+ * R1 = [(x4*128+32 mod P'(x) << 32)]' << 1
+ * R2 = [(x4*128-32 mod P'(x) << 32)]' << 1
+ * R3 = [(x128+32 mod P'(x) << 32)]' << 1
+ * R4 = [(x128-32 mod P'(x) << 32)]' << 1
+ * R5 = [(x64 mod P'(x) << 32)]' << 1
+ * R6 = [(x32 mod P'(x) << 32)]' << 1
+ *
+ * The bitreflected Barret reduction constant, u', is defined as
+ * the bit reversal of floor(x**64 / P(x)).
+ *
+ * where P(x) is the polynomial in the normal domain and the P'(x) is the
+ * polynomial in the reversed (bitreflected) domain.
+ *
+ * CRC-32 (IEEE 802.3 Ethernet, ...) polynomials:
+ *
+ * P(x) = 0x04C11DB7
+ * P'(x) = 0xEDB88320
+ */
+ const uv16qi perm_le2be = {15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0}; /* BE->LE mask */
+ const uv2di r2r1 = {0x1C6E41596, 0x154442BD4}; /* R2, R1 */
+ const uv2di r4r3 = {0x0CCAA009E, 0x1751997D0}; /* R4, R3 */
+ const uv2di r5 = {0, 0x163CD6124}; /* R5 */
+ const uv2di ru_poly = {0, 0x1F7011641}; /* u' */
+ const uv2di crc_poly = {0, 0x1DB710641}; /* P'(x) << 1 */
+
+ /*
+ * Load the initial CRC value.
+ *
+ * The CRC value is loaded into the rightmost word of the
+ * vector register and is later XORed with the LSB portion
+ * of the loaded input data.
+ */
+ uv2di v0 = {0, 0};
+ v0 = (uv2di)vec_insert(crc, (uv4si)v0, 3);
+
+ /* Load a 64-byte data chunk and XOR with CRC */
+ uv2di v1 = vec_perm(((uv2di *)buf)[0], ((uv2di *)buf)[0], perm_le2be);
+ uv2di v2 = vec_perm(((uv2di *)buf)[1], ((uv2di *)buf)[1], perm_le2be);
+ uv2di v3 = vec_perm(((uv2di *)buf)[2], ((uv2di *)buf)[2], perm_le2be);
+ uv2di v4 = vec_perm(((uv2di *)buf)[3], ((uv2di *)buf)[3], perm_le2be);
+
+ v1 ^= v0;
+ buf += 64;
+ len -= 64;
+
+ while (len >= 64) {
+ /* Load the next 64-byte data chunk */
+ uv16qi part1 = vec_perm(((uv16qi *)buf)[0], ((uv16qi *)buf)[0], perm_le2be);
+ uv16qi part2 = vec_perm(((uv16qi *)buf)[1], ((uv16qi *)buf)[1], perm_le2be);
+ uv16qi part3 = vec_perm(((uv16qi *)buf)[2], ((uv16qi *)buf)[2], perm_le2be);
+ uv16qi part4 = vec_perm(((uv16qi *)buf)[3], ((uv16qi *)buf)[3], perm_le2be);
+
+ /*
+ * Perform a GF(2) multiplication of the doublewords in V1 with
+ * the R1 and R2 reduction constants in V0. The intermediate result
+ * is then folded (accumulated) with the next data chunk in PART1 and
+ * stored in V1. Repeat this step for the register contents
+ * in V2, V3, and V4 respectively.
+ */
+ v1 = (uv2di)vec_gfmsum_accum_128(r2r1, v1, part1);
+ v2 = (uv2di)vec_gfmsum_accum_128(r2r1, v2, part2);
+ v3 = (uv2di)vec_gfmsum_accum_128(r2r1, v3, part3);
+ v4 = (uv2di)vec_gfmsum_accum_128(r2r1, v4, part4);
+
+ buf += 64;
+ len -= 64;
+ }
+
+ /*
+ * Fold V1 to V4 into a single 128-bit value in V1. Multiply V1 with R3
+ * and R4 and accumulating the next 128-bit chunk until a single 128-bit
+ * value remains.
+ */
+ v1 = (uv2di)vec_gfmsum_accum_128(r4r3, v1, (uv16qi)v2);
+ v1 = (uv2di)vec_gfmsum_accum_128(r4r3, v1, (uv16qi)v3);
+ v1 = (uv2di)vec_gfmsum_accum_128(r4r3, v1, (uv16qi)v4);
+
+ while (len >= 16) {
+ /* Load next data chunk */
+ v2 = vec_perm(*(uv2di *)buf, *(uv2di *)buf, perm_le2be);
+
+ /* Fold next data chunk */
+ v1 = (uv2di)vec_gfmsum_accum_128(r4r3, v1, (uv16qi)v2);
+
+ buf += 16;
+ len -= 16;
+ }
+
+ /*
+ * Set up a vector register for byte shifts. The shift value must
+ * be loaded in bits 1-4 in byte element 7 of a vector register.
+ * Shift by 8 bytes: 0x40
+ * Shift by 4 bytes: 0x20
+ */
+ uv16qi v9 = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+ v9 = vec_insert((unsigned char)0x40, v9, 7);
+
+ /*
+ * Prepare V0 for the next GF(2) multiplication: shift V0 by 8 bytes
+ * to move R4 into the rightmost doubleword and set the leftmost
+ * doubleword to 0x1.
+ */
+ v0 = vec_srb(r4r3, (uv2di)v9);
+ v0[0] = 1;
+
+ /*
+ * Compute GF(2) product of V1 and V0. The rightmost doubleword
+ * of V1 is multiplied with R4. The leftmost doubleword of V1 is
+ * multiplied by 0x1 and is then XORed with rightmost product.
+ * Implicitly, the intermediate leftmost product becomes padded
+ */
+ v1 = (uv2di)vec_gfmsum_128(v0, v1);
+
+ /*
+ * Now do the final 32-bit fold by multiplying the rightmost word
+ * in V1 with R5 and XOR the result with the remaining bits in V1.
+ *
+ * To achieve this by a single VGFMAG, right shift V1 by a word
+ * and store the result in V2 which is then accumulated. Use the
+ * vector unpack instruction to load the rightmost half of the
+ * doubleword into the rightmost doubleword element of V1; the other
+ * half is loaded in the leftmost doubleword.
+ * The vector register with CONST_R5 contains the R5 constant in the
+ * rightmost doubleword and the leftmost doubleword is zero to ignore
+ * the leftmost product of V1.
+ */
+ v9 = vec_insert((unsigned char)0x20, v9, 7);
+ v2 = vec_srb(v1, (uv2di)v9);
+ v1 = vec_unpackl((uv4si)v1); /* Split rightmost doubleword */
+ v1 = (uv2di)vec_gfmsum_accum_128(r5, v1, (uv16qi)v2);
+
+ /*
+ * Apply a Barret reduction to compute the final 32-bit CRC value.
+ *
+ * The input values to the Barret reduction are the degree-63 polynomial
+ * in V1 (R(x)), degree-32 generator polynomial, and the reduction
+ * constant u. The Barret reduction result is the CRC value of R(x) mod
+ * P(x).
+ *
+ * The Barret reduction algorithm is defined as:
+ *
+ * 1. T1(x) = floor( R(x) / x^32 ) GF2MUL u
+ * 2. T2(x) = floor( T1(x) / x^32 ) GF2MUL P(x)
+ * 3. C(x) = R(x) XOR T2(x) mod x^32
+ *
+ * Note: The leftmost doubleword of vector register containing
+ * CONST_RU_POLY is zero and, thus, the intermediate GF(2) product
+ * is zero and does not contribute to the final result.
+ */
+
+ /* T1(x) = floor( R(x) / x^32 ) GF2MUL u */
+ v2 = vec_unpackl((uv4si)v1);
+ v2 = (uv2di)vec_gfmsum_128(ru_poly, v2);
+
+ /*
+ * Compute the GF(2) product of the CRC polynomial with T1(x) in
+ * V2 and XOR the intermediate result, T2(x), with the value in V1.
+ * The final result is stored in word element 2 of V2.
+ */
+ v2 = vec_unpackl((uv4si)v2);
+ v2 = (uv2di)vec_gfmsum_accum_128(crc_poly, v2, (uv16qi)v1);
+
+ return ((uv4si)v2)[2];
+}
Index: zlib-1.2.13/contrib/s390/crc32_z_resolver.c
===================================================================
--- /dev/null
+++ zlib-1.2.13/contrib/s390/crc32_z_resolver.c
@@ -0,0 +1,41 @@
+#include <sys/auxv.h>
+#include "../gcc/zifunc.h"
+
+#define VX_MIN_LEN 64
+#define VX_ALIGNMENT 16L
+#define VX_ALIGN_MASK (VX_ALIGNMENT - 1)
+
+unsigned int crc32_le_vgfm_16(unsigned int crc, const unsigned char FAR *buf, z_size_t len);
+
+local unsigned long s390_crc32_vx(unsigned long crc, const unsigned char FAR *buf, z_size_t len)
+{
+ uintptr_t prealign, aligned, remaining;
+
+ if (buf == Z_NULL) return 0UL;
+
+ if (len < VX_MIN_LEN + VX_ALIGN_MASK)
+ return crc32_z_default(crc, buf, len);
+
+ if ((uintptr_t)buf & VX_ALIGN_MASK) {
+ prealign = VX_ALIGNMENT - ((uintptr_t)buf & VX_ALIGN_MASK);
+ len -= prealign;
+ crc = crc32_z_default(crc, buf, prealign);
+ buf += prealign;
+ }
+ aligned = len & ~VX_ALIGN_MASK;
+ remaining = len & VX_ALIGN_MASK;
+
+ crc = crc32_le_vgfm_16(crc ^ 0xffffffff, buf, (size_t)aligned) ^ 0xffffffff;
+
+ if (remaining)
+ crc = crc32_z_default(crc, buf + aligned, remaining);
+
+ return crc;
+}
+
+Z_IFUNC(crc32_z)
+{
+ if (hwcap & HWCAP_S390_VX)
+ return s390_crc32_vx;
+ return crc32_z_default;
+}
Index: zlib-1.2.13/crc32.c
===================================================================
--- zlib-1.2.13.orig/crc32.c
+++ zlib-1.2.13/crc32.c
@@ -745,12 +745,12 @@ local z_word_t crc_word_big(data)
#endif
/* ========================================================================= */
-#ifdef Z_POWER_OPT
+#if defined(Z_POWER_OPT) || defined(HAVE_S390X_VX)
/* Rename function so resolver can use its symbol. The default version will be
* returned by the resolver if the host has no support for an optimized version.
*/
#define crc32_z crc32_z_default
-#endif /* Z_POWER_OPT */
+#endif /* defined(Z_POWER_OPT) || defined(HAVE_S390X_VX) */
unsigned long ZEXPORT crc32_z(crc, buf, len)
unsigned long crc;
@@ -1073,10 +1073,15 @@ unsigned long ZEXPORT crc32_z(crc, buf,
return crc ^ 0xffffffff;
}
-#ifdef Z_POWER_OPT
+#if defined(Z_POWER_OPT) || defined(HAVE_S390X_VX)
#undef crc32_z
+#ifdef Z_POWER_OPT
#include "contrib/power/crc32_z_resolver.c"
#endif /* Z_POWER_OPT */
+#ifdef HAVE_S390X_VX
+#include "contrib/s390/crc32_z_resolver.c"
+#endif /* HAVE_S390X_VX */
+#endif /* defined(Z_POWER_OPT) || defined(HAVE_S390X_VX) */
#endif

View File

@ -1,27 +0,0 @@
From e554695638228b846d49657f31eeff0ca4680e8a Mon Sep 17 00:00:00 2001
From: Mark Adler <madler@alumni.caltech.edu>
Date: Thu, 15 Dec 2022 09:07:13 -0800
Subject: [PATCH] Fix bug in deflateBound() for level 0 and memLevel 9.
memLevel 9 would cause deflateBound() to assume the use of fixed
blocks, even if the compression level was 0, which forces stored
blocks. That could result in a bound less than the size of the
compressed data. Now level 0 always uses the stored blocks bound.
---
deflate.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/deflate.c b/deflate.c
index cd538b8ac..4a512e1f9 100644
--- a/deflate.c
+++ b/deflate.c
@@ -752,7 +752,8 @@ uLong ZEXPORT deflateBound(strm, sourceLen)
/* if not default parameters, return one of the conservative bounds */
if (s->w_bits != 15 || s->hash_bits != 8 + 7)
- return (s->w_bits <= s->hash_bits ? fixedlen : storelen) + wraplen;
+ return (s->w_bits <= s->hash_bits && s->level ? fixedlen : storelen) +
+ wraplen;
/* default settings: return tight bound for that case -- ~0.03% overhead
plus a small constant */

View File

@ -2,16 +2,12 @@ Index: deflate.c
=================================================================== ===================================================================
--- deflate.c.orig --- deflate.c.orig
+++ deflate.c +++ deflate.c
@@ -1233,15 +1233,16 @@ local void lm_init (s) @@ -1233,12 +1233,13 @@ local void lm_init (s)
* string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1 * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
* OUT assertion: the match length is not greater than s->lookahead. * OUT assertion: the match length is not greater than s->lookahead.
*/ */
-local uInt longest_match(s, cur_match) -local uInt longest_match(deflate_state *s, IPos cur_match) {
+local uInt longest_match(s, pcur_match) +local uInt longest_match(deflate_state *s, IPos pcur_match) {
deflate_state *s;
- IPos cur_match; /* current match */
+ IPos pcur_match; /* current match */
{
+ ptrdiff_t cur_match = pcur_match; /* extend to pointer width */ + ptrdiff_t cur_match = pcur_match; /* extend to pointer width */
unsigned chain_length = s->max_chain_length;/* max hash chain length */ unsigned chain_length = s->max_chain_length;/* max hash chain length */
register Bytef *scan = s->window + s->strstart; /* current string */ register Bytef *scan = s->window + s->strstart; /* current string */

BIN
zlib-1.2.13.tar.gz (Stored with Git LFS)

Binary file not shown.

View File

@ -1,7 +0,0 @@
-----BEGIN PGP SIGNATURE-----
Comment: GPGTools - http://gpgtools.org
iF0EABECAB0WIQRe1GpnIdNlWHeR4qp4P82OWLyvugUCY0h42QAKCRB4P82OWLyv
upvZAKCF7EgWGaMEfO78WnkA8hivLlBMlACgyI7Vm2A5BI2jI+h23yqrKjgQC5s=
=umRA
-----END PGP SIGNATURE-----

3
zlib-1.3.tar.gz Normal file
View File

@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:ff0ba4c292013dbc27530b3a81e1f9a813cd39de01ca5e0f8bf355702efa593e
size 1495873

7
zlib-1.3.tar.gz.asc Normal file
View File

@ -0,0 +1,7 @@
-----BEGIN PGP SIGNATURE-----
Comment: GPGTools - http://gpgtools.org
iF0EABECAB0WIQRe1GpnIdNlWHeR4qp4P82OWLyvugUCZN8+EgAKCRB4P82OWLyv
usBmAKC6ixPJLSVYgQivrqK4KBw4gTGFGwCgxJ9SfDFGqI3uqjyR99/13L7vn3o=
=TwN5
-----END PGP SIGNATURE-----

View File

@ -9,8 +9,8 @@ Index: zlib.h
is returned, and the error state is set to Z_STREAM_ERROR. is returned, and the error state is set to Z_STREAM_ERROR.
*/ */
-ZEXTERN int ZEXPORTVA gzprintf Z_ARG((gzFile file, const char *format, ...)); -ZEXTERN int ZEXPORTVA gzprintf(gzFile file, const char *format, ...);
+ZEXTERN int ZEXPORTVA gzprintf Z_ARG((gzFile file, const char *format, ...)) +ZEXTERN int ZEXPORTVA gzprintf(gzFile file, const char *format, ...)
+#ifdef __GNUC__ +#ifdef __GNUC__
+ __attribute__((__format__(__printf__,2,3))) + __attribute__((__format__(__printf__,2,3)))
+#endif +#endif

View File

@ -1,9 +1,9 @@
Index: zlib-1.2.12/infback.c Index: zlib-1.3/infback.c
=================================================================== ===================================================================
--- zlib-1.2.12.orig/infback.c --- zlib-1.3.orig/infback.c
+++ zlib-1.2.12/infback.c +++ zlib-1.3/infback.c
@@ -34,9 +34,6 @@ int stream_size; @@ -27,9 +27,6 @@ int ZEXPORT inflateBackInit_(z_streamp s
{ int stream_size) {
struct inflate_state FAR *state; struct inflate_state FAR *state;
- if (version == Z_NULL || version[0] != ZLIB_VERSION[0] || - if (version == Z_NULL || version[0] != ZLIB_VERSION[0] ||
@ -12,11 +12,11 @@ Index: zlib-1.2.12/infback.c
if (strm == Z_NULL || window == Z_NULL || if (strm == Z_NULL || window == Z_NULL ||
windowBits < 8 || windowBits > 15) windowBits < 8 || windowBits > 15)
return Z_STREAM_ERROR; return Z_STREAM_ERROR;
Index: zlib-1.2.12/inflate.c Index: zlib-1.3/inflate.c
=================================================================== ===================================================================
--- zlib-1.2.12.orig/inflate.c --- zlib-1.3.orig/inflate.c
+++ zlib-1.2.12/inflate.c +++ zlib-1.3/inflate.c
@@ -202,9 +202,6 @@ int stream_size; @@ -180,9 +180,6 @@ int ZEXPORT inflateInit2_(z_streamp strm
int ret; int ret;
struct inflate_state FAR *state; struct inflate_state FAR *state;
@ -26,12 +26,12 @@ Index: zlib-1.2.12/inflate.c
if (strm == Z_NULL) return Z_STREAM_ERROR; if (strm == Z_NULL) return Z_STREAM_ERROR;
strm->msg = Z_NULL; /* in case we return an error */ strm->msg = Z_NULL; /* in case we return an error */
if (strm->zalloc == (alloc_func)0) { if (strm->zalloc == (alloc_func)0) {
Index: zlib-1.2.12/deflate.c Index: zlib-1.3/deflate.c
=================================================================== ===================================================================
--- zlib-1.2.12.orig/deflate.c --- zlib-1.3.orig/deflate.c
+++ zlib-1.2.12/deflate.c +++ zlib-1.3/deflate.c
@@ -253,12 +253,7 @@ int ZEXPORT deflateInit2_(strm, level, m @@ -384,12 +384,7 @@ int ZEXPORT deflateInit2_(z_streamp strm
{ const char *version, int stream_size) {
deflate_state *s; deflate_state *s;
int wrap = 1; int wrap = 1;
- static const char my_version[] = ZLIB_VERSION; - static const char my_version[] = ZLIB_VERSION;

View File

@ -1,3 +1,38 @@
-------------------------------------------------------------------
Thu Oct 19 16:00:31 UTC 2023 - Danilo Spinella <oss@danyspin97.org>
- Update to 1.3:
* Building using K&R (pre-ANSI) function definitions is no longer supported.
* Fixed a bug in deflateBound() for level 0 and memLevel 9.
* Fixed a bug when gzungetc() is used immediately after gzopen().
* Fixed a bug when using gzflush() with a very small buffer.
* Fixed a crash when gzsetparams() is attempted for a transparent write.
* Fixed test/example.c to work with FORCE_STORED.
* Fixed minizip to allow it to open an empty zip file.
* Fixed reading disk number start on zip64 files in minizip.
* Fixed a logic error in minizip argument processing.
- Added patches:
* zlib-1.3-IBM-Z-hw-accelerated-deflate-s390x.patch
- Refreshed patches:
* zlib-1.2.12-add-optimized-slide_hash-for-power.patch
* zlib-1.2.12-add-vectorized-longest_match-for-power.patch
* zlib-1.2.12-adler32-vector-optimizations-for-power.patch
* zlib-1.2.13-optimized-s390.patch
* zlib-format.patch
* zlib-no-version-check.patch
- Removed patches:
* bsc1210593.patch
* zlib-1.2.13-fix-bug-deflateBound.patch
* zlib-1.2.12-s390-vectorize-crc32.patch
* zlib-1.2.13-IBM-Z-hw-accelerated-deflate-s390x.patch
* zlib-1.2.12-add-optimized-slide_hash-for-power.patch
* zlib-1.2.12-fix-invalid-memory-access-on-ppc-and-ppc64.patch
* zlib-1.2.12-add-vectorized-longest_match-for-power.patch
* zlib-1.2.12-adler32-vector-optimizations-for-power.patch
- Fix CVE-2023-45853, integer overflow and resultant heap-based buffer
overflow in zipOpenNewFileInZip4_6, bsc#1216378
* CVE-2023-45853.patch
------------------------------------------------------------------- -------------------------------------------------------------------
Fri May 5 09:56:31 UTC 2023 - Danilo Spinella <danilo.spinella@suse.com> Fri May 5 09:56:31 UTC 2023 - Danilo Spinella <danilo.spinella@suse.com>

View File

@ -17,7 +17,7 @@
Name: zlib Name: zlib
Version: 1.2.13 Version: 1.3
Release: 0 Release: 0
Summary: Library implementing the DEFLATE compression algorithm Summary: Library implementing the DEFLATE compression algorithm
License: Zlib License: Zlib
@ -37,25 +37,17 @@ Patch2: 0001-Do-not-try-to-store-negative-values-in-unsigned-int.patch
Patch3: zlib-no-version-check.patch Patch3: zlib-no-version-check.patch
#PATCH-FIX-SUSE https://github.com/madler/zlib/pull/229 #PATCH-FIX-SUSE https://github.com/madler/zlib/pull/229
Patch4: minizip-dont-install-crypt-header.patch Patch4: minizip-dont-install-crypt-header.patch
# PATCH-FIX-UPSTREAM https://github.com/madler/zlib/commit/e554695638228b846d49657f31eeff0ca4680e8a
Patch5: zlib-1.2.13-fix-bug-deflateBound.patch
#PATCH-FIX-SUSE https://github.com/madler/zlib/pull/410 #PATCH-FIX-SUSE https://github.com/madler/zlib/pull/410
Patch6: zlib-1.2.13-IBM-Z-hw-accelerated-deflate-s390x.patch Patch6: zlib-1.3-IBM-Z-hw-accelerated-deflate-s390x.patch
# Patches taken from https://github.com/iii-i/zlib/releases/tag/crc32vx-v3 # Patches taken from https://github.com/iii-i/zlib/releases/tag/crc32vx-v3
Patch7: zlib-1.2.5-minizip-fixuncrypt.patch Patch7: zlib-1.2.5-minizip-fixuncrypt.patch
Patch8: zlib-1.2.13-optimized-s390.patch Patch8: zlib-1.2.13-optimized-s390.patch
# https://github.com/iii-i/zlib/commit/171d0ff3c9ed40da0ac14085ab16b766b1162069 # https://github.com/iii-i/zlib/commit/171d0ff3c9ed40da0ac14085ab16b766b1162069
Patch10: zlib-1.2.11-covscan-issues.patch Patch10: zlib-1.2.11-covscan-issues.patch
Patch11: zlib-1.2.11-covscan-issues-rhel9.patch Patch11: zlib-1.2.11-covscan-issues-rhel9.patch
Patch14: zlib-1.2.12-s390-vectorize-crc32.patch # PATCh-FIX-SECURITY CVE-2023-45853.patch bsc#1216378 CVE-2023-45853 danilo.spinella@suse.com
# The following patches are taken from https://github.com/mscastanho/zlib/commits/power-optimizations-1.2.12 # integer overflow and resultant heap-based buffer overflow in zipOpenNewFileInZip4_6
Patch15: zlib-1.2.12-adler32-vector-optimizations-for-power.patch Patch12: CVE-2023-45853.patch
Patch16: zlib-1.2.12-fix-invalid-memory-access-on-ppc-and-ppc64.patch
Patch17: zlib-1.2.12-add-optimized-slide_hash-for-power.patch
Patch18: zlib-1.2.12-add-vectorized-longest_match-for-power.patch
# PATCH-FIX-UPSTREAM danilo.spinella@suse.com bsc#1210593 bsc#1211005
# Fix deflateBound() before deflateInit()
Patch19: bsc1210593.patch
BuildRequires: autoconf BuildRequires: autoconf
BuildRequires: automake BuildRequires: automake
BuildRequires: libtool BuildRequires: libtool
@ -147,18 +139,12 @@ It should exit 0
%patch2 -p1 %patch2 -p1
%patch3 -p1 %patch3 -p1
%patch4 -p1 %patch4 -p1
%patch5 -p1
%patch6 -p1 %patch6 -p1
%patch7 -p1 %patch7 -p1
%patch8 %patch8
%patch10 -p1 %patch10 -p1
%patch11 -p1 %patch11 -p1
%patch14 -p1 %patch12 -p1
%patch15 -p1
%patch16 -p1
%patch17 -p1
%patch18 -p1
%patch19 -p1
cp %{SOURCE4} . cp %{SOURCE4} .
%build %build
@ -221,7 +207,7 @@ find %{buildroot} -type f -name "*.la" -delete -print
%files -n libz1 %files -n libz1
%license LICENSE %license LICENSE
%{_libdir}/libz.so.1.2.* %{_libdir}/libz.so.1.3
%{_libdir}/libz.so.1 %{_libdir}/libz.so.1
%files devel %files devel