From: Cameron Cawley <ccawley2011@gmail.com>
Date: Fri, 30 Sep 2022 15:06:42 +0000 (+0100)
Subject: Use size_t instead of uint64_t for len in all crc32 functions
X-Git-Tag: 2.1.0-beta1~99
X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=23e43059328b0ae99f86a61aba6d8652d08e3ccc;p=thirdparty%2Fzlib-ng.git

Use size_t instead of uint64_t for len in all crc32 functions
---

diff --git a/arch/arm/crc32_acle.c b/arch/arm/crc32_acle.c
index ef18ae81..701c9925 100644
--- a/arch/arm/crc32_acle.c
+++ b/arch/arm/crc32_acle.c
@@ -13,7 +13,7 @@
 #endif
 #include "../../zbuild.h"
 
-uint32_t crc32_acle(uint32_t crc, const uint8_t *buf, uint64_t len) {
+uint32_t crc32_acle(uint32_t crc, const uint8_t *buf, size_t len) {
     Z_REGISTER uint32_t c;
     Z_REGISTER const uint16_t *buf2;
     Z_REGISTER const uint32_t *buf4;
diff --git a/arch/power/crc32_power8.c b/arch/power/crc32_power8.c
index fbb64721..1cb5f299 100644
--- a/arch/power/crc32_power8.c
+++ b/arch/power/crc32_power8.c
@@ -48,7 +48,7 @@ static unsigned int crc32_align(unsigned int crc, const unsigned char *p, unsign
 
 static unsigned int ALIGNED_(32) __crc32_vpmsum(unsigned int crc, const void* p, unsigned long len);
 
-Z_INTERNAL uint32_t crc32_power8(uint32_t crc, const unsigned char *p, uint64_t _len) {
+Z_INTERNAL uint32_t crc32_power8(uint32_t crc, const unsigned char *p, size_t _len) {
     unsigned int prealign;
     unsigned int tail;
 
diff --git a/arch/s390/crc32-vx.c b/arch/s390/crc32-vx.c
index a64d6291..78c0be5c 100644
--- a/arch/s390/crc32-vx.c
+++ b/arch/s390/crc32-vx.c
@@ -21,7 +21,7 @@ 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)));
 
-static uint32_t crc32_le_vgfm_16(uint32_t crc, const uint8_t *buf, uint64_t len) {
+static uint32_t crc32_le_vgfm_16(uint32_t crc, const uint8_t *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.
@@ -198,8 +198,8 @@ static uint32_t crc32_le_vgfm_16(uint32_t crc, const uint8_t *buf, uint64_t len)
 #define VX_ALIGNMENT 16L
 #define VX_ALIGN_MASK (VX_ALIGNMENT - 1)
 
-uint32_t Z_INTERNAL PREFIX(s390_crc32_vx)(uint32_t crc, const unsigned char *buf, uint64_t len) {
-    uint64_t prealign, aligned, remaining;
+uint32_t Z_INTERNAL PREFIX(s390_crc32_vx)(uint32_t crc, const unsigned char *buf, size_t len) {
+    size_t prealign, aligned, remaining;
 
     if (len < VX_MIN_LEN + VX_ALIGN_MASK)
         return PREFIX(crc32_braid)(crc, buf, len);
@@ -213,7 +213,7 @@ uint32_t Z_INTERNAL PREFIX(s390_crc32_vx)(uint32_t crc, const unsigned char *buf
     aligned = len & ~VX_ALIGN_MASK;
     remaining = len & VX_ALIGN_MASK;
 
-    crc = crc32_le_vgfm_16(crc ^ 0xffffffff, buf, (size_t)aligned) ^ 0xffffffff;
+    crc = crc32_le_vgfm_16(crc ^ 0xffffffff, buf, aligned) ^ 0xffffffff;
 
     if (remaining)
         crc = PREFIX(crc32_braid)(crc, buf + aligned, remaining);
diff --git a/arch/x86/crc32_fold_pclmulqdq.c b/arch/x86/crc32_fold_pclmulqdq.c
index c6f04842..ecee0c57 100644
--- a/arch/x86/crc32_fold_pclmulqdq.c
+++ b/arch/x86/crc32_fold_pclmulqdq.c
@@ -32,11 +32,11 @@
 #include <assert.h>
 
 #ifdef X86_VPCLMULQDQ_CRC
-extern uint64_t fold_16_vpclmulqdq(__m128i *xmm_crc0, __m128i *xmm_crc1,
-    __m128i *xmm_crc2, __m128i *xmm_crc3, const uint8_t *src, uint64_t len, __m128i init_crc,
+extern size_t fold_16_vpclmulqdq(__m128i *xmm_crc0, __m128i *xmm_crc1,
+    __m128i *xmm_crc2, __m128i *xmm_crc3, const uint8_t *src, size_t len, __m128i init_crc,
     int32_t first);
-extern uint64_t fold_16_vpclmulqdq_copy(__m128i *xmm_crc0, __m128i *xmm_crc1,
-    __m128i *xmm_crc2, __m128i *xmm_crc3, uint8_t *dst, const uint8_t *src, uint64_t len);
+extern size_t fold_16_vpclmulqdq_copy(__m128i *xmm_crc0, __m128i *xmm_crc1,
+    __m128i *xmm_crc2, __m128i *xmm_crc3, uint8_t *dst, const uint8_t *src, size_t len);
 #endif
 
 static void fold_1(__m128i *xmm_crc0, __m128i *xmm_crc1, __m128i *xmm_crc2, __m128i *xmm_crc3) {
@@ -185,7 +185,7 @@ static const unsigned ALIGNED_(32) pshufb_shf_table[60] = {
     0x0201008f, 0x06050403, 0x0a090807, 0x0e0d0c0b  /* shl  1 (16 -15)/shr15*/
 };
 
-static void partial_fold(const uint64_t len, __m128i *xmm_crc0, __m128i *xmm_crc1, __m128i *xmm_crc2,
+static void partial_fold(const size_t len, __m128i *xmm_crc0, __m128i *xmm_crc1, __m128i *xmm_crc2,
                          __m128i *xmm_crc3, __m128i *xmm_crc_part) {
 
     const __m128i xmm_fold4 = _mm_set_epi32( 0x00000001, 0x54442bd4,
@@ -342,7 +342,7 @@ Z_INTERNAL uint32_t crc32_fold_pclmulqdq_final(crc32_fold *crc) {
     return crc->value;
 }
 
-Z_INTERNAL uint32_t crc32_pclmulqdq(uint32_t crc32, const uint8_t *buf, uint64_t len) {
+Z_INTERNAL uint32_t crc32_pclmulqdq(uint32_t crc32, const uint8_t *buf, size_t len) {
     /* For lens < 64, crc32_braid method is faster. The CRC32 instruction for
      * these short lengths might also prove to be effective */
     if (len < 64)
diff --git a/arch/x86/crc32_fold_pclmulqdq_tpl.h b/arch/x86/crc32_fold_pclmulqdq_tpl.h
index 0799638e..47bbc011 100644
--- a/arch/x86/crc32_fold_pclmulqdq_tpl.h
+++ b/arch/x86/crc32_fold_pclmulqdq_tpl.h
@@ -20,9 +20,9 @@
 #ifdef X86_PCLMULQDQ_CRC
 
 #ifdef COPY
-Z_INTERNAL void crc32_fold_pclmulqdq_copy(crc32_fold *crc, uint8_t *dst, const uint8_t *src, uint64_t len) {
+Z_INTERNAL void crc32_fold_pclmulqdq_copy(crc32_fold *crc, uint8_t *dst, const uint8_t *src, size_t len) {
 #else
-Z_INTERNAL void crc32_fold_pclmulqdq(crc32_fold *crc, const uint8_t *src, uint64_t len, uint32_t init_crc) {
+Z_INTERNAL void crc32_fold_pclmulqdq(crc32_fold *crc, const uint8_t *src, size_t len, uint32_t init_crc) {
 #endif
     unsigned long algn_diff;
     __m128i xmm_t0, xmm_t1, xmm_t2, xmm_t3;
@@ -47,9 +47,9 @@ Z_INTERNAL void crc32_fold_pclmulqdq(crc32_fold *crc, const uint8_t *src, uint64
         if (len == 0)
             return;
 
-        memcpy(partial_buf, src, (size_t)len);
+        memcpy(partial_buf, src, len);
         xmm_crc_part = _mm_load_si128((const __m128i *)partial_buf);
-        memcpy(dst, partial_buf, (size_t)len);
+        memcpy(dst, partial_buf, len);
 #endif
         goto partial;
     }
@@ -82,10 +82,10 @@ Z_INTERNAL void crc32_fold_pclmulqdq(crc32_fold *crc, const uint8_t *src, uint64
 #ifdef X86_VPCLMULQDQ_CRC
     if (x86_cpu_has_vpclmulqdq && x86_cpu_has_avx512 && (len >= 256)) {
 #ifdef COPY
-        uint64_t n = fold_16_vpclmulqdq_copy(&xmm_crc0, &xmm_crc1, &xmm_crc2, &xmm_crc3, dst, src, len);
+        size_t n = fold_16_vpclmulqdq_copy(&xmm_crc0, &xmm_crc1, &xmm_crc2, &xmm_crc3, dst, src, len);
         dst += n;
 #else
-        uint64_t n = fold_16_vpclmulqdq(&xmm_crc0, &xmm_crc1, &xmm_crc2, &xmm_crc3, src, len,
+        size_t n = fold_16_vpclmulqdq(&xmm_crc0, &xmm_crc1, &xmm_crc2, &xmm_crc3, src, len,
             xmm_initial, first);
         first = 0;
 #endif
@@ -176,12 +176,12 @@ Z_INTERNAL void crc32_fold_pclmulqdq(crc32_fold *crc, const uint8_t *src, uint64
 
 partial:
     if (len) {
-        memcpy(&xmm_crc_part, src, (size_t)len);
+        memcpy(&xmm_crc_part, src, len);
 #ifdef COPY
         _mm_storeu_si128((__m128i *)partial_buf, xmm_crc_part);
-        memcpy(dst, partial_buf, (size_t)len);
+        memcpy(dst, partial_buf, len);
 #endif
-        partial_fold((size_t)len, &xmm_crc0, &xmm_crc1, &xmm_crc2, &xmm_crc3, &xmm_crc_part);
+        partial_fold(len, &xmm_crc0, &xmm_crc1, &xmm_crc2, &xmm_crc3, &xmm_crc_part);
     }
 
     crc32_fold_save((__m128i *)crc->fold, &xmm_crc0, &xmm_crc1, &xmm_crc2, &xmm_crc3);
diff --git a/arch/x86/crc32_fold_vpclmulqdq_tpl.h b/arch/x86/crc32_fold_vpclmulqdq_tpl.h
index 421b64da..3d27cb3d 100644
--- a/arch/x86/crc32_fold_vpclmulqdq_tpl.h
+++ b/arch/x86/crc32_fold_vpclmulqdq_tpl.h
@@ -4,18 +4,18 @@
  */
 
 #ifdef COPY
-Z_INTERNAL uint64_t fold_16_vpclmulqdq_copy(__m128i *xmm_crc0, __m128i *xmm_crc1,
-    __m128i *xmm_crc2, __m128i *xmm_crc3, uint8_t *dst, const uint8_t *src, uint64_t len) {
+Z_INTERNAL size_t fold_16_vpclmulqdq_copy(__m128i *xmm_crc0, __m128i *xmm_crc1,
+    __m128i *xmm_crc2, __m128i *xmm_crc3, uint8_t *dst, const uint8_t *src, size_t len) {
 #else
-Z_INTERNAL uint64_t fold_16_vpclmulqdq(__m128i *xmm_crc0, __m128i *xmm_crc1,
-    __m128i *xmm_crc2, __m128i *xmm_crc3, const uint8_t *src, uint64_t len,
+Z_INTERNAL size_t fold_16_vpclmulqdq(__m128i *xmm_crc0, __m128i *xmm_crc1,
+    __m128i *xmm_crc2, __m128i *xmm_crc3, const uint8_t *src, size_t len,
     __m128i init_crc, int32_t first) {
     __m512i zmm_initial = _mm512_zextsi128_si512(init_crc);
 #endif
     __m512i zmm_t0, zmm_t1, zmm_t2, zmm_t3;
     __m512i zmm_crc0, zmm_crc1, zmm_crc2, zmm_crc3;
     __m512i z0, z1, z2, z3;
-    uint64_t len_tmp = len;
+    size_t len_tmp = len;
     const __m512i zmm_fold4 = _mm512_set4_epi32(
         0x00000001, 0x54442bd4, 0x00000001, 0xc6e41596);
     const __m512i zmm_fold16 = _mm512_set4_epi32(
diff --git a/cpu_features.h b/cpu_features.h
index b8de4b71..9d10ce7d 100644
--- a/cpu_features.h
+++ b/cpu_features.h
@@ -65,10 +65,10 @@ extern uint32_t adler32_fold_copy_avx512_vnni(uint32_t adler, uint8_t *dst, cons
 /* CRC32 folding */
 #ifdef X86_PCLMULQDQ_CRC
 extern uint32_t crc32_fold_pclmulqdq_reset(crc32_fold *crc);
-extern void     crc32_fold_pclmulqdq_copy(crc32_fold *crc, uint8_t *dst, const uint8_t *src, uint64_t len);
-extern void     crc32_fold_pclmulqdq(crc32_fold *crc, const uint8_t *src, uint64_t len, uint32_t init_crc);
+extern void     crc32_fold_pclmulqdq_copy(crc32_fold *crc, uint8_t *dst, const uint8_t *src, size_t len);
+extern void     crc32_fold_pclmulqdq(crc32_fold *crc, const uint8_t *src, size_t len, uint32_t init_crc);
 extern uint32_t crc32_fold_pclmulqdq_final(crc32_fold *crc);
-extern uint32_t crc32_pclmulqdq(uint32_t crc32, const uint8_t *buf, uint64_t len);
+extern uint32_t crc32_pclmulqdq(uint32_t crc32, const uint8_t *buf, size_t len);
 #endif
 
 /* memory chunking */
@@ -111,15 +111,15 @@ extern uint8_t* chunkmemset_safe_power8(uint8_t *out, unsigned dist, unsigned le
 #endif
 
 /* CRC32 */
-typedef uint32_t (*crc32_func)(uint32_t crc32, const uint8_t *buf, uint64_t len);
+typedef uint32_t (*crc32_func)(uint32_t crc32, const uint8_t *buf, size_t len);
 
-extern uint32_t PREFIX(crc32_braid)(uint32_t crc, const uint8_t *buf, uint64_t len);
+extern uint32_t PREFIX(crc32_braid)(uint32_t crc, const uint8_t *buf, size_t len);
 #ifdef ARM_ACLE_CRC_HASH
-extern uint32_t crc32_acle(uint32_t crc, const uint8_t *buf, uint64_t len);
+extern uint32_t crc32_acle(uint32_t crc, const uint8_t *buf, size_t len);
 #elif defined(POWER8_VSX_CRC32)
-extern uint32_t crc32_power8(uint32_t crc, const uint8_t *buf, uint64_t len);
+extern uint32_t crc32_power8(uint32_t crc, const uint8_t *buf, size_t len);
 #elif defined(S390_CRC32_VX)
-extern uint32_t PREFIX(s390_crc32_vx)(uint32_t crc, const uint8_t *buf, uint64_t len);
+extern uint32_t PREFIX(s390_crc32_vx)(uint32_t crc, const uint8_t *buf, size_t len);
 #endif
 
 /* compare256 */
diff --git a/crc32_braid.c b/crc32_braid.c
index a7b9b7eb..96754b53 100644
--- a/crc32_braid.c
+++ b/crc32_braid.c
@@ -111,7 +111,7 @@ static z_word_t crc_word(z_word_t data) {
 #endif /* W */
 
 /* ========================================================================= */
-Z_INTERNAL uint32_t PREFIX(crc32_braid)(uint32_t crc, const uint8_t *buf, uint64_t len) {
+Z_INTERNAL uint32_t PREFIX(crc32_braid)(uint32_t crc, const uint8_t *buf, size_t len) {
     Z_REGISTER uint32_t c;
 
     /* Pre-condition the CRC */
@@ -120,7 +120,7 @@ Z_INTERNAL uint32_t PREFIX(crc32_braid)(uint32_t crc, const uint8_t *buf, uint64
 #ifdef W
     /* If provided enough bytes, do a braided CRC calculation. */
     if (len >= N * W + W - 1) {
-        uint64_t blks;
+        size_t blks;
         z_word_t const *words;
         int k;
 
diff --git a/crc32_braid_p.h b/crc32_braid_p.h
index 26906b64..1d8a0706 100644
--- a/crc32_braid_p.h
+++ b/crc32_braid_p.h
@@ -45,6 +45,6 @@
 /* CRC polynomial. */
 #define POLY 0xedb88320         /* p(x) reflected, with x^32 implied */
 
-extern uint32_t PREFIX(crc32_braid)(uint32_t crc, const uint8_t *buf, uint64_t len);
+extern uint32_t PREFIX(crc32_braid)(uint32_t crc, const uint8_t *buf, size_t len);
 
 #endif /* CRC32_BRAID_P_H_ */
diff --git a/crc32_fold.c b/crc32_fold.c
index b1011deb..5b3c7c45 100644
--- a/crc32_fold.c
+++ b/crc32_fold.c
@@ -14,23 +14,12 @@ Z_INTERNAL uint32_t crc32_fold_reset_c(crc32_fold *crc) {
     return crc->value;
 }
 
-Z_INTERNAL void crc32_fold_copy_c(crc32_fold *crc, uint8_t *dst, const uint8_t *src, uint64_t len) {
+Z_INTERNAL void crc32_fold_copy_c(crc32_fold *crc, uint8_t *dst, const uint8_t *src, size_t len) {
     crc->value = functable.crc32(crc->value, src, len);
-/* Test that we don't try to copy more than actually fits in available address space */
-#if INTPTR_MAX > SSIZE_MAX
-    while (len > SSIZE_MAX) {
-        memcpy(dst, src, SSIZE_MAX);
-        dst += SSIZE_MAX;
-        src += SSIZE_MAX;
-        len -= SSIZE_MAX;
-    }
-#endif
-    if (len) {
-        memcpy(dst, src, (size_t)len);
-    }
+    memcpy(dst, src, len);
 }
 
-Z_INTERNAL void crc32_fold_c(crc32_fold *crc, const uint8_t *src, uint64_t len, uint32_t init_crc) {
+Z_INTERNAL void crc32_fold_c(crc32_fold *crc, const uint8_t *src, size_t len, uint32_t init_crc) {
     /* Note: while this is basically the same thing as the vanilla CRC function, we still need
      * a functable entry for it so that we can generically dispatch to this function with the
      * same arguments for the versions that _do_ do a folding CRC but we don't want a copy. The
diff --git a/crc32_fold.h b/crc32_fold.h
index 6aef5467..0d2ff669 100644
--- a/crc32_fold.h
+++ b/crc32_fold.h
@@ -14,8 +14,8 @@ typedef struct crc32_fold_s {
 } crc32_fold;
 
 Z_INTERNAL uint32_t crc32_fold_reset_c(crc32_fold *crc);
-Z_INTERNAL void     crc32_fold_copy_c(crc32_fold *crc, uint8_t *dst, const uint8_t *src, uint64_t len);
-Z_INTERNAL void     crc32_fold_c(crc32_fold *crc, const uint8_t *src, uint64_t len, uint32_t init_crc);
+Z_INTERNAL void     crc32_fold_copy_c(crc32_fold *crc, uint8_t *dst, const uint8_t *src, size_t len);
+Z_INTERNAL void     crc32_fold_c(crc32_fold *crc, const uint8_t *src, size_t len, uint32_t init_crc);
 Z_INTERNAL uint32_t crc32_fold_final_c(crc32_fold *crc);
 
 #endif
diff --git a/functable.c b/functable.c
index 4af8a8a7..5dd5e54d 100644
--- a/functable.c
+++ b/functable.c
@@ -233,7 +233,7 @@ Z_INTERNAL uint32_t crc32_fold_reset_stub(crc32_fold *crc) {
     return functable.crc32_fold_reset(crc);
 }
 
-Z_INTERNAL void crc32_fold_copy_stub(crc32_fold *crc, uint8_t *dst, const uint8_t *src, uint64_t len) {
+Z_INTERNAL void crc32_fold_copy_stub(crc32_fold *crc, uint8_t *dst, const uint8_t *src, size_t len) {
     functable.crc32_fold_copy = &crc32_fold_copy_c;
     cpu_check_features();
 #ifdef X86_PCLMULQDQ_CRC
@@ -243,7 +243,7 @@ Z_INTERNAL void crc32_fold_copy_stub(crc32_fold *crc, uint8_t *dst, const uint8_
     functable.crc32_fold_copy(crc, dst, src, len);
 }
 
-Z_INTERNAL void crc32_fold_stub(crc32_fold *crc, const uint8_t *src, uint64_t len, uint32_t init_crc) {
+Z_INTERNAL void crc32_fold_stub(crc32_fold *crc, const uint8_t *src, size_t len, uint32_t init_crc) {
     functable.crc32_fold = &crc32_fold_c;
     cpu_check_features();
 #ifdef X86_PCLMULQDQ_CRC
@@ -403,10 +403,7 @@ Z_INTERNAL uint8_t* chunkmemset_safe_stub(uint8_t *out, unsigned dist, unsigned
     return functable.chunkmemset_safe(out, dist, len, left);
 }
 
-Z_INTERNAL uint32_t crc32_stub(uint32_t crc, const uint8_t *buf, uint64_t len) {
-    Assert(sizeof(uint64_t) >= sizeof(size_t),
-           "crc32_z takes size_t but internally we have a uint64_t len");
-
+Z_INTERNAL uint32_t crc32_stub(uint32_t crc, const uint8_t *buf, size_t len) {
     functable.crc32 = &PREFIX(crc32_braid);
     cpu_check_features();
 #ifdef ARM_ACLE_CRC_HASH
diff --git a/functable.h b/functable.h
index a5690871..9b6ec3b5 100644
--- a/functable.h
+++ b/functable.h
@@ -13,10 +13,10 @@
 struct functable_s {
     uint32_t (* adler32)            (uint32_t adler, const uint8_t *buf, uint64_t len);
     uint32_t (* adler32_fold_copy)  (uint32_t adler, uint8_t *dst, const uint8_t *src, uint64_t len);
-    uint32_t (* crc32)              (uint32_t crc, const uint8_t *buf, uint64_t len);
+    uint32_t (* crc32)              (uint32_t crc, const uint8_t *buf, size_t len);
     uint32_t (* crc32_fold_reset)   (struct crc32_fold_s *crc);
-    void     (* crc32_fold_copy)    (struct crc32_fold_s *crc, uint8_t *dst, const uint8_t *src, uint64_t len);
-    void     (* crc32_fold)         (struct crc32_fold_s *crc, const uint8_t *src, uint64_t len, uint32_t init_crc);
+    void     (* crc32_fold_copy)    (struct crc32_fold_s *crc, uint8_t *dst, const uint8_t *src, size_t len);
+    void     (* crc32_fold)         (struct crc32_fold_s *crc, const uint8_t *src, size_t len, uint32_t init_crc);
     uint32_t (* crc32_fold_final)   (struct crc32_fold_s *crc);
     uint32_t (* compare256)         (const uint8_t *src0, const uint8_t *src1);
     uint32_t (* chunksize)          (void);