git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH 3/3] Makefile: add USE_SHA1DC knob

From
Jeff King <peff@peff.net>
Date
Feb 23, 2017, 23:06 UTC
Message-ID
<20170223230621.43anex65ndoqbgnf@sigill.intra.peff.net>
In-Reply-To
<20170223230507.kuxjqtg3ghcfskc6@sigill.intra.peff.net>
This knob lets you use the sha1dc implementation from:
      https://github.com/cr-marcstevens/sha1collisiondetection

which can detect certain types of collision attacks (even when we only see half of the colliding pair).

The big downside is that it's slower than either the openssl or block-sha1 implementations.

Here are some timings based off of linux.git:
  - compute sha1 over whole packfile
    before: 1.349s
     after: 5.067s
    change: +275%
  - rev-list --all
    before: 5.742s
     after: 5.730s
    change: -0.2%
  - rev-list --all --objects
    before: 33.257s
     after: 33.392s
    change: +0.4%
  - index-pack --verify
    before: 2m20s
     after: 5m43s
    change: +145%
  - git log --no-merges -10000 -p
    before: 9.532s
     after: 9.683s
    change: +1.5%

So overall the sha1 computation is about 3-4x slower. But of course most operations do more than just sha1. Accessing commits and trees isn't slowed at all (both the +/- changes there are well within the run-to-run noise). Accessing the blobs is a little slower, but mostly drowned out by the cost of things like actually generating patches.

The most-affected operation is `index-pack --verify`, which is essentially just computing the sha1 on every object. It's a bit worse than twice as slow, which means every push and every fetch is going to experience that.

Signed-off-by: Jeff King <peff@peff.net>
---
 Makefile      | 10 ++++++++++
 sha1dc/sha1.c | 22 ++++++++++++++++++++++
 sha1dc/sha1.h | 16 ++++++++++++++++
 3 files changed, 48 insertions(+)
diff --git a/Makefile b/Makefile
index 8e4081e06..7c4906250 100644
--- a/Makefile
+++ b/Makefile
@@ -142,6 +142,10 @@ all::
 # Define PPC_SHA1 environment variable when running make to make use of
 # a bundled SHA1 routine optimized for PowerPC.
 #
+# Define USE_SHA1DC to unconditionally enable the collision-detecting sha1
+# algorithm. This is slower, but may detect attempted collision attacks.
+# Takes priority over other *_SHA1 knobs.
+#
 # Define SHA1_MAX_BLOCK_SIZE to limit the amount of data that will be hashed
 # in one call to the platform's SHA1_Update(). e.g. APPLE_COMMON_CRYPTO
 # wants 'SHA1_MAX_BLOCK_SIZE=1024L*1024L*1024L' defined.
@@ -1386,6 +1390,11 @@ ifdef APPLE_COMMON_CRYPTO
 	SHA1_MAX_BLOCK_SIZE = 1024L*1024L*1024L
 endif
 
+ifdef USE_SHA1DC
+	SHA1_HEADER = "sha1dc/sha1.h"
+	LIB_OBJS += sha1dc/sha1.o
+	LIB_OBJS += sha1dc/ubc_check.o
+else
 ifdef BLK_SHA1
 	SHA1_HEADER = "block-sha1/sha1.h"
 	LIB_OBJS += block-sha1/sha1.o
@@ -1403,6 +1412,7 @@ else
 endif
 endif
 endif
+endif
 
 ifdef SHA1_MAX_BLOCK_SIZE
 	LIB_OBJS += compat/sha1-chunked.o
diff --git a/sha1dc/sha1.c b/sha1dc/sha1.c
index 762c6fff8..1566ec4c7 100644
--- a/sha1dc/sha1.c
+++ b/sha1dc/sha1.c
@@ -1141,3 +1141,25 @@ int SHA1DCFinal(unsigned char output[20], SHA1_CTX *ctx)
 	output[19] = (unsigned char)(ctx->ihv[4]);
 	return ctx->found_collision;
 }
+
+static const char collision_message[] =
+"The SHA1 computation detected evidence of a collision attack;\n"
+"refusing to process the contents.";
+
+void git_SHA1DCFinal(unsigned char hash[20], SHA1_CTX *ctx)
+{
+	if (SHA1DCFinal(hash, ctx))
+		die(collision_message);
+}
+
+void git_SHA1DCUpdate(SHA1_CTX *ctx, const void *vdata, unsigned long len)
+{
+	const char *data = vdata;
+	/* We expect an unsigned long, but sha1dc only takes an int */
+	while (len > INT_MAX) {
+		SHA1DCUpdate(ctx, data, INT_MAX);
+		data += INT_MAX;
+		len -= INT_MAX;
+	}
+	SHA1DCUpdate(ctx, data, len);
+}
diff --git a/sha1dc/sha1.h b/sha1dc/sha1.h
index ce5390397..1bb0ace99 100644
--- a/sha1dc/sha1.h
+++ b/sha1dc/sha1.h
@@ -90,3 +90,19 @@ void SHA1DCUpdate(SHA1_CTX*, const char*, unsigned);
 // obtain SHA-1 hash from SHA-1 context
 // returns: 0 = no collision detected, otherwise = collision found => warn user for active attack
 int  SHA1DCFinal(unsigned char[20], SHA1_CTX*); 
+
+
+/*
+ * Same as SHA1DCFinal, but convert collision attack case into a verbose die().
+ */
+void git_SHA1DCFinal(unsigned char [20], SHA1_CTX *);
+
+/*
+ * Same as SHA1DCUpdate, but adjust types to match git's usual interface.
+ */
+void git_SHA1DCUpdate(SHA1_CTX *ctx, const void *data, unsigned long len);
+
+#define platform_SHA_CTX SHA1_CTX
+#define platform_SHA1_Init SHA1DCInit
+#define platform_SHA1_Update git_SHA1DCUpdate
+#define platform_SHA1_Final git_SHA1DCFinal
-- 
2.12.0.rc2.629.ga7951ed82
Previous: Jeff KingNext: HW42
Message 122 of 136 in “SHA1 collisions found”
  1. Joey HessFeb 23, 2017
  2. Junio C HamanoFeb 23, 2017
  3. Junio C HamanoFeb 23, 2017
  4. David LangFeb 23, 2017
  5. Jakub NarębskiFeb 23, 2017
  6. Jeff KingFeb 23, 2017
  7. Joey HessFeb 23, 2017
  8. Linus TorvaldsFeb 23, 2017
  9. Joey HessFeb 23, 2017
  10. Linus TorvaldsFeb 23, 2017
  11. Jeff KingFeb 23, 2017
  12. Linus TorvaldsFeb 23, 2017
  13. Jeff KingFeb 23, 2017
  14. Linus TorvaldsFeb 23, 2017
  15. Jeff KingFeb 23, 2017
  16. Øyvind A. HolmFeb 23, 2017
  17. Joey HessFeb 23, 2017
  18. Joey HessFeb 23, 2017
  19. Morten WelinderFeb 23, 2017
  20. Geert UytterhoevenFeb 24, 2017
  21. Jeff KingFeb 23, 2017
  22. David LangFeb 23, 2017
  23. David LangFeb 23, 2017
  24. David LangFeb 23, 2017
  25. Linus TorvaldsFeb 23, 2017
  26. Linus TorvaldsFeb 23, 2017
  27. Joey HessFeb 23, 2017
  28. Linus TorvaldsFeb 23, 2017
  29. Junio C HamanoFeb 23, 2017
  30. Duy NguyenFeb 24, 2017
  31. brian m. carlsonFeb 25, 2017
  32. René ScharfeFeb 27, 2017
  33. brian m. carlsonFeb 28, 2017
  34. Ian JacksonFeb 24, 2017
  35. ankostisFeb 24, 2017
  36. Junio C HamanoFeb 24, 2017
  37. David LangFeb 24, 2017
  38. Junio C HamanoFeb 24, 2017
  39. Stefan BellerFeb 24, 2017
  40. Junio C HamanoFeb 24, 2017
  41. Junio C HamanoFeb 24, 2017
  42. ankostisFeb 24, 2017
  43. Junio C HamanoFeb 24, 2017
  44. ankostisFeb 25, 2017
  45. Jason CooperFeb 26, 2017
  46. brian m. carlsonFeb 26, 2017
  47. Linus TorvaldsFeb 26, 2017
  48. Ævar Arnfjörð BjarmasonFeb 26, 2017
  49. Jeff KingFeb 26, 2017
  50. Transition plan for git to move to a new hash functionIan Jackson, Feb 27, 2017
  51. Markus TrippelsdorfFeb 27, 2017
  52. Ian JacksonFeb 27, 2017
  53. Tony FinchFeb 27, 2017
  54. brian m. carlsonFeb 28, 2017
  55. Ian JacksonMar 2, 2017
  56. brian m. carlsonMar 4, 2017
  57. Ian JacksonMar 5, 2017
  58. brian m. carlsonMar 5, 2017
  59. Philip OakleyFeb 24, 2017
  60. Jeff KingFeb 24, 2017
  61. Linus TorvaldsFeb 25, 2017
  62. Linus TorvaldsFeb 25, 2017
  63. Jeff KingFeb 25, 2017
  64. Junio C HamanoFeb 26, 2017
  65. Junio C HamanoFeb 25, 2017
  66. Jason CooperFeb 26, 2017
  67. Jeff KingFeb 26, 2017
  68. brian m. carlsonFeb 26, 2017
  69. Brandon WilliamsMar 2, 2017
  70. Jeff KingMar 3, 2017
  71. Ian JacksonMar 3, 2017
  72. Jeff KingMar 3, 2017
  73. Linus TorvaldsMar 2, 2017
  74. Junio C HamanoMar 2, 2017
  75. Linus TorvaldsMar 2, 2017
  76. Joey HessMar 2, 2017
  77. Linus TorvaldsMar 2, 2017
  78. Mike HommeyMar 3, 2017
  79. Linus TorvaldsMar 3, 2017
  80. Jeff KingMar 3, 2017
  81. Stefan BellerMar 3, 2017
  82. David LangFeb 25, 2017
  83. Stefan BellerFeb 25, 2017
  84. Jeff KingFeb 25, 2017
  85. David LangFeb 25, 2017
  86. Jeff KingFeb 25, 2017
  87. David LangFeb 25, 2017
  88. Jacob KellerFeb 25, 2017
  89. Jacob KellerFeb 25, 2017
  90. grarpampFeb 25, 2017
  91. Ian JacksonFeb 24, 2017
  92. Ian JacksonFeb 25, 2017
  93. brian m. carlsonFeb 25, 2017
  94. Jeff KingFeb 25, 2017
  95. Mike HommeyFeb 25, 2017
  96. brian m. carlsonFeb 26, 2017
  97. Jason CooperFeb 24, 2017
  98. ankostisFeb 25, 2017
  99. Jakub NarębskiFeb 24, 2017
  100. Santiago TorresFeb 24, 2017
  101. Jakub NarębskiFeb 24, 2017
  102. Øyvind A. HolmFeb 24, 2017
  103. Jeff KingFeb 24, 2017
  104. Jakub NarębskiFeb 24, 2017
  105. Lars SchneiderFeb 25, 2017
  106. Jeff KingFeb 26, 2017
  107. Junio C HamanoFeb 26, 2017
  108. Thomas BraunFeb 26, 2017
  109. Jeff KingFeb 26, 2017
  110. Geert UytterhoevenFeb 27, 2017
  111. Jeff KingFeb 27, 2017
  112. Morten WelinderFeb 27, 2017
  113. Jeff KingFeb 23, 2017
  114. Linus TorvaldsFeb 23, 2017
  115. Jeff KingFeb 23, 2017
  116. 1/3 add collision-detecting sha1 implementationJeff King, Feb 23, 2017
  117. Stefan BellerFeb 23, 2017
  118. Jeff KingFeb 24, 2017
  119. Linus TorvaldsFeb 24, 2017
  120. Jeff KingFeb 24, 2017
  121. 2/3 sha1dc: adjust header includes for gitJeff King, Feb 23, 2017
  122. 3/3 Makefile: add USE_SHA1DC knobJeff King, Feb 23, 2017
  123. HW42Feb 24, 2017
  124. Jeff KingFeb 24, 2017
  125. Linus TorvaldsFeb 23, 2017
  126. Junio C HamanoFeb 28, 2017
  127. Junio C HamanoFeb 28, 2017
  128. Jeff KingFeb 28, 2017
  129. Dan ShumowMar 1, 2017
  130. Linus TorvaldsFeb 28, 2017
  131. Shawn PearceFeb 28, 2017
  132. Linus TorvaldsFeb 28, 2017
  133. Dan ShumowFeb 28, 2017
  134. Marc StevensFeb 28, 2017
  135. Linus TorvaldsFeb 28, 2017
  136. Jeff KingMar 1, 2017

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.