Sfoglia il codice sorgente

Merge branch 'maint-0.2.4' into maint-0.2.5

Nick Mathewson 7 anni fa
parent
commit
332543baed
2 ha cambiato i file con 16 aggiunte e 0 eliminazioni
  1. 6 0
      changes/bug18089
  2. 10 0
      src/common/crypto.c

+ 6 - 0
changes/bug18089

@@ -0,0 +1,6 @@
+  o Minor fixes (security):
+    - Make memwipe() do nothing when passed a NULL pointer
+      or zero size. Check size argument to memwipe() for underflow.
+      Closes bug #18089. Reported by "gk", patch by "teor".
+      Bugfix on 0.2.3.25 and 0.2.4.6-alpha (#7352),
+      commit 49dd5ef3 on 7 Nov 2012.

+ 10 - 0
src/common/crypto.c

@@ -3056,6 +3056,7 @@ secret_to_key(char *key_out, size_t key_out_len, const char *secret,
 /**
  * Destroy the <b>sz</b> bytes of data stored at <b>mem</b>, setting them to
  * the value <b>byte</b>.
+ * If <b>mem</b> is NULL or <b>sz</b> is zero, nothing happens.
  *
  * This function is preferable to memset, since many compilers will happily
  * optimize out memset() when they can convince themselves that the data being
@@ -3073,6 +3074,15 @@ secret_to_key(char *key_out, size_t key_out_len, const char *secret,
 void
 memwipe(void *mem, uint8_t byte, size_t sz)
 {
+  if (sz == 0) {
+    return;
+  }
+  /* If sz is nonzero, then mem must not be NULL. */
+  tor_assert(mem != NULL);
+
+  /* Data this large is likely to be an underflow. */
+  tor_assert(sz < SIZE_T_CEILING);
+
   /* Because whole-program-optimization exists, we may not be able to just
    * have this function call "memset".  A smart compiler could inline it, then
    * eliminate dead memsets, and declare itself to be clever. */