Publishing nanopb-0.3.1

Possible consequences of bug:
1) Denial of service by causing a crash
   Possible when all of the following apply:
      - Untrusted data is passed to pb_decode()
      - The top-level message contains a static string field as the first field.
   Causes a single write of '0' byte to 1 byte before the message struct.

2) Remote code execution
   Possible when all of the following apply:
      - 64-bit platform
      - The message or a submessage contains a static/pointer string field.
      - Decoding directly from a custom pb_istream_t
      - bytes_left on the stream is set to larger than 4 GB
   Causes a write of up to 4 GB of data past the string field.

3) Possible heap corruption or remote code execution
   Possible when all of the following apply:
      - less than 64-bit platform
      - The message or a submessage contains a pointer-type bytes field.
   Causes a write of sizeof(pb_size_t) bytes of data past a 0-byte long
   malloc()ed buffer. On many malloc() implementations, this causes at
   most a crash. However, remote code execution through a controlled jump
   cannot be ruled out.

--

Detailed analysis follows

In the following consideration, I define "platform bitness" as equal to
number of bits in size_t datatype. Therefore most 8-bit platforms are
regarded as 16-bit for the purposes of this discussion.

1. The overflow in pb_dec_string

The overflow happens in this computation:

uint32_t size;
size_t alloc_size;
alloc_size = size + 1;

There are two ways in which the overflow can occur: In the uint32_t
addition, or in the cast to size_t. This depends on the platform
bitness.

On 32- and 64-bit platforms, the size has to be UINT32_MAX for the
overflow to occur. In that case alloc_size will be 0.

On 16-bit platforms, overflow will happen whenever size is more than
UINT16_MAX, and resulting alloc_size is attacker controlled.

For static fields, the alloc_size value is just checked against the
field data size. For pointer fields, the alloc_size value is passed to
malloc(). End result in both cases is the same, the storage is 0 or
just a few bytes in length.

On 16-bit platforms, another overflow occurs in the call to pb_read(),
when passing the original size. An attacker will want the passed value
to be larger than the alloc_size, therefore the only reasonable choice
is to have size = UINT16_MAX and alloc_size = 0. Any larger multiple
will truncate to the same values.

At this point we have read atleast the tag and the string length of the
message, i.e. atleast 3 bytes. The maximum initial value for stream
bytes_left is SIZE_MAX, thus at this point at most SIZE_MAX-3 bytes are
remaining.

On 32-bit and 16-bit platforms this means that the size passed to
pb_read() is always larger than the number of remaining bytes. This
causes pb_read() to fail immediately, before reading any bytes.

On 64-bit platforms, it is possible for the bytes_left value to be set
to a value larger than UINT32_MAX, which is the wraparound point in
size calculation. In this case pb_read() will succeed and write up to 4
GB of attacker controlled data over the RAM that comes after the string
field.

On all platforms, there is an unconditional write of a terminating null
byte. Because the size of size_t typically reflects the size of the
processor address space, a write at UINT16_MAX or UINT32_MAX bytes
after the string field actually wraps back to before the string field.
Consequently, on 32-bit and 16-bit platforms, the bug causes a single
write of '0' byte at one byte before the string field.

If the string field is in the middle of a message, this will just
corrupt other data in the message struct. Because the message contents
is attacker controlled anyway, this is a non-issue. However, if the
string field is the first field in the top-level message, it can
corrupt other data on the stack/heap before it. Typically a single '0'
write at a location not controlled by attacker is enough only for a
denial-of-service attack.

When using pointer fields and malloc(), the attacker controlled
alloc_size will cause a 0-size allocation to happen. By the same logic
as before, on 32-bit and 16-bit platforms this causes a '0' byte write
only. On 64-bit platforms, however, it will again allow up to 4 GB of
malicious data to be written over memory, if the stream length allows
the read.

2. The overflow in pb_dec_bytes

This overflow happens in the PB_BYTES_ARRAY_T_ALLOCSIZE macro:

The computation is done in size_t data type this time. This means that
an overflow is possible only when n is larger than SIZE_MAX -
offsetof(..). The offsetof value in this case is equal to
sizeof(pb_size_t) bytes.

Because the incoming size value is limited to 32 bits, no overflow can
happen here on 64-bit platforms.

The size will be passed to pb_read(). Like before, on 32-bit and 16-bit
platforms the read will always fail before writing anything.

This leaves only the write of bdest->size as exploitable. On statically
allocated fields, the size field will always be allocated, regardless
of alloc_size. In this case, no buffer overflow is possible here, but
user code could possibly use the attacker controlled size value and
read past a buffer.

If the field is allocated through malloc(), this will allow a write of
sizeof(pb_size_t) attacker controlled bytes to past a 0-byte long
buffer. In typical malloc implementations, this will either fit in
unused alignment padding area, or cause a heap corruption and a crash.
Under very exceptional situation it could allow attacker to influence
the behaviour of malloc(), possibly jumping into an attacker-controlled
location and thus leading to remote code execution.

CHANGELOG.txt

@@ -1,3 +1,11 @@
+nanopb-0.3.1 (2014-09-11)
+ Fix security issue due to size_t overflows. (issue 132)
+ Fix memory leak with duplicated fields and PB_ENABLE_MALLOC
+ Fix crash if pb_release() is called twice.
+ Fix cyclic message support (issue 130)
+ Fix error in generated initializers for repeated pointer fields.
+ Improve tests (issues 113, 126)
+
 nanopb-0.3.0 (2014-08-26)
  NOTE: See docs/migration.html or online at
  http://koti.kapsi.fi/~jpa/nanopb/docs/migration.html

docs/migration.rst

@@ -151,12 +151,12 @@ Callback function signature
 as *void\**. This allowed passing of any data, but made it unnecessarily
 complex to return a pointer from callback.
 
-**Changes:** The callback function parameter was changed to *void\**.
+**Changes:** The callback function parameter was changed to *void\*\**.
 
 **Required actions:** You can continue using the old callback style by
 defining *PB_OLD_CALLBACK_STYLE*. Recommended action is to:
 
-  * Change the callback signatures to contain *void\** for decoders and
+  * Change the callback signatures to contain *void\*\** for decoders and
     *void \* const \** for encoders.
   * Change the callback function body to use *\*arg* instead of *arg*.
 

docs/security.rst

@@ -26,9 +26,9 @@ The following data is regarded as **trusted**. It must be under the control of
 the application writer. Malicious data in these structures could cause
 security issues, such as execution of arbitrary code:
 
-1. Callback and extension fields in message structures given to pb_encode()
+1. Callback, pointer and extension fields in message structures given to
-   and pb_decode(). These fields are memory pointers, and are generated
+   pb_encode() and pb_decode(). These fields are memory pointers, and are
-   depending on the .proto file.
+   generated depending on the message definition in the .proto file.
 2. The automatically generated field definitions, i.e. *pb_field_t* lists.
 3. Contents of the *pb_istream_t* and *pb_ostream_t* structures (this does not
    mean the contents of the stream itself, just the stream definition).
@@ -38,7 +38,7 @@ these will cause "garbage in, garbage out" behaviour. It will not cause
 buffer overflows, information disclosure or other security problems:
 
 1. All data read from *pb_istream_t*.
-2. All fields in message structures, except callbacks and extensions.
+2. All fields in message structures, except callbacks, pointers and extensions.
    (Beginning with nanopb-0.2.4, in earlier versions the field sizes are partially unchecked.)
 
 Invariants
@@ -76,4 +76,6 @@ The following list is not comprehensive:
    stop a denial of service attack from using an infinite message.
 4. If using network sockets as streams, a timeout should be set to stop
    denial of service attacks.
-
+5. If using *malloc()* support, some method of limiting memory use should be
+   employed. This can be done by defining custom *pb_realloc()* function.
+   Nanopb will properly detect and handle failed memory allocations.

generator/nanopb_generator.py

@@ -1,7 +1,7 @@
 #!/usr/bin/python
 
 '''Generate header file for nanopb from a ProtoBuf FileDescriptorSet.'''
-nanopb_version = "nanopb-0.3.0"
+nanopb_version = "nanopb-0.3.1"
 
 import sys
 
@@ -546,7 +546,7 @@ class Message:
 
     def get_dependencies(self):
         '''Get list of type names that this structure refers to.'''
-        return [str(field.ctype) for field in self.fields]
+        return [str(field.ctype) for field in self.fields if field.allocation == 'STATIC']
     
     def __str__(self):
         result = 'typedef struct _%s {\n' % self.name
@@ -596,6 +596,8 @@ class Message:
                 else:
                     parts.append(field.get_initializer(null_init))
             elif field.allocation == 'POINTER':
+                if field.rules == 'REPEATED':
+                    parts.append('0')
                 parts.append('NULL')
             elif field.allocation == 'CALLBACK':
                 if field.pbtype == 'EXTENSION':

pb.h

@@ -46,7 +46,7 @@
 
 /* Version of the nanopb library. Just in case you want to check it in
  * your own program. */
-#define NANOPB_VERSION nanopb-0.3.0
+#define NANOPB_VERSION nanopb-0.3.1
 
 /* Include all the system headers needed by nanopb. You will need the
  * definitions of the following:

pb_decode.c

@@ -44,6 +44,11 @@ static bool checkreturn pb_dec_submessage(pb_istream_t *stream, const pb_field_t
 static bool checkreturn pb_skip_varint(pb_istream_t *stream);
 static bool checkreturn pb_skip_string(pb_istream_t *stream);
 
+#ifdef PB_ENABLE_MALLOC
+static bool checkreturn allocate_field(pb_istream_t *stream, void *pData, size_t data_size, size_t array_size);
+static void pb_release_single_field(const pb_field_iter_t *iter);
+#endif
+
 /* --- Function pointers to field decoders ---
  * Order in the array must match pb_action_t LTYPE numbering.
  */
@@ -400,18 +405,23 @@ static bool checkreturn allocate_field(pb_istream_t *stream, void *pData, size_t
 {    
     void *ptr = *(void**)pData;
     
+    if (data_size == 0 || array_size == 0)
+        PB_RETURN_ERROR(stream, "invalid size");
+    
     /* Check for multiplication overflows.
      * This code avoids the costly division if the sizes are small enough.
      * Multiplication is safe as long as only half of bits are set
      * in either multiplicand.
      */
-    const size_t check_limit = (size_t)1 << (sizeof(size_t) * 4);
-    if (data_size >= check_limit || array_size >= check_limit)
     {
-        const size_t size_max = (size_t)-1;
+        const size_t check_limit = (size_t)1 << (sizeof(size_t) * 4);
-        if (size_max / array_size < data_size)
+        if (data_size >= check_limit || array_size >= check_limit)
         {
-            PB_RETURN_ERROR(stream, "size too large");
+            const size_t size_max = (size_t)-1;
+            if (size_max / array_size < data_size)
+            {
+                PB_RETURN_ERROR(stream, "size too large");
+            }
         }
     }
     
@@ -458,6 +468,13 @@ static bool checkreturn decode_pointer_field(pb_istream_t *stream, pb_wire_type_
     {
         case PB_HTYPE_REQUIRED:
         case PB_HTYPE_OPTIONAL:
+            if (PB_LTYPE(type) == PB_LTYPE_SUBMESSAGE &&
+                *(void**)iter->pData != NULL)
+            {
+                /* Duplicate field, have to release the old allocation first. */
+                pb_release_single_field(iter);
+            }
+        
             if (PB_LTYPE(type) == PB_LTYPE_STRING ||
                 PB_LTYPE(type) == PB_LTYPE_BYTES)
             {
@@ -869,6 +886,55 @@ bool pb_decode_delimited(pb_istream_t *stream, const pb_field_t fields[], void *
 }
 
 #ifdef PB_ENABLE_MALLOC
+static void pb_release_single_field(const pb_field_iter_t *iter)
+{
+    pb_type_t type;
+    type = iter->pos->type;
+
+    if (PB_ATYPE(type) == PB_ATYPE_POINTER)
+    {
+        if (PB_HTYPE(type) == PB_HTYPE_REPEATED &&
+            (PB_LTYPE(type) == PB_LTYPE_STRING ||
+             PB_LTYPE(type) == PB_LTYPE_BYTES))
+        {
+            /* Release entries in repeated string or bytes array */
+            void **pItem = *(void***)iter->pData;
+            pb_size_t count = *(pb_size_t*)iter->pSize;
+            while (count--)
+            {
+                pb_free(*pItem);
+                *pItem++ = NULL;
+            }
+            *(pb_size_t*)iter->pSize = 0;
+        }
+        else if (PB_LTYPE(type) == PB_LTYPE_SUBMESSAGE)
+        {
+            /* Release fields in submessages */
+            void *pItem = *(void**)iter->pData;
+            if (pItem)
+            {
+                pb_size_t count = 1;
+                
+                if (PB_HTYPE(type) == PB_HTYPE_REPEATED)
+                {
+                    count = *(pb_size_t*)iter->pSize;
+                    *(pb_size_t*)iter->pSize = 0;
+                }
+                
+                while (count--)
+                {
+                    pb_release((const pb_field_t*)iter->pos->ptr, pItem);
+                    pItem = (uint8_t*)pItem + iter->pos->data_size;
+                }
+            }
+        }
+        
+        /* Release main item */
+        pb_free(*(void**)iter->pData);
+        *(void**)iter->pData = NULL;
+    }
+}
+
 void pb_release(const pb_field_t fields[], void *dest_struct)
 {
     pb_field_iter_t iter;
@@ -878,46 +944,7 @@ void pb_release(const pb_field_t fields[], void *dest_struct)
     
     do
     {
-        pb_type_t type;
+        pb_release_single_field(&iter);
-        type = iter.pos->type;
-    
-        if (PB_ATYPE(type) == PB_ATYPE_POINTER)
-        {
-            if (PB_HTYPE(type) == PB_HTYPE_REPEATED &&
-                (PB_LTYPE(type) == PB_LTYPE_STRING ||
-                 PB_LTYPE(type) == PB_LTYPE_BYTES))
-            {
-                /* Release entries in repeated string or bytes array */
-                void **pItem = *(void***)iter.pData;
-                pb_size_t count = *(pb_size_t*)iter.pSize;
-                while (count--)
-                {
-                    pb_free(*pItem);
-                    *pItem++ = NULL;
-                }
-            }
-            else if (PB_LTYPE(type) == PB_LTYPE_SUBMESSAGE)
-            {
-                /* Release fields in submessages */
-                void *pItem = *(void**)iter.pData;
-                pb_size_t count = (pItem ? 1 : 0);
-                
-                if (PB_HTYPE(type) == PB_HTYPE_REPEATED)
-                {
-                    count = *(pb_size_t*)iter.pSize;   
-                }
-                
-                while (count--)
-                {
-                    pb_release((const pb_field_t*)iter.pos->ptr, pItem);
-                    pItem = (uint8_t*)pItem + iter.pos->data_size;
-                }
-            }
-            
-            /* Release main item */
-            pb_free(*(void**)iter.pData);
-            *(void**)iter.pData = NULL;
-        }
     } while (pb_field_iter_next(&iter));
 }
 #endif
@@ -1045,33 +1072,36 @@ static bool checkreturn pb_dec_fixed64(pb_istream_t *stream, const pb_field_t *f
 static bool checkreturn pb_dec_bytes(pb_istream_t *stream, const pb_field_t *field, void *dest)
 {
     uint32_t size;
+    size_t alloc_size;
     pb_bytes_array_t *bdest;
     
     if (!pb_decode_varint32(stream, &size))
         return false;
     
+    if (size > PB_SIZE_MAX)
+        PB_RETURN_ERROR(stream, "bytes overflow");
+    
+    alloc_size = PB_BYTES_ARRAY_T_ALLOCSIZE(size);
+    if (size > alloc_size)
+        PB_RETURN_ERROR(stream, "size too large");
+    
     if (PB_ATYPE(field->type) == PB_ATYPE_POINTER)
     {
 #ifndef PB_ENABLE_MALLOC
         PB_RETURN_ERROR(stream, "no malloc support");
 #else
-        if (!allocate_field(stream, dest, PB_BYTES_ARRAY_T_ALLOCSIZE(size), 1))
+        if (!allocate_field(stream, dest, alloc_size, 1))
             return false;
         bdest = *(pb_bytes_array_t**)dest;
 #endif
     }
     else
     {
-        if (PB_BYTES_ARRAY_T_ALLOCSIZE(size) > field->data_size)
+        if (alloc_size > field->data_size)
             PB_RETURN_ERROR(stream, "bytes overflow");
         bdest = (pb_bytes_array_t*)dest;
     }
 
-    if (size > PB_SIZE_MAX)
-    {
-        PB_RETURN_ERROR(stream, "bytes overflow");
-    }
-
     bdest->size = (pb_size_t)size;
     return pb_read(stream, bdest->bytes, size);
 }
@@ -1087,6 +1117,9 @@ static bool checkreturn pb_dec_string(pb_istream_t *stream, const pb_field_t *fi
     /* Space for null terminator */
     alloc_size = size + 1;
     
+    if (alloc_size < size)
+        PB_RETURN_ERROR(stream, "size too large");
+    
     if (PB_ATYPE(field->type) == PB_ATYPE_POINTER)
     {
 #ifndef PB_ENABLE_MALLOC

tests/Makefile

@@ -10,12 +10,12 @@ coverage:
 	# LCOV does not like the newer gcov format
 	scons CC=gcc-4.6 CXX=gcc-4.6
 
-	# We are only interested in pb_encode.o and pb_decode.o
-	find build -name '*.gcda' -and \! \( -name '*pb_encode*' -or -name '*pb_decode*' \) -exec rm '{}' \;
-
 	# Collect the data
 	mkdir build/coverage
 	lcov --base-directory . --directory build/ --gcov-tool gcov-4.6 -c -o build/coverage/nanopb.info
 
+	# Remove the test code from results
+	lcov -r build/coverage/nanopb.info '*tests*' -o build/coverage/nanopb.info
+
 	# Generate HTML
 	genhtml -o build/coverage build/coverage/nanopb.info

tests/SConstruct

@@ -57,6 +57,7 @@ if not env.GetOption('clean'):
     if not stdbool or not stdint or not stddef or not string:
         conf.env.Append(CPPDEFINES = {'PB_SYSTEM_HEADER': '\\"pb_syshdr.h\\"'})
         conf.env.Append(CPPPATH = "#../extra")
+	conf.env.Append(SYSHDR = '\\"pb_syshdr.h\\"')
         
         if stdbool: conf.env.Append(CPPDEFINES = {'HAVE_STDBOOL_H': 1})
         if stdint: conf.env.Append(CPPDEFINES = {'HAVE_STDINT_H': 1})
@@ -101,7 +102,7 @@ if 'gcc' in env['CC']:
     # GNU Compiler Collection
     
     # Debug info, warnings as errors
-    env.Append(CFLAGS = '-ansi -pedantic -g -Wall -Werror -fprofile-arcs -ftest-coverage -fstack-protector-all')
+    env.Append(CFLAGS = '-ansi -pedantic -g -Wall -Werror -fprofile-arcs -ftest-coverage ')
     env.Append(CORECFLAGS = '-Wextra')
     env.Append(LINKFLAGS = '-g --coverage')
     

tests/common/unittestproto.proto

@@ -34,3 +34,8 @@ message CallbackContainer {
 message CallbackContainerContainer {
     required CallbackContainer submsg = 1;
 }
+
+message StringPointerContainer {
+    repeated string rep_str = 1 [(nanopb).type = FT_POINTER];
+}
+

tests/cyclic_messages/SConscript

@@ -0,0 +1,11 @@
+Import("env")
+
+# Encode cyclic messages with callback fields
+
+c = Copy("$TARGET", "$SOURCE")
+env.Command("cyclic_callback.proto", "cyclic.proto", c)
+env.NanopbProto(["cyclic_callback", "cyclic_callback.options"])
+
+enc_callback = env.Program(["encode_cyclic_callback.c", "cyclic_callback.pb.c", "$COMMON/pb_encode.o", "$COMMON/pb_common.o"])
+
+

tests/cyclic_messages/cyclic.proto

@@ -0,0 +1,25 @@
+// Test structures with cyclic references.
+// These can only be handled in pointer/callback mode,
+// see associated .options files.
+
+message TreeNode
+{
+    optional int32 leaf = 1;
+    optional TreeNode left = 2;
+    optional TreeNode right = 3;
+}
+
+message Dictionary
+{
+    repeated KeyValuePair dictItem = 1;
+}
+
+message KeyValuePair
+{
+    required string key = 1;
+    optional string stringValue = 2;
+    optional int32 intValue = 3;
+    optional Dictionary dictValue = 4;
+    optional TreeNode treeValue = 5;
+}
+

tests/cyclic_messages/cyclic_callback.options

@@ -0,0 +1,7 @@
+TreeNode.left               type:FT_CALLBACK
+TreeNode.right              type:FT_CALLBACK
+
+Dictionary.data             type:FT_CALLBACK
+KeyValuePair.key            max_size:8
+KeyValuePair.stringValue    max_size:8
+KeyValuePair.treeValue      type:FT_CALLBACK

tests/cyclic_messages/encode_cyclic_callback.c

@@ -0,0 +1,148 @@
+/* This program parses an input string in a format a bit like JSON:
+ * {'foobar': 1234, 'xyz': 'abc', 'tree': [[[1, 2], 3], [4, 5]]}
+ * and encodes it as protobuf
+ *
+ * Note: The string parsing here is not in any way intended to be robust
+ *       nor safe against buffer overflows. It is just for this test.
+ */
+
+#include <pb_encode.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "cyclic_callback.pb.h"
+
+static char *find_end_of_item(char *p)
+{
+    int depth = 0;
+    do {
+        if (*p == '[' || *p == '{') depth++;
+        if (*p == ']' || *p == '}') depth--;
+        p++;
+    } while (depth > 0 || (*p != ',' && *p != '}'));
+    
+    if (*p == '}')
+        return p; /* End of parent dict */
+    
+    p++;
+    while (*p == ' ') p++;
+    return p;
+}
+
+/* Parse a tree in format [[1 2] 3] and encode it directly to protobuf */
+static bool encode_tree(pb_ostream_t *stream, const pb_field_t *field, void * const *arg)
+{
+    TreeNode tree = TreeNode_init_zero;
+    char *p = (char*)*arg;
+    
+    if (*p == '[')
+    {
+        /* This is a tree branch */
+        p++;
+        tree.left.funcs.encode = encode_tree;
+        tree.left.arg = p;
+        
+        p = find_end_of_item(p);
+        tree.right.funcs.encode = encode_tree;
+        tree.right.arg = p;
+    }
+    else
+    {
+        /* This is a leaf node */
+        tree.has_leaf = true;
+        tree.leaf = atoi(p);
+    }
+    
+    return pb_encode_tag_for_field(stream, field) &&
+           pb_encode_submessage(stream, TreeNode_fields, &tree);
+}
+
+/* Parse a dictionary in format {'name': value} and encode it directly to protobuf */
+static bool encode_dictionary(pb_ostream_t *stream, const pb_field_t *field, void * const *arg)
+{
+    int textlen;
+    char *p = (char*)*arg;
+    if (*p == '{') p++;
+    while (*p != '}')
+    {
+        KeyValuePair pair = KeyValuePair_init_zero;
+        
+        if (*p != '\'')
+            PB_RETURN_ERROR(stream, "invalid key, missing quote");
+        
+        p++; /* Starting quote of key */
+        textlen = strchr(p, '\'') - p;
+        strncpy(pair.key, p, textlen);
+        pair.key[textlen] = 0;
+        p += textlen + 2;
+        
+        while (*p == ' ') p++;
+        
+        if (*p == '[')
+        {
+            /* Value is a tree */
+            pair.treeValue.funcs.encode = encode_tree;
+            pair.treeValue.arg = p;
+        }
+        else if (*p == '\'')
+        {
+            /* Value is a string */
+            pair.has_stringValue = true;
+            p++;
+            textlen = strchr(p, '\'') - p;
+            strncpy(pair.stringValue, p, textlen);
+            pair.stringValue[textlen] = 0;
+        }
+        else if (*p == '{')
+        {
+            /* Value is a dictionary */
+            pair.has_dictValue = true;
+            pair.dictValue.dictItem.funcs.encode = encode_dictionary;
+            pair.dictValue.dictItem.arg = p;
+        }
+        else
+        {
+            /* Value is integer */
+            pair.has_intValue = true;
+            pair.intValue = atoi(p);
+        }
+        
+        p = find_end_of_item(p);
+        
+        if (!pb_encode_tag_for_field(stream, field))
+            return false;
+        
+        if (!pb_encode_submessage(stream, KeyValuePair_fields, &pair))
+            return false;
+    }
+
+    return true;
+}
+
+
+int main(int argc, char *argv[])
+{
+    uint8_t buffer[256];
+    pb_ostream_t stream = pb_ostream_from_buffer(buffer, sizeof(buffer));
+    Dictionary dict = Dictionary_init_zero;
+    
+    if (argc <= 1)
+    {
+        fprintf(stderr, "Usage: %s \"{'foobar': 1234, ...}\"\n", argv[0]);
+        return 1;
+    }
+    
+    dict.dictItem.funcs.encode = encode_dictionary;
+    dict.dictItem.arg = argv[1];
+
+    if (!pb_encode(&stream, Dictionary_fields, &dict))
+    {
+        fprintf(stderr, "Encoding error: %s\n", PB_GET_ERROR(&stream));
+        return 1;
+    }
+    
+    fwrite(buffer, 1, stream.bytes_written, stdout);
+    return 0;
+}
+
+

tests/decode_unittests/decode_unittests.c

@@ -87,6 +87,20 @@ int main()
               pb_decode_varint(&s, (uint64_t*)&i) && i == -1));
         TEST((s = S("\xFF\xFF\xFF\xFF\xFF\xFF\xFF\xFF\xFF\x01"),
               pb_decode_varint(&s, &u) && u == UINT64_MAX));
+        TEST((s = S("\xFF\xFF\xFF\xFF\xFF\xFF\xFF\xFF\xFF\xFF\x01"),
+              !pb_decode_varint(&s, &u)));
+    }
+    
+    {
+        pb_istream_t s;
+        uint32_t u;
+        
+        COMMENT("Test pb_decode_varint32");
+        TEST((s = S("\x00"), pb_decode_varint32(&s, &u) && u == 0));
+        TEST((s = S("\x01"), pb_decode_varint32(&s, &u) && u == 1));
+        TEST((s = S("\xAC\x02"), pb_decode_varint32(&s, &u) && u == 300));
+        TEST((s = S("\xFF\xFF\xFF\xFF\x0F"), pb_decode_varint32(&s, &u) && u == UINT32_MAX));
+        TEST((s = S("\xFF\xFF\xFF\xFF\xFF\x01"), !pb_decode_varint32(&s, &u)));
     }
     
     {

tests/encode_unittests/encode_unittests.c

@@ -331,6 +331,23 @@ int main()
         TEST(s.bytes_written == StringMessage_size);
     }
     
+    {
+        uint8_t buffer[128];
+        pb_ostream_t s;
+        StringPointerContainer msg = StringPointerContainer_init_zero;
+        char *strs[1] = {NULL};
+        char zstr[] = "Z";
+        
+        COMMENT("Test string pointer encoding.");
+        
+        msg.rep_str = strs;
+        msg.rep_str_count = 1;
+        TEST(WRITES(pb_encode(&s, StringPointerContainer_fields, &msg), "\x0a\x00"))
+        
+        strs[0] = zstr;
+        TEST(WRITES(pb_encode(&s, StringPointerContainer_fields, &msg), "\x0a\x01Z"))
+    }
+    
     if (status != 0)
         fprintf(stdout, "\n\nSome tests FAILED!\n");
     

tests/fuzztest/SConscript

@@ -0,0 +1,58 @@
+# Run a fuzz test to verify robustness against corrupted/malicious data.
+
+Import("env")
+
+# We need our own pb_decode.o for the malloc support
+env = env.Clone()
+env.Append(CPPDEFINES = {'PB_ENABLE_MALLOC': 1,
+                         'PB_SYSTEM_HEADER': '\\"fuzz_syshdr.h\\"'})
+env.Append(CPPPATH = ".")
+
+if 'SYSHDR' in env:
+    env.Append(CPPDEFINES = {'PB_OLD_SYSHDR': env['SYSHDR']})
+
+# Disable libmudflap, because it will confuse valgrind
+# and other memory leak detection tools.
+if '-fmudflap' in env["CCFLAGS"]:
+    env["CCFLAGS"].remove("-fmudflap")
+    env["LINKFLAGS"].remove("-fmudflap")
+    env["LIBS"].remove("mudflap")
+
+strict = env.Clone()
+strict.Append(CFLAGS = strict['CORECFLAGS'])
+strict.Object("pb_decode_with_malloc.o", "$NANOPB/pb_decode.c")
+strict.Object("pb_encode_with_malloc.o", "$NANOPB/pb_encode.c")
+strict.Object("pb_common_with_malloc.o", "$NANOPB/pb_common.c")
+
+# We want both pointer and static versions of the AllTypes message
+env.Command("alltypes_static.proto", "#alltypes/alltypes.proto",
+            lambda target, source, env:
+                open(str(target[0]), 'w').write("package alltypes_static;\n"
+                                                + open(str(source[0])).read()))
+env.Command("alltypes_pointer.proto", "#alltypes/alltypes.proto",
+            lambda target, source, env:
+                open(str(target[0]), 'w').write("package alltypes_pointer;\n"
+                                                + open(str(source[0])).read()))
+
+p1 = env.NanopbProto(["alltypes_pointer", "alltypes_pointer.options"])
+p2 = env.NanopbProto(["alltypes_static", "alltypes_static.options"])
+fuzz = env.Program(["fuzztest.c",
+                    "alltypes_pointer.pb.c",
+                    "alltypes_static.pb.c",
+                    "pb_encode_with_malloc.o",
+                    "pb_decode_with_malloc.o",
+                    "pb_common_with_malloc.o",
+                    "malloc_wrappers.c"])
+Depends([p1, p2, fuzz], ["fuzz_syshdr.h", "malloc_wrappers.h"])
+
+env.RunTest(fuzz)
+
+fuzzstub = env.Program(["fuzzstub.c",
+                    "alltypes_pointer.pb.c",
+                    "alltypes_static.pb.c",
+                    "pb_encode_with_malloc.o",
+                    "pb_decode_with_malloc.o",
+                    "pb_common_with_malloc.o",
+                    "malloc_wrappers.c"])
+
+

tests/fuzztest/alltypes_pointer.options

@@ -0,0 +1,3 @@
+# Generate all fields as pointers.
+* type:FT_POINTER
+

tests/fuzztest/alltypes_static.options

@@ -0,0 +1,3 @@
+* max_size:32
+* max_count:8
+*.extensions type:FT_IGNORE

tests/fuzztest/fuzz_syshdr.h

@@ -0,0 +1,15 @@
+/* This is just a wrapper in order to get our own malloc wrappers into nanopb core. */
+
+#define pb_realloc(ptr,size) counting_realloc(ptr,size)
+#define pb_free(ptr) counting_free(ptr)
+
+#ifdef PB_OLD_SYSHDR
+#include PB_OLD_SYSHDR
+#else
+#include <stdint.h>
+#include <stddef.h>
+#include <stdbool.h>
+#include <string.h>
+#endif
+
+#include <malloc_wrappers.h>

tests/fuzztest/fuzzstub.c

@@ -0,0 +1,189 @@
+/* Fuzz testing for the nanopb core.
+ * This can be used with external fuzzers, e.g. radamsa.
+ * It performs most of the same checks as fuzztest, but does not feature data generation.
+ */
+
+#include <pb_decode.h>
+#include <pb_encode.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <assert.h>
+#include <time.h>
+#include "malloc_wrappers.h"
+#include "alltypes_static.pb.h"
+#include "alltypes_pointer.pb.h"
+
+#define BUFSIZE 4096
+
+static bool do_static_decode(uint8_t *buffer, size_t msglen, bool assert_success)
+{
+    pb_istream_t stream;
+    bool status;
+    
+    alltypes_static_AllTypes *msg = malloc_with_check(sizeof(alltypes_static_AllTypes));
+    stream = pb_istream_from_buffer(buffer, msglen);
+    status = pb_decode(&stream, alltypes_static_AllTypes_fields, msg);
+    
+    if (!status && assert_success)
+    {
+        /* Anything that was successfully encoded, should be decodeable.
+         * One exception: strings without null terminator are encoded up
+         * to end of buffer, but refused on decode because the terminator
+         * would not fit. */
+        if (strcmp(stream.errmsg, "string overflow") != 0)
+            assert(status);
+    }
+    
+    free_with_check(msg);
+    return status;
+}
+
+static bool do_pointer_decode(uint8_t *buffer, size_t msglen, bool assert_success)
+{
+    pb_istream_t stream;
+    bool status;
+    alltypes_pointer_AllTypes *msg;
+    
+    msg = malloc_with_check(sizeof(alltypes_pointer_AllTypes));
+    memset(msg, 0, sizeof(alltypes_pointer_AllTypes));
+    stream = pb_istream_from_buffer(buffer, msglen);
+
+    assert(get_alloc_count() == 0);
+    status = pb_decode(&stream, alltypes_pointer_AllTypes_fields, msg);
+    
+    if (assert_success)
+        assert(status);
+    
+    pb_release(alltypes_pointer_AllTypes_fields, msg);    
+    assert(get_alloc_count() == 0);
+    
+    free_with_check(msg);
+
+    return status;
+}
+
+/* Do a decode -> encode -> decode -> encode roundtrip */
+static void do_static_roundtrip(uint8_t *buffer, size_t msglen)
+{
+    bool status;
+    uint8_t *buf2 = malloc_with_check(BUFSIZE);
+    uint8_t *buf3 = malloc_with_check(BUFSIZE);
+    size_t msglen2, msglen3;
+    alltypes_static_AllTypes *msg1 = malloc_with_check(sizeof(alltypes_static_AllTypes));
+    alltypes_static_AllTypes *msg2 = malloc_with_check(sizeof(alltypes_static_AllTypes));
+    memset(msg1, 0, sizeof(alltypes_static_AllTypes));
+    memset(msg2, 0, sizeof(alltypes_static_AllTypes));
+    
+    {
+        pb_istream_t stream = pb_istream_from_buffer(buffer, msglen);
+        status = pb_decode(&stream, alltypes_static_AllTypes_fields, msg1);
+        assert(status);
+    }
+    
+    {
+        pb_ostream_t stream = pb_ostream_from_buffer(buf2, BUFSIZE);
+        status = pb_encode(&stream, alltypes_static_AllTypes_fields, msg1);
+        assert(status);
+        msglen2 = stream.bytes_written;
+    }
+    
+    {
+        pb_istream_t stream = pb_istream_from_buffer(buf2, msglen2);
+        status = pb_decode(&stream, alltypes_static_AllTypes_fields, msg2);
+        assert(status);
+    }
+    
+    {
+        pb_ostream_t stream = pb_ostream_from_buffer(buf3, BUFSIZE);
+        status = pb_encode(&stream, alltypes_static_AllTypes_fields, msg2);
+        assert(status);
+        msglen3 = stream.bytes_written;
+    }
+    
+    assert(msglen2 == msglen3);
+    assert(memcmp(buf2, buf3, msglen2) == 0);
+    
+    free_with_check(msg1);
+    free_with_check(msg2);
+    free_with_check(buf2);
+    free_with_check(buf3);
+}
+
+/* Do decode -> encode -> decode -> encode roundtrip */
+static void do_pointer_roundtrip(uint8_t *buffer, size_t msglen)
+{
+    bool status;
+    uint8_t *buf2 = malloc_with_check(BUFSIZE);
+    uint8_t *buf3 = malloc_with_check(BUFSIZE);
+    size_t msglen2, msglen3;
+    alltypes_pointer_AllTypes *msg1 = malloc_with_check(sizeof(alltypes_pointer_AllTypes));
+    alltypes_pointer_AllTypes *msg2 = malloc_with_check(sizeof(alltypes_pointer_AllTypes));
+    memset(msg1, 0, sizeof(alltypes_pointer_AllTypes));
+    memset(msg2, 0, sizeof(alltypes_pointer_AllTypes));
+    
+    {
+        pb_istream_t stream = pb_istream_from_buffer(buffer, msglen);
+        status = pb_decode(&stream, alltypes_pointer_AllTypes_fields, msg1);
+        assert(status);
+    }
+    
+    {
+        pb_ostream_t stream = pb_ostream_from_buffer(buf2, BUFSIZE);
+        status = pb_encode(&stream, alltypes_pointer_AllTypes_fields, msg1);
+        assert(status);
+        msglen2 = stream.bytes_written;
+    }
+    
+    {
+        pb_istream_t stream = pb_istream_from_buffer(buf2, msglen2);
+        status = pb_decode(&stream, alltypes_pointer_AllTypes_fields, msg2);
+        assert(status);
+    }
+    
+    {
+        pb_ostream_t stream = pb_ostream_from_buffer(buf3, BUFSIZE);
+        status = pb_encode(&stream, alltypes_pointer_AllTypes_fields, msg2);
+        assert(status);
+        msglen3 = stream.bytes_written;
+    }
+    
+    assert(msglen2 == msglen3);
+    assert(memcmp(buf2, buf3, msglen2) == 0);
+    
+    pb_release(alltypes_pointer_AllTypes_fields, msg1);
+    pb_release(alltypes_pointer_AllTypes_fields, msg2);
+    free_with_check(msg1);
+    free_with_check(msg2);
+    free_with_check(buf2);
+    free_with_check(buf3);
+}
+
+static void run_iteration()
+{
+    uint8_t *buffer = malloc_with_check(BUFSIZE);
+    size_t msglen;
+    bool status;
+    
+    msglen = fread(buffer, BUFSIZE, 1, stdin);
+
+    status = do_static_decode(buffer, msglen, false);
+    
+    if (status)
+        do_static_roundtrip(buffer, msglen);
+    
+    status = do_pointer_decode(buffer, msglen, false);
+    
+    if (status)
+        do_pointer_roundtrip(buffer, msglen);
+    
+    free_with_check(buffer);
+}
+
+int main(int argc, char **argv)
+{
+    run_iteration();
+    
+    return 0;
+}
+

tests/fuzztest/fuzztest.c

@@ -0,0 +1,431 @@
+/* Fuzz testing for the nanopb core.
+ * Attempts to verify all the properties defined in the security model document.
+ */
+
+#include <pb_decode.h>
+#include <pb_encode.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <assert.h>
+#include <time.h>
+#include "malloc_wrappers.h"
+#include "alltypes_static.pb.h"
+#include "alltypes_pointer.pb.h"
+
+static uint64_t random_seed;
+
+/* Uses xorshift64 here instead of rand() for both speed and
+ * reproducibility across platforms. */
+static uint32_t rand_word()
+{
+    random_seed ^= random_seed >> 12;
+    random_seed ^= random_seed << 25;
+    random_seed ^= random_seed >> 27;
+    return random_seed * 2685821657736338717ULL;
+}
+
+/* Get a random integer in range, with approximately flat distribution. */
+static int rand_int(int min, int max)
+{
+    return rand_word() % (max + 1 - min) + min;
+}
+
+static bool rand_bool()
+{
+    return rand_word() & 1;
+}
+
+/* Get a random byte, with skewed distribution.
+ * Important corner cases like 0xFF, 0x00 and 0xFE occur more
+ * often than other values. */
+static uint8_t rand_byte()
+{
+    uint32_t w = rand_word();
+    uint8_t b = w & 0xFF;
+    if (w & 0x100000)
+        b >>= (w >> 8) & 7;
+    if (w & 0x200000)
+        b <<= (w >> 12) & 7;
+    if (w & 0x400000)
+        b ^= 0xFF;
+    return b;
+}
+
+/* Get a random length, with skewed distribution.
+ * Favors the shorter lengths, but always atleast 1. */
+static size_t rand_len(size_t max)
+{
+    uint32_t w = rand_word();
+    size_t s;
+    if (w & 0x800000)
+        w &= 3;
+    else if (w & 0x400000)
+        w &= 15;
+    else if (w & 0x200000)
+        w &= 255;
+
+    s = (w % max);
+    if (s == 0)
+        s = 1;
+    
+    return s;
+}
+
+/* Fills a buffer with random data with skewed distribution. */
+static void rand_fill(uint8_t *buf, size_t count)
+{
+    while (count--)
+        *buf++ = rand_byte();
+}
+
+/* Fill with random protobuf-like data */
+static size_t rand_fill_protobuf(uint8_t *buf, size_t min_bytes, size_t max_bytes, int min_tag)
+{
+    pb_ostream_t stream = pb_ostream_from_buffer(buf, max_bytes);
+
+    while(stream.bytes_written < min_bytes)
+    {
+        pb_wire_type_t wt = rand_int(0, 3);
+        if (wt == 3) wt = 5; /* Gap in values */
+        
+        if (!pb_encode_tag(&stream, wt, rand_int(min_tag, min_tag + 512)))
+            break;
+    
+        if (wt == PB_WT_VARINT)
+        {
+            uint64_t value;
+            rand_fill((uint8_t*)&value, sizeof(value));
+            pb_encode_varint(&stream, value);
+        }
+        else if (wt == PB_WT_64BIT)
+        {
+            uint64_t value;
+            rand_fill((uint8_t*)&value, sizeof(value));
+            pb_encode_fixed64(&stream, &value);
+        }
+        else if (wt == PB_WT_32BIT)
+        {
+            uint32_t value;
+            rand_fill((uint8_t*)&value, sizeof(value));
+            pb_encode_fixed32(&stream, &value);
+        }
+        else if (wt == PB_WT_STRING)
+        {
+            size_t len;
+            uint8_t *buf;
+            
+            if (min_bytes > stream.bytes_written)
+                len = rand_len(min_bytes - stream.bytes_written);
+            else
+                len = 0;
+            
+            buf = malloc(len);
+            pb_encode_varint(&stream, len);
+            rand_fill(buf, len);
+            pb_write(&stream, buf, len);
+            free(buf);
+        }
+    }
+    
+    return stream.bytes_written;
+}
+
+/* Given a buffer of data, mess it up a bit */
+static void rand_mess(uint8_t *buf, size_t count)
+{
+    int m = rand_int(0, 3);
+    
+    if (m == 0)
+    {
+        /* Replace random substring */
+        int s = rand_int(0, count - 1);
+        int l = rand_len(count - s);
+        rand_fill(buf + s, l);
+    }
+    else if (m == 1)
+    {
+        /* Swap random bytes */
+        int a = rand_int(0, count - 1);
+        int b = rand_int(0, count - 1);
+        int x = buf[a];
+        buf[a] = buf[b];
+        buf[b] = x;
+    }
+    else if (m == 2)
+    {
+        /* Duplicate substring */
+        int s = rand_int(0, count - 2);
+        int l = rand_len((count - s) / 2);
+        memcpy(buf + s + l, buf + s, l);
+    }
+    else if (m == 3)
+    {
+        /* Add random protobuf noise */
+        int s = rand_int(0, count - 1);
+        int l = rand_len(count - s);
+        rand_fill_protobuf(buf + s, l, count - s, 1);
+    }
+}
+
+/* Some default data to put in the message */
+static const alltypes_static_AllTypes initval = alltypes_static_AllTypes_init_default;
+
+#define BUFSIZE 4096
+
+static bool do_static_encode(uint8_t *buffer, size_t *msglen)
+{
+    pb_ostream_t stream;
+    bool status;
+
+    /* Allocate a message and fill it with defaults */
+    alltypes_static_AllTypes *msg = malloc_with_check(sizeof(alltypes_static_AllTypes));
+    memcpy(msg, &initval, sizeof(initval));
+
+    /* Apply randomness to the data before encoding */
+    while (rand_int(0, 7))
+        rand_mess((uint8_t*)msg, sizeof(alltypes_static_AllTypes));
+
+    stream = pb_ostream_from_buffer(buffer, BUFSIZE);
+    status = pb_encode(&stream, alltypes_static_AllTypes_fields, msg);
+    assert(stream.bytes_written <= BUFSIZE);
+    assert(stream.bytes_written <= alltypes_static_AllTypes_size);
+    
+    *msglen = stream.bytes_written;
+    free_with_check(msg);
+    
+    return status;
+}
+
+/* Append or prepend protobuf noise */
+static void do_protobuf_noise(uint8_t *buffer, size_t *msglen)
+{
+    int m = rand_int(0, 2);
+    size_t max_size = BUFSIZE - 32 - *msglen;
+    if (m == 1)
+    {
+        /* Prepend */
+        uint8_t *tmp = malloc_with_check(BUFSIZE);
+        size_t s = rand_fill_protobuf(tmp, rand_len(max_size), BUFSIZE - *msglen, 512);
+        memmove(buffer + s, buffer, *msglen);
+        memcpy(buffer, tmp, s);
+        free_with_check(tmp);
+        *msglen += s;
+    }
+    else if (m == 2)
+    {
+        /* Append */
+        size_t s = rand_fill_protobuf(buffer + *msglen, rand_len(max_size), BUFSIZE - *msglen, 512);
+        *msglen += s;
+    }
+}
+
+static bool do_static_decode(uint8_t *buffer, size_t msglen, bool assert_success)
+{
+    pb_istream_t stream;
+    bool status;
+    
+    alltypes_static_AllTypes *msg = malloc_with_check(sizeof(alltypes_static_AllTypes));
+    rand_fill((uint8_t*)msg, sizeof(alltypes_static_AllTypes));
+    stream = pb_istream_from_buffer(buffer, msglen);
+    status = pb_decode(&stream, alltypes_static_AllTypes_fields, msg);
+    
+    if (!status && assert_success)
+    {
+        /* Anything that was successfully encoded, should be decodeable.
+         * One exception: strings without null terminator are encoded up
+         * to end of buffer, but refused on decode because the terminator
+         * would not fit. */
+        if (strcmp(stream.errmsg, "string overflow") != 0)
+            assert(status);
+    }
+    
+    free_with_check(msg);
+    return status;
+}
+
+static bool do_pointer_decode(uint8_t *buffer, size_t msglen, bool assert_success)
+{
+    pb_istream_t stream;
+    bool status;
+    alltypes_pointer_AllTypes *msg;
+    
+    msg = malloc_with_check(sizeof(alltypes_pointer_AllTypes));
+    memset(msg, 0, sizeof(alltypes_pointer_AllTypes));
+    stream = pb_istream_from_buffer(buffer, msglen);
+
+    assert(get_alloc_count() == 0);
+    status = pb_decode(&stream, alltypes_pointer_AllTypes_fields, msg);
+    
+    if (assert_success)
+        assert(status);
+    
+    pb_release(alltypes_pointer_AllTypes_fields, msg);    
+    assert(get_alloc_count() == 0);
+    
+    free_with_check(msg);
+
+    return status;
+}
+
+/* Do a decode -> encode -> decode -> encode roundtrip */
+static void do_static_roundtrip(uint8_t *buffer, size_t msglen)
+{
+    bool status;
+    uint8_t *buf2 = malloc_with_check(BUFSIZE);
+    uint8_t *buf3 = malloc_with_check(BUFSIZE);
+    size_t msglen2, msglen3;
+    alltypes_static_AllTypes *msg1 = malloc_with_check(sizeof(alltypes_static_AllTypes));
+    alltypes_static_AllTypes *msg2 = malloc_with_check(sizeof(alltypes_static_AllTypes));
+    memset(msg1, 0, sizeof(alltypes_static_AllTypes));
+    memset(msg2, 0, sizeof(alltypes_static_AllTypes));
+    
+    {
+        pb_istream_t stream = pb_istream_from_buffer(buffer, msglen);
+        status = pb_decode(&stream, alltypes_static_AllTypes_fields, msg1);
+        assert(status);
+    }
+    
+    {
+        pb_ostream_t stream = pb_ostream_from_buffer(buf2, BUFSIZE);
+        status = pb_encode(&stream, alltypes_static_AllTypes_fields, msg1);
+        assert(status);
+        msglen2 = stream.bytes_written;
+    }
+    
+    {
+        pb_istream_t stream = pb_istream_from_buffer(buf2, msglen2);
+        status = pb_decode(&stream, alltypes_static_AllTypes_fields, msg2);
+        assert(status);
+    }
+    
+    {
+        pb_ostream_t stream = pb_ostream_from_buffer(buf3, BUFSIZE);
+        status = pb_encode(&stream, alltypes_static_AllTypes_fields, msg2);
+        assert(status);
+        msglen3 = stream.bytes_written;
+    }
+    
+    assert(msglen2 == msglen3);
+    assert(memcmp(buf2, buf3, msglen2) == 0);
+    
+    free_with_check(msg1);
+    free_with_check(msg2);
+    free_with_check(buf2);
+    free_with_check(buf3);
+}
+
+/* Do decode -> encode -> decode -> encode roundtrip */
+static void do_pointer_roundtrip(uint8_t *buffer, size_t msglen)
+{
+    bool status;
+    uint8_t *buf2 = malloc_with_check(BUFSIZE);
+    uint8_t *buf3 = malloc_with_check(BUFSIZE);
+    size_t msglen2, msglen3;
+    alltypes_pointer_AllTypes *msg1 = malloc_with_check(sizeof(alltypes_pointer_AllTypes));
+    alltypes_pointer_AllTypes *msg2 = malloc_with_check(sizeof(alltypes_pointer_AllTypes));
+    memset(msg1, 0, sizeof(alltypes_pointer_AllTypes));
+    memset(msg2, 0, sizeof(alltypes_pointer_AllTypes));
+    
+    {
+        pb_istream_t stream = pb_istream_from_buffer(buffer, msglen);
+        status = pb_decode(&stream, alltypes_pointer_AllTypes_fields, msg1);
+        assert(status);
+    }
+    
+    {
+        pb_ostream_t stream = pb_ostream_from_buffer(buf2, BUFSIZE);
+        status = pb_encode(&stream, alltypes_pointer_AllTypes_fields, msg1);
+        assert(status);
+        msglen2 = stream.bytes_written;
+    }
+    
+    {
+        pb_istream_t stream = pb_istream_from_buffer(buf2, msglen2);
+        status = pb_decode(&stream, alltypes_pointer_AllTypes_fields, msg2);
+        assert(status);
+    }
+    
+    {
+        pb_ostream_t stream = pb_ostream_from_buffer(buf3, BUFSIZE);
+        status = pb_encode(&stream, alltypes_pointer_AllTypes_fields, msg2);
+        assert(status);
+        msglen3 = stream.bytes_written;
+    }
+    
+    assert(msglen2 == msglen3);
+    assert(memcmp(buf2, buf3, msglen2) == 0);
+    
+    pb_release(alltypes_pointer_AllTypes_fields, msg1);
+    pb_release(alltypes_pointer_AllTypes_fields, msg2);
+    free_with_check(msg1);
+    free_with_check(msg2);
+    free_with_check(buf2);
+    free_with_check(buf3);
+}
+
+static void run_iteration()
+{
+    uint8_t *buffer = malloc_with_check(BUFSIZE);
+    size_t msglen;
+    bool status;
+    
+    rand_fill(buffer, BUFSIZE);
+
+    if (do_static_encode(buffer, &msglen))
+    {
+        do_protobuf_noise(buffer, &msglen);
+    
+        status = do_static_decode(buffer, msglen, true);
+        
+        if (status)
+            do_static_roundtrip(buffer, msglen);
+        
+        status = do_pointer_decode(buffer, msglen, true);
+        
+        if (status)
+            do_pointer_roundtrip(buffer, msglen);
+        
+        /* Apply randomness to the encoded data */
+        while (rand_bool())
+            rand_mess(buffer, BUFSIZE);
+        
+        /* Apply randomness to encoded data length */
+        if (rand_bool())
+            msglen = rand_int(0, BUFSIZE);
+        
+        status = do_static_decode(buffer, msglen, false);
+        do_pointer_decode(buffer, msglen, status);
+        
+        if (status)
+        {
+            do_static_roundtrip(buffer, msglen);
+            do_pointer_roundtrip(buffer, msglen);
+        }
+    }
+    
+    free_with_check(buffer);
+}
+
+int main(int argc, char **argv)
+{
+    int i;
+    if (argc > 1)
+    {
+        random_seed = atol(argv[1]);
+    }
+    else
+    {
+        random_seed = time(NULL);
+    }
+    
+    fprintf(stderr, "Random seed: %llu\n", (long long unsigned)random_seed);
+    
+    for (i = 0; i < 10000; i++)
+    {
+        run_iteration();
+    }
+    
+    return 0;
+}
+

tests/fuzztest/malloc_wrappers.c

@@ -0,0 +1,54 @@
+#include "malloc_wrappers.h"
+#include <stdint.h>
+#include <assert.h>
+#include <string.h>
+
+static size_t alloc_count = 0;
+
+/* Allocate memory and place check values before and after. */
+void* malloc_with_check(size_t size)
+{
+    size_t size32 = (size + 3) / 4 + 3;
+    uint32_t *buf = malloc(size32 * sizeof(uint32_t));
+    buf[0] = size32;
+    buf[1] = 0xDEADBEEF;
+    buf[size32 - 1] = 0xBADBAD;
+    return buf + 2;
+}
+
+/* Free memory allocated with malloc_with_check() and do the checks. */
+void free_with_check(void *mem)
+{
+    uint32_t *buf = (uint32_t*)mem - 2;
+    assert(buf[1] == 0xDEADBEEF);
+    assert(buf[buf[0] - 1] == 0xBADBAD);
+    free(buf);
+}
+
+/* Track memory usage */
+void* counting_realloc(void *ptr, size_t size)
+{
+    /* Don't allocate crazy amounts of RAM when fuzzing */
+    if (size > 1000000)
+        return NULL;
+
+    if (!ptr && size)
+        alloc_count++;
+    
+    return realloc(ptr, size);
+}
+
+void counting_free(void *ptr)
+{
+    if (ptr)
+    {
+        assert(alloc_count > 0);
+        alloc_count--;
+        free(ptr);
+    }
+}
+
+size_t get_alloc_count()
+{
+    return alloc_count;
+}

tests/fuzztest/malloc_wrappers.h

@@ -0,0 +1,7 @@
+#include <stdlib.h>
+
+void* malloc_with_check(size_t size);
+void free_with_check(void *mem);
+void* counting_realloc(void *ptr, size_t size);
+void counting_free(void *ptr);
+size_t get_alloc_count();

tests/fuzztest/run_radamsa.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+
+TMP=`tempfile`
+
+echo $TMP
+while true
+do
+   radamsa sample_data/* > $TMP
+   $1 < $TMP
+   test $? -gt 127 && break
+done
+ 

tests/io_errors/SConscript

@@ -0,0 +1,15 @@
+# Simulate io errors when encoding and decoding
+
+Import("env")
+
+c = Copy("$TARGET", "$SOURCE")
+env.Command("alltypes.proto", "#alltypes/alltypes.proto", c)
+
+env.NanopbProto(["alltypes", "alltypes.options"])
+
+ioerr = env.Program(["io_errors.c", "alltypes.pb.c",
+                     "$COMMON/pb_encode.o", "$COMMON/pb_decode.o", "$COMMON/pb_common.o"])
+
+env.RunTest("io_errors.output", [ioerr, "$BUILD/alltypes/encode_alltypes.output"])
+
+

tests/io_errors/alltypes.options

@@ -0,0 +1,3 @@
+* max_size:16
+* max_count:5
+

tests/io_errors/io_errors.c

@@ -0,0 +1,140 @@
+/* Simulate IO errors after each byte in a stream.
+ * Verifies proper error propagation.
+ */
+
+#include <stdio.h>
+#include <pb_decode.h>
+#include <pb_encode.h>
+#include "alltypes.pb.h"
+#include "test_helpers.h"
+
+typedef struct
+{
+    uint8_t *buffer;
+    size_t fail_after;
+} faulty_stream_t;
+
+bool read_callback(pb_istream_t *stream, uint8_t *buf, size_t count)
+{
+    faulty_stream_t *state = stream->state;
+    
+    while (count--)
+    {
+        if (state->fail_after == 0)
+            PB_RETURN_ERROR(stream, "simulated");
+        state->fail_after--;
+        *buf++ = *state->buffer++;
+    }
+    
+    return true;
+}
+bool write_callback(pb_ostream_t *stream, const uint8_t *buf, size_t count)
+{
+    faulty_stream_t *state = stream->state;
+    
+    while (count--)
+    {
+        if (state->fail_after == 0)
+            PB_RETURN_ERROR(stream, "simulated");
+        state->fail_after--;
+        *state->buffer++ = *buf++;
+    }
+    
+    return true;
+}
+
+int main()
+{
+    uint8_t buffer[2048];
+    size_t msglen;
+    AllTypes msg = AllTypes_init_zero;
+    
+    /* Get some base data to run the tests with */
+    SET_BINARY_MODE(stdin);
+    msglen = fread(buffer, 1, sizeof(buffer), stdin);
+    
+    /* Test IO errors on decoding */
+    {
+        bool status;
+        pb_istream_t stream = {&read_callback, NULL, SIZE_MAX};
+        faulty_stream_t fs;
+        size_t i;
+        
+        for (i = 0; i < msglen; i++)
+        {
+            stream.bytes_left = msglen;
+            stream.state = &fs;
+            fs.buffer = buffer;
+            fs.fail_after = i;
+
+            status = pb_decode(&stream, AllTypes_fields, &msg);
+            if (status != false)
+            {
+                fprintf(stderr, "Unexpected success in decode\n");
+                return 2;
+            }
+            else if (strcmp(stream.errmsg, "simulated") != 0)
+            {
+                fprintf(stderr, "Wrong error in decode: %s\n", stream.errmsg);
+                return 3;
+            }
+        }
+        
+        stream.bytes_left = msglen;
+        stream.state = &fs;
+        fs.buffer = buffer;
+        fs.fail_after = msglen;
+        status = pb_decode(&stream, AllTypes_fields, &msg);
+        
+        if (!status)
+        {
+            fprintf(stderr, "Decoding failed: %s\n", stream.errmsg);
+            return 4;
+        }
+    }
+    
+    /* Test IO errors on encoding */
+    {
+        bool status;
+        pb_ostream_t stream = {&write_callback, NULL, SIZE_MAX, 0};
+        faulty_stream_t fs;
+        size_t i;
+        
+        for (i = 0; i < msglen; i++)
+        {
+            stream.max_size = msglen;
+            stream.bytes_written = 0;
+            stream.state = &fs;
+            fs.buffer = buffer;
+            fs.fail_after = i;
+
+            status = pb_encode(&stream, AllTypes_fields, &msg);
+            if (status != false)
+            {
+                fprintf(stderr, "Unexpected success in encode\n");
+                return 5;
+            }
+            else if (strcmp(stream.errmsg, "simulated") != 0)
+            {
+                fprintf(stderr, "Wrong error in encode: %s\n", stream.errmsg);
+                return 6;
+            }
+        }
+        
+        stream.max_size = msglen;
+        stream.bytes_written = 0;
+        stream.state = &fs;
+        fs.buffer = buffer;
+        fs.fail_after = msglen;
+        status = pb_encode(&stream, AllTypes_fields, &msg);
+        
+        if (!status)
+        {
+            fprintf(stderr, "Encoding failed: %s\n", stream.errmsg);
+            return 7;
+        }
+    }
+
+    return 0;   
+}
+

tests/io_errors_pointers/SConscript

@@ -0,0 +1,42 @@
+# Simulate io errors when encoding and decoding
+
+Import("env")
+
+# We need our own pb_decode.o for the malloc support
+env = env.Clone()
+env.Append(CPPDEFINES = {'PB_ENABLE_MALLOC': 1});
+
+# Disable libmudflap, because it will confuse valgrind
+# and other memory leak detection tools.
+if '-fmudflap' in env["CCFLAGS"]:
+    env["CCFLAGS"].remove("-fmudflap")
+    env["LINKFLAGS"].remove("-fmudflap")
+    env["LIBS"].remove("mudflap")
+
+strict = env.Clone()
+strict.Append(CFLAGS = strict['CORECFLAGS'])
+strict.Object("pb_decode_with_malloc.o", "$NANOPB/pb_decode.c")
+strict.Object("pb_encode_with_malloc.o", "$NANOPB/pb_encode.c")
+strict.Object("pb_common_with_malloc.o", "$NANOPB/pb_common.c")
+
+c = Copy("$TARGET", "$SOURCE")
+env.Command("alltypes.proto", "#alltypes/alltypes.proto", c)
+env.Command("io_errors.c", "#io_errors/io_errors.c", c)
+
+env.NanopbProto(["alltypes", "alltypes.options"])
+
+ioerr = env.Program(["io_errors.c", "alltypes.pb.c",
+                     "pb_encode_with_malloc.o",
+                     "pb_decode_with_malloc.o",
+                     "pb_common_with_malloc.o"])
+
+# Run tests under valgrind if available
+valgrind = env.WhereIs('valgrind')
+kwargs = {}
+if valgrind:
+    kwargs['COMMAND'] = valgrind
+    kwargs['ARGS'] = ["-q", ioerr[0].abspath]
+
+env.RunTest("io_errors.output", [ioerr, "$BUILD/alltypes/encode_alltypes.output"], **kwargs)
+
+

tests/io_errors_pointers/alltypes.options

@@ -0,0 +1,3 @@
+# Generate all fields as pointers.
+* type:FT_POINTER
+