]> arthur.barton.de Git - bup.git/blobdiff - lib/bup/_helpers.c
bitmatch: check for overflow via intprops
[bup.git] / lib / bup / _helpers.c
index 4ab4d4e92bd3ba5662e0928925a63fa35b99537d..b5a3c799489e72d051e4735e2207c75e91196a11 100644 (file)
@@ -66,6 +66,7 @@
 #endif
 
 #include "bupsplit.h"
+#include "bup/intprops.h"
 
 #if defined(FS_IOC_GETFLAGS) && defined(FS_IOC_SETFLAGS)
 #define BUP_HAVE_FILE_ATTRS 1
@@ -124,16 +125,10 @@ static void *checked_calloc(size_t n, size_t size)
     return result;
 }
 
-#ifndef BUP_HAVE_BUILTIN_MUL_OVERFLOW
-
-#define checked_malloc checked_calloc
-
-#else // defined BUP_HAVE_BUILTIN_MUL_OVERFLOW
-
 static void *checked_malloc(size_t n, size_t size)
 {
     size_t total;
-    if (__builtin_mul_overflow(n, size, &total))
+    if (!INT_MULTIPLY_OK(n, size, &total))
     {
         PyErr_Format(PyExc_OverflowError,
                      "request to allocate %zu items of size %zu is too large",
@@ -146,8 +141,6 @@ static void *checked_malloc(size_t n, size_t size)
     return result;
 }
 
-#endif // defined BUP_HAVE_BUILTIN_MUL_OVERFLOW
-
 
 #ifndef htonll
 // This function should technically be macro'd out if it's going to be used
@@ -163,37 +156,10 @@ static uint64_t htonll(uint64_t value)
 }
 #endif
 
+#define INTEGRAL_ASSIGNMENT_FITS(dest, src) INT_ADD_OK(src, 0, dest)
 
-// Disabling sign-compare here should be fine since we're explicitly
-// checking for a sign mismatch, i.e. if the signs don't match, then
-// it doesn't matter what the value comparison says.
-// FIXME: ... so should we reverse the order?
-#define INTEGRAL_ASSIGNMENT_FITS(dest, src)                             \
-    ({                                                                  \
-        _Pragma("GCC diagnostic push");                                 \
-        _Pragma("GCC diagnostic ignored \"-Wsign-compare\"");           \
-        _Pragma("clang diagnostic push");                               \
-        _Pragma("clang diagnostic ignored \"-Wshorten-64-to-32\"");     \
-        *(dest) = (src);                                                \
-        int result = *(dest) == (src) && (*(dest) < 1) == ((src) < 1);  \
-        _Pragma("clang diagnostic pop");                                \
-        _Pragma("GCC diagnostic pop");                                  \
-        result;                                                         \
-    })
-
-
-#define INTEGER_TO_PY(x)                                                \
-    ({                                                                  \
-        _Pragma("GCC diagnostic push");                                 \
-        _Pragma("GCC diagnostic ignored \"-Wtype-limits\"");   \
-        _Pragma("clang diagnostic push");                               \
-        _Pragma("clang diagnostic ignored \"-Wtautological-compare\""); \
-        PyObject *result = ((x) >= 0) ? PyLong_FromUnsignedLongLong(x) : PyLong_FromLongLong(x); \
-        _Pragma("clang diagnostic pop");                                \
-        _Pragma("GCC diagnostic pop");                                  \
-        result;                                                         \
-    })
-
+#define INTEGER_TO_PY(x) \
+    EXPR_SIGNED(x) ? PyLong_FromLongLong(x) : PyLong_FromUnsignedLongLong(x)
 
 #if PY_MAJOR_VERSION < 3
 static int bup_ulong_from_pyint(unsigned long *x, PyObject *py,
@@ -372,15 +338,11 @@ static int write_all(int fd, const void *buf, const size_t count)
 }
 
 
-static int uadd(unsigned long long *dest,
-                const unsigned long long x,
-                const unsigned long long y)
+static inline int uadd(unsigned long long *dest,
+                       const unsigned long long x,
+                       const unsigned long long y)
 {
-    const unsigned long long result = x + y;
-    if (result < x || result < y)
-        return 0;
-    *dest = result;
-    return 1;
+    return INT_ADD_OK(x, y, dest);
 }
 
 
@@ -658,8 +620,14 @@ static PyObject *bitmatch(PyObject *self, PyObject *args)
        }
     }
     
-    assert(byte <= (INT_MAX >> 3));
-    return Py_BuildValue("i", byte*8 + bit);
+    unsigned long long result;
+    if (!INT_MULTIPLY_OK(byte, 8, &result)
+        || !INT_ADD_OK(result, bit, &result))
+    {
+        PyErr_Format(PyExc_OverflowError, "bitmatch bit count too large");
+        return NULL;
+    }
+    return PyLong_FromUnsignedLongLong(result);
 }
 
 
@@ -1701,6 +1669,81 @@ static PyObject *bup_mincore(PyObject *self, PyObject *args)
 }
 #endif /* def BUP_MINCORE_BUF_TYPE */
 
+static unsigned int vuint_encode(long long val, char *buf)
+{
+    unsigned int len = 0;
+
+    if (val < 0) {
+        PyErr_SetString(PyExc_Exception, "vuints must not be negative");
+        return 0;
+    }
+
+    do {
+        buf[len] = val & 0x7f;
+
+        val >>= 7;
+        if (val)
+            buf[len] |= 0x80;
+
+        len++;
+    } while (val);
+
+    return len;
+}
+
+static unsigned int vint_encode(long long val, char *buf)
+{
+    unsigned int len = 1;
+    char sign = 0;
+
+    if (val < 0) {
+        sign = 0x40;
+        val = -val;
+    }
+
+    buf[0] = (val & 0x3f) | sign;
+    val >>= 6;
+    if (val)
+        buf[0] |= 0x80;
+
+    while (val) {
+        buf[len] = val & 0x7f;
+        val >>= 7;
+        if (val)
+            buf[len] |= 0x80;
+        len++;
+    }
+
+    return len;
+}
+
+static PyObject *bup_vuint_encode(PyObject *self, PyObject *args)
+{
+    long long val;
+    // size the buffer appropriately - need 8 bits to encode each 7
+    char buf[(sizeof(val) + 1) / 7 * 8];
+
+    if (!PyArg_ParseTuple(args, "L", &val))
+       return NULL;
+
+    unsigned int len = vuint_encode(val, buf);
+    if (!len)
+        return NULL;
+
+    return PyBytes_FromStringAndSize(buf, len);
+}
+
+static PyObject *bup_vint_encode(PyObject *self, PyObject *args)
+{
+    long long val;
+    // size the buffer appropriately - need 8 bits to encode each 7
+    char buf[(sizeof(val) + 1) / 7 * 8];
+
+    if (!PyArg_ParseTuple(args, "L", &val))
+       return NULL;
+
+    return PyBytes_FromStringAndSize(buf, vint_encode(val, buf));
+}
 
 static PyObject *tuple_from_cstrs(char **cstrs)
 {
@@ -1865,10 +1908,18 @@ static char *cstr_from_bytes(PyObject *bytes)
     int rc = PyBytes_AsStringAndSize(bytes, &buf, &length);
     if (rc == -1)
         return NULL;
-    char *result = checked_malloc(length, sizeof(char));
+    size_t c_len;
+    if (!INT_ADD_OK(length, 1, &c_len)) {
+        PyErr_Format(PyExc_OverflowError,
+                     "Cannot convert ssize_t sized bytes object (%zd) to C string",
+                     length);
+        return NULL;
+    }
+    char *result = checked_malloc(c_len, sizeof(char));
     if (!result)
         return NULL;
     memcpy(result, buf, length);
+    result[length] = 0;
     return result;
 }
 
@@ -2193,6 +2244,95 @@ static PyObject *bup_apply_acl(PyObject *self, PyObject *args)
 }
 #endif
 
+static PyObject *bup_limited_vint_pack(PyObject *self, PyObject *args)
+{
+    const char *fmt;
+    PyObject *packargs, *result;
+    Py_ssize_t sz, i, bufsz;
+    char *buf, *pos, *end;
+
+    if (!PyArg_ParseTuple(args, "sO", &fmt, &packargs))
+        return NULL;
+
+    if (!PyTuple_Check(packargs))
+        return PyErr_Format(PyExc_Exception, "pack() arg must be tuple");
+
+    sz = PyTuple_GET_SIZE(packargs);
+    if (sz != (Py_ssize_t)strlen(fmt))
+        return PyErr_Format(PyExc_Exception,
+                            "number of arguments (%ld) does not match format string (%ld)",
+                            (unsigned long)sz, (unsigned long)strlen(fmt));
+
+    if (sz > INT_MAX / 20)
+        return PyErr_Format(PyExc_Exception, "format is far too long");
+
+    // estimate no more than 20 bytes for each on average, the maximum
+    // vint/vuint we can encode is anyway 10 bytes, so this gives us
+    // some headroom for a few strings before we need to realloc ...
+    bufsz = sz * 20;
+    buf = malloc(bufsz);
+    if (!buf)
+        return PyErr_NoMemory();
+
+    pos = buf;
+    end = buf + bufsz;
+    for (i = 0; i < sz; i++) {
+        PyObject *item = PyTuple_GET_ITEM(packargs, i);
+        const char *bytes;
+
+        switch (fmt[i]) {
+        case 'V': {
+            long long val = PyLong_AsLongLong(item);
+            if (val == -1 && PyErr_Occurred())
+                return PyErr_Format(PyExc_OverflowError,
+                                    "pack arg %d invalid", (int)i);
+            if (end - pos < 10)
+                goto overflow;
+           pos += vuint_encode(val, pos);
+            break;
+        }
+        case 'v': {
+            long long val = PyLong_AsLongLong(item);
+            if (val == -1 && PyErr_Occurred())
+                return PyErr_Format(PyExc_OverflowError,
+                                    "pack arg %d invalid", (int)i);
+            if (end - pos < 10)
+                goto overflow;
+            pos += vint_encode(val, pos);
+            break;
+        }
+        case 's': {
+            bytes = PyBytes_AsString(item);
+            if (!bytes)
+                goto error;
+            if (end - pos < 10)
+                goto overflow;
+            Py_ssize_t val = PyBytes_GET_SIZE(item);
+            pos += vuint_encode(val, pos);
+            if (end - pos < val)
+                goto overflow;
+            memcpy(pos, bytes, val);
+            pos += val;
+            break;
+        }
+        default:
+            PyErr_Format(PyExc_Exception, "unknown xpack format string item %c",
+                         fmt[i]);
+            goto error;
+        }
+    }
+
+    result = PyBytes_FromStringAndSize(buf, pos - buf);
+    free(buf);
+    return result;
+
+ overflow:
+    PyErr_SetString(PyExc_OverflowError, "buffer (potentially) overflowed");
+ error:
+    free(buf);
+    return NULL;
+}
+
 static PyMethodDef helper_methods[] = {
     { "write_sparsely", bup_write_sparsely, METH_VARARGS,
       "Write buf excepting zeros at the end. Return trailing zero count." },
@@ -2312,6 +2452,10 @@ static PyMethodDef helper_methods[] = {
       "apply_acl(name, acl, def=None)\n\n"
       "Given a file/dirname (bytes) and the ACLs to restore, do that." },
 #endif /* HAVE_ACLS */
+    { "vuint_encode", bup_vuint_encode, METH_VARARGS, "encode an int to vuint" },
+    { "vint_encode", bup_vint_encode, METH_VARARGS, "encode an int to vint" },
+    { "limited_vint_pack", bup_limited_vint_pack, METH_VARARGS,
+      "Try to pack vint/vuint/str, throwing OverflowError when unable." },
     { NULL, NULL, 0, NULL },  // sentinel
 };
 
@@ -2357,6 +2501,9 @@ static int setup_module(PyObject *m)
     // Just be sure (relevant when passing timestamps back to Python above).
     assert(sizeof(PY_LONG_LONG) <= sizeof(long long));
     assert(sizeof(unsigned PY_LONG_LONG) <= sizeof(unsigned long long));
+    // At least for INTEGER_TO_PY
+    assert(sizeof(intmax_t) <= sizeof(long long));
+    assert(sizeof(uintmax_t) <= sizeof(unsigned long long));
 
     test_integral_assignment_fits();