#include <errno.h>
#include <fcntl.h>
#include <arpa/inet.h>
+#include <stddef.h>
#include <stdint.h>
#include <stdlib.h>
#include <stdio.h>
-#include <sys/mman.h>
+#include <string.h>
+#ifdef HAVE_SYS_MMAN_H
+#include <sys/mman.h>
+#endif
#ifdef HAVE_SYS_TYPES_H
#include <sys/types.h>
#endif
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif
+#ifdef HAVE_SYS_TIME_H
+#include <sys/time.h>
+#endif
#ifdef HAVE_LINUX_FS_H
#include <linux/fs.h>
#include <sys/ioctl.h>
#endif
+#ifdef HAVE_TM_TM_GMTOFF
+#include <time.h>
+#endif
+
#include "bupsplit.h"
#if defined(FS_IOC_GETFLAGS) && defined(FS_IOC_SETFLAGS)
#define FS_NOCOW_FL 0
#endif
-static int istty2 = 0;
+
+typedef unsigned char byte;
+
+
+typedef struct {
+ int istty2;
+} state_t;
+
+// cstr_argf: for byte vectors without null characters (e.g. paths)
+// rbuf_argf: for read-only byte vectors
+// wbuf_argf: for mutable byte vectors
+
+#if PY_MAJOR_VERSION < 3
+static state_t state;
+# define get_state(x) (&state)
+# define cstr_argf "s"
+# define rbuf_argf "s#"
+# define wbuf_argf "s*"
+#else
+# define get_state(x) ((state_t *) PyModule_GetState(x))
+# define cstr_argf "y"
+# define rbuf_argf "y#"
+# define wbuf_argf "y*"
+#endif // PY_MAJOR_VERSION >= 3
+
+
+static void *checked_malloc(size_t n, size_t size)
+{
+ size_t total;
+ if (__builtin_mul_overflow(n, size, &total))
+ {
+ PyErr_Format(PyExc_OverflowError,
+ "request to allocate %lu items of size %lu is too large",
+ n, size);
+ return NULL;
+ }
+ void *result = malloc(total);
+ if (!result)
+ return PyErr_NoMemory();
+ return result;
+}
+
+static void *checked_calloc(size_t n, size_t size)
+{
+ void *result = calloc(n, size);
+ if (!result)
+ PyErr_NoMemory();
+ return result;
+}
#ifndef htonll
#endif
+#ifdef __clang__
#define INTEGRAL_ASSIGNMENT_FITS(dest, src) \
({ \
*(dest) = (src); \
*(dest) == (src) && (*(dest) < 1) == ((src) < 1); \
})
+#else
+// 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\""); \
+ *(dest) = (src); \
+ *(dest) == (src) && (*(dest) < 1) == ((src) < 1); \
+ _Pragma("GCC diagnostic pop"); \
+ })
+#endif
-// At the moment any code that calls INTGER_TO_PY() will have to
+// At the moment any code that calls INTEGER_TO_PY() will have to
// disable -Wtautological-compare for clang. See below.
#define INTEGER_TO_PY(x) \
(((x) >= 0) ? PyLong_FromUnsignedLongLong(x) : PyLong_FromLongLong(x))
+
+#if PY_MAJOR_VERSION < 3
static int bup_ulong_from_pyint(unsigned long *x, PyObject *py,
const char *name)
{
*x = tmp;
return 1;
}
+#endif
static int bup_ulong_from_py(unsigned long *x, PyObject *py, const char *name)
{
+#if PY_MAJOR_VERSION < 3
if (PyInt_Check(py))
return bup_ulong_from_pyint(x, py, name);
+#endif
if (!PyLong_Check(py))
{
static int bup_ullong_from_py(unsigned PY_LONG_LONG *x, PyObject *py,
const char *name)
{
+#if PY_MAJOR_VERSION < 3
if (PyInt_Check(py))
{
unsigned long tmp;
}
return 0;
}
+#endif
if (!PyLong_Check(py))
{
}
+static PyObject *bup_bytescmp(PyObject *self, PyObject *args)
+{
+ PyObject *py_s1, *py_s2; // This is really a PyBytes/PyString
+ if (!PyArg_ParseTuple(args, "SS", &py_s1, &py_s2))
+ return NULL;
+ char *s1, *s2;
+ Py_ssize_t s1_len, s2_len;
+ if (PyBytes_AsStringAndSize(py_s1, &s1, &s1_len) == -1)
+ return NULL;
+ if (PyBytes_AsStringAndSize(py_s2, &s2, &s2_len) == -1)
+ return NULL;
+ const Py_ssize_t n = (s1_len < s2_len) ? s1_len : s2_len;
+ const int cmp = memcmp(s1, s2, n);
+ if (cmp != 0)
+ return PyLong_FromLong(cmp);
+ if (s1_len == s2_len)
+ return PyLong_FromLong(0);;
+ return PyLong_FromLong((s1_len < s2_len) ? -1 : 1);
+}
+
+
// Probably we should use autoconf or something and set HAVE_PY_GETARGCARGV...
#if __WIN32__ || __CYGWIN__
#endif // not __WIN32__ or __CYGWIN__
-static unsigned long long count_leading_zeros(const unsigned char * const buf,
- unsigned long long len)
-{
- const unsigned char *cur = buf;
- while(len-- && *cur == 0)
- cur++;
- return cur - buf;
-}
-
-
static int write_all(int fd, const void *buf, const size_t count)
{
size_t written = 0;
}
+static PyObject *append_sparse_region(const int fd, unsigned long long n)
+{
+ while (n)
+ {
+ off_t new_off;
+ if (!INTEGRAL_ASSIGNMENT_FITS(&new_off, n))
+ new_off = INT_MAX;
+ const off_t off = lseek(fd, new_off, SEEK_CUR);
+ if (off == (off_t) -1)
+ return PyErr_SetFromErrno(PyExc_IOError);
+ n -= new_off;
+ }
+ return NULL;
+}
+
+
+static PyObject *record_sparse_zeros(unsigned long long *new_pending,
+ const int fd,
+ unsigned long long prev_pending,
+ const unsigned long long count)
+{
+ // Add count additional sparse zeros to prev_pending and store the
+ // result in new_pending, or if the total won't fit in
+ // new_pending, write some of the zeros to fd sparsely, and store
+ // the remaining sum in new_pending.
+ if (!uadd(new_pending, prev_pending, count))
+ {
+ PyObject *err = append_sparse_region(fd, prev_pending);
+ if (err != NULL)
+ return err;
+ *new_pending = count;
+ }
+ return NULL;
+}
+
+
+static byte* find_not_zero(const byte * const start, const byte * const end)
+{
+ // Return a pointer to first non-zero byte between start and end,
+ // or end if there isn't one.
+ assert(start <= end);
+ const unsigned char *cur = start;
+ while (cur < end && *cur == 0)
+ cur++;
+ return (byte *) cur;
+}
+
+
+static byte* find_trailing_zeros(const byte * const start,
+ const byte * const end)
+{
+ // Return a pointer to the start of any trailing run of zeros, or
+ // end if there isn't one.
+ assert(start <= end);
+ if (start == end)
+ return (byte *) end;
+ const byte * cur = end;
+ while (cur > start && *--cur == 0) {}
+ if (*cur == 0)
+ return (byte *) cur;
+ else
+ return (byte *) (cur + 1);
+}
+
+
+static byte *find_non_sparse_end(const byte * const start,
+ const byte * const end,
+ const ptrdiff_t min_len)
+{
+ // Return the first pointer to a min_len sparse block in [start,
+ // end) if there is one, otherwise a pointer to the start of any
+ // trailing run of zeros. If there are no trailing zeros, return
+ // end.
+ if (start == end)
+ return (byte *) end;
+ assert(start < end);
+ assert(min_len);
+ // Probe in min_len jumps, searching backward from the jump
+ // destination for a non-zero byte. If such a byte is found, move
+ // just past it and try again.
+ const byte *candidate = start;
+ // End of any run of zeros, starting at candidate, that we've already seen
+ const byte *end_of_known_zeros = candidate;
+ while (end - candidate >= min_len) // Handle all min_len candidate blocks
+ {
+ const byte * const probe_end = candidate + min_len;
+ const byte * const trailing_zeros =
+ find_trailing_zeros(end_of_known_zeros, probe_end);
+ if (trailing_zeros == probe_end)
+ end_of_known_zeros = candidate = probe_end;
+ else if (trailing_zeros == end_of_known_zeros)
+ {
+ assert(candidate >= start);
+ assert(candidate <= end);
+ assert(*candidate == 0);
+ return (byte *) candidate;
+ }
+ else
+ {
+ candidate = trailing_zeros;
+ end_of_known_zeros = probe_end;
+ }
+ }
+
+ if (candidate == end)
+ return (byte *) end;
+
+ // No min_len sparse run found, search backward from end
+ const byte * const trailing_zeros = find_trailing_zeros(end_of_known_zeros,
+ end);
+
+ if (trailing_zeros == end_of_known_zeros)
+ {
+ assert(candidate >= start);
+ assert(candidate < end);
+ assert(*candidate == 0);
+ assert(end - candidate < min_len);
+ return (byte *) candidate;
+ }
+
+ if (trailing_zeros == end)
+ {
+ assert(*(end - 1) != 0);
+ return (byte *) end;
+ }
+
+ assert(end - trailing_zeros < min_len);
+ assert(trailing_zeros >= start);
+ assert(trailing_zeros < end);
+ assert(*trailing_zeros == 0);
+ return (byte *) trailing_zeros;
+}
+
+
static PyObject *bup_write_sparsely(PyObject *self, PyObject *args)
{
int fd;
unsigned char *buf = NULL;
Py_ssize_t sbuf_len;
PyObject *py_min_sparse_len, *py_prev_sparse_len;
- if (!PyArg_ParseTuple(args, "it#OO",
+ if (!PyArg_ParseTuple(args, "i" rbuf_argf "OO",
&fd, &buf, &sbuf_len,
&py_min_sparse_len, &py_prev_sparse_len))
return NULL;
- unsigned long long min_sparse_len, prev_sparse_len, buf_len;
- if (!bup_ullong_from_py(&min_sparse_len, py_min_sparse_len, "min_sparse_len"))
+ ptrdiff_t min_sparse_len;
+ unsigned long long prev_sparse_len, buf_len, ul_min_sparse_len;
+ if (!bup_ullong_from_py(&ul_min_sparse_len, py_min_sparse_len, "min_sparse_len"))
return NULL;
+ if (!INTEGRAL_ASSIGNMENT_FITS(&min_sparse_len, ul_min_sparse_len))
+ return PyErr_Format(PyExc_OverflowError, "min_sparse_len too large");
if (!bup_ullong_from_py(&prev_sparse_len, py_prev_sparse_len, "prev_sparse_len"))
return NULL;
if (sbuf_len < 0)
if (!INTEGRAL_ASSIGNMENT_FITS(&buf_len, sbuf_len))
return PyErr_Format(PyExc_OverflowError, "buffer length too large");
- // For now, there are some cases where we just give up if the
- // values are too large, but we could try to break up the relevant
- // operations into chunks.
-
- // Deal with preceding zeros. Just make them sparse, along with
- // any leading zeros in buf, even if the region's not >= min,
- // since the alternative is a potentially extra small write.
- if (prev_sparse_len)
+ const byte * block = buf; // Start of pending block
+ const byte * const end = buf + buf_len;
+ unsigned long long zeros = prev_sparse_len;
+ while (1)
{
- const unsigned long long zeros = count_leading_zeros(buf, buf_len);
- unsigned long long new_sparse_len = 0;
- if (!uadd(&new_sparse_len, prev_sparse_len, zeros))
- return PyErr_Format (PyExc_OverflowError, "sparse region too large");
- if (zeros == buf_len)
- return PyLong_FromUnsignedLongLong(new_sparse_len);
-
- off_t new_off;
- if (!INTEGRAL_ASSIGNMENT_FITS(&new_off, new_sparse_len))
- return PyErr_Format(PyExc_OverflowError,
- "sparse region too large for seek");
- const off_t off = lseek(fd, new_off, SEEK_CUR);
- if (off == -1)
- return PyErr_SetFromErrno(PyExc_IOError);
- buf += zeros;
- buf_len -= zeros;
- }
+ assert(block <= end);
+ if (block == end)
+ return PyLong_FromUnsignedLongLong(zeros);
- int rc;
- unsigned long long unexamined = buf_len;
- unsigned char *block_start = buf, *cur = buf;
- while(unexamined)
- {
- const unsigned long long zeros = count_leading_zeros(cur, unexamined);
- assert(zeros <= unexamined);
- unexamined -= zeros;
- if (unexamined == 0) // Runs off the end.
+ if (*block != 0)
{
- rc = write_all(fd, block_start, cur - block_start);
+ // Look for the end of block, i.e. the next sparse run of
+ // at least min_sparse_len zeros, or the end of the
+ // buffer.
+ const byte * const probe = find_non_sparse_end(block + 1, end,
+ min_sparse_len);
+ // Either at end of block, or end of non-sparse; write pending data
+ PyObject *err = append_sparse_region(fd, zeros);
+ if (err != NULL)
+ return err;
+ int rc = write_all(fd, block, probe - block);
if (rc)
return PyErr_SetFromErrno(PyExc_IOError);
- return PyLong_FromUnsignedLongLong(zeros);
- }
- cur += zeros;
- if (zeros >= min_sparse_len)
- {
- off_t new_off;
- if (!INTEGRAL_ASSIGNMENT_FITS(&new_off, zeros))
- return PyErr_Format(PyExc_ValueError,
- "zero count overflows off_t");
- off_t off = lseek(fd, new_off, SEEK_CUR);
- if (off == -1)
- return PyErr_SetFromErrno(PyExc_IOError);
- block_start = cur;
+
+ if (end - probe < min_sparse_len)
+ zeros = end - probe;
+ else
+ zeros = min_sparse_len;
+ block = probe + zeros;
}
- while (unexamined && *cur != 0)
+ else // *block == 0
{
- cur++; unexamined--;
+ // Should be in the first loop iteration, a sparse run of
+ // zeros, or nearly at the end of the block (within
+ // min_sparse_len).
+ const byte * const zeros_end = find_not_zero(block, end);
+ PyObject *err = record_sparse_zeros(&zeros, fd,
+ zeros, zeros_end - block);
+ if (err != NULL)
+ return err;
+ assert(block <= zeros_end);
+ block = zeros_end;
}
}
- rc = write_all(fd, block_start, cur - block_start);
- if (rc)
- return PyErr_SetFromErrno(PyExc_IOError);
- return PyInt_FromLong(0);
}
static PyObject *splitbuf(PyObject *self, PyObject *args)
{
- unsigned char *buf = NULL;
- Py_ssize_t len = 0;
+ // We stick to buffers in python 2 because they appear to be
+ // substantially smaller than memoryviews, and because
+ // zlib.compress() in python 2 can't accept a memoryview
+ // (cf. hashsplit.py).
int out = 0, bits = -1;
-
- if (!PyArg_ParseTuple(args, "t#", &buf, &len))
- return NULL;
- assert(len <= INT_MAX);
- out = bupsplit_find_ofs(buf, len, &bits);
+ if (PY_MAJOR_VERSION > 2)
+ {
+ Py_buffer buf;
+ if (!PyArg_ParseTuple(args, "y*", &buf))
+ return NULL;
+ assert(buf.len <= INT_MAX);
+ out = bupsplit_find_ofs(buf.buf, buf.len, &bits);
+ PyBuffer_Release(&buf);
+ }
+ else
+ {
+ unsigned char *buf = NULL;
+ Py_ssize_t len = 0;
+ if (!PyArg_ParseTuple(args, "t#", &buf, &len))
+ return NULL;
+ assert(len <= INT_MAX);
+ out = bupsplit_find_ofs(buf, len, &bits);
+ }
if (out) assert(bits >= BUP_BLOBBITS);
return Py_BuildValue("ii", out, bits);
}
Py_ssize_t byte;
int bit;
- if (!PyArg_ParseTuple(args, "t#t#", &buf1, &len1, &buf2, &len2))
+ if (!PyArg_ParseTuple(args, rbuf_argf rbuf_argf, &buf1, &len1, &buf2, &len2))
return NULL;
bit = 0;
Py_ssize_t len = 0;
uint32_t v;
- if (!PyArg_ParseTuple(args, "t#", &buf, &len))
+ if (!PyArg_ParseTuple(args, rbuf_argf, &buf, &len))
return NULL;
if (len < 4)
static PyObject *bloom_add(PyObject *self, PyObject *args)
{
- unsigned char *sha = NULL, *bloom = NULL;
- unsigned char *end;
- Py_ssize_t len = 0, blen = 0;
+ Py_buffer bloom, sha;
int nbits = 0, k = 0;
+ if (!PyArg_ParseTuple(args, wbuf_argf wbuf_argf "ii",
+ &bloom, &sha, &nbits, &k))
+ return NULL;
- if (!PyArg_ParseTuple(args, "w#s#ii", &bloom, &blen, &sha, &len, &nbits, &k))
- return NULL;
+ PyObject *result = NULL;
- if (blen < 16+(1<<nbits) || len % 20 != 0)
- return NULL;
+ if (bloom.len < 16+(1<<nbits) || sha.len % 20 != 0)
+ goto clean_and_return;
if (k == 5)
{
- if (nbits > 29)
- return NULL;
- for (end = sha + len; sha < end; sha += 20/k)
- bloom_set_bit5(bloom, sha, nbits);
+ if (nbits > 29)
+ goto clean_and_return;
+ unsigned char *cur = sha.buf;
+ unsigned char *end;
+ for (end = cur + sha.len; cur < end; cur += 20/k)
+ bloom_set_bit5(bloom.buf, cur, nbits);
}
else if (k == 4)
{
- if (nbits > 37)
- return NULL;
- for (end = sha + len; sha < end; sha += 20/k)
- bloom_set_bit4(bloom, sha, nbits);
+ if (nbits > 37)
+ goto clean_and_return;
+ unsigned char *cur = sha.buf;
+ unsigned char *end = cur + sha.len;
+ for (; cur < end; cur += 20/k)
+ bloom_set_bit4(bloom.buf, cur, nbits);
}
else
- return NULL;
+ goto clean_and_return;
+ result = Py_BuildValue("n", sha.len / 20);
- return Py_BuildValue("n", len/20);
+ clean_and_return:
+ PyBuffer_Release(&bloom);
+ PyBuffer_Release(&sha);
+ return result;
}
static PyObject *bloom_contains(PyObject *self, PyObject *args)
{
- unsigned char *sha = NULL, *bloom = NULL;
- Py_ssize_t len = 0, blen = 0;
+ Py_buffer bloom;
+ unsigned char *sha = NULL;
+ Py_ssize_t len = 0;
int nbits = 0, k = 0;
- unsigned char *end;
- int steps;
+ if (!PyArg_ParseTuple(args, wbuf_argf rbuf_argf "ii",
+ &bloom, &sha, &len, &nbits, &k))
+ return NULL;
- if (!PyArg_ParseTuple(args, "t#s#ii", &bloom, &blen, &sha, &len, &nbits, &k))
- return NULL;
+ PyObject *result = NULL;
if (len != 20)
- return NULL;
+ goto clean_and_return;
if (k == 5)
{
- if (nbits > 29)
- return NULL;
- for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
- if (!bloom_get_bit5(bloom, sha, nbits))
- return Py_BuildValue("Oi", Py_None, steps);
+ if (nbits > 29)
+ goto clean_and_return;
+ int steps;
+ unsigned char *end;
+ for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
+ if (!bloom_get_bit5(bloom.buf, sha, nbits))
+ {
+ result = Py_BuildValue("Oi", Py_None, steps);
+ goto clean_and_return;
+ }
}
else if (k == 4)
{
- if (nbits > 37)
- return NULL;
- for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
- if (!bloom_get_bit4(bloom, sha, nbits))
- return Py_BuildValue("Oi", Py_None, steps);
+ if (nbits > 37)
+ goto clean_and_return;
+ int steps;
+ unsigned char *end;
+ for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
+ if (!bloom_get_bit4(bloom.buf, sha, nbits))
+ {
+ result = Py_BuildValue("Oi", Py_None, steps);
+ goto clean_and_return;
+ }
}
else
- return NULL;
+ goto clean_and_return;
+
+ result = Py_BuildValue("ii", 1, k);
- return Py_BuildValue("ii", 1, k);
+ clean_and_return:
+ PyBuffer_Release(&bloom);
+ return result;
}
Py_ssize_t len = 0;
int nbits = 0;
- if (!PyArg_ParseTuple(args, "t#i", &buf, &len, &nbits))
+ if (!PyArg_ParseTuple(args, rbuf_argf "i", &buf, &len, &nbits))
return NULL;
if (len < 4)
unsigned char bytes[20];
};
+static inline int _cmp_sha(const struct sha *sha1, const struct sha *sha2)
+{
+ return memcmp(sha1->bytes, sha2->bytes, sizeof(sha1->bytes));
+}
+
struct idx {
unsigned char *map;
int name_base;
};
-
-static int _cmp_sha(const struct sha *sha1, const struct sha *sha2)
-{
- int i;
- for (i = 0; i < sizeof(struct sha); i++)
- if (sha1->bytes[i] != sha2->bytes[i])
- return sha1->bytes[i] - sha2->bytes[i];
- return 0;
-}
-
-
-static void _fix_idx_order(struct idx **idxs, int *last_i)
+static void _fix_idx_order(struct idx **idxs, Py_ssize_t *last_i)
{
struct idx *idx;
int low, mid, high, c = 0;
static PyObject *merge_into(PyObject *self, PyObject *args)
{
- PyObject *py_total, *ilist = NULL;
- unsigned char *fmap = NULL;
struct sha *sha_ptr, *sha_start = NULL;
uint32_t *table_ptr, *name_ptr, *name_start;
- struct idx **idxs = NULL;
- Py_ssize_t flen = 0;
- int bits = 0, i;
+ int i;
unsigned int total;
uint32_t count, prefix;
- int num_i;
- int last_i;
- if (!PyArg_ParseTuple(args, "w#iOO",
- &fmap, &flen, &bits, &py_total, &ilist))
+
+ Py_buffer fmap;
+ int bits;;
+ PyObject *py_total, *ilist = NULL;
+ if (!PyArg_ParseTuple(args, wbuf_argf "iOO",
+ &fmap, &bits, &py_total, &ilist))
return NULL;
+ PyObject *result = NULL;
+ struct idx **idxs = NULL;
+ Py_ssize_t num_i = 0;
+ int *idx_buf_init = NULL;
+ Py_buffer *idx_buf = NULL;
+
if (!bup_uint_from_py(&total, py_total, "total"))
- return NULL;
+ goto clean_and_return;
num_i = PyList_Size(ilist);
- idxs = (struct idx **)PyMem_Malloc(num_i * sizeof(struct idx *));
+
+ if (!(idxs = checked_malloc(num_i, sizeof(struct idx *))))
+ goto clean_and_return;
+ if (!(idx_buf_init = checked_calloc(num_i, sizeof(int))))
+ goto clean_and_return;
+ if (!(idx_buf = checked_malloc(num_i, sizeof(Py_buffer))))
+ goto clean_and_return;
for (i = 0; i < num_i; i++)
{
long len, sha_ofs, name_map_ofs;
- idxs[i] = (struct idx *)PyMem_Malloc(sizeof(struct idx));
+ if (!(idxs[i] = checked_malloc(1, sizeof(struct idx))))
+ goto clean_and_return;
PyObject *itup = PyList_GetItem(ilist, i);
- if (!PyArg_ParseTuple(itup, "t#llli", &idxs[i]->map, &idxs[i]->bytes,
- &len, &sha_ofs, &name_map_ofs, &idxs[i]->name_base))
+ if (!PyArg_ParseTuple(itup, wbuf_argf "llli",
+ &(idx_buf[i]), &len, &sha_ofs, &name_map_ofs,
+ &idxs[i]->name_base))
return NULL;
+ idx_buf_init[i] = 1;
+ idxs[i]->map = idx_buf[i].buf;
+ idxs[i]->bytes = idx_buf[i].len;
idxs[i]->cur = (struct sha *)&idxs[i]->map[sha_ofs];
idxs[i]->end = &idxs[i]->cur[len];
if (name_map_ofs)
else
idxs[i]->cur_name = NULL;
}
- table_ptr = (uint32_t *)&fmap[MIDX4_HEADERLEN];
+ table_ptr = (uint32_t *) &((unsigned char *) fmap.buf)[MIDX4_HEADERLEN];
sha_start = sha_ptr = (struct sha *)&table_ptr[1<<bits];
name_start = name_ptr = (uint32_t *)&sha_ptr[total];
- last_i = num_i-1;
+ Py_ssize_t last_i = num_i - 1;
count = 0;
prefix = 0;
while (last_i >= 0)
{
struct idx *idx;
uint32_t new_prefix;
- if (count % 102424 == 0 && istty2)
+ if (count % 102424 == 0 && get_state(self)->istty2)
fprintf(stderr, "midx: writing %.2f%% (%d/%d)\r",
count*100.0/total, count, total);
idx = idxs[last_i];
_fix_idx_order(idxs, &last_i);
++count;
}
- while (prefix < (1<<bits))
+ while (prefix < ((uint32_t) 1 << bits))
table_ptr[prefix++] = htonl(count);
assert(count == total);
- assert(prefix == (1<<bits));
+ assert(prefix == ((uint32_t) 1 << bits));
assert(sha_ptr == sha_start+count);
assert(name_ptr == name_start+count);
- PyMem_Free(idxs);
- return PyLong_FromUnsignedLong(count);
+ result = PyLong_FromUnsignedLong(count);
+
+ clean_and_return:
+ if (idx_buf_init)
+ {
+ for (i = 0; i < num_i; i++)
+ if (idx_buf_init[i])
+ PyBuffer_Release(&(idx_buf[i]));
+ free(idx_buf_init);
+ free(idx_buf);
+ }
+ if (idxs)
+ {
+ for (i = 0; i < num_i; i++)
+ free(idxs[i]);
+ free(idxs);
+ }
+ PyBuffer_Release(&fmap);
+ return result;
}
#define FAN_ENTRIES 256
char *filename = NULL;
PyObject *py_total, *idx = NULL;
PyObject *part;
- unsigned char *fmap = NULL;
- Py_ssize_t flen = 0;
unsigned int total = 0;
uint32_t count;
int i, j, ofs64_count;
uint64_t *ofs64_ptr;
struct sha *sha_ptr;
- if (!PyArg_ParseTuple(args, "sw#OO",
- &filename, &fmap, &flen, &idx, &py_total))
+ Py_buffer fmap;
+ if (!PyArg_ParseTuple(args, cstr_argf wbuf_argf "OO",
+ &filename, &fmap, &idx, &py_total))
return NULL;
+ PyObject *result = NULL;
+
if (!bup_uint_from_py(&total, py_total, "total"))
- return NULL;
+ goto clean_and_return;
if (PyList_Size (idx) != FAN_ENTRIES) // Check for list of the right length.
- return PyErr_Format (PyExc_TypeError, "idx must contain %d entries",
- FAN_ENTRIES);
+ {
+ result = PyErr_Format (PyExc_TypeError, "idx must contain %d entries",
+ FAN_ENTRIES);
+ goto clean_and_return;
+ }
const char idx_header[] = "\377tOc\0\0\0\002";
- memcpy (fmap, idx_header, sizeof(idx_header) - 1);
+ memcpy (fmap.buf, idx_header, sizeof(idx_header) - 1);
- fan_ptr = (uint32_t *)&fmap[sizeof(idx_header) - 1];
+ fan_ptr = (uint32_t *)&((unsigned char *)fmap.buf)[sizeof(idx_header) - 1];
sha_ptr = (struct sha *)&fan_ptr[FAN_ENTRIES];
crc_ptr = (uint32_t *)&sha_ptr[total];
ofs_ptr = (uint32_t *)&crc_ptr[total];
unsigned int crc;
unsigned PY_LONG_LONG ofs_ull;
uint64_t ofs;
- if (!PyArg_ParseTuple(PyList_GET_ITEM(part, j), "t#OO",
+ if (!PyArg_ParseTuple(PyList_GET_ITEM(part, j), rbuf_argf "OO",
&sha, &sha_len, &crc_py, &ofs_py))
- return NULL;
+ goto clean_and_return;
if(!bup_uint_from_py(&crc, crc_py, "crc"))
- return NULL;
+ goto clean_and_return;
if(!bup_ullong_from_py(&ofs_ull, ofs_py, "ofs"))
- return NULL;
+ goto clean_and_return;
assert(crc <= UINT32_MAX);
assert(ofs_ull <= UINT64_MAX);
ofs = ofs_ull;
if (sha_len != sizeof(struct sha))
- return NULL;
+ goto clean_and_return;
memcpy(sha_ptr++, sha, sizeof(struct sha));
*crc_ptr++ = htonl(crc);
if (ofs > 0x7fffffff)
}
}
- int rc = msync(fmap, flen, MS_ASYNC);
+ int rc = msync(fmap.buf, fmap.len, MS_ASYNC);
if (rc != 0)
- return PyErr_SetFromErrnoWithFilename(PyExc_IOError, filename);
+ {
+ result = PyErr_SetFromErrnoWithFilename(PyExc_IOError, filename);
+ goto clean_and_return;
+ }
- return PyLong_FromUnsignedLong(count);
+ result = PyLong_FromUnsignedLong(count);
+
+ clean_and_return:
+ PyBuffer_Release(&fmap);
+ return result;
}
{
char *filename = NULL;
int fd;
- if (!PyArg_ParseTuple(args, "s", &filename))
+ if (!PyArg_ParseTuple(args, cstr_argf, &filename))
return NULL;
fd = _open_noatime(filename, 0);
if (fd < 0)
static PyObject *fadvise_done(PyObject *self, PyObject *args)
{
int fd = -1;
- long long ofs = 0;
- if (!PyArg_ParseTuple(args, "iL", &fd, &ofs))
+ long long llofs, lllen = 0;
+ if (!PyArg_ParseTuple(args, "iLL", &fd, &llofs, &lllen))
return NULL;
+ off_t ofs, len;
+ if (!INTEGRAL_ASSIGNMENT_FITS(&ofs, llofs))
+ return PyErr_Format(PyExc_OverflowError,
+ "fadvise offset overflows off_t");
+ if (!INTEGRAL_ASSIGNMENT_FITS(&len, lllen))
+ return PyErr_Format(PyExc_OverflowError,
+ "fadvise length overflows off_t");
#ifdef POSIX_FADV_DONTNEED
- posix_fadvise(fd, 0, ofs, POSIX_FADV_DONTNEED);
+ posix_fadvise(fd, ofs, len, POSIX_FADV_DONTNEED);
#endif
return Py_BuildValue("");
}
char *path;
int fd;
- if (!PyArg_ParseTuple(args, "s", &path))
+ if (!PyArg_ParseTuple(args, cstr_argf, &path))
return NULL;
fd = _open_noatime(path, O_NONBLOCK);
PyObject *py_attr;
int fd;
- if (!PyArg_ParseTuple(args, "sO", &path, &py_attr))
+ if (!PyArg_ParseTuple(args, cstr_argf "O", &path, &py_attr))
return NULL;
if (!bup_uint_from_py(&attr, py_attr, "attr"))
// The extents flag can't be removed, so don't (see chattr(1) and chattr.c).
orig_attr = 0; // Handle int/long mismatch (see above)
rc = ioctl(fd, FS_IOC_GETFLAGS, &orig_attr);
- assert(orig_attr <= UINT_MAX); // Kernel type is actually int
if (rc == -1)
{
close(fd);
return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
}
+ assert(orig_attr <= UINT_MAX); // Kernel type is actually int
attr |= ((unsigned int) orig_attr) & FS_EXTENT_FL;
rc = ioctl(fd, FS_IOC_SETFLAGS, &attr);
PyObject *access_py, *modification_py;
struct timespec ts[2];
- if (!PyArg_ParseTuple(args, "is((Ol)(Ol))i",
+ if (!PyArg_ParseTuple(args, "i" cstr_argf "((Ol)(Ol))i",
&fd,
&path,
&access_py, &(ts[0].tv_nsec),
PyObject *access_py, *modification_py;
long long access_us, modification_us; // POSIX guarantees tv_usec is signed.
- if (!PyArg_ParseTuple(args, "s((OL)(OL))",
+ if (!PyArg_ParseTuple(args, cstr_argf "((OL)(OL))",
path,
&access_py, &access_us,
&modification_py, &modification_us))
// compile time, but not (easily) the unspecified types, so handle
// those via INTEGER_TO_PY(). Assumes ns values will fit in a
// long.
- return Py_BuildValue("OKOOOOOL(Ol)(Ol)(Ol)",
+ return Py_BuildValue("NKNNNNNL(Nl)(Nl)(Nl)",
INTEGER_TO_PY(st->st_mode),
(unsigned PY_LONG_LONG) st->st_ino,
INTEGER_TO_PY(st->st_dev),
int rc;
char *filename;
- if (!PyArg_ParseTuple(args, "s", &filename))
+ if (!PyArg_ParseTuple(args, cstr_argf, &filename))
return NULL;
struct stat st;
int rc;
char *filename;
- if (!PyArg_ParseTuple(args, "s", &filename))
+ if (!PyArg_ParseTuple(args, cstr_argf, &filename))
return NULL;
struct stat st;
}
+#ifdef HAVE_TM_TM_GMTOFF
+static PyObject *bup_localtime(PyObject *self, PyObject *args)
+{
+ long long lltime;
+ time_t ttime;
+ if (!PyArg_ParseTuple(args, "L", &lltime))
+ return NULL;
+ if (!INTEGRAL_ASSIGNMENT_FITS(&ttime, lltime))
+ return PyErr_Format(PyExc_OverflowError, "time value too large");
+
+ struct tm tm;
+ tzset();
+ if(localtime_r(&ttime, &tm) == NULL)
+ return PyErr_SetFromErrno(PyExc_OSError);
+
+ // Match the Python struct_time values.
+ return Py_BuildValue("[i,i,i,i,i,i,i,i,i,i,s]",
+ 1900 + tm.tm_year, tm.tm_mon + 1, tm.tm_mday,
+ tm.tm_hour, tm.tm_min, tm.tm_sec,
+ tm.tm_wday, tm.tm_yday + 1,
+ tm.tm_isdst, tm.tm_gmtoff, tm.tm_zone);
+}
+#endif /* def HAVE_TM_TM_GMTOFF */
+
+
+#ifdef BUP_MINCORE_BUF_TYPE
+static PyObject *bup_mincore(PyObject *self, PyObject *args)
+{
+ Py_buffer src, dest;
+ PyObject *py_src_n, *py_src_off, *py_dest_off;
+
+ if (!PyArg_ParseTuple(args, cstr_argf "*OOw*O",
+ &src, &py_src_n, &py_src_off,
+ &dest, &py_dest_off))
+ return NULL;
+
+ PyObject *result = NULL;
+
+ unsigned long long src_n, src_off, dest_off;
+ if (!(bup_ullong_from_py(&src_n, py_src_n, "src_n")
+ && bup_ullong_from_py(&src_off, py_src_off, "src_off")
+ && bup_ullong_from_py(&dest_off, py_dest_off, "dest_off")))
+ goto clean_and_return;
+
+ unsigned long long src_region_end;
+ if (!uadd(&src_region_end, src_off, src_n)) {
+ result = PyErr_Format(PyExc_OverflowError, "(src_off + src_n) too large");
+ goto clean_and_return;
+ }
+ assert(src.len >= 0);
+ if (src_region_end > (unsigned long long) src.len) {
+ result = PyErr_Format(PyExc_OverflowError, "region runs off end of src");
+ goto clean_and_return;
+ }
+
+ unsigned long long dest_size;
+ if (!INTEGRAL_ASSIGNMENT_FITS(&dest_size, dest.len)) {
+ result = PyErr_Format(PyExc_OverflowError, "invalid dest size");
+ goto clean_and_return;
+ }
+ if (dest_off > dest_size) {
+ result = PyErr_Format(PyExc_OverflowError, "region runs off end of dest");
+ goto clean_and_return;
+ }
+
+ size_t length;
+ if (!INTEGRAL_ASSIGNMENT_FITS(&length, src_n)) {
+ result = PyErr_Format(PyExc_OverflowError, "src_n overflows size_t");
+ goto clean_and_return;
+ }
+ int rc = mincore((void *)(src.buf + src_off), src_n,
+ (BUP_MINCORE_BUF_TYPE *) (dest.buf + dest_off));
+ if (rc != 0) {
+ result = PyErr_SetFromErrno(PyExc_OSError);
+ goto clean_and_return;
+ }
+ result = Py_BuildValue("O", Py_None);
+
+ clean_and_return:
+ PyBuffer_Release(&src);
+ PyBuffer_Release(&dest);
+ return result;
+}
+#endif /* def BUP_MINCORE_BUF_TYPE */
+
+
static PyMethodDef helper_methods[] = {
{ "write_sparsely", bup_write_sparsely, METH_VARARGS,
"Write buf excepting zeros at the end. Return trailing zero count." },
"Extended version of lstat." },
{ "fstat", bup_fstat, METH_VARARGS,
"Extended version of fstat." },
+#ifdef HAVE_TM_TM_GMTOFF
+ { "localtime", bup_localtime, METH_VARARGS,
+ "Return struct_time elements plus the timezone offset and name." },
+#endif
+ { "bytescmp", bup_bytescmp, METH_VARARGS,
+ "Return a negative value if x < y, zero if equal, positive otherwise."},
+#ifdef BUP_MINCORE_BUF_TYPE
+ { "mincore", bup_mincore, METH_VARARGS,
+ "For mincore(src, src_n, src_off, dest, dest_off)"
+ " call the system mincore(src + src_off, src_n, &dest[dest_off])." },
+#endif
{ NULL, NULL, 0, NULL }, // sentinel
};
+static void test_integral_assignment_fits(void)
+{
+ assert(sizeof(signed short) == sizeof(unsigned short));
+ assert(sizeof(signed short) < sizeof(signed long long));
+ assert(sizeof(signed short) < sizeof(unsigned long long));
+ assert(sizeof(unsigned short) < sizeof(signed long long));
+ assert(sizeof(unsigned short) < sizeof(unsigned long long));
+ assert(sizeof(Py_ssize_t) <= sizeof(size_t));
+ {
+ signed short ss, ssmin = SHRT_MIN, ssmax = SHRT_MAX;
+ unsigned short us, usmax = USHRT_MAX;
+ signed long long sllmin = LLONG_MIN, sllmax = LLONG_MAX;
+ unsigned long long ullmax = ULLONG_MAX;
+
+ assert(INTEGRAL_ASSIGNMENT_FITS(&ss, ssmax));
+ assert(INTEGRAL_ASSIGNMENT_FITS(&ss, ssmin));
+ assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, usmax));
+ assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, sllmin));
+ assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, sllmax));
+ assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, ullmax));
+
+ assert(INTEGRAL_ASSIGNMENT_FITS(&us, usmax));
+ assert(!INTEGRAL_ASSIGNMENT_FITS(&us, ssmin));
+ assert(!INTEGRAL_ASSIGNMENT_FITS(&us, sllmin));
+ assert(!INTEGRAL_ASSIGNMENT_FITS(&us, sllmax));
+ assert(!INTEGRAL_ASSIGNMENT_FITS(&us, ullmax));
+ }
+}
-PyMODINIT_FUNC init_helpers(void)
+static int setup_module(PyObject *m)
{
- // FIXME: migrate these tests to configure. Check against the
- // type we're going to use when passing to python. Other stat
- // types are tested at runtime.
+ // FIXME: migrate these tests to configure, or at least don't
+ // possibly crash the whole application. Check against the type
+ // we're going to use when passing to python. Other stat types
+ // are tested at runtime.
assert(sizeof(ino_t) <= sizeof(unsigned PY_LONG_LONG));
assert(sizeof(off_t) <= sizeof(PY_LONG_LONG));
assert(sizeof(blksize_t) <= sizeof(PY_LONG_LONG));
assert(sizeof(PY_LONG_LONG) <= sizeof(long long));
assert(sizeof(unsigned PY_LONG_LONG) <= sizeof(unsigned long long));
- char *e;
- PyObject *m = Py_InitModule("_helpers", helper_methods);
- if (m == NULL)
- return;
+ test_integral_assignment_fits();
+ // Originally required by append_sparse_region()
+ {
+ off_t probe;
+ if (!INTEGRAL_ASSIGNMENT_FITS(&probe, INT_MAX))
+ {
+ fprintf(stderr, "off_t can't hold INT_MAX; please report.\n");
+ exit(1);
+ }
+ }
+
+ char *e;
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wtautological-compare" // For INTEGER_TO_PY().
+ {
+ PyObject *value;
+ value = INTEGER_TO_PY(INT_MAX);
+ PyObject_SetAttrString(m, "INT_MAX", value);
+ Py_DECREF(value);
+ value = INTEGER_TO_PY(UINT_MAX);
+ PyObject_SetAttrString(m, "UINT_MAX", value);
+ Py_DECREF(value);
+ }
#ifdef HAVE_UTIMENSAT
{
PyObject *value;
Py_DECREF(value);
}
#endif
+#ifdef BUP_HAVE_MINCORE_INCORE
{
PyObject *value;
- const long arg_max = sysconf(_SC_ARG_MAX);
- if (arg_max == -1)
- {
- fprintf(stderr, "Cannot find SC_ARG_MAX, please report a bug.\n");
- exit(1);
- }
- value = INTEGER_TO_PY(arg_max);
- PyObject_SetAttrString(m, "SC_ARG_MAX", value);
+ value = INTEGER_TO_PY(MINCORE_INCORE);
+ PyObject_SetAttrString(m, "MINCORE_INCORE", value);
Py_DECREF(value);
}
+#endif
#pragma clang diagnostic pop // ignored "-Wtautological-compare"
e = getenv("BUP_FORCE_TTY");
- istty2 = isatty(2) || (atoi(e ? e : "0") & 2);
+ get_state(m)->istty2 = isatty(2) || (atoi(e ? e : "0") & 2);
unpythonize_argv();
+ return 1;
}
+
+
+#if PY_MAJOR_VERSION < 3
+
+PyMODINIT_FUNC init_helpers(void)
+{
+ PyObject *m = Py_InitModule("_helpers", helper_methods);
+ if (m == NULL)
+ return;
+
+ if (!setup_module(m))
+ {
+ Py_DECREF(m);
+ return;
+ }
+}
+
+# else // PY_MAJOR_VERSION >= 3
+
+static struct PyModuleDef helpers_def = {
+ PyModuleDef_HEAD_INIT,
+ "_helpers",
+ NULL,
+ sizeof(state_t),
+ helper_methods,
+ NULL,
+ NULL, // helpers_traverse,
+ NULL, // helpers_clear,
+ NULL
+};
+
+PyMODINIT_FUNC PyInit__helpers(void)
+{
+ PyObject *module = PyModule_Create(&helpers_def);
+ if (module == NULL)
+ return NULL;
+ if (!setup_module(module))
+ {
+ Py_DECREF(module);
+ return NULL;
+ }
+ return module;
+}
+
+#endif // PY_MAJOR_VERSION >= 3
projects / bup.git / blobdiff