1 #define _LARGEFILE64_SOURCE 1
2 #define PY_SSIZE_T_CLEAN 1
4 #include "../../config/config.h"
6 // According to Python, its header has to go first:
7 // http://docs.python.org/2/c-api/intro.html#include-files
8 // http://docs.python.org/3/c-api/intro.html#include-files
11 #include <arpa/inet.h>
23 #ifdef HAVE_SYS_MMAN_H
26 #ifdef HAVE_SYS_TYPES_H
27 #include <sys/types.h>
29 #ifdef HAVE_SYS_STAT_H
35 #ifdef HAVE_SYS_TIME_H
39 #ifdef HAVE_LINUX_FS_H
42 #ifdef HAVE_SYS_IOCTL_H
43 #include <sys/ioctl.h>
46 #ifdef HAVE_TM_TM_GMTOFF
50 #if defined(BUP_RL_EXPECTED_XOPEN_SOURCE) \
51 && (!defined(_XOPEN_SOURCE) || _XOPEN_SOURCE < BUP_RL_EXPECTED_XOPEN_SOURCE)
52 # warning "_XOPEN_SOURCE version is incorrect for readline"
55 #ifdef BUP_HAVE_READLINE
56 # pragma GCC diagnostic push
57 # pragma GCC diagnostic ignored "-Wstrict-prototypes"
58 # ifdef BUP_READLINE_INCLUDES_IN_SUBDIR
59 # include <readline/readline.h>
60 # include <readline/history.h>
62 # include <readline.h>
65 # pragma GCC diagnostic pop
70 #if defined(FS_IOC_GETFLAGS) && defined(FS_IOC_SETFLAGS)
71 #define BUP_HAVE_FILE_ATTRS 1
74 #if PY_MAJOR_VERSION > 2
75 # define BUP_USE_PYTHON_UTIME 1
78 #ifndef BUP_USE_PYTHON_UTIME // just for Python 2 now
80 * Check for incomplete UTIMENSAT support (NetBSD 6), and if so,
81 * pretend we don't have it.
83 #if !defined(AT_FDCWD) || !defined(AT_SYMLINK_NOFOLLOW)
86 #endif // defined BUP_USE_PYTHON_UTIME
89 // Of course, this assumes it's a bitfield value.
94 typedef unsigned char byte;
101 // cstr_argf: for byte vectors without null characters (e.g. paths)
102 // rbuf_argf: for read-only byte vectors
103 // wbuf_argf: for mutable byte vectors
105 #if PY_MAJOR_VERSION < 3
106 static state_t state;
107 # define get_state(x) (&state)
108 # define cstr_argf "s"
109 # define rbuf_argf "s#"
110 # define wbuf_argf "s*"
112 # define get_state(x) ((state_t *) PyModule_GetState(x))
113 # define cstr_argf "y"
114 # define rbuf_argf "y#"
115 # define wbuf_argf "y*"
116 #endif // PY_MAJOR_VERSION >= 3
119 static void *checked_calloc(size_t n, size_t size)
121 void *result = calloc(n, size);
127 #ifndef BUP_HAVE_BUILTIN_MUL_OVERFLOW
129 #define checked_malloc checked_calloc
131 #else // defined BUP_HAVE_BUILTIN_MUL_OVERFLOW
133 static void *checked_malloc(size_t n, size_t size)
136 if (__builtin_mul_overflow(n, size, &total))
138 PyErr_Format(PyExc_OverflowError,
139 "request to allocate %zu items of size %zu is too large",
143 void *result = malloc(total);
145 return PyErr_NoMemory();
149 #endif // defined BUP_HAVE_BUILTIN_MUL_OVERFLOW
153 // This function should technically be macro'd out if it's going to be used
154 // more than ocasionally. As of this writing, it'll actually never be called
155 // in real world bup scenarios (because our packs are < MAX_INT bytes).
156 static uint64_t htonll(uint64_t value)
158 static const int endian_test = 42;
160 if (*(char *)&endian_test == endian_test) // LSB-MSB
161 return ((uint64_t)htonl(value & 0xFFFFFFFF) << 32) | htonl(value >> 32);
162 return value; // already in network byte order MSB-LSB
167 // Disabling sign-compare here should be fine since we're explicitly
168 // checking for a sign mismatch, i.e. if the signs don't match, then
169 // it doesn't matter what the value comparison says.
170 // FIXME: ... so should we reverse the order?
171 #define INTEGRAL_ASSIGNMENT_FITS(dest, src) \
173 _Pragma("GCC diagnostic push"); \
174 _Pragma("GCC diagnostic ignored \"-Wsign-compare\""); \
175 _Pragma("clang diagnostic push"); \
176 _Pragma("clang diagnostic ignored \"-Wshorten-64-to-32\""); \
178 int result = *(dest) == (src) && (*(dest) < 1) == ((src) < 1); \
179 _Pragma("clang diagnostic pop"); \
180 _Pragma("GCC diagnostic pop"); \
185 #define INTEGER_TO_PY(x) \
187 _Pragma("GCC diagnostic push"); \
188 _Pragma("GCC diagnostic ignored \"-Wtype-limits\""); \
189 _Pragma("clang diagnostic push"); \
190 _Pragma("clang diagnostic ignored \"-Wtautological-compare\""); \
191 PyObject *result = ((x) >= 0) ? PyLong_FromUnsignedLongLong(x) : PyLong_FromLongLong(x); \
192 _Pragma("clang diagnostic pop"); \
193 _Pragma("GCC diagnostic pop"); \
198 #if PY_MAJOR_VERSION < 3
199 static int bup_ulong_from_pyint(unsigned long *x, PyObject *py,
202 const long tmp = PyInt_AsLong(py);
203 if (tmp == -1 && PyErr_Occurred())
205 if (PyErr_ExceptionMatches(PyExc_OverflowError))
206 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned long",
212 PyErr_Format(PyExc_OverflowError,
213 "negative %s cannot be converted to unsigned long", name);
222 static int bup_ulong_from_py(unsigned long *x, PyObject *py, const char *name)
224 #if PY_MAJOR_VERSION < 3
226 return bup_ulong_from_pyint(x, py, name);
229 if (!PyLong_Check(py))
231 PyErr_Format(PyExc_TypeError, "expected integer %s", name);
235 const unsigned long tmp = PyLong_AsUnsignedLong(py);
236 if (PyErr_Occurred())
238 if (PyErr_ExceptionMatches(PyExc_OverflowError))
239 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned long",
248 static int bup_uint_from_py(unsigned int *x, PyObject *py, const char *name)
251 if (!bup_ulong_from_py(&tmp, py, name))
256 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned int", name);
259 *x = (unsigned int) tmp;
263 static int bup_ullong_from_py(unsigned PY_LONG_LONG *x, PyObject *py,
266 #if PY_MAJOR_VERSION < 3
270 if (bup_ulong_from_pyint(&tmp, py, name))
279 if (!PyLong_Check(py))
281 PyErr_Format(PyExc_TypeError, "integer argument expected for %s", name);
285 const unsigned PY_LONG_LONG tmp = PyLong_AsUnsignedLongLong(py);
286 if (tmp == (unsigned long long) -1 && PyErr_Occurred())
288 if (PyErr_ExceptionMatches(PyExc_OverflowError))
289 PyErr_Format(PyExc_OverflowError,
290 "%s too big for unsigned long long", name);
298 static PyObject *bup_bytescmp(PyObject *self, PyObject *args)
300 PyObject *py_s1, *py_s2; // This is really a PyBytes/PyString
301 if (!PyArg_ParseTuple(args, "SS", &py_s1, &py_s2))
304 Py_ssize_t s1_len, s2_len;
305 if (PyBytes_AsStringAndSize(py_s1, &s1, &s1_len) == -1)
307 if (PyBytes_AsStringAndSize(py_s2, &s2, &s2_len) == -1)
309 const Py_ssize_t n = (s1_len < s2_len) ? s1_len : s2_len;
310 const int cmp = memcmp(s1, s2, n);
312 return PyLong_FromLong(cmp);
313 if (s1_len == s2_len)
314 return PyLong_FromLong(0);;
315 return PyLong_FromLong((s1_len < s2_len) ? -1 : 1);
319 static PyObject *bup_cat_bytes(PyObject *self, PyObject *args)
321 unsigned char *bufx = NULL, *bufy = NULL;
322 Py_ssize_t bufx_len, bufx_ofs, bufx_n;
323 Py_ssize_t bufy_len, bufy_ofs, bufy_n;
324 if (!PyArg_ParseTuple(args,
327 &bufx, &bufx_len, &bufx_ofs, &bufx_n,
328 &bufy, &bufy_len, &bufy_ofs, &bufy_n))
331 return PyErr_Format(PyExc_ValueError, "negative x offset");
333 return PyErr_Format(PyExc_ValueError, "negative x extent");
334 if (bufx_ofs > bufx_len)
335 return PyErr_Format(PyExc_ValueError, "x offset greater than length");
336 if (bufx_n > bufx_len - bufx_ofs)
337 return PyErr_Format(PyExc_ValueError, "x extent past end of buffer");
340 return PyErr_Format(PyExc_ValueError, "negative y offset");
342 return PyErr_Format(PyExc_ValueError, "negative y extent");
343 if (bufy_ofs > bufy_len)
344 return PyErr_Format(PyExc_ValueError, "y offset greater than length");
345 if (bufy_n > bufy_len - bufy_ofs)
346 return PyErr_Format(PyExc_ValueError, "y extent past end of buffer");
348 if (bufy_n > PY_SSIZE_T_MAX - bufx_n)
349 return PyErr_Format(PyExc_OverflowError, "result length too long");
351 PyObject *result = PyBytes_FromStringAndSize(NULL, bufx_n + bufy_n);
353 return PyErr_NoMemory();
354 char *buf = PyBytes_AS_STRING(result);
355 memcpy(buf, bufx + bufx_ofs, bufx_n);
356 memcpy(buf + bufx_n, bufy + bufy_ofs, bufy_n);
361 static int write_all(int fd, const void *buf, const size_t count)
364 while (written < count)
366 const ssize_t rc = write(fd, buf + written, count - written);
375 static int uadd(unsigned long long *dest,
376 const unsigned long long x,
377 const unsigned long long y)
379 const unsigned long long result = x + y;
380 if (result < x || result < y)
387 static PyObject *append_sparse_region(const int fd, unsigned long long n)
392 if (!INTEGRAL_ASSIGNMENT_FITS(&new_off, n))
394 const off_t off = lseek(fd, new_off, SEEK_CUR);
395 if (off == (off_t) -1)
396 return PyErr_SetFromErrno(PyExc_IOError);
403 static PyObject *record_sparse_zeros(unsigned long long *new_pending,
405 unsigned long long prev_pending,
406 const unsigned long long count)
408 // Add count additional sparse zeros to prev_pending and store the
409 // result in new_pending, or if the total won't fit in
410 // new_pending, write some of the zeros to fd sparsely, and store
411 // the remaining sum in new_pending.
412 if (!uadd(new_pending, prev_pending, count))
414 PyObject *err = append_sparse_region(fd, prev_pending);
417 *new_pending = count;
423 static byte* find_not_zero(const byte * const start, const byte * const end)
425 // Return a pointer to first non-zero byte between start and end,
426 // or end if there isn't one.
427 assert(start <= end);
428 const unsigned char *cur = start;
429 while (cur < end && *cur == 0)
435 static byte* find_trailing_zeros(const byte * const start,
436 const byte * const end)
438 // Return a pointer to the start of any trailing run of zeros, or
439 // end if there isn't one.
440 assert(start <= end);
443 const byte * cur = end;
444 while (cur > start && *--cur == 0) {}
448 return (byte *) (cur + 1);
452 static byte *find_non_sparse_end(const byte * const start,
453 const byte * const end,
454 const ptrdiff_t min_len)
456 // Return the first pointer to a min_len sparse block in [start,
457 // end) if there is one, otherwise a pointer to the start of any
458 // trailing run of zeros. If there are no trailing zeros, return
464 // Probe in min_len jumps, searching backward from the jump
465 // destination for a non-zero byte. If such a byte is found, move
466 // just past it and try again.
467 const byte *candidate = start;
468 // End of any run of zeros, starting at candidate, that we've already seen
469 const byte *end_of_known_zeros = candidate;
470 while (end - candidate >= min_len) // Handle all min_len candidate blocks
472 const byte * const probe_end = candidate + min_len;
473 const byte * const trailing_zeros =
474 find_trailing_zeros(end_of_known_zeros, probe_end);
475 if (trailing_zeros == probe_end)
476 end_of_known_zeros = candidate = probe_end;
477 else if (trailing_zeros == end_of_known_zeros)
479 assert(candidate >= start);
480 assert(candidate <= end);
481 assert(*candidate == 0);
482 return (byte *) candidate;
486 candidate = trailing_zeros;
487 end_of_known_zeros = probe_end;
491 if (candidate == end)
494 // No min_len sparse run found, search backward from end
495 const byte * const trailing_zeros = find_trailing_zeros(end_of_known_zeros,
498 if (trailing_zeros == end_of_known_zeros)
500 assert(candidate >= start);
501 assert(candidate < end);
502 assert(*candidate == 0);
503 assert(end - candidate < min_len);
504 return (byte *) candidate;
507 if (trailing_zeros == end)
509 assert(*(end - 1) != 0);
513 assert(end - trailing_zeros < min_len);
514 assert(trailing_zeros >= start);
515 assert(trailing_zeros < end);
516 assert(*trailing_zeros == 0);
517 return (byte *) trailing_zeros;
521 static PyObject *bup_write_sparsely(PyObject *self, PyObject *args)
524 unsigned char *buf = NULL;
526 PyObject *py_min_sparse_len, *py_prev_sparse_len;
527 if (!PyArg_ParseTuple(args, "i" rbuf_argf "OO",
528 &fd, &buf, &sbuf_len,
529 &py_min_sparse_len, &py_prev_sparse_len))
531 ptrdiff_t min_sparse_len;
532 unsigned long long prev_sparse_len, buf_len, ul_min_sparse_len;
533 if (!bup_ullong_from_py(&ul_min_sparse_len, py_min_sparse_len, "min_sparse_len"))
535 if (!INTEGRAL_ASSIGNMENT_FITS(&min_sparse_len, ul_min_sparse_len))
536 return PyErr_Format(PyExc_OverflowError, "min_sparse_len too large");
537 if (!bup_ullong_from_py(&prev_sparse_len, py_prev_sparse_len, "prev_sparse_len"))
540 return PyErr_Format(PyExc_ValueError, "negative bufer length");
541 if (!INTEGRAL_ASSIGNMENT_FITS(&buf_len, sbuf_len))
542 return PyErr_Format(PyExc_OverflowError, "buffer length too large");
544 const byte * block = buf; // Start of pending block
545 const byte * const end = buf + buf_len;
546 unsigned long long zeros = prev_sparse_len;
549 assert(block <= end);
551 return PyLong_FromUnsignedLongLong(zeros);
555 // Look for the end of block, i.e. the next sparse run of
556 // at least min_sparse_len zeros, or the end of the
558 const byte * const probe = find_non_sparse_end(block + 1, end,
560 // Either at end of block, or end of non-sparse; write pending data
561 PyObject *err = append_sparse_region(fd, zeros);
564 int rc = write_all(fd, block, probe - block);
566 return PyErr_SetFromErrno(PyExc_IOError);
568 if (end - probe < min_sparse_len)
571 zeros = min_sparse_len;
572 block = probe + zeros;
576 // Should be in the first loop iteration, a sparse run of
577 // zeros, or nearly at the end of the block (within
579 const byte * const zeros_end = find_not_zero(block, end);
580 PyObject *err = record_sparse_zeros(&zeros, fd,
581 zeros, zeros_end - block);
584 assert(block <= zeros_end);
591 static PyObject *selftest(PyObject *self, PyObject *args)
593 if (!PyArg_ParseTuple(args, ""))
596 return Py_BuildValue("i", !bupsplit_selftest());
600 static PyObject *blobbits(PyObject *self, PyObject *args)
602 if (!PyArg_ParseTuple(args, ""))
604 return Py_BuildValue("i", BUP_BLOBBITS);
608 static PyObject *splitbuf(PyObject *self, PyObject *args)
610 // We stick to buffers in python 2 because they appear to be
611 // substantially smaller than memoryviews, and because
612 // zlib.compress() in python 2 can't accept a memoryview
613 // (cf. hashsplit.py).
614 int out = 0, bits = -1;
615 if (PY_MAJOR_VERSION > 2)
618 if (!PyArg_ParseTuple(args, "y*", &buf))
620 assert(buf.len <= INT_MAX);
621 out = bupsplit_find_ofs(buf.buf, buf.len, &bits);
622 PyBuffer_Release(&buf);
626 unsigned char *buf = NULL;
628 if (!PyArg_ParseTuple(args, "t#", &buf, &len))
630 assert(len <= INT_MAX);
631 out = bupsplit_find_ofs(buf, (int) len, &bits);
633 if (out) assert(bits >= BUP_BLOBBITS);
634 return Py_BuildValue("ii", out, bits);
638 static PyObject *bitmatch(PyObject *self, PyObject *args)
640 unsigned char *buf1 = NULL, *buf2 = NULL;
641 Py_ssize_t len1 = 0, len2 = 0;
645 if (!PyArg_ParseTuple(args, rbuf_argf rbuf_argf, &buf1, &len1, &buf2, &len2))
649 for (byte = 0; byte < len1 && byte < len2; byte++)
651 int b1 = buf1[byte], b2 = buf2[byte];
654 for (bit = 0; bit < 8; bit++)
655 if ( (b1 & (0x80 >> bit)) != (b2 & (0x80 >> bit)) )
661 assert(byte <= (INT_MAX >> 3));
662 return Py_BuildValue("i", byte*8 + bit);
666 static PyObject *firstword(PyObject *self, PyObject *args)
668 unsigned char *buf = NULL;
672 if (!PyArg_ParseTuple(args, rbuf_argf, &buf, &len))
678 v = ntohl(*(uint32_t *)buf);
679 return PyLong_FromUnsignedLong(v);
683 #define BLOOM2_HEADERLEN 16
685 static void to_bloom_address_bitmask4(const unsigned char *buf,
686 const int nbits, uint64_t *v, unsigned char *bitmask)
692 memcpy(&high, buf, 4);
693 mask = (1<<nbits) - 1;
694 raw = (((uint64_t)ntohl(high) << 8) | buf[4]);
695 bit = (raw >> (37-nbits)) & 0x7;
696 *v = (raw >> (40-nbits)) & mask;
700 static void to_bloom_address_bitmask5(const unsigned char *buf,
701 const int nbits, uint32_t *v, unsigned char *bitmask)
707 memcpy(&high, buf, 4);
708 mask = (1<<nbits) - 1;
710 bit = (raw >> (29-nbits)) & 0x7;
711 *v = (raw >> (32-nbits)) & mask;
715 #define BLOOM_SET_BIT(name, address, otype) \
716 static void name(unsigned char *bloom, const unsigned char *buf, const int nbits)\
718 unsigned char bitmask;\
720 address(buf, nbits, &v, &bitmask);\
721 bloom[BLOOM2_HEADERLEN+v] |= bitmask;\
723 BLOOM_SET_BIT(bloom_set_bit4, to_bloom_address_bitmask4, uint64_t)
724 BLOOM_SET_BIT(bloom_set_bit5, to_bloom_address_bitmask5, uint32_t)
727 #define BLOOM_GET_BIT(name, address, otype) \
728 static int name(const unsigned char *bloom, const unsigned char *buf, const int nbits)\
730 unsigned char bitmask;\
732 address(buf, nbits, &v, &bitmask);\
733 return bloom[BLOOM2_HEADERLEN+v] & bitmask;\
735 BLOOM_GET_BIT(bloom_get_bit4, to_bloom_address_bitmask4, uint64_t)
736 BLOOM_GET_BIT(bloom_get_bit5, to_bloom_address_bitmask5, uint32_t)
739 static PyObject *bloom_add(PyObject *self, PyObject *args)
741 Py_buffer bloom, sha;
742 int nbits = 0, k = 0;
743 if (!PyArg_ParseTuple(args, wbuf_argf wbuf_argf "ii",
744 &bloom, &sha, &nbits, &k))
747 PyObject *result = NULL;
749 if (bloom.len < 16+(1<<nbits) || sha.len % 20 != 0)
750 goto clean_and_return;
755 goto clean_and_return;
756 unsigned char *cur = sha.buf;
758 for (end = cur + sha.len; cur < end; cur += 20/k)
759 bloom_set_bit5(bloom.buf, cur, nbits);
764 goto clean_and_return;
765 unsigned char *cur = sha.buf;
766 unsigned char *end = cur + sha.len;
767 for (; cur < end; cur += 20/k)
768 bloom_set_bit4(bloom.buf, cur, nbits);
771 goto clean_and_return;
773 result = Py_BuildValue("n", sha.len / 20);
776 PyBuffer_Release(&bloom);
777 PyBuffer_Release(&sha);
781 static PyObject *bloom_contains(PyObject *self, PyObject *args)
784 unsigned char *sha = NULL;
786 int nbits = 0, k = 0;
787 if (!PyArg_ParseTuple(args, wbuf_argf rbuf_argf "ii",
788 &bloom, &sha, &len, &nbits, &k))
791 PyObject *result = NULL;
794 goto clean_and_return;
799 goto clean_and_return;
802 for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
803 if (!bloom_get_bit5(bloom.buf, sha, nbits))
805 result = Py_BuildValue("Oi", Py_None, steps);
806 goto clean_and_return;
812 goto clean_and_return;
815 for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
816 if (!bloom_get_bit4(bloom.buf, sha, nbits))
818 result = Py_BuildValue("Oi", Py_None, steps);
819 goto clean_and_return;
823 goto clean_and_return;
825 result = Py_BuildValue("ii", 1, k);
828 PyBuffer_Release(&bloom);
833 static uint32_t _extract_bits(unsigned char *buf, int nbits)
837 mask = (1<<nbits) - 1;
838 v = ntohl(*(uint32_t *)buf);
839 v = (v >> (32-nbits)) & mask;
844 static PyObject *extract_bits(PyObject *self, PyObject *args)
846 unsigned char *buf = NULL;
850 if (!PyArg_ParseTuple(args, rbuf_argf "i", &buf, &len, &nbits))
856 return PyLong_FromUnsignedLong(_extract_bits(buf, nbits));
861 unsigned char bytes[20];
864 static inline int _cmp_sha(const struct sha *sha1, const struct sha *sha2)
866 return memcmp(sha1->bytes, sha2->bytes, sizeof(sha1->bytes));
879 static void _fix_idx_order(struct idx **idxs, Py_ssize_t *last_i)
882 Py_ssize_t low, mid, high;
886 if (idxs[*last_i]->cur >= idxs[*last_i]->end)
888 idxs[*last_i] = NULL;
901 mid = (low + high) / 2;
902 c = _cmp_sha(idx->cur, idxs[mid]->cur);
914 memmove(&idxs[mid+1], &idxs[mid], (*last_i-mid)*sizeof(struct idx *));
919 static uint32_t _get_idx_i(struct idx *idx)
921 if (idx->cur_name == NULL)
922 return idx->name_base;
923 return ntohl(*idx->cur_name) + idx->name_base;
926 #define MIDX4_HEADERLEN 12
928 static PyObject *merge_into(PyObject *self, PyObject *args)
930 struct sha *sha_ptr, *sha_start = NULL;
931 uint32_t *table_ptr, *name_ptr, *name_start;
934 uint32_t count, prefix;
939 PyObject *py_total, *ilist = NULL;
940 if (!PyArg_ParseTuple(args, wbuf_argf "iOO",
941 &fmap, &bits, &py_total, &ilist))
944 PyObject *result = NULL;
945 struct idx **idxs = NULL;
946 Py_ssize_t num_i = 0;
947 int *idx_buf_init = NULL;
948 Py_buffer *idx_buf = NULL;
950 if (!bup_uint_from_py(&total, py_total, "total"))
951 goto clean_and_return;
953 num_i = PyList_Size(ilist);
955 if (!(idxs = checked_malloc(num_i, sizeof(struct idx *))))
956 goto clean_and_return;
957 if (!(idx_buf_init = checked_calloc(num_i, sizeof(int))))
958 goto clean_and_return;
959 if (!(idx_buf = checked_malloc(num_i, sizeof(Py_buffer))))
960 goto clean_and_return;
962 for (i = 0; i < num_i; i++)
964 long len, sha_ofs, name_map_ofs;
965 if (!(idxs[i] = checked_malloc(1, sizeof(struct idx))))
966 goto clean_and_return;
967 PyObject *itup = PyList_GetItem(ilist, i);
968 if (!PyArg_ParseTuple(itup, wbuf_argf "llli",
969 &(idx_buf[i]), &len, &sha_ofs, &name_map_ofs,
970 &idxs[i]->name_base))
973 idxs[i]->map = idx_buf[i].buf;
974 idxs[i]->bytes = idx_buf[i].len;
975 idxs[i]->cur = (struct sha *)&idxs[i]->map[sha_ofs];
976 idxs[i]->end = &idxs[i]->cur[len];
978 idxs[i]->cur_name = (uint32_t *)&idxs[i]->map[name_map_ofs];
980 idxs[i]->cur_name = NULL;
982 table_ptr = (uint32_t *) &((unsigned char *) fmap.buf)[MIDX4_HEADERLEN];
983 sha_start = sha_ptr = (struct sha *)&table_ptr[1<<bits];
984 name_start = name_ptr = (uint32_t *)&sha_ptr[total];
986 Py_ssize_t last_i = num_i - 1;
993 if (count % 102424 == 0 && get_state(self)->istty2)
994 fprintf(stderr, "midx: writing %.2f%% (%d/%d)\r",
995 count*100.0/total, count, total);
997 new_prefix = _extract_bits((unsigned char *)idx->cur, bits);
998 while (prefix < new_prefix)
999 table_ptr[prefix++] = htonl(count);
1000 memcpy(sha_ptr++, idx->cur, sizeof(struct sha));
1001 *name_ptr++ = htonl(_get_idx_i(idx));
1003 if (idx->cur_name != NULL)
1005 _fix_idx_order(idxs, &last_i);
1008 while (prefix < ((uint32_t) 1 << bits))
1009 table_ptr[prefix++] = htonl(count);
1010 assert(count == total);
1011 assert(prefix == ((uint32_t) 1 << bits));
1012 assert(sha_ptr == sha_start+count);
1013 assert(name_ptr == name_start+count);
1015 result = PyLong_FromUnsignedLong(count);
1020 for (i = 0; i < num_i; i++)
1021 if (idx_buf_init[i])
1022 PyBuffer_Release(&(idx_buf[i]));
1028 for (i = 0; i < num_i; i++)
1032 PyBuffer_Release(&fmap);
1036 #define FAN_ENTRIES 256
1038 static PyObject *write_idx(PyObject *self, PyObject *args)
1040 char *filename = NULL;
1041 PyObject *py_total, *idx = NULL;
1043 unsigned int total = 0;
1046 uint32_t *fan_ptr, *crc_ptr, *ofs_ptr;
1047 uint64_t *ofs64_ptr;
1048 struct sha *sha_ptr;
1051 if (!PyArg_ParseTuple(args, cstr_argf wbuf_argf "OO",
1052 &filename, &fmap, &idx, &py_total))
1055 PyObject *result = NULL;
1057 if (!bup_uint_from_py(&total, py_total, "total"))
1058 goto clean_and_return;
1060 if (PyList_Size (idx) != FAN_ENTRIES) // Check for list of the right length.
1062 result = PyErr_Format (PyExc_TypeError, "idx must contain %d entries",
1064 goto clean_and_return;
1067 const char idx_header[] = "\377tOc\0\0\0\002";
1068 memcpy (fmap.buf, idx_header, sizeof(idx_header) - 1);
1070 fan_ptr = (uint32_t *)&((unsigned char *)fmap.buf)[sizeof(idx_header) - 1];
1071 sha_ptr = (struct sha *)&fan_ptr[FAN_ENTRIES];
1072 crc_ptr = (uint32_t *)&sha_ptr[total];
1073 ofs_ptr = (uint32_t *)&crc_ptr[total];
1074 ofs64_ptr = (uint64_t *)&ofs_ptr[total];
1077 uint32_t ofs64_count = 0;
1078 for (i = 0; i < FAN_ENTRIES; ++i)
1080 part = PyList_GET_ITEM(idx, i);
1083 if (!INTEGRAL_ASSIGNMENT_FITS(&plen, PyList_GET_SIZE(part))
1084 || UINT32_MAX - count < plen) {
1085 PyErr_Format(PyExc_OverflowError, "too many objects in index part");
1086 goto clean_and_return;
1089 *fan_ptr++ = htonl(count);
1091 for (j = 0; j < plen; ++j)
1093 unsigned char *sha = NULL;
1094 Py_ssize_t sha_len = 0;
1095 PyObject *crc_py, *ofs_py;
1097 unsigned PY_LONG_LONG ofs_ull;
1099 if (!PyArg_ParseTuple(PyList_GET_ITEM(part, j), rbuf_argf "OO",
1100 &sha, &sha_len, &crc_py, &ofs_py))
1101 goto clean_and_return;
1102 if(!bup_uint_from_py(&crc, crc_py, "crc"))
1103 goto clean_and_return;
1104 if(!bup_ullong_from_py(&ofs_ull, ofs_py, "ofs"))
1105 goto clean_and_return;
1106 assert(crc <= UINT32_MAX);
1107 assert(ofs_ull <= UINT64_MAX);
1109 if (sha_len != sizeof(struct sha))
1110 goto clean_and_return;
1111 memcpy(sha_ptr++, sha, sizeof(struct sha));
1112 *crc_ptr++ = htonl(crc);
1113 if (ofs > 0x7fffffff)
1115 *ofs64_ptr++ = htonll(ofs);
1116 ofs = 0x80000000 | ofs64_count++;
1118 *ofs_ptr++ = htonl((uint32_t)ofs);
1122 int rc = msync(fmap.buf, fmap.len, MS_ASYNC);
1125 result = PyErr_SetFromErrnoWithFilename(PyExc_IOError, filename);
1126 goto clean_and_return;
1129 result = PyLong_FromUnsignedLong(count);
1132 PyBuffer_Release(&fmap);
1137 // I would have made this a lower-level function that just fills in a buffer
1138 // with random values, and then written those values from python. But that's
1139 // about 20% slower in my tests, and since we typically generate random
1140 // numbers for benchmarking other parts of bup, any slowness in generating
1141 // random bytes will make our benchmarks inaccurate. Plus nobody wants
1142 // pseudorandom bytes much except for this anyway.
1143 static PyObject *write_random(PyObject *self, PyObject *args)
1145 uint32_t buf[1024/4];
1146 int fd = -1, seed = 0, verbose = 0;
1148 long long len = 0, kbytes = 0, written = 0;
1150 if (!PyArg_ParseTuple(args, "iLii", &fd, &len, &seed, &verbose))
1155 for (kbytes = 0; kbytes < len/1024; kbytes++)
1158 for (i = 0; i < sizeof(buf)/sizeof(buf[0]); i++)
1159 buf[i] = (uint32_t) random();
1160 ret = write(fd, buf, sizeof(buf));
1164 if (ret < (int)sizeof(buf))
1166 if (verbose && kbytes/1024 > 0 && !(kbytes%1024))
1167 fprintf(stderr, "Random: %lld Mbytes\r", kbytes/1024);
1170 // handle non-multiples of 1024
1174 for (i = 0; i < sizeof(buf)/sizeof(buf[0]); i++)
1175 buf[i] = (uint32_t) random();
1176 ret = write(fd, buf, len % 1024);
1182 if (kbytes/1024 > 0)
1183 fprintf(stderr, "Random: %lld Mbytes, done.\n", kbytes/1024);
1184 return Py_BuildValue("L", written);
1188 static PyObject *random_sha(PyObject *self, PyObject *args)
1190 static int seeded = 0;
1191 uint32_t shabuf[20/4];
1196 assert(sizeof(shabuf) == 20);
1197 srandom((unsigned int) time(NULL));
1201 if (!PyArg_ParseTuple(args, ""))
1204 memset(shabuf, 0, sizeof(shabuf));
1205 for (i=0; i < 20/4; i++)
1206 shabuf[i] = (uint32_t) random();
1207 return Py_BuildValue(rbuf_argf, shabuf, 20);
1211 static int _open_noatime(const char *filename, int attrs)
1213 int attrs_noatime, fd;
1216 attrs |= O_NOFOLLOW;
1219 attrs |= O_LARGEFILE;
1221 attrs_noatime = attrs;
1223 attrs_noatime |= O_NOATIME;
1225 fd = open(filename, attrs_noatime);
1226 if (fd < 0 && errno == EPERM)
1228 // older Linux kernels would return EPERM if you used O_NOATIME
1229 // and weren't the file's owner. This pointless restriction was
1230 // relaxed eventually, but we have to handle it anyway.
1231 // (VERY old kernels didn't recognized O_NOATIME, but they would
1232 // just harmlessly ignore it, so this branch won't trigger)
1233 fd = open(filename, attrs);
1239 static PyObject *open_noatime(PyObject *self, PyObject *args)
1241 char *filename = NULL;
1243 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1245 fd = _open_noatime(filename, 0);
1247 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1248 return Py_BuildValue("i", fd);
1252 static PyObject *fadvise_done(PyObject *self, PyObject *args)
1255 long long llofs, lllen = 0;
1256 if (!PyArg_ParseTuple(args, "iLL", &fd, &llofs, &lllen))
1259 if (!INTEGRAL_ASSIGNMENT_FITS(&ofs, llofs))
1260 return PyErr_Format(PyExc_OverflowError,
1261 "fadvise offset overflows off_t");
1262 if (!INTEGRAL_ASSIGNMENT_FITS(&len, lllen))
1263 return PyErr_Format(PyExc_OverflowError,
1264 "fadvise length overflows off_t");
1265 #ifdef POSIX_FADV_DONTNEED
1266 posix_fadvise(fd, ofs, len, POSIX_FADV_DONTNEED);
1268 return Py_BuildValue("");
1272 // Currently the Linux kernel and FUSE disagree over the type for
1273 // FS_IOC_GETFLAGS and FS_IOC_SETFLAGS. The kernel actually uses int,
1274 // but FUSE chose long (matching the declaration in linux/fs.h). So
1275 // if you use int, and then traverse a FUSE filesystem, you may
1276 // corrupt the stack. But if you use long, then you may get invalid
1277 // results on big-endian systems.
1279 // For now, we just use long, and then disable Linux attrs entirely
1280 // (with a warning) in helpers.py on systems that are affected.
1282 #ifdef BUP_HAVE_FILE_ATTRS
1283 static PyObject *bup_get_linux_file_attr(PyObject *self, PyObject *args)
1290 if (!PyArg_ParseTuple(args, cstr_argf, &path))
1293 fd = _open_noatime(path, O_NONBLOCK);
1295 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1297 attr = 0; // Handle int/long mismatch (see above)
1298 rc = ioctl(fd, FS_IOC_GETFLAGS, &attr);
1302 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1305 assert(attr <= UINT_MAX); // Kernel type is actually int
1306 return PyLong_FromUnsignedLong(attr);
1308 #endif /* def BUP_HAVE_FILE_ATTRS */
1312 #ifdef BUP_HAVE_FILE_ATTRS
1313 static PyObject *bup_set_linux_file_attr(PyObject *self, PyObject *args)
1316 unsigned long orig_attr;
1322 if (!PyArg_ParseTuple(args, cstr_argf "O", &path, &py_attr))
1325 if (!bup_uint_from_py(&attr, py_attr, "attr"))
1328 fd = open(path, O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_NOFOLLOW);
1330 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1332 // Restrict attr to modifiable flags acdeijstuADST -- see
1333 // chattr(1) and the e2fsprogs source. Letter to flag mapping is
1334 // in pf.c flags_array[].
1335 attr &= FS_APPEND_FL | FS_COMPR_FL | FS_NODUMP_FL | FS_EXTENT_FL
1336 | FS_IMMUTABLE_FL | FS_JOURNAL_DATA_FL | FS_SECRM_FL | FS_NOTAIL_FL
1337 | FS_UNRM_FL | FS_NOATIME_FL | FS_DIRSYNC_FL | FS_SYNC_FL
1338 | FS_TOPDIR_FL | FS_NOCOW_FL;
1340 // The extents flag can't be removed, so don't (see chattr(1) and chattr.c).
1341 orig_attr = 0; // Handle int/long mismatch (see above)
1342 rc = ioctl(fd, FS_IOC_GETFLAGS, &orig_attr);
1346 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1348 assert(orig_attr <= UINT_MAX); // Kernel type is actually int
1349 attr |= ((unsigned int) orig_attr) & FS_EXTENT_FL;
1351 rc = ioctl(fd, FS_IOC_SETFLAGS, &attr);
1355 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1359 return Py_BuildValue("O", Py_None);
1361 #endif /* def BUP_HAVE_FILE_ATTRS */
1364 #ifndef BUP_USE_PYTHON_UTIME // just for Python 2 now
1365 #ifndef HAVE_UTIMENSAT
1367 #error "cannot find utimensat or utimes()"
1369 #ifndef HAVE_LUTIMES
1370 #error "cannot find utimensat or lutimes()"
1373 #endif // defined BUP_USE_PYTHON_UTIME
1375 #define ASSIGN_PYLONG_TO_INTEGRAL(dest, pylong, overflow) \
1379 const long long lltmp = PyLong_AsLongLong(pylong); \
1380 if (lltmp == -1 && PyErr_Occurred()) \
1382 if (PyErr_ExceptionMatches(PyExc_OverflowError)) \
1384 const unsigned long long ulltmp = PyLong_AsUnsignedLongLong(pylong); \
1385 if (ulltmp == (unsigned long long) -1 && PyErr_Occurred()) \
1387 if (PyErr_ExceptionMatches(PyExc_OverflowError)) \
1393 if (INTEGRAL_ASSIGNMENT_FITS((dest), ulltmp)) \
1401 if (INTEGRAL_ASSIGNMENT_FITS((dest), lltmp)) \
1410 #ifndef BUP_USE_PYTHON_UTIME // just for Python 2 now
1411 #ifdef HAVE_UTIMENSAT
1413 static PyObject *bup_utimensat(PyObject *self, PyObject *args)
1418 PyObject *access_py, *modification_py;
1419 struct timespec ts[2];
1421 if (!PyArg_ParseTuple(args, "i" cstr_argf "((Ol)(Ol))i",
1424 &access_py, &(ts[0].tv_nsec),
1425 &modification_py, &(ts[1].tv_nsec),
1430 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(ts[0].tv_sec), access_py, &overflow))
1433 PyErr_SetString(PyExc_ValueError,
1434 "unable to convert access time seconds for utimensat");
1437 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(ts[1].tv_sec), modification_py, &overflow))
1440 PyErr_SetString(PyExc_ValueError,
1441 "unable to convert modification time seconds for utimensat");
1444 rc = utimensat(fd, path, ts, flag);
1446 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1448 return Py_BuildValue("O", Py_None);
1451 #endif /* def HAVE_UTIMENSAT */
1454 #if defined(HAVE_UTIMES) || defined(HAVE_LUTIMES)
1456 static int bup_parse_xutimes_args(char **path,
1457 struct timeval tv[2],
1460 PyObject *access_py, *modification_py;
1461 long long access_us, modification_us; // POSIX guarantees tv_usec is signed.
1463 if (!PyArg_ParseTuple(args, cstr_argf "((OL)(OL))",
1465 &access_py, &access_us,
1466 &modification_py, &modification_us))
1470 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(tv[0].tv_sec), access_py, &overflow))
1473 PyErr_SetString(PyExc_ValueError, "unable to convert access time seconds to timeval");
1476 if (!INTEGRAL_ASSIGNMENT_FITS(&(tv[0].tv_usec), access_us))
1478 PyErr_SetString(PyExc_ValueError, "unable to convert access time nanoseconds to timeval");
1481 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(tv[1].tv_sec), modification_py, &overflow))
1484 PyErr_SetString(PyExc_ValueError, "unable to convert modification time seconds to timeval");
1487 if (!INTEGRAL_ASSIGNMENT_FITS(&(tv[1].tv_usec), modification_us))
1489 PyErr_SetString(PyExc_ValueError, "unable to convert modification time nanoseconds to timeval");
1495 #endif /* defined(HAVE_UTIMES) || defined(HAVE_LUTIMES) */
1499 static PyObject *bup_utimes(PyObject *self, PyObject *args)
1502 struct timeval tv[2];
1503 if (!bup_parse_xutimes_args(&path, tv, args))
1505 int rc = utimes(path, tv);
1507 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1508 return Py_BuildValue("O", Py_None);
1510 #endif /* def HAVE_UTIMES */
1514 static PyObject *bup_lutimes(PyObject *self, PyObject *args)
1517 struct timeval tv[2];
1518 if (!bup_parse_xutimes_args(&path, tv, args))
1520 int rc = lutimes(path, tv);
1522 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1524 return Py_BuildValue("O", Py_None);
1526 #endif /* def HAVE_LUTIMES */
1528 #endif // defined BUP_USE_PYTHON_UTIME
1531 #ifdef HAVE_STAT_ST_ATIM
1532 # define BUP_STAT_ATIME_NS(st) (st)->st_atim.tv_nsec
1533 # define BUP_STAT_MTIME_NS(st) (st)->st_mtim.tv_nsec
1534 # define BUP_STAT_CTIME_NS(st) (st)->st_ctim.tv_nsec
1535 #elif defined HAVE_STAT_ST_ATIMENSEC
1536 # define BUP_STAT_ATIME_NS(st) (st)->st_atimespec.tv_nsec
1537 # define BUP_STAT_MTIME_NS(st) (st)->st_mtimespec.tv_nsec
1538 # define BUP_STAT_CTIME_NS(st) (st)->st_ctimespec.tv_nsec
1540 # define BUP_STAT_ATIME_NS(st) 0
1541 # define BUP_STAT_MTIME_NS(st) 0
1542 # define BUP_STAT_CTIME_NS(st) 0
1546 static PyObject *stat_struct_to_py(const struct stat *st,
1547 const char *filename,
1550 // We can check the known (via POSIX) signed and unsigned types at
1551 // compile time, but not (easily) the unspecified types, so handle
1552 // those via INTEGER_TO_PY(). Assumes ns values will fit in a
1554 return Py_BuildValue("NKNNNNNL(Nl)(Nl)(Nl)",
1555 INTEGER_TO_PY(st->st_mode),
1556 (unsigned PY_LONG_LONG) st->st_ino,
1557 INTEGER_TO_PY(st->st_dev),
1558 INTEGER_TO_PY(st->st_nlink),
1559 INTEGER_TO_PY(st->st_uid),
1560 INTEGER_TO_PY(st->st_gid),
1561 INTEGER_TO_PY(st->st_rdev),
1562 (PY_LONG_LONG) st->st_size,
1563 INTEGER_TO_PY(st->st_atime),
1564 (long) BUP_STAT_ATIME_NS(st),
1565 INTEGER_TO_PY(st->st_mtime),
1566 (long) BUP_STAT_MTIME_NS(st),
1567 INTEGER_TO_PY(st->st_ctime),
1568 (long) BUP_STAT_CTIME_NS(st));
1572 static PyObject *bup_stat(PyObject *self, PyObject *args)
1577 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1581 rc = stat(filename, &st);
1583 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1584 return stat_struct_to_py(&st, filename, 0);
1588 static PyObject *bup_lstat(PyObject *self, PyObject *args)
1593 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1597 rc = lstat(filename, &st);
1599 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1600 return stat_struct_to_py(&st, filename, 0);
1604 static PyObject *bup_fstat(PyObject *self, PyObject *args)
1608 if (!PyArg_ParseTuple(args, "i", &fd))
1612 rc = fstat(fd, &st);
1614 return PyErr_SetFromErrno(PyExc_OSError);
1615 return stat_struct_to_py(&st, NULL, fd);
1619 #ifdef HAVE_TM_TM_GMTOFF
1620 static PyObject *bup_localtime(PyObject *self, PyObject *args)
1624 if (!PyArg_ParseTuple(args, "L", &lltime))
1626 if (!INTEGRAL_ASSIGNMENT_FITS(&ttime, lltime))
1627 return PyErr_Format(PyExc_OverflowError, "time value too large");
1631 if(localtime_r(&ttime, &tm) == NULL)
1632 return PyErr_SetFromErrno(PyExc_OSError);
1634 // Match the Python struct_time values.
1635 return Py_BuildValue("[i,i,i,i,i,i,i,i,i,i,s]",
1636 1900 + tm.tm_year, tm.tm_mon + 1, tm.tm_mday,
1637 tm.tm_hour, tm.tm_min, tm.tm_sec,
1638 tm.tm_wday, tm.tm_yday + 1,
1639 tm.tm_isdst, tm.tm_gmtoff, tm.tm_zone);
1641 #endif /* def HAVE_TM_TM_GMTOFF */
1644 #ifdef BUP_MINCORE_BUF_TYPE
1645 static PyObject *bup_mincore(PyObject *self, PyObject *args)
1647 Py_buffer src, dest;
1648 PyObject *py_src_n, *py_src_off, *py_dest_off;
1650 if (!PyArg_ParseTuple(args, cstr_argf "*OOw*O",
1651 &src, &py_src_n, &py_src_off,
1652 &dest, &py_dest_off))
1655 PyObject *result = NULL;
1657 unsigned long long src_n, src_off, dest_off;
1658 if (!(bup_ullong_from_py(&src_n, py_src_n, "src_n")
1659 && bup_ullong_from_py(&src_off, py_src_off, "src_off")
1660 && bup_ullong_from_py(&dest_off, py_dest_off, "dest_off")))
1661 goto clean_and_return;
1663 unsigned long long src_region_end;
1664 if (!uadd(&src_region_end, src_off, src_n)) {
1665 result = PyErr_Format(PyExc_OverflowError, "(src_off + src_n) too large");
1666 goto clean_and_return;
1668 assert(src.len >= 0);
1669 if (src_region_end > (unsigned long long) src.len) {
1670 result = PyErr_Format(PyExc_OverflowError, "region runs off end of src");
1671 goto clean_and_return;
1674 unsigned long long dest_size;
1675 if (!INTEGRAL_ASSIGNMENT_FITS(&dest_size, dest.len)) {
1676 result = PyErr_Format(PyExc_OverflowError, "invalid dest size");
1677 goto clean_and_return;
1679 if (dest_off > dest_size) {
1680 result = PyErr_Format(PyExc_OverflowError, "region runs off end of dest");
1681 goto clean_and_return;
1685 if (!INTEGRAL_ASSIGNMENT_FITS(&length, src_n)) {
1686 result = PyErr_Format(PyExc_OverflowError, "src_n overflows size_t");
1687 goto clean_and_return;
1689 int rc = mincore((void *)(src.buf + src_off), length,
1690 (BUP_MINCORE_BUF_TYPE *) (dest.buf + dest_off));
1692 result = PyErr_SetFromErrno(PyExc_OSError);
1693 goto clean_and_return;
1695 result = Py_BuildValue("O", Py_None);
1698 PyBuffer_Release(&src);
1699 PyBuffer_Release(&dest);
1702 #endif /* def BUP_MINCORE_BUF_TYPE */
1705 static PyObject *tuple_from_cstrs(char **cstrs)
1707 // Assumes list is null terminated
1709 while(cstrs[n] != NULL)
1713 if (!INTEGRAL_ASSIGNMENT_FITS(&sn, n))
1714 return PyErr_Format(PyExc_OverflowError, "string array too large");
1716 PyObject *result = PyTuple_New(sn);
1718 for (i = 0; i < sn; i++)
1720 PyObject *gname = Py_BuildValue(cstr_argf, cstrs[i]);
1726 PyTuple_SET_ITEM(result, i, gname);
1731 static PyObject *appropriate_errno_ex(void)
1735 return PyErr_NoMemory();
1739 // In 3.3 IOError was merged into OSError.
1740 return PyErr_SetFromErrno(PyExc_IOError);
1742 return PyErr_SetFromErrno(PyExc_OSError);
1747 static PyObject *pwd_struct_to_py(const struct passwd *pwd)
1749 // We can check the known (via POSIX) signed and unsigned types at
1750 // compile time, but not (easily) the unspecified types, so handle
1751 // those via INTEGER_TO_PY().
1754 return Py_BuildValue(cstr_argf cstr_argf "OO"
1755 cstr_argf cstr_argf cstr_argf,
1758 INTEGER_TO_PY(pwd->pw_uid),
1759 INTEGER_TO_PY(pwd->pw_gid),
1765 static PyObject *bup_getpwuid(PyObject *self, PyObject *args)
1767 unsigned long long py_uid;
1768 if (!PyArg_ParseTuple(args, "K", &py_uid))
1771 if (!INTEGRAL_ASSIGNMENT_FITS(&uid, py_uid))
1772 return PyErr_Format(PyExc_OverflowError, "uid too large for uid_t");
1775 struct passwd *pwd = getpwuid(uid);
1777 return appropriate_errno_ex();
1778 return pwd_struct_to_py(pwd);
1781 static PyObject *bup_getpwnam(PyObject *self, PyObject *args)
1784 if (!PyArg_ParseTuple(args, "S", &py_name))
1787 char *name = PyBytes_AS_STRING(py_name);
1789 struct passwd *pwd = getpwnam(name);
1791 return appropriate_errno_ex();
1792 return pwd_struct_to_py(pwd);
1795 static PyObject *grp_struct_to_py(const struct group *grp)
1797 // We can check the known (via POSIX) signed and unsigned types at
1798 // compile time, but not (easily) the unspecified types, so handle
1799 // those via INTEGER_TO_PY().
1803 PyObject *members = tuple_from_cstrs(grp->gr_mem);
1804 if (members == NULL)
1806 return Py_BuildValue(cstr_argf cstr_argf "OO",
1809 INTEGER_TO_PY(grp->gr_gid),
1813 static PyObject *bup_getgrgid(PyObject *self, PyObject *args)
1815 unsigned long long py_gid;
1816 if (!PyArg_ParseTuple(args, "K", &py_gid))
1819 if (!INTEGRAL_ASSIGNMENT_FITS(&gid, py_gid))
1820 return PyErr_Format(PyExc_OverflowError, "gid too large for gid_t");
1823 struct group *grp = getgrgid(gid);
1825 return appropriate_errno_ex();
1826 return grp_struct_to_py(grp);
1829 static PyObject *bup_getgrnam(PyObject *self, PyObject *args)
1832 if (!PyArg_ParseTuple(args, "S", &py_name))
1835 char *name = PyBytes_AS_STRING(py_name);
1837 struct group *grp = getgrnam(name);
1839 return appropriate_errno_ex();
1840 return grp_struct_to_py(grp);
1844 static PyObject *bup_gethostname(PyObject *mod, PyObject *ignore)
1846 #ifdef HOST_NAME_MAX
1847 char buf[HOST_NAME_MAX + 1] = {};
1849 /* 'SUSv2 guarantees that "Host names are limited to 255 bytes".' */
1853 if (gethostname(buf, sizeof(buf) - 1))
1854 return PyErr_SetFromErrno(PyExc_IOError);
1855 return PyBytes_FromString(buf);
1859 #ifdef BUP_HAVE_READLINE
1861 static char *cstr_from_bytes(PyObject *bytes)
1865 int rc = PyBytes_AsStringAndSize(bytes, &buf, &length);
1868 char *result = checked_malloc(length, sizeof(char));
1871 memcpy(result, buf, length);
1875 static char **cstrs_from_seq(PyObject *seq)
1877 char **result = NULL;
1878 seq = PySequence_Fast(seq, "Cannot convert sequence items to C strings");
1882 const Py_ssize_t len = PySequence_Fast_GET_SIZE(seq);
1883 if (len > PY_SSIZE_T_MAX - 1) {
1884 PyErr_Format(PyExc_OverflowError,
1885 "Sequence length %zd too large for conversion to C array",
1889 result = checked_malloc(len + 1, sizeof(char *));
1893 for (i = 0; i < len; i++)
1895 PyObject *item = PySequence_Fast_GET_ITEM(seq, i);
1897 goto abandon_result;
1898 result[i] = cstr_from_bytes(item);
1901 goto abandon_result;
1919 static char* our_word_break_chars = NULL;
1922 bup_set_completer_word_break_characters(PyObject *self, PyObject *args)
1925 if (!PyArg_ParseTuple(args, cstr_argf, &bytes))
1927 char *prev = our_word_break_chars;
1928 char *next = strdup(bytes);
1930 return PyErr_NoMemory();
1931 our_word_break_chars = next;
1932 rl_completer_word_break_characters = next;
1939 bup_get_completer_word_break_characters(PyObject *self, PyObject *args)
1941 return PyBytes_FromString(rl_completer_word_break_characters);
1944 static PyObject *bup_get_line_buffer(PyObject *self, PyObject *args)
1946 return PyBytes_FromString(rl_line_buffer);
1950 bup_parse_and_bind(PyObject *self, PyObject *args)
1953 if (!PyArg_ParseTuple(args, cstr_argf ":parse_and_bind", &bytes))
1955 char *tmp = strdup(bytes); // Because it may modify the arg
1957 return PyErr_NoMemory();
1958 int rc = rl_parse_and_bind(tmp);
1961 return PyErr_Format(PyExc_OSError,
1962 "system rl_parse_and_bind failed (%d)", rc);
1967 static PyObject *py_on_attempted_completion;
1968 static char **prev_completions;
1970 static char **on_attempted_completion(const char *text, int start, int end)
1972 if (!py_on_attempted_completion)
1975 char **result = NULL;
1976 PyObject *py_result = PyObject_CallFunction(py_on_attempted_completion,
1981 if (py_result != Py_None) {
1982 result = cstrs_from_seq(py_result);
1983 free(prev_completions);
1984 prev_completions = result;
1986 Py_DECREF(py_result);
1991 bup_set_attempted_completion_function(PyObject *self, PyObject *args)
1993 PyObject *completer;
1994 if (!PyArg_ParseTuple(args, "O", &completer))
1997 PyObject *prev = py_on_attempted_completion;
1998 if (completer == Py_None)
2000 py_on_attempted_completion = NULL;
2001 rl_attempted_completion_function = NULL;
2003 py_on_attempted_completion = completer;
2004 rl_attempted_completion_function = on_attempted_completion;
2005 Py_INCREF(completer);
2012 static PyObject *py_on_completion_entry;
2014 static char *on_completion_entry(const char *text, int state)
2016 if (!py_on_completion_entry)
2019 PyObject *py_result = PyObject_CallFunction(py_on_completion_entry,
2020 cstr_argf "i", text, state);
2023 char *result = (py_result == Py_None) ? NULL : cstr_from_bytes(py_result);
2024 Py_DECREF(py_result);
2029 bup_set_completion_entry_function(PyObject *self, PyObject *args)
2031 PyObject *completer;
2032 if (!PyArg_ParseTuple(args, "O", &completer))
2035 PyObject *prev = py_on_completion_entry;
2036 if (completer == Py_None) {
2037 py_on_completion_entry = NULL;
2038 rl_completion_entry_function = NULL;
2040 py_on_completion_entry = completer;
2041 rl_completion_entry_function = on_completion_entry;
2042 Py_INCREF(completer);
2049 bup_readline(PyObject *self, PyObject *args)
2052 if (!PyArg_ParseTuple(args, cstr_argf, &prompt))
2054 char *line = readline(prompt);
2056 return PyErr_Format(PyExc_EOFError, "readline EOF");
2057 PyObject *result = PyBytes_FromString(line);
2062 #endif // defined BUP_HAVE_READLINE
2064 #if defined(HAVE_SYS_ACL_H) && \
2065 defined(HAVE_ACL_LIBACL_H) && \
2066 defined(HAVE_ACL_EXTENDED_FILE) && \
2067 defined(HAVE_ACL_GET_FILE) && \
2068 defined(HAVE_ACL_TO_ANY_TEXT) && \
2069 defined(HAVE_ACL_FROM_TEXT) && \
2070 defined(HAVE_ACL_SET_FILE)
2071 #define ACL_SUPPORT 1
2072 #include <sys/acl.h>
2073 #include <acl/libacl.h>
2077 // -1 for errors, with python exception set
2078 // -2 for ignored errors (not supported)
2079 static int bup_read_acl_to_text(const char *name, acl_type_t type,
2080 char **txt, char **num)
2084 acl = acl_get_file(name, type);
2086 if (errno == EOPNOTSUPP || errno == ENOSYS)
2088 PyErr_SetFromErrno(PyExc_IOError);
2093 *txt = acl_to_any_text(acl, "", '\n', TEXT_ABBREVIATE);
2095 *num = acl_to_any_text(acl, "", '\n', TEXT_ABBREVIATE | TEXT_NUMERIC_IDS);
2100 if (errno == ENOMEM)
2103 PyErr_SetFromErrno(PyExc_IOError);
2106 acl_free((acl_t)*txt);
2108 acl_free((acl_t)*num);
2113 static PyObject *bup_read_acl(PyObject *self, PyObject *args)
2117 PyObject *ret = NULL;
2118 char *acl_txt = NULL, *acl_num = NULL;
2120 if (!PyArg_ParseTuple(args, cstr_argf "i", &name, &isdir))
2123 if (!acl_extended_file(name))
2126 rv = bup_read_acl_to_text(name, ACL_TYPE_ACCESS, &acl_txt, &acl_num);
2131 char *def_txt = NULL, *def_num = NULL;
2133 rv = bup_read_acl_to_text(name, ACL_TYPE_DEFAULT, &def_txt, &def_num);
2137 ret = Py_BuildValue("[" cstr_argf cstr_argf cstr_argf cstr_argf "]",
2138 acl_txt, acl_num, def_txt, def_num);
2141 acl_free((acl_t)def_txt);
2143 acl_free((acl_t)def_num);
2145 ret = Py_BuildValue("[" cstr_argf cstr_argf "]",
2151 acl_free((acl_t)acl_txt);
2153 acl_free((acl_t)acl_num);
2159 static int bup_apply_acl_string(const char *name, const char *s)
2161 acl_t acl = acl_from_text(s);
2165 PyErr_SetFromErrno(PyExc_IOError);
2169 if (acl_set_file(name, ACL_TYPE_ACCESS, acl)) {
2170 PyErr_SetFromErrno(PyExc_IOError);
2179 static PyObject *bup_apply_acl(PyObject *self, PyObject *args)
2181 char *name, *acl, *def = NULL;
2183 if (!PyArg_ParseTuple(args, cstr_argf cstr_argf "|" cstr_argf, &name, &acl, &def))
2186 if (bup_apply_acl_string(name, acl))
2189 if (def && bup_apply_acl_string(name, def))
2196 static PyMethodDef helper_methods[] = {
2197 { "write_sparsely", bup_write_sparsely, METH_VARARGS,
2198 "Write buf excepting zeros at the end. Return trailing zero count." },
2199 { "selftest", selftest, METH_VARARGS,
2200 "Check that the rolling checksum rolls correctly (for unit tests)." },
2201 { "blobbits", blobbits, METH_VARARGS,
2202 "Return the number of bits in the rolling checksum." },
2203 { "splitbuf", splitbuf, METH_VARARGS,
2204 "Split a list of strings based on a rolling checksum." },
2205 { "bitmatch", bitmatch, METH_VARARGS,
2206 "Count the number of matching prefix bits between two strings." },
2207 { "firstword", firstword, METH_VARARGS,
2208 "Return an int corresponding to the first 32 bits of buf." },
2209 { "bloom_contains", bloom_contains, METH_VARARGS,
2210 "Check if a bloom filter of 2^nbits bytes contains an object" },
2211 { "bloom_add", bloom_add, METH_VARARGS,
2212 "Add an object to a bloom filter of 2^nbits bytes" },
2213 { "extract_bits", extract_bits, METH_VARARGS,
2214 "Take the first 'nbits' bits from 'buf' and return them as an int." },
2215 { "merge_into", merge_into, METH_VARARGS,
2216 "Merges a bunch of idx and midx files into a single midx." },
2217 { "write_idx", write_idx, METH_VARARGS,
2218 "Write a PackIdxV2 file from an idx list of lists of tuples" },
2219 { "write_random", write_random, METH_VARARGS,
2220 "Write random bytes to the given file descriptor" },
2221 { "random_sha", random_sha, METH_VARARGS,
2222 "Return a random 20-byte string" },
2223 { "open_noatime", open_noatime, METH_VARARGS,
2224 "open() the given filename for read with O_NOATIME if possible" },
2225 { "fadvise_done", fadvise_done, METH_VARARGS,
2226 "Inform the kernel that we're finished with earlier parts of a file" },
2227 #ifdef BUP_HAVE_FILE_ATTRS
2228 { "get_linux_file_attr", bup_get_linux_file_attr, METH_VARARGS,
2229 "Return the Linux attributes for the given file." },
2231 #ifdef BUP_HAVE_FILE_ATTRS
2232 { "set_linux_file_attr", bup_set_linux_file_attr, METH_VARARGS,
2233 "Set the Linux attributes for the given file." },
2236 #ifndef BUP_USE_PYTHON_UTIME // just for Python 2 now
2237 #ifdef HAVE_UTIMENSAT
2238 { "bup_utimensat", bup_utimensat, METH_VARARGS,
2239 "Change path timestamps with nanosecond precision (POSIX)." },
2242 { "bup_utimes", bup_utimes, METH_VARARGS,
2243 "Change path timestamps with microsecond precision." },
2246 { "bup_lutimes", bup_lutimes, METH_VARARGS,
2247 "Change path timestamps with microsecond precision;"
2248 " don't follow symlinks." },
2250 #endif // defined BUP_USE_PYTHON_UTIME
2252 { "stat", bup_stat, METH_VARARGS,
2253 "Extended version of stat." },
2254 { "lstat", bup_lstat, METH_VARARGS,
2255 "Extended version of lstat." },
2256 { "fstat", bup_fstat, METH_VARARGS,
2257 "Extended version of fstat." },
2258 #ifdef HAVE_TM_TM_GMTOFF
2259 { "localtime", bup_localtime, METH_VARARGS,
2260 "Return struct_time elements plus the timezone offset and name." },
2262 { "bytescmp", bup_bytescmp, METH_VARARGS,
2263 "Return a negative value if x < y, zero if equal, positive otherwise."},
2264 { "cat_bytes", bup_cat_bytes, METH_VARARGS,
2265 "For (x_bytes, x_ofs, x_n, y_bytes, y_ofs, y_n) arguments, return their concatenation."},
2266 #ifdef BUP_MINCORE_BUF_TYPE
2267 { "mincore", bup_mincore, METH_VARARGS,
2268 "For mincore(src, src_n, src_off, dest, dest_off)"
2269 " call the system mincore(src + src_off, src_n, &dest[dest_off])." },
2271 { "getpwuid", bup_getpwuid, METH_VARARGS,
2272 "Return the password database entry for the given numeric user id,"
2273 " as a tuple with all C strings as bytes(), or None if the user does"
2275 { "getpwnam", bup_getpwnam, METH_VARARGS,
2276 "Return the password database entry for the given user name,"
2277 " as a tuple with all C strings as bytes(), or None if the user does"
2279 { "getgrgid", bup_getgrgid, METH_VARARGS,
2280 "Return the group database entry for the given numeric group id,"
2281 " as a tuple with all C strings as bytes(), or None if the group does"
2283 { "getgrnam", bup_getgrnam, METH_VARARGS,
2284 "Return the group database entry for the given group name,"
2285 " as a tuple with all C strings as bytes(), or None if the group does"
2287 { "gethostname", bup_gethostname, METH_NOARGS,
2288 "Return the current hostname (as bytes)" },
2289 #ifdef BUP_HAVE_READLINE
2290 { "set_completion_entry_function", bup_set_completion_entry_function, METH_VARARGS,
2291 "Set rl_completion_entry_function. Called as f(text, state)." },
2292 { "set_attempted_completion_function", bup_set_attempted_completion_function, METH_VARARGS,
2293 "Set rl_attempted_completion_function. Called as f(text, start, end)." },
2294 { "parse_and_bind", bup_parse_and_bind, METH_VARARGS,
2295 "Call rl_parse_and_bind." },
2296 { "get_line_buffer", bup_get_line_buffer, METH_NOARGS,
2297 "Return rl_line_buffer." },
2298 { "get_completer_word_break_characters", bup_get_completer_word_break_characters, METH_NOARGS,
2299 "Return rl_completer_word_break_characters." },
2300 { "set_completer_word_break_characters", bup_set_completer_word_break_characters, METH_VARARGS,
2301 "Set rl_completer_word_break_characters." },
2302 { "readline", bup_readline, METH_VARARGS,
2303 "Call readline(prompt)." },
2304 #endif // defined BUP_HAVE_READLINE
2306 { "read_acl", bup_read_acl, METH_VARARGS,
2307 "read_acl(name, isdir)\n\n"
2308 "Read ACLs for the given file/dirname and return the correctly encoded"
2309 " list [txt, num, def_tx, def_num] (the def_* being empty bytestrings"
2310 " unless the second argument 'isdir' is True)." },
2311 { "apply_acl", bup_apply_acl, METH_VARARGS,
2312 "apply_acl(name, acl, def=None)\n\n"
2313 "Given a file/dirname (bytes) and the ACLs to restore, do that." },
2314 #endif /* HAVE_ACLS */
2315 { NULL, NULL, 0, NULL }, // sentinel
2318 static void test_integral_assignment_fits(void)
2320 assert(sizeof(signed short) == sizeof(unsigned short));
2321 assert(sizeof(signed short) < sizeof(signed long long));
2322 assert(sizeof(signed short) < sizeof(unsigned long long));
2323 assert(sizeof(unsigned short) < sizeof(signed long long));
2324 assert(sizeof(unsigned short) < sizeof(unsigned long long));
2325 assert(sizeof(Py_ssize_t) <= sizeof(size_t));
2327 signed short ss, ssmin = SHRT_MIN, ssmax = SHRT_MAX;
2328 unsigned short us, usmax = USHRT_MAX;
2329 signed long long sllmin = LLONG_MIN, sllmax = LLONG_MAX;
2330 unsigned long long ullmax = ULLONG_MAX;
2332 assert(INTEGRAL_ASSIGNMENT_FITS(&ss, ssmax));
2333 assert(INTEGRAL_ASSIGNMENT_FITS(&ss, ssmin));
2334 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, usmax));
2335 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, sllmin));
2336 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, sllmax));
2337 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, ullmax));
2339 assert(INTEGRAL_ASSIGNMENT_FITS(&us, usmax));
2340 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, ssmin));
2341 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, sllmin));
2342 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, sllmax));
2343 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, ullmax));
2347 static int setup_module(PyObject *m)
2349 // FIXME: migrate these tests to configure, or at least don't
2350 // possibly crash the whole application. Check against the type
2351 // we're going to use when passing to python. Other stat types
2352 // are tested at runtime.
2353 assert(sizeof(ino_t) <= sizeof(unsigned PY_LONG_LONG));
2354 assert(sizeof(off_t) <= sizeof(PY_LONG_LONG));
2355 assert(sizeof(blksize_t) <= sizeof(PY_LONG_LONG));
2356 assert(sizeof(blkcnt_t) <= sizeof(PY_LONG_LONG));
2357 // Just be sure (relevant when passing timestamps back to Python above).
2358 assert(sizeof(PY_LONG_LONG) <= sizeof(long long));
2359 assert(sizeof(unsigned PY_LONG_LONG) <= sizeof(unsigned long long));
2361 test_integral_assignment_fits();
2363 // Originally required by append_sparse_region()
2366 if (!INTEGRAL_ASSIGNMENT_FITS(&probe, INT_MAX))
2368 fprintf(stderr, "off_t can't hold INT_MAX; please report.\n");
2376 value = INTEGER_TO_PY(INT_MAX);
2377 PyObject_SetAttrString(m, "INT_MAX", value);
2379 value = INTEGER_TO_PY(UINT_MAX);
2380 PyObject_SetAttrString(m, "UINT_MAX", value);
2384 #ifndef BUP_USE_PYTHON_UTIME // just for Python 2 now
2385 #ifdef HAVE_UTIMENSAT
2388 value = INTEGER_TO_PY(AT_FDCWD);
2389 PyObject_SetAttrString(m, "AT_FDCWD", value);
2391 value = INTEGER_TO_PY(AT_SYMLINK_NOFOLLOW);
2392 PyObject_SetAttrString(m, "AT_SYMLINK_NOFOLLOW", value);
2394 value = INTEGER_TO_PY(UTIME_NOW);
2395 PyObject_SetAttrString(m, "UTIME_NOW", value);
2399 #endif // defined BUP_USE_PYTHON_UTIME
2401 #ifdef BUP_HAVE_MINCORE_INCORE
2404 value = INTEGER_TO_PY(MINCORE_INCORE);
2405 PyObject_SetAttrString(m, "MINCORE_INCORE", value);
2410 e = getenv("BUP_FORCE_TTY");
2411 get_state(m)->istty2 = isatty(2) || (atoi(e ? e : "0") & 2);
2416 #if PY_MAJOR_VERSION < 3
2418 PyMODINIT_FUNC init_helpers(void)
2420 PyObject *m = Py_InitModule("_helpers", helper_methods);
2422 PyErr_SetString(PyExc_RuntimeError, "bup._helpers init failed");
2425 if (!setup_module(m))
2427 PyErr_SetString(PyExc_RuntimeError, "bup._helpers set up failed");
2433 # else // PY_MAJOR_VERSION >= 3
2435 static struct PyModuleDef helpers_def = {
2436 PyModuleDef_HEAD_INIT,
2442 NULL, // helpers_traverse,
2443 NULL, // helpers_clear,
2447 PyMODINIT_FUNC PyInit__helpers(void)
2449 PyObject *module = PyModule_Create(&helpers_def);
2452 if (!setup_module(module))
2460 #endif // PY_MAJOR_VERSION >= 3