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
10 #include <arpa/inet.h>
22 #ifdef HAVE_SYS_MMAN_H
25 #ifdef HAVE_SYS_TYPES_H
26 #include <sys/types.h>
28 #ifdef HAVE_SYS_STAT_H
34 #ifdef HAVE_SYS_TIME_H
38 #ifdef HAVE_LINUX_FS_H
41 #ifdef HAVE_SYS_IOCTL_H
42 #include <sys/ioctl.h>
45 #ifdef HAVE_TM_TM_GMTOFF
49 #if defined(BUP_RL_EXPECTED_XOPEN_SOURCE) \
50 && (!defined(_XOPEN_SOURCE) || _XOPEN_SOURCE < BUP_RL_EXPECTED_XOPEN_SOURCE)
51 # warning "_XOPEN_SOURCE version is incorrect for readline"
54 #ifdef BUP_HAVE_READLINE
55 # pragma GCC diagnostic push
56 # pragma GCC diagnostic ignored "-Wstrict-prototypes"
57 # ifdef BUP_READLINE_INCLUDES_IN_SUBDIR
58 # include <readline/readline.h>
59 # include <readline/history.h>
61 # include <readline.h>
64 # pragma GCC diagnostic pop
69 #if defined(FS_IOC_GETFLAGS) && defined(FS_IOC_SETFLAGS)
70 #define BUP_HAVE_FILE_ATTRS 1
74 * Check for incomplete UTIMENSAT support (NetBSD 6), and if so,
75 * pretend we don't have it.
77 #if !defined(AT_FDCWD) || !defined(AT_SYMLINK_NOFOLLOW)
82 // Of course, this assumes it's a bitfield value.
87 typedef unsigned char byte;
94 // cstr_argf: for byte vectors without null characters (e.g. paths)
95 // rbuf_argf: for read-only byte vectors
96 // wbuf_argf: for mutable byte vectors
98 #if PY_MAJOR_VERSION < 3
100 # define get_state(x) (&state)
101 # define cstr_argf "s"
102 # define rbuf_argf "s#"
103 # define wbuf_argf "s*"
105 # define get_state(x) ((state_t *) PyModule_GetState(x))
106 # define cstr_argf "y"
107 # define rbuf_argf "y#"
108 # define wbuf_argf "y*"
109 #endif // PY_MAJOR_VERSION >= 3
112 static void *checked_calloc(size_t n, size_t size)
114 void *result = calloc(n, size);
120 #ifndef BUP_HAVE_BUILTIN_MUL_OVERFLOW
122 #define checked_malloc checked_calloc
124 #else // defined BUP_HAVE_BUILTIN_MUL_OVERFLOW
126 static void *checked_malloc(size_t n, size_t size)
129 if (__builtin_mul_overflow(n, size, &total))
131 PyErr_Format(PyExc_OverflowError,
132 "request to allocate %zu items of size %zu is too large",
136 void *result = malloc(total);
138 return PyErr_NoMemory();
142 #endif // defined BUP_HAVE_BUILTIN_MUL_OVERFLOW
146 // This function should technically be macro'd out if it's going to be used
147 // more than ocasionally. As of this writing, it'll actually never be called
148 // in real world bup scenarios (because our packs are < MAX_INT bytes).
149 static uint64_t htonll(uint64_t value)
151 static const int endian_test = 42;
153 if (*(char *)&endian_test == endian_test) // LSB-MSB
154 return ((uint64_t)htonl(value & 0xFFFFFFFF) << 32) | htonl(value >> 32);
155 return value; // already in network byte order MSB-LSB
160 // Disabling sign-compare here should be fine since we're explicitly
161 // checking for a sign mismatch, i.e. if the signs don't match, then
162 // it doesn't matter what the value comparison says.
163 // FIXME: ... so should we reverse the order?
164 #define INTEGRAL_ASSIGNMENT_FITS(dest, src) \
166 _Pragma("GCC diagnostic push"); \
167 _Pragma("GCC diagnostic ignored \"-Wsign-compare\""); \
168 _Pragma("clang diagnostic push"); \
169 _Pragma("clang diagnostic ignored \"-Wshorten-64-to-32\""); \
171 int result = *(dest) == (src) && (*(dest) < 1) == ((src) < 1); \
172 _Pragma("clang diagnostic pop"); \
173 _Pragma("GCC diagnostic pop"); \
178 #define INTEGER_TO_PY(x) \
180 _Pragma("GCC diagnostic push"); \
181 _Pragma("GCC diagnostic ignored \"-Wtype-limits\""); \
182 _Pragma("clang diagnostic push"); \
183 _Pragma("clang diagnostic ignored \"-Wtautological-compare\""); \
184 PyObject *result = ((x) >= 0) ? PyLong_FromUnsignedLongLong(x) : PyLong_FromLongLong(x); \
185 _Pragma("clang diagnostic pop"); \
186 _Pragma("GCC diagnostic pop"); \
191 #if PY_MAJOR_VERSION < 3
192 static int bup_ulong_from_pyint(unsigned long *x, PyObject *py,
195 const long tmp = PyInt_AsLong(py);
196 if (tmp == -1 && PyErr_Occurred())
198 if (PyErr_ExceptionMatches(PyExc_OverflowError))
199 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned long",
205 PyErr_Format(PyExc_OverflowError,
206 "negative %s cannot be converted to unsigned long", name);
215 static int bup_ulong_from_py(unsigned long *x, PyObject *py, const char *name)
217 #if PY_MAJOR_VERSION < 3
219 return bup_ulong_from_pyint(x, py, name);
222 if (!PyLong_Check(py))
224 PyErr_Format(PyExc_TypeError, "expected integer %s", name);
228 const unsigned long tmp = PyLong_AsUnsignedLong(py);
229 if (PyErr_Occurred())
231 if (PyErr_ExceptionMatches(PyExc_OverflowError))
232 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned long",
241 static int bup_uint_from_py(unsigned int *x, PyObject *py, const char *name)
244 if (!bup_ulong_from_py(&tmp, py, name))
249 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned int", name);
252 *x = (unsigned int) tmp;
256 static int bup_ullong_from_py(unsigned PY_LONG_LONG *x, PyObject *py,
259 #if PY_MAJOR_VERSION < 3
263 if (bup_ulong_from_pyint(&tmp, py, name))
272 if (!PyLong_Check(py))
274 PyErr_Format(PyExc_TypeError, "integer argument expected for %s", name);
278 const unsigned PY_LONG_LONG tmp = PyLong_AsUnsignedLongLong(py);
279 if (tmp == (unsigned long long) -1 && PyErr_Occurred())
281 if (PyErr_ExceptionMatches(PyExc_OverflowError))
282 PyErr_Format(PyExc_OverflowError,
283 "%s too big for unsigned long long", name);
291 static PyObject *bup_bytescmp(PyObject *self, PyObject *args)
293 PyObject *py_s1, *py_s2; // This is really a PyBytes/PyString
294 if (!PyArg_ParseTuple(args, "SS", &py_s1, &py_s2))
297 Py_ssize_t s1_len, s2_len;
298 if (PyBytes_AsStringAndSize(py_s1, &s1, &s1_len) == -1)
300 if (PyBytes_AsStringAndSize(py_s2, &s2, &s2_len) == -1)
302 const Py_ssize_t n = (s1_len < s2_len) ? s1_len : s2_len;
303 const int cmp = memcmp(s1, s2, n);
305 return PyLong_FromLong(cmp);
306 if (s1_len == s2_len)
307 return PyLong_FromLong(0);;
308 return PyLong_FromLong((s1_len < s2_len) ? -1 : 1);
312 static PyObject *bup_cat_bytes(PyObject *self, PyObject *args)
314 unsigned char *bufx = NULL, *bufy = NULL;
315 Py_ssize_t bufx_len, bufx_ofs, bufx_n;
316 Py_ssize_t bufy_len, bufy_ofs, bufy_n;
317 if (!PyArg_ParseTuple(args,
320 &bufx, &bufx_len, &bufx_ofs, &bufx_n,
321 &bufy, &bufy_len, &bufy_ofs, &bufy_n))
324 return PyErr_Format(PyExc_ValueError, "negative x offset");
326 return PyErr_Format(PyExc_ValueError, "negative x extent");
327 if (bufx_ofs > bufx_len)
328 return PyErr_Format(PyExc_ValueError, "x offset greater than length");
329 if (bufx_n > bufx_len - bufx_ofs)
330 return PyErr_Format(PyExc_ValueError, "x extent past end of buffer");
333 return PyErr_Format(PyExc_ValueError, "negative y offset");
335 return PyErr_Format(PyExc_ValueError, "negative y extent");
336 if (bufy_ofs > bufy_len)
337 return PyErr_Format(PyExc_ValueError, "y offset greater than length");
338 if (bufy_n > bufy_len - bufy_ofs)
339 return PyErr_Format(PyExc_ValueError, "y extent past end of buffer");
341 if (bufy_n > PY_SSIZE_T_MAX - bufx_n)
342 return PyErr_Format(PyExc_OverflowError, "result length too long");
344 PyObject *result = PyBytes_FromStringAndSize(NULL, bufx_n + bufy_n);
346 return PyErr_NoMemory();
347 char *buf = PyBytes_AS_STRING(result);
348 memcpy(buf, bufx + bufx_ofs, bufx_n);
349 memcpy(buf + bufx_n, bufy + bufy_ofs, bufy_n);
355 // Probably we should use autoconf or something and set HAVE_PY_GETARGCARGV...
356 #if __WIN32__ || __CYGWIN__
358 // There's no 'ps' on win32 anyway, and Py_GetArgcArgv() isn't available.
359 static void unpythonize_argv(void) { }
361 #else // not __WIN32__
363 // For some reason this isn't declared in Python.h
364 extern void Py_GetArgcArgv(int *argc, char ***argv);
366 static void unpythonize_argv(void)
371 Py_GetArgcArgv(&argc, &argv);
373 for (i = 0; i < argc-1; i++)
375 if (argv[i] + strlen(argv[i]) + 1 != argv[i+1])
377 // The argv block doesn't work the way we expected; it's unsafe
383 arge = argv[argc-1] + strlen(argv[argc-1]) + 1;
385 if (strstr(argv[0], "python") && argv[1] == argv[0] + strlen(argv[0]) + 1)
389 p = strrchr(argv[1], '/');
395 memmove(argv[0], p, len);
396 memset(arge - diff, 0, diff);
397 for (i = 0; i < argc; i++)
398 argv[i] = argv[i+1] ? argv[i+1]-diff : NULL;
403 #endif // not __WIN32__ or __CYGWIN__
406 static int write_all(int fd, const void *buf, const size_t count)
409 while (written < count)
411 const ssize_t rc = write(fd, buf + written, count - written);
420 static int uadd(unsigned long long *dest,
421 const unsigned long long x,
422 const unsigned long long y)
424 const unsigned long long result = x + y;
425 if (result < x || result < y)
432 static PyObject *append_sparse_region(const int fd, unsigned long long n)
437 if (!INTEGRAL_ASSIGNMENT_FITS(&new_off, n))
439 const off_t off = lseek(fd, new_off, SEEK_CUR);
440 if (off == (off_t) -1)
441 return PyErr_SetFromErrno(PyExc_IOError);
448 static PyObject *record_sparse_zeros(unsigned long long *new_pending,
450 unsigned long long prev_pending,
451 const unsigned long long count)
453 // Add count additional sparse zeros to prev_pending and store the
454 // result in new_pending, or if the total won't fit in
455 // new_pending, write some of the zeros to fd sparsely, and store
456 // the remaining sum in new_pending.
457 if (!uadd(new_pending, prev_pending, count))
459 PyObject *err = append_sparse_region(fd, prev_pending);
462 *new_pending = count;
468 static byte* find_not_zero(const byte * const start, const byte * const end)
470 // Return a pointer to first non-zero byte between start and end,
471 // or end if there isn't one.
472 assert(start <= end);
473 const unsigned char *cur = start;
474 while (cur < end && *cur == 0)
480 static byte* find_trailing_zeros(const byte * const start,
481 const byte * const end)
483 // Return a pointer to the start of any trailing run of zeros, or
484 // end if there isn't one.
485 assert(start <= end);
488 const byte * cur = end;
489 while (cur > start && *--cur == 0) {}
493 return (byte *) (cur + 1);
497 static byte *find_non_sparse_end(const byte * const start,
498 const byte * const end,
499 const ptrdiff_t min_len)
501 // Return the first pointer to a min_len sparse block in [start,
502 // end) if there is one, otherwise a pointer to the start of any
503 // trailing run of zeros. If there are no trailing zeros, return
509 // Probe in min_len jumps, searching backward from the jump
510 // destination for a non-zero byte. If such a byte is found, move
511 // just past it and try again.
512 const byte *candidate = start;
513 // End of any run of zeros, starting at candidate, that we've already seen
514 const byte *end_of_known_zeros = candidate;
515 while (end - candidate >= min_len) // Handle all min_len candidate blocks
517 const byte * const probe_end = candidate + min_len;
518 const byte * const trailing_zeros =
519 find_trailing_zeros(end_of_known_zeros, probe_end);
520 if (trailing_zeros == probe_end)
521 end_of_known_zeros = candidate = probe_end;
522 else if (trailing_zeros == end_of_known_zeros)
524 assert(candidate >= start);
525 assert(candidate <= end);
526 assert(*candidate == 0);
527 return (byte *) candidate;
531 candidate = trailing_zeros;
532 end_of_known_zeros = probe_end;
536 if (candidate == end)
539 // No min_len sparse run found, search backward from end
540 const byte * const trailing_zeros = find_trailing_zeros(end_of_known_zeros,
543 if (trailing_zeros == end_of_known_zeros)
545 assert(candidate >= start);
546 assert(candidate < end);
547 assert(*candidate == 0);
548 assert(end - candidate < min_len);
549 return (byte *) candidate;
552 if (trailing_zeros == end)
554 assert(*(end - 1) != 0);
558 assert(end - trailing_zeros < min_len);
559 assert(trailing_zeros >= start);
560 assert(trailing_zeros < end);
561 assert(*trailing_zeros == 0);
562 return (byte *) trailing_zeros;
566 static PyObject *bup_write_sparsely(PyObject *self, PyObject *args)
569 unsigned char *buf = NULL;
571 PyObject *py_min_sparse_len, *py_prev_sparse_len;
572 if (!PyArg_ParseTuple(args, "i" rbuf_argf "OO",
573 &fd, &buf, &sbuf_len,
574 &py_min_sparse_len, &py_prev_sparse_len))
576 ptrdiff_t min_sparse_len;
577 unsigned long long prev_sparse_len, buf_len, ul_min_sparse_len;
578 if (!bup_ullong_from_py(&ul_min_sparse_len, py_min_sparse_len, "min_sparse_len"))
580 if (!INTEGRAL_ASSIGNMENT_FITS(&min_sparse_len, ul_min_sparse_len))
581 return PyErr_Format(PyExc_OverflowError, "min_sparse_len too large");
582 if (!bup_ullong_from_py(&prev_sparse_len, py_prev_sparse_len, "prev_sparse_len"))
585 return PyErr_Format(PyExc_ValueError, "negative bufer length");
586 if (!INTEGRAL_ASSIGNMENT_FITS(&buf_len, sbuf_len))
587 return PyErr_Format(PyExc_OverflowError, "buffer length too large");
589 const byte * block = buf; // Start of pending block
590 const byte * const end = buf + buf_len;
591 unsigned long long zeros = prev_sparse_len;
594 assert(block <= end);
596 return PyLong_FromUnsignedLongLong(zeros);
600 // Look for the end of block, i.e. the next sparse run of
601 // at least min_sparse_len zeros, or the end of the
603 const byte * const probe = find_non_sparse_end(block + 1, end,
605 // Either at end of block, or end of non-sparse; write pending data
606 PyObject *err = append_sparse_region(fd, zeros);
609 int rc = write_all(fd, block, probe - block);
611 return PyErr_SetFromErrno(PyExc_IOError);
613 if (end - probe < min_sparse_len)
616 zeros = min_sparse_len;
617 block = probe + zeros;
621 // Should be in the first loop iteration, a sparse run of
622 // zeros, or nearly at the end of the block (within
624 const byte * const zeros_end = find_not_zero(block, end);
625 PyObject *err = record_sparse_zeros(&zeros, fd,
626 zeros, zeros_end - block);
629 assert(block <= zeros_end);
636 static PyObject *selftest(PyObject *self, PyObject *args)
638 if (!PyArg_ParseTuple(args, ""))
641 return Py_BuildValue("i", !bupsplit_selftest());
645 static PyObject *blobbits(PyObject *self, PyObject *args)
647 if (!PyArg_ParseTuple(args, ""))
649 return Py_BuildValue("i", BUP_BLOBBITS);
653 static PyObject *splitbuf(PyObject *self, PyObject *args)
655 // We stick to buffers in python 2 because they appear to be
656 // substantially smaller than memoryviews, and because
657 // zlib.compress() in python 2 can't accept a memoryview
658 // (cf. hashsplit.py).
659 int out = 0, bits = -1;
660 if (PY_MAJOR_VERSION > 2)
663 if (!PyArg_ParseTuple(args, "y*", &buf))
665 assert(buf.len <= INT_MAX);
666 out = bupsplit_find_ofs(buf.buf, buf.len, &bits);
667 PyBuffer_Release(&buf);
671 unsigned char *buf = NULL;
673 if (!PyArg_ParseTuple(args, "t#", &buf, &len))
675 assert(len <= INT_MAX);
676 out = bupsplit_find_ofs(buf, (int) len, &bits);
678 if (out) assert(bits >= BUP_BLOBBITS);
679 return Py_BuildValue("ii", out, bits);
683 static PyObject *bitmatch(PyObject *self, PyObject *args)
685 unsigned char *buf1 = NULL, *buf2 = NULL;
686 Py_ssize_t len1 = 0, len2 = 0;
690 if (!PyArg_ParseTuple(args, rbuf_argf rbuf_argf, &buf1, &len1, &buf2, &len2))
694 for (byte = 0; byte < len1 && byte < len2; byte++)
696 int b1 = buf1[byte], b2 = buf2[byte];
699 for (bit = 0; bit < 8; bit++)
700 if ( (b1 & (0x80 >> bit)) != (b2 & (0x80 >> bit)) )
706 assert(byte <= (INT_MAX >> 3));
707 return Py_BuildValue("i", byte*8 + bit);
711 static PyObject *firstword(PyObject *self, PyObject *args)
713 unsigned char *buf = NULL;
717 if (!PyArg_ParseTuple(args, rbuf_argf, &buf, &len))
723 v = ntohl(*(uint32_t *)buf);
724 return PyLong_FromUnsignedLong(v);
728 #define BLOOM2_HEADERLEN 16
730 static void to_bloom_address_bitmask4(const unsigned char *buf,
731 const int nbits, uint64_t *v, unsigned char *bitmask)
737 memcpy(&high, buf, 4);
738 mask = (1<<nbits) - 1;
739 raw = (((uint64_t)ntohl(high) << 8) | buf[4]);
740 bit = (raw >> (37-nbits)) & 0x7;
741 *v = (raw >> (40-nbits)) & mask;
745 static void to_bloom_address_bitmask5(const unsigned char *buf,
746 const int nbits, uint32_t *v, unsigned char *bitmask)
752 memcpy(&high, buf, 4);
753 mask = (1<<nbits) - 1;
755 bit = (raw >> (29-nbits)) & 0x7;
756 *v = (raw >> (32-nbits)) & mask;
760 #define BLOOM_SET_BIT(name, address, otype) \
761 static void name(unsigned char *bloom, const unsigned char *buf, const int nbits)\
763 unsigned char bitmask;\
765 address(buf, nbits, &v, &bitmask);\
766 bloom[BLOOM2_HEADERLEN+v] |= bitmask;\
768 BLOOM_SET_BIT(bloom_set_bit4, to_bloom_address_bitmask4, uint64_t)
769 BLOOM_SET_BIT(bloom_set_bit5, to_bloom_address_bitmask5, uint32_t)
772 #define BLOOM_GET_BIT(name, address, otype) \
773 static int name(const unsigned char *bloom, const unsigned char *buf, const int nbits)\
775 unsigned char bitmask;\
777 address(buf, nbits, &v, &bitmask);\
778 return bloom[BLOOM2_HEADERLEN+v] & bitmask;\
780 BLOOM_GET_BIT(bloom_get_bit4, to_bloom_address_bitmask4, uint64_t)
781 BLOOM_GET_BIT(bloom_get_bit5, to_bloom_address_bitmask5, uint32_t)
784 static PyObject *bloom_add(PyObject *self, PyObject *args)
786 Py_buffer bloom, sha;
787 int nbits = 0, k = 0;
788 if (!PyArg_ParseTuple(args, wbuf_argf wbuf_argf "ii",
789 &bloom, &sha, &nbits, &k))
792 PyObject *result = NULL;
794 if (bloom.len < 16+(1<<nbits) || sha.len % 20 != 0)
795 goto clean_and_return;
800 goto clean_and_return;
801 unsigned char *cur = sha.buf;
803 for (end = cur + sha.len; cur < end; cur += 20/k)
804 bloom_set_bit5(bloom.buf, cur, nbits);
809 goto clean_and_return;
810 unsigned char *cur = sha.buf;
811 unsigned char *end = cur + sha.len;
812 for (; cur < end; cur += 20/k)
813 bloom_set_bit4(bloom.buf, cur, nbits);
816 goto clean_and_return;
818 result = Py_BuildValue("n", sha.len / 20);
821 PyBuffer_Release(&bloom);
822 PyBuffer_Release(&sha);
826 static PyObject *bloom_contains(PyObject *self, PyObject *args)
829 unsigned char *sha = NULL;
831 int nbits = 0, k = 0;
832 if (!PyArg_ParseTuple(args, wbuf_argf rbuf_argf "ii",
833 &bloom, &sha, &len, &nbits, &k))
836 PyObject *result = NULL;
839 goto clean_and_return;
844 goto clean_and_return;
847 for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
848 if (!bloom_get_bit5(bloom.buf, sha, nbits))
850 result = Py_BuildValue("Oi", Py_None, steps);
851 goto clean_and_return;
857 goto clean_and_return;
860 for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
861 if (!bloom_get_bit4(bloom.buf, sha, nbits))
863 result = Py_BuildValue("Oi", Py_None, steps);
864 goto clean_and_return;
868 goto clean_and_return;
870 result = Py_BuildValue("ii", 1, k);
873 PyBuffer_Release(&bloom);
878 static uint32_t _extract_bits(unsigned char *buf, int nbits)
882 mask = (1<<nbits) - 1;
883 v = ntohl(*(uint32_t *)buf);
884 v = (v >> (32-nbits)) & mask;
889 static PyObject *extract_bits(PyObject *self, PyObject *args)
891 unsigned char *buf = NULL;
895 if (!PyArg_ParseTuple(args, rbuf_argf "i", &buf, &len, &nbits))
901 return PyLong_FromUnsignedLong(_extract_bits(buf, nbits));
906 unsigned char bytes[20];
909 static inline int _cmp_sha(const struct sha *sha1, const struct sha *sha2)
911 return memcmp(sha1->bytes, sha2->bytes, sizeof(sha1->bytes));
924 static void _fix_idx_order(struct idx **idxs, Py_ssize_t *last_i)
927 Py_ssize_t low, mid, high;
931 if (idxs[*last_i]->cur >= idxs[*last_i]->end)
933 idxs[*last_i] = NULL;
946 mid = (low + high) / 2;
947 c = _cmp_sha(idx->cur, idxs[mid]->cur);
959 memmove(&idxs[mid+1], &idxs[mid], (*last_i-mid)*sizeof(struct idx *));
964 static uint32_t _get_idx_i(struct idx *idx)
966 if (idx->cur_name == NULL)
967 return idx->name_base;
968 return ntohl(*idx->cur_name) + idx->name_base;
971 #define MIDX4_HEADERLEN 12
973 static PyObject *merge_into(PyObject *self, PyObject *args)
975 struct sha *sha_ptr, *sha_start = NULL;
976 uint32_t *table_ptr, *name_ptr, *name_start;
979 uint32_t count, prefix;
984 PyObject *py_total, *ilist = NULL;
985 if (!PyArg_ParseTuple(args, wbuf_argf "iOO",
986 &fmap, &bits, &py_total, &ilist))
989 PyObject *result = NULL;
990 struct idx **idxs = NULL;
991 Py_ssize_t num_i = 0;
992 int *idx_buf_init = NULL;
993 Py_buffer *idx_buf = NULL;
995 if (!bup_uint_from_py(&total, py_total, "total"))
996 goto clean_and_return;
998 num_i = PyList_Size(ilist);
1000 if (!(idxs = checked_malloc(num_i, sizeof(struct idx *))))
1001 goto clean_and_return;
1002 if (!(idx_buf_init = checked_calloc(num_i, sizeof(int))))
1003 goto clean_and_return;
1004 if (!(idx_buf = checked_malloc(num_i, sizeof(Py_buffer))))
1005 goto clean_and_return;
1007 for (i = 0; i < num_i; i++)
1009 long len, sha_ofs, name_map_ofs;
1010 if (!(idxs[i] = checked_malloc(1, sizeof(struct idx))))
1011 goto clean_and_return;
1012 PyObject *itup = PyList_GetItem(ilist, i);
1013 if (!PyArg_ParseTuple(itup, wbuf_argf "llli",
1014 &(idx_buf[i]), &len, &sha_ofs, &name_map_ofs,
1015 &idxs[i]->name_base))
1017 idx_buf_init[i] = 1;
1018 idxs[i]->map = idx_buf[i].buf;
1019 idxs[i]->bytes = idx_buf[i].len;
1020 idxs[i]->cur = (struct sha *)&idxs[i]->map[sha_ofs];
1021 idxs[i]->end = &idxs[i]->cur[len];
1023 idxs[i]->cur_name = (uint32_t *)&idxs[i]->map[name_map_ofs];
1025 idxs[i]->cur_name = NULL;
1027 table_ptr = (uint32_t *) &((unsigned char *) fmap.buf)[MIDX4_HEADERLEN];
1028 sha_start = sha_ptr = (struct sha *)&table_ptr[1<<bits];
1029 name_start = name_ptr = (uint32_t *)&sha_ptr[total];
1031 Py_ssize_t last_i = num_i - 1;
1037 uint32_t new_prefix;
1038 if (count % 102424 == 0 && get_state(self)->istty2)
1039 fprintf(stderr, "midx: writing %.2f%% (%d/%d)\r",
1040 count*100.0/total, count, total);
1042 new_prefix = _extract_bits((unsigned char *)idx->cur, bits);
1043 while (prefix < new_prefix)
1044 table_ptr[prefix++] = htonl(count);
1045 memcpy(sha_ptr++, idx->cur, sizeof(struct sha));
1046 *name_ptr++ = htonl(_get_idx_i(idx));
1048 if (idx->cur_name != NULL)
1050 _fix_idx_order(idxs, &last_i);
1053 while (prefix < ((uint32_t) 1 << bits))
1054 table_ptr[prefix++] = htonl(count);
1055 assert(count == total);
1056 assert(prefix == ((uint32_t) 1 << bits));
1057 assert(sha_ptr == sha_start+count);
1058 assert(name_ptr == name_start+count);
1060 result = PyLong_FromUnsignedLong(count);
1065 for (i = 0; i < num_i; i++)
1066 if (idx_buf_init[i])
1067 PyBuffer_Release(&(idx_buf[i]));
1073 for (i = 0; i < num_i; i++)
1077 PyBuffer_Release(&fmap);
1081 #define FAN_ENTRIES 256
1083 static PyObject *write_idx(PyObject *self, PyObject *args)
1085 char *filename = NULL;
1086 PyObject *py_total, *idx = NULL;
1088 unsigned int total = 0;
1091 uint32_t *fan_ptr, *crc_ptr, *ofs_ptr;
1092 uint64_t *ofs64_ptr;
1093 struct sha *sha_ptr;
1096 if (!PyArg_ParseTuple(args, cstr_argf wbuf_argf "OO",
1097 &filename, &fmap, &idx, &py_total))
1100 PyObject *result = NULL;
1102 if (!bup_uint_from_py(&total, py_total, "total"))
1103 goto clean_and_return;
1105 if (PyList_Size (idx) != FAN_ENTRIES) // Check for list of the right length.
1107 result = PyErr_Format (PyExc_TypeError, "idx must contain %d entries",
1109 goto clean_and_return;
1112 const char idx_header[] = "\377tOc\0\0\0\002";
1113 memcpy (fmap.buf, idx_header, sizeof(idx_header) - 1);
1115 fan_ptr = (uint32_t *)&((unsigned char *)fmap.buf)[sizeof(idx_header) - 1];
1116 sha_ptr = (struct sha *)&fan_ptr[FAN_ENTRIES];
1117 crc_ptr = (uint32_t *)&sha_ptr[total];
1118 ofs_ptr = (uint32_t *)&crc_ptr[total];
1119 ofs64_ptr = (uint64_t *)&ofs_ptr[total];
1122 uint32_t ofs64_count = 0;
1123 for (i = 0; i < FAN_ENTRIES; ++i)
1125 part = PyList_GET_ITEM(idx, i);
1128 if (!INTEGRAL_ASSIGNMENT_FITS(&plen, PyList_GET_SIZE(part))
1129 || UINT32_MAX - count < plen) {
1130 PyErr_Format(PyExc_OverflowError, "too many objects in index part");
1131 goto clean_and_return;
1134 *fan_ptr++ = htonl(count);
1136 for (j = 0; j < plen; ++j)
1138 unsigned char *sha = NULL;
1139 Py_ssize_t sha_len = 0;
1140 PyObject *crc_py, *ofs_py;
1142 unsigned PY_LONG_LONG ofs_ull;
1144 if (!PyArg_ParseTuple(PyList_GET_ITEM(part, j), rbuf_argf "OO",
1145 &sha, &sha_len, &crc_py, &ofs_py))
1146 goto clean_and_return;
1147 if(!bup_uint_from_py(&crc, crc_py, "crc"))
1148 goto clean_and_return;
1149 if(!bup_ullong_from_py(&ofs_ull, ofs_py, "ofs"))
1150 goto clean_and_return;
1151 assert(crc <= UINT32_MAX);
1152 assert(ofs_ull <= UINT64_MAX);
1154 if (sha_len != sizeof(struct sha))
1155 goto clean_and_return;
1156 memcpy(sha_ptr++, sha, sizeof(struct sha));
1157 *crc_ptr++ = htonl(crc);
1158 if (ofs > 0x7fffffff)
1160 *ofs64_ptr++ = htonll(ofs);
1161 ofs = 0x80000000 | ofs64_count++;
1163 *ofs_ptr++ = htonl((uint32_t)ofs);
1167 int rc = msync(fmap.buf, fmap.len, MS_ASYNC);
1170 result = PyErr_SetFromErrnoWithFilename(PyExc_IOError, filename);
1171 goto clean_and_return;
1174 result = PyLong_FromUnsignedLong(count);
1177 PyBuffer_Release(&fmap);
1182 // I would have made this a lower-level function that just fills in a buffer
1183 // with random values, and then written those values from python. But that's
1184 // about 20% slower in my tests, and since we typically generate random
1185 // numbers for benchmarking other parts of bup, any slowness in generating
1186 // random bytes will make our benchmarks inaccurate. Plus nobody wants
1187 // pseudorandom bytes much except for this anyway.
1188 static PyObject *write_random(PyObject *self, PyObject *args)
1190 uint32_t buf[1024/4];
1191 int fd = -1, seed = 0, verbose = 0;
1193 long long len = 0, kbytes = 0, written = 0;
1195 if (!PyArg_ParseTuple(args, "iLii", &fd, &len, &seed, &verbose))
1200 for (kbytes = 0; kbytes < len/1024; kbytes++)
1203 for (i = 0; i < sizeof(buf)/sizeof(buf[0]); i++)
1204 buf[i] = (uint32_t) random();
1205 ret = write(fd, buf, sizeof(buf));
1209 if (ret < (int)sizeof(buf))
1211 if (verbose && kbytes/1024 > 0 && !(kbytes%1024))
1212 fprintf(stderr, "Random: %lld Mbytes\r", kbytes/1024);
1215 // handle non-multiples of 1024
1219 for (i = 0; i < sizeof(buf)/sizeof(buf[0]); i++)
1220 buf[i] = (uint32_t) random();
1221 ret = write(fd, buf, len % 1024);
1227 if (kbytes/1024 > 0)
1228 fprintf(stderr, "Random: %lld Mbytes, done.\n", kbytes/1024);
1229 return Py_BuildValue("L", written);
1233 static PyObject *random_sha(PyObject *self, PyObject *args)
1235 static int seeded = 0;
1236 uint32_t shabuf[20/4];
1241 assert(sizeof(shabuf) == 20);
1242 srandom((unsigned int) time(NULL));
1246 if (!PyArg_ParseTuple(args, ""))
1249 memset(shabuf, 0, sizeof(shabuf));
1250 for (i=0; i < 20/4; i++)
1251 shabuf[i] = (uint32_t) random();
1252 return Py_BuildValue(rbuf_argf, shabuf, 20);
1256 static int _open_noatime(const char *filename, int attrs)
1258 int attrs_noatime, fd;
1261 attrs |= O_NOFOLLOW;
1264 attrs |= O_LARGEFILE;
1266 attrs_noatime = attrs;
1268 attrs_noatime |= O_NOATIME;
1270 fd = open(filename, attrs_noatime);
1271 if (fd < 0 && errno == EPERM)
1273 // older Linux kernels would return EPERM if you used O_NOATIME
1274 // and weren't the file's owner. This pointless restriction was
1275 // relaxed eventually, but we have to handle it anyway.
1276 // (VERY old kernels didn't recognized O_NOATIME, but they would
1277 // just harmlessly ignore it, so this branch won't trigger)
1278 fd = open(filename, attrs);
1284 static PyObject *open_noatime(PyObject *self, PyObject *args)
1286 char *filename = NULL;
1288 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1290 fd = _open_noatime(filename, 0);
1292 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1293 return Py_BuildValue("i", fd);
1297 static PyObject *fadvise_done(PyObject *self, PyObject *args)
1300 long long llofs, lllen = 0;
1301 if (!PyArg_ParseTuple(args, "iLL", &fd, &llofs, &lllen))
1304 if (!INTEGRAL_ASSIGNMENT_FITS(&ofs, llofs))
1305 return PyErr_Format(PyExc_OverflowError,
1306 "fadvise offset overflows off_t");
1307 if (!INTEGRAL_ASSIGNMENT_FITS(&len, lllen))
1308 return PyErr_Format(PyExc_OverflowError,
1309 "fadvise length overflows off_t");
1310 #ifdef POSIX_FADV_DONTNEED
1311 posix_fadvise(fd, ofs, len, POSIX_FADV_DONTNEED);
1313 return Py_BuildValue("");
1317 // Currently the Linux kernel and FUSE disagree over the type for
1318 // FS_IOC_GETFLAGS and FS_IOC_SETFLAGS. The kernel actually uses int,
1319 // but FUSE chose long (matching the declaration in linux/fs.h). So
1320 // if you use int, and then traverse a FUSE filesystem, you may
1321 // corrupt the stack. But if you use long, then you may get invalid
1322 // results on big-endian systems.
1324 // For now, we just use long, and then disable Linux attrs entirely
1325 // (with a warning) in helpers.py on systems that are affected.
1327 #ifdef BUP_HAVE_FILE_ATTRS
1328 static PyObject *bup_get_linux_file_attr(PyObject *self, PyObject *args)
1335 if (!PyArg_ParseTuple(args, cstr_argf, &path))
1338 fd = _open_noatime(path, O_NONBLOCK);
1340 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1342 attr = 0; // Handle int/long mismatch (see above)
1343 rc = ioctl(fd, FS_IOC_GETFLAGS, &attr);
1347 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1350 assert(attr <= UINT_MAX); // Kernel type is actually int
1351 return PyLong_FromUnsignedLong(attr);
1353 #endif /* def BUP_HAVE_FILE_ATTRS */
1357 #ifdef BUP_HAVE_FILE_ATTRS
1358 static PyObject *bup_set_linux_file_attr(PyObject *self, PyObject *args)
1361 unsigned long orig_attr;
1367 if (!PyArg_ParseTuple(args, cstr_argf "O", &path, &py_attr))
1370 if (!bup_uint_from_py(&attr, py_attr, "attr"))
1373 fd = open(path, O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_NOFOLLOW);
1375 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1377 // Restrict attr to modifiable flags acdeijstuADST -- see
1378 // chattr(1) and the e2fsprogs source. Letter to flag mapping is
1379 // in pf.c flags_array[].
1380 attr &= FS_APPEND_FL | FS_COMPR_FL | FS_NODUMP_FL | FS_EXTENT_FL
1381 | FS_IMMUTABLE_FL | FS_JOURNAL_DATA_FL | FS_SECRM_FL | FS_NOTAIL_FL
1382 | FS_UNRM_FL | FS_NOATIME_FL | FS_DIRSYNC_FL | FS_SYNC_FL
1383 | FS_TOPDIR_FL | FS_NOCOW_FL;
1385 // The extents flag can't be removed, so don't (see chattr(1) and chattr.c).
1386 orig_attr = 0; // Handle int/long mismatch (see above)
1387 rc = ioctl(fd, FS_IOC_GETFLAGS, &orig_attr);
1391 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1393 assert(orig_attr <= UINT_MAX); // Kernel type is actually int
1394 attr |= ((unsigned int) orig_attr) & FS_EXTENT_FL;
1396 rc = ioctl(fd, FS_IOC_SETFLAGS, &attr);
1400 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1404 return Py_BuildValue("O", Py_None);
1406 #endif /* def BUP_HAVE_FILE_ATTRS */
1409 #ifndef HAVE_UTIMENSAT
1411 #error "cannot find utimensat or utimes()"
1413 #ifndef HAVE_LUTIMES
1414 #error "cannot find utimensat or lutimes()"
1418 #define ASSIGN_PYLONG_TO_INTEGRAL(dest, pylong, overflow) \
1422 const long long lltmp = PyLong_AsLongLong(pylong); \
1423 if (lltmp == -1 && PyErr_Occurred()) \
1425 if (PyErr_ExceptionMatches(PyExc_OverflowError)) \
1427 const unsigned long long ulltmp = PyLong_AsUnsignedLongLong(pylong); \
1428 if (ulltmp == (unsigned long long) -1 && PyErr_Occurred()) \
1430 if (PyErr_ExceptionMatches(PyExc_OverflowError)) \
1436 if (INTEGRAL_ASSIGNMENT_FITS((dest), ulltmp)) \
1444 if (INTEGRAL_ASSIGNMENT_FITS((dest), lltmp)) \
1453 #ifdef HAVE_UTIMENSAT
1455 static PyObject *bup_utimensat(PyObject *self, PyObject *args)
1460 PyObject *access_py, *modification_py;
1461 struct timespec ts[2];
1463 if (!PyArg_ParseTuple(args, "i" cstr_argf "((Ol)(Ol))i",
1466 &access_py, &(ts[0].tv_nsec),
1467 &modification_py, &(ts[1].tv_nsec),
1472 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(ts[0].tv_sec), access_py, &overflow))
1475 PyErr_SetString(PyExc_ValueError,
1476 "unable to convert access time seconds for utimensat");
1479 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(ts[1].tv_sec), modification_py, &overflow))
1482 PyErr_SetString(PyExc_ValueError,
1483 "unable to convert modification time seconds for utimensat");
1486 rc = utimensat(fd, path, ts, flag);
1488 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1490 return Py_BuildValue("O", Py_None);
1493 #endif /* def HAVE_UTIMENSAT */
1496 #if defined(HAVE_UTIMES) || defined(HAVE_LUTIMES)
1498 static int bup_parse_xutimes_args(char **path,
1499 struct timeval tv[2],
1502 PyObject *access_py, *modification_py;
1503 long long access_us, modification_us; // POSIX guarantees tv_usec is signed.
1505 if (!PyArg_ParseTuple(args, cstr_argf "((OL)(OL))",
1507 &access_py, &access_us,
1508 &modification_py, &modification_us))
1512 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(tv[0].tv_sec), access_py, &overflow))
1515 PyErr_SetString(PyExc_ValueError, "unable to convert access time seconds to timeval");
1518 if (!INTEGRAL_ASSIGNMENT_FITS(&(tv[0].tv_usec), access_us))
1520 PyErr_SetString(PyExc_ValueError, "unable to convert access time nanoseconds to timeval");
1523 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(tv[1].tv_sec), modification_py, &overflow))
1526 PyErr_SetString(PyExc_ValueError, "unable to convert modification time seconds to timeval");
1529 if (!INTEGRAL_ASSIGNMENT_FITS(&(tv[1].tv_usec), modification_us))
1531 PyErr_SetString(PyExc_ValueError, "unable to convert modification time nanoseconds to timeval");
1537 #endif /* defined(HAVE_UTIMES) || defined(HAVE_LUTIMES) */
1541 static PyObject *bup_utimes(PyObject *self, PyObject *args)
1544 struct timeval tv[2];
1545 if (!bup_parse_xutimes_args(&path, tv, args))
1547 int rc = utimes(path, tv);
1549 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1550 return Py_BuildValue("O", Py_None);
1552 #endif /* def HAVE_UTIMES */
1556 static PyObject *bup_lutimes(PyObject *self, PyObject *args)
1559 struct timeval tv[2];
1560 if (!bup_parse_xutimes_args(&path, tv, args))
1562 int rc = lutimes(path, tv);
1564 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1566 return Py_BuildValue("O", Py_None);
1568 #endif /* def HAVE_LUTIMES */
1571 #ifdef HAVE_STAT_ST_ATIM
1572 # define BUP_STAT_ATIME_NS(st) (st)->st_atim.tv_nsec
1573 # define BUP_STAT_MTIME_NS(st) (st)->st_mtim.tv_nsec
1574 # define BUP_STAT_CTIME_NS(st) (st)->st_ctim.tv_nsec
1575 #elif defined HAVE_STAT_ST_ATIMENSEC
1576 # define BUP_STAT_ATIME_NS(st) (st)->st_atimespec.tv_nsec
1577 # define BUP_STAT_MTIME_NS(st) (st)->st_mtimespec.tv_nsec
1578 # define BUP_STAT_CTIME_NS(st) (st)->st_ctimespec.tv_nsec
1580 # define BUP_STAT_ATIME_NS(st) 0
1581 # define BUP_STAT_MTIME_NS(st) 0
1582 # define BUP_STAT_CTIME_NS(st) 0
1586 static PyObject *stat_struct_to_py(const struct stat *st,
1587 const char *filename,
1590 // We can check the known (via POSIX) signed and unsigned types at
1591 // compile time, but not (easily) the unspecified types, so handle
1592 // those via INTEGER_TO_PY(). Assumes ns values will fit in a
1594 return Py_BuildValue("NKNNNNNL(Nl)(Nl)(Nl)",
1595 INTEGER_TO_PY(st->st_mode),
1596 (unsigned PY_LONG_LONG) st->st_ino,
1597 INTEGER_TO_PY(st->st_dev),
1598 INTEGER_TO_PY(st->st_nlink),
1599 INTEGER_TO_PY(st->st_uid),
1600 INTEGER_TO_PY(st->st_gid),
1601 INTEGER_TO_PY(st->st_rdev),
1602 (PY_LONG_LONG) st->st_size,
1603 INTEGER_TO_PY(st->st_atime),
1604 (long) BUP_STAT_ATIME_NS(st),
1605 INTEGER_TO_PY(st->st_mtime),
1606 (long) BUP_STAT_MTIME_NS(st),
1607 INTEGER_TO_PY(st->st_ctime),
1608 (long) BUP_STAT_CTIME_NS(st));
1612 static PyObject *bup_stat(PyObject *self, PyObject *args)
1617 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1621 rc = stat(filename, &st);
1623 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1624 return stat_struct_to_py(&st, filename, 0);
1628 static PyObject *bup_lstat(PyObject *self, PyObject *args)
1633 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1637 rc = lstat(filename, &st);
1639 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1640 return stat_struct_to_py(&st, filename, 0);
1644 static PyObject *bup_fstat(PyObject *self, PyObject *args)
1648 if (!PyArg_ParseTuple(args, "i", &fd))
1652 rc = fstat(fd, &st);
1654 return PyErr_SetFromErrno(PyExc_OSError);
1655 return stat_struct_to_py(&st, NULL, fd);
1659 #ifdef HAVE_TM_TM_GMTOFF
1660 static PyObject *bup_localtime(PyObject *self, PyObject *args)
1664 if (!PyArg_ParseTuple(args, "L", &lltime))
1666 if (!INTEGRAL_ASSIGNMENT_FITS(&ttime, lltime))
1667 return PyErr_Format(PyExc_OverflowError, "time value too large");
1671 if(localtime_r(&ttime, &tm) == NULL)
1672 return PyErr_SetFromErrno(PyExc_OSError);
1674 // Match the Python struct_time values.
1675 return Py_BuildValue("[i,i,i,i,i,i,i,i,i,i,s]",
1676 1900 + tm.tm_year, tm.tm_mon + 1, tm.tm_mday,
1677 tm.tm_hour, tm.tm_min, tm.tm_sec,
1678 tm.tm_wday, tm.tm_yday + 1,
1679 tm.tm_isdst, tm.tm_gmtoff, tm.tm_zone);
1681 #endif /* def HAVE_TM_TM_GMTOFF */
1684 #ifdef BUP_MINCORE_BUF_TYPE
1685 static PyObject *bup_mincore(PyObject *self, PyObject *args)
1687 Py_buffer src, dest;
1688 PyObject *py_src_n, *py_src_off, *py_dest_off;
1690 if (!PyArg_ParseTuple(args, cstr_argf "*OOw*O",
1691 &src, &py_src_n, &py_src_off,
1692 &dest, &py_dest_off))
1695 PyObject *result = NULL;
1697 unsigned long long src_n, src_off, dest_off;
1698 if (!(bup_ullong_from_py(&src_n, py_src_n, "src_n")
1699 && bup_ullong_from_py(&src_off, py_src_off, "src_off")
1700 && bup_ullong_from_py(&dest_off, py_dest_off, "dest_off")))
1701 goto clean_and_return;
1703 unsigned long long src_region_end;
1704 if (!uadd(&src_region_end, src_off, src_n)) {
1705 result = PyErr_Format(PyExc_OverflowError, "(src_off + src_n) too large");
1706 goto clean_and_return;
1708 assert(src.len >= 0);
1709 if (src_region_end > (unsigned long long) src.len) {
1710 result = PyErr_Format(PyExc_OverflowError, "region runs off end of src");
1711 goto clean_and_return;
1714 unsigned long long dest_size;
1715 if (!INTEGRAL_ASSIGNMENT_FITS(&dest_size, dest.len)) {
1716 result = PyErr_Format(PyExc_OverflowError, "invalid dest size");
1717 goto clean_and_return;
1719 if (dest_off > dest_size) {
1720 result = PyErr_Format(PyExc_OverflowError, "region runs off end of dest");
1721 goto clean_and_return;
1725 if (!INTEGRAL_ASSIGNMENT_FITS(&length, src_n)) {
1726 result = PyErr_Format(PyExc_OverflowError, "src_n overflows size_t");
1727 goto clean_and_return;
1729 int rc = mincore((void *)(src.buf + src_off), length,
1730 (BUP_MINCORE_BUF_TYPE *) (dest.buf + dest_off));
1732 result = PyErr_SetFromErrno(PyExc_OSError);
1733 goto clean_and_return;
1735 result = Py_BuildValue("O", Py_None);
1738 PyBuffer_Release(&src);
1739 PyBuffer_Release(&dest);
1742 #endif /* def BUP_MINCORE_BUF_TYPE */
1745 static PyObject *tuple_from_cstrs(char **cstrs)
1747 // Assumes list is null terminated
1749 while(cstrs[n] != NULL)
1753 if (!INTEGRAL_ASSIGNMENT_FITS(&sn, n))
1754 return PyErr_Format(PyExc_OverflowError, "string array too large");
1756 PyObject *result = PyTuple_New(sn);
1758 for (i = 0; i < sn; i++)
1760 PyObject *gname = Py_BuildValue(cstr_argf, cstrs[i]);
1766 PyTuple_SET_ITEM(result, i, gname);
1771 static PyObject *appropriate_errno_ex(void)
1775 return PyErr_NoMemory();
1779 // In 3.3 IOError was merged into OSError.
1780 return PyErr_SetFromErrno(PyExc_IOError);
1782 return PyErr_SetFromErrno(PyExc_OSError);
1787 static PyObject *pwd_struct_to_py(const struct passwd *pwd)
1789 // We can check the known (via POSIX) signed and unsigned types at
1790 // compile time, but not (easily) the unspecified types, so handle
1791 // those via INTEGER_TO_PY().
1794 return Py_BuildValue(cstr_argf cstr_argf "OO"
1795 cstr_argf cstr_argf cstr_argf,
1798 INTEGER_TO_PY(pwd->pw_uid),
1799 INTEGER_TO_PY(pwd->pw_gid),
1805 static PyObject *bup_getpwuid(PyObject *self, PyObject *args)
1807 unsigned long long py_uid;
1808 if (!PyArg_ParseTuple(args, "K", &py_uid))
1811 if (!INTEGRAL_ASSIGNMENT_FITS(&uid, py_uid))
1812 return PyErr_Format(PyExc_OverflowError, "uid too large for uid_t");
1815 struct passwd *pwd = getpwuid(uid);
1817 return appropriate_errno_ex();
1818 return pwd_struct_to_py(pwd);
1821 static PyObject *bup_getpwnam(PyObject *self, PyObject *args)
1824 if (!PyArg_ParseTuple(args, "S", &py_name))
1827 char *name = PyBytes_AS_STRING(py_name);
1829 struct passwd *pwd = getpwnam(name);
1831 return appropriate_errno_ex();
1832 return pwd_struct_to_py(pwd);
1835 static PyObject *grp_struct_to_py(const struct group *grp)
1837 // We can check the known (via POSIX) signed and unsigned types at
1838 // compile time, but not (easily) the unspecified types, so handle
1839 // those via INTEGER_TO_PY().
1843 PyObject *members = tuple_from_cstrs(grp->gr_mem);
1844 if (members == NULL)
1846 return Py_BuildValue(cstr_argf cstr_argf "OO",
1849 INTEGER_TO_PY(grp->gr_gid),
1853 static PyObject *bup_getgrgid(PyObject *self, PyObject *args)
1855 unsigned long long py_gid;
1856 if (!PyArg_ParseTuple(args, "K", &py_gid))
1859 if (!INTEGRAL_ASSIGNMENT_FITS(&gid, py_gid))
1860 return PyErr_Format(PyExc_OverflowError, "gid too large for gid_t");
1863 struct group *grp = getgrgid(gid);
1865 return appropriate_errno_ex();
1866 return grp_struct_to_py(grp);
1869 static PyObject *bup_getgrnam(PyObject *self, PyObject *args)
1872 if (!PyArg_ParseTuple(args, "S", &py_name))
1875 char *name = PyBytes_AS_STRING(py_name);
1877 struct group *grp = getgrnam(name);
1879 return appropriate_errno_ex();
1880 return grp_struct_to_py(grp);
1884 static PyObject *bup_gethostname(PyObject *mod, PyObject *ignore)
1886 #ifdef HOST_NAME_MAX
1887 char buf[HOST_NAME_MAX + 1] = {};
1889 /* 'SUSv2 guarantees that "Host names are limited to 255 bytes".' */
1893 if (gethostname(buf, sizeof(buf) - 1))
1894 return PyErr_SetFromErrno(PyExc_IOError);
1895 return PyBytes_FromString(buf);
1899 #ifdef BUP_HAVE_READLINE
1901 static char *cstr_from_bytes(PyObject *bytes)
1905 int rc = PyBytes_AsStringAndSize(bytes, &buf, &length);
1908 char *result = checked_malloc(length, sizeof(char));
1911 memcpy(result, buf, length);
1915 static char **cstrs_from_seq(PyObject *seq)
1917 char **result = NULL;
1918 seq = PySequence_Fast(seq, "Cannot convert sequence items to C strings");
1922 const Py_ssize_t len = PySequence_Fast_GET_SIZE(seq);
1923 if (len > PY_SSIZE_T_MAX - 1) {
1924 PyErr_Format(PyExc_OverflowError,
1925 "Sequence length %zd too large for conversion to C array",
1929 result = checked_malloc(len + 1, sizeof(char *));
1933 for (i = 0; i < len; i++)
1935 PyObject *item = PySequence_Fast_GET_ITEM(seq, i);
1937 goto abandon_result;
1938 result[i] = cstr_from_bytes(item);
1941 goto abandon_result;
1959 static char* our_word_break_chars = NULL;
1962 bup_set_completer_word_break_characters(PyObject *self, PyObject *args)
1965 if (!PyArg_ParseTuple(args, cstr_argf, &bytes))
1967 char *prev = our_word_break_chars;
1968 char *next = strdup(bytes);
1970 return PyErr_NoMemory();
1971 our_word_break_chars = next;
1972 rl_completer_word_break_characters = next;
1979 bup_get_completer_word_break_characters(PyObject *self, PyObject *args)
1981 if (!PyArg_ParseTuple(args, ""))
1983 return PyBytes_FromString(rl_completer_word_break_characters);
1986 static PyObject *bup_get_line_buffer(PyObject *self, PyObject *args)
1988 if (!PyArg_ParseTuple(args, ""))
1990 return PyBytes_FromString(rl_line_buffer);
1994 bup_parse_and_bind(PyObject *self, PyObject *args)
1997 if (!PyArg_ParseTuple(args, cstr_argf ":parse_and_bind", &bytes))
1999 char *tmp = strdup(bytes); // Because it may modify the arg
2001 return PyErr_NoMemory();
2002 int rc = rl_parse_and_bind(tmp);
2005 return PyErr_Format(PyExc_OSError,
2006 "system rl_parse_and_bind failed (%d)", rc);
2011 static PyObject *py_on_attempted_completion;
2012 static char **prev_completions;
2014 static char **on_attempted_completion(const char *text, int start, int end)
2016 if (!py_on_attempted_completion)
2019 char **result = NULL;
2020 PyObject *py_result = PyObject_CallFunction(py_on_attempted_completion,
2025 if (py_result != Py_None) {
2026 result = cstrs_from_seq(py_result);
2027 free(prev_completions);
2028 prev_completions = result;
2030 Py_DECREF(py_result);
2035 bup_set_attempted_completion_function(PyObject *self, PyObject *args)
2037 PyObject *completer;
2038 if (!PyArg_ParseTuple(args, "O", &completer))
2041 PyObject *prev = py_on_attempted_completion;
2042 if (completer == Py_None)
2044 py_on_attempted_completion = NULL;
2045 rl_attempted_completion_function = NULL;
2047 py_on_attempted_completion = completer;
2048 rl_attempted_completion_function = on_attempted_completion;
2049 Py_INCREF(completer);
2056 static PyObject *py_on_completion_entry;
2058 static char *on_completion_entry(const char *text, int state)
2060 if (!py_on_completion_entry)
2063 PyObject *py_result = PyObject_CallFunction(py_on_completion_entry,
2064 cstr_argf "i", text, state);
2067 char *result = (py_result == Py_None) ? NULL : cstr_from_bytes(py_result);
2068 Py_DECREF(py_result);
2073 bup_set_completion_entry_function(PyObject *self, PyObject *args)
2075 PyObject *completer;
2076 if (!PyArg_ParseTuple(args, "O", &completer))
2079 PyObject *prev = py_on_completion_entry;
2080 if (completer == Py_None) {
2081 py_on_completion_entry = NULL;
2082 rl_completion_entry_function = NULL;
2084 py_on_completion_entry = completer;
2085 rl_completion_entry_function = on_completion_entry;
2086 Py_INCREF(completer);
2093 bup_readline(PyObject *self, PyObject *args)
2096 if (!PyArg_ParseTuple(args, cstr_argf, &prompt))
2098 char *line = readline(prompt);
2100 return PyErr_Format(PyExc_EOFError, "readline EOF");
2101 PyObject *result = PyBytes_FromString(line);
2106 #endif // defined BUP_HAVE_READLINE
2108 #if defined(HAVE_SYS_ACL_H) && \
2109 defined(HAVE_ACL_LIBACL_H) && \
2110 defined(HAVE_ACL_EXTENDED_FILE) && \
2111 defined(HAVE_ACL_GET_FILE) && \
2112 defined(HAVE_ACL_TO_ANY_TEXT) && \
2113 defined(HAVE_ACL_FROM_TEXT) && \
2114 defined(HAVE_ACL_SET_FILE)
2115 #define ACL_SUPPORT 1
2116 #include <sys/acl.h>
2117 #include <acl/libacl.h>
2121 // -1 for errors, with python exception set
2122 // -2 for ignored errors (not supported)
2123 static int bup_read_acl_to_text(const char *name, acl_type_t type,
2124 char **txt, char **num)
2128 acl = acl_get_file(name, type);
2130 if (errno == EOPNOTSUPP || errno == ENOSYS)
2132 PyErr_SetFromErrno(PyExc_IOError);
2137 *txt = acl_to_any_text(acl, "", '\n', TEXT_ABBREVIATE);
2139 *num = acl_to_any_text(acl, "", '\n', TEXT_ABBREVIATE | TEXT_NUMERIC_IDS);
2144 if (errno == ENOMEM)
2147 PyErr_SetFromErrno(PyExc_IOError);
2150 acl_free((acl_t)*txt);
2152 acl_free((acl_t)*num);
2157 static PyObject *bup_read_acl(PyObject *self, PyObject *args)
2161 PyObject *ret = NULL;
2162 char *acl_txt = NULL, *acl_num = NULL;
2164 if (!PyArg_ParseTuple(args, cstr_argf "i", &name, &isdir))
2167 if (!acl_extended_file(name))
2170 rv = bup_read_acl_to_text(name, ACL_TYPE_ACCESS, &acl_txt, &acl_num);
2175 char *def_txt = NULL, *def_num = NULL;
2177 rv = bup_read_acl_to_text(name, ACL_TYPE_DEFAULT, &def_txt, &def_num);
2181 ret = Py_BuildValue("[" cstr_argf cstr_argf cstr_argf cstr_argf "]",
2182 acl_txt, acl_num, def_txt, def_num);
2185 acl_free((acl_t)def_txt);
2187 acl_free((acl_t)def_num);
2189 ret = Py_BuildValue("[" cstr_argf cstr_argf "]",
2195 acl_free((acl_t)acl_txt);
2197 acl_free((acl_t)acl_num);
2203 static int bup_apply_acl_string(const char *name, const char *s)
2205 acl_t acl = acl_from_text(s);
2209 PyErr_SetFromErrno(PyExc_IOError);
2213 if (acl_set_file(name, ACL_TYPE_ACCESS, acl)) {
2214 PyErr_SetFromErrno(PyExc_IOError);
2223 static PyObject *bup_apply_acl(PyObject *self, PyObject *args)
2225 char *name, *acl, *def = NULL;
2227 if (!PyArg_ParseTuple(args, cstr_argf cstr_argf "|" cstr_argf, &name, &acl, &def))
2230 if (bup_apply_acl_string(name, acl))
2233 if (def && bup_apply_acl_string(name, def))
2240 static PyMethodDef helper_methods[] = {
2241 { "write_sparsely", bup_write_sparsely, METH_VARARGS,
2242 "Write buf excepting zeros at the end. Return trailing zero count." },
2243 { "selftest", selftest, METH_VARARGS,
2244 "Check that the rolling checksum rolls correctly (for unit tests)." },
2245 { "blobbits", blobbits, METH_VARARGS,
2246 "Return the number of bits in the rolling checksum." },
2247 { "splitbuf", splitbuf, METH_VARARGS,
2248 "Split a list of strings based on a rolling checksum." },
2249 { "bitmatch", bitmatch, METH_VARARGS,
2250 "Count the number of matching prefix bits between two strings." },
2251 { "firstword", firstword, METH_VARARGS,
2252 "Return an int corresponding to the first 32 bits of buf." },
2253 { "bloom_contains", bloom_contains, METH_VARARGS,
2254 "Check if a bloom filter of 2^nbits bytes contains an object" },
2255 { "bloom_add", bloom_add, METH_VARARGS,
2256 "Add an object to a bloom filter of 2^nbits bytes" },
2257 { "extract_bits", extract_bits, METH_VARARGS,
2258 "Take the first 'nbits' bits from 'buf' and return them as an int." },
2259 { "merge_into", merge_into, METH_VARARGS,
2260 "Merges a bunch of idx and midx files into a single midx." },
2261 { "write_idx", write_idx, METH_VARARGS,
2262 "Write a PackIdxV2 file from an idx list of lists of tuples" },
2263 { "write_random", write_random, METH_VARARGS,
2264 "Write random bytes to the given file descriptor" },
2265 { "random_sha", random_sha, METH_VARARGS,
2266 "Return a random 20-byte string" },
2267 { "open_noatime", open_noatime, METH_VARARGS,
2268 "open() the given filename for read with O_NOATIME if possible" },
2269 { "fadvise_done", fadvise_done, METH_VARARGS,
2270 "Inform the kernel that we're finished with earlier parts of a file" },
2271 #ifdef BUP_HAVE_FILE_ATTRS
2272 { "get_linux_file_attr", bup_get_linux_file_attr, METH_VARARGS,
2273 "Return the Linux attributes for the given file." },
2275 #ifdef BUP_HAVE_FILE_ATTRS
2276 { "set_linux_file_attr", bup_set_linux_file_attr, METH_VARARGS,
2277 "Set the Linux attributes for the given file." },
2279 #ifdef HAVE_UTIMENSAT
2280 { "bup_utimensat", bup_utimensat, METH_VARARGS,
2281 "Change path timestamps with nanosecond precision (POSIX)." },
2284 { "bup_utimes", bup_utimes, METH_VARARGS,
2285 "Change path timestamps with microsecond precision." },
2288 { "bup_lutimes", bup_lutimes, METH_VARARGS,
2289 "Change path timestamps with microsecond precision;"
2290 " don't follow symlinks." },
2292 { "stat", bup_stat, METH_VARARGS,
2293 "Extended version of stat." },
2294 { "lstat", bup_lstat, METH_VARARGS,
2295 "Extended version of lstat." },
2296 { "fstat", bup_fstat, METH_VARARGS,
2297 "Extended version of fstat." },
2298 #ifdef HAVE_TM_TM_GMTOFF
2299 { "localtime", bup_localtime, METH_VARARGS,
2300 "Return struct_time elements plus the timezone offset and name." },
2302 { "bytescmp", bup_bytescmp, METH_VARARGS,
2303 "Return a negative value if x < y, zero if equal, positive otherwise."},
2304 { "cat_bytes", bup_cat_bytes, METH_VARARGS,
2305 "For (x_bytes, x_ofs, x_n, y_bytes, y_ofs, y_n) arguments, return their concatenation."},
2306 #ifdef BUP_MINCORE_BUF_TYPE
2307 { "mincore", bup_mincore, METH_VARARGS,
2308 "For mincore(src, src_n, src_off, dest, dest_off)"
2309 " call the system mincore(src + src_off, src_n, &dest[dest_off])." },
2311 { "getpwuid", bup_getpwuid, METH_VARARGS,
2312 "Return the password database entry for the given numeric user id,"
2313 " as a tuple with all C strings as bytes(), or None if the user does"
2315 { "getpwnam", bup_getpwnam, METH_VARARGS,
2316 "Return the password database entry for the given user name,"
2317 " as a tuple with all C strings as bytes(), or None if the user does"
2319 { "getgrgid", bup_getgrgid, METH_VARARGS,
2320 "Return the group database entry for the given numeric group id,"
2321 " as a tuple with all C strings as bytes(), or None if the group does"
2323 { "getgrnam", bup_getgrnam, METH_VARARGS,
2324 "Return the group database entry for the given group name,"
2325 " as a tuple with all C strings as bytes(), or None if the group does"
2327 { "gethostname", bup_gethostname, METH_NOARGS,
2328 "Return the current hostname (as bytes)" },
2329 #ifdef BUP_HAVE_READLINE
2330 { "set_completion_entry_function", bup_set_completion_entry_function, METH_VARARGS,
2331 "Set rl_completion_entry_function. Called as f(text, state)." },
2332 { "set_attempted_completion_function", bup_set_attempted_completion_function, METH_VARARGS,
2333 "Set rl_attempted_completion_function. Called as f(text, start, end)." },
2334 { "parse_and_bind", bup_parse_and_bind, METH_VARARGS,
2335 "Call rl_parse_and_bind." },
2336 { "get_line_buffer", bup_get_line_buffer, METH_NOARGS,
2337 "Return rl_line_buffer." },
2338 { "get_completer_word_break_characters", bup_get_completer_word_break_characters, METH_NOARGS,
2339 "Return rl_completer_word_break_characters." },
2340 { "set_completer_word_break_characters", bup_set_completer_word_break_characters, METH_VARARGS,
2341 "Set rl_completer_word_break_characters." },
2342 { "readline", bup_readline, METH_VARARGS,
2343 "Call readline(prompt)." },
2344 #endif // defined BUP_HAVE_READLINE
2346 { "read_acl", bup_read_acl, METH_VARARGS,
2347 "read_acl(name, isdir)\n\n"
2348 "Read ACLs for the given file/dirname and return the correctly encoded"
2349 " list [txt, num, def_tx, def_num] (the def_* being empty bytestrings"
2350 " unless the second argument 'isdir' is True)." },
2351 { "apply_acl", bup_apply_acl, METH_VARARGS,
2352 "apply_acl(name, acl, def=None)\n\n"
2353 "Given a file/dirname (bytes) and the ACLs to restore, do that." },
2354 #endif /* HAVE_ACLS */
2355 { NULL, NULL, 0, NULL }, // sentinel
2358 static void test_integral_assignment_fits(void)
2360 assert(sizeof(signed short) == sizeof(unsigned short));
2361 assert(sizeof(signed short) < sizeof(signed long long));
2362 assert(sizeof(signed short) < sizeof(unsigned long long));
2363 assert(sizeof(unsigned short) < sizeof(signed long long));
2364 assert(sizeof(unsigned short) < sizeof(unsigned long long));
2365 assert(sizeof(Py_ssize_t) <= sizeof(size_t));
2367 signed short ss, ssmin = SHRT_MIN, ssmax = SHRT_MAX;
2368 unsigned short us, usmax = USHRT_MAX;
2369 signed long long sllmin = LLONG_MIN, sllmax = LLONG_MAX;
2370 unsigned long long ullmax = ULLONG_MAX;
2372 assert(INTEGRAL_ASSIGNMENT_FITS(&ss, ssmax));
2373 assert(INTEGRAL_ASSIGNMENT_FITS(&ss, ssmin));
2374 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, usmax));
2375 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, sllmin));
2376 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, sllmax));
2377 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, ullmax));
2379 assert(INTEGRAL_ASSIGNMENT_FITS(&us, usmax));
2380 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, ssmin));
2381 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, sllmin));
2382 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, sllmax));
2383 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, ullmax));
2387 static int setup_module(PyObject *m)
2389 // FIXME: migrate these tests to configure, or at least don't
2390 // possibly crash the whole application. Check against the type
2391 // we're going to use when passing to python. Other stat types
2392 // are tested at runtime.
2393 assert(sizeof(ino_t) <= sizeof(unsigned PY_LONG_LONG));
2394 assert(sizeof(off_t) <= sizeof(PY_LONG_LONG));
2395 assert(sizeof(blksize_t) <= sizeof(PY_LONG_LONG));
2396 assert(sizeof(blkcnt_t) <= sizeof(PY_LONG_LONG));
2397 // Just be sure (relevant when passing timestamps back to Python above).
2398 assert(sizeof(PY_LONG_LONG) <= sizeof(long long));
2399 assert(sizeof(unsigned PY_LONG_LONG) <= sizeof(unsigned long long));
2401 test_integral_assignment_fits();
2403 // Originally required by append_sparse_region()
2406 if (!INTEGRAL_ASSIGNMENT_FITS(&probe, INT_MAX))
2408 fprintf(stderr, "off_t can't hold INT_MAX; please report.\n");
2416 value = INTEGER_TO_PY(INT_MAX);
2417 PyObject_SetAttrString(m, "INT_MAX", value);
2419 value = INTEGER_TO_PY(UINT_MAX);
2420 PyObject_SetAttrString(m, "UINT_MAX", value);
2423 #ifdef HAVE_UTIMENSAT
2426 value = INTEGER_TO_PY(AT_FDCWD);
2427 PyObject_SetAttrString(m, "AT_FDCWD", value);
2429 value = INTEGER_TO_PY(AT_SYMLINK_NOFOLLOW);
2430 PyObject_SetAttrString(m, "AT_SYMLINK_NOFOLLOW", value);
2432 value = INTEGER_TO_PY(UTIME_NOW);
2433 PyObject_SetAttrString(m, "UTIME_NOW", value);
2437 #ifdef BUP_HAVE_MINCORE_INCORE
2440 value = INTEGER_TO_PY(MINCORE_INCORE);
2441 PyObject_SetAttrString(m, "MINCORE_INCORE", value);
2446 e = getenv("BUP_FORCE_TTY");
2447 get_state(m)->istty2 = isatty(2) || (atoi(e ? e : "0") & 2);
2453 #if PY_MAJOR_VERSION < 3
2455 PyMODINIT_FUNC init_helpers(void)
2457 PyObject *m = Py_InitModule("_helpers", helper_methods);
2461 if (!setup_module(m))
2468 # else // PY_MAJOR_VERSION >= 3
2470 static struct PyModuleDef helpers_def = {
2471 PyModuleDef_HEAD_INIT,
2477 NULL, // helpers_traverse,
2478 NULL, // helpers_clear,
2482 PyMODINIT_FUNC PyInit__helpers(void)
2484 PyObject *module = PyModule_Create(&helpers_def);
2487 if (!setup_module(module))
2495 #endif // PY_MAJOR_VERSION >= 3