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
73 #ifndef BUP_USE_PYTHON_UTIME // just for Python 2 now
75 * Check for incomplete UTIMENSAT support (NetBSD 6), and if so,
76 * pretend we don't have it.
78 #if !defined(AT_FDCWD) || !defined(AT_SYMLINK_NOFOLLOW)
81 #endif // defined BUP_USE_PYTHON_UTIME
84 // Of course, this assumes it's a bitfield value.
89 typedef unsigned char byte;
96 // cstr_argf: for byte vectors without null characters (e.g. paths)
97 // rbuf_argf: for read-only byte vectors
98 // wbuf_argf: for mutable byte vectors
100 #if PY_MAJOR_VERSION < 3
101 static state_t state;
102 # define get_state(x) (&state)
103 # define cstr_argf "s"
104 # define rbuf_argf "s#"
105 # define wbuf_argf "s*"
107 # define get_state(x) ((state_t *) PyModule_GetState(x))
108 # define cstr_argf "y"
109 # define rbuf_argf "y#"
110 # define wbuf_argf "y*"
111 #endif // PY_MAJOR_VERSION >= 3
114 static void *checked_calloc(size_t n, size_t size)
116 void *result = calloc(n, size);
122 #ifndef BUP_HAVE_BUILTIN_MUL_OVERFLOW
124 #define checked_malloc checked_calloc
126 #else // defined BUP_HAVE_BUILTIN_MUL_OVERFLOW
128 static void *checked_malloc(size_t n, size_t size)
131 if (__builtin_mul_overflow(n, size, &total))
133 PyErr_Format(PyExc_OverflowError,
134 "request to allocate %zu items of size %zu is too large",
138 void *result = malloc(total);
140 return PyErr_NoMemory();
144 #endif // defined BUP_HAVE_BUILTIN_MUL_OVERFLOW
148 // This function should technically be macro'd out if it's going to be used
149 // more than ocasionally. As of this writing, it'll actually never be called
150 // in real world bup scenarios (because our packs are < MAX_INT bytes).
151 static uint64_t htonll(uint64_t value)
153 static const int endian_test = 42;
155 if (*(char *)&endian_test == endian_test) // LSB-MSB
156 return ((uint64_t)htonl(value & 0xFFFFFFFF) << 32) | htonl(value >> 32);
157 return value; // already in network byte order MSB-LSB
162 // Disabling sign-compare here should be fine since we're explicitly
163 // checking for a sign mismatch, i.e. if the signs don't match, then
164 // it doesn't matter what the value comparison says.
165 // FIXME: ... so should we reverse the order?
166 #define INTEGRAL_ASSIGNMENT_FITS(dest, src) \
168 _Pragma("GCC diagnostic push"); \
169 _Pragma("GCC diagnostic ignored \"-Wsign-compare\""); \
170 _Pragma("clang diagnostic push"); \
171 _Pragma("clang diagnostic ignored \"-Wshorten-64-to-32\""); \
173 int result = *(dest) == (src) && (*(dest) < 1) == ((src) < 1); \
174 _Pragma("clang diagnostic pop"); \
175 _Pragma("GCC diagnostic pop"); \
180 #define INTEGER_TO_PY(x) \
182 _Pragma("GCC diagnostic push"); \
183 _Pragma("GCC diagnostic ignored \"-Wtype-limits\""); \
184 _Pragma("clang diagnostic push"); \
185 _Pragma("clang diagnostic ignored \"-Wtautological-compare\""); \
186 PyObject *result = ((x) >= 0) ? PyLong_FromUnsignedLongLong(x) : PyLong_FromLongLong(x); \
187 _Pragma("clang diagnostic pop"); \
188 _Pragma("GCC diagnostic pop"); \
193 #if PY_MAJOR_VERSION < 3
194 static int bup_ulong_from_pyint(unsigned long *x, PyObject *py,
197 const long tmp = PyInt_AsLong(py);
198 if (tmp == -1 && PyErr_Occurred())
200 if (PyErr_ExceptionMatches(PyExc_OverflowError))
201 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned long",
207 PyErr_Format(PyExc_OverflowError,
208 "negative %s cannot be converted to unsigned long", name);
217 static int bup_ulong_from_py(unsigned long *x, PyObject *py, const char *name)
219 #if PY_MAJOR_VERSION < 3
221 return bup_ulong_from_pyint(x, py, name);
224 if (!PyLong_Check(py))
226 PyErr_Format(PyExc_TypeError, "expected integer %s", name);
230 const unsigned long tmp = PyLong_AsUnsignedLong(py);
231 if (PyErr_Occurred())
233 if (PyErr_ExceptionMatches(PyExc_OverflowError))
234 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned long",
243 static int bup_uint_from_py(unsigned int *x, PyObject *py, const char *name)
246 if (!bup_ulong_from_py(&tmp, py, name))
251 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned int", name);
254 *x = (unsigned int) tmp;
258 static int bup_ullong_from_py(unsigned PY_LONG_LONG *x, PyObject *py,
261 #if PY_MAJOR_VERSION < 3
265 if (bup_ulong_from_pyint(&tmp, py, name))
274 if (!PyLong_Check(py))
276 PyErr_Format(PyExc_TypeError, "integer argument expected for %s", name);
280 const unsigned PY_LONG_LONG tmp = PyLong_AsUnsignedLongLong(py);
281 if (tmp == (unsigned long long) -1 && PyErr_Occurred())
283 if (PyErr_ExceptionMatches(PyExc_OverflowError))
284 PyErr_Format(PyExc_OverflowError,
285 "%s too big for unsigned long long", name);
293 static PyObject *bup_bytescmp(PyObject *self, PyObject *args)
295 PyObject *py_s1, *py_s2; // This is really a PyBytes/PyString
296 if (!PyArg_ParseTuple(args, "SS", &py_s1, &py_s2))
299 Py_ssize_t s1_len, s2_len;
300 if (PyBytes_AsStringAndSize(py_s1, &s1, &s1_len) == -1)
302 if (PyBytes_AsStringAndSize(py_s2, &s2, &s2_len) == -1)
304 const Py_ssize_t n = (s1_len < s2_len) ? s1_len : s2_len;
305 const int cmp = memcmp(s1, s2, n);
307 return PyLong_FromLong(cmp);
308 if (s1_len == s2_len)
309 return PyLong_FromLong(0);;
310 return PyLong_FromLong((s1_len < s2_len) ? -1 : 1);
314 static PyObject *bup_cat_bytes(PyObject *self, PyObject *args)
316 unsigned char *bufx = NULL, *bufy = NULL;
317 Py_ssize_t bufx_len, bufx_ofs, bufx_n;
318 Py_ssize_t bufy_len, bufy_ofs, bufy_n;
319 if (!PyArg_ParseTuple(args,
322 &bufx, &bufx_len, &bufx_ofs, &bufx_n,
323 &bufy, &bufy_len, &bufy_ofs, &bufy_n))
326 return PyErr_Format(PyExc_ValueError, "negative x offset");
328 return PyErr_Format(PyExc_ValueError, "negative x extent");
329 if (bufx_ofs > bufx_len)
330 return PyErr_Format(PyExc_ValueError, "x offset greater than length");
331 if (bufx_n > bufx_len - bufx_ofs)
332 return PyErr_Format(PyExc_ValueError, "x extent past end of buffer");
335 return PyErr_Format(PyExc_ValueError, "negative y offset");
337 return PyErr_Format(PyExc_ValueError, "negative y extent");
338 if (bufy_ofs > bufy_len)
339 return PyErr_Format(PyExc_ValueError, "y offset greater than length");
340 if (bufy_n > bufy_len - bufy_ofs)
341 return PyErr_Format(PyExc_ValueError, "y extent past end of buffer");
343 if (bufy_n > PY_SSIZE_T_MAX - bufx_n)
344 return PyErr_Format(PyExc_OverflowError, "result length too long");
346 PyObject *result = PyBytes_FromStringAndSize(NULL, bufx_n + bufy_n);
348 return PyErr_NoMemory();
349 char *buf = PyBytes_AS_STRING(result);
350 memcpy(buf, bufx + bufx_ofs, bufx_n);
351 memcpy(buf + bufx_n, bufy + bufy_ofs, bufy_n);
357 // Probably we should use autoconf or something and set HAVE_PY_GETARGCARGV...
358 #if __WIN32__ || __CYGWIN__ || PY_VERSION_HEX >= 0x03090000
360 // There's no 'ps' on win32 anyway, and Py_GetArgcArgv() isn't available.
361 static void unpythonize_argv(void) { }
363 #else // not __WIN32__
365 // For some reason this isn't declared in Python.h
366 extern void Py_GetArgcArgv(int *argc, char ***argv);
368 static void unpythonize_argv(void)
373 Py_GetArgcArgv(&argc, &argv);
375 for (i = 0; i < argc-1; i++)
377 if (argv[i] + strlen(argv[i]) + 1 != argv[i+1])
379 // The argv block doesn't work the way we expected; it's unsafe
385 arge = argv[argc-1] + strlen(argv[argc-1]) + 1;
387 if (strstr(argv[0], "python") && argv[1] == argv[0] + strlen(argv[0]) + 1)
391 p = strrchr(argv[1], '/');
397 memmove(argv[0], p, len);
398 memset(arge - diff, 0, diff);
399 for (i = 0; i < argc; i++)
400 argv[i] = argv[i+1] ? argv[i+1]-diff : NULL;
405 #endif // not __WIN32__ or __CYGWIN__
408 static int write_all(int fd, const void *buf, const size_t count)
411 while (written < count)
413 const ssize_t rc = write(fd, buf + written, count - written);
422 static int uadd(unsigned long long *dest,
423 const unsigned long long x,
424 const unsigned long long y)
426 const unsigned long long result = x + y;
427 if (result < x || result < y)
434 static PyObject *append_sparse_region(const int fd, unsigned long long n)
439 if (!INTEGRAL_ASSIGNMENT_FITS(&new_off, n))
441 const off_t off = lseek(fd, new_off, SEEK_CUR);
442 if (off == (off_t) -1)
443 return PyErr_SetFromErrno(PyExc_IOError);
450 static PyObject *record_sparse_zeros(unsigned long long *new_pending,
452 unsigned long long prev_pending,
453 const unsigned long long count)
455 // Add count additional sparse zeros to prev_pending and store the
456 // result in new_pending, or if the total won't fit in
457 // new_pending, write some of the zeros to fd sparsely, and store
458 // the remaining sum in new_pending.
459 if (!uadd(new_pending, prev_pending, count))
461 PyObject *err = append_sparse_region(fd, prev_pending);
464 *new_pending = count;
470 static byte* find_not_zero(const byte * const start, const byte * const end)
472 // Return a pointer to first non-zero byte between start and end,
473 // or end if there isn't one.
474 assert(start <= end);
475 const unsigned char *cur = start;
476 while (cur < end && *cur == 0)
482 static byte* find_trailing_zeros(const byte * const start,
483 const byte * const end)
485 // Return a pointer to the start of any trailing run of zeros, or
486 // end if there isn't one.
487 assert(start <= end);
490 const byte * cur = end;
491 while (cur > start && *--cur == 0) {}
495 return (byte *) (cur + 1);
499 static byte *find_non_sparse_end(const byte * const start,
500 const byte * const end,
501 const ptrdiff_t min_len)
503 // Return the first pointer to a min_len sparse block in [start,
504 // end) if there is one, otherwise a pointer to the start of any
505 // trailing run of zeros. If there are no trailing zeros, return
511 // Probe in min_len jumps, searching backward from the jump
512 // destination for a non-zero byte. If such a byte is found, move
513 // just past it and try again.
514 const byte *candidate = start;
515 // End of any run of zeros, starting at candidate, that we've already seen
516 const byte *end_of_known_zeros = candidate;
517 while (end - candidate >= min_len) // Handle all min_len candidate blocks
519 const byte * const probe_end = candidate + min_len;
520 const byte * const trailing_zeros =
521 find_trailing_zeros(end_of_known_zeros, probe_end);
522 if (trailing_zeros == probe_end)
523 end_of_known_zeros = candidate = probe_end;
524 else if (trailing_zeros == end_of_known_zeros)
526 assert(candidate >= start);
527 assert(candidate <= end);
528 assert(*candidate == 0);
529 return (byte *) candidate;
533 candidate = trailing_zeros;
534 end_of_known_zeros = probe_end;
538 if (candidate == end)
541 // No min_len sparse run found, search backward from end
542 const byte * const trailing_zeros = find_trailing_zeros(end_of_known_zeros,
545 if (trailing_zeros == end_of_known_zeros)
547 assert(candidate >= start);
548 assert(candidate < end);
549 assert(*candidate == 0);
550 assert(end - candidate < min_len);
551 return (byte *) candidate;
554 if (trailing_zeros == end)
556 assert(*(end - 1) != 0);
560 assert(end - trailing_zeros < min_len);
561 assert(trailing_zeros >= start);
562 assert(trailing_zeros < end);
563 assert(*trailing_zeros == 0);
564 return (byte *) trailing_zeros;
568 static PyObject *bup_write_sparsely(PyObject *self, PyObject *args)
571 unsigned char *buf = NULL;
573 PyObject *py_min_sparse_len, *py_prev_sparse_len;
574 if (!PyArg_ParseTuple(args, "i" rbuf_argf "OO",
575 &fd, &buf, &sbuf_len,
576 &py_min_sparse_len, &py_prev_sparse_len))
578 ptrdiff_t min_sparse_len;
579 unsigned long long prev_sparse_len, buf_len, ul_min_sparse_len;
580 if (!bup_ullong_from_py(&ul_min_sparse_len, py_min_sparse_len, "min_sparse_len"))
582 if (!INTEGRAL_ASSIGNMENT_FITS(&min_sparse_len, ul_min_sparse_len))
583 return PyErr_Format(PyExc_OverflowError, "min_sparse_len too large");
584 if (!bup_ullong_from_py(&prev_sparse_len, py_prev_sparse_len, "prev_sparse_len"))
587 return PyErr_Format(PyExc_ValueError, "negative bufer length");
588 if (!INTEGRAL_ASSIGNMENT_FITS(&buf_len, sbuf_len))
589 return PyErr_Format(PyExc_OverflowError, "buffer length too large");
591 const byte * block = buf; // Start of pending block
592 const byte * const end = buf + buf_len;
593 unsigned long long zeros = prev_sparse_len;
596 assert(block <= end);
598 return PyLong_FromUnsignedLongLong(zeros);
602 // Look for the end of block, i.e. the next sparse run of
603 // at least min_sparse_len zeros, or the end of the
605 const byte * const probe = find_non_sparse_end(block + 1, end,
607 // Either at end of block, or end of non-sparse; write pending data
608 PyObject *err = append_sparse_region(fd, zeros);
611 int rc = write_all(fd, block, probe - block);
613 return PyErr_SetFromErrno(PyExc_IOError);
615 if (end - probe < min_sparse_len)
618 zeros = min_sparse_len;
619 block = probe + zeros;
623 // Should be in the first loop iteration, a sparse run of
624 // zeros, or nearly at the end of the block (within
626 const byte * const zeros_end = find_not_zero(block, end);
627 PyObject *err = record_sparse_zeros(&zeros, fd,
628 zeros, zeros_end - block);
631 assert(block <= zeros_end);
638 static PyObject *selftest(PyObject *self, PyObject *args)
640 if (!PyArg_ParseTuple(args, ""))
643 return Py_BuildValue("i", !bupsplit_selftest());
647 static PyObject *blobbits(PyObject *self, PyObject *args)
649 if (!PyArg_ParseTuple(args, ""))
651 return Py_BuildValue("i", BUP_BLOBBITS);
655 static PyObject *splitbuf(PyObject *self, PyObject *args)
657 // We stick to buffers in python 2 because they appear to be
658 // substantially smaller than memoryviews, and because
659 // zlib.compress() in python 2 can't accept a memoryview
660 // (cf. hashsplit.py).
661 int out = 0, bits = -1;
662 if (PY_MAJOR_VERSION > 2)
665 if (!PyArg_ParseTuple(args, "y*", &buf))
667 assert(buf.len <= INT_MAX);
668 out = bupsplit_find_ofs(buf.buf, buf.len, &bits);
669 PyBuffer_Release(&buf);
673 unsigned char *buf = NULL;
675 if (!PyArg_ParseTuple(args, "t#", &buf, &len))
677 assert(len <= INT_MAX);
678 out = bupsplit_find_ofs(buf, (int) len, &bits);
680 if (out) assert(bits >= BUP_BLOBBITS);
681 return Py_BuildValue("ii", out, bits);
685 static PyObject *bitmatch(PyObject *self, PyObject *args)
687 unsigned char *buf1 = NULL, *buf2 = NULL;
688 Py_ssize_t len1 = 0, len2 = 0;
692 if (!PyArg_ParseTuple(args, rbuf_argf rbuf_argf, &buf1, &len1, &buf2, &len2))
696 for (byte = 0; byte < len1 && byte < len2; byte++)
698 int b1 = buf1[byte], b2 = buf2[byte];
701 for (bit = 0; bit < 8; bit++)
702 if ( (b1 & (0x80 >> bit)) != (b2 & (0x80 >> bit)) )
708 assert(byte <= (INT_MAX >> 3));
709 return Py_BuildValue("i", byte*8 + bit);
713 static PyObject *firstword(PyObject *self, PyObject *args)
715 unsigned char *buf = NULL;
719 if (!PyArg_ParseTuple(args, rbuf_argf, &buf, &len))
725 v = ntohl(*(uint32_t *)buf);
726 return PyLong_FromUnsignedLong(v);
730 #define BLOOM2_HEADERLEN 16
732 static void to_bloom_address_bitmask4(const unsigned char *buf,
733 const int nbits, uint64_t *v, unsigned char *bitmask)
739 memcpy(&high, buf, 4);
740 mask = (1<<nbits) - 1;
741 raw = (((uint64_t)ntohl(high) << 8) | buf[4]);
742 bit = (raw >> (37-nbits)) & 0x7;
743 *v = (raw >> (40-nbits)) & mask;
747 static void to_bloom_address_bitmask5(const unsigned char *buf,
748 const int nbits, uint32_t *v, unsigned char *bitmask)
754 memcpy(&high, buf, 4);
755 mask = (1<<nbits) - 1;
757 bit = (raw >> (29-nbits)) & 0x7;
758 *v = (raw >> (32-nbits)) & mask;
762 #define BLOOM_SET_BIT(name, address, otype) \
763 static void name(unsigned char *bloom, const unsigned char *buf, const int nbits)\
765 unsigned char bitmask;\
767 address(buf, nbits, &v, &bitmask);\
768 bloom[BLOOM2_HEADERLEN+v] |= bitmask;\
770 BLOOM_SET_BIT(bloom_set_bit4, to_bloom_address_bitmask4, uint64_t)
771 BLOOM_SET_BIT(bloom_set_bit5, to_bloom_address_bitmask5, uint32_t)
774 #define BLOOM_GET_BIT(name, address, otype) \
775 static int name(const unsigned char *bloom, const unsigned char *buf, const int nbits)\
777 unsigned char bitmask;\
779 address(buf, nbits, &v, &bitmask);\
780 return bloom[BLOOM2_HEADERLEN+v] & bitmask;\
782 BLOOM_GET_BIT(bloom_get_bit4, to_bloom_address_bitmask4, uint64_t)
783 BLOOM_GET_BIT(bloom_get_bit5, to_bloom_address_bitmask5, uint32_t)
786 static PyObject *bloom_add(PyObject *self, PyObject *args)
788 Py_buffer bloom, sha;
789 int nbits = 0, k = 0;
790 if (!PyArg_ParseTuple(args, wbuf_argf wbuf_argf "ii",
791 &bloom, &sha, &nbits, &k))
794 PyObject *result = NULL;
796 if (bloom.len < 16+(1<<nbits) || sha.len % 20 != 0)
797 goto clean_and_return;
802 goto clean_and_return;
803 unsigned char *cur = sha.buf;
805 for (end = cur + sha.len; cur < end; cur += 20/k)
806 bloom_set_bit5(bloom.buf, cur, nbits);
811 goto clean_and_return;
812 unsigned char *cur = sha.buf;
813 unsigned char *end = cur + sha.len;
814 for (; cur < end; cur += 20/k)
815 bloom_set_bit4(bloom.buf, cur, nbits);
818 goto clean_and_return;
820 result = Py_BuildValue("n", sha.len / 20);
823 PyBuffer_Release(&bloom);
824 PyBuffer_Release(&sha);
828 static PyObject *bloom_contains(PyObject *self, PyObject *args)
831 unsigned char *sha = NULL;
833 int nbits = 0, k = 0;
834 if (!PyArg_ParseTuple(args, wbuf_argf rbuf_argf "ii",
835 &bloom, &sha, &len, &nbits, &k))
838 PyObject *result = NULL;
841 goto clean_and_return;
846 goto clean_and_return;
849 for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
850 if (!bloom_get_bit5(bloom.buf, sha, nbits))
852 result = Py_BuildValue("Oi", Py_None, steps);
853 goto clean_and_return;
859 goto clean_and_return;
862 for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
863 if (!bloom_get_bit4(bloom.buf, sha, nbits))
865 result = Py_BuildValue("Oi", Py_None, steps);
866 goto clean_and_return;
870 goto clean_and_return;
872 result = Py_BuildValue("ii", 1, k);
875 PyBuffer_Release(&bloom);
880 static uint32_t _extract_bits(unsigned char *buf, int nbits)
884 mask = (1<<nbits) - 1;
885 v = ntohl(*(uint32_t *)buf);
886 v = (v >> (32-nbits)) & mask;
891 static PyObject *extract_bits(PyObject *self, PyObject *args)
893 unsigned char *buf = NULL;
897 if (!PyArg_ParseTuple(args, rbuf_argf "i", &buf, &len, &nbits))
903 return PyLong_FromUnsignedLong(_extract_bits(buf, nbits));
908 unsigned char bytes[20];
911 static inline int _cmp_sha(const struct sha *sha1, const struct sha *sha2)
913 return memcmp(sha1->bytes, sha2->bytes, sizeof(sha1->bytes));
926 static void _fix_idx_order(struct idx **idxs, Py_ssize_t *last_i)
929 Py_ssize_t low, mid, high;
933 if (idxs[*last_i]->cur >= idxs[*last_i]->end)
935 idxs[*last_i] = NULL;
948 mid = (low + high) / 2;
949 c = _cmp_sha(idx->cur, idxs[mid]->cur);
961 memmove(&idxs[mid+1], &idxs[mid], (*last_i-mid)*sizeof(struct idx *));
966 static uint32_t _get_idx_i(struct idx *idx)
968 if (idx->cur_name == NULL)
969 return idx->name_base;
970 return ntohl(*idx->cur_name) + idx->name_base;
973 #define MIDX4_HEADERLEN 12
975 static PyObject *merge_into(PyObject *self, PyObject *args)
977 struct sha *sha_ptr, *sha_start = NULL;
978 uint32_t *table_ptr, *name_ptr, *name_start;
981 uint32_t count, prefix;
986 PyObject *py_total, *ilist = NULL;
987 if (!PyArg_ParseTuple(args, wbuf_argf "iOO",
988 &fmap, &bits, &py_total, &ilist))
991 PyObject *result = NULL;
992 struct idx **idxs = NULL;
993 Py_ssize_t num_i = 0;
994 int *idx_buf_init = NULL;
995 Py_buffer *idx_buf = NULL;
997 if (!bup_uint_from_py(&total, py_total, "total"))
998 goto clean_and_return;
1000 num_i = PyList_Size(ilist);
1002 if (!(idxs = checked_malloc(num_i, sizeof(struct idx *))))
1003 goto clean_and_return;
1004 if (!(idx_buf_init = checked_calloc(num_i, sizeof(int))))
1005 goto clean_and_return;
1006 if (!(idx_buf = checked_malloc(num_i, sizeof(Py_buffer))))
1007 goto clean_and_return;
1009 for (i = 0; i < num_i; i++)
1011 long len, sha_ofs, name_map_ofs;
1012 if (!(idxs[i] = checked_malloc(1, sizeof(struct idx))))
1013 goto clean_and_return;
1014 PyObject *itup = PyList_GetItem(ilist, i);
1015 if (!PyArg_ParseTuple(itup, wbuf_argf "llli",
1016 &(idx_buf[i]), &len, &sha_ofs, &name_map_ofs,
1017 &idxs[i]->name_base))
1019 idx_buf_init[i] = 1;
1020 idxs[i]->map = idx_buf[i].buf;
1021 idxs[i]->bytes = idx_buf[i].len;
1022 idxs[i]->cur = (struct sha *)&idxs[i]->map[sha_ofs];
1023 idxs[i]->end = &idxs[i]->cur[len];
1025 idxs[i]->cur_name = (uint32_t *)&idxs[i]->map[name_map_ofs];
1027 idxs[i]->cur_name = NULL;
1029 table_ptr = (uint32_t *) &((unsigned char *) fmap.buf)[MIDX4_HEADERLEN];
1030 sha_start = sha_ptr = (struct sha *)&table_ptr[1<<bits];
1031 name_start = name_ptr = (uint32_t *)&sha_ptr[total];
1033 Py_ssize_t last_i = num_i - 1;
1039 uint32_t new_prefix;
1040 if (count % 102424 == 0 && get_state(self)->istty2)
1041 fprintf(stderr, "midx: writing %.2f%% (%d/%d)\r",
1042 count*100.0/total, count, total);
1044 new_prefix = _extract_bits((unsigned char *)idx->cur, bits);
1045 while (prefix < new_prefix)
1046 table_ptr[prefix++] = htonl(count);
1047 memcpy(sha_ptr++, idx->cur, sizeof(struct sha));
1048 *name_ptr++ = htonl(_get_idx_i(idx));
1050 if (idx->cur_name != NULL)
1052 _fix_idx_order(idxs, &last_i);
1055 while (prefix < ((uint32_t) 1 << bits))
1056 table_ptr[prefix++] = htonl(count);
1057 assert(count == total);
1058 assert(prefix == ((uint32_t) 1 << bits));
1059 assert(sha_ptr == sha_start+count);
1060 assert(name_ptr == name_start+count);
1062 result = PyLong_FromUnsignedLong(count);
1067 for (i = 0; i < num_i; i++)
1068 if (idx_buf_init[i])
1069 PyBuffer_Release(&(idx_buf[i]));
1075 for (i = 0; i < num_i; i++)
1079 PyBuffer_Release(&fmap);
1083 #define FAN_ENTRIES 256
1085 static PyObject *write_idx(PyObject *self, PyObject *args)
1087 char *filename = NULL;
1088 PyObject *py_total, *idx = NULL;
1090 unsigned int total = 0;
1093 uint32_t *fan_ptr, *crc_ptr, *ofs_ptr;
1094 uint64_t *ofs64_ptr;
1095 struct sha *sha_ptr;
1098 if (!PyArg_ParseTuple(args, cstr_argf wbuf_argf "OO",
1099 &filename, &fmap, &idx, &py_total))
1102 PyObject *result = NULL;
1104 if (!bup_uint_from_py(&total, py_total, "total"))
1105 goto clean_and_return;
1107 if (PyList_Size (idx) != FAN_ENTRIES) // Check for list of the right length.
1109 result = PyErr_Format (PyExc_TypeError, "idx must contain %d entries",
1111 goto clean_and_return;
1114 const char idx_header[] = "\377tOc\0\0\0\002";
1115 memcpy (fmap.buf, idx_header, sizeof(idx_header) - 1);
1117 fan_ptr = (uint32_t *)&((unsigned char *)fmap.buf)[sizeof(idx_header) - 1];
1118 sha_ptr = (struct sha *)&fan_ptr[FAN_ENTRIES];
1119 crc_ptr = (uint32_t *)&sha_ptr[total];
1120 ofs_ptr = (uint32_t *)&crc_ptr[total];
1121 ofs64_ptr = (uint64_t *)&ofs_ptr[total];
1124 uint32_t ofs64_count = 0;
1125 for (i = 0; i < FAN_ENTRIES; ++i)
1127 part = PyList_GET_ITEM(idx, i);
1130 if (!INTEGRAL_ASSIGNMENT_FITS(&plen, PyList_GET_SIZE(part))
1131 || UINT32_MAX - count < plen) {
1132 PyErr_Format(PyExc_OverflowError, "too many objects in index part");
1133 goto clean_and_return;
1136 *fan_ptr++ = htonl(count);
1138 for (j = 0; j < plen; ++j)
1140 unsigned char *sha = NULL;
1141 Py_ssize_t sha_len = 0;
1142 PyObject *crc_py, *ofs_py;
1144 unsigned PY_LONG_LONG ofs_ull;
1146 if (!PyArg_ParseTuple(PyList_GET_ITEM(part, j), rbuf_argf "OO",
1147 &sha, &sha_len, &crc_py, &ofs_py))
1148 goto clean_and_return;
1149 if(!bup_uint_from_py(&crc, crc_py, "crc"))
1150 goto clean_and_return;
1151 if(!bup_ullong_from_py(&ofs_ull, ofs_py, "ofs"))
1152 goto clean_and_return;
1153 assert(crc <= UINT32_MAX);
1154 assert(ofs_ull <= UINT64_MAX);
1156 if (sha_len != sizeof(struct sha))
1157 goto clean_and_return;
1158 memcpy(sha_ptr++, sha, sizeof(struct sha));
1159 *crc_ptr++ = htonl(crc);
1160 if (ofs > 0x7fffffff)
1162 *ofs64_ptr++ = htonll(ofs);
1163 ofs = 0x80000000 | ofs64_count++;
1165 *ofs_ptr++ = htonl((uint32_t)ofs);
1169 int rc = msync(fmap.buf, fmap.len, MS_ASYNC);
1172 result = PyErr_SetFromErrnoWithFilename(PyExc_IOError, filename);
1173 goto clean_and_return;
1176 result = PyLong_FromUnsignedLong(count);
1179 PyBuffer_Release(&fmap);
1184 // I would have made this a lower-level function that just fills in a buffer
1185 // with random values, and then written those values from python. But that's
1186 // about 20% slower in my tests, and since we typically generate random
1187 // numbers for benchmarking other parts of bup, any slowness in generating
1188 // random bytes will make our benchmarks inaccurate. Plus nobody wants
1189 // pseudorandom bytes much except for this anyway.
1190 static PyObject *write_random(PyObject *self, PyObject *args)
1192 uint32_t buf[1024/4];
1193 int fd = -1, seed = 0, verbose = 0;
1195 long long len = 0, kbytes = 0, written = 0;
1197 if (!PyArg_ParseTuple(args, "iLii", &fd, &len, &seed, &verbose))
1202 for (kbytes = 0; kbytes < len/1024; kbytes++)
1205 for (i = 0; i < sizeof(buf)/sizeof(buf[0]); i++)
1206 buf[i] = (uint32_t) random();
1207 ret = write(fd, buf, sizeof(buf));
1211 if (ret < (int)sizeof(buf))
1213 if (verbose && kbytes/1024 > 0 && !(kbytes%1024))
1214 fprintf(stderr, "Random: %lld Mbytes\r", kbytes/1024);
1217 // handle non-multiples of 1024
1221 for (i = 0; i < sizeof(buf)/sizeof(buf[0]); i++)
1222 buf[i] = (uint32_t) random();
1223 ret = write(fd, buf, len % 1024);
1229 if (kbytes/1024 > 0)
1230 fprintf(stderr, "Random: %lld Mbytes, done.\n", kbytes/1024);
1231 return Py_BuildValue("L", written);
1235 static PyObject *random_sha(PyObject *self, PyObject *args)
1237 static int seeded = 0;
1238 uint32_t shabuf[20/4];
1243 assert(sizeof(shabuf) == 20);
1244 srandom((unsigned int) time(NULL));
1248 if (!PyArg_ParseTuple(args, ""))
1251 memset(shabuf, 0, sizeof(shabuf));
1252 for (i=0; i < 20/4; i++)
1253 shabuf[i] = (uint32_t) random();
1254 return Py_BuildValue(rbuf_argf, shabuf, 20);
1258 static int _open_noatime(const char *filename, int attrs)
1260 int attrs_noatime, fd;
1263 attrs |= O_NOFOLLOW;
1266 attrs |= O_LARGEFILE;
1268 attrs_noatime = attrs;
1270 attrs_noatime |= O_NOATIME;
1272 fd = open(filename, attrs_noatime);
1273 if (fd < 0 && errno == EPERM)
1275 // older Linux kernels would return EPERM if you used O_NOATIME
1276 // and weren't the file's owner. This pointless restriction was
1277 // relaxed eventually, but we have to handle it anyway.
1278 // (VERY old kernels didn't recognized O_NOATIME, but they would
1279 // just harmlessly ignore it, so this branch won't trigger)
1280 fd = open(filename, attrs);
1286 static PyObject *open_noatime(PyObject *self, PyObject *args)
1288 char *filename = NULL;
1290 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1292 fd = _open_noatime(filename, 0);
1294 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1295 return Py_BuildValue("i", fd);
1299 static PyObject *fadvise_done(PyObject *self, PyObject *args)
1302 long long llofs, lllen = 0;
1303 if (!PyArg_ParseTuple(args, "iLL", &fd, &llofs, &lllen))
1306 if (!INTEGRAL_ASSIGNMENT_FITS(&ofs, llofs))
1307 return PyErr_Format(PyExc_OverflowError,
1308 "fadvise offset overflows off_t");
1309 if (!INTEGRAL_ASSIGNMENT_FITS(&len, lllen))
1310 return PyErr_Format(PyExc_OverflowError,
1311 "fadvise length overflows off_t");
1312 #ifdef POSIX_FADV_DONTNEED
1313 posix_fadvise(fd, ofs, len, POSIX_FADV_DONTNEED);
1315 return Py_BuildValue("");
1319 // Currently the Linux kernel and FUSE disagree over the type for
1320 // FS_IOC_GETFLAGS and FS_IOC_SETFLAGS. The kernel actually uses int,
1321 // but FUSE chose long (matching the declaration in linux/fs.h). So
1322 // if you use int, and then traverse a FUSE filesystem, you may
1323 // corrupt the stack. But if you use long, then you may get invalid
1324 // results on big-endian systems.
1326 // For now, we just use long, and then disable Linux attrs entirely
1327 // (with a warning) in helpers.py on systems that are affected.
1329 #ifdef BUP_HAVE_FILE_ATTRS
1330 static PyObject *bup_get_linux_file_attr(PyObject *self, PyObject *args)
1337 if (!PyArg_ParseTuple(args, cstr_argf, &path))
1340 fd = _open_noatime(path, O_NONBLOCK);
1342 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1344 attr = 0; // Handle int/long mismatch (see above)
1345 rc = ioctl(fd, FS_IOC_GETFLAGS, &attr);
1349 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1352 assert(attr <= UINT_MAX); // Kernel type is actually int
1353 return PyLong_FromUnsignedLong(attr);
1355 #endif /* def BUP_HAVE_FILE_ATTRS */
1359 #ifdef BUP_HAVE_FILE_ATTRS
1360 static PyObject *bup_set_linux_file_attr(PyObject *self, PyObject *args)
1363 unsigned long orig_attr;
1369 if (!PyArg_ParseTuple(args, cstr_argf "O", &path, &py_attr))
1372 if (!bup_uint_from_py(&attr, py_attr, "attr"))
1375 fd = open(path, O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_NOFOLLOW);
1377 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1379 // Restrict attr to modifiable flags acdeijstuADST -- see
1380 // chattr(1) and the e2fsprogs source. Letter to flag mapping is
1381 // in pf.c flags_array[].
1382 attr &= FS_APPEND_FL | FS_COMPR_FL | FS_NODUMP_FL | FS_EXTENT_FL
1383 | FS_IMMUTABLE_FL | FS_JOURNAL_DATA_FL | FS_SECRM_FL | FS_NOTAIL_FL
1384 | FS_UNRM_FL | FS_NOATIME_FL | FS_DIRSYNC_FL | FS_SYNC_FL
1385 | FS_TOPDIR_FL | FS_NOCOW_FL;
1387 // The extents flag can't be removed, so don't (see chattr(1) and chattr.c).
1388 orig_attr = 0; // Handle int/long mismatch (see above)
1389 rc = ioctl(fd, FS_IOC_GETFLAGS, &orig_attr);
1393 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1395 assert(orig_attr <= UINT_MAX); // Kernel type is actually int
1396 attr |= ((unsigned int) orig_attr) & FS_EXTENT_FL;
1398 rc = ioctl(fd, FS_IOC_SETFLAGS, &attr);
1402 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1406 return Py_BuildValue("O", Py_None);
1408 #endif /* def BUP_HAVE_FILE_ATTRS */
1411 #ifndef BUP_USE_PYTHON_UTIME // just for Python 2 now
1412 #ifndef HAVE_UTIMENSAT
1414 #error "cannot find utimensat or utimes()"
1416 #ifndef HAVE_LUTIMES
1417 #error "cannot find utimensat or lutimes()"
1420 #endif // defined BUP_USE_PYTHON_UTIME
1422 #define ASSIGN_PYLONG_TO_INTEGRAL(dest, pylong, overflow) \
1426 const long long lltmp = PyLong_AsLongLong(pylong); \
1427 if (lltmp == -1 && PyErr_Occurred()) \
1429 if (PyErr_ExceptionMatches(PyExc_OverflowError)) \
1431 const unsigned long long ulltmp = PyLong_AsUnsignedLongLong(pylong); \
1432 if (ulltmp == (unsigned long long) -1 && PyErr_Occurred()) \
1434 if (PyErr_ExceptionMatches(PyExc_OverflowError)) \
1440 if (INTEGRAL_ASSIGNMENT_FITS((dest), ulltmp)) \
1448 if (INTEGRAL_ASSIGNMENT_FITS((dest), lltmp)) \
1457 #ifndef BUP_USE_PYTHON_UTIME // just for Python 2 now
1458 #ifdef HAVE_UTIMENSAT
1460 static PyObject *bup_utimensat(PyObject *self, PyObject *args)
1465 PyObject *access_py, *modification_py;
1466 struct timespec ts[2];
1468 if (!PyArg_ParseTuple(args, "i" cstr_argf "((Ol)(Ol))i",
1471 &access_py, &(ts[0].tv_nsec),
1472 &modification_py, &(ts[1].tv_nsec),
1477 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(ts[0].tv_sec), access_py, &overflow))
1480 PyErr_SetString(PyExc_ValueError,
1481 "unable to convert access time seconds for utimensat");
1484 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(ts[1].tv_sec), modification_py, &overflow))
1487 PyErr_SetString(PyExc_ValueError,
1488 "unable to convert modification time seconds for utimensat");
1491 rc = utimensat(fd, path, ts, flag);
1493 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1495 return Py_BuildValue("O", Py_None);
1498 #endif /* def HAVE_UTIMENSAT */
1501 #if defined(HAVE_UTIMES) || defined(HAVE_LUTIMES)
1503 static int bup_parse_xutimes_args(char **path,
1504 struct timeval tv[2],
1507 PyObject *access_py, *modification_py;
1508 long long access_us, modification_us; // POSIX guarantees tv_usec is signed.
1510 if (!PyArg_ParseTuple(args, cstr_argf "((OL)(OL))",
1512 &access_py, &access_us,
1513 &modification_py, &modification_us))
1517 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(tv[0].tv_sec), access_py, &overflow))
1520 PyErr_SetString(PyExc_ValueError, "unable to convert access time seconds to timeval");
1523 if (!INTEGRAL_ASSIGNMENT_FITS(&(tv[0].tv_usec), access_us))
1525 PyErr_SetString(PyExc_ValueError, "unable to convert access time nanoseconds to timeval");
1528 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(tv[1].tv_sec), modification_py, &overflow))
1531 PyErr_SetString(PyExc_ValueError, "unable to convert modification time seconds to timeval");
1534 if (!INTEGRAL_ASSIGNMENT_FITS(&(tv[1].tv_usec), modification_us))
1536 PyErr_SetString(PyExc_ValueError, "unable to convert modification time nanoseconds to timeval");
1542 #endif /* defined(HAVE_UTIMES) || defined(HAVE_LUTIMES) */
1546 static PyObject *bup_utimes(PyObject *self, PyObject *args)
1549 struct timeval tv[2];
1550 if (!bup_parse_xutimes_args(&path, tv, args))
1552 int rc = utimes(path, tv);
1554 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1555 return Py_BuildValue("O", Py_None);
1557 #endif /* def HAVE_UTIMES */
1561 static PyObject *bup_lutimes(PyObject *self, PyObject *args)
1564 struct timeval tv[2];
1565 if (!bup_parse_xutimes_args(&path, tv, args))
1567 int rc = lutimes(path, tv);
1569 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1571 return Py_BuildValue("O", Py_None);
1573 #endif /* def HAVE_LUTIMES */
1575 #endif // defined BUP_USE_PYTHON_UTIME
1578 #ifdef HAVE_STAT_ST_ATIM
1579 # define BUP_STAT_ATIME_NS(st) (st)->st_atim.tv_nsec
1580 # define BUP_STAT_MTIME_NS(st) (st)->st_mtim.tv_nsec
1581 # define BUP_STAT_CTIME_NS(st) (st)->st_ctim.tv_nsec
1582 #elif defined HAVE_STAT_ST_ATIMENSEC
1583 # define BUP_STAT_ATIME_NS(st) (st)->st_atimespec.tv_nsec
1584 # define BUP_STAT_MTIME_NS(st) (st)->st_mtimespec.tv_nsec
1585 # define BUP_STAT_CTIME_NS(st) (st)->st_ctimespec.tv_nsec
1587 # define BUP_STAT_ATIME_NS(st) 0
1588 # define BUP_STAT_MTIME_NS(st) 0
1589 # define BUP_STAT_CTIME_NS(st) 0
1593 static PyObject *stat_struct_to_py(const struct stat *st,
1594 const char *filename,
1597 // We can check the known (via POSIX) signed and unsigned types at
1598 // compile time, but not (easily) the unspecified types, so handle
1599 // those via INTEGER_TO_PY(). Assumes ns values will fit in a
1601 return Py_BuildValue("NKNNNNNL(Nl)(Nl)(Nl)",
1602 INTEGER_TO_PY(st->st_mode),
1603 (unsigned PY_LONG_LONG) st->st_ino,
1604 INTEGER_TO_PY(st->st_dev),
1605 INTEGER_TO_PY(st->st_nlink),
1606 INTEGER_TO_PY(st->st_uid),
1607 INTEGER_TO_PY(st->st_gid),
1608 INTEGER_TO_PY(st->st_rdev),
1609 (PY_LONG_LONG) st->st_size,
1610 INTEGER_TO_PY(st->st_atime),
1611 (long) BUP_STAT_ATIME_NS(st),
1612 INTEGER_TO_PY(st->st_mtime),
1613 (long) BUP_STAT_MTIME_NS(st),
1614 INTEGER_TO_PY(st->st_ctime),
1615 (long) BUP_STAT_CTIME_NS(st));
1619 static PyObject *bup_stat(PyObject *self, PyObject *args)
1624 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1628 rc = stat(filename, &st);
1630 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1631 return stat_struct_to_py(&st, filename, 0);
1635 static PyObject *bup_lstat(PyObject *self, PyObject *args)
1640 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1644 rc = lstat(filename, &st);
1646 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1647 return stat_struct_to_py(&st, filename, 0);
1651 static PyObject *bup_fstat(PyObject *self, PyObject *args)
1655 if (!PyArg_ParseTuple(args, "i", &fd))
1659 rc = fstat(fd, &st);
1661 return PyErr_SetFromErrno(PyExc_OSError);
1662 return stat_struct_to_py(&st, NULL, fd);
1666 #ifdef HAVE_TM_TM_GMTOFF
1667 static PyObject *bup_localtime(PyObject *self, PyObject *args)
1671 if (!PyArg_ParseTuple(args, "L", &lltime))
1673 if (!INTEGRAL_ASSIGNMENT_FITS(&ttime, lltime))
1674 return PyErr_Format(PyExc_OverflowError, "time value too large");
1678 if(localtime_r(&ttime, &tm) == NULL)
1679 return PyErr_SetFromErrno(PyExc_OSError);
1681 // Match the Python struct_time values.
1682 return Py_BuildValue("[i,i,i,i,i,i,i,i,i,i,s]",
1683 1900 + tm.tm_year, tm.tm_mon + 1, tm.tm_mday,
1684 tm.tm_hour, tm.tm_min, tm.tm_sec,
1685 tm.tm_wday, tm.tm_yday + 1,
1686 tm.tm_isdst, tm.tm_gmtoff, tm.tm_zone);
1688 #endif /* def HAVE_TM_TM_GMTOFF */
1691 #ifdef BUP_MINCORE_BUF_TYPE
1692 static PyObject *bup_mincore(PyObject *self, PyObject *args)
1694 Py_buffer src, dest;
1695 PyObject *py_src_n, *py_src_off, *py_dest_off;
1697 if (!PyArg_ParseTuple(args, cstr_argf "*OOw*O",
1698 &src, &py_src_n, &py_src_off,
1699 &dest, &py_dest_off))
1702 PyObject *result = NULL;
1704 unsigned long long src_n, src_off, dest_off;
1705 if (!(bup_ullong_from_py(&src_n, py_src_n, "src_n")
1706 && bup_ullong_from_py(&src_off, py_src_off, "src_off")
1707 && bup_ullong_from_py(&dest_off, py_dest_off, "dest_off")))
1708 goto clean_and_return;
1710 unsigned long long src_region_end;
1711 if (!uadd(&src_region_end, src_off, src_n)) {
1712 result = PyErr_Format(PyExc_OverflowError, "(src_off + src_n) too large");
1713 goto clean_and_return;
1715 assert(src.len >= 0);
1716 if (src_region_end > (unsigned long long) src.len) {
1717 result = PyErr_Format(PyExc_OverflowError, "region runs off end of src");
1718 goto clean_and_return;
1721 unsigned long long dest_size;
1722 if (!INTEGRAL_ASSIGNMENT_FITS(&dest_size, dest.len)) {
1723 result = PyErr_Format(PyExc_OverflowError, "invalid dest size");
1724 goto clean_and_return;
1726 if (dest_off > dest_size) {
1727 result = PyErr_Format(PyExc_OverflowError, "region runs off end of dest");
1728 goto clean_and_return;
1732 if (!INTEGRAL_ASSIGNMENT_FITS(&length, src_n)) {
1733 result = PyErr_Format(PyExc_OverflowError, "src_n overflows size_t");
1734 goto clean_and_return;
1736 int rc = mincore((void *)(src.buf + src_off), length,
1737 (BUP_MINCORE_BUF_TYPE *) (dest.buf + dest_off));
1739 result = PyErr_SetFromErrno(PyExc_OSError);
1740 goto clean_and_return;
1742 result = Py_BuildValue("O", Py_None);
1745 PyBuffer_Release(&src);
1746 PyBuffer_Release(&dest);
1749 #endif /* def BUP_MINCORE_BUF_TYPE */
1752 static PyObject *tuple_from_cstrs(char **cstrs)
1754 // Assumes list is null terminated
1756 while(cstrs[n] != NULL)
1760 if (!INTEGRAL_ASSIGNMENT_FITS(&sn, n))
1761 return PyErr_Format(PyExc_OverflowError, "string array too large");
1763 PyObject *result = PyTuple_New(sn);
1765 for (i = 0; i < sn; i++)
1767 PyObject *gname = Py_BuildValue(cstr_argf, cstrs[i]);
1773 PyTuple_SET_ITEM(result, i, gname);
1778 static PyObject *appropriate_errno_ex(void)
1782 return PyErr_NoMemory();
1786 // In 3.3 IOError was merged into OSError.
1787 return PyErr_SetFromErrno(PyExc_IOError);
1789 return PyErr_SetFromErrno(PyExc_OSError);
1794 static PyObject *pwd_struct_to_py(const struct passwd *pwd)
1796 // We can check the known (via POSIX) signed and unsigned types at
1797 // compile time, but not (easily) the unspecified types, so handle
1798 // those via INTEGER_TO_PY().
1801 return Py_BuildValue(cstr_argf cstr_argf "OO"
1802 cstr_argf cstr_argf cstr_argf,
1805 INTEGER_TO_PY(pwd->pw_uid),
1806 INTEGER_TO_PY(pwd->pw_gid),
1812 static PyObject *bup_getpwuid(PyObject *self, PyObject *args)
1814 unsigned long long py_uid;
1815 if (!PyArg_ParseTuple(args, "K", &py_uid))
1818 if (!INTEGRAL_ASSIGNMENT_FITS(&uid, py_uid))
1819 return PyErr_Format(PyExc_OverflowError, "uid too large for uid_t");
1822 struct passwd *pwd = getpwuid(uid);
1824 return appropriate_errno_ex();
1825 return pwd_struct_to_py(pwd);
1828 static PyObject *bup_getpwnam(PyObject *self, PyObject *args)
1831 if (!PyArg_ParseTuple(args, "S", &py_name))
1834 char *name = PyBytes_AS_STRING(py_name);
1836 struct passwd *pwd = getpwnam(name);
1838 return appropriate_errno_ex();
1839 return pwd_struct_to_py(pwd);
1842 static PyObject *grp_struct_to_py(const struct group *grp)
1844 // We can check the known (via POSIX) signed and unsigned types at
1845 // compile time, but not (easily) the unspecified types, so handle
1846 // those via INTEGER_TO_PY().
1850 PyObject *members = tuple_from_cstrs(grp->gr_mem);
1851 if (members == NULL)
1853 return Py_BuildValue(cstr_argf cstr_argf "OO",
1856 INTEGER_TO_PY(grp->gr_gid),
1860 static PyObject *bup_getgrgid(PyObject *self, PyObject *args)
1862 unsigned long long py_gid;
1863 if (!PyArg_ParseTuple(args, "K", &py_gid))
1866 if (!INTEGRAL_ASSIGNMENT_FITS(&gid, py_gid))
1867 return PyErr_Format(PyExc_OverflowError, "gid too large for gid_t");
1870 struct group *grp = getgrgid(gid);
1872 return appropriate_errno_ex();
1873 return grp_struct_to_py(grp);
1876 static PyObject *bup_getgrnam(PyObject *self, PyObject *args)
1879 if (!PyArg_ParseTuple(args, "S", &py_name))
1882 char *name = PyBytes_AS_STRING(py_name);
1884 struct group *grp = getgrnam(name);
1886 return appropriate_errno_ex();
1887 return grp_struct_to_py(grp);
1891 static PyObject *bup_gethostname(PyObject *mod, PyObject *ignore)
1893 #ifdef HOST_NAME_MAX
1894 char buf[HOST_NAME_MAX + 1] = {};
1896 /* 'SUSv2 guarantees that "Host names are limited to 255 bytes".' */
1900 if (gethostname(buf, sizeof(buf) - 1))
1901 return PyErr_SetFromErrno(PyExc_IOError);
1902 return PyBytes_FromString(buf);
1906 #ifdef BUP_HAVE_READLINE
1908 static char *cstr_from_bytes(PyObject *bytes)
1912 int rc = PyBytes_AsStringAndSize(bytes, &buf, &length);
1915 char *result = checked_malloc(length, sizeof(char));
1918 memcpy(result, buf, length);
1922 static char **cstrs_from_seq(PyObject *seq)
1924 char **result = NULL;
1925 seq = PySequence_Fast(seq, "Cannot convert sequence items to C strings");
1929 const Py_ssize_t len = PySequence_Fast_GET_SIZE(seq);
1930 if (len > PY_SSIZE_T_MAX - 1) {
1931 PyErr_Format(PyExc_OverflowError,
1932 "Sequence length %zd too large for conversion to C array",
1936 result = checked_malloc(len + 1, sizeof(char *));
1940 for (i = 0; i < len; i++)
1942 PyObject *item = PySequence_Fast_GET_ITEM(seq, i);
1944 goto abandon_result;
1945 result[i] = cstr_from_bytes(item);
1948 goto abandon_result;
1966 static char* our_word_break_chars = NULL;
1969 bup_set_completer_word_break_characters(PyObject *self, PyObject *args)
1972 if (!PyArg_ParseTuple(args, cstr_argf, &bytes))
1974 char *prev = our_word_break_chars;
1975 char *next = strdup(bytes);
1977 return PyErr_NoMemory();
1978 our_word_break_chars = next;
1979 rl_completer_word_break_characters = next;
1986 bup_get_completer_word_break_characters(PyObject *self, PyObject *args)
1988 if (!PyArg_ParseTuple(args, ""))
1990 return PyBytes_FromString(rl_completer_word_break_characters);
1993 static PyObject *bup_get_line_buffer(PyObject *self, PyObject *args)
1995 if (!PyArg_ParseTuple(args, ""))
1997 return PyBytes_FromString(rl_line_buffer);
2001 bup_parse_and_bind(PyObject *self, PyObject *args)
2004 if (!PyArg_ParseTuple(args, cstr_argf ":parse_and_bind", &bytes))
2006 char *tmp = strdup(bytes); // Because it may modify the arg
2008 return PyErr_NoMemory();
2009 int rc = rl_parse_and_bind(tmp);
2012 return PyErr_Format(PyExc_OSError,
2013 "system rl_parse_and_bind failed (%d)", rc);
2018 static PyObject *py_on_attempted_completion;
2019 static char **prev_completions;
2021 static char **on_attempted_completion(const char *text, int start, int end)
2023 if (!py_on_attempted_completion)
2026 char **result = NULL;
2027 PyObject *py_result = PyObject_CallFunction(py_on_attempted_completion,
2032 if (py_result != Py_None) {
2033 result = cstrs_from_seq(py_result);
2034 free(prev_completions);
2035 prev_completions = result;
2037 Py_DECREF(py_result);
2042 bup_set_attempted_completion_function(PyObject *self, PyObject *args)
2044 PyObject *completer;
2045 if (!PyArg_ParseTuple(args, "O", &completer))
2048 PyObject *prev = py_on_attempted_completion;
2049 if (completer == Py_None)
2051 py_on_attempted_completion = NULL;
2052 rl_attempted_completion_function = NULL;
2054 py_on_attempted_completion = completer;
2055 rl_attempted_completion_function = on_attempted_completion;
2056 Py_INCREF(completer);
2063 static PyObject *py_on_completion_entry;
2065 static char *on_completion_entry(const char *text, int state)
2067 if (!py_on_completion_entry)
2070 PyObject *py_result = PyObject_CallFunction(py_on_completion_entry,
2071 cstr_argf "i", text, state);
2074 char *result = (py_result == Py_None) ? NULL : cstr_from_bytes(py_result);
2075 Py_DECREF(py_result);
2080 bup_set_completion_entry_function(PyObject *self, PyObject *args)
2082 PyObject *completer;
2083 if (!PyArg_ParseTuple(args, "O", &completer))
2086 PyObject *prev = py_on_completion_entry;
2087 if (completer == Py_None) {
2088 py_on_completion_entry = NULL;
2089 rl_completion_entry_function = NULL;
2091 py_on_completion_entry = completer;
2092 rl_completion_entry_function = on_completion_entry;
2093 Py_INCREF(completer);
2100 bup_readline(PyObject *self, PyObject *args)
2103 if (!PyArg_ParseTuple(args, cstr_argf, &prompt))
2105 char *line = readline(prompt);
2107 return PyErr_Format(PyExc_EOFError, "readline EOF");
2108 PyObject *result = PyBytes_FromString(line);
2113 #endif // defined BUP_HAVE_READLINE
2115 #if defined(HAVE_SYS_ACL_H) && \
2116 defined(HAVE_ACL_LIBACL_H) && \
2117 defined(HAVE_ACL_EXTENDED_FILE) && \
2118 defined(HAVE_ACL_GET_FILE) && \
2119 defined(HAVE_ACL_TO_ANY_TEXT) && \
2120 defined(HAVE_ACL_FROM_TEXT) && \
2121 defined(HAVE_ACL_SET_FILE)
2122 #define ACL_SUPPORT 1
2123 #include <sys/acl.h>
2124 #include <acl/libacl.h>
2128 // -1 for errors, with python exception set
2129 // -2 for ignored errors (not supported)
2130 static int bup_read_acl_to_text(const char *name, acl_type_t type,
2131 char **txt, char **num)
2135 acl = acl_get_file(name, type);
2137 if (errno == EOPNOTSUPP || errno == ENOSYS)
2139 PyErr_SetFromErrno(PyExc_IOError);
2144 *txt = acl_to_any_text(acl, "", '\n', TEXT_ABBREVIATE);
2146 *num = acl_to_any_text(acl, "", '\n', TEXT_ABBREVIATE | TEXT_NUMERIC_IDS);
2151 if (errno == ENOMEM)
2154 PyErr_SetFromErrno(PyExc_IOError);
2157 acl_free((acl_t)*txt);
2159 acl_free((acl_t)*num);
2164 static PyObject *bup_read_acl(PyObject *self, PyObject *args)
2168 PyObject *ret = NULL;
2169 char *acl_txt = NULL, *acl_num = NULL;
2171 if (!PyArg_ParseTuple(args, cstr_argf "i", &name, &isdir))
2174 if (!acl_extended_file(name))
2177 rv = bup_read_acl_to_text(name, ACL_TYPE_ACCESS, &acl_txt, &acl_num);
2182 char *def_txt = NULL, *def_num = NULL;
2184 rv = bup_read_acl_to_text(name, ACL_TYPE_DEFAULT, &def_txt, &def_num);
2188 ret = Py_BuildValue("[" cstr_argf cstr_argf cstr_argf cstr_argf "]",
2189 acl_txt, acl_num, def_txt, def_num);
2192 acl_free((acl_t)def_txt);
2194 acl_free((acl_t)def_num);
2196 ret = Py_BuildValue("[" cstr_argf cstr_argf "]",
2202 acl_free((acl_t)acl_txt);
2204 acl_free((acl_t)acl_num);
2210 static int bup_apply_acl_string(const char *name, const char *s)
2212 acl_t acl = acl_from_text(s);
2216 PyErr_SetFromErrno(PyExc_IOError);
2220 if (acl_set_file(name, ACL_TYPE_ACCESS, acl)) {
2221 PyErr_SetFromErrno(PyExc_IOError);
2230 static PyObject *bup_apply_acl(PyObject *self, PyObject *args)
2232 char *name, *acl, *def = NULL;
2234 if (!PyArg_ParseTuple(args, cstr_argf cstr_argf "|" cstr_argf, &name, &acl, &def))
2237 if (bup_apply_acl_string(name, acl))
2240 if (def && bup_apply_acl_string(name, def))
2247 static PyMethodDef helper_methods[] = {
2248 { "write_sparsely", bup_write_sparsely, METH_VARARGS,
2249 "Write buf excepting zeros at the end. Return trailing zero count." },
2250 { "selftest", selftest, METH_VARARGS,
2251 "Check that the rolling checksum rolls correctly (for unit tests)." },
2252 { "blobbits", blobbits, METH_VARARGS,
2253 "Return the number of bits in the rolling checksum." },
2254 { "splitbuf", splitbuf, METH_VARARGS,
2255 "Split a list of strings based on a rolling checksum." },
2256 { "bitmatch", bitmatch, METH_VARARGS,
2257 "Count the number of matching prefix bits between two strings." },
2258 { "firstword", firstword, METH_VARARGS,
2259 "Return an int corresponding to the first 32 bits of buf." },
2260 { "bloom_contains", bloom_contains, METH_VARARGS,
2261 "Check if a bloom filter of 2^nbits bytes contains an object" },
2262 { "bloom_add", bloom_add, METH_VARARGS,
2263 "Add an object to a bloom filter of 2^nbits bytes" },
2264 { "extract_bits", extract_bits, METH_VARARGS,
2265 "Take the first 'nbits' bits from 'buf' and return them as an int." },
2266 { "merge_into", merge_into, METH_VARARGS,
2267 "Merges a bunch of idx and midx files into a single midx." },
2268 { "write_idx", write_idx, METH_VARARGS,
2269 "Write a PackIdxV2 file from an idx list of lists of tuples" },
2270 { "write_random", write_random, METH_VARARGS,
2271 "Write random bytes to the given file descriptor" },
2272 { "random_sha", random_sha, METH_VARARGS,
2273 "Return a random 20-byte string" },
2274 { "open_noatime", open_noatime, METH_VARARGS,
2275 "open() the given filename for read with O_NOATIME if possible" },
2276 { "fadvise_done", fadvise_done, METH_VARARGS,
2277 "Inform the kernel that we're finished with earlier parts of a file" },
2278 #ifdef BUP_HAVE_FILE_ATTRS
2279 { "get_linux_file_attr", bup_get_linux_file_attr, METH_VARARGS,
2280 "Return the Linux attributes for the given file." },
2282 #ifdef BUP_HAVE_FILE_ATTRS
2283 { "set_linux_file_attr", bup_set_linux_file_attr, METH_VARARGS,
2284 "Set the Linux attributes for the given file." },
2287 #ifndef BUP_USE_PYTHON_UTIME // just for Python 2 now
2288 #ifdef HAVE_UTIMENSAT
2289 { "bup_utimensat", bup_utimensat, METH_VARARGS,
2290 "Change path timestamps with nanosecond precision (POSIX)." },
2293 { "bup_utimes", bup_utimes, METH_VARARGS,
2294 "Change path timestamps with microsecond precision." },
2297 { "bup_lutimes", bup_lutimes, METH_VARARGS,
2298 "Change path timestamps with microsecond precision;"
2299 " don't follow symlinks." },
2301 #endif // defined BUP_USE_PYTHON_UTIME
2303 { "stat", bup_stat, METH_VARARGS,
2304 "Extended version of stat." },
2305 { "lstat", bup_lstat, METH_VARARGS,
2306 "Extended version of lstat." },
2307 { "fstat", bup_fstat, METH_VARARGS,
2308 "Extended version of fstat." },
2309 #ifdef HAVE_TM_TM_GMTOFF
2310 { "localtime", bup_localtime, METH_VARARGS,
2311 "Return struct_time elements plus the timezone offset and name." },
2313 { "bytescmp", bup_bytescmp, METH_VARARGS,
2314 "Return a negative value if x < y, zero if equal, positive otherwise."},
2315 { "cat_bytes", bup_cat_bytes, METH_VARARGS,
2316 "For (x_bytes, x_ofs, x_n, y_bytes, y_ofs, y_n) arguments, return their concatenation."},
2317 #ifdef BUP_MINCORE_BUF_TYPE
2318 { "mincore", bup_mincore, METH_VARARGS,
2319 "For mincore(src, src_n, src_off, dest, dest_off)"
2320 " call the system mincore(src + src_off, src_n, &dest[dest_off])." },
2322 { "getpwuid", bup_getpwuid, METH_VARARGS,
2323 "Return the password database entry for the given numeric user id,"
2324 " as a tuple with all C strings as bytes(), or None if the user does"
2326 { "getpwnam", bup_getpwnam, METH_VARARGS,
2327 "Return the password database entry for the given user name,"
2328 " as a tuple with all C strings as bytes(), or None if the user does"
2330 { "getgrgid", bup_getgrgid, METH_VARARGS,
2331 "Return the group database entry for the given numeric group id,"
2332 " as a tuple with all C strings as bytes(), or None if the group does"
2334 { "getgrnam", bup_getgrnam, METH_VARARGS,
2335 "Return the group database entry for the given group name,"
2336 " as a tuple with all C strings as bytes(), or None if the group does"
2338 { "gethostname", bup_gethostname, METH_NOARGS,
2339 "Return the current hostname (as bytes)" },
2340 #ifdef BUP_HAVE_READLINE
2341 { "set_completion_entry_function", bup_set_completion_entry_function, METH_VARARGS,
2342 "Set rl_completion_entry_function. Called as f(text, state)." },
2343 { "set_attempted_completion_function", bup_set_attempted_completion_function, METH_VARARGS,
2344 "Set rl_attempted_completion_function. Called as f(text, start, end)." },
2345 { "parse_and_bind", bup_parse_and_bind, METH_VARARGS,
2346 "Call rl_parse_and_bind." },
2347 { "get_line_buffer", bup_get_line_buffer, METH_NOARGS,
2348 "Return rl_line_buffer." },
2349 { "get_completer_word_break_characters", bup_get_completer_word_break_characters, METH_NOARGS,
2350 "Return rl_completer_word_break_characters." },
2351 { "set_completer_word_break_characters", bup_set_completer_word_break_characters, METH_VARARGS,
2352 "Set rl_completer_word_break_characters." },
2353 { "readline", bup_readline, METH_VARARGS,
2354 "Call readline(prompt)." },
2355 #endif // defined BUP_HAVE_READLINE
2357 { "read_acl", bup_read_acl, METH_VARARGS,
2358 "read_acl(name, isdir)\n\n"
2359 "Read ACLs for the given file/dirname and return the correctly encoded"
2360 " list [txt, num, def_tx, def_num] (the def_* being empty bytestrings"
2361 " unless the second argument 'isdir' is True)." },
2362 { "apply_acl", bup_apply_acl, METH_VARARGS,
2363 "apply_acl(name, acl, def=None)\n\n"
2364 "Given a file/dirname (bytes) and the ACLs to restore, do that." },
2365 #endif /* HAVE_ACLS */
2366 { NULL, NULL, 0, NULL }, // sentinel
2369 static void test_integral_assignment_fits(void)
2371 assert(sizeof(signed short) == sizeof(unsigned short));
2372 assert(sizeof(signed short) < sizeof(signed long long));
2373 assert(sizeof(signed short) < sizeof(unsigned long long));
2374 assert(sizeof(unsigned short) < sizeof(signed long long));
2375 assert(sizeof(unsigned short) < sizeof(unsigned long long));
2376 assert(sizeof(Py_ssize_t) <= sizeof(size_t));
2378 signed short ss, ssmin = SHRT_MIN, ssmax = SHRT_MAX;
2379 unsigned short us, usmax = USHRT_MAX;
2380 signed long long sllmin = LLONG_MIN, sllmax = LLONG_MAX;
2381 unsigned long long ullmax = ULLONG_MAX;
2383 assert(INTEGRAL_ASSIGNMENT_FITS(&ss, ssmax));
2384 assert(INTEGRAL_ASSIGNMENT_FITS(&ss, ssmin));
2385 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, usmax));
2386 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, sllmin));
2387 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, sllmax));
2388 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, ullmax));
2390 assert(INTEGRAL_ASSIGNMENT_FITS(&us, usmax));
2391 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, ssmin));
2392 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, sllmin));
2393 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, sllmax));
2394 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, ullmax));
2398 static int setup_module(PyObject *m)
2400 // FIXME: migrate these tests to configure, or at least don't
2401 // possibly crash the whole application. Check against the type
2402 // we're going to use when passing to python. Other stat types
2403 // are tested at runtime.
2404 assert(sizeof(ino_t) <= sizeof(unsigned PY_LONG_LONG));
2405 assert(sizeof(off_t) <= sizeof(PY_LONG_LONG));
2406 assert(sizeof(blksize_t) <= sizeof(PY_LONG_LONG));
2407 assert(sizeof(blkcnt_t) <= sizeof(PY_LONG_LONG));
2408 // Just be sure (relevant when passing timestamps back to Python above).
2409 assert(sizeof(PY_LONG_LONG) <= sizeof(long long));
2410 assert(sizeof(unsigned PY_LONG_LONG) <= sizeof(unsigned long long));
2412 test_integral_assignment_fits();
2414 // Originally required by append_sparse_region()
2417 if (!INTEGRAL_ASSIGNMENT_FITS(&probe, INT_MAX))
2419 fprintf(stderr, "off_t can't hold INT_MAX; please report.\n");
2427 value = INTEGER_TO_PY(INT_MAX);
2428 PyObject_SetAttrString(m, "INT_MAX", value);
2430 value = INTEGER_TO_PY(UINT_MAX);
2431 PyObject_SetAttrString(m, "UINT_MAX", value);
2435 #ifndef BUP_USE_PYTHON_UTIME // just for Python 2 now
2436 #ifdef HAVE_UTIMENSAT
2439 value = INTEGER_TO_PY(AT_FDCWD);
2440 PyObject_SetAttrString(m, "AT_FDCWD", value);
2442 value = INTEGER_TO_PY(AT_SYMLINK_NOFOLLOW);
2443 PyObject_SetAttrString(m, "AT_SYMLINK_NOFOLLOW", value);
2445 value = INTEGER_TO_PY(UTIME_NOW);
2446 PyObject_SetAttrString(m, "UTIME_NOW", value);
2450 #endif // defined BUP_USE_PYTHON_UTIME
2452 #ifdef BUP_HAVE_MINCORE_INCORE
2455 value = INTEGER_TO_PY(MINCORE_INCORE);
2456 PyObject_SetAttrString(m, "MINCORE_INCORE", value);
2461 e = getenv("BUP_FORCE_TTY");
2462 get_state(m)->istty2 = isatty(2) || (atoi(e ? e : "0") & 2);
2468 #if PY_MAJOR_VERSION < 3
2470 PyMODINIT_FUNC init_helpers(void)
2472 PyObject *m = Py_InitModule("_helpers", helper_methods);
2476 if (!setup_module(m))
2483 # else // PY_MAJOR_VERSION >= 3
2485 static struct PyModuleDef helpers_def = {
2486 PyModuleDef_HEAD_INIT,
2492 NULL, // helpers_traverse,
2493 NULL, // helpers_clear,
2497 PyMODINIT_FUNC PyInit__helpers(void)
2499 PyObject *module = PyModule_Create(&helpers_def);
2502 if (!setup_module(module))
2510 #endif // PY_MAJOR_VERSION >= 3