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 // At the moment any code that calls INTEGER_TO_PY() will have to
179 // disable -Wtautological-compare for clang. See below.
181 #define INTEGER_TO_PY(x) \
182 (((x) >= 0) ? PyLong_FromUnsignedLongLong(x) : PyLong_FromLongLong(x))
186 #if PY_MAJOR_VERSION < 3
187 static int bup_ulong_from_pyint(unsigned long *x, PyObject *py,
190 const long tmp = PyInt_AsLong(py);
191 if (tmp == -1 && PyErr_Occurred())
193 if (PyErr_ExceptionMatches(PyExc_OverflowError))
194 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned long",
200 PyErr_Format(PyExc_OverflowError,
201 "negative %s cannot be converted to unsigned long", name);
210 static int bup_ulong_from_py(unsigned long *x, PyObject *py, const char *name)
212 #if PY_MAJOR_VERSION < 3
214 return bup_ulong_from_pyint(x, py, name);
217 if (!PyLong_Check(py))
219 PyErr_Format(PyExc_TypeError, "expected integer %s", name);
223 const unsigned long tmp = PyLong_AsUnsignedLong(py);
224 if (PyErr_Occurred())
226 if (PyErr_ExceptionMatches(PyExc_OverflowError))
227 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned long",
236 static int bup_uint_from_py(unsigned int *x, PyObject *py, const char *name)
239 if (!bup_ulong_from_py(&tmp, py, name))
244 PyErr_Format(PyExc_OverflowError, "%s too big for unsigned int", name);
247 *x = (unsigned int) tmp;
251 static int bup_ullong_from_py(unsigned PY_LONG_LONG *x, PyObject *py,
254 #if PY_MAJOR_VERSION < 3
258 if (bup_ulong_from_pyint(&tmp, py, name))
267 if (!PyLong_Check(py))
269 PyErr_Format(PyExc_TypeError, "integer argument expected for %s", name);
273 const unsigned PY_LONG_LONG tmp = PyLong_AsUnsignedLongLong(py);
274 if (tmp == (unsigned long long) -1 && PyErr_Occurred())
276 if (PyErr_ExceptionMatches(PyExc_OverflowError))
277 PyErr_Format(PyExc_OverflowError,
278 "%s too big for unsigned long long", name);
286 static PyObject *bup_bytescmp(PyObject *self, PyObject *args)
288 PyObject *py_s1, *py_s2; // This is really a PyBytes/PyString
289 if (!PyArg_ParseTuple(args, "SS", &py_s1, &py_s2))
292 Py_ssize_t s1_len, s2_len;
293 if (PyBytes_AsStringAndSize(py_s1, &s1, &s1_len) == -1)
295 if (PyBytes_AsStringAndSize(py_s2, &s2, &s2_len) == -1)
297 const Py_ssize_t n = (s1_len < s2_len) ? s1_len : s2_len;
298 const int cmp = memcmp(s1, s2, n);
300 return PyLong_FromLong(cmp);
301 if (s1_len == s2_len)
302 return PyLong_FromLong(0);;
303 return PyLong_FromLong((s1_len < s2_len) ? -1 : 1);
307 static PyObject *bup_cat_bytes(PyObject *self, PyObject *args)
309 unsigned char *bufx = NULL, *bufy = NULL;
310 Py_ssize_t bufx_len, bufx_ofs, bufx_n;
311 Py_ssize_t bufy_len, bufy_ofs, bufy_n;
312 if (!PyArg_ParseTuple(args,
315 &bufx, &bufx_len, &bufx_ofs, &bufx_n,
316 &bufy, &bufy_len, &bufy_ofs, &bufy_n))
319 return PyErr_Format(PyExc_ValueError, "negative x offset");
321 return PyErr_Format(PyExc_ValueError, "negative x extent");
322 if (bufx_ofs > bufx_len)
323 return PyErr_Format(PyExc_ValueError, "x offset greater than length");
324 if (bufx_n > bufx_len - bufx_ofs)
325 return PyErr_Format(PyExc_ValueError, "x extent past end of buffer");
328 return PyErr_Format(PyExc_ValueError, "negative y offset");
330 return PyErr_Format(PyExc_ValueError, "negative y extent");
331 if (bufy_ofs > bufy_len)
332 return PyErr_Format(PyExc_ValueError, "y offset greater than length");
333 if (bufy_n > bufy_len - bufy_ofs)
334 return PyErr_Format(PyExc_ValueError, "y extent past end of buffer");
336 if (bufy_n > PY_SSIZE_T_MAX - bufx_n)
337 return PyErr_Format(PyExc_OverflowError, "result length too long");
339 PyObject *result = PyBytes_FromStringAndSize(NULL, bufx_n + bufy_n);
341 return PyErr_NoMemory();
342 char *buf = PyBytes_AS_STRING(result);
343 memcpy(buf, bufx + bufx_ofs, bufx_n);
344 memcpy(buf + bufx_n, bufy + bufy_ofs, bufy_n);
350 // Probably we should use autoconf or something and set HAVE_PY_GETARGCARGV...
351 #if __WIN32__ || __CYGWIN__
353 // There's no 'ps' on win32 anyway, and Py_GetArgcArgv() isn't available.
354 static void unpythonize_argv(void) { }
356 #else // not __WIN32__
358 // For some reason this isn't declared in Python.h
359 extern void Py_GetArgcArgv(int *argc, char ***argv);
361 static void unpythonize_argv(void)
366 Py_GetArgcArgv(&argc, &argv);
368 for (i = 0; i < argc-1; i++)
370 if (argv[i] + strlen(argv[i]) + 1 != argv[i+1])
372 // The argv block doesn't work the way we expected; it's unsafe
378 arge = argv[argc-1] + strlen(argv[argc-1]) + 1;
380 if (strstr(argv[0], "python") && argv[1] == argv[0] + strlen(argv[0]) + 1)
384 p = strrchr(argv[1], '/');
390 memmove(argv[0], p, len);
391 memset(arge - diff, 0, diff);
392 for (i = 0; i < argc; i++)
393 argv[i] = argv[i+1] ? argv[i+1]-diff : NULL;
398 #endif // not __WIN32__ or __CYGWIN__
401 static int write_all(int fd, const void *buf, const size_t count)
404 while (written < count)
406 const ssize_t rc = write(fd, buf + written, count - written);
415 static int uadd(unsigned long long *dest,
416 const unsigned long long x,
417 const unsigned long long y)
419 const unsigned long long result = x + y;
420 if (result < x || result < y)
427 static PyObject *append_sparse_region(const int fd, unsigned long long n)
432 if (!INTEGRAL_ASSIGNMENT_FITS(&new_off, n))
434 const off_t off = lseek(fd, new_off, SEEK_CUR);
435 if (off == (off_t) -1)
436 return PyErr_SetFromErrno(PyExc_IOError);
443 static PyObject *record_sparse_zeros(unsigned long long *new_pending,
445 unsigned long long prev_pending,
446 const unsigned long long count)
448 // Add count additional sparse zeros to prev_pending and store the
449 // result in new_pending, or if the total won't fit in
450 // new_pending, write some of the zeros to fd sparsely, and store
451 // the remaining sum in new_pending.
452 if (!uadd(new_pending, prev_pending, count))
454 PyObject *err = append_sparse_region(fd, prev_pending);
457 *new_pending = count;
463 static byte* find_not_zero(const byte * const start, const byte * const end)
465 // Return a pointer to first non-zero byte between start and end,
466 // or end if there isn't one.
467 assert(start <= end);
468 const unsigned char *cur = start;
469 while (cur < end && *cur == 0)
475 static byte* find_trailing_zeros(const byte * const start,
476 const byte * const end)
478 // Return a pointer to the start of any trailing run of zeros, or
479 // end if there isn't one.
480 assert(start <= end);
483 const byte * cur = end;
484 while (cur > start && *--cur == 0) {}
488 return (byte *) (cur + 1);
492 static byte *find_non_sparse_end(const byte * const start,
493 const byte * const end,
494 const ptrdiff_t min_len)
496 // Return the first pointer to a min_len sparse block in [start,
497 // end) if there is one, otherwise a pointer to the start of any
498 // trailing run of zeros. If there are no trailing zeros, return
504 // Probe in min_len jumps, searching backward from the jump
505 // destination for a non-zero byte. If such a byte is found, move
506 // just past it and try again.
507 const byte *candidate = start;
508 // End of any run of zeros, starting at candidate, that we've already seen
509 const byte *end_of_known_zeros = candidate;
510 while (end - candidate >= min_len) // Handle all min_len candidate blocks
512 const byte * const probe_end = candidate + min_len;
513 const byte * const trailing_zeros =
514 find_trailing_zeros(end_of_known_zeros, probe_end);
515 if (trailing_zeros == probe_end)
516 end_of_known_zeros = candidate = probe_end;
517 else if (trailing_zeros == end_of_known_zeros)
519 assert(candidate >= start);
520 assert(candidate <= end);
521 assert(*candidate == 0);
522 return (byte *) candidate;
526 candidate = trailing_zeros;
527 end_of_known_zeros = probe_end;
531 if (candidate == end)
534 // No min_len sparse run found, search backward from end
535 const byte * const trailing_zeros = find_trailing_zeros(end_of_known_zeros,
538 if (trailing_zeros == end_of_known_zeros)
540 assert(candidate >= start);
541 assert(candidate < end);
542 assert(*candidate == 0);
543 assert(end - candidate < min_len);
544 return (byte *) candidate;
547 if (trailing_zeros == end)
549 assert(*(end - 1) != 0);
553 assert(end - trailing_zeros < min_len);
554 assert(trailing_zeros >= start);
555 assert(trailing_zeros < end);
556 assert(*trailing_zeros == 0);
557 return (byte *) trailing_zeros;
561 static PyObject *bup_write_sparsely(PyObject *self, PyObject *args)
564 unsigned char *buf = NULL;
566 PyObject *py_min_sparse_len, *py_prev_sparse_len;
567 if (!PyArg_ParseTuple(args, "i" rbuf_argf "OO",
568 &fd, &buf, &sbuf_len,
569 &py_min_sparse_len, &py_prev_sparse_len))
571 ptrdiff_t min_sparse_len;
572 unsigned long long prev_sparse_len, buf_len, ul_min_sparse_len;
573 if (!bup_ullong_from_py(&ul_min_sparse_len, py_min_sparse_len, "min_sparse_len"))
575 if (!INTEGRAL_ASSIGNMENT_FITS(&min_sparse_len, ul_min_sparse_len))
576 return PyErr_Format(PyExc_OverflowError, "min_sparse_len too large");
577 if (!bup_ullong_from_py(&prev_sparse_len, py_prev_sparse_len, "prev_sparse_len"))
580 return PyErr_Format(PyExc_ValueError, "negative bufer length");
581 if (!INTEGRAL_ASSIGNMENT_FITS(&buf_len, sbuf_len))
582 return PyErr_Format(PyExc_OverflowError, "buffer length too large");
584 const byte * block = buf; // Start of pending block
585 const byte * const end = buf + buf_len;
586 unsigned long long zeros = prev_sparse_len;
589 assert(block <= end);
591 return PyLong_FromUnsignedLongLong(zeros);
595 // Look for the end of block, i.e. the next sparse run of
596 // at least min_sparse_len zeros, or the end of the
598 const byte * const probe = find_non_sparse_end(block + 1, end,
600 // Either at end of block, or end of non-sparse; write pending data
601 PyObject *err = append_sparse_region(fd, zeros);
604 int rc = write_all(fd, block, probe - block);
606 return PyErr_SetFromErrno(PyExc_IOError);
608 if (end - probe < min_sparse_len)
611 zeros = min_sparse_len;
612 block = probe + zeros;
616 // Should be in the first loop iteration, a sparse run of
617 // zeros, or nearly at the end of the block (within
619 const byte * const zeros_end = find_not_zero(block, end);
620 PyObject *err = record_sparse_zeros(&zeros, fd,
621 zeros, zeros_end - block);
624 assert(block <= zeros_end);
631 static PyObject *selftest(PyObject *self, PyObject *args)
633 if (!PyArg_ParseTuple(args, ""))
636 return Py_BuildValue("i", !bupsplit_selftest());
640 static PyObject *blobbits(PyObject *self, PyObject *args)
642 if (!PyArg_ParseTuple(args, ""))
644 return Py_BuildValue("i", BUP_BLOBBITS);
648 static PyObject *splitbuf(PyObject *self, PyObject *args)
650 // We stick to buffers in python 2 because they appear to be
651 // substantially smaller than memoryviews, and because
652 // zlib.compress() in python 2 can't accept a memoryview
653 // (cf. hashsplit.py).
654 int out = 0, bits = -1;
655 if (PY_MAJOR_VERSION > 2)
658 if (!PyArg_ParseTuple(args, "y*", &buf))
660 assert(buf.len <= INT_MAX);
661 out = bupsplit_find_ofs(buf.buf, buf.len, &bits);
662 PyBuffer_Release(&buf);
666 unsigned char *buf = NULL;
668 if (!PyArg_ParseTuple(args, "t#", &buf, &len))
670 assert(len <= INT_MAX);
671 out = bupsplit_find_ofs(buf, (int) len, &bits);
673 if (out) assert(bits >= BUP_BLOBBITS);
674 return Py_BuildValue("ii", out, bits);
678 static PyObject *bitmatch(PyObject *self, PyObject *args)
680 unsigned char *buf1 = NULL, *buf2 = NULL;
681 Py_ssize_t len1 = 0, len2 = 0;
685 if (!PyArg_ParseTuple(args, rbuf_argf rbuf_argf, &buf1, &len1, &buf2, &len2))
689 for (byte = 0; byte < len1 && byte < len2; byte++)
691 int b1 = buf1[byte], b2 = buf2[byte];
694 for (bit = 0; bit < 8; bit++)
695 if ( (b1 & (0x80 >> bit)) != (b2 & (0x80 >> bit)) )
701 assert(byte <= (INT_MAX >> 3));
702 return Py_BuildValue("i", byte*8 + bit);
706 static PyObject *firstword(PyObject *self, PyObject *args)
708 unsigned char *buf = NULL;
712 if (!PyArg_ParseTuple(args, rbuf_argf, &buf, &len))
718 v = ntohl(*(uint32_t *)buf);
719 return PyLong_FromUnsignedLong(v);
723 #define BLOOM2_HEADERLEN 16
725 static void to_bloom_address_bitmask4(const unsigned char *buf,
726 const int nbits, uint64_t *v, unsigned char *bitmask)
732 memcpy(&high, buf, 4);
733 mask = (1<<nbits) - 1;
734 raw = (((uint64_t)ntohl(high) << 8) | buf[4]);
735 bit = (raw >> (37-nbits)) & 0x7;
736 *v = (raw >> (40-nbits)) & mask;
740 static void to_bloom_address_bitmask5(const unsigned char *buf,
741 const int nbits, uint32_t *v, unsigned char *bitmask)
747 memcpy(&high, buf, 4);
748 mask = (1<<nbits) - 1;
750 bit = (raw >> (29-nbits)) & 0x7;
751 *v = (raw >> (32-nbits)) & mask;
755 #define BLOOM_SET_BIT(name, address, otype) \
756 static void name(unsigned char *bloom, const unsigned char *buf, const int nbits)\
758 unsigned char bitmask;\
760 address(buf, nbits, &v, &bitmask);\
761 bloom[BLOOM2_HEADERLEN+v] |= bitmask;\
763 BLOOM_SET_BIT(bloom_set_bit4, to_bloom_address_bitmask4, uint64_t)
764 BLOOM_SET_BIT(bloom_set_bit5, to_bloom_address_bitmask5, uint32_t)
767 #define BLOOM_GET_BIT(name, address, otype) \
768 static int name(const unsigned char *bloom, const unsigned char *buf, const int nbits)\
770 unsigned char bitmask;\
772 address(buf, nbits, &v, &bitmask);\
773 return bloom[BLOOM2_HEADERLEN+v] & bitmask;\
775 BLOOM_GET_BIT(bloom_get_bit4, to_bloom_address_bitmask4, uint64_t)
776 BLOOM_GET_BIT(bloom_get_bit5, to_bloom_address_bitmask5, uint32_t)
779 static PyObject *bloom_add(PyObject *self, PyObject *args)
781 Py_buffer bloom, sha;
782 int nbits = 0, k = 0;
783 if (!PyArg_ParseTuple(args, wbuf_argf wbuf_argf "ii",
784 &bloom, &sha, &nbits, &k))
787 PyObject *result = NULL;
789 if (bloom.len < 16+(1<<nbits) || sha.len % 20 != 0)
790 goto clean_and_return;
795 goto clean_and_return;
796 unsigned char *cur = sha.buf;
798 for (end = cur + sha.len; cur < end; cur += 20/k)
799 bloom_set_bit5(bloom.buf, cur, nbits);
804 goto clean_and_return;
805 unsigned char *cur = sha.buf;
806 unsigned char *end = cur + sha.len;
807 for (; cur < end; cur += 20/k)
808 bloom_set_bit4(bloom.buf, cur, nbits);
811 goto clean_and_return;
813 result = Py_BuildValue("n", sha.len / 20);
816 PyBuffer_Release(&bloom);
817 PyBuffer_Release(&sha);
821 static PyObject *bloom_contains(PyObject *self, PyObject *args)
824 unsigned char *sha = NULL;
826 int nbits = 0, k = 0;
827 if (!PyArg_ParseTuple(args, wbuf_argf rbuf_argf "ii",
828 &bloom, &sha, &len, &nbits, &k))
831 PyObject *result = NULL;
834 goto clean_and_return;
839 goto clean_and_return;
842 for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
843 if (!bloom_get_bit5(bloom.buf, sha, nbits))
845 result = Py_BuildValue("Oi", Py_None, steps);
846 goto clean_and_return;
852 goto clean_and_return;
855 for (steps = 1, end = sha + 20; sha < end; sha += 20/k, steps++)
856 if (!bloom_get_bit4(bloom.buf, sha, nbits))
858 result = Py_BuildValue("Oi", Py_None, steps);
859 goto clean_and_return;
863 goto clean_and_return;
865 result = Py_BuildValue("ii", 1, k);
868 PyBuffer_Release(&bloom);
873 static uint32_t _extract_bits(unsigned char *buf, int nbits)
877 mask = (1<<nbits) - 1;
878 v = ntohl(*(uint32_t *)buf);
879 v = (v >> (32-nbits)) & mask;
884 static PyObject *extract_bits(PyObject *self, PyObject *args)
886 unsigned char *buf = NULL;
890 if (!PyArg_ParseTuple(args, rbuf_argf "i", &buf, &len, &nbits))
896 return PyLong_FromUnsignedLong(_extract_bits(buf, nbits));
901 unsigned char bytes[20];
904 static inline int _cmp_sha(const struct sha *sha1, const struct sha *sha2)
906 return memcmp(sha1->bytes, sha2->bytes, sizeof(sha1->bytes));
919 static void _fix_idx_order(struct idx **idxs, Py_ssize_t *last_i)
922 Py_ssize_t low, mid, high;
926 if (idxs[*last_i]->cur >= idxs[*last_i]->end)
928 idxs[*last_i] = NULL;
941 mid = (low + high) / 2;
942 c = _cmp_sha(idx->cur, idxs[mid]->cur);
954 memmove(&idxs[mid+1], &idxs[mid], (*last_i-mid)*sizeof(struct idx *));
959 static uint32_t _get_idx_i(struct idx *idx)
961 if (idx->cur_name == NULL)
962 return idx->name_base;
963 return ntohl(*idx->cur_name) + idx->name_base;
966 #define MIDX4_HEADERLEN 12
968 static PyObject *merge_into(PyObject *self, PyObject *args)
970 struct sha *sha_ptr, *sha_start = NULL;
971 uint32_t *table_ptr, *name_ptr, *name_start;
974 uint32_t count, prefix;
979 PyObject *py_total, *ilist = NULL;
980 if (!PyArg_ParseTuple(args, wbuf_argf "iOO",
981 &fmap, &bits, &py_total, &ilist))
984 PyObject *result = NULL;
985 struct idx **idxs = NULL;
986 Py_ssize_t num_i = 0;
987 int *idx_buf_init = NULL;
988 Py_buffer *idx_buf = NULL;
990 if (!bup_uint_from_py(&total, py_total, "total"))
991 goto clean_and_return;
993 num_i = PyList_Size(ilist);
995 if (!(idxs = checked_malloc(num_i, sizeof(struct idx *))))
996 goto clean_and_return;
997 if (!(idx_buf_init = checked_calloc(num_i, sizeof(int))))
998 goto clean_and_return;
999 if (!(idx_buf = checked_malloc(num_i, sizeof(Py_buffer))))
1000 goto clean_and_return;
1002 for (i = 0; i < num_i; i++)
1004 long len, sha_ofs, name_map_ofs;
1005 if (!(idxs[i] = checked_malloc(1, sizeof(struct idx))))
1006 goto clean_and_return;
1007 PyObject *itup = PyList_GetItem(ilist, i);
1008 if (!PyArg_ParseTuple(itup, wbuf_argf "llli",
1009 &(idx_buf[i]), &len, &sha_ofs, &name_map_ofs,
1010 &idxs[i]->name_base))
1012 idx_buf_init[i] = 1;
1013 idxs[i]->map = idx_buf[i].buf;
1014 idxs[i]->bytes = idx_buf[i].len;
1015 idxs[i]->cur = (struct sha *)&idxs[i]->map[sha_ofs];
1016 idxs[i]->end = &idxs[i]->cur[len];
1018 idxs[i]->cur_name = (uint32_t *)&idxs[i]->map[name_map_ofs];
1020 idxs[i]->cur_name = NULL;
1022 table_ptr = (uint32_t *) &((unsigned char *) fmap.buf)[MIDX4_HEADERLEN];
1023 sha_start = sha_ptr = (struct sha *)&table_ptr[1<<bits];
1024 name_start = name_ptr = (uint32_t *)&sha_ptr[total];
1026 Py_ssize_t last_i = num_i - 1;
1032 uint32_t new_prefix;
1033 if (count % 102424 == 0 && get_state(self)->istty2)
1034 fprintf(stderr, "midx: writing %.2f%% (%d/%d)\r",
1035 count*100.0/total, count, total);
1037 new_prefix = _extract_bits((unsigned char *)idx->cur, bits);
1038 while (prefix < new_prefix)
1039 table_ptr[prefix++] = htonl(count);
1040 memcpy(sha_ptr++, idx->cur, sizeof(struct sha));
1041 *name_ptr++ = htonl(_get_idx_i(idx));
1043 if (idx->cur_name != NULL)
1045 _fix_idx_order(idxs, &last_i);
1048 while (prefix < ((uint32_t) 1 << bits))
1049 table_ptr[prefix++] = htonl(count);
1050 assert(count == total);
1051 assert(prefix == ((uint32_t) 1 << bits));
1052 assert(sha_ptr == sha_start+count);
1053 assert(name_ptr == name_start+count);
1055 result = PyLong_FromUnsignedLong(count);
1060 for (i = 0; i < num_i; i++)
1061 if (idx_buf_init[i])
1062 PyBuffer_Release(&(idx_buf[i]));
1068 for (i = 0; i < num_i; i++)
1072 PyBuffer_Release(&fmap);
1076 #define FAN_ENTRIES 256
1078 static PyObject *write_idx(PyObject *self, PyObject *args)
1080 char *filename = NULL;
1081 PyObject *py_total, *idx = NULL;
1083 unsigned int total = 0;
1085 int i, j, ofs64_count;
1086 uint32_t *fan_ptr, *crc_ptr, *ofs_ptr;
1087 uint64_t *ofs64_ptr;
1088 struct sha *sha_ptr;
1091 if (!PyArg_ParseTuple(args, cstr_argf wbuf_argf "OO",
1092 &filename, &fmap, &idx, &py_total))
1095 PyObject *result = NULL;
1097 if (!bup_uint_from_py(&total, py_total, "total"))
1098 goto clean_and_return;
1100 if (PyList_Size (idx) != FAN_ENTRIES) // Check for list of the right length.
1102 result = PyErr_Format (PyExc_TypeError, "idx must contain %d entries",
1104 goto clean_and_return;
1107 const char idx_header[] = "\377tOc\0\0\0\002";
1108 memcpy (fmap.buf, idx_header, sizeof(idx_header) - 1);
1110 fan_ptr = (uint32_t *)&((unsigned char *)fmap.buf)[sizeof(idx_header) - 1];
1111 sha_ptr = (struct sha *)&fan_ptr[FAN_ENTRIES];
1112 crc_ptr = (uint32_t *)&sha_ptr[total];
1113 ofs_ptr = (uint32_t *)&crc_ptr[total];
1114 ofs64_ptr = (uint64_t *)&ofs_ptr[total];
1118 for (i = 0; i < FAN_ENTRIES; ++i)
1121 part = PyList_GET_ITEM(idx, i);
1123 plen = PyList_GET_SIZE(part);
1124 if (plen > UINT32_MAX || UINT32_MAX - count < plen) {
1125 PyErr_Format(PyExc_OverflowError, "too many objects in index part");
1126 goto clean_and_return;
1128 count += (uint32_t) plen;
1129 *fan_ptr++ = htonl(count);
1130 for (j = 0; j < plen; ++j)
1132 unsigned char *sha = NULL;
1133 Py_ssize_t sha_len = 0;
1134 PyObject *crc_py, *ofs_py;
1136 unsigned PY_LONG_LONG ofs_ull;
1138 if (!PyArg_ParseTuple(PyList_GET_ITEM(part, j), rbuf_argf "OO",
1139 &sha, &sha_len, &crc_py, &ofs_py))
1140 goto clean_and_return;
1141 if(!bup_uint_from_py(&crc, crc_py, "crc"))
1142 goto clean_and_return;
1143 if(!bup_ullong_from_py(&ofs_ull, ofs_py, "ofs"))
1144 goto clean_and_return;
1145 assert(crc <= UINT32_MAX);
1146 assert(ofs_ull <= UINT64_MAX);
1148 if (sha_len != sizeof(struct sha))
1149 goto clean_and_return;
1150 memcpy(sha_ptr++, sha, sizeof(struct sha));
1151 *crc_ptr++ = htonl(crc);
1152 if (ofs > 0x7fffffff)
1154 *ofs64_ptr++ = htonll(ofs);
1155 ofs = 0x80000000 | ofs64_count++;
1157 *ofs_ptr++ = htonl((uint32_t)ofs);
1161 int rc = msync(fmap.buf, fmap.len, MS_ASYNC);
1164 result = PyErr_SetFromErrnoWithFilename(PyExc_IOError, filename);
1165 goto clean_and_return;
1168 result = PyLong_FromUnsignedLong(count);
1171 PyBuffer_Release(&fmap);
1176 // I would have made this a lower-level function that just fills in a buffer
1177 // with random values, and then written those values from python. But that's
1178 // about 20% slower in my tests, and since we typically generate random
1179 // numbers for benchmarking other parts of bup, any slowness in generating
1180 // random bytes will make our benchmarks inaccurate. Plus nobody wants
1181 // pseudorandom bytes much except for this anyway.
1182 static PyObject *write_random(PyObject *self, PyObject *args)
1184 uint32_t buf[1024/4];
1185 int fd = -1, seed = 0, verbose = 0;
1187 long long len = 0, kbytes = 0, written = 0;
1189 if (!PyArg_ParseTuple(args, "iLii", &fd, &len, &seed, &verbose))
1194 for (kbytes = 0; kbytes < len/1024; kbytes++)
1197 for (i = 0; i < sizeof(buf)/sizeof(buf[0]); i++)
1198 buf[i] = (uint32_t) random();
1199 ret = write(fd, buf, sizeof(buf));
1203 if (ret < (int)sizeof(buf))
1205 if (verbose && kbytes/1024 > 0 && !(kbytes%1024))
1206 fprintf(stderr, "Random: %lld Mbytes\r", kbytes/1024);
1209 // handle non-multiples of 1024
1213 for (i = 0; i < sizeof(buf)/sizeof(buf[0]); i++)
1214 buf[i] = (uint32_t) random();
1215 ret = write(fd, buf, len % 1024);
1221 if (kbytes/1024 > 0)
1222 fprintf(stderr, "Random: %lld Mbytes, done.\n", kbytes/1024);
1223 return Py_BuildValue("L", written);
1227 static PyObject *random_sha(PyObject *self, PyObject *args)
1229 static int seeded = 0;
1230 uint32_t shabuf[20/4];
1235 assert(sizeof(shabuf) == 20);
1236 srandom((unsigned int) time(NULL));
1240 if (!PyArg_ParseTuple(args, ""))
1243 memset(shabuf, 0, sizeof(shabuf));
1244 for (i=0; i < 20/4; i++)
1245 shabuf[i] = (uint32_t) random();
1246 return Py_BuildValue(rbuf_argf, shabuf, 20);
1250 static int _open_noatime(const char *filename, int attrs)
1252 int attrs_noatime, fd;
1255 attrs |= O_NOFOLLOW;
1258 attrs |= O_LARGEFILE;
1260 attrs_noatime = attrs;
1262 attrs_noatime |= O_NOATIME;
1264 fd = open(filename, attrs_noatime);
1265 if (fd < 0 && errno == EPERM)
1267 // older Linux kernels would return EPERM if you used O_NOATIME
1268 // and weren't the file's owner. This pointless restriction was
1269 // relaxed eventually, but we have to handle it anyway.
1270 // (VERY old kernels didn't recognized O_NOATIME, but they would
1271 // just harmlessly ignore it, so this branch won't trigger)
1272 fd = open(filename, attrs);
1278 static PyObject *open_noatime(PyObject *self, PyObject *args)
1280 char *filename = NULL;
1282 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1284 fd = _open_noatime(filename, 0);
1286 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1287 return Py_BuildValue("i", fd);
1291 static PyObject *fadvise_done(PyObject *self, PyObject *args)
1294 long long llofs, lllen = 0;
1295 if (!PyArg_ParseTuple(args, "iLL", &fd, &llofs, &lllen))
1298 if (!INTEGRAL_ASSIGNMENT_FITS(&ofs, llofs))
1299 return PyErr_Format(PyExc_OverflowError,
1300 "fadvise offset overflows off_t");
1301 if (!INTEGRAL_ASSIGNMENT_FITS(&len, lllen))
1302 return PyErr_Format(PyExc_OverflowError,
1303 "fadvise length overflows off_t");
1304 #ifdef POSIX_FADV_DONTNEED
1305 posix_fadvise(fd, ofs, len, POSIX_FADV_DONTNEED);
1307 return Py_BuildValue("");
1311 // Currently the Linux kernel and FUSE disagree over the type for
1312 // FS_IOC_GETFLAGS and FS_IOC_SETFLAGS. The kernel actually uses int,
1313 // but FUSE chose long (matching the declaration in linux/fs.h). So
1314 // if you use int, and then traverse a FUSE filesystem, you may
1315 // corrupt the stack. But if you use long, then you may get invalid
1316 // results on big-endian systems.
1318 // For now, we just use long, and then disable Linux attrs entirely
1319 // (with a warning) in helpers.py on systems that are affected.
1321 #ifdef BUP_HAVE_FILE_ATTRS
1322 static PyObject *bup_get_linux_file_attr(PyObject *self, PyObject *args)
1329 if (!PyArg_ParseTuple(args, cstr_argf, &path))
1332 fd = _open_noatime(path, O_NONBLOCK);
1334 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1336 attr = 0; // Handle int/long mismatch (see above)
1337 rc = ioctl(fd, FS_IOC_GETFLAGS, &attr);
1341 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1344 assert(attr <= UINT_MAX); // Kernel type is actually int
1345 return PyLong_FromUnsignedLong(attr);
1347 #endif /* def BUP_HAVE_FILE_ATTRS */
1351 #ifdef BUP_HAVE_FILE_ATTRS
1352 static PyObject *bup_set_linux_file_attr(PyObject *self, PyObject *args)
1355 unsigned long orig_attr;
1361 if (!PyArg_ParseTuple(args, cstr_argf "O", &path, &py_attr))
1364 if (!bup_uint_from_py(&attr, py_attr, "attr"))
1367 fd = open(path, O_RDONLY | O_NONBLOCK | O_LARGEFILE | O_NOFOLLOW);
1369 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1371 // Restrict attr to modifiable flags acdeijstuADST -- see
1372 // chattr(1) and the e2fsprogs source. Letter to flag mapping is
1373 // in pf.c flags_array[].
1374 attr &= FS_APPEND_FL | FS_COMPR_FL | FS_NODUMP_FL | FS_EXTENT_FL
1375 | FS_IMMUTABLE_FL | FS_JOURNAL_DATA_FL | FS_SECRM_FL | FS_NOTAIL_FL
1376 | FS_UNRM_FL | FS_NOATIME_FL | FS_DIRSYNC_FL | FS_SYNC_FL
1377 | FS_TOPDIR_FL | FS_NOCOW_FL;
1379 // The extents flag can't be removed, so don't (see chattr(1) and chattr.c).
1380 orig_attr = 0; // Handle int/long mismatch (see above)
1381 rc = ioctl(fd, FS_IOC_GETFLAGS, &orig_attr);
1385 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1387 assert(orig_attr <= UINT_MAX); // Kernel type is actually int
1388 attr |= ((unsigned int) orig_attr) & FS_EXTENT_FL;
1390 rc = ioctl(fd, FS_IOC_SETFLAGS, &attr);
1394 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1398 return Py_BuildValue("O", Py_None);
1400 #endif /* def BUP_HAVE_FILE_ATTRS */
1403 #ifndef HAVE_UTIMENSAT
1405 #error "cannot find utimensat or utimes()"
1407 #ifndef HAVE_LUTIMES
1408 #error "cannot find utimensat or lutimes()"
1412 #define ASSIGN_PYLONG_TO_INTEGRAL(dest, pylong, overflow) \
1416 const long long lltmp = PyLong_AsLongLong(pylong); \
1417 if (lltmp == -1 && PyErr_Occurred()) \
1419 if (PyErr_ExceptionMatches(PyExc_OverflowError)) \
1421 const unsigned long long ulltmp = PyLong_AsUnsignedLongLong(pylong); \
1422 if (ulltmp == (unsigned long long) -1 && PyErr_Occurred()) \
1424 if (PyErr_ExceptionMatches(PyExc_OverflowError)) \
1430 if (INTEGRAL_ASSIGNMENT_FITS((dest), ulltmp)) \
1438 if (INTEGRAL_ASSIGNMENT_FITS((dest), lltmp)) \
1447 #ifdef HAVE_UTIMENSAT
1449 static PyObject *bup_utimensat(PyObject *self, PyObject *args)
1454 PyObject *access_py, *modification_py;
1455 struct timespec ts[2];
1457 if (!PyArg_ParseTuple(args, "i" cstr_argf "((Ol)(Ol))i",
1460 &access_py, &(ts[0].tv_nsec),
1461 &modification_py, &(ts[1].tv_nsec),
1466 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(ts[0].tv_sec), access_py, &overflow))
1469 PyErr_SetString(PyExc_ValueError,
1470 "unable to convert access time seconds for utimensat");
1473 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(ts[1].tv_sec), modification_py, &overflow))
1476 PyErr_SetString(PyExc_ValueError,
1477 "unable to convert modification time seconds for utimensat");
1480 rc = utimensat(fd, path, ts, flag);
1482 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1484 return Py_BuildValue("O", Py_None);
1487 #endif /* def HAVE_UTIMENSAT */
1490 #if defined(HAVE_UTIMES) || defined(HAVE_LUTIMES)
1492 static int bup_parse_xutimes_args(char **path,
1493 struct timeval tv[2],
1496 PyObject *access_py, *modification_py;
1497 long long access_us, modification_us; // POSIX guarantees tv_usec is signed.
1499 if (!PyArg_ParseTuple(args, cstr_argf "((OL)(OL))",
1501 &access_py, &access_us,
1502 &modification_py, &modification_us))
1506 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(tv[0].tv_sec), access_py, &overflow))
1509 PyErr_SetString(PyExc_ValueError, "unable to convert access time seconds to timeval");
1512 if (!INTEGRAL_ASSIGNMENT_FITS(&(tv[0].tv_usec), access_us))
1514 PyErr_SetString(PyExc_ValueError, "unable to convert access time nanoseconds to timeval");
1517 if (!ASSIGN_PYLONG_TO_INTEGRAL(&(tv[1].tv_sec), modification_py, &overflow))
1520 PyErr_SetString(PyExc_ValueError, "unable to convert modification time seconds to timeval");
1523 if (!INTEGRAL_ASSIGNMENT_FITS(&(tv[1].tv_usec), modification_us))
1525 PyErr_SetString(PyExc_ValueError, "unable to convert modification time nanoseconds to timeval");
1531 #endif /* defined(HAVE_UTIMES) || defined(HAVE_LUTIMES) */
1535 static PyObject *bup_utimes(PyObject *self, PyObject *args)
1538 struct timeval tv[2];
1539 if (!bup_parse_xutimes_args(&path, tv, args))
1541 int rc = utimes(path, tv);
1543 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1544 return Py_BuildValue("O", Py_None);
1546 #endif /* def HAVE_UTIMES */
1550 static PyObject *bup_lutimes(PyObject *self, PyObject *args)
1553 struct timeval tv[2];
1554 if (!bup_parse_xutimes_args(&path, tv, args))
1556 int rc = lutimes(path, tv);
1558 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, path);
1560 return Py_BuildValue("O", Py_None);
1562 #endif /* def HAVE_LUTIMES */
1565 #ifdef HAVE_STAT_ST_ATIM
1566 # define BUP_STAT_ATIME_NS(st) (st)->st_atim.tv_nsec
1567 # define BUP_STAT_MTIME_NS(st) (st)->st_mtim.tv_nsec
1568 # define BUP_STAT_CTIME_NS(st) (st)->st_ctim.tv_nsec
1569 #elif defined HAVE_STAT_ST_ATIMENSEC
1570 # define BUP_STAT_ATIME_NS(st) (st)->st_atimespec.tv_nsec
1571 # define BUP_STAT_MTIME_NS(st) (st)->st_mtimespec.tv_nsec
1572 # define BUP_STAT_CTIME_NS(st) (st)->st_ctimespec.tv_nsec
1574 # define BUP_STAT_ATIME_NS(st) 0
1575 # define BUP_STAT_MTIME_NS(st) 0
1576 # define BUP_STAT_CTIME_NS(st) 0
1580 #pragma clang diagnostic push
1581 #pragma clang diagnostic ignored "-Wtautological-compare" // For INTEGER_TO_PY().
1583 static PyObject *stat_struct_to_py(const struct stat *st,
1584 const char *filename,
1587 // We can check the known (via POSIX) signed and unsigned types at
1588 // compile time, but not (easily) the unspecified types, so handle
1589 // those via INTEGER_TO_PY(). Assumes ns values will fit in a
1591 return Py_BuildValue("NKNNNNNL(Nl)(Nl)(Nl)",
1592 INTEGER_TO_PY(st->st_mode),
1593 (unsigned PY_LONG_LONG) st->st_ino,
1594 INTEGER_TO_PY(st->st_dev),
1595 INTEGER_TO_PY(st->st_nlink),
1596 INTEGER_TO_PY(st->st_uid),
1597 INTEGER_TO_PY(st->st_gid),
1598 INTEGER_TO_PY(st->st_rdev),
1599 (PY_LONG_LONG) st->st_size,
1600 INTEGER_TO_PY(st->st_atime),
1601 (long) BUP_STAT_ATIME_NS(st),
1602 INTEGER_TO_PY(st->st_mtime),
1603 (long) BUP_STAT_MTIME_NS(st),
1604 INTEGER_TO_PY(st->st_ctime),
1605 (long) BUP_STAT_CTIME_NS(st));
1608 #pragma clang diagnostic pop // ignored "-Wtautological-compare"
1610 static PyObject *bup_stat(PyObject *self, PyObject *args)
1615 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1619 rc = stat(filename, &st);
1621 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1622 return stat_struct_to_py(&st, filename, 0);
1626 static PyObject *bup_lstat(PyObject *self, PyObject *args)
1631 if (!PyArg_ParseTuple(args, cstr_argf, &filename))
1635 rc = lstat(filename, &st);
1637 return PyErr_SetFromErrnoWithFilename(PyExc_OSError, filename);
1638 return stat_struct_to_py(&st, filename, 0);
1642 static PyObject *bup_fstat(PyObject *self, PyObject *args)
1646 if (!PyArg_ParseTuple(args, "i", &fd))
1650 rc = fstat(fd, &st);
1652 return PyErr_SetFromErrno(PyExc_OSError);
1653 return stat_struct_to_py(&st, NULL, fd);
1657 #ifdef HAVE_TM_TM_GMTOFF
1658 static PyObject *bup_localtime(PyObject *self, PyObject *args)
1662 if (!PyArg_ParseTuple(args, "L", &lltime))
1664 if (!INTEGRAL_ASSIGNMENT_FITS(&ttime, lltime))
1665 return PyErr_Format(PyExc_OverflowError, "time value too large");
1669 if(localtime_r(&ttime, &tm) == NULL)
1670 return PyErr_SetFromErrno(PyExc_OSError);
1672 // Match the Python struct_time values.
1673 return Py_BuildValue("[i,i,i,i,i,i,i,i,i,i,s]",
1674 1900 + tm.tm_year, tm.tm_mon + 1, tm.tm_mday,
1675 tm.tm_hour, tm.tm_min, tm.tm_sec,
1676 tm.tm_wday, tm.tm_yday + 1,
1677 tm.tm_isdst, tm.tm_gmtoff, tm.tm_zone);
1679 #endif /* def HAVE_TM_TM_GMTOFF */
1682 #ifdef BUP_MINCORE_BUF_TYPE
1683 static PyObject *bup_mincore(PyObject *self, PyObject *args)
1685 Py_buffer src, dest;
1686 PyObject *py_src_n, *py_src_off, *py_dest_off;
1688 if (!PyArg_ParseTuple(args, cstr_argf "*OOw*O",
1689 &src, &py_src_n, &py_src_off,
1690 &dest, &py_dest_off))
1693 PyObject *result = NULL;
1695 unsigned long long src_n, src_off, dest_off;
1696 if (!(bup_ullong_from_py(&src_n, py_src_n, "src_n")
1697 && bup_ullong_from_py(&src_off, py_src_off, "src_off")
1698 && bup_ullong_from_py(&dest_off, py_dest_off, "dest_off")))
1699 goto clean_and_return;
1701 unsigned long long src_region_end;
1702 if (!uadd(&src_region_end, src_off, src_n)) {
1703 result = PyErr_Format(PyExc_OverflowError, "(src_off + src_n) too large");
1704 goto clean_and_return;
1706 assert(src.len >= 0);
1707 if (src_region_end > (unsigned long long) src.len) {
1708 result = PyErr_Format(PyExc_OverflowError, "region runs off end of src");
1709 goto clean_and_return;
1712 unsigned long long dest_size;
1713 if (!INTEGRAL_ASSIGNMENT_FITS(&dest_size, dest.len)) {
1714 result = PyErr_Format(PyExc_OverflowError, "invalid dest size");
1715 goto clean_and_return;
1717 if (dest_off > dest_size) {
1718 result = PyErr_Format(PyExc_OverflowError, "region runs off end of dest");
1719 goto clean_and_return;
1723 if (!INTEGRAL_ASSIGNMENT_FITS(&length, src_n)) {
1724 result = PyErr_Format(PyExc_OverflowError, "src_n overflows size_t");
1725 goto clean_and_return;
1727 int rc = mincore((void *)(src.buf + src_off), length,
1728 (BUP_MINCORE_BUF_TYPE *) (dest.buf + dest_off));
1730 result = PyErr_SetFromErrno(PyExc_OSError);
1731 goto clean_and_return;
1733 result = Py_BuildValue("O", Py_None);
1736 PyBuffer_Release(&src);
1737 PyBuffer_Release(&dest);
1740 #endif /* def BUP_MINCORE_BUF_TYPE */
1743 static PyObject *tuple_from_cstrs(char **cstrs)
1745 // Assumes list is null terminated
1747 while(cstrs[n] != NULL)
1751 if (!INTEGRAL_ASSIGNMENT_FITS(&sn, n))
1752 return PyErr_Format(PyExc_OverflowError, "string array too large");
1754 PyObject *result = PyTuple_New(sn);
1756 for (i = 0; i < sn; i++)
1758 PyObject *gname = Py_BuildValue(cstr_argf, cstrs[i]);
1764 PyTuple_SET_ITEM(result, i, gname);
1769 static PyObject *appropriate_errno_ex(void)
1773 return PyErr_NoMemory();
1777 // In 3.3 IOError was merged into OSError.
1778 return PyErr_SetFromErrno(PyExc_IOError);
1780 return PyErr_SetFromErrno(PyExc_OSError);
1785 static PyObject *pwd_struct_to_py(const struct passwd *pwd)
1787 // We can check the known (via POSIX) signed and unsigned types at
1788 // compile time, but not (easily) the unspecified types, so handle
1789 // those via INTEGER_TO_PY().
1792 return Py_BuildValue(cstr_argf cstr_argf "OO"
1793 cstr_argf cstr_argf cstr_argf,
1796 INTEGER_TO_PY(pwd->pw_uid),
1797 INTEGER_TO_PY(pwd->pw_gid),
1803 static PyObject *bup_getpwuid(PyObject *self, PyObject *args)
1805 unsigned long long py_uid;
1806 if (!PyArg_ParseTuple(args, "K", &py_uid))
1809 if (!INTEGRAL_ASSIGNMENT_FITS(&uid, py_uid))
1810 return PyErr_Format(PyExc_OverflowError, "uid too large for uid_t");
1813 struct passwd *pwd = getpwuid(uid);
1815 return appropriate_errno_ex();
1816 return pwd_struct_to_py(pwd);
1819 static PyObject *bup_getpwnam(PyObject *self, PyObject *args)
1822 if (!PyArg_ParseTuple(args, "S", &py_name))
1825 char *name = PyBytes_AS_STRING(py_name);
1827 struct passwd *pwd = getpwnam(name);
1829 return appropriate_errno_ex();
1830 return pwd_struct_to_py(pwd);
1833 static PyObject *grp_struct_to_py(const struct group *grp)
1835 // We can check the known (via POSIX) signed and unsigned types at
1836 // compile time, but not (easily) the unspecified types, so handle
1837 // those via INTEGER_TO_PY().
1841 PyObject *members = tuple_from_cstrs(grp->gr_mem);
1842 if (members == NULL)
1844 return Py_BuildValue(cstr_argf cstr_argf "OO",
1847 INTEGER_TO_PY(grp->gr_gid),
1851 static PyObject *bup_getgrgid(PyObject *self, PyObject *args)
1853 unsigned long long py_gid;
1854 if (!PyArg_ParseTuple(args, "K", &py_gid))
1857 if (!INTEGRAL_ASSIGNMENT_FITS(&gid, py_gid))
1858 return PyErr_Format(PyExc_OverflowError, "gid too large for gid_t");
1861 struct group *grp = getgrgid(gid);
1863 return appropriate_errno_ex();
1864 return grp_struct_to_py(grp);
1867 static PyObject *bup_getgrnam(PyObject *self, PyObject *args)
1870 if (!PyArg_ParseTuple(args, "S", &py_name))
1873 char *name = PyBytes_AS_STRING(py_name);
1875 struct group *grp = getgrnam(name);
1877 return appropriate_errno_ex();
1878 return grp_struct_to_py(grp);
1882 static PyObject *bup_gethostname(PyObject *mod, PyObject *ignore)
1884 #ifdef HOST_NAME_MAX
1885 char buf[HOST_NAME_MAX + 1] = {};
1887 /* 'SUSv2 guarantees that "Host names are limited to 255 bytes".' */
1891 if (gethostname(buf, sizeof(buf) - 1))
1892 return PyErr_SetFromErrno(PyExc_IOError);
1893 return PyBytes_FromString(buf);
1897 #ifdef BUP_HAVE_READLINE
1899 static char *cstr_from_bytes(PyObject *bytes)
1903 int rc = PyBytes_AsStringAndSize(bytes, &buf, &length);
1906 char *result = checked_malloc(length, sizeof(char));
1909 memcpy(result, buf, length);
1913 static char **cstrs_from_seq(PyObject *seq)
1915 char **result = NULL;
1916 seq = PySequence_Fast(seq, "Cannot convert sequence items to C strings");
1920 const Py_ssize_t len = PySequence_Fast_GET_SIZE(seq);
1921 if (len > PY_SSIZE_T_MAX - 1) {
1922 PyErr_Format(PyExc_OverflowError,
1923 "Sequence length %zd too large for conversion to C array",
1927 result = checked_malloc(len + 1, sizeof(char *));
1931 for (i = 0; i < len; i++)
1933 PyObject *item = PySequence_Fast_GET_ITEM(seq, i);
1935 goto abandon_result;
1936 result[i] = cstr_from_bytes(item);
1939 goto abandon_result;
1957 static char* our_word_break_chars = NULL;
1960 bup_set_completer_word_break_characters(PyObject *self, PyObject *args)
1963 if (!PyArg_ParseTuple(args, cstr_argf, &bytes))
1965 char *prev = our_word_break_chars;
1966 char *next = strdup(bytes);
1968 return PyErr_NoMemory();
1969 our_word_break_chars = next;
1970 rl_completer_word_break_characters = next;
1977 bup_get_completer_word_break_characters(PyObject *self, PyObject *args)
1979 if (!PyArg_ParseTuple(args, ""))
1981 return PyBytes_FromString(rl_completer_word_break_characters);
1984 static PyObject *bup_get_line_buffer(PyObject *self, PyObject *args)
1986 if (!PyArg_ParseTuple(args, ""))
1988 return PyBytes_FromString(rl_line_buffer);
1992 bup_parse_and_bind(PyObject *self, PyObject *args)
1995 if (!PyArg_ParseTuple(args, cstr_argf ":parse_and_bind", &bytes))
1997 char *tmp = strdup(bytes); // Because it may modify the arg
1999 return PyErr_NoMemory();
2000 int rc = rl_parse_and_bind(tmp);
2003 return PyErr_Format(PyExc_OSError,
2004 "system rl_parse_and_bind failed (%d)", rc);
2009 static PyObject *py_on_attempted_completion;
2010 static char **prev_completions;
2012 static char **on_attempted_completion(const char *text, int start, int end)
2014 if (!py_on_attempted_completion)
2017 char **result = NULL;
2018 PyObject *py_result = PyObject_CallFunction(py_on_attempted_completion,
2023 if (py_result != Py_None) {
2024 result = cstrs_from_seq(py_result);
2025 free(prev_completions);
2026 prev_completions = result;
2028 Py_DECREF(py_result);
2033 bup_set_attempted_completion_function(PyObject *self, PyObject *args)
2035 PyObject *completer;
2036 if (!PyArg_ParseTuple(args, "O", &completer))
2039 PyObject *prev = py_on_attempted_completion;
2040 if (completer == Py_None)
2042 py_on_attempted_completion = NULL;
2043 rl_attempted_completion_function = NULL;
2045 py_on_attempted_completion = completer;
2046 rl_attempted_completion_function = on_attempted_completion;
2047 Py_INCREF(completer);
2054 static PyObject *py_on_completion_entry;
2056 static char *on_completion_entry(const char *text, int state)
2058 if (!py_on_completion_entry)
2061 PyObject *py_result = PyObject_CallFunction(py_on_completion_entry,
2062 cstr_argf "i", text, state);
2065 char *result = (py_result == Py_None) ? NULL : cstr_from_bytes(py_result);
2066 Py_DECREF(py_result);
2071 bup_set_completion_entry_function(PyObject *self, PyObject *args)
2073 PyObject *completer;
2074 if (!PyArg_ParseTuple(args, "O", &completer))
2077 PyObject *prev = py_on_completion_entry;
2078 if (completer == Py_None) {
2079 py_on_completion_entry = NULL;
2080 rl_completion_entry_function = NULL;
2082 py_on_completion_entry = completer;
2083 rl_completion_entry_function = on_completion_entry;
2084 Py_INCREF(completer);
2091 bup_readline(PyObject *self, PyObject *args)
2094 if (!PyArg_ParseTuple(args, cstr_argf, &prompt))
2096 char *line = readline(prompt);
2098 return PyErr_Format(PyExc_EOFError, "readline EOF");
2099 PyObject *result = PyBytes_FromString(line);
2104 #endif // defined BUP_HAVE_READLINE
2106 #if defined(HAVE_SYS_ACL_H) && \
2107 defined(HAVE_ACL_LIBACL_H) && \
2108 defined(HAVE_ACL_EXTENDED_FILE) && \
2109 defined(HAVE_ACL_GET_FILE) && \
2110 defined(HAVE_ACL_TO_ANY_TEXT) && \
2111 defined(HAVE_ACL_FROM_TEXT) && \
2112 defined(HAVE_ACL_SET_FILE)
2113 #define ACL_SUPPORT 1
2114 #include <sys/acl.h>
2115 #include <acl/libacl.h>
2119 // -1 for errors, with python exception set
2120 // -2 for ignored errors (not supported)
2121 static int bup_read_acl_to_text(const char *name, acl_type_t type,
2122 char **txt, char **num)
2126 acl = acl_get_file(name, type);
2128 if (errno == EOPNOTSUPP || errno == ENOSYS)
2130 PyErr_SetFromErrno(PyExc_IOError);
2135 *txt = acl_to_any_text(acl, "", '\n', TEXT_ABBREVIATE);
2137 *num = acl_to_any_text(acl, "", '\n', TEXT_ABBREVIATE | TEXT_NUMERIC_IDS);
2142 if (errno == ENOMEM)
2145 PyErr_SetFromErrno(PyExc_IOError);
2148 acl_free((acl_t)*txt);
2150 acl_free((acl_t)*num);
2155 static PyObject *bup_read_acl(PyObject *self, PyObject *args)
2159 PyObject *ret = NULL;
2160 char *acl_txt = NULL, *acl_num = NULL;
2162 if (!PyArg_ParseTuple(args, cstr_argf "i", &name, &isdir))
2165 if (!acl_extended_file(name))
2168 rv = bup_read_acl_to_text(name, ACL_TYPE_ACCESS, &acl_txt, &acl_num);
2173 char *def_txt = NULL, *def_num = NULL;
2175 rv = bup_read_acl_to_text(name, ACL_TYPE_DEFAULT, &def_txt, &def_num);
2179 ret = Py_BuildValue("[" cstr_argf cstr_argf cstr_argf cstr_argf "]",
2180 acl_txt, acl_num, def_txt, def_num);
2183 acl_free((acl_t)def_txt);
2185 acl_free((acl_t)def_num);
2187 ret = Py_BuildValue("[" cstr_argf cstr_argf "]",
2193 acl_free((acl_t)acl_txt);
2195 acl_free((acl_t)acl_num);
2201 static int bup_apply_acl_string(const char *name, const char *s)
2203 acl_t acl = acl_from_text(s);
2207 PyErr_SetFromErrno(PyExc_IOError);
2211 if (acl_set_file(name, ACL_TYPE_ACCESS, acl)) {
2212 PyErr_SetFromErrno(PyExc_IOError);
2221 static PyObject *bup_apply_acl(PyObject *self, PyObject *args)
2223 char *name, *acl, *def = NULL;
2225 if (!PyArg_ParseTuple(args, cstr_argf cstr_argf "|" cstr_argf, &name, &acl, &def))
2228 if (bup_apply_acl_string(name, acl))
2231 if (def && bup_apply_acl_string(name, def))
2238 static PyMethodDef helper_methods[] = {
2239 { "write_sparsely", bup_write_sparsely, METH_VARARGS,
2240 "Write buf excepting zeros at the end. Return trailing zero count." },
2241 { "selftest", selftest, METH_VARARGS,
2242 "Check that the rolling checksum rolls correctly (for unit tests)." },
2243 { "blobbits", blobbits, METH_VARARGS,
2244 "Return the number of bits in the rolling checksum." },
2245 { "splitbuf", splitbuf, METH_VARARGS,
2246 "Split a list of strings based on a rolling checksum." },
2247 { "bitmatch", bitmatch, METH_VARARGS,
2248 "Count the number of matching prefix bits between two strings." },
2249 { "firstword", firstword, METH_VARARGS,
2250 "Return an int corresponding to the first 32 bits of buf." },
2251 { "bloom_contains", bloom_contains, METH_VARARGS,
2252 "Check if a bloom filter of 2^nbits bytes contains an object" },
2253 { "bloom_add", bloom_add, METH_VARARGS,
2254 "Add an object to a bloom filter of 2^nbits bytes" },
2255 { "extract_bits", extract_bits, METH_VARARGS,
2256 "Take the first 'nbits' bits from 'buf' and return them as an int." },
2257 { "merge_into", merge_into, METH_VARARGS,
2258 "Merges a bunch of idx and midx files into a single midx." },
2259 { "write_idx", write_idx, METH_VARARGS,
2260 "Write a PackIdxV2 file from an idx list of lists of tuples" },
2261 { "write_random", write_random, METH_VARARGS,
2262 "Write random bytes to the given file descriptor" },
2263 { "random_sha", random_sha, METH_VARARGS,
2264 "Return a random 20-byte string" },
2265 { "open_noatime", open_noatime, METH_VARARGS,
2266 "open() the given filename for read with O_NOATIME if possible" },
2267 { "fadvise_done", fadvise_done, METH_VARARGS,
2268 "Inform the kernel that we're finished with earlier parts of a file" },
2269 #ifdef BUP_HAVE_FILE_ATTRS
2270 { "get_linux_file_attr", bup_get_linux_file_attr, METH_VARARGS,
2271 "Return the Linux attributes for the given file." },
2273 #ifdef BUP_HAVE_FILE_ATTRS
2274 { "set_linux_file_attr", bup_set_linux_file_attr, METH_VARARGS,
2275 "Set the Linux attributes for the given file." },
2277 #ifdef HAVE_UTIMENSAT
2278 { "bup_utimensat", bup_utimensat, METH_VARARGS,
2279 "Change path timestamps with nanosecond precision (POSIX)." },
2282 { "bup_utimes", bup_utimes, METH_VARARGS,
2283 "Change path timestamps with microsecond precision." },
2286 { "bup_lutimes", bup_lutimes, METH_VARARGS,
2287 "Change path timestamps with microsecond precision;"
2288 " don't follow symlinks." },
2290 { "stat", bup_stat, METH_VARARGS,
2291 "Extended version of stat." },
2292 { "lstat", bup_lstat, METH_VARARGS,
2293 "Extended version of lstat." },
2294 { "fstat", bup_fstat, METH_VARARGS,
2295 "Extended version of fstat." },
2296 #ifdef HAVE_TM_TM_GMTOFF
2297 { "localtime", bup_localtime, METH_VARARGS,
2298 "Return struct_time elements plus the timezone offset and name." },
2300 { "bytescmp", bup_bytescmp, METH_VARARGS,
2301 "Return a negative value if x < y, zero if equal, positive otherwise."},
2302 { "cat_bytes", bup_cat_bytes, METH_VARARGS,
2303 "For (x_bytes, x_ofs, x_n, y_bytes, y_ofs, y_n) arguments, return their concatenation."},
2304 #ifdef BUP_MINCORE_BUF_TYPE
2305 { "mincore", bup_mincore, METH_VARARGS,
2306 "For mincore(src, src_n, src_off, dest, dest_off)"
2307 " call the system mincore(src + src_off, src_n, &dest[dest_off])." },
2309 { "getpwuid", bup_getpwuid, METH_VARARGS,
2310 "Return the password database entry for the given numeric user id,"
2311 " as a tuple with all C strings as bytes(), or None if the user does"
2313 { "getpwnam", bup_getpwnam, METH_VARARGS,
2314 "Return the password database entry for the given user name,"
2315 " as a tuple with all C strings as bytes(), or None if the user does"
2317 { "getgrgid", bup_getgrgid, METH_VARARGS,
2318 "Return the group database entry for the given numeric group id,"
2319 " as a tuple with all C strings as bytes(), or None if the group does"
2321 { "getgrnam", bup_getgrnam, METH_VARARGS,
2322 "Return the group database entry for the given group name,"
2323 " as a tuple with all C strings as bytes(), or None if the group does"
2325 { "gethostname", bup_gethostname, METH_NOARGS,
2326 "Return the current hostname (as bytes)" },
2327 #ifdef BUP_HAVE_READLINE
2328 { "set_completion_entry_function", bup_set_completion_entry_function, METH_VARARGS,
2329 "Set rl_completion_entry_function. Called as f(text, state)." },
2330 { "set_attempted_completion_function", bup_set_attempted_completion_function, METH_VARARGS,
2331 "Set rl_attempted_completion_function. Called as f(text, start, end)." },
2332 { "parse_and_bind", bup_parse_and_bind, METH_VARARGS,
2333 "Call rl_parse_and_bind." },
2334 { "get_line_buffer", bup_get_line_buffer, METH_NOARGS,
2335 "Return rl_line_buffer." },
2336 { "get_completer_word_break_characters", bup_get_completer_word_break_characters, METH_NOARGS,
2337 "Return rl_completer_word_break_characters." },
2338 { "set_completer_word_break_characters", bup_set_completer_word_break_characters, METH_VARARGS,
2339 "Set rl_completer_word_break_characters." },
2340 { "readline", bup_readline, METH_VARARGS,
2341 "Call readline(prompt)." },
2342 #endif // defined BUP_HAVE_READLINE
2344 { "read_acl", bup_read_acl, METH_VARARGS,
2345 "read_acl(name, isdir)\n\n"
2346 "Read ACLs for the given file/dirname and return the correctly encoded"
2347 " list [txt, num, def_tx, def_num] (the def_* being empty bytestrings"
2348 " unless the second argument 'isdir' is True)." },
2349 { "apply_acl", bup_apply_acl, METH_VARARGS,
2350 "apply_acl(name, acl, def=None)\n\n"
2351 "Given a file/dirname (bytes) and the ACLs to restore, do that." },
2352 #endif /* HAVE_ACLS */
2353 { NULL, NULL, 0, NULL }, // sentinel
2356 static void test_integral_assignment_fits(void)
2358 assert(sizeof(signed short) == sizeof(unsigned short));
2359 assert(sizeof(signed short) < sizeof(signed long long));
2360 assert(sizeof(signed short) < sizeof(unsigned long long));
2361 assert(sizeof(unsigned short) < sizeof(signed long long));
2362 assert(sizeof(unsigned short) < sizeof(unsigned long long));
2363 assert(sizeof(Py_ssize_t) <= sizeof(size_t));
2365 signed short ss, ssmin = SHRT_MIN, ssmax = SHRT_MAX;
2366 unsigned short us, usmax = USHRT_MAX;
2367 signed long long sllmin = LLONG_MIN, sllmax = LLONG_MAX;
2368 unsigned long long ullmax = ULLONG_MAX;
2370 assert(INTEGRAL_ASSIGNMENT_FITS(&ss, ssmax));
2371 assert(INTEGRAL_ASSIGNMENT_FITS(&ss, ssmin));
2372 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, usmax));
2373 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, sllmin));
2374 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, sllmax));
2375 assert(!INTEGRAL_ASSIGNMENT_FITS(&ss, ullmax));
2377 assert(INTEGRAL_ASSIGNMENT_FITS(&us, usmax));
2378 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, ssmin));
2379 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, sllmin));
2380 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, sllmax));
2381 assert(!INTEGRAL_ASSIGNMENT_FITS(&us, ullmax));
2385 static int setup_module(PyObject *m)
2387 // FIXME: migrate these tests to configure, or at least don't
2388 // possibly crash the whole application. Check against the type
2389 // we're going to use when passing to python. Other stat types
2390 // are tested at runtime.
2391 assert(sizeof(ino_t) <= sizeof(unsigned PY_LONG_LONG));
2392 assert(sizeof(off_t) <= sizeof(PY_LONG_LONG));
2393 assert(sizeof(blksize_t) <= sizeof(PY_LONG_LONG));
2394 assert(sizeof(blkcnt_t) <= sizeof(PY_LONG_LONG));
2395 // Just be sure (relevant when passing timestamps back to Python above).
2396 assert(sizeof(PY_LONG_LONG) <= sizeof(long long));
2397 assert(sizeof(unsigned PY_LONG_LONG) <= sizeof(unsigned long long));
2399 test_integral_assignment_fits();
2401 // Originally required by append_sparse_region()
2404 if (!INTEGRAL_ASSIGNMENT_FITS(&probe, INT_MAX))
2406 fprintf(stderr, "off_t can't hold INT_MAX; please report.\n");
2412 #pragma clang diagnostic push
2413 #pragma clang diagnostic ignored "-Wtautological-compare" // For INTEGER_TO_PY().
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);
2445 #pragma clang diagnostic pop // ignored "-Wtautological-compare"
2447 e = getenv("BUP_FORCE_TTY");
2448 get_state(m)->istty2 = isatty(2) || (atoi(e ? e : "0") & 2);
2454 #if PY_MAJOR_VERSION < 3
2456 PyMODINIT_FUNC init_helpers(void)
2458 PyObject *m = Py_InitModule("_helpers", helper_methods);
2462 if (!setup_module(m))
2469 # else // PY_MAJOR_VERSION >= 3
2471 static struct PyModuleDef helpers_def = {
2472 PyModuleDef_HEAD_INIT,
2478 NULL, // helpers_traverse,
2479 NULL, // helpers_clear,
2483 PyMODINIT_FUNC PyInit__helpers(void)
2485 PyObject *module = PyModule_Create(&helpers_def);
2488 if (!setup_module(module))
2496 #endif // PY_MAJOR_VERSION >= 3