static int
archive_read_format_ar_bid(struct archive_read *a, int best_bid)
{
const void *h;
(void)best_bid; /* UNUSED */
/*
* Verify the 8-byte file signature.
* TODO: Do we need to check more than this?
*/
if ((h = __archive_read_ahead(a, 8, NULL)) == NULL)
return (-1);
if (memcmp(h, "!<arch>\n", 8) == 0) {
return (64);
}
return (-1);
}
static int
mtree_bid(struct archive_read *a, int best_bid)
{
const char *signature = "#mtree";
const char *p;
(void)best_bid; /* UNUSED */
/* Now let's look at the actual header and see if it matches. */
p = __archive_read_ahead(a, strlen(signature), NULL);
if (p == NULL)
return (-1);
if (memcmp(p, signature, strlen(signature)) == 0)
return (8 * (int)strlen(signature));
/*
* There is not a mtree signature. Let's try to detect mtree format.
*/
return (detect_form(a, NULL));
}
static int
tk_archive_read_format_ar_bid(struct archive_read *a)
{
const void *h;
if (a->archive.archive_format != 0 &&
(a->archive.archive_format & ARCHIVE_FORMAT_BASE_MASK) !=
ARCHIVE_FORMAT_AR)
return(0);
/*
* Verify the 8-byte file signature.
* TODO: Do we need to check more than this?
*/
if ((h = __archive_read_ahead(a, 8, NULL)) == NULL)
return (-1);
if (strncmp((const char*)h, "!<arch>\n", 8) == 0) {
return (64);
}
return (-1);
}
static int
archive_read_format_raw_read_data(struct archive_read *a,
const void **buff, size_t *size, int64_t *offset)
{
struct raw_info *info;
ssize_t avail;
info = (struct raw_info *)(a->format->data);
/* Consume the bytes we read last time. */
if (info->unconsumed) {
__archive_read_consume(a, info->unconsumed);
info->unconsumed = 0;
}
if (info->end_of_file)
return (ARCHIVE_EOF);
/* Get whatever bytes are immediately available. */
*buff = __archive_read_ahead(a, 1, &avail);
if (avail > 0) {
/* Return the bytes we just read */
*size = avail;
*offset = info->offset;
info->offset += *size;
info->unconsumed = avail;
return (ARCHIVE_OK);
} else if (0 == avail) {
/* Record and return end-of-file. */
info->end_of_file = 1;
*size = 0;
*offset = info->offset;
return (ARCHIVE_EOF);
} else {
/* Record and return an error. */
*size = 0;
*offset = info->offset;
return ((int)avail);
}
}
/*
* NOTE: if a filename suffix is ".z", it is the file gziped by afio.
* it would be nice that we can show uncompressed file size and we can
* uncompressed file contents automatically, unfortunately we have nothing
* to get a uncompressed file size while reading each header. It means
* we also cannot uncompress file contents under our framework.
*/
static int
header_afiol(struct archive_read *a, struct cpio *cpio,
struct archive_entry *entry, size_t *namelength, size_t *name_pad)
{
const void *h;
const char *header;
a->archive.archive_format = ARCHIVE_FORMAT_CPIO_AFIO_LARGE;
a->archive.archive_format_name = "afio large ASCII";
/* Read fixed-size portion of header. */
h = __archive_read_ahead(a, afiol_header_size, NULL);
if (h == NULL)
return (ARCHIVE_FATAL);
/* Parse out octal fields. */
header = (const char *)h;
archive_entry_set_dev(entry,
(dev_t)atol16(header + afiol_dev_offset, afiol_dev_size));
archive_entry_set_ino(entry, atol16(header + afiol_ino_offset, afiol_ino_size));
archive_entry_set_mode(entry,
(mode_t)atol8(header + afiol_mode_offset, afiol_mode_size));
archive_entry_set_uid(entry, atol16(header + afiol_uid_offset, afiol_uid_size));
archive_entry_set_gid(entry, atol16(header + afiol_gid_offset, afiol_gid_size));
archive_entry_set_nlink(entry,
(unsigned int)atol16(header + afiol_nlink_offset, afiol_nlink_size));
archive_entry_set_rdev(entry,
(dev_t)atol16(header + afiol_rdev_offset, afiol_rdev_size));
archive_entry_set_mtime(entry, atol16(header + afiol_mtime_offset, afiol_mtime_size), 0);
*namelength = (size_t)atol16(header + afiol_namesize_offset, afiol_namesize_size);
*name_pad = 0; /* No padding of filename. */
cpio->entry_bytes_remaining =
atol16(header + afiol_filesize_offset, afiol_filesize_size);
archive_entry_set_size(entry, cpio->entry_bytes_remaining);
cpio->entry_padding = 0;
__archive_read_consume(a, afiol_header_size);
return (ARCHIVE_OK);
}
/*
* Read "uncompressed" data. According to the current specification,
* if ZIP_LENGTH_AT_END is specified, then the size fields in the
* initial file header are supposed to be set to zero. This would, of
* course, make it impossible for us to read the archive, since we
* couldn't determine the end of the file data. Info-ZIP seems to
* include the real size fields both before and after the data in this
* case (the CRC only appears afterwards), so this works as you would
* expect.
*
* Returns ARCHIVE_OK if successful, ARCHIVE_FATAL otherwise, sets
* zip->end_of_entry if it consumes all of the data.
*/
static int
zip_read_data_none(struct archive_read *a, const void **buff,
size_t *size, off_t *offset)
{
struct zip *zip;
ssize_t bytes_avail;
zip = (struct zip *)(a->format->data);
if (zip->entry_bytes_remaining == 0) {
*buff = NULL;
*size = 0;
*offset = zip->entry_offset;
zip->end_of_entry = 1;
return (ARCHIVE_OK);
}
/*
* Note: '1' here is a performance optimization.
* Recall that the decompression layer returns a count of
* available bytes; asking for more than that forces the
* decompressor to combine reads by copying data.
*/
*buff = __archive_read_ahead(a, 1, &bytes_avail);
if (bytes_avail <= 0) {
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
"Truncated ZIP file data");
return (ARCHIVE_FATAL);
}
if (bytes_avail > zip->entry_bytes_remaining)
bytes_avail = zip->entry_bytes_remaining;
__archive_read_consume(a, bytes_avail);
*size = bytes_avail;
*offset = zip->entry_offset;
zip->entry_offset += *size;
zip->entry_bytes_remaining -= *size;
zip->entry_uncompressed_bytes_read += *size;
zip->entry_compressed_bytes_read += *size;
return (ARCHIVE_OK);
}
static int
header_bin_be(struct archive_read *a, struct cpio *cpio,
struct archive_entry *entry, size_t *namelength, size_t *name_pad)
{
const void *h;
const unsigned char *header;
a->archive.archive_format = ARCHIVE_FORMAT_CPIO_BIN_BE;
a->archive.archive_format_name = "cpio (big-endian binary)";
/* Read fixed-size portion of header. */
h = __archive_read_ahead(a, bin_header_size, NULL);
if (h == NULL) {
archive_set_error(&a->archive, 0,
"End of file trying to read next cpio header");
return (ARCHIVE_FATAL);
}
/* Parse out binary fields. */
header = (const unsigned char *)h;
archive_entry_set_dev(entry, header[bin_dev_offset] * 256 + header[bin_dev_offset + 1]);
archive_entry_set_ino(entry, header[bin_ino_offset] * 256 + header[bin_ino_offset + 1]);
archive_entry_set_mode(entry, header[bin_mode_offset] * 256 + header[bin_mode_offset + 1]);
archive_entry_set_uid(entry, header[bin_uid_offset] * 256 + header[bin_uid_offset + 1]);
archive_entry_set_gid(entry, header[bin_gid_offset] * 256 + header[bin_gid_offset + 1]);
archive_entry_set_nlink(entry, header[bin_nlink_offset] * 256 + header[bin_nlink_offset + 1]);
archive_entry_set_rdev(entry, header[bin_rdev_offset] * 256 + header[bin_rdev_offset + 1]);
archive_entry_set_mtime(entry, be4(header + bin_mtime_offset), 0);
*namelength = header[bin_namesize_offset] * 256 + header[bin_namesize_offset + 1];
*name_pad = *namelength & 1; /* Pad to even. */
cpio->entry_bytes_remaining = be4(header + bin_filesize_offset);
archive_entry_set_size(entry, cpio->entry_bytes_remaining);
cpio->entry_padding = cpio->entry_bytes_remaining & 1; /* Pad to even. */
__archive_read_consume(a, bin_header_size);
return (ARCHIVE_OK);
}
static int
archive_read_format_cpio_read_data(struct archive_read *a,
const void **buff, size_t *size, int64_t *offset)
{
ssize_t bytes_read;
struct cpio *cpio;
cpio = (struct cpio *)(a->format->data);
if (cpio->entry_bytes_unconsumed) {
__archive_read_consume(a, cpio->entry_bytes_unconsumed);
cpio->entry_bytes_unconsumed = 0;
}
if (cpio->entry_bytes_remaining > 0) {
*buff = __archive_read_ahead(a, 1, &bytes_read);
if (bytes_read <= 0)
return (ARCHIVE_FATAL);
if (bytes_read > cpio->entry_bytes_remaining)
bytes_read = (ssize_t)cpio->entry_bytes_remaining;
*size = bytes_read;
cpio->entry_bytes_unconsumed = bytes_read;
*offset = cpio->entry_offset;
cpio->entry_offset += bytes_read;
cpio->entry_bytes_remaining -= bytes_read;
return (ARCHIVE_OK);
} else {
if (cpio->entry_padding !=
__archive_read_consume(a, cpio->entry_padding)) {
return (ARCHIVE_FATAL);
}
cpio->entry_padding = 0;
*buff = NULL;
*size = 0;
*offset = cpio->entry_offset;
return (ARCHIVE_EOF);
}
}
static int
_warc_read(struct archive_read *a, const void **buf, size_t *bsz, int64_t *off)
{
struct warc_s *w = a->format->data;
const char *rab;
ssize_t nrd;
if (w->cntoff >= w->cntlen) {
eof:
/* it's our lucky day, no work, we can leave early */
*buf = NULL;
*bsz = 0U;
*off = w->cntoff + 4U/*for \r\n\r\n separator*/;
w->unconsumed = 0U;
return (ARCHIVE_EOF);
}
rab = __archive_read_ahead(a, 1U, &nrd);
if (nrd < 0) {
*bsz = 0U;
/* big catastrophe */
return (int)nrd;
} else if (nrd == 0) {
goto eof;
} else if ((size_t)nrd > w->cntlen - w->cntoff) {
/* clamp to content-length */
nrd = w->cntlen - w->cntoff;
}
*off = w->cntoff;
*bsz = nrd;
*buf = rab;
w->cntoff += nrd;
w->unconsumed = (size_t)nrd;
return (ARCHIVE_OK);
}
static int
_warc_rdhdr(struct archive_read *a, struct archive_entry *entry)
{
#define HDR_PROBE_LEN (12U)
struct warc_s *w = a->format->data;
unsigned int ver;
const char *buf;
ssize_t nrd;
const char *eoh;
/* for the file name, saves some strndup()'ing */
warc_string_t fnam;
/* warc record type, not that we really use it a lot */
warc_type_t ftyp;
/* content-length+error monad */
ssize_t cntlen;
/* record time is the WARC-Date time we reinterpret it as ctime */
time_t rtime;
/* mtime is the Last-Modified time which will be the entry's mtime */
time_t mtime;
start_over:
/* just use read_ahead() they keep track of unconsumed
* bits and bobs for us; no need to put an extra shift in
* and reproduce that functionality here */
buf = __archive_read_ahead(a, HDR_PROBE_LEN, &nrd);
if (nrd < 0) {
/* no good */
archive_set_error(
&a->archive, ARCHIVE_ERRNO_MISC,
"Bad record header");
return (ARCHIVE_FATAL);
} else if (buf == NULL) {
/* there should be room for at least WARC/bla\r\n
* must be EOF therefore */
return (ARCHIVE_EOF);
}
/* looks good so far, try and find the end of the header now */
eoh = _warc_find_eoh(buf, nrd);
if (eoh == NULL) {
/* still no good, the header end might be beyond the
* probe we've requested, but then again who'd cram
* so much stuff into the header *and* be 28500-compliant */
archive_set_error(
&a->archive, ARCHIVE_ERRNO_MISC,
"Bad record header");
return (ARCHIVE_FATAL);
} else if ((ver = _warc_rdver(buf, eoh - buf)) > 10000U) {
/* nawww, I wish they promised backward compatibility
* anyhoo, in their infinite wisdom the 28500 guys might
* come up with something we can't possibly handle so
* best end things here */
archive_set_error(
&a->archive, ARCHIVE_ERRNO_MISC,
"Unsupported record version");
return (ARCHIVE_FATAL);
} else if ((cntlen = _warc_rdlen(buf, eoh - buf)) < 0) {
/* nightmare! the specs say content-length is mandatory
* so I don't feel overly bad stopping the reader here */
archive_set_error(
&a->archive, EINVAL,
"Bad content length");
return (ARCHIVE_FATAL);
} else if ((rtime = _warc_rdrtm(buf, eoh - buf)) == (time_t)-1) {
/* record time is mandatory as per WARC/1.0,
* so just barf here, fast and loud */
archive_set_error(
&a->archive, EINVAL,
"Bad record time");
return (ARCHIVE_FATAL);
}
/* let the world know we're a WARC archive */
a->archive.archive_format = ARCHIVE_FORMAT_WARC;
if (ver != w->pver) {
/* stringify this entry's version */
archive_string_sprintf(&w->sver,
"WARC/%u.%u", ver / 10000, ver % 10000);
/* remember the version */
w->pver = ver;
}
/* start off with the type */
ftyp = _warc_rdtyp(buf, eoh - buf);
/* and let future calls know about the content */
w->cntlen = cntlen;
w->cntoff = 0U;
mtime = 0;/* Avoid compiling error on some platform. */
switch (ftyp) {
case WT_RSRC:
case WT_RSP:
/* only try and read the filename in the cases that are
* guaranteed to have one */
fnam = _warc_rduri(buf, eoh - buf);
/* check the last character in the URI to avoid creating
* directory endpoints as files, see Todo above */
if (fnam.len == 0 || fnam.str[fnam.len - 1] == '/') {
/* break here for now */
fnam.len = 0U;
fnam.str = NULL;
break;
}
/* bang to our string pool, so we save a
* malloc()+free() roundtrip */
if (fnam.len + 1U > w->pool.len) {
w->pool.len = ((fnam.len + 64U) / 64U) * 64U;
w->pool.str = realloc(w->pool.str, w->pool.len);
}
memcpy(w->pool.str, fnam.str, fnam.len);
w->pool.str[fnam.len] = '\0';
/* let noone else know about the pool, it's a secret, shhh */
fnam.str = w->pool.str;
/* snarf mtime or deduce from rtime
* this is a custom header added by our writer, it's quite
* hard to believe anyone else would go through with it
* (apart from being part of some http responses of course) */
if ((mtime = _warc_rdmtm(buf, eoh - buf)) == (time_t)-1) {
mtime = rtime;
}
break;
default:
fnam.len = 0U;
fnam.str = NULL;
break;
}
/* now eat some of those delicious buffer bits */
__archive_read_consume(a, eoh - buf);
switch (ftyp) {
case WT_RSRC:
case WT_RSP:
if (fnam.len > 0U) {
/* populate entry object */
archive_entry_set_filetype(entry, AE_IFREG);
archive_entry_copy_pathname(entry, fnam.str);
archive_entry_set_size(entry, cntlen);
archive_entry_set_perm(entry, 0644);
/* rtime is the new ctime, mtime stays mtime */
archive_entry_set_ctime(entry, rtime, 0L);
archive_entry_set_mtime(entry, mtime, 0L);
break;
}
/* FALLTHROUGH */
default:
/* consume the content and start over */
_warc_skip(a);
goto start_over;
}
return (ARCHIVE_OK);
}
static int
slurp_central_directory(struct archive_read *a, struct zip *zip)
{
unsigned i;
int64_t correction;
static const struct archive_rb_tree_ops rb_ops = {
&cmp_node, &cmp_key
};
static const struct archive_rb_tree_ops rb_rsrc_ops = {
&rsrc_cmp_node, &rsrc_cmp_key
};
/*
* Consider the archive file we are reading may be SFX.
* So we have to calculate a SFX header size to revise
* ZIP header offsets.
*/
correction = zip->end_of_central_directory_offset -
(zip->central_directory_offset + zip->central_directory_size);
/* The central directory offset is relative value, and so
* we revise this offset for SFX. */
zip->central_directory_offset += correction;
__archive_read_seek(a, zip->central_directory_offset, SEEK_SET);
zip->offset = zip->central_directory_offset;
__archive_rb_tree_init(&zip->tree, &rb_ops);
__archive_rb_tree_init(&zip->tree_rsrc, &rb_rsrc_ops);
zip->zip_entries = calloc(zip->central_directory_entries,
sizeof(struct zip_entry));
for (i = 0; i < zip->central_directory_entries; ++i) {
struct zip_entry *zip_entry = &zip->zip_entries[i];
size_t filename_length, extra_length, comment_length;
uint32_t external_attributes;
const char *name, *p, *r;
if ((p = __archive_read_ahead(a, 46, NULL)) == NULL)
return ARCHIVE_FATAL;
if (memcmp(p, "PK\001\002", 4) != 0) {
archive_set_error(&a->archive,
-1, "Invalid central directory signature");
return ARCHIVE_FATAL;
}
zip->have_central_directory = 1;
/* version = p[4]; */
zip_entry->system = p[5];
/* version_required = archive_le16dec(p + 6); */
zip_entry->flags = archive_le16dec(p + 8);
zip_entry->compression = (char)archive_le16dec(p + 10);
zip_entry->mtime = zip_time(p + 12);
zip_entry->crc32 = archive_le32dec(p + 16);
zip_entry->compressed_size = archive_le32dec(p + 20);
zip_entry->uncompressed_size = archive_le32dec(p + 24);
filename_length = archive_le16dec(p + 28);
extra_length = archive_le16dec(p + 30);
comment_length = archive_le16dec(p + 32);
/* disk_start = archive_le16dec(p + 34); */ /* Better be zero. */
/* internal_attributes = archive_le16dec(p + 36); */ /* text bit */
external_attributes = archive_le32dec(p + 38);
zip_entry->local_header_offset =
archive_le32dec(p + 42) + correction;
/* If we can't guess the mode, leave it zero here;
when we read the local file header we might get
more information. */
zip_entry->mode = 0;
if (zip_entry->system == 3) {
zip_entry->mode = external_attributes >> 16;
}
/*
* Mac resource fork files are stored under the
* "__MACOSX/" directory, so we should check if
* it is.
*/
/* Make sure we have the file name. */
if ((p = __archive_read_ahead(a, 46 + filename_length, NULL))
== NULL)
return ARCHIVE_FATAL;
name = p + 46;
r = rsrc_basename(name, filename_length);
if (filename_length >= 9 &&
strncmp("__MACOSX/", name, 9) == 0) {
/* If this file is not a resource fork nor
* a directory. We should treat it as a non
* resource fork file to expose it. */
if (name[filename_length-1] != '/' &&
(r - name < 3 || r[0] != '.' || r[1] != '_')) {
__archive_rb_tree_insert_node(&zip->tree,
&zip_entry->node);
/* Expose its parent directories. */
expose_parent_dirs(zip, name, filename_length);
} else {
/* This file is a resource fork file or
* a directory. */
archive_strncpy(&(zip_entry->rsrcname), name,
filename_length);
__archive_rb_tree_insert_node(&zip->tree_rsrc,
&zip_entry->node);
}
} else {
/* Generate resource fork name to find its resource
* file at zip->tree_rsrc. */
archive_strcpy(&(zip_entry->rsrcname), "__MACOSX/");
archive_strncat(&(zip_entry->rsrcname), name, r - name);
archive_strcat(&(zip_entry->rsrcname), "._");
archive_strncat(&(zip_entry->rsrcname),
name + (r - name), filename_length - (r - name));
/* Register an entry to RB tree to sort it by
* file offset. */
__archive_rb_tree_insert_node(&zip->tree,
&zip_entry->node);
}
/* We don't read the filename until we get to the
local file header. Reading it here would speed up
table-of-contents operations (removing the need to
find and read local file header to get the
filename) at the cost of requiring a lot of extra
space. */
/* We don't read the extra block here. We assume it
will be duplicated at the local file header. */
__archive_read_consume(a,
46 + filename_length + extra_length + comment_length);
}
static int
tk_archive_read_format_zip_read_header(struct archive_read *a,
struct archive_entry *entry)
{
const void *h;
const char *signature;
struct zip *zip;
int r = ARCHIVE_OK, r1;
a->archive.archive_format = ARCHIVE_FORMAT_ZIP;
if (a->archive.archive_format_name == NULL)
a->archive.archive_format_name = "ZIP";
zip = (struct zip *)(a->format->data);
zip->decompress_init = 0;
zip->end_of_entry = 0;
zip->entry_uncompressed_bytes_read = 0;
zip->entry_compressed_bytes_read = 0;
zip->entry_crc32 = crc32(0, NULL, 0);
if ((h = __archive_read_ahead(a, 4, NULL)) == NULL)
return (ARCHIVE_FATAL);
signature = (const char *)h;
if (signature[0] == 'M' && signature[1] == 'Z') {
/* This is an executable? Must be self-extracting... */
r = skip_sfx(a);
if (r < ARCHIVE_WARN)
return (r);
if ((h = __archive_read_ahead(a, 4, NULL)) == NULL)
return (ARCHIVE_FATAL);
signature = (const char *)h;
}
if (signature[0] != 'P' || signature[1] != 'K') {
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
"Bad ZIP file");
return (ARCHIVE_FATAL);
}
/*
* "PK00" signature is used for "split" archives that
* only have a single segment. This means we can just
* skip the PK00; the first real file header should follow.
*/
if (signature[2] == '0' && signature[3] == '0') {
__archive_read_consume(a, 4);
if ((h = __archive_read_ahead(a, 4, NULL)) == NULL)
return (ARCHIVE_FATAL);
signature = (const char *)h;
if (signature[0] != 'P' || signature[1] != 'K') {
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
"Bad ZIP file");
return (ARCHIVE_FATAL);
}
}
if (signature[2] == '\001' && signature[3] == '\002') {
/* Beginning of central directory. */
return (ARCHIVE_EOF);
}
if (signature[2] == '\003' && signature[3] == '\004') {
/* Regular file entry. */
r1 = zip_read_file_header(a, entry, zip);
if (r1 != ARCHIVE_OK)
return (r1);
return (r);
}
if (signature[2] == '\005' && signature[3] == '\006') {
/* End-of-archive record. */
return (ARCHIVE_EOF);
}
if (signature[2] == '\007' && signature[3] == '\010') {
/*
* We should never encounter this record here;
* see ZIP_LENGTH_AT_END handling below for details.
*/
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"Bad ZIP file: Unexpected end-of-entry record");
return (ARCHIVE_FATAL);
}
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
"Damaged ZIP file or unsupported format variant (%d,%d)",
signature[2], signature[3]);
return (ARCHIVE_FATAL);
}
static int
tk_archive_read_format_zip_bid(struct archive_read *a)
{
const char *p;
const void *buff;
ssize_t bytes_avail, offset;
if ((p = __archive_read_ahead(a, 4, NULL)) == NULL)
return (-1);
/*
* Bid of 30 here is: 16 bits for "PK",
* next 16-bit field has four options (-2 bits).
* 16 + 16-2 = 30.
*/
if (p[0] == 'P' && p[1] == 'K') {
if ((p[2] == '\001' && p[3] == '\002')
|| (p[2] == '\003' && p[3] == '\004')
|| (p[2] == '\005' && p[3] == '\006')
|| (p[2] == '\007' && p[3] == '\010')
|| (p[2] == '0' && p[3] == '0'))
return (30);
}
/*
* Attempt to handle self-extracting archives
* by noting a PE header and searching forward
* up to 128k for a 'PK\003\004' marker.
*/
if (p[0] == 'M' && p[1] == 'Z') {
/*
* TODO: Optimize by initializing 'offset' to an
* estimate of the likely start of the archive data
* based on values in the PE header. Note that we
* don't need to be exact, but we mustn't skip too
* far. The search below will compensate if we
* undershoot.
*/
offset = 0;
while (offset < 124000) {
/* Get 4k of data beyond where we stopped. */
buff = __archive_read_ahead(a, offset + 4096,
&bytes_avail);
if (buff == NULL)
break;
p = (const char *)buff + offset;
while (p + 9 < (const char *)buff + bytes_avail) {
if (p[0] == 'P' && p[1] == 'K' /* signature */
&& p[2] == 3 && p[3] == 4 /* File entry */
&& p[8] == 8 /* compression == deflate */
&& p[9] == 0 /* High byte of compression */
)
{
return (30);
}
++p;
}
offset = p - (const char *)buff;
}
}
return (0);
}
static int
header_newc(struct archive_read *a, struct cpio *cpio,
struct archive_entry *entry, size_t *namelength, size_t *name_pad)
{
const void *h;
const char *header;
int r;
r = find_newc_header(a);
if (r < ARCHIVE_WARN)
return (r);
/* Read fixed-size portion of header. */
h = __archive_read_ahead(a, newc_header_size, NULL);
if (h == NULL)
return (ARCHIVE_FATAL);
/* Parse out hex fields. */
header = (const char *)h;
if (memcmp(header + newc_magic_offset, "070701", 6) == 0) {
a->archive.archive_format = ARCHIVE_FORMAT_CPIO_SVR4_NOCRC;
a->archive.archive_format_name = "ASCII cpio (SVR4 with no CRC)";
} else if (memcmp(header + newc_magic_offset, "070702", 6) == 0) {
a->archive.archive_format = ARCHIVE_FORMAT_CPIO_SVR4_CRC;
a->archive.archive_format_name = "ASCII cpio (SVR4 with CRC)";
} else {
/* TODO: Abort here? */
}
archive_entry_set_devmajor(entry,
(dev_t)atol16(header + newc_devmajor_offset, newc_devmajor_size));
archive_entry_set_devminor(entry,
(dev_t)atol16(header + newc_devminor_offset, newc_devminor_size));
archive_entry_set_ino(entry, atol16(header + newc_ino_offset, newc_ino_size));
archive_entry_set_mode(entry,
(mode_t)atol16(header + newc_mode_offset, newc_mode_size));
archive_entry_set_uid(entry, atol16(header + newc_uid_offset, newc_uid_size));
archive_entry_set_gid(entry, atol16(header + newc_gid_offset, newc_gid_size));
archive_entry_set_nlink(entry,
(unsigned int)atol16(header + newc_nlink_offset, newc_nlink_size));
archive_entry_set_rdevmajor(entry,
(dev_t)atol16(header + newc_rdevmajor_offset, newc_rdevmajor_size));
archive_entry_set_rdevminor(entry,
(dev_t)atol16(header + newc_rdevminor_offset, newc_rdevminor_size));
archive_entry_set_mtime(entry, atol16(header + newc_mtime_offset, newc_mtime_size), 0);
*namelength = (size_t)atol16(header + newc_namesize_offset, newc_namesize_size);
/* Pad name to 2 more than a multiple of 4. */
*name_pad = (2 - *namelength) & 3;
/*
* Note: entry_bytes_remaining is at least 64 bits and
* therefore guaranteed to be big enough for a 33-bit file
* size.
*/
cpio->entry_bytes_remaining =
atol16(header + newc_filesize_offset, newc_filesize_size);
archive_entry_set_size(entry, cpio->entry_bytes_remaining);
/* Pad file contents to a multiple of 4. */
cpio->entry_padding = 3 & -cpio->entry_bytes_remaining;
__archive_read_consume(a, newc_header_size);
return (r);
}
static int
find_odc_header(struct archive_read *a)
{
const void *h;
const char *p, *q;
size_t skip, skipped = 0;
ssize_t bytes;
for (;;) {
h = __archive_read_ahead(a, odc_header_size, &bytes);
if (h == NULL)
return (ARCHIVE_FATAL);
p = h;
q = p + bytes;
/* Try the typical case first, then go into the slow search.*/
if (memcmp("070707", p, 6) == 0 && is_octal(p, odc_header_size))
return (ARCHIVE_OK);
if (memcmp("070727", p, 6) == 0 && is_afio_large(p, bytes)) {
a->archive.archive_format = ARCHIVE_FORMAT_CPIO_AFIO_LARGE;
return (ARCHIVE_OK);
}
/*
* Scan ahead until we find something that looks
* like an odc header.
*/
while (p + odc_header_size <= q) {
switch (p[5]) {
case '7':
if ((memcmp("070707", p, 6) == 0
&& is_octal(p, odc_header_size))
|| (memcmp("070727", p, 6) == 0
&& is_afio_large(p, q - p))) {
skip = p - (const char *)h;
__archive_read_consume(a, skip);
skipped += skip;
if (p[4] == '2')
a->archive.archive_format =
ARCHIVE_FORMAT_CPIO_AFIO_LARGE;
if (skipped > 0) {
archive_set_error(&a->archive,
0,
"Skipped %d bytes before "
"finding valid header",
(int)skipped);
return (ARCHIVE_WARN);
}
return (ARCHIVE_OK);
}
p += 2;
break;
case '0':
p++;
break;
default:
p += 6;
break;
}
}
skip = p - (const char *)h;
__archive_read_consume(a, skip);
skipped += skip;
}
}
static int
detect_form(struct archive_read *a, int *is_form_d)
{
const char *p;
ssize_t avail, ravail;
ssize_t detected_bytes = 0, len, nl;
int entry_cnt = 0, multiline = 0;
int form_D = 0;/* The archive is generated by `NetBSD mtree -D'
* (In this source we call it `form D') . */
if (is_form_d != NULL)
*is_form_d = 0;
p = __archive_read_ahead(a, 1, &avail);
if (p == NULL)
return (-1);
ravail = avail;
for (;;) {
len = next_line(a, &p, &avail, &ravail, &nl);
/* The terminal character of the line should be
* a new line character, '\r\n' or '\n'. */
if (len <= 0 || nl == 0)
break;
if (!multiline) {
/* Leading whitespace is never significant,
* ignore it. */
while (len > 0 && (*p == ' ' || *p == '\t')) {
++p;
--avail;
--len;
}
/* Skip comment or empty line. */
if (p[0] == '#' || p[0] == '\n' || p[0] == '\r') {
p += len;
avail -= len;
continue;
}
} else {
/* A continuance line; the terminal
* character of previous line was '\' character. */
if (bid_keyword_list(p, len, 0, 0) <= 0)
break;
if (multiline == 1)
detected_bytes += len;
if (p[len-nl-1] != '\\') {
if (multiline == 1 &&
++entry_cnt >= MAX_BID_ENTRY)
break;
multiline = 0;
}
p += len;
avail -= len;
continue;
}
if (p[0] != '/') {
int last_is_path, keywords;
keywords = bid_entry(p, len, nl, &last_is_path);
if (keywords >= 0) {
detected_bytes += len;
if (form_D == 0) {
if (last_is_path)
form_D = 1;
else if (keywords > 0)
/* This line is not `form D'. */
form_D = -1;
} else if (form_D == 1) {
if (!last_is_path && keywords > 0)
/* This this is not `form D'
* and We cannot accept mixed
* format. */
break;
}
if (!last_is_path && p[len-nl-1] == '\\')
/* This line continues. */
multiline = 1;
else {
/* We've got plenty of correct lines
* to assume that this file is a mtree
* format. */
if (++entry_cnt >= MAX_BID_ENTRY)
break;
}
} else
break;
} else if (strncmp(p, "/set", 4) == 0) {
if (bid_keyword_list(p+4, len-4, 0, 0) <= 0)
break;
/* This line continues. */
if (p[len-nl-1] == '\\')
multiline = 2;
} else if (strncmp(p, "/unset", 6) == 0) {
if (bid_keyword_list(p+6, len-6, 1, 0) <= 0)
break;
/* This line continues. */
if (p[len-nl-1] == '\\')
multiline = 2;
} else
break;
/* Test next line. */
p += len;
avail -= len;
}
if (entry_cnt >= MAX_BID_ENTRY || (entry_cnt > 0 && len == 0)) {
if (is_form_d != NULL) {
if (form_D == 1)
*is_form_d = 1;
}
return (32);
}
return (0);
}
static int
archive_read_format_zip_read_data(struct archive_read *a,
const void **buff, size_t *size, off_t *offset)
{
int r;
struct zip *zip;
zip = (struct zip *)(a->format->data);
/*
* If we hit end-of-entry last time, clean up and return
* ARCHIVE_EOF this time.
*/
if (zip->end_of_entry) {
if (!zip->end_of_entry_cleanup) {
if (zip->flags & ZIP_LENGTH_AT_END) {
const char *p;
if ((p = __archive_read_ahead(a, 16)) == NULL) {
archive_set_error(&a->archive,
ARCHIVE_ERRNO_FILE_FORMAT,
"Truncated ZIP end-of-file record");
return (ARCHIVE_FATAL);
}
zip->crc32 = archive_le32dec(p + 4);
zip->compressed_size = archive_le32dec(p + 8);
zip->uncompressed_size = archive_le32dec(p + 12);
(a->decompressor->consume)(a, 16);
}
/* Check file size, CRC against these values. */
if (zip->compressed_size != zip->entry_compressed_bytes_read) {
archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"ZIP compressed data is wrong size");
return (ARCHIVE_WARN);
}
/* Size field only stores the lower 32 bits of the actual size. */
if ((zip->uncompressed_size & UINT32_MAX)
!= (zip->entry_uncompressed_bytes_read & UINT32_MAX)) {
archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"ZIP uncompressed data is wrong size");
return (ARCHIVE_WARN);
}
/* TODO: Compute CRC. */
/*
if (zip->crc32 != zip->entry_crc32_calculated) {
archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"ZIP data CRC error");
return (ARCHIVE_WARN);
}
*/
/* End-of-entry cleanup done. */
zip->end_of_entry_cleanup = 1;
}
*offset = zip->entry_uncompressed_bytes_read;
*size = 0;
*buff = NULL;
return (ARCHIVE_EOF);
}
switch(zip->compression) {
case 0: /* No compression. */
r = zip_read_data_none(a, buff, size, offset);
break;
case 8: /* Deflate compression. */
r = zip_read_data_deflate(a, buff, size, offset);
break;
default: /* Unsupported compression. */
*buff = NULL;
*size = 0;
*offset = 0;
/* Return a warning. */
archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
"Unsupported ZIP compression method (%s)",
zip->compression_name);
if (zip->flags & ZIP_LENGTH_AT_END) {
/*
* ZIP_LENGTH_AT_END requires us to
* decompress the entry in order to
* skip it, but we don't know this
* compression method, so we give up.
*/
r = ARCHIVE_FATAL;
} else {
/* We can't decompress this entry, but we will
* be able to skip() it and try the next entry. */
r = ARCHIVE_WARN;
}
break;
}
return (r);
}
/*
* Returns length of line (including trailing newline)
* or negative on error. 'start' argument is updated to
* point to first character of line.
*/
static ssize_t
readline(struct archive_read *a, struct mtree *mtree, char **start, ssize_t limit)
{
ssize_t bytes_read;
ssize_t total_size = 0;
ssize_t find_off = 0;
const void *t;
const char *s;
void *p;
char *u;
/* Accumulate line in a line buffer. */
for (;;) {
/* Read some more. */
t = __archive_read_ahead(a, 1, &bytes_read);
if (t == NULL)
return (0);
if (bytes_read < 0)
return (ARCHIVE_FATAL);
s = t; /* Start of line? */
p = memchr(t, '\n', bytes_read);
/* If we found '\n', trim the read. */
if (p != NULL) {
bytes_read = 1 + ((const char *)p) - s;
}
if (total_size + bytes_read + 1 > limit) {
archive_set_error(&a->archive,
ARCHIVE_ERRNO_FILE_FORMAT,
"Line too long");
return (ARCHIVE_FATAL);
}
if (archive_string_ensure(&mtree->line,
total_size + bytes_read + 1) == NULL) {
archive_set_error(&a->archive, ENOMEM,
"Can't allocate working buffer");
return (ARCHIVE_FATAL);
}
memcpy(mtree->line.s + total_size, t, bytes_read);
__archive_read_consume(a, bytes_read);
total_size += bytes_read;
/* Null terminate. */
mtree->line.s[total_size] = '\0';
/* If we found an unescaped '\n', clean up and return. */
for (u = mtree->line.s + find_off; *u; ++u) {
if (u[0] == '\n') {
*start = mtree->line.s;
return total_size;
}
if (u[0] == '#') {
if (p == NULL)
break;
*start = mtree->line.s;
return total_size;
}
if (u[0] != '\\')
continue;
if (u[1] == '\\') {
++u;
continue;
}
if (u[1] == '\n') {
memmove(u, u + 1,
total_size - (u - mtree->line.s) + 1);
--total_size;
++u;
break;
}
if (u[1] == '\0')
break;
}
find_off = u - mtree->line.s;
}
}
static int
mtree_bid(struct archive_read *a, int best_bid)
{
const char *signature = "#mtree";
const char *p;
ssize_t avail, ravail;
ssize_t len, nl;
int detected_bytes = 0, entry_cnt = 0, multiline = 0;
(void)best_bid; /* UNUSED */
/* Now let's look at the actual header and see if it matches. */
p = __archive_read_ahead(a, strlen(signature), &avail);
if (p == NULL)
return (-1);
if (memcmp(p, signature, strlen(signature)) == 0)
return (8 * (int)strlen(signature));
/*
* There is not a mtree signature. Let's try to detect mtree format.
*/
ravail = avail;
for (;;) {
len = next_line(a, &p, &avail, &ravail, &nl);
/* The terminal character of the line should be
* a new line character, '\r\n' or '\n'. */
if (len <= 0 || nl == 0)
break;
if (!multiline) {
/* Leading whitespace is never significant,
* ignore it. */
while (len > 0 && (*p == ' ' || *p == '\t')) {
++p;
--avail;
--len;
}
/* Skip comment or empty line. */
if (p[0] == '#' || p[0] == '\n' || p[0] == '\r') {
p += len;
avail -= len;
continue;
}
} else {
/* A continuance line; the terminal
* character of previous line was '\' character. */
if (bid_keyword_list(p, len, 0) <= 0)
break;
if (multiline == 1)
detected_bytes += len;
if (p[len-nl-1] != '\\') {
if (multiline == 1 &&
++entry_cnt >= MAX_BID_ENTRY)
break;
multiline = 0;
}
p += len;
avail -= len;
continue;
}
if (p[0] != '/') {
if (bid_entry(p, len) >= 0) {
detected_bytes += len;
if (p[len-nl-1] == '\\')
/* This line continues. */
multiline = 1;
else {
/* We've got plenty of correct lines
* to assume that this file is a mtree
* format. */
if (++entry_cnt >= MAX_BID_ENTRY)
break;
}
} else
break;
} else if (strncmp(p, "/set", 4) == 0) {
if (bid_keyword_list(p+4, len-4, 0) <= 0)
break;
/* This line continues. */
if (p[len-nl-1] == '\\')
multiline = 2;
} else if (strncmp(p, "/unset", 6) == 0) {
if (bid_keyword_list(p+6, len-6, 1) <= 0)
break;
/* This line continues. */
if (p[len-nl-1] == '\\')
multiline = 2;
} else
break;
/* Test next line. */
p += len;
avail -= len;
}
if (entry_cnt >= MAX_BID_ENTRY || (entry_cnt > 0 && len == 0))
return (32);
return (0);
}
static int
zip_read_data_deflate(struct archive_read *a, const void **buff,
size_t *size, off_t *offset)
{
struct zip *zip;
ssize_t bytes_avail;
const void *compressed_buff;
int r;
zip = (struct zip *)(a->format->data);
/* If the buffer hasn't been allocated, allocate it now. */
if (zip->uncompressed_buffer == NULL) {
zip->uncompressed_buffer_size = 32 * 1024;
zip->uncompressed_buffer
= (unsigned char *)malloc(zip->uncompressed_buffer_size);
if (zip->uncompressed_buffer == NULL) {
tk_archive_set_error(&a->archive, ENOMEM,
"No memory for ZIP decompression");
return (ARCHIVE_FATAL);
}
}
/* If we haven't yet read any data, initialize the decompressor. */
if (!zip->decompress_init) {
if (zip->stream_valid)
r = inflateReset(&zip->stream);
else
r = inflateInit2(&zip->stream,
-15 /* Don't check for zlib header */);
if (r != Z_OK) {
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"Can't initialize ZIP decompression.");
return (ARCHIVE_FATAL);
}
/* Stream structure has been set up. */
zip->stream_valid = 1;
/* We've initialized decompression for this stream. */
zip->decompress_init = 1;
}
/*
* Note: '1' here is a performance optimization.
* Recall that the decompression layer returns a count of
* available bytes; asking for more than that forces the
* decompressor to combine reads by copying data.
*/
compressed_buff = __archive_read_ahead(a, 1, &bytes_avail);
if (bytes_avail <= 0) {
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
"Truncated ZIP file body");
return (ARCHIVE_FATAL);
}
/*
* A bug in zlib.h: stream.next_in should be marked 'const'
* but isn't (the library never alters data through the
* next_in pointer, only reads it). The result: this ugly
* cast to remove 'const'.
*/
zip->stream.next_in = (Bytef *)(uintptr_t)(const void *)compressed_buff;
zip->stream.avail_in = bytes_avail;
zip->stream.total_in = 0;
zip->stream.next_out = zip->uncompressed_buffer;
zip->stream.avail_out = zip->uncompressed_buffer_size;
zip->stream.total_out = 0;
r = inflate(&zip->stream, 0);
switch (r) {
case Z_OK:
break;
case Z_STREAM_END:
zip->end_of_entry = 1;
break;
case Z_MEM_ERROR:
tk_archive_set_error(&a->archive, ENOMEM,
"Out of memory for ZIP decompression");
return (ARCHIVE_FATAL);
default:
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"ZIP decompression failed (%d)", r);
return (ARCHIVE_FATAL);
}
/* Consume as much as the compressor actually used. */
bytes_avail = zip->stream.total_in;
__archive_read_consume(a, bytes_avail);
zip->entry_bytes_remaining -= bytes_avail;
zip->entry_compressed_bytes_read += bytes_avail;
*offset = zip->entry_offset;
*size = zip->stream.total_out;
zip->entry_uncompressed_bytes_read += *size;
*buff = zip->uncompressed_buffer;
zip->entry_offset += *size;
return (ARCHIVE_OK);
}
static int
zip_read_file_header(struct archive_read *a, struct archive_entry *entry,
struct zip *zip)
{
const struct zip_file_header *p;
const void *h;
if ((p = __archive_read_ahead(a, sizeof *p, NULL)) == NULL) {
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
"Truncated ZIP file header");
return (ARCHIVE_FATAL);
}
zip->version = p->version[0];
zip->system = p->version[1];
zip->flags = tk_archive_le16dec(p->flags);
zip->compression = tk_archive_le16dec(p->compression);
if (zip->compression <
sizeof(compression_names)/sizeof(compression_names[0]))
zip->compression_name = compression_names[zip->compression];
else
zip->compression_name = "??";
zip->mtime = zip_time(p->timedate);
zip->ctime = 0;
zip->atime = 0;
zip->mode = 0;
zip->uid = 0;
zip->gid = 0;
zip->crc32 = tk_archive_le32dec(p->crc32);
zip->filename_length = tk_archive_le16dec(p->filename_length);
zip->extra_length = tk_archive_le16dec(p->extra_length);
zip->uncompressed_size = tk_archive_le32dec(p->uncompressed_size);
zip->compressed_size = tk_archive_le32dec(p->compressed_size);
__archive_read_consume(a, sizeof(struct zip_file_header));
/* Read the filename. */
if ((h = __archive_read_ahead(a, zip->filename_length, NULL)) == NULL) {
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
"Truncated ZIP file header");
return (ARCHIVE_FATAL);
}
if (tk_archive_string_ensure(&zip->pathname, zip->filename_length) == NULL)
__archive_errx(1, "Out of memory");
tk_archive_strncpy(&zip->pathname, h, zip->filename_length);
__archive_read_consume(a, zip->filename_length);
tk_archive_entry_set_pathname(entry, zip->pathname.s);
if (zip->pathname.s[tk_archive_strlen(&zip->pathname) - 1] == '/')
zip->mode = AE_IFDIR | 0777;
else
zip->mode = AE_IFREG | 0777;
/* Read the extra data. */
if ((h = __archive_read_ahead(a, zip->extra_length, NULL)) == NULL) {
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
"Truncated ZIP file header");
return (ARCHIVE_FATAL);
}
process_extra(h, zip);
__archive_read_consume(a, zip->extra_length);
/* Populate some additional entry fields: */
tk_archive_entry_set_mode(entry, zip->mode);
tk_archive_entry_set_uid(entry, zip->uid);
tk_archive_entry_set_gid(entry, zip->gid);
tk_archive_entry_set_mtime(entry, zip->mtime, 0);
tk_archive_entry_set_ctime(entry, zip->ctime, 0);
tk_archive_entry_set_atime(entry, zip->atime, 0);
/* Set the size only if it's meaningful. */
if (0 == (zip->flags & ZIP_LENGTH_AT_END))
tk_archive_entry_set_size(entry, zip->uncompressed_size);
zip->entry_bytes_remaining = zip->compressed_size;
zip->entry_offset = 0;
/* If there's no body, force read_data() to return EOF immediately. */
if (0 == (zip->flags & ZIP_LENGTH_AT_END)
&& zip->entry_bytes_remaining < 1)
zip->end_of_entry = 1;
/* Set up a more descriptive format name. */
sprintf(zip->format_name, "ZIP %d.%d (%s)",
zip->version / 10, zip->version % 10,
zip->compression_name);
a->archive.archive_format_name = zip->format_name;
return (ARCHIVE_OK);
}
static int
tk_archive_read_format_zip_read_data(struct archive_read *a,
const void **buff, size_t *size, off_t *offset)
{
int r;
struct zip *zip;
zip = (struct zip *)(a->format->data);
/*
* If we hit end-of-entry last time, clean up and return
* ARCHIVE_EOF this time.
*/
if (zip->end_of_entry) {
*offset = zip->entry_uncompressed_bytes_read;
*size = 0;
*buff = NULL;
return (ARCHIVE_EOF);
}
switch(zip->compression) {
case 0: /* No compression. */
r = zip_read_data_none(a, buff, size, offset);
break;
case 8: /* Deflate compression. */
r = zip_read_data_deflate(a, buff, size, offset);
break;
default: /* Unsupported compression. */
*buff = NULL;
*size = 0;
*offset = 0;
/* Return a warning. */
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
"Unsupported ZIP compression method (%s)",
zip->compression_name);
if (zip->flags & ZIP_LENGTH_AT_END) {
/*
* ZIP_LENGTH_AT_END requires us to
* decompress the entry in order to
* skip it, but we don't know this
* compression method, so we give up.
*/
r = ARCHIVE_FATAL;
} else {
/* We can't decompress this entry, but we will
* be able to skip() it and try the next entry. */
r = ARCHIVE_WARN;
}
break;
}
if (r != ARCHIVE_OK)
return (r);
/* Update checksum */
if (*size)
zip->entry_crc32 = crc32(zip->entry_crc32, *buff, *size);
/* If we hit the end, swallow any end-of-data marker. */
if (zip->end_of_entry) {
if (zip->flags & ZIP_LENGTH_AT_END) {
const char *p;
if ((p = __archive_read_ahead(a, 16, NULL)) == NULL) {
tk_archive_set_error(&a->archive,
ARCHIVE_ERRNO_FILE_FORMAT,
"Truncated ZIP end-of-file record");
return (ARCHIVE_FATAL);
}
zip->crc32 = tk_archive_le32dec(p + 4);
zip->compressed_size = tk_archive_le32dec(p + 8);
zip->uncompressed_size = tk_archive_le32dec(p + 12);
__archive_read_consume(a, 16);
}
/* Check file size, CRC against these values. */
if (zip->compressed_size != zip->entry_compressed_bytes_read) {
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"ZIP compressed data is wrong size");
return (ARCHIVE_WARN);
}
/* Size field only stores the lower 32 bits of the actual size. */
if ((zip->uncompressed_size & UINT32_MAX)
!= (zip->entry_uncompressed_bytes_read & UINT32_MAX)) {
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"ZIP uncompressed data is wrong size");
return (ARCHIVE_WARN);
}
/* Check computed CRC against header */
if (zip->crc32 != zip->entry_crc32) {
tk_archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"ZIP bad CRC: 0x%lx should be 0x%lx",
zip->entry_crc32, zip->crc32);
return (ARCHIVE_WARN);
}
}
/* Return EOF immediately if this is a non-regular file. */
if (AE_IFREG != (zip->mode & AE_IFMT))
return (ARCHIVE_EOF);
return (ARCHIVE_OK);
}
/*
* TODO: This is a performance sink because it forces the read core to
* drop buffered data from the start of file, which will then have to
* be re-read again if this bidder loses.
*
* We workaround this a little by passing in the best bid so far so
* that later bidders can do nothing if they know they'll never
* outbid. But we can certainly do better...
*/
static int
archive_read_format_zip_seekable_bid(struct archive_read *a, int best_bid)
{
struct zip *zip = (struct zip *)a->format->data;
int64_t filesize;
const char *p;
/* If someone has already bid more than 32, then avoid
trashing the look-ahead buffers with a seek. */
if (best_bid > 32)
return (-1);
filesize = __archive_read_seek(a, -22, SEEK_END);
/* If we can't seek, then we can't bid. */
if (filesize <= 0)
return 0;
/* TODO: More robust search for end of central directory record. */
if ((p = __archive_read_ahead(a, 22, NULL)) == NULL)
return 0;
/* First four bytes are signature for end of central directory
record. Four zero bytes ensure this isn't a multi-volume
Zip file (which we don't yet support). */
if (memcmp(p, "PK\005\006\000\000\000\000", 8) != 0) {
int64_t i, tail;
int found;
/*
* If there is a comment in end of central directory
* record, 22 bytes are too short. we have to read more
* to properly detect the record. Hopefully, a length
* of the comment is not longer than 16362 bytes(16K-22).
*/
if (filesize + 22 > 1024 * 16) {
tail = 1024 * 16;
filesize = __archive_read_seek(a, tail * -1, SEEK_END);
} else {
tail = filesize + 22;
filesize = __archive_read_seek(a, 0, SEEK_SET);
}
if (filesize < 0)
return 0;
if ((p = __archive_read_ahead(a, (size_t)tail, NULL)) == NULL)
return 0;
for (found = 0, i = 0;!found && i < tail - 22;) {
switch (p[i]) {
case 'P':
if (memcmp(p+i,
"PK\005\006\000\000\000\000", 8) == 0) {
p += i;
filesize += tail -
(22 + archive_le16dec(p+20));
found = 1;
} else
i += 8;
break;
case 'K': i += 7; break;
case 005: i += 6; break;
case 006: i += 5; break;
default: i += 1; break;
}
}
if (!found)
return 0;
}
/* Since we've already done the hard work of finding the
end of central directory record, let's save the important
information. */
zip->central_directory_entries = archive_le16dec(p + 10);
zip->central_directory_size = archive_le32dec(p + 12);
zip->central_directory_offset = archive_le32dec(p + 16);
zip->end_of_central_directory_offset = filesize;
/* Just one volume, so central dir must all be on this volume. */
if (zip->central_directory_entries != archive_le16dec(p + 8))
return 0;
/* Central directory can't extend beyond end of this file. */
if (zip->central_directory_offset +
(int64_t)zip->central_directory_size > filesize)
return 0;
/* This is just a tiny bit higher than the maximum returned by
the streaming Zip bidder. This ensures that the more accurate
seeking Zip parser wins whenever seek is available. */
return 32;
}
/*
* Returns length of line (including trailing newline)
* or negative on error. 'start' argument is updated to
* point to first character of line.
*/
static ssize_t
readline(struct archive_read *a, struct mtree *mtree, char **start,
ssize_t limit)
{
ssize_t bytes_read;
ssize_t total_size = 0;
ssize_t find_off = 0;
const void *t;
void *nl;
char *u;
/* Accumulate line in a line buffer. */
for (;;) {
/* Read some more. */
t = __archive_read_ahead(a, 1, &bytes_read);
if (t == NULL)
return (0);
if (bytes_read < 0)
return (ARCHIVE_FATAL);
nl = memchr(t, '\n', bytes_read);
/* If we found '\n', trim the read to end exactly there. */
if (nl != NULL) {
bytes_read = ((const char *)nl) - ((const char *)t) + 1;
}
if (total_size + bytes_read + 1 > limit) {
archive_set_error(&a->archive,
ARCHIVE_ERRNO_FILE_FORMAT,
"Line too long");
return (ARCHIVE_FATAL);
}
if (archive_string_ensure(&mtree->line,
total_size + bytes_read + 1) == NULL) {
archive_set_error(&a->archive, ENOMEM,
"Can't allocate working buffer");
return (ARCHIVE_FATAL);
}
/* Append new bytes to string. */
memcpy(mtree->line.s + total_size, t, bytes_read);
__archive_read_consume(a, bytes_read);
total_size += bytes_read;
mtree->line.s[total_size] = '\0';
for (u = mtree->line.s + find_off; *u; ++u) {
if (u[0] == '\n') {
/* Ends with unescaped newline. */
*start = mtree->line.s;
return total_size;
} else if (u[0] == '#') {
/* Ends with comment sequence #...\n */
if (nl == NULL) {
/* But we've not found the \n yet */
break;
}
} else if (u[0] == '\\') {
if (u[1] == '\n') {
/* Trim escaped newline. */
total_size -= 2;
mtree->line.s[total_size] = '\0';
break;
} else if (u[1] != '\0') {
/* Skip the two-char escape sequence */
++u;
}
}
}
find_off = u - mtree->line.s;
}
}
static int
archive_read_format_cpio_bid(struct archive_read *a, int best_bid)
{
const unsigned char *p;
struct cpio *cpio;
int bid;
(void)best_bid; /* UNUSED */
cpio = (struct cpio *)(a->format->data);
if ((p = __archive_read_ahead(a, 6, NULL)) == NULL)
return (-1);
bid = 0;
if (memcmp(p, "070707", 6) == 0) {
/* ASCII cpio archive (odc, POSIX.1) */
cpio->read_header = header_odc;
bid += 48;
/*
* XXX TODO: More verification; Could check that only octal
* digits appear in appropriate header locations. XXX
*/
} else if (memcmp(p, "070727", 6) == 0) {
/* afio large ASCII cpio archive */
cpio->read_header = header_odc;
bid += 48;
/*
* XXX TODO: More verification; Could check that almost hex
* digits appear in appropriate header locations. XXX
*/
} else if (memcmp(p, "070701", 6) == 0) {
/* ASCII cpio archive (SVR4 without CRC) */
cpio->read_header = header_newc;
bid += 48;
/*
* XXX TODO: More verification; Could check that only hex
* digits appear in appropriate header locations. XXX
*/
} else if (memcmp(p, "070702", 6) == 0) {
/* ASCII cpio archive (SVR4 with CRC) */
/* XXX TODO: Flag that we should check the CRC. XXX */
cpio->read_header = header_newc;
bid += 48;
/*
* XXX TODO: More verification; Could check that only hex
* digits appear in appropriate header locations. XXX
*/
} else if (p[0] * 256 + p[1] == 070707) {
/* big-endian binary cpio archives */
cpio->read_header = header_bin_be;
bid += 16;
/* Is more verification possible here? */
} else if (p[0] + p[1] * 256 == 070707) {
/* little-endian binary cpio archives */
cpio->read_header = header_bin_le;
bid += 16;
/* Is more verification possible here? */
} else
return (ARCHIVE_WARN);
return (bid);
}
static int
archive_read_format_zip_bid(struct archive_read *a)
{
const char *p;
const void *buff;
size_t bytes_avail;
if ((p = __archive_read_ahead(a, 4)) == NULL)
return (-1);
/*
* Bid of 30 here is: 16 bits for "PK",
* next 16-bit field has four options (-2 bits).
* 16 + 16-2 = 30.
*/
if (p[0] == 'P' && p[1] == 'K') {
if ((p[2] == '\001' && p[3] == '\002')
|| (p[2] == '\003' && p[3] == '\004')
|| (p[2] == '\005' && p[3] == '\006')
|| (p[2] == '\007' && p[3] == '\010')
|| (p[2] == '0' && p[3] == '0'))
return (30);
}
/*
* Attempt to handle self-extracting archives
* by noting a PE header and searching forward
* up to 64k for a 'PK\003\004' marker.
*/
if (p[0] == 'M' && p[1] == 'Z') {
/*
* TODO: Additional checks that this really is a PE
* file before we invoke the 128k lookahead below.
* No point in allocating a bigger lookahead buffer
* if we don't need to.
*/
/*
* TODO: Of course, the compression layer lookahead
* buffers aren't dynamically sized yet; they should be.
*/
bytes_avail = (a->decompressor->read_ahead)(a, &buff, 128*1024);
p = (const char *)buff;
/*
* TODO: Optimize by jumping forward based on values
* in the PE header. Note that we don't need to be
* exact, but we mustn't skip too far. The search
* below will compensate if we undershoot. Skipping
* will also reduce the chance of false positives
* (which is not really all that high to begin with,
* so maybe skipping isn't really necessary).
*/
while (p < bytes_avail + (const char *)buff) {
if (p[0] == 'P' && p[1] == 'K' /* "PK" signature */
&& p[2] == 3 && p[3] == 4 /* File entry */
&& p[8] == 8 /* compression == deflate */
&& p[9] == 0 /* High byte of compression */
)
{
return (30);
}
++p;
}
}
return (0);
}
static int
_ar_read_header(struct archive_read *a, struct archive_entry *entry,
struct ar *ar, const char *h, size_t *unconsumed)
{
char filename[AR_name_size + 1];
uint64_t number; /* Used to hold parsed numbers before validation. */
size_t bsd_name_length, entry_size;
char *p, *st;
const void *b;
int r;
/* Verify the magic signature on the file header. */
if (strncmp(h + AR_fmag_offset, "`\n", 2) != 0) {
archive_set_error(&a->archive, EINVAL,
"Incorrect file header signature");
return (ARCHIVE_FATAL);
}
/* Copy filename into work buffer. */
strncpy(filename, h + AR_name_offset, AR_name_size);
filename[AR_name_size] = '\0';
/*
* Guess the format variant based on the filename.
*/
if (a->archive.archive_format == ARCHIVE_FORMAT_AR) {
/* We don't already know the variant, so let's guess. */
/*
* Biggest clue is presence of '/': GNU starts special
* filenames with '/', appends '/' as terminator to
* non-special names, so anything with '/' should be
* GNU except for BSD long filenames.
*/
if (strncmp(filename, "#1/", 3) == 0)
a->archive.archive_format = ARCHIVE_FORMAT_AR_BSD;
else if (strchr(filename, '/') != NULL)
a->archive.archive_format = ARCHIVE_FORMAT_AR_GNU;
else if (strncmp(filename, "__.SYMDEF", 9) == 0)
a->archive.archive_format = ARCHIVE_FORMAT_AR_BSD;
/*
* XXX Do GNU/SVR4 'ar' programs ever omit trailing '/'
* if name exactly fills 16-byte field? If so, we
* can't assume entries without '/' are BSD. XXX
*/
}
/* Update format name from the code. */
if (a->archive.archive_format == ARCHIVE_FORMAT_AR_GNU)
a->archive.archive_format_name = "ar (GNU/SVR4)";
else if (a->archive.archive_format == ARCHIVE_FORMAT_AR_BSD)
a->archive.archive_format_name = "ar (BSD)";
else
a->archive.archive_format_name = "ar";
/*
* Remove trailing spaces from the filename. GNU and BSD
* variants both pad filename area out with spaces.
* This will only be wrong if GNU/SVR4 'ar' implementations
* omit trailing '/' for 16-char filenames and we have
* a 16-char filename that ends in ' '.
*/
p = filename + AR_name_size - 1;
while (p >= filename && *p == ' ') {
*p = '\0';
p--;
}
/*
* Remove trailing slash unless first character is '/'.
* (BSD entries never end in '/', so this will only trim
* GNU-format entries. GNU special entries start with '/'
* and are not terminated in '/', so we don't trim anything
* that starts with '/'.)
*/
if (filename[0] != '/' && p > filename && *p == '/') {
*p = '\0';
}
if (p < filename) {
archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"Found entry with empty filename");
return (ARCHIVE_FATAL);
}
/*
* '//' is the GNU filename table.
* Later entries can refer to names in this table.
*/
if (strcmp(filename, "//") == 0) {
/* This must come before any call to _read_ahead. */
ar_parse_common_header(ar, entry, h);
archive_entry_copy_pathname(entry, filename);
archive_entry_set_filetype(entry, AE_IFREG);
/* Get the size of the filename table. */
number = ar_atol10(h + AR_size_offset, AR_size_size);
if (number > SIZE_MAX || number > 1024 * 1024 * 1024) {
archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"Filename table too large");
return (ARCHIVE_FATAL);
}
entry_size = (size_t)number;
if (entry_size == 0) {
archive_set_error(&a->archive, EINVAL,
"Invalid string table");
return (ARCHIVE_FATAL);
}
if (ar->strtab != NULL) {
archive_set_error(&a->archive, EINVAL,
"More than one string tables exist");
return (ARCHIVE_FATAL);
}
/* Read the filename table into memory. */
st = malloc(entry_size);
if (st == NULL) {
archive_set_error(&a->archive, ENOMEM,
"Can't allocate filename table buffer");
return (ARCHIVE_FATAL);
}
ar->strtab = st;
ar->strtab_size = entry_size;
if (*unconsumed) {
__archive_read_consume(a, *unconsumed);
*unconsumed = 0;
}
if ((b = __archive_read_ahead(a, entry_size, NULL)) == NULL)
return (ARCHIVE_FATAL);
memcpy(st, b, entry_size);
__archive_read_consume(a, entry_size);
/* All contents are consumed. */
ar->entry_bytes_remaining = 0;
archive_entry_set_size(entry, ar->entry_bytes_remaining);
/* Parse the filename table. */
return (ar_parse_gnu_filename_table(a));
}
/*
* GNU variant handles long filenames by storing /<number>
* to indicate a name stored in the filename table.
* XXX TODO: Verify that it's all digits... Don't be fooled
* by "/9xyz" XXX
*/
if (filename[0] == '/' && filename[1] >= '0' && filename[1] <= '9') {
number = ar_atol10(h + AR_name_offset + 1, AR_name_size - 1);
/*
* If we can't look up the real name, warn and return
* the entry with the wrong name.
*/
if (ar->strtab == NULL || number >= ar->strtab_size) {
archive_set_error(&a->archive, EINVAL,
"Can't find long filename for GNU/SVR4 archive entry");
archive_entry_copy_pathname(entry, filename);
/* Parse the time, owner, mode, size fields. */
ar_parse_common_header(ar, entry, h);
return (ARCHIVE_FATAL);
}
archive_entry_copy_pathname(entry, &ar->strtab[(size_t)number]);
/* Parse the time, owner, mode, size fields. */
return (ar_parse_common_header(ar, entry, h));
}
/*
* BSD handles long filenames by storing "#1/" followed by the
* length of filename as a decimal number, then prepends the
* the filename to the file contents.
*/
if (strncmp(filename, "#1/", 3) == 0) {
/* Parse the time, owner, mode, size fields. */
/* This must occur before _read_ahead is called again. */
ar_parse_common_header(ar, entry, h);
/* Parse the size of the name, adjust the file size. */
number = ar_atol10(h + AR_name_offset + 3, AR_name_size - 3);
/* Sanity check the filename length:
* = Must be <= SIZE_MAX - 1
* = Must be <= 1MB
* = Cannot be bigger than the entire entry
*/
if (number > SIZE_MAX - 1
|| number > 1024 * 1024
|| (int64_t)number > ar->entry_bytes_remaining) {
archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"Bad input file size");
return (ARCHIVE_FATAL);
}
bsd_name_length = (size_t)number;
ar->entry_bytes_remaining -= bsd_name_length;
/* Adjust file size reported to client. */
archive_entry_set_size(entry, ar->entry_bytes_remaining);
if (*unconsumed) {
__archive_read_consume(a, *unconsumed);
*unconsumed = 0;
}
/* Read the long name into memory. */
if ((b = __archive_read_ahead(a, bsd_name_length, NULL)) == NULL) {
archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC,
"Truncated input file");
return (ARCHIVE_FATAL);
}
/* Store it in the entry. */
p = (char *)malloc(bsd_name_length + 1);
if (p == NULL) {
archive_set_error(&a->archive, ENOMEM,
"Can't allocate fname buffer");
return (ARCHIVE_FATAL);
}
strncpy(p, b, bsd_name_length);
p[bsd_name_length] = '\0';
__archive_read_consume(a, bsd_name_length);
archive_entry_copy_pathname(entry, p);
free(p);
return (ARCHIVE_OK);
}
/*
* "/" is the SVR4/GNU archive symbol table.
* "/SYM64/" is the SVR4/GNU 64-bit variant archive symbol table.
*/
if (strcmp(filename, "/") == 0 || strcmp(filename, "/SYM64/") == 0) {
archive_entry_copy_pathname(entry, filename);
/* Parse the time, owner, mode, size fields. */
r = ar_parse_common_header(ar, entry, h);
/* Force the file type to a regular file. */
archive_entry_set_filetype(entry, AE_IFREG);
return (r);
}
/*
* "__.SYMDEF" is a BSD archive symbol table.
*/
if (strcmp(filename, "__.SYMDEF") == 0) {
archive_entry_copy_pathname(entry, filename);
/* Parse the time, owner, mode, size fields. */
return (ar_parse_common_header(ar, entry, h));
}
/*
* Otherwise, this is a standard entry. The filename
* has already been trimmed as much as possible, based
* on our current knowledge of the format.
*/
archive_entry_copy_pathname(entry, filename);
return (ar_parse_common_header(ar, entry, h));
}
static int
archive_read_format_cpio_read_header(struct archive_read *a,
struct archive_entry *entry)
{
struct cpio *cpio;
const void *h, *hl;
struct archive_string_conv *sconv;
size_t namelength;
size_t name_pad;
int r;
cpio = (struct cpio *)(a->format->data);
sconv = cpio->opt_sconv;
if (sconv == NULL) {
if (!cpio->init_default_conversion) {
cpio->sconv_default =
archive_string_default_conversion_for_read(
&(a->archive));
cpio->init_default_conversion = 1;
}
sconv = cpio->sconv_default;
}
r = (cpio->read_header(a, cpio, entry, &namelength, &name_pad));
if (r < ARCHIVE_WARN)
return (r);
/* Read name from buffer. */
h = __archive_read_ahead(a, namelength + name_pad, NULL);
if (h == NULL)
return (ARCHIVE_FATAL);
if (archive_entry_copy_pathname_l(entry,
(const char *)h, namelength, sconv) != 0) {
if (errno == ENOMEM) {
archive_set_error(&a->archive, ENOMEM,
"Can't allocate memory for Pathname");
return (ARCHIVE_FATAL);
}
archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT,
"Pathname can't be converted from %s to current locale.",
archive_string_conversion_charset_name(sconv));
r = ARCHIVE_WARN;
}
cpio->entry_offset = 0;
__archive_read_consume(a, namelength + name_pad);
/* If this is a symlink, read the link contents. */
if (archive_entry_filetype(entry) == AE_IFLNK) {
if (cpio->entry_bytes_remaining > 1024 * 1024) {
archive_set_error(&a->archive, ENOMEM,
"Rejecting malformed cpio archive: symlink contents exceed 1 megabyte");
return (ARCHIVE_FATAL);
}
hl = __archive_read_ahead(a,
(size_t)cpio->entry_bytes_remaining, NULL);
if (hl == NULL)
return (ARCHIVE_FATAL);
if (archive_entry_copy_symlink_l(entry, (const char *)hl,
(size_t)cpio->entry_bytes_remaining, sconv) != 0) {
if (errno == ENOMEM) {
archive_set_error(&a->archive, ENOMEM,
"Can't allocate memory for Linkname");
return (ARCHIVE_FATAL);
}
archive_set_error(&a->archive,
ARCHIVE_ERRNO_FILE_FORMAT,
"Linkname can't be converted from %s to "
"current locale.",
archive_string_conversion_charset_name(sconv));
r = ARCHIVE_WARN;
}
__archive_read_consume(a, cpio->entry_bytes_remaining);
cpio->entry_bytes_remaining = 0;
}
/* XXX TODO: If the full mode is 0160200, then this is a Solaris
* ACL description for the following entry. Read this body
* and parse it as a Solaris-style ACL, then read the next
* header. XXX */
/* Compare name to "TRAILER!!!" to test for end-of-archive. */
if (namelength == 11 && strncmp((const char *)h, "TRAILER!!!",
11) == 0) {
/* TODO: Store file location of start of block. */
archive_clear_error(&a->archive);
return (ARCHIVE_EOF);
}
/* Detect and record hardlinks to previously-extracted entries. */
if (record_hardlink(a, cpio, entry) != ARCHIVE_OK) {
return (ARCHIVE_FATAL);
}
return (r);
}
