WTAP_ERR_UNSUPPORTED_ENCAP means "I can't *write* that particular
encapsulation type to a file of this format", which mainly means "that
file format simply can't handle packets of that type";
WTAP_ERR_UNSUPPORTED means "this file can't currently be supported by
Wireshark, as there's some feature in the file - such as a file or
per-packet encapsulation type - that we don't (yet) handle".
Change-Id: I53cadf9913d20efb2bccb29f61877b71d53807be
Reviewed-on: https://code.wireshark.org/review/5794
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Clean up some things we ran across while making those changes.
Change-Id: Ic0d8943d36e6e120d7af0a6148fad98015d1e83e
Reviewed-on: https://code.wireshark.org/review/4581
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Unlike the standard I/O routines, the code we introduced that supports
fast random seeking on gzipped files will always supply some specific
error code for read errors, so we don't need WTAP_ERR_CANT_READ.
Add WTAP_ERR_CANT_WRITE for writing, as we're still using the standard
I/O routines for that. Set errno to WTAP_ERR_CANT_WRITE before calling
fwrite() in wtap_dump_file_write(), so that it's used if fwrite() fails
without setting errno.
Change-Id: I6bf066a6838284a532737aa65fd0c9bb3639ad63
Reviewed-on: https://code.wireshark.org/review/4540
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Add wtap_read_bytes(), which takes a FILE_T, a pointer, a byte count, an
error number pointer, and an error string pointer as arguments, and that
treats a short read of any sort, including a read that returns 0 bytes,
as a WTAP_ERR_SHORT_READ error, and that returns the error number and
string through its last two arguments.
Add wtap_read_bytes_or_eof(), which is similar, but that treats a read
that returns 0 bytes as an EOF, supplying an error number of 0 as an EOF
indication.
Use those in file readers; that simplifies the code and makes it less
likely that somebody will fail to supply the error number and error
string on a file read error.
Change-Id: Ia5dba2a6f81151e87b614461349d611cffc16210
Reviewed-on: https://code.wireshark.org/review/4512
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Pcap-ng files don't have a per-file time stamp resolution, they have a
per-interface time stamp resolution. Add new time stamp resolution
types of "unknown" and "per-packet", add the time stamp resolution to
struct wtap_pkthdr, have the libwiretap core initialize it to the
per-file time stamp resolution, and have pcap-ng do the same thing with
the resolution that it does with the packet encapsulation.
Get rid of the TS_PREC_AUTO_XXX values; just have TS_PREC_AUTO, which
means "use the packet's resolution to determine how many significant
digits to display". Rename all the WTAP_FILE_TSPREC_XXX values to
WTAP_TSPREC_XXX, as they're also used for per-packet values.
Change-Id: If9fd8f799b19836a5104aaa0870a951498886c69
Reviewed-on: https://code.wireshark.org/review/4349
Reviewed-by: Guy Harris <guy@alum.mit.edu>
If it fails due to, for example, the file being gzipped and having a bad
gzip CRC, the error returned is WTAP_ERR_DECOMPRESS and, for that error,
err_info is expected to be set to a string giving details of the
problem, so we need to pass back to our caller the string in question.
Bug: 10484
Change-Id: I3aa2a92d04fcc08946ff073a40efa708079bbb3e
Reviewed-on: https://code.wireshark.org/review/4201
Reviewed-by: Guy Harris <guy@alum.mit.edu>
When trying to guess what type of capture a file is, look for as many
bogosities (caplen > len, microseconds >= 10^6/nanoseconds >= 10^9,
too-high caplen, too-high original len, caplen > snapshort length), to
increase the chances of guessing correctly.
(Every time somebody uses 0xa1b2c3d4 as the magic number for a capture
file that isn't standard pcap format, God kills a kitten. Please, think
of the kittens.)
Change-Id: I3f397d598ed61dc82e2832be30452ebe8ace98e8
Reviewed-on: https://code.wireshark.org/review/3808
Reviewed-by: Guy Harris <guy@alum.mit.edu>
In particular, epan/wslua/lrexlib.c has its own buffer_ routines,
causing some linker warnings on some platforms, as reported in bug
10332.
(Not to be backported to 1.12, as that would change the API and ABI of
libwsutil and libwiretap. We should also make the buffer_ routines in
epan/wslua/lrexlib.c static, which should also address this problem, but
the name change avoids other potential namespace collisions.)
Change-Id: I1d42c7d1778c7e4c019deb2608d476c52001ce28
Reviewed-on: https://code.wireshark.org/review/3351
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Otherwise, if you link with both libwiretap and libfiletap, it's
anybody's guess which one you get. That means you're wasting memory
with two copies of its routines if they're identical, and means
surprising behavior if they're not (which showed up when I was debugging
a double-free crash - fixing libwiretap's buffer_free() didn't fix the
problem, because Wireshark happened to be calling libfiletap' unfixed
buffer_free()).
There's nothing *tap-specific about Buffers, anyway, so it really
belongs in wsutil.
Change-Id: I91537e46917e91277981f8f3365a2c0873152870
Reviewed-on: https://code.wireshark.org/review/3066
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Also, make the block of code containing that comment intended
consistently with spaces.
Change-Id: I8e8eb346833662f15c53ece5869b12cc430bad11
Reviewed-on: https://code.wireshark.org/review/2661
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Add a "record type" field to "struct wtap_pkthdr"; currently, it can be
REC_TYPE_PACKET, for a record containing a packet, or
REC_TYPE_FILE_TYPE_SPECIFIC, for records containing file-type-specific
data.
Modify code that reads packets to be able to handle non-packet records,
even if that just means ignoring them.
Rename some routines to indicate that they handle more than just
packets.
We don't yet have any libwiretap code that supplies records other than
REC_TYPE_PACKET or that supporting writing records other than
REC_TYPE_PACKET, or any code to support plugins for handling
REC_TYPE_FILE_TYPE_SPECIFIC records; this is just the first step for bug
8590.
Change-Id: Idb40b78f17c2c3aea72031bcd252abf9bc11c813
Reviewed-on: https://code.wireshark.org/review/1773
Reviewed-by: Guy Harris <guy@alum.mit.edu>
This reverts commit c0c480d08c.
A better way to do this is to have the record type be part of struct wtap_pkthdr; that keeps the metadata for the record together and requires fewer API changes. That is in-progress.
Change-Id: Ic558f163a48e2c6d0df7f55e81a35a5e24b53bc6
Reviewed-on: https://code.wireshark.org/review/1741
Reviewed-by: Guy Harris <guy@alum.mit.edu>
This is the first step towards implementing the mechanisms requestd in
bug 8590; currently, we don't return any records other than packet
records from libwiretap, and just ignore non-packet records in the rest
of Wireshark, but this at least gets the ball rolling.
Change-Id: I34a45b54dd361f69fdad1a758d8ca4f42d67d574
Reviewed-on: https://code.wireshark.org/review/1736
Reviewed-by: Guy Harris <guy@alum.mit.edu>
This reverts commit 1abeb277f5.
This isn't building, and looks as if it requires significant work to fix.
Change-Id: I622b1bb243e353e874883a302ab419532b7601f2
Reviewed-on: https://code.wireshark.org/review/1568
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Start of refactoring Wiretap and breaking structures down into "generally useful fields for dissection" and "capture specific". Since this in intended as a "base" for Wiretap and Filetap, the "wft" prefix is used for "common" functionality.
The "architectural" changes can be found in cfile.h, wtap.h, wtap-int.h and (new file) wftap-int.h. Most of the other (painstaking) changes were really just the result of compiling those new architecture changes.
bug:9607
Change-Id: Ife858a61760d7a8a03be073546c0e7e582cab2ae
Reviewed-on: https://code.wireshark.org/review/1485
Reviewed-by: Michael Mann <mmann78@netscape.net>
(Using sed : sed -i '/^ \* \$Id\$/,+1 d')
Fix manually some typo (in export_object_dicom.c and crc16-plain.c)
Change-Id: I4c1ae68d1c4afeace8cb195b53c715cf9e1227a8
Reviewed-on: https://code.wireshark.org/review/497
Reviewed-by: Anders Broman <a.broman58@gmail.com>
For some routines that take multiple arguments that come from a struct
wtap_pkthdr, pass a pointer to the struct wtap_pkthdr in question,
rather than the separate arguments. Do this even if we're passing
expressions that were earlier assigned to the struct wtap_pkthdr fields
in question. This simplifies the calling sequences and ensures that the
right values are picked up by the called routine; in at least one case
we were *not* passing the right values (the code to handle Simple Packet
Blocks in pcap-ng files).
Also, call the byte-swapping routines for pseudo-header fields only if
we need to do byte-swapping.
Change-Id: I3a8badfcfeb0237dfc1d1014185a67f18c0f2ebe
Reviewed-on: https://code.wireshark.org/review/119
Reviewed-by: Evan Huus <eapache@gmail.com>
Reviewed-by: Guy Harris <guy@alum.mit.edu>
willing to read or that's bigger than will fit in the file format;
instead, report an error.
For the "I can't write a packet of that type in that file type" error,
report the file type in question.
svn path=/trunk/; revision=54882
heuristic cases broken in r49999 when we permitted packets > 64KB, since that
relaxed so severely the definition of a valid packet header.
64MB is an arbitrary and perhaps suboptimal number, but it seems to do the right
thing in all the examples I have handy.
Fixes https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=9634
svn path=/trunk/; revision=54812
XXX, people are not aware that expression of this macros might be evaluated multiple times, like:
- BSWAP16(tvb_get_letohs(tvb, off)) : \
+ GUINT16_SWAP_LE_BE(tvb_get_letohs(tvb, off)) : \
Should be tvb_get_ntohs() called?
svn path=/trunk/; revision=53653
subtypes, e.g. Network Monitor version 1 and Network Monitor version 2
are separate "file types", even though they both come from Network
Monitor.
Rename various functions, #defines, and variables appropriately.
svn path=/trunk/; revision=53166
out there (especially over USB) and we should be able to load them as long as
they are snapped to a sane length.
Also validate that packets do not specify a snapshot length larger than the one
in the file header, though only make it a warning, as this is not necessarily a
fatally corrupt packet.
https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=8808
svn path=/trunk/; revision=49999
as the "where to put the packet data" argument.
This lets more of the libwiretap code be common between the read and
seek-read code paths, and also allows for more flexibility in the "fill
in the data" path - we can expand the buffer as needed in both cases.
svn path=/trunk/; revision=49949
routines are passed a separate struct wtap_pkthdr to be filled in.
Get rid of the pseudo_header member of the wblock structure - the
pseudo-header is part of the struct wtap_pkthdr.
Get rid of the union wtap_pseudo_header * argument to
pcap_process_pseudo_header() - it's passed a pointer to a struct
pcap_pkthdr, and that structure contains the union in question.
Have libpcap_read_header() take a FILE_T argument, rather than using
only the "sequential" handle of the wtap it's handed. Have the libpcap
read routine return the offset of the beginning of the pcap record, and
have the seek-read routine read the header and fill in the struct
wtap_pkthdr handed to it.
svn path=/trunk/; revision=49401
leads to a double-free in wtap_close. Fix all the instances I found via
manual code review, and add a brief comment to the list of open routines in
file_access.c
Fixes https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=8518
svn path=/trunk/; revision=48552
wtap_file_read_expected_bytes() from an open routine - open routines are
supposed to return -1 on error, 0 if the file doesn't appear to be a
file of the specified type, or 1 if the file does appear to be a file of
the specified type, but those macros will cause the caller to return
FALSE on errors (so that, even if there's an I/O error, it reports "the
file isn't a file of the specified type" rather than "we got an error
trying to read the file").
When doing reads in an open routine before we've concluded that the file
is probably of the right type, return 0, rather than -1, if we get
WTAP_ERR_SHORT_READ - if we don't have enough data to check whether a
file is of a given type, we should keep trying other types, not give up.
For reads done *after* we've concluded the file is probably of the right
type, if a read doesn't return the number of bytes we asked for, but
returns an error of 0, return WTAP_ERR_SHORT_READ - the file is
apparently cut short.
For NetMon and NetXRay/Windows Sniffer files, use a #define for the
magic number size, and use that for both magic numbers.
svn path=/trunk/; revision=46803
(otherwise, nobody sees them); do so.
Fix some cases where we weren't doing the proper post-processing after
doing the heuristics for format changes that didn't involve magic-number
changes (discovered because constructing the interface list is now being
done as part of that post-processing).
Fixes bug 7287.
svn path=/trunk/; revision=42872
which could use lseek() and were thus expensive due to system call
overhead. To avoid making a system call for every packet on a
sequential read, we maintained a data_offset field in the wtap structure
for sequential reads.
It's now a routine that just returns information from the FILE_T data
structure, so it's cheap. Use it, rather than maintaining the data_offset
field.
Readers for some file formats need to maintain file offset themselves;
have them do so in their private data structures.
svn path=/trunk/; revision=42423
by Wiretap, to indicate whether certain fields in that structure
actually have data in them.
Use the "time stamp present" flag to omit showing time stamp information
for packets (and "packets") that don't have time stamps; don't bother
working very hard to "fake" a time stamp for data files.
Use the "interface ID present" flag to omit the interface ID for packets
that don't have an interface ID.
We don't use the "captured length, separate from packet length, present"
flag to omit the captured length; that flag might be present but equal
to the packet length, and if you want to know if a packet was cut short
by a snapshot length, comparing the values would be the way to do that.
More work is needed to have wiretap/pcapng.c properly report the flags,
e.g. reporting no time stamp being present for a Simple Packet Block.
svn path=/trunk/; revision=41185
form of corruption/bogosity in a file, including in a file header as
well as in records in the file. Change the error message
wtap_strerror() returns for it to reflect that.
Use it for some file header problems for which it wasn't already being
used - WTAP_ERR_UNSUPPORTED shouldn't be used for that, it should only
be used for files that we have no reason to believe are invalid but that
have a version number we don't know about or some other
non-link-layer-encapsulation-type value we don't know about.
svn path=/trunk/; revision=40175
same.
Add to wiretap/pcap-common.c a routine to fill in the pseudo-header for
ATM (by looking at the VPI, VCI, and packet data, and guessing) and
Ethernet (setting the FCS length appropriately). Use it for both pcap
and pcap-ng files.
svn path=/trunk/; revision=38840
by the gunzipping code. Have it also supply a err_info string, and
report it. Have file_error() supply an err_info string.
Put "the file" - or, for WTAP_ERR_DECOMPRESS, "the compressed file", to
suggest a decompression error - into the rawshark and tshark errors,
along the lines of what other programs print.
Fix a case in the Netscaler code where we weren't fetching the error
code on a read failure.
svn path=/trunk/; revision=36748
can't be saved in compress form" are both equivalent to "this file file
format requires seeking when writing it". Change the "can compress"
Boolean in the file format table to "writing requires seeking", give all
the entries the proper value, and do the checks for attempting to write
a file format to a pipe or write it in compressed format to common code.
This means we don't need to pass the "can't seek" flag to the dump open
routines.
svn path=/trunk/; revision=36575
file_read(buf, bsize, count, file) macro is compilant with fread
function and takes elements count+ size of each element, however to make
it compilant with gzread() it always returns number of bytes.
In wiretap file_read() this is not really used, file_read is called
either with bsize set to 1 or count to 1.
Attached patch remove bsize argument from macro.
svn path=/trunk/; revision=36491
everybody use it; the places using the old wtap_dump_file_write() were
using it in the same way the old wtap_dump_file_write_all() did.
That also lets us get rid of wtap_dump_file_ferror().
Also, have the new wtap_dump_file_write() check for errors from
gzwrite() and fwrite() differently - the former returns 0 on error, the
latter can return a short write on error.
svn path=/trunk/; revision=33113
Guy Harris