the string tokenisation must be performed by the caller (which is
generally easy because it's almost always a static command).
* change do_conversion() to return a char *argv[] instead of char *cmd.
tokenisation of the command is done internally.
* change retrieve() to take char *argv[] instead of char *cmd.
(to take advantage of the above changes). fixes [bin/8173]
* use fparseln() instead of fgetln()
* store conversions in listed order (rather than reverse order)
* use stringlists instead of handrolling code to manage an argv.
connection destination, hoping this to help ftpd's behavior with
scoped IPv6 addresses.
I'm not sure if it is the right way, but it is the best way available to us.
LPRT or EPRT command gives no information about which interface (or scope)
to be used for new data connection.
ftp(1): On data connection establishment, warn if scoped address is used.
If peer (ftp daemon) does not handle scoped address, data connection
may not work right.
This seems to be sort of protocol spec hole, not implementation issue.
(this was broken in the last commit). problem noticed by simonb@
* don't display the stderr output of the internal ls.
* modify usage of lreply so that generally only one `XXX-' code per
`block' is displayed; the rest of the lines have four spaces instead.
i find this easier to read.
* fix a couple places where byte accounting wasn't correct
* implement xferstats. full stats are displayed for `STAT', and a
summary is displayed upon exit (and syslogged). inspired by wu-ftpd.
* wrap data xfers in {send,receive}_data with alarm() timeouts. this
should remove the majority of the `hanging ftpd' problems that
people were still seeing. inspired by wu-ftpd.
* link with ../../bin/ls, so that bin/ls is not required under a
chroot()ed area for `LIST' to work. based on [bin/4497] from
"Soren S. Jorvang" <soren@t.dk>
* migrate code from util.c into ftpd.c, so that it doesn't conflict
with ls' util.c.
* remove man page comment about ~ftp/bin/ls being necessary.
* bump version to 7.2.0.
* syslog xfer time with xfer stats.
* if appropriate, syslog error message with command.
internal code stuff:
* change arguments of various functions from `char *' to `const char *'.
* define PLURAL(x) macro, which returns `' if x == 1, `s' otherwise.
use macro appropriately
* lreply(): a code of -1 means ``send line as is''. a code of 0
means ``send line with 4 space prefix''. don't print a space after
the `-' for any other code.
* logcmd(): add `const struct timeval *elapsed' and `const char *error'
for more flexible error reporting
solution with less code replication. use realpath() in logcmd() so
that all logged filenames are sane.
* support `REST STREAM' in `FEAT' reply (from draft-ietf-ftpext-mlst-05)
* in 'HELP', suffix unimplemented commands with `-' instead of `*'; the
former is easier to differentiate from `+'.
* deprecate curdir() now that logcmd() doesn't use it.
* ensure all filename buffers are at least MAXPATHLEN+1 in size.
* move jmp_buf errcatch out of extern.h, removing need to #include <setjmp.h>
in every file.
* implement FEAT and OPTS from RFC2389. FEAT returns SIZE and MDTM.
OPTS only works on NOOP (as a test).
* extend format of /etc/ftpchroot similar to /etc/ftpusers; each entry
can take an optional trailing `yes' or `no' which indicates if
chroot should be done (defaults to `yes').
based on patches from Ty Sarna <tsarna@endicor.com> in [bin/4769]
cleanups/bugs:
* reorder and reformat entries in yacc parser to match cmdtab[].
add a blank line between each rule.
* add short hasopts and char *options to struct tab, to support OPTS.
* deprecate upper(); use strcasecmp() instead of strcmp()
* remove unnecessary for (;;) { } in yylex();
* replace copy() and sgetsave() with xstrdup()
* fix a couple of `hasyyerrored = 1' that were accidently removed.
problem; move `hasyyerrored' state flag out of yylex() so that
check_{login,modify} can also set it.
* check result of check_login for PORT command
* set initial timeout before the "setjmp(); for(;;) yyparse()",
otherwise an invalid command after login incorrectly sets the timeout
to 5 minutes (rather than what was set in ftpd.conf)
* replace (char *)0 with NULL
* move yyerror() from ftpd.c to ftpcmd.y
* remove need for -Dunix, by using the version string from ftpd.c
(instead of `BSD-199506')
* move all extern-ed vars into extern.h
fixes long standing ftpd bug where two replies would be returned
to the client if a command was flagged as accepting `ARGS' but the
parser didn't know how to cope. obvious symptom of this would be
ftp client is always one error message `behind' the server.
* consistently refer to the RFC as `RFC 959' not `RFC959' or `RFC-959',
and replace refs to RFC 765 with RFC 959.
* change order of commands in cmdtab[] to: RFC 959, BSD extras, and obsolete.
* whitespace police, deprecate register, replace malloc/strcpy with strdup
- ensure hostname from gethostname() is nul-terminated in all cases
- minor KNF
- use MAXHOSTNAMELEN over various other values/defines
- be safe will buffers that hold hostnames
the user at the end of a transfer.
this generates a file in /tmp, so anonymous requires a
writable ~ftp/tmp, which you may not want to do (because it may
allow people to unwanted upload files).
XXX: a better method of storing the stderr output would be nice, but
is a lot more effort to implement. this feature can at least be
used temporarily whilst debugging why an ftp conversion doesn't
work.
want to timeout using this value that are executed before the conf file
is parsed will work. fixes ``stale ftpd stuck in connection phase''
* in PASV mode, wrap accept() in an alarm timeout. fixes ``stale ftpd because
of client disappearing in pasv mode (usually browsers)''
* main() doesn't need envp argument
* display current setting of checkportcmd in STAT
* ensure that curclass.checkportcmd is initialised to 0
* document default setting of checkportcmd in ftpd(8)
* cleanup code a bit, putting code for "checkportcmd" in alphabetical order
(0 == user allowed in /etc/ftpusers, 1 == user denied in /etc/ftpusers).
from Jim Bernard <jbernard@tater.mines.edu> in [security/4061] with mods
- getopt returns -1 not EOF
- in lostcon(), call dologout(1) not dologout(-1);
257 "dirname" some message
(any "s in dirname should be doubled, per the RFC)
- don't put an extra / in the output of NLST if the last char in the
directory is a /
- bump the version to 7.01 because of these fixes
from Tatoku Ogaito <tacha@tera.fukui-med.ac.jp> in [bin/3967]
* fgetln() doesn't \0 terminate its string. look for the \n and replace
it with \0 (if no \n, ignore the line - it's most likely corrupt)
* more intensive checks on strdup() returns (not a current mem leak,
but depended upon code elsewhere to cleanup - not good)
* cleanup some syslog error messages
printf format strings, you've got to make sure you cast quantities
passed to %qd to long long because on 64-bit machines they're often
just long, which is not the same, even when it's the same size.
controllable on a per class (which is one of: real, chroot, guest,
all or none) basis:
* on-the-fly execution of a command to build the file (a ``conversion''),
providing support for "get dirname.tar" and the like.
* displaying the contents of a file when a directory is entered
for the first time.
* maximum value for timeout (replaces -T).
* control usage of CHMOD, DELE, MKD, RMD, UMASK; replacing -DINSECURE_GUEST.
* notifying the user of the existance of a files matching a glob
pattern when a directory is entered for the first time.
* default value for timeout (replaces -t).
* default umask (replaces -DGUEST_CMASK and -u).
The conversion, display, and notify functionality was based on code by
Simon Burge <simonb@telstra.com.au>.
* clean up and re-order parts of the man page into subsections.
* STAT displays the settings defined for the class of the current user.
* bump version from 6.00 to 7.00, because of ftpd.conf.
* deprecate -DGUEST_CMASK and -DINSECURE_GUEST in the Makefile, and
-t, -T and -u, as ftpd.conf allows finer control of these.
* add "nostderr" argument to ftpd_popen(), because you don't want the
stderr stream mixing with the stdout stream during a conversion,
as this can corrupt the stream.
information in the same file by following the username with `allow'
or `deny'. Also, the user `*' can be used to set the default for
users not listed in the file. This is entirely backward compatable
with old /etc/ftpusers files.
Also, do the /etc/ftpusers and the valid login shell checks after
the password is verified, rather than before, so as not to give away
whether or not a particular user ID is present on the system.
* Set umask to 707;
* Disable UMASK, CHMOD, DELE, RMD and MKD commands.
Compile-time options let you change that umask and go back to the
old, insecure way if you like.
itojun