It's now possible to run commands as other users without shell
interpolation by using "--exec":
Read /etc/shadow as root without specifying user:
```
su --exec /bin/cat -- /etc/shadow
```
Or specify user:
```
su --exec /bin/cat root -- /etc/shadow
```
Mechanical rename distinguishing this variable from intended changes
supporting executing commands without using an interpretive shell
(i.e. no '/bin/sh -c').
In preparation for supporting --exec I was testing the robustness
of "--" handling and it became apparent that things are currently
a bit broken in `su`.
Since "--" is currently of limited utility, as the subsequent
words are simply passed to the shell after "-c","command_string",
it seems to have gone unnoticed for ages.
However, with --exec, it's expected that "--" would be an almost
required separator with every such usage, considering the
following flags must be passed verbatim to execve() and will
likely begin with hyphens looking indistinguishable from any
other flags in lieu of shell interpolation to worry about.
For some practical context of the existing situation, this
invocation doesn't work today:
```
$ su --command ls -- flags for shell
No passwd entry for user 'flags'
$
```
This should just run ls as root with "flags","for","shell"
forwarded to the shell after "-c","ls".
The "--" should block "flags" from being treated as the user.
That particular issue isn't a getopt one per-se, it's arguably
just a bug in su.c's implementation.
It *seemed* like an easy fix for this would be to add a check if
argv[optind-1] were "--" before treating argv[optind] as USER.
But testing that fix revealed getopt was rearranging things when
encountering "--", the "--" would always separate the handled
opts from the unhandled ones. USER would become shifted to
*after* "--" even when it occurred before it!
If we change the command to specify the user, it works as-is:
```
$ su --command ls root -- flags for shell
Password:
testfile
$
```
But what's rather surprising is how that works; the argv winds up:
"su","--command","ls","--","root","flags","for","shell"
with optind pointing at "root".
That arrangement of argv is indistinguishable from omitting the
user and having "root","flags","for","shell" as the stuff after
"--".
This makes it non-trivial to fix the bug of omitting user
treating the first word after "--" as the user, which one could
argue is a potentially serious security bug if you omit the user,
expect the command to run as root, and the first word after "--"
is a valid user, and what follows that something valid and
potentially destructive not only running in unintended form but
as whatever user happened to be the first word after "--".
So, it seems like something important to fix, and getopt seems to
be getting in the way of fixing it properly without being more
trouble than replacing getopt.
In disbelief of what I was seeing getopt doing with argv here, I
took a glance at the getopt source and found the following:
```
/* The special ARGV-element '--' means premature end of options.
Skip it like a null option,
then exchange with previous non-options as if it were an option,
then skip everything else like a non-option. */
if (d->optind != argc && !strcmp (argv[d->optind], "--"))
```
I basically never use getopt personally because ages ago it
annoyed me with its terrible API for what little it brought to
the table, and this brings it to a whole new level of awful.
This is a stability fix, not a security fix, because the affected -o
option can only be used by root and it takes a modified passwd file.
If a gecos field for a user has BUFSIZ characters without commas and an
equals sign (i.e. a huge slop/extra field) and chfn is called with -o,
then a buffer overflow occurs.
It is not possible to trigger this with shadow tools. Therefore, the
passwd file must be modified manually.
I have fixed this unlikely case the easiest and cleanest way possible.
Since chfn bails out if more than 80 characters excluding commas are
supposed to be written into gecos field, we can stop processing early on
if -o argument is too long.
Signed-off-by: Tobias Stoeckmann <tobias@stoeckmann.org>
This is merely a stability fix, not a security fix.
As the root user, it is possible to set time values which later on
result in signed integer overflows.
For this to work, an sgetspent implementation must be used which
supports long values (glibc on amd64 only parses 32 bit, not 64).
Either use musl or simply call configure with following environment
variable:
$ ac_cv_func_sgetspent=no ./configure
Also it is recommended to compile with -fsanitize=undefined or
-ftrapv to see these issues easily.
Examples to trigger issues when calling "chage -l user":
$ chage -d 9223372036854775807 user
$ chage -d 106751991167300 user
$ chage -M 9999 user
$ chage -d 90000000000000 user
$ chage -I 90000000000000 user
$ chage -M 9999 user
$ chage -E 9223372036854775807 user
While at it, I fixed casting issues which could lead to signed integer
overflows on systems which still have a 32 bit time_t.
Signed-off-by: Tobias Stoeckmann <tobias@stoeckmann.org>
Closes#154
Currently this has three functions: one which returns the
list of subuid ranges for a user, one returning the subgids,
and one which frees the ranges lists.
I might be mistaken about what -disable-man means; some of
the code suggests it means just don't re-generate them, but
not totally ignore them. But that doesn't seem to really work,
so let's just ignore man/ when -disable-man.
Remove --disable-shared. I'm not sure why it was there, but it stems
from long, long ago, and I suspect it comes from some ancient
toolchain bug.
Create a tests/run_some, a shorter version of run_all. I'll
slowly add tests to this as I verify they work, then I can
work on fixing the once which don't.
Also, don't touch man/ if not -enable-man.
Changelog:
Apr 22: change the subid list api as recomended by Dan Walsh.
Apr 23: implement get_subid_owner
Apr 24: implement range add/release
Apr 25: finish tests and rebase
May 10: make @owner const
Signed-off-by: Serge Hallyn <serge@hallyn.com>
Now that travis supports more architectures let's make sure we test on
all of them and that we enable Coverity too.
Signed-off-by: Christian Brauner <christian@brauner.io>
