The error handling is performed after the loop. By just calling break it
is possible to reuse the error handling if status is not ERANGE.
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
A failure of DUP_FUNCTION is already handled for non-reentrant
function wrapper. Perform the check for reentrant version as well.
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
OpenSSH and coreutils' chroot call chroot first and then chdir. Doing it
this way is a bit safer because otherwise something could happen between
chdir and chroot to the specified path (like exchange of links) so the
working directory would not end up within the chroot environment.
This is a purely defensive measure.
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
If a user has home directory "/" and login shell "*" then login and su
enter an endless loop by constantly switching to the next subsystem.
This could also be achieved with a layered approach so just checking
for "/" as home directory is not enough to protect against such a
misconfiguration.
Just break the loop if it progressed too far. I doubt that this has
negative impact on any real setup.
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
Using --prefix in a setuid binary is quite dangerous. An unprivileged
user could prepare a custom shadow file in home directory. During a data
race the user could exchange directories with links which could lead to
exchange of shadow file in system's /etc directory.
This could be used for local privilege escalation.
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
Some strings are first written into static char arrays before passed to
functions which expect a const char pointer anyway.
It is easier to pass these strings directly as arguments.
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
The only user of login_prompt is the login tool. This implies that the
first argument is always the same.
It is much easier to verify printf's format string and its argument if
both are next to each other.
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
Parsing optional environment variables after a login name is a feature
which is neither documented nor available in util-linux or busybox
login which are other wide spread login utilities used in Linux
distributions as reference.
Removing this feature resolves two issues:
- A memory leak exists if variables without an equal sign are used,
because set_env creates copies on its own. This could lead to OOM
situations in privileged part of login or may lead to heap spraying.
- Environment variables are not reset between login attempts. This
could lead to additional environment variables set for a user who
never intended to do so.
Proof of Concept on a system with shadow login without PAM and
util-linux agetty:
1. Provoke an invalid login, e.g. user `noone` and password `invalid`.
This starts shadow login and subsequent inputs are passed through
the function login_prompt.
2. Provoke an invalid login with environment variables, e.g.
user `noone HISTFILE=/tmp/owo` and password `invalid`.
3. Log in correctly with user `root`.
Now you can see with `echo $HISTFILE` that `/tmp/owo` has been set for
the root user.
This requires a malicious failed login attempt and a successful login
within the configured login timeout (default 60 seconds).
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
The getline function does not return a pointer but the amount of read
characters. The error return value to check for is -1.
Set buf to NULL to avoid dereference of an uninitialized stack value.
The getline function returns -1 if size argument is NULL. Always use
a valid pointer even if size is unimportant.
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
Add the relevant XKCD to the passwd(1) manual page. It already explains
most of the rationale behind this patch.
Add also reference to makepasswd(1), which is a good way to generate
strong passwords. Personally, I commonly run `makepasswd --chars 64` to
create my passwords, or 32 for passwords I need to type interactively
often.
The strength of a password is an exponential formula, where the base is
the size of the character set, and the exponent is the length of the
password. That already shows why long passwords of just lowercase
letters are better than short Pa$sw0rdZ3. But an even more important
point is that humans, when forced to use symbols in a password, are more
likely to do trivial substitutions on simple passwords, which doesn't
increase strength, and can instead give a false sense of strength, which
is dangerous.
Closes: <https://github.com/shadow-maint/shadow/issues/688>
Link: <https://xkcd.com/936/>
Cc: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
getline(3) is much more readable than manually looping. It has some
overhead due to the allocation of a buffer, but that shouldn't be a
problem here. If that was a problem, we could reuse the buffer (thus
making the function non-reentrant), but I don't think that's worth the
extra complexity.
Using rpmatch(3) instead of a simple y/n test provides i18n to the
response checking. We have a fall-back minimalistic implementation for
systems that lack this function (e.g., musl libc).
While we're at it, apply some other minor improvements to this file:
- Remove comment saying which files use this function. That's likely
to get outdated. And anyway, it's just a grep(1) away, so it doesn't
really add any value.
- Remove unnecessary casts to (void) that were used to verbosely ignore
errors from stdio calls. They add clutter without really adding much
value to the code (or I don't see it).
- Remove comments from the function body. They make the function less
readable. Instead, centralize the description of the function into a
man-page-like comment before the function definition. This keeps the
function body short and sweet.
- Add '#include <stdbool.h>', which was missing.
- Minor whitespace style changes (it doesn't hurt the diff at this
point, since most of the affected lines were already touched by other
changes, so I applied my preferred style :).
Acked-by: Samanta Navarro <ferivoz@riseup.net>
Cc: Serge Hallyn <serge@hallyn.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
Do not stop after 79 characters. Read the complete line to avoid
arbitrary limitations.
Proof of Concept:
```
cat > passwd-poc << EOF
root❌0:0:root:/root:/bin/bash
root❌0:0:root:/root:/bin/bash
root❌0:0:root:/root:/bin/bash
EOF
python -c "print(80*'y')" | pwck passwd-poc
```
Two lines should still be within the file because we agreed only once
to remove a duplicated line.
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
Reviewed-by: Alejandro Colomar <alx@kernel.org>
Reviewed-by: Serge Hallyn <serge@hallyn.com>
This commit will serve to document why we shouldn't worry about the
truncation in the call to strlcpy(3). Since we have one more byte in
tmptty than in full_tty, truncation will produce a string that is at
least one byte longer than full_tty. Such a string could never compare
equal, so we're actually handling the truncation in a clever way. Maybe
too clever, but that's why I'm documenting it here.
Now, about the simplification itself:
Since we made sure that both full_tty and tmptty are null-terminated, we
can call strcmp(3) instead of strncmp(3). We can also simplify the
return logic avoiding one branch.
Cc: Paul Eggert <eggert@cs.ucla.edu>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
Reviewed-by: Iker Pedrosa <ipedrosa@redhat.com>
* libmisc/utmp.c (is_my_tty): Declare the parameter as a char array,
not char *, as it is not necessarily null-terminated.
Avoid a read overrun when reading 'tty', which comes from
'ut_utname'.
Reported-by: Paul Eggert <eggert@cs.ucla.edu>
Co-developed-by: Paul Eggert <eggert@cs.ucla.edu>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
Reviewed-by: Iker Pedrosa <ipedrosa@redhat.com>
* libmisc/date_to_str.c (date_to_str): Do not crash if gmtime(3)
returns NULL because the timestamp is far in the future.
Reported-by: Paul Eggert <eggert@cs.ucla.edu>
Co-developed-by: Paul Eggert <eggert@cs.ucla.edu>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
Reviewed-by: Iker Pedrosa <ipedrosa@redhat.com>
gettime.c:25:30: warning: a function declaration without a prototype is deprecated in all versions of C [-Wstrict-prototypes]
/*@observer@*/time_t gettime ()
^
void
We can't use a pointer that was input to realloc(3), nor any pointers
that point to reallocated memory, without making sure that the memory
wasn't moved. If we do, the Behavior is Undefined.
Signed-off-by: Alejandro Colomar <alx@kernel.org>
Use of these macros, apart from the benefits mentioned in the commit
that adds the macros, has some other good side effects:
- Consistency in getting the size of the object from sizeof(type),
instead of a mix of sizeof(type) sometimes and sizeof(*p) other
times.
- More readable code: no casts, and no sizeof(), so also shorter lines
that we don't need to cut.
- Consistency in using array allocation calls for allocations of arrays
of objects, even when the object size is 1.
Cc: Valentin V. Bartenev <vbartenev@gmail.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
We'll expand the contents in a following commit, so let's move the file
to a more generic name, have a dedicated header, and update includes.
Signed-off-by: Alejandro Colomar <alx@kernel.org>
Use the new header for xstrdup()
Signed-off-by: Alejandro Colomar <alx@kernel.org>
On 2/19/23 18:09, David Mudrich wrote:
> I am working on a RAM based Linux OS from source, and try to use
> latest versions of all software. I found shadow needs libbsd's
> readpassphrase(3) as superior alternative to getpass(3). While
> considering if I a) include libbsd, or include libbsd's code of
> readpassphrase(3) into shadow, found, that libbsd's readpassphrase(3)
> never returns \n or \r
> <https://cgit.freedesktop.org/libbsd/tree/src/readpassphrase.c>
> line 122, while agetpass() uses a check for \n in agetpass.c line 108.
> I assume it always fails.
Indeed, it always failed. I made a mistake when writing agetpass(),
assuming that readpassphrase(3) would keep newlines.
>
> I propose a check of len == PASS_MAX - 1, with false positive error for
> exactly PASS_MAX - 1 long passwords.
Instead, I added an extra byte to the allocation to allow a maximum
password length of PASS_MAX (which is the maximum for getpass(3), which
we're replacing.
While doing that, I notice that my previous implementation also had
another bug (minor): The maximum password length was PASS_MAX - 1
instead of PASS_MAX. That's also fixed in this commit.
Reported-by: David Mudrich <dmudrich@gmx.de>
Fixes: 155c9421b9 ("libmisc: agetpass(), erase_pass(): Add functions for getting passwords safely")
Cc: Iker Pedrosa <ipedrosa@redhat.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
Some programs don't support `(uint16_t) -1` or `(uint32_t) -1` as user
or group IDs. This is because `-1` is used as an error code or as an
unspecified ID, e.g. in `chown(2)` parameters, and in the past, `gid_t`
and `uid_t` have changed width. For legacy reasons, those values have
been kept reserved in programs today (for example systemd does this; see
the documentation in the link below).
This should not be confused with catching overflow in the ID values,
since that is already caught by our ERANGE checks. This is about not
using reserved values that have been reserved for legacy reasons.
Link: <https://systemd.io/UIDS-GIDS/>
Reviewed-by: Alejandro Colomar <alx@kernel.org>
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
These comments should indicate which functions they really wrap.
An alternative would be to remove the line completely to avoid
future copy&paste mistakes.
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
strncat(3), strlcpy(3), and many other functions are often misused for
catenating strings, when they should never be used for that. strlcat(3)
is good. However, there's no equivalent to strlcat(3) similar to
snprintf(3). Let's add stpecpy(), which is similar to strlcat(3), but
it is also the only function compatible with stpeprintf(), which makes
it more useful than strlcat(3).
Signed-off-by: Alejandro Colomar <alx@kernel.org>
This function allows reducing error checking (since errors are
propagated across chained calls), and also simplifies the calculation of
the start and end of the buffer where the string should be written.
Moreover, the new code is more optimized, since many calls to strlen(3)
have been removed.
Signed-off-by: Alejandro Colomar <alx@kernel.org>
[v]stpeprintf() are similar to [v]snprintf(3), but they allow chaining.
[v]snprintf(3) are very dangerous for catenating strings, since the
obvious ways to do it invoke Undefined Behavior, and the ways that avoid
UB are very error-prone.
Cc: Iker Pedrosa <ipedrosa@redhat.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
The intention of the code is just to not report an error message when
'typefile' doesn't exist. If we call access(2) and then fopen(2),
there's a race. It's not a huge problem, and the worst thing that can
happen is reporting an error when the file has been removed after
access(2). It's not a problem, but we can fix the race and at the same
time clarify the intention of not warning about ENOENT and also remove
one syscall. Seems like a win-win.
Suggested-by: Christian Göttsche <cgzones@googlemail.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
- Every non-const pointer converts automatically to void *.
- Every pointer converts automatically to void *.
- void * converts to any other pointer.
- const void * converts to any other const pointer.
- Integer variables convert to each other.
I changed the declaration of a few variables in order to allow removing
a cast.
However, I didn't attempt to edit casts inside comparisons, since they
are very delicate. I also kept casts in variadic functions, since they
are necessary, and in allocation functions, because I have other plans
for them.
I also changed a few casts to int that are better as ptrdiff_t.
This change has triggered some warnings about const correctness issues,
which have also been fixed in this patch (see for example src/login.c).
Signed-off-by: Alejandro Colomar <alx@kernel.org>
In variadic functions we still do the cast. In POSIX, it's not
necessary, since NULL is required to be of type 'void *', and 'void *'
is guaranteed to have the same alignment and representation as 'char *'.
However, since ISO C still doesn't mandate that, and moreover they're
doing dubious stuff by adding nullptr, let's be on the cautious side.
Also, C++ requires that NULL is _not_ 'void *', but either plain 0 or
some magic stuff.
Signed-off-by: Alejandro Colomar <alx@kernel.org>
If the file referenced by ENV_TZ has a zero length string, then an out
of boundary write occurs. Also the result can be wrong because it is
assumed that the file will always end with a newline.
Only override a newline character with '\0' to avoid these cases.
This cannot be considered to be security relevant because login.defs
and its contained references to system files should be trusted to begin
with.
Proof of Concept:
1. Compile shadow's su with address sanitizer and --without-libpam
2. Setup your /etc/login.defs to contain ENV_TZ=/etc/tzname
3. Prepare /etc/tzname to contain a '\0' byte at the beginning
`python -c "print('\x00')" > /etc/tzname`
4. Use su
`su -l`
You can see the following output:
`tz.c:45:8: runtime error: index 18446744073709551615 out of bounds for type 'char [8192]'`
Signed-off-by: Samanta Navarro <ferivoz@riseup.net>
__int128, which is needed for optimizing that part of the range, is not
always available. We need the unoptimized version for portability
reasons.
Closes: <https://github.com/shadow-maint/shadow/issues/634>
Fixes: 1a0e13f94e ("Optimize csrand_uniform()")
Reported-by: Adam Sampson <ats@offog.org>
Cc: Iker Pedrosa <ipedrosa@redhat.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
We do need the unoptimized version of csrand_uniform() for high values
of `n`, since the optimized version depends on having __int128, and it's
not available on several platforms, including ARMv7, IA32, and MK68k.
This reverts commit 848f53c1d3c1362c86d3baab6906e1e4419d2634; however,
I applied some tweaks to the reverted commit.
Reported-by: Adam Sampson <ats@offog.org>
Cc: Iker Pedrosa <ipedrosa@redhat.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
Now that we optimized csrand_uniform(), we don't need these functions.
This reverts commit 7c8fe291b1260e127c10562bfd7616961013730f.
Signed-off-by: Alejandro Colomar <alx@kernel.org>
Use a different algorithm to minimize rejection. This is essentially
the same algorithm implemented in the Linux kernel for
__get_random_u32_below(), but written in a more readable way, and
avoiding microopimizations that make it less readable.
Which (the Linux kernel implementation) is itself based on Daniel
Lemire's algorithm from "Fast Random Integer Generation in an Interval",
linked below. However, I couldn't really understand that paper very
much, so I had to reconstruct the proofs from scratch, just from what I
could understand from the Linux kernel implementation source code.
I constructed some graphical explanation of how it works, and why it
is optimal, because I needed to visualize it to understand it. It is
published in the GitHub pull request linked below.
Here goes a wordy explanation of why this algorithm based on
multiplication is better optimized than my original implementation based
on masking.
masking:
It discards the extra bits of entropy that are not necessary for
this operation. This works as if dividing the entire space of
possible csrand() values into smaller spaces of a size that is
a smaller power of 2. Each of those smaller spaces has a
rejection band, so we get as many rejection bands as spaces
there are. For smaller values of 'n', the size of each
rejection band is smaller, but having more rejection bands
compensates for this, and results in the same inefficiency as
for large values of 'n'.
multiplication:
It divides the entire space of possible random numbers in
chunks of size exactly 'n', so that there is only one rejection
band that is the remainder of `2^64 % n`. The worst case is
still similar to the masking algorithm, a rejection band that is
almost half the entire space (n = 2^63 + 1), but for lower
values of 'n', by only having one small rejection band, it is
much faster than the masking algorithm.
This algorithm, however, has one caveat: the implementation
is harder to read, since it relies on several bitwise tricky
operations to perform operations like `2^64 % n`, `mult % 2^64`,
and `mult / 2^64`. And those operations are different depending
on the number of bits of the maximum possible random number
generated by the function. This means that while this algorithm
could also be applied to get uniform random numbers in the range
[0, n-1] quickly from a function like rand(3), which only
produces 31 bits of (non-CS) random numbers, it would need to be
implemented differently. However, that's not a concern for us,
it's just a note so that nobody picks this code and expects it
to just work with rand(3) (which BTW I tried for testing it, and
got a bit confused until I realized this).
Finally, here's some light testing of this implementation, just to know
that I didn't goof it. I pasted this function into a standalone
program, and run it many times to find if it has any bias (I tested also
to see how many iterations it performs, and it's also almost always 1,
but that test is big enough to not paste it here).
int main(int argc, char *argv[])
{
printf("%lu\n", csrand_uniform(atoi(argv[1])));
}
$ seq 1 1000 | while read _; do ./a.out 3; done | grep 1 | wc -l
341
$ seq 1 1000 | while read _; do ./a.out 3; done | grep 1 | wc -l
339
$ seq 1 1000 | while read _; do ./a.out 3; done | grep 1 | wc -l
338
$ seq 1 1000 | while read _; do ./a.out 3; done | grep 2 | wc -l
336
$ seq 1 1000 | while read _; do ./a.out 3; done | grep 2 | wc -l
328
$ seq 1 1000 | while read _; do ./a.out 3; done | grep 2 | wc -l
335
$ seq 1 1000 | while read _; do ./a.out 3; done | grep 0 | wc -l
332
$ seq 1 1000 | while read _; do ./a.out 3; done | grep 0 | wc -l
331
$ seq 1 1000 | while read _; do ./a.out 3; done | grep 0 | wc -l
327
This isn't a complete test for a cryptographically-secure random number
generator, of course, but I leave that for interested parties.
Link: <https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=e9a688bcb19348862afe30d7c85bc37c4c293471>
Link: <https://github.com/shadow-maint/shadow/pull/624#discussion_r1059574358>
Link: <https://arxiv.org/abs/1805.10941>
Cc: "Jason A. Donenfeld" <Jason@zx2c4.com>
Cc: Cristian Rodríguez <crrodriguez@opensuse.org>
Cc: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Cc: Björn Esser <besser82@fedoraproject.org>
Cc: Yann Droneaud <ydroneaud@opteya.com>
Cc: Joseph Myers <joseph@codesourcery.com>
Cc: Sam James <sam@gentoo.org>
Cc: Serge Hallyn <serge@hallyn.com>
Cc: Iker Pedrosa <ipedrosa@redhat.com>
[Daniel Lemire: Added link to research paper in source code]
Cc: Daniel Lemire <daniel@lemire.me>
Signed-off-by: Alejandro Colomar <alx@kernel.org>
Samanta Navarro
Alejandro Colomar
skyler-ferrante
Christian Göttsche
Martin Kletzander