The nstype argument of setns() is an int, but it is passed to the kernel
in a 64-bit register, and the kernel truncates it to the low 32 bits.
The rule that blocks setns() with nstype == 0 compared the full 64-bit
register with SCMP_CMP_EQ, so setting any bit above bit 31 let the call
through, and the kernel then joined whichever namespace the fd referred
to, even one that RestrictNamespaces= is supposed to deny.
Use SCMP_CMP_MASKED_EQ with a 32-bit mask so we match on the same value
the kernel does.
Add a test that issues setns() via syscall() with an upper bit set.
Follow-up for #44134
Signed-off-by: Ricardo Branco <rbranco@suse.de>
(cherry picked from commit da72537164910f9050793055fef030711c00b954)
The domain argument of socket() is an int, but it is passed to the kernel
in a 64-bit register, and the kernel truncates it to the low 32 bits.
The deny-list rules we install compared the full 64-bit register with
SCMP_CMP_EQ, so setting any bit above bit 31 (e.g. AF_VSOCK | 1UL << 32)
let a denied family through unfiltered.
Use SCMP_CMP_MASKED_EQ with a 32-bit mask so we match on the same value
the kernel does.
The allow-list path is not affected: it relies on the LT/GT range checks,
which compare as unsigned 64-bit, so values with upper bits set are
always rejected.
Add a test that issues socket() via syscall() with an upper bit set.
Fixes: #44134
Signed-off-by: Ricardo Branco <rbranco@suse.de>
(cherry picked from commit 62eb6cd8eab6a35a408b2c6547526645bda55ffa)
PTRACE_POKETEXT and PTRACE_POKEDATA use FOLL_FORCE to forcibly write
into a tracee's memory. That includes executable read-only pages. If we
restrict execution to trusted binaries PID 1 must enforce that it's
impossible to abuse this to get untrusted code execution. Note, that
usermodehelpers are not subject to PID 1's seccomp filters.
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
When --suppress-sync=yes was introduced in
4a4654e024 it filtered out openat2()
completely, as we can't check its "flags" argument because it's hidden
in an indirect struct. This was perfectly fine at that time, as
openat2() was quite new and software shipped with a fallback to
open()/openat() if the syscall wasn't present.
However, today the situation is different and an increasing number of
software is moving to openat2() without any fallback - tar [0] being the
most common one in the recent reports and workarounds, and attr recently
fixed a CVE by switching to openat2() [1] as well, to name a few.
Given that we already block all the sync-family syscalls and calling
openat2() with O_SYNC is relatively rare, let's just blanket-enable it
in the --suppress-sync=yes mode. This means that we might issue a
synchronous write when something calls openat2() with O_SYNC, but not
breaking the apps here feels like a reasonable trade-off, at least until
a better solution pops up.
Note that the same situation is in seccomp_restrict_sxid(), but allowing
the openat2() syscall there could actually have some security-related
implications under certain circumstances.
Resolves: #41868
[0] https://cgit.git.savannah.gnu.org/cgit/tar.git/commit/?id=75b03fdff48916bd0654677ed21379bdb0db016d
[1] https://cgit.git.savannah.gnu.org/cgit/attr.git/commit/?id=1cea2fc5e8b5c8a5f29abf1141399e01a57a3154
Since all executables now manage their required dlopen notes directly
within their own source code with explicit priority levels, it is no
longer necessary to declare high-priority dlopen notes in the shared
libraries themselves.
This also switches all these notes to be defined via SD_ELF_NOTE_DLOPEN_ANCHORED().
Consequently, any notes added within unreachable or unused functions will be
automatically garbage-collected by the linker (--gc-sections) instead of
bloating the final binary.
E.g. unused p11-kit library dependency is now dropped from
systemd-repart.standalone binary.
Before:
```
$ systemd-analyze dlopen-metadata build/systemd-repart.standalone
FEATURE DESCRIPTION SONAME PRIORITY
cryptsetup Support for disk encryption, integrity, and authentication libcryptsetup.so.12 recommended
blkid Support for block device identification libblkid.so.1 required
libcrypto Support for cryptographic operations libcrypto.so.4 libcrypto.so.3 recommended
mount Support for mount enumeration libmount.so.1 required
fdisk Support for reading and writing partition tables libfdisk.so.1 required
blkid Support for block device identification libblkid.so.1 recommended
libcrypto Support for cryptographic operations libcrypto.so.4 libcrypto.so.3 suggested
cryptsetup Support for disk encryption, integrity, and authentication libcryptsetup.so.12 suggested
fdisk Support for reading and writing partition tables libfdisk.so.1 suggested
idn Support for internationalized domain names libidn2.so.0 suggested
mount Support for mount enumeration libmount.so.1 recommended
selinux Support for SELinux libselinux.so.1 recommended
tpm Support for TPM libtss2-esys.so.0 suggested
tpm Support for TPM libtss2-rc.so.0 suggested
tpm Support for TPM libtss2-mu.so.0 suggested
tpm Support for TPM libtss2-tcti-device.so.0 suggested
p11-kit Support for PKCS11 hardware tokens libp11-kit.so.0 suggested
```
After:
```
$ systemd-analyze dlopen-metadata build/systemd-repart.standalone
FEATURE DESCRIPTION SONAME PRIORITY
cryptsetup Support for disk encryption, integrity, and authentication libcryptsetup.so.12 recommended
blkid Support for block device identification libblkid.so.1 required
libcrypto Support for cryptographic operations libcrypto.so.4 libcrypto.so.3 recommended
mount Support for mount enumeration libmount.so.1 required
fdisk Support for reading and writing partition tables libfdisk.so.1 required
blkid Support for block device identification libblkid.so.1 recommended
libcrypto Support for cryptographic operations libcrypto.so.4 libcrypto.so.3 suggested
cryptsetup Support for disk encryption, integrity, and authentication libcryptsetup.so.12 suggested
fdisk Support for reading and writing partition tables libfdisk.so.1 suggested
mount Support for mount enumeration libmount.so.1 recommended
selinux Support for SELinux libselinux.so.1 recommended
tpm Support for TPM libtss2-esys.so.0 suggested
tpm Support for TPM libtss2-rc.so.0 suggested
tpm Support for TPM libtss2-mu.so.0 suggested
tpm Support for TPM libtss2-tcti-device.so.0 suggested
```
This macro emits an SD_ELF_NOTE_DLOPEN note so that calling sites,
and not just libsystemd-shared, get the ELF stamped with the notes
documenting the dependency. Currently most of the dlopen notes end
up in libsystemd-shared, and executables have no way to document
the optional dependencies they need. With this new macro, this
issue can be fixed.
Co-developed-by: Claude Opus 4.8 <noreply@anthropic.com>
We currently log at warning level:
/usr/lib/systemd/system/systemd-udevd.service:56: System call bpf cannot be
resolved as libseccomp is not available, ignoring: Operation not supported
In the initrd, or in minimal installations, not installing libseccomp is
reasonable, depending on various other choices. We shouldn't make fuss
about this.
C.f. https://bugzilla.redhat.com/show_bug.cgi?id=2463540.
Each dlopen_*() wrapper kept its dl handle as a file-scope
'static void *xxx_dl = NULL;' even though only the wrapper itself
ever referenced it. Move each one inside the corresponding function
so its scope matches its actual use, leaving the rest of each
translation unit free of the unused file-scope name.
In pcre2-util.c the assert(pcre2_dl) in pattern_matches_and_log()
becomes assert(sym_pcre2_match), which carries the same invariant
(pattern_compile_and_log() ran dlopen_pcre2()).
Every dlopen_xxx() helper now takes an int log_level argument. It is
passed through to dlopen_many_sym_or_warn() (which in turn propagates it
to dlopen_verbose() for the library-not-installed case), and is used by
the fallback stub when support for the library is not compiled in to
emit a "<lib> support is not compiled in." message at the caller's level.
Callers pass LOG_DEBUG when gracefully degrading, or a higher level when
the failure should surface, and no longer need to log redundantly at the
call site.
As part of this, dlopen_bpf_full() (which already took a log_level) is
merged into dlopen_bpf() rather than keeping both.
The static inline fallbacks used to live in the headers, which required
pulling log.h in from every header that declared a dlopen_xxx(). Move
them into the .c files instead: the declaration is always outside the
#if HAVE_XXX block, the impl sits outside the outer #if HAVE_XXX wrap
with its own internal #if HAVE_XXX/#else/#endif, and apparmor-util.c,
idn-util.c, libmount-util.c and pam-util.c are now always compiled so
they can host their stubs.
The updated parsing_hacks.h file uncovered a bunch of extra
things that the check-pointer-deref coccinelle script flags.
This commit fixes them to make the tree check-pointer-deref clean.
systemd now has a system call wrapper that does a long series of #ifdef's to
differentiate between architectures and ABIs. This wrapper has two problems.
1. On mips, it needs to differentiate between O32, N32, N64 ABI. It does that
via a code block in src/include/override/sys/generate-syscall.py (and derived
files):
76 # elif defined(_MIPS_SIM)
77 # if _MIPS_SIM == _MIPS_SIM_ABI32
78 # define systemd_NR_{syscall} {nr_mipso32}
79 # elif _MIPS_SIM == _MIPS_SIM_NABI32
80 # define systemd_NR_{syscall} {nr_mips64n32}
81 # elif _MIPS_SIM == _MIPS_SIM_ABI64
82 # define systemd_NR_{syscall} {nr_mips64}
83 # else
84 # error "Unknown MIPS ABI"
85 # endif
86 # elif defined(__hppa__)
Now the _MIPS_SIM* constants stem from a vendor-specific header file sgidefs.h,
which is included with glibc, but not with musl. It is however always present
in the Linux kernel headers as asm/sgidefs.h ...
2. To work around this, the syscall wrapper already has a block
47 #ifdef ARCH_MIPS
48 #include <asm/sgidefs.h>
49 #endif
Turns out, ARCH_MIPS is defined nowhere in Gentoo, neither on glibc nor on musl.
As a result the code (by accident, probably sgidefs.h is included transitively
somehow) works on glibc, but not on musl.
The simplest fix is to replace line 47 in the generator and the derived file
with
47 #ifdef __mips__
Two other source code files require a similar fix since they rely on the
constants.
Bug: https://github.com/systemd/systemd/issues/41239
Bug: https://bugs.gentoo.org/971376
Signed-off-by: Andreas K. Hüttel <dilfridge@gentoo.org>
Switch all internal callers from the private ELF_NOTE_DLOPEN() macro to
the new public SD_ELF_NOTE_DLOPEN() API from sd-dlopen.h, and remove
the now-redundant macro definitions from dlfcn-util.h.
Signed-off-by: Christian Brauner <brauner@kernel.org>
These syscalls are part of a newer kernel API to replace interaction
with /proc/self/attr, with the goal of allowing LSM stacking. These are
being used now by e.g. libapparmor, so should be more easily available
to services using seccomp filtering.
Major distributions already have libseccomp 2.5.x or newer.
Let's bump to the required minimum version to 2.4.0, which provides
SCMP_ACT_KILL_PROCESS, SCMP_ACT_LOG, SCMP_ARCH_PARISC, and
SCMP_ARCH_PARISC64.
Note, libseccomp 2.4.0 was released on 2019-03-15.
See also #38608.
Latest glibc uses _Generic to have strstr() and other functions return
const char* or char* based on whether the input is a const char* or a
char*. This causes build failures as we previously always expected a char*.
Let's fix the compilation failures and add our own macros similar to glibc's
to have string functions that return a mutable or const pointer depending on
the input.
../src/shared/seccomp-util.c: In function ‘seccomp_restrict_sxid’:
../src/shared/seccomp-util.c:2228:25: error: ‘__NR_fchmodat2’ undeclared (first use in this function); did you mean ‘fchmodat2’?
2228 | __NR_fchmodat2,
| ^~~~~~~~~~~~~~
| fchmodat2
The override/sys/syscalls.h needs to be included before the seccomp
headers, otherwise the internal seccomp preprocessor machinery will
not see the local definitions, so the local ifdef will be true but
the seccomp own definitions will be empty
We had errno_to_name() which works for "known" errnos, and returns NULL for
unknown ones, and then ERRNO_NAME which always returns an answer, possibly just
a number as a string, but requires a helper buffer.
It is possible for the kernel to add a new errno. We recently learned that some
architectures define custom errno names. Or for some function to unexpectedly
return a bogus errno value. In almost all cases it's better to print that value
rather than "n/a" or "(null)". So let's use ERRNO_NAME is most error handling
code. Noteably, our code wasn't very good in handling the potential NULL, so
in various places we could print "(null)". Since this is supposed to be used
most of the time, let's shorten the names to ERRNO_NAME/errno_name.
There are a few places where we don't want to use the fallback path, in
particular for D-Bus error names or when saving the error name. Let's rename
errno_to_name() to errno_name_no_fallback() to make the distinction clearer.
The header provides _MIPS_SIM_ABI32 and friends. Glibc indirectly includes
the header through sys/syscall.h or unistd.h, but let's explicitly include
the header where we use _MIPS_SIM_ABI32 and friends.
Now that we make use of open_tree() in places we previously used
openat() with O_PATH, it makes sense to move it from @mount to
@file-system. Without the OPEN_TREE_CLONE flag open_tree() is after all
unprivileged.
Note that open_tree_attr() I left in @mount, since it's purpose is
really to set mount options when cloning, and that's clearly a mount
related thing, not so much something you could use unpriv.
Follow-up for: c5de7b14ae
This addresses an issue tracked down by Antonio Feijoo: since the commit
that started to use open_tree() various apps started to crash because
they used seccomp filters and sd-device started to use open_tree()
internally.
Now that the necessary functions from log.h have been moved to macro.h,
we can stop including log.h in macro.h. This requires modifying source
files all over the tree to include log.h instead.
This is a new syscall provided by the kernel used to implement faster
uprobes. It's not supposed to be called by userspace, but only by kernel
generated uprobe code.
It should be fine to allow this, as the kernel authenticates the
invocation itself, and we shouldn't break compat with things.
Note that this allowlisting is not sufficient to make ureprobe() work.
libseccomp must be tought the syscall too, but this can happen
independently.
Fixes: #34615
Every services and containers should be able to protect their users and
limit the impact of security bugs thanks to the security syscalls
provided by seccomp and Landlock. The goal of these syscalls is to
improve security with additional restrictions. They are designed to be
safely used by unprivileged (and then potentially malicious) users.
Remove the now-redundant "seccomp" entry for nspawn.
The glibc API is behind the wrapper is called "secure_getenv()", hence
our wrapper really should keep the order too, otherwise things are just
too confusing.
Follow-up for 2331c02d06.
Note, currently, the function is always called with SCMP_ACT_ALLOW as
the default action, except for the test. So, this should not change
anything in the runtime code.
glibc starting using fchmodat2 to implement fchmod with flags [1], but
current version of libseccomp does not support fchmodat2 [2]. This is
causing problems with programs sandboxed by systemd. libseccomp needs to know
a syscall to be able to set any kind of filter for it, so for syscalls unknown
by libseccomp we would always do the default action, i.e. either return the
errno set by SystemCallErrorNumber or send a fatal signal. For glibc to ignore
the unknown syscall and gracefully fall back to the older implementation,
we need to return ENOSYS. In particular, tar now fails with the default
SystemCallFilter="@system-service" sandbox [3].
This is of course a wider problem: any time the kernel gains new syscalls,
before libseccomp and systemd have caught up, we'd behave incorrectly. Let's
do the same as we already were doing in nspawn since
3573e032f2, and do the "default action" only
for syscalls which are known by us and libseccomp, and return ENOSYS for
anything else. This means that users can start using a sandbox with the new
syscalls only after libseccomp and systemd have been updated, but before that
happens they behaviour that is backwards-compatible.
[1] https://github.com/bminor/glibc/commit/65341f7bbea824d2ff9d37db15d8be162df42bd3
[2] https://github.com/seccomp/libseccomp/issues/406
[2] https://github.com/systemd/systemd/issues/30250
Fixes https://github.com/systemd/systemd/issues/30250.
In seccomp_restrict_sxid() there's a chunk conditionalized with
'#if defined(__SNR_fchmodat2)'. We need to kep that because seccomp_restrict_sxid()
seccomp_restrict_suid_sgid() uses SCMP_ACT_ALLOW as the default action.
If there is no libseccomp support, just ban the entire syscall instead
so wrappers will fall back to older, supported syscalls.
Also reflect all of this in `test-seccomp.c`.
Ricardo Branco
Christian Brauner
Frantisek Sumsal
Yu Watanabe
Lennart Poettering
Luca Boccassi
Zbigniew Jędrzejewski-Szmek
Daan De Meyer
Michael Vogt
Andreas K. Hüttel
Nick Rosbrook
Daan De Meyer
Lennart Poettering
Zbigniew Jędrzejewski-Szmek
Integral
Mickaël Salaün
Arseny Maslennikov
Xiaotian Wu