A cross-architecture build might pass a target-specific CC on 'make',
and not on 'make install', and this is what happens in Debian
cross-qa tests.
Given that we select binaries to be installed depending on the target
architecture, this means we would build AVX2 binaries in any case on
a x86_64 build machine.
By overriding TARGET in package build rules, we can tell the Makefile
about the target architecture, also for the 'install' (Makefile)
target.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Debian cross-building automatic checks:
http://crossqa.debian.net/src/passt
currently fail because we don't use the right target architecture and
compiler while building the system call lists and resolving their
numbers in seccomp.sh. Pass ARCH and CC to seccomp.sh and use them.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Recently, commit 4ddbcb9c0c ("tcp: Disable optimisations
for tcp_hash()") worked around yet another issue we hit with gcc 12
and '-flto -O2': some stores affecting the input data to siphash_20b()
were omitted altogether, and tcp_hash() wouldn't get the correct hash
for incoming connections.
Digging further into this revealed that, at least according to gcc's
interpretation of C99 aliasing rules, passing pointers to functions
with different types compared to the effective type of the object
(for example, a uint8_t pointer to an anonymous struct, as it happens
in tcp_hash()), doesn't guarantee that stores are not reordered
across the function call.
This means that, in general, our checksum and hash functions might
not see parts of input data that was intended to be provided by
callers.
Not even switching from uint8_t to character types, which should be
appropriate here, according to C99 (ISO/IEC 9899, TC3, draft N1256),
section 6.5, "Expressions", paragraph 7:
An object shall have its stored value accessed only by an lvalue
expression that has one of the following types:
[...]
— a character type.
does the trick. I guess this is also subject to interpretation:
casting passed pointers to character types, and then using those as
different types, might still violate (dubious) aliasing rules.
Disable gcc strict aliasing rules for potentially affected functions,
which, in turn, disables gcc's Type-Based Alias Analysis (TBAA)
optimisations based on those function arguments.
Drop the existing workarounds. Also the (seemingly?) bogus
'maybe-uninitialized' warning on the tcp_tap_handler() > tcp_hash() >
siphash_20b() path goes away with -fno-strict-aliasing on
siphash_20b().
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Future compilers will not support implicit ints by default, causing
the probe to always fail.
Link: https://bugs.passt.top/show_bug.cgi?id=42
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
These files are left around by emacs amongst other editors.
Signed-off-by: Richard W.M. Jones <rjones@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
If you run the build several times it will fail unnecessarily with:
ln -s passt pasta
ln: failed to create symbolic link 'pasta': File exists
make: *** [Makefile:134: pasta] Error 1
Signed-off-by: Richard W.M. Jones <rjones@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Currently, when instructed to open an IPv6 socket, sock_l4() explicitly
sets the IPV6_V6ONLY socket option so that the socket will only respond to
IPv6 connections. Linux (and probably other platforms) allow "dual stack"
sockets: IPv6 sockets which can also accept IPv4 connections.
Extend sock_l4() to be able to make such sockets, by passing AF_UNSPEC as
the address family and no bind address (binding to a specific address would
defeat the purpose). We add a Makefile define 'DUAL_STACK_SOCKETS' to
indicate availability of this feature on the target platform.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
struct tcp_conn stores an address which could be IPv6 or IPv4 using a
union. We can do this without an additional tag by encoding IPv4 addresses
as IPv4-mapped IPv6 addresses.
This approach is useful wider than the specific place in tcp_conn, so
expose a new 'union inany_addr' like this from a new inany.h. Along with
that create a number of helper functions to make working with these "inany"
addresses easier.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Currently spliced and non-spliced connections use completely independent
tracking structures. We want to unify these, so as a preliminary step move
the definitions for both variants into a new tcp_conn.h header, shared by
tcp.c and tcp_splice.c.
This requires renaming some #defines with the same name but different
meanings between the two cases. In the process we correct some places that
are slightly out of sync between the comments and the code for various
event bit names.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
clang-tools 15.0.0 appears to have added a new warning that will always
complain about assignments in if statements, which we use in a number of
places in passt/pasta. Encountered on Fedora 37 with
clang-tools-extra-15.0.0-3.fc37.x86_64.
Suppress the new warning so that we can compile and test.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
The AUDIT_ARCH defines in seccomp.h corresponding to HPPA are
AUDIT_ARCH_PARISC and AUDIT_ARCH_PARISC64.
Build error spotted in Debian's buildd log on
phantom.physik.fu-berlin.de.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
On mips64el, gcc -dumpmachine correctly reports mips64el as
architecture prefix, but for some reason seccomp.h defines
AUDIT_ARCH_MIPSEL64 and not AUDIT_ARCH_MIPS64EL. Mangle AUDIT_ARCH
accordingly.
Build error spotted in Debian's buildd logs from Loongson build.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Drop it from the internal FLAGS variable, but honour -O2 if passed in
CFLAGS. In Debian packages, dpkg-buildflags uses it as hardening
flag, and we get a QA warning if we drop it:
https://qa.debian.org/bls/bytag/W-dpkg-buildflags-missing.html
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
CPPFLAGS allow the user to pass pre-processor flags. This is unlikely
to be needed at the moment, but the Debian Hardening Walkthrough
reasonably requests it to be handled in order to fully support
hardened build flags:
https://wiki.debian.org/HardeningWalkthrough#Handling_dpkg-buildflags_in_your_upstream_build_system
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
In some environments, such as KubeVirt pods, we might not have a
system logger available. We could choose to run in foreground, but
this takes away the convenient synchronisation mechanism derived from
forking to background when interfaces are ready.
Add optional logging to file with -l/--log-file and --log-size.
Unfortunately, this means we need to duplicate features that are more
appropriately implemented by a system logger, such as rotation. Keep
that reasonably simple, by using fallocate() with range collapsing
where supported (Linux kernel >= 3.15, extent-based ext4 and XFS) and
falling back to an unsophisticated block-by-block moving of entries
toward the beginning of the file once we reach the (mandatory) size
limit.
While at it, clarify the role of LOG_EMERG in passt.c.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Logging to file is going to add some further complexity that we don't
want to squeeze into util.c.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
With gcc 11 and 12, passing -flto, or -flto=auto, and -O2,
intra-procedural optimisation gets rid of a fundamental bit in ndp():
the store of hop_limit in the IPv6 header, before the checksum is
calculated, which on x86_64 looks like this:
ip6hr->hop_limit = IPPROTO_ICMPV6;
b8c0: c6 44 24 35 3a movb $0x3a,0x35(%rsp)
Here, hop_limit is temporarily set to the protocol number, to
conveniently get the IPv6 pseudo-header for ICMPv6 checksum
calculation in memory.
With LTO, the assignment just disappears from the binary.
This is rather visible as NDP messages get a wrong checksum, namely
the expected checksum plus 58, and they're ignored by the guest or
in the namespace, meaning we can't get any IPv6 routes, as reported
by Wenli Quan.
The issue affects a significant number of distribution builds,
including the ones for CentOS Stream 9, EPEL 9, Fedora >= 35,
Mageia Cauldron, and openSUSE Tumbleweed.
As a quick workaround, declare csum_unaligned() as "noipa" for gcc
11 and 12, with -flto and -O2. This disables inlining and cloning,
which causes the assignment to be compiled again.
Leave a TODO item: we should figure out if a gcc issue has already
been reported, and report one otherwise. There's no apparent
justification as to why the store could go away.
Reported-by: Wenli Quan <wquan@redhat.com>
Link: https://bugzilla.redhat.com/show_bug.cgi?id=2129713
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Commit 1a563a0cbd ("passt: Address gcc 11 warnings") works around an
issue where the remote address passed to hash functions is seen as
uninitialised by gcc, with -flto and -O2.
It turns out we get the same exact behaviour on gcc 12.1 and 12.2, so
extend the applicability of the same workaround to gcc 12.
Don't go further than that, though: should the issue reported at:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78993
happen to be fixed in a later version of gcc, we won't need the
noinline attributes anymore. Otherwise, we'll notice.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
It's unclear what original suppressions these unmatchedSuppression
suppressions were supposed to go with. They don't trigger any warnings on
the current code that I can tell, so remove them. If we find a problem
with some cppcheck versions in future, replace them with inline
suppressions so it's clearer exactly where the issue is originating.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
We have a couple of functions that are unused (for now) by design.
Although at least one has a flag so that gcc doesn't warn, cppcheck has its
own warnings about this. Add specific inline suppressions for these rather
than a blanket suppression in the Makefile.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
I can't get this warning to trigger, even without the suppression, so
remove it. If it shows up again on some cppcheck version, we can replace
it with inline suppressions so it's clear where the issue lay.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
I can't get these warnings to trigger on the cppcheck versions I have, so
remove them. If we find in future we need to replace these, they should be
replaced with inline suppressions so its clear what's the section of code
at issue.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
I can't get this warning to trigger, so I think this suppression must be
out of date. Whether that's because we've changed our code to no longer
have the problem, or because cppcheck itself has been updated to remove a
false positive I don't know.
If we find that we do need a suppression like this for some cppcheck
version, we should replace it with an inline suppression so it's clear
what exactly is triggering the warning.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
seccomp.sh generates seccomp.h, so if we change it, we should re-build
seccomp.h as well. Add this to the make dependencies so it happens.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Since bf95322f "conf: Make the argument to --pcap option mandatory" we
no longer call localtime() in pcap.c, so we no longer need the matching
cppcheck suppression.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
In a number of places in passt we use structures to represent over the wire
or in-file data with a fixed layout. After initialization we don't access
the fields individually and just write the structure as a whole to its
destination.
Unfortunately cppcheck doesn't cope with this pattern and thinks all the
structure members are unused. We already have suppressions for this in
pcap.c and dhcp.c However, it also appears in dhcp.c and netlink.c at
least. Since this is likely to be common, it seems wiser to just suppress
the error globally.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
strtok() is non-reentrant and old-fashioned, so cppcheck would complains
about its use in conf.c if it weren't suppressed. We're single threaded
and strtok() is convenient though, so it's not really worth reworking at
this time. Convert this to an inline suppression so it's adjacent to the
code its annotating.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
qrap.c uses several old-fashioned functions that cppcheck complains about.
Since it's headed for obselesence anyway, just suppress these rather than
attempting to modernize the code.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
We define our own ffsl() as a weak symbol, in case our C library doesn't
include it. On glibc systems which *do* include it, this causes a cppcheck
warning because unsurprisingly our version doesn't pick the same argument
names. Convert the suppression for this into an inline suppression.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
conf_runas() handles several of the different possible cases for the
--runas argument in a slightly odd order. Although it can parse both
numeric UIDs/GIDs and user/group names, it can't parse a numeric UID
combined with a group name or vice versa. That's not obviously useful, but
it's slightly surprising gap to have.
Rework the parsing to be more systematic: first split the option into
user and (optional) group parts, then separately parse each part as either
numeric or a name. As a bonus this removes some clang-tidy warnings.
While we're there also add cppcheck suppressions for getpwnam() and
getgrnam(). It complains about those because they're not reentrant.
passt is single threaded though, and is always likely to be during
this initialization code, even if we multithread later.
There were some existing suppressions for these in the cppcheck
invocation but they're no longer up to date. Replace them with inline
suppressions which, being next to the code, are more likely to stay
correct.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Debian and similar distros put target specific header files in
/usr/include/<arch-vendor-os>, rather than directly in /usr/include. Add
this directory to the includes for cppcheck so it can find them.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
We do some manipulation of the output of cc -v to get the target triple
for the platform, to locate headers for cppcheck. However, we can get
this more easily with cc -dumpmachine - and in fact we do so elsewhere in
the Makefile.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Adding the --quiet option to cppcheck makes the actual errors and warnings
easier to find.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
make cppcheck takes a long time, because it checks a large number of
different configurations. It's assembling this very large set of
configurations not because of conditionals in the passt code itself,
but from those in the system headers. By adding --config-exclude
directives to stop considering those configs, make cppcheck becomes
around 60x faster on my system.
Similarly, any problems that are found in the system headers are not our
problem, and so we can uniformly suppress them, rather than having specific
suppressions for particular problems in particular files (which might not
be correct for all different distro / version combinations either).
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
This check complains about any identifier of less than 3 characters. For
locals and parameters this is often pointlessly verbose. Disable it.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
enum conf_port_type is local to conf.c and is used to track the port
forwarding mode during configuration. We don't keep it around in the
context structure, however the 'init_detect_ports' and 'ns_detect_ports'
fields in the context are based solely on this. Rather than changing
encoding, just include the forwarding mode into the context structure.
Move the type definition to a new port_fwd.h, which is kind of trivial at
the moment but will have more stuff later.
While we're there, "conf_port_type" doesn't really convey that this enum is
describing how port forwarding is configured. Rename it to port_fwd_mode.
The variables (now fields) of this type also have mildly confusing names
since it's not immediately obvious whether 'ns' and 'init' refer to the
source or destination of the packets. Use "in" (host to guest / init to
ns) and "out" (guest to host / ns to init) instead.
This has the added bonus that we no longer have locals 'udp_init' and
'tcp_init' which shadow global functions.
In addition, add a typedef 'port_fwd_map' for a bitmap of each port number,
which is used in several places.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Define the target machine architecture in lowercase.
The name of the executable qemu-system-* is defined from the build flags
and should be in lowercase:
( "qemu-system-" ARCH ),
I.e. qemu-system-x86_64 instead of qemu-system-X86_64. Otherwise, the
exec call will fail.
Signed-off-by: Vasiliy Ulyanov <vulyanov@suse.de>
Signed-off-by: Dario Faggioli <dfaggioli@suse.com>
Targets running static checkers (cppcheck and clang-tidy) need
seccomp.h, but the latter is not included in HEADERS. Add it.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
If we append CFLAGS to the ones passed via command line (if any),
-D options we append will override -D options passed on command line
(if any).
For example, OpenSUSE build flags include -D_FORTIFY_SOURCE=3, and we
want to have -D_FORTIFY_SOURCE=2, if and only if not overridden. The
current behaviour implies we redefine _FORTIFY_SOURCE as 2, though.
Instead of appending CFLAGS, prepend them by adding all the default
build flags to another variable, a simply expanded one (defined with
:=), named FLAGS, and pass that *before* CFLAGS in targets, so that
defines from command line can override default flags.
Reported-by: Dario Faggioli <dfaggioli@suse.com>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Tested-by: Dario Faggioli <dfaggioli@suse.com>
passt/pasta contains a number of routines designed to isolate passt from
the rest of the system for security. These are spread through util.c and
passt.c. Move them together into a new isolation.c file.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
We don't set any, but we should use them if they are passed in the
environment. On a Fedora Rawhide package build, annocheck
(https://sourceware.org/annobin/) reports:
Hardened: /usr/bin/passt: FAIL: bind-now test because not linked with -Wl,-z,now
...despite the build system exporting -Wl,-z,now in LDFLAGS.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
It turns out that, while on most distributions "docdir" would be
/usr/share/doc, it's /usr/share/doc/packages/ on OpenSUSE Tumbleweed.
Use an explicit docdir as shown in:
https://en.opensuse.org/openSUSE:Build_Service_cross_distribution_howto
and don't unnecessarily hardcode directory variables in the Makefile.
Otherwise, RPM builds for OpenSUSE will fail now that we have a README
there.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Distribution packages reasonably expect to have a human-readable
Markdown version of the README under /usr/share/doc/, but all we have
right now is a heavily web-oriented version.
Introduce a ugly hack to strip web-oriented parts from the current
README and install it.
It should probably work the other way around: a human-readable README
could be used as a source for the web page. But cgit needs a file
that's in the tree, not something that can be built, and
https://passt.top/ is based on cgit. It should eventually be doable
to work around this in cgit, instead.
Reported-by: Benson Muite <benson_muite@emailplus.org>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Some versions of valgrind (such as the version on my Fedora laptop -
valgrind-3.19.0-3.fc36.x86_64) use futexes. But futex is currently not
allowed in the seccomp filter, even with the extra calls added for
valgrind builds. Add it, to avoid spurious valgrind failures.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
We handle SIGQUIT and SIGTERM calling exit(), which is usually
implemented with the exit_group() system call.
If we don't allow exit_group(), we'll get a SIGSYS while handling
SIGQUIT and SIGTERM, which means a misleading non-zero exit code.
Reported-by: Wenli Quan <wquan@redhat.com>
Link: https://bugzilla.redhat.com/show_bug.cgi?id=2101990
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
If the tests are interrupted at the right point a passt.pid file can be
left over. Clean it up with "make clean" and add it to .gitignore so it
doesn't get accidentally committed.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Two places in passt need to read files line by line (one parsing
resolv.conf, the other parsing /proc/net/*. They can't use fgets()
because in glibc that can allocate memory. Instead they use an
implementation line_read() in util.c. This has some problems:
* It has two completely separate modes of operation, one buffering
and one not, the relation between these and how they're activated
is subtle and confusing
* At least in non-buffered mode, it will mishandle an empty line,
folding them onto the start of the next non-empty line
* In non-buffered mode it will use lseek() which prevents using this
on non-regular files (we don't need that at present, but it's a
surprising limitation)
* It has a lot of difficult to read pointer mangling
Add a new cleaner implementation of allocation-free line-by-line
reading in lineread.c. This one always buffers, using a state
structure to keep track of what we need. This is larger than I'd
like, but it turns out handling all the edge cases of line-by-line
reading in C is surprisingly hard.
This just adds the code, subsequent patches will change the existing
users of line_read() to the new implementation.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
