We're creating a system for tests to more reliably execute commands in
various contexts (e.g. host, guest, namespace). That transition is going
to happen over a number of steps though, so in the meantime we need to deal
with both the old-style issuing of commands via typing into and screen
scraping tmux panels, and the new-style system for executing commands in
context.
Introduce some transitional helpers which will issue a command via context
if the requested context is initialized, but will otherwise fall back to
the old style tmux panel based method. Re-implement the various test DSL
commands in terms of these new helpers.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
We're moving to a new way of the tests dispatching commands to running in
contexts (host, guest, namespace, etc.). As we make this transition,
though, we still want the user to be able to watch the commands running
in a context, as they previously could from the commands issued in the
pane.
Add a helper to set up a pane to watch a context's log to allow this. In
some cases we currently issue commands from several different logical
contexts in the same pane, so allow a pane to watch several contexts at
once. Also use tail's --retry option to allow starting the watch before
we've initialized the context which will be useful in some cases.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
For the tests, we need to run commands in various contexts: in the host,
in a guest or in a namespace. Currently we do this by running each context
in a tmux pane, and using tmux commands to type the commands into the
relevant pane, then screen-scrape the output for the results if we need
them.
This is very fragile, because we have to make various assumptions to parse
the output. Those can break if a shell doesn't have the prompt we expect,
if the tmux pane is too small or in various other conditions.
This starts some library functions for a new "context" system, that
provides a common way to invoke commands in a given context, in a way that
properly preserves stdout, stderr and the process return code.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
test_iperf3() is a pretty inscrutable mess of nested background processes.
It has a number of ugly sleeps needed to wait for things to complete.
Rewrite it to be cleaner:
* Use the construct (a & b & wait) to run 'a' and 'b' in parallel, but
then wait for them both to complete before continuing
* This allows us to wait for both the server and client to finish, rather
than sleeping
* Use jq to do all the math we need to get the final result, rather than
jq followed by some complicated 'bc' mangling
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Currently all the throughput tests are run for 30s. This is reflected in
both the actual parameters given to the iperf commands, but also in the
matching sleeps in test_iperf3.
Allow this to be adjusted more easily with a new parameter to test_iperf3.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
[sbrivio: Reflect new parameter in comment to test_iperf3()]
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
These two commands in the DSL to run an iperf client and server are always
used together, and some of the parameters must match between them. The
iperf3s must also be run more or less immediately after iperf3c, since
iperf3c will run a client in the background after a sleep and requires a
server to be running before it will work.
A bunch of things can be made cleaner if we make a single DSL command that
runs both sides of the test. For now make the combined command work
exactly like the two commands together did, warts and all.
This does lose the ability for the DSL scripts to give additional options
to the iperf3 server, but we weren't using that anyway.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Currently in at least some of the testcases we kill qemu processes we're
done with by issuing a Control-C to the tmux panel it's running in. That
makes things harder as we try to move towards allowing "headless" testing
without tmux.
So, instead always use an explicit kill on a pid derived from a pidfile
for killing qemu. Note that we don't need to remove the pidfiles
afterwards, because qemu does that itself when terminated.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
For the passt and passt_in_ns tests we have a "shutdown" testcase that
checks for any errors from the passt process we were using (including
valgrind warnings). Do the same for pasta tests, so that we catch any
error codes from the pasta process.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The "valgrind" test cases are designed to pick up errors reported when
passt is running under valgrind. But what it actually does is just kill
the passt process, then see if it had a non-zero exit code. That means it
will equally well pick up any other problems which caused passt to exit
with an error status: either something detected within passt or as a result
of passt being killed by an unexpected signal.
The fact that the "valgrind" test is actually responsible for shutting down
the passt process is non-obvious and can lead to problems when selectively
running tests during debugging.
Rename the "valgrind" tests to "shutdown" tests and run it regardless of
whether we're using valgrind or not. This allows us to remove an ugly
speacial case in the passt_in_ns teardown code.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Currently the build tests and distro tests share a common setup function.
That works for now, but changes we want to make will mean they need
slightly different setup, so split the setup functions in preparation.
Currently, neither build nor distro tests have any teardown function.
Again, future changes are going to mean we need to do some teardown, so
create some empty for now teardown functions in preparation.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The DEMO_XTERM and CI_XTERM variables defined in test/lib/term aren't used
anywhere. Remove them.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The iperf based test commands create a bunch of .bw and .pid files for
each iperf client and server. The server side .bw files are cleaned
up afterwards, but the pid files are not, and none of the client side
files are cleaned up. The latter doesn't really matter when the
client is run on ephemeral guests, but sometimes we run it in a
namespace that shares the filesystem with the host.
Clean up all of these files after the tests.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Before starting the guests, these tests configure addresses in a pasta
namespace using dhclient. However, because it's a user namespace, it's
not running as "real" root and can't write to the dhclient pid file.
This doesn't stop it working, but causes an ugly error message which we
can avoid by using the --no-pid option.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
teardown_passt_in_ns() sends a ^D to the NS pane, which appears to be
intended to terminate the nsenter running there, leaving the namespace.
However, we've also sent a ^D to the PASST pane which will exit the pasta
instance which created the namespace. With the namespace destroyed the
nsenter in the NS pane will be killed, so it does not need to be exited
explicitly.
In fact sending the extra ^D can be harmful, since it will exit the shell
in which the nsenter was run, causing the whole pane to be closed. That
can then mean that the "pane_wait NS" hangs indefinitely. I believe this
will sometimes work, because there's a race between the various options
here, but it should be more reliable without the extra ^D.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The dhcp/passt and dhcp/passt_in_ns tests at least, and maybe others
use 'hout' commands that need to be able to detect empty output.
However, we don't set PS1, which means the screen-scraping logic which
detects this may not be reliable. In addition, if the host is using a
recent bash, it will have bracketed paste mode enabled which will also
add escape codes which will mess up the empty output detection.
Set the prompt and disable bracketed paste mode from the passt and
passt_in_ns setups to avoid these problems.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Similar case as the one fixed by David's patch "tests: Remove
unnecessary ^D in passt_in_ns teardown": we happen to pseudo-randomly
close panes by unnecessarily exiting the parent shells there, and
subsequent pane_wait directives hang.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Currently test/run uses wildcards to run all of the tests in a directory.
However, that wildcard list is filtered down by the "onlyfor" directives
in the test files... usually to a single file.
Therefore, just explicitly list the files we *really* want to run for this
test mode. This makes it easier to see at the top level what tests will
be executed, and to change that list temporarily while debugging specific
failures.
This means the "onlyfor" directive no longer has any purpose, and we can
remove it. "onlyfor" was also the only used of the $MODE variable, so we
can remove that too.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The top level listing control of which tests to run is in test/run, however
it uses the test() function which runs an entire directory of test files,
filtered by some criteria. This makes it awkward to narrow down to a
subset of tests when debugging a specific failure.
To make this easier, have test() take an explicit list of test files to
run, and have the caller in test/run handle the directory traversal. The
construct we use for this is pretty awkward to handle the fact that we're
in the source tree root directory rather than test/ at this point in
test/run. Later cleanups will improve that.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The test scripts support a "req" directive which requires one test script
to be run before another. It's implemented by doing a topological sort
based on these directives in the runner scripts, which is about as awkward
as you'd expect in Bourne shell.
It turns out we only use this functionality in one place - to make the
"make install" test run after the plain "make" test. We also already have
a simpler way of making sure tests run in a specific order: just put them
into the same test script file.
So, remove support for the "req" directive and just fold the build/all and
build/install test scripts together.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Move the download of mbuto and using it to create a sample initramfs to
the asset build makefile, rather than embedding it in the test scripts
themselves.
The two_guests tests used to use two separate copies of the mbuto
image. As an initramfs the mbuto image is strictly readonly though,
so that's not necessary. So, also use the same image for both guests.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
A number of passt/pasta testcases have initial steps which are just about
building images or other assets we need for the test proper. Repeating
these for each test run can be quite costly.
This patch makes a start on moving this sort of test asset building to
a separate phase before running the tests proper. For now just add a
Makefile to handle the asset building (although it doesn't build
anything yet), and make the path where we'll be building the assets
available to the tests.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
A lot of tests and examples invoke qemu with the command "kvm". However,
as far as I can tell, "kvm" being aliased to the appropriate qemu system
binary is Debian specific. The binary names from qemu upstream -
qemu-system-$ARCH - also aren't universal, but they are more common (they
should be good for both Debian and Fedora at least).
In order to still get KVM acceleration when available, we use the option
"-M accel=kvm:tcg" to tell qemu to try using either KVM or TCG in that
order
A number of the places we invoked "kvm" are expecting specifically an x86
guest, and so it's also safer to explicitly invoke qemu-system-x86_64.
Some others appear to be independent of the target arch (just wanting the
same arch as the host to allow KVM acceleration). Although I suspect there
may be more subtle x86 specific options in the qemu command lines, attempt
to preserve arch independence by using $(uname -m).
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This depends on a future change in mbuto to accept external profile
files. Add a file defining what we need for tests and demos, dropping
udhcpc and script as they're not needed anymore, and switch to it.
Suggested-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
For some reason, the passt/pasta tests and examples use dhclient for
DHCPv6, but in most cases use udhcpc for DHCPv4. Change it to use dhclient
for both DHCPv4 and DHCPv6. This means one less tool we need for testing,
plus dhclient is easily available on Fedora whereas udhcpc is not.
Note that the passt tests still rely on udhcpc indirectly because mbuto
wants to put it into the guest images it generates.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
A number of tests and examples use dhclient in both IPv4 and IPv6 modes.
We use "dhclient -6" for IPv6, but usually just "dhclient" for IPv4. Add
an explicit "-4" argument to make it more clear and explicit.
In addition, when dhclient is run from within pasta it usually won't be
"real" root, and so will not have access to write the default global pid
file. This results in a mostly harmless but irritating error:
Can't create /var/run/dhclient.pid: Permission denied
We can avoid that by using the --no-pid flag to dhclient.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
ip(8)'s ability to take abbreviated arguments (e.g. "li sh" instead of
"link show") is very handy when using it interactively, but it doesn't make
for very readable scripts and examples when shown that way.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Having all those 'echo $?' is rather distracting in demos.
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Sefano Brivio <sbrivio@redhat.com>
...there are no 'test' directives in demo, and this causes a
script failure.
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Now that we have pane_status to check the success of commands issued to
panes, we can more easily check for the success of the 'which' commands
used to check tool availability, rather than constructing, then parsing
special "skip" output.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
When we use pane_wait to wait for a command issued to a tmux pane to finish
we have no idea whether the command succeeded or not. This means that the
test scripts can keep running long after the point something vital has
failed, making it difficult to work out what went wrong.
Add a new pane_status command that checks for success of the issued command
and use it in most places instead of pane_wait. We still need explicit
pane_wait where we're gathering explicit output with pane_parse, because
the way we check the status with 'echo $?' means we lose track of that
output.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
[sbrivio:
- instead of quitting the script, make a test fail if a command
issued in a pane fails during a test, and loop until the status code is
numeric in pane_status() as a hack to make it a bit more robust
- retain usage of pane_wait() in iperf3 and teardown functions as we
interrupt iperf3, passt, and pasta, so a non-zero exit code is expected
- drop bogus ns_{1,2}_wait() calls in teardown_two_guests(), those
functions were never implemented
- use pane_status() for "guest" test directives too
]
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Most commands issued during the testing scripts aren't explicitly checked
for errors. Therefore, if they fail, the shell will just keep on
executing. This makes it difficult to figure out where things started
going wrong if things fall over.
Run the whole script with the set -e mode so that it will exit in the case
of any (unchecked) failing command. To make this work we do need to add
explicit checks / fallbacks for some commands which we expect to fail.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
[sbrivio: use sh -e instead of setting -e later, so that we don't miss
anything before set -e is issued]
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
pane_parse() attempts to grab the output from the last command issued
into a tmux pane. It strips out control characters using tr, which in
particular includes the final \r\n. However, this won't fully strip
out terminal escape sequences. In particular this breaks if the shell
in the pane is bash, with enable-bracketed-paste enabled in readline.
That issues terminal sequences to enable and disable bracketed paste
mode around every shell prompt.
We can work around this because these escapes are followed by a \r
(CR). More generally, it seems reasonable to assume that any terminal
shenanigans followed by a CR, but not an LF is supposed to be hidden.
So, use sed to strip everything before the second last CR. We still
need the tr to remove the final \r\n from the string (sed processes a
line at a time, and doesn't consider the CRLF part of the buffer it's
processing).
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
[sbrivio: modify regexp to keep foo\r\r\n unchanged, by matching on at
least one CR and a non-CR afterwards: that's the usual output pattern
for bash on Debian 8 and Debian 9]
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
run_term() uses tmux set-option -g to globally set the default shell.
Unfortunately this hits a chicken-and-egg problem that's common with many
of tmux's session options. If there isn't already a tmux server running,
we can't connect to set the option. If we attempt this after starting our
session (and therefore the server), then the session will already be
started with the previous default shell.
In any case it's not a good idea to set tmux global options, since that
might interfere with whatever else the user is doing in tmux. So, instead
set the default-shell option locally to the session after starting it. To
make sure we get the right shell for our initial script, explicitly invoke
/bin/sh to interpret it.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The semantics of tmux's update-environment option are a bit confusing.
It says it means the given variables are copied into the session
environment from the source environment, but it's not entirely clear
what the "source" environment means.
From my experimentation it appeast to be the environment from which
the tmux *server* is launched, not the one issuing the 'new-session'
command. That makes it pretty much useles, certainly in our case where
we have no way of knowing if the user has pre-existing tmux sessions.
Instead use the new-session -e option to explicitly pass in the variables
we want to propagate.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The DEBUG option for tests/run enables debugging options to passt/pasta,
however that doesn't help with debugging the test scripts themselves, which
are fairly fragile.
Extend the DEBUG option so it also prints information on each command in
the test scripts to make it easier to work out where things are falling
over.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Pass to seccomp.sh a list of additional syscalls valgrind needs as
EXTRA_SYSCALLS in a new 'valgrind' make target, and add corresponding
support in seccomp.sh itself.
In test setup functions, start passt with valgrind, but not for
performance tests.
Add tests checking that valgrind exits without errors after all the
other tests in the group are done.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
--debug can be a bit too noisy, especially as single packets or
socket messages are logged: implement a new option, --trace,
implying --debug, that enables all debug messages.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
For demos, cool-retro-term(1) looked fancier, but several threads of
that and ffmpeg(1) are just messing up with performance testing.
The CI videos started getting really big as well, and they were
difficult to read.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
...showing setup steps, some peculiarities as --net option, and a
general side-to-side comparison with slirp4netns(1), including
"quick" TCP and UDP throughput and latency benchmarks.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
To reach (at least) a conceptually equivalent security level as
implemented by --enable-sandbox in slirp4netns, we need to create a
new mount namespace and pivot_root() into a new (empty) mountpoint, so
that passt and pasta can't access any filesystem resource after
initialisation.
While at it, also detach IPC, PID (only for passt, to prevent
vulnerabilities based on the knowledge of a target PID), and UTS
namespaces.
With this approach, if we apply the seccomp filters right after the
configuration step, the number of allowed syscalls grows further. To
prevent this, defer the application of seccomp policies after the
initialisation phase, before the main loop, that's where we expect bad
things to happen, potentially. This way, we get back to 22 allowed
syscalls for passt and 34 for pasta, on x86_64.
While at it, move #syscalls notes to specific code paths wherever it
conceptually makes sense.
We have to open all the file handles we'll ever need before
sandboxing:
- the packet capture file can only be opened once, drop instance
numbers from the default path and use the (pre-sandbox) PID instead
- /proc/net/tcp{,v6} and /proc/net/udp{,v6}, for automatic detection
of bound ports in pasta mode, are now opened only once, before
sandboxing, and their handles are stored in the execution context
- the UNIX domain socket for passt is also bound only once, before
sandboxing: to reject clients after the first one, instead of
closing the listening socket, keep it open, accept and immediately
discard new connection if we already have a valid one
Clarify the (unchanged) behaviour for --netns-only in the man page.
To actually make passt and pasta processes run in a separate PID
namespace, we need to unshare(CLONE_NEWPID) before forking to
background (if configured to do so). Introduce a small daemon()
implementation, __daemon(), that additionally saves the PID file
before forking. While running in foreground, the process itself can't
move to a new PID namespace (a process can't change the notion of its
own PID): mention that in the man page.
For some reason, fork() in a detached PID namespace causes SIGTERM
and SIGQUIT to be ignored, even if the handler is still reported as
SIG_DFL: add a signal handler that just exits.
We can now drop most of the pasta_child_handler() implementation,
that took care of terminating all processes running in the same
namespace, if pasta started a shell: the shell itself is now the
init process in that namespace, and all children will terminate
once the init process exits.
Issuing 'echo $$' in a detached PID namespace won't return the
actual namespace PID as seen from the init namespace: adapt
demo and test setup scripts to reflect that.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
For distribution tests, we'll repeat some tests frequently. Add a
'def' directive that starts a block, ended by 'endef', whose
execution can then be triggered by simply giving its name as a
directive itself.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
We might have highlighting and slightly different prompts across
different distributions, allow a more reasonable set of prompt
strings to be accepted as prompts.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
With recent improvements, we're not CPU-bound at all while testing
UDP performance. Give the VM more memory and CPUs, forward two
additional ports, start up to four threads in parallel, and give
single iperf3 threads higher bandwidth targets.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
