- dev - passt

[PATCH] Makefile: Honour LDFLAGS for binary targets
by Stefano Brivio 07 Sep '22

07 Sep '22

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(a)redhat.com> --- Makefile | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/Makefile b/Makefile index 26e64d5..644a541 100644 --- a/Makefile +++ b/Makefile @@ -105,11 +105,11 @@ seccomp.h: $(PASST_SRCS) $(PASST_HEADERS) @ EXTRA_SYSCALLS=$(EXTRA_SYSCALLS) ./seccomp.sh $^ passt: $(PASST_SRCS) $(PASST_HEADERS) seccomp.h - $(CC) $(CFLAGS) $(PASST_SRCS) -o passt + $(CC) $(CFLAGS) $(PASST_SRCS) -o passt $(LDFLAGS) passt.avx2: CFLAGS += -Ofast -mavx2 -ftree-vectorize -funroll-loops passt.avx2: $(PASST_SRCS) $(PASST_HEADERS) seccomp.h - $(CC) $(filter-out -O2,$(CFLAGS)) $(PASST_SRCS) -o passt.avx2 + $(CC) $(filter-out -O2,$(CFLAGS)) $(PASST_SRCS) -o passt.avx2 $(LDFLAGS) passt.avx2: passt @@ -117,7 +117,7 @@ pasta.avx2 pasta.1 pasta: pasta%: passt% ln -s $< $@ qrap: $(QRAP_SRCS) passt.h - $(CC) $(CFLAGS) $(QRAP_SRCS) -o qrap + $(CC) $(CFLAGS) $(QRAP_SRCS) -o qrap $(LDFLAGS) valgrind: EXTRA_SYSCALLS="rt_sigprocmask rt_sigtimedwait rt_sigaction \ getpid gettid kill clock_gettime mmap munmap open \ -- 2.35.1

2 1

[PATCH] fedora: Escape % characters in spec file's changelog
by Stefano Brivio 06 Sep '22

06 Sep '22

...rpmbuild otherwise expands valid macro names in changelog entries. Signed-off-by: Stefano Brivio <sbrivio(a)redhat.com> --- contrib/fedora/rpkg.macros | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/contrib/fedora/rpkg.macros b/contrib/fedora/rpkg.macros index df9dfc5..06359ab 100644 --- a/contrib/fedora/rpkg.macros +++ b/contrib/fedora/rpkg.macros @@ -36,7 +36,7 @@ function passt_git_changelog_entry { IFS=' ' for l in $(git log ${__from}..${__to} --pretty=format:"- %s" -- contrib/fedora); do - printf "%s%s\n" '-' "${l#*:}" + printf "%s%s\n" '-' "${l#*:}" | sed s/%/%%/g done unset IFS -- 2.35.1

1 0

[PATCH] test: Wait for systemd-resolved to be ready on Ubuntu 22.04 for s390x
by Stefano Brivio 05 Sep '22

05 Sep '22

On new Ubuntu 22.04 images, stopping systemd-resolved to get the dhclient script override resolv.conf doesn't work anymore. I originally used that hack to avoid introducing a delay which is needed when running it on TCG. Keep systemd-resolved running instead, and wait for it to be ready by retrying to resolve a domain a few times before installing packages, so that we don't add another ugly delay that might unnecessarily slow down things even further. Signed-off-by: Stefano Brivio <sbrivio(a)redhat.com> --- test/distro/ubuntu | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/test/distro/ubuntu b/test/distro/ubuntu index 6666f38..5e67a46 100644 --- a/test/distro/ubuntu +++ b/test/distro/ubuntu @@ -54,6 +54,11 @@ hout INIT_MSG cat /tmp/init_msg check [ __INIT_MSG__ = "from_ns" ] endef +# With systemd-resolved and TCG, DNS might take a while to work +def dns_ready_wait +host r=10; while [ \${r} -gt 0 ]; do host ubuntu.com && break; sleep 5; r=\$((r - 1)); done +endef + # Start passt, set common variables hostb ./passt -P __PIDFILE__ & sleep 1 @@ -185,11 +190,10 @@ test Ubuntu 22.04 (Jammy Jellyfish), s390x host ./qrap 5 qemu-system-s390x -m 2048 -smp 2 -serial stdio -nodefaults -nographic __BASEPATH__/prepared-jammy-server-cloudimg-s390x.img -net socket,fd=5 -net nic,model=virtio -device virtio-rng-ccw -snapshot -host service systemd-resolved stop host export DEBIAN_FRONTEND=noninteractive host apt-get -y remove needrestart snapd host dhclient -4 -sleep 2 +dns_ready_wait host apt-get update host apt-get -y install make gcc socat -- 2.35.1

1 0

[PATCH] conf: Fix getopt_long() optstring for current semantics of -D, -S, -p
by Stefano Brivio 03 Sep '22

03 Sep '22

Declaring them as required_argument in the longopts array specifies validation, but doesn't affect how optind is increased after parsing their values. Currently, passing one of these options as last option causes pasta to handle their own values as path to a binary to execute. Fixes: aae2a9bbf7d1 ("conf: Use "-D none" and "-S none" instead of missing empty option arguments") Fixes: bf95322fc1ef ("conf: Make the argument to --pcap option mandatory") Signed-off-by: Stefano Brivio <sbrivio(a)redhat.com> --- conf.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/conf.c b/conf.c index e6d1c62..2edb4ae 100644 --- a/conf.c +++ b/conf.c @@ -1094,9 +1094,9 @@ void conf(struct ctx *c, int argc, char **argv) const char *optstring; if (c->mode == MODE_PASST) - optstring = "dqfehs:p::P:m:a:n:M:g:i:D::S::46t:u:"; + optstring = "dqfehs:p:P:m:a:n:M:g:i:D:S:46t:u:"; else - optstring = "dqfehI:p::P:m:a:n:M:g:i:D::S::46t:u:T:U:"; + optstring = "dqfehI:p:P:m:a:n:M:g:i:D:S:46t:u:T:U:"; name = getopt_long(argc, argv, optstring, options, NULL); -- 2.35.1

2 1

[PATCH v2 0/2] Two further fixes for Fedora's spec file issues
by Stefano Brivio 02 Sep '22

02 Sep '22

...both reported by Daniel. v2: Details in single patch messages Stefano Brivio (2): fedora: Add %dir entries for own SELinux policy directory and documentation fedora: Add selinux-policy Requires: tag contrib/fedora/passt.spec | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) -- 2.35.1

1 2

[PATCH 0/2] Two further fixes for Fedora's spec file issues
by Stefano Brivio 02 Sep '22

02 Sep '22

...both reported by Daniel. Stefano Brivio (2): fedora: Add %dir entries for own SELinux policy directory and documentation fedora: Add selinux-policy Requires: tag contrib/fedora/passt.spec | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) -- 2.35.1

2 7

[PATCH 00/11] Improve command dispatch in test scripts
by David Gibson 02 Sep '22

02 Sep '22

The test scripts need to dispatch commands in various different contexts: on the host, in namespaces, and in guests. Currently this is done by running each context in a tmux pane and simulating typing into that pane using tmux commands. This has a number of problems: * It's slow * Getting the results from programs is tricky and error prone. We can misparse output if: * The window isn't large enough * Commands produce output which looks like a shell prompt * The shell doesn't have the prompt we expect * A accidental keypress while the tests are running can mess up the test run * We have to issue explicit "echo $?" commands to check if things succeeded This series adds a new subsystem to the test scripts for dispatching commands which accurately preserves output (both stdout and stderr), and exit codes. Most of the testsuite is converted to this new system. For now the distro tests still rely on the old approach, since they have some complications that require additional handling. That does, alas, mean we have some ugly transitional code remaining. Oh well, we should be able to clean that up at some point. This series is pased on the previous series cleaning up the performance tests. David Gibson (11): test: Correctly match "background" with "wait" commands test: Context execution helpers test: Allow a tmux pane to watch commands executed in contexts test: Integration of old-style pane execution and new context execution test: Issue host commands via context for most tests test: Use new-style contexts for passt pane in the pasta and passt tests test: Add nsholder utility test: Extend context system to run commands in namespace for pasta tests test: Use context system for guest commands test: Use context system for two_guests tests test: Use new-style command issue for passt_in_ns tests test/.gitignore | 3 + test/Makefile | 13 +- test/lib/context | 127 ++++++++++++++++++ test/lib/layout | 58 +++----- test/lib/setup | 274 +++++++++++++++++++------------------- test/lib/term | 84 ++++++++++++ test/lib/test | 138 ++++++++----------- test/nsholder.c | 117 ++++++++++++++++ test/passt.mbuto | 32 ++++- test/run | 1 + test/shutdown/passt | 4 +- test/shutdown/passt_in_ns | 4 +- test/tcp/passt | 2 + test/tcp/passt_in_ns | 17 ++- test/tcp/pasta | 1 - test/udp/passt | 1 + test/udp/passt_in_ns | 5 +- test/udp/pasta | 1 - 18 files changed, 608 insertions(+), 274 deletions(-) create mode 100644 test/lib/context create mode 100644 test/nsholder.c -- 2.37.2

1 11

[PATCH 0/4] Cleanups to performance tests
by David Gibson 01 Sep '22

01 Sep '22

A handful of changes to the performance tests to make them more robust and faster. David Gibson (4): gitignore pidfiles other than passt.pid test: Combine iperf3c and iperf3s into a single DSL command test: Parameterize run time for throughput performance tests test: Rewrite test_iperf3 .gitignore | 3 +- test/lib/test | 113 +++++++++++++++++++------------------------- test/perf/passt_tcp | 68 ++++++++++---------------- test/perf/passt_udp | 63 +++++++++--------------- test/perf/pasta_tcp | 59 +++++++++-------------- test/perf/pasta_udp | 61 ++++++++++-------------- 6 files changed, 143 insertions(+), 224 deletions(-) -- 2.37.2

1 4

[PATCH 0/8] Allow pasta to take a command to spawn instead of shell
by David Gibson 01 Sep '22

01 Sep '22

When not attaching to an existing network namespace, pasta always spawns an interactive shell in a new namespace to attach to. Most commands which can issue a shell in a modified environment can also issue other commands as well (e.g. env, strace). We want to allow pasta to do the same. Because of the way the non-option argument to pasta is currently overloaded, allowing this requires some other changes to the way we parse the command line. David Gibson (8): conf: Make the argument to --pcap option mandatory conf: Use "-D none" and "-S none" instead of missing empty option arguments Correct manpage for --userns Remove --nsrun-dir option Move ENOENT error message into conf_ns_opt() More deterministic detection of whether argument is a PID, PATH or NAME Use explicit --netns option rather than multiplexing with PID Allow pasta to take a command to execute conf.c | 269 +++++++++++++++++++++++++++++--------------------------- passt.1 | 54 ++++++------ pasta.c | 33 ++++--- pasta.h | 2 +- pcap.c | 28 ------ 5 files changed, 191 insertions(+), 195 deletions(-) -- 2.37.2

2 17

[PATCH] test/README: Requirements for socket buffer sizes and hardware performance events
by Stefano Brivio 01 Sep '22

01 Sep '22

Performance tests use iperf3(1) with large windows, and these sysctl entries are needed to run them unmodified. The passt demo uses perf(1) to report syscall overhead, and that needs access to hardware performance counters for unprivileged users. Reported-by: David Gibson <david(a)gibson.dropbear.id.au> Signed-off-by: Stefano Brivio <sbrivio(a)redhat.com> --- test/README.md | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) diff --git a/test/README.md b/test/README.md index 581949a..72a4986 100644 --- a/test/README.md +++ b/test/README.md @@ -47,6 +47,22 @@ Virtual machine images are built during test executions using [mbuto](https://mbuto.lameexcu.se/), the shell script is sourced via _git_ as needed, so there's no need to actually install it. +### Kernel parameters + +Performance tests use iperf3 with rather large TCP receiving and sending +windows, to decrease the likelihood of iperf3 itself becoming the bottleneck. +These values need to be allowed by the kernel of the host running the tests. +Example for /etc/sysctl.conf: + + net.core.rmem_max = 134217728 + net.core.wmem_max = 134217728 + +Further, the passt demo uses perf(1), relying on hardware events for performance +counters, to display syscall overhead. The kernel needs to allow unprivileged +users to access these events. Suggested entry for /etc/sysctl.conf: + + kernel.perf_event_paranoid = -1 + ### Special requirements for continuous integration and demo modes Running the test suite as continuous integration or demo modes will record the -- 2.35.1

2 1