port forward range test: fix an oops

[edsantiago]

Wrong variable.

Fixes: #14854

Signed-off-by: Ed Santiago [email protected]

None

[openshift-ci]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: edsantiago

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [edsantiago]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[rhatdan]

LGTM
@containers/podman-maintainers PTAL

[vrothberg]

/lgtm
/hold

[edsantiago]

@Luap99 PTAL. The obvious question is, how did this ever work?

[Luap99]

It simply did not work. We just never looped in for port.

But this is not even the actual problem I was reporting. The test picked 5354-5356 which is wrong, 5355 is always blocked. So the free port logic has a bug as well.

[edsantiago]

Testing this on my laptop, rootless, I'm reliably getting:

$ bin/podman run --network bridge -p 5933-5955:5933-5955 -d quay.io/libpod/testimage:20220615 top
Error: failed to mount runtime directory for rootless netns: no such file or directory

Looks like #13703, but this is f36, and it's not an SELinux issue (fails with setenforce 0). This morning's dnf upgrade brought in a new glibc, could that be it?

[Luap99]

Testing this on my laptop, rootless, I'm reliably getting:

$ bin/podman run --network bridge -p 5933-5955:5933-5955 -d quay.io/libpod/testimage:20220615 top
Error: failed to mount runtime directory for rootless netns: no such file or directory

Looks like #13703, but this is f36, and it's not an SELinux issue (fails with setenforce 0). This morning's dnf upgrade brought in a new glibc, could that be it?

That error can happen because of many different reasons. Try to cleanup the netns rm -f $XDG_RUNTIME_DIR/netns/rootless-netns*

[edsantiago]

Thanks, that fixed it. I wonder if the error message could be fixed so it offers that command (rm -f ...) as a suggestion? That will be helpful to people who get that error message without having a clue what happened or how to fix it. Like me.

[Luap99]

Thanks, that fixed it. I wonder if the error message could be fixed so it offers that command (rm -f ...) as a suggestion? That will be helpful to people who get that error message without having a clue what happened or how to fix it. Like me.

Well there are downsides to this, for testing it is fine but in production you should not do it. The problem is that without the namespace newly started containers can not talk to containers started before you removed it.

This whole thing is fragile and needs to be improved. Podman should be able to recover from this on its own without user intervention.

[Luap99]

@edsantiago I think we also need:

diff --git a/test/system/helpers.bash b/test/system/helpers.bash
index 273e8d2f5..e4d44c4a9 100644
--- a/test/system/helpers.bash
+++ b/test/system/helpers.bash
@@ -300,7 +300,7 @@ function random_free_port_range() {
     while [[ $maxtries -gt 0 ]]; do
         local firstport=$(random_free_port)
         local all_ports_free=1
-        for i in $(seq 2 $size); do
+        for i in $(seq 1 $(($size - 1))); do
             if ! port_is_free $((firstport + $i)); then
                 all_ports_free=
                 break

right now it tests port, port + 2, port + 3 so it failed to check port + 1

[edsantiago]

Yes, but there's also something else seriously broken. The whole exec /dev/tcp thing is completely broken. I'm trying to figure out why.

[edsantiago]

This blew up. New changes are much, much more complicated than originally intended. However, the end result is IMO cleaner.

[edsantiago]

SUBTLE BUT IMPORTANT! Note the change of dash-asterisk to asterisk-star in this one!

[Luap99]

/lgtm

[edsantiago]

/hold cancel