Re: [PATCH net-next v2 5/7] selftests: rds: Fix gcov and pcap collection

[Allison Henderson <achender@kernel.org>]

On Wed, 2026-04-29 at 19:48 -0700, Jakub Kicinski wrote:
> On Wed, 29 Apr 2026 19:42:06 -0700 Jakub Kicinski wrote:
> > > +# tcpdump saves pcaps to /tmp because it requires chown to save the
> > > +# pcap but chown is not supported by 9p.  Mount tmpfs on /tmp if it is
> > > +# not already a separate filesystem
> > > +if ! mountpoint -q /tmp 2>/dev/null; then
> > > +	mount -t tmpfs tmpfs /tmp
> > > +fi
> > > +  
> > 
> > Could this introduce a regression when the test is run directly on a host
> > workstation rather than inside an isolated VM?
> > 
> > If /tmp is a regular directory on the root filesystem, mounting tmpfs over it
> > will instantly hide all existing files and UNIX domain sockets in /tmp.
> > 
> > Since there is no cleanup trap to unmount it on exit or failure, does this
> > leave the host system in a degraded state?
> 
> I share Sashiko's mixed feelings here. Do other tests mount /tmp ?
> Seems like something that belongs in the "CI setup", external
> to ksft itself.
Ok, I did some looking around, and while I dont see any tests that mount /tmp, I did notice that mptcp_join.sh and
simult_flows.sh use a $SUDO_USER, presumably to get away from the chown issue. But without getting off the 9p fs
with some sort of scratch area, we still end up with empty pcaps.

> 
> point #2 my vng does have an overlayfs mounted over /tmp so the
> mountpoint check doesn't trigger IDK if this is what you meant 
> or I have a different version.

I did try using --overlay-rwdir, and I think that gives rw to the guest, but it's ro to the host, so we dont get to keep
the pcaps post mortem.

What we can do, if it sounds ok to you, is set up a temp scratch area in the rds_logs folder so it's not mounting over
/tmp, and then run.sh can handle the mount with a cleanup trap addressing Sashiko's concern

Let me know what you think?

Thank you for the reviews!
Allison