Date: Fri, 31 Mar 2023 09:15:27 +0900 From: Kristof Provost <kp@FreeBSD.org> To: Mark Johnston <markj@freebsd.org> Cc: src-committers@freebsd.org, dev-commits-src-all@freebsd.org, dev-commits-src-main@freebsd.org Subject: Re: git: b60600ceeb68 - main - pf tests: Serialize Message-ID: <DFBE4D6A-4CA1-4E24-AA67-A1564AF96968@FreeBSD.org> In-Reply-To: <ZCYj50rKEtzIgqc0@nuc> References: <202303302336.32UNatba067594@gitrepo.freebsd.org> <3A26AB82-357F-421F-853E-07320387ACBE@FreeBSD.org> <ZCYj50rKEtzIgqc0@nuc>
next in thread | previous in thread | raw e-mail | index | archive | help
On 31 Mar 2023, at 9:05, Mark Johnston wrote: > On Fri, Mar 31, 2023 at 08:56:56AM +0900, Kristof Provost wrote: >> On 31 Mar 2023, at 8:36, Mark Johnston wrote: >>> The branch main has been updated by markj: >>> >>> URL: https://cgit.FreeBSD.org/src/commit/?id=3Db60600ceeb68d1001d6122= 2830e0be3441ef0885 >>> >>> commit b60600ceeb68d1001d61222830e0be3441ef0885 >>> Author: Mark Johnston <markj@FreeBSD.org> >>> AuthorDate: 2023-03-25 12:55:41 +0000 >>> Commit: Mark Johnston <markj@FreeBSD.org> >>> CommitDate: 2023-03-30 23:35:59 +0000 >>> >>> pf tests: Serialize >>> >>> These tests reuse jail names and cannot run in parallel. Until t= his is >>> fixed - which is desirable since these takes take a while to run = - tell >>> kyua to serialize them. >>> >> >> The tests also recycle IP ranges, so merely changing the jail names is= insufficient. > > Could you please give an example? I looked at some of the tests but ca= n > only see cases where addresses are assigned within the vnet jail(s) > created by the tests, in which case I'd expect no problems. > Altq:hfsq assigns 192.0.2.1/24 on an epair on the host, so does altq:matc= h, altq:cbq_vlan, altq:codel_bridge, dup:dup_to, ether:mac, ether:proto a= nd .. then I got bored looking. It=E2=80=99s not just the pf tests either. For example the netinet/forwar= d:fwd_ip_icmp_iface_fast_success test does that too (well, 192.0.2.1/29, = but anyway). There are going to be more, this is just from a very quick l= ook. >> Realistically the easiest way to get these to run in parallel would be= to run each test in its own vnet so both overlapping IP ranges and name = conflicts don=E2=80=99t matter. > > Yeah, I was wondering whether it'd be possible to have kyua handle the > creation and teardown of a per-test jail, if only to avoid having to go= > through all of the tests which use static jail names. That=E2=80=99s what I was thinking as well, yes. We might be able to get to a point where all the tests have unique jail n= ames, but we=E2=80=99re never going to manage to de-conflict all of the I= P assignments. Even if we could, it=E2=80=99d make writing tests harder a= nd that=E2=80=99s counterproductive. The only issue I can think of with having the tests run in their own vnet= is that there may be some tests that play with non-vneted sysctls, so we= will have to make that vnet feature optional. Best regards, Kristof
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?DFBE4D6A-4CA1-4E24-AA67-A1564AF96968>