Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 9 Dec 2024 19:19:20 +0100
From:      FreeBSD User <freebsd@walstatt-de.de>
To:        Tomoaki AOKI <junchoon@dec.sakura.ne.jp>
Cc:        Juraj Lutter <otis@FreeBSD.org>, Ronald Klop <ronald@FreeBSD.org>, freebsd-current@freebsd.org
Subject:   Re: (ipfw) Re: HELP! fetch: stuck forever OR error: RPC failed: curl 56 recv failure: Operation timed out
Message-ID:  <20241209191947.39ac4843@thor.intern.walstatt.dynvpn.de>
In-Reply-To: <20241210022710.88c9087dd7cb09774507f232@dec.sakura.ne.jp>
References:  <20241206034709.4dd32cc5@thor.intern.walstatt.dynvpn.de> <279848701.11738.1733510402875@localhost> <20241206210947.3ae835e4@thor.intern.walstatt.dynvpn.de> <f8952585-4b68-4cfd-a60f-1ebbd7f2545f@FreeBSD.org> <8E43EAA1-BA3E-4655-ACE1-2E4523E901DE@FreeBSD.org> <20241209214314.2443b590d774423a2b97f0a8@dec.sakura.ne.jp> <20241209174541.39c286f5@thor.intern.walstatt.dynvpn.de> <20241210022710.88c9087dd7cb09774507f232@dec.sakura.ne.jp>

next in thread | previous in thread | raw e-mail | index | archive | help
Am Tue, 10 Dec 2024 02:27:10 +0900
Tomoaki AOKI <junchoon@dec.sakura.ne.jp> schrieb:

My apology for topposting.

The host I first realised the problems is updated on an almost daily basis =
and the issue
reported started last weekend.

A possible candidate could be

https://cgit.freebsd.org/src/commit/sys/netpfil/ipfw?id=3D0fc7bdc978366abb4=
351b0b76b50a5848cc5d982

since the other, younger, seem innocent. I try to revert the patch mentione=
d and see ...

> On Mon, 9 Dec 2024 17:45:14 +0100
> FreeBSD User <freebsd@walstatt-de.de> wrote:
>=20
> > Am Mon, 9 Dec 2024 21:43:14 +0900
> > Tomoaki AOKI <junchoon@dec.sakura.ne.jp> schrieb:
> >  =20
> > > On Mon, 9 Dec 2024 11:09:14 +0100
> > > Juraj Lutter <otis@FreeBSD.org> wrote:
> > >  =20
> > > > > On 8 Dec 2024, at 20:30, Ronald Klop <ronald@FreeBSD.org> wrote:
> > > > >=20
> > > > > Hi,
> > > > >=20
> > > > > I can reproduce your error.
> > > > >=20
> > > > > A cronjob which does a scp to another server didn't work anymore.=
 When I go back to
> > > > > the previous BE it works fine again. Ipfw disable firewall also m=
akes the scp work.
> > > > >=20
> > > > > Scp also seems to work fine if I replace the statefull firewall r=
ules with stateless
> > > > > "pass all from any to any".   =20
> > > >=20
> > > > Have you tried to allow ICMP in both directions explicitly, in case=
 of stateful rules?
> > > >=20
> > > > =E2=80=94
> > > > Juraj Lutter
> > > > otis@FreeBSD.org   =20
> > >=20
> > > I think would usually work for clients with some limited services
> > > exposed to outside. IIUC, it basically allow all sessions from inside
> > > and allows limited serivices configured with variables
> > > via /etc/rc.conf[.local].
> > >=20
> > > Some notes.
> > >   *Last actual changes in /usr/src/libexec/rc/rc.firewall was at
> > >    Jul.23, 2020.
> > >      https://github.com/freebsd/freebsd-src/commits/main/libexec/rc/r=
c.firewall
> > >        [cgit.freebsd.org seems to be unstable now.]
> > >=20
> > >   *Variable firewall_logif currently does not exist.
> > >=20
> > >   *Don't you need allowing 22/UDP, too, like below?
> > >      firewall_myservices=3D"22/tcp 22/udp"
> > >=20
> > > And if you're creating kernel config from scratch (such as copying fr=
om
> > > GENERIC at some point and editing it), it's no longer adviced.
> > > It's not robust for changes in GENERIC.
> > >=20
> > > Instead, include GENERIC and describe changes you want.
> > >=20
> > > An example (one of my test kernel config for a bit old stable).
> > >=20
> > >    =3D=3D=3D=3D=3D Start example =3D=3D=3D=3D=3D
> > > =20
> > > include GENERIC
> > >=20
> > > ident   TEST15
> > >=20
> > > nooptions       DDB
> > > nooptions       GDB=20
> > > nooptions       INVARIANTS
> > > nooptions       INVARIANT_SUPPORT
> > > nooptions       WITNESS
> > > nooptions       WITNESS_SKIPSPIN
> > > nooptions       DEADLKRES
> > >=20
> > > options         CAM_IOSCHED_DYNAMIC
> > >=20
> > > device          sg
> > >=20
> > >    =3D=3D=3D=3D=3D End example =3D=3D=3D=3D=3D
> > >=20
> > >  =20
> >=20
> > Thank you very much for the advice - but I do this kind of confugration=
 now since, I guess,
> > 2020 or 2021. consider the host's kernel name to be "THOR", then /etc/s=
rc.conf has lines
> >=20
> > KERNCONF=3D       THOR
> > KERNCONFDIR=3D    /etc/config/amd64/kernel_conf/
> >=20
> > and the target's config file "/etc/config/amd64/kernel_conf/THOR" conta=
ins
> >=20
> > include         GENERIC
> > include         NODEVICE-THOR
> > include         "std.nodebug"
> > include         ADDON-THOR
> >=20
> > This concept isn't bullet proof, since I had trouble with the relativel=
y recent introduced
> > "std.nodebug". As you mentioned, NODEVICE contains ALL "nooptions" and =
"nodevice" and ADDON
> > contains some extra options not contained in GENERIC. GENERIC is a symb=
olic link to the
> > original GENERIC in the appropriate sys folder.
> >=20
> > Thanks to FReeBSD's sophisticated kernel configuration, this hierarchic=
al scheme prevents
> > most accidents triggered by significant GENERIC changes.
> >=20
> > Do you suspect a misconfiguration due to uncaught changes in GENERIC?=20
> >=20
> > --=20
> > O. Hartmann =20
>=20
> Just a possibility. Not sure when it was and which firewall (ipfw? pf?
> or else?) it was, but IIRC, I've seen some excessive configs to enable
> non-default options/devices built into kernel caused unwanted changes
> in default before when I was searching unrelated things of FreeBSD.
>=20
> And why I've mentioned about kernel configuration file is because I've
> bitten by the changes in GENERIC before.
>=20
> In contrast with rc.firewall, kernel side of ipfw received some changes
> this year.
>=20
>   https://cgit.freebsd.org/src/log/sys/netpfil/ipfw
>=20
> So anything committed after your previous build could cause the issue.
> Sorry, I have not enough time to dig into deeper. Other points to worth
> checking would be:
>=20
>   https://cgit.freebsd.org/src/log/sys/netgraph
>=20
>   https://cgit.freebsd.org/src/tree/sys/net
>=20
> The latter should be limited with bpf related (otherwise too bload).
> net and netinet would be too bload, too. Bloader codes should be
> checked when narrow and close codes are analysed to be NOT affecting.
> Otherwise, there would be too much confusions.
>=20



--=20
O. Hartmann



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20241209191947.39ac4843>