MIME-Version: 1.0
References: <20230317113427.302162-1-noltari@gmail.com> <20230317113427.302162-3-noltari@gmail.com>
 <20230317115115.s32r52rz3svuj4ed@skbuf> <CAKR-sGe3xHkN-1+aLn0ixnskctPK4GTzfXu8O_dkFhHyY1nTeg@mail.gmail.com>
 <20230317130434.7cbzk5gxx5guarcz@skbuf> <CAKR-sGeFZLnuqH=4Gok1URJEvrQKxbk203Q8zdMd9830G_XD7A@mail.gmail.com>
 <20230317142919.hhjd64juws35j47o@skbuf>
In-Reply-To: <20230317142919.hhjd64juws35j47o@skbuf>
From:   =?UTF-8?B?w4FsdmFybyBGZXJuw6FuZGV6IFJvamFz?= <noltari@gmail.com>
Date:   Fri, 17 Mar 2023 17:23:34 +0100
Message-ID: <CAKR-sGc7u346XqoihOuDse3q=d8HG6er3H6R1NCm_pQeNW7edA@mail.gmail.com>
Subject: Re: [PATCH 2/3] net: dsa: b53: mmap: register MDIO Mux bus controller
To:     Vladimir Oltean <olteanv@gmail.com>
Cc:     davem@davemloft.net, edumazet@google.com, kuba@kernel.org,
        pabeni@redhat.com, robh+dt@kernel.org,
        krzysztof.kozlowski+dt@linaro.org, f.fainelli@gmail.com,
        jonas.gorski@gmail.com, andrew@lunn.ch, hkallweit1@gmail.com,
        linux@armlinux.org.uk, netdev@vger.kernel.org,
        devicetree@vger.kernel.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

El vie, 17 mar 2023 a las 15:29, Vladimir Oltean (<olteanv@gmail.com>) escr=
ibi=C3=B3:
>
> On Fri, Mar 17, 2023 at 03:17:12PM +0100, =C3=81lvaro Fern=C3=A1ndez Roja=
s wrote:
> > > The proposed solution is too radical for a problem that was not prope=
rly
> > > characterized yet, so this patch set has my temporary NACK.
> >
> > Forgive me, but why do you consider this solution too radical?
>
> Because it involves changing device tree bindings (stable ABI) in an
> incompatible way.
>
> > >
> > > > But maybe Florian or Jonas can give some more details about the iss=
ue...
> > >
> > > I think you also have the tools necessary to investigate this further=
.
> > > We need to know what resource belonging to the switch is it that the
> > > MDIO mux needs. Where is the earliest place you can add the call to
> > > b53_mmap_mdiomux_init() such that your board works reliably? Note tha=
t
> > > b53_switch_register() indirectly calls b53_setup(). By placing this
> > > function where you have, the entirety of b53_setup() has finished
> > > execution, and we don't know exactly what is it from there that is
> > > needed.
> >
> > In the following link you will find different bootlogs related to
> > different scenarios all of them with the same result: any attempt of
> > calling b53_mmap_mdiomux_init() earlier than b53_switch_register()
> > will either result in a kernel panic or a device hang:
> > https://gist.github.com/Noltari/b0bd6d5211160ac7bf349d998d21e7f7
> >
> > 1. before b53_switch_register():
> >
> > 2. before dsa_register_switch():
> >
> > 3. before b53_switch_init():
>
> Did you read what I said?

Yes, but I didn't get your point, sorry for that.

>
> | Note that b53_switch_register() indirectly calls b53_setup(). By placin=
g
> | this function where you have, the entirety of b53_setup() has finished
> | execution, and we don't know exactly what is it from there that is
> | needed.
>
> Can you place the b53_mmap_mdiomux_init() in various places within
> b53_setup() to restrict the search further?

I tried and these are the results:
https://gist.github.com/Noltari/d5bdba66b8f2e392c9e4c2759661d862

All of them hang when dsa_tree_setup() is called for DSA tree 1
(external switch) without having completely setup DSA tree 0 (internal
switch):
[ 1.471345] b53-switch 10e00000.switch: found switch: BCM63xx, rev 0
[ 1.481099] bcm6368-enetsw 1000d800.ethernet: IRQ tx not found
[ 1.506752] bcm6368-enetsw 1000d800.ethernet: mtd mac 4c:60:de:86:52:12
[ 1.594365] bcm7038-wdt 1000005c.watchdog: Registered BCM7038 Watchdog
[ 1.612008] NET: Registered PF_INET6 protocol family
[ 1.645617] Segment Routing with IPv6
[ 1.649547] In-situ OAM (IOAM) with IPv6
[ 1.653948] NET: Registered PF_PACKET protocol family
[ 1.659984] 8021q: 802.1Q VLAN Support v1.8
[ 1.699193] b53-switch 10e00000.switch: found switch: BCM63xx, rev 0
[ 2.124257] bcm53xx 0.1:1e: found switch: BCM53125, rev 4
*** Device hang ***

I don't know if there's a way to defer the probe of DSA tree 1 (the
external switch) until DSA tree 0 (the internal switch) is completely
setup, because that would probably be the only alternative way of
fixing this.

--
=C3=81lvaro.