linux-pci.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
To: Rob Herring <robh@kernel.org>
Cc: PCI <linux-pci@vger.kernel.org>, Marc Zyngier <maz@kernel.org>,
	Bjorn Helgaas <bhelgaas@google.com>,
	Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Subject: Re: [PATCH] PCI: of: Fix MSI domain lookup with child bus nodes
Date: Mon, 16 Aug 2021 10:42:52 +0200	[thread overview]
Message-ID: <20210816104252.045a7b75@coco.lan> (raw)
In-Reply-To: <CAL_JsqL+aPRh9CJ0iKbzGCyE3wDR6QMEKAEC8p=1ZVRDSb6JDA@mail.gmail.com>

Em Sat, 14 Aug 2021 12:32:56 -0500
Rob Herring <robh@kernel.org> escreveu:

> On Fri, Aug 13, 2021 at 12:19 PM Mauro Carvalho Chehab
> <mchehab+huawei@kernel.org> wrote:
> >
> > Em Fri, 13 Aug 2021 11:02:57 -0500
> > Rob Herring <robh@kernel.org> escreveu:
> >  
> > > When a DT contains PCI child bus nodes, lookup of the MSI domain on PCI
> > > buses fails resulting in the following warnings:
> > >
> > > WARNING: CPU: 4 PID: 7 at include/linux/msi.h:256 __pci_enable_msi_range+0x398/0x59c
> > >
> > > The issue is that pci_host_bridge_of_msi_domain() will check the DT node of
> > > the passed in bus even if it's not the host bridge's bus. Based on the
> > > name of the function, that's clearly not what we want. Fix this by
> > > walking the bus parents to the root bus.
> > >
> > > Reported-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
> > > Cc: Marc Zyngier <maz@kernel.org>
> > > Cc: Bjorn Helgaas <bhelgaas@google.com>
> > > Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> > > Signed-off-by: Rob Herring <robh@kernel.org>
> > > ---
> > > Compile tested only. Mauro, Can you see if this fixes your issue.
> > >
> > >  drivers/pci/of.c | 4 ++++
> > >  1 file changed, 4 insertions(+)
> > >
> > > diff --git a/drivers/pci/of.c b/drivers/pci/of.c
> > > index a143b02b2dcd..ea70aede054c 100644
> > > --- a/drivers/pci/of.c
> > > +++ b/drivers/pci/of.c
> > > @@ -84,6 +84,10 @@ struct irq_domain *pci_host_bridge_of_msi_domain(struct pci_bus *bus)
> > >       if (!bus->dev.of_node)
> > >               return NULL;
> > >
> > > +     /* Find the host bridge bus */
> > > +     while (!pci_is_root_bus(bus))
> > > +             bus = bus->parent;
> > > +
> > >       /* Start looking for a phandle to an MSI controller. */
> > >       d = of_msi_get_domain(&bus->dev, bus->dev.of_node, DOMAIN_BUS_PCI_MSI);
> > >       if (d)  
> >
> > Nope, it didn't solve the issue.  
> 
> Can you try adding some prints of the domain, pci dev, and DT node to
> pci_set_bus_msi_domain(). Comparing those when having child nodes or
> not would be helpful.

Debug patch enclosed.

This is what happens when msi-parent is at /soc/pcie@f4000000/pcie@0,0:

[    4.305247] pci_bus 0000:01: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0: IRQ domain: ffff000104a2f6c0
[    4.442212] pci_bus 0000:02: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0: IRQ domain: ffff000104a2f6c0
[    4.688145] pci_bus 0000:03: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0/pcie@1,0: IRQ domain: ffff000104a2f6c0
[    4.800613] pci_bus 0000:04: pci_set_bus_msi_domain: of_node (null): IRQ domain: ffff000104a2f6c0
[    4.856254] pci_bus 0000:05: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0/pcie@5,0: IRQ domain: ffff000104a2f6c0
[    4.922117] pci_bus 0000:06: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0/pcie@7,0: IRQ domain: ffff000104a2f6c0
[    5.032708] pci_bus 0000:07: pci_set_bus_msi_domain: of_node (null): IRQ domain: ffff000104a2f6c0

And this is what happens when msi-parent is at /soc/pcie@f4000000
(either with or without your patch applied):

[    4.120328] pci_bus 0000:01: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0: IRQ domain: 0000000000000000
[    4.214541] pci_bus 0000:02: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0: IRQ domain: 0000000000000000
[    4.226054] pci_bus 0000:01: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0: IRQ domain: 0000000000000000
[    4.478858] pci_bus 0000:03: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0/pcie@1,0: IRQ domain: 0000000000000000
[    4.491218] pci_bus 0000:02: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0: IRQ domain: 0000000000000000
[    4.502793] pci_bus 0000:01: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0: IRQ domain: 0000000000000000
[    4.588196] pci_bus 0000:04: pci_set_bus_msi_domain: of_node (null): IRQ domain: 0000000000000000
[    4.597161] pci_bus 0000:02: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0: IRQ domain: 0000000000000000
[    4.608747] pci_bus 0000:01: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0: IRQ domain: 0000000000000000
[    4.658892] pci_bus 0000:05: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0/pcie@5,0: IRQ domain: 0000000000000000
[    4.671241] pci_bus 0000:02: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0: IRQ domain: 0000000000000000
[    4.682869] pci_bus 0000:01: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0: IRQ domain: 0000000000000000
[    4.732938] pci_bus 0000:06: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0/pcie@7,0: IRQ domain: 0000000000000000
[    4.745305] pci_bus 0000:02: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0: IRQ domain: 0000000000000000
[    4.756880] pci_bus 0000:01: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0: IRQ domain: 0000000000000000
[    4.850363] pci_bus 0000:07: pci_set_bus_msi_domain: of_node (null): IRQ domain: 0000000000000000
[    4.859322] pci_bus 0000:02: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0/pcie@0,0: IRQ domain: 0000000000000000
[    4.870895] pci_bus 0000:01: pci_set_bus_msi_domain: of_node /soc/pcie@f4000000/pcie@0,0: IRQ domain: 0000000000000000

Btw, despite lspci works on both cases, the Ethernet adapter stops
working when msi-parent is at /soc/pcie@f4000000. I didn't notice it
before, as (up to last week) I was using WiFi to connect to this board.

Thanks,
Mauro

[PATCH] PCI: probe: Add a debug printk at pci_set_bus_msi_domain

That helps to discover problems when trying to get the MSI IRQ
domain.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
index c5dfc1afb1d3..f73bd81922b3 100644
--- a/drivers/pci/probe.c
+++ b/drivers/pci/probe.c
@@ -866,6 +866,8 @@ static void pci_set_bus_msi_domain(struct pci_bus *bus)
        for (b = bus, d = NULL; !d && !pci_is_root_bus(b); b = b->parent) {
                if (b->self)
                        d = dev_get_msi_domain(&b->self->dev);
+               dev_dbg(&b->dev, "%s: of_node %pOF: IRQ domain: %px\n",
+                       __func__, b->dev.of_node, d);
        }
 
        if (!d)


  reply	other threads:[~2021-08-16  8:42 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-08-13 16:02 [PATCH] PCI: of: Fix MSI domain lookup with child bus nodes Rob Herring
2021-08-13 17:19 ` Mauro Carvalho Chehab
2021-08-14 17:32   ` Rob Herring
2021-08-16  8:42     ` Mauro Carvalho Chehab [this message]
2021-08-16 13:36       ` Mauro Carvalho Chehab
2021-08-16 19:01         ` Rob Herring
2021-08-16 19:47           ` Mauro Carvalho Chehab

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210816104252.045a7b75@coco.lan \
    --to=mchehab+huawei@kernel.org \
    --cc=bhelgaas@google.com \
    --cc=linux-pci@vger.kernel.org \
    --cc=lorenzo.pieralisi@arm.com \
    --cc=maz@kernel.org \
    --cc=robh@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).