[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20240809181434.GA204249@bhelgaas>
Date: Fri, 9 Aug 2024 13:14:34 -0500
From: Bjorn Helgaas <helgaas@...nel.org>
To: Guilherme Giácomo Simões <trintaeoitogc@...il.com>
Cc: Ilpo Järvinen <ilpo.jarvinen@...ux.intel.com>,
bhelgaas@...gle.com, linux-pci@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] PCI: remove type return
On Thu, Aug 08, 2024 at 06:05:54PM -0300, Guilherme Giácomo Simões wrote:
> Bjorn Helgaas <helgaas@...nel.org> writes:
> >
> > On Tue, Aug 06, 2024 at 05:54:15PM -0300, Guilherme Giácomo Simões wrote:
> > > Ilpo Järvinen <ilpo.jarvinen@...ux.intel.com> wrote:
> > > > On Sat, 3 Aug 2024, Guilherme Giacomo Simoes wrote:
> > > >
> > > > > I can see that the function pci_hp_add_brigde have a int return
> > > > > propagation.
> > > ...
> >
> > > > The lack of return value checking seems to be on the list in
> > > > pci_hp_add_bridge(). So perhaps the right course of action would be to
> > > > handle return values correctly.
> > >
> > > Ok, so if the right course is for the driver to handle return value,
> > > then this is a
> > > task for the driver developers, because only they know what to do when
> > > pci_hp_add_bridge() doesn't work correctly, right?
> >
> > pci_hp_add_bridge() is only for hotplug drivers, so the list of
> > callers is short and completely under our control. There's plenty of
> > opportunity for improving this. Beyond just the return value, all the
> > callers of pci_hp_add_bridge() should be doing much of the same work
> > that could potentially be factored out.
>
> Okay, then what the action that the drivers must be do when the add
> bridge is failed?
pci_hp_add_bridge() fails when there's no bus number available to
assign to new hot-added devices. When that happens, there's really
nothing the hotplug drivers can do to improve the situation.
pci_hp_add_bridge() already logs a message for one of the failure
cases. It may be that it should also log a message for the other
failure case. The end result is that we can't use the hot-added
devices because there's no space for them in the PCI bus number space,
so we can't address them.
Bjorn
Powered by blists - more mailing lists