Re: 3.16rc3 multiplatform, Armada 370 and IOMMU: unbootable kernel

[Jason Cooper <jason@lakedaemon.net>]

Ezequiel,

On Mon, Jul 07, 2014 at 08:37:58PM -0300, Ezequiel Garcia wrote:
> On 07 Jul 11:30 AM, Greg Kroah-Hartman wrote:
> > On Mon, Jul 07, 2014 at 07:58:18AM -0300, Ezequiel Garcia wrote:
> [..]
> > > 
> > > I guess I snipped the thread and lost most of the information about the panic.
> > > Here's the original bug report:
> > > 
> > > http://www.spinics.net/lists/arm-kernel/msg344059.html
> > > 
> > > The problem reported involves enabling OMAP IOMMU driver and not any other IOMMU
> > > driver. Doing some tracing and adding a few prints, we found that
> > > omap_iommu_init() sets a bus notifier for the platform bus type:
> > > 
> > > omap_iommu_init -> bus_set_iommu -> iommu_bus_init:
> > > 
> > > static void iommu_bus_init(struct bus_type *bus, struct iommu_ops *ops)
> > > {
> > >         bus_register_notifier(bus, &iommu_bus_nb);
> > >         bus_for_each_dev(bus, NULL, ops, add_iommu_group);
> > > }
> > > 
> > > But the iommu bus notifier gets called for the 'pci' bus type, which
> > > has the iommu_ops field NULL (since it hasn't been set for iommu).
> > 
> > So this is what needs to be figured out, how is the notifier being
> > called with a PCI device?  Who else called iommu_bus_init() for the PCI
> > bus?
> > 
> 
> Nobody. The only caller of iommu_bus_init() the above OMAP IOMMU.
> 
> However, I found something interesting. Tried to bisect this without much
> luck; I did two bisects and ended up in the same merge commit:
> 
> # good: [0f16aa3c24a216d14d7f0587e1cbd2c1b51a38f3] Merge tag 'samsung-dt-3' of git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung into next/dt
> # first bad commit: [755a9ba7bf24a45b6dbf8bb15a5a56c8ed12461a] Merge tag 'dt-for-3.16' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc into next
> 
> So, after doing a few diff's between that good and bad and searching for
> "bus_notifier" changes, saw something strange in arch/arm/mach-mvebu/coherency.c.
> 
> It seems bus_register_notifier() is been called for platform and pci devices
> with the *same* notifier block. Haven't looked close enough, but you mentioned
> that could cause trouble?
> 
> This patch fixes the issue here:

Thanks for digging in to this!  Please submit this as patch on it's own
and I'll merge it in.

thx,

Jason.