Re: [PATCH] PCI: imx6: install the fault handler only if we are really running on a compatible device

[H. Nikolaus Schaller]

Hi Bjorn,

> Am 08.03.2023 um 19:49 schrieb Bjorn Helgaas <helgaas@xxxxxxxxxx>:
> 
> On Tue, Feb 28, 2023 at 09:43:54AM +0100, H. Nikolaus Schaller wrote:
>> commit bb38919ec56e ("PCI: imx6: Add support for i.MX6 PCIe controller")
>> added a fault hook to this driver in the probe function. So it was only
>> installed if needed.
>> 
>> commit bde4a5a00e76 ("PCI: imx6: Allow probe deferral by reset GPIO")
>> moved it from probe to driver init which installs the hook unconditionally
>> as soon as the driver is compiled into a kernel.
>> 
>> When this driver is compiled as a module, the hook is not registered
>> until after the driver has been matched with a .compatible and
>> loaded.
>> 
>> commit 415b6185c541 ("PCI: imx6: Fix config read timeout handling")
>> extended the fault handling code.
>> 
>> commit 2d8ed461dbc9 ("PCI: imx6: Add support for i.MX8MQ")
>> added some protection for non-ARM architectures, but this does not
>> protect non-i.MX ARM architectures.
> 
> Are *all* these commits relevant?

Yes, it was correct when introduced by commit bb38919ec56e for a goo reason.
And it was broken by bde4a5a00e76 an all attempts later made it worse.

>  Question also applies to Fixes:
> below.

It fixes all between bde4a5a00e76 and HEAD. Well, one can argue that
commit bde4a5a00e76 could be sufficient for Fixes:

I don't know if it is a problem because I have no overview over side-effects.

> 
>> Since fault handlers can be triggered on any architecture for different
>> reasons, there is no guarantee that they will be triggered only for the
>> assumed situation, leading to improper error handling (i.MX6-specific
>> imx6q_pcie_abort_handler) on foreign systems.
>> 
>> I had seen strange L3 imprecise external abort messages several times on
>> OMAP4 and OMAP5 devices and couldn't make sense of them until I realized
>> they were related to this unused imx6q driver because I had
>> CONFIG_PCI_IMX6=y.
> 
> Apparently imx6q_pcie_abort_handler() assumes it is always called
> because of a PCI abort?  If so, that sounds problematic.

> 
> If non-PCI imprecise aborts happen on OMAP4 and OMAP5 where imx6q is
> unused and imx6q_pcie_abort_handler() is not appropriate, I assume
> similar non-PCI aborts can also happen on systems where imx6q *is*
> used.

As far as I know the reasons why imprecise aborts occur may be SoC specific.

So I have no experience with i.MX6 to judge this. My goal is to shield other
architectures from this fault handler may it be correct or wrong.

> So imx6q_pcie_abort_handler() may be trying to fixup non-PCI aborts
> when it shouldn't?

Yes, at least if it is triggered on OMAP4/OMAP5 by accessing non-existing
registers in some subsystems (e.g. through devmem2).

> 
>> Note that CONFIG_PCI_IMX6=y is useful for kernel binaries that are designed
>> to run on different ARM SoC and be differentiated only by device tree
>> binaries. So turning off CONFIG_PCI_IMX6 is not a solution.
>> 
>> Therefore we check the compatible in the init function before registering
>> the fault handler.
>> 
>> Fixes: bde4a5a00e76 ("PCI: imx6: Allow probe deferral by reset GPIO")
>> Fixes: 415b6185c541 ("PCI: imx6: Fix config read timeout handling")
>> Fixes: 2d8ed461dbc9 ("PCI: imx6: Add support for i.MX8MQ")
>> 
>> Signed-off-by: H. Nikolaus Schaller <hns@xxxxxxxxxxxxx>
>> ---
>> drivers/pci/controller/dwc/pci-imx6.c | 9 +++++++++
>> 1 file changed, 9 insertions(+)
>> 
>> diff --git a/drivers/pci/controller/dwc/pci-imx6.c b/drivers/pci/controller/dwc/pci-imx6.c
>> index 1dde5c579edc8..89774aa187ae8 100644
>> --- a/drivers/pci/controller/dwc/pci-imx6.c
>> +++ b/drivers/pci/controller/dwc/pci-imx6.c
>> @@ -1402,6 +1402,15 @@ DECLARE_PCI_FIXUP_CLASS_HEADER(PCI_VENDOR_ID_SYNOPSYS, 0xabcd,
>> static int __init imx6_pcie_init(void)
>> {
>> #ifdef CONFIG_ARM
>> +	const struct of_device_id *reboot_id;
>> +	struct device_node *np;
>> +
>> +	np = of_find_matching_node_and_match(NULL, imx6_pcie_of_match,
>> +					     &reboot_id);
> 
> Since you don't need reboot_id, I think you should use
> of_find_matching_node() instead.

Well, I used it for debugging, but for production code it has indeed no benefit.

of_find_matching_node it is just a static inline wrapper for
of_find_matching_node_and_match with NULL parameter, but we can save one stack position.

I'll send a v2 soon.

> 
>> +	if (!np)
>> +		return -ENODEV;
>> +	of_node_put(np);
>> +
>> 	/*
>> 	 * Since probe() can be deferred we need to make sure that
>> 	 * hook_fault_code is not called after __init memory is freed
>> -- 
>> 2.38.1
>> 

BR and thanks,
Nikolaus