linux-kernel - Re: [PATCH] PCI/MSI: pci-xgene-msi: Enable MSI support in ACPI boot for X-Gene v1

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <56CDD767.6020103@arm.com>
Date:	Wed, 24 Feb 2016 16:16:39 +0000
From:	Marc Zyngier <marc.zyngier@....com>
To:	Mark Salter <msalter@...hat.com>, Duc Dang <dhdang@....com>,
	Bjorn Helgaas <bhelgaas@...gle.com>
Cc:	Feng Kan <fkan@....com>, linux-pci@...r.kernel.org,
	patches@....com, linux-kernel@...r.kernel.org,
	Loc Ho <lho@....com>, linux-arm-kernel@...ts.infradead.org,
	Tanmay Inamdar <tinamdar@....com>
Subject: Re: [PATCH] PCI/MSI: pci-xgene-msi: Enable MSI support in ACPI boot
 for X-Gene v1

On 24/02/16 16:09, Mark Salter wrote:
> On Tue, 2016-02-09 at 17:56 -0800, Duc Dang wrote:
>> This patch makes pci-xgene-msi driver ACPI-aware and provides
>> MSI capability for X-Gene v1 PCIe controllers in ACPI boot mode.
>>
>> Signed-off-by: Duc Dang <dhdang@....com>
>> ---
>>  drivers/pci/host/pci-xgene-msi.c | 35 ++++++++++++++++++++++++++++++++---
>>  1 file changed, 32 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/pci/host/pci-xgene-msi.c b/drivers/pci/host/pci-xgene-msi.c
>> index a6456b5..466aa93 100644
>> --- a/drivers/pci/host/pci-xgene-msi.c
>> +++ b/drivers/pci/host/pci-xgene-msi.c
>> @@ -24,6 +24,7 @@
>>  #include <linux/pci.h>
>>  #include <linux/platform_device.h>
>>  #include <linux/of_pci.h>
>> +#include <linux/acpi.h>
>>  
>>  #define MSI_IR0			0x000000
>>  #define MSI_INT0		0x800000
>> @@ -39,7 +40,7 @@ struct xgene_msi_group {
>>  };
>>  
>>  struct xgene_msi {
>> -	struct device_node	*node;
>> +	struct fwnode_handle	*fwnode;
>>  	struct irq_domain	*inner_domain;
>>  	struct irq_domain	*msi_domain;
>>  	u64			msi_addr;
>> @@ -249,6 +250,13 @@ static const struct irq_domain_ops msi_domain_ops = {
>>  	.free   = xgene_irq_domain_free,
>>  };
>>  
>> +#ifdef CONFIG_ACPI
>> +static struct fwnode_handle *xgene_msi_get_fwnode(struct device *dev)
>> +{
>> +	return xgene_msi_ctrl.fwnode;
>> +}
>> +#endif
>> +
>>  static int xgene_allocate_domains(struct xgene_msi *msi)
>>  {
>>  	msi->inner_domain = irq_domain_add_linear(NULL, NR_MSI_VEC,
>> @@ -256,7 +264,7 @@ static int xgene_allocate_domains(struct xgene_msi *msi)
>>  	if (!msi->inner_domain)
>>  		return -ENOMEM;
>>  
>> -	msi->msi_domain = pci_msi_create_irq_domain(of_node_to_fwnode(msi->node),
>> +	msi->msi_domain = pci_msi_create_irq_domain(msi->fwnode,
>>  						    &xgene_msi_domain_info,
>>  						    msi->inner_domain);
> 
> This doesn't work for me (ACPI probing on Mustang) unless I change this
> to be pci_msi_create_default_irq_domain(). The problem seems to be that
> the MSI probe happens after the PCIe RC is probed so there is no MSI domain
> at the time the PCIe root is initialized by ACPI.

pci_msi_create_default_irq_domain is the wrong thing do use, specially
if you have multiple MSI controllers in the system. I certainly wouldn't
want to see it being used on arm64.

This is the usual dependency hell. You try moving the probing earlier,
but that may break something else in the process.

	M.
-- 
Jazz is not dead. It just smells funny...