[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20170221175932.508c628a@free-electrons.com>
Date: Tue, 21 Feb 2017 17:59:32 +0100
From: Thomas Petazzoni <thomas.petazzoni@...e-electrons.com>
To: Arnd Bergmann <arnd@...db.de>
Cc: Thomas Gleixner <tglx@...utronix.de>,
Jason Cooper <jason@...edaemon.net>,
Marc Zyngier <marc.zyngier@....com>,
linux-kernel@...r.kernel.org,
Grégory Clement
<gregory.clement@...e-electrons.com>,
Jason Cooper <jason@...edaemon.net>,
Sebastian Hesselbarth <sebastian.hesselbarth@...il.com>,
Andrew Lunn <andrew@...n.ch>
Subject: Re: [PATCH] irqchip: mvebu-odmi: select GENERIC_MSI_IRQ_DOMAIN
gHello,
On Tue, 14 Feb 2017 22:24:36 +0100, Arnd Bergmann wrote:
> This driver uses the MSI domain but has no strict dependency on PCI_MSI, so we
> may run into a build failure when CONFIG_GENERIC_MSI_IRQ_DOMAIN is disabled:
>
> drivers/irqchip/irq-mvebu-odmi.c:152:15: error: variable 'odmi_msi_ops' has initializer but incomplete type
> static struct msi_domain_ops odmi_msi_ops = {
> ^~~~~~~~~~~~~~
> drivers/irqchip/irq-mvebu-odmi.c:155:15: error: variable 'odmi_msi_domain_info' has initializer but incomplete type
> static struct msi_domain_info odmi_msi_domain_info = {
> ^~~~~~~~~~~~~~~
> drivers/irqchip/irq-mvebu-odmi.c:156:3: error: 'struct msi_domain_info' has no member named 'flags'
> .flags = (MSI_FLAG_USE_DEF_DOM_OPS | MSI_FLAG_USE_DEF_CHIP_OPS),
> ^~~~~
> drivers/irqchip/irq-mvebu-odmi.c:156:12: error: 'MSI_FLAG_USE_DEF_DOM_OPS' undeclared here (not in a function)
> .flags = (MSI_FLAG_USE_DEF_DOM_OPS | MSI_FLAG_USE_DEF_CHIP_OPS),
> ^~~~~~~~~~~~~~~~~~~~~~~~
> drivers/irqchip/irq-mvebu-odmi.c:156:39: error: 'MSI_FLAG_USE_DEF_CHIP_OPS' undeclared here (not in a function); did you mean 'MSI_FLAG_USE_DEF_DOM_OPS'?
>
> Selecting the option from this driver seems to solve this nicely, though I could
> not find any other instance of this in irqchip drivers.
Here is my explanation of why MVEBU_ODMI is the only one affected:
because it's the only one that provides only platform MSI and not PCI
MSI.
Let's look at the other irqchip drivers using msi_domain_ops:
irq-gic-v2m.c:static struct msi_domain_ops gicv2m_pmsi_ops = {
irq-gic-v3-its-pci-msi.c:static struct msi_domain_ops its_pci_msi_ops = {
irq-gic-v3-its-platform-msi.c:static struct msi_domain_ops its_pmsi_ops = {
irq-gic-v3-its.c:static struct msi_domain_ops its_msi_domain_ops = {
irq-mvebu-odmi.c:static struct msi_domain_ops odmi_msi_ops = {
So, besides mvebu-odmi, we've got GICv2m and GIv3 ITS. Their Kconfig
declaration is:
config ARM_GIC_V2M
bool
depends on PCI
select ARM_GIC
select PCI_MSI
config ARM_GIC_V3_ITS
bool
depends on PCI
depends on PCI_MSI
select ACPI_IORT if ACPI
so both of them "select PCI_MSI". Then we have:
config PCI_MSI_IRQ_DOMAIN
def_bool ARC || ARM || ARM64 || X86
depends on PCI_MSI
select GENERIC_MSI_IRQ_DOMAIN
So, when we are on ARC, ARM, ARM64 or x86, and PCI_MSI is enabled,
GENERIC_MSI_IRQ_DOMAIN is automatically enabled. So basically the fact
that GICv2m and GICv3 "select PCI_MSI" ensure that
"GENERIC_MSI_IRQ_DOMAIN" will be enabled.
Therefore:
> Signed-off-by: Arnd Bergmann <arnd@...db.de>
Acked-by: Thomas Petazzoni <thomas.petazzoni@...e-electrons.com>
Thanks!
Thomas
--
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com
Powered by blists - more mailing lists