[PATCH v2 05/15] PCI: xilinx: Convert to MSI domains

Bharat Kumar Gogada bharatku at xilinx.com
Wed Mar 24 13:56:16 GMT 2021


> > Hi Marc,
> >
> > Thanks for the patch.
> >
> > > Subject: [PATCH v2 05/15] PCI: xilinx: Convert to MSI domains
> > >
> > > In anticipation of the removal of the msi_controller structure,
> > > convert the ancient xilinx host controller driver to MSI domains.
> > >
> > > We end-up with the usual two domain structure, the top one being a
> > > generic PCI/MSI domain, the bottom one being xilinx-specific and
> > > handling the actual HW interrupt allocation.
> > >
> > > This allows us to fix some of the most appalling MSI programming,
> > > where the message programmed in the device is the virtual IRQ number
> > > instead of the allocated vector number. The allocator is also made
> > > safe with a mutex. This should allow support for MultiMSI, but I
> > > decided not to even try, since I cannot test it.
> > >
> > > Acked-by: Bjorn Helgaas <bhelgaas at google.com>
> > > Signed-off-by: Marc Zyngier <maz at kernel.org>
> > > ---
> > >  drivers/pci/controller/Kconfig       |   2 +-
> > >  drivers/pci/controller/pcie-xilinx.c | 234
> > > +++++++++++----------------
> > >  2 files changed, 97 insertions(+), 139 deletions(-)
> > >
> > > diff --git a/drivers/pci/controller/Kconfig
> > > b/drivers/pci/controller/Kconfig index 5cc07d28a3a0..60045f7aafc5
> > > 100644
> > ...
> >
> >
> > > +static struct irq_chip xilinx_msi_bottom_chip = {
> > > +	.name			= "Xilinx MSI",
> > > +	.irq_set_affinity 	= xilinx_msi_set_affinity,
> > > +	.irq_compose_msi_msg	= xilinx_compose_msi_msg,
> > > +};
> > >
> > I see a crash while testing MSI in handle_edge_irq [<c015bdd4>]
> > (handle_edge_irq) from [<c0157164>] (generic_handle_irq+0x28/0x38)
> > [<c0157164>] (generic_handle_irq) from [<c03a9714>]
> > (xilinx_pcie_intr_handler+0x17c/0x2b0)
> > [<c03a9714>] (xilinx_pcie_intr_handler) from [<c0157d94>]
> > (__handle_irq_event_percpu+0x3c/0xc0)
> > [<c0157d94>] (__handle_irq_event_percpu) from [<c0157e44>]
> > (handle_irq_event_percpu+0x2c/0x7c)
> > [<c0157e44>] (handle_irq_event_percpu) from [<c0157ecc>]
> > (handle_irq_event+0x38/0x5c) [<c0157ecc>] (handle_irq_event) from
> > [<c015bc8c>] (handle_fasteoi_irq+0x9c/0x114)
> 
> Thanks for that. Can you please try the following patch and let me know if it
> helps?
> 
> Thanks,
> 
> 	M.
> 
> diff --git a/drivers/pci/controller/pcie-xilinx.c b/drivers/pci/controller/pcie-
> xilinx.c
> index ad9abf405167..14001febf59a 100644
> --- a/drivers/pci/controller/pcie-xilinx.c
> +++ b/drivers/pci/controller/pcie-xilinx.c
> @@ -194,8 +194,18 @@ static struct pci_ops xilinx_pcie_ops = {
> 
>  /* MSI functions */
> 
> +static void xilinx_msi_top_irq_ack(struct irq_data *d) {
> +	/*
> +	 * xilinx_pcie_intr_handler() will have performed the Ack.
> +	 * Eventually, this should be fixed and the Ack be moved in
> +	 * the respective callbacks for INTx and MSI.
> +	 */
> +}
> +
>  static struct irq_chip xilinx_msi_top_chip = {
>  	.name		= "PCIe MSI",
> +	.irq_ack	= xilinx_msi_top_irq_ack,
>  };
> 
>  static int xilinx_msi_set_affinity(struct irq_data *d, const struct cpumask
> *mask, bool force) @@ -206,7 +216,7 @@ static int
> xilinx_msi_set_affinity(struct irq_data *d, const struct cpumask *mas  static
> void xilinx_compose_msi_msg(struct irq_data *data, struct msi_msg *msg)  {
>  	struct xilinx_pcie_port *pcie = irq_data_get_irq_chip_data(data);
> -	phys_addr_t pa = virt_to_phys(pcie);
> +	phys_addr_t pa = ALIGN_DOWN(virt_to_phys(pcie), SZ_4K);
> 
>  	msg->address_lo = lower_32_bits(pa);
>  	msg->address_hi = upper_32_bits(pa);
> @@ -468,7 +478,7 @@ static int xilinx_pcie_init_irq_domain(struct
> xilinx_pcie_port *port)
> 
>  	/* Setup MSI */
>  	if (IS_ENABLED(CONFIG_PCI_MSI)) {
> -		phys_addr_t pa = virt_to_phys(port);
> +		phys_addr_t pa = ALIGN_DOWN(virt_to_phys(port), SZ_4K);
> 
>  		ret = xilinx_allocate_msi_domains(port);
>  		if (ret)
> 
Thanks Marc.
With above patch now everything works fine, tested a Samsung NVMe SSD. 
tst~# lspci
00:00.0 PCI bridge: Xilinx Corporation Device 0706
01:00.0 Non-Volatile memory controller: Samsung Electronics Co Ltd NVMe SSD Controller 172Xa/172Xb (rev 01)

Regards,
Bharat



More information about the Linux-mediatek mailing list