Skip to contentSkip to conversation

arm/pci: update interrupt-map only for legacy interrupts

PATCH Applied Sathyam Panda 3 participants 3 messages Nov 11, 2021 – Dec 14, 2021 KVM

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 arm/pci: update interrupt-map only for legacy interrupts

The interrupt pin cell in "interrupt-map" property is defined only for legacy interrupts with a valid range in [1-4] corrspoding to INTA#..INTD#. And the PCI endpoint devices that support advance interrupt mechanism like MSI or MSI-X should not have an entry with value 0 in "interrupt-map". This patch takes care of this problem by avoiding redundant entries.

---
 arm/pci.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
arm/pci.c +10 -0
index 2251f62..e44e453 100644
@@ -80,6 +80,16 @@ void pci__generate_fdt_nodes(void *fdt)
 		u8 irq = pci_hdr->irq_line;
 		u32 irq_flags = pci_hdr->irq_type;
 
+		/*
+		 * Avoid adding entries in "interrupt-map" for devices that
+		 * will be using advance interrupt mechanisms like MSI or
+		 * MSI-X instead of legacy interrupt pins INTA#..INTD#
+		 */
+		if (pin == 0) {
+			dev_hdr = device__next_dev(dev_hdr);
+			continue;
+		}
+
 		*entry = (struct of_interrupt_map_entry) {
 			.pci_irq_mask = {
 				.pci_addr = {
Signature · 1 line
2.25.1
Reviewed-by on v1 1/1
#1 lore raw

Hi Sathyam,

Quoted 21 lines · Sathyam Panda

The interrupt pin cell in "interrupt-map" property is defined only for legacy interrupts with a valid range in [1-4] corrspoding to INTA#..INTD#. And the PCI endpoint devices that support advance interrupt mechanism like MSI or MSI-X should not have an entry with value 0 in "interrupt-map". This patch takes care of this problem by avoiding redundant entries.

---
 arm/pci.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
arm/pci.c +0 -0
index 2251f62..e44e453 100644
@@ -80,6 +80,16 @@ void pci__generate_fdt_nodes(void *fdt)
 		u8 irq = pci_hdr->irq_line;
 		u32 irq_flags = pci_hdr->irq_type;

arm/pci.c void pci__generate_fdt_nodes(void *fdt) · line 91

Quoted 10 lines
arm/pci.c +9 -0
 
+		/*
+		 * Avoid adding entries in "interrupt-map" for devices that
+		 * will be using advance interrupt mechanisms like MSI or
+		 * MSI-X instead of legacy interrupt pins INTA#..INTD#
+		 */
+		if (pin == 0) {
+			dev_hdr = device__next_dev(dev_hdr);
+			continue;
+		}

I found this in "Open Firmware Recommended Practive: Interrupt Mapping" [1] (referenced by Documentation/devicetree/bindings/pci/pci.txt in the Linux source tree), in the PCI bus examples section (page 8):

"The PCI binding defines an "interrupts" property to consist of one cell, which encodes wheth- er the PCI device’s interrupt is connected to the PCI connector’s INTA#...INTD# pins, with values 1...4, respectively (assuming that the device is on a plug-in PCI card)".

The patch makes sense to me, if the interrupt pin is not present a the device, don't describe it in the DTB:

[1] https://www.devicetree.org/open-firmware/practice/imap/imap0_9d.pdf

Thanks,
Alex

Quoted 7 lines
arm/pci.c +1 -0
+
 		*entry = (struct of_interrupt_map_entry) {
 			.pci_irq_mask = {
 				.pci_addr = {
Signature · 1 line
2.25.1
Applied
#2 lore raw
In reply to Sathyam Panda
Quoted 9 lines · Sathyam Panda

The interrupt pin cell in "interrupt-map" property is defined only for legacy interrupts with a valid range in [1-4] corrspoding to INTA#..INTD#. And the PCI endpoint devices that support advance interrupt mechanism like MSI or MSI-X should not have an entry with value 0 in "interrupt-map". This patch takes care of this problem by avoiding redundant entries.

[...]

Applied to kvmtool (master), thanks!

[1/1] arm/pci: update interrupt-map only for legacy interrupts

      https://git.kernel.org/will/kvmtool/c/7a60af05c183

Cheers,

Signature · 5 lines
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev