Skip to contentSkip to conversation

x86/pci: don't set acpi stuff if !CONFIG_ACPI

PATCH In review Adam Borowski 3 participants 6 messages Aug 20, 2020 – Aug 21, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86/pci: don't set acpi stuff if !CONFIG_ACPI

Not that x86 without ACPI sees any real use...

---
Found by randconfig builds.

 arch/x86/pci/intel_mid_pci.c | 2 ++
 arch/x86/pci/xen.c           | 2 ++
 2 files changed, 4 insertions(+)
arch/x86/pci/intel_mid_pci.c +2 -0
index 00c62115f39c..f14a911f0d06 100644
@@ -299,8 +299,10 @@ int __init intel_mid_pci_init(void)
 	pcibios_disable_irq = intel_mid_pci_irq_disable;
 	pci_root_ops = intel_mid_pci_ops;
 	pci_soc_mode = 1;
+#ifdef CONFIG_ACPI
 	/* Continue with standard init */
 	acpi_noirq_set();
+#endif
 	return 1;
 }
 
arch/x86/pci/xen.c +2 -0
index 9f9aad42ccff..681eb5c34c03 100644
@@ -406,8 +406,10 @@ int __init pci_xen_init(void)
 	pcibios_enable_irq = xen_pcifront_enable_irq;
 	pcibios_disable_irq = NULL;
 
+#ifdef CONFIG_ACPI
 	/* Keep ACPI out of the picture */
 	acpi_noirq_set();
+#endif
 
 #ifdef CONFIG_PCI_MSI
 	x86_msi.setup_msi_irqs = xen_setup_msi_irqs;
Signature · 1 line
2.28.0
#1 lore raw
Quoted 5 lines · Adam Borowski

Not that x86 without ACPI sees any real use...

---
Found by randconfig builds.

Note that
#include <asm/acpi.h>
has a stub for acpi_noirq_set() when ACPI is not set/enabled. That would be better. And I have submitted that for arch/x86/pci/xen.c -- and a different patch for intel_mid_pci.c

But I didn't submit them to the X86 maintainers because the MAINTAINERS file pointed me to the PCI maintainer and to the XEN PCI maintainer....

Quoted 36 lines
 arch/x86/pci/intel_mid_pci.c | 2 ++
 arch/x86/pci/xen.c           | 2 ++
 2 files changed, 4 insertions(+)
arch/x86/pci/intel_mid_pci.c +2 -0
index 00c62115f39c..f14a911f0d06 100644
@@ -299,8 +299,10 @@ int __init intel_mid_pci_init(void)
 	pcibios_disable_irq = intel_mid_pci_irq_disable;
 	pci_root_ops = intel_mid_pci_ops;
 	pci_soc_mode = 1;
+#ifdef CONFIG_ACPI
 	/* Continue with standard init */
 	acpi_noirq_set();
+#endif
 	return 1;
 }
 
arch/x86/pci/xen.c +2 -0
index 9f9aad42ccff..681eb5c34c03 100644
@@ -406,8 +406,10 @@ int __init pci_xen_init(void)
 	pcibios_enable_irq = xen_pcifront_enable_irq;
 	pcibios_disable_irq = NULL;
 
+#ifdef CONFIG_ACPI
 	/* Keep ACPI out of the picture */
 	acpi_noirq_set();
+#endif
 
 #ifdef CONFIG_PCI_MSI
 	x86_msi.setup_msi_irqs = xen_setup_msi_irqs;
Signature · 2 lines
~Randy
Reported-by: Randy Dunlap <[email protected]>
#2 lore raw
In reply to Adam Borowski
Quoted 15 lines · Adam Borowski

Not that x86 without ACPI sees any real use...

---
Found by randconfig builds.

 arch/x86/pci/intel_mid_pci.c | 2 ++
 arch/x86/pci/xen.c           | 2 ++
 2 files changed, 4 insertions(+)
arch/x86/pci/intel_mid_pci.c +0 -0
index 00c62115f39c..f14a911f0d06 100644

arch/x86/pci/intel_mid_pci.c int __init intel_mid_pci_init(void) · line 305

Quoted 7 lines
arch/x86/pci/intel_mid_pci.c +2 -0
@@ -299,8 +299,10 @@ int __init intel_mid_pci_init(void)
 	pcibios_disable_irq = intel_mid_pci_irq_disable;
 	pci_root_ops = intel_mid_pci_ops;
 	pci_soc_mode = 1;
+#ifdef CONFIG_ACPI
 	/* Continue with standard init */
 	acpi_noirq_set();
+#endif

If CONFIG_ACPI=n then acpi_noirq_set() is an empty stub inline. So I'm not sure what you are trying to solve here.

Ah, I see with CONFIG_ACPI=n linux/acpi.h does not include asm/acpi.h so the stubs are unreachable. So that needs to be fixed and not papered over with #ifdeffery

Thanks,

tglx

#3 lore raw
Quoted 17 lines · Thomas Gleixner

arch/x86/pci/intel_mid_pci.c int __init intel_mid_pci_init(void) · line 302

Quoted 12 lines · Adam Borowski

Found by randconfig builds.

 arch/x86/pci/intel_mid_pci.c | 2 ++
 arch/x86/pci/xen.c           | 2 ++
arch/x86/pci/intel_mid_pci.c +2 -0
@@ -299,8 +299,10 @@ int __init intel_mid_pci_init(void)
+#ifdef CONFIG_ACPI
 	/* Continue with standard init */
 	acpi_noirq_set();
+#endif

If CONFIG_ACPI=n then acpi_noirq_set() is an empty stub inline. So I'm not sure what you are trying to solve here.

Quoted 3 lines

Ah, I see with CONFIG_ACPI=n linux/acpi.h does not include asm/acpi.h so the stubs are unreachable. So that needs to be fixed and not papered over with #ifdeffery

If I understand Randy Dunlap correctly, he already sent a pair of patches that do what you want.

Meow.

Signature · 4 lines
⢀⣴⠾⠻⢶⣦⠀
⣾⠁⢠⠒⠀⣿⡁
⢿⡄⠘⠷⠚⠋⠀ It's time to migrate your Imaginary Protocol from version 4i to 6i.
⠈⠳⣄⠀⠀⠀⠀
#4 lore raw
Quoted 18 lines · Adam Borowski
Quoted 17 lines · Thomas Gleixner

arch/x86/pci/intel_mid_pci.c int __init intel_mid_pci_init(void) · line 302

Quoted 12 lines · Adam Borowski

Found by randconfig builds.

 arch/x86/pci/intel_mid_pci.c | 2 ++
 arch/x86/pci/xen.c           | 2 ++
arch/x86/pci/intel_mid_pci.c +2 -0
@@ -299,8 +299,10 @@ int __init intel_mid_pci_init(void)
+#ifdef CONFIG_ACPI
 	/* Continue with standard init */
 	acpi_noirq_set();
+#endif

If CONFIG_ACPI=n then acpi_noirq_set() is an empty stub inline. So I'm not sure what you are trying to solve here.

Quoted 6 lines
Quoted 3 lines

Ah, I see with CONFIG_ACPI=n linux/acpi.h does not include asm/acpi.h so the stubs are unreachable. So that needs to be fixed and not papered over with #ifdeffery

If I understand Randy Dunlap correctly, he already sent a pair of patches that do what you want.

I did, but I sent them to the Xen and PCI maintainers, not the x86 maintainers, but I will happily resend this patch. The Xen patch has already been applied whereas the patch to intel_mid_pci.c is in limbo. :(

Thomas, do you want me to send it to you/X86 people? (with 2 Reviewed-by: additions)

thanks.

Signature · 1 line
~Randy
#5 lore raw
Quoted 3 lines · Randy Dunlap
Quoted 2 lines · Adam Borowski

If I understand Randy Dunlap correctly, he already sent a pair of patches that do what you want.

I replied before reading Randy's reply. Old habit of reading stuff from top and not getting biased by other peoples replies before doing so. Is most of the time the correct approach, but sometimes it would be better to do it the other way round :)

Quoted 7 lines

I did, but I sent them to the Xen and PCI maintainers, not the x86 maintainers, but I will happily resend this patch. The Xen patch has already been applied whereas the patch to intel_mid_pci.c is in limbo. :(

Thomas, do you want me to send it to you/X86 people? (with 2 Reviewed-by: additions)

Sure, but usually Bjorn handles the x86/pci/ stuff.

As I trust you, here is a blind

just in case.

Thanks,

tglx