Skip to contentSkip to conversation

S390: Fix libphy link failure

PATCH In review Jean Delvare 3 participants 5 messages May 6, 2013

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

PATCH 1/1 S390: Fix libphy link failure

On S390 basic IRQ functions are provided by the PCI subsystem, so code which calls these cannot be built if PCI support was not selected.

Cc "David S. Miller" <[email protected]>
Cc Heiko Carstens <[email protected]>

---
I know this isn't nice, but in the current state of S390 support this is all I can offer.

 drivers/net/phy/Kconfig |    1 +
 1 file changed, 1 insertion(+)
drivers/net/phy/Kconfig +1 -0
@@ -5,6 +5,7 @@
 menuconfig PHYLIB
 	tristate "PHY Device support and infrastructure"
 	depends on NETDEVICES
+	depends on !S390 || PCI
 	help
 	  Ethernet controllers are usually attached to PHY
 	  devices.  This option provides infrastructure for
Signature · 2 lines
Jean Delvare
Suse L3
maintainer
#1 lore raw

From: Jean Delvare <[email protected]>
Date: Mon, 06 May 2013 16:53:39 +0200

Quoted 10 lines

On S390 basic IRQ functions are provided by the PCI subsystem, so code which calls these cannot be built if PCI support was not selected.

Cc "David S. Miller" <[email protected]>
Cc Heiko Carstens <[email protected]>

---
I know this isn't nice, but in the current state of S390 support this is all I can offer.

I really am tired of all of these special S390 dependencies.

You can't say this is a new problem, because we've been hitting this issue repeatedly for more than a decade. And I don't want to hear any excuses, we've all already invested enough time into this to prove that catering to S390's weird build limitations is a never ending burdon for non-s390 people.

Really, our interfaces are abstract enough that every driver ought to be buildable on every architecture.

It's time to put the burdon on your side, if you don't support some IRQ interface or whatever, provide a stub that returns an error rather than being that "special architecture" that everything else has to specially cater for with ugly Kconfig dependencies.

Thanks.

#2 lore raw

Hi David,

Quoted 22 lines · David Miller

From: Jean Delvare <[email protected]>
Date: Mon, 06 May 2013 16:53:39 +0200

Quoted 10 lines

On S390 basic IRQ functions are provided by the PCI subsystem, so code which calls these cannot be built if PCI support was not selected.

Cc "David S. Miller" <[email protected]>
Cc Heiko Carstens <[email protected]>

---
I know this isn't nice, but in the current state of S390 support this is all I can offer.

I really am tired of all of these special S390 dependencies.

You can't say this is a new problem, because we've been hitting this issue repeatedly for more than a decade. And I don't want to hear any excuses, we've all already invested enough time into this to prove that catering to S390's weird build limitations is a never ending burdon for non-s390 people.

Quoted 7 lines

Really, our interfaces are abstract enough that every driver ought to be buildable on every architecture.

It's time to put the burdon on your side, if you don't support some IRQ interface or whatever, provide a stub that returns an error rather than being that "special architecture" that everything else has to specially cater for with ugly Kconfig dependencies.

Fair enough, I tend to agree with you. S390 isn't my thing in the first place. I was only trying to help fix a build regression (make allmodconfig fails on 32-bit S390.)

Signature · 2 lines
Jean Delvare
Suse L3
#3 lore raw
In reply to David Miller
Quoted 14 lines · David Miller

From: Jean Delvare <[email protected]>
Date: Mon, 06 May 2013 16:53:39 +0200

Quoted 10 lines

On S390 basic IRQ functions are provided by the PCI subsystem, so code which calls these cannot be built if PCI support was not selected.

Cc "David S. Miller" <[email protected]>
Cc Heiko Carstens <[email protected]>

---
I know this isn't nice, but in the current state of S390 support this is all I can offer.

Quoted 7 lines

I really am tired of all of these special S390 dependencies.

You can't say this is a new problem, because we've been hitting this issue repeatedly for more than a decade. And I don't want to hear any excuses, we've all already invested enough time into this to prove that catering to S390's weird build limitations is a never ending burdon for non-s390 people.

Agreed. And you won't hear any excuses since I'm tired of adding these Kconfig dependencies as well.
Unfortunately it got worse instead of better since we support CONFIG_PCI.

Quoted 7 lines

Really, our interfaces are abstract enough that every driver ought to be buildable on every architecture.

It's time to put the burdon on your side, if you don't support some IRQ interface or whatever, provide a stub that returns an error rather than being that "special architecture" that everything else has to specially cater for with ugly Kconfig dependencies.

Actually I'm in favor of supplying some stupid IRQ stubs as well. Unless our PCI guys think that we can go for GENERIC_HARDIRQS. But that will only work for 64 bit configs, not 31 bit.

So you wouldn't mind some stupid stubs like

request_irq(unsigned int irq, irq_handler_t handler, unsigned long flags,

            const char *name, void *dev) {
	    return -EWHATEVER;
}

?

maintainer
#4 lore raw

From: Heiko Carstens <[email protected]>
Date: Mon, 6 May 2013 17:49:31 +0200

Quoted 8 lines

So you wouldn't mind some stupid stubs like

request_irq(unsigned int irq, irq_handler_t handler, unsigned long flags,

            const char *name, void *dev) {
	    return -EWHATEVER;
}

?

It's your architecture, so I really don't care.

All I care is that we don't see S390 blobs any more in driver/net/ Kconfig's :-)