Skip to contentSkip to conversation

Which architectures need 64-bit resources?

DISCUSSION Matthew Wilcox 5 participants 8 messages Nov 6, 2006 – Nov 7, 2006

Subsystems (7)
lore
Share & tools
lore
#0 lore raw

I have a patch sitting in the parisc tree which does:

config RESOURCES_64BIT
- bool "64 bit Memory and IO resources (EXPERIMENTAL)" if (!64BIT && EXPER
IMENTAL)
+ bool "64 bit Memory and IO resources (EXPERIMENTAL)" if (!64BIT && !PARI
SC && EXPERIMENTAL)
default 64BIT

This is a bit selfish really, and it'll lead to something pretty ugly and unmaintainable ...

+ bool "64 bit Memory and IO resources (EXPERIMENTAL)" if (!64BIT && !PARISC && !M68K && !AVR && !CRIS && !FRV && !H8300 && !M32R && !S390 && !SUPERH && !SPARC && !V850 && !XTENSA && EXPERIMENTAL)

at least if my understanding of those arches needs is correct. Much better to have the dependencies be positive:

+ bool "64 bit Memory and IO resources (EXPERIMENTAL)" if (!64BIT && (ARM || X86 || MIPS || PPC) && EXPERIMENTAL)

Do I have that set of architectures right? Any others that need 64 bit resource types? Any there that shouldn't be there? Should we ditch the EXPERIMENTAL tag at the same time?

maintainer
#1 lore raw
Quoted 2 lines · Matthew Wilcox

+       bool "64 bit Memory and IO resources (EXPERIMENTAL)" if (!64BIT && (ARM || X86 || MIPS || PPC) && EXPERIMENTAL)

How about reversing the logic to make it

	bool "64 bit Memory and IO resources (EXPERIMENTAL)" if (RESOURCES_64BIT_ENABLE && EXPERIMENTAL)

Then each architecture could either have

config RESOURCES_64BIT_ENABLE
def_bool y if !64BIT

or even

select RESOURCES_64BIT

to switch it on unconditionally.

	Arnd <><
#2 lore raw
In reply to Matthew Wilcox
Quoted 1 line · Matthew Wilcox

+ bool "64 bit Memory and IO resources (EXPERIMENTAL)" if (!64BIT && (ARM || X86 || MIPS || PPC) && EXPERIMENTAL)

For MIPS at least it would be needed for a few platforms only, so I'd like to do "select RESOURCES_64BIT" as suitable for a platform. in the big machine selection menu in arch/mips/Kconfig.

Ralf

#3 lore raw
Quoted 7 lines · Ralf Baechle
Quoted 1 line · Matthew Wilcox

+ bool "64 bit Memory and IO resources (EXPERIMENTAL)" if (!64BIT && (ARM || X86 || MIPS || PPC) && EXPERIMENTAL)

For MIPS at least it would be needed for a few platforms only, so I'd like to do "select RESOURCES_64BIT" as suitable for a platform. in the big machine selection menu in arch/mips/Kconfig.

I don't see why you shouldn't do that anyway ... seems like ARM wants to do the same thing. In fact, I'd go so far as to say this isn't an option that users should ever be asked. You can tell by the help text for it:

help
This option allows memory and IO resources to be 64 bit.

How ... helpful.

Here's a proposed patch (which needs buy-in from the ARM, MIPS and PowerPC guys, plus anyone who knows about x86-32 machines with 64-bit IO). Grepping the arch/*/configs/* doesn't show anyone else using it by default.

I got the info for ARM from Vince Sanders, and made my best guess for MIPS & PPC. This patch is entirely untested.

arch/arm/Kconfig +1 -0
index adb05de..d73cf59 100644
@@ -241,6 +241,7 @@ config ARCH_IXP23XX
  	bool "IXP23XX-based"
 	depends on MMU
  	select PCI
+	select RESOURCES_64BIT
 	help
 	  Support for Intel's IXP23xx (XScale) family of processors.
 
arch/mips/Kconfig +1 -0
index 8a49884..f0205e0 100644
@@ -1556,6 +1556,7 @@ config SB1_PASS_2_1_WORKAROUNDS
 config 64BIT_PHYS_ADDR
 	bool "Support for 64-bit physical address space"
 	depends on (CPU_R4X00 || CPU_R5000 || CPU_RM7000 || CPU_RM9000 || CPU_R10000 || CPU_SB1 || CPU_MIPS32 || CPU_MIPS64) && 32BIT
+	select RESOURCES_64BIT
 
 config CPU_HAS_LLSC
 	bool
arch/powerpc/Kconfig +2 -1
index 8b69104..29ab1e0 100644
@@ -227,10 +227,11 @@ config PTE_64BIT
 config PHYS_64BIT
 	bool 'Large physical address support' if E500
 	depends on 44x || E500
+	select RESOURCES_64BIT
 	default y if 44x
 	---help---
 	  This option enables kernel support for larger than 32-bit physical
-	  addresses.  This features is not be available on all e500 cores.
+	  addresses.  This features is not available on all e500 cores.
 
 	  If in doubt, say N here.
 
lib/Kconfig +6 -1
index 734ce95..613942d 100644
@@ -92,9 +92,14 @@ config TEXTSEARCH_FSM
 	tristate
 
 #
-# plist support is select#ed if needed
+# plist support is selected if needed
 #
 config PLIST
 	boolean
 
+# 64-bit resources are selected by individual architectures
+config RESOURCES_64BIT
+	bool
+	default 64BIT
+
 endmenu
mm/Kconfig +0 -6
index db7c55d..089513f 100644
@@ -151,9 +151,3 @@ config MIGRATION
 	  while the virtual addresses are not changed. This is useful for
 	  example on NUMA systems to put pages nearer to the processors accessing
 	  the page.
-
-config RESOURCES_64BIT
-	bool "64 bit Memory and IO resources (EXPERIMENTAL)" if (!64BIT && EXPERIMENTAL)
-	default 64BIT
-	help
-	  This option allows memory and IO resources to be 64 bit.
#4 lore raw
Quoted 7 lines · Matthew Wilcox

Here's a proposed patch (which needs buy-in from the ARM, MIPS and PowerPC guys, plus anyone who knows about x86-32 machines with 64-bit IO). Grepping the arch/*/configs/* doesn't show anyone else using it by default.

I got the info for ARM from Vince Sanders, and made my best guess for MIPS & PPC. This patch is entirely untested.

The MIPS segment looks good, so feel free to add an Acked-by me.

Ralf

#5 lore raw
In reply to Matthew Wilcox
Quoted 5 lines · Matthew Wilcox
Quoted 4 lines · Ralf Baechle
Quoted 1 line · Matthew Wilcox

+ bool "64 bit Memory and IO resources (EXPERIMENTAL)" if (!64BIT && (ARM || X86 || MIPS || PPC) && EXPERIMENTAL)

Quoted 7 lines
Quoted 3 lines

For MIPS at least it would be needed for a few platforms only, so I'd like to do "select RESOURCES_64BIT" as suitable for a platform. in the big machine selection menu in arch/mips/Kconfig.

I don't see why you shouldn't do that anyway ... seems like ARM wants to do the same thing. In fact, I'd go so far as to say this isn't an option that users should ever be asked. You can tell by the help text for it:

Ah, good that the possibility' of 64-bit resources on m68k goes away, and I haven't fixed' all those printf() format warnings yet...

Speaking about that: how many recently-added silencing casts should be removed on platforms without 64-bit resources?

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

#6 lore raw
Quoted 5 lines · Geert Uytterhoeven

Ah, good that the possibility' of 64-bit resources on m68k goes away, and I haven't fixed' all those printf() format warnings yet...

Speaking about that: how many recently-added silencing casts should be removed on platforms without 64-bit resources?

Absolutely none, I'm afraid.

struct resource {
        resource_size_t start;
	resource_size_t end;

...

GCC complains about resource_size_t, no matter what you typedef it to be. It'd be nice if we could tell GCC that %r meant it'd consume a resource_size_t, but afaict, there's no ability to define additions to the format attribute.

Viro's mused on disabling the printk checking entirely with gcc and just using sparse to verify format strings (so we can add our own types, etc), but I don't think that's realistic unless a lot more people start installing sparse.

And widespread adoption of sparse seems unlikely, given that it's under OSL 1.1 which means it can't go into Debian [1].

GCC's in Stage 1 ... anyone want to write a patch to c-format.c?

[1] http://wiki.debian.org/DFSGLicenses#head-2b7e4b3507dcc5657dabaf145df20d6c7d41f159

Acked-by
#7 lore raw
In reply to Matthew Wilcox
Quoted 5 lines · Matthew Wilcox
Quoted 4 lines · Ralf Baechle
Quoted 1 line · Matthew Wilcox

+ bool "64 bit Memory and IO resources (EXPERIMENTAL)" if (!64BIT && (ARM || X86 || MIPS || PPC) && EXPERIMENTAL)

Quoted 7 lines
Quoted 3 lines

For MIPS at least it would be needed for a few platforms only, so I'd like to do "select RESOURCES_64BIT" as suitable for a platform. in the big machine selection menu in arch/mips/Kconfig.

I don't see why you shouldn't do that anyway ... seems like ARM wants to do the same thing. In fact, I'd go so far as to say this isn't an option that users should ever be asked. You can tell by the help text for it:

For the ARM bits...

Thanks.

Signature · 3 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core