* [PATCH] intel-x86 model config cleanup
@ 2002-05-29 14:35 J.A. Magallon
2002-05-29 20:37 ` Jeff Garzik
0 siblings, 1 reply; 12+ messages in thread
From: J.A. Magallon @ 2002-05-29 14:35 UTC (permalink / raw)
To: Lista Linux-Kernel
Hi all..
One other try.
This is an attempt to clean up the CPU model flags. Changes:
- change names for CONFIG_Mxxxx, trying to make them more intuitive
- split PII from PPro
- introduce X86_F00F config flag and fixmap fix.
- kill CONFIG_M586 as independent flag, and make it just an
extra flag for 586 (I couldnt get a better name for
MGEN586, suggestions wellcome...)
- kill CONFIG_M686 as independent flag, and make it just an
extra flag for anything >= PPro.
- change option order definition to avoid duplicates, like:
if [ "$CONFIG_MPENTIUM" = "y" ]; then
define_bool CONFIG_M586 y <==============
define_bool CONFIG_X86_TSC y
fi
if [ "$CONFIG_MPENTIUMMMX" = "y" ]; then
define_bool CONFIG_M586 y <==============
define_bool CONFIG_X86_TSC y
define_bool CONFIG_X86_GOOD_APIC y
fi
if [ "$CONFIG_M586" = "y" ]; then <======== common things here
define_int CONFIG_X86_L1_CACHE_SHIFT 5
define_bool CONFIG_X86_USE_STRING_486 y
define_bool CONFIG_X86_ALIGNMENT_16 y
define_bool CONFIG_X86_PPRO_FENCE y
define_bool CONFIG_X86_F00F_BUG y
fi
Patch follows:
diff -ruN linux-2.4.19-pre9-jam1/Documentation/Configure.help linux-2.4.19-pre9-jam1-arch/Documentation/Configure.help
--- linux-2.4.19-pre9-jam1/Documentation/Configure.help Wed May 29 11:50:43 2002
+++ linux-2.4.19-pre9-jam1-arch/Documentation/Configure.help Wed May 29 11:52:18 2002
@@ -3943,16 +3943,17 @@
a PPro, but not necessarily on a i486.
Here are the settings recommended for greatest speed:
- - "386" for the AMD/Cyrix/Intel 386DX/DXL/SL/SLC/SX, Cyrix/TI
+ - "Generic-386" for the AMD/Cyrix/Intel 386DX/DXL/SL/SLC/SX, Cyrix/TI
486DLC/DLC2, UMC 486SX-S and NexGen Nx586. Only "386" kernels
will run on a 386 class machine.
- - "486" for the AMD/Cyrix/IBM/Intel 486DX/DX2/DX4 or
+ - "Generic-486" for the AMD/Cyrix/IBM/Intel 486DX/DX2/DX4 or
SL/SLC/SLC2/SLC3/SX/SX2 and UMC U5D or U5S.
- - "586" for generic Pentium CPUs, possibly lacking the TSC
+ - "Generic-586" for generic Pentium CPUs, possibly lacking the TSC
(time stamp counter) register.
- - "Pentium-Classic" for the Intel Pentium.
+ - "Pentium" for the Intel Pentium.
- "Pentium-MMX" for the Intel Pentium MMX.
- - "Pentium-Pro" for the Intel Pentium Pro/Celeron/Pentium II.
+ - "Pentium-Pro" for the Intel Pentium Pro.
+ - "Pentium-II" for the Intel Pentium II / Celeron.
- "Pentium-III" for the Intel Pentium III
and Celerons based on the Coppermine core.
- "Pentium-4" for the Intel Pentium 4.
diff -ruN linux-2.4.19-pre9-jam1/arch/i386/Makefile linux-2.4.19-pre9-jam1-arch/arch/i386/Makefile
--- linux-2.4.19-pre9-jam1/arch/i386/Makefile Wed May 29 11:50:43 2002
+++ linux-2.4.19-pre9-jam1-arch/arch/i386/Makefile Wed May 29 11:55:39 2002
@@ -34,23 +34,27 @@
CFLAGS += -march=i486
endif
-ifdef CONFIG_M586
+ifdef CONFIG_MGEN586
CFLAGS += -march=i586
endif
-ifdef CONFIG_M586TSC
+ifdef CONFIG_MPENTIUM
CFLAGS += -march=i586
endif
-ifdef CONFIG_M586MMX
+ifdef CONFIG_MPENTIUMMMX
CFLAGS += -march=i586
endif
-ifdef CONFIG_M686
+ifdef CONFIG_MPENTIUMPRO
CFLAGS += -march=i686
endif
-ifdef CONFIG_MPENTIUMIII
+ifdef CONFIG_MPENTIUM2
+CFLAGS += -march=i686
+endif
+
+ifdef CONFIG_MPENTIUM3
CFLAGS += -march=i686
endif
diff -ruN linux-2.4.19-pre9-jam1/arch/i386/config.in linux-2.4.19-pre9-jam1-arch/arch/i386/config.in
--- linux-2.4.19-pre9-jam1/arch/i386/config.in Wed May 29 11:50:43 2002
+++ linux-2.4.19-pre9-jam1-arch/arch/i386/config.in Wed May 29 11:54:25 2002
@@ -27,22 +27,23 @@
mainmenu_option next_comment
comment 'Processor type and features'
choice 'Processor family' \
- "386 CONFIG_M386 \
- 486 CONFIG_M486 \
- 586/K5/5x86/6x86/6x86MX CONFIG_M586 \
- Pentium-Classic CONFIG_M586TSC \
- Pentium-MMX CONFIG_M586MMX \
- Pentium-Pro/Celeron/Pentium-II CONFIG_M686 \
- Pentium-III/Celeron(Coppermine) CONFIG_MPENTIUMIII \
- Pentium-4 CONFIG_MPENTIUM4 \
- K6/K6-II/K6-III CONFIG_MK6 \
- Athlon/Duron/K7 CONFIG_MK7 \
- Elan CONFIG_MELAN \
- Crusoe CONFIG_MCRUSOE \
- Winchip-C6 CONFIG_MWINCHIPC6 \
- Winchip-2 CONFIG_MWINCHIP2 \
- Winchip-2A/Winchip-3 CONFIG_MWINCHIP3D \
- CyrixIII/VIA-C3/VIA-C5 CONFIG_MCYRIXIII" Pentium-Pro
+ "Generic-386 CONFIG_M386 \
+ Generic-486 CONFIG_M486 \
+ Generic-586 CONFIG_MGEN586 \
+ Pentium CONFIG_MPENTIUM \
+ Pentium-MMX CONFIG_MPENTIUMMMX \
+ Pentium-Pro CONFIG_MPENTIUMPRO \
+ Pentium-II/Celeron CONFIG_MPENTIUM2 \
+ Pentium-III/Celeron(Coppermine) CONFIG_MPENTIUM3 \
+ Pentium-4 CONFIG_MPENTIUM4 \
+ K6/K6-II/K6-III CONFIG_MK6 \
+ Athlon/Duron/K7 CONFIG_MK7 \
+ Elan CONFIG_MELAN \
+ Crusoe CONFIG_MCRUSOE \
+ Winchip-C6 CONFIG_MWINCHIPC6 \
+ Winchip-2 CONFIG_MWINCHIP2 \
+ Winchip-2A/Winchip-3 CONFIG_MWINCHIP3D \
+ CyrixIII/VIA-C3/VIA-C5 CONFIG_MCYRIXIII" Pentium-III
#
# Define implied options from the CPU selection here
#
@@ -54,6 +55,7 @@
define_bool CONFIG_RWSEM_GENERIC_SPINLOCK y
define_bool CONFIG_RWSEM_XCHGADD_ALGORITHM n
define_bool CONFIG_X86_PPRO_FENCE y
+ define_bool CONFIG_X86_F00F_BUG y
else
define_bool CONFIG_X86_WP_WORKS_OK y
define_bool CONFIG_X86_INVLPG y
@@ -64,55 +66,59 @@
define_bool CONFIG_RWSEM_GENERIC_SPINLOCK n
define_bool CONFIG_RWSEM_XCHGADD_ALGORITHM y
fi
+
if [ "$CONFIG_M486" = "y" ]; then
define_int CONFIG_X86_L1_CACHE_SHIFT 4
define_bool CONFIG_X86_USE_STRING_486 y
define_bool CONFIG_X86_ALIGNMENT_16 y
define_bool CONFIG_X86_PPRO_FENCE y
+ define_bool CONFIG_X86_F00F_BUG y
fi
-if [ "$CONFIG_M586" = "y" ]; then
- define_int CONFIG_X86_L1_CACHE_SHIFT 5
- define_bool CONFIG_X86_USE_STRING_486 y
- define_bool CONFIG_X86_ALIGNMENT_16 y
- define_bool CONFIG_X86_PPRO_FENCE y
+
+if [ "$CONFIG_MGEN586" = "y" ]; then
+ define_bool CONFIG_M586 y
fi
-if [ "$CONFIG_M586TSC" = "y" ]; then
- define_int CONFIG_X86_L1_CACHE_SHIFT 5
- define_bool CONFIG_X86_USE_STRING_486 y
- define_bool CONFIG_X86_ALIGNMENT_16 y
+if [ "$CONFIG_MPENTIUM" = "y" ]; then
+ define_bool CONFIG_M586 y
define_bool CONFIG_X86_TSC y
- define_bool CONFIG_X86_PPRO_FENCE y
fi
-if [ "$CONFIG_M586MMX" = "y" ]; then
+if [ "$CONFIG_MPENTIUMMMX" = "y" ]; then
+ define_bool CONFIG_M586 y
+ define_bool CONFIG_X86_TSC y
+ define_bool CONFIG_X86_GOOD_APIC y
+fi
+if [ "$CONFIG_M586" = "y" ]; then
define_int CONFIG_X86_L1_CACHE_SHIFT 5
define_bool CONFIG_X86_USE_STRING_486 y
define_bool CONFIG_X86_ALIGNMENT_16 y
- define_bool CONFIG_X86_TSC y
- define_bool CONFIG_X86_GOOD_APIC y
define_bool CONFIG_X86_PPRO_FENCE y
+ define_bool CONFIG_X86_F00F_BUG y
fi
-if [ "$CONFIG_M686" = "y" ]; then
+
+if [ "$CONFIG_MPENTIUMPRO" = "y" ]; then
+ define_bool CONFIG_M686 y
define_int CONFIG_X86_L1_CACHE_SHIFT 5
- define_bool CONFIG_X86_TSC y
- define_bool CONFIG_X86_GOOD_APIC y
- define_bool CONFIG_X86_PGE y
- define_bool CONFIG_X86_USE_PPRO_CHECKSUM y
define_bool CONFIG_X86_PPRO_FENCE y
fi
-if [ "$CONFIG_MPENTIUMIII" = "y" ]; then
+if [ "$CONFIG_MPENTIUM2" = "y" ]; then
+ define_bool CONFIG_M686 y
+ define_int CONFIG_X86_L1_CACHE_SHIFT 5
+fi
+if [ "$CONFIG_MPENTIUM3" = "y" ]; then
+ define_bool CONFIG_M686 y
define_int CONFIG_X86_L1_CACHE_SHIFT 5
- define_bool CONFIG_X86_TSC y
- define_bool CONFIG_X86_GOOD_APIC y
- define_bool CONFIG_X86_PGE y
- define_bool CONFIG_X86_USE_PPRO_CHECKSUM y
fi
if [ "$CONFIG_MPENTIUM4" = "y" ]; then
+ define_bool CONFIG_M686 y
define_int CONFIG_X86_L1_CACHE_SHIFT 7
+fi
+if [ "$CONFIG_M686" = "y" ]; then
define_bool CONFIG_X86_TSC y
define_bool CONFIG_X86_GOOD_APIC y
define_bool CONFIG_X86_PGE y
define_bool CONFIG_X86_USE_PPRO_CHECKSUM y
fi
+
if [ "$CONFIG_MK6" = "y" ]; then
define_int CONFIG_X86_L1_CACHE_SHIFT 5
define_bool CONFIG_X86_ALIGNMENT_16 y
diff -ruN linux-2.4.19-pre9-jam1/arch/i386/defconfig linux-2.4.19-pre9-jam1-arch/arch/i386/defconfig
--- linux-2.4.19-pre9-jam1/arch/i386/defconfig Wed May 29 11:50:43 2002
+++ linux-2.4.19-pre9-jam1-arch/arch/i386/defconfig Wed May 29 11:54:59 2002
@@ -23,11 +23,12 @@
#
# CONFIG_M386 is not set
# CONFIG_M486 is not set
-# CONFIG_M586 is not set
-# CONFIG_M586TSC is not set
-# CONFIG_M586MMX is not set
-# CONFIG_M686 is not set
-CONFIG_MPENTIUMIII=y
+# CONFIG_MGEN586 is not set
+# CONFIG_MPENTIUM is not set
+# CONFIG_MPENTIUMMMX is not set
+# CONFIG_MPENTIUMPRO is not set
+# CONFIG_MPENTIUM2 is not set
+CONFIG_MPENTIUM3=y
# CONFIG_MPENTIUM4 is not set
# CONFIG_MK6 is not set
# CONFIG_MK7 is not set
diff -ruN linux-2.4.19-pre9-jam1/arch/i386/kernel/setup.c linux-2.4.19-pre9-jam1-arch/arch/i386/kernel/setup.c
--- linux-2.4.19-pre9-jam1/arch/i386/kernel/setup.c Wed May 29 11:50:43 2002
+++ linux-2.4.19-pre9-jam1-arch/arch/i386/kernel/setup.c Wed May 29 11:52:18 2002
@@ -2114,13 +2114,10 @@
static void __init init_intel(struct cpuinfo_x86 *c)
{
-#ifndef CONFIG_M686
- static int f00f_workaround_enabled = 0;
-#endif
char *p = NULL;
unsigned int l1i = 0, l1d = 0, l2 = 0, l3 = 0; /* Cache sizes */
-#ifndef CONFIG_M686
+#ifdef CONFIG_X86_F00F_BUG
/*
* All current models of Pentium and Pentium with MMX technology CPUs
* have the F0 0F bug, which lets nonpriviledged users lock up the system.
@@ -2128,6 +2125,8 @@
*/
c->f00f_bug = 0;
if ( c->x86 == 5 ) {
+ static int f00f_workaround_enabled = 0;
+
c->f00f_bug = 1;
if ( !f00f_workaround_enabled ) {
trap_init_f00f_bug();
diff -ruN linux-2.4.19-pre9-jam1/arch/i386/kernel/traps.c linux-2.4.19-pre9-jam1-arch/arch/i386/kernel/traps.c
--- linux-2.4.19-pre9-jam1/arch/i386/kernel/traps.c Wed May 29 11:50:43 2002
+++ linux-2.4.19-pre9-jam1-arch/arch/i386/kernel/traps.c Wed May 29 11:52:18 2002
@@ -757,35 +757,17 @@
#endif /* CONFIG_MATH_EMULATION */
-#ifndef CONFIG_M686
+#ifdef CONFIG_X86_F00F_BUG
void __init trap_init_f00f_bug(void)
{
- unsigned long page;
- pgd_t * pgd;
- pmd_t * pmd;
- pte_t * pte;
-
- /*
- * Allocate a new page in virtual address space,
- * move the IDT into it and write protect this page.
- */
- page = (unsigned long) vmalloc(PAGE_SIZE);
- pgd = pgd_offset(&init_mm, page);
- pmd = pmd_offset(pgd, page);
- pte = pte_offset(pmd, page);
- __free_page(pte_page(*pte));
- *pte = mk_pte_phys(__pa(&idt_table), PAGE_KERNEL_RO);
- /*
- * Not that any PGE-capable kernel should have the f00f bug ...
- */
- __flush_tlb_all();
+ __set_fixmap(FIX_F00F_IDT, __pa(&idt_table), PAGE_KERNEL_RO);
/*
* "idt" is magic - it overlaps the idt_descr
* variable so that updating idt will automatically
* update the idt descriptor..
*/
- idt = (struct desc_struct *)page;
+ idt = (struct desc_struct *) fix_to_virt(FIX_F00F_IDT);
__asm__ __volatile__("lidt %0": "=m" (idt_descr));
}
#endif
diff -ruN linux-2.4.19-pre9-jam1/arch/i386/mm/fault.c linux-2.4.19-pre9-jam1-arch/arch/i386/mm/fault.c
--- linux-2.4.19-pre9-jam1/arch/i386/mm/fault.c Wed May 29 11:50:43 2002
+++ linux-2.4.19-pre9-jam1-arch/arch/i386/mm/fault.c Wed May 29 11:52:18 2002
@@ -281,6 +281,7 @@
return;
}
+#ifdef CONFIG_X86_F00F_BUG
/*
* Pentium F0 0F C7 C8 bug workaround.
*/
@@ -294,6 +295,7 @@
return;
}
}
+#endif
no_context:
/* Are we prepared to handle this kernel fault? */
diff -ruN linux-2.4.19-pre9-jam1/include/asm-i386/fixmap.h linux-2.4.19-pre9-jam1-arch/include/asm-i386/fixmap.h
--- linux-2.4.19-pre9-jam1/include/asm-i386/fixmap.h Wed May 29 11:50:43 2002
+++ linux-2.4.19-pre9-jam1-arch/include/asm-i386/fixmap.h Wed May 29 11:52:18 2002
@@ -61,6 +61,9 @@
FIX_LI_PCIA, /* Lithium PCI Bridge A */
FIX_LI_PCIB, /* Lithium PCI Bridge B */
#endif
+#ifdef CONFIG_X86_F00F_BUG
+ FIX_F00F_IDT, /* Virtual mapping for IDT */
+#endif
#ifdef CONFIG_HIGHMEM
FIX_KMAP_BEGIN, /* reserved pte's for temporary kernel mappings */
FIX_KMAP_END = FIX_KMAP_BEGIN+(KM_TYPE_NR*NR_CPUS)-1,
--
J.A. Magallon # Let the source be with you...
mailto:jamagallon@able.es
Mandrake Linux release 8.3 (Cooker) for i586
Linux werewolf 2.4.19-pre9-jam1 #1 SMP mié may 29 02:20:48 CEST 2002 i686
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] intel-x86 model config cleanup
2002-05-29 14:35 [PATCH] intel-x86 model config cleanup J.A. Magallon
@ 2002-05-29 20:37 ` Jeff Garzik
2002-05-29 20:38 ` Jeff Garzik
0 siblings, 1 reply; 12+ messages in thread
From: Jeff Garzik @ 2002-05-29 20:37 UTC (permalink / raw)
To: J.A. Magallon; +Cc: Lista Linux-Kernel, Alan Cox, davej, Linus Torvalds
J.A. Magallon wrote:
>Hi all..
>
>One other try.
>
>This is an attempt to clean up the CPU model flags. Changes:
>
>- change names for CONFIG_Mxxxx, trying to make them more intuitive
>- split PII from PPro
>- introduce X86_F00F config flag and fixmap fix.
>- kill CONFIG_M586 as independent flag, and make it just an
> extra flag for 586 (I couldnt get a better name for
> MGEN586, suggestions wellcome...)
>- kill CONFIG_M686 as independent flag, and make it just an
> extra flag for anything >= PPro.
>- change option order definition to avoid duplicates, like:
>
>if [ "$CONFIG_MPENTIUM" = "y" ]; then
> define_bool CONFIG_M586 y <==============
> define_bool CONFIG_X86_TSC y
>fi
>if [ "$CONFIG_MPENTIUMMMX" = "y" ]; then
> define_bool CONFIG_M586 y <==============
> define_bool CONFIG_X86_TSC y
> define_bool CONFIG_X86_GOOD_APIC y
>fi
>if [ "$CONFIG_M586" = "y" ]; then <======== common things here
> define_int CONFIG_X86_L1_CACHE_SHIFT 5
> define_bool CONFIG_X86_USE_STRING_486 y
> define_bool CONFIG_X86_ALIGNMENT_16 y
> define_bool CONFIG_X86_PPRO_FENCE y
> define_bool CONFIG_X86_F00F_BUG y
>fi
>
>Patch follows:
>
>diff -ruN linux-2.4.19-pre9-jam1/Documentation/Configure.help linux-2.4.19-pre9-jam1-arch/Documentation/Configure.help
>--- linux-2.4.19-pre9-jam1/Documentation/Configure.help Wed May 29 11:50:43 2002
>+++ linux-2.4.19-pre9-jam1-arch/Documentation/Configure.help Wed May 29 11:52:18 2002
>@@ -3943,16 +3943,17 @@
> a PPro, but not necessarily on a i486.
>
> Here are the settings recommended for greatest speed:
>- - "386" for the AMD/Cyrix/Intel 386DX/DXL/SL/SLC/SX, Cyrix/TI
>+ - "Generic-386" for the AMD/Cyrix/Intel 386DX/DXL/SL/SLC/SX, Cyrix/TI
> 486DLC/DLC2, UMC 486SX-S and NexGen Nx586. Only "386" kernels
> will run on a 386 class machine.
>- - "486" for the AMD/Cyrix/IBM/Intel 486DX/DX2/DX4 or
>+ - "Generic-486" for the AMD/Cyrix/IBM/Intel 486DX/DX2/DX4 or
> SL/SLC/SLC2/SLC3/SX/SX2 and UMC U5D or U5S.
>- - "586" for generic Pentium CPUs, possibly lacking the TSC
>+ - "Generic-586" for generic Pentium CPUs, possibly lacking the TSC
> (time stamp counter) register.
>- - "Pentium-Classic" for the Intel Pentium.
>+ - "Pentium" for the Intel Pentium.
> - "Pentium-MMX" for the Intel Pentium MMX.
>- - "Pentium-Pro" for the Intel Pentium Pro/Celeron/Pentium II.
>+ - "Pentium-Pro" for the Intel Pentium Pro.
>+ - "Pentium-II" for the Intel Pentium II / Celeron.
> - "Pentium-III" for the Intel Pentium III
> and Celerons based on the Coppermine core.
> - "Pentium-4" for the Intel Pentium 4.
>
>
Since you are playing in this area, I wonder if you would consider doing
something that has been needed for a while:
Individual CPU selection.
This implies separating the concept of a "generic x86 kernel that
supports N CPU types" from "kernel supports one CPU type and one only."
The i386/config.in is currently a mishmash of both. Dave Jones did
some work along these lines in his "cpuchoice" diff, which I have attached.
Basically, the general direction IMO should be: a user can select the
CPU they own (Pentium-II), and none of the options for Pentium
3/4/Athlon will be enabled. Generic kernels would ask for a minimal CPU
level to support, I imagine. Support >=486, >=586, etc. Pretty much
what we have now.
Jeff
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] intel-x86 model config cleanup
2002-05-29 20:37 ` Jeff Garzik
@ 2002-05-29 20:38 ` Jeff Garzik
2002-05-29 22:05 ` Albert D. Cahalan
2002-05-29 22:44 ` J.A. Magallon
0 siblings, 2 replies; 12+ messages in thread
From: Jeff Garzik @ 2002-05-29 20:38 UTC (permalink / raw)
To: J.A. Magallon; +Cc: Lista Linux-Kernel, Alan Cox, davej, Linus Torvalds
[-- Attachment #1: Type: text/plain, Size: 352 bytes --]
Jeff Garzik wrote:
>
> This implies separating the concept of a "generic x86 kernel that
> supports N CPU types" from "kernel supports one CPU type and one
> only." The i386/config.in is currently a mishmash of both. Dave Jones
> did some work along these lines in his "cpuchoice" diff, which I have
> attached.
....or rather, attached here.
[-- Attachment #2: cpu-choice-2.diff --]
[-- Type: text/plain, Size: 2544 bytes --]
diff -urN --exclude-from=/home/davej/.exclude linux/arch/i386/config.in linux-dj/arch/i386/config.in
--- linux/arch/i386/config.in Mon Nov 12 19:58:08 2001
+++ linux-dj/arch/i386/config.in Tue Dec 4 00:40:28 2001
@@ -26,22 +26,75 @@
mainmenu_option next_comment
comment 'Processor type and features'
-choice 'Processor family' \
+choice 'Processor vendor' \
+ "AMD CONFIG_VENDOR_AMD \
+ Cyrix CONFIG_VENDOR_CYRIX \
+ Generic CONFIG_VENDOR_GENERIC \
+ IDT CONFIG_VENDOR_IDT \
+ Intel CONFIG_VENDOR_INTEL \
+ NationalSemiconductor CONFIG_VENDOR_NATSEMI \
+ RiSE CONFIG_VENDOR_RISE \
+ Transmeta CONFIG_VENDOR_TRANSMETA \
+ VIA CONFIG_VENDOR_VIA" Generic
+
+if [ "$CONFIG_VENDOR_AMD" = "y" ]; then
+ choice 'Processor family' \
+ "386 CONFIG_M386 \
+ 486 CONFIG_M486 \
+ K5/5x86 CONFIG_M586 \
+ K6/K6-II/K6-III CONFIG_MK6 \
+ Athlon/Duron CONFIG_MK7" Athlon
+fi
+
+if [ "$CONFIG_VENDOR_CYRIX" = "y" ]; then
+ choice 'Processor family' \
+ "386 CONFIG_M386 \
+ 486 CONFIG_M486 \
+ 586/5x86/6x86/6x86MX CONFIG_M586" 586
+fi
+
+if [ "$CONFIG_VENDOR_GENERIC" = "y" ]; then
+ choice 'Minimum spec CPU to generate code for' \
+ "386 CONFIG_M386 \
+ 486 CONFIG_M486 \
+ 586 CONFIG_M586 \
+ 686 CONFIG_M686" 386
+fi
+
+if [ "$CONFIG_VENDOR_IDT" = "y" ]; then
+ choice 'Processor family' \
+ "Winchip-C6 CONFIG_MWINCHIPC6 \
+ Winchip-2 CONFIG_MWINCHIP2 \
+ Winchip-2A/Winchip-3 CONFIG_MWINCHIP3D" Winchip-C6
+fi
+
+if [ "$CONFIG_VENDOR_INTEL" = "y" ]; then
+ choice 'Processor family' \
"386 CONFIG_M386 \
486 CONFIG_M486 \
- 586/K5/5x86/6x86/6x86MX CONFIG_M586 \
Pentium-Classic CONFIG_M586TSC \
Pentium-MMX CONFIG_M586MMX \
Pentium-Pro/Celeron/Pentium-II CONFIG_M686 \
Pentium-III/Celeron(Coppermine) CONFIG_MPENTIUMIII \
- Pentium-4 CONFIG_MPENTIUM4 \
- K6/K6-II/K6-III CONFIG_MK6 \
- Athlon/Duron/K7 CONFIG_MK7 \
- Crusoe CONFIG_MCRUSOE \
- Winchip-C6 CONFIG_MWINCHIPC6 \
- Winchip-2 CONFIG_MWINCHIP2 \
- Winchip-2A/Winchip-3 CONFIG_MWINCHIP3D \
- CyrixIII/C3 CONFIG_MCYRIXIII" Pentium-Pro
+ Pentium-4 CONFIG_MPENTIUM4" Pentium-Pro
+fi
+
+if [ "$CONFIG_VENDOR_NATSEMI" = "y" ]; then
+ define_bool CONFIG_M586 y
+fi
+
+if [ "$CONFIG_VENDOR_RISE" = "y" ]; then
+ define_bool CONFIG_M586 y
+fi
+
+if [ "$CONFIG_VENDOR_TRANSMETA" = "y" ]; then
+ define_bool CONFIG_MCRUSOE y
+fi
+
+if [ "$CONFIG_VENDOR_VIA" = "y" ]; then
+ define_bool CONFIG_MCYRIXIII y
+fi
+
#
# Define implied options from the CPU selection here
#
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] intel-x86 model config cleanup
2002-05-29 20:38 ` Jeff Garzik
@ 2002-05-29 22:05 ` Albert D. Cahalan
2002-05-30 0:11 ` Dave Jones
2002-05-29 22:44 ` J.A. Magallon
1 sibling, 1 reply; 12+ messages in thread
From: Albert D. Cahalan @ 2002-05-29 22:05 UTC (permalink / raw)
To: Jeff Garzik
Cc: J.A. Magallon, Lista Linux-Kernel, Alan Cox, davej, Linus Torvalds
Jeff Garzik writes:
> +choice 'Processor vendor' \
> + "AMD CONFIG_VENDOR_AMD \
> + Cyrix CONFIG_VENDOR_CYRIX \
> + Generic CONFIG_VENDOR_GENERIC \
> + IDT CONFIG_VENDOR_IDT \
> + Intel CONFIG_VENDOR_INTEL \
> + NationalSemiconductor CONFIG_VENDOR_NATSEMI \
> + RiSE CONFIG_VENDOR_RISE \
> + Transmeta CONFIG_VENDOR_TRANSMETA \
> + VIA CONFIG_VENDOR_VIA" Generic
> +
> +if [ "$CONFIG_VENDOR_AMD" = "y" ]; then
> + choice 'Processor family' \
> + "386 CONFIG_M386 \
> + 486 CONFIG_M486 \
> + K5/5x86 CONFIG_M586 \
> + K6/K6-II/K6-III CONFIG_MK6 \
> + Athlon/Duron CONFIG_MK7" Athlon
> +fi
This is still a mess. It's better to have one boolean
per processor, and order the processors by the year
in which they were most commonly sold.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] intel-x86 model config cleanup
2002-05-29 20:38 ` Jeff Garzik
2002-05-29 22:05 ` Albert D. Cahalan
@ 2002-05-29 22:44 ` J.A. Magallon
2002-05-29 22:54 ` Jeff Garzik
1 sibling, 1 reply; 12+ messages in thread
From: J.A. Magallon @ 2002-05-29 22:44 UTC (permalink / raw)
To: Jeff Garzik; +Cc: Lista Linux-Kernel, Alan Cox, davej
On 2002.05.29 Jeff Garzik wrote:
>Jeff Garzik wrote:
>
>>
>> This implies separating the concept of a "generic x86 kernel that
>> supports N CPU types" from "kernel supports one CPU type and one
>> only." The i386/config.in is currently a mishmash of both. Dave Jones
>> did some work along these lines in his "cpuchoice" diff, which I have
>> attached.
>
Could this be organized this way ?
CONFIG_VENDOR_ CONFIG_CPU_ CONFIG_M
GENERIC GENERIC_386 386
GENERIC_486 486
GENERIC_586 586
GENERIC_686 686
INTEL INTEL_386 386
INTEL_486 486
INTEL_PENTIUM 586
INTEL_PENTIUMMMX 586
INTEL_PENTIUMPRO 686
INTEL_PENTIUM2 686
INTEL_CELERON2 686
INTEL_PENTIUM3 686
INTEL_CELERON3 686
INTEL_PENTIUM4 686
AMD AMD_386 386
AMD_486 486
AMD_K5 586
AMD_K6 K6
AMD_K6II K6
AMD_K6III K6
AMD_K7 K7
AMD_DURON K7
AMD_ATHLON K7
CYRIX CYRIX_386 386
CYRIX_486 486
CYRIX_586 586
CYRIX_686 686
VIA CYRIX_III 686
TRANSMETA TRANSMETA_CRUSOE ???
NATSEMI NATSEMI_586 586
RISE RISE_586 586
IDT IDT_WINCHIPC6 ???
IDT_WINCHIP2 ???
IDT_WINCHIP2A ???
IDT_WINCHIP3 ???
Then for each model you would define its generic CONFIG_M<arch>, and
the specific features not contained in the generic. And then define
the rest of features based on generic.
The CONFIG_M<arch> would serve as a flag for 'this cpu has all features
of a generic xxx'.
Or if you are worried about namespace pollution these could be named
CONFIG_CPU_VENDOR_, CONFIG_CPU_, CONFIG_CPU_M.
BTW: any more stars in processor vendor/model sky ?
Comments awaited...
--
J.A. Magallon # Let the source be with you...
mailto:jamagallon@able.es
Mandrake Linux release 8.3 (Cooker) for i586
Linux werewolf 2.4.19-pre9-jam1 #1 SMP mié may 29 02:20:48 CEST 2002 i686
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] intel-x86 model config cleanup
2002-05-29 22:44 ` J.A. Magallon
@ 2002-05-29 22:54 ` Jeff Garzik
2002-05-29 23:06 ` J.A. Magallon
0 siblings, 1 reply; 12+ messages in thread
From: Jeff Garzik @ 2002-05-29 22:54 UTC (permalink / raw)
To: J.A. Magallon; +Cc: Lista Linux-Kernel, Alan Cox, davej
J.A. Magallon wrote:
>
>
>Then for each model you would define its generic CONFIG_M<arch>, and
>the specific features not contained in the generic. And then define
>the rest of features based on generic.
>The CONFIG_M<arch> would serve as a flag for 'this cpu has all features
>of a generic xxx'.
>
>Or if you are worried about namespace pollution these could be named
>CONFIG_CPU_VENDOR_, CONFIG_CPU_, CONFIG_CPU_M.
>
>
Your division of categories (snipped from above quoted) seems ok.
The basic thing to remember is that "generic_foo" or "cpu_intel_foo"
options should very rarely, if ever, appear in the config.in or sources.
We simply want to use the generic or cpu-specific user selection to
determine (a) compiler flags, (b) CONFIG_xxx symbols for specific CPU
features and optimizations, [like CONFIG_X86_F00F_BUG] and maybe (c)
enable and disable CPU-specific drivers. (c) will be a special case,
since very few drivers should require a specific CPU type... but some
drivers simply don't work on 386.
Jeff
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] intel-x86 model config cleanup
2002-05-29 22:54 ` Jeff Garzik
@ 2002-05-29 23:06 ` J.A. Magallon
2002-05-29 23:12 ` Jeff Garzik
0 siblings, 1 reply; 12+ messages in thread
From: J.A. Magallon @ 2002-05-29 23:06 UTC (permalink / raw)
To: Jeff Garzik; +Cc: J.A. Magallon, Lista Linux-Kernel, Alan Cox, davej
On 2002.05.30 Jeff Garzik wrote:
>
>The basic thing to remember is that "generic_foo" or "cpu_intel_foo"
>options should very rarely, if ever, appear in the config.in or sources.
> We simply want to use the generic or cpu-specific user selection to
>determine (a) compiler flags, (b) CONFIG_xxx symbols for specific CPU
>features and optimizations, [like CONFIG_X86_F00F_BUG] and maybe (c)
>enable and disable CPU-specific drivers. (c) will be a special case,
>since very few drivers should require a specific CPU type... but some
>drivers simply don't work on 386.
>
Grep on the tree showed this:
drivers/char/serial.c:
#if defined(__i386__) && (defined(CONFIG_M386) || defined(CONFIG_M486))
#define SERIAL_INLINE
#endif
include/asm-i386/processor.h:
/* Prefetch instructions for Pentium III and AMD Athlon */
#ifdef CONFIG_MPENTIUMIII
#define ARCH_HAS_PREFETCH
extern inline void prefetch(const void *x)
{
__asm__ __volatile__ ("prefetchnta (%0)" : : "r"(x));
}
#elif CONFIG_X86_USE_3DNOW
#define ARCH_HAS_PREFETCH
#define ARCH_HAS_PREFETCHW
#define ARCH_HAS_SPINLOCK_PREFETCH
More candidates for CONFIG_X86_xxxxx.
But these spawn over other architextures:
include/asm-alpha/processor.h:#define ARCH_HAS_PREFETCH
include/asm-ppc/processor.h:#define ARCH_HAS_PREFETCH
...
--
J.A. Magallon # Let the source be with you...
mailto:jamagallon@able.es
Mandrake Linux release 8.3 (Cooker) for i586
Linux werewolf 2.4.19-pre9-jam1 #1 SMP mié may 29 02:20:48 CEST 2002 i686
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] intel-x86 model config cleanup
2002-05-29 23:06 ` J.A. Magallon
@ 2002-05-29 23:12 ` Jeff Garzik
0 siblings, 0 replies; 12+ messages in thread
From: Jeff Garzik @ 2002-05-29 23:12 UTC (permalink / raw)
To: J.A. Magallon; +Cc: Lista Linux-Kernel, Alan Cox, davej
J.A. Magallon wrote:
>On 2002.05.30 Jeff Garzik wrote:
>
>
>>The basic thing to remember is that "generic_foo" or "cpu_intel_foo"
>>options should very rarely, if ever, appear in the config.in or sources.
>>We simply want to use the generic or cpu-specific user selection to
>>determine (a) compiler flags, (b) CONFIG_xxx symbols for specific CPU
>>features and optimizations, [like CONFIG_X86_F00F_BUG] and maybe (c)
>>enable and disable CPU-specific drivers. (c) will be a special case,
>>since very few drivers should require a specific CPU type... but some
>>drivers simply don't work on 386.
>>
>>
>>
>
>Grep on the tree showed this:
>
>drivers/char/serial.c:
>
>#if defined(__i386__) && (defined(CONFIG_M386) || defined(CONFIG_M486))
>#define SERIAL_INLINE
>#endif
>
>include/asm-i386/processor.h:
>
>/* Prefetch instructions for Pentium III and AMD Athlon */
>#ifdef CONFIG_MPENTIUMIII
>
>#define ARCH_HAS_PREFETCH
>
>
ARCH_HAS_foo are feature symbols, just like CONFIG_X86_F00F_BUG, so
those are fine. They're just cross-architecture, but perform the same
task: feature dis/enabling. For the serial example, you'd just need to
make sure the net effect of the code is the same, both before and after
an x86 config.in cleanup.
Jeff
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] intel-x86 model config cleanup
2002-05-29 22:05 ` Albert D. Cahalan
@ 2002-05-30 0:11 ` Dave Jones
2002-05-30 2:43 ` Albert D. Cahalan
0 siblings, 1 reply; 12+ messages in thread
From: Dave Jones @ 2002-05-30 0:11 UTC (permalink / raw)
To: Albert D. Cahalan
Cc: Jeff Garzik, J.A. Magallon, Lista Linux-Kernel, Alan Cox, Linus Torvalds
On Wed, May 29, 2002 at 06:05:40PM -0400, Albert D. Cahalan wrote:
>
> This is still a mess. It's better to have one boolean
> per processor, and order the processors by the year
> in which they were most commonly sold.
The information hiding of irrelevant options was one of the
motivations behind that original patch. If I know I have
an AMD Athlon, showing me all the Intel CPUs just gets in the way.
A zillion options in a scrolly list is no better than what
we currently have imo.
--
| Dave Jones. http://www.codemonkey.org.uk
| SuSE Labs
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] intel-x86 model config cleanup
2002-05-30 0:11 ` Dave Jones
@ 2002-05-30 2:43 ` Albert D. Cahalan
2002-05-30 4:34 ` Jeff Garzik
2002-05-30 11:09 ` Dave Jones
0 siblings, 2 replies; 12+ messages in thread
From: Albert D. Cahalan @ 2002-05-30 2:43 UTC (permalink / raw)
To: Dave Jones
Cc: Albert D. Cahalan, Jeff Garzik, J.A. Magallon,
Lista Linux-Kernel, Alan Cox, Linus Torvalds
Dave Jones writes:
> On Wed, May 29, 2002 at 06:05:40PM -0400, Albert D. Cahalan wrote:
>> This is still a mess. It's better to have one boolean
>> per processor, and order the processors by the year
>> in which they were most commonly sold.
>
> The information hiding of irrelevant options was one of the
> motivations behind that original patch. If I know I have
> an AMD Athlon, showing me all the Intel CPUs just gets in the way.
No, it's like this:
I want one kernel. I have a Pentium-MMX and a Pentium Pro.
I don't need support for a 386, 486, Athlon, or Xeon.
It's also like this:
We have a lab full of Athlon and Pentium III boxes.
There's not a Pentium 4 in sight, and no Pentium II
either. It's too much work to manage multiple kernels;
every box must boot from the same disk image.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] intel-x86 model config cleanup
2002-05-30 2:43 ` Albert D. Cahalan
@ 2002-05-30 4:34 ` Jeff Garzik
2002-05-30 11:09 ` Dave Jones
1 sibling, 0 replies; 12+ messages in thread
From: Jeff Garzik @ 2002-05-30 4:34 UTC (permalink / raw)
To: Albert D. Cahalan
Cc: Dave Jones, J.A. Magallon, Lista Linux-Kernel, Alan Cox, Linus Torvalds
Albert D. Cahalan wrote:
>
>
>I want one kernel. I have a Pentium-MMX and a Pentium Pro.
>I don't need support for a 386, 486, Athlon, or Xeon.
>
>
It depends on how much work the patch author wants to do. This is a
perfectly reasonable request, but increases the complexity of the
overall problem somewhat.
Your current solution, whatever it is, should map directly onto one of
the 'generic kernel' selections, hopefully. So, what you are really
asking for is an RFE to add on to this RFE currently being discussed :)
Jeff
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [PATCH] intel-x86 model config cleanup
2002-05-30 2:43 ` Albert D. Cahalan
2002-05-30 4:34 ` Jeff Garzik
@ 2002-05-30 11:09 ` Dave Jones
1 sibling, 0 replies; 12+ messages in thread
From: Dave Jones @ 2002-05-30 11:09 UTC (permalink / raw)
To: Albert D. Cahalan
Cc: Jeff Garzik, J.A. Magallon, Lista Linux-Kernel, Alan Cox, Linus Torvalds
On Wed, May 29, 2002 at 10:43:18PM -0400, Albert D. Cahalan wrote:
> No, it's like this:
> I want one kernel. I have a Pentium-MMX and a Pentium Pro.
> I don't need support for a 386, 486, Athlon, or Xeon.
>
> It's also like this:
>
> We have a lab full of Athlon and Pentium III boxes.
> There's not a Pentium 4 in sight, and no Pentium II
> either. It's too much work to manage multiple kernels;
> every box must boot from the same disk image.
Ok, I think I see what you're getting at. Instead of a multi-list
giving one result, you'd like to tick every box that kernel is
going run on, and have it spit out CONFIG_Mxxx for each of those
processors.
It's doable, and it keeps things like vendor kernels possible
(they just select all CPUs), it's probably quite a bit of work
though.
The trickiest bit is getting the Makefile right to choose
the lowest-chosen-denominator in the list for its gcc options, and that's
probably not that hard, just a few extra ifdefs in an already ifdef heavy area.
Other than that, dealing with the cases like "this cpu has TSC" "this
doesn't" and "this has this bug" "this doesn't have this bug" need to be
done too, but thats again, just a few more ifdefs away..
Dave.
--
| Dave Jones. http://www.codemonkey.org.uk
| SuSE Labs
^ permalink raw reply [flat|nested] 12+ messages in thread
end of thread, other threads:[~2002-05-30 11:09 UTC | newest]
Thread overview: 12+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2002-05-29 14:35 [PATCH] intel-x86 model config cleanup J.A. Magallon
2002-05-29 20:37 ` Jeff Garzik
2002-05-29 20:38 ` Jeff Garzik
2002-05-29 22:05 ` Albert D. Cahalan
2002-05-30 0:11 ` Dave Jones
2002-05-30 2:43 ` Albert D. Cahalan
2002-05-30 4:34 ` Jeff Garzik
2002-05-30 11:09 ` Dave Jones
2002-05-29 22:44 ` J.A. Magallon
2002-05-29 22:54 ` Jeff Garzik
2002-05-29 23:06 ` J.A. Magallon
2002-05-29 23:12 ` Jeff Garzik
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).