From: Andi Kleen We still need some kind of early CPU detection, e.g. for the AMD768 workaround and for the slab allocator to size its slabs correctly for the cache line. Also some other code already had private early CPU routines. This patch takes a new approach compared to the previous patch which caused Andrew so much grief. It only fills in a few selected fields in boot_cpu_data (only the data needed to identify the CPU type and the cache alignment). In particular the feature masks are not filled in, and the other fields are also not touched to prevent unwanted side effects. Also convert the ppro workaround to use standard cpu data now. Please consider applying to -mm. I'm not sure if slab still has the necessary support to use the cache line size early; previously Manfred showed some serious memory saving with this for kernels that are compiled for a bigger cache line size than the CPU (is often the case on distribution kernels). This code could be reenable now with this patch. --- 25-akpm/arch/i386/kernel/cpu/common.c | 49 ++++++++++++++++++++++++++++++---- 25-akpm/arch/i386/kernel/cpu/cpu.h | 3 ++ 25-akpm/arch/i386/kernel/cpu/intel.c | 47 +++++++++----------------------- 3 files changed, 61 insertions(+), 38 deletions(-) diff -puN arch/i386/kernel/cpu/common.c~new-version-of-early-cpu-detect arch/i386/kernel/cpu/common.c --- 25/arch/i386/kernel/cpu/common.c~new-version-of-early-cpu-detect 2004-04-22 22:54:23.207094088 -0700 +++ 25-akpm/arch/i386/kernel/cpu/common.c 2004-04-22 22:54:23.214093024 -0700 @@ -137,8 +137,7 @@ static char __init *table_lookup_model(s } - -void __init get_cpu_vendor(struct cpuinfo_x86 *c) +void __init get_cpu_vendor(struct cpuinfo_x86 *c, int early) { char *v = c->x86_vendor_id; int i; @@ -149,7 +148,8 @@ void __init get_cpu_vendor(struct cpuinf (cpu_devs[i]->c_ident[1] && !strcmp(v,cpu_devs[i]->c_ident[1]))) { c->x86_vendor = i; - this_cpu = cpu_devs[i]; + if (!early) + this_cpu = cpu_devs[i]; break; } } @@ -193,6 +193,44 @@ int __init have_cpuid_p(void) return flag_is_changeable_p(X86_EFLAGS_ID); } +/* Do minimum CPU detection early. + Fields really needed: vendor, cpuid_level, family, model, mask, cache alignment. + The others are not touched to avoid unwanted side effects. */ +void __init early_cpu_detect(void) +{ + struct cpuinfo_x86 *c = &boot_cpu_data; + + if (!have_cpuid_p()) + return; + + /* Get vendor name */ + cpuid(0x00000000, &c->cpuid_level, + (int *)&c->x86_vendor_id[0], + (int *)&c->x86_vendor_id[8], + (int *)&c->x86_vendor_id[4]); + + get_cpu_vendor(c, 1); + + c->x86 = 4; + c->x86_cache_alignment = 32; + + if (c->cpuid_level >= 0x00000001) { + u32 junk, tfms, cap0, misc; + cpuid(0x00000001, &tfms, &misc, &junk, &cap0); + c->x86 = (tfms >> 8) & 15; + c->x86_model = (tfms >> 4) & 15; + if (c->x86 == 0xf) { + c->x86 += (tfms >> 20) & 0xff; + c->x86_model += ((tfms >> 16) & 0xF) << 4; + } + c->x86_mask = tfms & 15; + if (cap0 & (1<<19)) + c->x86_cache_alignment = ((misc >> 8) & 0xff) * 8; + } + + early_intel_workaround(c); +} + void __init generic_identify(struct cpuinfo_x86 * c) { u32 tfms, xlvl; @@ -205,7 +243,7 @@ void __init generic_identify(struct cpui (int *)&c->x86_vendor_id[8], (int *)&c->x86_vendor_id[4]); - get_cpu_vendor(c); + get_cpu_vendor(c, 0); /* Initialize the standard set of capabilities */ /* Note that the vendor-specific code below might override */ @@ -383,7 +421,6 @@ void __init identify_cpu(struct cpuinfo_ void __init dodgy_tsc(void) { - get_cpu_vendor(&boot_cpu_data); if (( boot_cpu_data.x86_vendor == X86_VENDOR_CYRIX ) || ( boot_cpu_data.x86_vendor == X86_VENDOR_NSC )) cpu_devs[X86_VENDOR_CYRIX]->c_init(&boot_cpu_data); @@ -431,9 +468,11 @@ extern int transmeta_init_cpu(void); extern int rise_init_cpu(void); extern int nexgen_init_cpu(void); extern int umc_init_cpu(void); +void early_cpu_detect(void); void __init early_cpu_init(void) { + early_cpu_detect(); intel_cpu_init(); cyrix_init_cpu(); nsc_init_cpu(); diff -puN arch/i386/kernel/cpu/cpu.h~new-version-of-early-cpu-detect arch/i386/kernel/cpu/cpu.h --- 25/arch/i386/kernel/cpu/cpu.h~new-version-of-early-cpu-detect 2004-04-22 22:54:23.208093936 -0700 +++ 25-akpm/arch/i386/kernel/cpu/cpu.h 2004-04-22 22:54:23.214093024 -0700 @@ -26,3 +26,6 @@ extern void display_cacheinfo(struct cpu extern void generic_identify(struct cpuinfo_x86 * c); extern int have_cpuid_p(void); + +extern void early_intel_workaround(struct cpuinfo_x86 *c); + diff -puN arch/i386/kernel/cpu/intel.c~new-version-of-early-cpu-detect arch/i386/kernel/cpu/intel.c --- 25/arch/i386/kernel/cpu/intel.c~new-version-of-early-cpu-detect 2004-04-22 22:54:23.209093784 -0700 +++ 25-akpm/arch/i386/kernel/cpu/intel.c 2004-04-22 22:54:23.215092872 -0700 @@ -28,6 +28,15 @@ extern int trap_init_f00f_bug(void); struct movsl_mask movsl_mask; #endif +void __init early_intel_workaround(struct cpuinfo_x86 *c) +{ + if (c->x86_vendor != X86_VENDOR_INTEL) + return; + /* Netburst reports 64 bytes clflush size, but does IO in 128 bytes */ + if (c->x86 == 15 && c->x86_cache_alignment == 64) + c->x86_cache_alignment = 128; +} + /* * Early probe support logic for ppro memory erratum #50 * @@ -36,42 +45,14 @@ struct movsl_mask movsl_mask; int __init ppro_with_ram_bug(void) { - char vendor_id[16]; - int ident; - - /* Must have CPUID */ - if(!have_cpuid_p()) - return 0; - if(cpuid_eax(0)<1) - return 0; - - /* Must be Intel */ - cpuid(0, &ident, - (int *)&vendor_id[0], - (int *)&vendor_id[8], - (int *)&vendor_id[4]); - - if(memcmp(vendor_id, "IntelInside", 12)) - return 0; - - ident = cpuid_eax(1); - - /* Model 6 */ - - if(((ident>>8)&15)!=6) - return 0; - - /* Pentium Pro */ - - if(((ident>>4)&15)!=1) - return 0; - - if((ident&15) < 8) - { + /* Uses data from early_cpu_detect now */ + if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL && + boot_cpu_data.x86 == 6 && + boot_cpu_data.x86_model == 1 && + boot_cpu_data.x86_mask < 8) { printk(KERN_INFO "Pentium Pro with Errata#50 detected. Taking evasive action.\n"); return 1; } - printk(KERN_INFO "Your Pentium Pro seems ok.\n"); return 0; } _