Index: head/sys/i386/i386/initcpu.c =================================================================== --- head/sys/i386/i386/initcpu.c (revision 271407) +++ head/sys/i386/i386/initcpu.c (revision 271408) @@ -1,1084 +1,1106 @@ /*- * Copyright (c) KATO Takenori, 1997, 1998. * * All rights reserved. Unpublished rights reserved under the copyright * laws of Japan. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer as * the first lines of this file unmodified. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_cpu.h" #include #include #include #include #include #include #include #include #include #if !defined(CPU_DISABLE_SSE) && defined(I686_CPU) #define CPU_ENABLE_SSE #endif #ifdef I486_CPU static void init_5x86(void); static void init_bluelightning(void); static void init_486dlc(void); static void init_cy486dx(void); #ifdef CPU_I486_ON_386 static void init_i486_on_386(void); #endif static void init_6x86(void); #endif /* I486_CPU */ #if defined(I586_CPU) && defined(CPU_WT_ALLOC) static void enable_K5_wt_alloc(void); static void enable_K6_wt_alloc(void); static void enable_K6_2_wt_alloc(void); #endif #ifdef I686_CPU static void init_6x86MX(void); static void init_ppro(void); static void init_mendocino(void); #endif static int hw_instruction_sse; SYSCTL_INT(_hw, OID_AUTO, instruction_sse, CTLFLAG_RD, &hw_instruction_sse, 0, "SIMD/MMX2 instructions available in CPU"); /* * -1: automatic (default) * 0: keep enable CLFLUSH * 1: force disable CLFLUSH */ static int hw_clflush_disable = -1; int cpu; /* Are we 386, 386sx, 486, etc? */ u_int cpu_feature; /* Feature flags */ u_int cpu_feature2; /* Feature flags */ u_int amd_feature; /* AMD feature flags */ u_int amd_feature2; /* AMD feature flags */ u_int amd_pminfo; /* AMD advanced power management info */ u_int via_feature_rng; /* VIA RNG features */ u_int via_feature_xcrypt; /* VIA ACE features */ u_int cpu_high; /* Highest arg to CPUID */ u_int cpu_exthigh; /* Highest arg to extended CPUID */ u_int cpu_id; /* Stepping ID */ u_int cpu_procinfo; /* HyperThreading Info / Brand Index / CLFUSH */ u_int cpu_procinfo2; /* Multicore info */ char cpu_vendor[20]; /* CPU Origin code */ u_int cpu_vendor_id; /* CPU vendor ID */ #ifdef CPU_ENABLE_SSE u_int cpu_fxsr; /* SSE enabled */ u_int cpu_mxcsr_mask; /* Valid bits in mxcsr */ #endif u_int cpu_clflush_line_size = 32; u_int cpu_stdext_feature; u_int cpu_mon_mwait_flags; /* MONITOR/MWAIT flags (CPUID.05H.ECX) */ u_int cpu_mon_min_size; /* MONITOR minimum range size, bytes */ u_int cpu_mon_max_size; /* MONITOR minimum range size, bytes */ u_int cyrix_did; /* Device ID of Cyrix CPU */ SYSCTL_UINT(_hw, OID_AUTO, via_feature_rng, CTLFLAG_RD, &via_feature_rng, 0, "VIA RNG feature available in CPU"); SYSCTL_UINT(_hw, OID_AUTO, via_feature_xcrypt, CTLFLAG_RD, &via_feature_xcrypt, 0, "VIA xcrypt feature available in CPU"); #ifdef I486_CPU /* * IBM Blue Lightning */ static void init_bluelightning(void) { register_t saveintr; #if defined(PC98) && !defined(CPU_UPGRADE_HW_CACHE) need_post_dma_flush = 1; #endif saveintr = intr_disable(); load_cr0(rcr0() | CR0_CD | CR0_NW); invd(); #ifdef CPU_BLUELIGHTNING_FPU_OP_CACHE wrmsr(0x1000, 0x9c92LL); /* FP operand can be cacheable on Cyrix FPU */ #else wrmsr(0x1000, 0x1c92LL); /* Intel FPU */ #endif /* Enables 13MB and 0-640KB cache. */ wrmsr(0x1001, (0xd0LL << 32) | 0x3ff); #ifdef CPU_BLUELIGHTNING_3X wrmsr(0x1002, 0x04000000LL); /* Enables triple-clock mode. */ #else wrmsr(0x1002, 0x03000000LL); /* Enables double-clock mode. */ #endif /* Enable caching in CR0. */ load_cr0(rcr0() & ~(CR0_CD | CR0_NW)); /* CD = 0 and NW = 0 */ invd(); intr_restore(saveintr); } /* * Cyrix 486SLC/DLC/SR/DR series */ static void init_486dlc(void) { register_t saveintr; u_char ccr0; saveintr = intr_disable(); invd(); ccr0 = read_cyrix_reg(CCR0); #ifndef CYRIX_CACHE_WORKS ccr0 |= CCR0_NC1 | CCR0_BARB; write_cyrix_reg(CCR0, ccr0); invd(); #else ccr0 &= ~CCR0_NC0; #ifndef CYRIX_CACHE_REALLY_WORKS ccr0 |= CCR0_NC1 | CCR0_BARB; #else ccr0 |= CCR0_NC1; #endif #ifdef CPU_DIRECT_MAPPED_CACHE ccr0 |= CCR0_CO; /* Direct mapped mode. */ #endif write_cyrix_reg(CCR0, ccr0); /* Clear non-cacheable region. */ write_cyrix_reg(NCR1+2, NCR_SIZE_0K); write_cyrix_reg(NCR2+2, NCR_SIZE_0K); write_cyrix_reg(NCR3+2, NCR_SIZE_0K); write_cyrix_reg(NCR4+2, NCR_SIZE_0K); write_cyrix_reg(0, 0); /* dummy write */ /* Enable caching in CR0. */ load_cr0(rcr0() & ~(CR0_CD | CR0_NW)); /* CD = 0 and NW = 0 */ invd(); #endif /* !CYRIX_CACHE_WORKS */ intr_restore(saveintr); } /* * Cyrix 486S/DX series */ static void init_cy486dx(void) { register_t saveintr; u_char ccr2; saveintr = intr_disable(); invd(); ccr2 = read_cyrix_reg(CCR2); #ifdef CPU_SUSP_HLT ccr2 |= CCR2_SUSP_HLT; #endif #ifdef PC98 /* Enables WB cache interface pin and Lock NW bit in CR0. */ ccr2 |= CCR2_WB | CCR2_LOCK_NW; /* Unlock NW bit in CR0. */ write_cyrix_reg(CCR2, ccr2 & ~CCR2_LOCK_NW); load_cr0((rcr0() & ~CR0_CD) | CR0_NW); /* CD = 0, NW = 1 */ #endif write_cyrix_reg(CCR2, ccr2); intr_restore(saveintr); } /* * Cyrix 5x86 */ static void init_5x86(void) { register_t saveintr; u_char ccr2, ccr3, ccr4, pcr0; saveintr = intr_disable(); load_cr0(rcr0() | CR0_CD | CR0_NW); wbinvd(); (void)read_cyrix_reg(CCR3); /* dummy */ /* Initialize CCR2. */ ccr2 = read_cyrix_reg(CCR2); ccr2 |= CCR2_WB; #ifdef CPU_SUSP_HLT ccr2 |= CCR2_SUSP_HLT; #else ccr2 &= ~CCR2_SUSP_HLT; #endif ccr2 |= CCR2_WT1; write_cyrix_reg(CCR2, ccr2); /* Initialize CCR4. */ ccr3 = read_cyrix_reg(CCR3); write_cyrix_reg(CCR3, CCR3_MAPEN0); ccr4 = read_cyrix_reg(CCR4); ccr4 |= CCR4_DTE; ccr4 |= CCR4_MEM; #ifdef CPU_FASTER_5X86_FPU ccr4 |= CCR4_FASTFPE; #else ccr4 &= ~CCR4_FASTFPE; #endif ccr4 &= ~CCR4_IOMASK; /******************************************************************** * WARNING: The "BIOS Writers Guide" mentions that I/O recovery time * should be 0 for errata fix. ********************************************************************/ #ifdef CPU_IORT ccr4 |= CPU_IORT & CCR4_IOMASK; #endif write_cyrix_reg(CCR4, ccr4); /* Initialize PCR0. */ /**************************************************************** * WARNING: RSTK_EN and LOOP_EN could make your system unstable. * BTB_EN might make your system unstable. ****************************************************************/ pcr0 = read_cyrix_reg(PCR0); #ifdef CPU_RSTK_EN pcr0 |= PCR0_RSTK; #else pcr0 &= ~PCR0_RSTK; #endif #ifdef CPU_BTB_EN pcr0 |= PCR0_BTB; #else pcr0 &= ~PCR0_BTB; #endif #ifdef CPU_LOOP_EN pcr0 |= PCR0_LOOP; #else pcr0 &= ~PCR0_LOOP; #endif /**************************************************************** * WARNING: if you use a memory mapped I/O device, don't use * DISABLE_5X86_LSSER option, which may reorder memory mapped * I/O access. * IF YOUR MOTHERBOARD HAS PCI BUS, DON'T DISABLE LSSER. ****************************************************************/ #ifdef CPU_DISABLE_5X86_LSSER pcr0 &= ~PCR0_LSSER; #else pcr0 |= PCR0_LSSER; #endif write_cyrix_reg(PCR0, pcr0); /* Restore CCR3. */ write_cyrix_reg(CCR3, ccr3); (void)read_cyrix_reg(0x80); /* dummy */ /* Unlock NW bit in CR0. */ write_cyrix_reg(CCR2, read_cyrix_reg(CCR2) & ~CCR2_LOCK_NW); load_cr0((rcr0() & ~CR0_CD) | CR0_NW); /* CD = 0, NW = 1 */ /* Lock NW bit in CR0. */ write_cyrix_reg(CCR2, read_cyrix_reg(CCR2) | CCR2_LOCK_NW); intr_restore(saveintr); } #ifdef CPU_I486_ON_386 /* * There are i486 based upgrade products for i386 machines. * In this case, BIOS doesn't enable CPU cache. */ static void init_i486_on_386(void) { register_t saveintr; #if defined(PC98) && !defined(CPU_UPGRADE_HW_CACHE) need_post_dma_flush = 1; #endif saveintr = intr_disable(); load_cr0(rcr0() & ~(CR0_CD | CR0_NW)); /* CD = 0, NW = 0 */ intr_restore(saveintr); } #endif /* * Cyrix 6x86 * * XXX - What should I do here? Please let me know. */ static void init_6x86(void) { register_t saveintr; u_char ccr3, ccr4; saveintr = intr_disable(); load_cr0(rcr0() | CR0_CD | CR0_NW); wbinvd(); /* Initialize CCR0. */ write_cyrix_reg(CCR0, read_cyrix_reg(CCR0) | CCR0_NC1); /* Initialize CCR1. */ #ifdef CPU_CYRIX_NO_LOCK write_cyrix_reg(CCR1, read_cyrix_reg(CCR1) | CCR1_NO_LOCK); #else write_cyrix_reg(CCR1, read_cyrix_reg(CCR1) & ~CCR1_NO_LOCK); #endif /* Initialize CCR2. */ #ifdef CPU_SUSP_HLT write_cyrix_reg(CCR2, read_cyrix_reg(CCR2) | CCR2_SUSP_HLT); #else write_cyrix_reg(CCR2, read_cyrix_reg(CCR2) & ~CCR2_SUSP_HLT); #endif ccr3 = read_cyrix_reg(CCR3); write_cyrix_reg(CCR3, CCR3_MAPEN0); /* Initialize CCR4. */ ccr4 = read_cyrix_reg(CCR4); ccr4 |= CCR4_DTE; ccr4 &= ~CCR4_IOMASK; #ifdef CPU_IORT write_cyrix_reg(CCR4, ccr4 | (CPU_IORT & CCR4_IOMASK)); #else write_cyrix_reg(CCR4, ccr4 | 7); #endif /* Initialize CCR5. */ #ifdef CPU_WT_ALLOC write_cyrix_reg(CCR5, read_cyrix_reg(CCR5) | CCR5_WT_ALLOC); #endif /* Restore CCR3. */ write_cyrix_reg(CCR3, ccr3); /* Unlock NW bit in CR0. */ write_cyrix_reg(CCR2, read_cyrix_reg(CCR2) & ~CCR2_LOCK_NW); /* * Earlier revision of the 6x86 CPU could crash the system if * L1 cache is in write-back mode. */ if ((cyrix_did & 0xff00) > 0x1600) load_cr0(rcr0() & ~(CR0_CD | CR0_NW)); /* CD = 0 and NW = 0 */ else { /* Revision 2.6 and lower. */ #ifdef CYRIX_CACHE_REALLY_WORKS load_cr0(rcr0() & ~(CR0_CD | CR0_NW)); /* CD = 0 and NW = 0 */ #else load_cr0((rcr0() & ~CR0_CD) | CR0_NW); /* CD = 0 and NW = 1 */ #endif } /* Lock NW bit in CR0. */ write_cyrix_reg(CCR2, read_cyrix_reg(CCR2) | CCR2_LOCK_NW); intr_restore(saveintr); } #endif /* I486_CPU */ #ifdef I586_CPU /* * Rise mP6 */ static void init_rise(void) { /* * The CMPXCHG8B instruction is always available but hidden. */ cpu_feature |= CPUID_CX8; } /* * IDT WinChip C6/2/2A/2B/3 * * http://www.centtech.com/winchip_bios_writers_guide_v4_0.pdf */ static void init_winchip(void) { u_int regs[4]; uint64_t fcr; fcr = rdmsr(0x0107); /* * Set ECX8, DSMC, DTLOCK/EDCTLB, EMMX, and ERETSTK and clear DPDC. */ fcr |= (1 << 1) | (1 << 7) | (1 << 8) | (1 << 9) | (1 << 16); fcr &= ~(1ULL << 11); /* * Additioanlly, set EBRPRED, E2MMX and EAMD3D for WinChip 2 and 3. */ if (CPUID_TO_MODEL(cpu_id) >= 8) fcr |= (1 << 12) | (1 << 19) | (1 << 20); wrmsr(0x0107, fcr); do_cpuid(1, regs); cpu_feature = regs[3]; } #endif #ifdef I686_CPU /* * Cyrix 6x86MX (code-named M2) * * XXX - What should I do here? Please let me know. */ static void init_6x86MX(void) { register_t saveintr; u_char ccr3, ccr4; saveintr = intr_disable(); load_cr0(rcr0() | CR0_CD | CR0_NW); wbinvd(); /* Initialize CCR0. */ write_cyrix_reg(CCR0, read_cyrix_reg(CCR0) | CCR0_NC1); /* Initialize CCR1. */ #ifdef CPU_CYRIX_NO_LOCK write_cyrix_reg(CCR1, read_cyrix_reg(CCR1) | CCR1_NO_LOCK); #else write_cyrix_reg(CCR1, read_cyrix_reg(CCR1) & ~CCR1_NO_LOCK); #endif /* Initialize CCR2. */ #ifdef CPU_SUSP_HLT write_cyrix_reg(CCR2, read_cyrix_reg(CCR2) | CCR2_SUSP_HLT); #else write_cyrix_reg(CCR2, read_cyrix_reg(CCR2) & ~CCR2_SUSP_HLT); #endif ccr3 = read_cyrix_reg(CCR3); write_cyrix_reg(CCR3, CCR3_MAPEN0); /* Initialize CCR4. */ ccr4 = read_cyrix_reg(CCR4); ccr4 &= ~CCR4_IOMASK; #ifdef CPU_IORT write_cyrix_reg(CCR4, ccr4 | (CPU_IORT & CCR4_IOMASK)); #else write_cyrix_reg(CCR4, ccr4 | 7); #endif /* Initialize CCR5. */ #ifdef CPU_WT_ALLOC write_cyrix_reg(CCR5, read_cyrix_reg(CCR5) | CCR5_WT_ALLOC); #endif /* Restore CCR3. */ write_cyrix_reg(CCR3, ccr3); /* Unlock NW bit in CR0. */ write_cyrix_reg(CCR2, read_cyrix_reg(CCR2) & ~CCR2_LOCK_NW); load_cr0(rcr0() & ~(CR0_CD | CR0_NW)); /* CD = 0 and NW = 0 */ /* Lock NW bit in CR0. */ write_cyrix_reg(CCR2, read_cyrix_reg(CCR2) | CCR2_LOCK_NW); intr_restore(saveintr); } +static int ppro_apic_used = -1; + static void init_ppro(void) { u_int64_t apicbase; /* * Local APIC should be disabled if it is not going to be used. */ - apicbase = rdmsr(MSR_APICBASE); - apicbase &= ~APICBASE_ENABLED; - wrmsr(MSR_APICBASE, apicbase); + if (ppro_apic_used != 1) { + apicbase = rdmsr(MSR_APICBASE); + apicbase &= ~APICBASE_ENABLED; + wrmsr(MSR_APICBASE, apicbase); + ppro_apic_used = 0; + } +} + +/* + * If the local APIC is going to be used after being disabled above, + * re-enable it and don't disable it in the future. + */ +void +ppro_reenable_apic(void) +{ + u_int64_t apicbase; + + if (ppro_apic_used == 0) { + apicbase = rdmsr(MSR_APICBASE); + apicbase |= APICBASE_ENABLED; + wrmsr(MSR_APICBASE, apicbase); + ppro_apic_used = 1; + } } /* * Initialize BBL_CR_CTL3 (Control register 3: used to configure the * L2 cache). */ static void init_mendocino(void) { #ifdef CPU_PPRO2CELERON register_t saveintr; u_int64_t bbl_cr_ctl3; saveintr = intr_disable(); load_cr0(rcr0() | CR0_CD | CR0_NW); wbinvd(); bbl_cr_ctl3 = rdmsr(MSR_BBL_CR_CTL3); /* If the L2 cache is configured, do nothing. */ if (!(bbl_cr_ctl3 & 1)) { bbl_cr_ctl3 = 0x134052bLL; /* Set L2 Cache Latency (Default: 5). */ #ifdef CPU_CELERON_L2_LATENCY #if CPU_L2_LATENCY > 15 #error invalid CPU_L2_LATENCY. #endif bbl_cr_ctl3 |= CPU_L2_LATENCY << 1; #else bbl_cr_ctl3 |= 5 << 1; #endif wrmsr(MSR_BBL_CR_CTL3, bbl_cr_ctl3); } load_cr0(rcr0() & ~(CR0_CD | CR0_NW)); intr_restore(saveintr); #endif /* CPU_PPRO2CELERON */ } /* * Initialize special VIA features */ static void init_via(void) { u_int regs[4], val; uint64_t fcr; /* * Explicitly enable CX8 and PGE on C3. * * http://www.via.com.tw/download/mainboards/6/13/VIA_C3_EBGA%20datasheet110.pdf */ if (CPUID_TO_MODEL(cpu_id) <= 9) fcr = (1 << 1) | (1 << 7); else fcr = 0; /* * Check extended CPUID for PadLock features. * * http://www.via.com.tw/en/downloads/whitepapers/initiatives/padlock/programming_guide.pdf */ do_cpuid(0xc0000000, regs); if (regs[0] >= 0xc0000001) { do_cpuid(0xc0000001, regs); val = regs[3]; } else val = 0; /* Enable RNG if present. */ if ((val & VIA_CPUID_HAS_RNG) != 0) { via_feature_rng = VIA_HAS_RNG; wrmsr(0x110B, rdmsr(0x110B) | VIA_CPUID_DO_RNG); } /* Enable PadLock if present. */ if ((val & VIA_CPUID_HAS_ACE) != 0) via_feature_xcrypt |= VIA_HAS_AES; if ((val & VIA_CPUID_HAS_ACE2) != 0) via_feature_xcrypt |= VIA_HAS_AESCTR; if ((val & VIA_CPUID_HAS_PHE) != 0) via_feature_xcrypt |= VIA_HAS_SHA; if ((val & VIA_CPUID_HAS_PMM) != 0) via_feature_xcrypt |= VIA_HAS_MM; if (via_feature_xcrypt != 0) fcr |= 1 << 28; wrmsr(0x1107, rdmsr(0x1107) | fcr); } #endif /* I686_CPU */ #if defined(I586_CPU) || defined(I686_CPU) static void init_transmeta(void) { u_int regs[0]; /* Expose all hidden features. */ wrmsr(0x80860004, rdmsr(0x80860004) | ~0UL); do_cpuid(1, regs); cpu_feature = regs[3]; } #endif /* * Initialize CR4 (Control register 4) to enable SSE instructions. */ void enable_sse(void) { #if defined(CPU_ENABLE_SSE) if ((cpu_feature & CPUID_XMM) && (cpu_feature & CPUID_FXSR)) { load_cr4(rcr4() | CR4_FXSR | CR4_XMM); cpu_fxsr = hw_instruction_sse = 1; } #endif } extern int elf32_nxstack; void initializecpu(void) { switch (cpu) { #ifdef I486_CPU case CPU_BLUE: init_bluelightning(); break; case CPU_486DLC: init_486dlc(); break; case CPU_CY486DX: init_cy486dx(); break; case CPU_M1SC: init_5x86(); break; #ifdef CPU_I486_ON_386 case CPU_486: init_i486_on_386(); break; #endif case CPU_M1: init_6x86(); break; #endif /* I486_CPU */ #ifdef I586_CPU case CPU_586: switch (cpu_vendor_id) { case CPU_VENDOR_AMD: #ifdef CPU_WT_ALLOC if (((cpu_id & 0x0f0) > 0) && ((cpu_id & 0x0f0) < 0x60) && ((cpu_id & 0x00f) > 3)) enable_K5_wt_alloc(); else if (((cpu_id & 0x0f0) > 0x80) || (((cpu_id & 0x0f0) == 0x80) && (cpu_id & 0x00f) > 0x07)) enable_K6_2_wt_alloc(); else if ((cpu_id & 0x0f0) > 0x50) enable_K6_wt_alloc(); #endif if ((cpu_id & 0xf0) == 0xa0) /* * Make sure the TSC runs through * suspension, otherwise we can't use * it as timecounter */ wrmsr(0x1900, rdmsr(0x1900) | 0x20ULL); break; case CPU_VENDOR_CENTAUR: init_winchip(); break; case CPU_VENDOR_TRANSMETA: init_transmeta(); break; case CPU_VENDOR_RISE: init_rise(); break; } break; #endif #ifdef I686_CPU case CPU_M2: init_6x86MX(); break; case CPU_686: switch (cpu_vendor_id) { case CPU_VENDOR_INTEL: switch (cpu_id & 0xff0) { case 0x610: init_ppro(); break; case 0x660: init_mendocino(); break; } break; #ifdef CPU_ATHLON_SSE_HACK case CPU_VENDOR_AMD: /* * Sometimes the BIOS doesn't enable SSE instructions. * According to AMD document 20734, the mobile * Duron, the (mobile) Athlon 4 and the Athlon MP * support SSE. These correspond to cpu_id 0x66X * or 0x67X. */ if ((cpu_feature & CPUID_XMM) == 0 && ((cpu_id & ~0xf) == 0x660 || (cpu_id & ~0xf) == 0x670 || (cpu_id & ~0xf) == 0x680)) { u_int regs[4]; wrmsr(MSR_HWCR, rdmsr(MSR_HWCR) & ~0x08000); do_cpuid(1, regs); cpu_feature = regs[3]; } break; #endif case CPU_VENDOR_CENTAUR: init_via(); break; case CPU_VENDOR_TRANSMETA: init_transmeta(); break; } #ifdef PAE if ((amd_feature & AMDID_NX) != 0) { uint64_t msr; msr = rdmsr(MSR_EFER) | EFER_NXE; wrmsr(MSR_EFER, msr); pg_nx = PG_NX; elf32_nxstack = 1; } #endif break; #endif default: break; } enable_sse(); /* * CPUID with %eax = 1, %ebx returns * Bits 15-8: CLFLUSH line size * (Value * 8 = cache line size in bytes) */ if ((cpu_feature & CPUID_CLFSH) != 0) cpu_clflush_line_size = ((cpu_procinfo >> 8) & 0xff) * 8; /* * XXXKIB: (temporary) hack to work around traps generated * when CLFLUSHing APIC register window under virtualization * environments. These environments tend to disable the * CPUID_SS feature even though the native CPU supports it. */ TUNABLE_INT_FETCH("hw.clflush_disable", &hw_clflush_disable); if (vm_guest != VM_GUEST_NO && hw_clflush_disable == -1) cpu_feature &= ~CPUID_CLFSH; /* * Allow to disable CLFLUSH feature manually by * hw.clflush_disable tunable. */ if (hw_clflush_disable == 1) cpu_feature &= ~CPUID_CLFSH; #if defined(PC98) && !defined(CPU_UPGRADE_HW_CACHE) /* * OS should flush L1 cache by itself because no PC-98 supports * non-Intel CPUs. Use wbinvd instruction before DMA transfer * when need_pre_dma_flush = 1, use invd instruction after DMA * transfer when need_post_dma_flush = 1. If your CPU upgrade * product supports hardware cache control, you can add the * CPU_UPGRADE_HW_CACHE option in your kernel configuration file. * This option eliminates unneeded cache flush instruction(s). */ if (cpu_vendor_id == CPU_VENDOR_CYRIX) { switch (cpu) { #ifdef I486_CPU case CPU_486DLC: need_post_dma_flush = 1; break; case CPU_M1SC: need_pre_dma_flush = 1; break; case CPU_CY486DX: need_pre_dma_flush = 1; #ifdef CPU_I486_ON_386 need_post_dma_flush = 1; #endif break; #endif default: break; } } else if (cpu_vendor_id == CPU_VENDOR_AMD) { switch (cpu_id & 0xFF0) { case 0x470: /* Enhanced Am486DX2 WB */ case 0x490: /* Enhanced Am486DX4 WB */ case 0x4F0: /* Am5x86 WB */ need_pre_dma_flush = 1; break; } } else if (cpu_vendor_id == CPU_VENDOR_IBM) { need_post_dma_flush = 1; } else { #ifdef CPU_I486_ON_386 need_pre_dma_flush = 1; #endif } #endif /* PC98 && !CPU_UPGRADE_HW_CACHE */ } #if defined(I586_CPU) && defined(CPU_WT_ALLOC) /* * Enable write allocate feature of AMD processors. * Following two functions require the Maxmem variable being set. */ static void enable_K5_wt_alloc(void) { u_int64_t msr; register_t saveintr; /* * Write allocate is supported only on models 1, 2, and 3, with * a stepping of 4 or greater. */ if (((cpu_id & 0xf0) > 0) && ((cpu_id & 0x0f) > 3)) { saveintr = intr_disable(); msr = rdmsr(0x83); /* HWCR */ wrmsr(0x83, msr & !(0x10)); /* * We have to tell the chip where the top of memory is, * since video cards could have frame bufferes there, * memory-mapped I/O could be there, etc. */ if(Maxmem > 0) msr = Maxmem / 16; else msr = 0; msr |= AMD_WT_ALLOC_TME | AMD_WT_ALLOC_FRE; #ifdef PC98 if (!(inb(0x43b) & 4)) { wrmsr(0x86, 0x0ff00f0); msr |= AMD_WT_ALLOC_PRE; } #else /* * There is no way to know wheter 15-16M hole exists or not. * Therefore, we disable write allocate for this range. */ wrmsr(0x86, 0x0ff00f0); msr |= AMD_WT_ALLOC_PRE; #endif wrmsr(0x85, msr); msr=rdmsr(0x83); wrmsr(0x83, msr|0x10); /* enable write allocate */ intr_restore(saveintr); } } static void enable_K6_wt_alloc(void) { quad_t size; u_int64_t whcr; register_t saveintr; saveintr = intr_disable(); wbinvd(); #ifdef CPU_DISABLE_CACHE /* * Certain K6-2 box becomes unstable when write allocation is * enabled. */ /* * The AMD-K6 processer provides the 64-bit Test Register 12(TR12), * but only the Cache Inhibit(CI) (bit 3 of TR12) is suppported. * All other bits in TR12 have no effect on the processer's operation. * The I/O Trap Restart function (bit 9 of TR12) is always enabled * on the AMD-K6. */ wrmsr(0x0000000e, (u_int64_t)0x0008); #endif /* Don't assume that memory size is aligned with 4M. */ if (Maxmem > 0) size = ((Maxmem >> 8) + 3) >> 2; else size = 0; /* Limit is 508M bytes. */ if (size > 0x7f) size = 0x7f; whcr = (rdmsr(0xc0000082) & ~(0x7fLL << 1)) | (size << 1); #if defined(PC98) || defined(NO_MEMORY_HOLE) if (whcr & (0x7fLL << 1)) { #ifdef PC98 /* * If bit 2 of port 0x43b is 0, disable wrte allocate for the * 15-16M range. */ if (!(inb(0x43b) & 4)) whcr &= ~0x0001LL; else #endif whcr |= 0x0001LL; } #else /* * There is no way to know wheter 15-16M hole exists or not. * Therefore, we disable write allocate for this range. */ whcr &= ~0x0001LL; #endif wrmsr(0x0c0000082, whcr); intr_restore(saveintr); } static void enable_K6_2_wt_alloc(void) { quad_t size; u_int64_t whcr; register_t saveintr; saveintr = intr_disable(); wbinvd(); #ifdef CPU_DISABLE_CACHE /* * Certain K6-2 box becomes unstable when write allocation is * enabled. */ /* * The AMD-K6 processer provides the 64-bit Test Register 12(TR12), * but only the Cache Inhibit(CI) (bit 3 of TR12) is suppported. * All other bits in TR12 have no effect on the processer's operation. * The I/O Trap Restart function (bit 9 of TR12) is always enabled * on the AMD-K6. */ wrmsr(0x0000000e, (u_int64_t)0x0008); #endif /* Don't assume that memory size is aligned with 4M. */ if (Maxmem > 0) size = ((Maxmem >> 8) + 3) >> 2; else size = 0; /* Limit is 4092M bytes. */ if (size > 0x3fff) size = 0x3ff; whcr = (rdmsr(0xc0000082) & ~(0x3ffLL << 22)) | (size << 22); #if defined(PC98) || defined(NO_MEMORY_HOLE) if (whcr & (0x3ffLL << 22)) { #ifdef PC98 /* * If bit 2 of port 0x43b is 0, disable wrte allocate for the * 15-16M range. */ if (!(inb(0x43b) & 4)) whcr &= ~(1LL << 16); else #endif whcr |= 1LL << 16; } #else /* * There is no way to know wheter 15-16M hole exists or not. * Therefore, we disable write allocate for this range. */ whcr &= ~(1LL << 16); #endif wrmsr(0x0c0000082, whcr); intr_restore(saveintr); } #endif /* I585_CPU && CPU_WT_ALLOC */ #include "opt_ddb.h" #ifdef DDB #include DB_SHOW_COMMAND(cyrixreg, cyrixreg) { register_t saveintr; u_int cr0; u_char ccr1, ccr2, ccr3; u_char ccr0 = 0, ccr4 = 0, ccr5 = 0, pcr0 = 0; cr0 = rcr0(); if (cpu_vendor_id == CPU_VENDOR_CYRIX) { saveintr = intr_disable(); if ((cpu != CPU_M1SC) && (cpu != CPU_CY486DX)) { ccr0 = read_cyrix_reg(CCR0); } ccr1 = read_cyrix_reg(CCR1); ccr2 = read_cyrix_reg(CCR2); ccr3 = read_cyrix_reg(CCR3); if ((cpu == CPU_M1SC) || (cpu == CPU_M1) || (cpu == CPU_M2)) { write_cyrix_reg(CCR3, CCR3_MAPEN0); ccr4 = read_cyrix_reg(CCR4); if ((cpu == CPU_M1) || (cpu == CPU_M2)) ccr5 = read_cyrix_reg(CCR5); else pcr0 = read_cyrix_reg(PCR0); write_cyrix_reg(CCR3, ccr3); /* Restore CCR3. */ } intr_restore(saveintr); if ((cpu != CPU_M1SC) && (cpu != CPU_CY486DX)) printf("CCR0=%x, ", (u_int)ccr0); printf("CCR1=%x, CCR2=%x, CCR3=%x", (u_int)ccr1, (u_int)ccr2, (u_int)ccr3); if ((cpu == CPU_M1SC) || (cpu == CPU_M1) || (cpu == CPU_M2)) { printf(", CCR4=%x, ", (u_int)ccr4); if (cpu == CPU_M1SC) printf("PCR0=%x\n", pcr0); else printf("CCR5=%x\n", ccr5); } } printf("CR0=%x\n", cr0); } #endif /* DDB */ Index: head/sys/i386/include/md_var.h =================================================================== --- head/sys/i386/include/md_var.h (revision 271407) +++ head/sys/i386/include/md_var.h (revision 271408) @@ -1,117 +1,118 @@ /*- * Copyright (c) 1995 Bruce D. Evans. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the author nor the names of contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #ifndef _MACHINE_MD_VAR_H_ #define _MACHINE_MD_VAR_H_ /* * Miscellaneous machine-dependent declarations. */ extern long Maxmem; extern u_int basemem; /* PA of original top of base memory */ extern int busdma_swi_pending; extern u_int cpu_exthigh; extern u_int cpu_feature; extern u_int cpu_feature2; extern u_int amd_feature; extern u_int amd_feature2; extern u_int amd_pminfo; extern u_int via_feature_rng; extern u_int via_feature_xcrypt; extern u_int cpu_clflush_line_size; extern u_int cpu_stdext_feature; extern u_int cpu_fxsr; extern u_int cpu_high; extern u_int cpu_id; extern u_int cpu_mxcsr_mask; extern u_int cpu_procinfo; extern u_int cpu_procinfo2; extern char cpu_vendor[]; extern u_int cpu_vendor_id; extern u_int cpu_mon_mwait_flags; extern u_int cpu_mon_min_size; extern u_int cpu_mon_max_size; extern u_int cyrix_did; #if defined(I586_CPU) && !defined(NO_F00F_HACK) extern int has_f00f_bug; #endif extern char kstack[]; extern char sigcode[]; extern int szsigcode; #ifdef COMPAT_FREEBSD4 extern int szfreebsd4_sigcode; #endif #ifdef COMPAT_43 extern int szosigcode; #endif extern uint32_t *vm_page_dump; extern int vm_page_dump_size; extern int workaround_erratum383; typedef void alias_for_inthand_t(u_int cs, u_int ef, u_int esp, u_int ss); struct thread; struct reg; struct fpreg; struct dbreg; struct dumperinfo; void bcopyb(const void *from, void *to, size_t len); void busdma_swi(void); void cpu_setregs(void); void cpu_switch_load_gs(void) __asm(__STRING(cpu_switch_load_gs)); void doreti_iret(void) __asm(__STRING(doreti_iret)); void doreti_iret_fault(void) __asm(__STRING(doreti_iret_fault)); void doreti_popl_ds(void) __asm(__STRING(doreti_popl_ds)); void doreti_popl_ds_fault(void) __asm(__STRING(doreti_popl_ds_fault)); void doreti_popl_es(void) __asm(__STRING(doreti_popl_es)); void doreti_popl_es_fault(void) __asm(__STRING(doreti_popl_es_fault)); void doreti_popl_fs(void) __asm(__STRING(doreti_popl_fs)); void doreti_popl_fs_fault(void) __asm(__STRING(doreti_popl_fs_fault)); void dump_add_page(vm_paddr_t); void dump_drop_page(vm_paddr_t); void finishidentcpu(void); void enable_sse(void); void fillw(int /*u_short*/ pat, void *base, size_t cnt); void initializecpu(void); void i686_pagezero(void *addr); void sse2_pagezero(void *addr); void init_AMD_Elan_sc520(void); int is_physical_memory(vm_paddr_t addr); int isa_nmi(int cd); vm_paddr_t kvtop(void *addr); void panicifcpuunsupported(void); +void ppro_reenable_apic(void); void printcpuinfo(void); void setidt(int idx, alias_for_inthand_t *func, int typ, int dpl, int selec); int user_dbreg_trap(void); void minidumpsys(struct dumperinfo *); #endif /* !_MACHINE_MD_VAR_H_ */ Index: head/sys/x86/x86/local_apic.c =================================================================== --- head/sys/x86/x86/local_apic.c (revision 271407) +++ head/sys/x86/x86/local_apic.c (revision 271408) @@ -1,1591 +1,1582 @@ /*- * Copyright (c) 2003 John Baldwin * Copyright (c) 1996, by Steve Passe * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. The name of the developer may NOT be used to endorse or promote products * derived from this software without specific prior written permission. * 3. Neither the name of the author nor the names of any co-contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * Local APIC support on Pentium and later processors. */ #include __FBSDID("$FreeBSD$"); #include "opt_atpic.h" #include "opt_hwpmc_hooks.h" #include "opt_ddb.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include -#include #include #include #include #include #include #include #include #include #ifdef DDB #include #include #endif #ifdef __amd64__ #define SDT_APIC SDT_SYSIGT #define SDT_APICT SDT_SYSIGT #define GSEL_APIC 0 #else #define SDT_APIC SDT_SYS386IGT #define SDT_APICT SDT_SYS386TGT #define GSEL_APIC GSEL(GCODE_SEL, SEL_KPL) #endif /* Sanity checks on IDT vectors. */ CTASSERT(APIC_IO_INTS + APIC_NUM_IOINTS == APIC_TIMER_INT); CTASSERT(APIC_TIMER_INT < APIC_LOCAL_INTS); CTASSERT(APIC_LOCAL_INTS == 240); CTASSERT(IPI_STOP < APIC_SPURIOUS_INT); /* Magic IRQ values for the timer and syscalls. */ #define IRQ_TIMER (NUM_IO_INTS + 1) #define IRQ_SYSCALL (NUM_IO_INTS + 2) #define IRQ_DTRACE_RET (NUM_IO_INTS + 3) #define IRQ_EVTCHN (NUM_IO_INTS + 4) /* * Support for local APICs. Local APICs manage interrupts on each * individual processor as opposed to I/O APICs which receive interrupts * from I/O devices and then forward them on to the local APICs. * * Local APICs can also send interrupts to each other thus providing the * mechanism for IPIs. */ struct lvt { u_int lvt_edgetrigger:1; u_int lvt_activehi:1; u_int lvt_masked:1; u_int lvt_active:1; u_int lvt_mode:16; u_int lvt_vector:8; }; struct lapic { struct lvt la_lvts[APIC_LVT_MAX + 1]; u_int la_id:8; u_int la_cluster:4; u_int la_cluster_id:2; u_int la_present:1; u_long *la_timer_count; u_long la_timer_period; u_int la_timer_mode; uint32_t lvt_timer_cache; /* Include IDT_SYSCALL to make indexing easier. */ int la_ioint_irqs[APIC_NUM_IOINTS + 1]; } static lapics[MAX_APIC_ID + 1]; /* Global defaults for local APIC LVT entries. */ static struct lvt lvts[APIC_LVT_MAX + 1] = { { 1, 1, 1, 1, APIC_LVT_DM_EXTINT, 0 }, /* LINT0: masked ExtINT */ { 1, 1, 0, 1, APIC_LVT_DM_NMI, 0 }, /* LINT1: NMI */ { 1, 1, 1, 1, APIC_LVT_DM_FIXED, APIC_TIMER_INT }, /* Timer */ { 1, 1, 0, 1, APIC_LVT_DM_FIXED, APIC_ERROR_INT }, /* Error */ { 1, 1, 1, 1, APIC_LVT_DM_NMI, 0 }, /* PMC */ { 1, 1, 1, 1, APIC_LVT_DM_FIXED, APIC_THERMAL_INT }, /* Thermal */ { 1, 1, 1, 1, APIC_LVT_DM_FIXED, APIC_CMC_INT }, /* CMCI */ }; static inthand_t *ioint_handlers[] = { NULL, /* 0 - 31 */ IDTVEC(apic_isr1), /* 32 - 63 */ IDTVEC(apic_isr2), /* 64 - 95 */ IDTVEC(apic_isr3), /* 96 - 127 */ IDTVEC(apic_isr4), /* 128 - 159 */ IDTVEC(apic_isr5), /* 160 - 191 */ IDTVEC(apic_isr6), /* 192 - 223 */ IDTVEC(apic_isr7), /* 224 - 255 */ }; static u_int32_t lapic_timer_divisors[] = { APIC_TDCR_1, APIC_TDCR_2, APIC_TDCR_4, APIC_TDCR_8, APIC_TDCR_16, APIC_TDCR_32, APIC_TDCR_64, APIC_TDCR_128 }; extern inthand_t IDTVEC(rsvd); volatile lapic_t *lapic; vm_paddr_t lapic_paddr; static u_long lapic_timer_divisor; static struct eventtimer lapic_et; static void lapic_enable(void); static void lapic_resume(struct pic *pic, bool suspend_cancelled); static void lapic_timer_oneshot(struct lapic *, u_int count, int enable_int); static void lapic_timer_periodic(struct lapic *, u_int count, int enable_int); static void lapic_timer_stop(struct lapic *); static void lapic_timer_set_divisor(u_int divisor); static uint32_t lvt_mode(struct lapic *la, u_int pin, uint32_t value); static int lapic_et_start(struct eventtimer *et, sbintime_t first, sbintime_t period); static int lapic_et_stop(struct eventtimer *et); static u_int apic_idt_to_irq(u_int apic_id, u_int vector); static void lapic_set_tpr(u_int vector); struct pic lapic_pic = { .pic_resume = lapic_resume }; /* Forward declarations for apic_ops */ static void native_lapic_create(u_int apic_id, int boot_cpu); static void native_lapic_init(vm_paddr_t addr); static void native_lapic_setup(int boot); static void native_lapic_dump(const char *str); static void native_lapic_disable(void); static void native_lapic_eoi(void); static int native_lapic_id(void); static int native_lapic_intr_pending(u_int vector); static u_int native_apic_cpuid(u_int apic_id); static u_int native_apic_alloc_vector(u_int apic_id, u_int irq); static u_int native_apic_alloc_vectors(u_int apic_id, u_int *irqs, u_int count, u_int align); static void native_apic_disable_vector(u_int apic_id, u_int vector); static void native_apic_enable_vector(u_int apic_id, u_int vector); static void native_apic_free_vector(u_int apic_id, u_int vector, u_int irq); static void native_lapic_set_logical_id(u_int apic_id, u_int cluster, u_int cluster_id); static int native_lapic_enable_pmc(void); static void native_lapic_disable_pmc(void); static void native_lapic_reenable_pmc(void); static void native_lapic_enable_cmc(void); static void native_lapic_ipi_raw(register_t icrlo, u_int dest); static void native_lapic_ipi_vectored(u_int vector, int dest); static int native_lapic_ipi_wait(int delay); static int native_lapic_set_lvt_mask(u_int apic_id, u_int lvt, u_char masked); static int native_lapic_set_lvt_mode(u_int apic_id, u_int lvt, uint32_t mode); static int native_lapic_set_lvt_polarity(u_int apic_id, u_int lvt, enum intr_polarity pol); static int native_lapic_set_lvt_triggermode(u_int apic_id, u_int lvt, enum intr_trigger trigger); struct apic_ops apic_ops = { .create = native_lapic_create, .init = native_lapic_init, .setup = native_lapic_setup, .dump = native_lapic_dump, .disable = native_lapic_disable, .eoi = native_lapic_eoi, .id = native_lapic_id, .intr_pending = native_lapic_intr_pending, .set_logical_id = native_lapic_set_logical_id, .cpuid = native_apic_cpuid, .alloc_vector = native_apic_alloc_vector, .alloc_vectors = native_apic_alloc_vectors, .enable_vector = native_apic_enable_vector, .disable_vector = native_apic_disable_vector, .free_vector = native_apic_free_vector, .enable_pmc = native_lapic_enable_pmc, .disable_pmc = native_lapic_disable_pmc, .reenable_pmc = native_lapic_reenable_pmc, .enable_cmc = native_lapic_enable_cmc, #ifdef SMP .ipi_raw = native_lapic_ipi_raw, .ipi_vectored = native_lapic_ipi_vectored, .ipi_wait = native_lapic_ipi_wait, #endif .set_lvt_mask = native_lapic_set_lvt_mask, .set_lvt_mode = native_lapic_set_lvt_mode, .set_lvt_polarity = native_lapic_set_lvt_polarity, .set_lvt_triggermode = native_lapic_set_lvt_triggermode, }; static uint32_t lvt_mode(struct lapic *la, u_int pin, uint32_t value) { struct lvt *lvt; KASSERT(pin <= APIC_LVT_MAX, ("%s: pin %u out of range", __func__, pin)); if (la->la_lvts[pin].lvt_active) lvt = &la->la_lvts[pin]; else lvt = &lvts[pin]; value &= ~(APIC_LVT_M | APIC_LVT_TM | APIC_LVT_IIPP | APIC_LVT_DM | APIC_LVT_VECTOR); if (lvt->lvt_edgetrigger == 0) value |= APIC_LVT_TM; if (lvt->lvt_activehi == 0) value |= APIC_LVT_IIPP_INTALO; if (lvt->lvt_masked) value |= APIC_LVT_M; value |= lvt->lvt_mode; switch (lvt->lvt_mode) { case APIC_LVT_DM_NMI: case APIC_LVT_DM_SMI: case APIC_LVT_DM_INIT: case APIC_LVT_DM_EXTINT: if (!lvt->lvt_edgetrigger && bootverbose) { printf("lapic%u: Forcing LINT%u to edge trigger\n", la->la_id, pin); value |= APIC_LVT_TM; } /* Use a vector of 0. */ break; case APIC_LVT_DM_FIXED: value |= lvt->lvt_vector; break; default: panic("bad APIC LVT delivery mode: %#x\n", value); } return (value); } /* * Map the local APIC and setup necessary interrupt vectors. */ static void native_lapic_init(vm_paddr_t addr) { u_int regs[4]; int i, arat; /* Map the local APIC and setup the spurious interrupt handler. */ KASSERT(trunc_page(addr) == addr, ("local APIC not aligned on a page boundary")); lapic_paddr = addr; lapic = pmap_mapdev(addr, sizeof(lapic_t)); setidt(APIC_SPURIOUS_INT, IDTVEC(spuriousint), SDT_APIC, SEL_KPL, GSEL_APIC); /* Perform basic initialization of the BSP's local APIC. */ lapic_enable(); /* Set BSP's per-CPU local APIC ID. */ PCPU_SET(apic_id, lapic_id()); /* Local APIC timer interrupt. */ setidt(APIC_TIMER_INT, IDTVEC(timerint), SDT_APIC, SEL_KPL, GSEL_APIC); /* Local APIC error interrupt. */ setidt(APIC_ERROR_INT, IDTVEC(errorint), SDT_APIC, SEL_KPL, GSEL_APIC); /* XXX: Thermal interrupt */ /* Local APIC CMCI. */ setidt(APIC_CMC_INT, IDTVEC(cmcint), SDT_APICT, SEL_KPL, GSEL_APIC); if ((resource_int_value("apic", 0, "clock", &i) != 0 || i != 0)) { arat = 0; /* Intel CPUID 0x06 EAX[2] set if APIC timer runs in C3. */ if (cpu_vendor_id == CPU_VENDOR_INTEL && cpu_high >= 6) { do_cpuid(0x06, regs); if ((regs[0] & CPUTPM1_ARAT) != 0) arat = 1; } bzero(&lapic_et, sizeof(lapic_et)); lapic_et.et_name = "LAPIC"; lapic_et.et_flags = ET_FLAGS_PERIODIC | ET_FLAGS_ONESHOT | ET_FLAGS_PERCPU; lapic_et.et_quality = 600; if (!arat) { lapic_et.et_flags |= ET_FLAGS_C3STOP; lapic_et.et_quality -= 200; } lapic_et.et_frequency = 0; /* We don't know frequency yet, so trying to guess. */ lapic_et.et_min_period = 0x00001000LL; lapic_et.et_max_period = SBT_1S; lapic_et.et_start = lapic_et_start; lapic_et.et_stop = lapic_et_stop; lapic_et.et_priv = NULL; et_register(&lapic_et); } } /* * Create a local APIC instance. */ static void native_lapic_create(u_int apic_id, int boot_cpu) { int i; if (apic_id > MAX_APIC_ID) { printf("APIC: Ignoring local APIC with ID %d\n", apic_id); if (boot_cpu) panic("Can't ignore BSP"); return; } KASSERT(!lapics[apic_id].la_present, ("duplicate local APIC %u", apic_id)); /* * Assume no local LVT overrides and a cluster of 0 and * intra-cluster ID of 0. */ lapics[apic_id].la_present = 1; lapics[apic_id].la_id = apic_id; for (i = 0; i <= APIC_LVT_MAX; i++) { lapics[apic_id].la_lvts[i] = lvts[i]; lapics[apic_id].la_lvts[i].lvt_active = 0; } for (i = 0; i <= APIC_NUM_IOINTS; i++) lapics[apic_id].la_ioint_irqs[i] = -1; lapics[apic_id].la_ioint_irqs[IDT_SYSCALL - APIC_IO_INTS] = IRQ_SYSCALL; lapics[apic_id].la_ioint_irqs[APIC_TIMER_INT - APIC_IO_INTS] = IRQ_TIMER; #ifdef KDTRACE_HOOKS lapics[apic_id].la_ioint_irqs[IDT_DTRACE_RET - APIC_IO_INTS] = IRQ_DTRACE_RET; #endif #ifdef XENHVM lapics[apic_id].la_ioint_irqs[IDT_EVTCHN - APIC_IO_INTS] = IRQ_EVTCHN; #endif #ifdef SMP cpu_add(apic_id, boot_cpu); #endif } /* * Dump contents of local APIC registers */ static void native_lapic_dump(const char* str) { uint32_t maxlvt; maxlvt = (lapic->version & APIC_VER_MAXLVT) >> MAXLVTSHIFT; printf("cpu%d %s:\n", PCPU_GET(cpuid), str); printf(" ID: 0x%08x VER: 0x%08x LDR: 0x%08x DFR: 0x%08x\n", lapic->id, lapic->version, lapic->ldr, lapic->dfr); printf(" lint0: 0x%08x lint1: 0x%08x TPR: 0x%08x SVR: 0x%08x\n", lapic->lvt_lint0, lapic->lvt_lint1, lapic->tpr, lapic->svr); printf(" timer: 0x%08x therm: 0x%08x err: 0x%08x", lapic->lvt_timer, lapic->lvt_thermal, lapic->lvt_error); if (maxlvt >= APIC_LVT_PMC) printf(" pmc: 0x%08x", lapic->lvt_pcint); printf("\n"); if (maxlvt >= APIC_LVT_CMCI) printf(" cmci: 0x%08x\n", lapic->lvt_cmci); } static void native_lapic_setup(int boot) { struct lapic *la; u_int32_t maxlvt; register_t saveintr; char buf[MAXCOMLEN + 1]; la = &lapics[lapic_id()]; KASSERT(la->la_present, ("missing APIC structure")); saveintr = intr_disable(); maxlvt = (lapic->version & APIC_VER_MAXLVT) >> MAXLVTSHIFT; /* Initialize the TPR to allow all interrupts. */ lapic_set_tpr(0); /* Setup spurious vector and enable the local APIC. */ lapic_enable(); /* Program LINT[01] LVT entries. */ lapic->lvt_lint0 = lvt_mode(la, APIC_LVT_LINT0, lapic->lvt_lint0); lapic->lvt_lint1 = lvt_mode(la, APIC_LVT_LINT1, lapic->lvt_lint1); /* Program the PMC LVT entry if present. */ if (maxlvt >= APIC_LVT_PMC) lapic->lvt_pcint = lvt_mode(la, APIC_LVT_PMC, lapic->lvt_pcint); /* Program timer LVT and setup handler. */ la->lvt_timer_cache = lapic->lvt_timer = lvt_mode(la, APIC_LVT_TIMER, lapic->lvt_timer); if (boot) { snprintf(buf, sizeof(buf), "cpu%d:timer", PCPU_GET(cpuid)); intrcnt_add(buf, &la->la_timer_count); } /* Setup the timer if configured. */ if (la->la_timer_mode != 0) { KASSERT(la->la_timer_period != 0, ("lapic%u: zero divisor", lapic_id())); lapic_timer_set_divisor(lapic_timer_divisor); if (la->la_timer_mode == 1) lapic_timer_periodic(la, la->la_timer_period, 1); else lapic_timer_oneshot(la, la->la_timer_period, 1); } /* Program error LVT and clear any existing errors. */ lapic->lvt_error = lvt_mode(la, APIC_LVT_ERROR, lapic->lvt_error); lapic->esr = 0; /* XXX: Thermal LVT */ /* Program the CMCI LVT entry if present. */ if (maxlvt >= APIC_LVT_CMCI) lapic->lvt_cmci = lvt_mode(la, APIC_LVT_CMCI, lapic->lvt_cmci); intr_restore(saveintr); } static void native_lapic_reenable_pmc(void) { #ifdef HWPMC_HOOKS uint32_t value; value = lapic->lvt_pcint; value &= ~APIC_LVT_M; lapic->lvt_pcint = value; #endif } #ifdef HWPMC_HOOKS static void lapic_update_pmc(void *dummy) { struct lapic *la; la = &lapics[lapic_id()]; lapic->lvt_pcint = lvt_mode(la, APIC_LVT_PMC, lapic->lvt_pcint); } #endif static int native_lapic_enable_pmc(void) { #ifdef HWPMC_HOOKS u_int32_t maxlvt; /* Fail if the local APIC is not present. */ if (lapic == NULL) return (0); /* Fail if the PMC LVT is not present. */ maxlvt = (lapic->version & APIC_VER_MAXLVT) >> MAXLVTSHIFT; if (maxlvt < APIC_LVT_PMC) return (0); lvts[APIC_LVT_PMC].lvt_masked = 0; #ifdef SMP /* * If hwpmc was loaded at boot time then the APs may not be * started yet. In that case, don't forward the request to * them as they will program the lvt when they start. */ if (smp_started) smp_rendezvous(NULL, lapic_update_pmc, NULL, NULL); else #endif lapic_update_pmc(NULL); return (1); #else return (0); #endif } static void native_lapic_disable_pmc(void) { #ifdef HWPMC_HOOKS u_int32_t maxlvt; /* Fail if the local APIC is not present. */ if (lapic == NULL) return; /* Fail if the PMC LVT is not present. */ maxlvt = (lapic->version & APIC_VER_MAXLVT) >> MAXLVTSHIFT; if (maxlvt < APIC_LVT_PMC) return; lvts[APIC_LVT_PMC].lvt_masked = 1; #ifdef SMP /* The APs should always be started when hwpmc is unloaded. */ KASSERT(mp_ncpus == 1 || smp_started, ("hwpmc unloaded too early")); #endif smp_rendezvous(NULL, lapic_update_pmc, NULL, NULL); #endif } static int lapic_et_start(struct eventtimer *et, sbintime_t first, sbintime_t period) { struct lapic *la; u_long value; la = &lapics[PCPU_GET(apic_id)]; if (et->et_frequency == 0) { /* Start off with a divisor of 2 (power on reset default). */ lapic_timer_divisor = 2; /* Try to calibrate the local APIC timer. */ do { lapic_timer_set_divisor(lapic_timer_divisor); lapic_timer_oneshot(la, APIC_TIMER_MAX_COUNT, 0); DELAY(1000000); value = APIC_TIMER_MAX_COUNT - lapic->ccr_timer; if (value != APIC_TIMER_MAX_COUNT) break; lapic_timer_divisor <<= 1; } while (lapic_timer_divisor <= 128); if (lapic_timer_divisor > 128) panic("lapic: Divisor too big"); if (bootverbose) printf("lapic: Divisor %lu, Frequency %lu Hz\n", lapic_timer_divisor, value); et->et_frequency = value; et->et_min_period = (0x00000002LLU << 32) / et->et_frequency; et->et_max_period = (0xfffffffeLLU << 32) / et->et_frequency; } if (la->la_timer_mode == 0) lapic_timer_set_divisor(lapic_timer_divisor); if (period != 0) { la->la_timer_mode = 1; la->la_timer_period = ((uint32_t)et->et_frequency * period) >> 32; lapic_timer_periodic(la, la->la_timer_period, 1); } else { la->la_timer_mode = 2; la->la_timer_period = ((uint32_t)et->et_frequency * first) >> 32; lapic_timer_oneshot(la, la->la_timer_period, 1); } return (0); } static int lapic_et_stop(struct eventtimer *et) { struct lapic *la = &lapics[PCPU_GET(apic_id)]; la->la_timer_mode = 0; lapic_timer_stop(la); return (0); } static void native_lapic_disable(void) { uint32_t value; /* Software disable the local APIC. */ value = lapic->svr; value &= ~APIC_SVR_SWEN; lapic->svr = value; } static void lapic_enable(void) { u_int32_t value; /* Program the spurious vector to enable the local APIC. */ value = lapic->svr; value &= ~(APIC_SVR_VECTOR | APIC_SVR_FOCUS); value |= (APIC_SVR_FEN | APIC_SVR_SWEN | APIC_SPURIOUS_INT); lapic->svr = value; } /* Reset the local APIC on the BSP during resume. */ static void lapic_resume(struct pic *pic, bool suspend_cancelled) { lapic_setup(0); } static int native_lapic_id(void) { KASSERT(lapic != NULL, ("local APIC is not mapped")); return (lapic->id >> APIC_ID_SHIFT); } static int native_lapic_intr_pending(u_int vector) { volatile u_int32_t *irr; /* * The IRR registers are an array of 128-bit registers each of * which only describes 32 interrupts in the low 32 bits.. Thus, * we divide the vector by 32 to get the 128-bit index. We then * multiply that index by 4 to get the equivalent index from * treating the IRR as an array of 32-bit registers. Finally, we * modulus the vector by 32 to determine the individual bit to * test. */ irr = &lapic->irr0; return (irr[(vector / 32) * 4] & 1 << (vector % 32)); } static void native_lapic_set_logical_id(u_int apic_id, u_int cluster, u_int cluster_id) { struct lapic *la; KASSERT(lapics[apic_id].la_present, ("%s: APIC %u doesn't exist", __func__, apic_id)); KASSERT(cluster <= APIC_MAX_CLUSTER, ("%s: cluster %u too big", __func__, cluster)); KASSERT(cluster_id <= APIC_MAX_INTRACLUSTER_ID, ("%s: intra cluster id %u too big", __func__, cluster_id)); la = &lapics[apic_id]; la->la_cluster = cluster; la->la_cluster_id = cluster_id; } static int native_lapic_set_lvt_mask(u_int apic_id, u_int pin, u_char masked) { if (pin > APIC_LVT_MAX) return (EINVAL); if (apic_id == APIC_ID_ALL) { lvts[pin].lvt_masked = masked; if (bootverbose) printf("lapic:"); } else { KASSERT(lapics[apic_id].la_present, ("%s: missing APIC %u", __func__, apic_id)); lapics[apic_id].la_lvts[pin].lvt_masked = masked; lapics[apic_id].la_lvts[pin].lvt_active = 1; if (bootverbose) printf("lapic%u:", apic_id); } if (bootverbose) printf(" LINT%u %s\n", pin, masked ? "masked" : "unmasked"); return (0); } static int native_lapic_set_lvt_mode(u_int apic_id, u_int pin, u_int32_t mode) { struct lvt *lvt; if (pin > APIC_LVT_MAX) return (EINVAL); if (apic_id == APIC_ID_ALL) { lvt = &lvts[pin]; if (bootverbose) printf("lapic:"); } else { KASSERT(lapics[apic_id].la_present, ("%s: missing APIC %u", __func__, apic_id)); lvt = &lapics[apic_id].la_lvts[pin]; lvt->lvt_active = 1; if (bootverbose) printf("lapic%u:", apic_id); } lvt->lvt_mode = mode; switch (mode) { case APIC_LVT_DM_NMI: case APIC_LVT_DM_SMI: case APIC_LVT_DM_INIT: case APIC_LVT_DM_EXTINT: lvt->lvt_edgetrigger = 1; lvt->lvt_activehi = 1; if (mode == APIC_LVT_DM_EXTINT) lvt->lvt_masked = 1; else lvt->lvt_masked = 0; break; default: panic("Unsupported delivery mode: 0x%x\n", mode); } if (bootverbose) { printf(" Routing "); switch (mode) { case APIC_LVT_DM_NMI: printf("NMI"); break; case APIC_LVT_DM_SMI: printf("SMI"); break; case APIC_LVT_DM_INIT: printf("INIT"); break; case APIC_LVT_DM_EXTINT: printf("ExtINT"); break; } printf(" -> LINT%u\n", pin); } return (0); } static int native_lapic_set_lvt_polarity(u_int apic_id, u_int pin, enum intr_polarity pol) { if (pin > APIC_LVT_MAX || pol == INTR_POLARITY_CONFORM) return (EINVAL); if (apic_id == APIC_ID_ALL) { lvts[pin].lvt_activehi = (pol == INTR_POLARITY_HIGH); if (bootverbose) printf("lapic:"); } else { KASSERT(lapics[apic_id].la_present, ("%s: missing APIC %u", __func__, apic_id)); lapics[apic_id].la_lvts[pin].lvt_active = 1; lapics[apic_id].la_lvts[pin].lvt_activehi = (pol == INTR_POLARITY_HIGH); if (bootverbose) printf("lapic%u:", apic_id); } if (bootverbose) printf(" LINT%u polarity: %s\n", pin, pol == INTR_POLARITY_HIGH ? "high" : "low"); return (0); } static int native_lapic_set_lvt_triggermode(u_int apic_id, u_int pin, enum intr_trigger trigger) { if (pin > APIC_LVT_MAX || trigger == INTR_TRIGGER_CONFORM) return (EINVAL); if (apic_id == APIC_ID_ALL) { lvts[pin].lvt_edgetrigger = (trigger == INTR_TRIGGER_EDGE); if (bootverbose) printf("lapic:"); } else { KASSERT(lapics[apic_id].la_present, ("%s: missing APIC %u", __func__, apic_id)); lapics[apic_id].la_lvts[pin].lvt_edgetrigger = (trigger == INTR_TRIGGER_EDGE); lapics[apic_id].la_lvts[pin].lvt_active = 1; if (bootverbose) printf("lapic%u:", apic_id); } if (bootverbose) printf(" LINT%u trigger: %s\n", pin, trigger == INTR_TRIGGER_EDGE ? "edge" : "level"); return (0); } /* * Adjust the TPR of the current CPU so that it blocks all interrupts below * the passed in vector. */ static void lapic_set_tpr(u_int vector) { #ifdef CHEAP_TPR lapic->tpr = vector; #else u_int32_t tpr; tpr = lapic->tpr & ~APIC_TPR_PRIO; tpr |= vector; lapic->tpr = tpr; #endif } static void native_lapic_eoi(void) { lapic->eoi = 0; } void lapic_handle_intr(int vector, struct trapframe *frame) { struct intsrc *isrc; isrc = intr_lookup_source(apic_idt_to_irq(PCPU_GET(apic_id), vector)); intr_execute_handlers(isrc, frame); } void lapic_handle_timer(struct trapframe *frame) { struct lapic *la; struct trapframe *oldframe; struct thread *td; /* Send EOI first thing. */ lapic_eoi(); #if defined(SMP) && !defined(SCHED_ULE) /* * Don't do any accounting for the disabled HTT cores, since it * will provide misleading numbers for the userland. * * No locking is necessary here, since even if we lose the race * when hlt_cpus_mask changes it is not a big deal, really. * * Don't do that for ULE, since ULE doesn't consider hlt_cpus_mask * and unlike other schedulers it actually schedules threads to * those CPUs. */ if (CPU_ISSET(PCPU_GET(cpuid), &hlt_cpus_mask)) return; #endif /* Look up our local APIC structure for the tick counters. */ la = &lapics[PCPU_GET(apic_id)]; (*la->la_timer_count)++; critical_enter(); if (lapic_et.et_active) { td = curthread; td->td_intr_nesting_level++; oldframe = td->td_intr_frame; td->td_intr_frame = frame; lapic_et.et_event_cb(&lapic_et, lapic_et.et_arg); td->td_intr_frame = oldframe; td->td_intr_nesting_level--; } critical_exit(); } static void lapic_timer_set_divisor(u_int divisor) { KASSERT(powerof2(divisor), ("lapic: invalid divisor %u", divisor)); KASSERT(ffs(divisor) <= sizeof(lapic_timer_divisors) / sizeof(u_int32_t), ("lapic: invalid divisor %u", divisor)); lapic->dcr_timer = lapic_timer_divisors[ffs(divisor) - 1]; } static void lapic_timer_oneshot(struct lapic *la, u_int count, int enable_int) { u_int32_t value; value = la->lvt_timer_cache; value &= ~APIC_LVTT_TM; value |= APIC_LVTT_TM_ONE_SHOT; if (enable_int) value &= ~APIC_LVT_M; lapic->lvt_timer = value; lapic->icr_timer = count; } static void lapic_timer_periodic(struct lapic *la, u_int count, int enable_int) { u_int32_t value; value = la->lvt_timer_cache; value &= ~APIC_LVTT_TM; value |= APIC_LVTT_TM_PERIODIC; if (enable_int) value &= ~APIC_LVT_M; lapic->lvt_timer = value; lapic->icr_timer = count; } static void lapic_timer_stop(struct lapic *la) { u_int32_t value; value = la->lvt_timer_cache; value &= ~APIC_LVTT_TM; value |= APIC_LVT_M; lapic->lvt_timer = value; } void lapic_handle_cmc(void) { lapic_eoi(); cmc_intr(); } /* * Called from the mca_init() to activate the CMC interrupt if this CPU is * responsible for monitoring any MC banks for CMC events. Since mca_init() * is called prior to lapic_setup() during boot, this just needs to unmask * this CPU's LVT_CMCI entry. */ static void native_lapic_enable_cmc(void) { u_int apic_id; #ifdef DEV_ATPIC if (lapic == NULL) return; #endif apic_id = PCPU_GET(apic_id); KASSERT(lapics[apic_id].la_present, ("%s: missing APIC %u", __func__, apic_id)); lapics[apic_id].la_lvts[APIC_LVT_CMCI].lvt_masked = 0; lapics[apic_id].la_lvts[APIC_LVT_CMCI].lvt_active = 1; if (bootverbose) printf("lapic%u: CMCI unmasked\n", apic_id); } void lapic_handle_error(void) { u_int32_t esr; /* * Read the contents of the error status register. Write to * the register first before reading from it to force the APIC * to update its value to indicate any errors that have * occurred since the previous write to the register. */ lapic->esr = 0; esr = lapic->esr; printf("CPU%d: local APIC error 0x%x\n", PCPU_GET(cpuid), esr); lapic_eoi(); } static u_int native_apic_cpuid(u_int apic_id) { #ifdef SMP return apic_cpuids[apic_id]; #else return 0; #endif } /* Request a free IDT vector to be used by the specified IRQ. */ static u_int native_apic_alloc_vector(u_int apic_id, u_int irq) { u_int vector; KASSERT(irq < NUM_IO_INTS, ("Invalid IRQ %u", irq)); /* * Search for a free vector. Currently we just use a very simple * algorithm to find the first free vector. */ mtx_lock_spin(&icu_lock); for (vector = 0; vector < APIC_NUM_IOINTS; vector++) { if (lapics[apic_id].la_ioint_irqs[vector] != -1) continue; lapics[apic_id].la_ioint_irqs[vector] = irq; mtx_unlock_spin(&icu_lock); return (vector + APIC_IO_INTS); } mtx_unlock_spin(&icu_lock); return (0); } /* * Request 'count' free contiguous IDT vectors to be used by 'count' * IRQs. 'count' must be a power of two and the vectors will be * aligned on a boundary of 'align'. If the request cannot be * satisfied, 0 is returned. */ static u_int native_apic_alloc_vectors(u_int apic_id, u_int *irqs, u_int count, u_int align) { u_int first, run, vector; KASSERT(powerof2(count), ("bad count")); KASSERT(powerof2(align), ("bad align")); KASSERT(align >= count, ("align < count")); #ifdef INVARIANTS for (run = 0; run < count; run++) KASSERT(irqs[run] < NUM_IO_INTS, ("Invalid IRQ %u at index %u", irqs[run], run)); #endif /* * Search for 'count' free vectors. As with apic_alloc_vector(), * this just uses a simple first fit algorithm. */ run = 0; first = 0; mtx_lock_spin(&icu_lock); for (vector = 0; vector < APIC_NUM_IOINTS; vector++) { /* Vector is in use, end run. */ if (lapics[apic_id].la_ioint_irqs[vector] != -1) { run = 0; first = 0; continue; } /* Start a new run if run == 0 and vector is aligned. */ if (run == 0) { if ((vector & (align - 1)) != 0) continue; first = vector; } run++; /* Keep looping if the run isn't long enough yet. */ if (run < count) continue; /* Found a run, assign IRQs and return the first vector. */ for (vector = 0; vector < count; vector++) lapics[apic_id].la_ioint_irqs[first + vector] = irqs[vector]; mtx_unlock_spin(&icu_lock); return (first + APIC_IO_INTS); } mtx_unlock_spin(&icu_lock); printf("APIC: Couldn't find APIC vectors for %u IRQs\n", count); return (0); } /* * Enable a vector for a particular apic_id. Since all lapics share idt * entries and ioint_handlers this enables the vector on all lapics. lapics * which do not have the vector configured would report spurious interrupts * should it fire. */ static void native_apic_enable_vector(u_int apic_id, u_int vector) { KASSERT(vector != IDT_SYSCALL, ("Attempt to overwrite syscall entry")); KASSERT(ioint_handlers[vector / 32] != NULL, ("No ISR handler for vector %u", vector)); #ifdef KDTRACE_HOOKS KASSERT(vector != IDT_DTRACE_RET, ("Attempt to overwrite DTrace entry")); #endif setidt(vector, ioint_handlers[vector / 32], SDT_APIC, SEL_KPL, GSEL_APIC); } static void native_apic_disable_vector(u_int apic_id, u_int vector) { KASSERT(vector != IDT_SYSCALL, ("Attempt to overwrite syscall entry")); #ifdef KDTRACE_HOOKS KASSERT(vector != IDT_DTRACE_RET, ("Attempt to overwrite DTrace entry")); #endif KASSERT(ioint_handlers[vector / 32] != NULL, ("No ISR handler for vector %u", vector)); #ifdef notyet /* * We can not currently clear the idt entry because other cpus * may have a valid vector at this offset. */ setidt(vector, &IDTVEC(rsvd), SDT_APICT, SEL_KPL, GSEL_APIC); #endif } /* Release an APIC vector when it's no longer in use. */ static void native_apic_free_vector(u_int apic_id, u_int vector, u_int irq) { struct thread *td; KASSERT(vector >= APIC_IO_INTS && vector != IDT_SYSCALL && vector <= APIC_IO_INTS + APIC_NUM_IOINTS, ("Vector %u does not map to an IRQ line", vector)); KASSERT(irq < NUM_IO_INTS, ("Invalid IRQ %u", irq)); KASSERT(lapics[apic_id].la_ioint_irqs[vector - APIC_IO_INTS] == irq, ("IRQ mismatch")); #ifdef KDTRACE_HOOKS KASSERT(vector != IDT_DTRACE_RET, ("Attempt to overwrite DTrace entry")); #endif /* * Bind us to the cpu that owned the vector before freeing it so * we don't lose an interrupt delivery race. */ td = curthread; if (!rebooting) { thread_lock(td); if (sched_is_bound(td)) panic("apic_free_vector: Thread already bound.\n"); sched_bind(td, apic_cpuid(apic_id)); thread_unlock(td); } mtx_lock_spin(&icu_lock); lapics[apic_id].la_ioint_irqs[vector - APIC_IO_INTS] = -1; mtx_unlock_spin(&icu_lock); if (!rebooting) { thread_lock(td); sched_unbind(td); thread_unlock(td); } } /* Map an IDT vector (APIC) to an IRQ (interrupt source). */ static u_int apic_idt_to_irq(u_int apic_id, u_int vector) { int irq; KASSERT(vector >= APIC_IO_INTS && vector != IDT_SYSCALL && vector <= APIC_IO_INTS + APIC_NUM_IOINTS, ("Vector %u does not map to an IRQ line", vector)); #ifdef KDTRACE_HOOKS KASSERT(vector != IDT_DTRACE_RET, ("Attempt to overwrite DTrace entry")); #endif irq = lapics[apic_id].la_ioint_irqs[vector - APIC_IO_INTS]; if (irq < 0) irq = 0; return (irq); } #ifdef DDB /* * Dump data about APIC IDT vector mappings. */ DB_SHOW_COMMAND(apic, db_show_apic) { struct intsrc *isrc; int i, verbose; u_int apic_id; u_int irq; if (strcmp(modif, "vv") == 0) verbose = 2; else if (strcmp(modif, "v") == 0) verbose = 1; else verbose = 0; for (apic_id = 0; apic_id <= MAX_APIC_ID; apic_id++) { if (lapics[apic_id].la_present == 0) continue; db_printf("Interrupts bound to lapic %u\n", apic_id); for (i = 0; i < APIC_NUM_IOINTS + 1 && !db_pager_quit; i++) { irq = lapics[apic_id].la_ioint_irqs[i]; if (irq == -1 || irq == IRQ_SYSCALL) continue; #ifdef KDTRACE_HOOKS if (irq == IRQ_DTRACE_RET) continue; #endif #ifdef XENHVM if (irq == IRQ_EVTCHN) continue; #endif db_printf("vec 0x%2x -> ", i + APIC_IO_INTS); if (irq == IRQ_TIMER) db_printf("lapic timer\n"); else if (irq < NUM_IO_INTS) { isrc = intr_lookup_source(irq); if (isrc == NULL || verbose == 0) db_printf("IRQ %u\n", irq); else db_dump_intr_event(isrc->is_event, verbose == 2); } else db_printf("IRQ %u ???\n", irq); } } } static void dump_mask(const char *prefix, uint32_t v, int base) { int i, first; first = 1; for (i = 0; i < 32; i++) if (v & (1 << i)) { if (first) { db_printf("%s:", prefix); first = 0; } db_printf(" %02x", base + i); } if (!first) db_printf("\n"); } /* Show info from the lapic regs for this CPU. */ DB_SHOW_COMMAND(lapic, db_show_lapic) { uint32_t v; db_printf("lapic ID = %d\n", lapic_id()); v = lapic->version; db_printf("version = %d.%d\n", (v & APIC_VER_VERSION) >> 4, v & 0xf); db_printf("max LVT = %d\n", (v & APIC_VER_MAXLVT) >> MAXLVTSHIFT); v = lapic->svr; db_printf("SVR = %02x (%s)\n", v & APIC_SVR_VECTOR, v & APIC_SVR_ENABLE ? "enabled" : "disabled"); db_printf("TPR = %02x\n", lapic->tpr); #define dump_field(prefix, index) \ dump_mask(__XSTRING(prefix ## index), lapic->prefix ## index, \ index * 32) db_printf("In-service Interrupts:\n"); dump_field(isr, 0); dump_field(isr, 1); dump_field(isr, 2); dump_field(isr, 3); dump_field(isr, 4); dump_field(isr, 5); dump_field(isr, 6); dump_field(isr, 7); db_printf("TMR Interrupts:\n"); dump_field(tmr, 0); dump_field(tmr, 1); dump_field(tmr, 2); dump_field(tmr, 3); dump_field(tmr, 4); dump_field(tmr, 5); dump_field(tmr, 6); dump_field(tmr, 7); db_printf("IRR Interrupts:\n"); dump_field(irr, 0); dump_field(irr, 1); dump_field(irr, 2); dump_field(irr, 3); dump_field(irr, 4); dump_field(irr, 5); dump_field(irr, 6); dump_field(irr, 7); #undef dump_field } #endif /* * APIC probing support code. This includes code to manage enumerators. */ static SLIST_HEAD(, apic_enumerator) enumerators = SLIST_HEAD_INITIALIZER(enumerators); static struct apic_enumerator *best_enum; void apic_register_enumerator(struct apic_enumerator *enumerator) { #ifdef INVARIANTS struct apic_enumerator *apic_enum; SLIST_FOREACH(apic_enum, &enumerators, apic_next) { if (apic_enum == enumerator) panic("%s: Duplicate register of %s", __func__, enumerator->apic_name); } #endif SLIST_INSERT_HEAD(&enumerators, enumerator, apic_next); } /* * We have to look for CPU's very, very early because certain subsystems * want to know how many CPU's we have extremely early on in the boot * process. */ static void apic_init(void *dummy __unused) { struct apic_enumerator *enumerator; -#ifndef __amd64__ - uint64_t apic_base; -#endif int retval, best; /* We only support built in local APICs. */ if (!(cpu_feature & CPUID_APIC)) return; /* Don't probe if APIC mode is disabled. */ if (resource_disabled("apic", 0)) return; /* Probe all the enumerators to find the best match. */ best_enum = NULL; best = 0; SLIST_FOREACH(enumerator, &enumerators, apic_next) { retval = enumerator->apic_probe(); if (retval > 0) continue; if (best_enum == NULL || best < retval) { best_enum = enumerator; best = retval; } } if (best_enum == NULL) { if (bootverbose) printf("APIC: Could not find any APICs.\n"); #ifndef DEV_ATPIC panic("running without device atpic requires a local APIC"); #endif return; } if (bootverbose) printf("APIC: Using the %s enumerator.\n", best_enum->apic_name); #ifndef __amd64__ /* * To work around an errata, we disable the local APIC on some * CPUs during early startup. We need to turn the local APIC back * on on such CPUs now. */ - if (cpu == CPU_686 && cpu_vendor_id == CPU_VENDOR_INTEL && - (cpu_id & 0xff0) == 0x610) { - apic_base = rdmsr(MSR_APICBASE); - apic_base |= APICBASE_ENABLED; - wrmsr(MSR_APICBASE, apic_base); - } + ppro_reenable_apic(); #endif /* Probe the CPU's in the system. */ retval = best_enum->apic_probe_cpus(); if (retval != 0) printf("%s: Failed to probe CPUs: returned %d\n", best_enum->apic_name, retval); } SYSINIT(apic_init, SI_SUB_TUNABLES - 1, SI_ORDER_SECOND, apic_init, NULL); /* * Setup the local APIC. We have to do this prior to starting up the APs * in the SMP case. */ static void apic_setup_local(void *dummy __unused) { int retval; if (best_enum == NULL) return; /* Initialize the local APIC. */ retval = best_enum->apic_setup_local(); if (retval != 0) printf("%s: Failed to setup the local APIC: returned %d\n", best_enum->apic_name, retval); } SYSINIT(apic_setup_local, SI_SUB_CPU, SI_ORDER_SECOND, apic_setup_local, NULL); /* * Setup the I/O APICs. */ static void apic_setup_io(void *dummy __unused) { int retval; if (best_enum == NULL) return; /* * Local APIC must be registered before other PICs and pseudo PICs * for proper suspend/resume order. */ #ifndef XEN intr_register_pic(&lapic_pic); #endif retval = best_enum->apic_setup_io(); if (retval != 0) printf("%s: Failed to setup I/O APICs: returned %d\n", best_enum->apic_name, retval); #ifdef XEN return; #endif /* * Finish setting up the local APIC on the BSP once we know how to * properly program the LINT pins. */ lapic_setup(1); if (bootverbose) lapic_dump("BSP"); /* Enable the MSI "pic". */ msi_init(); } SYSINIT(apic_setup_io, SI_SUB_INTR, SI_ORDER_THIRD, apic_setup_io, NULL); #ifdef SMP /* * Inter Processor Interrupt functions. The lapic_ipi_*() functions are * private to the MD code. The public interface for the rest of the * kernel is defined in mp_machdep.c. */ static int native_lapic_ipi_wait(int delay) { int x, incr; /* * Wait delay loops for IPI to be sent. This is highly bogus * since this is sensitive to CPU clock speed. If delay is * -1, we wait forever. */ if (delay == -1) { incr = 0; delay = 1; } else incr = 1; for (x = 0; x < delay; x += incr) { if ((lapic->icr_lo & APIC_DELSTAT_MASK) == APIC_DELSTAT_IDLE) return (1); ia32_pause(); } return (0); } static void native_lapic_ipi_raw(register_t icrlo, u_int dest) { register_t value, saveintr; /* XXX: Need more sanity checking of icrlo? */ KASSERT(lapic != NULL, ("%s called too early", __func__)); KASSERT((dest & ~(APIC_ID_MASK >> APIC_ID_SHIFT)) == 0, ("%s: invalid dest field", __func__)); KASSERT((icrlo & APIC_ICRLO_RESV_MASK) == 0, ("%s: reserved bits set in ICR LO register", __func__)); /* Set destination in ICR HI register if it is being used. */ saveintr = intr_disable(); if ((icrlo & APIC_DEST_MASK) == APIC_DEST_DESTFLD) { value = lapic->icr_hi; value &= ~APIC_ID_MASK; value |= dest << APIC_ID_SHIFT; lapic->icr_hi = value; } /* Program the contents of the IPI and dispatch it. */ value = lapic->icr_lo; value &= APIC_ICRLO_RESV_MASK; value |= icrlo; lapic->icr_lo = value; intr_restore(saveintr); } #define BEFORE_SPIN 1000000 #ifdef DETECT_DEADLOCK #define AFTER_SPIN 1000 #endif static void native_lapic_ipi_vectored(u_int vector, int dest) { register_t icrlo, destfield; KASSERT((vector & ~APIC_VECTOR_MASK) == 0, ("%s: invalid vector %d", __func__, vector)); icrlo = APIC_DESTMODE_PHY | APIC_TRIGMOD_EDGE; /* * IPI_STOP_HARD is just a "fake" vector used to send a NMI. * Use special rules regard NMI if passed, otherwise specify * the vector. */ if (vector == IPI_STOP_HARD) icrlo |= APIC_DELMODE_NMI | APIC_LEVEL_ASSERT; else icrlo |= vector | APIC_DELMODE_FIXED | APIC_LEVEL_DEASSERT; destfield = 0; switch (dest) { case APIC_IPI_DEST_SELF: icrlo |= APIC_DEST_SELF; break; case APIC_IPI_DEST_ALL: icrlo |= APIC_DEST_ALLISELF; break; case APIC_IPI_DEST_OTHERS: icrlo |= APIC_DEST_ALLESELF; break; default: KASSERT((dest & ~(APIC_ID_MASK >> APIC_ID_SHIFT)) == 0, ("%s: invalid destination 0x%x", __func__, dest)); destfield = dest; } /* Wait for an earlier IPI to finish. */ if (!lapic_ipi_wait(BEFORE_SPIN)) { if (panicstr != NULL) return; else panic("APIC: Previous IPI is stuck"); } lapic_ipi_raw(icrlo, destfield); #ifdef DETECT_DEADLOCK /* Wait for IPI to be delivered. */ if (!lapic_ipi_wait(AFTER_SPIN)) { #ifdef needsattention /* * XXX FIXME: * * The above function waits for the message to actually be * delivered. It breaks out after an arbitrary timeout * since the message should eventually be delivered (at * least in theory) and that if it wasn't we would catch * the failure with the check above when the next IPI is * sent. * * We could skip this wait entirely, EXCEPT it probably * protects us from other routines that assume that the * message was delivered and acted upon when this function * returns. */ printf("APIC: IPI might be stuck\n"); #else /* !needsattention */ /* Wait until mesage is sent without a timeout. */ while (lapic->icr_lo & APIC_DELSTAT_PEND) ia32_pause(); #endif /* needsattention */ } #endif /* DETECT_DEADLOCK */ } #endif /* SMP */