The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/amd64/amd64/cpu_switch.S

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 2003 Peter Wemm.
    3  * Copyright (c) 1990 The Regents of the University of California.
    4  * All rights reserved.
    5  *
    6  * This code is derived from software contributed to Berkeley by
    7  * William Jolitz.
    8  *
    9  * Redistribution and use in source and binary forms, with or without
   10  * modification, are permitted provided that the following conditions
   11  * are met:
   12  * 1. Redistributions of source code must retain the above copyright
   13  *    notice, this list of conditions and the following disclaimer.
   14  * 2. Redistributions in binary form must reproduce the above copyright
   15  *    notice, this list of conditions and the following disclaimer in the
   16  *    documentation and/or other materials provided with the distribution.
   17  * 4. Neither the name of the University nor the names of its contributors
   18  *    may be used to endorse or promote products derived from this software
   19  *    without specific prior written permission.
   20  *
   21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   31  * SUCH DAMAGE.
   32  *
   33  * $FreeBSD: releng/8.2/sys/amd64/amd64/cpu_switch.S 215513 2010-11-19 09:49:14Z kib $
   34  */
   35 
   36 #include <machine/asmacros.h>
   37 #include <machine/specialreg.h>
   38 
   39 #include "assym.s"
   40 #include "opt_sched.h"
   41 
   42 /*****************************************************************************/
   43 /* Scheduling                                                                */
   44 /*****************************************************************************/
   45 
   46         .text
   47 
   48 #ifdef SMP
   49 #define LK      lock ;
   50 #else
   51 #define LK
   52 #endif
   53 
   54 #if defined(SCHED_ULE) && defined(SMP)
   55 #define SETLK   xchgq
   56 #else
   57 #define SETLK   movq
   58 #endif
   59 
   60 /*
   61  * cpu_throw()
   62  *
   63  * This is the second half of cpu_switch(). It is used when the current
   64  * thread is either a dummy or slated to die, and we no longer care
   65  * about its state.  This is only a slight optimization and is probably
   66  * not worth it anymore.  Note that we need to clear the pm_active bits so
   67  * we do need the old proc if it still exists.
   68  * %rdi = oldtd
   69  * %rsi = newtd
   70  */
   71 ENTRY(cpu_throw)
   72         movl    PCPU(CPUID),%eax
   73         testq   %rdi,%rdi
   74         jz      1f
   75         /* release bit from old pm_active */
   76         movq    PCPU(CURPMAP),%rdx
   77         LK btrl %eax,PM_ACTIVE(%rdx)            /* clear old */
   78 1:
   79         movq    TD_PCB(%rsi),%r8                /* newtd->td_proc */
   80         movq    PCB_CR3(%r8),%rdx
   81         movq    %rdx,%cr3                       /* new address space */
   82         jmp     swact
   83 END(cpu_throw)
   84 
   85 /*
   86  * cpu_switch(old, new, mtx)
   87  *
   88  * Save the current thread state, then select the next thread to run
   89  * and load its state.
   90  * %rdi = oldtd
   91  * %rsi = newtd
   92  * %rdx = mtx
   93  */
   94 ENTRY(cpu_switch)
   95         /* Switch to new thread.  First, save context. */
   96         movq    TD_PCB(%rdi),%r8
   97         movb    $1,PCB_FULL_IRET(%r8)
   98 
   99         movq    (%rsp),%rax                     /* Hardware registers */
  100         movq    %r15,PCB_R15(%r8)
  101         movq    %r14,PCB_R14(%r8)
  102         movq    %r13,PCB_R13(%r8)
  103         movq    %r12,PCB_R12(%r8)
  104         movq    %rbp,PCB_RBP(%r8)
  105         movq    %rsp,PCB_RSP(%r8)
  106         movq    %rbx,PCB_RBX(%r8)
  107         movq    %rax,PCB_RIP(%r8)
  108 
  109         testl   $PCB_DBREGS,PCB_FLAGS(%r8)
  110         jnz     store_dr                        /* static predict not taken */
  111 done_store_dr:
  112 
  113         /* have we used fp, and need a save? */
  114         cmpq    %rdi,PCPU(FPCURTHREAD)
  115         jne     1f
  116         movq    PCB_SAVEFPU(%r8),%r8
  117         clts
  118         fxsave  (%r8)
  119         smsw    %ax
  120         orb     $CR0_TS,%al
  121         lmsw    %ax
  122         xorl    %eax,%eax
  123         movq    %rax,PCPU(FPCURTHREAD)
  124 1:
  125 
  126         /* Save is done.  Now fire up new thread. Leave old vmspace. */
  127         movq    TD_PCB(%rsi),%r8
  128 
  129         /* switch address space */
  130         movq    PCB_CR3(%r8),%rcx
  131         movq    %cr3,%rax
  132         cmpq    %rcx,%rax                       /* Same address space? */
  133         jne     swinact
  134         SETLK   %rdx, TD_LOCK(%rdi)             /* Release the old thread */
  135         jmp     sw1
  136 swinact:
  137         movq    %rcx,%cr3                       /* new address space */
  138         movl    PCPU(CPUID), %eax
  139         /* Release bit from old pmap->pm_active */
  140         movq    PCPU(CURPMAP),%rcx
  141         LK btrl %eax,PM_ACTIVE(%rcx)            /* clear old */
  142         SETLK   %rdx, TD_LOCK(%rdi)             /* Release the old thread */
  143 swact:
  144         /* Set bit in new pmap->pm_active */
  145         movq    TD_PROC(%rsi),%rdx              /* newproc */
  146         movq    P_VMSPACE(%rdx), %rdx
  147         addq    $VM_PMAP,%rdx
  148         LK btsl %eax,PM_ACTIVE(%rdx)            /* set new */
  149         movq    %rdx,PCPU(CURPMAP)
  150 
  151 sw1:
  152 #if defined(SCHED_ULE) && defined(SMP)
  153         /* Wait for the new thread to become unblocked */
  154         movq    $blocked_lock, %rdx
  155 1:
  156         movq    TD_LOCK(%rsi),%rcx
  157         cmpq    %rcx, %rdx
  158         pause
  159         je      1b
  160 #endif
  161         /*
  162          * At this point, we've switched address spaces and are ready
  163          * to load up the rest of the next context.
  164          */
  165 
  166         /* Skip loading user fsbase/gsbase for kthreads */
  167         testl   $TDP_KTHREAD,TD_PFLAGS(%rsi)
  168         jnz     do_kthread
  169 
  170         /*
  171          * Load ldt register
  172          */
  173         movq    TD_PROC(%rsi),%rcx
  174         cmpq    $0, P_MD+MD_LDT(%rcx)
  175         jne     do_ldt
  176         xorl    %eax,%eax
  177 ld_ldt: lldt    %ax
  178 
  179         /* Restore fs base in GDT */
  180         movl    PCB_FSBASE(%r8),%eax
  181         movq    PCPU(FS32P),%rdx
  182         movw    %ax,2(%rdx)
  183         shrl    $16,%eax
  184         movb    %al,4(%rdx)
  185         shrl    $8,%eax
  186         movb    %al,7(%rdx)
  187 
  188         /* Restore gs base in GDT */
  189         movl    PCB_GSBASE(%r8),%eax
  190         movq    PCPU(GS32P),%rdx
  191         movw    %ax,2(%rdx)
  192         shrl    $16,%eax
  193         movb    %al,4(%rdx)
  194         shrl    $8,%eax
  195         movb    %al,7(%rdx)
  196 
  197 do_kthread:
  198         /* Do we need to reload tss ? */
  199         movq    PCPU(TSSP),%rax
  200         movq    PCB_TSSP(%r8),%rdx
  201         testq   %rdx,%rdx
  202         cmovzq  PCPU(COMMONTSSP),%rdx
  203         cmpq    %rax,%rdx
  204         jne     do_tss
  205 done_tss:
  206         movq    %r8,PCPU(RSP0)
  207         movq    %r8,PCPU(CURPCB)
  208         /* Update the TSS_RSP0 pointer for the next interrupt */
  209         movq    %r8,COMMON_TSS_RSP0(%rdx)
  210         movq    %rsi,PCPU(CURTHREAD)            /* into next thread */
  211 
  212         /* Test if debug registers should be restored. */
  213         testl   $PCB_DBREGS,PCB_FLAGS(%r8)
  214         jnz     load_dr                         /* static predict not taken */
  215 done_load_dr:
  216 
  217         /* Restore context. */
  218         movq    PCB_R15(%r8),%r15
  219         movq    PCB_R14(%r8),%r14
  220         movq    PCB_R13(%r8),%r13
  221         movq    PCB_R12(%r8),%r12
  222         movq    PCB_RBP(%r8),%rbp
  223         movq    PCB_RSP(%r8),%rsp
  224         movq    PCB_RBX(%r8),%rbx
  225         movq    PCB_RIP(%r8),%rax
  226         movq    %rax,(%rsp)
  227         ret
  228 
  229         /*
  230          * We order these strangely for several reasons.
  231          * 1: I wanted to use static branch prediction hints
  232          * 2: Most athlon64/opteron cpus don't have them.  They define
  233          *    a forward branch as 'predict not taken'.  Intel cores have
  234          *    the 'rep' prefix to invert this.
  235          * So, to make it work on both forms of cpu we do the detour.
  236          * We use jumps rather than call in order to avoid the stack.
  237          */
  238 
  239 store_dr:
  240         movq    %dr7,%rax                       /* yes, do the save */
  241         movq    %dr0,%r15
  242         movq    %dr1,%r14
  243         movq    %dr2,%r13
  244         movq    %dr3,%r12
  245         movq    %dr6,%r11
  246         movq    %r15,PCB_DR0(%r8)
  247         movq    %r14,PCB_DR1(%r8)
  248         movq    %r13,PCB_DR2(%r8)
  249         movq    %r12,PCB_DR3(%r8)
  250         movq    %r11,PCB_DR6(%r8)
  251         movq    %rax,PCB_DR7(%r8)
  252         andq    $0x0000fc00, %rax               /* disable all watchpoints */
  253         movq    %rax,%dr7
  254         jmp     done_store_dr
  255 
  256 load_dr:
  257         movq    %dr7,%rax
  258         movq    PCB_DR0(%r8),%r15
  259         movq    PCB_DR1(%r8),%r14
  260         movq    PCB_DR2(%r8),%r13
  261         movq    PCB_DR3(%r8),%r12
  262         movq    PCB_DR6(%r8),%r11
  263         movq    PCB_DR7(%r8),%rcx
  264         movq    %r15,%dr0
  265         movq    %r14,%dr1
  266         /* Preserve reserved bits in %dr7 */
  267         andq    $0x0000fc00,%rax
  268         andq    $~0x0000fc00,%rcx
  269         movq    %r13,%dr2
  270         movq    %r12,%dr3
  271         orq     %rcx,%rax
  272         movq    %r11,%dr6
  273         movq    %rax,%dr7
  274         jmp     done_load_dr
  275 
  276 do_tss: movq    %rdx,PCPU(TSSP)
  277         movq    %rdx,%rcx
  278         movq    PCPU(TSS),%rax
  279         movw    %rcx,2(%rax)
  280         shrq    $16,%rcx
  281         movb    %cl,4(%rax)
  282         shrq    $8,%rcx
  283         movb    %cl,7(%rax)
  284         shrq    $8,%rcx
  285         movl    %ecx,8(%rax)
  286         movb    $0x89,5(%rax)   /* unset busy */
  287         movl    $TSSSEL,%eax
  288         ltr     %ax
  289         jmp     done_tss
  290 
  291 do_ldt: movq    PCPU(LDT),%rax
  292         movq    P_MD+MD_LDT_SD(%rcx),%rdx
  293         movq    %rdx,(%rax)
  294         movq    P_MD+MD_LDT_SD+8(%rcx),%rdx
  295         movq    %rdx,8(%rax)
  296         movl    $LDTSEL,%eax
  297         jmp     ld_ldt
  298 END(cpu_switch)
  299 
  300 /*
  301  * savectx(pcb)
  302  * Update pcb, saving current processor state.
  303  */
  304 ENTRY(savectx)
  305         /* Save caller's return address. */
  306         movq    (%rsp),%rax
  307         movq    %rax,PCB_RIP(%rdi)
  308 
  309         movq    %rbx,PCB_RBX(%rdi)
  310         movq    %rsp,PCB_RSP(%rdi)
  311         movq    %rbp,PCB_RBP(%rdi)
  312         movq    %r12,PCB_R12(%rdi)
  313         movq    %r13,PCB_R13(%rdi)
  314         movq    %r14,PCB_R14(%rdi)
  315         movq    %r15,PCB_R15(%rdi)
  316 
  317         movq    %cr0,%rsi
  318         movq    %rsi,PCB_CR0(%rdi)
  319         movq    %cr2,%rax
  320         movq    %rax,PCB_CR2(%rdi)
  321         movq    %cr3,%rax
  322         movq    %rax,PCB_CR3(%rdi)
  323         movq    %cr4,%rax
  324         movq    %rax,PCB_CR4(%rdi)
  325 
  326         movq    %dr0,%rax
  327         movq    %rax,PCB_DR0(%rdi)
  328         movq    %dr1,%rax
  329         movq    %rax,PCB_DR1(%rdi)
  330         movq    %dr2,%rax
  331         movq    %rax,PCB_DR2(%rdi)
  332         movq    %dr3,%rax
  333         movq    %rax,PCB_DR3(%rdi)
  334         movq    %dr6,%rax
  335         movq    %rax,PCB_DR6(%rdi)
  336         movq    %dr7,%rax
  337         movq    %rax,PCB_DR7(%rdi)
  338 
  339         movl    $MSR_FSBASE,%ecx
  340         rdmsr
  341         movl    %eax,PCB_FSBASE(%rdi)
  342         movl    %edx,PCB_FSBASE+4(%rdi)
  343         movl    $MSR_GSBASE,%ecx
  344         rdmsr
  345         movl    %eax,PCB_GSBASE(%rdi)
  346         movl    %edx,PCB_GSBASE+4(%rdi)
  347         movl    $MSR_KGSBASE,%ecx
  348         rdmsr
  349         movl    %eax,PCB_KGSBASE(%rdi)
  350         movl    %edx,PCB_KGSBASE+4(%rdi)
  351 
  352         sgdt    PCB_GDT(%rdi)
  353         sidt    PCB_IDT(%rdi)
  354         sldt    PCB_LDT(%rdi)
  355         str     PCB_TR(%rdi)
  356 
  357         clts
  358         fxsave  PCB_USERFPU(%rdi)
  359         movq    %rsi,%cr0       /* The previous %cr0 is saved in %rsi. */
  360 
  361         movl    $1,%eax
  362         ret
  363 END(savectx)

Cache object: 2ca4d629554ece8726b1fb6b32ba200c


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.