2 * Copyright (c) 2003,2004 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * Copyright (c) 1990 The Regents of the University of California.
35 * All rights reserved.
37 * This code is derived from software contributed to Berkeley by
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
43 * 1. Redistributions of source code must retain the above copyright
44 * notice, this list of conditions and the following disclaimer.
45 * 2. Redistributions in binary form must reproduce the above copyright
46 * notice, this list of conditions and the following disclaimer in the
47 * documentation and/or other materials provided with the distribution.
48 * 3. All advertising materials mentioning features or use of this software
49 * must display the following acknowledgement:
50 * This product includes software developed by the University of
51 * California, Berkeley and its contributors.
52 * 4. Neither the name of the University nor the names of its contributors
53 * may be used to endorse or promote products derived from this software
54 * without specific prior written permission.
56 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
57 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
58 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
59 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
60 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
61 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
62 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
63 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
64 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
65 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
68 * $FreeBSD: src/sys/i386/i386/swtch.s,v 1.89.2.10 2003/01/23 03:36:24 ps Exp $
69 * $DragonFly: src/sys/platform/vkernel/i386/swtch.s,v 1.8 2007/07/01 02:51:43 dillon Exp $
74 #include <sys/rtprio.h>
76 #include <machine/asmacros.h>
77 #include <machine/segments.h>
79 #include <machine/pmap.h>
80 #include <machine/lock.h>
85 #define MPLOCKED lock ;
93 .globl lwkt_switch_return
95 #if defined(SWTCH_OPTIM_STATS)
96 .globl swtch_optim_stats, tlb_flush_count
97 swtch_optim_stats: .long 0 /* number of _swtch_optims */
98 tlb_flush_count: .long 0
105 * cpu_heavy_switch(next_thread)
107 * Switch from the current thread to a new thread. This entry
108 * is normally called via the thread->td_switch function, and will
109 * only be called when the current thread is a heavy weight process.
111 * Some instructions have been reordered to reduce pipeline stalls.
113 * YYY disable interrupts once giant is removed.
115 ENTRY(cpu_heavy_switch)
119 movl PCPU(curthread),%ecx
120 movl (%esp),%eax /* (reorder optimization) */
121 movl TD_PCB(%ecx),%edx /* EDX = PCB */
122 movl %eax,PCB_EIP(%edx) /* return PC may be modified */
123 movl %ebx,PCB_EBX(%edx)
124 movl %esp,PCB_ESP(%edx)
125 movl %ebp,PCB_EBP(%edx)
126 movl %esi,PCB_ESI(%edx)
127 movl %edi,PCB_EDI(%edx)
129 movl %ecx,%ebx /* EBX = curthread */
130 movl TD_LWP(%ecx),%ecx
131 movl PCPU(cpuid), %eax
132 movl LWP_VMSPACE(%ecx), %ecx /* ECX = vmspace */
133 MPLOCKED btrl %eax, VM_PMAP+PM_ACTIVE(%ecx)
136 * Push the LWKT switch restore function, which resumes a heavy
137 * weight process. Note that the LWKT switcher is based on
138 * TD_SP, while the heavy weight process switcher is based on
139 * PCB_ESP. TD_SP is usually two ints pushed relative to
140 * PCB_ESP. We push the flags for later restore by cpu_heavy_restore.
143 pushl $cpu_heavy_restore
144 movl %esp,TD_SP(%ebx)
147 * Save debug regs if necessary
149 movb PCB_FLAGS(%edx),%al
151 jz 1f /* no, skip over */
152 movl %dr7,%eax /* yes, do the save */
153 movl %eax,PCB_DR7(%edx)
154 andl $0x0000fc00, %eax /* disable all watchpoints */
157 movl %eax,PCB_DR6(%edx)
159 movl %eax,PCB_DR3(%edx)
161 movl %eax,PCB_DR2(%edx)
163 movl %eax,PCB_DR1(%edx)
165 movl %eax,PCB_DR0(%edx)
170 * Save the FP state if we have used the FP. Note that calling
171 * npxsave will NULL out PCPU(npxthread).
173 cmpl %ebx,PCPU(npxthread)
175 pushl TD_SAVEFPU(%ebx)
176 call npxsave /* do it in a big C function */
177 addl $4,%esp /* EAX, ECX, EDX trashed */
179 #endif /* NNPX > 0 */
182 * Switch to the next thread, which was passed as an argument
183 * to cpu_heavy_switch(). Due to the eflags and switch-restore
184 * function we pushed, the argument is at 12(%esp). Set the current
185 * thread, load the stack pointer, and 'ret' into the switch-restore
188 * The switch restore function expects the new thread to be in %eax
189 * and the old one to be in %ebx.
191 * There is a one-instruction window where curthread is the new
192 * thread but %esp still points to the old thread's stack, but
193 * we are protected by a critical section so it is ok.
195 movl 12(%esp),%eax /* EAX = newtd, EBX = oldtd */
196 movl %eax,PCPU(curthread)
197 movl TD_SP(%eax),%esp
203 * The switch function is changed to this when a thread is going away
204 * for good. We have to ensure that the MMU state is not cached, and
205 * we don't bother saving the existing thread state before switching.
207 * At this point we are in a critical section and this cpu owns the
208 * thread's token, which serves as an interlock until the switchout is
211 ENTRY(cpu_exit_switch)
213 * Get us out of the vmspace
223 movl PCPU(curthread),%ebx
226 * If this is a process/lwp, deactivate the pmap after we've
229 movl TD_LWP(%ebx),%ecx
232 movl PCPU(cpuid), %eax
233 movl LWP_VMSPACE(%ecx), %ecx /* ECX = vmspace */
234 MPLOCKED btrl %eax, VM_PMAP+PM_ACTIVE(%ecx)
237 * Switch to the next thread. RET into the restore function, which
238 * expects the new thread in EAX and the old in EBX.
240 * There is a one-instruction window where curthread is the new
241 * thread but %esp still points to the old thread's stack, but
242 * we are protected by a critical section so it is ok.
245 movl %eax,PCPU(curthread)
246 movl TD_SP(%eax),%esp
250 * cpu_heavy_restore() (current thread in %eax on entry)
252 * Restore the thread after an LWKT switch. This entry is normally
253 * called via the LWKT switch restore function, which was pulled
254 * off the thread stack and jumped to.
256 * This entry is only called if the thread was previously saved
257 * using cpu_heavy_switch() (the heavy weight process thread switcher),
258 * or when a new process is initially scheduled.
260 * NOTE: The lwp may be in any state, not necessarily LSRUN, because
261 * a preemption switch may interrupt the process and then return via
264 * YYY theoretically we do not have to restore everything here, a lot
265 * of this junk can wait until we return to usermode. But for now
266 * we restore everything.
268 * YYY the PCB crap is really crap, it makes startup a bitch because
269 * we can't switch away.
271 * YYY note: spl check is done in mi_switch when it splx()'s.
274 ENTRY(cpu_heavy_restore)
276 movl TD_PCB(%eax),%edx /* EDX = PCB */
277 movl TD_LWP(%eax),%ecx
279 #if defined(SWTCH_OPTIM_STATS)
280 incl _swtch_optim_stats
283 * Tell the pmap that our cpu is using the VMSPACE now. We cannot
284 * safely test/reload %cr3 until after we have set the bit in the
285 * pmap (remember, we do not hold the MP lock in the switch code).
287 movl LWP_VMSPACE(%ecx), %ecx /* ECX = vmspace */
288 movl PCPU(cpuid), %esi
289 MPLOCKED btsl %esi, VM_PMAP+PM_ACTIVE(%ecx)
292 * Restore the MMU address space. If it is the same as the last
293 * thread we don't have to invalidate the tlb (i.e. reload cr3).
294 * YYY which naturally also means that the PM_ACTIVE bit had better
295 * already have been set before we set it above, check? YYY
299 movl PCB_CR3(%edx),%ecx
302 #if defined(SWTCH_OPTIM_STATS)
303 decl _swtch_optim_stats
304 incl _tlb_flush_count
310 * NOTE: %ebx is the previous thread and %eax is the new thread.
311 * %ebx is retained throughout so we can return it.
313 * lwkt_switch[_return] is responsible for handling TDF_RUNNING.
317 * Deal with the PCB extension, restore the private tss
319 movl PCB_EXT(%edx),%edi /* check for a PCB extension */
320 movl $1,%ecx /* maybe mark use of a private tss */
325 * Going back to the common_tss. We may need to update TSS_ESP0
326 * which sets the top of the supervisor stack when entering from
327 * usermode. The PCB is at the top of the stack but we need another
328 * 16 bytes to take vm86 into account.
331 movl %ecx, PCPU(common_tss) + TSS_ESP0
333 cmpl $0,PCPU(private_tss) /* don't have to reload if */
334 je 3f /* already using the common TSS */
336 subl %ecx,%ecx /* unmark use of private tss */
339 * Get the address of the common TSS descriptor for the ltr.
340 * There is no way to get the address of a segment-accessed variable
341 * so we store a self-referential pointer at the base of the per-cpu
342 * data area and add the appropriate offset.
344 movl $gd_common_tssd, %edi
348 * Move the correct TSS descriptor into the GDT slot, then reload
352 movl %ecx,PCPU(private_tss) /* mark/unmark private tss */
353 movl PCPU(tss_gdt), %ecx /* entry in GDT */
358 movl $GPROC0_SEL*8, %esi /* GSEL(entry, SEL_KPL) */
363 * Restore general registers. %ebx is restored later.
365 movl PCB_ESP(%edx),%esp
366 movl PCB_EBP(%edx),%ebp
367 movl PCB_ESI(%edx),%esi
368 movl PCB_EDI(%edx),%edi
369 movl PCB_EIP(%edx),%eax
374 * Restore the user LDT if we have one
376 cmpl $0, PCB_USERLDT(%edx)
378 movl _default_ldt,%eax
379 cmpl PCPU(currentldt),%eax
382 movl %eax,PCPU(currentldt)
391 * Restore the user TLS if we have one
399 * Restore the DEBUG register state if necessary.
401 movb PCB_FLAGS(%edx),%al
403 jz 1f /* no, skip over */
404 movl PCB_DR6(%edx),%eax /* yes, do the restore */
406 movl PCB_DR3(%edx),%eax
408 movl PCB_DR2(%edx),%eax
410 movl PCB_DR1(%edx),%eax
412 movl PCB_DR0(%edx),%eax
414 movl %dr7,%eax /* load dr7 so as not to disturb */
415 andl $0x0000fc00,%eax /* reserved bits */
416 movl PCB_DR7(%edx),%ecx
417 andl $~0x0000fc00,%ecx
421 movl %ebx,%eax /* return previous thread */
422 movl PCB_EBX(%edx),%ebx
428 * Update pcb, saving current processor state.
434 /* caller's return address - child won't execute this routine */
436 movl %eax,PCB_EIP(%ecx)
437 movl %ebx,PCB_EBX(%ecx)
438 movl %esp,PCB_ESP(%ecx)
439 movl %ebp,PCB_EBP(%ecx)
440 movl %esi,PCB_ESI(%ecx)
441 movl %edi,PCB_EDI(%ecx)
445 * If npxthread == NULL, then the npx h/w state is irrelevant and the
446 * state had better already be in the pcb. This is true for forks
447 * but not for dumps (the old book-keeping with FP flags in the pcb
448 * always lost for dumps because the dump pcb has 0 flags).
450 * If npxthread != NULL, then we have to save the npx h/w state to
451 * npxthread's pcb and copy it to the requested pcb, or save to the
452 * requested pcb and reload. Copying is easier because we would
453 * have to handle h/w bugs for reloading. We used to lose the
454 * parent's npx state for forks by forgetting to reload.
456 movl PCPU(npxthread),%eax
460 pushl %ecx /* target pcb */
461 movl TD_SAVEFPU(%eax),%eax /* originating savefpu area */
471 pushl $PCB_SAVEFPU_SIZE
472 leal PCB_SAVEFPU(%ecx),%ecx
477 #endif /* NNPX > 0 */
483 * cpu_idle_restore() (current thread in %eax on entry) (one-time execution)
485 * Don't bother setting up any regs other then %ebp so backtraces
486 * don't die. This restore function is used to bootstrap into the
487 * cpu_idle() LWKT only, after that cpu_lwkt_*() will be used for
490 * Clear TDF_RUNNING in old thread only after we've cleaned up %cr3.
491 * This only occurs during system boot so no special handling is
492 * required for migration.
494 * If we are an AP we have to call ap_init() before jumping to
495 * cpu_idle(). ap_init() will synchronize with the BP and finish
496 * setting up various ncpu-dependant globaldata fields. This may
497 * happen on UP as well as SMP if we happen to be simulating multiple
500 ENTRY(cpu_idle_restore)
504 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
506 orl $TDF_RUNNING,TD_FLAGS(%eax)
518 * cpu_kthread_restore() (current thread is %eax on entry) (one-time execution)
520 * Don't bother setting up any regs other then %ebp so backtraces
521 * don't die. This restore function is used to bootstrap into an
522 * LWKT based kernel thread only. cpu_lwkt_switch() will be used
525 * Since all of our context is on the stack we are reentrant and
526 * we can release our critical section and enable interrupts early.
528 * Because this switch target does not 'return' to lwkt_switch()
529 * we have to call lwkt_switch_return(otd) to clean up otd.
532 ENTRY(cpu_kthread_restore)
534 movl TD_PCB(%eax),%esi
538 pushl %ebx /* argument to lwkt_switch_return */
539 call lwkt_switch_return
543 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
544 orl $TDF_RUNNING,TD_FLAGS(%eax)
546 decl TD_CRITCOUNT(%eax)
547 popl %eax /* kthread exit function */
548 pushl PCB_EBX(%esi) /* argument to ESI function */
549 pushl %eax /* set exit func as return address */
550 movl PCB_ESI(%esi),%eax
556 * Standard LWKT switching function. Only non-scratch registers are
557 * saved and we don't bother with the MMU state or anything else.
559 * This function is always called while in a critical section.
561 * There is a one-instruction window where curthread is the new
562 * thread but %esp still points to the old thread's stack, but
563 * we are protected by a critical section so it is ok.
567 ENTRY(cpu_lwkt_switch)
568 pushl %ebp /* note: GDB hacked to locate ebp relative to td_sp */
570 movl PCPU(curthread),%ebx
574 /* warning: adjust movl into %eax below if you change the pushes */
578 * Save the FP state if we have used the FP. Note that calling
579 * npxsave will NULL out PCPU(npxthread).
581 * We have to deal with the FP state for LWKT threads in case they
582 * happen to get preempted or block while doing an optimized
583 * bzero/bcopy/memcpy.
585 cmpl %ebx,PCPU(npxthread)
587 pushl TD_SAVEFPU(%ebx)
588 call npxsave /* do it in a big C function */
589 addl $4,%esp /* EAX, ECX, EDX trashed */
591 #endif /* NNPX > 0 */
593 movl 4+20(%esp),%eax /* switch to this thread */
594 pushl $cpu_lwkt_restore
595 movl %esp,TD_SP(%ebx)
596 movl %eax,PCPU(curthread)
597 movl TD_SP(%eax),%esp
600 * eax contains new thread, ebx contains old thread.
605 * cpu_lwkt_restore() (current thread in %eax on entry)
607 * Standard LWKT restore function. This function is always called
608 * while in a critical section.
610 * Warning: due to preemption the restore function can be used to
611 * 'return' to the original thread. Interrupt disablement must be
612 * protected through the switch so we cannot run splz here.
614 ENTRY(cpu_lwkt_restore)
616 * NOTE: %ebx is the previous thread and %eax is the new thread.
617 * %ebx is retained throughout so we can return it.
619 * lwkt_switch[_return] is responsible for handling TDF_RUNNING.
632 * Make AP become the idle loop.
634 ENTRY(bootstrap_idle)
635 movl PCPU(curthread),%eax
637 movl TD_SP(%eax),%esp