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/pc32/i386/swtch.s,v 1.43 2007/01/08 03:33:42 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_base/apic/apicreg.h>
81 #include <machine/lock.h>
86 #define MPLOCKED lock ;
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_PROC(%ecx),%ecx
131 movl PCPU(cpuid), %eax
132 movl P_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
221 movl PCPU(curthread),%ebx
223 * Switch to the next thread. RET into the restore function, which
224 * expects the new thread in EAX and the old in EBX.
226 * There is a one-instruction window where curthread is the new
227 * thread but %esp still points to the old thread's stack, but
228 * we are protected by a critical section so it is ok.
231 movl %eax,PCPU(curthread)
232 movl TD_SP(%eax),%esp
236 * cpu_heavy_restore() (current thread in %eax on entry)
238 * Restore the thread after an LWKT switch. This entry is normally
239 * called via the LWKT switch restore function, which was pulled
240 * off the thread stack and jumped to.
242 * This entry is only called if the thread was previously saved
243 * using cpu_heavy_switch() (the heavy weight process thread switcher),
244 * or when a new process is initially scheduled. The first thing we
245 * do is clear the TDF_RUNNING bit in the old thread and set it in the
248 * NOTE: The process may be in any state, not necessarily SRUN, because
249 * a preemption switch may interrupt the process and then return via
252 * YYY theoretically we do not have to restore everything here, a lot
253 * of this junk can wait until we return to usermode. But for now
254 * we restore everything.
256 * YYY the PCB crap is really crap, it makes startup a bitch because
257 * we can't switch away.
259 * YYY note: spl check is done in mi_switch when it splx()'s.
262 ENTRY(cpu_heavy_restore)
264 movl TD_PCB(%eax),%edx /* EDX = PCB */
265 movl TD_PROC(%eax),%ecx
267 #if defined(SWTCH_OPTIM_STATS)
268 incl _swtch_optim_stats
271 * Tell the pmap that our cpu is using the VMSPACE now. We cannot
272 * safely test/reload %cr3 until after we have set the bit in the
273 * pmap (remember, we do not hold the MP lock in the switch code).
275 movl P_VMSPACE(%ecx), %ecx /* ECX = vmspace */
276 movl PCPU(cpuid), %esi
277 MPLOCKED btsl %esi, VM_PMAP+PM_ACTIVE(%ecx)
280 * Restore the MMU address space. If it is the same as the last
281 * thread we don't have to invalidate the tlb (i.e. reload cr3).
282 * YYY which naturally also means that the PM_ACTIVE bit had better
283 * already have been set before we set it above, check? YYY
286 movl PCB_CR3(%edx),%ecx
289 #if defined(SWTCH_OPTIM_STATS)
290 decl _swtch_optim_stats
291 incl _tlb_flush_count
296 * Clear TDF_RUNNING flag in old thread only after cleaning up
297 * %cr3. The target thread is already protected by being TDF_RUNQ
298 * so setting TDF_RUNNING isn't as big a deal.
300 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
301 orl $TDF_RUNNING,TD_FLAGS(%eax)
304 * Deal with the PCB extension, restore the private tss
306 movl PCB_EXT(%edx),%edi /* check for a PCB extension */
307 movl $1,%ebx /* maybe mark use of a private tss */
312 * Going back to the common_tss. We may need to update TSS_ESP0
313 * which sets the top of the supervisor stack when entering from
314 * usermode. The PCB is at the top of the stack but we need another
315 * 16 bytes to take vm86 into account.
318 movl %ebx, PCPU(common_tss) + TSS_ESP0
320 cmpl $0,PCPU(private_tss) /* don't have to reload if */
321 je 3f /* already using the common TSS */
323 subl %ebx,%ebx /* unmark use of private tss */
326 * Get the address of the common TSS descriptor for the ltr.
327 * There is no way to get the address of a segment-accessed variable
328 * so we store a self-referential pointer at the base of the per-cpu
329 * data area and add the appropriate offset.
331 movl $gd_common_tssd, %edi
335 * Move the correct TSS descriptor into the GDT slot, then reload
339 movl %ebx,PCPU(private_tss) /* mark/unmark private tss */
340 movl PCPU(tss_gdt), %ebx /* entry in GDT */
345 movl $GPROC0_SEL*8, %esi /* GSEL(entry, SEL_KPL) */
350 * Restore general registers.
352 movl PCB_EBX(%edx),%ebx
353 movl PCB_ESP(%edx),%esp
354 movl PCB_EBP(%edx),%ebp
355 movl PCB_ESI(%edx),%esi
356 movl PCB_EDI(%edx),%edi
357 movl PCB_EIP(%edx),%eax
361 * Restore the user LDT if we have one
363 cmpl $0, PCB_USERLDT(%edx)
365 movl _default_ldt,%eax
366 cmpl PCPU(currentldt),%eax
369 movl %eax,PCPU(currentldt)
376 * Restore the user TLS if we have one
383 * Restore the DEBUG register state if necessary.
385 movb PCB_FLAGS(%edx),%al
387 jz 1f /* no, skip over */
388 movl PCB_DR6(%edx),%eax /* yes, do the restore */
390 movl PCB_DR3(%edx),%eax
392 movl PCB_DR2(%edx),%eax
394 movl PCB_DR1(%edx),%eax
396 movl PCB_DR0(%edx),%eax
398 movl %dr7,%eax /* load dr7 so as not to disturb */
399 andl $0x0000fc00,%eax /* reserved bits */
401 movl PCB_DR7(%edx),%ebx
402 andl $~0x0000fc00,%ebx
413 * Update pcb, saving current processor state.
419 /* caller's return address - child won't execute this routine */
421 movl %eax,PCB_EIP(%ecx)
424 movl %eax,PCB_CR3(%ecx)
426 movl %ebx,PCB_EBX(%ecx)
427 movl %esp,PCB_ESP(%ecx)
428 movl %ebp,PCB_EBP(%ecx)
429 movl %esi,PCB_ESI(%ecx)
430 movl %edi,PCB_EDI(%ecx)
434 * If npxthread == NULL, then the npx h/w state is irrelevant and the
435 * state had better already be in the pcb. This is true for forks
436 * but not for dumps (the old book-keeping with FP flags in the pcb
437 * always lost for dumps because the dump pcb has 0 flags).
439 * If npxthread != NULL, then we have to save the npx h/w state to
440 * npxthread's pcb and copy it to the requested pcb, or save to the
441 * requested pcb and reload. Copying is easier because we would
442 * have to handle h/w bugs for reloading. We used to lose the
443 * parent's npx state for forks by forgetting to reload.
445 movl PCPU(npxthread),%eax
449 pushl %ecx /* target pcb */
450 movl TD_SAVEFPU(%eax),%eax /* originating savefpu area */
460 pushl $PCB_SAVEFPU_SIZE
461 leal PCB_SAVEFPU(%ecx),%ecx
466 #endif /* NNPX > 0 */
472 * cpu_idle_restore() (current thread in %eax on entry) (one-time execution)
474 * Don't bother setting up any regs other then %ebp so backtraces
475 * don't die. This restore function is used to bootstrap into the
476 * cpu_idle() LWKT only, after that cpu_lwkt_*() will be used for
479 * Clear TDF_RUNNING in old thread only after we've cleaned up %cr3.
481 * If we are an AP we have to call ap_init() before jumping to
482 * cpu_idle(). ap_init() will synchronize with the BP and finish
483 * setting up various ncpu-dependant globaldata fields. This may
484 * happen on UP as well as SMP if we happen to be simulating multiple
487 ENTRY(cpu_idle_restore)
493 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
494 orl $TDF_RUNNING,TD_FLAGS(%eax)
505 * cpu_kthread_restore() (current thread is %eax on entry) (one-time execution)
507 * Don't bother setting up any regs other then %ebp so backtraces
508 * don't die. This restore function is used to bootstrap into an
509 * LWKT based kernel thread only. cpu_lwkt_switch() will be used
512 * Since all of our context is on the stack we are reentrant and
513 * we can release our critical section and enable interrupts early.
515 ENTRY(cpu_kthread_restore)
518 movl TD_PCB(%eax),%edx
521 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
522 orl $TDF_RUNNING,TD_FLAGS(%eax)
523 subl $TDPRI_CRIT,TD_PRI(%eax)
524 popl %eax /* kthread exit function */
525 pushl PCB_EBX(%edx) /* argument to ESI function */
526 pushl %eax /* set exit func as return address */
527 movl PCB_ESI(%edx),%eax
533 * Standard LWKT switching function. Only non-scratch registers are
534 * saved and we don't bother with the MMU state or anything else.
536 * This function is always called while in a critical section.
538 * There is a one-instruction window where curthread is the new
539 * thread but %esp still points to the old thread's stack, but
540 * we are protected by a critical section so it is ok.
544 ENTRY(cpu_lwkt_switch)
545 pushl %ebp /* note: GDB hacked to locate ebp relative to td_sp */
547 movl PCPU(curthread),%ebx
551 /* warning: adjust movl into %eax below if you change the pushes */
555 * Save the FP state if we have used the FP. Note that calling
556 * npxsave will NULL out PCPU(npxthread).
558 * We have to deal with the FP state for LWKT threads in case they
559 * happen to get preempted or block while doing an optimized
560 * bzero/bcopy/memcpy.
562 cmpl %ebx,PCPU(npxthread)
564 pushl TD_SAVEFPU(%ebx)
565 call npxsave /* do it in a big C function */
566 addl $4,%esp /* EAX, ECX, EDX trashed */
568 #endif /* NNPX > 0 */
570 movl 4+20(%esp),%eax /* switch to this thread */
571 pushl $cpu_lwkt_restore
572 movl %esp,TD_SP(%ebx)
573 movl %eax,PCPU(curthread)
574 movl TD_SP(%eax),%esp
577 * eax contains new thread, ebx contains old thread.
582 * cpu_lwkt_restore() (current thread in %eax on entry)
584 * Standard LWKT restore function. This function is always called
585 * while in a critical section.
587 * Warning: due to preemption the restore function can be used to
588 * 'return' to the original thread. Interrupt disablement must be
589 * protected through the switch so we cannot run splz here.
591 * YYY we theoretically do not need to load IdlePTD into cr3, but if
592 * so we need a way to detect when the PTD we are using is being
593 * deleted due to a process exiting.
595 ENTRY(cpu_lwkt_restore)
596 movl IdlePTD,%ecx /* YYY borrow but beware desched/cpuchg/exit */
602 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
603 orl $TDF_RUNNING,TD_FLAGS(%eax)