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 1.37 2005/02/21 21:40:53 dillon Exp $
74 #include <sys/rtprio.h>
76 #include <machine/asmacros.h>
77 #include <machine/ipl.h>
79 #include <machine/pmap.h>
80 #include <machine/smptests.h> /** GRAB_LOPRIO */
81 #include <machine/apicreg.h>
82 #include <machine/lock.h>
87 #define MPLOCKED lock ;
96 #if defined(SWTCH_OPTIM_STATS)
97 .globl swtch_optim_stats, tlb_flush_count
98 swtch_optim_stats: .long 0 /* number of _swtch_optims */
99 tlb_flush_count: .long 0
106 * cpu_heavy_switch(next_thread)
108 * Switch from the current thread to a new thread. This entry
109 * is normally called via the thread->td_switch function, and will
110 * only be called when the current thread is a heavy weight process.
112 * Some instructions have been reordered to reduce pipeline stalls.
114 * YYY disable interrupts once giant is removed.
116 ENTRY(cpu_heavy_switch)
120 movl PCPU(curthread),%ecx
121 movl (%esp),%eax /* (reorder optimization) */
122 movl TD_PCB(%ecx),%edx /* EDX = PCB */
123 movl %eax,PCB_EIP(%edx) /* return PC may be modified */
124 movl %ebx,PCB_EBX(%edx)
125 movl %esp,PCB_ESP(%edx)
126 movl %ebp,PCB_EBP(%edx)
127 movl %esi,PCB_ESI(%edx)
128 movl %edi,PCB_EDI(%edx)
129 movl %gs,PCB_GS(%edx)
131 movl %ecx,%ebx /* EBX = curthread */
132 movl TD_PROC(%ecx),%ecx
133 movl PCPU(cpuid), %eax
134 movl P_VMSPACE(%ecx), %ecx /* ECX = vmspace */
135 MPLOCKED btrl %eax, VM_PMAP+PM_ACTIVE(%ecx)
138 * Push the LWKT switch restore function, which resumes a heavy
139 * weight process. Note that the LWKT switcher is based on
140 * TD_SP, while the heavy weight process switcher is based on
141 * PCB_ESP. TD_SP is usually two ints pushed relative to
142 * PCB_ESP. We push the flags for later restore by cpu_heavy_restore.
145 pushl $cpu_heavy_restore
146 movl %esp,TD_SP(%ebx)
149 * Save debug regs if necessary
151 movb PCB_FLAGS(%edx),%al
153 jz 1f /* no, skip over */
154 movl %dr7,%eax /* yes, do the save */
155 movl %eax,PCB_DR7(%edx)
156 andl $0x0000fc00, %eax /* disable all watchpoints */
159 movl %eax,PCB_DR6(%edx)
161 movl %eax,PCB_DR3(%edx)
163 movl %eax,PCB_DR2(%edx)
165 movl %eax,PCB_DR1(%edx)
167 movl %eax,PCB_DR0(%edx)
172 * Save the FP state if we have used the FP. Note that calling
173 * npxsave will NULL out PCPU(npxthread).
175 cmpl %ebx,PCPU(npxthread)
177 pushl TD_SAVEFPU(%ebx)
178 call npxsave /* do it in a big C function */
179 addl $4,%esp /* EAX, ECX, EDX trashed */
181 #endif /* NNPX > 0 */
184 * Switch to the next thread, which was passed as an argument
185 * to cpu_heavy_switch(). Due to the eflags and switch-restore
186 * function we pushed, the argument is at 12(%esp). Set the current
187 * thread, load the stack pointer, and 'ret' into the switch-restore
190 * The switch restore function expects the new thread to be in %eax
191 * and the old one to be in %ebx.
193 * There is a one-instruction window where curthread is the new
194 * thread but %esp still points to the old thread's stack, but
195 * we are protected by a critical section so it is ok.
197 movl 12(%esp),%eax /* EAX = newtd, EBX = oldtd */
198 movl %eax,PCPU(curthread)
199 movl TD_SP(%eax),%esp
205 * The switch function is changed to this when a thread is going away
206 * for good. We have to ensure that the MMU state is not cached, and
207 * we don't bother saving the existing thread state before switching.
209 * At this point we are in a critical section and this cpu owns the
210 * thread's token, which serves as an interlock until the switchout is
213 ENTRY(cpu_exit_switch)
215 * Get us out of the vmspace
223 movl PCPU(curthread),%ebx
225 * Switch to the next thread. RET into the restore function, which
226 * expects the new thread in EAX and the old in EBX.
228 * There is a one-instruction window where curthread is the new
229 * thread but %esp still points to the old thread's stack, but
230 * we are protected by a critical section so it is ok.
233 movl %eax,PCPU(curthread)
234 movl TD_SP(%eax),%esp
238 * cpu_heavy_restore() (current thread in %eax on entry)
240 * Restore the thread after an LWKT switch. This entry is normally
241 * called via the LWKT switch restore function, which was pulled
242 * off the thread stack and jumped to.
244 * This entry is only called if the thread was previously saved
245 * using cpu_heavy_switch() (the heavy weight process thread switcher),
246 * or when a new process is initially scheduled. The first thing we
247 * do is clear the TDF_RUNNING bit in the old thread and set it in the
250 * NOTE: The process may be in any state, not necessarily SRUN, because
251 * a preemption switch may interrupt the process and then return via
254 * YYY theoretically we do not have to restore everything here, a lot
255 * of this junk can wait until we return to usermode. But for now
256 * we restore everything.
258 * YYY the PCB crap is really crap, it makes startup a bitch because
259 * we can't switch away.
261 * YYY note: spl check is done in mi_switch when it splx()'s.
264 ENTRY(cpu_heavy_restore)
266 movl TD_PCB(%eax),%edx /* EDX = PCB */
267 movl TD_PROC(%eax),%ecx
269 #if defined(SWTCH_OPTIM_STATS)
270 incl _swtch_optim_stats
273 * Tell the pmap that our cpu is using the VMSPACE now. We cannot
274 * safely test/reload %cr3 until after we have set the bit in the
275 * pmap (remember, we do not hold the MP lock in the switch code).
277 movl P_VMSPACE(%ecx), %ecx /* ECX = vmspace */
278 movl PCPU(cpuid), %esi
279 MPLOCKED btsl %esi, VM_PMAP+PM_ACTIVE(%ecx)
282 * Restore the MMU address space. If it is the same as the last
283 * thread we don't have to invalidate the tlb (i.e. reload cr3).
284 * YYY which naturally also means that the PM_ACTIVE bit had better
285 * already have been set before we set it above, check? YYY
288 movl PCB_CR3(%edx),%ecx
291 #if defined(SWTCH_OPTIM_STATS)
292 decl _swtch_optim_stats
293 incl _tlb_flush_count
298 * Clear TDF_RUNNING flag in old thread only after cleaning up
299 * %cr3. The target thread is already protected by being TDF_RUNQ
300 * so setting TDF_RUNNING isn't as big a deal.
302 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
303 orl $TDF_RUNNING,TD_FLAGS(%eax)
306 * Deal with the PCB extension, restore the private tss
308 movl PCB_EXT(%edx),%edi /* check for a PCB extension */
309 movl $1,%ebx /* maybe mark use of a private tss */
314 * Going back to the common_tss. We may need to update TSS_ESP0
315 * which sets the top of the supervisor stack when entering from
316 * usermode. The PCB is at the top of the stack but we need another
317 * 16 bytes to take vm86 into account.
320 movl %ebx, PCPU(common_tss) + TSS_ESP0
322 cmpl $0,PCPU(private_tss) /* don't have to reload if */
323 je 3f /* already using the common TSS */
325 subl %ebx,%ebx /* unmark use of private tss */
328 * Get the address of the common TSS descriptor for the ltr.
329 * There is no way to get the address of a segment-accessed variable
330 * so we store a self-referential pointer at the base of the per-cpu
331 * data area and add the appropriate offset.
333 movl $gd_common_tssd, %edi
337 * Move the correct TSS descriptor into the GDT slot, then reload
341 movl %ebx,PCPU(private_tss) /* mark/unmark private tss */
342 movl PCPU(tss_gdt), %ebx /* entry in GDT */
347 movl $GPROC0_SEL*8, %esi /* GSEL(entry, SEL_KPL) */
352 * Restore general registers.
354 movl PCB_EBX(%edx),%ebx
355 movl PCB_ESP(%edx),%esp
356 movl PCB_EBP(%edx),%ebp
357 movl PCB_ESI(%edx),%esi
358 movl PCB_EDI(%edx),%edi
359 movl PCB_EIP(%edx),%eax
363 * Restore the user LDT if we have one
365 cmpl $0, PCB_USERLDT(%edx)
367 movl _default_ldt,%eax
368 cmpl PCPU(currentldt),%eax
371 movl %eax,PCPU(currentldt)
378 * Restore the user TLS if we have one
384 * Restore the %gs segment register, which must be done after
385 * loading the user LDT. Since user processes can modify the
386 * register via procfs, this may result in a fault which is
387 * detected by checking the fault address against cpu_switch_load_gs
388 * in i386/i386/trap.c
390 .globl cpu_switch_load_gs
392 movl PCB_GS(%edx),%gs
395 * Restore the DEBUG register state if necessary.
397 movb PCB_FLAGS(%edx),%al
399 jz 1f /* no, skip over */
400 movl PCB_DR6(%edx),%eax /* yes, do the restore */
402 movl PCB_DR3(%edx),%eax
404 movl PCB_DR2(%edx),%eax
406 movl PCB_DR1(%edx),%eax
408 movl PCB_DR0(%edx),%eax
410 movl %dr7,%eax /* load dr7 so as not to disturb */
411 andl $0x0000fc00,%eax /* reserved bits */
413 movl PCB_DR7(%edx),%ebx
414 andl $~0x0000fc00,%ebx
425 * Update pcb, saving current processor state.
431 /* caller's return address - child won't execute this routine */
433 movl %eax,PCB_EIP(%ecx)
436 movl %eax,PCB_CR3(%ecx)
438 movl %ebx,PCB_EBX(%ecx)
439 movl %esp,PCB_ESP(%ecx)
440 movl %ebp,PCB_EBP(%ecx)
441 movl %esi,PCB_ESI(%ecx)
442 movl %edi,PCB_EDI(%ecx)
443 movl %gs,PCB_GS(%ecx)
447 * If npxthread == NULL, then the npx h/w state is irrelevant and the
448 * state had better already be in the pcb. This is true for forks
449 * but not for dumps (the old book-keeping with FP flags in the pcb
450 * always lost for dumps because the dump pcb has 0 flags).
452 * If npxthread != NULL, then we have to save the npx h/w state to
453 * npxthread's pcb and copy it to the requested pcb, or save to the
454 * requested pcb and reload. Copying is easier because we would
455 * have to handle h/w bugs for reloading. We used to lose the
456 * parent's npx state for forks by forgetting to reload.
458 movl PCPU(npxthread),%eax
462 pushl %ecx /* target pcb */
463 movl TD_SAVEFPU(%eax),%eax /* originating savefpu area */
473 pushl $PCB_SAVEFPU_SIZE
474 leal PCB_SAVEFPU(%ecx),%ecx
479 #endif /* NNPX > 0 */
485 * cpu_idle_restore() (current thread in %eax on entry) (one-time execution)
487 * Don't bother setting up any regs other then %ebp so backtraces
488 * don't die. This restore function is used to bootstrap into the
489 * cpu_idle() LWKT only, after that cpu_lwkt_*() will be used for
492 * Clear TDF_RUNNING in old thread only after we've cleaned up %cr3.
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)
506 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
507 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 ENTRY(cpu_kthread_restore)
531 movl TD_PCB(%eax),%edx
534 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
535 orl $TDF_RUNNING,TD_FLAGS(%eax)
536 subl $TDPRI_CRIT,TD_PRI(%eax)
537 popl %eax /* kthread exit function */
538 pushl PCB_EBX(%edx) /* argument to ESI function */
539 pushl %eax /* set exit func as return address */
540 movl PCB_ESI(%edx),%eax
546 * Standard LWKT switching function. Only non-scratch registers are
547 * saved and we don't bother with the MMU state or anything else.
549 * This function is always called while in a critical section.
551 * There is a one-instruction window where curthread is the new
552 * thread but %esp still points to the old thread's stack, but
553 * we are protected by a critical section so it is ok.
557 ENTRY(cpu_lwkt_switch)
558 pushl %ebp /* note: GDB hacked to locate ebp relative to td_sp */
560 movl PCPU(curthread),%ebx
564 /* warning: adjust movl into %eax below if you change the pushes */
568 * Save the FP state if we have used the FP. Note that calling
569 * npxsave will NULL out PCPU(npxthread).
571 * We have to deal with the FP state for LWKT threads in case they
572 * happen to get preempted or block while doing an optimized
573 * bzero/bcopy/memcpy.
575 cmpl %ebx,PCPU(npxthread)
577 pushl TD_SAVEFPU(%ebx)
578 call npxsave /* do it in a big C function */
579 addl $4,%esp /* EAX, ECX, EDX trashed */
581 #endif /* NNPX > 0 */
583 movl 4+20(%esp),%eax /* switch to this thread */
584 pushl $cpu_lwkt_restore
585 movl %esp,TD_SP(%ebx)
586 movl %eax,PCPU(curthread)
587 movl TD_SP(%eax),%esp
590 * eax contains new thread, ebx contains old thread.
595 * cpu_lwkt_restore() (current thread in %eax on entry)
597 * Standard LWKT restore function. This function is always called
598 * while in a critical section.
600 * Warning: due to preemption the restore function can be used to
601 * 'return' to the original thread. Interrupt disablement must be
602 * protected through the switch so we cannot run splz here.
604 * YYY we theoretically do not need to load IdlePTD into cr3, but if
605 * so we need a way to detect when the PTD we are using is being
606 * deleted due to a process exiting.
608 ENTRY(cpu_lwkt_restore)
609 movl IdlePTD,%ecx /* YYY borrow but beware desched/cpuchg/exit */
615 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
616 orl $TDF_RUNNING,TD_FLAGS(%eax)