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 1.42 2006/11/07 18:50:07 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)
128 movl %gs,PCB_GS(%edx)
130 movl %ecx,%ebx /* EBX = curthread */
131 movl TD_PROC(%ecx),%ecx
132 movl PCPU(cpuid), %eax
133 movl P_VMSPACE(%ecx), %ecx /* ECX = vmspace */
134 MPLOCKED btrl %eax, VM_PMAP+PM_ACTIVE(%ecx)
137 * Push the LWKT switch restore function, which resumes a heavy
138 * weight process. Note that the LWKT switcher is based on
139 * TD_SP, while the heavy weight process switcher is based on
140 * PCB_ESP. TD_SP is usually two ints pushed relative to
141 * PCB_ESP. We push the flags for later restore by cpu_heavy_restore.
144 pushl $cpu_heavy_restore
145 movl %esp,TD_SP(%ebx)
148 * Save debug regs if necessary
150 movb PCB_FLAGS(%edx),%al
152 jz 1f /* no, skip over */
153 movl %dr7,%eax /* yes, do the save */
154 movl %eax,PCB_DR7(%edx)
155 andl $0x0000fc00, %eax /* disable all watchpoints */
158 movl %eax,PCB_DR6(%edx)
160 movl %eax,PCB_DR3(%edx)
162 movl %eax,PCB_DR2(%edx)
164 movl %eax,PCB_DR1(%edx)
166 movl %eax,PCB_DR0(%edx)
171 * Save the FP state if we have used the FP. Note that calling
172 * npxsave will NULL out PCPU(npxthread).
174 cmpl %ebx,PCPU(npxthread)
176 pushl TD_SAVEFPU(%ebx)
177 call npxsave /* do it in a big C function */
178 addl $4,%esp /* EAX, ECX, EDX trashed */
180 #endif /* NNPX > 0 */
183 * Switch to the next thread, which was passed as an argument
184 * to cpu_heavy_switch(). Due to the eflags and switch-restore
185 * function we pushed, the argument is at 12(%esp). Set the current
186 * thread, load the stack pointer, and 'ret' into the switch-restore
189 * The switch restore function expects the new thread to be in %eax
190 * and the old one to be in %ebx.
192 * There is a one-instruction window where curthread is the new
193 * thread but %esp still points to the old thread's stack, but
194 * we are protected by a critical section so it is ok.
196 movl 12(%esp),%eax /* EAX = newtd, EBX = oldtd */
197 movl %eax,PCPU(curthread)
198 movl TD_SP(%eax),%esp
204 * The switch function is changed to this when a thread is going away
205 * for good. We have to ensure that the MMU state is not cached, and
206 * we don't bother saving the existing thread state before switching.
208 * At this point we are in a critical section and this cpu owns the
209 * thread's token, which serves as an interlock until the switchout is
212 ENTRY(cpu_exit_switch)
214 * Get us out of the vmspace
222 movl PCPU(curthread),%ebx
224 * Switch to the next thread. RET into the restore function, which
225 * expects the new thread in EAX and the old in EBX.
227 * There is a one-instruction window where curthread is the new
228 * thread but %esp still points to the old thread's stack, but
229 * we are protected by a critical section so it is ok.
232 movl %eax,PCPU(curthread)
233 movl TD_SP(%eax),%esp
237 * cpu_heavy_restore() (current thread in %eax on entry)
239 * Restore the thread after an LWKT switch. This entry is normally
240 * called via the LWKT switch restore function, which was pulled
241 * off the thread stack and jumped to.
243 * This entry is only called if the thread was previously saved
244 * using cpu_heavy_switch() (the heavy weight process thread switcher),
245 * or when a new process is initially scheduled. The first thing we
246 * do is clear the TDF_RUNNING bit in the old thread and set it in the
249 * NOTE: The process may be in any state, not necessarily SRUN, because
250 * a preemption switch may interrupt the process and then return via
253 * YYY theoretically we do not have to restore everything here, a lot
254 * of this junk can wait until we return to usermode. But for now
255 * we restore everything.
257 * YYY the PCB crap is really crap, it makes startup a bitch because
258 * we can't switch away.
260 * YYY note: spl check is done in mi_switch when it splx()'s.
263 ENTRY(cpu_heavy_restore)
265 movl TD_PCB(%eax),%edx /* EDX = PCB */
266 movl TD_PROC(%eax),%ecx
268 #if defined(SWTCH_OPTIM_STATS)
269 incl _swtch_optim_stats
272 * Tell the pmap that our cpu is using the VMSPACE now. We cannot
273 * safely test/reload %cr3 until after we have set the bit in the
274 * pmap (remember, we do not hold the MP lock in the switch code).
276 movl P_VMSPACE(%ecx), %ecx /* ECX = vmspace */
277 movl PCPU(cpuid), %esi
278 MPLOCKED btsl %esi, VM_PMAP+PM_ACTIVE(%ecx)
281 * Restore the MMU address space. If it is the same as the last
282 * thread we don't have to invalidate the tlb (i.e. reload cr3).
283 * YYY which naturally also means that the PM_ACTIVE bit had better
284 * already have been set before we set it above, check? YYY
287 movl PCB_CR3(%edx),%ecx
290 #if defined(SWTCH_OPTIM_STATS)
291 decl _swtch_optim_stats
292 incl _tlb_flush_count
297 * Clear TDF_RUNNING flag in old thread only after cleaning up
298 * %cr3. The target thread is already protected by being TDF_RUNQ
299 * so setting TDF_RUNNING isn't as big a deal.
301 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
302 orl $TDF_RUNNING,TD_FLAGS(%eax)
305 * Deal with the PCB extension, restore the private tss
307 movl PCB_EXT(%edx),%edi /* check for a PCB extension */
308 movl $1,%ebx /* maybe mark use of a private tss */
313 * Going back to the common_tss. We may need to update TSS_ESP0
314 * which sets the top of the supervisor stack when entering from
315 * usermode. The PCB is at the top of the stack but we need another
316 * 16 bytes to take vm86 into account.
319 movl %ebx, PCPU(common_tss) + TSS_ESP0
321 cmpl $0,PCPU(private_tss) /* don't have to reload if */
322 je 3f /* already using the common TSS */
324 subl %ebx,%ebx /* unmark use of private tss */
327 * Get the address of the common TSS descriptor for the ltr.
328 * There is no way to get the address of a segment-accessed variable
329 * so we store a self-referential pointer at the base of the per-cpu
330 * data area and add the appropriate offset.
332 movl $gd_common_tssd, %edi
336 * Move the correct TSS descriptor into the GDT slot, then reload
340 movl %ebx,PCPU(private_tss) /* mark/unmark private tss */
341 movl PCPU(tss_gdt), %ebx /* entry in GDT */
346 movl $GPROC0_SEL*8, %esi /* GSEL(entry, SEL_KPL) */
351 * Restore general registers.
353 movl PCB_EBX(%edx),%ebx
354 movl PCB_ESP(%edx),%esp
355 movl PCB_EBP(%edx),%ebp
356 movl PCB_ESI(%edx),%esi
357 movl PCB_EDI(%edx),%edi
358 movl PCB_EIP(%edx),%eax
362 * Restore the user LDT if we have one
364 cmpl $0, PCB_USERLDT(%edx)
366 movl _default_ldt,%eax
367 cmpl PCPU(currentldt),%eax
370 movl %eax,PCPU(currentldt)
377 * Restore the user TLS if we have one
383 * Restore the %gs segment register, which must be done after
384 * loading the user LDT. Since user processes can modify the
385 * register via procfs, this may result in a fault which is
386 * detected by checking the fault address against cpu_switch_load_gs
387 * in i386/i386/trap.c
389 .globl cpu_switch_load_gs
391 movl PCB_GS(%edx),%gs
394 * Restore the DEBUG register state if necessary.
396 movb PCB_FLAGS(%edx),%al
398 jz 1f /* no, skip over */
399 movl PCB_DR6(%edx),%eax /* yes, do the restore */
401 movl PCB_DR3(%edx),%eax
403 movl PCB_DR2(%edx),%eax
405 movl PCB_DR1(%edx),%eax
407 movl PCB_DR0(%edx),%eax
409 movl %dr7,%eax /* load dr7 so as not to disturb */
410 andl $0x0000fc00,%eax /* reserved bits */
412 movl PCB_DR7(%edx),%ebx
413 andl $~0x0000fc00,%ebx
424 * Update pcb, saving current processor state.
430 /* caller's return address - child won't execute this routine */
432 movl %eax,PCB_EIP(%ecx)
435 movl %eax,PCB_CR3(%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)
442 movl %gs,PCB_GS(%ecx)
446 * If npxthread == NULL, then the npx h/w state is irrelevant and the
447 * state had better already be in the pcb. This is true for forks
448 * but not for dumps (the old book-keeping with FP flags in the pcb
449 * always lost for dumps because the dump pcb has 0 flags).
451 * If npxthread != NULL, then we have to save the npx h/w state to
452 * npxthread's pcb and copy it to the requested pcb, or save to the
453 * requested pcb and reload. Copying is easier because we would
454 * have to handle h/w bugs for reloading. We used to lose the
455 * parent's npx state for forks by forgetting to reload.
457 movl PCPU(npxthread),%eax
461 pushl %ecx /* target pcb */
462 movl TD_SAVEFPU(%eax),%eax /* originating savefpu area */
472 pushl $PCB_SAVEFPU_SIZE
473 leal PCB_SAVEFPU(%ecx),%ecx
478 #endif /* NNPX > 0 */
484 * cpu_idle_restore() (current thread in %eax on entry) (one-time execution)
486 * Don't bother setting up any regs other then %ebp so backtraces
487 * don't die. This restore function is used to bootstrap into the
488 * cpu_idle() LWKT only, after that cpu_lwkt_*() will be used for
491 * Clear TDF_RUNNING in old thread only after we've cleaned up %cr3.
493 * If we are an AP we have to call ap_init() before jumping to
494 * cpu_idle(). ap_init() will synchronize with the BP and finish
495 * setting up various ncpu-dependant globaldata fields. This may
496 * happen on UP as well as SMP if we happen to be simulating multiple
499 ENTRY(cpu_idle_restore)
505 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
506 orl $TDF_RUNNING,TD_FLAGS(%eax)
517 * cpu_kthread_restore() (current thread is %eax on entry) (one-time execution)
519 * Don't bother setting up any regs other then %ebp so backtraces
520 * don't die. This restore function is used to bootstrap into an
521 * LWKT based kernel thread only. cpu_lwkt_switch() will be used
524 * Since all of our context is on the stack we are reentrant and
525 * we can release our critical section and enable interrupts early.
527 ENTRY(cpu_kthread_restore)
530 movl TD_PCB(%eax),%edx
533 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
534 orl $TDF_RUNNING,TD_FLAGS(%eax)
535 subl $TDPRI_CRIT,TD_PRI(%eax)
536 popl %eax /* kthread exit function */
537 pushl PCB_EBX(%edx) /* argument to ESI function */
538 pushl %eax /* set exit func as return address */
539 movl PCB_ESI(%edx),%eax
545 * Standard LWKT switching function. Only non-scratch registers are
546 * saved and we don't bother with the MMU state or anything else.
548 * This function is always called while in a critical section.
550 * There is a one-instruction window where curthread is the new
551 * thread but %esp still points to the old thread's stack, but
552 * we are protected by a critical section so it is ok.
556 ENTRY(cpu_lwkt_switch)
557 pushl %ebp /* note: GDB hacked to locate ebp relative to td_sp */
559 movl PCPU(curthread),%ebx
563 /* warning: adjust movl into %eax below if you change the pushes */
567 * Save the FP state if we have used the FP. Note that calling
568 * npxsave will NULL out PCPU(npxthread).
570 * We have to deal with the FP state for LWKT threads in case they
571 * happen to get preempted or block while doing an optimized
572 * bzero/bcopy/memcpy.
574 cmpl %ebx,PCPU(npxthread)
576 pushl TD_SAVEFPU(%ebx)
577 call npxsave /* do it in a big C function */
578 addl $4,%esp /* EAX, ECX, EDX trashed */
580 #endif /* NNPX > 0 */
582 movl 4+20(%esp),%eax /* switch to this thread */
583 pushl $cpu_lwkt_restore
584 movl %esp,TD_SP(%ebx)
585 movl %eax,PCPU(curthread)
586 movl TD_SP(%eax),%esp
589 * eax contains new thread, ebx contains old thread.
594 * cpu_lwkt_restore() (current thread in %eax on entry)
596 * Standard LWKT restore function. This function is always called
597 * while in a critical section.
599 * Warning: due to preemption the restore function can be used to
600 * 'return' to the original thread. Interrupt disablement must be
601 * protected through the switch so we cannot run splz here.
603 * YYY we theoretically do not need to load IdlePTD into cr3, but if
604 * so we need a way to detect when the PTD we are using is being
605 * deleted due to a process exiting.
607 ENTRY(cpu_lwkt_restore)
608 movl IdlePTD,%ecx /* YYY borrow but beware desched/cpuchg/exit */
614 andl $~TDF_RUNNING,TD_FLAGS(%ebx)
615 orl $TDF_RUNNING,TD_FLAGS(%eax)