summaryrefslogtreecommitdiffstats
path: root/sys/i386/isa/npx.c
blob: fbeb8244bdd30ab9c51031be2c8b96034a7fe6da (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
/*-
 * Copyright (c) 1990 William Jolitz.
 * Copyright (c) 1991 The Regents of the University of California.
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	from: @(#)npx.c	7.2 (Berkeley) 5/12/91
 *	$Id: npx.c,v 1.28 1996/01/04 19:51:14 wollman Exp $
 */

#include "npx.h"
#if NNPX > 0

#include "opt_math_emulate.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/conf.h>
#include <sys/file.h>
#include <sys/proc.h>
#include <sys/devconf.h>
#include <sys/ioctl.h>
#include <sys/syslog.h>
#include <sys/signalvar.h>

#include <machine/cpu.h>
#include <machine/pcb.h>
#include <machine/trap.h>
#include <machine/clock.h>
#include <machine/specialreg.h>

#include <i386/isa/icu.h>
#include <i386/isa/isa_device.h>
#include <i386/isa/isa.h>

/*
 * 387 and 287 Numeric Coprocessor Extension (NPX) Driver.
 */

#ifdef	__GNUC__

#define	fldcw(addr)		__asm("fldcw %0" : : "m" (*(addr)))
#define	fnclex()		__asm("fnclex")
#define	fninit()		__asm("fninit")
#define	fnop()			__asm("fnop")
#define	fnsave(addr)		__asm("fnsave %0" : "=m" (*(addr)))
#define	fnstcw(addr)		__asm("fnstcw %0" : "=m" (*(addr)))
#define	fnstsw(addr)		__asm("fnstsw %0" : "=m" (*(addr)))
#define	fp_divide_by_0()	__asm("fldz; fld1; fdiv %st,%st(1); fnop")
#define	frstor(addr)		__asm("frstor %0" : : "m" (*(addr)))
#define	start_emulating()	__asm("smsw %%ax; orb %0,%%al; lmsw %%ax" \
				      : : "n" (CR0_TS) : "ax")
#define	stop_emulating()	__asm("clts")

#else	/* not __GNUC__ */

void	fldcw		__P((caddr_t addr));
void	fnclex		__P((void));
void	fninit		__P((void));
void	fnop		__P((void));
void	fnsave		__P((caddr_t addr));
void	fnstcw		__P((caddr_t addr));
void	fnstsw		__P((caddr_t addr));
void	fp_divide_by_0	__P((void));
void	frstor		__P((caddr_t addr));
void	start_emulating	__P((void));
void	stop_emulating	__P((void));

#endif	/* __GNUC__ */

typedef u_char bool_t;

static	int	npxattach	__P((struct isa_device *dvp));
static	int	npxprobe	__P((struct isa_device *dvp));
static	int	npxprobe1	__P((struct isa_device *dvp));

struct	isa_driver npxdriver = {
	npxprobe, npxattach, "npx",
};

int	hw_float;		/* XXX currently just alias for npx_exists */

SYSCTL_INT(_hw,HW_FLOATINGPT, floatingpoint,
	CTLFLAG_RD, &hw_float, 0, 
	"Floatingpoint instructions executed in hardware");

static u_int	npx0_imask = SWI_CLOCK_MASK;
struct proc	*npxproc;

static	bool_t			npx_ex16;
static	bool_t			npx_exists;
static	struct gate_descriptor	npx_idt_probeintr;
static	int			npx_intrno;
static	volatile u_int		npx_intrs_while_probing;
static	bool_t			npx_irq13;
static	volatile u_int		npx_traps_while_probing;

/*
 * Special interrupt handlers.  Someday intr0-intr15 will be used to count
 * interrupts.  We'll still need a special exception 16 handler.  The busy
 * latch stuff in probeintr() can be moved to npxprobe().
 */
inthand_t probeintr;
asm
("
	.text
_probeintr:
	ss
	incl	_npx_intrs_while_probing
	pushl	%eax
	movb	$0x20,%al	# EOI (asm in strings loses cpp features)
	outb	%al,$0xa0	# IO_ICU2
	outb	%al,$0x20	# IO_ICU1
	movb	$0,%al
	outb	%al,$0xf0	# clear BUSY# latch
	popl	%eax
	iret
");

inthand_t probetrap;
asm
("
	.text
_probetrap:
	ss
	incl	_npx_traps_while_probing
	fnclex
	iret
");

static struct kern_devconf kdc_npx[NNPX] = { {
	0, 0, 0,		/* filled in by dev_attach */
	"npx", 0, { MDDT_ISA, 0 },
	isa_generic_externalize, 0, 0, ISA_EXTERNALLEN,
	&kdc_isa0,		/* parent */
	0,			/* parentdata */
	DC_UNCONFIGURED,	/* state */
	"Floating-point unit",
	DC_CLS_MISC		/* class */
} };

static inline void
npx_registerdev(struct isa_device *id)
{
	int	unit;

	unit = id->id_unit;
	if (unit != 0)
		kdc_npx[unit] = kdc_npx[0];
	kdc_npx[unit].kdc_unit = unit;
	kdc_npx[unit].kdc_isa = id;
	dev_attach(&kdc_npx[unit]);
}

/*
 * Probe routine.  Initialize cr0 to give correct behaviour for [f]wait
 * whether the device exists or not (XXX should be elsewhere).  Set flags
 * to tell npxattach() what to do.  Modify device struct if npx doesn't
 * need to use interrupts.  Return 1 if device exists.
 */
static int
npxprobe(dvp)
	struct isa_device *dvp;
{
	int	result;
	u_long	save_eflags;
	u_char	save_icu1_mask;
	u_char	save_icu2_mask;
	struct	gate_descriptor save_idt_npxintr;
	struct	gate_descriptor save_idt_npxtrap;
	/*
	 * This routine is now just a wrapper for npxprobe1(), to install
	 * special npx interrupt and trap handlers, to enable npx interrupts
	 * and to disable other interrupts.  Someday isa_configure() will
	 * install suitable handlers and run with interrupts enabled so we
	 * won't need to do so much here.
	 */
	npx_registerdev(dvp);
	npx_intrno = NRSVIDT + ffs(dvp->id_irq) - 1;
	save_eflags = read_eflags();
	disable_intr();
	save_icu1_mask = inb(IO_ICU1 + 1);
	save_icu2_mask = inb(IO_ICU2 + 1);
	save_idt_npxintr = idt[npx_intrno];
	save_idt_npxtrap = idt[16];
	outb(IO_ICU1 + 1, ~(IRQ_SLAVE | dvp->id_irq));
	outb(IO_ICU2 + 1, ~(dvp->id_irq >> 8));
	setidt(16, probetrap, SDT_SYS386TGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
	setidt(npx_intrno, probeintr, SDT_SYS386IGT, SEL_KPL, GSEL(GCODE_SEL, SEL_KPL));
	npx_idt_probeintr = idt[npx_intrno];
	enable_intr();
	result = npxprobe1(dvp);
	disable_intr();
	outb(IO_ICU1 + 1, save_icu1_mask);
	outb(IO_ICU2 + 1, save_icu2_mask);
	idt[npx_intrno] = save_idt_npxintr;
	idt[16] = save_idt_npxtrap;
	write_eflags(save_eflags);
	return (result);
}

static int
npxprobe1(dvp)
	struct isa_device *dvp;
{
	u_short control;
	u_short status;

	/*
	 * Partially reset the coprocessor, if any.  Some BIOS's don't reset
	 * it after a warm boot.
	 */
	outb(0xf1, 0);		/* full reset on some systems, NOP on others */
	outb(0xf0, 0);		/* clear BUSY# latch */
	/*
	 * Prepare to trap all ESC (i.e., NPX) instructions and all WAIT
	 * instructions.  We must set the CR0_MP bit and use the CR0_TS
	 * bit to control the trap, because setting the CR0_EM bit does
	 * not cause WAIT instructions to trap.  It's important to trap
	 * WAIT instructions - otherwise the "wait" variants of no-wait
	 * control instructions would degenerate to the "no-wait" variants
	 * after FP context switches but work correctly otherwise.  It's
	 * particularly important to trap WAITs when there is no NPX -
	 * otherwise the "wait" variants would always degenerate.
	 *
	 * Try setting CR0_NE to get correct error reporting on 486DX's.
	 * Setting it should fail or do nothing on lesser processors.
	 */
	load_cr0(rcr0() | CR0_MP | CR0_NE);
	/*
	 * But don't trap while we're probing.
	 */
	stop_emulating();
	/*
	 * Finish resetting the coprocessor, if any.  If there is an error
	 * pending, then we may get a bogus IRQ13, but probeintr() will handle
	 * it OK.  Bogus halts have never been observed, but we enabled
	 * IRQ13 and cleared the BUSY# latch early to handle them anyway.
	 */
	fninit();
	/*
	 * Don't use fwait here because it might hang.
	 * Don't use fnop here because it usually hangs if there is no FPU.
	 */
	DELAY(1000);		/* wait for any IRQ13 */
#ifdef DIAGNOSTIC
	if (npx_intrs_while_probing != 0)
		printf("fninit caused %u bogus npx interrupt(s)\n",
		       npx_intrs_while_probing);
	if (npx_traps_while_probing != 0)
		printf("fninit caused %u bogus npx trap(s)\n",
		       npx_traps_while_probing);
#endif
	/*
	 * Check for a status of mostly zero.
	 */
	status = 0x5a5a;
	fnstsw(&status);
	if ((status & 0xb8ff) == 0) {
		/*
		 * Good, now check for a proper control word.
		 */
		control = 0x5a5a;
		fnstcw(&control);
		if ((control & 0x1f3f) == 0x033f) {
			hw_float = npx_exists = 1;
			/*
			 * We have an npx, now divide by 0 to see if exception
			 * 16 works.
			 */
			control &= ~(1 << 2);	/* enable divide by 0 trap */
			fldcw(&control);
			npx_traps_while_probing = npx_intrs_while_probing = 0;
			fp_divide_by_0();
			if (npx_traps_while_probing != 0) {
				/*
				 * Good, exception 16 works.
				 */
				npx_ex16 = 1;
				dvp->id_irq = 0;	/* zap the interrupt */
				/*
				 * special return value to flag that we do not
				 * actually use any I/O registers
				 */
				return (-1);
			}
			if (npx_intrs_while_probing != 0) {
				/*
				 * Bad, we are stuck with IRQ13.
				 */
				npx_irq13 = 1;
				/*
				 * npxattach would be too late to set npx0_imask.
				 */
				npx0_imask |= dvp->id_irq;
				return (IO_NPXSIZE);
			}
			/*
			 * Worse, even IRQ13 is broken.  Use emulator.
			 */
		}
	}
	/*
	 * Probe failed, but we want to get to npxattach to initialize the
	 * emulator and say that it has been installed.  XXX handle devices
	 * that aren't really devices better.
	 */
	dvp->id_irq = 0;
	/*
	 * special return value to flag that we do not
	 * actually use any I/O registers
	 */
	return (-1);
}

/*
 * Attach routine - announce which it is, and wire into system
 */
int
npxattach(dvp)
	struct isa_device *dvp;
{
	if (npx_ex16)
		printf("npx%d: INT 16 interface\n", dvp->id_unit);
	else if (npx_irq13)
		;		/* higher level has printed "irq 13" */
#if defined(MATH_EMULATE) || defined(GPL_MATH_EMULATE)
	else if (npx_exists) {
		printf("npx%d: error reporting broken; using 387 emulator\n",
			dvp->id_unit);
		npx_exists = 0;
	} else
		printf("npx%d: 387 emulator\n",dvp->id_unit);
#else
	else
		printf("npx%d: no 387 emulator in kernel!\n", dvp->id_unit);
#endif
	npxinit(__INITIAL_NPXCW__);
	if (npx_exists) {
		kdc_npx[dvp->id_unit].kdc_state = DC_BUSY;
	}
	return (1);		/* XXX unused */
}

/*
 * Initialize floating point unit.
 */
void
npxinit(control)
	u_short control;
{
	struct save87 dummy;

	if (!npx_exists)
		return;
	/*
	 * fninit has the same h/w bugs as fnsave.  Use the detoxified
	 * fnsave to throw away any junk in the fpu.  npxsave() initializes
	 * the fpu and sets npxproc = NULL as important side effects.
	 */
	npxsave(&dummy);
	stop_emulating();
	fldcw(&control);
	if (curpcb != NULL)
		fnsave(&curpcb->pcb_savefpu);
	start_emulating();
}

/*
 * Free coprocessor (if we have it).
 */
void
npxexit(p)
	struct proc *p;
{

	if (p == npxproc)
		npxsave(&curpcb->pcb_savefpu);
	if (npx_exists) {
		u_int	masked_exceptions;

		masked_exceptions = curpcb->pcb_savefpu.sv_env.en_cw
				    & curpcb->pcb_savefpu.sv_env.en_sw & 0x7f;
		/*
		 * Overflow, divde by 0, and invalid operand would have
		 * caused a trap in 1.1.5.
		 */
		if (masked_exceptions & 0x0d)
			log(LOG_ERR,
	"pid %d (%s) exited with masked floating point exceptions 0x%02x\n",
			    p->p_pid, p->p_comm, masked_exceptions);
	}
}

/*
 * Preserve the FP status word, clear FP exceptions, then generate a SIGFPE.
 *
 * Clearing exceptions is necessary mainly to avoid IRQ13 bugs.  We now
 * depend on longjmp() restoring a usable state.  Restoring the state
 * or examining it might fail if we didn't clear exceptions.
 *
 * XXX there is no standard way to tell SIGFPE handlers about the error
 * state.  The old interface:
 *
 *	void handler(int sig, int code, struct sigcontext *scp);
 *
 * is broken because it is non-ANSI and because the FP state is not in
 * struct sigcontext.
 *
 * XXX the FP state is not preserved across signal handlers.  So signal
 * handlers cannot afford to do FP unless they preserve the state or
 * longjmp() out.  Both preserving the state and longjmp()ing may be
 * destroyed by IRQ13 bugs.  Clearing FP exceptions is not an acceptable
 * solution for signals other than SIGFPE.
 */
void
npxintr(unit)
	int unit;
{
	int code;
	struct intrframe *frame;

	if (npxproc == NULL || !npx_exists) {
		printf("npxintr: npxproc = %p, curproc = %p, npx_exists = %d\n",
		       npxproc, curproc, npx_exists);
		panic("npxintr from nowhere");
	}
	if (npxproc != curproc) {
		printf("npxintr: npxproc = %p, curproc = %p, npx_exists = %d\n",
		       npxproc, curproc, npx_exists);
		panic("npxintr from non-current process");
	}

	outb(0xf0, 0);
	fnstsw(&curpcb->pcb_savefpu.sv_ex_sw);
	fnclex();
	fnop();

	/*
	 * Pass exception to process.
	 */
	frame = (struct intrframe *)&unit;	/* XXX */
	if (ISPL(frame->if_cs) == SEL_UPL) {
		/*
		 * Interrupt is essentially a trap, so we can afford to call
		 * the SIGFPE handler (if any) as soon as the interrupt
		 * returns.
		 *
		 * XXX little or nothing is gained from this, and plenty is
		 * lost - the interrupt frame has to contain the trap frame
		 * (this is otherwise only necessary for the rescheduling trap
		 * in doreti, and the frame for that could easily be set up
		 * just before it is used).
		 */
		curproc->p_md.md_regs = &frame->if_es;
#ifdef notyet
		/*
		 * Encode the appropriate code for detailed information on
		 * this exception.
		 */
		code = XXX_ENCODE(curpcb->pcb_savefpu.sv_ex_sw);
#else
		code = 0;	/* XXX */
#endif
		trapsignal(curproc, SIGFPE, code);
	} else {
		/*
		 * Nested interrupt.  These losers occur when:
		 *	o an IRQ13 is bogusly generated at a bogus time, e.g.:
		 *		o immediately after an fnsave or frstor of an
		 *		  error state.
		 *		o a couple of 386 instructions after
		 *		  "fstpl _memvar" causes a stack overflow.
		 *	  These are especially nasty when combined with a
		 *	  trace trap.
		 *	o an IRQ13 occurs at the same time as another higher-
		 *	  priority interrupt.
		 *
		 * Treat them like a true async interrupt.
		 */
		psignal(curproc, SIGFPE);
	}
}

/*
 * Implement device not available (DNA) exception
 *
 * It would be better to switch FP context here (if curproc != npxproc)
 * and not necessarily for every context switch, but it is too hard to
 * access foreign pcb's.
 */
int
npxdna()
{
	if (!npx_exists)
		return (0);
	if (npxproc != NULL) {
		printf("npxdna: npxproc = %p, curproc = %p\n",
		       npxproc, curproc);
		panic("npxdna");
	}
	stop_emulating();
	/*
	 * Record new context early in case frstor causes an IRQ13.
	 */
	npxproc = curproc;
	curpcb->pcb_savefpu.sv_ex_sw = 0;
	/*
	 * The following frstor may cause an IRQ13 when the state being
	 * restored has a pending error.  The error will appear to have been
	 * triggered by the current (npx) user instruction even when that
	 * instruction is a no-wait instruction that should not trigger an
	 * error (e.g., fnclex).  On at least one 486 system all of the
	 * no-wait instructions are broken the same as frstor, so our
	 * treatment does not amplify the breakage.  On at least one
	 * 386/Cyrix 387 system, fnclex works correctly while frstor and
	 * fnsave are broken, so our treatment breaks fnclex if it is the
	 * first FPU instruction after a context switch.
	 */
	frstor(&curpcb->pcb_savefpu);

	return (1);
}

/*
 * Wrapper for fnsave instruction to handle h/w bugs.  If there is an error
 * pending, then fnsave generates a bogus IRQ13 on some systems.  Force
 * any IRQ13 to be handled immediately, and then ignore it.  This routine is
 * often called at splhigh so it must not use many system services.  In
 * particular, it's much easier to install a special handler than to
 * guarantee that it's safe to use npxintr() and its supporting code.
 */
void
npxsave(addr)
	struct save87 *addr;
{
	u_char	icu1_mask;
	u_char	icu2_mask;
	u_char	old_icu1_mask;
	u_char	old_icu2_mask;
	struct gate_descriptor	save_idt_npxintr;

	disable_intr();
	old_icu1_mask = inb(IO_ICU1 + 1);
	old_icu2_mask = inb(IO_ICU2 + 1);
	save_idt_npxintr = idt[npx_intrno];
	outb(IO_ICU1 + 1, old_icu1_mask & ~(IRQ_SLAVE | npx0_imask));
	outb(IO_ICU2 + 1, old_icu2_mask & ~(npx0_imask >> 8));
	idt[npx_intrno] = npx_idt_probeintr;
	enable_intr();
	stop_emulating();
	fnsave(addr);
	fnop();
	start_emulating();
	npxproc = NULL;
	disable_intr();
	icu1_mask = inb(IO_ICU1 + 1);	/* masks may have changed */
	icu2_mask = inb(IO_ICU2 + 1);
	outb(IO_ICU1 + 1,
	     (icu1_mask & ~npx0_imask) | (old_icu1_mask & npx0_imask));
	outb(IO_ICU2 + 1,
	     (icu2_mask & ~(npx0_imask >> 8))
	     | (old_icu2_mask & (npx0_imask >> 8)));
	idt[npx_intrno] = save_idt_npxintr;
	enable_intr();		/* back to usual state */
}

#endif /* NNPX > 0 */
OpenPOWER on IntegriCloud