1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * linux/arch/arm/vfp/vfpmodule.c
4 *
5 * Copyright (C) 2004 ARM Limited.
6 * Written by Deep Blue Solutions Limited.
7 */
8 #include <linux/types.h>
9 #include <linux/cpu.h>
10 #include <linux/cpu_pm.h>
11 #include <linux/hardirq.h>
12 #include <linux/kernel.h>
13 #include <linux/notifier.h>
14 #include <linux/signal.h>
15 #include <linux/sched/signal.h>
16 #include <linux/smp.h>
17 #include <linux/init.h>
18 #include <linux/uaccess.h>
19 #include <linux/user.h>
20 #include <linux/export.h>
21
22 #include <asm/cp15.h>
23 #include <asm/cputype.h>
24 #include <asm/system_info.h>
25 #include <asm/thread_notify.h>
26 #include <asm/traps.h>
27 #include <asm/vfp.h>
28
29 #include "vfpinstr.h"
30 #include "vfp.h"
31
32 /*
33 * Our undef handlers (in entry.S)
34 */
35 asmlinkage void vfp_support_entry(void);
36 asmlinkage void vfp_null_entry(void);
37
38 asmlinkage void (*vfp_vector)(void) = vfp_null_entry;
39
40 /*
41 * Dual-use variable.
42 * Used in startup: set to non-zero if VFP checks fail
43 * After startup, holds VFP architecture
44 */
45 static unsigned int __initdata VFP_arch;
46
47 /*
48 * The pointer to the vfpstate structure of the thread which currently
49 * owns the context held in the VFP hardware, or NULL if the hardware
50 * context is invalid.
51 *
52 * For UP, this is sufficient to tell which thread owns the VFP context.
53 * However, for SMP, we also need to check the CPU number stored in the
54 * saved state too to catch migrations.
55 */
56 union vfp_state *vfp_current_hw_state[NR_CPUS];
57
58 /*
59 * Is 'thread's most up to date state stored in this CPUs hardware?
60 * Must be called from non-preemptible context.
61 */
vfp_state_in_hw(unsigned int cpu,struct thread_info * thread)62 static bool vfp_state_in_hw(unsigned int cpu, struct thread_info *thread)
63 {
64 #ifdef CONFIG_SMP
65 if (thread->vfpstate.hard.cpu != cpu)
66 return false;
67 #endif
68 return vfp_current_hw_state[cpu] == &thread->vfpstate;
69 }
70
71 /*
72 * Force a reload of the VFP context from the thread structure. We do
73 * this by ensuring that access to the VFP hardware is disabled, and
74 * clear vfp_current_hw_state. Must be called from non-preemptible context.
75 */
vfp_force_reload(unsigned int cpu,struct thread_info * thread)76 static void vfp_force_reload(unsigned int cpu, struct thread_info *thread)
77 {
78 if (vfp_state_in_hw(cpu, thread)) {
79 fmxr(FPEXC, fmrx(FPEXC) & ~FPEXC_EN);
80 vfp_current_hw_state[cpu] = NULL;
81 }
82 #ifdef CONFIG_SMP
83 thread->vfpstate.hard.cpu = NR_CPUS;
84 #endif
85 }
86
87 /*
88 * Per-thread VFP initialization.
89 */
vfp_thread_flush(struct thread_info * thread)90 static void vfp_thread_flush(struct thread_info *thread)
91 {
92 union vfp_state *vfp = &thread->vfpstate;
93 unsigned int cpu;
94
95 /*
96 * Disable VFP to ensure we initialize it first. We must ensure
97 * that the modification of vfp_current_hw_state[] and hardware
98 * disable are done for the same CPU and without preemption.
99 *
100 * Do this first to ensure that preemption won't overwrite our
101 * state saving should access to the VFP be enabled at this point.
102 */
103 cpu = get_cpu();
104 if (vfp_current_hw_state[cpu] == vfp)
105 vfp_current_hw_state[cpu] = NULL;
106 fmxr(FPEXC, fmrx(FPEXC) & ~FPEXC_EN);
107 put_cpu();
108
109 memset(vfp, 0, sizeof(union vfp_state));
110
111 vfp->hard.fpexc = FPEXC_EN;
112 vfp->hard.fpscr = FPSCR_ROUND_NEAREST;
113 #ifdef CONFIG_SMP
114 vfp->hard.cpu = NR_CPUS;
115 #endif
116 }
117
vfp_thread_exit(struct thread_info * thread)118 static void vfp_thread_exit(struct thread_info *thread)
119 {
120 /* release case: Per-thread VFP cleanup. */
121 union vfp_state *vfp = &thread->vfpstate;
122 unsigned int cpu = get_cpu();
123
124 if (vfp_current_hw_state[cpu] == vfp)
125 vfp_current_hw_state[cpu] = NULL;
126 put_cpu();
127 }
128
vfp_thread_copy(struct thread_info * thread)129 static void vfp_thread_copy(struct thread_info *thread)
130 {
131 struct thread_info *parent = current_thread_info();
132
133 vfp_sync_hwstate(parent);
134 thread->vfpstate = parent->vfpstate;
135 #ifdef CONFIG_SMP
136 thread->vfpstate.hard.cpu = NR_CPUS;
137 #endif
138 }
139
140 /*
141 * When this function is called with the following 'cmd's, the following
142 * is true while this function is being run:
143 * THREAD_NOFTIFY_SWTICH:
144 * - the previously running thread will not be scheduled onto another CPU.
145 * - the next thread to be run (v) will not be running on another CPU.
146 * - thread->cpu is the local CPU number
147 * - not preemptible as we're called in the middle of a thread switch
148 * THREAD_NOTIFY_FLUSH:
149 * - the thread (v) will be running on the local CPU, so
150 * v === current_thread_info()
151 * - thread->cpu is the local CPU number at the time it is accessed,
152 * but may change at any time.
153 * - we could be preempted if tree preempt rcu is enabled, so
154 * it is unsafe to use thread->cpu.
155 * THREAD_NOTIFY_EXIT
156 * - we could be preempted if tree preempt rcu is enabled, so
157 * it is unsafe to use thread->cpu.
158 */
vfp_notifier(struct notifier_block * self,unsigned long cmd,void * v)159 static int vfp_notifier(struct notifier_block *self, unsigned long cmd, void *v)
160 {
161 struct thread_info *thread = v;
162 u32 fpexc;
163 #ifdef CONFIG_SMP
164 unsigned int cpu;
165 #endif
166
167 switch (cmd) {
168 case THREAD_NOTIFY_SWITCH:
169 fpexc = fmrx(FPEXC);
170
171 #ifdef CONFIG_SMP
172 cpu = thread->cpu;
173
174 /*
175 * On SMP, if VFP is enabled, save the old state in
176 * case the thread migrates to a different CPU. The
177 * restoring is done lazily.
178 */
179 if ((fpexc & FPEXC_EN) && vfp_current_hw_state[cpu])
180 vfp_save_state(vfp_current_hw_state[cpu], fpexc);
181 #endif
182
183 /*
184 * Always disable VFP so we can lazily save/restore the
185 * old state.
186 */
187 fmxr(FPEXC, fpexc & ~FPEXC_EN);
188 break;
189
190 case THREAD_NOTIFY_FLUSH:
191 vfp_thread_flush(thread);
192 break;
193
194 case THREAD_NOTIFY_EXIT:
195 vfp_thread_exit(thread);
196 break;
197
198 case THREAD_NOTIFY_COPY:
199 vfp_thread_copy(thread);
200 break;
201 }
202
203 return NOTIFY_DONE;
204 }
205
206 static struct notifier_block vfp_notifier_block = {
207 .notifier_call = vfp_notifier,
208 };
209
210 /*
211 * Raise a SIGFPE for the current process.
212 * sicode describes the signal being raised.
213 */
vfp_raise_sigfpe(unsigned int sicode,struct pt_regs * regs)214 static void vfp_raise_sigfpe(unsigned int sicode, struct pt_regs *regs)
215 {
216 /*
217 * This is the same as NWFPE, because it's not clear what
218 * this is used for
219 */
220 current->thread.error_code = 0;
221 current->thread.trap_no = 6;
222
223 send_sig_fault(SIGFPE, sicode,
224 (void __user *)(instruction_pointer(regs) - 4),
225 current);
226 }
227
vfp_panic(char * reason,u32 inst)228 static void vfp_panic(char *reason, u32 inst)
229 {
230 int i;
231
232 pr_err("VFP: Error: %s\n", reason);
233 pr_err("VFP: EXC 0x%08x SCR 0x%08x INST 0x%08x\n",
234 fmrx(FPEXC), fmrx(FPSCR), inst);
235 for (i = 0; i < 32; i += 2)
236 pr_err("VFP: s%2u: 0x%08x s%2u: 0x%08x\n",
237 i, vfp_get_float(i), i+1, vfp_get_float(i+1));
238 }
239
240 /*
241 * Process bitmask of exception conditions.
242 */
vfp_raise_exceptions(u32 exceptions,u32 inst,u32 fpscr,struct pt_regs * regs)243 static void vfp_raise_exceptions(u32 exceptions, u32 inst, u32 fpscr, struct pt_regs *regs)
244 {
245 int si_code = 0;
246
247 pr_debug("VFP: raising exceptions %08x\n", exceptions);
248
249 if (exceptions == VFP_EXCEPTION_ERROR) {
250 vfp_panic("unhandled bounce", inst);
251 vfp_raise_sigfpe(FPE_FLTINV, regs);
252 return;
253 }
254
255 /*
256 * If any of the status flags are set, update the FPSCR.
257 * Comparison instructions always return at least one of
258 * these flags set.
259 */
260 if (exceptions & (FPSCR_N|FPSCR_Z|FPSCR_C|FPSCR_V))
261 fpscr &= ~(FPSCR_N|FPSCR_Z|FPSCR_C|FPSCR_V);
262
263 fpscr |= exceptions;
264
265 fmxr(FPSCR, fpscr);
266
267 #define RAISE(stat,en,sig) \
268 if (exceptions & stat && fpscr & en) \
269 si_code = sig;
270
271 /*
272 * These are arranged in priority order, least to highest.
273 */
274 RAISE(FPSCR_DZC, FPSCR_DZE, FPE_FLTDIV);
275 RAISE(FPSCR_IXC, FPSCR_IXE, FPE_FLTRES);
276 RAISE(FPSCR_UFC, FPSCR_UFE, FPE_FLTUND);
277 RAISE(FPSCR_OFC, FPSCR_OFE, FPE_FLTOVF);
278 RAISE(FPSCR_IOC, FPSCR_IOE, FPE_FLTINV);
279
280 if (si_code)
281 vfp_raise_sigfpe(si_code, regs);
282 }
283
284 /*
285 * Emulate a VFP instruction.
286 */
vfp_emulate_instruction(u32 inst,u32 fpscr,struct pt_regs * regs)287 static u32 vfp_emulate_instruction(u32 inst, u32 fpscr, struct pt_regs *regs)
288 {
289 u32 exceptions = VFP_EXCEPTION_ERROR;
290
291 pr_debug("VFP: emulate: INST=0x%08x SCR=0x%08x\n", inst, fpscr);
292
293 if (INST_CPRTDO(inst)) {
294 if (!INST_CPRT(inst)) {
295 /*
296 * CPDO
297 */
298 if (vfp_single(inst)) {
299 exceptions = vfp_single_cpdo(inst, fpscr);
300 } else {
301 exceptions = vfp_double_cpdo(inst, fpscr);
302 }
303 } else {
304 /*
305 * A CPRT instruction can not appear in FPINST2, nor
306 * can it cause an exception. Therefore, we do not
307 * have to emulate it.
308 */
309 }
310 } else {
311 /*
312 * A CPDT instruction can not appear in FPINST2, nor can
313 * it cause an exception. Therefore, we do not have to
314 * emulate it.
315 */
316 }
317 return exceptions & ~VFP_NAN_FLAG;
318 }
319
320 /*
321 * Package up a bounce condition.
322 */
VFP_bounce(u32 trigger,u32 fpexc,struct pt_regs * regs)323 void VFP_bounce(u32 trigger, u32 fpexc, struct pt_regs *regs)
324 {
325 u32 fpscr, orig_fpscr, fpsid, exceptions;
326
327 pr_debug("VFP: bounce: trigger %08x fpexc %08x\n", trigger, fpexc);
328
329 /*
330 * At this point, FPEXC can have the following configuration:
331 *
332 * EX DEX IXE
333 * 0 1 x - synchronous exception
334 * 1 x 0 - asynchronous exception
335 * 1 x 1 - sychronous on VFP subarch 1 and asynchronous on later
336 * 0 0 1 - synchronous on VFP9 (non-standard subarch 1
337 * implementation), undefined otherwise
338 *
339 * Clear various bits and enable access to the VFP so we can
340 * handle the bounce.
341 */
342 fmxr(FPEXC, fpexc & ~(FPEXC_EX|FPEXC_DEX|FPEXC_FP2V|FPEXC_VV|FPEXC_TRAP_MASK));
343
344 fpsid = fmrx(FPSID);
345 orig_fpscr = fpscr = fmrx(FPSCR);
346
347 /*
348 * Check for the special VFP subarch 1 and FPSCR.IXE bit case
349 */
350 if ((fpsid & FPSID_ARCH_MASK) == (1 << FPSID_ARCH_BIT)
351 && (fpscr & FPSCR_IXE)) {
352 /*
353 * Synchronous exception, emulate the trigger instruction
354 */
355 goto emulate;
356 }
357
358 if (fpexc & FPEXC_EX) {
359 #ifndef CONFIG_CPU_FEROCEON
360 /*
361 * Asynchronous exception. The instruction is read from FPINST
362 * and the interrupted instruction has to be restarted.
363 */
364 trigger = fmrx(FPINST);
365 regs->ARM_pc -= 4;
366 #endif
367 } else if (!(fpexc & FPEXC_DEX)) {
368 /*
369 * Illegal combination of bits. It can be caused by an
370 * unallocated VFP instruction but with FPSCR.IXE set and not
371 * on VFP subarch 1.
372 */
373 vfp_raise_exceptions(VFP_EXCEPTION_ERROR, trigger, fpscr, regs);
374 goto exit;
375 }
376
377 /*
378 * Modify fpscr to indicate the number of iterations remaining.
379 * If FPEXC.EX is 0, FPEXC.DEX is 1 and the FPEXC.VV bit indicates
380 * whether FPEXC.VECITR or FPSCR.LEN is used.
381 */
382 if (fpexc & (FPEXC_EX | FPEXC_VV)) {
383 u32 len;
384
385 len = fpexc + (1 << FPEXC_LENGTH_BIT);
386
387 fpscr &= ~FPSCR_LENGTH_MASK;
388 fpscr |= (len & FPEXC_LENGTH_MASK) << (FPSCR_LENGTH_BIT - FPEXC_LENGTH_BIT);
389 }
390
391 /*
392 * Handle the first FP instruction. We used to take note of the
393 * FPEXC bounce reason, but this appears to be unreliable.
394 * Emulate the bounced instruction instead.
395 */
396 exceptions = vfp_emulate_instruction(trigger, fpscr, regs);
397 if (exceptions)
398 vfp_raise_exceptions(exceptions, trigger, orig_fpscr, regs);
399
400 /*
401 * If there isn't a second FP instruction, exit now. Note that
402 * the FPEXC.FP2V bit is valid only if FPEXC.EX is 1.
403 */
404 if ((fpexc & (FPEXC_EX | FPEXC_FP2V)) != (FPEXC_EX | FPEXC_FP2V))
405 goto exit;
406
407 /*
408 * The barrier() here prevents fpinst2 being read
409 * before the condition above.
410 */
411 barrier();
412 trigger = fmrx(FPINST2);
413
414 emulate:
415 exceptions = vfp_emulate_instruction(trigger, orig_fpscr, regs);
416 if (exceptions)
417 vfp_raise_exceptions(exceptions, trigger, orig_fpscr, regs);
418 exit:
419 local_bh_enable();
420 }
421
vfp_enable(void * unused)422 static void vfp_enable(void *unused)
423 {
424 u32 access;
425
426 BUG_ON(preemptible());
427 access = get_copro_access();
428
429 /*
430 * Enable full access to VFP (cp10 and cp11)
431 */
432 set_copro_access(access | CPACC_FULL(10) | CPACC_FULL(11));
433 }
434
435 /* Called by platforms on which we want to disable VFP because it may not be
436 * present on all CPUs within a SMP complex. Needs to be called prior to
437 * vfp_init().
438 */
vfp_disable(void)439 void __init vfp_disable(void)
440 {
441 if (VFP_arch) {
442 pr_debug("%s: should be called prior to vfp_init\n", __func__);
443 return;
444 }
445 VFP_arch = 1;
446 }
447
448 #ifdef CONFIG_CPU_PM
vfp_pm_suspend(void)449 static int vfp_pm_suspend(void)
450 {
451 struct thread_info *ti = current_thread_info();
452 u32 fpexc = fmrx(FPEXC);
453
454 /* if vfp is on, then save state for resumption */
455 if (fpexc & FPEXC_EN) {
456 pr_debug("%s: saving vfp state\n", __func__);
457 vfp_save_state(&ti->vfpstate, fpexc);
458
459 /* disable, just in case */
460 fmxr(FPEXC, fmrx(FPEXC) & ~FPEXC_EN);
461 } else if (vfp_current_hw_state[ti->cpu]) {
462 #ifndef CONFIG_SMP
463 fmxr(FPEXC, fpexc | FPEXC_EN);
464 vfp_save_state(vfp_current_hw_state[ti->cpu], fpexc);
465 fmxr(FPEXC, fpexc);
466 #endif
467 }
468
469 /* clear any information we had about last context state */
470 vfp_current_hw_state[ti->cpu] = NULL;
471
472 return 0;
473 }
474
vfp_pm_resume(void)475 static void vfp_pm_resume(void)
476 {
477 /* ensure we have access to the vfp */
478 vfp_enable(NULL);
479
480 /* and disable it to ensure the next usage restores the state */
481 fmxr(FPEXC, fmrx(FPEXC) & ~FPEXC_EN);
482 }
483
vfp_cpu_pm_notifier(struct notifier_block * self,unsigned long cmd,void * v)484 static int vfp_cpu_pm_notifier(struct notifier_block *self, unsigned long cmd,
485 void *v)
486 {
487 switch (cmd) {
488 case CPU_PM_ENTER:
489 vfp_pm_suspend();
490 break;
491 case CPU_PM_ENTER_FAILED:
492 case CPU_PM_EXIT:
493 vfp_pm_resume();
494 break;
495 }
496 return NOTIFY_OK;
497 }
498
499 static struct notifier_block vfp_cpu_pm_notifier_block = {
500 .notifier_call = vfp_cpu_pm_notifier,
501 };
502
vfp_pm_init(void)503 static void vfp_pm_init(void)
504 {
505 cpu_pm_register_notifier(&vfp_cpu_pm_notifier_block);
506 }
507
508 #else
vfp_pm_init(void)509 static inline void vfp_pm_init(void) { }
510 #endif /* CONFIG_CPU_PM */
511
512 /*
513 * Ensure that the VFP state stored in 'thread->vfpstate' is up to date
514 * with the hardware state.
515 */
vfp_sync_hwstate(struct thread_info * thread)516 void vfp_sync_hwstate(struct thread_info *thread)
517 {
518 unsigned int cpu = get_cpu();
519
520 local_bh_disable();
521
522 if (vfp_state_in_hw(cpu, thread)) {
523 u32 fpexc = fmrx(FPEXC);
524
525 /*
526 * Save the last VFP state on this CPU.
527 */
528 fmxr(FPEXC, fpexc | FPEXC_EN);
529 vfp_save_state(&thread->vfpstate, fpexc | FPEXC_EN);
530 fmxr(FPEXC, fpexc);
531 }
532
533 local_bh_enable();
534 put_cpu();
535 }
536
537 /* Ensure that the thread reloads the hardware VFP state on the next use. */
vfp_flush_hwstate(struct thread_info * thread)538 void vfp_flush_hwstate(struct thread_info *thread)
539 {
540 unsigned int cpu = get_cpu();
541
542 vfp_force_reload(cpu, thread);
543
544 put_cpu();
545 }
546
547 /*
548 * Save the current VFP state into the provided structures and prepare
549 * for entry into a new function (signal handler).
550 */
vfp_preserve_user_clear_hwstate(struct user_vfp * ufp,struct user_vfp_exc * ufp_exc)551 int vfp_preserve_user_clear_hwstate(struct user_vfp *ufp,
552 struct user_vfp_exc *ufp_exc)
553 {
554 struct thread_info *thread = current_thread_info();
555 struct vfp_hard_struct *hwstate = &thread->vfpstate.hard;
556
557 /* Ensure that the saved hwstate is up-to-date. */
558 vfp_sync_hwstate(thread);
559
560 /*
561 * Copy the floating point registers. There can be unused
562 * registers see asm/hwcap.h for details.
563 */
564 memcpy(&ufp->fpregs, &hwstate->fpregs, sizeof(hwstate->fpregs));
565
566 /*
567 * Copy the status and control register.
568 */
569 ufp->fpscr = hwstate->fpscr;
570
571 /*
572 * Copy the exception registers.
573 */
574 ufp_exc->fpexc = hwstate->fpexc;
575 ufp_exc->fpinst = hwstate->fpinst;
576 ufp_exc->fpinst2 = hwstate->fpinst2;
577
578 /* Ensure that VFP is disabled. */
579 vfp_flush_hwstate(thread);
580
581 /*
582 * As per the PCS, clear the length and stride bits for function
583 * entry.
584 */
585 hwstate->fpscr &= ~(FPSCR_LENGTH_MASK | FPSCR_STRIDE_MASK);
586 return 0;
587 }
588
589 /* Sanitise and restore the current VFP state from the provided structures. */
vfp_restore_user_hwstate(struct user_vfp * ufp,struct user_vfp_exc * ufp_exc)590 int vfp_restore_user_hwstate(struct user_vfp *ufp, struct user_vfp_exc *ufp_exc)
591 {
592 struct thread_info *thread = current_thread_info();
593 struct vfp_hard_struct *hwstate = &thread->vfpstate.hard;
594 unsigned long fpexc;
595
596 /* Disable VFP to avoid corrupting the new thread state. */
597 vfp_flush_hwstate(thread);
598
599 /*
600 * Copy the floating point registers. There can be unused
601 * registers see asm/hwcap.h for details.
602 */
603 memcpy(&hwstate->fpregs, &ufp->fpregs, sizeof(hwstate->fpregs));
604 /*
605 * Copy the status and control register.
606 */
607 hwstate->fpscr = ufp->fpscr;
608
609 /*
610 * Sanitise and restore the exception registers.
611 */
612 fpexc = ufp_exc->fpexc;
613
614 /* Ensure the VFP is enabled. */
615 fpexc |= FPEXC_EN;
616
617 /* Ensure FPINST2 is invalid and the exception flag is cleared. */
618 fpexc &= ~(FPEXC_EX | FPEXC_FP2V);
619 hwstate->fpexc = fpexc;
620
621 hwstate->fpinst = ufp_exc->fpinst;
622 hwstate->fpinst2 = ufp_exc->fpinst2;
623
624 return 0;
625 }
626
627 /*
628 * VFP hardware can lose all context when a CPU goes offline.
629 * As we will be running in SMP mode with CPU hotplug, we will save the
630 * hardware state at every thread switch. We clear our held state when
631 * a CPU has been killed, indicating that the VFP hardware doesn't contain
632 * a threads VFP state. When a CPU starts up, we re-enable access to the
633 * VFP hardware. The callbacks below are called on the CPU which
634 * is being offlined/onlined.
635 */
vfp_dying_cpu(unsigned int cpu)636 static int vfp_dying_cpu(unsigned int cpu)
637 {
638 vfp_current_hw_state[cpu] = NULL;
639 return 0;
640 }
641
vfp_starting_cpu(unsigned int unused)642 static int vfp_starting_cpu(unsigned int unused)
643 {
644 vfp_enable(NULL);
645 return 0;
646 }
647
648 #ifdef CONFIG_KERNEL_MODE_NEON
649
vfp_kmode_exception(struct pt_regs * regs,unsigned int instr)650 static int vfp_kmode_exception(struct pt_regs *regs, unsigned int instr)
651 {
652 /*
653 * If we reach this point, a floating point exception has been raised
654 * while running in kernel mode. If the NEON/VFP unit was enabled at the
655 * time, it means a VFP instruction has been issued that requires
656 * software assistance to complete, something which is not currently
657 * supported in kernel mode.
658 * If the NEON/VFP unit was disabled, and the location pointed to below
659 * is properly preceded by a call to kernel_neon_begin(), something has
660 * caused the task to be scheduled out and back in again. In this case,
661 * rebuilding and running with CONFIG_DEBUG_ATOMIC_SLEEP enabled should
662 * be helpful in localizing the problem.
663 */
664 if (fmrx(FPEXC) & FPEXC_EN)
665 pr_crit("BUG: unsupported FP instruction in kernel mode\n");
666 else
667 pr_crit("BUG: FP instruction issued in kernel mode with FP unit disabled\n");
668 pr_crit("FPEXC == 0x%08x\n", fmrx(FPEXC));
669 return 1;
670 }
671
672 static struct undef_hook vfp_kmode_exception_hook[] = {{
673 .instr_mask = 0xfe000000,
674 .instr_val = 0xf2000000,
675 .cpsr_mask = MODE_MASK | PSR_T_BIT,
676 .cpsr_val = SVC_MODE,
677 .fn = vfp_kmode_exception,
678 }, {
679 .instr_mask = 0xff100000,
680 .instr_val = 0xf4000000,
681 .cpsr_mask = MODE_MASK | PSR_T_BIT,
682 .cpsr_val = SVC_MODE,
683 .fn = vfp_kmode_exception,
684 }, {
685 .instr_mask = 0xef000000,
686 .instr_val = 0xef000000,
687 .cpsr_mask = MODE_MASK | PSR_T_BIT,
688 .cpsr_val = SVC_MODE | PSR_T_BIT,
689 .fn = vfp_kmode_exception,
690 }, {
691 .instr_mask = 0xff100000,
692 .instr_val = 0xf9000000,
693 .cpsr_mask = MODE_MASK | PSR_T_BIT,
694 .cpsr_val = SVC_MODE | PSR_T_BIT,
695 .fn = vfp_kmode_exception,
696 }, {
697 .instr_mask = 0x0c000e00,
698 .instr_val = 0x0c000a00,
699 .cpsr_mask = MODE_MASK,
700 .cpsr_val = SVC_MODE,
701 .fn = vfp_kmode_exception,
702 }};
703
vfp_kmode_exception_hook_init(void)704 static int __init vfp_kmode_exception_hook_init(void)
705 {
706 int i;
707
708 for (i = 0; i < ARRAY_SIZE(vfp_kmode_exception_hook); i++)
709 register_undef_hook(&vfp_kmode_exception_hook[i]);
710 return 0;
711 }
712 subsys_initcall(vfp_kmode_exception_hook_init);
713
714 /*
715 * Kernel-side NEON support functions
716 */
kernel_neon_begin(void)717 void kernel_neon_begin(void)
718 {
719 struct thread_info *thread = current_thread_info();
720 unsigned int cpu;
721 u32 fpexc;
722
723 local_bh_disable();
724
725 /*
726 * Kernel mode NEON is only allowed outside of hardirq context with
727 * preemption and softirq processing disabled. This will make sure that
728 * the kernel mode NEON register contents never need to be preserved.
729 */
730 BUG_ON(in_hardirq());
731 cpu = __smp_processor_id();
732
733 fpexc = fmrx(FPEXC) | FPEXC_EN;
734 fmxr(FPEXC, fpexc);
735
736 /*
737 * Save the userland NEON/VFP state. Under UP,
738 * the owner could be a task other than 'current'
739 */
740 if (vfp_state_in_hw(cpu, thread))
741 vfp_save_state(&thread->vfpstate, fpexc);
742 #ifndef CONFIG_SMP
743 else if (vfp_current_hw_state[cpu] != NULL)
744 vfp_save_state(vfp_current_hw_state[cpu], fpexc);
745 #endif
746 vfp_current_hw_state[cpu] = NULL;
747 }
748 EXPORT_SYMBOL(kernel_neon_begin);
749
kernel_neon_end(void)750 void kernel_neon_end(void)
751 {
752 /* Disable the NEON/VFP unit. */
753 fmxr(FPEXC, fmrx(FPEXC) & ~FPEXC_EN);
754 local_bh_enable();
755 }
756 EXPORT_SYMBOL(kernel_neon_end);
757
758 #endif /* CONFIG_KERNEL_MODE_NEON */
759
vfp_detect(struct pt_regs * regs,unsigned int instr)760 static int __init vfp_detect(struct pt_regs *regs, unsigned int instr)
761 {
762 VFP_arch = UINT_MAX; /* mark as not present */
763 regs->ARM_pc += 4;
764 return 0;
765 }
766
767 static struct undef_hook vfp_detect_hook __initdata = {
768 .instr_mask = 0x0c000e00,
769 .instr_val = 0x0c000a00,
770 .cpsr_mask = MODE_MASK,
771 .cpsr_val = SVC_MODE,
772 .fn = vfp_detect,
773 };
774
775 /*
776 * VFP support code initialisation.
777 */
vfp_init(void)778 static int __init vfp_init(void)
779 {
780 unsigned int vfpsid;
781 unsigned int cpu_arch = cpu_architecture();
782 unsigned int isar6;
783
784 /*
785 * Enable the access to the VFP on all online CPUs so the
786 * following test on FPSID will succeed.
787 */
788 if (cpu_arch >= CPU_ARCH_ARMv6)
789 on_each_cpu(vfp_enable, NULL, 1);
790
791 /*
792 * First check that there is a VFP that we can use.
793 * The handler is already setup to just log calls, so
794 * we just need to read the VFPSID register.
795 */
796 register_undef_hook(&vfp_detect_hook);
797 barrier();
798 vfpsid = fmrx(FPSID);
799 barrier();
800 unregister_undef_hook(&vfp_detect_hook);
801 vfp_vector = vfp_null_entry;
802
803 pr_info("VFP support v0.3: ");
804 if (VFP_arch) {
805 pr_cont("not present\n");
806 return 0;
807 /* Extract the architecture on CPUID scheme */
808 } else if ((read_cpuid_id() & 0x000f0000) == 0x000f0000) {
809 VFP_arch = vfpsid & FPSID_CPUID_ARCH_MASK;
810 VFP_arch >>= FPSID_ARCH_BIT;
811 /*
812 * Check for the presence of the Advanced SIMD
813 * load/store instructions, integer and single
814 * precision floating point operations. Only check
815 * for NEON if the hardware has the MVFR registers.
816 */
817 if (IS_ENABLED(CONFIG_NEON) &&
818 (fmrx(MVFR1) & 0x000fff00) == 0x00011100)
819 elf_hwcap |= HWCAP_NEON;
820
821 if (IS_ENABLED(CONFIG_VFPv3)) {
822 u32 mvfr0 = fmrx(MVFR0);
823 if (((mvfr0 & MVFR0_DP_MASK) >> MVFR0_DP_BIT) == 0x2 ||
824 ((mvfr0 & MVFR0_SP_MASK) >> MVFR0_SP_BIT) == 0x2) {
825 elf_hwcap |= HWCAP_VFPv3;
826 /*
827 * Check for VFPv3 D16 and VFPv4 D16. CPUs in
828 * this configuration only have 16 x 64bit
829 * registers.
830 */
831 if ((mvfr0 & MVFR0_A_SIMD_MASK) == 1)
832 /* also v4-D16 */
833 elf_hwcap |= HWCAP_VFPv3D16;
834 else
835 elf_hwcap |= HWCAP_VFPD32;
836 }
837
838 if ((fmrx(MVFR1) & 0xf0000000) == 0x10000000)
839 elf_hwcap |= HWCAP_VFPv4;
840 if (((fmrx(MVFR1) & MVFR1_ASIMDHP_MASK) >> MVFR1_ASIMDHP_BIT) == 0x2)
841 elf_hwcap |= HWCAP_ASIMDHP;
842 if (((fmrx(MVFR1) & MVFR1_FPHP_MASK) >> MVFR1_FPHP_BIT) == 0x3)
843 elf_hwcap |= HWCAP_FPHP;
844 }
845
846 /*
847 * Check for the presence of Advanced SIMD Dot Product
848 * instructions.
849 */
850 isar6 = read_cpuid_ext(CPUID_EXT_ISAR6);
851 if (cpuid_feature_extract_field(isar6, 4) == 0x1)
852 elf_hwcap |= HWCAP_ASIMDDP;
853 /*
854 * Check for the presence of Advanced SIMD Floating point
855 * half-precision multiplication instructions.
856 */
857 if (cpuid_feature_extract_field(isar6, 8) == 0x1)
858 elf_hwcap |= HWCAP_ASIMDFHM;
859 /*
860 * Check for the presence of Advanced SIMD Bfloat16
861 * floating point instructions.
862 */
863 if (cpuid_feature_extract_field(isar6, 20) == 0x1)
864 elf_hwcap |= HWCAP_ASIMDBF16;
865 /*
866 * Check for the presence of Advanced SIMD and floating point
867 * Int8 matrix multiplication instructions instructions.
868 */
869 if (cpuid_feature_extract_field(isar6, 24) == 0x1)
870 elf_hwcap |= HWCAP_I8MM;
871
872 /* Extract the architecture version on pre-cpuid scheme */
873 } else {
874 if (vfpsid & FPSID_NODOUBLE) {
875 pr_cont("no double precision support\n");
876 return 0;
877 }
878
879 VFP_arch = (vfpsid & FPSID_ARCH_MASK) >> FPSID_ARCH_BIT;
880 }
881
882 cpuhp_setup_state_nocalls(CPUHP_AP_ARM_VFP_STARTING,
883 "arm/vfp:starting", vfp_starting_cpu,
884 vfp_dying_cpu);
885
886 vfp_vector = vfp_support_entry;
887
888 thread_register_notifier(&vfp_notifier_block);
889 vfp_pm_init();
890
891 /*
892 * We detected VFP, and the support code is
893 * in place; report VFP support to userspace.
894 */
895 elf_hwcap |= HWCAP_VFP;
896
897 pr_cont("implementor %02x architecture %d part %02x variant %x rev %x\n",
898 (vfpsid & FPSID_IMPLEMENTER_MASK) >> FPSID_IMPLEMENTER_BIT,
899 VFP_arch,
900 (vfpsid & FPSID_PART_MASK) >> FPSID_PART_BIT,
901 (vfpsid & FPSID_VARIANT_MASK) >> FPSID_VARIANT_BIT,
902 (vfpsid & FPSID_REV_MASK) >> FPSID_REV_BIT);
903
904 return 0;
905 }
906
907 core_initcall(vfp_init);
908