arch/arm64/kernel/fpsimd.c

   1 /*
   2  * FP/SIMD context switching and fault handling
   3  *
   4  * Copyright (C) 2012 ARM Ltd.
   5  * Author: Catalin Marinas <catalin.marinas@arm.com>
   6  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License version 2 as
   9  * published by the Free Software Foundation.
  10  *
  11  * This program is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14  * GNU General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
  18  */
  19
  20 #include <linux/cpu.h>
  21 #include <linux/cpu_pm.h>
  22 #include <linux/kernel.h>
  23 #include <linux/init.h>
  24 #include <linux/sched/signal.h>
  25 #include <linux/signal.h>
  26 #include <linux/hardirq.h>
  27
  28 #include <asm/fpsimd.h>
  29 #include <asm/cputype.h>
  30
  31 #define FPEXC_IOF       (1 << 0)
  32 #define FPEXC_DZF       (1 << 1)
  33 #define FPEXC_OFF       (1 << 2)
  34 #define FPEXC_UFF       (1 << 3)
  35 #define FPEXC_IXF       (1 << 4)
  36 #define FPEXC_IDF       (1 << 7)
  37
  38 /*
  39  * In order to reduce the number of times the FPSIMD state is needlessly saved
  40  * and restored, we need to keep track of two things:
  41  * (a) for each task, we need to remember which CPU was the last one to have
  42  *     the task's FPSIMD state loaded into its FPSIMD registers;
  43  * (b) for each CPU, we need to remember which task's userland FPSIMD state has
  44  *     been loaded into its FPSIMD registers most recently, or whether it has
  45  *     been used to perform kernel mode NEON in the meantime.
  46  *
  47  * For (a), we add a 'cpu' field to struct fpsimd_state, which gets updated to
  48  * the id of the current CPU every time the state is loaded onto a CPU. For (b),
  49  * we add the per-cpu variable 'fpsimd_last_state' (below), which contains the
  50  * address of the userland FPSIMD state of the task that was loaded onto the CPU
  51  * the most recently, or NULL if kernel mode NEON has been performed after that.
  52  *
  53  * With this in place, we no longer have to restore the next FPSIMD state right
  54  * when switching between tasks. Instead, we can defer this check to userland
  55  * resume, at which time we verify whether the CPU's fpsimd_last_state and the
  56  * task's fpsimd_state.cpu are still mutually in sync. If this is the case, we
  57  * can omit the FPSIMD restore.
  58  *
  59  * As an optimization, we use the thread_info flag TIF_FOREIGN_FPSTATE to
  60  * indicate whether or not the userland FPSIMD state of the current task is
  61  * present in the registers. The flag is set unless the FPSIMD registers of this
  62  * CPU currently contain the most recent userland FPSIMD state of the current
  63  * task.
  64  *
  65  * For a certain task, the sequence may look something like this:
  66  * - the task gets scheduled in; if both the task's fpsimd_state.cpu field
  67  *   contains the id of the current CPU, and the CPU's fpsimd_last_state per-cpu
  68  *   variable points to the task's fpsimd_state, the TIF_FOREIGN_FPSTATE flag is
  69  *   cleared, otherwise it is set;
  70  *
  71  * - the task returns to userland; if TIF_FOREIGN_FPSTATE is set, the task's
  72  *   userland FPSIMD state is copied from memory to the registers, the task's
  73  *   fpsimd_state.cpu field is set to the id of the current CPU, the current
  74  *   CPU's fpsimd_last_state pointer is set to this task's fpsimd_state and the
  75  *   TIF_FOREIGN_FPSTATE flag is cleared;
  76  *
  77  * - the task executes an ordinary syscall; upon return to userland, the
  78  *   TIF_FOREIGN_FPSTATE flag will still be cleared, so no FPSIMD state is
  79  *   restored;
  80  *
  81  * - the task executes a syscall which executes some NEON instructions; this is
  82  *   preceded by a call to kernel_neon_begin(), which copies the task's FPSIMD
  83  *   register contents to memory, clears the fpsimd_last_state per-cpu variable
  84  *   and sets the TIF_FOREIGN_FPSTATE flag;
  85  *
  86  * - the task gets preempted after kernel_neon_end() is called; as we have not
  87  *   returned from the 2nd syscall yet, TIF_FOREIGN_FPSTATE is still set so
  88  *   whatever is in the FPSIMD registers is not saved to memory, but discarded.
  89  */
  90 static DEFINE_PER_CPU(struct fpsimd_state *, fpsimd_last_state);
  91
  92 /*
  93  * Trapped FP/ASIMD access.
  94  */
  95 void do_fpsimd_acc(unsigned int esr, struct pt_regs *regs)
  96 {
  97         /* TODO: implement lazy context saving/restoring */
  98         WARN_ON(1);
  99 }
 100
 101 /*
 102  * Raise a SIGFPE for the current process.
 103  */
 104 void do_fpsimd_exc(unsigned int esr, struct pt_regs *regs)
 105 {
 106         siginfo_t info;
 107         unsigned int si_code = 0;
 108
 109         if (esr & FPEXC_IOF)
 110                 si_code = FPE_FLTINV;
 111         else if (esr & FPEXC_DZF)
 112                 si_code = FPE_FLTDIV;
 113         else if (esr & FPEXC_OFF)
 114                 si_code = FPE_FLTOVF;
 115         else if (esr & FPEXC_UFF)
 116                 si_code = FPE_FLTUND;
 117         else if (esr & FPEXC_IXF)
 118                 si_code = FPE_FLTRES;
 119
 120         memset(&info, 0, sizeof(info));
 121         info.si_signo = SIGFPE;
 122         info.si_code = si_code;
 123         info.si_addr = (void __user *)instruction_pointer(regs);
 124
 125         send_sig_info(SIGFPE, &info, current);
 126 }
 127
 128 void fpsimd_thread_switch(struct task_struct *next)
 129 {
 130         if (!system_supports_fpsimd())
 131                 return;
 132         /*
 133          * Save the current FPSIMD state to memory, but only if whatever is in
 134          * the registers is in fact the most recent userland FPSIMD state of
 135          * 'current'.
 136          */
 137         if (current->mm && !test_thread_flag(TIF_FOREIGN_FPSTATE))
 138                 fpsimd_save_state(&current->thread.fpsimd_state);
 139
 140         if (next->mm) {
 141                 /*
 142                  * If we are switching to a task whose most recent userland
 143                  * FPSIMD state is already in the registers of *this* cpu,
 144                  * we can skip loading the state from memory. Otherwise, set
 145                  * the TIF_FOREIGN_FPSTATE flag so the state will be loaded
 146                  * upon the next return to userland.
 147                  */
 148                 struct fpsimd_state *st = &next->thread.fpsimd_state;
 149
 150                 if (__this_cpu_read(fpsimd_last_state) == st
 151                     && st->cpu == smp_processor_id())
 152                         clear_ti_thread_flag(task_thread_info(next),
 153                                              TIF_FOREIGN_FPSTATE);
 154                 else
 155                         set_ti_thread_flag(task_thread_info(next),
 156                                            TIF_FOREIGN_FPSTATE);
 157         }
 158 }
 159
 160 void fpsimd_flush_thread(void)
 161 {
 162         if (!system_supports_fpsimd())
 163                 return;
 164         memset(&current->thread.fpsimd_state, 0, sizeof(struct fpsimd_state));
 165         fpsimd_flush_task_state(current);
 166         set_thread_flag(TIF_FOREIGN_FPSTATE);
 167 }
 168
 169 /*
 170  * Save the userland FPSIMD state of 'current' to memory, but only if the state
 171  * currently held in the registers does in fact belong to 'current'
 172  */
 173 void fpsimd_preserve_current_state(void)
 174 {
 175         if (!system_supports_fpsimd())
 176                 return;
 177         preempt_disable();
 178         if (!test_thread_flag(TIF_FOREIGN_FPSTATE))
 179                 fpsimd_save_state(&current->thread.fpsimd_state);
 180         preempt_enable();
 181 }
 182
 183 /*
 184  * Load the userland FPSIMD state of 'current' from memory, but only if the
 185  * FPSIMD state already held in the registers is /not/ the most recent FPSIMD
 186  * state of 'current'
 187  */
 188 void fpsimd_restore_current_state(void)
 189 {
 190         if (!system_supports_fpsimd())
 191                 return;
 192         preempt_disable();
 193         if (test_and_clear_thread_flag(TIF_FOREIGN_FPSTATE)) {
 194                 struct fpsimd_state *st = &current->thread.fpsimd_state;
 195
 196                 fpsimd_load_state(st);
 197                 this_cpu_write(fpsimd_last_state, st);
 198                 st->cpu = smp_processor_id();
 199         }
 200         preempt_enable();
 201 }
 202
 203 /*
 204  * Load an updated userland FPSIMD state for 'current' from memory and set the
 205  * flag that indicates that the FPSIMD register contents are the most recent
 206  * FPSIMD state of 'current'
 207  */
 208 void fpsimd_update_current_state(struct fpsimd_state *state)
 209 {
 210         if (!system_supports_fpsimd())
 211                 return;
 212         preempt_disable();
 213         fpsimd_load_state(state);
 214         if (test_and_clear_thread_flag(TIF_FOREIGN_FPSTATE)) {
 215                 struct fpsimd_state *st = &current->thread.fpsimd_state;
 216
 217                 this_cpu_write(fpsimd_last_state, st);
 218                 st->cpu = smp_processor_id();
 219         }
 220         preempt_enable();
 221 }
 222
 223 /*
 224  * Invalidate live CPU copies of task t's FPSIMD state
 225  */
 226 void fpsimd_flush_task_state(struct task_struct *t)
 227 {
 228         t->thread.fpsimd_state.cpu = NR_CPUS;
 229 }
 230
 231 #ifdef CONFIG_KERNEL_MODE_NEON
 232
 233 static DEFINE_PER_CPU(struct fpsimd_partial_state, hardirq_fpsimdstate);
 234 static DEFINE_PER_CPU(struct fpsimd_partial_state, softirq_fpsimdstate);
 235
 236 /*
 237  * Kernel-side NEON support functions
 238  */
 239 void kernel_neon_begin_partial(u32 num_regs)
 240 {
 241         if (WARN_ON(!system_supports_fpsimd()))
 242                 return;
 243         if (in_interrupt()) {
 244                 struct fpsimd_partial_state *s = this_cpu_ptr(
 245                         in_irq() ? &hardirq_fpsimdstate : &softirq_fpsimdstate);
 246
 247                 BUG_ON(num_regs > 32);
 248                 fpsimd_save_partial_state(s, roundup(num_regs, 2));
 249         } else {
 250                 /*
 251                  * Save the userland FPSIMD state if we have one and if we
 252                  * haven't done so already. Clear fpsimd_last_state to indicate
 253                  * that there is no longer userland FPSIMD state in the
 254                  * registers.
 255                  */
 256                 preempt_disable();
 257                 if (current->mm &&
 258                     !test_and_set_thread_flag(TIF_FOREIGN_FPSTATE))
 259                         fpsimd_save_state(&current->thread.fpsimd_state);
 260                 this_cpu_write(fpsimd_last_state, NULL);
 261         }
 262 }
 263 EXPORT_SYMBOL(kernel_neon_begin_partial);
 264
 265 void kernel_neon_end(void)
 266 {
 267         if (!system_supports_fpsimd())
 268                 return;
 269         if (in_interrupt()) {
 270                 struct fpsimd_partial_state *s = this_cpu_ptr(
 271                         in_irq() ? &hardirq_fpsimdstate : &softirq_fpsimdstate);
 272                 fpsimd_load_partial_state(s);
 273         } else {
 274                 preempt_enable();
 275         }
 276 }
 277 EXPORT_SYMBOL(kernel_neon_end);
 278
 279 #endif /* CONFIG_KERNEL_MODE_NEON */
 280
 281 #ifdef CONFIG_CPU_PM
 282 static int fpsimd_cpu_pm_notifier(struct notifier_block *self,
 283                                   unsigned long cmd, void *v)
 284 {
 285         switch (cmd) {
 286         case CPU_PM_ENTER:
 287                 if (current->mm && !test_thread_flag(TIF_FOREIGN_FPSTATE))
 288                         fpsimd_save_state(&current->thread.fpsimd_state);
 289                 this_cpu_write(fpsimd_last_state, NULL);
 290                 break;
 291         case CPU_PM_EXIT:
 292                 if (current->mm)
 293                         set_thread_flag(TIF_FOREIGN_FPSTATE);
 294                 break;
 295         case CPU_PM_ENTER_FAILED:
 296         default:
 297                 return NOTIFY_DONE;
 298         }
 299         return NOTIFY_OK;
 300 }
 301
 302 static struct notifier_block fpsimd_cpu_pm_notifier_block = {
 303         .notifier_call = fpsimd_cpu_pm_notifier,
 304 };
 305
 306 static void __init fpsimd_pm_init(void)
 307 {
 308         cpu_pm_register_notifier(&fpsimd_cpu_pm_notifier_block);
 309 }
 310
 311 #else
 312 static inline void fpsimd_pm_init(void) { }
 313 #endif /* CONFIG_CPU_PM */
 314
 315 #ifdef CONFIG_HOTPLUG_CPU
 316 static int fpsimd_cpu_dead(unsigned int cpu)
 317 {
 318         per_cpu(fpsimd_last_state, cpu) = NULL;
 319         return 0;
 320 }
 321
 322 static inline void fpsimd_hotplug_init(void)
 323 {
 324         cpuhp_setup_state_nocalls(CPUHP_ARM64_FPSIMD_DEAD, "arm64/fpsimd:dead",
 325                                   NULL, fpsimd_cpu_dead);
 326 }
 327
 328 #else
 329 static inline void fpsimd_hotplug_init(void) { }
 330 #endif
 331
 332 /*
 333  * FP/SIMD support code initialisation.
 334  */
 335 static int __init fpsimd_init(void)
 336 {
 337         if (elf_hwcap & HWCAP_FP) {
 338                 fpsimd_pm_init();
 339                 fpsimd_hotplug_init();
 340         } else {
 341                 pr_notice("Floating-point is not implemented\n");
 342         }
 343
 344         if (!(elf_hwcap & HWCAP_ASIMD))
 345                 pr_notice("Advanced SIMD is not implemented\n");
 346
 347         return 0;
 348 }
 349 late_initcall(fpsimd_init);