process.c 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769
  1. // SPDX-License-Identifier: GPL-2.0-only
  2. /*
  3. * Based on arch/arm/kernel/process.c
  4. *
  5. * Original Copyright (C) 1995 Linus Torvalds
  6. * Copyright (C) 1996-2000 Russell King - Converted to ARM.
  7. * Copyright (C) 2012 ARM Ltd.
  8. */
  9. #include <linux/compat.h>
  10. #include <linux/efi.h>
  11. #include <linux/elf.h>
  12. #include <linux/export.h>
  13. #include <linux/sched.h>
  14. #include <linux/sched/debug.h>
  15. #include <linux/sched/task.h>
  16. #include <linux/sched/task_stack.h>
  17. #include <linux/kernel.h>
  18. #include <linux/mman.h>
  19. #include <linux/mm.h>
  20. #include <linux/nospec.h>
  21. #include <linux/stddef.h>
  22. #include <linux/sysctl.h>
  23. #include <linux/unistd.h>
  24. #include <linux/user.h>
  25. #include <linux/delay.h>
  26. #include <linux/reboot.h>
  27. #include <linux/interrupt.h>
  28. #include <linux/init.h>
  29. #include <linux/cpu.h>
  30. #include <linux/elfcore.h>
  31. #include <linux/pm.h>
  32. #include <linux/tick.h>
  33. #include <linux/utsname.h>
  34. #include <linux/uaccess.h>
  35. #include <linux/random.h>
  36. #include <linux/hw_breakpoint.h>
  37. #include <linux/personality.h>
  38. #include <linux/notifier.h>
  39. #include <trace/events/power.h>
  40. #include <linux/percpu.h>
  41. #include <linux/thread_info.h>
  42. #include <linux/prctl.h>
  43. #include <linux/stacktrace.h>
  44. #include <trace/hooks/fpsimd.h>
  45. #include <trace/hooks/mpam.h>
  46. #include <asm/alternative.h>
  47. #include <asm/compat.h>
  48. #include <asm/cpufeature.h>
  49. #include <asm/cacheflush.h>
  50. #include <asm/exec.h>
  51. #include <asm/fpsimd.h>
  52. #include <asm/mmu_context.h>
  53. #include <asm/mte.h>
  54. #include <asm/processor.h>
  55. #include <asm/pointer_auth.h>
  56. #include <asm/stacktrace.h>
  57. #include <asm/switch_to.h>
  58. #include <asm/system_misc.h>
  59. #if defined(CONFIG_STACKPROTECTOR) && !defined(CONFIG_STACKPROTECTOR_PER_TASK)
  60. #include <linux/stackprotector.h>
  61. unsigned long __stack_chk_guard __ro_after_init;
  62. EXPORT_SYMBOL(__stack_chk_guard);
  63. #endif
  64. /*
  65. * Function pointers to optional machine specific functions
  66. */
  67. void (*pm_power_off)(void);
  68. EXPORT_SYMBOL_GPL(pm_power_off);
  69. #ifdef CONFIG_HOTPLUG_CPU
  70. void arch_cpu_idle_dead(void)
  71. {
  72. cpu_die();
  73. }
  74. #endif
  75. /*
  76. * Called by kexec, immediately prior to machine_kexec().
  77. *
  78. * This must completely disable all secondary CPUs; simply causing those CPUs
  79. * to execute e.g. a RAM-based pin loop is not sufficient. This allows the
  80. * kexec'd kernel to use any and all RAM as it sees fit, without having to
  81. * avoid any code or data used by any SW CPU pin loop. The CPU hotplug
  82. * functionality embodied in smpt_shutdown_nonboot_cpus() to achieve this.
  83. */
  84. void machine_shutdown(void)
  85. {
  86. smp_shutdown_nonboot_cpus(reboot_cpu);
  87. }
  88. /*
  89. * Halting simply requires that the secondary CPUs stop performing any
  90. * activity (executing tasks, handling interrupts). smp_send_stop()
  91. * achieves this.
  92. */
  93. void machine_halt(void)
  94. {
  95. local_irq_disable();
  96. smp_send_stop();
  97. while (1);
  98. }
  99. /*
  100. * Power-off simply requires that the secondary CPUs stop performing any
  101. * activity (executing tasks, handling interrupts). smp_send_stop()
  102. * achieves this. When the system power is turned off, it will take all CPUs
  103. * with it.
  104. */
  105. void machine_power_off(void)
  106. {
  107. local_irq_disable();
  108. smp_send_stop();
  109. do_kernel_power_off();
  110. }
  111. /*
  112. * Restart requires that the secondary CPUs stop performing any activity
  113. * while the primary CPU resets the system. Systems with multiple CPUs must
  114. * provide a HW restart implementation, to ensure that all CPUs reset at once.
  115. * This is required so that any code running after reset on the primary CPU
  116. * doesn't have to co-ordinate with other CPUs to ensure they aren't still
  117. * executing pre-reset code, and using RAM that the primary CPU's code wishes
  118. * to use. Implementing such co-ordination would be essentially impossible.
  119. */
  120. void machine_restart(char *cmd)
  121. {
  122. /* Disable interrupts first */
  123. local_irq_disable();
  124. smp_send_stop();
  125. /*
  126. * UpdateCapsule() depends on the system being reset via
  127. * ResetSystem().
  128. */
  129. if (efi_enabled(EFI_RUNTIME_SERVICES))
  130. efi_reboot(reboot_mode, NULL);
  131. /* Now call the architecture specific reboot code. */
  132. do_kernel_restart(cmd);
  133. /*
  134. * Whoops - the architecture was unable to reboot.
  135. */
  136. printk("Reboot failed -- System halted\n");
  137. while (1);
  138. }
  139. #define bstr(suffix, str) [PSR_BTYPE_ ## suffix >> PSR_BTYPE_SHIFT] = str
  140. static const char *const btypes[] = {
  141. bstr(NONE, "--"),
  142. bstr( JC, "jc"),
  143. bstr( C, "-c"),
  144. bstr( J , "j-")
  145. };
  146. #undef bstr
  147. static void print_pstate(struct pt_regs *regs)
  148. {
  149. u64 pstate = regs->pstate;
  150. if (compat_user_mode(regs)) {
  151. printk("pstate: %08llx (%c%c%c%c %c %s %s %c%c%c %cDIT %cSSBS)\n",
  152. pstate,
  153. pstate & PSR_AA32_N_BIT ? 'N' : 'n',
  154. pstate & PSR_AA32_Z_BIT ? 'Z' : 'z',
  155. pstate & PSR_AA32_C_BIT ? 'C' : 'c',
  156. pstate & PSR_AA32_V_BIT ? 'V' : 'v',
  157. pstate & PSR_AA32_Q_BIT ? 'Q' : 'q',
  158. pstate & PSR_AA32_T_BIT ? "T32" : "A32",
  159. pstate & PSR_AA32_E_BIT ? "BE" : "LE",
  160. pstate & PSR_AA32_A_BIT ? 'A' : 'a',
  161. pstate & PSR_AA32_I_BIT ? 'I' : 'i',
  162. pstate & PSR_AA32_F_BIT ? 'F' : 'f',
  163. pstate & PSR_AA32_DIT_BIT ? '+' : '-',
  164. pstate & PSR_AA32_SSBS_BIT ? '+' : '-');
  165. } else {
  166. const char *btype_str = btypes[(pstate & PSR_BTYPE_MASK) >>
  167. PSR_BTYPE_SHIFT];
  168. printk("pstate: %08llx (%c%c%c%c %c%c%c%c %cPAN %cUAO %cTCO %cDIT %cSSBS BTYPE=%s)\n",
  169. pstate,
  170. pstate & PSR_N_BIT ? 'N' : 'n',
  171. pstate & PSR_Z_BIT ? 'Z' : 'z',
  172. pstate & PSR_C_BIT ? 'C' : 'c',
  173. pstate & PSR_V_BIT ? 'V' : 'v',
  174. pstate & PSR_D_BIT ? 'D' : 'd',
  175. pstate & PSR_A_BIT ? 'A' : 'a',
  176. pstate & PSR_I_BIT ? 'I' : 'i',
  177. pstate & PSR_F_BIT ? 'F' : 'f',
  178. pstate & PSR_PAN_BIT ? '+' : '-',
  179. pstate & PSR_UAO_BIT ? '+' : '-',
  180. pstate & PSR_TCO_BIT ? '+' : '-',
  181. pstate & PSR_DIT_BIT ? '+' : '-',
  182. pstate & PSR_SSBS_BIT ? '+' : '-',
  183. btype_str);
  184. }
  185. }
  186. void __show_regs(struct pt_regs *regs)
  187. {
  188. int i, top_reg;
  189. u64 lr, sp;
  190. if (compat_user_mode(regs)) {
  191. lr = regs->compat_lr;
  192. sp = regs->compat_sp;
  193. top_reg = 12;
  194. } else {
  195. lr = regs->regs[30];
  196. sp = regs->sp;
  197. top_reg = 29;
  198. }
  199. show_regs_print_info(KERN_DEFAULT);
  200. print_pstate(regs);
  201. if (!user_mode(regs)) {
  202. printk("pc : %pS\n", (void *)regs->pc);
  203. printk("lr : %pS\n", (void *)ptrauth_strip_insn_pac(lr));
  204. } else {
  205. printk("pc : %016llx\n", regs->pc);
  206. printk("lr : %016llx\n", lr);
  207. }
  208. printk("sp : %016llx\n", sp);
  209. if (system_uses_irq_prio_masking())
  210. printk("pmr_save: %08llx\n", regs->pmr_save);
  211. i = top_reg;
  212. while (i >= 0) {
  213. printk("x%-2d: %016llx", i, regs->regs[i]);
  214. while (i-- % 3)
  215. pr_cont(" x%-2d: %016llx", i, regs->regs[i]);
  216. pr_cont("\n");
  217. }
  218. }
  219. void show_regs(struct pt_regs *regs)
  220. {
  221. __show_regs(regs);
  222. dump_backtrace(regs, NULL, KERN_DEFAULT);
  223. }
  224. EXPORT_SYMBOL_GPL(show_regs);
  225. static void tls_thread_flush(void)
  226. {
  227. write_sysreg(0, tpidr_el0);
  228. if (system_supports_tpidr2())
  229. write_sysreg_s(0, SYS_TPIDR2_EL0);
  230. if (is_compat_task()) {
  231. current->thread.uw.tp_value = 0;
  232. /*
  233. * We need to ensure ordering between the shadow state and the
  234. * hardware state, so that we don't corrupt the hardware state
  235. * with a stale shadow state during context switch.
  236. */
  237. barrier();
  238. write_sysreg(0, tpidrro_el0);
  239. }
  240. }
  241. static void flush_tagged_addr_state(void)
  242. {
  243. if (IS_ENABLED(CONFIG_ARM64_TAGGED_ADDR_ABI))
  244. clear_thread_flag(TIF_TAGGED_ADDR);
  245. }
  246. void flush_thread(void)
  247. {
  248. fpsimd_flush_thread();
  249. tls_thread_flush();
  250. flush_ptrace_hw_breakpoint(current);
  251. flush_tagged_addr_state();
  252. }
  253. void arch_release_task_struct(struct task_struct *tsk)
  254. {
  255. fpsimd_release_task(tsk);
  256. }
  257. int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
  258. {
  259. if (current->mm)
  260. fpsimd_preserve_current_state();
  261. *dst = *src;
  262. /* We rely on the above assignment to initialize dst's thread_flags: */
  263. BUILD_BUG_ON(!IS_ENABLED(CONFIG_THREAD_INFO_IN_TASK));
  264. /*
  265. * Detach src's sve_state (if any) from dst so that it does not
  266. * get erroneously used or freed prematurely. dst's copies
  267. * will be allocated on demand later on if dst uses SVE.
  268. * For consistency, also clear TIF_SVE here: this could be done
  269. * later in copy_process(), but to avoid tripping up future
  270. * maintainers it is best not to leave TIF flags and buffers in
  271. * an inconsistent state, even temporarily.
  272. */
  273. dst->thread.sve_state = NULL;
  274. clear_tsk_thread_flag(dst, TIF_SVE);
  275. /*
  276. * In the unlikely event that we create a new thread with ZA
  277. * enabled we should retain the ZA state so duplicate it here.
  278. * This may be shortly freed if we exec() or if CLONE_SETTLS
  279. * but it's simpler to do it here. To avoid confusing the rest
  280. * of the code ensure that we have a sve_state allocated
  281. * whenever za_state is allocated.
  282. */
  283. if (thread_za_enabled(&src->thread)) {
  284. dst->thread.sve_state = kzalloc(sve_state_size(src),
  285. GFP_KERNEL);
  286. if (!dst->thread.sve_state)
  287. return -ENOMEM;
  288. dst->thread.za_state = kmemdup(src->thread.za_state,
  289. za_state_size(src),
  290. GFP_KERNEL);
  291. if (!dst->thread.za_state) {
  292. kfree(dst->thread.sve_state);
  293. dst->thread.sve_state = NULL;
  294. return -ENOMEM;
  295. }
  296. } else {
  297. dst->thread.za_state = NULL;
  298. clear_tsk_thread_flag(dst, TIF_SME);
  299. }
  300. /* clear any pending asynchronous tag fault raised by the parent */
  301. clear_tsk_thread_flag(dst, TIF_MTE_ASYNC_FAULT);
  302. return 0;
  303. }
  304. asmlinkage void ret_from_fork(void) asm("ret_from_fork");
  305. int copy_thread(struct task_struct *p, const struct kernel_clone_args *args)
  306. {
  307. unsigned long clone_flags = args->flags;
  308. unsigned long stack_start = args->stack;
  309. unsigned long tls = args->tls;
  310. struct pt_regs *childregs = task_pt_regs(p);
  311. memset(&p->thread.cpu_context, 0, sizeof(struct cpu_context));
  312. /*
  313. * In case p was allocated the same task_struct pointer as some
  314. * other recently-exited task, make sure p is disassociated from
  315. * any cpu that may have run that now-exited task recently.
  316. * Otherwise we could erroneously skip reloading the FPSIMD
  317. * registers for p.
  318. */
  319. fpsimd_flush_task_state(p);
  320. ptrauth_thread_init_kernel(p);
  321. if (likely(!args->fn)) {
  322. *childregs = *current_pt_regs();
  323. childregs->regs[0] = 0;
  324. /*
  325. * Read the current TLS pointer from tpidr_el0 as it may be
  326. * out-of-sync with the saved value.
  327. */
  328. *task_user_tls(p) = read_sysreg(tpidr_el0);
  329. if (system_supports_tpidr2())
  330. p->thread.tpidr2_el0 = read_sysreg_s(SYS_TPIDR2_EL0);
  331. if (stack_start) {
  332. if (is_compat_thread(task_thread_info(p)))
  333. childregs->compat_sp = stack_start;
  334. else
  335. childregs->sp = stack_start;
  336. }
  337. /*
  338. * If a TLS pointer was passed to clone, use it for the new
  339. * thread. We also reset TPIDR2 if it's in use.
  340. */
  341. if (clone_flags & CLONE_SETTLS) {
  342. p->thread.uw.tp_value = tls;
  343. p->thread.tpidr2_el0 = 0;
  344. }
  345. } else {
  346. /*
  347. * A kthread has no context to ERET to, so ensure any buggy
  348. * ERET is treated as an illegal exception return.
  349. *
  350. * When a user task is created from a kthread, childregs will
  351. * be initialized by start_thread() or start_compat_thread().
  352. */
  353. memset(childregs, 0, sizeof(struct pt_regs));
  354. childregs->pstate = PSR_MODE_EL1h | PSR_IL_BIT;
  355. p->thread.cpu_context.x19 = (unsigned long)args->fn;
  356. p->thread.cpu_context.x20 = (unsigned long)args->fn_arg;
  357. }
  358. p->thread.cpu_context.pc = (unsigned long)ret_from_fork;
  359. p->thread.cpu_context.sp = (unsigned long)childregs;
  360. /*
  361. * For the benefit of the unwinder, set up childregs->stackframe
  362. * as the final frame for the new task.
  363. */
  364. p->thread.cpu_context.fp = (unsigned long)childregs->stackframe;
  365. ptrace_hw_copy_thread(p);
  366. return 0;
  367. }
  368. void tls_preserve_current_state(void)
  369. {
  370. *task_user_tls(current) = read_sysreg(tpidr_el0);
  371. if (system_supports_tpidr2() && !is_compat_task())
  372. current->thread.tpidr2_el0 = read_sysreg_s(SYS_TPIDR2_EL0);
  373. }
  374. static void tls_thread_switch(struct task_struct *next)
  375. {
  376. tls_preserve_current_state();
  377. if (is_compat_thread(task_thread_info(next)))
  378. write_sysreg(next->thread.uw.tp_value, tpidrro_el0);
  379. else if (!arm64_kernel_unmapped_at_el0())
  380. write_sysreg(0, tpidrro_el0);
  381. write_sysreg(*task_user_tls(next), tpidr_el0);
  382. if (system_supports_tpidr2())
  383. write_sysreg_s(next->thread.tpidr2_el0, SYS_TPIDR2_EL0);
  384. }
  385. /*
  386. * Force SSBS state on context-switch, since it may be lost after migrating
  387. * from a CPU which treats the bit as RES0 in a heterogeneous system.
  388. */
  389. static void ssbs_thread_switch(struct task_struct *next)
  390. {
  391. /*
  392. * Nothing to do for kernel threads, but 'regs' may be junk
  393. * (e.g. idle task) so check the flags and bail early.
  394. */
  395. if (unlikely(next->flags & PF_KTHREAD))
  396. return;
  397. /*
  398. * If all CPUs implement the SSBS extension, then we just need to
  399. * context-switch the PSTATE field.
  400. */
  401. if (cpus_have_const_cap(ARM64_SSBS))
  402. return;
  403. spectre_v4_enable_task_mitigation(next);
  404. }
  405. /*
  406. * We store our current task in sp_el0, which is clobbered by userspace. Keep a
  407. * shadow copy so that we can restore this upon entry from userspace.
  408. *
  409. * This is *only* for exception entry from EL0, and is not valid until we
  410. * __switch_to() a user task.
  411. */
  412. DEFINE_PER_CPU(struct task_struct *, __entry_task);
  413. static void entry_task_switch(struct task_struct *next)
  414. {
  415. __this_cpu_write(__entry_task, next);
  416. }
  417. /*
  418. * ARM erratum 1418040 handling, affecting the 32bit view of CNTVCT.
  419. * Ensure access is disabled when switching to a 32bit task, ensure
  420. * access is enabled when switching to a 64bit task.
  421. */
  422. static void erratum_1418040_thread_switch(struct task_struct *next)
  423. {
  424. if (!IS_ENABLED(CONFIG_ARM64_ERRATUM_1418040) ||
  425. !this_cpu_has_cap(ARM64_WORKAROUND_1418040))
  426. return;
  427. if (is_compat_thread(task_thread_info(next)))
  428. sysreg_clear_set(cntkctl_el1, ARCH_TIMER_USR_VCT_ACCESS_EN, 0);
  429. else
  430. sysreg_clear_set(cntkctl_el1, 0, ARCH_TIMER_USR_VCT_ACCESS_EN);
  431. }
  432. static void erratum_1418040_new_exec(void)
  433. {
  434. preempt_disable();
  435. erratum_1418040_thread_switch(current);
  436. preempt_enable();
  437. }
  438. /*
  439. * __switch_to() checks current->thread.sctlr_user as an optimisation. Therefore
  440. * this function must be called with preemption disabled and the update to
  441. * sctlr_user must be made in the same preemption disabled block so that
  442. * __switch_to() does not see the variable update before the SCTLR_EL1 one.
  443. */
  444. void update_sctlr_el1(u64 sctlr)
  445. {
  446. /*
  447. * EnIA must not be cleared while in the kernel as this is necessary for
  448. * in-kernel PAC. It will be cleared on kernel exit if needed.
  449. */
  450. sysreg_clear_set(sctlr_el1, SCTLR_USER_MASK & ~SCTLR_ELx_ENIA, sctlr);
  451. /* ISB required for the kernel uaccess routines when setting TCF0. */
  452. isb();
  453. }
  454. /*
  455. * Thread switching.
  456. */
  457. __notrace_funcgraph __sched
  458. struct task_struct *__switch_to(struct task_struct *prev,
  459. struct task_struct *next)
  460. {
  461. struct task_struct *last;
  462. fpsimd_thread_switch(next);
  463. tls_thread_switch(next);
  464. hw_breakpoint_thread_switch(next);
  465. contextidr_thread_switch(next);
  466. entry_task_switch(next);
  467. ssbs_thread_switch(next);
  468. erratum_1418040_thread_switch(next);
  469. ptrauth_thread_switch_user(next);
  470. /*
  471. * vendor hook is needed before the dsb(),
  472. * because MPAM is related to cache maintenance.
  473. */
  474. trace_android_vh_mpam_set(prev, next);
  475. /*
  476. * Complete any pending TLB or cache maintenance on this CPU in case
  477. * the thread migrates to a different CPU.
  478. * This full barrier is also required by the membarrier system
  479. * call.
  480. */
  481. dsb(ish);
  482. /*
  483. * MTE thread switching must happen after the DSB above to ensure that
  484. * any asynchronous tag check faults have been logged in the TFSR*_EL1
  485. * registers.
  486. */
  487. mte_thread_switch(next);
  488. /* avoid expensive SCTLR_EL1 accesses if no change */
  489. if (prev->thread.sctlr_user != next->thread.sctlr_user)
  490. update_sctlr_el1(next->thread.sctlr_user);
  491. trace_android_vh_is_fpsimd_save(prev, next);
  492. /* the actual thread switch */
  493. last = cpu_switch_to(prev, next);
  494. return last;
  495. }
  496. struct wchan_info {
  497. unsigned long pc;
  498. int count;
  499. };
  500. static bool get_wchan_cb(void *arg, unsigned long pc)
  501. {
  502. struct wchan_info *wchan_info = arg;
  503. if (!in_sched_functions(pc)) {
  504. wchan_info->pc = pc;
  505. return false;
  506. }
  507. return wchan_info->count++ < 16;
  508. }
  509. unsigned long __get_wchan(struct task_struct *p)
  510. {
  511. struct wchan_info wchan_info = {
  512. .pc = 0,
  513. .count = 0,
  514. };
  515. if (!try_get_task_stack(p))
  516. return 0;
  517. arch_stack_walk(get_wchan_cb, &wchan_info, p, NULL);
  518. put_task_stack(p);
  519. return wchan_info.pc;
  520. }
  521. EXPORT_SYMBOL_GPL(get_wchan);
  522. unsigned long arch_align_stack(unsigned long sp)
  523. {
  524. if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
  525. sp -= prandom_u32_max(PAGE_SIZE);
  526. return sp & ~0xf;
  527. }
  528. #ifdef CONFIG_COMPAT
  529. int compat_elf_check_arch(const struct elf32_hdr *hdr)
  530. {
  531. if (!system_supports_32bit_el0())
  532. return false;
  533. if ((hdr)->e_machine != EM_ARM)
  534. return false;
  535. if (!((hdr)->e_flags & EF_ARM_EABI_MASK))
  536. return false;
  537. /*
  538. * Prevent execve() of a 32-bit program from a deadline task
  539. * if the restricted affinity mask would be inadmissible on an
  540. * asymmetric system.
  541. */
  542. return !static_branch_unlikely(&arm64_mismatched_32bit_el0) ||
  543. !dl_task_check_affinity(current, system_32bit_el0_cpumask());
  544. }
  545. #endif
  546. /*
  547. * Called from setup_new_exec() after (COMPAT_)SET_PERSONALITY.
  548. */
  549. void arch_setup_new_exec(void)
  550. {
  551. unsigned long mmflags = 0;
  552. if (is_compat_task()) {
  553. mmflags = MMCF_AARCH32;
  554. /*
  555. * Restrict the CPU affinity mask for a 32-bit task so that
  556. * it contains only 32-bit-capable CPUs.
  557. *
  558. * From the perspective of the task, this looks similar to
  559. * what would happen if the 64-bit-only CPUs were hot-unplugged
  560. * at the point of execve(), although we try a bit harder to
  561. * honour the cpuset hierarchy.
  562. */
  563. if (static_branch_unlikely(&arm64_mismatched_32bit_el0))
  564. force_compatible_cpus_allowed_ptr(current);
  565. } else if (static_branch_unlikely(&arm64_mismatched_32bit_el0)) {
  566. relax_compatible_cpus_allowed_ptr(current);
  567. }
  568. current->mm->context.flags = mmflags;
  569. ptrauth_thread_init_user();
  570. mte_thread_init_user();
  571. erratum_1418040_new_exec();
  572. if (task_spec_ssb_noexec(current)) {
  573. arch_prctl_spec_ctrl_set(current, PR_SPEC_STORE_BYPASS,
  574. PR_SPEC_ENABLE);
  575. }
  576. }
  577. #ifdef CONFIG_ARM64_TAGGED_ADDR_ABI
  578. /*
  579. * Control the relaxed ABI allowing tagged user addresses into the kernel.
  580. */
  581. static unsigned int tagged_addr_disabled;
  582. long set_tagged_addr_ctrl(struct task_struct *task, unsigned long arg)
  583. {
  584. unsigned long valid_mask = PR_TAGGED_ADDR_ENABLE;
  585. struct thread_info *ti = task_thread_info(task);
  586. if (is_compat_thread(ti))
  587. return -EINVAL;
  588. if (system_supports_mte())
  589. valid_mask |= PR_MTE_TCF_SYNC | PR_MTE_TCF_ASYNC \
  590. | PR_MTE_TAG_MASK;
  591. if (arg & ~valid_mask)
  592. return -EINVAL;
  593. /*
  594. * Do not allow the enabling of the tagged address ABI if globally
  595. * disabled via sysctl abi.tagged_addr_disabled.
  596. */
  597. if (arg & PR_TAGGED_ADDR_ENABLE && tagged_addr_disabled)
  598. return -EINVAL;
  599. if (set_mte_ctrl(task, arg) != 0)
  600. return -EINVAL;
  601. update_ti_thread_flag(ti, TIF_TAGGED_ADDR, arg & PR_TAGGED_ADDR_ENABLE);
  602. return 0;
  603. }
  604. long get_tagged_addr_ctrl(struct task_struct *task)
  605. {
  606. long ret = 0;
  607. struct thread_info *ti = task_thread_info(task);
  608. if (is_compat_thread(ti))
  609. return -EINVAL;
  610. if (test_ti_thread_flag(ti, TIF_TAGGED_ADDR))
  611. ret = PR_TAGGED_ADDR_ENABLE;
  612. ret |= get_mte_ctrl(task);
  613. return ret;
  614. }
  615. /*
  616. * Global sysctl to disable the tagged user addresses support. This control
  617. * only prevents the tagged address ABI enabling via prctl() and does not
  618. * disable it for tasks that already opted in to the relaxed ABI.
  619. */
  620. static struct ctl_table tagged_addr_sysctl_table[] = {
  621. {
  622. .procname = "tagged_addr_disabled",
  623. .mode = 0644,
  624. .data = &tagged_addr_disabled,
  625. .maxlen = sizeof(int),
  626. .proc_handler = proc_dointvec_minmax,
  627. .extra1 = SYSCTL_ZERO,
  628. .extra2 = SYSCTL_ONE,
  629. },
  630. { }
  631. };
  632. static int __init tagged_addr_init(void)
  633. {
  634. if (!register_sysctl("abi", tagged_addr_sysctl_table))
  635. return -EINVAL;
  636. return 0;
  637. }
  638. core_initcall(tagged_addr_init);
  639. #endif /* CONFIG_ARM64_TAGGED_ADDR_ABI */
  640. #ifdef CONFIG_BINFMT_ELF
  641. int arch_elf_adjust_prot(int prot, const struct arch_elf_state *state,
  642. bool has_interp, bool is_interp)
  643. {
  644. /*
  645. * For dynamically linked executables the interpreter is
  646. * responsible for setting PROT_BTI on everything except
  647. * itself.
  648. */
  649. if (is_interp != has_interp)
  650. return prot;
  651. if (!(state->flags & ARM64_ELF_BTI))
  652. return prot;
  653. if (prot & PROT_EXEC)
  654. prot |= PROT_BTI;
  655. return prot;
  656. }
  657. #endif