Merge tag 'uml-for-linux-6.19-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/uml/linux

Pull UML updates from Johannes Berg:
 "Apart from the usual small churn, we have

   - initial SMP support (only kernel)

   - major vDSO cleanups (and fixes for 32-bit)"

* tag 'uml-for-linux-6.19-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/uml/linux: (33 commits)
  um: Disable KASAN_INLINE when STATIC_LINK is selected
  um: Don't rename vmap to kernel_vmap
  um: drivers: virtio: use string choices helper
  um: Always set up AT_HWCAP and AT_PLATFORM
  x86/um: Remove FIXADDR_USER_START and FIXADDR_USE_END
  um: Remove __access_ok_vsyscall()
  um: Remove redundant range check from __access_ok_vsyscall()
  um: Remove fixaddr_user_init()
  x86/um: Drop gate area handling
  x86/um: Do not inherit vDSO from host
  um: Split out default elf_aux_hwcap
  x86/um: Move ELF_PLATFORM fallback to x86-specific code
  um: Split out default elf_aux_platform
  um: Avoid circular dependency on asm-offsets in pgtable.h
  um: Enable SMP support on x86
  asm-generic: percpu: Add assembly guard
  um: vdso: Remove getcpu support on x86
  um: Add initial SMP support
  um: Define timers on a per-CPU basis
  um: Determine sleep based on need_resched()
  ...

1 files changed, 60 insertions, 18 deletions

diff --git a/arch/um/os-Linux/time.c b/arch/um/os-Linux/time.c
index 4d5591d96d8c..13ebc86918d4 100644
--- a/arch/um/os-Linux/time.c
+++ b/arch/um/os-Linux/time.c
@@ -11,12 +11,15 @@
 #include <errno.h>
 #include <signal.h>
 #include <time.h>
+#include <sys/signalfd.h>
 #include <sys/time.h>
 #include <kern_util.h>
 #include <os.h>
+#include <smp.h>
 #include <string.h>
+#include "internal.h"
 
-static timer_t event_high_res_timer = 0;
+static timer_t event_high_res_timer[CONFIG_NR_CPUS] = { 0 };
 
 static inline long long timespec_to_ns(const struct timespec *ts)
 {
@@ -31,20 +34,31 @@ long long os_persistent_clock_emulation(void)
 	return timespec_to_ns(&realtime_tp);
 }
 
+#ifndef sigev_notify_thread_id
+#define sigev_notify_thread_id _sigev_un._tid
+#endif
+
 /**
  * os_timer_create() - create an new posix (interval) timer
  */
 int os_timer_create(void)
 {
-	timer_t *t = &event_high_res_timer;
+	int cpu = uml_curr_cpu();
+	timer_t *t = &event_high_res_timer[cpu];
+	struct sigevent sev = {
+		.sigev_notify = SIGEV_THREAD_ID,
+		.sigev_signo = SIGALRM,
+		.sigev_value.sival_ptr = t,
+		.sigev_notify_thread_id = gettid(),
+	};
 
-	if (timer_create(CLOCK_MONOTONIC, NULL, t) == -1)
+	if (timer_create(CLOCK_MONOTONIC, &sev, t) == -1)
 		return -1;
 
 	return 0;
 }
 
-int os_timer_set_interval(unsigned long long nsecs)
+int os_timer_set_interval(int cpu, unsigned long long nsecs)
 {
 	struct itimerspec its;
 
@@ -54,13 +68,13 @@ int os_timer_set_interval(unsigned long long nsecs)
 	its.it_interval.tv_sec = nsecs / UM_NSEC_PER_SEC;
 	its.it_interval.tv_nsec = nsecs % UM_NSEC_PER_SEC;
 
-	if (timer_settime(event_high_res_timer, 0, &its, NULL) == -1)
+	if (timer_settime(event_high_res_timer[cpu], 0, &its, NULL) == -1)
 		return -errno;
 
 	return 0;
 }
 
-int os_timer_one_shot(unsigned long long nsecs)
+int os_timer_one_shot(int cpu, unsigned long long nsecs)
 {
 	struct itimerspec its = {
 		.it_value.tv_sec = nsecs / UM_NSEC_PER_SEC,
@@ -70,19 +84,20 @@ int os_timer_one_shot(unsigned long long nsecs)
 		.it_interval.tv_nsec = 0, // we cheat here
 	};
 
-	timer_settime(event_high_res_timer, 0, &its, NULL);
+	timer_settime(event_high_res_timer[cpu], 0, &its, NULL);
 	return 0;
 }
 
 /**
  * os_timer_disable() - disable the posix (interval) timer
+ * @cpu: the CPU for which the timer is to be disabled
  */
-void os_timer_disable(void)
+void os_timer_disable(int cpu)
 {
 	struct itimerspec its;
 
 	memset(&its, 0, sizeof(struct itimerspec));
-	timer_settime(event_high_res_timer, 0, &its, NULL);
+	timer_settime(event_high_res_timer[cpu], 0, &its, NULL);
 }
 
 long long os_nsecs(void)
@@ -93,23 +108,50 @@ long long os_nsecs(void)
 	return timespec_to_ns(&ts);
 }
 
+static __thread int wake_signals;
+
+void os_idle_prepare(void)
+{
+	sigset_t set;
+
+	sigemptyset(&set);
+	sigaddset(&set, SIGALRM);
+	sigaddset(&set, IPI_SIGNAL);
+
+	/*
+	 * We need to use signalfd rather than sigsuspend in idle sleep
+	 * because the IPI signal is a real-time signal that carries data,
+	 * and unlike handling SIGALRM, we cannot simply flag it in
+	 * signals_pending.
+	 */
+	wake_signals = signalfd(-1, &set, SFD_CLOEXEC);
+	if (wake_signals < 0)
+		panic("Failed to create signal FD, errno = %d", errno);
+}
+
 /**
  * os_idle_sleep() - sleep until interrupted
  */
 void os_idle_sleep(void)
 {
-	struct itimerspec its;
-	sigset_t set, old;
+	sigset_t set;
 
-	/* block SIGALRM while we analyze the timer state */
+	/*
+	 * Block SIGALRM while performing the need_resched check.
+	 * Note that, because IRQs are disabled, the IPI signal is
+	 * already blocked.
+	 */
 	sigemptyset(&set);
 	sigaddset(&set, SIGALRM);
-	sigprocmask(SIG_BLOCK, &set, &old);
+	sigprocmask(SIG_BLOCK, &set, NULL);
+
+	/*
+	 * Because disabling IRQs does not block SIGALRM, it is also
+	 * necessary to check for any pending timer alarms.
+	 */
+	if (!uml_need_resched() && !timer_alarm_pending())
+		os_poll(1, &wake_signals);
 
-	/* check the timer, and if it'll fire then wait for it */
-	timer_gettime(event_high_res_timer, &its);
-	if (its.it_value.tv_sec || its.it_value.tv_nsec)
-		sigsuspend(&old);
-	/* either way, restore the signal mask */
+	/* Restore the signal mask. */
 	sigprocmask(SIG_UNBLOCK, &set, NULL);
 }