Merge tag 'timers-urgent-2020-08-14' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

torvalds · torvalds · commit b923f1247b72 · 2020-08-14T14:26:08.000-07:00
Pull timekeeping updates from Thomas Gleixner:
 "A set of timekeeping/VDSO updates:

   - Preparatory work to allow S390 to switch over to the generic VDSO
     implementation.

     S390 requires that the VDSO data pointer is handed in to the
     counter read function when time namespace support is enabled.
     Adding the pointer is a NOOP for all other architectures because
     the compiler is supposed to optimize that out when it is unused in
     the architecture specific inline. The change also solved a similar
     problem for MIPS which fortunately has time namespaces not yet
     enabled.

     S390 needs to update clock related VDSO data independent of the
     timekeeping updates. This was solved so far with yet another
     sequence counter in the S390 implementation. A better solution is
     to utilize the already existing VDSO sequence count for this. The
     core code now exposes helper functions which allow to serialize
     against the timekeeper code and against concurrent readers.

     S390 needs extra data for their clock readout function. The initial
     common VDSO data structure did not provide a way to add that. It
     now has an embedded architecture specific struct embedded which
     defaults to an empty struct.

     Doing this now avoids tree dependencies and conflicts post rc1 and
     allows all other architectures which work on generic VDSO support
     to work from a common upstream base.

   - A trivial comment fix"

* tag 'timers-urgent-2020-08-14' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  time: Delete repeated words in comments
  lib/vdso: Allow to add architecture-specific vdso data
  timekeeping/vsyscall: Provide vdso_update_begin/end()
  vdso/treewide: Add vdso_data pointer argument to __arch_get_hw_counter()
diff --git a/arch/Kconfig b/arch/Kconfig
@@ -972,6 +972,9 @@ config HAVE_SPARSE_SYSCALL_NR
 	  entries at 4000, 5000 and 6000 locations. This option turns on syscall
 	  related optimizations for a given architecture.
 
+config ARCH_HAS_VDSO_DATA
+	bool
+
 source "kernel/gcov/Kconfig"
 
 source "scripts/gcc-plugins/Kconfig"
diff --git a/arch/arm/include/asm/vdso/gettimeofday.h b/arch/arm/include/asm/vdso/gettimeofday.h
@@ -113,7 +113,8 @@ static inline bool arm_vdso_hres_capable(void)
 }
 #define __arch_vdso_hres_capable arm_vdso_hres_capable
 
-static __always_inline u64 __arch_get_hw_counter(int clock_mode)
+static __always_inline u64 __arch_get_hw_counter(int clock_mode,
+						 const struct vdso_data *vd)
 {
 #ifdef CONFIG_ARM_ARCH_TIMER
 	u64 cycle_now;
diff --git a/arch/arm64/include/asm/vdso/compat_gettimeofday.h b/arch/arm64/include/asm/vdso/compat_gettimeofday.h
@@ -103,7 +103,8 @@ int clock_getres32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
 	return ret;
 }
 
-static __always_inline u64 __arch_get_hw_counter(s32 clock_mode)
+static __always_inline u64 __arch_get_hw_counter(s32 clock_mode,
+						 const struct vdso_data *vd)
 {
 	u64 res;
 
diff --git a/arch/arm64/include/asm/vdso/gettimeofday.h b/arch/arm64/include/asm/vdso/gettimeofday.h
@@ -64,7 +64,8 @@ int clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
 	return ret;
 }
 
-static __always_inline u64 __arch_get_hw_counter(s32 clock_mode)
+static __always_inline u64 __arch_get_hw_counter(s32 clock_mode,
+						 const struct vdso_data *vd)
 {
 	u64 res;
 
diff --git a/arch/mips/include/asm/vdso/gettimeofday.h b/arch/mips/include/asm/vdso/gettimeofday.h
@@ -167,15 +167,16 @@ static __always_inline u64 read_gic_count(const struct vdso_data *data)
 
 #endif
 
-static __always_inline u64 __arch_get_hw_counter(s32 clock_mode)
+static __always_inline u64 __arch_get_hw_counter(s32 clock_mode,
+						 const struct vdso_data *vd)
 {
 #ifdef CONFIG_CSRC_R4K
 	if (clock_mode == VDSO_CLOCKMODE_R4K)
 		return read_r4k_count();
 #endif
 #ifdef CONFIG_CLKSRC_MIPS_GIC
 	if (clock_mode == VDSO_CLOCKMODE_GIC)
-		return read_gic_count(get_vdso_data());
+		return read_gic_count(vd);
 #endif
 	/*
 	 * Core checks mode already. So this raced against a concurrent
diff --git a/arch/riscv/include/asm/vdso/gettimeofday.h b/arch/riscv/include/asm/vdso/gettimeofday.h
@@ -60,7 +60,8 @@ int clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
 	return ret;
 }
 
-static __always_inline u64 __arch_get_hw_counter(s32 clock_mode)
+static __always_inline u64 __arch_get_hw_counter(s32 clock_mode,
+						 const struct vdso_data *vd)
 {
 	/*
 	 * The purpose of csr_read(CSR_TIME) is to trap the system into
diff --git a/arch/x86/include/asm/vdso/gettimeofday.h b/arch/x86/include/asm/vdso/gettimeofday.h
@@ -241,7 +241,8 @@ static u64 vread_hvclock(void)
 }
 #endif
 
-static inline u64 __arch_get_hw_counter(s32 clock_mode)
+static inline u64 __arch_get_hw_counter(s32 clock_mode,
+					const struct vdso_data *vd)
 {
 	if (likely(clock_mode == VDSO_CLOCKMODE_TSC))
 		return (u64)rdtsc_ordered();
diff --git a/include/vdso/datapage.h b/include/vdso/datapage.h
@@ -19,6 +19,12 @@
 #include <vdso/time32.h>
 #include <vdso/time64.h>
 
+#ifdef CONFIG_ARCH_HAS_VDSO_DATA
+#include <asm/vdso/data.h>
+#else
+struct arch_vdso_data {};
+#endif
+
 #define VDSO_BASES	(CLOCK_TAI + 1)
 #define VDSO_HRES	(BIT(CLOCK_REALTIME)		| \
 			 BIT(CLOCK_MONOTONIC)		| \
@@ -64,6 +70,8 @@ struct vdso_timestamp {
  * @tz_dsttime:		type of DST correction
  * @hrtimer_res:	hrtimer resolution
  * @__unused:		unused
+ * @arch_data:		architecture specific data (optional, defaults
+ *			to an empty struct)
  *
  * vdso_data will be accessed by 64 bit and compat code at the same time
  * so we should be careful before modifying this structure.
@@ -97,6 +105,8 @@ struct vdso_data {
 	s32			tz_dsttime;
 	u32			hrtimer_res;
 	u32			__unused;
+
+	struct arch_vdso_data	arch_data;
 };
 
 /*
diff --git a/include/vdso/vsyscall.h b/include/vdso/vsyscall.h
@@ -6,6 +6,9 @@
 
 #include <asm/vdso/vsyscall.h>
 
+unsigned long vdso_update_begin(void);
+void vdso_update_end(unsigned long flags);
+
 #endif /* !__ASSEMBLY__ */
 
 #endif /* __VDSO_VSYSCALL_H */
diff --git a/kernel/time/alarmtimer.c b/kernel/time/alarmtimer.c
@@ -192,7 +192,7 @@ static void alarmtimer_dequeue(struct alarm_base *base, struct alarm *alarm)
  * When a alarm timer fires, this runs through the timerqueue to
  * see which alarms expired, and runs those. If there are more alarm
  * timers queued for the future, we set the hrtimer to fire when
- * when the next future alarm timer expires.
+ * the next future alarm timer expires.
  */
 static enum hrtimer_restart alarmtimer_fired(struct hrtimer *timer)
 {
diff --git a/kernel/time/sched_clock.c b/kernel/time/sched_clock.c
@@ -229,7 +229,7 @@ void __init generic_sched_clock_init(void)
 {
 	/*
 	 * If no sched_clock() function has been provided at that point,
-	 * make it the final one one.
+	 * make it the final one.
 	 */
 	if (cd.actual_read_sched_clock == jiffy_sched_clock_read)
 		sched_clock_register(jiffy_sched_clock_read, BITS_PER_LONG, HZ);
diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
@@ -39,7 +39,7 @@ enum timekeeping_adv_mode {
 	TK_ADV_FREQ
 };
 
-static DEFINE_RAW_SPINLOCK(timekeeper_lock);
+DEFINE_RAW_SPINLOCK(timekeeper_lock);
 
 /*
  * The most important data for readout fits into a single 64 byte
@@ -2004,7 +2004,7 @@ static inline unsigned int accumulate_nsecs_to_secs(struct timekeeper *tk)
  * logarithmic_accumulation - shifted accumulation of cycles
  *
  * This functions accumulates a shifted interval of cycles into
- * into a shifted interval nanoseconds. Allows for O(log) accumulation
+ * a shifted interval nanoseconds. Allows for O(log) accumulation
  * loop.
  *
  * Returns the unconsumed cycles.
diff --git a/kernel/time/timekeeping_internal.h b/kernel/time/timekeeping_internal.h
@@ -1,12 +1,14 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 #ifndef _TIMEKEEPING_INTERNAL_H
 #define _TIMEKEEPING_INTERNAL_H
-/*
- * timekeeping debug functions
- */
+
 #include <linux/clocksource.h>
+#include <linux/spinlock.h>
 #include <linux/time.h>
 
+/*
+ * timekeeping debug functions
+ */
 #ifdef CONFIG_DEBUG_FS
 extern void tk_debug_account_sleep_time(const struct timespec64 *t);
 #else
@@ -31,4 +33,7 @@ static inline u64 clocksource_delta(u64 now, u64 last, u64 mask)
 }
 #endif
 
+/* Semi public for serialization of non timekeeper VDSO updates. */
+extern raw_spinlock_t timekeeper_lock;
+
 #endif /* _TIMEKEEPING_INTERNAL_H */
diff --git a/kernel/time/vsyscall.c b/kernel/time/vsyscall.c
@@ -13,6 +13,8 @@
 #include <vdso/helpers.h>
 #include <vdso/vsyscall.h>
 
+#include "timekeeping_internal.h"
+
 static inline void update_vdso_data(struct vdso_data *vdata,
 				    struct timekeeper *tk)
 {
@@ -127,3 +129,42 @@ void update_vsyscall_tz(void)
 
 	__arch_sync_vdso_data(vdata);
 }
+
+/**
+ * vdso_update_begin - Start of a VDSO update section
+ *
+ * Allows architecture code to safely update the architecture specific VDSO
+ * data. Disables interrupts, acquires timekeeper lock to serialize against
+ * concurrent updates from timekeeping and invalidates the VDSO data
+ * sequence counter to prevent concurrent readers from accessing
+ * inconsistent data.
+ *
+ * Returns: Saved interrupt flags which need to be handed in to
+ * vdso_update_end().
+ */
+unsigned long vdso_update_begin(void)
+{
+	struct vdso_data *vdata = __arch_get_k_vdso_data();
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&timekeeper_lock, flags);
+	vdso_write_begin(vdata);
+	return flags;
+}
+
+/**
+ * vdso_update_end - End of a VDSO update section
+ * @flags:	Interrupt flags as returned from vdso_update_begin()
+ *
+ * Pairs with vdso_update_begin(). Marks vdso data consistent, invokes data
+ * synchronization if the architecture requires it, drops timekeeper lock
+ * and restores interrupt flags.
+ */
+void vdso_update_end(unsigned long flags)
+{
+	struct vdso_data *vdata = __arch_get_k_vdso_data();
+
+	vdso_write_end(vdata);
+	__arch_sync_vdso_data(vdata);
+	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
+}
diff --git a/lib/vdso/gettimeofday.c b/lib/vdso/gettimeofday.c
@@ -68,7 +68,7 @@ static int do_hres_timens(const struct vdso_data *vdns, clockid_t clk,
 		if (unlikely(!vdso_clocksource_ok(vd)))
 			return -1;
 
-		cycles = __arch_get_hw_counter(vd->clock_mode);
+		cycles = __arch_get_hw_counter(vd->clock_mode, vd);
 		if (unlikely(!vdso_cycles_ok(cycles)))
 			return -1;
 		ns = vdso_ts->nsec;
@@ -138,7 +138,7 @@ static __always_inline int do_hres(const struct vdso_data *vd, clockid_t clk,
 		if (unlikely(!vdso_clocksource_ok(vd)))
 			return -1;
 
-		cycles = __arch_get_hw_counter(vd->clock_mode);
+		cycles = __arch_get_hw_counter(vd->clock_mode, vd);
 		if (unlikely(!vdso_cycles_ok(cycles)))
 			return -1;
 		ns = vdso_ts->nsec;

Original file line number	Diff line number	Diff line change
`@@ -113,7 +113,8 @@ static inline bool arm_vdso_hres_capable(void)`
`113`	`113`	`}`
`114`	`114`	`#define __arch_vdso_hres_capable arm_vdso_hres_capable`
`115`	`115`
`116`		`-static __always_inline u64 __arch_get_hw_counter(int clock_mode)`
	`116`	`+static __always_inline u64 __arch_get_hw_counter(int clock_mode,`
	`117`	`+ const struct vdso_data *vd)`
`117`	`118`	`{`
`118`	`119`	`#ifdef CONFIG_ARM_ARCH_TIMER`
`119`	`120`	`u64 cycle_now;`
Original file line number	Diff line number	Diff line change
`@@ -103,7 +103,8 @@ int clock_getres32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)`
`103`	`103`	`return ret;`
`104`	`104`	`}`
`105`	`105`
`106`		`-static __always_inline u64 __arch_get_hw_counter(s32 clock_mode)`
	`106`	`+static __always_inline u64 __arch_get_hw_counter(s32 clock_mode,`
	`107`	`+ const struct vdso_data *vd)`
`107`	`108`	`{`
`108`	`109`	`u64 res;`
`109`	`110`
Original file line number	Diff line number	Diff line change
`@@ -64,7 +64,8 @@ int clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)`
`64`	`64`	`return ret;`
`65`	`65`	`}`
`66`	`66`
`67`		`-static __always_inline u64 __arch_get_hw_counter(s32 clock_mode)`
	`67`	`+static __always_inline u64 __arch_get_hw_counter(s32 clock_mode,`
	`68`	`+ const struct vdso_data *vd)`
`68`	`69`	`{`
`69`	`70`	`u64 res;`
`70`	`71`
Original file line number	Diff line number	Diff line change
`@@ -60,7 +60,8 @@ int clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)`
`60`	`60`	`return ret;`
`61`	`61`	`}`
`62`	`62`
`63`		`-static __always_inline u64 __arch_get_hw_counter(s32 clock_mode)`
	`63`	`+static __always_inline u64 __arch_get_hw_counter(s32 clock_mode,`
	`64`	`+ const struct vdso_data *vd)`
`64`	`65`	`{`
`65`	`66`	`/*`
`66`	`67`	`* The purpose of csr_read(CSR_TIME) is to trap the system into`
Original file line number	Diff line number	Diff line change
`@@ -241,7 +241,8 @@ static u64 vread_hvclock(void)`
`241`	`241`	`}`
`242`	`242`	`#endif`
`243`	`243`
`244`		`-static inline u64 __arch_get_hw_counter(s32 clock_mode)`
	`244`	`+static inline u64 __arch_get_hw_counter(s32 clock_mode,`
	`245`	`+ const struct vdso_data *vd)`
`245`	`246`	`{`
`246`	`247`	`if (likely(clock_mode == VDSO_CLOCKMODE_TSC))`
`247`	`248`	`return (u64)rdtsc_ordered();`
Original file line number	Diff line number	Diff line change
`@@ -192,7 +192,7 @@ static void alarmtimer_dequeue(struct alarm_base base, struct alarm alarm)`
`192`	`192`	`* When a alarm timer fires, this runs through the timerqueue to`
`193`	`193`	`* see which alarms expired, and runs those. If there are more alarm`
`194`	`194`	`* timers queued for the future, we set the hrtimer to fire when`
`195`		`- * when the next future alarm timer expires.`
	`195`	`+ * the next future alarm timer expires.`
`196`	`196`	`*/`
`197`	`197`	`static enum hrtimer_restart alarmtimer_fired(struct hrtimer *timer)`
`198`	`198`	`{`