From 8d3e994241e6bcc7ead2b918c4f15b7683afa90a Mon Sep 17 00:00:00 2001 From: Suzuki K Poulose Date: Tue, 10 Jul 2018 09:57:58 +0100 Subject: arm_pmu: Clean up maximum period handling Each PMU defines their max_period of the counter as the maximum value that can be counted. Since all the PMU backends support 32bit counters by default, let us remove the redundant field. No functional changes. Cc: Will Deacon Acked-by: Mark Rutland Reviewed-by: Julien Thierry Signed-off-by: Suzuki K Poulose Signed-off-by: Will Deacon --- include/linux/perf/arm_pmu.h | 1 - 1 file changed, 1 deletion(-) (limited to 'include/linux') diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h index ad5444491975..12c30a22fc8d 100644 --- a/include/linux/perf/arm_pmu.h +++ b/include/linux/perf/arm_pmu.h @@ -94,7 +94,6 @@ struct arm_pmu { void (*reset)(void *); int (*map_event)(struct perf_event *event); int num_events; - u64 max_period; bool secure_access; /* 32-bit ARM only */ #define ARMV8_PMUV3_MAX_COMMON_EVENTS 0x40 DECLARE_BITMAP(pmceid_bitmap, ARMV8_PMUV3_MAX_COMMON_EVENTS); -- cgit v1.2.3-71-gd317 From 3a95200d3f89afd8b67f39d88d36cc7ec96ce385 Mon Sep 17 00:00:00 2001 From: Suzuki K Poulose Date: Tue, 10 Jul 2018 09:57:59 +0100 Subject: arm_pmu: Change API to support 64bit counter values Convert the {read/write}_counter APIs to handle 64bit values to enable supporting chained event counters. The backends still use 32bit values and we pass them 32bit values only. So in effect there are no functional changes. Cc: Will Deacon Acked-by: Mark Rutland Reviewed-by: Julien Thierry Signed-off-by: Suzuki K Poulose Signed-off-by: Will Deacon --- arch/arm/kernel/perf_event_v6.c | 4 ++-- arch/arm/kernel/perf_event_v7.c | 4 ++-- arch/arm/kernel/perf_event_xscale.c | 8 ++++---- arch/arm64/kernel/perf_event.c | 9 ++++----- include/linux/perf/arm_pmu.h | 4 ++-- 5 files changed, 14 insertions(+), 15 deletions(-) (limited to 'include/linux') diff --git a/arch/arm/kernel/perf_event_v6.c b/arch/arm/kernel/perf_event_v6.c index f64a6bfebcec..0729f9841ef4 100644 --- a/arch/arm/kernel/perf_event_v6.c +++ b/arch/arm/kernel/perf_event_v6.c @@ -233,7 +233,7 @@ armv6_pmcr_counter_has_overflowed(unsigned long pmcr, return ret; } -static inline u32 armv6pmu_read_counter(struct perf_event *event) +static inline u64 armv6pmu_read_counter(struct perf_event *event) { struct hw_perf_event *hwc = &event->hw; int counter = hwc->idx; @@ -251,7 +251,7 @@ static inline u32 armv6pmu_read_counter(struct perf_event *event) return value; } -static inline void armv6pmu_write_counter(struct perf_event *event, u32 value) +static inline void armv6pmu_write_counter(struct perf_event *event, u64 value) { struct hw_perf_event *hwc = &event->hw; int counter = hwc->idx; diff --git a/arch/arm/kernel/perf_event_v7.c b/arch/arm/kernel/perf_event_v7.c index 2cf1ca2925c8..973043dd6187 100644 --- a/arch/arm/kernel/perf_event_v7.c +++ b/arch/arm/kernel/perf_event_v7.c @@ -743,7 +743,7 @@ static inline void armv7_pmnc_select_counter(int idx) isb(); } -static inline u32 armv7pmu_read_counter(struct perf_event *event) +static inline u64 armv7pmu_read_counter(struct perf_event *event) { struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); struct hw_perf_event *hwc = &event->hw; @@ -763,7 +763,7 @@ static inline u32 armv7pmu_read_counter(struct perf_event *event) return value; } -static inline void armv7pmu_write_counter(struct perf_event *event, u32 value) +static inline void armv7pmu_write_counter(struct perf_event *event, u64 value) { struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); struct hw_perf_event *hwc = &event->hw; diff --git a/arch/arm/kernel/perf_event_xscale.c b/arch/arm/kernel/perf_event_xscale.c index c4f029458b52..942230fe0426 100644 --- a/arch/arm/kernel/perf_event_xscale.c +++ b/arch/arm/kernel/perf_event_xscale.c @@ -316,7 +316,7 @@ static void xscale1pmu_stop(struct arm_pmu *cpu_pmu) raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static inline u32 xscale1pmu_read_counter(struct perf_event *event) +static inline u64 xscale1pmu_read_counter(struct perf_event *event) { struct hw_perf_event *hwc = &event->hw; int counter = hwc->idx; @@ -337,7 +337,7 @@ static inline u32 xscale1pmu_read_counter(struct perf_event *event) return val; } -static inline void xscale1pmu_write_counter(struct perf_event *event, u32 val) +static inline void xscale1pmu_write_counter(struct perf_event *event, u64 val) { struct hw_perf_event *hwc = &event->hw; int counter = hwc->idx; @@ -678,7 +678,7 @@ static void xscale2pmu_stop(struct arm_pmu *cpu_pmu) raw_spin_unlock_irqrestore(&events->pmu_lock, flags); } -static inline u32 xscale2pmu_read_counter(struct perf_event *event) +static inline u64 xscale2pmu_read_counter(struct perf_event *event) { struct hw_perf_event *hwc = &event->hw; int counter = hwc->idx; @@ -705,7 +705,7 @@ static inline u32 xscale2pmu_read_counter(struct perf_event *event) return val; } -static inline void xscale2pmu_write_counter(struct perf_event *event, u32 val) +static inline void xscale2pmu_write_counter(struct perf_event *event, u64 val) { struct hw_perf_event *hwc = &event->hw; int counter = hwc->idx; diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c index 678ecffd3724..66a2ffdca6dd 100644 --- a/arch/arm64/kernel/perf_event.c +++ b/arch/arm64/kernel/perf_event.c @@ -512,7 +512,7 @@ static inline int armv8pmu_select_counter(int idx) return idx; } -static inline u32 armv8pmu_read_counter(struct perf_event *event) +static inline u64 armv8pmu_read_counter(struct perf_event *event) { struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); struct hw_perf_event *hwc = &event->hw; @@ -530,7 +530,7 @@ static inline u32 armv8pmu_read_counter(struct perf_event *event) return value; } -static inline void armv8pmu_write_counter(struct perf_event *event, u32 value) +static inline void armv8pmu_write_counter(struct perf_event *event, u64 value) { struct arm_pmu *cpu_pmu = to_arm_pmu(event->pmu); struct hw_perf_event *hwc = &event->hw; @@ -545,9 +545,8 @@ static inline void armv8pmu_write_counter(struct perf_event *event, u32 value) * count using the lower 32bits and we want an interrupt when * it overflows. */ - u64 value64 = 0xffffffff00000000ULL | value; - - write_sysreg(value64, pmccntr_el0); + value |= 0xffffffff00000000ULL; + write_sysreg(value, pmccntr_el0); } else if (armv8pmu_select_counter(idx) == idx) write_sysreg(value, pmxevcntr_el0); } diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h index 12c30a22fc8d..f7126a21df30 100644 --- a/include/linux/perf/arm_pmu.h +++ b/include/linux/perf/arm_pmu.h @@ -87,8 +87,8 @@ struct arm_pmu { struct perf_event *event); int (*set_event_filter)(struct hw_perf_event *evt, struct perf_event_attr *attr); - u32 (*read_counter)(struct perf_event *event); - void (*write_counter)(struct perf_event *event, u32 val); + u64 (*read_counter)(struct perf_event *event); + void (*write_counter)(struct perf_event *event, u64 val); void (*start)(struct arm_pmu *); void (*stop)(struct arm_pmu *); void (*reset)(void *); -- cgit v1.2.3-71-gd317 From e2da97d328d4951d25f6634eda7213f7257417b6 Mon Sep 17 00:00:00 2001 From: Suzuki K Poulose Date: Tue, 10 Jul 2018 09:58:00 +0100 Subject: arm_pmu: Add support for 64bit event counters Each PMU has a set of 32bit event counters. But in some special cases, the events could be counted using counters which are effectively 64bit wide. e.g, Arm V8 PMUv3 has a 64 bit cycle counter which can count only the CPU cycles. Also, the PMU can chain the event counters to effectively count as a 64bit counter. Add support for tracking the events that uses 64bit counters. This only affects the periods set for each counter in the core driver. Cc: Will Deacon Reviewed-by: Julien Thierry Acked-by: Mark Rutland Signed-off-by: Suzuki K Poulose Signed-off-by: Will Deacon --- drivers/perf/arm_pmu.c | 16 ++++++++++------ include/linux/perf/arm_pmu.h | 6 ++++++ 2 files changed, 16 insertions(+), 6 deletions(-) (limited to 'include/linux') diff --git a/drivers/perf/arm_pmu.c b/drivers/perf/arm_pmu.c index 6ddc00da5373..8cad6b535a2c 100644 --- a/drivers/perf/arm_pmu.c +++ b/drivers/perf/arm_pmu.c @@ -28,9 +28,12 @@ static DEFINE_PER_CPU(struct arm_pmu *, cpu_armpmu); static DEFINE_PER_CPU(int, cpu_irq); -static inline u64 arm_pmu_max_period(void) +static inline u64 arm_pmu_event_max_period(struct perf_event *event) { - return (1ULL << 32) - 1; + if (event->hw.flags & ARMPMU_EVT_64BIT) + return GENMASK_ULL(63, 0); + else + return GENMASK_ULL(31, 0); } static int @@ -122,7 +125,7 @@ int armpmu_event_set_period(struct perf_event *event) u64 max_period; int ret = 0; - max_period = arm_pmu_max_period(); + max_period = arm_pmu_event_max_period(event); if (unlikely(left <= -period)) { left = period; local64_set(&hwc->period_left, left); @@ -148,7 +151,7 @@ int armpmu_event_set_period(struct perf_event *event) local64_set(&hwc->prev_count, (u64)-left); - armpmu->write_counter(event, (u64)(-left) & 0xffffffff); + armpmu->write_counter(event, (u64)(-left) & max_period); perf_event_update_userpage(event); @@ -160,7 +163,7 @@ u64 armpmu_event_update(struct perf_event *event) struct arm_pmu *armpmu = to_arm_pmu(event->pmu); struct hw_perf_event *hwc = &event->hw; u64 delta, prev_raw_count, new_raw_count; - u64 max_period = arm_pmu_max_period(); + u64 max_period = arm_pmu_event_max_period(event); again: prev_raw_count = local64_read(&hwc->prev_count); @@ -368,6 +371,7 @@ __hw_perf_event_init(struct perf_event *event) struct hw_perf_event *hwc = &event->hw; int mapping; + hwc->flags = 0; mapping = armpmu->map_event(event); if (mapping < 0) { @@ -410,7 +414,7 @@ __hw_perf_event_init(struct perf_event *event) * is far less likely to overtake the previous one unless * you have some serious IRQ latency issues. */ - hwc->sample_period = arm_pmu_max_period() >> 1; + hwc->sample_period = arm_pmu_event_max_period(event) >> 1; hwc->last_period = hwc->sample_period; local64_set(&hwc->period_left, hwc->sample_period); } diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h index f7126a21df30..10f92e1d8e7b 100644 --- a/include/linux/perf/arm_pmu.h +++ b/include/linux/perf/arm_pmu.h @@ -25,6 +25,12 @@ */ #define ARMPMU_MAX_HWEVENTS 32 +/* + * ARM PMU hw_event flags + */ +/* Event uses a 64bit counter */ +#define ARMPMU_EVT_64BIT 1 + #define HW_OP_UNSUPPORTED 0xFFFF #define C(_x) PERF_COUNT_HW_CACHE_##_x #define CACHE_OP_UNSUPPORTED 0xFFFF -- cgit v1.2.3-71-gd317 From bf1c77b4644f46d2986b7ca5e43e012f0fc8984b Mon Sep 17 00:00:00 2001 From: Mark Rutland Date: Wed, 11 Jul 2018 14:56:50 +0100 Subject: kernel: add ksys_personality() Using this helper allows us to avoid the in-kernel call to the sys_personality() syscall. The ksys_ prefix denotes that this function is meant as a drop-in replacement for the syscall. In particular, it uses the same calling convention as sys_personality(). Since ksys_personality is trivial, it is implemented directly in , as we do for ksys_close() and friends. This helper is necessary to enable conversion of arm64's syscall handling to use pt_regs wrappers. Signed-off-by: Mark Rutland Reviewed-by: Dominik Brodowski Cc: Al Viro Cc: Christoph Hellwig Cc: Dave Martin Signed-off-by: Will Deacon --- include/linux/syscalls.h | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'include/linux') diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h index a368a68cb667..abfe12d8a9c5 100644 --- a/include/linux/syscalls.h +++ b/include/linux/syscalls.h @@ -80,6 +80,7 @@ union bpf_attr; #include #include #include +#include #include #ifdef CONFIG_ARCH_HAS_SYSCALL_WRAPPER @@ -1281,4 +1282,14 @@ static inline long ksys_truncate(const char __user *pathname, loff_t length) return do_sys_truncate(pathname, length); } +static inline unsigned int ksys_personality(unsigned int personality) +{ + unsigned int old = current->personality; + + if (personality != 0xffffffff) + set_personality(personality); + + return old; +} + #endif -- cgit v1.2.3-71-gd317 From 9b54bf9d6a5b30e2cc22b18793e9a4158c5b4882 Mon Sep 17 00:00:00 2001 From: Mark Rutland Date: Wed, 11 Jul 2018 14:56:51 +0100 Subject: kernel: add kcompat_sys_{f,}statfs64() Using this helper allows us to avoid the in-kernel calls to the compat_sys_{f,}statfs64() sycalls, as are necessary for parameter mangling in arm64's compat handling. Following the example of ksys_* functions, kcompat_sys_* functions are intended to be a drop-in replacement for their compat_sys_* counterparts, with the same calling convention. This is necessary to enable conversion of arm64's syscall handling to use pt_regs wrappers. Signed-off-by: Mark Rutland Reviewed-by: Dominik Brodowski Cc: Al Viro Cc: linux-fsdevel@vger.kernel.org Signed-off-by: Will Deacon --- fs/statfs.c | 14 ++++++++++++-- include/linux/compat.h | 11 +++++++++++ 2 files changed, 23 insertions(+), 2 deletions(-) (limited to 'include/linux') diff --git a/fs/statfs.c b/fs/statfs.c index 5b2a24f0f263..f0216629621d 100644 --- a/fs/statfs.c +++ b/fs/statfs.c @@ -335,7 +335,7 @@ static int put_compat_statfs64(struct compat_statfs64 __user *ubuf, struct kstat return 0; } -COMPAT_SYSCALL_DEFINE3(statfs64, const char __user *, pathname, compat_size_t, sz, struct compat_statfs64 __user *, buf) +int kcompat_sys_statfs64(const char __user * pathname, compat_size_t sz, struct compat_statfs64 __user * buf) { struct kstatfs tmp; int error; @@ -349,7 +349,12 @@ COMPAT_SYSCALL_DEFINE3(statfs64, const char __user *, pathname, compat_size_t, s return error; } -COMPAT_SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, compat_size_t, sz, struct compat_statfs64 __user *, buf) +COMPAT_SYSCALL_DEFINE3(statfs64, const char __user *, pathname, compat_size_t, sz, struct compat_statfs64 __user *, buf) +{ + return kcompat_sys_statfs64(pathname, sz, buf); +} + +int kcompat_sys_fstatfs64(unsigned int fd, compat_size_t sz, struct compat_statfs64 __user * buf) { struct kstatfs tmp; int error; @@ -363,6 +368,11 @@ COMPAT_SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, compat_size_t, sz, struct co return error; } +COMPAT_SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, compat_size_t, sz, struct compat_statfs64 __user *, buf) +{ + return kcompat_sys_fstatfs64(fd, sz, buf); +} + /* * This is a copy of sys_ustat, just dealing with a structure layout. * Given how simple this syscall is that apporach is more maintainable diff --git a/include/linux/compat.h b/include/linux/compat.h index c68acc47da57..43f4ed44c5d5 100644 --- a/include/linux/compat.h +++ b/include/linux/compat.h @@ -1028,6 +1028,17 @@ static inline struct compat_timeval ns_to_compat_timeval(s64 nsec) return ctv; } +/* + * Kernel code should not call compat syscalls (i.e., compat_sys_xyzyyz()) + * directly. Instead, use one of the functions which work equivalently, such + * as the kcompat_sys_xyzyyz() functions prototyped below. + */ + +int kcompat_sys_statfs64(const char __user * pathname, compat_size_t sz, + struct compat_statfs64 __user * buf); +int kcompat_sys_fstatfs64(unsigned int fd, compat_size_t sz, + struct compat_statfs64 __user * buf); + #else /* !CONFIG_COMPAT */ #define is_compat_task() (0) -- cgit v1.2.3-71-gd317