linux/arch/s390/kernel/perf_event.c
Heiko Carstens 504b73d00a s390/perf: implement perf_callchain_user()
Daan De Meyer and Neal Gompa reported that s390 does not support perf user
stack unwinding.

This was never implemented since this requires user space to be compiled
with the -mbackchain compile option, which until now no distribution
did. However this is going to change with Fedora. Therefore provide a
perf_callchain_user() implementation.

Note that due to the way s390 sets up stack frames the provided call chains
can contain invalid values. This is especially true for the first stack
frame, where it is not possible to tell if the return address has been
written to the stack already or not.

Reported-by: Daan De Meyer <daan.j.demeyer@gmail.com>
Reported-by: Neal Gompa <ngompa@fedoraproject.org>
Closes: https://lore.kernel.org/all/CAO8sHcn3+_qrnvp0580aK7jN0Wion5F7KYeBAa4MnCY4mqABPA@mail.gmail.com/
Link: https://lore.kernel.org/all/20231030123558.10816-A-hca@linux.ibm.com
Reviewed-by: Neal Gompa <ngompa@fedoraproject.org>
Acked-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
2023-11-05 22:34:57 +01:00

265 lines
6.5 KiB
C

// SPDX-License-Identifier: GPL-2.0
/*
* Performance event support for s390x
*
* Copyright IBM Corp. 2012, 2013
* Author(s): Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
*/
#define KMSG_COMPONENT "perf"
#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
#include <linux/kernel.h>
#include <linux/perf_event.h>
#include <linux/kvm_host.h>
#include <linux/percpu.h>
#include <linux/export.h>
#include <linux/seq_file.h>
#include <linux/spinlock.h>
#include <linux/uaccess.h>
#include <linux/compat.h>
#include <linux/sysfs.h>
#include <asm/stacktrace.h>
#include <asm/irq.h>
#include <asm/cpu_mf.h>
#include <asm/lowcore.h>
#include <asm/processor.h>
#include <asm/sysinfo.h>
#include <asm/unwind.h>
static struct kvm_s390_sie_block *sie_block(struct pt_regs *regs)
{
struct stack_frame *stack = (struct stack_frame *) regs->gprs[15];
if (!stack)
return NULL;
return (struct kvm_s390_sie_block *)stack->sie_control_block;
}
static bool is_in_guest(struct pt_regs *regs)
{
if (user_mode(regs))
return false;
#if IS_ENABLED(CONFIG_KVM)
return instruction_pointer(regs) == (unsigned long) &sie_exit;
#else
return false;
#endif
}
static unsigned long guest_is_user_mode(struct pt_regs *regs)
{
return sie_block(regs)->gpsw.mask & PSW_MASK_PSTATE;
}
static unsigned long instruction_pointer_guest(struct pt_regs *regs)
{
return sie_block(regs)->gpsw.addr;
}
unsigned long perf_instruction_pointer(struct pt_regs *regs)
{
return is_in_guest(regs) ? instruction_pointer_guest(regs)
: instruction_pointer(regs);
}
static unsigned long perf_misc_guest_flags(struct pt_regs *regs)
{
return guest_is_user_mode(regs) ? PERF_RECORD_MISC_GUEST_USER
: PERF_RECORD_MISC_GUEST_KERNEL;
}
static unsigned long perf_misc_flags_sf(struct pt_regs *regs)
{
struct perf_sf_sde_regs *sde_regs;
unsigned long flags;
sde_regs = (struct perf_sf_sde_regs *) &regs->int_parm_long;
if (sde_regs->in_guest)
flags = user_mode(regs) ? PERF_RECORD_MISC_GUEST_USER
: PERF_RECORD_MISC_GUEST_KERNEL;
else
flags = user_mode(regs) ? PERF_RECORD_MISC_USER
: PERF_RECORD_MISC_KERNEL;
return flags;
}
unsigned long perf_misc_flags(struct pt_regs *regs)
{
/* Check if the cpum_sf PMU has created the pt_regs structure.
* In this case, perf misc flags can be easily extracted. Otherwise,
* do regular checks on the pt_regs content.
*/
if (regs->int_code == 0x1407 && regs->int_parm == CPU_MF_INT_SF_PRA)
if (!regs->gprs[15])
return perf_misc_flags_sf(regs);
if (is_in_guest(regs))
return perf_misc_guest_flags(regs);
return user_mode(regs) ? PERF_RECORD_MISC_USER
: PERF_RECORD_MISC_KERNEL;
}
static void print_debug_cf(void)
{
struct cpumf_ctr_info cf_info;
int cpu = smp_processor_id();
memset(&cf_info, 0, sizeof(cf_info));
if (!qctri(&cf_info))
pr_info("CPU[%i] CPUM_CF: ver=%u.%u A=%04x E=%04x C=%04x\n",
cpu, cf_info.cfvn, cf_info.csvn,
cf_info.auth_ctl, cf_info.enable_ctl, cf_info.act_ctl);
}
static void print_debug_sf(void)
{
struct hws_qsi_info_block si;
int cpu = smp_processor_id();
memset(&si, 0, sizeof(si));
if (qsi(&si))
return;
pr_info("CPU[%i] CPUM_SF: basic=%i diag=%i min=%lu max=%lu cpu_speed=%u\n",
cpu, si.as, si.ad, si.min_sampl_rate, si.max_sampl_rate,
si.cpu_speed);
if (si.as)
pr_info("CPU[%i] CPUM_SF: Basic-sampling: a=%i e=%i c=%i"
" bsdes=%i tear=%016lx dear=%016lx\n", cpu,
si.as, si.es, si.cs, si.bsdes, si.tear, si.dear);
if (si.ad)
pr_info("CPU[%i] CPUM_SF: Diagnostic-sampling: a=%i e=%i c=%i"
" dsdes=%i tear=%016lx dear=%016lx\n", cpu,
si.ad, si.ed, si.cd, si.dsdes, si.tear, si.dear);
}
void perf_event_print_debug(void)
{
unsigned long flags;
local_irq_save(flags);
if (cpum_cf_avail())
print_debug_cf();
if (cpum_sf_avail())
print_debug_sf();
local_irq_restore(flags);
}
/* Service level infrastructure */
static void sl_print_counter(struct seq_file *m)
{
struct cpumf_ctr_info ci;
memset(&ci, 0, sizeof(ci));
if (qctri(&ci))
return;
seq_printf(m, "CPU-MF: Counter facility: version=%u.%u "
"authorization=%04x\n", ci.cfvn, ci.csvn, ci.auth_ctl);
}
static void sl_print_sampling(struct seq_file *m)
{
struct hws_qsi_info_block si;
memset(&si, 0, sizeof(si));
if (qsi(&si))
return;
if (!si.as && !si.ad)
return;
seq_printf(m, "CPU-MF: Sampling facility: min_rate=%lu max_rate=%lu"
" cpu_speed=%u\n", si.min_sampl_rate, si.max_sampl_rate,
si.cpu_speed);
if (si.as)
seq_printf(m, "CPU-MF: Sampling facility: mode=basic"
" sample_size=%u\n", si.bsdes);
if (si.ad)
seq_printf(m, "CPU-MF: Sampling facility: mode=diagnostic"
" sample_size=%u\n", si.dsdes);
}
static void service_level_perf_print(struct seq_file *m,
struct service_level *sl)
{
if (cpum_cf_avail())
sl_print_counter(m);
if (cpum_sf_avail())
sl_print_sampling(m);
}
static struct service_level service_level_perf = {
.seq_print = service_level_perf_print,
};
static int __init service_level_perf_register(void)
{
return register_service_level(&service_level_perf);
}
arch_initcall(service_level_perf_register);
void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry,
struct pt_regs *regs)
{
struct unwind_state state;
unsigned long addr;
unwind_for_each_frame(&state, current, regs, 0) {
addr = unwind_get_return_address(&state);
if (!addr || perf_callchain_store(entry, addr))
return;
}
}
void perf_callchain_user(struct perf_callchain_entry_ctx *entry,
struct pt_regs *regs)
{
struct stack_frame_user __user *sf;
unsigned long ip, sp;
bool first = true;
if (is_compat_task())
return;
perf_callchain_store(entry, instruction_pointer(regs));
sf = (void __user *)user_stack_pointer(regs);
pagefault_disable();
while (entry->nr < entry->max_stack) {
if (__get_user(sp, &sf->back_chain))
break;
if (__get_user(ip, &sf->gprs[8]))
break;
if (ip & 0x1) {
/*
* If the instruction address is invalid, and this
* is the first stack frame, assume r14 has not
* been written to the stack yet. Otherwise exit.
*/
if (first && !(regs->gprs[14] & 0x1))
ip = regs->gprs[14];
else
break;
}
perf_callchain_store(entry, ip);
/* Sanity check: ABI requires SP to be aligned 8 bytes. */
if (!sp || sp & 0x7)
break;
sf = (void __user *)sp;
first = false;
}
pagefault_enable();
}
/* Perf definitions for PMU event attributes in sysfs */
ssize_t cpumf_events_sysfs_show(struct device *dev,
struct device_attribute *attr, char *page)
{
struct perf_pmu_events_attr *pmu_attr;
pmu_attr = container_of(attr, struct perf_pmu_events_attr, attr);
return sprintf(page, "event=0x%04llx\n", pmu_attr->id);
}