Juergen Gross
2016-Apr-06 14:17 UTC
[PATCH v5 0/6] Support calling functions on dedicated physical cpu
Some hardware (e.g. Dell Studio laptops) require special functions to be called on physical cpu 0 in order to avoid occasional hangs. When running as dom0 under Xen this could be achieved only via special boot parameters (vcpu pinning) limiting the hypervisor in it's scheduling decisions. This patch series is adding a generic function to be able to temporarily pin a (virtual) cpu to a dedicated physical cpu for executing above mentioned functions on that specific cpu. The drivers (dcdbas and i8k) requiring this functionality are modified accordingly. Changes in V5: - patch 3: rename and reshuffle parameters of smp_call_on_cpu() as requested by Peter Zijlstra - patch 3: test target cpu to be online as requested by Peter Zijlstra - patch 4: less wordy messages as requested by David Vrabel Changes in V4: - move patches 5 and 6 further up in the series - patch 2 (was 5): WARN_ONCE in case platform doesn't support pinning as requested by Peter Zijlstra - patch 3 (was 2): change return value in case of illegal cpu as requested by Peter Zijlstra - patch 3 (was 2): make pinning of vcpu an option as suggested by Peter Zijlstra - patches 5 and 6 (were 3 and 4): add call to get_online_cpus() Changes in V3: - use get_cpu()/put_cpu() as suggested by David Vrabel Changes in V2: - instead of manipulating the allowed set of cpus use cpu specific workqueue as requested by Peter Zijlstra - add include/linux/hypervisor.h to hide architecture specific stuff from generic kernel code Juergen Gross (6): xen: sync xen header virt, sched: add generic vcpu pinning support smp: add function to execute a function synchronously on a cpu xen: add xen_pin_vcpu() to support calling functions on a dedicated pcpu dcdbas: make use of smp_call_on_cpu() hwmon: use smp_call_on_cpu() for dell-smm i8k MAINTAINERS | 1 + arch/x86/include/asm/hypervisor.h | 4 ++ arch/x86/kernel/cpu/hypervisor.c | 11 +++++ arch/x86/xen/enlighten.c | 40 +++++++++++++++ drivers/firmware/dcdbas.c | 51 +++++++++---------- drivers/hwmon/dell-smm-hwmon.c | 35 +++++++------ include/linux/hypervisor.h | 17 +++++++ include/linux/smp.h | 3 ++ include/xen/interface/sched.h | 100 +++++++++++++++++++++++++++++++------- kernel/smp.c | 51 +++++++++++++++++++ kernel/up.c | 18 +++++++ 11 files changed, 273 insertions(+), 58 deletions(-) create mode 100644 include/linux/hypervisor.h -- 2.6.6
Import the actual version of include/xen/interface/sched.h from Xen. Signed-off-by: Juergen Gross <jgross at suse.com> Acked-by: David Vrabel <david.vrabel at citrix.com> --- include/xen/interface/sched.h | 100 ++++++++++++++++++++++++++++++++++-------- 1 file changed, 82 insertions(+), 18 deletions(-) diff --git a/include/xen/interface/sched.h b/include/xen/interface/sched.h index f184909..a4c4d73 100644 --- a/include/xen/interface/sched.h +++ b/include/xen/interface/sched.h @@ -3,6 +3,24 @@ * * Scheduler state interactions * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal in the Software without restriction, including without limitation the + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or + * sell copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * * Copyright (c) 2005, Keir Fraser <keir at xensource.com> */ @@ -12,18 +30,30 @@ #include <xen/interface/event_channel.h> /* + * Guest Scheduler Operations + * + * The SCHEDOP interface provides mechanisms for a guest to interact + * with the scheduler, including yield, blocking and shutting itself + * down. + */ + +/* * The prototype for this hypercall is: - * long sched_op_new(int cmd, void *arg) + * long HYPERVISOR_sched_op(enum sched_op cmd, void *arg, ...) + * * @cmd == SCHEDOP_??? (scheduler operation). * @arg == Operation-specific extra argument(s), as described below. + * ... == Additional Operation-specific extra arguments, described below. * - * **NOTE**: - * Versions of Xen prior to 3.0.2 provide only the following legacy version + * Versions of Xen prior to 3.0.2 provided only the following legacy version * of this hypercall, supporting only the commands yield, block and shutdown: * long sched_op(int cmd, unsigned long arg) * @cmd == SCHEDOP_??? (scheduler operation). * @arg == 0 (SCHEDOP_yield and SCHEDOP_block) * == SHUTDOWN_* code (SCHEDOP_shutdown) + * + * This legacy version is available to new guests as: + * long HYPERVISOR_sched_op_compat(enum sched_op cmd, unsigned long arg) */ /* @@ -44,12 +74,17 @@ /* * Halt execution of this domain (all VCPUs) and notify the system controller. * @arg == pointer to sched_shutdown structure. + * + * If the sched_shutdown_t reason is SHUTDOWN_suspend then + * x86 PV guests must also set RDX (EDX for 32-bit guests) to the MFN + * of the guest's start info page. RDX/EDX is the third hypercall + * argument. + * + * In addition, which reason is SHUTDOWN_suspend this hypercall + * returns 1 if suspend was cancelled or the domain was merely + * checkpointed, and 0 if it is resuming in a new domain. */ #define SCHEDOP_shutdown 2 -struct sched_shutdown { - unsigned int reason; /* SHUTDOWN_* */ -}; -DEFINE_GUEST_HANDLE_STRUCT(sched_shutdown); /* * Poll a set of event-channel ports. Return when one or more are pending. An @@ -57,12 +92,6 @@ DEFINE_GUEST_HANDLE_STRUCT(sched_shutdown); * @arg == pointer to sched_poll structure. */ #define SCHEDOP_poll 3 -struct sched_poll { - GUEST_HANDLE(evtchn_port_t) ports; - unsigned int nr_ports; - uint64_t timeout; -}; -DEFINE_GUEST_HANDLE_STRUCT(sched_poll); /* * Declare a shutdown for another domain. The main use of this function is @@ -71,15 +100,11 @@ DEFINE_GUEST_HANDLE_STRUCT(sched_poll); * @arg == pointer to sched_remote_shutdown structure. */ #define SCHEDOP_remote_shutdown 4 -struct sched_remote_shutdown { - domid_t domain_id; /* Remote domain ID */ - unsigned int reason; /* SHUTDOWN_xxx reason */ -}; /* * Latch a shutdown code, so that when the domain later shuts down it * reports this code to the control tools. - * @arg == as for SCHEDOP_shutdown. + * @arg == sched_shutdown, as for SCHEDOP_shutdown. */ #define SCHEDOP_shutdown_code 5 @@ -92,10 +117,47 @@ struct sched_remote_shutdown { * With id != 0 and timeout != 0, poke watchdog timer and set new timeout. */ #define SCHEDOP_watchdog 6 + +/* + * Override the current vcpu affinity by pinning it to one physical cpu or + * undo this override restoring the previous affinity. + * @arg == pointer to sched_pin_override structure. + * + * A negative pcpu value will undo a previous pin override and restore the + * previous cpu affinity. + * This call is allowed for the hardware domain only and requires the cpu + * to be part of the domain's cpupool. + */ +#define SCHEDOP_pin_override 7 + +struct sched_shutdown { + unsigned int reason; /* SHUTDOWN_* => shutdown reason */ +}; +DEFINE_GUEST_HANDLE_STRUCT(sched_shutdown); + +struct sched_poll { + GUEST_HANDLE(evtchn_port_t) ports; + unsigned int nr_ports; + uint64_t timeout; +}; +DEFINE_GUEST_HANDLE_STRUCT(sched_poll); + +struct sched_remote_shutdown { + domid_t domain_id; /* Remote domain ID */ + unsigned int reason; /* SHUTDOWN_* => shutdown reason */ +}; +DEFINE_GUEST_HANDLE_STRUCT(sched_remote_shutdown); + struct sched_watchdog { uint32_t id; /* watchdog ID */ uint32_t timeout; /* timeout */ }; +DEFINE_GUEST_HANDLE_STRUCT(sched_watchdog); + +struct sched_pin_override { + int32_t pcpu; +}; +DEFINE_GUEST_HANDLE_STRUCT(sched_pin_override); /* * Reason codes for SCHEDOP_shutdown. These may be interpreted by control @@ -107,6 +169,7 @@ struct sched_watchdog { #define SHUTDOWN_suspend 2 /* Clean up, save suspend info, kill. */ #define SHUTDOWN_crash 3 /* Tell controller we've crashed. */ #define SHUTDOWN_watchdog 4 /* Restart because watchdog time expired. */ + /* * Domain asked to perform 'soft reset' for it. The expected behavior is to * reset internal Xen state for the domain returning it to the point where it @@ -115,5 +178,6 @@ struct sched_watchdog { * interfaces again. */ #define SHUTDOWN_soft_reset 5 +#define SHUTDOWN_MAX 5 /* Maximum valid shutdown reason. */ #endif /* __XEN_PUBLIC_SCHED_H__ */ -- 2.6.6
Juergen Gross
2016-Apr-06 14:17 UTC
[PATCH v5 2/6] virt, sched: add generic vcpu pinning support
Add generic virtualization support for pinning the current vcpu to a specified physical cpu. As this operation isn't performance critical (a very limited set of operations like BIOS calls and SMIs is expected to need this) just add a hypervisor specific indirection. Signed-off-by: Juergen Gross <jgross at suse.com> --- V4: move this patch some places up in the series WARN_ONCE in case platform doesn't support pinning as requested by Peter Zijlstra V3: use getc_cpu()/put_cpu() as suggested by David Vrabel V2: adapt to using workqueues add include/linux/hypervisor.h to hide architecture specific stuff from generic kernel code In case paravirt maintainers don't want to be responsible for include/linux/hypervisor.h I could take it. --- MAINTAINERS | 1 + arch/x86/include/asm/hypervisor.h | 4 ++++ arch/x86/kernel/cpu/hypervisor.c | 11 +++++++++++ include/linux/hypervisor.h | 17 +++++++++++++++++ kernel/smp.c | 1 + kernel/up.c | 1 + 6 files changed, 35 insertions(+) create mode 100644 include/linux/hypervisor.h diff --git a/MAINTAINERS b/MAINTAINERS index 40eb1db..d3bde4f 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -8330,6 +8330,7 @@ S: Supported F: Documentation/virtual/paravirt_ops.txt F: arch/*/kernel/paravirt* F: arch/*/include/asm/paravirt.h +F: include/linux/hypervisor.h PARIDE DRIVERS FOR PARALLEL PORT IDE DEVICES M: Tim Waugh <tim at cyberelk.net> diff --git a/arch/x86/include/asm/hypervisor.h b/arch/x86/include/asm/hypervisor.h index 055ea99..67942b6 100644 --- a/arch/x86/include/asm/hypervisor.h +++ b/arch/x86/include/asm/hypervisor.h @@ -43,6 +43,9 @@ struct hypervisor_x86 { /* X2APIC detection (run once per boot) */ bool (*x2apic_available)(void); + + /* pin current vcpu to specified physical cpu (run rarely) */ + void (*pin_vcpu)(int); }; extern const struct hypervisor_x86 *x86_hyper; @@ -56,6 +59,7 @@ extern const struct hypervisor_x86 x86_hyper_kvm; extern void init_hypervisor(struct cpuinfo_x86 *c); extern void init_hypervisor_platform(void); extern bool hypervisor_x2apic_available(void); +extern void hypervisor_pin_vcpu(int cpu); #else static inline void init_hypervisor(struct cpuinfo_x86 *c) { } static inline void init_hypervisor_platform(void) { } diff --git a/arch/x86/kernel/cpu/hypervisor.c b/arch/x86/kernel/cpu/hypervisor.c index 73d391a..ff108f8 100644 --- a/arch/x86/kernel/cpu/hypervisor.c +++ b/arch/x86/kernel/cpu/hypervisor.c @@ -85,3 +85,14 @@ bool __init hypervisor_x2apic_available(void) x86_hyper->x2apic_available && x86_hyper->x2apic_available(); } + +void hypervisor_pin_vcpu(int cpu) +{ + if (!x86_hyper) + return; + + if (x86_hyper->pin_vcpu) + x86_hyper->pin_vcpu(cpu); + else + WARN_ONCE(1, "vcpu pinning requested but not supported!\n"); +} diff --git a/include/linux/hypervisor.h b/include/linux/hypervisor.h new file mode 100644 index 0000000..3fa5ef2 --- /dev/null +++ b/include/linux/hypervisor.h @@ -0,0 +1,17 @@ +#ifndef __LINUX_HYPEVISOR_H +#define __LINUX_HYPEVISOR_H + +/* + * Generic Hypervisor support + * Juergen Gross <jgross at suse.com> + */ + +#ifdef CONFIG_HYPERVISOR_GUEST +#include <asm/hypervisor.h> +#else +static inline void hypervisor_pin_vcpu(int cpu) +{ +} +#endif + +#endif /* __LINUX_HYPEVISOR_H */ diff --git a/kernel/smp.c b/kernel/smp.c index 7416544..9388064 100644 --- a/kernel/smp.c +++ b/kernel/smp.c @@ -14,6 +14,7 @@ #include <linux/smp.h> #include <linux/cpu.h> #include <linux/sched.h> +#include <linux/hypervisor.h> #include "smpboot.h" diff --git a/kernel/up.c b/kernel/up.c index 1760bf3..3ccee2b 100644 --- a/kernel/up.c +++ b/kernel/up.c @@ -6,6 +6,7 @@ #include <linux/kernel.h> #include <linux/export.h> #include <linux/smp.h> +#include <linux/hypervisor.h> int smp_call_function_single(int cpu, void (*func) (void *info), void *info, int wait) -- 2.6.6
Juergen Gross
2016-Apr-06 14:17 UTC
[PATCH v5 3/6] smp: add function to execute a function synchronously on a cpu
On some hardware models (e.g. Dell Studio 1555 laptop) some hardware related functions (e.g. SMIs) are to be executed on physical cpu 0 only. Instead of open coding such a functionality multiple times in the kernel add a service function for this purpose. This will enable the possibility to take special measures in virtualized environments like Xen, too. Signed-off-by: Juergen Gross <jgross at suse.com> --- V5: rename and reshuffle parameters of smp_call_on_cpu() as requested by Peter Zijlstra test target cpu to be online as requested by Peter Zijlstra V4: change return value in case of illegal cpu as requested by Peter Zijlstra make pinning of vcpu an option as suggested by Peter Zijlstra V2: instead of manipulating the allowed set of cpus use cpu specific workqueue as requested by Peter Zijlstra --- include/linux/smp.h | 3 +++ kernel/smp.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++++ kernel/up.c | 17 +++++++++++++++++ 3 files changed, 70 insertions(+) diff --git a/include/linux/smp.h b/include/linux/smp.h index c441407..4f1c475 100644 --- a/include/linux/smp.h +++ b/include/linux/smp.h @@ -196,4 +196,7 @@ extern void arch_enable_nonboot_cpus_end(void); void smp_setup_processor_id(void); +int smp_call_on_cpu(unsigned int cpu, int (*func)(void *), void *par, + bool phys); + #endif /* __LINUX_SMP_H */ diff --git a/kernel/smp.c b/kernel/smp.c index 9388064..f671948 100644 --- a/kernel/smp.c +++ b/kernel/smp.c @@ -740,3 +740,53 @@ void wake_up_all_idle_cpus(void) preempt_enable(); } EXPORT_SYMBOL_GPL(wake_up_all_idle_cpus); + +/** + * smp_call_on_cpu - Call a function on a specific cpu + * + * Used to call a function on a specific cpu and wait for it to return. + * Optionally make sure the call is done on a specified physical cpu via vcpu + * pinning in order to support virtualized environments. + */ +struct smp_call_on_cpu_struct { + struct work_struct work; + struct completion done; + int (*func)(void *); + void *data; + int ret; + int cpu; +}; + +static void smp_call_on_cpu_callback(struct work_struct *work) +{ + struct smp_call_on_cpu_struct *sscs; + + sscs = container_of(work, struct smp_call_on_cpu_struct, work); + if (sscs->cpu >= 0) + hypervisor_pin_vcpu(sscs->cpu); + sscs->ret = sscs->func(sscs->data); + if (sscs->cpu >= 0) + hypervisor_pin_vcpu(-1); + + complete(&sscs->done); +} + +int smp_call_on_cpu(unsigned int cpu, int (*func)(void *), void *par, bool phys) +{ + struct smp_call_on_cpu_struct sscs = { + .work = __WORK_INITIALIZER(sscs.work, smp_call_on_cpu_callback), + .done = COMPLETION_INITIALIZER_ONSTACK(sscs.done), + .func = func, + .data = par, + .cpu = phys ? cpu : -1, + }; + + if (cpu >= nr_cpu_ids || !cpu_online(cpu)) + return -ENXIO; + + queue_work_on(cpu, system_wq, &sscs.work); + wait_for_completion(&sscs.done); + + return sscs.ret; +} +EXPORT_SYMBOL_GPL(smp_call_on_cpu); diff --git a/kernel/up.c b/kernel/up.c index 3ccee2b..ee81ac9 100644 --- a/kernel/up.c +++ b/kernel/up.c @@ -83,3 +83,20 @@ void on_each_cpu_cond(bool (*cond_func)(int cpu, void *info), preempt_enable(); } EXPORT_SYMBOL(on_each_cpu_cond); + +int smp_call_on_cpu(unsigned int cpu, int (*func)(void *), void *par, bool phys) +{ + int ret; + + if (cpu != 0) + return -ENXIO; + + if (phys) + hypervisor_pin_vcpu(0); + ret = func(par); + if (phys) + hypervisor_pin_vcpu(-1); + + return ret; +} +EXPORT_SYMBOL_GPL(smp_call_on_cpu); -- 2.6.6
Juergen Gross
2016-Apr-06 14:17 UTC
[PATCH v5 4/6] xen: add xen_pin_vcpu() to support calling functions on a dedicated pcpu
Some hardware models (e.g. Dell Studio 1555 laptops) require calls to the firmware to be issued on cpu 0 only. As Dom0 might have to use these calls, add xen_pin_vcpu() to achieve this functionality. In case either the domain doesn't have the privilege to make the related hypercall or the hypervisor isn't supporting it, issue a warning once and disable further pinning attempts. Signed-off-by: Juergen Gross <jgross at suse.com> Acked-by: David Vrabel <david.vrabel at citrix.com> --- V5: less wordy messages as requested by David Vrabel --- arch/x86/xen/enlighten.c | 40 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 40 insertions(+) diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c index 880862c..bfa2f21 100644 --- a/arch/x86/xen/enlighten.c +++ b/arch/x86/xen/enlighten.c @@ -1885,6 +1885,45 @@ static void xen_set_cpu_features(struct cpuinfo_x86 *c) } } +static void xen_pin_vcpu(int cpu) +{ + static bool disable_pinning; + struct sched_pin_override pin_override; + int ret; + + if (disable_pinning) + return; + + pin_override.pcpu = cpu; + ret = HYPERVISOR_sched_op(SCHEDOP_pin_override, &pin_override); + + /* Ignore errors when removing override. */ + if (cpu < 0) + return; + + switch (ret) { + case -ENOSYS: + pr_warn("Unable to pin on physical cpu %d. In case of problems consider vcpu pinning.\n", + cpu); + disable_pinning = true; + break; + case -EPERM: + WARN(1, "Trying to pin vcpu without having privilege to do so\n"); + disable_pinning = true; + break; + case -EINVAL: + case -EBUSY: + pr_warn("Physical cpu %d not available for pinning. Check Xen cpu configuration.\n", + cpu); + break; + case 0: + break; + default: + WARN(1, "rc %d while trying to pin vcpu\n", ret); + disable_pinning = true; + } +} + const struct hypervisor_x86 x86_hyper_xen = { .name = "Xen", .detect = xen_platform, @@ -1893,6 +1932,7 @@ const struct hypervisor_x86 x86_hyper_xen = { #endif .x2apic_available = xen_x2apic_para_available, .set_cpu_features = xen_set_cpu_features, + .pin_vcpu = xen_pin_vcpu, }; EXPORT_SYMBOL(x86_hyper_xen); -- 2.6.6
Use smp_call_on_cpu() to raise SMI on cpu 0. Make call secure by adding get_online_cpus() to avoid e.g. suspend resume cycles in between. Signed-off-by: Juergen Gross <jgross at suse.com> --- V4: add call to get_online_cpus() --- drivers/firmware/dcdbas.c | 51 ++++++++++++++++++++++++----------------------- 1 file changed, 26 insertions(+), 25 deletions(-) diff --git a/drivers/firmware/dcdbas.c b/drivers/firmware/dcdbas.c index 829eec8..2fe1a13 100644 --- a/drivers/firmware/dcdbas.c +++ b/drivers/firmware/dcdbas.c @@ -23,6 +23,7 @@ #include <linux/platform_device.h> #include <linux/dma-mapping.h> #include <linux/errno.h> +#include <linux/cpu.h> #include <linux/gfp.h> #include <linux/init.h> #include <linux/kernel.h> @@ -238,33 +239,14 @@ static ssize_t host_control_on_shutdown_store(struct device *dev, return count; } -/** - * dcdbas_smi_request: generate SMI request - * - * Called with smi_data_lock. - */ -int dcdbas_smi_request(struct smi_cmd *smi_cmd) +static int raise_smi(void *par) { - cpumask_var_t old_mask; - int ret = 0; + struct smi_cmd *smi_cmd = par; - if (smi_cmd->magic != SMI_CMD_MAGIC) { - dev_info(&dcdbas_pdev->dev, "%s: invalid magic value\n", - __func__); - return -EBADR; - } - - /* SMI requires CPU 0 */ - if (!alloc_cpumask_var(&old_mask, GFP_KERNEL)) - return -ENOMEM; - - cpumask_copy(old_mask, ¤t->cpus_allowed); - set_cpus_allowed_ptr(current, cpumask_of(0)); if (smp_processor_id() != 0) { dev_dbg(&dcdbas_pdev->dev, "%s: failed to get CPU 0\n", __func__); - ret = -EBUSY; - goto out; + return -EBUSY; } /* generate SMI */ @@ -280,9 +262,28 @@ int dcdbas_smi_request(struct smi_cmd *smi_cmd) : "memory" ); -out: - set_cpus_allowed_ptr(current, old_mask); - free_cpumask_var(old_mask); + return 0; +} +/** + * dcdbas_smi_request: generate SMI request + * + * Called with smi_data_lock. + */ +int dcdbas_smi_request(struct smi_cmd *smi_cmd) +{ + int ret; + + if (smi_cmd->magic != SMI_CMD_MAGIC) { + dev_info(&dcdbas_pdev->dev, "%s: invalid magic value\n", + __func__); + return -EBADR; + } + + /* SMI requires CPU 0 */ + get_online_cpus(); + ret = smp_call_on_cpu(0, raise_smi, smi_cmd, true); + put_online_cpus(); + return ret; } -- 2.6.6
Juergen Gross
2016-Apr-06 14:17 UTC
[PATCH v5 6/6] hwmon: use smp_call_on_cpu() for dell-smm i8k
Use the smp_call_on_cpu() function to call system management mode on cpu 0. Make call secure by adding get_online_cpus() to avoid e.g. suspend resume cycles in between. Signed-off-by: Juergen Gross <jgross at suse.com> --- V4: add call to get_online_cpus() --- drivers/hwmon/dell-smm-hwmon.c | 35 ++++++++++++++++++++--------------- 1 file changed, 20 insertions(+), 15 deletions(-) diff --git a/drivers/hwmon/dell-smm-hwmon.c b/drivers/hwmon/dell-smm-hwmon.c index c43318d..c91bf35 100644 --- a/drivers/hwmon/dell-smm-hwmon.c +++ b/drivers/hwmon/dell-smm-hwmon.c @@ -21,6 +21,7 @@ #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt +#include <linux/cpu.h> #include <linux/delay.h> #include <linux/module.h> #include <linux/types.h> @@ -35,6 +36,7 @@ #include <linux/uaccess.h> #include <linux/io.h> #include <linux/sched.h> +#include <linux/smp.h> #include <linux/i8k.h> @@ -130,23 +132,15 @@ static inline const char *i8k_get_dmi_data(int field) /* * Call the System Management Mode BIOS. Code provided by Jonathan Buzzard. */ -static int i8k_smm(struct smm_regs *regs) +static int i8k_smm_func(void *par) { int rc; + struct smm_regs *regs = par; int eax = regs->eax; - cpumask_var_t old_mask; /* SMM requires CPU 0 */ - if (!alloc_cpumask_var(&old_mask, GFP_KERNEL)) - return -ENOMEM; - cpumask_copy(old_mask, ¤t->cpus_allowed); - rc = set_cpus_allowed_ptr(current, cpumask_of(0)); - if (rc) - goto out; - if (smp_processor_id() != 0) { - rc = -EBUSY; - goto out; - } + if (smp_processor_id() != 0) + return -EBUSY; #if defined(CONFIG_X86_64) asm volatile("pushq %%rax\n\t" @@ -204,13 +198,24 @@ static int i8k_smm(struct smm_regs *regs) if (rc != 0 || (regs->eax & 0xffff) == 0xffff || regs->eax == eax) rc = -EINVAL; -out: - set_cpus_allowed_ptr(current, old_mask); - free_cpumask_var(old_mask); return rc; } /* + * Call the System Management Mode BIOS. + */ +static int i8k_smm(struct smm_regs *regs) +{ + int ret; + + get_online_cpus(); + ret = smp_call_on_cpu(0, i8k_smm_func, regs, true); + put_online_cpus(); + + return ret; +} + +/* * Read the fan status. */ static int i8k_get_fan_status(int fan) -- 2.6.6
Peter Zijlstra
2016-Apr-06 18:04 UTC
[PATCH v5 3/6] smp: add function to execute a function synchronously on a cpu
On Wed, Apr 06, 2016 at 04:17:43PM +0200, Juergen Gross wrote:> On some hardware models (e.g. Dell Studio 1555 laptop) some hardware > related functions (e.g. SMIs) are to be executed on physical cpu 0 > only. Instead of open coding such a functionality multiple times in > the kernel add a service function for this purpose. This will enable > the possibility to take special measures in virtualized environments > like Xen, too. > > Signed-off-by: Juergen Gross <jgross at suse.com>Thanks! Acked-by: Peter Zijlstra (Intel) <peterz at infradead.org>
Juergen Gross
2016-Apr-13 08:49 UTC
[PATCH v5 0/6] Support calling functions on dedicated physical cpu
On 06/04/16 16:17, Juergen Gross wrote:> Some hardware (e.g. Dell Studio laptops) require special functions to > be called on physical cpu 0 in order to avoid occasional hangs. When > running as dom0 under Xen this could be achieved only via special boot > parameters (vcpu pinning) limiting the hypervisor in it's scheduling > decisions. > > This patch series is adding a generic function to be able to temporarily > pin a (virtual) cpu to a dedicated physical cpu for executing above > mentioned functions on that specific cpu. The drivers (dcdbas and i8k) > requiring this functionality are modified accordingly. > > Changes in V5: > - patch 3: rename and reshuffle parameters of smp_call_on_cpu() as requested > by Peter Zijlstra > - patch 3: test target cpu to be online as requested by Peter Zijlstra > - patch 4: less wordy messages as requested by David Vrabel > > Changes in V4: > - move patches 5 and 6 further up in the series > - patch 2 (was 5): WARN_ONCE in case platform doesn't support pinning > as requested by Peter Zijlstra > - patch 3 (was 2): change return value in case of illegal cpu as > requested by Peter Zijlstra > - patch 3 (was 2): make pinning of vcpu an option as suggested by > Peter Zijlstra > - patches 5 and 6 (were 3 and 4): add call to get_online_cpus() > > Changes in V3: > - use get_cpu()/put_cpu() as suggested by David Vrabel > > Changes in V2: > - instead of manipulating the allowed set of cpus use cpu specific > workqueue as requested by Peter Zijlstra > - add include/linux/hypervisor.h to hide architecture specific stuff > from generic kernel code > > Juergen Gross (6): > xen: sync xen header > virt, sched: add generic vcpu pinning support > smp: add function to execute a function synchronously on a cpu > xen: add xen_pin_vcpu() to support calling functions on a dedicated > pcpu > dcdbas: make use of smp_call_on_cpu() > hwmon: use smp_call_on_cpu() for dell-smm i8k > > MAINTAINERS | 1 + > arch/x86/include/asm/hypervisor.h | 4 ++ > arch/x86/kernel/cpu/hypervisor.c | 11 +++++ > arch/x86/xen/enlighten.c | 40 +++++++++++++++ > drivers/firmware/dcdbas.c | 51 +++++++++---------- > drivers/hwmon/dell-smm-hwmon.c | 35 +++++++------ > include/linux/hypervisor.h | 17 +++++++ > include/linux/smp.h | 3 ++ > include/xen/interface/sched.h | 100 +++++++++++++++++++++++++++++++------- > kernel/smp.c | 51 +++++++++++++++++++ > kernel/up.c | 18 +++++++ > 11 files changed, 273 insertions(+), 58 deletions(-) > create mode 100644 include/linux/hypervisor.h >So patches 1, 3 and 4 are acked. Any comments regarding patches 2, 5 and 6? Juergen
Juergen Gross
2016-May-17 11:21 UTC
[PATCH v5 0/6] Support calling functions on dedicated physical cpu
On 13/04/16 10:49, Juergen Gross wrote:> On 06/04/16 16:17, Juergen Gross wrote: >> Some hardware (e.g. Dell Studio laptops) require special functions to >> be called on physical cpu 0 in order to avoid occasional hangs. When >> running as dom0 under Xen this could be achieved only via special boot >> parameters (vcpu pinning) limiting the hypervisor in it's scheduling >> decisions. >> >> This patch series is adding a generic function to be able to temporarily >> pin a (virtual) cpu to a dedicated physical cpu for executing above >> mentioned functions on that specific cpu. The drivers (dcdbas and i8k) >> requiring this functionality are modified accordingly. >> >> Changes in V5: >> - patch 3: rename and reshuffle parameters of smp_call_on_cpu() as requested >> by Peter Zijlstra >> - patch 3: test target cpu to be online as requested by Peter Zijlstra >> - patch 4: less wordy messages as requested by David Vrabel >> >> Changes in V4: >> - move patches 5 and 6 further up in the series >> - patch 2 (was 5): WARN_ONCE in case platform doesn't support pinning >> as requested by Peter Zijlstra >> - patch 3 (was 2): change return value in case of illegal cpu as >> requested by Peter Zijlstra >> - patch 3 (was 2): make pinning of vcpu an option as suggested by >> Peter Zijlstra >> - patches 5 and 6 (were 3 and 4): add call to get_online_cpus() >> >> Changes in V3: >> - use get_cpu()/put_cpu() as suggested by David Vrabel >> >> Changes in V2: >> - instead of manipulating the allowed set of cpus use cpu specific >> workqueue as requested by Peter Zijlstra >> - add include/linux/hypervisor.h to hide architecture specific stuff >> from generic kernel code >> >> Juergen Gross (6): >> xen: sync xen header >> virt, sched: add generic vcpu pinning support >> smp: add function to execute a function synchronously on a cpu >> xen: add xen_pin_vcpu() to support calling functions on a dedicated >> pcpu >> dcdbas: make use of smp_call_on_cpu() >> hwmon: use smp_call_on_cpu() for dell-smm i8k >> >> MAINTAINERS | 1 + >> arch/x86/include/asm/hypervisor.h | 4 ++ >> arch/x86/kernel/cpu/hypervisor.c | 11 +++++ >> arch/x86/xen/enlighten.c | 40 +++++++++++++++ >> drivers/firmware/dcdbas.c | 51 +++++++++---------- >> drivers/hwmon/dell-smm-hwmon.c | 35 +++++++------ >> include/linux/hypervisor.h | 17 +++++++ >> include/linux/smp.h | 3 ++ >> include/xen/interface/sched.h | 100 +++++++++++++++++++++++++++++++------- >> kernel/smp.c | 51 +++++++++++++++++++ >> kernel/up.c | 18 +++++++ >> 11 files changed, 273 insertions(+), 58 deletions(-) >> create mode 100644 include/linux/hypervisor.h >> > > So patches 1, 3 and 4 are acked. Any comments regarding patches 2, 5 > and 6?Patch 6 gained another Ack. What about 2 and 5? Juergen
Pali Rohár
2016-Jun-13 19:05 UTC
[PATCH v5 6/6] hwmon: use smp_call_on_cpu() for dell-smm i8k
On Wednesday 06 April 2016 16:17:46 Juergen Gross wrote:> Use the smp_call_on_cpu() function to call system management > mode on cpu 0. > Make call secure by adding get_online_cpus() to avoid e.g. suspend > resume cycles in between. > > Signed-off-by: Juergen Gross <jgross at suse.com> > --- > V4: add call to get_online_cpus() > --- > drivers/hwmon/dell-smm-hwmon.c | 35As mentioned in discussion under V4, you can add my Acked-by. -- Pali Roh?r pali.rohar at gmail.com -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 198 bytes Desc: This is a digitally signed message part. URL: <http://lists.linuxfoundation.org/pipermail/virtualization/attachments/20160613/fcc98df1/attachment.sig>
Juergen Gross
2016-Jun-15 11:18 UTC
[PATCH v5 2/6] virt, sched: add generic vcpu pinning support
On 06/04/16 16:17, Juergen Gross wrote:> Add generic virtualization support for pinning the current vcpu to a > specified physical cpu. As this operation isn't performance critical > (a very limited set of operations like BIOS calls and SMIs is expected > to need this) just add a hypervisor specific indirection. > > Signed-off-by: Juergen Gross <jgross at suse.com>Could please some maintainer comment on this patch? Juergen> --- > V4: move this patch some places up in the series > WARN_ONCE in case platform doesn't support pinning as requested by > Peter Zijlstra > > V3: use getc_cpu()/put_cpu() as suggested by David Vrabel > > V2: adapt to using workqueues > add include/linux/hypervisor.h to hide architecture specific stuff > from generic kernel code > > In case paravirt maintainers don't want to be responsible for > include/linux/hypervisor.h I could take it. > --- > MAINTAINERS | 1 + > arch/x86/include/asm/hypervisor.h | 4 ++++ > arch/x86/kernel/cpu/hypervisor.c | 11 +++++++++++ > include/linux/hypervisor.h | 17 +++++++++++++++++ > kernel/smp.c | 1 + > kernel/up.c | 1 + > 6 files changed, 35 insertions(+) > create mode 100644 include/linux/hypervisor.h > > diff --git a/MAINTAINERS b/MAINTAINERS > index 40eb1db..d3bde4f 100644 > --- a/MAINTAINERS > +++ b/MAINTAINERS > @@ -8330,6 +8330,7 @@ S: Supported > F: Documentation/virtual/paravirt_ops.txt > F: arch/*/kernel/paravirt* > F: arch/*/include/asm/paravirt.h > +F: include/linux/hypervisor.h > > PARIDE DRIVERS FOR PARALLEL PORT IDE DEVICES > M: Tim Waugh <tim at cyberelk.net> > diff --git a/arch/x86/include/asm/hypervisor.h b/arch/x86/include/asm/hypervisor.h > index 055ea99..67942b6 100644 > --- a/arch/x86/include/asm/hypervisor.h > +++ b/arch/x86/include/asm/hypervisor.h > @@ -43,6 +43,9 @@ struct hypervisor_x86 { > > /* X2APIC detection (run once per boot) */ > bool (*x2apic_available)(void); > + > + /* pin current vcpu to specified physical cpu (run rarely) */ > + void (*pin_vcpu)(int); > }; > > extern const struct hypervisor_x86 *x86_hyper; > @@ -56,6 +59,7 @@ extern const struct hypervisor_x86 x86_hyper_kvm; > extern void init_hypervisor(struct cpuinfo_x86 *c); > extern void init_hypervisor_platform(void); > extern bool hypervisor_x2apic_available(void); > +extern void hypervisor_pin_vcpu(int cpu); > #else > static inline void init_hypervisor(struct cpuinfo_x86 *c) { } > static inline void init_hypervisor_platform(void) { } > diff --git a/arch/x86/kernel/cpu/hypervisor.c b/arch/x86/kernel/cpu/hypervisor.c > index 73d391a..ff108f8 100644 > --- a/arch/x86/kernel/cpu/hypervisor.c > +++ b/arch/x86/kernel/cpu/hypervisor.c > @@ -85,3 +85,14 @@ bool __init hypervisor_x2apic_available(void) > x86_hyper->x2apic_available && > x86_hyper->x2apic_available(); > } > + > +void hypervisor_pin_vcpu(int cpu) > +{ > + if (!x86_hyper) > + return; > + > + if (x86_hyper->pin_vcpu) > + x86_hyper->pin_vcpu(cpu); > + else > + WARN_ONCE(1, "vcpu pinning requested but not supported!\n"); > +} > diff --git a/include/linux/hypervisor.h b/include/linux/hypervisor.h > new file mode 100644 > index 0000000..3fa5ef2 > --- /dev/null > +++ b/include/linux/hypervisor.h > @@ -0,0 +1,17 @@ > +#ifndef __LINUX_HYPEVISOR_H > +#define __LINUX_HYPEVISOR_H > + > +/* > + * Generic Hypervisor support > + * Juergen Gross <jgross at suse.com> > + */ > + > +#ifdef CONFIG_HYPERVISOR_GUEST > +#include <asm/hypervisor.h> > +#else > +static inline void hypervisor_pin_vcpu(int cpu) > +{ > +} > +#endif > + > +#endif /* __LINUX_HYPEVISOR_H */ > diff --git a/kernel/smp.c b/kernel/smp.c > index 7416544..9388064 100644 > --- a/kernel/smp.c > +++ b/kernel/smp.c > @@ -14,6 +14,7 @@ > #include <linux/smp.h> > #include <linux/cpu.h> > #include <linux/sched.h> > +#include <linux/hypervisor.h> > > #include "smpboot.h" > > diff --git a/kernel/up.c b/kernel/up.c > index 1760bf3..3ccee2b 100644 > --- a/kernel/up.c > +++ b/kernel/up.c > @@ -6,6 +6,7 @@ > #include <linux/kernel.h> > #include <linux/export.h> > #include <linux/smp.h> > +#include <linux/hypervisor.h> > > int smp_call_function_single(int cpu, void (*func) (void *info), void *info, > int wait) >
On 06/04/16 16:17, Juergen Gross wrote:> Use smp_call_on_cpu() to raise SMI on cpu 0. > Make call secure by adding get_online_cpus() to avoid e.g. suspend > resume cycles in between. > > Signed-off-by: Juergen Gross <jgross at suse.com>Could please some maintainer comment on this patch? Juergen> --- > V4: add call to get_online_cpus() > --- > drivers/firmware/dcdbas.c | 51 ++++++++++++++++++++++++----------------------- > 1 file changed, 26 insertions(+), 25 deletions(-) > > diff --git a/drivers/firmware/dcdbas.c b/drivers/firmware/dcdbas.c > index 829eec8..2fe1a13 100644 > --- a/drivers/firmware/dcdbas.c > +++ b/drivers/firmware/dcdbas.c > @@ -23,6 +23,7 @@ > #include <linux/platform_device.h> > #include <linux/dma-mapping.h> > #include <linux/errno.h> > +#include <linux/cpu.h> > #include <linux/gfp.h> > #include <linux/init.h> > #include <linux/kernel.h> > @@ -238,33 +239,14 @@ static ssize_t host_control_on_shutdown_store(struct device *dev, > return count; > } > > -/** > - * dcdbas_smi_request: generate SMI request > - * > - * Called with smi_data_lock. > - */ > -int dcdbas_smi_request(struct smi_cmd *smi_cmd) > +static int raise_smi(void *par) > { > - cpumask_var_t old_mask; > - int ret = 0; > + struct smi_cmd *smi_cmd = par; > > - if (smi_cmd->magic != SMI_CMD_MAGIC) { > - dev_info(&dcdbas_pdev->dev, "%s: invalid magic value\n", > - __func__); > - return -EBADR; > - } > - > - /* SMI requires CPU 0 */ > - if (!alloc_cpumask_var(&old_mask, GFP_KERNEL)) > - return -ENOMEM; > - > - cpumask_copy(old_mask, ¤t->cpus_allowed); > - set_cpus_allowed_ptr(current, cpumask_of(0)); > if (smp_processor_id() != 0) { > dev_dbg(&dcdbas_pdev->dev, "%s: failed to get CPU 0\n", > __func__); > - ret = -EBUSY; > - goto out; > + return -EBUSY; > } > > /* generate SMI */ > @@ -280,9 +262,28 @@ int dcdbas_smi_request(struct smi_cmd *smi_cmd) > : "memory" > ); > > -out: > - set_cpus_allowed_ptr(current, old_mask); > - free_cpumask_var(old_mask); > + return 0; > +} > +/** > + * dcdbas_smi_request: generate SMI request > + * > + * Called with smi_data_lock. > + */ > +int dcdbas_smi_request(struct smi_cmd *smi_cmd) > +{ > + int ret; > + > + if (smi_cmd->magic != SMI_CMD_MAGIC) { > + dev_info(&dcdbas_pdev->dev, "%s: invalid magic value\n", > + __func__); > + return -EBADR; > + } > + > + /* SMI requires CPU 0 */ > + get_online_cpus(); > + ret = smp_call_on_cpu(0, raise_smi, smi_cmd, true); > + put_online_cpus(); > + > return ret; > } > >
Juergen Gross
2016-Jun-20 08:40 UTC
[PATCH v5 0/6] Support calling functions on dedicated physical cpu
On 06/04/16 16:17, Juergen Gross wrote:> Some hardware (e.g. Dell Studio laptops) require special functions to > be called on physical cpu 0 in order to avoid occasional hangs. When > running as dom0 under Xen this could be achieved only via special boot > parameters (vcpu pinning) limiting the hypervisor in it's scheduling > decisions. > > This patch series is adding a generic function to be able to temporarily > pin a (virtual) cpu to a dedicated physical cpu for executing above > mentioned functions on that specific cpu. The drivers (dcdbas and i8k) > requiring this functionality are modified accordingly.Peter, Ingo, is one of you considering to take this series for 4.8? I know not all patches got an Ack, OTOH maintainers had 2.5 months time to object. I'd really appreciate this series would make it into 4.8. Juergen> > Changes in V5: > - patch 3: rename and reshuffle parameters of smp_call_on_cpu() as requested > by Peter Zijlstra > - patch 3: test target cpu to be online as requested by Peter Zijlstra > - patch 4: less wordy messages as requested by David Vrabel > > Changes in V4: > - move patches 5 and 6 further up in the series > - patch 2 (was 5): WARN_ONCE in case platform doesn't support pinning > as requested by Peter Zijlstra > - patch 3 (was 2): change return value in case of illegal cpu as > requested by Peter Zijlstra > - patch 3 (was 2): make pinning of vcpu an option as suggested by > Peter Zijlstra > - patches 5 and 6 (were 3 and 4): add call to get_online_cpus() > > Changes in V3: > - use get_cpu()/put_cpu() as suggested by David Vrabel > > Changes in V2: > - instead of manipulating the allowed set of cpus use cpu specific > workqueue as requested by Peter Zijlstra > - add include/linux/hypervisor.h to hide architecture specific stuff > from generic kernel code > > Juergen Gross (6): > xen: sync xen header > virt, sched: add generic vcpu pinning support > smp: add function to execute a function synchronously on a cpu > xen: add xen_pin_vcpu() to support calling functions on a dedicated > pcpu > dcdbas: make use of smp_call_on_cpu() > hwmon: use smp_call_on_cpu() for dell-smm i8k > > MAINTAINERS | 1 + > arch/x86/include/asm/hypervisor.h | 4 ++ > arch/x86/kernel/cpu/hypervisor.c | 11 +++++ > arch/x86/xen/enlighten.c | 40 +++++++++++++++ > drivers/firmware/dcdbas.c | 51 +++++++++---------- > drivers/hwmon/dell-smm-hwmon.c | 35 +++++++------ > include/linux/hypervisor.h | 17 +++++++ > include/linux/smp.h | 3 ++ > include/xen/interface/sched.h | 100 +++++++++++++++++++++++++++++++------- > kernel/smp.c | 51 +++++++++++++++++++ > kernel/up.c | 18 +++++++ > 11 files changed, 273 insertions(+), 58 deletions(-) > create mode 100644 include/linux/hypervisor.h >
Juergen Gross
2016-Jun-27 10:34 UTC
[PATCH v5 0/6] Support calling functions on dedicated physical cpu
On 06/04/16 16:17, Juergen Gross wrote:> Some hardware (e.g. Dell Studio laptops) require special functions to > be called on physical cpu 0 in order to avoid occasional hangs. When > running as dom0 under Xen this could be achieved only via special boot > parameters (vcpu pinning) limiting the hypervisor in it's scheduling > decisions. > > This patch series is adding a generic function to be able to temporarily > pin a (virtual) cpu to a dedicated physical cpu for executing above > mentioned functions on that specific cpu. The drivers (dcdbas and i8k) > requiring this functionality are modified accordingly.David, as I've got no reactions when asking for the last outstanding Acks (patches 2 and 5) and neither Ingo nor Peter seem to be willing to take the patches, I'd like to suggest we carry the series via the xen tree. Are you okay with this suggestion? Juergen> > Changes in V5: > - patch 3: rename and reshuffle parameters of smp_call_on_cpu() as requested > by Peter Zijlstra > - patch 3: test target cpu to be online as requested by Peter Zijlstra > - patch 4: less wordy messages as requested by David Vrabel > > Changes in V4: > - move patches 5 and 6 further up in the series > - patch 2 (was 5): WARN_ONCE in case platform doesn't support pinning > as requested by Peter Zijlstra > - patch 3 (was 2): change return value in case of illegal cpu as > requested by Peter Zijlstra > - patch 3 (was 2): make pinning of vcpu an option as suggested by > Peter Zijlstra > - patches 5 and 6 (were 3 and 4): add call to get_online_cpus() > > Changes in V3: > - use get_cpu()/put_cpu() as suggested by David Vrabel > > Changes in V2: > - instead of manipulating the allowed set of cpus use cpu specific > workqueue as requested by Peter Zijlstra > - add include/linux/hypervisor.h to hide architecture specific stuff > from generic kernel code > > Juergen Gross (6): > xen: sync xen header > virt, sched: add generic vcpu pinning support > smp: add function to execute a function synchronously on a cpu > xen: add xen_pin_vcpu() to support calling functions on a dedicated > pcpu > dcdbas: make use of smp_call_on_cpu() > hwmon: use smp_call_on_cpu() for dell-smm i8k > > MAINTAINERS | 1 + > arch/x86/include/asm/hypervisor.h | 4 ++ > arch/x86/kernel/cpu/hypervisor.c | 11 +++++ > arch/x86/xen/enlighten.c | 40 +++++++++++++++ > drivers/firmware/dcdbas.c | 51 +++++++++---------- > drivers/hwmon/dell-smm-hwmon.c | 35 +++++++------ > include/linux/hypervisor.h | 17 +++++++ > include/linux/smp.h | 3 ++ > include/xen/interface/sched.h | 100 +++++++++++++++++++++++++++++++------- > kernel/smp.c | 51 +++++++++++++++++++ > kernel/up.c | 18 +++++++ > 11 files changed, 273 insertions(+), 58 deletions(-) > create mode 100644 include/linux/hypervisor.h >
Maybe Matching Threads
- [PATCH v4 6/6] hwmon: use smp_call_on_cpu() for dell-smm i8k
- [PATCH v5 6/6] hwmon: use smp_call_on_cpu() for dell-smm i8k
- [PATCH v4 6/6] hwmon: use smp_call_on_cpu() for dell-smm i8k
- [PATCH v4 6/6] hwmon: use smp_call_on_cpu() for dell-smm i8k
- [PATCH v5 0/6] Support calling functions on dedicated physical cpu