| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Fri, 21 Apr 2023 16:17:16 +0200
From: Peter Newman <peternewman@...gle.com>
To: Fenghua Yu <fenghua.yu@...el.com>,
Reinette Chatre <reinette.chatre@...el.com>
Cc: Babu Moger <babu.moger@....com>,
Thomas Gleixner <tglx@...utronix.de>,
Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
Dave Hansen <dave.hansen@...ux.intel.com>, x86@...nel.org,
"H. Peter Anvin" <hpa@...or.com>,
Stephane Eranian <eranian@...gle.com>,
James Morse <james.morse@....com>,
linux-kernel@...r.kernel.org, linux-kselftest@...r.kernel.org,
Peter Newman <peternewman@...gle.com>
Subject: [PATCH v1 2/9] x86/resctrl: Hold a spinlock in __rmid_read() on AMD
From: Stephane Eranian <eranian@...gle.com>
In AMD PQoS Versions 1.0 and 2.0, IA32_QM_EVTSEL MSR is shared by all
processors in a QOS domain. So there's a chance it can read a different
event when two processors are reading the counter concurrently. Add a
spinlock to prevent this race.
Co-developed-by: Peter Newman <peternewman@...gle.com>
Signed-off-by: Peter Newman <peternewman@...gle.com>
Signed-off-by: Stephane Eranian <eranian@...gle.com>
---
arch/x86/kernel/cpu/resctrl/core.c | 41 ++++++++++++++++++++++++++
arch/x86/kernel/cpu/resctrl/internal.h | 5 ++++
arch/x86/kernel/cpu/resctrl/monitor.c | 14 +++++++--
3 files changed, 57 insertions(+), 3 deletions(-)
diff --git a/arch/x86/kernel/cpu/resctrl/core.c b/arch/x86/kernel/cpu/resctrl/core.c
index 030d3b409768..47b1c37a81f8 100644
--- a/arch/x86/kernel/cpu/resctrl/core.c
+++ b/arch/x86/kernel/cpu/resctrl/core.c
@@ -25,6 +25,8 @@
#include <asm/resctrl.h>
#include "internal.h"
+DEFINE_STATIC_KEY_FALSE(rmid_read_locked);
+
/* Mutex to protect rdtgroup access. */
DEFINE_MUTEX(rdtgroup_mutex);
@@ -529,6 +531,8 @@ static void domain_add_cpu(int cpu, struct rdt_resource *r)
d->id = id;
cpumask_set_cpu(cpu, &d->cpu_mask);
+ raw_spin_lock_init(&hw_dom->evtsel_lock);
+
rdt_domain_reconfigure_cdp(r);
if (r->alloc_capable && domain_setup_ctrlval(r, d)) {
@@ -829,6 +833,41 @@ static __init bool get_rdt_mon_resources(void)
return !rdt_get_mon_l3_config(r);
}
+static __init bool amd_shared_qm_evtsel(void)
+{
+ /*
+ * From AMD64 Technology Platform Quality of Service Extensions,
+ * Revision 1.03:
+ *
+ * "For PQoS Version 1.0 and 2.0, as identified by Family/Model, the
+ * QM_EVTSEL register is shared by all the processors in a QOS domain."
+ *
+ * Check the inclusive Family/Model ranges for PQoS Extension versions
+ * 1.0 and 2.0 from the PQoS Extension Versions table.
+ */
+ if (boot_cpu_data.x86 == 0x17)
+ /* V1.0 */
+ return boot_cpu_data.x86_model >= 0x30 &&
+ boot_cpu_data.x86_model <= 0x9f;
+
+ if (boot_cpu_data.x86 == 0x19)
+ /* V2.0 */
+ return (boot_cpu_data.x86_model <= 0xf) ||
+ ((boot_cpu_data.x86_model >= 0x20) &&
+ (boot_cpu_data.x86_model <= 0x5f));
+
+ return false;
+}
+
+static __init void __check_quirks_amd(void)
+{
+ if (rdt_cpu_has(X86_FEATURE_CQM_MBM_TOTAL) ||
+ rdt_cpu_has(X86_FEATURE_CQM_MBM_LOCAL)) {
+ if (amd_shared_qm_evtsel())
+ static_branch_enable(&rmid_read_locked);
+ }
+}
+
static __init void __check_quirks_intel(void)
{
switch (boot_cpu_data.x86_model) {
@@ -852,6 +891,8 @@ static __init void check_quirks(void)
{
if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL)
__check_quirks_intel();
+ else if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
+ __check_quirks_amd();
}
static __init bool get_rdt_resources(void)
diff --git a/arch/x86/kernel/cpu/resctrl/internal.h b/arch/x86/kernel/cpu/resctrl/internal.h
index 85ceaf9a31ac..02a062558c67 100644
--- a/arch/x86/kernel/cpu/resctrl/internal.h
+++ b/arch/x86/kernel/cpu/resctrl/internal.h
@@ -325,6 +325,7 @@ struct arch_mbm_state {
* @ctrl_val: array of cache or mem ctrl values (indexed by CLOSID)
* @arch_mbm_total: arch private state for MBM total bandwidth
* @arch_mbm_local: arch private state for MBM local bandwidth
+ * @lock: serializes counter reads when QM_EVTSEL MSR is shared per-domain
*
* Members of this structure are accessed via helpers that provide abstraction.
*/
@@ -333,6 +334,7 @@ struct rdt_hw_domain {
u32 *ctrl_val;
struct arch_mbm_state *arch_mbm_total;
struct arch_mbm_state *arch_mbm_local;
+ raw_spinlock_t evtsel_lock;
};
static inline struct rdt_hw_domain *resctrl_to_arch_dom(struct rdt_domain *r)
@@ -428,6 +430,9 @@ extern struct rdt_hw_resource rdt_resources_all[];
extern struct rdtgroup rdtgroup_default;
DECLARE_STATIC_KEY_FALSE(rdt_alloc_enable_key);
+/* Serialization required in resctrl_arch_rmid_read(). */
+DECLARE_STATIC_KEY_FALSE(rmid_read_locked);
+
extern struct dentry *debugfs_resctrl;
enum resctrl_res_level {
diff --git a/arch/x86/kernel/cpu/resctrl/monitor.c b/arch/x86/kernel/cpu/resctrl/monitor.c
index 20952419be75..2de8397f91cd 100644
--- a/arch/x86/kernel/cpu/resctrl/monitor.c
+++ b/arch/x86/kernel/cpu/resctrl/monitor.c
@@ -146,10 +146,15 @@ static inline struct rmid_entry *__rmid_entry(u32 rmid)
return entry;
}
-static int __rmid_read(u32 rmid, enum resctrl_event_id eventid, u64 *val)
+static int __rmid_read(struct rdt_hw_domain *hw_dom, u32 rmid,
+ enum resctrl_event_id eventid, u64 *val)
{
+ unsigned long flags;
u64 msr_val;
+ if (static_branch_likely(&rmid_read_locked))
+ raw_spin_lock_irqsave(&hw_dom->evtsel_lock, flags);
+
/*
* As per the SDM, when IA32_QM_EVTSEL.EvtID (bits 7:0) is configured
* with a valid event code for supported resource type and the bits
@@ -161,6 +166,9 @@ static int __rmid_read(u32 rmid, enum resctrl_event_id eventid, u64 *val)
wrmsr(MSR_IA32_QM_EVTSEL, eventid, rmid);
rdmsrl(MSR_IA32_QM_CTR, msr_val);
+ if (static_branch_likely(&rmid_read_locked))
+ raw_spin_unlock_irqrestore(&hw_dom->evtsel_lock, flags);
+
if (msr_val & RMID_VAL_ERROR)
return -EIO;
if (msr_val & RMID_VAL_UNAVAIL)
@@ -200,7 +208,7 @@ void resctrl_arch_reset_rmid(struct rdt_resource *r, struct rdt_domain *d,
memset(am, 0, sizeof(*am));
/* Record any initial, non-zero count value. */
- __rmid_read(rmid, eventid, &am->prev_msr);
+ __rmid_read(hw_dom, rmid, eventid, &am->prev_msr);
}
}
@@ -241,7 +249,7 @@ int resctrl_arch_rmid_read(struct rdt_resource *r, struct rdt_domain *d,
if (!cpumask_test_cpu(smp_processor_id(), &d->cpu_mask))
return -EINVAL;
- ret = __rmid_read(rmid, eventid, &msr_val);
+ ret = __rmid_read(hw_dom, rmid, eventid, &msr_val);
if (ret)
return ret;
--
2.40.0.634.g4ca3ef3211-goog
Powered by blists - more mailing lists