LKML Archive on lore.kernel.org
 help / Atom feed
* [PATCH v2 0/8] CR4 handling improvements
@ 2014-10-24 22:58 Andy Lutomirski
  2014-10-24 22:58 ` [PATCH v2 1/8] perf: Clean up pmu::event_idx Andy Lutomirski
                   ` (7 more replies)
  0 siblings, 8 replies; 31+ messages in thread
From: Andy Lutomirski @ 2014-10-24 22:58 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, Andy Lutomirski

This little series tightens up rdpmc permissions.  With it applied,
rdpmc can only be used if a perf_event is actually mmapped.  For now,
this is only really useful for seccomp.

At some point this could be further tightened up to only allow rdpmc
if an actual self-monitoring perf event that is compatible with
rdpmc is mapped.

This should add <50ns to context switches between rdpmc-capable and
rdpmc-non-capable mms.  I suspect that this is well under 10%
overhead, given that perf already adds some context switch latency.

I think that patches 1-3 are a good idea regardless of any rdpmc changes.

AMD Uncore userspace rdpmc is broken by these patches (cap_user_rdpmc
will be zero), but it was broken anyway.

Changes from v1 (aka RFC):
 - Rebased on top of the KVM CR4 fix.  This applies to a very recent -linus.
 - Renamed the cr4 helpers (Peter, Borislav)
 - Fixed buggy cr4 helpers (Hilf)
 - Improved lots of comments (everyone)
 - Renamed read_cr4 and write_cr4 to make sure I didn't miss anything.
   (NB: This will introduce conflicts with Andi's FSGSBASE work.  This is
    a good thing.)

Andy Lutomirski (7):
  x86: Clean up cr4 manipulation
  x86: Store a per-cpu shadow copy of CR4
  x86: Add a comment clarifying LDT context switching
  perf: Add pmu callbacks to track event mapping and unmapping
  perf: Pass the event to arch_perf_update_userpage
  x86, perf: Only allow rdpmc if a perf_event is mapped
  x86, perf: Add /sys/devices/cpu/rdpmc=2 to allow rdpmc for all tasks

Peter Zijlstra (1):
  perf: Clean up pmu::event_idx

 arch/powerpc/perf/hv-24x7.c          |  6 ---
 arch/powerpc/perf/hv-gpci.c          |  6 ---
 arch/s390/kernel/perf_cpum_sf.c      |  6 ---
 arch/x86/include/asm/mmu.h           |  2 +
 arch/x86/include/asm/mmu_context.h   | 32 ++++++++++++++-
 arch/x86/include/asm/paravirt.h      |  6 +--
 arch/x86/include/asm/processor.h     | 33 ----------------
 arch/x86/include/asm/special_insns.h |  6 +--
 arch/x86/include/asm/tlbflush.h      | 77 ++++++++++++++++++++++++++++++++----
 arch/x86/include/asm/virtext.h       |  5 ++-
 arch/x86/kernel/acpi/sleep.c         |  2 +-
 arch/x86/kernel/cpu/common.c         | 17 +++++---
 arch/x86/kernel/cpu/mcheck/mce.c     |  3 +-
 arch/x86/kernel/cpu/mcheck/p5.c      |  3 +-
 arch/x86/kernel/cpu/mcheck/winchip.c |  3 +-
 arch/x86/kernel/cpu/mtrr/cyrix.c     |  6 +--
 arch/x86/kernel/cpu/mtrr/generic.c   |  6 +--
 arch/x86/kernel/cpu/perf_event.c     | 76 ++++++++++++++++++++++++++---------
 arch/x86/kernel/cpu/perf_event.h     |  2 +
 arch/x86/kernel/head32.c             |  1 +
 arch/x86/kernel/head64.c             |  2 +
 arch/x86/kernel/i387.c               |  3 +-
 arch/x86/kernel/process.c            |  5 ++-
 arch/x86/kernel/process_32.c         |  2 +-
 arch/x86/kernel/process_64.c         |  2 +-
 arch/x86/kernel/setup.c              |  2 +-
 arch/x86/kernel/xsave.c              |  3 +-
 arch/x86/kvm/svm.c                   |  2 +-
 arch/x86/kvm/vmx.c                   | 10 ++---
 arch/x86/mm/fault.c                  |  2 +-
 arch/x86/mm/init.c                   | 12 +++++-
 arch/x86/mm/tlb.c                    |  3 --
 arch/x86/power/cpu.c                 | 11 ++----
 arch/x86/realmode/init.c             |  2 +-
 arch/x86/xen/enlighten.c             |  4 +-
 drivers/lguest/x86/core.c            |  4 +-
 include/linux/perf_event.h           |  7 ++++
 kernel/events/core.c                 | 29 ++++++--------
 kernel/events/hw_breakpoint.c        |  7 ----
 39 files changed, 256 insertions(+), 154 deletions(-)

-- 
1.9.3


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [PATCH v2 1/8] perf: Clean up pmu::event_idx
  2014-10-24 22:58 [PATCH v2 0/8] CR4 handling improvements Andy Lutomirski
@ 2014-10-24 22:58 ` Andy Lutomirski
  2014-10-24 22:58 ` [PATCH v2 2/8] x86: Clean up cr4 manipulation Andy Lutomirski
                   ` (6 subsequent siblings)
  7 siblings, 0 replies; 31+ messages in thread
From: Andy Lutomirski @ 2014-10-24 22:58 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj

From: Peter Zijlstra <peterz@infradead.org>

Andy reported that the current state of event_idx is rather confused.
So remove all but the x86_pmu implementation and change the default to
return 0 (the safe option).

Reported-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/powerpc/perf/hv-24x7.c     |  6 ------
 arch/powerpc/perf/hv-gpci.c     |  6 ------
 arch/s390/kernel/perf_cpum_sf.c |  6 ------
 kernel/events/core.c            | 15 +--------------
 kernel/events/hw_breakpoint.c   |  7 -------
 5 files changed, 1 insertion(+), 39 deletions(-)

diff --git a/arch/powerpc/perf/hv-24x7.c b/arch/powerpc/perf/hv-24x7.c
index 6c8710dd90c9..dba34088da28 100644
--- a/arch/powerpc/perf/hv-24x7.c
+++ b/arch/powerpc/perf/hv-24x7.c
@@ -417,11 +417,6 @@ static int h_24x7_event_add(struct perf_event *event, int flags)
 	return 0;
 }
 
-static int h_24x7_event_idx(struct perf_event *event)
-{
-	return 0;
-}
-
 static struct pmu h_24x7_pmu = {
 	.task_ctx_nr = perf_invalid_context,
 
@@ -433,7 +428,6 @@ static struct pmu h_24x7_pmu = {
 	.start       = h_24x7_event_start,
 	.stop        = h_24x7_event_stop,
 	.read        = h_24x7_event_update,
-	.event_idx   = h_24x7_event_idx,
 };
 
 static int hv_24x7_init(void)
diff --git a/arch/powerpc/perf/hv-gpci.c b/arch/powerpc/perf/hv-gpci.c
index 15fc76c93022..a051fe946c63 100644
--- a/arch/powerpc/perf/hv-gpci.c
+++ b/arch/powerpc/perf/hv-gpci.c
@@ -246,11 +246,6 @@ static int h_gpci_event_init(struct perf_event *event)
 	return 0;
 }
 
-static int h_gpci_event_idx(struct perf_event *event)
-{
-	return 0;
-}
-
 static struct pmu h_gpci_pmu = {
 	.task_ctx_nr = perf_invalid_context,
 
@@ -262,7 +257,6 @@ static struct pmu h_gpci_pmu = {
 	.start       = h_gpci_event_start,
 	.stop        = h_gpci_event_stop,
 	.read        = h_gpci_event_update,
-	.event_idx   = h_gpci_event_idx,
 };
 
 static int hv_gpci_init(void)
diff --git a/arch/s390/kernel/perf_cpum_sf.c b/arch/s390/kernel/perf_cpum_sf.c
index 08e761318c17..b878f12a9597 100644
--- a/arch/s390/kernel/perf_cpum_sf.c
+++ b/arch/s390/kernel/perf_cpum_sf.c
@@ -1411,11 +1411,6 @@ static void cpumsf_pmu_del(struct perf_event *event, int flags)
 	perf_pmu_enable(event->pmu);
 }
 
-static int cpumsf_pmu_event_idx(struct perf_event *event)
-{
-	return event->hw.idx;
-}
-
 CPUMF_EVENT_ATTR(SF, SF_CYCLES_BASIC, PERF_EVENT_CPUM_SF);
 CPUMF_EVENT_ATTR(SF, SF_CYCLES_BASIC_DIAG, PERF_EVENT_CPUM_SF_DIAG);
 
@@ -1458,7 +1453,6 @@ static struct pmu cpumf_sampling = {
 	.stop	      = cpumsf_pmu_stop,
 	.read	      = cpumsf_pmu_read,
 
-	.event_idx    = cpumsf_pmu_event_idx,
 	.attr_groups  = cpumsf_pmu_attr_groups,
 };
 
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 1425d07018de..2b02c9fda790 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -6071,11 +6071,6 @@ static int perf_swevent_init(struct perf_event *event)
 	return 0;
 }
 
-static int perf_swevent_event_idx(struct perf_event *event)
-{
-	return 0;
-}
-
 static struct pmu perf_swevent = {
 	.task_ctx_nr	= perf_sw_context,
 
@@ -6085,8 +6080,6 @@ static struct pmu perf_swevent = {
 	.start		= perf_swevent_start,
 	.stop		= perf_swevent_stop,
 	.read		= perf_swevent_read,
-
-	.event_idx	= perf_swevent_event_idx,
 };
 
 #ifdef CONFIG_EVENT_TRACING
@@ -6204,8 +6197,6 @@ static struct pmu perf_tracepoint = {
 	.start		= perf_swevent_start,
 	.stop		= perf_swevent_stop,
 	.read		= perf_swevent_read,
-
-	.event_idx	= perf_swevent_event_idx,
 };
 
 static inline void perf_tp_register(void)
@@ -6431,8 +6422,6 @@ static struct pmu perf_cpu_clock = {
 	.start		= cpu_clock_event_start,
 	.stop		= cpu_clock_event_stop,
 	.read		= cpu_clock_event_read,
-
-	.event_idx	= perf_swevent_event_idx,
 };
 
 /*
@@ -6511,8 +6500,6 @@ static struct pmu perf_task_clock = {
 	.start		= task_clock_event_start,
 	.stop		= task_clock_event_stop,
 	.read		= task_clock_event_read,
-
-	.event_idx	= perf_swevent_event_idx,
 };
 
 static void perf_pmu_nop_void(struct pmu *pmu)
@@ -6542,7 +6529,7 @@ static void perf_pmu_cancel_txn(struct pmu *pmu)
 
 static int perf_event_idx_default(struct perf_event *event)
 {
-	return event->hw.idx + 1;
+	return 0;
 }
 
 /*
diff --git a/kernel/events/hw_breakpoint.c b/kernel/events/hw_breakpoint.c
index 1559fb0b9296..9803a6600d49 100644
--- a/kernel/events/hw_breakpoint.c
+++ b/kernel/events/hw_breakpoint.c
@@ -605,11 +605,6 @@ static void hw_breakpoint_stop(struct perf_event *bp, int flags)
 	bp->hw.state = PERF_HES_STOPPED;
 }
 
-static int hw_breakpoint_event_idx(struct perf_event *bp)
-{
-	return 0;
-}
-
 static struct pmu perf_breakpoint = {
 	.task_ctx_nr	= perf_sw_context, /* could eventually get its own */
 
@@ -619,8 +614,6 @@ static struct pmu perf_breakpoint = {
 	.start		= hw_breakpoint_start,
 	.stop		= hw_breakpoint_stop,
 	.read		= hw_breakpoint_pmu_read,
-
-	.event_idx	= hw_breakpoint_event_idx,
 };
 
 int __init init_hw_breakpoint(void)
-- 
1.9.3


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [PATCH v2 2/8] x86: Clean up cr4 manipulation
  2014-10-24 22:58 [PATCH v2 0/8] CR4 handling improvements Andy Lutomirski
  2014-10-24 22:58 ` [PATCH v2 1/8] perf: Clean up pmu::event_idx Andy Lutomirski
@ 2014-10-24 22:58 ` Andy Lutomirski
  2014-11-01 19:56   ` Thomas Gleixner
  2015-02-04 14:41   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
  2014-10-24 22:58 ` [PATCH v2 3/8] x86: Store a per-cpu shadow copy of CR4 Andy Lutomirski
                   ` (5 subsequent siblings)
  7 siblings, 2 replies; 31+ messages in thread
From: Andy Lutomirski @ 2014-10-24 22:58 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, Andy Lutomirski

CR4 manipulation was split, seemingly at random, between direct
(write_cr4) and using a helper (set/clear_in_cr4).  Unfortunately,
the set_in_cr4 and clear_in_cr4 helpers also poke at the boot code,
which only a small subset of users actually wanted.

This patch replaces all cr4 access in functions that don't leave cr4
exactly the way they found it with new helpers cr4_set_bits,
cr4_clear_bits, and cr4_set_bits_and_update_boot.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
---
 arch/x86/include/asm/processor.h     | 33 --------------------------------
 arch/x86/include/asm/tlbflush.h      | 37 ++++++++++++++++++++++++++++++++++++
 arch/x86/include/asm/virtext.h       |  3 ++-
 arch/x86/kernel/cpu/common.c         | 10 +++++-----
 arch/x86/kernel/cpu/mcheck/mce.c     |  3 ++-
 arch/x86/kernel/cpu/mcheck/p5.c      |  3 ++-
 arch/x86/kernel/cpu/mcheck/winchip.c |  3 ++-
 arch/x86/kernel/cpu/perf_event.c     |  7 ++++---
 arch/x86/kernel/i387.c               |  3 ++-
 arch/x86/kernel/process.c            |  5 +++--
 arch/x86/kernel/xsave.c              |  3 ++-
 arch/x86/kvm/vmx.c                   |  4 ++--
 arch/x86/mm/init.c                   |  4 ++--
 arch/x86/xen/enlighten.c             |  4 ++--
 drivers/lguest/x86/core.c            |  4 ++--
 15 files changed, 69 insertions(+), 57 deletions(-)

diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index eb71ec794732..ddd8d13a010f 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -578,39 +578,6 @@ static inline void load_sp0(struct tss_struct *tss,
 #define set_iopl_mask native_set_iopl_mask
 #endif /* CONFIG_PARAVIRT */
 
-/*
- * Save the cr4 feature set we're using (ie
- * Pentium 4MB enable and PPro Global page
- * enable), so that any CPU's that boot up
- * after us can get the correct flags.
- */
-extern unsigned long mmu_cr4_features;
-extern u32 *trampoline_cr4_features;
-
-static inline void set_in_cr4(unsigned long mask)
-{
-	unsigned long cr4;
-
-	mmu_cr4_features |= mask;
-	if (trampoline_cr4_features)
-		*trampoline_cr4_features = mmu_cr4_features;
-	cr4 = read_cr4();
-	cr4 |= mask;
-	write_cr4(cr4);
-}
-
-static inline void clear_in_cr4(unsigned long mask)
-{
-	unsigned long cr4;
-
-	mmu_cr4_features &= ~mask;
-	if (trampoline_cr4_features)
-		*trampoline_cr4_features = mmu_cr4_features;
-	cr4 = read_cr4();
-	cr4 &= ~mask;
-	write_cr4(cr4);
-}
-
 typedef struct {
 	unsigned long		seg;
 } mm_segment_t;
diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h
index 04905bfc508b..fc0c4bc356ce 100644
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -15,6 +15,43 @@
 #define __flush_tlb_single(addr) __native_flush_tlb_single(addr)
 #endif
 
+/* Set in this cpu's CR4. */
+static inline void cr4_set_bits(unsigned long mask)
+{
+	unsigned long cr4;
+
+	cr4 = read_cr4();
+	cr4 |= mask;
+	write_cr4(cr4);
+}
+
+/* Clear in this cpu's CR4. */
+static inline void cr4_clear_bits(unsigned long mask)
+{
+	unsigned long cr4;
+
+	cr4 = read_cr4();
+	cr4 &= ~mask;
+	write_cr4(cr4);
+}
+
+/*
+ * Save some of cr4 feature set we're using (e.g.  Pentium 4MB
+ * enable and PPro Global page enable), so that any CPU's that boot
+ * up after us can get the correct flags.  This should only be used
+ * during boot on the boot cpu.
+ */
+extern unsigned long mmu_cr4_features;
+extern u32 *trampoline_cr4_features;
+
+static inline void cr4_set_bits_and_update_boot(unsigned long mask)
+{
+	mmu_cr4_features |= mask;
+	if (trampoline_cr4_features)
+		*trampoline_cr4_features = mmu_cr4_features;
+	cr4_set_bits(mask);
+}
+
 static inline void __native_flush_tlb(void)
 {
 	native_write_cr3(native_read_cr3());
diff --git a/arch/x86/include/asm/virtext.h b/arch/x86/include/asm/virtext.h
index 5da71c27cc59..f41e19ca717b 100644
--- a/arch/x86/include/asm/virtext.h
+++ b/arch/x86/include/asm/virtext.h
@@ -19,6 +19,7 @@
 
 #include <asm/vmx.h>
 #include <asm/svm.h>
+#include <asm/tlbflush.h>
 
 /*
  * VMX functions:
@@ -40,7 +41,7 @@ static inline int cpu_has_vmx(void)
 static inline void cpu_vmxoff(void)
 {
 	asm volatile (ASM_VMX_VMXOFF : : : "cc");
-	write_cr4(read_cr4() & ~X86_CR4_VMXE);
+	cr4_clear_bits(X86_CR4_VMXE);
 }
 
 static inline int cpu_vmx_enabled(void)
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 4b4f78c9ba19..847d62b78431 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -276,7 +276,7 @@ __setup("nosmep", setup_disable_smep);
 static __always_inline void setup_smep(struct cpuinfo_x86 *c)
 {
 	if (cpu_has(c, X86_FEATURE_SMEP))
-		set_in_cr4(X86_CR4_SMEP);
+		cr4_set_bits(X86_CR4_SMEP);
 }
 
 static __init int setup_disable_smap(char *arg)
@@ -296,9 +296,9 @@ static __always_inline void setup_smap(struct cpuinfo_x86 *c)
 
 	if (cpu_has(c, X86_FEATURE_SMAP)) {
 #ifdef CONFIG_X86_SMAP
-		set_in_cr4(X86_CR4_SMAP);
+		cr4_set_bits(X86_CR4_SMAP);
 #else
-		clear_in_cr4(X86_CR4_SMAP);
+		cr4_clear_bits(X86_CR4_SMAP);
 #endif
 	}
 }
@@ -1320,7 +1320,7 @@ void cpu_init(void)
 
 	pr_debug("Initializing CPU#%d\n", cpu);
 
-	clear_in_cr4(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
+	cr4_clear_bits(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
 
 	/*
 	 * Initialize the per-CPU GDT with the boot GDT,
@@ -1401,7 +1401,7 @@ void cpu_init(void)
 	printk(KERN_INFO "Initializing CPU#%d\n", cpu);
 
 	if (cpu_feature_enabled(X86_FEATURE_VME) || cpu_has_tsc || cpu_has_de)
-		clear_in_cr4(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
+		cr4_clear_bits(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
 
 	load_current_idt();
 	switch_to_new_gdt(cpu);
diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index 61a9668cebfd..617bc12ef575 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -43,6 +43,7 @@
 #include <linux/export.h>
 
 #include <asm/processor.h>
+#include <asm/tlbflush.h>
 #include <asm/mce.h>
 #include <asm/msr.h>
 
@@ -1455,7 +1456,7 @@ static void __mcheck_cpu_init_generic(void)
 	bitmap_fill(all_banks, MAX_NR_BANKS);
 	machine_check_poll(MCP_UC | m_fl, &all_banks);
 
-	set_in_cr4(X86_CR4_MCE);
+	cr4_set_bits(X86_CR4_MCE);
 
 	rdmsrl(MSR_IA32_MCG_CAP, cap);
 	if (cap & MCG_CTL_P)
diff --git a/arch/x86/kernel/cpu/mcheck/p5.c b/arch/x86/kernel/cpu/mcheck/p5.c
index a3042989398c..30692ac88d1e 100644
--- a/arch/x86/kernel/cpu/mcheck/p5.c
+++ b/arch/x86/kernel/cpu/mcheck/p5.c
@@ -8,6 +8,7 @@
 #include <linux/smp.h>
 
 #include <asm/processor.h>
+#include <asm/tlbflush.h>
 #include <asm/mce.h>
 #include <asm/msr.h>
 
@@ -59,7 +60,7 @@ void intel_p5_mcheck_init(struct cpuinfo_x86 *c)
 	       "Intel old style machine check architecture supported.\n");
 
 	/* Enable MCE: */
-	set_in_cr4(X86_CR4_MCE);
+	cr4_set_bits(X86_CR4_MCE);
 	printk(KERN_INFO
 	       "Intel old style machine check reporting enabled on CPU#%d.\n",
 	       smp_processor_id());
diff --git a/arch/x86/kernel/cpu/mcheck/winchip.c b/arch/x86/kernel/cpu/mcheck/winchip.c
index 7dc5564d0cdf..590cc753ba8f 100644
--- a/arch/x86/kernel/cpu/mcheck/winchip.c
+++ b/arch/x86/kernel/cpu/mcheck/winchip.c
@@ -7,6 +7,7 @@
 #include <linux/types.h>
 
 #include <asm/processor.h>
+#include <asm/tlbflush.h>
 #include <asm/mce.h>
 #include <asm/msr.h>
 
@@ -31,7 +32,7 @@ void winchip_mcheck_init(struct cpuinfo_x86 *c)
 	lo &= ~(1<<4);	/* Enable MCE */
 	wrmsr(MSR_IDT_FCR1, lo, hi);
 
-	set_in_cr4(X86_CR4_MCE);
+	cr4_set_bits(X86_CR4_MCE);
 
 	printk(KERN_INFO
 	       "Winchip machine check reporting enabled on CPU#0.\n");
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 1b8299dd3d91..2176647fdbb7 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -31,6 +31,7 @@
 #include <asm/nmi.h>
 #include <asm/smp.h>
 #include <asm/alternative.h>
+#include <asm/tlbflush.h>
 #include <asm/timer.h>
 #include <asm/desc.h>
 #include <asm/ldt.h>
@@ -1336,7 +1337,7 @@ x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
 
 	case CPU_STARTING:
 		if (x86_pmu.attr_rdpmc)
-			set_in_cr4(X86_CR4_PCE);
+			cr4_set_bits(X86_CR4_PCE);
 		if (x86_pmu.cpu_starting)
 			x86_pmu.cpu_starting(cpu);
 		break;
@@ -1842,9 +1843,9 @@ static void change_rdpmc(void *info)
 	bool enable = !!(unsigned long)info;
 
 	if (enable)
-		set_in_cr4(X86_CR4_PCE);
+		cr4_set_bits(X86_CR4_PCE);
 	else
-		clear_in_cr4(X86_CR4_PCE);
+		cr4_clear_bits(X86_CR4_PCE);
 }
 
 static ssize_t set_attr_rdpmc(struct device *cdev,
diff --git a/arch/x86/kernel/i387.c b/arch/x86/kernel/i387.c
index a9a4229f6161..87727b03196d 100644
--- a/arch/x86/kernel/i387.c
+++ b/arch/x86/kernel/i387.c
@@ -13,6 +13,7 @@
 #include <asm/sigcontext.h>
 #include <asm/processor.h>
 #include <asm/math_emu.h>
+#include <asm/tlbflush.h>
 #include <asm/uaccess.h>
 #include <asm/ptrace.h>
 #include <asm/i387.h>
@@ -180,7 +181,7 @@ void fpu_init(void)
 	if (cpu_has_xmm)
 		cr4_mask |= X86_CR4_OSXMMEXCPT;
 	if (cr4_mask)
-		set_in_cr4(cr4_mask);
+		cr4_set_bits(cr4_mask);
 
 	cr0 = read_cr0();
 	cr0 &= ~(X86_CR0_TS|X86_CR0_EM); /* clear TS and EM */
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index e127ddaa2d5a..046e2d620bbe 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -28,6 +28,7 @@
 #include <asm/fpu-internal.h>
 #include <asm/debugreg.h>
 #include <asm/nmi.h>
+#include <asm/tlbflush.h>
 
 /*
  * per-CPU TSS segments. Threads are completely 'soft' on Linux,
@@ -141,7 +142,7 @@ void flush_thread(void)
 
 static void hard_disable_TSC(void)
 {
-	write_cr4(read_cr4() | X86_CR4_TSD);
+	cr4_set_bits(X86_CR4_TSD);
 }
 
 void disable_TSC(void)
@@ -158,7 +159,7 @@ void disable_TSC(void)
 
 static void hard_enable_TSC(void)
 {
-	write_cr4(read_cr4() & ~X86_CR4_TSD);
+	cr4_clear_bits(X86_CR4_TSD);
 }
 
 static void enable_TSC(void)
diff --git a/arch/x86/kernel/xsave.c b/arch/x86/kernel/xsave.c
index 4c540c4719d8..9828fdff6d59 100644
--- a/arch/x86/kernel/xsave.c
+++ b/arch/x86/kernel/xsave.c
@@ -12,6 +12,7 @@
 #include <asm/i387.h>
 #include <asm/fpu-internal.h>
 #include <asm/sigframe.h>
+#include <asm/tlbflush.h>
 #include <asm/xcr.h>
 
 /*
@@ -453,7 +454,7 @@ static void prepare_fx_sw_frame(void)
  */
 static inline void xstate_enable(void)
 {
-	set_in_cr4(X86_CR4_OSXSAVE);
+	cr4_set_bits(X86_CR4_OSXSAVE);
 	xsetbv(XCR_XFEATURE_ENABLED_MASK, pcntxt_mask);
 }
 
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 0acac81f198b..93e2649fb786 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -2765,7 +2765,7 @@ static int hardware_enable(void)
 		/* enable and lock */
 		wrmsrl(MSR_IA32_FEATURE_CONTROL, old | test_bits);
 	}
-	write_cr4(read_cr4() | X86_CR4_VMXE); /* FIXME: not cpu hotplug safe */
+	cr4_set_bits(X86_CR4_VMXE);
 
 	if (vmm_exclusive) {
 		kvm_cpu_vmxon(phys_addr);
@@ -2802,7 +2802,7 @@ static void hardware_disable(void)
 		vmclear_local_loaded_vmcss();
 		kvm_cpu_vmxoff();
 	}
-	write_cr4(read_cr4() & ~X86_CR4_VMXE);
+	cr4_clear_bits(X86_CR4_VMXE);
 }
 
 static __init int adjust_vmx_controls(u32 ctl_min, u32 ctl_opt,
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index 66dba36f2343..73b37cd0d253 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -144,11 +144,11 @@ static void __init probe_page_size_mask(void)
 
 	/* Enable PSE if available */
 	if (cpu_has_pse)
-		set_in_cr4(X86_CR4_PSE);
+		cr4_set_bits_and_update_boot(X86_CR4_PSE);
 
 	/* Enable PGE if available */
 	if (cpu_has_pge) {
-		set_in_cr4(X86_CR4_PGE);
+		cr4_set_bits_and_update_boot(X86_CR4_PGE);
 		__supported_pte_mask |= _PAGE_GLOBAL;
 	}
 }
diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 1a3f0445432a..03920ef71f9c 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -1483,10 +1483,10 @@ static void xen_pvh_set_cr_flags(int cpu)
 	 * set them here. For all, OSFXSR OSXMMEXCPT are set in fpu_init.
 	*/
 	if (cpu_has_pse)
-		set_in_cr4(X86_CR4_PSE);
+		cr4_set_bits_and_update_boot(X86_CR4_PSE);
 
 	if (cpu_has_pge)
-		set_in_cr4(X86_CR4_PGE);
+		cr4_set_bits_and_update_boot(X86_CR4_PGE);
 }
 
 /*
diff --git a/drivers/lguest/x86/core.c b/drivers/lguest/x86/core.c
index 922a1acbf652..e1700ff370dd 100644
--- a/drivers/lguest/x86/core.c
+++ b/drivers/lguest/x86/core.c
@@ -452,9 +452,9 @@ void lguest_arch_handle_trap(struct lg_cpu *cpu)
 static void adjust_pge(void *on)
 {
 	if (on)
-		write_cr4(read_cr4() | X86_CR4_PGE);
+		cr4_set_bits(X86_CR4_PGE);
 	else
-		write_cr4(read_cr4() & ~X86_CR4_PGE);
+		cr4_clear_bits(X86_CR4_PGE);
 }
 
 /*H:020
-- 
1.9.3


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [PATCH v2 3/8] x86: Store a per-cpu shadow copy of CR4
  2014-10-24 22:58 [PATCH v2 0/8] CR4 handling improvements Andy Lutomirski
  2014-10-24 22:58 ` [PATCH v2 1/8] perf: Clean up pmu::event_idx Andy Lutomirski
  2014-10-24 22:58 ` [PATCH v2 2/8] x86: Clean up cr4 manipulation Andy Lutomirski
@ 2014-10-24 22:58 ` Andy Lutomirski
  2014-11-01 19:56   ` Thomas Gleixner
  2015-02-04 14:41   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
  2014-10-24 22:58 ` [PATCH v2 4/8] x86: Add a comment clarifying LDT context switching Andy Lutomirski
                   ` (4 subsequent siblings)
  7 siblings, 2 replies; 31+ messages in thread
From: Andy Lutomirski @ 2014-10-24 22:58 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, Andy Lutomirski

Context switches and TLB flushes can change individual bits of CR4.
CR4 reads take several cycles, so store a shadow copy of CR4 in a
per-cpu variable.

To avoid wasting a cache line, I added the CR4 shadow to
cpu_tlbstate, which is already touched in switch_mm.  The heaviest
users of the cr4 shadow will be switch_mm and __switch_to_xtra, and
__switch_to_xtra is called shortly after switch_mm during context
switch, so the cacheline is likely to be hot.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
---
 arch/x86/include/asm/paravirt.h      |  6 ++---
 arch/x86/include/asm/special_insns.h |  6 ++---
 arch/x86/include/asm/tlbflush.h      | 52 +++++++++++++++++++++++++++---------
 arch/x86/include/asm/virtext.h       |  2 +-
 arch/x86/kernel/acpi/sleep.c         |  2 +-
 arch/x86/kernel/cpu/common.c         |  7 +++++
 arch/x86/kernel/cpu/mtrr/cyrix.c     |  6 ++---
 arch/x86/kernel/cpu/mtrr/generic.c   |  6 ++---
 arch/x86/kernel/head32.c             |  1 +
 arch/x86/kernel/head64.c             |  2 ++
 arch/x86/kernel/process_32.c         |  2 +-
 arch/x86/kernel/process_64.c         |  2 +-
 arch/x86/kernel/setup.c              |  2 +-
 arch/x86/kvm/svm.c                   |  2 +-
 arch/x86/kvm/vmx.c                   |  6 ++---
 arch/x86/mm/fault.c                  |  2 +-
 arch/x86/mm/init.c                   |  8 ++++++
 arch/x86/mm/tlb.c                    |  3 ---
 arch/x86/power/cpu.c                 | 11 +++-----
 arch/x86/realmode/init.c             |  2 +-
 20 files changed, 84 insertions(+), 46 deletions(-)

diff --git a/arch/x86/include/asm/paravirt.h b/arch/x86/include/asm/paravirt.h
index cd6e1610e29e..a9d76e02301b 100644
--- a/arch/x86/include/asm/paravirt.h
+++ b/arch/x86/include/asm/paravirt.h
@@ -80,16 +80,16 @@ static inline void write_cr3(unsigned long x)
 	PVOP_VCALL1(pv_mmu_ops.write_cr3, x);
 }
 
-static inline unsigned long read_cr4(void)
+static inline unsigned long __read_cr4(void)
 {
 	return PVOP_CALL0(unsigned long, pv_cpu_ops.read_cr4);
 }
-static inline unsigned long read_cr4_safe(void)
+static inline unsigned long __read_cr4_safe(void)
 {
 	return PVOP_CALL0(unsigned long, pv_cpu_ops.read_cr4_safe);
 }
 
-static inline void write_cr4(unsigned long x)
+static inline void __write_cr4(unsigned long x)
 {
 	PVOP_VCALL1(pv_cpu_ops.write_cr4, x);
 }
diff --git a/arch/x86/include/asm/special_insns.h b/arch/x86/include/asm/special_insns.h
index e820c080a4e9..6a4b00fafb00 100644
--- a/arch/x86/include/asm/special_insns.h
+++ b/arch/x86/include/asm/special_insns.h
@@ -137,17 +137,17 @@ static inline void write_cr3(unsigned long x)
 	native_write_cr3(x);
 }
 
-static inline unsigned long read_cr4(void)
+static inline unsigned long __read_cr4(void)
 {
 	return native_read_cr4();
 }
 
-static inline unsigned long read_cr4_safe(void)
+static inline unsigned long __read_cr4_safe(void)
 {
 	return native_read_cr4_safe();
 }
 
-static inline void write_cr4(unsigned long x)
+static inline void __write_cr4(unsigned long x)
 {
 	native_write_cr4(x);
 }
diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h
index fc0c4bc356ce..cd791948b286 100644
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -15,14 +15,37 @@
 #define __flush_tlb_single(addr) __native_flush_tlb_single(addr)
 #endif
 
+struct tlb_state {
+#ifdef CONFIG_SMP
+	struct mm_struct *active_mm;
+	int state;
+#endif
+
+	/*
+	 * Access to this CR4 shadow and to H/W CR4 is protected by
+	 * disabling interrupts when modifying either one.
+	 */
+	unsigned long cr4;
+};
+DECLARE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate);
+
+/* Initialize cr4 shadow for this CPU. */
+static inline void cr4_init_shadow(void)
+{
+	this_cpu_write(cpu_tlbstate.cr4, __read_cr4());
+}
+
 /* Set in this cpu's CR4. */
 static inline void cr4_set_bits(unsigned long mask)
 {
 	unsigned long cr4;
 
-	cr4 = read_cr4();
-	cr4 |= mask;
-	write_cr4(cr4);
+	cr4 = this_cpu_read(cpu_tlbstate.cr4);
+	if ((cr4 | mask) != cr4) {
+		cr4 |= mask;
+		this_cpu_write(cpu_tlbstate.cr4, cr4);
+		__write_cr4(cr4);
+	}
 }
 
 /* Clear in this cpu's CR4. */
@@ -30,9 +53,18 @@ static inline void cr4_clear_bits(unsigned long mask)
 {
 	unsigned long cr4;
 
-	cr4 = read_cr4();
-	cr4 &= ~mask;
-	write_cr4(cr4);
+	cr4 = this_cpu_read(cpu_tlbstate.cr4);
+	if ((cr4 & ~mask) != cr4) {
+		cr4 &= ~mask;
+		this_cpu_write(cpu_tlbstate.cr4, cr4);
+		__write_cr4(cr4);
+	}
+}
+
+/* Read the CR4 shadow. */
+static inline unsigned long cr4_read_shadow(void)
+{
+	return this_cpu_read(cpu_tlbstate.cr4);
 }
 
 /*
@@ -61,7 +93,7 @@ static inline void __native_flush_tlb_global_irq_disabled(void)
 {
 	unsigned long cr4;
 
-	cr4 = native_read_cr4();
+	cr4 = this_cpu_read(cpu_tlbstate.cr4);
 	/* clear PGE */
 	native_write_cr4(cr4 & ~X86_CR4_PGE);
 	/* write old PGE again and flush TLBs */
@@ -221,12 +253,6 @@ void native_flush_tlb_others(const struct cpumask *cpumask,
 #define TLBSTATE_OK	1
 #define TLBSTATE_LAZY	2
 
-struct tlb_state {
-	struct mm_struct *active_mm;
-	int state;
-};
-DECLARE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate);
-
 static inline void reset_lazy_tlbstate(void)
 {
 	this_cpu_write(cpu_tlbstate.state, 0);
diff --git a/arch/x86/include/asm/virtext.h b/arch/x86/include/asm/virtext.h
index f41e19ca717b..cce9ee68e335 100644
--- a/arch/x86/include/asm/virtext.h
+++ b/arch/x86/include/asm/virtext.h
@@ -46,7 +46,7 @@ static inline void cpu_vmxoff(void)
 
 static inline int cpu_vmx_enabled(void)
 {
-	return read_cr4() & X86_CR4_VMXE;
+	return __read_cr4() & X86_CR4_VMXE;
 }
 
 /** Disable VMX if it is enabled on the current CPU
diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
index 31368207837c..d1daead5fcdd 100644
--- a/arch/x86/kernel/acpi/sleep.c
+++ b/arch/x86/kernel/acpi/sleep.c
@@ -78,7 +78,7 @@ int x86_acpi_suspend_lowlevel(void)
 
 	header->pmode_cr0 = read_cr0();
 	if (__this_cpu_read(cpu_info.cpuid_level) >= 0) {
-		header->pmode_cr4 = read_cr4();
+		header->pmode_cr4 = __read_cr4();
 		header->pmode_behavior |= (1 << WAKEUP_BEHAVIOR_RESTORE_CR4);
 	}
 	if (!rdmsr_safe(MSR_IA32_MISC_ENABLE,
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 847d62b78431..1b008f8e5219 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -19,6 +19,7 @@
 #include <asm/archrandom.h>
 #include <asm/hypervisor.h>
 #include <asm/processor.h>
+#include <asm/tlbflush.h>
 #include <asm/debugreg.h>
 #include <asm/sections.h>
 #include <asm/vsyscall.h>
@@ -1302,6 +1303,12 @@ void cpu_init(void)
 	wait_for_master_cpu(cpu);
 
 	/*
+	 * Initialize the CR4 shadow before doing anything that could
+	 * try to read it.
+	 */
+	cr4_init_shadow();
+
+	/*
 	 * Load microcode on this cpu if a valid microcode is available.
 	 * This is early microcode loading procedure.
 	 */
diff --git a/arch/x86/kernel/cpu/mtrr/cyrix.c b/arch/x86/kernel/cpu/mtrr/cyrix.c
index 9e451b0876b5..f8c81ba0b465 100644
--- a/arch/x86/kernel/cpu/mtrr/cyrix.c
+++ b/arch/x86/kernel/cpu/mtrr/cyrix.c
@@ -138,8 +138,8 @@ static void prepare_set(void)
 
 	/*  Save value of CR4 and clear Page Global Enable (bit 7)  */
 	if (cpu_has_pge) {
-		cr4 = read_cr4();
-		write_cr4(cr4 & ~X86_CR4_PGE);
+		cr4 = __read_cr4();
+		__write_cr4(cr4 & ~X86_CR4_PGE);
 	}
 
 	/*
@@ -171,7 +171,7 @@ static void post_set(void)
 
 	/* Restore value of CR4 */
 	if (cpu_has_pge)
-		write_cr4(cr4);
+		__write_cr4(cr4);
 }
 
 static void cyrix_set_arr(unsigned int reg, unsigned long base,
diff --git a/arch/x86/kernel/cpu/mtrr/generic.c b/arch/x86/kernel/cpu/mtrr/generic.c
index 0e25a1bc5ab5..7d74f7b3c6ba 100644
--- a/arch/x86/kernel/cpu/mtrr/generic.c
+++ b/arch/x86/kernel/cpu/mtrr/generic.c
@@ -678,8 +678,8 @@ static void prepare_set(void) __acquires(set_atomicity_lock)
 
 	/* Save value of CR4 and clear Page Global Enable (bit 7) */
 	if (cpu_has_pge) {
-		cr4 = read_cr4();
-		write_cr4(cr4 & ~X86_CR4_PGE);
+		cr4 = __read_cr4();
+		__write_cr4(cr4 & ~X86_CR4_PGE);
 	}
 
 	/* Flush all TLBs via a mov %cr3, %reg; mov %reg, %cr3 */
@@ -708,7 +708,7 @@ static void post_set(void) __releases(set_atomicity_lock)
 
 	/* Restore value of CR4 */
 	if (cpu_has_pge)
-		write_cr4(cr4);
+		__write_cr4(cr4);
 	raw_spin_unlock(&set_atomicity_lock);
 }
 
diff --git a/arch/x86/kernel/head32.c b/arch/x86/kernel/head32.c
index d6c1b9836995..2911ef3a9f1c 100644
--- a/arch/x86/kernel/head32.c
+++ b/arch/x86/kernel/head32.c
@@ -31,6 +31,7 @@ static void __init i386_default_early_setup(void)
 
 asmlinkage __visible void __init i386_start_kernel(void)
 {
+	cr4_init_shadow();
 	sanitize_boot_params(&boot_params);
 
 	/* Call the subarch specific early setup function */
diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c
index eda1a865641e..3b241f0ca005 100644
--- a/arch/x86/kernel/head64.c
+++ b/arch/x86/kernel/head64.c
@@ -155,6 +155,8 @@ asmlinkage __visible void __init x86_64_start_kernel(char * real_mode_data)
 				(__START_KERNEL & PGDIR_MASK)));
 	BUILD_BUG_ON(__fix_to_virt(__end_of_fixed_addresses) <= MODULES_END);
 
+	cr4_init_shadow();
+
 	/* Kill off the identity-map trampoline */
 	reset_early_page_tables();
 
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
index 8f3ebfe710d0..603c4f99cb5a 100644
--- a/arch/x86/kernel/process_32.c
+++ b/arch/x86/kernel/process_32.c
@@ -101,7 +101,7 @@ void __show_regs(struct pt_regs *regs, int all)
 	cr0 = read_cr0();
 	cr2 = read_cr2();
 	cr3 = read_cr3();
-	cr4 = read_cr4_safe();
+	cr4 = __read_cr4_safe();
 	printk(KERN_DEFAULT "CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
 			cr0, cr2, cr3, cr4);
 
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 3ed4a68d4013..7e2810252d88 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -93,7 +93,7 @@ void __show_regs(struct pt_regs *regs, int all)
 	cr0 = read_cr0();
 	cr2 = read_cr2();
 	cr3 = read_cr3();
-	cr4 = read_cr4();
+	cr4 = __read_cr4();
 
 	printk(KERN_DEFAULT "FS:  %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n",
 	       fs, fsindex, gs, gsindex, shadowgs);
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index 235cfd39e0d7..2d7cab9adaf6 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -1176,7 +1176,7 @@ void __init setup_arch(char **cmdline_p)
 
 	if (boot_cpu_data.cpuid_level >= 0) {
 		/* A CPU has %cr4 if and only if it has CPUID */
-		mmu_cr4_features = read_cr4();
+		mmu_cr4_features = __read_cr4();
 		if (trampoline_cr4_features)
 			*trampoline_cr4_features = mmu_cr4_features;
 	}
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index 65510f624dfe..42c3a90281ee 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -1581,7 +1581,7 @@ static void svm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 
 static int svm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 {
-	unsigned long host_cr4_mce = read_cr4() & X86_CR4_MCE;
+	unsigned long host_cr4_mce = cr4_read_shadow() & X86_CR4_MCE;
 	unsigned long old_cr4 = to_svm(vcpu)->vmcb->save.cr4;
 
 	if (cr4 & X86_CR4_VMXE)
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 93e2649fb786..de464ddcff20 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -2738,7 +2738,7 @@ static int hardware_enable(void)
 	u64 phys_addr = __pa(per_cpu(vmxarea, cpu));
 	u64 old, test_bits;
 
-	if (read_cr4() & X86_CR4_VMXE)
+	if (cr4_read_shadow() & X86_CR4_VMXE)
 		return -EBUSY;
 
 	INIT_LIST_HEAD(&per_cpu(loaded_vmcss_on_cpu, cpu));
@@ -4274,7 +4274,7 @@ static void vmx_set_constant_host_state(struct vcpu_vmx *vmx)
 	vmcs_writel(HOST_CR3, read_cr3());  /* 22.2.3  FIXME: shadow tables */
 
 	/* Save the most likely value for this task's CR4 in the VMCS. */
-	cr4 = read_cr4();
+	cr4 = cr4_read_shadow();
 	vmcs_writel(HOST_CR4, cr4);			/* 22.2.3, 22.2.5 */
 	vmx->host_state.vmcs_host_cr4 = cr4;
 
@@ -7546,7 +7546,7 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
 	if (test_bit(VCPU_REGS_RIP, (unsigned long *)&vcpu->arch.regs_dirty))
 		vmcs_writel(GUEST_RIP, vcpu->arch.regs[VCPU_REGS_RIP]);
 
-	cr4 = read_cr4();
+	cr4 = cr4_read_shadow();
 	if (unlikely(cr4 != vmx->host_state.vmcs_host_cr4)) {
 		vmcs_writel(HOST_CR4, cr4);
 		vmx->host_state.vmcs_host_cr4 = cr4;
diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index d973e61e450d..5239ac2e189a 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -600,7 +600,7 @@ show_fault_oops(struct pt_regs *regs, unsigned long error_code,
 			printk(nx_warning, from_kuid(&init_user_ns, current_uid()));
 		if (pte && pte_present(*pte) && pte_exec(*pte) &&
 				(pgd_flags(*pgd) & _PAGE_USER) &&
-				(read_cr4() & X86_CR4_SMEP))
+				(__read_cr4() & X86_CR4_SMEP))
 			printk(smep_warning, from_kuid(&init_user_ns, current_uid()));
 	}
 
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index 73b37cd0d253..0a59a63bcdad 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -687,3 +687,11 @@ void __init zone_sizes_init(void)
 	free_area_init_nodes(max_zone_pfns);
 }
 
+DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) = {
+#ifdef CONFIG_SMP
+	.active_mm = &init_mm,
+	.state = 0,
+#endif
+	.cr4 = ~0UL,	/* fail hard if we screw up cr4 shadow initialization */
+};
+EXPORT_SYMBOL_GPL(cpu_tlbstate);
diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c
index ee61c36d64f8..3250f2371aea 100644
--- a/arch/x86/mm/tlb.c
+++ b/arch/x86/mm/tlb.c
@@ -14,9 +14,6 @@
 #include <asm/uv/uv.h>
 #include <linux/debugfs.h>
 
-DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate)
-			= { &init_mm, 0, };
-
 /*
  *	Smarter SMP flushing macros.
  *		c/o Linus Torvalds.
diff --git a/arch/x86/power/cpu.c b/arch/x86/power/cpu.c
index 6ec7910f59bf..3e32ed5648a0 100644
--- a/arch/x86/power/cpu.c
+++ b/arch/x86/power/cpu.c
@@ -105,11 +105,8 @@ static void __save_processor_state(struct saved_context *ctxt)
 	ctxt->cr0 = read_cr0();
 	ctxt->cr2 = read_cr2();
 	ctxt->cr3 = read_cr3();
-#ifdef CONFIG_X86_32
-	ctxt->cr4 = read_cr4_safe();
-#else
-/* CONFIG_X86_64 */
-	ctxt->cr4 = read_cr4();
+	ctxt->cr4 = __read_cr4_safe();
+#ifdef CONFIG_X86_64
 	ctxt->cr8 = read_cr8();
 #endif
 	ctxt->misc_enable_saved = !rdmsrl_safe(MSR_IA32_MISC_ENABLE,
@@ -175,12 +172,12 @@ static void notrace __restore_processor_state(struct saved_context *ctxt)
 	/* cr4 was introduced in the Pentium CPU */
 #ifdef CONFIG_X86_32
 	if (ctxt->cr4)
-		write_cr4(ctxt->cr4);
+		__write_cr4(ctxt->cr4);
 #else
 /* CONFIG X86_64 */
 	wrmsrl(MSR_EFER, ctxt->efer);
 	write_cr8(ctxt->cr8);
-	write_cr4(ctxt->cr4);
+	__write_cr4(ctxt->cr4);
 #endif
 	write_cr3(ctxt->cr3);
 	write_cr2(ctxt->cr2);
diff --git a/arch/x86/realmode/init.c b/arch/x86/realmode/init.c
index bad628a620c4..0b7a63d98440 100644
--- a/arch/x86/realmode/init.c
+++ b/arch/x86/realmode/init.c
@@ -81,7 +81,7 @@ void __init setup_real_mode(void)
 
 	trampoline_header->start = (u64) secondary_startup_64;
 	trampoline_cr4_features = &trampoline_header->cr4;
-	*trampoline_cr4_features = read_cr4();
+	*trampoline_cr4_features = __read_cr4();
 
 	trampoline_pgd = (u64 *) __va(real_mode_header->trampoline_pgd);
 	trampoline_pgd[0] = init_level4_pgt[pgd_index(__PAGE_OFFSET)].pgd;
-- 
1.9.3


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [PATCH v2 4/8] x86: Add a comment clarifying LDT context switching
  2014-10-24 22:58 [PATCH v2 0/8] CR4 handling improvements Andy Lutomirski
                   ` (2 preceding siblings ...)
  2014-10-24 22:58 ` [PATCH v2 3/8] x86: Store a per-cpu shadow copy of CR4 Andy Lutomirski
@ 2014-10-24 22:58 ` Andy Lutomirski
  2014-11-01 19:56   ` Thomas Gleixner
  2015-02-04 14:41   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
  2014-10-24 22:58 ` [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping Andy Lutomirski
                   ` (3 subsequent siblings)
  7 siblings, 2 replies; 31+ messages in thread
From: Andy Lutomirski @ 2014-10-24 22:58 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, Andy Lutomirski

The code is correct, but only for a rather subtle reason.  This
confused me for quite a while when I read switch_mm, so clarify the
code to avoid confusing other people, too.

TBH, I wouldn't be surprised if this code was only correct by
accident.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
---
 arch/x86/include/asm/mmu_context.h | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h
index 166af2a8e865..23697f74b372 100644
--- a/arch/x86/include/asm/mmu_context.h
+++ b/arch/x86/include/asm/mmu_context.h
@@ -53,7 +53,18 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 		/* Stop flush ipis for the previous mm */
 		cpumask_clear_cpu(cpu, mm_cpumask(prev));
 
-		/* Load the LDT, if the LDT is different: */
+		/*
+		 * Load the LDT, if the LDT is different.
+		 *
+		 * It's possible that prev->context.ldt doesn't match
+		 * the LDT register.  This can happen if leave_mm(prev)
+		 * was called and then modify_ldt changed
+		 * prev->context.ldt but suppressed an IPI to this CPU.
+		 * In this case, prev->context.ldt != NULL, because we
+		 * never free an LDT while the mm still exists.  That
+		 * means that next->context.ldt != prev->context.ldt,
+		 * because mms never share an LDT.
+		 */
 		if (unlikely(prev->context.ldt != next->context.ldt))
 			load_LDT_nolock(&next->context);
 	}
-- 
1.9.3


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping
  2014-10-24 22:58 [PATCH v2 0/8] CR4 handling improvements Andy Lutomirski
                   ` (3 preceding siblings ...)
  2014-10-24 22:58 ` [PATCH v2 4/8] x86: Add a comment clarifying LDT context switching Andy Lutomirski
@ 2014-10-24 22:58 ` Andy Lutomirski
  2014-11-01 19:59   ` Thomas Gleixner
  2015-02-04 14:42   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
  2014-10-24 22:58 ` [PATCH v2 7/8] x86, perf: Only allow rdpmc if a perf_event is mapped Andy Lutomirski
                   ` (2 subsequent siblings)
  7 siblings, 2 replies; 31+ messages in thread
From: Andy Lutomirski @ 2014-10-24 22:58 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, Andy Lutomirski

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
---
 include/linux/perf_event.h | 7 +++++++
 kernel/events/core.c       | 9 +++++++++
 2 files changed, 16 insertions(+)

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 893a0d07986f..c93aff25a459 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -207,6 +207,13 @@ struct pmu {
 	 */
 	int (*event_init)		(struct perf_event *event);
 
+	/*
+	 * Notification that the event was mapped or unmapped.  Called
+	 * in the context of the mapping task.
+	 */
+	void (*event_mapped)		(struct perf_event *event); /*optional*/
+	void (*event_unmapped)		(struct perf_event *event); /*optional*/
+
 #define PERF_EF_START	0x01		/* start the counter when adding    */
 #define PERF_EF_RELOAD	0x02		/* reload the counter when starting */
 #define PERF_EF_UPDATE	0x04		/* update the counter when stopping */
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 2b02c9fda790..3b1ec69d189f 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -4139,6 +4139,9 @@ static void perf_mmap_open(struct vm_area_struct *vma)
 
 	atomic_inc(&event->mmap_count);
 	atomic_inc(&event->rb->mmap_count);
+
+	if (event->pmu->event_mapped)
+		event->pmu->event_mapped(event);
 }
 
 /*
@@ -4158,6 +4161,9 @@ static void perf_mmap_close(struct vm_area_struct *vma)
 	int mmap_locked = rb->mmap_locked;
 	unsigned long size = perf_data_size(rb);
 
+	if (event->pmu->event_unmapped)
+		event->pmu->event_unmapped(event);
+
 	atomic_dec(&rb->mmap_count);
 
 	if (!atomic_dec_and_mutex_lock(&event->mmap_count, &event->mmap_mutex))
@@ -4359,6 +4365,9 @@ unlock:
 	vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND | VM_DONTDUMP;
 	vma->vm_ops = &perf_mmap_vmops;
 
+	if (event->pmu->event_mapped)
+		event->pmu->event_mapped(event);
+
 	return ret;
 }
 
-- 
1.9.3


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [PATCH v2 7/8] x86, perf: Only allow rdpmc if a perf_event is mapped
  2014-10-24 22:58 [PATCH v2 0/8] CR4 handling improvements Andy Lutomirski
                   ` (4 preceding siblings ...)
  2014-10-24 22:58 ` [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping Andy Lutomirski
@ 2014-10-24 22:58 ` Andy Lutomirski
  2014-10-31 17:54   ` Paolo Bonzini
  2015-02-04 14:42   ` [tip:perf/x86] perf/x86: " tip-bot for Andy Lutomirski
  2014-10-31 15:09 ` [PATCH v2 0/8] CR4 handling improvements Peter Zijlstra
  2014-11-12 23:38 ` Andy Lutomirski
  7 siblings, 2 replies; 31+ messages in thread
From: Andy Lutomirski @ 2014-10-24 22:58 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, Andy Lutomirski

We currently allow any process to use rdpmc.  This significantly
weakens the protection offered by PR_TSC_DISABLED, and it could be
helpful to users attempting to exploit timing attacks.

Since we can't enable access to individual counters, use a very
coarse heuristic to limit access to rdpmc: allow access only when
a perf_event is mmapped.  This protects seccomp sandboxes.

There is plenty of room to further tighen these restrictions.  For
example, this allows rdpmc for any x86_pmu event, but it's only
useful for self-monitoring tasks.

As a side effect, cap_user_rdpmc will now be false for AMD uncore
events.  This isn't a real regression, since .event_idx is disabled
for these events anyway for the time being.  Whenever that gets
re-added, the cap_user_rdpmc code can be adjusted or refactored
accordingly.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
---
 arch/x86/include/asm/mmu.h         |  2 ++
 arch/x86/include/asm/mmu_context.h | 16 +++++++++++
 arch/x86/kernel/cpu/perf_event.c   | 57 +++++++++++++++++++++++++-------------
 arch/x86/kernel/cpu/perf_event.h   |  2 ++
 4 files changed, 58 insertions(+), 19 deletions(-)

diff --git a/arch/x86/include/asm/mmu.h b/arch/x86/include/asm/mmu.h
index 876e74e8eec7..09b9620a73b4 100644
--- a/arch/x86/include/asm/mmu.h
+++ b/arch/x86/include/asm/mmu.h
@@ -19,6 +19,8 @@ typedef struct {
 
 	struct mutex lock;
 	void __user *vdso;
+
+	atomic_t perf_rdpmc_allowed;	/* nonzero if rdpmc is allowed */
 } mm_context_t;
 
 #ifdef CONFIG_SMP
diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h
index 23697f74b372..ccad8d616038 100644
--- a/arch/x86/include/asm/mmu_context.h
+++ b/arch/x86/include/asm/mmu_context.h
@@ -19,6 +19,18 @@ static inline void paravirt_activate_mm(struct mm_struct *prev,
 }
 #endif	/* !CONFIG_PARAVIRT */
 
+#ifdef CONFIG_PERF_EVENTS
+static inline void load_mm_cr4(struct mm_struct *mm)
+{
+	if (atomic_read(&mm->context.perf_rdpmc_allowed))
+		cr4_set_bits(X86_CR4_PCE);
+	else
+		cr4_clear_bits(X86_CR4_PCE);
+}
+#else
+static inline void load_mm_cr4(struct mm_struct *mm) {}
+#endif
+
 /*
  * Used for LDT copy/destruction.
  */
@@ -53,6 +65,9 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 		/* Stop flush ipis for the previous mm */
 		cpumask_clear_cpu(cpu, mm_cpumask(prev));
 
+		/* Load per-mm CR4 state */
+		load_mm_cr4(next);
+
 		/*
 		 * Load the LDT, if the LDT is different.
 		 *
@@ -88,6 +103,7 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 			 */
 			load_cr3(next->pgd);
 			trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL);
+			load_mm_cr4(next);
 			load_LDT_nolock(&next->context);
 		}
 	}
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 00fbab7aa587..3e875b3b30f2 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -31,6 +31,7 @@
 #include <asm/nmi.h>
 #include <asm/smp.h>
 #include <asm/alternative.h>
+#include <asm/mmu_context.h>
 #include <asm/tlbflush.h>
 #include <asm/timer.h>
 #include <asm/desc.h>
@@ -1336,8 +1337,6 @@ x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
 		break;
 
 	case CPU_STARTING:
-		if (x86_pmu.attr_rdpmc)
-			cr4_set_bits(X86_CR4_PCE);
 		if (x86_pmu.cpu_starting)
 			x86_pmu.cpu_starting(cpu);
 		break;
@@ -1813,14 +1812,44 @@ static int x86_pmu_event_init(struct perf_event *event)
 			event->destroy(event);
 	}
 
+	if (ACCESS_ONCE(x86_pmu.attr_rdpmc))
+		event->hw.flags |= PERF_X86_EVENT_RDPMC_ALLOWED;
+
 	return err;
 }
 
+static void refresh_pce(void *ignored)
+{
+	if (current->mm)
+		load_mm_cr4(current->mm);
+}
+
+static void x86_pmu_event_mapped(struct perf_event *event)
+{
+	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
+		return;
+
+	if (atomic_inc_return(&current->mm->context.perf_rdpmc_allowed) == 1)
+		on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
+}
+
+static void x86_pmu_event_unmapped(struct perf_event *event)
+{
+	if (!current->mm)
+		return;
+
+	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
+		return;
+
+	if (atomic_dec_and_test(&current->mm->context.perf_rdpmc_allowed))
+		on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
+}
+
 static int x86_pmu_event_idx(struct perf_event *event)
 {
 	int idx = event->hw.idx;
 
-	if (!x86_pmu.attr_rdpmc)
+	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
 		return 0;
 
 	if (x86_pmu.num_counters_fixed && idx >= INTEL_PMC_IDX_FIXED) {
@@ -1838,16 +1867,6 @@ static ssize_t get_attr_rdpmc(struct device *cdev,
 	return snprintf(buf, 40, "%d\n", x86_pmu.attr_rdpmc);
 }
 
-static void change_rdpmc(void *info)
-{
-	bool enable = !!(unsigned long)info;
-
-	if (enable)
-		cr4_set_bits(X86_CR4_PCE);
-	else
-		cr4_clear_bits(X86_CR4_PCE);
-}
-
 static ssize_t set_attr_rdpmc(struct device *cdev,
 			      struct device_attribute *attr,
 			      const char *buf, size_t count)
@@ -1862,11 +1881,7 @@ static ssize_t set_attr_rdpmc(struct device *cdev,
 	if (x86_pmu.attr_rdpmc_broken)
 		return -ENOTSUPP;
 
-	if (!!val != !!x86_pmu.attr_rdpmc) {
-		x86_pmu.attr_rdpmc = !!val;
-		on_each_cpu(change_rdpmc, (void *)val, 1);
-	}
-
+	x86_pmu.attr_rdpmc = !!val;
 	return count;
 }
 
@@ -1909,6 +1924,9 @@ static struct pmu pmu = {
 
 	.event_init		= x86_pmu_event_init,
 
+	.event_mapped		= x86_pmu_event_mapped,
+	.event_unmapped		= x86_pmu_event_unmapped,
+
 	.add			= x86_pmu_add,
 	.del			= x86_pmu_del,
 	.start			= x86_pmu_start,
@@ -1930,7 +1948,8 @@ void arch_perf_update_userpage(struct perf_event *event,
 
 	userpg->cap_user_time = 0;
 	userpg->cap_user_time_zero = 0;
-	userpg->cap_user_rdpmc = x86_pmu.attr_rdpmc;
+	userpg->cap_user_rdpmc =
+		!!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED);
 	userpg->pmc_width = x86_pmu.cntval_bits;
 
 	if (!sched_clock_stable())
diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index d98a34d435d7..f6868186e67b 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -71,6 +71,8 @@ struct event_constraint {
 #define PERF_X86_EVENT_COMMITTED	0x8 /* event passed commit_txn */
 #define PERF_X86_EVENT_PEBS_LD_HSW	0x10 /* haswell style datala, load */
 #define PERF_X86_EVENT_PEBS_NA_HSW	0x20 /* haswell style datala, unknown */
+#define PERF_X86_EVENT_RDPMC_ALLOWED	0x40 /* grant rdpmc permission */
+
 
 struct amd_nb {
 	int nb_id;  /* NorthBridge id */
-- 
1.9.3


^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 0/8] CR4 handling improvements
  2014-10-24 22:58 [PATCH v2 0/8] CR4 handling improvements Andy Lutomirski
                   ` (5 preceding siblings ...)
  2014-10-24 22:58 ` [PATCH v2 7/8] x86, perf: Only allow rdpmc if a perf_event is mapped Andy Lutomirski
@ 2014-10-31 15:09 ` Peter Zijlstra
  2014-10-31 17:09   ` Andy Lutomirski
  2015-01-14  0:52   ` Andy Lutomirski
  2014-11-12 23:38 ` Andy Lutomirski
  7 siblings, 2 replies; 31+ messages in thread
From: Peter Zijlstra @ 2014-10-31 15:09 UTC (permalink / raw)
  To: Andy Lutomirski
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, x86



This looks ok I suppose, although the x86 people should pass verdict on
the first few patches.

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 0/8] CR4 handling improvements
  2014-10-31 15:09 ` [PATCH v2 0/8] CR4 handling improvements Peter Zijlstra
@ 2014-10-31 17:09   ` Andy Lutomirski
  2015-01-14  0:52   ` Andy Lutomirski
  1 sibling, 0 replies; 31+ messages in thread
From: Andy Lutomirski @ 2014-10-31 17:09 UTC (permalink / raw)
  To: Peter Zijlstra, kvm list, Paolo Bonzini, Gleb Natapov
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, X86 ML

On Fri, Oct 31, 2014 at 8:09 AM, Peter Zijlstra <peterz@infradead.org> wrote:
>
>
> This looks ok I suppose, although the x86 people should pass verdict on
> the first few patches.

Adding some KVM people, too.  Patches 2 and 3 affect KVM, and that
code has been buggy in the recent past.  They should mostly fix the
performance regression that the fix for CVE-2014-3690 caused.

For reference, the series is here:

http://thread.gmane.org/gmane.linux.kernel/1813271/focus=1818661

Patch 1 has already been applied.

--Andy

-- 
Andy Lutomirski
AMA Capital Management, LLC

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 7/8] x86, perf: Only allow rdpmc if a perf_event is mapped
  2014-10-24 22:58 ` [PATCH v2 7/8] x86, perf: Only allow rdpmc if a perf_event is mapped Andy Lutomirski
@ 2014-10-31 17:54   ` Paolo Bonzini
  2014-10-31 18:25     ` Andy Lutomirski
  2015-02-04 14:42   ` [tip:perf/x86] perf/x86: " tip-bot for Andy Lutomirski
  1 sibling, 1 reply; 31+ messages in thread
From: Paolo Bonzini @ 2014-10-31 17:54 UTC (permalink / raw)
  To: Andy Lutomirski, Peter Zijlstra
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj

On 25/10/2014 00:58, Andy Lutomirski wrote:
> We currently allow any process to use rdpmc.  This significantly
> weakens the protection offered by PR_TSC_DISABLED, and it could be
> helpful to users attempting to exploit timing attacks.
> 
> Since we can't enable access to individual counters, use a very
> coarse heuristic to limit access to rdpmc: allow access only when
> a perf_event is mmapped.  This protects seccomp sandboxes.
> 
> There is plenty of room to further tighen these restrictions.  For
> example, this allows rdpmc for any x86_pmu event, but it's only
> useful for self-monitoring tasks.
> 
> As a side effect, cap_user_rdpmc will now be false for AMD uncore
> events.  This isn't a real regression, since .event_idx is disabled
> for these events anyway for the time being.  Whenever that gets
> re-added, the cap_user_rdpmc code can be adjusted or refactored
> accordingly.
> 
> Signed-off-by: Andy Lutomirski <luto@amacapital.net>

What's the impact of this if the host doesn't have "x86,kvm,vmx:
Preserve CR4 across VM entry"?

Paolo

> ---
>  arch/x86/include/asm/mmu.h         |  2 ++
>  arch/x86/include/asm/mmu_context.h | 16 +++++++++++
>  arch/x86/kernel/cpu/perf_event.c   | 57 +++++++++++++++++++++++++-------------
>  arch/x86/kernel/cpu/perf_event.h   |  2 ++
>  4 files changed, 58 insertions(+), 19 deletions(-)
> 
> diff --git a/arch/x86/include/asm/mmu.h b/arch/x86/include/asm/mmu.h
> index 876e74e8eec7..09b9620a73b4 100644
> --- a/arch/x86/include/asm/mmu.h
> +++ b/arch/x86/include/asm/mmu.h
> @@ -19,6 +19,8 @@ typedef struct {
>  
>  	struct mutex lock;
>  	void __user *vdso;
> +
> +	atomic_t perf_rdpmc_allowed;	/* nonzero if rdpmc is allowed */
>  } mm_context_t;
>  
>  #ifdef CONFIG_SMP
> diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h
> index 23697f74b372..ccad8d616038 100644
> --- a/arch/x86/include/asm/mmu_context.h
> +++ b/arch/x86/include/asm/mmu_context.h
> @@ -19,6 +19,18 @@ static inline void paravirt_activate_mm(struct mm_struct *prev,
>  }
>  #endif	/* !CONFIG_PARAVIRT */
>  
> +#ifdef CONFIG_PERF_EVENTS
> +static inline void load_mm_cr4(struct mm_struct *mm)
> +{
> +	if (atomic_read(&mm->context.perf_rdpmc_allowed))
> +		cr4_set_bits(X86_CR4_PCE);
> +	else
> +		cr4_clear_bits(X86_CR4_PCE);
> +}
> +#else
> +static inline void load_mm_cr4(struct mm_struct *mm) {}
> +#endif
> +
>  /*
>   * Used for LDT copy/destruction.
>   */
> @@ -53,6 +65,9 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
>  		/* Stop flush ipis for the previous mm */
>  		cpumask_clear_cpu(cpu, mm_cpumask(prev));
>  
> +		/* Load per-mm CR4 state */
> +		load_mm_cr4(next);
> +
>  		/*
>  		 * Load the LDT, if the LDT is different.
>  		 *
> @@ -88,6 +103,7 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
>  			 */
>  			load_cr3(next->pgd);
>  			trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL);
> +			load_mm_cr4(next);
>  			load_LDT_nolock(&next->context);
>  		}
>  	}
> diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
> index 00fbab7aa587..3e875b3b30f2 100644
> --- a/arch/x86/kernel/cpu/perf_event.c
> +++ b/arch/x86/kernel/cpu/perf_event.c
> @@ -31,6 +31,7 @@
>  #include <asm/nmi.h>
>  #include <asm/smp.h>
>  #include <asm/alternative.h>
> +#include <asm/mmu_context.h>
>  #include <asm/tlbflush.h>
>  #include <asm/timer.h>
>  #include <asm/desc.h>
> @@ -1336,8 +1337,6 @@ x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
>  		break;
>  
>  	case CPU_STARTING:
> -		if (x86_pmu.attr_rdpmc)
> -			cr4_set_bits(X86_CR4_PCE);
>  		if (x86_pmu.cpu_starting)
>  			x86_pmu.cpu_starting(cpu);
>  		break;
> @@ -1813,14 +1812,44 @@ static int x86_pmu_event_init(struct perf_event *event)
>  			event->destroy(event);
>  	}
>  
> +	if (ACCESS_ONCE(x86_pmu.attr_rdpmc))
> +		event->hw.flags |= PERF_X86_EVENT_RDPMC_ALLOWED;
> +
>  	return err;
>  }
>  
> +static void refresh_pce(void *ignored)
> +{
> +	if (current->mm)
> +		load_mm_cr4(current->mm);
> +}
> +
> +static void x86_pmu_event_mapped(struct perf_event *event)
> +{
> +	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
> +		return;
> +
> +	if (atomic_inc_return(&current->mm->context.perf_rdpmc_allowed) == 1)
> +		on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
> +}
> +
> +static void x86_pmu_event_unmapped(struct perf_event *event)
> +{
> +	if (!current->mm)
> +		return;
> +
> +	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
> +		return;
> +
> +	if (atomic_dec_and_test(&current->mm->context.perf_rdpmc_allowed))
> +		on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
> +}
> +
>  static int x86_pmu_event_idx(struct perf_event *event)
>  {
>  	int idx = event->hw.idx;
>  
> -	if (!x86_pmu.attr_rdpmc)
> +	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
>  		return 0;
>  
>  	if (x86_pmu.num_counters_fixed && idx >= INTEL_PMC_IDX_FIXED) {
> @@ -1838,16 +1867,6 @@ static ssize_t get_attr_rdpmc(struct device *cdev,
>  	return snprintf(buf, 40, "%d\n", x86_pmu.attr_rdpmc);
>  }
>  
> -static void change_rdpmc(void *info)
> -{
> -	bool enable = !!(unsigned long)info;
> -
> -	if (enable)
> -		cr4_set_bits(X86_CR4_PCE);
> -	else
> -		cr4_clear_bits(X86_CR4_PCE);
> -}
> -
>  static ssize_t set_attr_rdpmc(struct device *cdev,
>  			      struct device_attribute *attr,
>  			      const char *buf, size_t count)
> @@ -1862,11 +1881,7 @@ static ssize_t set_attr_rdpmc(struct device *cdev,
>  	if (x86_pmu.attr_rdpmc_broken)
>  		return -ENOTSUPP;
>  
> -	if (!!val != !!x86_pmu.attr_rdpmc) {
> -		x86_pmu.attr_rdpmc = !!val;
> -		on_each_cpu(change_rdpmc, (void *)val, 1);
> -	}
> -
> +	x86_pmu.attr_rdpmc = !!val;
>  	return count;
>  }
>  
> @@ -1909,6 +1924,9 @@ static struct pmu pmu = {
>  
>  	.event_init		= x86_pmu_event_init,
>  
> +	.event_mapped		= x86_pmu_event_mapped,
> +	.event_unmapped		= x86_pmu_event_unmapped,
> +
>  	.add			= x86_pmu_add,
>  	.del			= x86_pmu_del,
>  	.start			= x86_pmu_start,
> @@ -1930,7 +1948,8 @@ void arch_perf_update_userpage(struct perf_event *event,
>  
>  	userpg->cap_user_time = 0;
>  	userpg->cap_user_time_zero = 0;
> -	userpg->cap_user_rdpmc = x86_pmu.attr_rdpmc;
> +	userpg->cap_user_rdpmc =
> +		!!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED);
>  	userpg->pmc_width = x86_pmu.cntval_bits;
>  
>  	if (!sched_clock_stable())
> diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
> index d98a34d435d7..f6868186e67b 100644
> --- a/arch/x86/kernel/cpu/perf_event.h
> +++ b/arch/x86/kernel/cpu/perf_event.h
> @@ -71,6 +71,8 @@ struct event_constraint {
>  #define PERF_X86_EVENT_COMMITTED	0x8 /* event passed commit_txn */
>  #define PERF_X86_EVENT_PEBS_LD_HSW	0x10 /* haswell style datala, load */
>  #define PERF_X86_EVENT_PEBS_NA_HSW	0x20 /* haswell style datala, unknown */
> +#define PERF_X86_EVENT_RDPMC_ALLOWED	0x40 /* grant rdpmc permission */
> +
>  
>  struct amd_nb {
>  	int nb_id;  /* NorthBridge id */
> 

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 7/8] x86, perf: Only allow rdpmc if a perf_event is mapped
  2014-10-31 17:54   ` Paolo Bonzini
@ 2014-10-31 18:25     ` Andy Lutomirski
  0 siblings, 0 replies; 31+ messages in thread
From: Andy Lutomirski @ 2014-10-31 18:25 UTC (permalink / raw)
  To: Paolo Bonzini
  Cc: Peter Zijlstra, Ingo Molnar, hillf.zj, Vince Weaver,
	linux-kernel, Paul Mackerras, Kees Cook,
	Arnaldo Carvalho de Melo, Andrea Arcangeli, Valdis Kletnieks

On Oct 31, 2014 10:54 AM, "Paolo Bonzini" <pbonzini@redhat.com> wrote:
>
> On 25/10/2014 00:58, Andy Lutomirski wrote:
> > We currently allow any process to use rdpmc.  This significantly
> > weakens the protection offered by PR_TSC_DISABLED, and it could be
> > helpful to users attempting to exploit timing attacks.
> >
> > Since we can't enable access to individual counters, use a very
> > coarse heuristic to limit access to rdpmc: allow access only when
> > a perf_event is mmapped.  This protects seccomp sandboxes.
> >
> > There is plenty of room to further tighen these restrictions.  For
> > example, this allows rdpmc for any x86_pmu event, but it's only
> > useful for self-monitoring tasks.
> >
> > As a side effect, cap_user_rdpmc will now be false for AMD uncore
> > events.  This isn't a real regression, since .event_idx is disabled
> > for these events anyway for the time being.  Whenever that gets
> > re-added, the cap_user_rdpmc code can be adjusted or refactored
> > accordingly.
> >
> > Signed-off-by: Andy Lutomirski <luto@amacapital.net>
>
> What's the impact of this if the host doesn't have "x86,kvm,vmx:
> Preserve CR4 across VM entry"?

Do you mean if these patches are applied in a guest and the host isn't
fixed?  I think it shoudn't make a difference.  As far as I know,
there was never anything wrong with KVM's handling of the guest's cr4
value, at least under VMX.  I haven't looked at SVM much.

If you apply this on the host without the fix, then you'll have some
conflicts, and, if you fix those conflicts, then you'll exacerbate the
VMX bug, because perf_event will trigger it, too.  You'll also have
unprivileged-host-user-triggerable IPIs that change cr4, resulting in
lovely cross-CPU races.  This is why I arranged the fix so that the
entire window during which a host cr4 change would get lost happens
with interrupts off.

This is actually how I found the bug in the first place.  I was trying
to convince myself that these patches weren't racy, so I looked at all
of the cr4-manipulation code I could find, and the VMX code made my
scratch my head. :)

--Andy

>
> Paolo
>
> > ---
> >  arch/x86/include/asm/mmu.h         |  2 ++
> >  arch/x86/include/asm/mmu_context.h | 16 +++++++++++
> >  arch/x86/kernel/cpu/perf_event.c   | 57 +++++++++++++++++++++++++-------------
> >  arch/x86/kernel/cpu/perf_event.h   |  2 ++
> >  4 files changed, 58 insertions(+), 19 deletions(-)
> >
> > diff --git a/arch/x86/include/asm/mmu.h b/arch/x86/include/asm/mmu.h
> > index 876e74e8eec7..09b9620a73b4 100644
> > --- a/arch/x86/include/asm/mmu.h
> > +++ b/arch/x86/include/asm/mmu.h
> > @@ -19,6 +19,8 @@ typedef struct {
> >
> >       struct mutex lock;
> >       void __user *vdso;
> > +
> > +     atomic_t perf_rdpmc_allowed;    /* nonzero if rdpmc is allowed */
> >  } mm_context_t;
> >
> >  #ifdef CONFIG_SMP
> > diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h
> > index 23697f74b372..ccad8d616038 100644
> > --- a/arch/x86/include/asm/mmu_context.h
> > +++ b/arch/x86/include/asm/mmu_context.h
> > @@ -19,6 +19,18 @@ static inline void paravirt_activate_mm(struct mm_struct *prev,
> >  }
> >  #endif       /* !CONFIG_PARAVIRT */
> >
> > +#ifdef CONFIG_PERF_EVENTS
> > +static inline void load_mm_cr4(struct mm_struct *mm)
> > +{
> > +     if (atomic_read(&mm->context.perf_rdpmc_allowed))
> > +             cr4_set_bits(X86_CR4_PCE);
> > +     else
> > +             cr4_clear_bits(X86_CR4_PCE);
> > +}
> > +#else
> > +static inline void load_mm_cr4(struct mm_struct *mm) {}
> > +#endif
> > +
> >  /*
> >   * Used for LDT copy/destruction.
> >   */
> > @@ -53,6 +65,9 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
> >               /* Stop flush ipis for the previous mm */
> >               cpumask_clear_cpu(cpu, mm_cpumask(prev));
> >
> > +             /* Load per-mm CR4 state */
> > +             load_mm_cr4(next);
> > +
> >               /*
> >                * Load the LDT, if the LDT is different.
> >                *
> > @@ -88,6 +103,7 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
> >                        */
> >                       load_cr3(next->pgd);
> >                       trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL);
> > +                     load_mm_cr4(next);
> >                       load_LDT_nolock(&next->context);
> >               }
> >       }
> > diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
> > index 00fbab7aa587..3e875b3b30f2 100644
> > --- a/arch/x86/kernel/cpu/perf_event.c
> > +++ b/arch/x86/kernel/cpu/perf_event.c
> > @@ -31,6 +31,7 @@
> >  #include <asm/nmi.h>
> >  #include <asm/smp.h>
> >  #include <asm/alternative.h>
> > +#include <asm/mmu_context.h>
> >  #include <asm/tlbflush.h>
> >  #include <asm/timer.h>
> >  #include <asm/desc.h>
> > @@ -1336,8 +1337,6 @@ x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
> >               break;
> >
> >       case CPU_STARTING:
> > -             if (x86_pmu.attr_rdpmc)
> > -                     cr4_set_bits(X86_CR4_PCE);
> >               if (x86_pmu.cpu_starting)
> >                       x86_pmu.cpu_starting(cpu);
> >               break;
> > @@ -1813,14 +1812,44 @@ static int x86_pmu_event_init(struct perf_event *event)
> >                       event->destroy(event);
> >       }
> >
> > +     if (ACCESS_ONCE(x86_pmu.attr_rdpmc))
> > +             event->hw.flags |= PERF_X86_EVENT_RDPMC_ALLOWED;
> > +
> >       return err;
> >  }
> >
> > +static void refresh_pce(void *ignored)
> > +{
> > +     if (current->mm)
> > +             load_mm_cr4(current->mm);
> > +}
> > +
> > +static void x86_pmu_event_mapped(struct perf_event *event)
> > +{
> > +     if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
> > +             return;
> > +
> > +     if (atomic_inc_return(&current->mm->context.perf_rdpmc_allowed) == 1)
> > +             on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
> > +}
> > +
> > +static void x86_pmu_event_unmapped(struct perf_event *event)
> > +{
> > +     if (!current->mm)
> > +             return;
> > +
> > +     if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
> > +             return;
> > +
> > +     if (atomic_dec_and_test(&current->mm->context.perf_rdpmc_allowed))
> > +             on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
> > +}
> > +
> >  static int x86_pmu_event_idx(struct perf_event *event)
> >  {
> >       int idx = event->hw.idx;
> >
> > -     if (!x86_pmu.attr_rdpmc)
> > +     if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
> >               return 0;
> >
> >       if (x86_pmu.num_counters_fixed && idx >= INTEL_PMC_IDX_FIXED) {
> > @@ -1838,16 +1867,6 @@ static ssize_t get_attr_rdpmc(struct device *cdev,
> >       return snprintf(buf, 40, "%d\n", x86_pmu.attr_rdpmc);
> >  }
> >
> > -static void change_rdpmc(void *info)
> > -{
> > -     bool enable = !!(unsigned long)info;
> > -
> > -     if (enable)
> > -             cr4_set_bits(X86_CR4_PCE);
> > -     else
> > -             cr4_clear_bits(X86_CR4_PCE);
> > -}
> > -
> >  static ssize_t set_attr_rdpmc(struct device *cdev,
> >                             struct device_attribute *attr,
> >                             const char *buf, size_t count)
> > @@ -1862,11 +1881,7 @@ static ssize_t set_attr_rdpmc(struct device *cdev,
> >       if (x86_pmu.attr_rdpmc_broken)
> >               return -ENOTSUPP;
> >
> > -     if (!!val != !!x86_pmu.attr_rdpmc) {
> > -             x86_pmu.attr_rdpmc = !!val;
> > -             on_each_cpu(change_rdpmc, (void *)val, 1);
> > -     }
> > -
> > +     x86_pmu.attr_rdpmc = !!val;
> >       return count;
> >  }
> >
> > @@ -1909,6 +1924,9 @@ static struct pmu pmu = {
> >
> >       .event_init             = x86_pmu_event_init,
> >
> > +     .event_mapped           = x86_pmu_event_mapped,
> > +     .event_unmapped         = x86_pmu_event_unmapped,
> > +
> >       .add                    = x86_pmu_add,
> >       .del                    = x86_pmu_del,
> >       .start                  = x86_pmu_start,
> > @@ -1930,7 +1948,8 @@ void arch_perf_update_userpage(struct perf_event *event,
> >
> >       userpg->cap_user_time = 0;
> >       userpg->cap_user_time_zero = 0;
> > -     userpg->cap_user_rdpmc = x86_pmu.attr_rdpmc;
> > +     userpg->cap_user_rdpmc =
> > +             !!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED);
> >       userpg->pmc_width = x86_pmu.cntval_bits;
> >
> >       if (!sched_clock_stable())
> > diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
> > index d98a34d435d7..f6868186e67b 100644
> > --- a/arch/x86/kernel/cpu/perf_event.h
> > +++ b/arch/x86/kernel/cpu/perf_event.h
> > @@ -71,6 +71,8 @@ struct event_constraint {
> >  #define PERF_X86_EVENT_COMMITTED     0x8 /* event passed commit_txn */
> >  #define PERF_X86_EVENT_PEBS_LD_HSW   0x10 /* haswell style datala, load */
> >  #define PERF_X86_EVENT_PEBS_NA_HSW   0x20 /* haswell style datala, unknown */
> > +#define PERF_X86_EVENT_RDPMC_ALLOWED 0x40 /* grant rdpmc permission */
> > +
> >
> >  struct amd_nb {
> >       int nb_id;  /* NorthBridge id */
> >

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 2/8] x86: Clean up cr4 manipulation
  2014-10-24 22:58 ` [PATCH v2 2/8] x86: Clean up cr4 manipulation Andy Lutomirski
@ 2014-11-01 19:56   ` Thomas Gleixner
  2015-02-04 14:41   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
  1 sibling, 0 replies; 31+ messages in thread
From: Thomas Gleixner @ 2014-11-01 19:56 UTC (permalink / raw)
  To: Andy Lutomirski
  Cc: Peter Zijlstra, Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, x86

On Fri, 24 Oct 2014, Andy Lutomirski wrote:

> CR4 manipulation was split, seemingly at random, between direct
> (write_cr4) and using a helper (set/clear_in_cr4).  Unfortunately,
> the set_in_cr4 and clear_in_cr4 helpers also poke at the boot code,
> which only a small subset of users actually wanted.
> 
> This patch replaces all cr4 access in functions that don't leave cr4
> exactly the way they found it with new helpers cr4_set_bits,
> cr4_clear_bits, and cr4_set_bits_and_update_boot.
> 
> Signed-off-by: Andy Lutomirski <luto@amacapital.net>

Reviewed-by: Thomas Gleixner <tglx@linutronix.de>


^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 3/8] x86: Store a per-cpu shadow copy of CR4
  2014-10-24 22:58 ` [PATCH v2 3/8] x86: Store a per-cpu shadow copy of CR4 Andy Lutomirski
@ 2014-11-01 19:56   ` Thomas Gleixner
  2015-02-04 14:41   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
  1 sibling, 0 replies; 31+ messages in thread
From: Thomas Gleixner @ 2014-11-01 19:56 UTC (permalink / raw)
  To: Andy Lutomirski
  Cc: Peter Zijlstra, Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, x86

On Fri, 24 Oct 2014, Andy Lutomirski wrote:

> Context switches and TLB flushes can change individual bits of CR4.
> CR4 reads take several cycles, so store a shadow copy of CR4 in a
> per-cpu variable.
> 
> To avoid wasting a cache line, I added the CR4 shadow to
> cpu_tlbstate, which is already touched in switch_mm.  The heaviest
> users of the cr4 shadow will be switch_mm and __switch_to_xtra, and
> __switch_to_xtra is called shortly after switch_mm during context
> switch, so the cacheline is likely to be hot.
> 
> Signed-off-by: Andy Lutomirski <luto@amacapital.net>

Reviewed-by: Thomas Gleixner <tglx@linutronix.de>

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 4/8] x86: Add a comment clarifying LDT context switching
  2014-10-24 22:58 ` [PATCH v2 4/8] x86: Add a comment clarifying LDT context switching Andy Lutomirski
@ 2014-11-01 19:56   ` Thomas Gleixner
  2015-02-04 14:41   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
  1 sibling, 0 replies; 31+ messages in thread
From: Thomas Gleixner @ 2014-11-01 19:56 UTC (permalink / raw)
  To: Andy Lutomirski
  Cc: Peter Zijlstra, Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, x86

On Fri, 24 Oct 2014, Andy Lutomirski wrote:
> The code is correct, but only for a rather subtle reason.  This
> confused me for quite a while when I read switch_mm, so clarify the
> code to avoid confusing other people, too.
> 
> TBH, I wouldn't be surprised if this code was only correct by
> accident.

:)
 
> Signed-off-by: Andy Lutomirski <luto@amacapital.net>

Reviewed-by: Thomas Gleixner <tglx@linutronix.de>

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping
  2014-10-24 22:58 ` [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping Andy Lutomirski
@ 2014-11-01 19:59   ` Thomas Gleixner
  2014-11-01 20:32     ` Andy Lutomirski
  2015-02-04 14:42   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
  1 sibling, 1 reply; 31+ messages in thread
From: Thomas Gleixner @ 2014-11-01 19:59 UTC (permalink / raw)
  To: Andy Lutomirski
  Cc: Peter Zijlstra, Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, x86

On Fri, 24 Oct 2014, Andy Lutomirski wrote:

I'm probably missing something.

Why is this tied to the mmap? If I just open a counter then I should
be able to read the counter from user space w/o mmapping something in
the first place.

Thanks,

	tglx



^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping
  2014-11-01 19:59   ` Thomas Gleixner
@ 2014-11-01 20:32     ` Andy Lutomirski
  2014-11-01 20:39       ` Thomas Gleixner
  0 siblings, 1 reply; 31+ messages in thread
From: Andy Lutomirski @ 2014-11-01 20:32 UTC (permalink / raw)
  To: Thomas Gleixner
  Cc: Peter Zijlstra, Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, X86 ML

On Sat, Nov 1, 2014 at 12:59 PM, Thomas Gleixner <tglx@linutronix.de> wrote:
> On Fri, 24 Oct 2014, Andy Lutomirski wrote:
>
> I'm probably missing something.
>
> Why is this tied to the mmap? If I just open a counter then I should
> be able to read the counter from user space w/o mmapping something in
> the first place.

You can read it with read(2), which this patch shouldn't affect.  If
you want to read it with rdpmc, then you need to know what rdpmc index
to use, and the API for that is to mmap the event, check the
userpage's cap_user_rdpmc, and then rdpmc on ->idx - 1 (assuming that
->idx != 0).  You can't safely make any assumptions about which rdpmc
index it will be without explicitly checking, because perf reserves
the right to change the index whenever it wants.

There's plenty of room to tighten up the restrictions further, but
this is, I think, a decent first step, and it solves the problem of
information leaking into seccomp sandboxes.

Does that help?  I tested this with the only reasonable test case I
could find, which was Andi Kleen's "self" program.  It still worked.

--Andy

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping
  2014-11-01 20:32     ` Andy Lutomirski
@ 2014-11-01 20:39       ` Thomas Gleixner
  2014-11-01 21:49         ` Andy Lutomirski
  0 siblings, 1 reply; 31+ messages in thread
From: Thomas Gleixner @ 2014-11-01 20:39 UTC (permalink / raw)
  To: Andy Lutomirski
  Cc: Peter Zijlstra, Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, X86 ML

On Sat, 1 Nov 2014, Andy Lutomirski wrote:
> On Sat, Nov 1, 2014 at 12:59 PM, Thomas Gleixner <tglx@linutronix.de> wrote:
> > On Fri, 24 Oct 2014, Andy Lutomirski wrote:
> >
> > I'm probably missing something.
> >
> > Why is this tied to the mmap? If I just open a counter then I should
> > be able to read the counter from user space w/o mmapping something in
> > the first place.
> 
> You can read it with read(2), which this patch shouldn't affect.  If
> you want to read it with rdpmc, then you need to know what rdpmc index
> to use, and the API for that is to mmap the event, check the
> userpage's cap_user_rdpmc, and then rdpmc on ->idx - 1 (assuming that
> ->idx != 0).  You can't safely make any assumptions about which rdpmc
> index it will be without explicitly checking, because perf reserves
> the right to change the index whenever it wants.

Got it. As I expected: I was missing something :)
 
> There's plenty of room to tighten up the restrictions further, but
> this is, I think, a decent first step, and it solves the problem of
> information leaking into seccomp sandboxes.

In which way?

Thanks,

	tglx


^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping
  2014-11-01 20:39       ` Thomas Gleixner
@ 2014-11-01 21:49         ` Andy Lutomirski
  2014-11-01 22:10           ` Thomas Gleixner
  0 siblings, 1 reply; 31+ messages in thread
From: Andy Lutomirski @ 2014-11-01 21:49 UTC (permalink / raw)
  To: Thomas Gleixner
  Cc: X86 ML, Peter Zijlstra, Ingo Molnar, hillf.zj, Vince Weaver,
	linux-kernel, Paul Mackerras, Kees Cook,
	Arnaldo Carvalho de Melo, Andrea Arcangeli, Valdis Kletnieks

On Nov 1, 2014 1:39 PM, "Thomas Gleixner" <tglx@linutronix.de> wrote:
>
> On Sat, 1 Nov 2014, Andy Lutomirski wrote:
> > On Sat, Nov 1, 2014 at 12:59 PM, Thomas Gleixner <tglx@linutronix.de> wrote:
> > > On Fri, 24 Oct 2014, Andy Lutomirski wrote:
> > >
> > > I'm probably missing something.
> > >
> > > Why is this tied to the mmap? If I just open a counter then I should
> > > be able to read the counter from user space w/o mmapping something in
> > > the first place.
> >
> > You can read it with read(2), which this patch shouldn't affect.  If
> > you want to read it with rdpmc, then you need to know what rdpmc index
> > to use, and the API for that is to mmap the event, check the
> > userpage's cap_user_rdpmc, and then rdpmc on ->idx - 1 (assuming that
> > ->idx != 0).  You can't safely make any assumptions about which rdpmc
> > index it will be without explicitly checking, because perf reserves
> > the right to change the index whenever it wants.
>
> Got it. As I expected: I was missing something :)
>
> > There's plenty of room to tighten up the restrictions further, but
> > this is, I think, a decent first step, and it solves the problem of
> > information leaking into seccomp sandboxes.
>
> In which way?

All the performance counters were readable without using any syscalls.
That leaks hints as to which events are in use, and it possibly leaks
interesting side channel information.   With this series applied, you
need a at least mmap an rdpmc-able event, which most seccomp sandboxes
won't allow.

Unfortunately, rdpmc access to counters can't be controlled
individually, so it's hard to do all that much better than this.

--Andy

>
> Thanks,
>
>         tglx
>

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping
  2014-11-01 21:49         ` Andy Lutomirski
@ 2014-11-01 22:10           ` Thomas Gleixner
  2014-11-02 20:15             ` Andy Lutomirski
  0 siblings, 1 reply; 31+ messages in thread
From: Thomas Gleixner @ 2014-11-01 22:10 UTC (permalink / raw)
  To: Andy Lutomirski
  Cc: X86 ML, Peter Zijlstra, Ingo Molnar, hillf.zj, Vince Weaver,
	linux-kernel, Paul Mackerras, Kees Cook,
	Arnaldo Carvalho de Melo, Andrea Arcangeli, Valdis Kletnieks

On Sat, 1 Nov 2014, Andy Lutomirski wrote:
> On Nov 1, 2014 1:39 PM, "Thomas Gleixner" <tglx@linutronix.de> wrote:
> > On Sat, 1 Nov 2014, Andy Lutomirski wrote:
> > > There's plenty of room to tighten up the restrictions further, but
> > > this is, I think, a decent first step, and it solves the problem of
> > > information leaking into seccomp sandboxes.
> >
> > In which way?
> 
> All the performance counters were readable without using any syscalls.
> That leaks hints as to which events are in use, and it possibly leaks
> interesting side channel information.   With this series applied, you
> need a at least mmap an rdpmc-able event, which most seccomp sandboxes
> won't allow.

Ok. So you are preventing the seccomp sandboxes to open/mmap a counter.
 
> Unfortunately, rdpmc access to counters can't be controlled
> individually, so it's hard to do all that much better than this.

Yeah, I know ...

Thanks,

	tglx

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping
  2014-11-01 22:10           ` Thomas Gleixner
@ 2014-11-02 20:15             ` Andy Lutomirski
  0 siblings, 0 replies; 31+ messages in thread
From: Andy Lutomirski @ 2014-11-02 20:15 UTC (permalink / raw)
  To: Thomas Gleixner
  Cc: X86 ML, Peter Zijlstra, Ingo Molnar, hillf.zj, Vince Weaver,
	linux-kernel, Paul Mackerras, Kees Cook,
	Arnaldo Carvalho de Melo, Andrea Arcangeli, Valdis Kletnieks

On Sat, Nov 1, 2014 at 3:10 PM, Thomas Gleixner <tglx@linutronix.de> wrote:
> On Sat, 1 Nov 2014, Andy Lutomirski wrote:
>> On Nov 1, 2014 1:39 PM, "Thomas Gleixner" <tglx@linutronix.de> wrote:
>> > On Sat, 1 Nov 2014, Andy Lutomirski wrote:
>> > > There's plenty of room to tighten up the restrictions further, but
>> > > this is, I think, a decent first step, and it solves the problem of
>> > > information leaking into seccomp sandboxes.
>> >
>> > In which way?
>>
>> All the performance counters were readable without using any syscalls.
>> That leaks hints as to which events are in use, and it possibly leaks
>> interesting side channel information.   With this series applied, you
>> need a at least mmap an rdpmc-able event, which most seccomp sandboxes
>> won't allow.
>
> Ok. So you are preventing the seccomp sandboxes to open/mmap a counter.
>

Yes.

Conversely, if someone lets perf_event_open through a seccomp filter,
then the sandboxed code can probably gather more interesting
information using perf_event_open the normal way than they can by
poking at rdpmc.

--Andy

>> Unfortunately, rdpmc access to counters can't be controlled
>> individually, so it's hard to do all that much better than this.
>
> Yeah, I know ...
>
> Thanks,
>
>         tglx



-- 
Andy Lutomirski
AMA Capital Management, LLC

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 0/8] CR4 handling improvements
  2014-10-24 22:58 [PATCH v2 0/8] CR4 handling improvements Andy Lutomirski
                   ` (6 preceding siblings ...)
  2014-10-31 15:09 ` [PATCH v2 0/8] CR4 handling improvements Peter Zijlstra
@ 2014-11-12 23:38 ` Andy Lutomirski
  7 siblings, 0 replies; 31+ messages in thread
From: Andy Lutomirski @ 2014-11-12 23:38 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, Andy Lutomirski

On Fri, Oct 24, 2014 at 3:58 PM, Andy Lutomirski <luto@amacapital.net> wrote:
> This little series tightens up rdpmc permissions.  With it applied,
> rdpmc can only be used if a perf_event is actually mmapped.  For now,
> this is only really useful for seccomp.
>
> At some point this could be further tightened up to only allow rdpmc
> if an actual self-monitoring perf event that is compatible with
> rdpmc is mapped.
>
> This should add <50ns to context switches between rdpmc-capable and
> rdpmc-non-capable mms.  I suspect that this is well under 10%
> overhead, given that perf already adds some context switch latency.
>
> I think that patches 1-3 are a good idea regardless of any rdpmc changes.
>
> AMD Uncore userspace rdpmc is broken by these patches (cap_user_rdpmc
> will be zero), but it was broken anyway.
>
> Changes from v1 (aka RFC):
>  - Rebased on top of the KVM CR4 fix.  This applies to a very recent -linus.
>  - Renamed the cr4 helpers (Peter, Borislav)
>  - Fixed buggy cr4 helpers (Hilf)
>  - Improved lots of comments (everyone)
>  - Renamed read_cr4 and write_cr4 to make sure I didn't miss anything.
>    (NB: This will introduce conflicts with Andi's FSGSBASE work.  This is
>     a good thing.)
>
> Andy Lutomirski (7):
>   x86: Clean up cr4 manipulation
>   x86: Store a per-cpu shadow copy of CR4
>   x86: Add a comment clarifying LDT context switching
>   perf: Add pmu callbacks to track event mapping and unmapping
>   perf: Pass the event to arch_perf_update_userpage
>   x86, perf: Only allow rdpmc if a perf_event is mapped
>   x86, perf: Add /sys/devices/cpu/rdpmc=2 to allow rdpmc for all tasks
>
> Peter Zijlstra (1):
>   perf: Clean up pmu::event_idx
>


What's the status of these?  I think that "perf: Clean up
pmu::event_idx" and "x86: Add a comment clarifying LDT context
switching" are in -tip, the two cr4 cleanups ("x86: Clean up cr4
manipulation" and "x86: Store a per-cpu shadow copy of CR4") are
reviewed but will conflict with Andi's fsgsbase work, and the rest are
waiting for review.

Thanks,
Andy


>  arch/powerpc/perf/hv-24x7.c          |  6 ---
>  arch/powerpc/perf/hv-gpci.c          |  6 ---
>  arch/s390/kernel/perf_cpum_sf.c      |  6 ---
>  arch/x86/include/asm/mmu.h           |  2 +
>  arch/x86/include/asm/mmu_context.h   | 32 ++++++++++++++-
>  arch/x86/include/asm/paravirt.h      |  6 +--
>  arch/x86/include/asm/processor.h     | 33 ----------------
>  arch/x86/include/asm/special_insns.h |  6 +--
>  arch/x86/include/asm/tlbflush.h      | 77 ++++++++++++++++++++++++++++++++----
>  arch/x86/include/asm/virtext.h       |  5 ++-
>  arch/x86/kernel/acpi/sleep.c         |  2 +-
>  arch/x86/kernel/cpu/common.c         | 17 +++++---
>  arch/x86/kernel/cpu/mcheck/mce.c     |  3 +-
>  arch/x86/kernel/cpu/mcheck/p5.c      |  3 +-
>  arch/x86/kernel/cpu/mcheck/winchip.c |  3 +-
>  arch/x86/kernel/cpu/mtrr/cyrix.c     |  6 +--
>  arch/x86/kernel/cpu/mtrr/generic.c   |  6 +--
>  arch/x86/kernel/cpu/perf_event.c     | 76 ++++++++++++++++++++++++++---------
>  arch/x86/kernel/cpu/perf_event.h     |  2 +
>  arch/x86/kernel/head32.c             |  1 +
>  arch/x86/kernel/head64.c             |  2 +
>  arch/x86/kernel/i387.c               |  3 +-
>  arch/x86/kernel/process.c            |  5 ++-
>  arch/x86/kernel/process_32.c         |  2 +-
>  arch/x86/kernel/process_64.c         |  2 +-
>  arch/x86/kernel/setup.c              |  2 +-
>  arch/x86/kernel/xsave.c              |  3 +-
>  arch/x86/kvm/svm.c                   |  2 +-
>  arch/x86/kvm/vmx.c                   | 10 ++---
>  arch/x86/mm/fault.c                  |  2 +-
>  arch/x86/mm/init.c                   | 12 +++++-
>  arch/x86/mm/tlb.c                    |  3 --
>  arch/x86/power/cpu.c                 | 11 ++----
>  arch/x86/realmode/init.c             |  2 +-
>  arch/x86/xen/enlighten.c             |  4 +-
>  drivers/lguest/x86/core.c            |  4 +-
>  include/linux/perf_event.h           |  7 ++++
>  kernel/events/core.c                 | 29 ++++++--------
>  kernel/events/hw_breakpoint.c        |  7 ----
>  39 files changed, 256 insertions(+), 154 deletions(-)
>
> --
> 1.9.3
>



-- 
Andy Lutomirski
AMA Capital Management, LLC

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 0/8] CR4 handling improvements
  2014-10-31 15:09 ` [PATCH v2 0/8] CR4 handling improvements Peter Zijlstra
  2014-10-31 17:09   ` Andy Lutomirski
@ 2015-01-14  0:52   ` Andy Lutomirski
  2015-01-22 22:42     ` Thomas Gleixner
  1 sibling, 1 reply; 31+ messages in thread
From: Andy Lutomirski @ 2015-01-14  0:52 UTC (permalink / raw)
  To: Peter Zijlstra
  Cc: Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, X86 ML

On Fri, Oct 31, 2014 at 8:09 AM, Peter Zijlstra <peterz@infradead.org> wrote:
>
>
> This looks ok I suppose, although the x86 people should pass verdict on
> the first few patches.

Hi all-

What's the status of this series?

Thanks,
Andy

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 0/8] CR4 handling improvements
  2015-01-14  0:52   ` Andy Lutomirski
@ 2015-01-22 22:42     ` Thomas Gleixner
  2015-01-23  8:37       ` Peter Zijlstra
  0 siblings, 1 reply; 31+ messages in thread
From: Thomas Gleixner @ 2015-01-22 22:42 UTC (permalink / raw)
  To: Andy Lutomirski
  Cc: Peter Zijlstra, Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, X86 ML

On Tue, 13 Jan 2015, Andy Lutomirski wrote:
> On Fri, Oct 31, 2014 at 8:09 AM, Peter Zijlstra <peterz@infradead.org> wrote:
> >
> >
> > This looks ok I suppose, although the x86 people should pass verdict on
> > the first few patches.
> 
> Hi all-
> 
> What's the status of this series?

I've reviewed the x86 parts except for the perf stuff. So if Peter is
happy with that he should take it via perf as I expect conflicts there
rather than in the non perf areas

Thanks,

	tglx


^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [PATCH v2 0/8] CR4 handling improvements
  2015-01-22 22:42     ` Thomas Gleixner
@ 2015-01-23  8:37       ` Peter Zijlstra
  0 siblings, 0 replies; 31+ messages in thread
From: Peter Zijlstra @ 2015-01-23  8:37 UTC (permalink / raw)
  To: Thomas Gleixner
  Cc: Andy Lutomirski, Valdis Kletnieks, linux-kernel, Paul Mackerras,
	Arnaldo Carvalho de Melo, Ingo Molnar, Kees Cook,
	Andrea Arcangeli, Vince Weaver, hillf.zj, X86 ML

On Thu, Jan 22, 2015 at 11:42:49PM +0100, Thomas Gleixner wrote:
> On Tue, 13 Jan 2015, Andy Lutomirski wrote:
> > On Fri, Oct 31, 2014 at 8:09 AM, Peter Zijlstra <peterz@infradead.org> wrote:
> > >
> > >
> > > This looks ok I suppose, although the x86 people should pass verdict on
> > > the first few patches.
> > 
> > Hi all-
> > 
> > What's the status of this series?
> 
> I've reviewed the x86 parts except for the perf stuff. So if Peter is
> happy with that he should take it via perf as I expect conflicts there
> rather than in the non perf areas

Ah indeed you did; let me queue the lot then. Thanks!

^ permalink raw reply	[flat|nested] 31+ messages in thread

* [tip:perf/x86] x86: Clean up cr4 manipulation
  2014-10-24 22:58 ` [PATCH v2 2/8] x86: Clean up cr4 manipulation Andy Lutomirski
  2014-11-01 19:56   ` Thomas Gleixner
@ 2015-02-04 14:41   ` " tip-bot for Andy Lutomirski
  1 sibling, 0 replies; 31+ messages in thread
From: tip-bot for Andy Lutomirski @ 2015-02-04 14:41 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: hpa, linux-kernel, luto, keescook, aarcange, paulus,
	Valdis.Kletnieks, hillf.zj, torvalds, vince, mingo, tglx, peterz,
	acme

Commit-ID:  375074cc736ab1d89a708c0a8d7baa4a70d5d476
Gitweb:     http://git.kernel.org/tip/375074cc736ab1d89a708c0a8d7baa4a70d5d476
Author:     Andy Lutomirski <luto@amacapital.net>
AuthorDate: Fri, 24 Oct 2014 15:58:07 -0700
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Wed, 4 Feb 2015 12:10:41 +0100

x86: Clean up cr4 manipulation

CR4 manipulation was split, seemingly at random, between direct
(write_cr4) and using a helper (set/clear_in_cr4).  Unfortunately,
the set_in_cr4 and clear_in_cr4 helpers also poke at the boot code,
which only a small subset of users actually wanted.

This patch replaces all cr4 access in functions that don't leave cr4
exactly the way they found it with new helpers cr4_set_bits,
cr4_clear_bits, and cr4_set_bits_and_update_boot.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/495a10bdc9e67016b8fd3945700d46cfd5c12c2f.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 arch/x86/include/asm/processor.h     | 33 --------------------------------
 arch/x86/include/asm/tlbflush.h      | 37 ++++++++++++++++++++++++++++++++++++
 arch/x86/include/asm/virtext.h       |  3 ++-
 arch/x86/kernel/cpu/common.c         | 10 +++++-----
 arch/x86/kernel/cpu/mcheck/mce.c     |  3 ++-
 arch/x86/kernel/cpu/mcheck/p5.c      |  3 ++-
 arch/x86/kernel/cpu/mcheck/winchip.c |  3 ++-
 arch/x86/kernel/cpu/perf_event.c     |  7 ++++---
 arch/x86/kernel/i387.c               |  3 ++-
 arch/x86/kernel/process.c            |  5 +++--
 arch/x86/kernel/xsave.c              |  3 ++-
 arch/x86/kvm/vmx.c                   |  4 ++--
 arch/x86/mm/init.c                   |  4 ++--
 arch/x86/xen/enlighten.c             |  4 ++--
 drivers/lguest/x86/core.c            |  5 +++--
 15 files changed, 70 insertions(+), 57 deletions(-)

diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index a092a0c..ec1c935 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -579,39 +579,6 @@ static inline void load_sp0(struct tss_struct *tss,
 #define set_iopl_mask native_set_iopl_mask
 #endif /* CONFIG_PARAVIRT */
 
-/*
- * Save the cr4 feature set we're using (ie
- * Pentium 4MB enable and PPro Global page
- * enable), so that any CPU's that boot up
- * after us can get the correct flags.
- */
-extern unsigned long mmu_cr4_features;
-extern u32 *trampoline_cr4_features;
-
-static inline void set_in_cr4(unsigned long mask)
-{
-	unsigned long cr4;
-
-	mmu_cr4_features |= mask;
-	if (trampoline_cr4_features)
-		*trampoline_cr4_features = mmu_cr4_features;
-	cr4 = read_cr4();
-	cr4 |= mask;
-	write_cr4(cr4);
-}
-
-static inline void clear_in_cr4(unsigned long mask)
-{
-	unsigned long cr4;
-
-	mmu_cr4_features &= ~mask;
-	if (trampoline_cr4_features)
-		*trampoline_cr4_features = mmu_cr4_features;
-	cr4 = read_cr4();
-	cr4 &= ~mask;
-	write_cr4(cr4);
-}
-
 typedef struct {
 	unsigned long		seg;
 } mm_segment_t;
diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h
index 04905bf..fc0c4bc 100644
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -15,6 +15,43 @@
 #define __flush_tlb_single(addr) __native_flush_tlb_single(addr)
 #endif
 
+/* Set in this cpu's CR4. */
+static inline void cr4_set_bits(unsigned long mask)
+{
+	unsigned long cr4;
+
+	cr4 = read_cr4();
+	cr4 |= mask;
+	write_cr4(cr4);
+}
+
+/* Clear in this cpu's CR4. */
+static inline void cr4_clear_bits(unsigned long mask)
+{
+	unsigned long cr4;
+
+	cr4 = read_cr4();
+	cr4 &= ~mask;
+	write_cr4(cr4);
+}
+
+/*
+ * Save some of cr4 feature set we're using (e.g.  Pentium 4MB
+ * enable and PPro Global page enable), so that any CPU's that boot
+ * up after us can get the correct flags.  This should only be used
+ * during boot on the boot cpu.
+ */
+extern unsigned long mmu_cr4_features;
+extern u32 *trampoline_cr4_features;
+
+static inline void cr4_set_bits_and_update_boot(unsigned long mask)
+{
+	mmu_cr4_features |= mask;
+	if (trampoline_cr4_features)
+		*trampoline_cr4_features = mmu_cr4_features;
+	cr4_set_bits(mask);
+}
+
 static inline void __native_flush_tlb(void)
 {
 	native_write_cr3(native_read_cr3());
diff --git a/arch/x86/include/asm/virtext.h b/arch/x86/include/asm/virtext.h
index 5da71c2..f41e19c 100644
--- a/arch/x86/include/asm/virtext.h
+++ b/arch/x86/include/asm/virtext.h
@@ -19,6 +19,7 @@
 
 #include <asm/vmx.h>
 #include <asm/svm.h>
+#include <asm/tlbflush.h>
 
 /*
  * VMX functions:
@@ -40,7 +41,7 @@ static inline int cpu_has_vmx(void)
 static inline void cpu_vmxoff(void)
 {
 	asm volatile (ASM_VMX_VMXOFF : : : "cc");
-	write_cr4(read_cr4() & ~X86_CR4_VMXE);
+	cr4_clear_bits(X86_CR4_VMXE);
 }
 
 static inline int cpu_vmx_enabled(void)
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index c604965..9d8fc49 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -278,7 +278,7 @@ __setup("nosmep", setup_disable_smep);
 static __always_inline void setup_smep(struct cpuinfo_x86 *c)
 {
 	if (cpu_has(c, X86_FEATURE_SMEP))
-		set_in_cr4(X86_CR4_SMEP);
+		cr4_set_bits(X86_CR4_SMEP);
 }
 
 static __init int setup_disable_smap(char *arg)
@@ -298,9 +298,9 @@ static __always_inline void setup_smap(struct cpuinfo_x86 *c)
 
 	if (cpu_has(c, X86_FEATURE_SMAP)) {
 #ifdef CONFIG_X86_SMAP
-		set_in_cr4(X86_CR4_SMAP);
+		cr4_set_bits(X86_CR4_SMAP);
 #else
-		clear_in_cr4(X86_CR4_SMAP);
+		cr4_clear_bits(X86_CR4_SMAP);
 #endif
 	}
 }
@@ -1312,7 +1312,7 @@ void cpu_init(void)
 
 	pr_debug("Initializing CPU#%d\n", cpu);
 
-	clear_in_cr4(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
+	cr4_clear_bits(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
 
 	/*
 	 * Initialize the per-CPU GDT with the boot GDT,
@@ -1393,7 +1393,7 @@ void cpu_init(void)
 	printk(KERN_INFO "Initializing CPU#%d\n", cpu);
 
 	if (cpu_feature_enabled(X86_FEATURE_VME) || cpu_has_tsc || cpu_has_de)
-		clear_in_cr4(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
+		cr4_clear_bits(X86_CR4_VME|X86_CR4_PVI|X86_CR4_TSD|X86_CR4_DE);
 
 	load_current_idt();
 	switch_to_new_gdt(cpu);
diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index d231799..15ad3ed 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -44,6 +44,7 @@
 
 #include <asm/processor.h>
 #include <asm/traps.h>
+#include <asm/tlbflush.h>
 #include <asm/mce.h>
 #include <asm/msr.h>
 
@@ -1449,7 +1450,7 @@ static void __mcheck_cpu_init_generic(void)
 	bitmap_fill(all_banks, MAX_NR_BANKS);
 	machine_check_poll(MCP_UC | m_fl, &all_banks);
 
-	set_in_cr4(X86_CR4_MCE);
+	cr4_set_bits(X86_CR4_MCE);
 
 	rdmsrl(MSR_IA32_MCG_CAP, cap);
 	if (cap & MCG_CTL_P)
diff --git a/arch/x86/kernel/cpu/mcheck/p5.c b/arch/x86/kernel/cpu/mcheck/p5.c
index ec2663a..737b0ad 100644
--- a/arch/x86/kernel/cpu/mcheck/p5.c
+++ b/arch/x86/kernel/cpu/mcheck/p5.c
@@ -9,6 +9,7 @@
 
 #include <asm/processor.h>
 #include <asm/traps.h>
+#include <asm/tlbflush.h>
 #include <asm/mce.h>
 #include <asm/msr.h>
 
@@ -65,7 +66,7 @@ void intel_p5_mcheck_init(struct cpuinfo_x86 *c)
 	       "Intel old style machine check architecture supported.\n");
 
 	/* Enable MCE: */
-	set_in_cr4(X86_CR4_MCE);
+	cr4_set_bits(X86_CR4_MCE);
 	printk(KERN_INFO
 	       "Intel old style machine check reporting enabled on CPU#%d.\n",
 	       smp_processor_id());
diff --git a/arch/x86/kernel/cpu/mcheck/winchip.c b/arch/x86/kernel/cpu/mcheck/winchip.c
index bd5d46a..44f1382 100644
--- a/arch/x86/kernel/cpu/mcheck/winchip.c
+++ b/arch/x86/kernel/cpu/mcheck/winchip.c
@@ -8,6 +8,7 @@
 
 #include <asm/processor.h>
 #include <asm/traps.h>
+#include <asm/tlbflush.h>
 #include <asm/mce.h>
 #include <asm/msr.h>
 
@@ -36,7 +37,7 @@ void winchip_mcheck_init(struct cpuinfo_x86 *c)
 	lo &= ~(1<<4);	/* Enable MCE */
 	wrmsr(MSR_IDT_FCR1, lo, hi);
 
-	set_in_cr4(X86_CR4_MCE);
+	cr4_set_bits(X86_CR4_MCE);
 
 	printk(KERN_INFO
 	       "Winchip machine check reporting enabled on CPU#0.\n");
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 143e5f5..6b5acd5 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -31,6 +31,7 @@
 #include <asm/nmi.h>
 #include <asm/smp.h>
 #include <asm/alternative.h>
+#include <asm/tlbflush.h>
 #include <asm/timer.h>
 #include <asm/desc.h>
 #include <asm/ldt.h>
@@ -1328,7 +1329,7 @@ x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
 
 	case CPU_STARTING:
 		if (x86_pmu.attr_rdpmc)
-			set_in_cr4(X86_CR4_PCE);
+			cr4_set_bits(X86_CR4_PCE);
 		if (x86_pmu.cpu_starting)
 			x86_pmu.cpu_starting(cpu);
 		break;
@@ -1834,9 +1835,9 @@ static void change_rdpmc(void *info)
 	bool enable = !!(unsigned long)info;
 
 	if (enable)
-		set_in_cr4(X86_CR4_PCE);
+		cr4_set_bits(X86_CR4_PCE);
 	else
-		clear_in_cr4(X86_CR4_PCE);
+		cr4_clear_bits(X86_CR4_PCE);
 }
 
 static ssize_t set_attr_rdpmc(struct device *cdev,
diff --git a/arch/x86/kernel/i387.c b/arch/x86/kernel/i387.c
index a9a4229..87727b0 100644
--- a/arch/x86/kernel/i387.c
+++ b/arch/x86/kernel/i387.c
@@ -13,6 +13,7 @@
 #include <asm/sigcontext.h>
 #include <asm/processor.h>
 #include <asm/math_emu.h>
+#include <asm/tlbflush.h>
 #include <asm/uaccess.h>
 #include <asm/ptrace.h>
 #include <asm/i387.h>
@@ -180,7 +181,7 @@ void fpu_init(void)
 	if (cpu_has_xmm)
 		cr4_mask |= X86_CR4_OSXMMEXCPT;
 	if (cr4_mask)
-		set_in_cr4(cr4_mask);
+		cr4_set_bits(cr4_mask);
 
 	cr0 = read_cr0();
 	cr0 &= ~(X86_CR0_TS|X86_CR0_EM); /* clear TS and EM */
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index e127dda..046e2d6 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -28,6 +28,7 @@
 #include <asm/fpu-internal.h>
 #include <asm/debugreg.h>
 #include <asm/nmi.h>
+#include <asm/tlbflush.h>
 
 /*
  * per-CPU TSS segments. Threads are completely 'soft' on Linux,
@@ -141,7 +142,7 @@ void flush_thread(void)
 
 static void hard_disable_TSC(void)
 {
-	write_cr4(read_cr4() | X86_CR4_TSD);
+	cr4_set_bits(X86_CR4_TSD);
 }
 
 void disable_TSC(void)
@@ -158,7 +159,7 @@ void disable_TSC(void)
 
 static void hard_enable_TSC(void)
 {
-	write_cr4(read_cr4() & ~X86_CR4_TSD);
+	cr4_clear_bits(X86_CR4_TSD);
 }
 
 static void enable_TSC(void)
diff --git a/arch/x86/kernel/xsave.c b/arch/x86/kernel/xsave.c
index 0de1fae..34f66e5 100644
--- a/arch/x86/kernel/xsave.c
+++ b/arch/x86/kernel/xsave.c
@@ -12,6 +12,7 @@
 #include <asm/i387.h>
 #include <asm/fpu-internal.h>
 #include <asm/sigframe.h>
+#include <asm/tlbflush.h>
 #include <asm/xcr.h>
 
 /*
@@ -453,7 +454,7 @@ static void prepare_fx_sw_frame(void)
  */
 static inline void xstate_enable(void)
 {
-	set_in_cr4(X86_CR4_OSXSAVE);
+	cr4_set_bits(X86_CR4_OSXSAVE);
 	xsetbv(XCR_XFEATURE_ENABLED_MASK, pcntxt_mask);
 }
 
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index d4c58d8..db77537 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -2812,7 +2812,7 @@ static int hardware_enable(void)
 		/* enable and lock */
 		wrmsrl(MSR_IA32_FEATURE_CONTROL, old | test_bits);
 	}
-	write_cr4(read_cr4() | X86_CR4_VMXE); /* FIXME: not cpu hotplug safe */
+	cr4_set_bits(X86_CR4_VMXE);
 
 	if (vmm_exclusive) {
 		kvm_cpu_vmxon(phys_addr);
@@ -2849,7 +2849,7 @@ static void hardware_disable(void)
 		vmclear_local_loaded_vmcss();
 		kvm_cpu_vmxoff();
 	}
-	write_cr4(read_cr4() & ~X86_CR4_VMXE);
+	cr4_clear_bits(X86_CR4_VMXE);
 }
 
 static __init int adjust_vmx_controls(u32 ctl_min, u32 ctl_opt,
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index 079c3b6..d4eddbd 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -173,11 +173,11 @@ static void __init probe_page_size_mask(void)
 
 	/* Enable PSE if available */
 	if (cpu_has_pse)
-		set_in_cr4(X86_CR4_PSE);
+		cr4_set_bits_and_update_boot(X86_CR4_PSE);
 
 	/* Enable PGE if available */
 	if (cpu_has_pge) {
-		set_in_cr4(X86_CR4_PGE);
+		cr4_set_bits_and_update_boot(X86_CR4_PGE);
 		__supported_pte_mask |= _PAGE_GLOBAL;
 	}
 }
diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 78a881b..bd8b845 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -1494,10 +1494,10 @@ static void xen_pvh_set_cr_flags(int cpu)
 	 * set them here. For all, OSFXSR OSXMMEXCPT are set in fpu_init.
 	*/
 	if (cpu_has_pse)
-		set_in_cr4(X86_CR4_PSE);
+		cr4_set_bits_and_update_boot(X86_CR4_PSE);
 
 	if (cpu_has_pge)
-		set_in_cr4(X86_CR4_PGE);
+		cr4_set_bits_and_update_boot(X86_CR4_PGE);
 }
 
 /*
diff --git a/drivers/lguest/x86/core.c b/drivers/lguest/x86/core.c
index 922a1ac..6adfd7b 100644
--- a/drivers/lguest/x86/core.c
+++ b/drivers/lguest/x86/core.c
@@ -47,6 +47,7 @@
 #include <asm/lguest.h>
 #include <asm/uaccess.h>
 #include <asm/i387.h>
+#include <asm/tlbflush.h>
 #include "../lg.h"
 
 static int cpu_had_pge;
@@ -452,9 +453,9 @@ void lguest_arch_handle_trap(struct lg_cpu *cpu)
 static void adjust_pge(void *on)
 {
 	if (on)
-		write_cr4(read_cr4() | X86_CR4_PGE);
+		cr4_set_bits(X86_CR4_PGE);
 	else
-		write_cr4(read_cr4() & ~X86_CR4_PGE);
+		cr4_clear_bits(X86_CR4_PGE);
 }
 
 /*H:020

^ permalink raw reply	[flat|nested] 31+ messages in thread

* [tip:perf/x86] x86: Store a per-cpu shadow copy of CR4
  2014-10-24 22:58 ` [PATCH v2 3/8] x86: Store a per-cpu shadow copy of CR4 Andy Lutomirski
  2014-11-01 19:56   ` Thomas Gleixner
@ 2015-02-04 14:41   ` " tip-bot for Andy Lutomirski
  1 sibling, 0 replies; 31+ messages in thread
From: tip-bot for Andy Lutomirski @ 2015-02-04 14:41 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: torvalds, linux-kernel, paulus, vince, tglx, hpa, peterz, mingo,
	hillf.zj, aarcange, acme, keescook, Valdis.Kletnieks, luto

Commit-ID:  1e02ce4cccdcb9688386e5b8d2c9fa4660b45389
Gitweb:     http://git.kernel.org/tip/1e02ce4cccdcb9688386e5b8d2c9fa4660b45389
Author:     Andy Lutomirski <luto@amacapital.net>
AuthorDate: Fri, 24 Oct 2014 15:58:08 -0700
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Wed, 4 Feb 2015 12:10:42 +0100

x86: Store a per-cpu shadow copy of CR4

Context switches and TLB flushes can change individual bits of CR4.
CR4 reads take several cycles, so store a shadow copy of CR4 in a
per-cpu variable.

To avoid wasting a cache line, I added the CR4 shadow to
cpu_tlbstate, which is already touched in switch_mm.  The heaviest
users of the cr4 shadow will be switch_mm and __switch_to_xtra, and
__switch_to_xtra is called shortly after switch_mm during context
switch, so the cacheline is likely to be hot.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/3a54dd3353fffbf84804398e00dfdc5b7c1afd7d.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 arch/x86/include/asm/paravirt.h      |  6 ++---
 arch/x86/include/asm/special_insns.h |  6 ++---
 arch/x86/include/asm/tlbflush.h      | 52 +++++++++++++++++++++++++++---------
 arch/x86/include/asm/virtext.h       |  2 +-
 arch/x86/kernel/acpi/sleep.c         |  2 +-
 arch/x86/kernel/cpu/common.c         |  7 +++++
 arch/x86/kernel/cpu/mtrr/cyrix.c     |  6 ++---
 arch/x86/kernel/cpu/mtrr/generic.c   |  6 ++---
 arch/x86/kernel/head32.c             |  1 +
 arch/x86/kernel/head64.c             |  2 ++
 arch/x86/kernel/process_32.c         |  2 +-
 arch/x86/kernel/process_64.c         |  2 +-
 arch/x86/kernel/setup.c              |  2 +-
 arch/x86/kvm/svm.c                   |  2 +-
 arch/x86/kvm/vmx.c                   |  6 ++---
 arch/x86/mm/fault.c                  |  2 +-
 arch/x86/mm/init.c                   |  9 +++++++
 arch/x86/mm/tlb.c                    |  3 ---
 arch/x86/power/cpu.c                 | 11 +++-----
 arch/x86/realmode/init.c             |  2 +-
 20 files changed, 85 insertions(+), 46 deletions(-)

diff --git a/arch/x86/include/asm/paravirt.h b/arch/x86/include/asm/paravirt.h
index 32444ae..965c47d 100644
--- a/arch/x86/include/asm/paravirt.h
+++ b/arch/x86/include/asm/paravirt.h
@@ -80,16 +80,16 @@ static inline void write_cr3(unsigned long x)
 	PVOP_VCALL1(pv_mmu_ops.write_cr3, x);
 }
 
-static inline unsigned long read_cr4(void)
+static inline unsigned long __read_cr4(void)
 {
 	return PVOP_CALL0(unsigned long, pv_cpu_ops.read_cr4);
 }
-static inline unsigned long read_cr4_safe(void)
+static inline unsigned long __read_cr4_safe(void)
 {
 	return PVOP_CALL0(unsigned long, pv_cpu_ops.read_cr4_safe);
 }
 
-static inline void write_cr4(unsigned long x)
+static inline void __write_cr4(unsigned long x)
 {
 	PVOP_VCALL1(pv_cpu_ops.write_cr4, x);
 }
diff --git a/arch/x86/include/asm/special_insns.h b/arch/x86/include/asm/special_insns.h
index e820c08..6a4b00f 100644
--- a/arch/x86/include/asm/special_insns.h
+++ b/arch/x86/include/asm/special_insns.h
@@ -137,17 +137,17 @@ static inline void write_cr3(unsigned long x)
 	native_write_cr3(x);
 }
 
-static inline unsigned long read_cr4(void)
+static inline unsigned long __read_cr4(void)
 {
 	return native_read_cr4();
 }
 
-static inline unsigned long read_cr4_safe(void)
+static inline unsigned long __read_cr4_safe(void)
 {
 	return native_read_cr4_safe();
 }
 
-static inline void write_cr4(unsigned long x)
+static inline void __write_cr4(unsigned long x)
 {
 	native_write_cr4(x);
 }
diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h
index fc0c4bc..cd79194 100644
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -15,14 +15,37 @@
 #define __flush_tlb_single(addr) __native_flush_tlb_single(addr)
 #endif
 
+struct tlb_state {
+#ifdef CONFIG_SMP
+	struct mm_struct *active_mm;
+	int state;
+#endif
+
+	/*
+	 * Access to this CR4 shadow and to H/W CR4 is protected by
+	 * disabling interrupts when modifying either one.
+	 */
+	unsigned long cr4;
+};
+DECLARE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate);
+
+/* Initialize cr4 shadow for this CPU. */
+static inline void cr4_init_shadow(void)
+{
+	this_cpu_write(cpu_tlbstate.cr4, __read_cr4());
+}
+
 /* Set in this cpu's CR4. */
 static inline void cr4_set_bits(unsigned long mask)
 {
 	unsigned long cr4;
 
-	cr4 = read_cr4();
-	cr4 |= mask;
-	write_cr4(cr4);
+	cr4 = this_cpu_read(cpu_tlbstate.cr4);
+	if ((cr4 | mask) != cr4) {
+		cr4 |= mask;
+		this_cpu_write(cpu_tlbstate.cr4, cr4);
+		__write_cr4(cr4);
+	}
 }
 
 /* Clear in this cpu's CR4. */
@@ -30,9 +53,18 @@ static inline void cr4_clear_bits(unsigned long mask)
 {
 	unsigned long cr4;
 
-	cr4 = read_cr4();
-	cr4 &= ~mask;
-	write_cr4(cr4);
+	cr4 = this_cpu_read(cpu_tlbstate.cr4);
+	if ((cr4 & ~mask) != cr4) {
+		cr4 &= ~mask;
+		this_cpu_write(cpu_tlbstate.cr4, cr4);
+		__write_cr4(cr4);
+	}
+}
+
+/* Read the CR4 shadow. */
+static inline unsigned long cr4_read_shadow(void)
+{
+	return this_cpu_read(cpu_tlbstate.cr4);
 }
 
 /*
@@ -61,7 +93,7 @@ static inline void __native_flush_tlb_global_irq_disabled(void)
 {
 	unsigned long cr4;
 
-	cr4 = native_read_cr4();
+	cr4 = this_cpu_read(cpu_tlbstate.cr4);
 	/* clear PGE */
 	native_write_cr4(cr4 & ~X86_CR4_PGE);
 	/* write old PGE again and flush TLBs */
@@ -221,12 +253,6 @@ void native_flush_tlb_others(const struct cpumask *cpumask,
 #define TLBSTATE_OK	1
 #define TLBSTATE_LAZY	2
 
-struct tlb_state {
-	struct mm_struct *active_mm;
-	int state;
-};
-DECLARE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate);
-
 static inline void reset_lazy_tlbstate(void)
 {
 	this_cpu_write(cpu_tlbstate.state, 0);
diff --git a/arch/x86/include/asm/virtext.h b/arch/x86/include/asm/virtext.h
index f41e19c..cce9ee6 100644
--- a/arch/x86/include/asm/virtext.h
+++ b/arch/x86/include/asm/virtext.h
@@ -46,7 +46,7 @@ static inline void cpu_vmxoff(void)
 
 static inline int cpu_vmx_enabled(void)
 {
-	return read_cr4() & X86_CR4_VMXE;
+	return __read_cr4() & X86_CR4_VMXE;
 }
 
 /** Disable VMX if it is enabled on the current CPU
diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
index 3136820..d1daead 100644
--- a/arch/x86/kernel/acpi/sleep.c
+++ b/arch/x86/kernel/acpi/sleep.c
@@ -78,7 +78,7 @@ int x86_acpi_suspend_lowlevel(void)
 
 	header->pmode_cr0 = read_cr0();
 	if (__this_cpu_read(cpu_info.cpuid_level) >= 0) {
-		header->pmode_cr4 = read_cr4();
+		header->pmode_cr4 = __read_cr4();
 		header->pmode_behavior |= (1 << WAKEUP_BEHAVIOR_RESTORE_CR4);
 	}
 	if (!rdmsr_safe(MSR_IA32_MISC_ENABLE,
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 9d8fc49..07f2fc3 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -19,6 +19,7 @@
 #include <asm/archrandom.h>
 #include <asm/hypervisor.h>
 #include <asm/processor.h>
+#include <asm/tlbflush.h>
 #include <asm/debugreg.h>
 #include <asm/sections.h>
 #include <asm/vsyscall.h>
@@ -1294,6 +1295,12 @@ void cpu_init(void)
 	wait_for_master_cpu(cpu);
 
 	/*
+	 * Initialize the CR4 shadow before doing anything that could
+	 * try to read it.
+	 */
+	cr4_init_shadow();
+
+	/*
 	 * Load microcode on this cpu if a valid microcode is available.
 	 * This is early microcode loading procedure.
 	 */
diff --git a/arch/x86/kernel/cpu/mtrr/cyrix.c b/arch/x86/kernel/cpu/mtrr/cyrix.c
index 9e451b0..f8c81ba 100644
--- a/arch/x86/kernel/cpu/mtrr/cyrix.c
+++ b/arch/x86/kernel/cpu/mtrr/cyrix.c
@@ -138,8 +138,8 @@ static void prepare_set(void)
 
 	/*  Save value of CR4 and clear Page Global Enable (bit 7)  */
 	if (cpu_has_pge) {
-		cr4 = read_cr4();
-		write_cr4(cr4 & ~X86_CR4_PGE);
+		cr4 = __read_cr4();
+		__write_cr4(cr4 & ~X86_CR4_PGE);
 	}
 
 	/*
@@ -171,7 +171,7 @@ static void post_set(void)
 
 	/* Restore value of CR4 */
 	if (cpu_has_pge)
-		write_cr4(cr4);
+		__write_cr4(cr4);
 }
 
 static void cyrix_set_arr(unsigned int reg, unsigned long base,
diff --git a/arch/x86/kernel/cpu/mtrr/generic.c b/arch/x86/kernel/cpu/mtrr/generic.c
index 0e25a1b..7d74f7b 100644
--- a/arch/x86/kernel/cpu/mtrr/generic.c
+++ b/arch/x86/kernel/cpu/mtrr/generic.c
@@ -678,8 +678,8 @@ static void prepare_set(void) __acquires(set_atomicity_lock)
 
 	/* Save value of CR4 and clear Page Global Enable (bit 7) */
 	if (cpu_has_pge) {
-		cr4 = read_cr4();
-		write_cr4(cr4 & ~X86_CR4_PGE);
+		cr4 = __read_cr4();
+		__write_cr4(cr4 & ~X86_CR4_PGE);
 	}
 
 	/* Flush all TLBs via a mov %cr3, %reg; mov %reg, %cr3 */
@@ -708,7 +708,7 @@ static void post_set(void) __releases(set_atomicity_lock)
 
 	/* Restore value of CR4 */
 	if (cpu_has_pge)
-		write_cr4(cr4);
+		__write_cr4(cr4);
 	raw_spin_unlock(&set_atomicity_lock);
 }
 
diff --git a/arch/x86/kernel/head32.c b/arch/x86/kernel/head32.c
index d6c1b983..2911ef3 100644
--- a/arch/x86/kernel/head32.c
+++ b/arch/x86/kernel/head32.c
@@ -31,6 +31,7 @@ static void __init i386_default_early_setup(void)
 
 asmlinkage __visible void __init i386_start_kernel(void)
 {
+	cr4_init_shadow();
 	sanitize_boot_params(&boot_params);
 
 	/* Call the subarch specific early setup function */
diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c
index eda1a86..3b241f0 100644
--- a/arch/x86/kernel/head64.c
+++ b/arch/x86/kernel/head64.c
@@ -155,6 +155,8 @@ asmlinkage __visible void __init x86_64_start_kernel(char * real_mode_data)
 				(__START_KERNEL & PGDIR_MASK)));
 	BUILD_BUG_ON(__fix_to_virt(__end_of_fixed_addresses) <= MODULES_END);
 
+	cr4_init_shadow();
+
 	/* Kill off the identity-map trampoline */
 	reset_early_page_tables();
 
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
index 8f3ebfe..603c4f9 100644
--- a/arch/x86/kernel/process_32.c
+++ b/arch/x86/kernel/process_32.c
@@ -101,7 +101,7 @@ void __show_regs(struct pt_regs *regs, int all)
 	cr0 = read_cr0();
 	cr2 = read_cr2();
 	cr3 = read_cr3();
-	cr4 = read_cr4_safe();
+	cr4 = __read_cr4_safe();
 	printk(KERN_DEFAULT "CR0: %08lx CR2: %08lx CR3: %08lx CR4: %08lx\n",
 			cr0, cr2, cr3, cr4);
 
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 5a2c029..67fcc43 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -93,7 +93,7 @@ void __show_regs(struct pt_regs *regs, int all)
 	cr0 = read_cr0();
 	cr2 = read_cr2();
 	cr3 = read_cr3();
-	cr4 = read_cr4();
+	cr4 = __read_cr4();
 
 	printk(KERN_DEFAULT "FS:  %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n",
 	       fs, fsindex, gs, gsindex, shadowgs);
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index ab4734e..04e6c62 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -1178,7 +1178,7 @@ void __init setup_arch(char **cmdline_p)
 
 	if (boot_cpu_data.cpuid_level >= 0) {
 		/* A CPU has %cr4 if and only if it has CPUID */
-		mmu_cr4_features = read_cr4();
+		mmu_cr4_features = __read_cr4();
 		if (trampoline_cr4_features)
 			*trampoline_cr4_features = mmu_cr4_features;
 	}
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index 41dd038..496a548 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -1583,7 +1583,7 @@ static void svm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 
 static int svm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 {
-	unsigned long host_cr4_mce = read_cr4() & X86_CR4_MCE;
+	unsigned long host_cr4_mce = cr4_read_shadow() & X86_CR4_MCE;
 	unsigned long old_cr4 = to_svm(vcpu)->vmcb->save.cr4;
 
 	if (cr4 & X86_CR4_VMXE)
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index db77537..8dca6cc 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -2785,7 +2785,7 @@ static int hardware_enable(void)
 	u64 phys_addr = __pa(per_cpu(vmxarea, cpu));
 	u64 old, test_bits;
 
-	if (read_cr4() & X86_CR4_VMXE)
+	if (cr4_read_shadow() & X86_CR4_VMXE)
 		return -EBUSY;
 
 	INIT_LIST_HEAD(&per_cpu(loaded_vmcss_on_cpu, cpu));
@@ -4255,7 +4255,7 @@ static void vmx_set_constant_host_state(struct vcpu_vmx *vmx)
 	vmcs_writel(HOST_CR3, read_cr3());  /* 22.2.3  FIXME: shadow tables */
 
 	/* Save the most likely value for this task's CR4 in the VMCS. */
-	cr4 = read_cr4();
+	cr4 = cr4_read_shadow();
 	vmcs_writel(HOST_CR4, cr4);			/* 22.2.3, 22.2.5 */
 	vmx->host_state.vmcs_host_cr4 = cr4;
 
@@ -7784,7 +7784,7 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
 	if (test_bit(VCPU_REGS_RIP, (unsigned long *)&vcpu->arch.regs_dirty))
 		vmcs_writel(GUEST_RIP, vcpu->arch.regs[VCPU_REGS_RIP]);
 
-	cr4 = read_cr4();
+	cr4 = cr4_read_shadow();
 	if (unlikely(cr4 != vmx->host_state.vmcs_host_cr4)) {
 		vmcs_writel(HOST_CR4, cr4);
 		vmx->host_state.vmcs_host_cr4 = cr4;
diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index e3ff27a..ede025f 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -600,7 +600,7 @@ show_fault_oops(struct pt_regs *regs, unsigned long error_code,
 			printk(nx_warning, from_kuid(&init_user_ns, current_uid()));
 		if (pte && pte_present(*pte) && pte_exec(*pte) &&
 				(pgd_flags(*pgd) & _PAGE_USER) &&
-				(read_cr4() & X86_CR4_SMEP))
+				(__read_cr4() & X86_CR4_SMEP))
 			printk(smep_warning, from_kuid(&init_user_ns, current_uid()));
 	}
 
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index d4eddbd..a74aa0f 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -713,6 +713,15 @@ void __init zone_sizes_init(void)
 	free_area_init_nodes(max_zone_pfns);
 }
 
+DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate) = {
+#ifdef CONFIG_SMP
+	.active_mm = &init_mm,
+	.state = 0,
+#endif
+	.cr4 = ~0UL,	/* fail hard if we screw up cr4 shadow initialization */
+};
+EXPORT_SYMBOL_GPL(cpu_tlbstate);
+
 void update_cache_mode_entry(unsigned entry, enum page_cache_mode cache)
 {
 	/* entry 0 MUST be WB (hardwired to speed up translations) */
diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c
index ee61c36..3250f23 100644
--- a/arch/x86/mm/tlb.c
+++ b/arch/x86/mm/tlb.c
@@ -14,9 +14,6 @@
 #include <asm/uv/uv.h>
 #include <linux/debugfs.h>
 
-DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state, cpu_tlbstate)
-			= { &init_mm, 0, };
-
 /*
  *	Smarter SMP flushing macros.
  *		c/o Linus Torvalds.
diff --git a/arch/x86/power/cpu.c b/arch/x86/power/cpu.c
index 6ec7910..3e32ed5 100644
--- a/arch/x86/power/cpu.c
+++ b/arch/x86/power/cpu.c
@@ -105,11 +105,8 @@ static void __save_processor_state(struct saved_context *ctxt)
 	ctxt->cr0 = read_cr0();
 	ctxt->cr2 = read_cr2();
 	ctxt->cr3 = read_cr3();
-#ifdef CONFIG_X86_32
-	ctxt->cr4 = read_cr4_safe();
-#else
-/* CONFIG_X86_64 */
-	ctxt->cr4 = read_cr4();
+	ctxt->cr4 = __read_cr4_safe();
+#ifdef CONFIG_X86_64
 	ctxt->cr8 = read_cr8();
 #endif
 	ctxt->misc_enable_saved = !rdmsrl_safe(MSR_IA32_MISC_ENABLE,
@@ -175,12 +172,12 @@ static void notrace __restore_processor_state(struct saved_context *ctxt)
 	/* cr4 was introduced in the Pentium CPU */
 #ifdef CONFIG_X86_32
 	if (ctxt->cr4)
-		write_cr4(ctxt->cr4);
+		__write_cr4(ctxt->cr4);
 #else
 /* CONFIG X86_64 */
 	wrmsrl(MSR_EFER, ctxt->efer);
 	write_cr8(ctxt->cr8);
-	write_cr4(ctxt->cr4);
+	__write_cr4(ctxt->cr4);
 #endif
 	write_cr3(ctxt->cr3);
 	write_cr2(ctxt->cr2);
diff --git a/arch/x86/realmode/init.c b/arch/x86/realmode/init.c
index bad628a..0b7a63d 100644
--- a/arch/x86/realmode/init.c
+++ b/arch/x86/realmode/init.c
@@ -81,7 +81,7 @@ void __init setup_real_mode(void)
 
 	trampoline_header->start = (u64) secondary_startup_64;
 	trampoline_cr4_features = &trampoline_header->cr4;
-	*trampoline_cr4_features = read_cr4();
+	*trampoline_cr4_features = __read_cr4();
 
 	trampoline_pgd = (u64 *) __va(real_mode_header->trampoline_pgd);
 	trampoline_pgd[0] = init_level4_pgt[pgd_index(__PAGE_OFFSET)].pgd;

^ permalink raw reply	[flat|nested] 31+ messages in thread

* [tip:perf/x86] x86: Add a comment clarifying LDT context switching
  2014-10-24 22:58 ` [PATCH v2 4/8] x86: Add a comment clarifying LDT context switching Andy Lutomirski
  2014-11-01 19:56   ` Thomas Gleixner
@ 2015-02-04 14:41   ` " tip-bot for Andy Lutomirski
  1 sibling, 0 replies; 31+ messages in thread
From: tip-bot for Andy Lutomirski @ 2015-02-04 14:41 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: hpa, hillf.zj, luto, linux-kernel, keescook, tglx, peterz,
	torvalds, aarcange, acme, Valdis.Kletnieks, paulus, mingo, vince

Commit-ID:  22c4bd9fa921c2b1b3f2420d7b9dabbe982f3059
Gitweb:     http://git.kernel.org/tip/22c4bd9fa921c2b1b3f2420d7b9dabbe982f3059
Author:     Andy Lutomirski <luto@amacapital.net>
AuthorDate: Fri, 24 Oct 2014 15:58:09 -0700
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Wed, 4 Feb 2015 12:10:43 +0100

x86: Add a comment clarifying LDT context switching

The code is correct, but only for a rather subtle reason.  This
confused me for quite a while when I read switch_mm, so clarify the
code to avoid confusing other people, too.

TBH, I wouldn't be surprised if this code was only correct by
accident.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/0db86397f968996fb772c443c251415b0b430ddd.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 arch/x86/include/asm/mmu_context.h | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h
index 4b75d59..52c1835 100644
--- a/arch/x86/include/asm/mmu_context.h
+++ b/arch/x86/include/asm/mmu_context.h
@@ -55,12 +55,14 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 		/*
 		 * Load the LDT, if the LDT is different.
 		 *
-		 * It's possible leave_mm(prev) has been called.  If so,
-		 * then prev->context.ldt could be out of sync with the
-		 * LDT descriptor or the LDT register.  This can only happen
-		 * if prev->context.ldt is non-null, since we never free
-		 * an LDT.  But LDTs can't be shared across mms, so
-		 * prev->context.ldt won't be equal to next->context.ldt.
+		 * It's possible that prev->context.ldt doesn't match
+		 * the LDT register.  This can happen if leave_mm(prev)
+		 * was called and then modify_ldt changed
+		 * prev->context.ldt but suppressed an IPI to this CPU.
+		 * In this case, prev->context.ldt != NULL, because we
+		 * never free an LDT while the mm still exists.  That
+		 * means that next->context.ldt != prev->context.ldt,
+		 * because mms never share an LDT.
 		 */
 		if (unlikely(prev->context.ldt != next->context.ldt))
 			load_LDT_nolock(&next->context);

^ permalink raw reply	[flat|nested] 31+ messages in thread

* [tip:perf/x86] perf: Add pmu callbacks to track event mapping and unmapping
  2014-10-24 22:58 ` [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping Andy Lutomirski
  2014-11-01 19:59   ` Thomas Gleixner
@ 2015-02-04 14:42   ` " tip-bot for Andy Lutomirski
  1 sibling, 0 replies; 31+ messages in thread
From: tip-bot for Andy Lutomirski @ 2015-02-04 14:42 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: aarcange, tglx, mingo, hpa, paulus, linux-kernel, torvalds,
	Valdis.Kletnieks, vince, hillf.zj, peterz, luto, acme, keescook

Commit-ID:  1e0fb9ec679c9273a641f1d6f3d25ea47baef2bb
Gitweb:     http://git.kernel.org/tip/1e0fb9ec679c9273a641f1d6f3d25ea47baef2bb
Author:     Andy Lutomirski <luto@amacapital.net>
AuthorDate: Fri, 24 Oct 2014 15:58:10 -0700
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Wed, 4 Feb 2015 12:10:45 +0100

perf: Add pmu callbacks to track event mapping and unmapping

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/266afcba1d1f91ea5501e4e16e94bbbc1a9339b6.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 include/linux/perf_event.h | 7 +++++++
 kernel/events/core.c       | 9 +++++++++
 2 files changed, 16 insertions(+)

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 5cad0e6..3326200 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -202,6 +202,13 @@ struct pmu {
 	 */
 	int (*event_init)		(struct perf_event *event);
 
+	/*
+	 * Notification that the event was mapped or unmapped.  Called
+	 * in the context of the mapping task.
+	 */
+	void (*event_mapped)		(struct perf_event *event); /*optional*/
+	void (*event_unmapped)		(struct perf_event *event); /*optional*/
+
 #define PERF_EF_START	0x01		/* start the counter when adding    */
 #define PERF_EF_RELOAD	0x02		/* reload the counter when starting */
 #define PERF_EF_UPDATE	0x04		/* update the counter when stopping */
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 7f2fbb8..cc14871 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -4293,6 +4293,9 @@ static void perf_mmap_open(struct vm_area_struct *vma)
 
 	atomic_inc(&event->mmap_count);
 	atomic_inc(&event->rb->mmap_count);
+
+	if (event->pmu->event_mapped)
+		event->pmu->event_mapped(event);
 }
 
 /*
@@ -4312,6 +4315,9 @@ static void perf_mmap_close(struct vm_area_struct *vma)
 	int mmap_locked = rb->mmap_locked;
 	unsigned long size = perf_data_size(rb);
 
+	if (event->pmu->event_unmapped)
+		event->pmu->event_unmapped(event);
+
 	atomic_dec(&rb->mmap_count);
 
 	if (!atomic_dec_and_mutex_lock(&event->mmap_count, &event->mmap_mutex))
@@ -4513,6 +4519,9 @@ unlock:
 	vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND | VM_DONTDUMP;
 	vma->vm_ops = &perf_mmap_vmops;
 
+	if (event->pmu->event_mapped)
+		event->pmu->event_mapped(event);
+
 	return ret;
 }
 

^ permalink raw reply	[flat|nested] 31+ messages in thread

* [tip:perf/x86] perf: Pass the event to arch_perf_update_userpage( )
  2014-10-24 22:58 [PATCH v2 6/8] perf: Pass the event to arch_perf_update_userpage Andy Lutomirski
@ 2015-02-04 14:42 ` tip-bot for Andy Lutomirski
  0 siblings, 0 replies; 31+ messages in thread
From: tip-bot for Andy Lutomirski @ 2015-02-04 14:42 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: torvalds, hpa, hillf.zj, peterz, luto, linux-kernel, mingo,
	keescook, tglx, paulus, acme, vince, Valdis.Kletnieks, aarcange

Commit-ID:  c1317ec2b906442930318d9d6e51425c5a69e9cb
Gitweb:     http://git.kernel.org/tip/c1317ec2b906442930318d9d6e51425c5a69e9cb
Author:     Andy Lutomirski <luto@amacapital.net>
AuthorDate: Fri, 24 Oct 2014 15:58:11 -0700
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Wed, 4 Feb 2015 12:10:46 +0100

perf: Pass the event to arch_perf_update_userpage()

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/0fea9a7fac3c1eea86cb0a5954184e74f4213666.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 arch/x86/kernel/cpu/perf_event.c | 3 ++-
 kernel/events/core.c             | 5 +++--
 2 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 6b5acd5..73e84a3 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1915,7 +1915,8 @@ static struct pmu pmu = {
 	.flush_branch_stack	= x86_pmu_flush_branch_stack,
 };
 
-void arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now)
+void arch_perf_update_userpage(struct perf_event *event,
+			       struct perf_event_mmap_page *userpg, u64 now)
 {
 	struct cyc2ns_data *data;
 
diff --git a/kernel/events/core.c b/kernel/events/core.c
index cc14871..13209a9 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -4101,7 +4101,8 @@ unlock:
 	rcu_read_unlock();
 }
 
-void __weak arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now)
+void __weak arch_perf_update_userpage(
+	struct perf_event *event, struct perf_event_mmap_page *userpg, u64 now)
 {
 }
 
@@ -4151,7 +4152,7 @@ void perf_event_update_userpage(struct perf_event *event)
 	userpg->time_running = running +
 			atomic64_read(&event->child_total_time_running);
 
-	arch_perf_update_userpage(userpg, now);
+	arch_perf_update_userpage(event, userpg, now);
 
 	barrier();
 	++userpg->lock;

^ permalink raw reply	[flat|nested] 31+ messages in thread

* [tip:perf/x86] perf/x86: Only allow rdpmc if a perf_event is mapped
  2014-10-24 22:58 ` [PATCH v2 7/8] x86, perf: Only allow rdpmc if a perf_event is mapped Andy Lutomirski
  2014-10-31 17:54   ` Paolo Bonzini
@ 2015-02-04 14:42   ` " tip-bot for Andy Lutomirski
  1 sibling, 0 replies; 31+ messages in thread
From: tip-bot for Andy Lutomirski @ 2015-02-04 14:42 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: peterz, vince, torvalds, luto, keescook, hillf.zj,
	Valdis.Kletnieks, mingo, acme, linux-kernel, tglx, aarcange, hpa

Commit-ID:  7911d3f7af14a614617e38245fedf98a724e46a9
Gitweb:     http://git.kernel.org/tip/7911d3f7af14a614617e38245fedf98a724e46a9
Author:     Andy Lutomirski <luto@amacapital.net>
AuthorDate: Fri, 24 Oct 2014 15:58:12 -0700
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Wed, 4 Feb 2015 12:10:47 +0100

perf/x86: Only allow rdpmc if a perf_event is mapped

We currently allow any process to use rdpmc.  This significantly
weakens the protection offered by PR_TSC_DISABLED, and it could be
helpful to users attempting to exploit timing attacks.

Since we can't enable access to individual counters, use a very
coarse heuristic to limit access to rdpmc: allow access only when
a perf_event is mmapped.  This protects seccomp sandboxes.

There is plenty of room to further tighen these restrictions.  For
example, this allows rdpmc for any x86_pmu event, but it's only
useful for self-monitoring tasks.

As a side effect, cap_user_rdpmc will now be false for AMD uncore
events.  This isn't a real regression, since .event_idx is disabled
for these events anyway for the time being.  Whenever that gets
re-added, the cap_user_rdpmc code can be adjusted or refactored
accordingly.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/a2bdb3cf3a1d70c26980d7c6dddfbaa69f3182bf.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 arch/x86/include/asm/mmu.h         |  2 ++
 arch/x86/include/asm/mmu_context.h | 16 +++++++++++
 arch/x86/kernel/cpu/perf_event.c   | 57 +++++++++++++++++++++++++-------------
 arch/x86/kernel/cpu/perf_event.h   |  2 ++
 4 files changed, 58 insertions(+), 19 deletions(-)

diff --git a/arch/x86/include/asm/mmu.h b/arch/x86/include/asm/mmu.h
index 876e74e..09b9620 100644
--- a/arch/x86/include/asm/mmu.h
+++ b/arch/x86/include/asm/mmu.h
@@ -19,6 +19,8 @@ typedef struct {
 
 	struct mutex lock;
 	void __user *vdso;
+
+	atomic_t perf_rdpmc_allowed;	/* nonzero if rdpmc is allowed */
 } mm_context_t;
 
 #ifdef CONFIG_SMP
diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h
index 52c1835..89c1fec 100644
--- a/arch/x86/include/asm/mmu_context.h
+++ b/arch/x86/include/asm/mmu_context.h
@@ -18,6 +18,18 @@ static inline void paravirt_activate_mm(struct mm_struct *prev,
 }
 #endif	/* !CONFIG_PARAVIRT */
 
+#ifdef CONFIG_PERF_EVENTS
+static inline void load_mm_cr4(struct mm_struct *mm)
+{
+	if (atomic_read(&mm->context.perf_rdpmc_allowed))
+		cr4_set_bits(X86_CR4_PCE);
+	else
+		cr4_clear_bits(X86_CR4_PCE);
+}
+#else
+static inline void load_mm_cr4(struct mm_struct *mm) {}
+#endif
+
 /*
  * Used for LDT copy/destruction.
  */
@@ -52,6 +64,9 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 		/* Stop flush ipis for the previous mm */
 		cpumask_clear_cpu(cpu, mm_cpumask(prev));
 
+		/* Load per-mm CR4 state */
+		load_mm_cr4(next);
+
 		/*
 		 * Load the LDT, if the LDT is different.
 		 *
@@ -87,6 +102,7 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 			 */
 			load_cr3(next->pgd);
 			trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL);
+			load_mm_cr4(next);
 			load_LDT_nolock(&next->context);
 		}
 	}
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 73e84a3..bec5cff 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -31,6 +31,7 @@
 #include <asm/nmi.h>
 #include <asm/smp.h>
 #include <asm/alternative.h>
+#include <asm/mmu_context.h>
 #include <asm/tlbflush.h>
 #include <asm/timer.h>
 #include <asm/desc.h>
@@ -1328,8 +1329,6 @@ x86_pmu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
 		break;
 
 	case CPU_STARTING:
-		if (x86_pmu.attr_rdpmc)
-			cr4_set_bits(X86_CR4_PCE);
 		if (x86_pmu.cpu_starting)
 			x86_pmu.cpu_starting(cpu);
 		break;
@@ -1805,14 +1804,44 @@ static int x86_pmu_event_init(struct perf_event *event)
 			event->destroy(event);
 	}
 
+	if (ACCESS_ONCE(x86_pmu.attr_rdpmc))
+		event->hw.flags |= PERF_X86_EVENT_RDPMC_ALLOWED;
+
 	return err;
 }
 
+static void refresh_pce(void *ignored)
+{
+	if (current->mm)
+		load_mm_cr4(current->mm);
+}
+
+static void x86_pmu_event_mapped(struct perf_event *event)
+{
+	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
+		return;
+
+	if (atomic_inc_return(&current->mm->context.perf_rdpmc_allowed) == 1)
+		on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
+}
+
+static void x86_pmu_event_unmapped(struct perf_event *event)
+{
+	if (!current->mm)
+		return;
+
+	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
+		return;
+
+	if (atomic_dec_and_test(&current->mm->context.perf_rdpmc_allowed))
+		on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
+}
+
 static int x86_pmu_event_idx(struct perf_event *event)
 {
 	int idx = event->hw.idx;
 
-	if (!x86_pmu.attr_rdpmc)
+	if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
 		return 0;
 
 	if (x86_pmu.num_counters_fixed && idx >= INTEL_PMC_IDX_FIXED) {
@@ -1830,16 +1859,6 @@ static ssize_t get_attr_rdpmc(struct device *cdev,
 	return snprintf(buf, 40, "%d\n", x86_pmu.attr_rdpmc);
 }
 
-static void change_rdpmc(void *info)
-{
-	bool enable = !!(unsigned long)info;
-
-	if (enable)
-		cr4_set_bits(X86_CR4_PCE);
-	else
-		cr4_clear_bits(X86_CR4_PCE);
-}
-
 static ssize_t set_attr_rdpmc(struct device *cdev,
 			      struct device_attribute *attr,
 			      const char *buf, size_t count)
@@ -1854,11 +1873,7 @@ static ssize_t set_attr_rdpmc(struct device *cdev,
 	if (x86_pmu.attr_rdpmc_broken)
 		return -ENOTSUPP;
 
-	if (!!val != !!x86_pmu.attr_rdpmc) {
-		x86_pmu.attr_rdpmc = !!val;
-		on_each_cpu(change_rdpmc, (void *)val, 1);
-	}
-
+	x86_pmu.attr_rdpmc = !!val;
 	return count;
 }
 
@@ -1901,6 +1916,9 @@ static struct pmu pmu = {
 
 	.event_init		= x86_pmu_event_init,
 
+	.event_mapped		= x86_pmu_event_mapped,
+	.event_unmapped		= x86_pmu_event_unmapped,
+
 	.add			= x86_pmu_add,
 	.del			= x86_pmu_del,
 	.start			= x86_pmu_start,
@@ -1922,7 +1940,8 @@ void arch_perf_update_userpage(struct perf_event *event,
 
 	userpg->cap_user_time = 0;
 	userpg->cap_user_time_zero = 0;
-	userpg->cap_user_rdpmc = x86_pmu.attr_rdpmc;
+	userpg->cap_user_rdpmc =
+		!!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED);
 	userpg->pmc_width = x86_pmu.cntval_bits;
 
 	if (!sched_clock_stable())
diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index 4e6cdb0..df525d2 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -71,6 +71,8 @@ struct event_constraint {
 #define PERF_X86_EVENT_COMMITTED	0x8 /* event passed commit_txn */
 #define PERF_X86_EVENT_PEBS_LD_HSW	0x10 /* haswell style datala, load */
 #define PERF_X86_EVENT_PEBS_NA_HSW	0x20 /* haswell style datala, unknown */
+#define PERF_X86_EVENT_RDPMC_ALLOWED	0x40 /* grant rdpmc permission */
+
 
 struct amd_nb {
 	int nb_id;  /* NorthBridge id */

^ permalink raw reply	[flat|nested] 31+ messages in thread

* [tip:perf/x86] perf/x86: Add /sys/devices/cpu/rdpmc= 2 to allow rdpmc for all tasks
  2014-10-24 22:58 [PATCH v2 8/8] x86, perf: Add /sys/devices/cpu/rdpmc=2 to allow rdpmc for all tasks Andy Lutomirski
@ 2015-02-04 14:43 ` " tip-bot for Andy Lutomirski
  0 siblings, 0 replies; 31+ messages in thread
From: tip-bot for Andy Lutomirski @ 2015-02-04 14:43 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: mingo, luto, keescook, Valdis.Kletnieks, tglx, hpa, torvalds,
	linux-kernel, paulus, hillf.zj, aarcange, peterz, vince, acme

Commit-ID:  a66734297f78707ce39d756b656bfae861d53f62
Gitweb:     http://git.kernel.org/tip/a66734297f78707ce39d756b656bfae861d53f62
Author:     Andy Lutomirski <luto@amacapital.net>
AuthorDate: Fri, 24 Oct 2014 15:58:13 -0700
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Wed, 4 Feb 2015 12:10:49 +0100

perf/x86: Add /sys/devices/cpu/rdpmc=2 to allow rdpmc for all tasks

While perfmon2 is a sufficiently evil library (it pokes MSRs
directly) that breaking it is fair game, it's still useful, so we
might as well try to support it.  This allows users to write 2 to
/sys/devices/cpu/rdpmc to disable all rdpmc protection so that hack
like perfmon2 can continue to work.

At some point, if perf_event becomes fast enough to replace
perfmon2, then this can go.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/caac3c1c707dcca48ecbc35f4def21495856f479.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 arch/x86/include/asm/mmu_context.h |  5 ++++-
 arch/x86/kernel/cpu/perf_event.c   | 21 ++++++++++++++++++++-
 2 files changed, 24 insertions(+), 2 deletions(-)

diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h
index 89c1fec..883f6b9 100644
--- a/arch/x86/include/asm/mmu_context.h
+++ b/arch/x86/include/asm/mmu_context.h
@@ -19,9 +19,12 @@ static inline void paravirt_activate_mm(struct mm_struct *prev,
 #endif	/* !CONFIG_PARAVIRT */
 
 #ifdef CONFIG_PERF_EVENTS
+extern struct static_key rdpmc_always_available;
+
 static inline void load_mm_cr4(struct mm_struct *mm)
 {
-	if (atomic_read(&mm->context.perf_rdpmc_allowed))
+	if (static_key_true(&rdpmc_always_available) ||
+	    atomic_read(&mm->context.perf_rdpmc_allowed))
 		cr4_set_bits(X86_CR4_PCE);
 	else
 		cr4_clear_bits(X86_CR4_PCE);
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index bec5cff..b71a7f8 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -45,6 +45,8 @@ DEFINE_PER_CPU(struct cpu_hw_events, cpu_hw_events) = {
 	.enabled = 1,
 };
 
+struct static_key rdpmc_always_available = STATIC_KEY_INIT_FALSE;
+
 u64 __read_mostly hw_cache_event_ids
 				[PERF_COUNT_HW_CACHE_MAX]
 				[PERF_COUNT_HW_CACHE_OP_MAX]
@@ -1870,10 +1872,27 @@ static ssize_t set_attr_rdpmc(struct device *cdev,
 	if (ret)
 		return ret;
 
+	if (val > 2)
+		return -EINVAL;
+
 	if (x86_pmu.attr_rdpmc_broken)
 		return -ENOTSUPP;
 
-	x86_pmu.attr_rdpmc = !!val;
+	if ((val == 2) != (x86_pmu.attr_rdpmc == 2)) {
+		/*
+		 * Changing into or out of always available, aka
+		 * perf-event-bypassing mode.  This path is extremely slow,
+		 * but only root can trigger it, so it's okay.
+		 */
+		if (val == 2)
+			static_key_slow_inc(&rdpmc_always_available);
+		else
+			static_key_slow_dec(&rdpmc_always_available);
+		on_each_cpu(refresh_pce, NULL, 1);
+	}
+
+	x86_pmu.attr_rdpmc = val;
+
 	return count;
 }
 

^ permalink raw reply	[flat|nested] 31+ messages in thread

end of thread, back to index

Thread overview: 31+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-10-24 22:58 [PATCH v2 0/8] CR4 handling improvements Andy Lutomirski
2014-10-24 22:58 ` [PATCH v2 1/8] perf: Clean up pmu::event_idx Andy Lutomirski
2014-10-24 22:58 ` [PATCH v2 2/8] x86: Clean up cr4 manipulation Andy Lutomirski
2014-11-01 19:56   ` Thomas Gleixner
2015-02-04 14:41   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
2014-10-24 22:58 ` [PATCH v2 3/8] x86: Store a per-cpu shadow copy of CR4 Andy Lutomirski
2014-11-01 19:56   ` Thomas Gleixner
2015-02-04 14:41   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
2014-10-24 22:58 ` [PATCH v2 4/8] x86: Add a comment clarifying LDT context switching Andy Lutomirski
2014-11-01 19:56   ` Thomas Gleixner
2015-02-04 14:41   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
2014-10-24 22:58 ` [PATCH v2 5/8] perf: Add pmu callbacks to track event mapping and unmapping Andy Lutomirski
2014-11-01 19:59   ` Thomas Gleixner
2014-11-01 20:32     ` Andy Lutomirski
2014-11-01 20:39       ` Thomas Gleixner
2014-11-01 21:49         ` Andy Lutomirski
2014-11-01 22:10           ` Thomas Gleixner
2014-11-02 20:15             ` Andy Lutomirski
2015-02-04 14:42   ` [tip:perf/x86] " tip-bot for Andy Lutomirski
2014-10-24 22:58 ` [PATCH v2 7/8] x86, perf: Only allow rdpmc if a perf_event is mapped Andy Lutomirski
2014-10-31 17:54   ` Paolo Bonzini
2014-10-31 18:25     ` Andy Lutomirski
2015-02-04 14:42   ` [tip:perf/x86] perf/x86: " tip-bot for Andy Lutomirski
2014-10-31 15:09 ` [PATCH v2 0/8] CR4 handling improvements Peter Zijlstra
2014-10-31 17:09   ` Andy Lutomirski
2015-01-14  0:52   ` Andy Lutomirski
2015-01-22 22:42     ` Thomas Gleixner
2015-01-23  8:37       ` Peter Zijlstra
2014-11-12 23:38 ` Andy Lutomirski
2014-10-24 22:58 [PATCH v2 6/8] perf: Pass the event to arch_perf_update_userpage Andy Lutomirski
2015-02-04 14:42 ` [tip:perf/x86] perf: Pass the event to arch_perf_update_userpage( ) tip-bot for Andy Lutomirski
2014-10-24 22:58 [PATCH v2 8/8] x86, perf: Add /sys/devices/cpu/rdpmc=2 to allow rdpmc for all tasks Andy Lutomirski
2015-02-04 14:43 ` [tip:perf/x86] perf/x86: Add /sys/devices/cpu/rdpmc= 2 " tip-bot for Andy Lutomirski

LKML Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/lkml/0 lkml/git/0.git
	git clone --mirror https://lore.kernel.org/lkml/1 lkml/git/1.git
	git clone --mirror https://lore.kernel.org/lkml/2 lkml/git/2.git
	git clone --mirror https://lore.kernel.org/lkml/3 lkml/git/3.git
	git clone --mirror https://lore.kernel.org/lkml/4 lkml/git/4.git
	git clone --mirror https://lore.kernel.org/lkml/5 lkml/git/5.git
	git clone --mirror https://lore.kernel.org/lkml/6 lkml/git/6.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 lkml lkml/ https://lore.kernel.org/lkml \
		linux-kernel@vger.kernel.org linux-kernel@archiver.kernel.org
	public-inbox-index lkml


Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-kernel


AGPL code for this site: git clone https://public-inbox.org/ public-inbox