Skip to contentSkip to conversation

x86/idt: Minor alloc_intr_gate() sanitization

PATCH Applied Vitaly Kuznetsov 3 participants 10 messages Apr 28, 2020 – May 14, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v2 cover x86/idt: Minor alloc_intr_gate() sanitization

This series is a successor of "[PATCH] x86/idt: Keep spurious entries unset in system_vectors".

The original issue I tried to address was that /proc/interrupts output was always containing all possible system vectors, including non allocated ones (e.g. 'Hypervisor callback interrupts' on bare metal). Thomas suggested to expand this cosmetic change to making alloc_intr_gate() __init.

Vitaly Kuznetsov (3):
x86/xen: Split HVM vector callback setup and interrupt gate allocation x86/idt: Annotate alloc_intr_gate() with __init
x86/idt: Keep spurious entries unset in system_vectors

 arch/x86/kernel/idt.c            | 22 ++++++++++++++++++----
 arch/x86/xen/suspend_hvm.c       |  2 +-
 arch/x86/xen/xen-ops.h           |  2 +-
 drivers/xen/events/events_base.c | 28 +++++++++++++++++-----------
 4 files changed, 37 insertions(+), 17 deletions(-)
Signature · 1 line
2.25.3
#1 lore raw

PATCH v2 1/3 x86/xen: Split HVM vector callback setup and interrupt gate allocation

As a preparatory change for making alloc_intr_gate() __init split xen_callback_vector() into callback vector setup via hypercall (xen_setup_callback_vector()) and interrupt gate allocation (xen_alloc_callback_vector()).

xen_setup_callback_vector() is being called twice: on init and upon system resume from xen_hvm_post_suspend(). alloc_intr_gate() only needs to be called once.

---
 arch/x86/xen/suspend_hvm.c       |  2 +-
 arch/x86/xen/xen-ops.h           |  2 +-
 drivers/xen/events/events_base.c | 28 +++++++++++++++++-----------
 3 files changed, 19 insertions(+), 13 deletions(-)
arch/x86/xen/suspend_hvm.c +1 -1
index e666b614cf6d..5152afe16876 100644
@@ -13,6 +13,6 @@ void xen_hvm_post_suspend(int suspend_cancelled)
 		xen_hvm_init_shared_info();
 		xen_vcpu_restore();
 	}
-	xen_callback_vector();
+	xen_setup_callback_vector();
 	xen_unplug_emulated_devices();
 }
arch/x86/xen/xen-ops.h +1 -1
index 45a441c33d6d..1cc1568bfe04 100644
@@ -55,7 +55,7 @@ void xen_enable_sysenter(void);
 void xen_enable_syscall(void);
 void xen_vcpu_restore(void);
 
-void xen_callback_vector(void);
+void xen_setup_callback_vector(void);
 void xen_hvm_init_shared_info(void);
 void xen_unplug_emulated_devices(void);
 
drivers/xen/events/events_base.c +17 -11
index 3a791c8485d0..eb35c3cda9a6 100644
@@ -1639,26 +1639,30 @@ EXPORT_SYMBOL_GPL(xen_set_callback_via);
 /* Vector callbacks are better than PCI interrupts to receive event
  * channel notifications because we can receive vector callbacks on any
  * vcpu and we don't need PCI support or APIC interactions. */
-void xen_callback_vector(void)
+void xen_setup_callback_vector(void)
 {
-	int rc;
 	uint64_t callback_via;
 
 	if (xen_have_vector_callback) {
 		callback_via = HVM_CALLBACK_VECTOR(HYPERVISOR_CALLBACK_VECTOR);
-		rc = xen_set_callback_via(callback_via);
-		if (rc) {
+		if (xen_set_callback_via(callback_via)) {
 			pr_err("Request for Xen HVM callback vector failed\n");
 			xen_have_vector_callback = 0;
-			return;
 		}
-		pr_info_once("Xen HVM callback vector for event delivery is enabled\n");
-		alloc_intr_gate(HYPERVISOR_CALLBACK_VECTOR,
-				xen_hvm_callback_vector);
 	}
 }
+
+static __init void xen_alloc_callback_vector(void)
+{
+	if (!xen_have_vector_callback)
+		return;
+
+	pr_info("Xen HVM callback vector for event delivery is enabled\n");
+	alloc_intr_gate(HYPERVISOR_CALLBACK_VECTOR, xen_hvm_callback_vector);
+}
 #else
-void xen_callback_vector(void) {}
+void xen_setup_callback_vector(void) {}
+static inline void xen_alloc_callback_vector(void) {}
 #endif
 
 #undef MODULE_PARAM_PREFIX
@@ -1692,8 +1696,10 @@ void __init xen_init_IRQ(void)
 		if (xen_initial_domain())
 			pci_xen_initial_domain();
 	}
-	if (xen_feature(XENFEAT_hvm_callback_vector))
-		xen_callback_vector();
+	if (xen_feature(XENFEAT_hvm_callback_vector)) {
+		xen_setup_callback_vector();
+		xen_alloc_callback_vector();
+	}
 
 	if (xen_hvm_domain()) {
 		native_init_IRQ();
Signature · 1 line
2.25.3
#2 lore raw

[tip: x86/entry] x86/xen: Split HVM vector callback setup and interrupt gate allocation

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     fad1940a6a856f59b073e8650e02052ce531154c
Gitweb:        https://git.kernel.org/tip/fad1940a6a856f59b073e8650e02052ce531154c
Author:        Vitaly Kuznetsov <[email protected]>
AuthorDate:    Tue, 28 Apr 2020 11:38:22 +02:00
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Wed, 13 May 2020 21:13:54 +02:00

x86/xen: Split HVM vector callback setup and interrupt gate allocation

As a preparatory change for making alloc_intr_gate() __init split xen_callback_vector() into callback vector setup via hypercall (xen_setup_callback_vector()) and interrupt gate allocation (xen_alloc_callback_vector()).

xen_setup_callback_vector() is being called twice: on init and upon system resume from xen_hvm_post_suspend(). alloc_intr_gate() only needs to be called once.

---
 arch/x86/xen/suspend_hvm.c       |  2 +-
 arch/x86/xen/xen-ops.h           |  2 +-
 drivers/xen/events/events_base.c | 28 +++++++++++++++++-----------
 3 files changed, 19 insertions(+), 13 deletions(-)
arch/x86/xen/suspend_hvm.c +1 -1
index e666b61..5152afe 100644
@@ -13,6 +13,6 @@ void xen_hvm_post_suspend(int suspend_cancelled)
 		xen_hvm_init_shared_info();
 		xen_vcpu_restore();
 	}
-	xen_callback_vector();
+	xen_setup_callback_vector();
 	xen_unplug_emulated_devices();
 }
arch/x86/xen/xen-ops.h +1 -1
index 45a441c..1cc1568 100644
@@ -55,7 +55,7 @@ void xen_enable_sysenter(void);
 void xen_enable_syscall(void);
 void xen_vcpu_restore(void);
 
-void xen_callback_vector(void);
+void xen_setup_callback_vector(void);
 void xen_hvm_init_shared_info(void);
 void xen_unplug_emulated_devices(void);
 
drivers/xen/events/events_base.c +17 -11
index 3a791c8..eb35c3c 100644
@@ -1639,26 +1639,30 @@ EXPORT_SYMBOL_GPL(xen_set_callback_via);
 /* Vector callbacks are better than PCI interrupts to receive event
  * channel notifications because we can receive vector callbacks on any
  * vcpu and we don't need PCI support or APIC interactions. */
-void xen_callback_vector(void)
+void xen_setup_callback_vector(void)
 {
-	int rc;
 	uint64_t callback_via;
 
 	if (xen_have_vector_callback) {
 		callback_via = HVM_CALLBACK_VECTOR(HYPERVISOR_CALLBACK_VECTOR);
-		rc = xen_set_callback_via(callback_via);
-		if (rc) {
+		if (xen_set_callback_via(callback_via)) {
 			pr_err("Request for Xen HVM callback vector failed\n");
 			xen_have_vector_callback = 0;
-			return;
 		}
-		pr_info_once("Xen HVM callback vector for event delivery is enabled\n");
-		alloc_intr_gate(HYPERVISOR_CALLBACK_VECTOR,
-				xen_hvm_callback_vector);
 	}
 }
+
+static __init void xen_alloc_callback_vector(void)
+{
+	if (!xen_have_vector_callback)
+		return;
+
+	pr_info("Xen HVM callback vector for event delivery is enabled\n");
+	alloc_intr_gate(HYPERVISOR_CALLBACK_VECTOR, xen_hvm_callback_vector);
+}
 #else
-void xen_callback_vector(void) {}
+void xen_setup_callback_vector(void) {}
+static inline void xen_alloc_callback_vector(void) {}
 #endif
 
 #undef MODULE_PARAM_PREFIX
@@ -1692,8 +1696,10 @@ void __init xen_init_IRQ(void)
 		if (xen_initial_domain())
 			pci_xen_initial_domain();
 	}
-	if (xen_feature(XENFEAT_hvm_callback_vector))
-		xen_callback_vector();
+	if (xen_feature(XENFEAT_hvm_callback_vector)) {
+		xen_setup_callback_vector();
+		xen_alloc_callback_vector();
+	}
 
 	if (xen_hvm_domain()) {
 		native_init_IRQ();
#3 lore raw

PATCH v2 2/3 x86/idt: Annotate alloc_intr_gate() with __init

In reply to Vitaly Kuznetsov

There seems to be no reason to allocate interrupt gates after init. Mark alloc_intr_gate() as __init and add WARN_ON() checks making sure it is only used before idt_setup_apic_and_irq_gates() finalizes IDT setup and maps all un-allocated entries to spurious entries.

---
 arch/x86/kernel/idt.c | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)
arch/x86/kernel/idt.c +13 -3
index 87ef69a72c52..f95c3be00e5a 100644
@@ -51,6 +51,9 @@ struct idt_data {
 #define TSKG(_vector, _gdt)				\
 	G(_vector, NULL, DEFAULT_STACK, GATE_TASK, DPL0, _gdt << 3)
 
+
+static bool idt_setup_done __initdata;
+
 /*
  * Early traps running on the DEFAULT_STACK because the other interrupt
  * stacks work only after cpu_init().
@@ -323,6 +326,7 @@ void __init idt_setup_apic_and_irq_gates(void)
 		set_intr_gate(i, entry);
 	}
 #endif
+	idt_setup_done = true;
 }
 
 /**
@@ -352,6 +356,7 @@ void idt_invalidate(void *addr)
 	load_idt(&idt);
 }
 
+/* This goes away once ASYNC_PF is sanitized */
 void __init update_intr_gate(unsigned int n, const void *addr)
 {
 	if (WARN_ON_ONCE(!test_bit(n, system_vectors)))
@@ -359,9 +364,14 @@ void __init update_intr_gate(unsigned int n, const void *addr)
 	set_intr_gate(n, addr);
 }
 
-void alloc_intr_gate(unsigned int n, const void *addr)
+void __init alloc_intr_gate(unsigned int n, const void *addr)
 {
-	BUG_ON(n < FIRST_SYSTEM_VECTOR);
-	if (!test_and_set_bit(n, system_vectors))
+	if (WARN_ON(n < FIRST_SYSTEM_VECTOR))
+		return;
+
+	if (WARN_ON(idt_setup_done))
+		return;
+
+	if (!WARN_ON(test_and_set_bit(n, system_vectors)))
 		set_intr_gate(n, addr);
 }
Signature · 1 line
2.25.3
#4 lore raw

[tip: x86/entry] x86/idt: Annotate alloc_intr_gate() with __init

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     4c74d51dab3dd655062a4740af150c1835e19cff
Gitweb:        https://git.kernel.org/tip/4c74d51dab3dd655062a4740af150c1835e19cff
Author:        Vitaly Kuznetsov <[email protected]>
AuthorDate:    Tue, 28 Apr 2020 11:38:23 +02:00
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Wed, 13 May 2020 21:13:54 +02:00

x86/idt: Annotate alloc_intr_gate() with __init

There seems to be no reason to allocate interrupt gates after init. Mark alloc_intr_gate() as __init and add WARN_ON() checks making sure it is only used before idt_setup_apic_and_irq_gates() finalizes IDT setup and maps all un-allocated entries to spurious entries.

---
 arch/x86/kernel/idt.c | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)
arch/x86/kernel/idt.c +13 -3
index 98bcb50..0e92051 100644
@@ -51,6 +51,9 @@ struct idt_data {
 #define TSKG(_vector, _gdt)				\
 	G(_vector, NULL, DEFAULT_STACK, GATE_TASK, DPL0, _gdt << 3)
 
+
+static bool idt_setup_done __initdata;
+
 /*
  * Early traps running on the DEFAULT_STACK because the other interrupt
  * stacks work only after cpu_init().
@@ -323,6 +326,7 @@ void __init idt_setup_apic_and_irq_gates(void)
 		set_intr_gate(i, entry);
 	}
 #endif
+	idt_setup_done = true;
 }
 
 /**
@@ -352,6 +356,7 @@ void idt_invalidate(void *addr)
 	load_idt(&idt);
 }
 
+/* This goes away once ASYNC_PF is sanitized */
 void __init update_intr_gate(unsigned int n, const void *addr)
 {
 	if (WARN_ON_ONCE(!test_bit(n, system_vectors)))
@@ -359,9 +364,14 @@ void __init update_intr_gate(unsigned int n, const void *addr)
 	set_intr_gate(n, addr);
 }
 
-void alloc_intr_gate(unsigned int n, const void *addr)
+void __init alloc_intr_gate(unsigned int n, const void *addr)
 {
-	BUG_ON(n < FIRST_SYSTEM_VECTOR);
-	if (!test_and_set_bit(n, system_vectors))
+	if (WARN_ON(n < FIRST_SYSTEM_VECTOR))
+		return;
+
+	if (WARN_ON(idt_setup_done))
+		return;
+
+	if (!WARN_ON(test_and_set_bit(n, system_vectors)))
 		set_intr_gate(n, addr);
 }
#5 lore raw

PATCH v2 3/3 x86/idt: Keep spurious entries unset in system_vectors

In reply to Vitaly Kuznetsov

With commit dc20b2d52653 ("x86/idt: Move interrupt gate initialization to IDT code") non assigned system vectors are also marked as used in 'used_vectors' (now 'system_vectors') bitmap. This makes checks in arch_show_interrupts() whether a particular system vector is allocated to always pass and e.g. 'Hyper-V reenlightenment interrupts' entry always shows up in /proc/interrupts.

Another side effect of having all unassigned system vectors marked as used is that irq_matrix_debug_show() will wrongly count them among 'System' vectors.

As we now make sure alloc_intr_gate() is not called after init, it is possible to leave unused entries in 'system_vectors' unset to fix the issues.

---
 arch/x86/kernel/idt.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
arch/x86/kernel/idt.c +5 -1
index f95c3be00e5a..28500ccb835c 100644
@@ -321,7 +321,11 @@ void __init idt_setup_apic_and_irq_gates(void)
 
 #ifdef CONFIG_X86_LOCAL_APIC
 	for_each_clear_bit_from(i, system_vectors, NR_VECTORS) {
-		set_bit(i, system_vectors);
+		/*
+		 * Don't set the non assigned system vectors in the
+		 * system_vectors bitmap. Otherwise they show up in
+		 * /proc/interrupts.
+		 */
 		entry = spurious_entries_start + 8 * (i - FIRST_SYSTEM_VECTOR);
 		set_intr_gate(i, entry);
 	}
Signature · 1 line
2.25.3
#6 lore raw

[tip: x86/entry] x86/idt: Keep spurious entries unset in system_vectors

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     82ff351052bcc4bf49dc966960f563d25f54d22b
Gitweb:        https://git.kernel.org/tip/82ff351052bcc4bf49dc966960f563d25f54d22b
Author:        Vitaly Kuznetsov <[email protected]>
AuthorDate:    Tue, 28 Apr 2020 11:38:24 +02:00
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Wed, 13 May 2020 21:13:55 +02:00

x86/idt: Keep spurious entries unset in system_vectors

With commit dc20b2d52653 ("x86/idt: Move interrupt gate initialization to IDT code") non assigned system vectors are also marked as used in 'used_vectors' (now 'system_vectors') bitmap. This makes checks in arch_show_interrupts() whether a particular system vector is allocated to always pass and e.g. 'Hyper-V reenlightenment interrupts' entry always shows up in /proc/interrupts.

Another side effect of having all unassigned system vectors marked as used is that irq_matrix_debug_show() will wrongly count them among 'System' vectors.

As it is now ensured that alloc_intr_gate() is not called after init, it is possible to leave unused entries in 'system_vectors' unset to fix these issues.

---
 arch/x86/kernel/idt.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
arch/x86/kernel/idt.c +5 -1
index 0e92051..36fef90 100644
@@ -321,7 +321,11 @@ void __init idt_setup_apic_and_irq_gates(void)
 
 #ifdef CONFIG_X86_LOCAL_APIC
 	for_each_clear_bit_from(i, system_vectors, NR_VECTORS) {
-		set_bit(i, system_vectors);
+		/*
+		 * Don't set the non assigned system vectors in the
+		 * system_vectors bitmap. Otherwise they show up in
+		 * /proc/interrupts.
+		 */
 		entry = spurious_entries_start + 8 * (i - FIRST_SYSTEM_VECTOR);
 		set_intr_gate(i, entry);
 	}
#7 lore raw
In reply to Vitaly Kuznetsov
Quoted 9 lines · Vitaly Kuznetsov

This series is a successor of "[PATCH] x86/idt: Keep spurious entries unset in system_vectors".

The original issue I tried to address was that /proc/interrupts output was always containing all possible system vectors, including non allocated ones (e.g. 'Hypervisor callback interrupts' on bare metal). Thomas suggested to expand this cosmetic change to making alloc_intr_gate() __init.

Quoted 5 lines

Vitaly Kuznetsov (3):
x86/xen: Split HVM vector callback setup and interrupt gate allocation x86/idt: Annotate alloc_intr_gate() with __init
x86/idt: Keep spurious entries unset in system_vectors

Ping?

Signature · 1 line
Vitaly
#8 lore raw
Quoted 11 lines · Vitaly Kuznetsov
Quoted 9 lines · Vitaly Kuznetsov

This series is a successor of "[PATCH] x86/idt: Keep spurious entries unset in system_vectors".

The original issue I tried to address was that /proc/interrupts output was always containing all possible system vectors, including non allocated ones (e.g. 'Hypervisor callback interrupts' on bare metal). Thomas suggested to expand this cosmetic change to making alloc_intr_gate() __init.

Quoted 7 lines
Quoted 5 lines

Vitaly Kuznetsov (3):
x86/xen: Split HVM vector callback setup and interrupt gate allocation x86/idt: Annotate alloc_intr_gate() with __init
x86/idt: Keep spurious entries unset in system_vectors

Ping?

It's in my pile and I just did not come around to merge it.

Thanks,

tglx

#9 lore raw
Quoted 18 lines · Thomas Gleixner
Quoted 17 lines · Vitaly Kuznetsov
Quoted 14 lines · Vitaly Kuznetsov

This series is a successor of "[PATCH] x86/idt: Keep spurious entries unset in system_vectors".

The original issue I tried to address was that /proc/interrupts output was always containing all possible system vectors, including non allocated ones (e.g. 'Hypervisor callback interrupts' on bare metal). Thomas suggested to expand this cosmetic change to making alloc_intr_gate() __init.

Vitaly Kuznetsov (3):
x86/xen: Split HVM vector callback setup and interrupt gate allocation x86/idt: Annotate alloc_intr_gate() with __init
x86/idt: Keep spurious entries unset in system_vectors

Quoted 4 lines
Quoted 1 line

Ping?

It's in my pile and I just did not come around to merge it.

No problem, just wanted to make sure it didn't slip through the cracks. I see tip-bot already queued it, thanks!

Signature · 1 line
Vitaly