diff mbox series

[RFC,v3,12/15] context_tracking,x86: Defer kernel text patching IPIs

Message ID 20241119153502.41361-13-vschneid@redhat.com (mailing list archive)
State RFC
Headers show
Series context_tracking,x86: Defer some IPIs until a user->kernel transition | expand

Checks

Context Check Description
netdev/tree_selection success Not a local patch

Commit Message

Valentin Schneider Nov. 19, 2024, 3:34 p.m. UTC 
text_poke_bp_batch() sends IPIs to all online CPUs to synchronize
them vs the newly patched instruction. CPUs that are executing in userspace
do not need this synchronization to happen immediately, and this is
actually harmful interference for NOHZ_FULL CPUs.

As the synchronization IPIs are sent using a blocking call, returning from
text_poke_bp_batch() implies all CPUs will observe the patched
instruction(s), and this should be preserved even if the IPI is deferred.
In other words, to safely defer this synchronization, any kernel
instruction leading to the execution of the deferred instruction
sync (ct_work_flush()) must *not* be mutable (patchable) at runtime.

This means we must pay attention to mutable instructions in the early entry
code:
- alternatives
- static keys
- all sorts of probes (kprobes/ftrace/bpf/???)

The early entry code leading to ct_work_flush() is noinstr, which gets rid
of the probes.

Alternatives are safe, because it's boot-time patching (before SMP is
even brought up) which is before any IPI deferral can happen.

This leaves us with static keys. Any static key used in early entry code
should be only forever-enabled at boot time, IOW __ro_after_init (pretty
much like alternatives). Exceptions are marked as forceful and will always
generate an IPI when flipped. Objtool is now able to point at static keys
that don't respect this, and all static keys used in early entry code have
now been verified as behaving like so.

Leverage the new context_tracking infrastructure to defer sync_core() IPIs
to a target CPU's next kernel entry.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Signed-off-by: Valentin Schneider <vschneid@redhat.com>
---
 arch/x86/include/asm/context_tracking_work.h |  6 ++--
 arch/x86/include/asm/text-patching.h         |  1 +
 arch/x86/kernel/alternative.c                | 33 +++++++++++++++++---
 arch/x86/kernel/kprobes/core.c               |  4 +--
 arch/x86/kernel/kprobes/opt.c                |  4 +--
 arch/x86/kernel/module.c                     |  2 +-
 include/linux/context_tracking_work.h        |  4 +--
 7 files changed, 41 insertions(+), 13 deletions(-)

Comments

Peter Zijlstra Nov. 20, 2024, 3:13 p.m. UTC | #1 
On Tue, Nov 19, 2024 at 04:34:59PM +0100, Valentin Schneider wrote:

> +static void __text_poke_sync(smp_cond_func_t cond_func)
> +{
> +	on_each_cpu_cond(cond_func, do_sync_core, NULL, 1);
> +}
> +
>  void text_poke_sync(void)
>  {
> -	on_each_cpu(do_sync_core, NULL, 1);
> +	__text_poke_sync(NULL);
> +}
> +
> +void text_poke_sync_deferrable(void)
> +{
> +	__text_poke_sync(do_sync_core_defer_cond);
>  }

How about we unwrap some of that like so:

>  /*
> @@ -2257,6 +2273,8 @@ static int tp_vec_nr;
>  static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries)
>  {
>  	unsigned char int3 = INT3_INSN_OPCODE;
> +	bool force_ipi = false;
> +	void (*sync_fn)(void);

	smp_cond_func_t cond = do_sync_core_defer_cond;

>  	unsigned int i;
>  	int do_sync;
>  
> @@ -2291,11 +2309,18 @@ static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries
>  	 * First step: add a int3 trap to the address that will be patched.
>  	 */
>  	for (i = 0; i < nr_entries; i++) {
> +		/*
> +		 * Record that we need to send the IPI if at least one location
> +		 * in the batch requires it.
> +		 */
> +		force_ipi |= tp[i].force_ipi;

		if (tp[i].force_ipi)
			cond = NULL;

>  		tp[i].old = *(u8 *)text_poke_addr(&tp[i]);
>  		text_poke(text_poke_addr(&tp[i]), &int3, INT3_INSN_SIZE);
>  	}
>  
> -	text_poke_sync();
> +	sync_fn = force_ipi ? text_poke_sync : text_poke_sync_deferrable;
> +
> +	sync_fn();

	__text_poke_sync(cond);
diff mbox series

Patch

diff --git a/arch/x86/include/asm/context_tracking_work.h b/arch/x86/include/asm/context_tracking_work.h
index 5bc29e6b2ed38..2c66687ce00e2 100644
--- a/arch/x86/include/asm/context_tracking_work.h
+++ b/arch/x86/include/asm/context_tracking_work.h
@@ -2,11 +2,13 @@ 
 #ifndef _ASM_X86_CONTEXT_TRACKING_WORK_H
 #define _ASM_X86_CONTEXT_TRACKING_WORK_H
 
+#include <asm/sync_core.h>
+
 static __always_inline void arch_context_tracking_work(int work)
 {
 	switch (work) {
-	case CONTEXT_WORK_n:
-		// Do work...
+	case CONTEXT_WORK_SYNC:
+		sync_core();
 		break;
 	}
 }
diff --git a/arch/x86/include/asm/text-patching.h b/arch/x86/include/asm/text-patching.h
index e34de36cab61e..37344e95f738f 100644
--- a/arch/x86/include/asm/text-patching.h
+++ b/arch/x86/include/asm/text-patching.h
@@ -33,6 +33,7 @@  extern void apply_relocation(u8 *buf, const u8 * const instr, size_t instrlen, u
  */
 extern void *text_poke(void *addr, const void *opcode, size_t len);
 extern void text_poke_sync(void);
+extern void text_poke_sync_deferrable(void);
 extern void *text_poke_kgdb(void *addr, const void *opcode, size_t len);
 extern void *text_poke_copy(void *addr, const void *opcode, size_t len);
 extern void *text_poke_copy_locked(void *addr, const void *opcode, size_t len, bool core_ok);
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 954c4c0f7fc58..4ce224d927b03 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -18,6 +18,7 @@ 
 #include <linux/mmu_context.h>
 #include <linux/bsearch.h>
 #include <linux/sync_core.h>
+#include <linux/context_tracking.h>
 #include <asm/text-patching.h>
 #include <asm/alternative.h>
 #include <asm/sections.h>
@@ -2080,9 +2081,24 @@  static void do_sync_core(void *info)
 	sync_core();
 }
 
+static bool do_sync_core_defer_cond(int cpu, void *info)
+{
+	return !ct_set_cpu_work(cpu, CONTEXT_WORK_SYNC);
+}
+
+static void __text_poke_sync(smp_cond_func_t cond_func)
+{
+	on_each_cpu_cond(cond_func, do_sync_core, NULL, 1);
+}
+
 void text_poke_sync(void)
 {
-	on_each_cpu(do_sync_core, NULL, 1);
+	__text_poke_sync(NULL);
+}
+
+void text_poke_sync_deferrable(void)
+{
+	__text_poke_sync(do_sync_core_defer_cond);
 }
 
 /*
@@ -2257,6 +2273,8 @@  static int tp_vec_nr;
 static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries)
 {
 	unsigned char int3 = INT3_INSN_OPCODE;
+	bool force_ipi = false;
+	void (*sync_fn)(void);
 	unsigned int i;
 	int do_sync;
 
@@ -2291,11 +2309,18 @@  static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries
 	 * First step: add a int3 trap to the address that will be patched.
 	 */
 	for (i = 0; i < nr_entries; i++) {
+		/*
+		 * Record that we need to send the IPI if at least one location
+		 * in the batch requires it.
+		 */
+		force_ipi |= tp[i].force_ipi;
 		tp[i].old = *(u8 *)text_poke_addr(&tp[i]);
 		text_poke(text_poke_addr(&tp[i]), &int3, INT3_INSN_SIZE);
 	}
 
-	text_poke_sync();
+	sync_fn = force_ipi ? text_poke_sync : text_poke_sync_deferrable;
+
+	sync_fn();
 
 	/*
 	 * Second step: update all but the first byte of the patched range.
@@ -2357,7 +2382,7 @@  static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries
 		 * not necessary and we'd be safe even without it. But
 		 * better safe than sorry (plus there's not only Intel).
 		 */
-		text_poke_sync();
+		sync_fn();
 	}
 
 	/*
@@ -2378,7 +2403,7 @@  static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries
 	}
 
 	if (do_sync)
-		text_poke_sync();
+		sync_fn();
 
 	/*
 	 * Remove and wait for refs to be zero.
diff --git a/arch/x86/kernel/kprobes/core.c b/arch/x86/kernel/kprobes/core.c
index 72e6a45e7ec24..c2fd2578fd5fc 100644
--- a/arch/x86/kernel/kprobes/core.c
+++ b/arch/x86/kernel/kprobes/core.c
@@ -817,7 +817,7 @@  void arch_arm_kprobe(struct kprobe *p)
 	u8 int3 = INT3_INSN_OPCODE;
 
 	text_poke(p->addr, &int3, 1);
-	text_poke_sync();
+	text_poke_sync_deferrable();
 	perf_event_text_poke(p->addr, &p->opcode, 1, &int3, 1);
 }
 
@@ -827,7 +827,7 @@  void arch_disarm_kprobe(struct kprobe *p)
 
 	perf_event_text_poke(p->addr, &int3, 1, &p->opcode, 1);
 	text_poke(p->addr, &p->opcode, 1);
-	text_poke_sync();
+	text_poke_sync_deferrable();
 }
 
 void arch_remove_kprobe(struct kprobe *p)
diff --git a/arch/x86/kernel/kprobes/opt.c b/arch/x86/kernel/kprobes/opt.c
index 36d6809c6c9e1..b2ce4d9c3ba56 100644
--- a/arch/x86/kernel/kprobes/opt.c
+++ b/arch/x86/kernel/kprobes/opt.c
@@ -513,11 +513,11 @@  void arch_unoptimize_kprobe(struct optimized_kprobe *op)
 	       JMP32_INSN_SIZE - INT3_INSN_SIZE);
 
 	text_poke(addr, new, INT3_INSN_SIZE);
-	text_poke_sync();
+	text_poke_sync_deferrable();
 	text_poke(addr + INT3_INSN_SIZE,
 		  new + INT3_INSN_SIZE,
 		  JMP32_INSN_SIZE - INT3_INSN_SIZE);
-	text_poke_sync();
+	text_poke_sync_deferrable();
 
 	perf_event_text_poke(op->kp.addr, old, JMP32_INSN_SIZE, new, JMP32_INSN_SIZE);
 }
diff --git a/arch/x86/kernel/module.c b/arch/x86/kernel/module.c
index 837450b6e882f..00e71ad30c01d 100644
--- a/arch/x86/kernel/module.c
+++ b/arch/x86/kernel/module.c
@@ -191,7 +191,7 @@  static int write_relocate_add(Elf64_Shdr *sechdrs,
 				   write, apply);
 
 	if (!early) {
-		text_poke_sync();
+		text_poke_sync_deferrable();
 		mutex_unlock(&text_mutex);
 	}
 
diff --git a/include/linux/context_tracking_work.h b/include/linux/context_tracking_work.h
index fb74db8876dd2..13fc97b395030 100644
--- a/include/linux/context_tracking_work.h
+++ b/include/linux/context_tracking_work.h
@@ -5,12 +5,12 @@ 
 #include <linux/bitops.h>
 
 enum {
-	CONTEXT_WORK_n_OFFSET,
+	CONTEXT_WORK_SYNC_OFFSET,
 	CONTEXT_WORK_MAX_OFFSET
 };
 
 enum ct_work {
-	CONTEXT_WORK_n        = BIT(CONTEXT_WORK_n_OFFSET),
+	CONTEXT_WORK_SYNC     = BIT(CONTEXT_WORK_SYNC_OFFSET),
 	CONTEXT_WORK_MAX      = BIT(CONTEXT_WORK_MAX_OFFSET)
 };