On March 27, 2025 1:53:53 PM PDT, Ingo Molnar <mingo@kernel.org> wrote:
>Unlike sync_core(), text_poke_sync() is a very heavy operation, as
>it sends an IPI to every online CPU in the system and waits for
>completion.
>
>Reflect this in the name.
>
>Signed-off-by: Ingo Molnar <mingo@kernel.org>
>---
> arch/x86/include/asm/text-patching.h | 2 +-
> arch/x86/kernel/alternative.c | 12 ++++++------
> arch/x86/kernel/kprobes/core.c | 4 ++--
> arch/x86/kernel/kprobes/opt.c | 4 ++--
> arch/x86/kernel/module.c | 2 +-
> 5 files changed, 12 insertions(+), 12 deletions(-)
>
>diff --git a/arch/x86/include/asm/text-patching.h b/arch/x86/include/asm/text-patching.h
>index 611957617278..ff30aa1d0c47 100644
>--- a/arch/x86/include/asm/text-patching.h
>+++ b/arch/x86/include/asm/text-patching.h
>@@ -32,7 +32,7 @@ extern void apply_relocation(u8 *buf, const u8 * const instr, size_t instrlen, u
> * an inconsistent instruction while you patch.
> */
> extern void *text_poke(void *addr, const void *opcode, size_t len);
>-extern void text_poke_sync(void);
>+extern void text_poke_sync_each_cpu(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);
> #define text_poke_copy text_poke_copy
>diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
>index 1b523496a2f6..32d3707d7963 100644
>--- a/arch/x86/kernel/alternative.c
>+++ b/arch/x86/kernel/alternative.c
>@@ -2445,7 +2445,7 @@ static void do_sync_core(void *info)
> sync_core();
> }
>
>-void text_poke_sync(void)
>+void text_poke_sync_each_cpu(void)
> {
> on_each_cpu(do_sync_core, NULL, 1);
> }
>@@ -2469,8 +2469,8 @@ struct text_poke_int3_loc {
> #define TP_ARRAY_NR_ENTRIES_MAX (PAGE_SIZE / sizeof(struct text_poke_int3_loc))
>
> static struct text_poke_int3_array {
>- int nr_entries;
> struct text_poke_int3_loc vec[TP_ARRAY_NR_ENTRIES_MAX];
>+ int nr_entries;
> } tp_array;
>
> static DEFINE_PER_CPU(atomic_t, tp_array_refs);
>@@ -2649,7 +2649,7 @@ static void text_poke_int3_batch_process(void)
> text_poke(text_poke_int3_addr(&tp_array.vec[i]), &int3, INT3_INSN_SIZE);
> }
>
>- text_poke_sync();
>+ text_poke_sync_each_cpu();
>
> /*
> * Second step: update all but the first byte of the patched range.
>@@ -2711,7 +2711,7 @@ static void text_poke_int3_batch_process(void)
> * not necessary and we'd be safe even without it. But
> * better safe than sorry (plus there's not only Intel).
> */
>- text_poke_sync();
>+ text_poke_sync_each_cpu();
> }
>
> /*
>@@ -2732,13 +2732,13 @@ static void text_poke_int3_batch_process(void)
> }
>
> if (do_sync)
>- text_poke_sync();
>+ text_poke_sync_each_cpu();
>
> /*
> * Remove and wait for refs to be zero.
> *
> * Notably, if after step-3 above the INT3 got removed, then the
>- * text_poke_sync() will have serialized against any running INT3
>+ * text_poke_sync_each_cpu() will have serialized against any running INT3
> * handlers and the below spin-wait will not happen.
> *
> * IOW. unless the replacement instruction is INT3, this case goes
>diff --git a/arch/x86/kernel/kprobes/core.c b/arch/x86/kernel/kprobes/core.c
>index 09608fd93687..5e35c95524dc 100644
>--- a/arch/x86/kernel/kprobes/core.c
>+++ b/arch/x86/kernel/kprobes/core.c
>@@ -808,7 +808,7 @@ void arch_arm_kprobe(struct kprobe *p)
> u8 int3 = INT3_INSN_OPCODE;
>
> text_poke(p->addr, &int3, 1);
>- text_poke_sync();
>+ text_poke_sync_each_cpu();
> perf_event_text_poke(p->addr, &p->opcode, 1, &int3, 1);
> }
>
>@@ -818,7 +818,7 @@ 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_each_cpu();
> }
>
> void arch_remove_kprobe(struct kprobe *p)
>diff --git a/arch/x86/kernel/kprobes/opt.c b/arch/x86/kernel/kprobes/opt.c
>index 54bc5e7c6886..5efa7b50bbb3 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_each_cpu();
> text_poke(addr + INT3_INSN_SIZE,
> new + INT3_INSN_SIZE,
> JMP32_INSN_SIZE - INT3_INSN_SIZE);
>- text_poke_sync();
>+ text_poke_sync_each_cpu();
>
> 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 a7998f351701..1c598c90e24d 100644
>--- a/arch/x86/kernel/module.c
>+++ b/arch/x86/kernel/module.c
>@@ -206,7 +206,7 @@ static int write_relocate_add(Elf64_Shdr *sechdrs,
> write, apply);
>
> if (!early) {
>- text_poke_sync();
>+ text_poke_sync_each_cpu();
> mutex_unlock(&text_mutex);
> }
>
Is that the only use case we have for syncing all CPUs?