Merge tag 'core-static_call-2020-10-12' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull static call support from Ingo Molnar:
 "This introduces static_call(), which is the idea of static_branch()
  applied to indirect function calls. Remove a data load (indirection)
  by modifying the text.

  They give the flexibility of function pointers, but with better
  performance. (This is especially important for cases where retpolines
  would otherwise be used, as retpolines can be pretty slow.)

  API overview:

      DECLARE_STATIC_CALL(name, func);
      DEFINE_STATIC_CALL(name, func);
      DEFINE_STATIC_CALL_NULL(name, typename);

      static_call(name)(args...);
      static_call_cond(name)(args...);
      static_call_update(name, func);

  x86 is supported via text patching, otherwise basic indirect calls are
  used, with function pointers.

  There's a second variant using inline code patching, inspired by
  jump-labels, implemented on x86 as well.

  The new APIs are utilized in the x86 perf code, a heavy user of
  function pointers, where static calls speed up the PMU handler by
  4.2% (!).

  The generic implementation is not really excercised on other
  architectures, outside of the trivial test_static_call_init()
  self-test"

* tag 'core-static_call-2020-10-12' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (21 commits)
  static_call: Fix return type of static_call_init
  tracepoint: Fix out of sync data passing by static caller
  tracepoint: Fix overly long tracepoint names
  x86/perf, static_call: Optimize x86_pmu methods
  tracepoint: Optimize using static_call()
  static_call: Allow early init
  static_call: Add some validation
  static_call: Handle tail-calls
  static_call: Add static_call_cond()
  x86/alternatives: Teach text_poke_bp() to emulate RET
  static_call: Add simple self-test for static calls
  x86/static_call: Add inline static call implementation for x86-64
  x86/static_call: Add out-of-line static call implementation
  static_call: Avoid kprobes on inline static_call()s
  static_call: Add inline static call infrastructure
  static_call: Add basic static call infrastructure
  compiler.h: Make __ADDRESSABLE() symbol truly unique
  jump_label,module: Fix module lifetime for __jump_label_mod_text_reserved()
  module: Properly propagate MODULE_STATE_COMING failure
  module: Fix up module_notifier return values
  ...

2 files changed, 59 insertions, 0 deletions

diff --git a/arch/x86/include/asm/static_call.h b/arch/x86/include/asm/static_call.h
new file mode 100644
index 000000000000..c37f11999d0c
--- /dev/null
+++ b/arch/x86/include/asm/static_call.h
@@ -0,0 +1,40 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_STATIC_CALL_H
+#define _ASM_STATIC_CALL_H
+
+#include <asm/text-patching.h>
+
+/*
+ * For CONFIG_HAVE_STATIC_CALL_INLINE, this is a temporary trampoline which
+ * uses the current value of the key->func pointer to do an indirect jump to
+ * the function.  This trampoline is only used during boot, before the call
+ * sites get patched by static_call_update().  The name of this trampoline has
+ * a magical aspect: objtool uses it to find static call sites so it can create
+ * the .static_call_sites section.
+ *
+ * For CONFIG_HAVE_STATIC_CALL, this is a permanent trampoline which
+ * does a direct jump to the function.  The direct jump gets patched by
+ * static_call_update().
+ *
+ * Having the trampoline in a special section forces GCC to emit a JMP.d32 when
+ * it does tail-call optimization on the call; since you cannot compute the
+ * relative displacement across sections.
+ */
+
+#define __ARCH_DEFINE_STATIC_CALL_TRAMP(name, insns)			\
+	asm(".pushsection .static_call.text, \"ax\"		\n"	\
+	    ".align 4						\n"	\
+	    ".globl " STATIC_CALL_TRAMP_STR(name) "		\n"	\
+	    STATIC_CALL_TRAMP_STR(name) ":			\n"	\
+	    insns "						\n"	\
+	    ".type " STATIC_CALL_TRAMP_STR(name) ", @function	\n"	\
+	    ".size " STATIC_CALL_TRAMP_STR(name) ", . - " STATIC_CALL_TRAMP_STR(name) " \n" \
+	    ".popsection					\n")
+
+#define ARCH_DEFINE_STATIC_CALL_TRAMP(name, func)			\
+	__ARCH_DEFINE_STATIC_CALL_TRAMP(name, ".byte 0xe9; .long " #func " - (. + 4)")
+
+#define ARCH_DEFINE_STATIC_CALL_NULL_TRAMP(name)			\
+	__ARCH_DEFINE_STATIC_CALL_TRAMP(name, "ret; nop; nop; nop; nop")
+
+#endif /* _ASM_STATIC_CALL_H */
diff --git a/arch/x86/include/asm/text-patching.h b/arch/x86/include/asm/text-patching.h
index 6593b42cb379..b7421780e4e9 100644
--- a/arch/x86/include/asm/text-patching.h
+++ b/arch/x86/include/asm/text-patching.h
@@ -53,6 +53,9 @@ extern void text_poke_finish(void);
 #define INT3_INSN_SIZE		1
 #define INT3_INSN_OPCODE	0xCC
 
+#define RET_INSN_SIZE		1
+#define RET_INSN_OPCODE		0xC3
+
 #define CALL_INSN_SIZE		5
 #define CALL_INSN_OPCODE	0xE8
 
@@ -73,6 +76,7 @@ static __always_inline int text_opcode_size(u8 opcode)
 
 	switch(opcode) {
 	__CASE(INT3);
+	__CASE(RET);
 	__CASE(CALL);
 	__CASE(JMP32);
 	__CASE(JMP8);
@@ -141,11 +145,26 @@ void int3_emulate_push(struct pt_regs *regs, unsigned long val)
 }
 
 static __always_inline
+unsigned long int3_emulate_pop(struct pt_regs *regs)
+{
+	unsigned long val = *(unsigned long *)regs->sp;
+	regs->sp += sizeof(unsigned long);
+	return val;
+}
+
+static __always_inline
 void int3_emulate_call(struct pt_regs *regs, unsigned long func)
 {
 	int3_emulate_push(regs, regs->ip - INT3_INSN_SIZE + CALL_INSN_SIZE);
 	int3_emulate_jmp(regs, func);
 }
+
+static __always_inline
+void int3_emulate_ret(struct pt_regs *regs)
+{
+	unsigned long ip = int3_emulate_pop(regs);
+	int3_emulate_jmp(regs, ip);
+}
 #endif /* !CONFIG_UML_X86 */
 
 #endif /* _ASM_X86_TEXT_PATCHING_H */