From: Peter Zijlstra <peterz@infradead.org>
To: x86@kernel.org
Cc: peterz@infradead.org, linux-kernel@vger.kernel.org,
rostedt@goodmis.org, mhiramat@kernel.org, bristot@redhat.com,
jbaron@akamai.com, torvalds@linux-foundation.org,
tglx@linutronix.de, mingo@kernel.org, namit@vmware.com,
hpa@zytor.com, luto@kernel.org, ard.biesheuvel@linaro.org,
jpoimboe@redhat.com, pbonzini@redhat.com,
mathieu.desnoyers@efficios.com, linux@rasmusvillemoes.dk
Subject: [PATCH v6 13/17] static_call: Add static_call_cond()
Date: Fri, 10 Jul 2020 15:38:44 +0200 [thread overview]
Message-ID: <20200710134336.918547865@infradead.org> (raw)
In-Reply-To: 20200710133831.943894387@infradead.org
Extend the static_call infrastructure to optimize the following common
pattern:
if (func_ptr)
func_ptr(args...)
For the trampoline (which is in effect a tail-call), we patch the
JMP.d32 into a RET, which then directly consumes the trampoline call.
For the in-line sites we replace the CALL with a NOP5.
NOTE: this is 'obviously' limited to functions with a 'void' return type.
NOTE: DEFINE_STATIC_COND_CALL() only requires a typename, as opposed
to a full function.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
arch/x86/include/asm/static_call.h | 12 +++--
arch/x86/kernel/static_call.c | 42 +++++++++++++-----
include/linux/static_call.h | 86 +++++++++++++++++++++++++++++++++++++
3 files changed, 127 insertions(+), 13 deletions(-)
--- a/arch/x86/include/asm/static_call.h
+++ b/arch/x86/include/asm/static_call.h
@@ -20,15 +20,21 @@
* it does tail-call optimization on the call; since you cannot compute the
* relative displacement across sections.
*/
-#define ARCH_DEFINE_STATIC_CALL_TRAMP(name, func) \
+
+#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" \
- " .byte 0xe9 # jmp.d32 \n" \
- " .long " #func " - (. + 4) \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 */
--- a/arch/x86/kernel/static_call.c
+++ b/arch/x86/kernel/static_call.c
@@ -4,19 +4,41 @@
#include <linux/bug.h>
#include <asm/text-patching.h>
-static void __static_call_transform(void *insn, u8 opcode, void *func)
+enum insn_type {
+ CALL = 0, /* site call */
+ NOP = 1, /* site cond-call */
+ JMP = 2, /* tramp / site tail-call */
+ RET = 3, /* tramp / site cond-tail-call */
+};
+
+static void __static_call_transform(void *insn, enum insn_type type, void *func)
{
- const void *code = text_gen_insn(opcode, insn, func);
+ int size = CALL_INSN_SIZE;
+ const void *code;
- if (WARN_ONCE(*(u8 *)insn != opcode,
- "unexpected static call insn opcode 0x%x at %pS\n",
- opcode, insn))
- return;
+ switch (type) {
+ case CALL:
+ code = text_gen_insn(CALL_INSN_OPCODE, insn, func);
+ break;
+
+ case NOP:
+ code = ideal_nops[NOP_ATOMIC5];
+ break;
+
+ case JMP:
+ code = text_gen_insn(JMP32_INSN_OPCODE, insn, func);
+ break;
+
+ case RET:
+ code = text_gen_insn(RET_INSN_OPCODE, insn, func);
+ size = RET_INSN_SIZE;
+ break;
+ }
- if (memcmp(insn, code, CALL_INSN_SIZE) == 0)
+ if (memcmp(insn, code, size) == 0)
return;
- text_poke_bp(insn, code, CALL_INSN_SIZE, NULL);
+ text_poke_bp(insn, code, size, NULL);
}
void arch_static_call_transform(void *site, void *tramp, void *func)
@@ -24,10 +46,10 @@ void arch_static_call_transform(void *si
mutex_lock(&text_mutex);
if (tramp)
- __static_call_transform(tramp, JMP32_INSN_OPCODE, func);
+ __static_call_transform(tramp, func ? JMP : RET, func);
if (IS_ENABLED(CONFIG_HAVE_STATIC_CALL_INLINE) && site)
- __static_call_transform(site, CALL_INSN_OPCODE, func);
+ __static_call_transform(site, func ? CALL : NOP, func);
mutex_unlock(&text_mutex);
}
--- a/include/linux/static_call.h
+++ b/include/linux/static_call.h
@@ -16,7 +16,9 @@
*
* 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);
*
* Usage example:
@@ -52,6 +54,43 @@
* rather than calling through the trampoline. This requires objtool or a
* compiler plugin to detect all the static_call() sites and annotate them
* in the .static_call_sites section.
+ *
+ *
+ * Notes on NULL function pointers:
+ *
+ * Static_call()s support NULL functions, with many of the caveats that
+ * regular function pointers have.
+ *
+ * Clearly calling a NULL function pointer is 'BAD', so too for
+ * static_call()s (although when HAVE_STATIC_CALL it might not be immediately
+ * fatal). A NULL static_call can be the result of:
+ *
+ * DECLARE_STATIC_CALL_NULL(my_static_call, void (*)(int));
+ *
+ * which is equivalent to declaring a NULL function pointer with just a
+ * typename:
+ *
+ * void (*my_func_ptr)(int arg1) = NULL;
+ *
+ * or using static_call_update() with a NULL function. In both cases the
+ * HAVE_STATIC_CALL implementation will patch the trampoline with a RET
+ * instruction, instead of an immediate tail-call JMP. HAVE_STATIC_CALL_INLINE
+ * architectures can patch the trampoline call to a NOP.
+ *
+ * In all cases, any argument evaluation is unconditional. Unlike a regular
+ * conditional function pointer call:
+ *
+ * if (my_func_ptr)
+ * my_func_ptr(arg1)
+ *
+ * where the argument evaludation also depends on the pointer value.
+ *
+ * When calling a static_call that can be NULL, use:
+ *
+ * static_call_cond(name)(arg1);
+ *
+ * which will include the required value tests to avoid NULL-pointer
+ * dereferences.
*/
#include <linux/types.h>
@@ -120,7 +159,16 @@ extern int static_call_text_reserved(voi
}; \
ARCH_DEFINE_STATIC_CALL_TRAMP(name, _func)
+#define DEFINE_STATIC_CALL_NULL(name, _func) \
+ DECLARE_STATIC_CALL(name, _func); \
+ struct static_call_key STATIC_CALL_KEY(name) = { \
+ .func = NULL, \
+ .type = 1, \
+ }; \
+ ARCH_DEFINE_STATIC_CALL_NULL_TRAMP(name)
+
#define static_call(name) __static_call(name)
+#define static_call_cond(name) (void)__static_call(name)
#define EXPORT_STATIC_CALL(name) \
EXPORT_SYMBOL(STATIC_CALL_KEY(name)); \
@@ -143,7 +191,15 @@ struct static_call_key {
}; \
ARCH_DEFINE_STATIC_CALL_TRAMP(name, _func)
+#define DEFINE_STATIC_CALL_NULL(name, _func) \
+ DECLARE_STATIC_CALL(name, _func); \
+ struct static_call_key STATIC_CALL_KEY(name) = { \
+ .func = NULL, \
+ }; \
+ ARCH_DEFINE_STATIC_CALL_NULL_TRAMP(name)
+
#define static_call(name) __static_call(name)
+#define static_call_cond(name) (void)__static_call(name)
static inline
void __static_call_update(struct static_call_key *key, void *tramp, void *func)
@@ -179,9 +235,39 @@ struct static_call_key {
.func = _func, \
}
+#define DEFINE_STATIC_CALL_NULL(name, _func) \
+ DECLARE_STATIC_CALL(name, _func); \
+ struct static_call_key STATIC_CALL_KEY(name) = { \
+ .func = NULL, \
+ }
+
#define static_call(name) \
((typeof(STATIC_CALL_TRAMP(name))*)(STATIC_CALL_KEY(name).func))
+static inline void __static_call_nop(void) { }
+
+/*
+ * This horrific hack takes care of two things:
+ *
+ * - it ensures the compiler will only load the function pointer ONCE,
+ * which avoids a reload race.
+ *
+ * - it ensures the argument evaluation is unconditional, similar
+ * to the HAVE_STATIC_CALL variant.
+ *
+ * Sadly current GCC/Clang (10 for both) do not optimize this properly
+ * and will emit an indirect call for the NULL case :-(
+ */
+#define __static_call_cond(name) \
+({ \
+ void *func = READ_ONCE(STATIC_CALL_KEY(name).func); \
+ if (!func) \
+ func = &__static_call_nop; \
+ (typeof(STATIC_CALL_TRAMP(name))*)func; \
+})
+
+#define static_call_cond(name) (void)__static_call_cond(name)
+
static inline
void __static_call_update(struct static_call_key *key, void *tramp, void *func)
{
next prev parent reply other threads:[~2020-07-10 13:46 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-07-10 13:38 [PATCH v6 00/17] Add static_call Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 01/17] notifier: Fix broken error handling pattern Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 02/17] module: Fix up module_notifier return values Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 03/17] module: Properly propagate MODULE_STATE_COMING failure Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 04/17] jump_label,module: Fix module lifetime for __jump_label_mod_text_reserved Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 05/17] compiler.h: Make __ADDRESSABLE() symbol truly unique Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 06/17] static_call: Add basic static call infrastructure Peter Zijlstra
2020-07-10 21:56 ` Steven Rostedt
2020-07-10 13:38 ` [PATCH v6 07/17] static_call: Add inline " Peter Zijlstra
2020-07-10 21:57 ` Steven Rostedt
2020-07-10 13:38 ` [PATCH v6 08/17] static_call: Avoid kprobes on inline static_call()s Peter Zijlstra
2020-07-10 22:00 ` Steven Rostedt
2020-07-11 10:30 ` Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 09/17] x86/static_call: Add out-of-line static call implementation Peter Zijlstra
2020-07-10 22:13 ` Steven Rostedt
2020-07-11 10:11 ` Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 10/17] x86/static_call: Add inline static call implementation for x86-64 Peter Zijlstra
2020-07-10 22:31 ` Steven Rostedt
2020-07-11 9:56 ` Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 11/17] static_call: Simple self-test Peter Zijlstra
2020-07-10 22:42 ` Steven Rostedt
2020-07-11 10:27 ` Peter Zijlstra
2020-07-13 20:26 ` Steven Rostedt
2020-07-10 13:38 ` [PATCH v6 12/17] x86/alternatives: Teach text_poke_bp() to emulate RET Peter Zijlstra
2020-07-10 22:44 ` Steven Rostedt
2020-07-10 13:38 ` Peter Zijlstra [this message]
2020-07-10 23:08 ` [PATCH v6 13/17] static_call: Add static_call_cond() Steven Rostedt
2020-07-11 5:09 ` Peter Zijlstra
2020-07-11 10:49 ` Peter Zijlstra
2020-07-13 20:32 ` Steven Rostedt
2020-07-14 9:53 ` Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 14/17] static_call: Handle tail-calls Peter Zijlstra
2020-07-11 0:23 ` Steven Rostedt
2020-07-11 5:06 ` Peter Zijlstra
2020-07-11 5:08 ` Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 15/17] static_call: Allow early init Peter Zijlstra
2020-07-11 1:14 ` Steven Rostedt
2020-07-11 5:08 ` Peter Zijlstra
2020-07-13 20:24 ` Steven Rostedt
2020-07-14 9:51 ` Peter Zijlstra
2020-07-14 14:16 ` Steven Rostedt
2020-07-14 15:54 ` Peter Zijlstra
2020-07-14 16:07 ` Steven Rostedt
2020-07-14 18:31 ` Peter Zijlstra
2020-07-14 19:38 ` Steven Rostedt
2020-07-10 13:38 ` [PATCH v6 16/17] tracepoint: Optimize using static_call() Peter Zijlstra
2020-07-10 13:38 ` [PATCH v6 17/17] x86/perf, static_call: Optimize x86_pmu methods Peter Zijlstra
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20200710134336.918547865@infradead.org \
--to=peterz@infradead.org \
--cc=ard.biesheuvel@linaro.org \
--cc=bristot@redhat.com \
--cc=hpa@zytor.com \
--cc=jbaron@akamai.com \
--cc=jpoimboe@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux@rasmusvillemoes.dk \
--cc=luto@kernel.org \
--cc=mathieu.desnoyers@efficios.com \
--cc=mhiramat@kernel.org \
--cc=mingo@kernel.org \
--cc=namit@vmware.com \
--cc=pbonzini@redhat.com \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.org \
--cc=x86@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).