From: Peter Zijlstra <peterz@infradead.org>
To: "H. Peter Anvin" <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
Ingo Molnar <mingo@kernel.org>, Andi Kleen <ak@linux.intel.com>,
Mike Galbraith <bitbucket@online.de>,
Thomas Gleixner <tglx@linutronix.de>,
Arjan van de Ven <arjan@linux.intel.com>,
linux-kernel@vger.kernel.org, linux-arch@vger.kernel.org
Subject: Re: [RFC][PATCH 0/5] preempt_count rework
Date: Wed, 14 Aug 2013 18:04:16 +0200 [thread overview]
Message-ID: <20130814160416.GI24092@twins.programming.kicks-ass.net> (raw)
In-Reply-To: <520B8A81.1080405@zytor.com>
On Wed, Aug 14, 2013 at 06:47:45AM -0700, H. Peter Anvin wrote:
> I still don't see this using a decrement of the percpu variable
> anywhere. The C compiler doesn't know how to generate those, so if I'm
> not completely wet we will end up relying on sub_preempt_count()...
> which, because it relies on taking the address of the percpu variable
> will generate absolutely horrific code.
>
> On x86, you never want to take the address of a percpu variable if you
> can avoid it, as you end up generating code like:
>
> movq %fs:0,%rax
> subl $1,(%rax)
>
Urgh,. yes you're right. I keep forgetting GCC doesn't know how to merge
those :/
OK, so something like the below would cure the worst of that I suppose.
It compiles but doesn't boot; must've done something wrong.
Someone please look at it because my asm-foo blows. I pretty much
copy/pasted this from asm/percpu.h.
---
--- a/arch/x86/include/asm/preempt.h
+++ b/arch/x86/include/asm/preempt.h
@@ -20,6 +20,28 @@ static __always_inline int *preempt_coun
return &__raw_get_cpu_var(__preempt_count);
}
+#define __preempt_count_add(x) do { \
+ asm("addl %1," __percpu_arg(0) \
+ : "+m" (__preempt_count) \
+ : "ri" ((int)x) \
+ : "memory"); \
+} while (0)
+
+#define __preempt_count_sub(x) do { \
+ asm("subl %1," __percpu_arg(0) \
+ : "+m" (__preempt_count) \
+ : "ri" ((int)x) \
+ : "memory"); \
+} while (0)
+
+#define preempt_enable() do { \
+ asm("\nsubl $1," __percpu_arg(0) \
+ "\njnz 1f" \
+ "\ncall preempt_schedule" \
+ "\n1:" : "+m" (__preempt_count) \
+ : : "memory"); \
+} while (0)
+
/*
* must be macros to avoid header recursion hell
*/
--- a/include/asm-generic/preempt.h
+++ b/include/asm-generic/preempt.h
@@ -17,6 +17,9 @@ static __always_inline int *preempt_coun
return ¤t_thread_info()->preempt_count;
}
+#define __preempt_count_add(x) do { current_thread_info()->preempt_count += (x); } while (0)
+#define __preempt_count_sub(x) __preempt_count_add(-(x))
+
/*
* must be macros to avoid header recursion hell
*/
--- a/include/linux/preempt.h
+++ b/include/linux/preempt.h
@@ -45,8 +45,8 @@ static __always_inline bool test_preempt
extern void add_preempt_count(int val);
extern void sub_preempt_count(int val);
#else
-# define add_preempt_count(val) do { *preempt_count_ptr() += (val); } while (0)
-# define sub_preempt_count(val) do { *preempt_count_ptr() -= (val); } while (0)
+# define add_preempt_count(val) __preempt_count_add(val)
+# define sub_preempt_count(val) __preempt_count_sub(val)
#endif
#define inc_preempt_count() add_preempt_count(1)
@@ -101,17 +101,17 @@ do { \
#define preempt_enable_no_resched() sched_preempt_enable_no_resched()
+#ifndef preempt_enable
#define preempt_enable() \
do { \
preempt_enable_no_resched(); \
preempt_check_resched(); \
} while (0)
+#endif
/* For debugging and tracer internals only! */
-#define add_preempt_count_notrace(val) \
- do { *preempt_count_ptr() += (val); } while (0)
-#define sub_preempt_count_notrace(val) \
- do { *preempt_count_ptr() -= (val); } while (0)
+#define add_preempt_count_notrace(val) __preempt_count_add(val)
+#define sub_preempt_count_notrace(val) __preempt_count_sub(val)
#define inc_preempt_count_notrace() add_preempt_count_notrace(1)
#define dec_preempt_count_notrace() sub_preempt_count_notrace(1)
next prev parent reply other threads:[~2013-08-14 16:04 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-08-14 13:15 [RFC][PATCH 0/5] preempt_count rework Peter Zijlstra
2013-08-14 13:15 ` [RFC][PATCH 1/5] sched: Introduce preempt_count accessor functions Peter Zijlstra
2013-08-14 13:15 ` [RFC][PATCH 2/5] sched: Add NEED_RESCHED to the preempt_count Peter Zijlstra
2013-08-14 13:15 ` [RFC][PATCH 3/5] sched, arch: Create asm/preempt.h Peter Zijlstra
2013-08-14 13:15 ` [RFC][PATCH 4/5] sched: Create more preempt_count accessors Peter Zijlstra
2013-08-14 13:15 ` [RFC][PATCH 5/5] sched, x86: Provide a per-cpu preempt_count implementation Peter Zijlstra
2013-08-14 13:47 ` [RFC][PATCH 0/5] preempt_count rework H. Peter Anvin
2013-08-14 15:39 ` Mike Galbraith
2013-08-14 15:43 ` H. Peter Anvin
2013-08-15 9:01 ` Mike Galbraith
2013-08-14 16:06 ` Peter Zijlstra
2013-08-14 16:14 ` H. Peter Anvin
2013-08-14 16:52 ` Peter Zijlstra
2013-08-14 16:58 ` H. Peter Anvin
2013-08-14 16:04 ` Peter Zijlstra [this message]
2013-08-14 17:31 ` Peter Zijlstra
2013-08-14 16:48 ` Andi Kleen
2013-08-14 16:55 ` Peter Zijlstra
2013-08-14 17:12 ` Andi Kleen
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=20130814160416.GI24092@twins.programming.kicks-ass.net \
--to=peterz@infradead.org \
--cc=ak@linux.intel.com \
--cc=arjan@linux.intel.com \
--cc=bitbucket@online.de \
--cc=hpa@zytor.com \
--cc=linux-arch@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@kernel.org \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.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