mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH v18 0/10] Add 32 bit VDSO time function support
@ 2014-02-16 21:52 Stefani Seibold
  2014-02-16 21:52 ` [PATCH v18 01/10] Make vsyscall_gtod_data handling x86 generic Stefani Seibold
                   ` (8 more replies)
  0 siblings, 9 replies; 32+ messages in thread
From: Stefani Seibold @ 2014-02-16 21:52 UTC (permalink / raw)
  To: gregkh, linux-kernel, x86, tglx, mingo, hpa, ak, aarcange,
	john.stultz, luto, xemul, gorcunov, andriy.shevchenko
  Cc: Martin.Runge, Andreas.Brief, Stefani Seibold

This patch add the functions vdso_gettimeofday(), vdso_clock_gettime()
and vdso_time() to the 32 bit VDSO.

The reason to do this was to get a fast reliable time stamp. Many developers
uses TSC to get a fast time stamp, without knowing the pitfalls. VDSO
time functions a fast and a reliable way, because the kernel knows the
best time source and the P- and C-state of the CPU.

The helper library to use the VDSO functions can be download at
http://http://seibold.net/vdso.c
The libary is very small, only 228 lines of code. Compile it with
gcc -Wall -O3 -fpic vdso.c -lrt -shared -o libvdso.so
and use it with LD_PRELOAD=<path>/libvdso.so

This kind of helper must be integrated into glibc, for x86 64 bit and
PowerPC it is already there.

Some linux 32 bit kernel benchmark results (all measurements are in nano
seconds):

Intel(R) Celeron(TM) CPU 400MHz

Average time kernel call:
 gettimeofday(): 1039
 clock_gettime(): 1578
 time(): 526
Average time VDSO call:
 gettimeofday(): 378
 clock_gettime(): 303
 time(): 60

Celeron(R) Dual-Core CPU T3100 1.90GHz

Average time kernel call:
 gettimeofday(): 209
 clock_gettime(): 406
 time(): 135
Average time VDSO call:
 gettimeofday(): 51
 clock_gettime(): 43
 time(): 10

So you can see a performance increase between 4 and 13, depending on the
CPU and the function.

The address layout of the VDSO has changed, because there is no fixed
address space available on a x86 32 bit kernel, despite the name. Because
someone decided to add an offset to the __FIXADDR_TOP for virtualization.

Also the IA32 Emulation uses the whole 4 GB address space, so there is no
fixed address available.

This was the reason not depend on this kind of address and change the layout
of the VDSO. The VDSO for a 32 bit application has now three pages:

^ Higher Address
|
+----------------------------------------+
+ VDSO page (includes code) ro+x         +
+----------------------------------------+
+ VVAR page (export kernel variables) ro +
+----------------------------------------+
+ HPET page (mapped registers) ro 
+----------------------------------------+
|
^ Lower Address

The VDSO page for a 32 bit resided still on 0xffffe000, the the VVAR and
HPET page are mapped before.

In the non compat mode the VMA of the VDSO is now 3 pages for a 32 bit kernel.
So this decrease the available logical address room by 2 pages.

The patch is against kernel 3.14 (e7651b819e90da924991d727d3c007200a18670d)

Changelog:
25.11.2012 - first release and proof of concept for linux 3.4
11.12.2012 - Port to linux 3.7 and code cleanup
12.12.2012 - fixes suggested by Andy Lutomirski
           - fixes suggested by John Stultz
           - use call VDSO32_vsyscall instead of int 80
           - code cleanup
17.12.2012 - support for IA32_EMULATION, this includes
             - code cleanup
             - include cleanup to fix compile warnings and errors
             - move out seqcount from seqlock, enable use in VDSO
             - map FIXMAP and HPET into the 32 bit address space
18.12.2012 - split into separate patches
30.01.2014 - revamp the code
             - code clean up
             - VDSO layout changed
             - no fixed addresses
             - port to 3.14
01.02.2014 - code cleanup
02.02.2014 - code cleanup
             - split into more patches
             - use HPET_COUNTER instead of hard coded value
             - fix changelog to the right year ;-)
02.02.2014 - reverse the mapping, this make the new VDSO 32 bit support
             full compatible.
03.02.2014 - code cleanup
             - fix comment
             - fix ABI break in vdso32.lds.S
04.02.2014 - revamp IA32 emulation support
             - introduce VVAR macro
             - rearranged vsyscall_gtod_data struture for IA32 emulation support
             - code cleanup
05.02.2014 - revamp IA32 emulation support
             - replace seqcount_t by an unsigned, to make the vsyscall_gtod_data
               structure independed of kernel config and functions.
08.02.2014 - revamp IA32 emulation support
             - replace all internal structures by fix size elements
10.02.2014 - code cleanup
             - add commets
             - revamp inline assembly
12.02.2014 - add conditional fixmap of vvar and hpet pages for 32 bit kernel
14.02.2014 - fix CONFIG_PARAVIRT_CLOCK, which is not supported in 32 bit VDSO
15.02.2014 - fix tsc
             code cleanup
             tested make ARCH=i386 allyesconfig and make allyesconfig
16.02.2014 - code cleanup
             - fix all C=1 warnings, also some one not introduced by this patch
             - hack to fix C=1 32 bit VDSO spinlock for a 64 bit kernel
             - fix VDSO Makefile for newer gcc
             tested for gcc 4.3.4 and 4.8.1
             tested ARCH=i386 allyesconfig, defconfig and allmodconfig
             tested X86_64 allyesconfig, defconfig and allmodconfig

^ permalink raw reply	[flat|nested] 32+ messages in thread
* [PATCH v19 09/10] Add 32 bit VDSO time support for 64 bit kernel
@ 2014-02-17 10:03 Stefani Seibold
  2014-02-17 15:58 ` [tip:x86/vdso] x86, vdso: Add 32-bit VDSO time support for the 64-bit kernel tip-bot for Stefani Seibold
  2014-02-17 19:03 ` tip-bot for Stefani Seibold
  0 siblings, 2 replies; 32+ messages in thread
From: Stefani Seibold @ 2014-02-17 10:03 UTC (permalink / raw)
  To: gregkh, linux-kernel, x86, tglx, mingo, hpa, ak, aarcange,
	john.stultz, luto, xemul, gorcunov, andriy.shevchenko
  Cc: Martin.Runge, Andreas.Brief, Stefani Seibold

This patch add the VDSO time support for the IA32 Emulation Layer.

Due the nature of the kernel headers and the LP64 compiler where the
size of a long and a pointer differs against a 32 bit compiler, there
is some type hacking necessary for optimal performance.

The vsyscall_gtod_data struture must be a rearranged to serve 32- and
64-bit code access at the same time:

- The seqcount_t was replaced by an unsigned, this makes the
  vsyscall_gtod_data intedepend of kernel configuration and internal functions.
- All kernel internal structures are replaced by fix size elements
  which works for 32- and 64-bit access
- The inner struct clock was removed to pack the whole struct.

The "unsigned seq" would be handled by functions derivated from seqcount_t.

Signed-off-by: Stefani Seibold <stefani@seibold.net>
---
 arch/x86/include/asm/vgtod.h          | 71 ++++++++++++++++++++++------
 arch/x86/include/asm/vvar.h           |  5 ++
 arch/x86/kernel/vsyscall_gtod.c       | 34 ++++++++-----
 arch/x86/vdso/vclock_gettime.c        | 89 +++++++++++++++++++----------------
 arch/x86/vdso/vdso32/vclock_gettime.c | 18 +++++++
 5 files changed, 150 insertions(+), 67 deletions(-)

diff --git a/arch/x86/include/asm/vgtod.h b/arch/x86/include/asm/vgtod.h
index 46e24d3..3c3366c 100644
--- a/arch/x86/include/asm/vgtod.h
+++ b/arch/x86/include/asm/vgtod.h
@@ -1,30 +1,73 @@
 #ifndef _ASM_X86_VGTOD_H
 #define _ASM_X86_VGTOD_H
 
-#include <asm/vsyscall.h>
+#include <linux/compiler.h>
 #include <linux/clocksource.h>
 
+#ifdef BUILD_VDSO32_64
+typedef u64 gtod_long_t;
+#else
+typedef unsigned long gtod_long_t;
+#endif
+/*
+ * vsyscall_gtod_data will be accessed by 32 and 64 bit code at the same time
+ * so be carefull by modifying this structure.
+ */
 struct vsyscall_gtod_data {
-	seqcount_t	seq;
+	unsigned seq;
 
-	struct { /* extract of a clocksource struct */
-		int vclock_mode;
-		cycle_t	cycle_last;
-		cycle_t	mask;
-		u32	mult;
-		u32	shift;
-	} clock;
+	int vclock_mode;
+	cycle_t	cycle_last;
+	cycle_t	mask;
+	u32	mult;
+	u32	shift;
 
 	/* open coded 'struct timespec' */
-	time_t		wall_time_sec;
 	u64		wall_time_snsec;
+	gtod_long_t	wall_time_sec;
+	gtod_long_t	monotonic_time_sec;
 	u64		monotonic_time_snsec;
-	time_t		monotonic_time_sec;
+	gtod_long_t	wall_time_coarse_sec;
+	gtod_long_t	wall_time_coarse_nsec;
+	gtod_long_t	monotonic_time_coarse_sec;
+	gtod_long_t	monotonic_time_coarse_nsec;
 
-	struct timezone sys_tz;
-	struct timespec wall_time_coarse;
-	struct timespec monotonic_time_coarse;
+	int		tz_minuteswest;
+	int		tz_dsttime;
 };
 extern struct vsyscall_gtod_data vsyscall_gtod_data;
 
+static inline unsigned gtod_read_begin(const struct vsyscall_gtod_data *s)
+{
+	unsigned ret;
+
+repeat:
+	ret = ACCESS_ONCE(s->seq);
+	if (unlikely(ret & 1)) {
+		cpu_relax();
+		goto repeat;
+	}
+	smp_rmb();
+	return ret;
+}
+
+static inline int gtod_read_retry(const struct vsyscall_gtod_data *s,
+					unsigned start)
+{
+	smp_rmb();
+	return unlikely(s->seq != start);
+}
+
+static inline void gtod_write_begin(struct vsyscall_gtod_data *s)
+{
+	++s->seq;
+	smp_wmb();
+}
+
+static inline void gtod_write_end(struct vsyscall_gtod_data *s)
+{
+	smp_wmb();
+	++s->seq;
+}
+
 #endif /* _ASM_X86_VGTOD_H */
diff --git a/arch/x86/include/asm/vvar.h b/arch/x86/include/asm/vvar.h
index 52c79ff..081d909 100644
--- a/arch/x86/include/asm/vvar.h
+++ b/arch/x86/include/asm/vvar.h
@@ -16,6 +16,9 @@
  * you mess up, the linker will catch it.)
  */
 
+#ifndef _ASM_X86_VVAR_H
+#define _ASM_X86_VVAR_H
+
 #if defined(__VVAR_KERNEL_LDS)
 
 /* The kernel linker script defines its own magic to put vvars in the
@@ -64,3 +67,5 @@ DECLARE_VVAR(16, int, vgetcpu_mode)
 DECLARE_VVAR(128, struct vsyscall_gtod_data, vsyscall_gtod_data)
 
 #undef DECLARE_VVAR
+
+#endif
diff --git a/arch/x86/kernel/vsyscall_gtod.c b/arch/x86/kernel/vsyscall_gtod.c
index 91862a4..eacd12f 100644
--- a/arch/x86/kernel/vsyscall_gtod.c
+++ b/arch/x86/kernel/vsyscall_gtod.c
@@ -4,6 +4,7 @@
  *
  *  Modified for x86 32 bit architecture by
  *  Stefani Seibold <stefani@seibold.net>
+ *  sponsored by Rohde & Schwarz GmbH & Co. KG Munich/Germany
  *
  *  Thanks to hpa@transmeta.com for some useful hint.
  *  Special thanks to Ingo Molnar for his early experience with
@@ -13,26 +14,28 @@
 
 #include <linux/timekeeper_internal.h>
 #include <asm/vgtod.h>
+#include <asm/vvar.h>
 
 DEFINE_VVAR(struct vsyscall_gtod_data, vsyscall_gtod_data);
 
 void update_vsyscall_tz(void)
 {
-	vsyscall_gtod_data.sys_tz = sys_tz;
+	vsyscall_gtod_data.tz_minuteswest = sys_tz.tz_minuteswest;
+	vsyscall_gtod_data.tz_dsttime = sys_tz.tz_dsttime;
 }
 
 void update_vsyscall(struct timekeeper *tk)
 {
 	struct vsyscall_gtod_data *vdata = &vsyscall_gtod_data;
 
-	write_seqcount_begin(&vdata->seq);
+	gtod_write_begin(vdata);
 
 	/* copy vsyscall data */
-	vdata->clock.vclock_mode	= tk->clock->archdata.vclock_mode;
-	vdata->clock.cycle_last		= tk->clock->cycle_last;
-	vdata->clock.mask		= tk->clock->mask;
-	vdata->clock.mult		= tk->mult;
-	vdata->clock.shift		= tk->shift;
+	vdata->vclock_mode	= tk->clock->archdata.vclock_mode;
+	vdata->cycle_last	= tk->clock->cycle_last;
+	vdata->mask		= tk->clock->mask;
+	vdata->mult		= tk->mult;
+	vdata->shift		= tk->shift;
 
 	vdata->wall_time_sec		= tk->xtime_sec;
 	vdata->wall_time_snsec		= tk->xtime_nsec;
@@ -49,12 +52,19 @@ void update_vsyscall(struct timekeeper *tk)
 		vdata->monotonic_time_sec++;
 	}
 
-	vdata->wall_time_coarse.tv_sec	= tk->xtime_sec;
-	vdata->wall_time_coarse.tv_nsec	= (long)(tk->xtime_nsec >> tk->shift);
+	vdata->wall_time_coarse_sec	= tk->xtime_sec;
+	vdata->wall_time_coarse_nsec	= (long)(tk->xtime_nsec >> tk->shift);
 
-	vdata->monotonic_time_coarse	= timespec_add(vdata->wall_time_coarse,
-							tk->wall_to_monotonic);
+	vdata->monotonic_time_coarse_sec =
+		vdata->wall_time_coarse_sec + tk->wall_to_monotonic.tv_sec;
+	vdata->monotonic_time_coarse_nsec =
+		vdata->wall_time_coarse_nsec + tk->wall_to_monotonic.tv_nsec;
 
-	write_seqcount_end(&vdata->seq);
+	while (vdata->monotonic_time_coarse_nsec >= NSEC_PER_SEC) {
+		vdata->monotonic_time_coarse_nsec -= NSEC_PER_SEC;
+		vdata->monotonic_time_coarse_sec++;
+	}
+
+	gtod_write_end(vdata);
 }
 
diff --git a/arch/x86/vdso/vclock_gettime.c b/arch/x86/vdso/vclock_gettime.c
index fcbc974..b82e7df 100644
--- a/arch/x86/vdso/vclock_gettime.c
+++ b/arch/x86/vdso/vclock_gettime.c
@@ -14,16 +14,14 @@
 /* Disable profiling for userspace code: */
 #define DISABLE_BRANCH_PROFILING
 
-#include <linux/kernel.h>
 #include <uapi/linux/time.h>
-#include <linux/string.h>
-#include <asm/vsyscall.h>
-#include <asm/fixmap.h>
 #include <asm/vgtod.h>
 #include <asm/hpet.h>
+#include <asm/vvar.h>
 #include <asm/unistd.h>
-#include <asm/io.h>
-#include <asm/pvclock.h>
+#include <asm/msr.h>
+#include <linux/math64.h>
+#include <linux/time.h>
 
 #define gtod (&VVAR(vsyscall_gtod_data))
 
@@ -31,11 +29,21 @@ extern int __vdso_clock_gettime(clockid_t clock, struct timespec *ts);
 extern int __vdso_gettimeofday(struct timeval *tv, struct timezone *tz);
 extern time_t __vdso_time(time_t *t);
 
+static inline u32 read_hpet_counter(const volatile void *addr)
+{
+	return *(const volatile u32 *) (addr + HPET_COUNTER);
+}
+
 #ifndef BUILD_VDSO32
 
+#include <linux/kernel.h>
+#include <asm/vsyscall.h>
+#include <asm/fixmap.h>
+#include <asm/pvclock.h>
+
 static notrace cycle_t vread_hpet(void)
 {
-	return readl((const void __iomem *)fix_to_virt(VSYSCALL_HPET) + HPET_COUNTER);
+	return read_hpet_counter((const void *)fix_to_virt(VSYSCALL_HPET));
 }
 
 notrace static long vdso_fallback_gettime(long clock, struct timespec *ts)
@@ -116,7 +124,7 @@ static notrace cycle_t vread_pvclock(int *mode)
 		*mode = VCLOCK_NONE;
 
 	/* refer to tsc.c read_tsc() comment for rationale */
-	last = gtod->clock.cycle_last;
+	last = gtod->cycle_last;
 
 	if (likely(ret >= last))
 		return ret;
@@ -133,7 +141,7 @@ extern u8 hpet_page
 #ifdef CONFIG_HPET_TIMER
 static notrace cycle_t vread_hpet(void)
 {
-	return readl((const void __iomem *)(&hpet_page + HPET_COUNTER));
+	return read_hpet_counter((const void *)(&hpet_page));
 }
 #endif
 
@@ -193,7 +201,7 @@ notrace static cycle_t vread_tsc(void)
 	rdtsc_barrier();
 	ret = (cycle_t)__native_read_tsc();
 
-	last = gtod->clock.cycle_last;
+	last = gtod->cycle_last;
 
 	if (likely(ret >= last))
 		return ret;
@@ -214,20 +222,21 @@ notrace static inline u64 vgetsns(int *mode)
 {
 	u64 v;
 	cycles_t cycles;
-	if (gtod->clock.vclock_mode == VCLOCK_TSC)
+
+	if (gtod->vclock_mode == VCLOCK_TSC)
 		cycles = vread_tsc();
 #ifdef CONFIG_HPET_TIMER
-	else if (gtod->clock.vclock_mode == VCLOCK_HPET)
+	else if (gtod->vclock_mode == VCLOCK_HPET)
 		cycles = vread_hpet();
 #endif
 #ifdef CONFIG_PARAVIRT_CLOCK
-	else if (gtod->clock.vclock_mode == VCLOCK_PVCLOCK)
+	else if (gtod->vclock_mode == VCLOCK_PVCLOCK)
 		cycles = vread_pvclock(mode);
 #endif
 	else
 		return 0;
-	v = (cycles - gtod->clock.cycle_last) & gtod->clock.mask;
-	return v * gtod->clock.mult;
+	v = (cycles - gtod->cycle_last) & gtod->mask;
+	return v * gtod->mult;
 }
 
 /* Code size doesn't matter (vdso is 4k anyway) and this is faster. */
@@ -237,17 +246,18 @@ notrace static int __always_inline do_realtime(struct timespec *ts)
 	u64 ns;
 	int mode;
 
-	ts->tv_nsec = 0;
 	do {
-		seq = raw_read_seqcount_begin(&gtod->seq);
-		mode = gtod->clock.vclock_mode;
+		seq = gtod_read_begin(gtod);
+		mode = gtod->vclock_mode;
 		ts->tv_sec = gtod->wall_time_sec;
 		ns = gtod->wall_time_snsec;
 		ns += vgetsns(&mode);
-		ns >>= gtod->clock.shift;
-	} while (unlikely(read_seqcount_retry(&gtod->seq, seq)));
+		ns >>= gtod->shift;
+	} while (unlikely(gtod_read_retry(gtod, seq)));
+
+	ts->tv_sec += __iter_div_u64_rem(ns, NSEC_PER_SEC, &ns);
+	ts->tv_nsec = ns;
 
-	timespec_add_ns(ts, ns);
 	return mode;
 }
 
@@ -257,16 +267,17 @@ notrace static int do_monotonic(struct timespec *ts)
 	u64 ns;
 	int mode;
 
-	ts->tv_nsec = 0;
 	do {
-		seq = raw_read_seqcount_begin(&gtod->seq);
-		mode = gtod->clock.vclock_mode;
+		seq = gtod_read_begin(gtod);
+		mode = gtod->vclock_mode;
 		ts->tv_sec = gtod->monotonic_time_sec;
 		ns = gtod->monotonic_time_snsec;
 		ns += vgetsns(&mode);
-		ns >>= gtod->clock.shift;
-	} while (unlikely(read_seqcount_retry(&gtod->seq, seq)));
-	timespec_add_ns(ts, ns);
+		ns >>= gtod->shift;
+	} while (unlikely(gtod_read_retry(gtod, seq)));
+
+	ts->tv_sec += __iter_div_u64_rem(ns, NSEC_PER_SEC, &ns);
+	ts->tv_nsec = ns;
 
 	return mode;
 }
@@ -275,20 +286,20 @@ notrace static void do_realtime_coarse(struct timespec *ts)
 {
 	unsigned long seq;
 	do {
-		seq = raw_read_seqcount_begin(&gtod->seq);
-		ts->tv_sec = gtod->wall_time_coarse.tv_sec;
-		ts->tv_nsec = gtod->wall_time_coarse.tv_nsec;
-	} while (unlikely(read_seqcount_retry(&gtod->seq, seq)));
+		seq = gtod_read_begin(gtod);
+		ts->tv_sec = gtod->wall_time_coarse_sec;
+		ts->tv_nsec = gtod->wall_time_coarse_nsec;
+	} while (unlikely(gtod_read_retry(gtod, seq)));
 }
 
 notrace static void do_monotonic_coarse(struct timespec *ts)
 {
 	unsigned long seq;
 	do {
-		seq = raw_read_seqcount_begin(&gtod->seq);
-		ts->tv_sec = gtod->monotonic_time_coarse.tv_sec;
-		ts->tv_nsec = gtod->monotonic_time_coarse.tv_nsec;
-	} while (unlikely(read_seqcount_retry(&gtod->seq, seq)));
+		seq = gtod_read_begin(gtod);
+		ts->tv_sec = gtod->monotonic_time_coarse_sec;
+		ts->tv_nsec = gtod->monotonic_time_coarse_nsec;
+	} while (unlikely(gtod_read_retry(gtod, seq)));
 }
 
 notrace int __vdso_clock_gettime(clockid_t clock, struct timespec *ts)
@@ -322,17 +333,13 @@ int clock_gettime(clockid_t, struct timespec *)
 notrace int __vdso_gettimeofday(struct timeval *tv, struct timezone *tz)
 {
 	if (likely(tv != NULL)) {
-		BUILD_BUG_ON(offsetof(struct timeval, tv_usec) !=
-			     offsetof(struct timespec, tv_nsec) ||
-			     sizeof(*tv) != sizeof(struct timespec));
 		if (unlikely(do_realtime((struct timespec *)tv) == VCLOCK_NONE))
 			return vdso_fallback_gtod(tv, tz);
 		tv->tv_usec /= 1000;
 	}
 	if (unlikely(tz != NULL)) {
-		/* Avoid memcpy. Some old compilers fail to inline it */
-		tz->tz_minuteswest = gtod->sys_tz.tz_minuteswest;
-		tz->tz_dsttime = gtod->sys_tz.tz_dsttime;
+		tz->tz_minuteswest = gtod->tz_minuteswest;
+		tz->tz_dsttime = gtod->tz_dsttime;
 	}
 
 	return 0;
diff --git a/arch/x86/vdso/vdso32/vclock_gettime.c b/arch/x86/vdso/vdso32/vclock_gettime.c
index d9bb9fc..d7c7fd7 100644
--- a/arch/x86/vdso/vdso32/vclock_gettime.c
+++ b/arch/x86/vdso/vdso32/vclock_gettime.c
@@ -1,4 +1,22 @@
 #define BUILD_VDSO32
 
+#ifdef CONFIG_X86_64
+
+/*
+ * in case of a 32 bit VDSO for a 64 bit kernel fake a 32 bit kernel
+ * configuration
+ */
+#undef CONFIG_64BIT
+#undef CONFIG_X86_64
+#undef CONFIG_ILLEGAL_POINTER_VALUE
+
+#define CONFIG_X86_32 1
+#define CONFIG_PAGE_OFFSET 0
+#define CONFIG_ILLEGAL_POINTER_VALUE 0
+
+#define BUILD_VDSO32_64
+
+#endif
+
 #include "../vclock_gettime.c"
 
-- 
1.8.5.5


^ permalink raw reply	[flat|nested] 32+ messages in thread
* [PATCH v20 09/10] Add 32 bit VDSO time support for 64 bit kernel
@ 2014-02-19  9:09 Stefani Seibold
  2014-02-19 22:30 ` [tip:x86/vdso] x86, vdso: Add 32-bit VDSO time support for the 64-bit kernel tip-bot for Stefani Seibold
  0 siblings, 1 reply; 32+ messages in thread
From: Stefani Seibold @ 2014-02-19  9:09 UTC (permalink / raw)
  To: gregkh, linux-kernel, x86, tglx, mingo, hpa, ak, aarcange,
	john.stultz, luto, xemul, gorcunov, andriy.shevchenko
  Cc: Martin.Runge, Andreas.Brief, Stefani Seibold

This patch add the VDSO time support for the IA32 Emulation Layer.

Due the nature of the kernel headers and the LP64 compiler where the
size of a long and a pointer differs against a 32 bit compiler, there
is some type hacking necessary for optimal performance.

The vsyscall_gtod_data struture must be a rearranged to serve 32- and
64-bit code access at the same time:

- The seqcount_t was replaced by an unsigned, this makes the
  vsyscall_gtod_data intedepend of kernel configuration and internal functions.
- All kernel internal structures are replaced by fix size elements
  which works for 32- and 64-bit access
- The inner struct clock was removed to pack the whole struct.

The "unsigned seq" would be handled by functions derivated from seqcount_t.

Signed-off-by: Stefani Seibold <stefani@seibold.net>
---
 arch/x86/include/asm/vgtod.h          | 71 +++++++++++++++++++++------
 arch/x86/include/asm/vvar.h           |  5 ++
 arch/x86/kernel/vsyscall_gtod.c       | 34 ++++++++-----
 arch/x86/vdso/vclock_gettime.c        | 91 +++++++++++++++++++----------------
 arch/x86/vdso/vdso32/vclock_gettime.c | 21 ++++++++
 5 files changed, 155 insertions(+), 67 deletions(-)

diff --git a/arch/x86/include/asm/vgtod.h b/arch/x86/include/asm/vgtod.h
index 46e24d3..3c3366c 100644
--- a/arch/x86/include/asm/vgtod.h
+++ b/arch/x86/include/asm/vgtod.h
@@ -1,30 +1,73 @@
 #ifndef _ASM_X86_VGTOD_H
 #define _ASM_X86_VGTOD_H
 
-#include <asm/vsyscall.h>
+#include <linux/compiler.h>
 #include <linux/clocksource.h>
 
+#ifdef BUILD_VDSO32_64
+typedef u64 gtod_long_t;
+#else
+typedef unsigned long gtod_long_t;
+#endif
+/*
+ * vsyscall_gtod_data will be accessed by 32 and 64 bit code at the same time
+ * so be carefull by modifying this structure.
+ */
 struct vsyscall_gtod_data {
-	seqcount_t	seq;
+	unsigned seq;
 
-	struct { /* extract of a clocksource struct */
-		int vclock_mode;
-		cycle_t	cycle_last;
-		cycle_t	mask;
-		u32	mult;
-		u32	shift;
-	} clock;
+	int vclock_mode;
+	cycle_t	cycle_last;
+	cycle_t	mask;
+	u32	mult;
+	u32	shift;
 
 	/* open coded 'struct timespec' */
-	time_t		wall_time_sec;
 	u64		wall_time_snsec;
+	gtod_long_t	wall_time_sec;
+	gtod_long_t	monotonic_time_sec;
 	u64		monotonic_time_snsec;
-	time_t		monotonic_time_sec;
+	gtod_long_t	wall_time_coarse_sec;
+	gtod_long_t	wall_time_coarse_nsec;
+	gtod_long_t	monotonic_time_coarse_sec;
+	gtod_long_t	monotonic_time_coarse_nsec;
 
-	struct timezone sys_tz;
-	struct timespec wall_time_coarse;
-	struct timespec monotonic_time_coarse;
+	int		tz_minuteswest;
+	int		tz_dsttime;
 };
 extern struct vsyscall_gtod_data vsyscall_gtod_data;
 
+static inline unsigned gtod_read_begin(const struct vsyscall_gtod_data *s)
+{
+	unsigned ret;
+
+repeat:
+	ret = ACCESS_ONCE(s->seq);
+	if (unlikely(ret & 1)) {
+		cpu_relax();
+		goto repeat;
+	}
+	smp_rmb();
+	return ret;
+}
+
+static inline int gtod_read_retry(const struct vsyscall_gtod_data *s,
+					unsigned start)
+{
+	smp_rmb();
+	return unlikely(s->seq != start);
+}
+
+static inline void gtod_write_begin(struct vsyscall_gtod_data *s)
+{
+	++s->seq;
+	smp_wmb();
+}
+
+static inline void gtod_write_end(struct vsyscall_gtod_data *s)
+{
+	smp_wmb();
+	++s->seq;
+}
+
 #endif /* _ASM_X86_VGTOD_H */
diff --git a/arch/x86/include/asm/vvar.h b/arch/x86/include/asm/vvar.h
index 52c79ff..081d909 100644
--- a/arch/x86/include/asm/vvar.h
+++ b/arch/x86/include/asm/vvar.h
@@ -16,6 +16,9 @@
  * you mess up, the linker will catch it.)
  */
 
+#ifndef _ASM_X86_VVAR_H
+#define _ASM_X86_VVAR_H
+
 #if defined(__VVAR_KERNEL_LDS)
 
 /* The kernel linker script defines its own magic to put vvars in the
@@ -64,3 +67,5 @@ DECLARE_VVAR(16, int, vgetcpu_mode)
 DECLARE_VVAR(128, struct vsyscall_gtod_data, vsyscall_gtod_data)
 
 #undef DECLARE_VVAR
+
+#endif
diff --git a/arch/x86/kernel/vsyscall_gtod.c b/arch/x86/kernel/vsyscall_gtod.c
index 91862a4..eacd12f 100644
--- a/arch/x86/kernel/vsyscall_gtod.c
+++ b/arch/x86/kernel/vsyscall_gtod.c
@@ -4,6 +4,7 @@
  *
  *  Modified for x86 32 bit architecture by
  *  Stefani Seibold <stefani@seibold.net>
+ *  sponsored by Rohde & Schwarz GmbH & Co. KG Munich/Germany
  *
  *  Thanks to hpa@transmeta.com for some useful hint.
  *  Special thanks to Ingo Molnar for his early experience with
@@ -13,26 +14,28 @@
 
 #include <linux/timekeeper_internal.h>
 #include <asm/vgtod.h>
+#include <asm/vvar.h>
 
 DEFINE_VVAR(struct vsyscall_gtod_data, vsyscall_gtod_data);
 
 void update_vsyscall_tz(void)
 {
-	vsyscall_gtod_data.sys_tz = sys_tz;
+	vsyscall_gtod_data.tz_minuteswest = sys_tz.tz_minuteswest;
+	vsyscall_gtod_data.tz_dsttime = sys_tz.tz_dsttime;
 }
 
 void update_vsyscall(struct timekeeper *tk)
 {
 	struct vsyscall_gtod_data *vdata = &vsyscall_gtod_data;
 
-	write_seqcount_begin(&vdata->seq);
+	gtod_write_begin(vdata);
 
 	/* copy vsyscall data */
-	vdata->clock.vclock_mode	= tk->clock->archdata.vclock_mode;
-	vdata->clock.cycle_last		= tk->clock->cycle_last;
-	vdata->clock.mask		= tk->clock->mask;
-	vdata->clock.mult		= tk->mult;
-	vdata->clock.shift		= tk->shift;
+	vdata->vclock_mode	= tk->clock->archdata.vclock_mode;
+	vdata->cycle_last	= tk->clock->cycle_last;
+	vdata->mask		= tk->clock->mask;
+	vdata->mult		= tk->mult;
+	vdata->shift		= tk->shift;
 
 	vdata->wall_time_sec		= tk->xtime_sec;
 	vdata->wall_time_snsec		= tk->xtime_nsec;
@@ -49,12 +52,19 @@ void update_vsyscall(struct timekeeper *tk)
 		vdata->monotonic_time_sec++;
 	}
 
-	vdata->wall_time_coarse.tv_sec	= tk->xtime_sec;
-	vdata->wall_time_coarse.tv_nsec	= (long)(tk->xtime_nsec >> tk->shift);
+	vdata->wall_time_coarse_sec	= tk->xtime_sec;
+	vdata->wall_time_coarse_nsec	= (long)(tk->xtime_nsec >> tk->shift);
 
-	vdata->monotonic_time_coarse	= timespec_add(vdata->wall_time_coarse,
-							tk->wall_to_monotonic);
+	vdata->monotonic_time_coarse_sec =
+		vdata->wall_time_coarse_sec + tk->wall_to_monotonic.tv_sec;
+	vdata->monotonic_time_coarse_nsec =
+		vdata->wall_time_coarse_nsec + tk->wall_to_monotonic.tv_nsec;
 
-	write_seqcount_end(&vdata->seq);
+	while (vdata->monotonic_time_coarse_nsec >= NSEC_PER_SEC) {
+		vdata->monotonic_time_coarse_nsec -= NSEC_PER_SEC;
+		vdata->monotonic_time_coarse_sec++;
+	}
+
+	gtod_write_end(vdata);
 }
 
diff --git a/arch/x86/vdso/vclock_gettime.c b/arch/x86/vdso/vclock_gettime.c
index fcbc974..828888e 100644
--- a/arch/x86/vdso/vclock_gettime.c
+++ b/arch/x86/vdso/vclock_gettime.c
@@ -14,16 +14,14 @@
 /* Disable profiling for userspace code: */
 #define DISABLE_BRANCH_PROFILING
 
-#include <linux/kernel.h>
 #include <uapi/linux/time.h>
-#include <linux/string.h>
-#include <asm/vsyscall.h>
-#include <asm/fixmap.h>
 #include <asm/vgtod.h>
 #include <asm/hpet.h>
+#include <asm/vvar.h>
 #include <asm/unistd.h>
-#include <asm/io.h>
-#include <asm/pvclock.h>
+#include <asm/msr.h>
+#include <linux/math64.h>
+#include <linux/time.h>
 
 #define gtod (&VVAR(vsyscall_gtod_data))
 
@@ -31,11 +29,23 @@ extern int __vdso_clock_gettime(clockid_t clock, struct timespec *ts);
 extern int __vdso_gettimeofday(struct timeval *tv, struct timezone *tz);
 extern time_t __vdso_time(time_t *t);
 
+#ifdef CONFIG_HPET_TIMER
+static inline u32 read_hpet_counter(const volatile void *addr)
+{
+	return *(const volatile u32 *) (addr + HPET_COUNTER);
+}
+#endif
+
 #ifndef BUILD_VDSO32
 
+#include <linux/kernel.h>
+#include <asm/vsyscall.h>
+#include <asm/fixmap.h>
+#include <asm/pvclock.h>
+
 static notrace cycle_t vread_hpet(void)
 {
-	return readl((const void __iomem *)fix_to_virt(VSYSCALL_HPET) + HPET_COUNTER);
+	return read_hpet_counter((const void *)fix_to_virt(VSYSCALL_HPET));
 }
 
 notrace static long vdso_fallback_gettime(long clock, struct timespec *ts)
@@ -116,7 +126,7 @@ static notrace cycle_t vread_pvclock(int *mode)
 		*mode = VCLOCK_NONE;
 
 	/* refer to tsc.c read_tsc() comment for rationale */
-	last = gtod->clock.cycle_last;
+	last = gtod->cycle_last;
 
 	if (likely(ret >= last))
 		return ret;
@@ -133,7 +143,7 @@ extern u8 hpet_page
 #ifdef CONFIG_HPET_TIMER
 static notrace cycle_t vread_hpet(void)
 {
-	return readl((const void __iomem *)(&hpet_page + HPET_COUNTER));
+	return read_hpet_counter((const void *)(&hpet_page));
 }
 #endif
 
@@ -193,7 +203,7 @@ notrace static cycle_t vread_tsc(void)
 	rdtsc_barrier();
 	ret = (cycle_t)__native_read_tsc();
 
-	last = gtod->clock.cycle_last;
+	last = gtod->cycle_last;
 
 	if (likely(ret >= last))
 		return ret;
@@ -214,20 +224,21 @@ notrace static inline u64 vgetsns(int *mode)
 {
 	u64 v;
 	cycles_t cycles;
-	if (gtod->clock.vclock_mode == VCLOCK_TSC)
+
+	if (gtod->vclock_mode == VCLOCK_TSC)
 		cycles = vread_tsc();
 #ifdef CONFIG_HPET_TIMER
-	else if (gtod->clock.vclock_mode == VCLOCK_HPET)
+	else if (gtod->vclock_mode == VCLOCK_HPET)
 		cycles = vread_hpet();
 #endif
 #ifdef CONFIG_PARAVIRT_CLOCK
-	else if (gtod->clock.vclock_mode == VCLOCK_PVCLOCK)
+	else if (gtod->vclock_mode == VCLOCK_PVCLOCK)
 		cycles = vread_pvclock(mode);
 #endif
 	else
 		return 0;
-	v = (cycles - gtod->clock.cycle_last) & gtod->clock.mask;
-	return v * gtod->clock.mult;
+	v = (cycles - gtod->cycle_last) & gtod->mask;
+	return v * gtod->mult;
 }
 
 /* Code size doesn't matter (vdso is 4k anyway) and this is faster. */
@@ -237,17 +248,18 @@ notrace static int __always_inline do_realtime(struct timespec *ts)
 	u64 ns;
 	int mode;
 
-	ts->tv_nsec = 0;
 	do {
-		seq = raw_read_seqcount_begin(&gtod->seq);
-		mode = gtod->clock.vclock_mode;
+		seq = gtod_read_begin(gtod);
+		mode = gtod->vclock_mode;
 		ts->tv_sec = gtod->wall_time_sec;
 		ns = gtod->wall_time_snsec;
 		ns += vgetsns(&mode);
-		ns >>= gtod->clock.shift;
-	} while (unlikely(read_seqcount_retry(&gtod->seq, seq)));
+		ns >>= gtod->shift;
+	} while (unlikely(gtod_read_retry(gtod, seq)));
+
+	ts->tv_sec += __iter_div_u64_rem(ns, NSEC_PER_SEC, &ns);
+	ts->tv_nsec = ns;
 
-	timespec_add_ns(ts, ns);
 	return mode;
 }
 
@@ -257,16 +269,17 @@ notrace static int do_monotonic(struct timespec *ts)
 	u64 ns;
 	int mode;
 
-	ts->tv_nsec = 0;
 	do {
-		seq = raw_read_seqcount_begin(&gtod->seq);
-		mode = gtod->clock.vclock_mode;
+		seq = gtod_read_begin(gtod);
+		mode = gtod->vclock_mode;
 		ts->tv_sec = gtod->monotonic_time_sec;
 		ns = gtod->monotonic_time_snsec;
 		ns += vgetsns(&mode);
-		ns >>= gtod->clock.shift;
-	} while (unlikely(read_seqcount_retry(&gtod->seq, seq)));
-	timespec_add_ns(ts, ns);
+		ns >>= gtod->shift;
+	} while (unlikely(gtod_read_retry(gtod, seq)));
+
+	ts->tv_sec += __iter_div_u64_rem(ns, NSEC_PER_SEC, &ns);
+	ts->tv_nsec = ns;
 
 	return mode;
 }
@@ -275,20 +288,20 @@ notrace static void do_realtime_coarse(struct timespec *ts)
 {
 	unsigned long seq;
 	do {
-		seq = raw_read_seqcount_begin(&gtod->seq);
-		ts->tv_sec = gtod->wall_time_coarse.tv_sec;
-		ts->tv_nsec = gtod->wall_time_coarse.tv_nsec;
-	} while (unlikely(read_seqcount_retry(&gtod->seq, seq)));
+		seq = gtod_read_begin(gtod);
+		ts->tv_sec = gtod->wall_time_coarse_sec;
+		ts->tv_nsec = gtod->wall_time_coarse_nsec;
+	} while (unlikely(gtod_read_retry(gtod, seq)));
 }
 
 notrace static void do_monotonic_coarse(struct timespec *ts)
 {
 	unsigned long seq;
 	do {
-		seq = raw_read_seqcount_begin(&gtod->seq);
-		ts->tv_sec = gtod->monotonic_time_coarse.tv_sec;
-		ts->tv_nsec = gtod->monotonic_time_coarse.tv_nsec;
-	} while (unlikely(read_seqcount_retry(&gtod->seq, seq)));
+		seq = gtod_read_begin(gtod);
+		ts->tv_sec = gtod->monotonic_time_coarse_sec;
+		ts->tv_nsec = gtod->monotonic_time_coarse_nsec;
+	} while (unlikely(gtod_read_retry(gtod, seq)));
 }
 
 notrace int __vdso_clock_gettime(clockid_t clock, struct timespec *ts)
@@ -322,17 +335,13 @@ int clock_gettime(clockid_t, struct timespec *)
 notrace int __vdso_gettimeofday(struct timeval *tv, struct timezone *tz)
 {
 	if (likely(tv != NULL)) {
-		BUILD_BUG_ON(offsetof(struct timeval, tv_usec) !=
-			     offsetof(struct timespec, tv_nsec) ||
-			     sizeof(*tv) != sizeof(struct timespec));
 		if (unlikely(do_realtime((struct timespec *)tv) == VCLOCK_NONE))
 			return vdso_fallback_gtod(tv, tz);
 		tv->tv_usec /= 1000;
 	}
 	if (unlikely(tz != NULL)) {
-		/* Avoid memcpy. Some old compilers fail to inline it */
-		tz->tz_minuteswest = gtod->sys_tz.tz_minuteswest;
-		tz->tz_dsttime = gtod->sys_tz.tz_dsttime;
+		tz->tz_minuteswest = gtod->tz_minuteswest;
+		tz->tz_dsttime = gtod->tz_dsttime;
 	}
 
 	return 0;
diff --git a/arch/x86/vdso/vdso32/vclock_gettime.c b/arch/x86/vdso/vdso32/vclock_gettime.c
index d9bb9fc..867c21d 100644
--- a/arch/x86/vdso/vdso32/vclock_gettime.c
+++ b/arch/x86/vdso/vdso32/vclock_gettime.c
@@ -1,4 +1,25 @@
 #define BUILD_VDSO32
 
+#ifdef CONFIG_X86_64
+
+/*
+ * in case of a 32 bit VDSO for a 64 bit kernel fake a 32 bit kernel
+ * configuration
+ */
+#undef CONFIG_64BIT
+#undef CONFIG_X86_64
+#undef CONFIG_ILLEGAL_POINTER_VALUE
+#undef CONFIG_SPARSEMEM_VMEMMAP
+#undef CONFIG_NR_CPUS
+
+#define CONFIG_X86_32 1
+#define CONFIG_PAGE_OFFSET 0
+#define CONFIG_ILLEGAL_POINTER_VALUE 0
+#define CONFIG_NR_CPUS 1
+
+#define BUILD_VDSO32_64
+
+#endif
+
 #include "../vclock_gettime.c"
 
-- 
1.8.5.5


^ permalink raw reply	[flat|nested] 32+ messages in thread
end of thread, other threads:[~2014-02-20  4:52 UTC | newest]

Thread overview: 32+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-02-16 21:52 [PATCH v18 0/10] Add 32 bit VDSO time function support Stefani Seibold
2014-02-16 21:52 ` [PATCH v18 01/10] Make vsyscall_gtod_data handling x86 generic Stefani Seibold
2014-02-17  0:52   ` [tip:x86/vdso] x86, vdso: " tip-bot for Stefani Seibold
2014-02-16 21:52 ` [PATCH v18 02/10] Add new func _install_special_mapping() to mmap.c Stefani Seibold
2014-02-17  0:52   ` [tip:x86/vdso] mm: " tip-bot for Stefani Seibold
2014-02-20  4:51   ` [tip:x86/vdso] mm: Clean up style in install_special_mapping() tip-bot for H. Peter Anvin
2014-02-16 21:52 ` [PATCH v18 03/10] revamp vclock_gettime.c Stefani Seibold
2014-02-17  0:52   ` [tip:x86/vdso] x86, vdso: Revamp vclock_gettime.c tip-bot for Stefani Seibold
2014-02-16 21:52 ` [PATCH v18 04/10] vclock_gettime.c __vdso_clock_gettime cleanup Stefani Seibold
2014-02-17  0:53   ` [tip:x86/vdso] x86, vdso: " tip-bot for Stefani Seibold
2014-02-16 21:52 ` [PATCH v18 05/10] replace VVAR(vsyscall_gtod_data) by gtod macro Stefani Seibold
2014-02-17  0:53   ` [tip:x86/vdso] x86, vdso: Replace VVAR(vsyscall_gtod_data) by the " tip-bot for Stefani Seibold
2014-02-16 21:52 ` [PATCH v18 06/10] cleanup __vdso_gettimeofday Stefani Seibold
2014-02-17  0:53   ` [tip:x86/vdso] x86, vdso: Cleanup __vdso_gettimeofday() tip-bot for Stefani Seibold
2014-02-16 21:52 ` [PATCH v18 07/10] introduce VVAR marco for vdso32 Stefani Seibold
2014-02-17  0:53   ` [tip:x86/vdso] x86, vdso: Introduce " tip-bot for Stefani Seibold
2014-02-16 21:52 ` [PATCH v18 08/10] Add 32 bit VDSO time support for 32 bit kernel Stefani Seibold
2014-02-17  0:53   ` [tip:x86/vdso] x86, vdso: Add 32-bit VDSO time support for the 32-bit kernel tip-bot for Stefani Seibold
2014-02-16 21:52 ` [PATCH v18 09/10] Add 32 bit VDSO time support for 64 bit kernel Stefani Seibold
2014-02-17  0:53   ` [tip:x86/vdso] x86, vdso: Add 32-bit VDSO time support for the 64-bit kernel tip-bot for Stefani Seibold
2014-02-17  0:54   ` [tip:x86/vdso] x86, vdso: Add a lot more dummy locking functions to vclock_gettime.c tip-bot for H. Peter Anvin
2014-02-17  1:48     ` H. Peter Anvin
2014-02-17  3:51   ` [tip:x86/vdso] x86, vdso: Instead of dummy functions, include < linux/spinlock_up.h> tip-bot for H. Peter Anvin
2014-02-17  4:06     ` H. Peter Anvin
2014-02-17  7:42       ` Stefani Seibold
2014-02-17  9:27         ` H. Peter Anvin
2014-02-17  9:46           ` Stefani Seibold
2014-02-17  9:50             ` H. Peter Anvin
2014-02-17 10:07               ` Stefani Seibold
2014-02-17 10:03 [PATCH v19 09/10] Add 32 bit VDSO time support for 64 bit kernel Stefani Seibold
2014-02-17 15:58 ` [tip:x86/vdso] x86, vdso: Add 32-bit VDSO time support for the 64-bit kernel tip-bot for Stefani Seibold
2014-02-17 19:03 ` tip-bot for Stefani Seibold
2014-02-19  9:09 [PATCH v20 09/10] Add 32 bit VDSO time support for 64 bit kernel Stefani Seibold
2014-02-19 22:30 ` [tip:x86/vdso] x86, vdso: Add 32-bit VDSO time support for the 64-bit kernel tip-bot for Stefani Seibold

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Powered by JetHome