* [PATCH] x86/vdso: Give the [ph]vclock_page declarations real types
@ 2019-06-22 22:08 Andy Lutomirski
2019-06-23 12:42 ` Thomas Gleixner
2019-06-23 23:57 ` [tip:timers/vdso] " tip-bot for Andy Lutomirski
0 siblings, 2 replies; 6+ messages in thread
From: Andy Lutomirski @ 2019-06-22 22:08 UTC (permalink / raw)
To: x86
Cc: LKML, Borislav Petkov, Andy Lutomirski, Peter Zijlstra, Linus Torvalds
Clean up the vDSO code a bit by giving pvclock_page and hvclock_page
their actual types instead of u8[PAGE_SIZE]. This shouldn't
materially affect the generated code.
Heavily based on a patch from Linus.
Cc: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
---
arch/x86/entry/vdso/vclock_gettime.c | 36 ++++++++++++++++++----------
1 file changed, 24 insertions(+), 12 deletions(-)
diff --git a/arch/x86/entry/vdso/vclock_gettime.c b/arch/x86/entry/vdso/vclock_gettime.c
index 4aed41f638bb..907efc5015ec 100644
--- a/arch/x86/entry/vdso/vclock_gettime.c
+++ b/arch/x86/entry/vdso/vclock_gettime.c
@@ -28,13 +28,33 @@ 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);
+/*
+ * Declare the memory-mapped vclock data pages. These come from hypervisors.
+ * If we ever reintroduce something like direct access to an MMIO clock like
+ * the HPET again, it will go here as well.
+ *
+ * A load from any of these pages will segfault if the clock in question is
+ * disabled, so appropriate compiler barriers and checks need to be used
+ * to prevent stray loads.
+ *
+ * These declarations MUST NOT be const. The compiler will assume that
+ * an extern const variable has genuinely constant contents, and the
+ * resulting code won't work, since the whole point is that these pages
+ * change over time, possibly while we're accessing them.
+ */
+
#ifdef CONFIG_PARAVIRT_CLOCK
-extern u8 pvclock_page[PAGE_SIZE]
+/*
+ * This is the vCPU 0 pvclock page. We only use pvclock from the vDSO
+ * if the hypervisor tells us that all vCPUs can get valid data from the
+ * vCPU 0 page.
+ */
+extern struct pvclock_vsyscall_time_info pvclock_page
__attribute__((visibility("hidden")));
#endif
#ifdef CONFIG_HYPERV_TSCPAGE
-extern u8 hvclock_page[PAGE_SIZE]
+extern struct ms_hyperv_tsc_page hvclock_page
__attribute__((visibility("hidden")));
#endif
@@ -69,14 +89,9 @@ notrace static long vdso_fallback_gettime(long clock, struct timespec *ts)
#endif
#ifdef CONFIG_PARAVIRT_CLOCK
-static notrace const struct pvclock_vsyscall_time_info *get_pvti0(void)
-{
- return (const struct pvclock_vsyscall_time_info *)&pvclock_page;
-}
-
static notrace u64 vread_pvclock(void)
{
- const struct pvclock_vcpu_time_info *pvti = &get_pvti0()->pvti;
+ const struct pvclock_vcpu_time_info *pvti = &pvclock_page.pvti;
u32 version;
u64 ret;
@@ -117,10 +132,7 @@ static notrace u64 vread_pvclock(void)
#ifdef CONFIG_HYPERV_TSCPAGE
static notrace u64 vread_hvclock(void)
{
- const struct ms_hyperv_tsc_page *tsc_pg =
- (const struct ms_hyperv_tsc_page *)&hvclock_page;
-
- return hv_read_tsc_page(tsc_pg);
+ return hv_read_tsc_page(&hvclock_page);
}
#endif
--
2.21.0
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH] x86/vdso: Give the [ph]vclock_page declarations real types
2019-06-22 22:08 [PATCH] x86/vdso: Give the [ph]vclock_page declarations real types Andy Lutomirski
@ 2019-06-23 12:42 ` Thomas Gleixner
[not found] ` <CAHk-=whywzng7FLV9X67RPmHNnygK+7VJV+zh4njT6BA+h9tCw@mail.gmail.com>
2019-06-23 23:57 ` [tip:timers/vdso] " tip-bot for Andy Lutomirski
1 sibling, 1 reply; 6+ messages in thread
From: Thomas Gleixner @ 2019-06-23 12:42 UTC (permalink / raw)
To: Andy Lutomirski
Cc: x86, LKML, Borislav Petkov, Peter Zijlstra, Linus Torvalds
On Sat, 22 Jun 2019, Andy Lutomirski wrote:
> Clean up the vDSO code a bit by giving pvclock_page and hvclock_page
> their actual types instead of u8[PAGE_SIZE]. This shouldn't
> materially affect the generated code.
>
> Heavily based on a patch from Linus.
>
> Cc: Borislav Petkov <bp@alien8.de>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
So either this wants a
From: Linus line
or this tag should be:
Originally-by: Linus ...
Hmm?
Thanks,
tglx
^ permalink raw reply [flat|nested] 6+ messages in thread
* [tip:timers/vdso] x86/vdso: Give the [ph]vclock_page declarations real types
2019-06-22 22:08 [PATCH] x86/vdso: Give the [ph]vclock_page declarations real types Andy Lutomirski
2019-06-23 12:42 ` Thomas Gleixner
@ 2019-06-23 23:57 ` tip-bot for Andy Lutomirski
1 sibling, 0 replies; 6+ messages in thread
From: tip-bot for Andy Lutomirski @ 2019-06-23 23:57 UTC (permalink / raw)
To: linux-tip-commits
Cc: torvalds, luto, linux-kernel, mingo, bp, peterz, hpa, tglx
Commit-ID: ecf9db3d1f1a8fd2c335148891c3b044e9ce0628
Gitweb: https://git.kernel.org/tip/ecf9db3d1f1a8fd2c335148891c3b044e9ce0628
Author: Andy Lutomirski <luto@kernel.org>
AuthorDate: Sat, 22 Jun 2019 15:08:18 -0700
Committer: Thomas Gleixner <tglx@linutronix.de>
CommitDate: Mon, 24 Jun 2019 01:21:31 +0200
x86/vdso: Give the [ph]vclock_page declarations real types
Clean up the vDSO code a bit by giving pvclock_page and hvclock_page
their actual types instead of u8[PAGE_SIZE]. This shouldn't
materially affect the generated code.
Heavily based on a patch from Linus.
[ tglx: Adapted to the unified VDSO code ]
Co-developed-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lkml.kernel.org/r/6920c5188f8658001af1fc56fd35b815706d300c.1561241273.git.luto@kernel.org
---
arch/x86/include/asm/vdso/gettimeofday.h | 36 +++++++++++++++++++++-----------
1 file changed, 24 insertions(+), 12 deletions(-)
diff --git a/arch/x86/include/asm/vdso/gettimeofday.h b/arch/x86/include/asm/vdso/gettimeofday.h
index f92752d6cbcf..5b63f1f78a1f 100644
--- a/arch/x86/include/asm/vdso/gettimeofday.h
+++ b/arch/x86/include/asm/vdso/gettimeofday.h
@@ -26,13 +26,33 @@
#define VDSO_HAS_CLOCK_GETRES 1
+/*
+ * Declare the memory-mapped vclock data pages. These come from hypervisors.
+ * If we ever reintroduce something like direct access to an MMIO clock like
+ * the HPET again, it will go here as well.
+ *
+ * A load from any of these pages will segfault if the clock in question is
+ * disabled, so appropriate compiler barriers and checks need to be used
+ * to prevent stray loads.
+ *
+ * These declarations MUST NOT be const. The compiler will assume that
+ * an extern const variable has genuinely constant contents, and the
+ * resulting code won't work, since the whole point is that these pages
+ * change over time, possibly while we're accessing them.
+ */
+
#ifdef CONFIG_PARAVIRT_CLOCK
-extern u8 pvclock_page[PAGE_SIZE]
+/*
+ * This is the vCPU 0 pvclock page. We only use pvclock from the vDSO
+ * if the hypervisor tells us that all vCPUs can get valid data from the
+ * vCPU 0 page.
+ */
+extern struct pvclock_vsyscall_time_info pvclock_page
__attribute__((visibility("hidden")));
#endif
#ifdef CONFIG_HYPERV_TSCPAGE
-extern u8 hvclock_page[PAGE_SIZE]
+extern struct ms_hyperv_tsc_page hvclock_page
__attribute__((visibility("hidden")));
#endif
@@ -131,14 +151,9 @@ clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
#endif
#ifdef CONFIG_PARAVIRT_CLOCK
-static const struct pvclock_vsyscall_time_info *get_pvti0(void)
-{
- return (const struct pvclock_vsyscall_time_info *)&pvclock_page;
-}
-
static u64 vread_pvclock(void)
{
- const struct pvclock_vcpu_time_info *pvti = &get_pvti0()->pvti;
+ const struct pvclock_vcpu_time_info *pvti = &pvclock_page.pvti;
u32 version;
u64 ret;
@@ -180,10 +195,7 @@ static u64 vread_pvclock(void)
#ifdef CONFIG_HYPERV_TSCPAGE
static u64 vread_hvclock(void)
{
- const struct ms_hyperv_tsc_page *tsc_pg =
- (const struct ms_hyperv_tsc_page *)&hvclock_page;
-
- return hv_read_tsc_page(tsc_pg);
+ return hv_read_tsc_page(&hvclock_page);
}
#endif
^ permalink raw reply related [flat|nested] 6+ messages in thread
end of thread, other threads:[~2019-06-24 3:01 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-06-22 22:08 [PATCH] x86/vdso: Give the [ph]vclock_page declarations real types Andy Lutomirski
2019-06-23 12:42 ` Thomas Gleixner
[not found] ` <CAHk-=whywzng7FLV9X67RPmHNnygK+7VJV+zh4njT6BA+h9tCw@mail.gmail.com>
2019-06-23 13:10 ` Thomas Gleixner
2019-06-23 16:03 ` Andy Lutomirski
2019-06-23 22:19 ` Thomas Gleixner
2019-06-23 23:57 ` [tip:timers/vdso] " tip-bot for Andy Lutomirski
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).