diff mbox series

[5/6] riscv: don't use the rdtime(h) pseudo-instructions

Message ID 20190821145837.3686-6-hch@lst.de (mailing list archive)
State New, archived
Headers show
Series [1/6] riscv: refactor the IPI code | expand

Commit Message

Christoph Hellwig Aug. 21, 2019, 2:58 p.m. UTC
If we just use the CSRs that these map to directly the code is simpler
and doesn't require extra inline assembly code.  Also fix up the top-level
comment in timer-riscv.c to not talk about the cycle count or mention
details of the clocksource interface, of which this file is just a
consumer.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/riscv/include/asm/timex.h    | 44 +++++++++++++++----------------
 drivers/clocksource/timer-riscv.c | 17 +++---------
 2 files changed, 25 insertions(+), 36 deletions(-)

Comments

Atish Patra Aug. 24, 2019, 12:37 a.m. UTC | #1
On Wed, 2019-08-21 at 23:58 +0900, Christoph Hellwig wrote:
> If we just use the CSRs that these map to directly the code is
> simpler
> and doesn't require extra inline assembly code.  Also fix up the top-
> level
> comment in timer-riscv.c to not talk about the cycle count or mention
> details of the clocksource interface, of which this file is just a
> consumer.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  arch/riscv/include/asm/timex.h    | 44 +++++++++++++++------------
> ----
>  drivers/clocksource/timer-riscv.c | 17 +++---------
>  2 files changed, 25 insertions(+), 36 deletions(-)
> 
> diff --git a/arch/riscv/include/asm/timex.h
> b/arch/riscv/include/asm/timex.h
> index 6a703ec9d796..c7ef131b9e4c 100644
> --- a/arch/riscv/include/asm/timex.h
> +++ b/arch/riscv/include/asm/timex.h
> @@ -6,43 +6,41 @@
>  #ifndef _ASM_RISCV_TIMEX_H
>  #define _ASM_RISCV_TIMEX_H
>  
> -#include <asm/param.h>
> +#include <asm/csr.h>
>  
>  typedef unsigned long cycles_t;
>  
> -static inline cycles_t get_cycles_inline(void)
> +static inline cycles_t get_cycles(void)
>  {
> -	cycles_t n;
> -
> -	__asm__ __volatile__ (
> -		"rdtime %0"
> -		: "=r" (n));
> -	return n;
> +	return csr_read(CSR_TIME);

Does this work correctly in QEMU ? I was looking at the qemu code and
it looks like it returns cpu_get_host_ticks which seems wrong to me.

https://github.com/qemu/qemu/blob/master/target/riscv/csr.c#L213


>  }
> -#define get_cycles get_cycles_inline
> +#define get_cycles get_cycles
>  
>  #ifdef CONFIG_64BIT
> -static inline uint64_t get_cycles64(void)
> +static inline u64 get_cycles64(void)
> +{
> +	return get_cycles();
> +}
> +#else /* CONFIG_64BIT */
> +static inline u32 get_cycles_hi(void)
>  {
> -        return get_cycles();
> +	return csr_read(CSR_TIMEH);
>  }
> -#else
> -static inline uint64_t get_cycles64(void)
> +
> +static inline u64 get_cycles64(void)
>  {
> -	u32 lo, hi, tmp;
> -	__asm__ __volatile__ (
> -		"1:\n"
> -		"rdtimeh %0\n"
> -		"rdtime %1\n"
> -		"rdtimeh %2\n"
> -		"bne %0, %2, 1b"
> -		: "=&r" (hi), "=&r" (lo), "=&r" (tmp));
> +	u32 hi, lo;
> +
> +	do {
> +		hi = get_cycles_hi();
> +		lo = get_cycles();
> +	} while (hi != get_cycles_hi());
> +
>  	return ((u64)hi << 32) | lo;
>  }
> -#endif
> +#endif /* CONFIG_64BIT */
>  
>  #define ARCH_HAS_READ_CURRENT_TIMER
> -
>  static inline int read_current_timer(unsigned long *timer_val)
>  {
>  	*timer_val = get_cycles();
> diff --git a/drivers/clocksource/timer-riscv.c
> b/drivers/clocksource/timer-riscv.c
> index 09e031176bc6..470c7ef02ea4 100644
> --- a/drivers/clocksource/timer-riscv.c
> +++ b/drivers/clocksource/timer-riscv.c
> @@ -2,6 +2,10 @@
>  /*
>   * Copyright (C) 2012 Regents of the University of California
>   * Copyright (C) 2017 SiFive
> + *
> + * All RISC-V systems have a timer attached to every hart.  These
> timers can be
> + * read from the "time" and "timeh" CSRs, and can use the SBI to
> setup
> + * events.
>   */
>  #include <linux/clocksource.h>
>  #include <linux/clockchips.h>
> @@ -12,19 +16,6 @@
>  #include <asm/smp.h>
>  #include <asm/sbi.h>
>  
> -/*
> - * All RISC-V systems have a timer attached to every hart.  These
> timers can be
> - * read by the 'rdcycle' pseudo instruction, and can use the SBI to
> setup
> - * events.  In order to abstract the architecture-specific timer
> reading and
> - * setting functions away from the clock event insertion code, we
> provide
> - * function pointers to the clockevent subsystem that perform two
> basic
> - * operations: rdtime() reads the timer on the current CPU, and
> - * next_event(delta) sets the next timer event to 'delta' cycles in
> the future.
> - * As the timers are inherently a per-cpu resource, these callbacks
> perform
> - * operations on the current hart.  There is guaranteed to be
> exactly one timer
> - * per hart on all RISC-V systems.
> - */
> -
>  static int riscv_clock_next_event(unsigned long delta,
>  		struct clock_event_device *ce)
>  {
Atish Patra Aug. 24, 2019, 12:43 a.m. UTC | #2
On Fri, 2019-08-23 at 17:37 -0700, Atish Patra wrote:
> On Wed, 2019-08-21 at 23:58 +0900, Christoph Hellwig wrote:
> > If we just use the CSRs that these map to directly the code is
> > simpler
> > and doesn't require extra inline assembly code.  Also fix up the
> > top-
> > level
> > comment in timer-riscv.c to not talk about the cycle count or
> > mention
> > details of the clocksource interface, of which this file is just a
> > consumer.
> > 
> > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > ---
> >  arch/riscv/include/asm/timex.h    | 44 +++++++++++++++------------
> > ----
> >  drivers/clocksource/timer-riscv.c | 17 +++---------
> >  2 files changed, 25 insertions(+), 36 deletions(-)
> > 
> > diff --git a/arch/riscv/include/asm/timex.h
> > b/arch/riscv/include/asm/timex.h
> > index 6a703ec9d796..c7ef131b9e4c 100644
> > --- a/arch/riscv/include/asm/timex.h
> > +++ b/arch/riscv/include/asm/timex.h
> > @@ -6,43 +6,41 @@
> >  #ifndef _ASM_RISCV_TIMEX_H
> >  #define _ASM_RISCV_TIMEX_H
> >  
> > -#include <asm/param.h>
> > +#include <asm/csr.h>
> >  
> >  typedef unsigned long cycles_t;
> >  
> > -static inline cycles_t get_cycles_inline(void)
> > +static inline cycles_t get_cycles(void)
> >  {
> > -	cycles_t n;
> > -
> > -	__asm__ __volatile__ (
> > -		"rdtime %0"
> > -		: "=r" (n));
> > -	return n;
> > +	return csr_read(CSR_TIME);
> 
> Does this work correctly in QEMU ? I was looking at the qemu code and
> it looks like it returns cpu_get_host_ticks which seems wrong to me.
> 
> https://github.com/qemu/qemu/blob/master/target/riscv/csr.c#L213
> 
> 

Nevermind. I missed the CONFIG_USER_ONLY and got confused.
csr_read will also trap and get the correct value.

Regards,
Atish
> >  }
> > -#define get_cycles get_cycles_inline
> > +#define get_cycles get_cycles
> >  
> >  #ifdef CONFIG_64BIT
> > -static inline uint64_t get_cycles64(void)
> > +static inline u64 get_cycles64(void)
> > +{
> > +	return get_cycles();
> > +}
> > +#else /* CONFIG_64BIT */
> > +static inline u32 get_cycles_hi(void)
> >  {
> > -        return get_cycles();
> > +	return csr_read(CSR_TIMEH);
> >  }
> > -#else
> > -static inline uint64_t get_cycles64(void)
> > +
> > +static inline u64 get_cycles64(void)
> >  {
> > -	u32 lo, hi, tmp;
> > -	__asm__ __volatile__ (
> > -		"1:\n"
> > -		"rdtimeh %0\n"
> > -		"rdtime %1\n"
> > -		"rdtimeh %2\n"
> > -		"bne %0, %2, 1b"
> > -		: "=&r" (hi), "=&r" (lo), "=&r" (tmp));
> > +	u32 hi, lo;
> > +
> > +	do {
> > +		hi = get_cycles_hi();
> > +		lo = get_cycles();
> > +	} while (hi != get_cycles_hi());
> > +
> >  	return ((u64)hi << 32) | lo;
> >  }
> > -#endif
> > +#endif /* CONFIG_64BIT */
> >  
> >  #define ARCH_HAS_READ_CURRENT_TIMER
> > -
> >  static inline int read_current_timer(unsigned long *timer_val)
> >  {
> >  	*timer_val = get_cycles();
> > diff --git a/drivers/clocksource/timer-riscv.c
> > b/drivers/clocksource/timer-riscv.c
> > index 09e031176bc6..470c7ef02ea4 100644
> > --- a/drivers/clocksource/timer-riscv.c
> > +++ b/drivers/clocksource/timer-riscv.c
> > @@ -2,6 +2,10 @@
> >  /*
> >   * Copyright (C) 2012 Regents of the University of California
> >   * Copyright (C) 2017 SiFive
> > + *
> > + * All RISC-V systems have a timer attached to every hart.  These
> > timers can be
> > + * read from the "time" and "timeh" CSRs, and can use the SBI to
> > setup
> > + * events.
> >   */
> >  #include <linux/clocksource.h>
> >  #include <linux/clockchips.h>
> > @@ -12,19 +16,6 @@
> >  #include <asm/smp.h>
> >  #include <asm/sbi.h>
> >  
> > -/*
> > - * All RISC-V systems have a timer attached to every hart.  These
> > timers can be
> > - * read by the 'rdcycle' pseudo instruction, and can use the SBI
> > to
> > setup
> > - * events.  In order to abstract the architecture-specific timer
> > reading and
> > - * setting functions away from the clock event insertion code, we
> > provide
> > - * function pointers to the clockevent subsystem that perform two
> > basic
> > - * operations: rdtime() reads the timer on the current CPU, and
> > - * next_event(delta) sets the next timer event to 'delta' cycles
> > in
> > the future.
> > - * As the timers are inherently a per-cpu resource, these
> > callbacks
> > perform
> > - * operations on the current hart.  There is guaranteed to be
> > exactly one timer
> > - * per hart on all RISC-V systems.
> > - */
> > -
> >  static int riscv_clock_next_event(unsigned long delta,
> >  		struct clock_event_device *ce)
> >  {
Christoph Hellwig Aug. 26, 2019, 11:30 a.m. UTC | #3
On Sat, Aug 24, 2019 at 12:37:02AM +0000, Atish Patra wrote:
> > -static inline cycles_t get_cycles_inline(void)
> > +static inline cycles_t get_cycles(void)
> >  {
> > -	cycles_t n;
> > -
> > -	__asm__ __volatile__ (
> > -		"rdtime %0"
> > -		: "=r" (n));
> > -	return n;
> > +	return csr_read(CSR_TIME);
> 
> Does this work correctly in QEMU ? I was looking at the qemu code and
> it looks like it returns cpu_get_host_ticks which seems wrong to me.

It better should.  rdtime is just a pseudo-instruction that the
assembler translates to a CSR read.  (in other words something totally
pointless, no idea why it even is in the spec).
Atish Patra Aug. 27, 2019, 6:50 p.m. UTC | #4
On Mon, 2019-08-26 at 13:30 +0200, hch@lst.de wrote:
> On Sat, Aug 24, 2019 at 12:37:02AM +0000, Atish Patra wrote:
> > > -static inline cycles_t get_cycles_inline(void)
> > > +static inline cycles_t get_cycles(void)
> > >  {
> > > -	cycles_t n;
> > > -
> > > -	__asm__ __volatile__ (
> > > -		"rdtime %0"
> > > -		: "=r" (n));
> > > -	return n;
> > > +	return csr_read(CSR_TIME);
> > 
> > Does this work correctly in QEMU ? I was looking at the qemu code
> > and
> > it looks like it returns cpu_get_host_ticks which seems wrong to
> > me.
> 
> It better should.  rdtime is just a pseudo-instruction that the
> assembler translates to a CSR read.  (in other words something
> totally
> pointless, no idea why it even is in the spec).

Yes. I did not see the usermode macro carefully in qemu code.

Reviewed-by: Atish Patra <atish.patra@wdc.com>
Paul Walmsley Sept. 5, 2019, 8:55 a.m. UTC | #5
On Wed, 21 Aug 2019, Christoph Hellwig wrote:

> If we just use the CSRs that these map to directly the code is simpler
> and doesn't require extra inline assembly code.  Also fix up the top-level
> comment in timer-riscv.c to not talk about the cycle count or mention
> details of the clocksource interface, of which this file is just a
> consumer.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>

Thanks, queued for v5.4-rc1 with Atish's Reviewed-by:.


- Paul
diff mbox series

Patch

diff --git a/arch/riscv/include/asm/timex.h b/arch/riscv/include/asm/timex.h
index 6a703ec9d796..c7ef131b9e4c 100644
--- a/arch/riscv/include/asm/timex.h
+++ b/arch/riscv/include/asm/timex.h
@@ -6,43 +6,41 @@ 
 #ifndef _ASM_RISCV_TIMEX_H
 #define _ASM_RISCV_TIMEX_H
 
-#include <asm/param.h>
+#include <asm/csr.h>
 
 typedef unsigned long cycles_t;
 
-static inline cycles_t get_cycles_inline(void)
+static inline cycles_t get_cycles(void)
 {
-	cycles_t n;
-
-	__asm__ __volatile__ (
-		"rdtime %0"
-		: "=r" (n));
-	return n;
+	return csr_read(CSR_TIME);
 }
-#define get_cycles get_cycles_inline
+#define get_cycles get_cycles
 
 #ifdef CONFIG_64BIT
-static inline uint64_t get_cycles64(void)
+static inline u64 get_cycles64(void)
+{
+	return get_cycles();
+}
+#else /* CONFIG_64BIT */
+static inline u32 get_cycles_hi(void)
 {
-        return get_cycles();
+	return csr_read(CSR_TIMEH);
 }
-#else
-static inline uint64_t get_cycles64(void)
+
+static inline u64 get_cycles64(void)
 {
-	u32 lo, hi, tmp;
-	__asm__ __volatile__ (
-		"1:\n"
-		"rdtimeh %0\n"
-		"rdtime %1\n"
-		"rdtimeh %2\n"
-		"bne %0, %2, 1b"
-		: "=&r" (hi), "=&r" (lo), "=&r" (tmp));
+	u32 hi, lo;
+
+	do {
+		hi = get_cycles_hi();
+		lo = get_cycles();
+	} while (hi != get_cycles_hi());
+
 	return ((u64)hi << 32) | lo;
 }
-#endif
+#endif /* CONFIG_64BIT */
 
 #define ARCH_HAS_READ_CURRENT_TIMER
-
 static inline int read_current_timer(unsigned long *timer_val)
 {
 	*timer_val = get_cycles();
diff --git a/drivers/clocksource/timer-riscv.c b/drivers/clocksource/timer-riscv.c
index 09e031176bc6..470c7ef02ea4 100644
--- a/drivers/clocksource/timer-riscv.c
+++ b/drivers/clocksource/timer-riscv.c
@@ -2,6 +2,10 @@ 
 /*
  * Copyright (C) 2012 Regents of the University of California
  * Copyright (C) 2017 SiFive
+ *
+ * All RISC-V systems have a timer attached to every hart.  These timers can be
+ * read from the "time" and "timeh" CSRs, and can use the SBI to setup
+ * events.
  */
 #include <linux/clocksource.h>
 #include <linux/clockchips.h>
@@ -12,19 +16,6 @@ 
 #include <asm/smp.h>
 #include <asm/sbi.h>
 
-/*
- * All RISC-V systems have a timer attached to every hart.  These timers can be
- * read by the 'rdcycle' pseudo instruction, and can use the SBI to setup
- * events.  In order to abstract the architecture-specific timer reading and
- * setting functions away from the clock event insertion code, we provide
- * function pointers to the clockevent subsystem that perform two basic
- * operations: rdtime() reads the timer on the current CPU, and
- * next_event(delta) sets the next timer event to 'delta' cycles in the future.
- * As the timers are inherently a per-cpu resource, these callbacks perform
- * operations on the current hart.  There is guaranteed to be exactly one timer
- * per hart on all RISC-V systems.
- */
-
 static int riscv_clock_next_event(unsigned long delta,
 		struct clock_event_device *ce)
 {