[PATCH] fio: fix cycles_start build issue

[PATCH] fio: fix cycles_start build issue

[Christian Ehrhardt]

Hi,
a recent checkout of fio-git revealed a build issue on s390.
For whatever reason my git send-email setup is broken so I paste it in 
the mail and hope it doesn't get mangled, let me know if that fails.

---

Subject: [PATCH] fio: fix cycles_start build issue

From: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>

There is an issue introduced with "commit 73df3e07 gettime: offset CPU cycle
counter by initial value". For architectures which define
ARCH_CPU_CLOCK_CYCLES_PER_USEC (currently only s390) this causes a build 
error

gettime.c:174:11: error: ‘cycles_start’ undeclared (first use in this 
function)
    if (t < cycles_start && !cycles_wrap)

To make sure variables and code are only compiled in the same cases I 
added a
clock definition called ARCH_CPU_CLOCK_WRAPS. We could merge
ARCH_CPU_CLOCK_WRAPS and the existing ARCH_CPU_CLOCK_CYCLES_PER_USEC 
into one
if you prefer, so far nobody else uses it.

To avoid cluttering all architecture headers I enabled it by default in
arch-generic.h, so any arch not needing the wrap handling can undef
ARCH_CPU_CLOCK_WRAPS later in their headers.

Signed-off-by: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
---

[diffstat]
  arch/arch-generic.h |    1 +
  arch/arch-s390.h    |    1 +
  gettime.c           |    7 +++++++
  3 files changed, 9 insertions(+)

[diff]
diff --git a/arch/arch-generic.h b/arch/arch-generic.h
index a0b71f8..825a829 100644
--- a/arch/arch-generic.h
+++ b/arch/arch-generic.h
@@ -2,6 +2,7 @@
  #define ARCH_GENERIC_H

  #define FIO_ARCH	(arch_generic)
+#define ARCH_CPU_CLOCK_WRAPS

  #define nop			do { } while (0)
  #define read_barrier()		__asm__ __volatile__("": : :"memory")
diff --git a/arch/arch-s390.h b/arch/arch-s390.h
index 169282b..0c93424 100644
--- a/arch/arch-s390.h
+++ b/arch/arch-s390.h
@@ -39,6 +39,7 @@ static inline unsigned long long get_cpu_clock(void)
  }

  #define ARCH_CPU_CLOCK_CYCLES_PER_USEC 1
+#undef ARCH_CPU_CLOCK_WRAPS
  #define ARCH_HAVE_CPU_CLOCK

  #define ARCH_HAVE_INIT
diff --git a/gettime.c b/gettime.c
index 6863ce3..29884a1 100644
--- a/gettime.c
+++ b/gettime.c
@@ -17,6 +17,8 @@
  static unsigned long cycles_per_usec;
  static unsigned long inv_cycles_per_usec;
  static uint64_t max_cycles_for_mult;
+#endif
+#ifdef ARCH_CPU_CLOCK_WRAPS
  static unsigned long long cycles_start, cycles_wrap;
  #endif
  int tsc_reliable = 0;
@@ -171,6 +173,7 @@ static void __fio_gettime(struct timeval *tp)
  #endif

  		t = get_cpu_clock();
+#ifdef ARCH_CPU_CLOCK_WRAPS
  		if (t < cycles_start && !cycles_wrap)
  			cycles_wrap = 1;
  		else if (cycles_wrap && t >= cycles_start && !tv->warned) {
@@ -179,6 +182,8 @@ static void __fio_gettime(struct timeval *tp)
  		}

  		t -= cycles_start;
+#endif
+
  		tv->last_cycles = t;
  		tv->last_tv_valid = 1;
  #ifdef ARCH_CPU_CLOCK_CYCLES_PER_USEC
@@ -311,8 +316,10 @@ static int calibrate_cpu_clock(void)
  	inv_cycles_per_usec = 16777216UL / cycles_per_usec;
  	max_cycles_for_mult = ~0ULL / inv_cycles_per_usec;
  	dprint(FD_TIME, "inv_cycles_per_usec=%lu\n", inv_cycles_per_usec);
+#ifdef ARCH_CPU_CLOCK_WRAPS
  	cycles_start = get_cpu_clock();
  	dprint(FD_TIME, "cycles_start=%llu\n", cycles_start);
+#endif
  	return 0;
  }
  #else

Re: [PATCH] fio: fix cycles_start build issue

[Jens Axboe]

On 02/27/2015 01:31 AM, Christian Ehrhardt wrote:
> Hi,
> a recent checkout of fio-git revealed a build issue on s390.
> For whatever reason my git send-email setup is broken so I paste it in
> the mail and hope it doesn't get mangled, let me know if that fails.

It mangled it horribly, but I just hand applied it. One issue, though:

> diff --git a/arch/arch-generic.h b/arch/arch-generic.h
> index a0b71f8..825a829 100644
> --- a/arch/arch-generic.h
> +++ b/arch/arch-generic.h
> @@ -2,6 +2,7 @@
>   #define ARCH_GENERIC_H
>
>   #define FIO_ARCH    (arch_generic)
> +#define ARCH_CPU_CLOCK_WRAPS

This wont work like I suspect you think it does. This define needs to go 
in the arch.h header, if you want all archs to catch it.

-- 
Jens Axboe

Re: [PATCH] fio: fix cycles_start build issue

[Christian Ehrhardt]

On 27/02/15 16:20, Jens Axboe wrote:
> On 02/27/2015 01:31 AM, Christian Ehrhardt wrote:
>> Hi,
>> a recent checkout of fio-git revealed a build issue on s390.
>> For whatever reason my git send-email setup is broken so I paste it in
>> the mail and hope it doesn't get mangled, let me know if that fails.
>
> It mangled it horribly, but I just hand applied it. One issue, though:
>
>> diff --git a/arch/arch-generic.h b/arch/arch-generic.h
>> index a0b71f8..825a829 100644
>> --- a/arch/arch-generic.h
>> +++ b/arch/arch-generic.h
>> @@ -2,6 +2,7 @@
>>   #define ARCH_GENERIC_H
>>
>>   #define FIO_ARCH    (arch_generic)
>> +#define ARCH_CPU_CLOCK_WRAPS
>
> This wont work like I suspect you think it does. This define needs to go
> in the arch.h header, if you want all archs to catch it.
>

I can only agree and thank you for unmangling the content.
Thanks !