Hello Thomas
I took Jon Masters' latest hwlat_detector patch for upstream, and grabbed
the bits that aren't currently in -rt. The patch is fairly small.
Please consider it for the next -rt release.
John
>From 97e12a96f58eea70c6a223ac849ec894165884b5 Mon Sep 17 00:00:00 2001
From: Jon Masters <[email protected]>
Date: Thu, 22 Oct 2009 19:33:56 +0200
Subject: [PATCH] Updates from Jon Masters for hwlat_detector.c
Signed-off-by: Jon Masters <[email protected]>
Signed-off-by: John Kacur <[email protected]>
---
drivers/misc/Kconfig | 2 +-
drivers/misc/hwlat_detector.c | 16 ++++++++++++----
2 files changed, 13 insertions(+), 5 deletions(-)
diff --git a/drivers/misc/Kconfig b/drivers/misc/Kconfig
index 15cd6bc..35f441c 100644
--- a/drivers/misc/Kconfig
+++ b/drivers/misc/Kconfig
@@ -95,7 +95,7 @@ config IBM_ASM
config HWLAT_DETECTOR
tristate "Testing module to detect hardware-induced latencies"
depends on DEBUG_FS
- select RING_BUFFER
+ depends on RING_BUFFER
default m
---help---
A simple hardware latency detector. Use this module to detect
diff --git a/drivers/misc/hwlat_detector.c b/drivers/misc/hwlat_detector.c
index be6553f..7191e17 100644
--- a/drivers/misc/hwlat_detector.c
+++ b/drivers/misc/hwlat_detector.c
@@ -262,6 +262,8 @@ static int get_sample(void *unused)
/* Keep a running maximum ever recorded hardware latency */
if (sample > data.max_sample)
data.max_sample = sample;
+
+ wake_up(&data.wq); /* wake up reader(s) */
}
ret = 0;
@@ -299,8 +301,6 @@ static int kthread_fn(void *unused)
goto err_out;
}
- wake_up(&data.wq); /* wake up reader(s) */
-
interval = data.sample_window - data.sample_width;
do_div(interval, USEC_PER_MSEC); /* modifies interval value */
@@ -607,7 +607,11 @@ static ssize_t debug_enable_fwrite(struct file *filp,
if (!enabled)
goto unlock;
enabled = 0;
- stop_kthread();
+ err = stop_kthread();
+ if (0 != err) {
+ printk(KERN_ERR BANNER "cannot stop kthread\n");
+ return -EFAULT;
+ }
wake_up(&data.wq); /* reader(s) should return */
}
unlock:
@@ -1194,9 +1198,13 @@ out:
*/
static void detector_exit(void)
{
+ int err;
+
if (enabled) {
enabled = 0;
- stop_kthread();
+ err = stop_kthread();
+ if (0 != err)
+ printk(KERN_ERR BANNER "cannot stop kthread\n");
}
free_debugfs();
--
1.6.0.6
Hi,
On Fri, 23 Oct 2009, John Kacur wrote:
> I took Jon Masters' latest hwlat_detector patch for upstream, and grabbed
> the bits that aren't currently in -rt. The patch is fairly small.
>
> Please consider it for the next -rt release.
Not really.
> diff --git a/drivers/misc/hwlat_detector.c b/drivers/misc/hwlat_detector.c
> index be6553f..7191e17 100644
> --- a/drivers/misc/hwlat_detector.c
> +++ b/drivers/misc/hwlat_detector.c
> @@ -262,6 +262,8 @@ static int get_sample(void *unused)
> /* Keep a running maximum ever recorded hardware latency */
> if (sample > data.max_sample)
> data.max_sample = sample;
> +
> + wake_up(&data.wq); /* wake up reader(s) */
> }
>
> ret = 0;
> @@ -299,8 +301,6 @@ static int kthread_fn(void *unused)
> goto err_out;
> }
>
> - wake_up(&data.wq); /* wake up reader(s) */
> -
This reverts -rt commit 5fc485f6341c177e9d5f641c6fb9e33ca465aa84.
CONFIG_DEBUG_SPINLOCK_SLEEP resp. CONFIG_DEBUG_PREEMPT will tell you
why this is a bad idea on -rt :)
Thanks,
tglx
On Mon, Nov 2, 2009 at 4:43 PM, Thomas Gleixner <[email protected]> wrote:
> Hi,
>
> On Fri, 23 Oct 2009, John Kacur wrote:
>> I took Jon Masters' latest hwlat_detector patch for upstream, and grabbed
>> the bits that aren't currently in -rt. The patch is fairly small.
>>
>> Please consider it for the next -rt release.
>
> Not really.
>
>> diff --git a/drivers/misc/hwlat_detector.c b/drivers/misc/hwlat_detector.c
>> index be6553f..7191e17 100644
>> --- a/drivers/misc/hwlat_detector.c
>> +++ b/drivers/misc/hwlat_detector.c
>> @@ -262,6 +262,8 @@ static int get_sample(void *unused)
>> ? ? ? ? ? ? ? /* Keep a running maximum ever recorded hardware latency */
>> ? ? ? ? ? ? ? if (sample > data.max_sample)
>> ? ? ? ? ? ? ? ? ? ? ? data.max_sample = sample;
>> +
>> + ? ? ? ? ? ? wake_up(&data.wq); /* wake up reader(s) */
>> ? ? ? }
>>
>> ? ? ? ret = 0;
>> @@ -299,8 +301,6 @@ static int kthread_fn(void *unused)
>> ? ? ? ? ? ? ? ? ? ? ? goto err_out;
>> ? ? ? ? ? ? ? }
>>
>> - ? ? ? ? ? ? wake_up(&data.wq); /* wake up reader(s) */
>> -
>
> This reverts -rt commit 5fc485f6341c177e9d5f641c6fb9e33ca465aa84.
>
> CONFIG_DEBUG_SPINLOCK_SLEEP resp. CONFIG_DEBUG_PREEMPT will tell you
> why this is a bad idea on -rt :)
>
> Thanks,
>
> ? ? ? ?tglx
> --
Ugh. Thanks for catching that. To be honest not sure how it happened,
so it must be my fault. :)
Updated patch just adds the error checking from Jon Masters then,
patched against -rt.
On Mon, 2 Nov 2009, John Kacur wrote:
> Updated patch just adds the error checking from Jon Masters then,
> patched against -rt.
Can you please inline patches ?
@@ -607,7 +607,11 @@ static ssize_t debug_enable_fwrite(stru
if (!enabled)
goto unlock;
enabled = 0;
- stop_kthread();
+ err = stop_kthread();
+ if (0 != err) {
Eeew !
Thanks,
tglx
On Tue, 3 Nov 2009, Thomas Gleixner wrote:
> On Mon, 2 Nov 2009, John Kacur wrote:
> > Updated patch just adds the error checking from Jon Masters then,
> > patched against -rt.
>
> Can you please inline patches ?
>
> @@ -607,7 +607,11 @@ static ssize_t debug_enable_fwrite(stru
> if (!enabled)
> goto unlock;
> enabled = 0;
> - stop_kthread();
> + err = stop_kthread();
> + if (0 != err) {
>
> Eeew !
Yeah, I saw that too and thought it a bit ugly. However, my goal was only
to make sure that the latest version of Jon's hwlat patch got into -rt, so
I didn't think it was my place to "fix" his style.
Note - even the much hated checkpatch program doesn't balk at that.
However that is hardly an authority.
If you read the introduction to "Expert C Programming - Deep C Secrets",
they mention that some programmers write
if (3==i)
instead of
if (i==3)
because if you mistakenly only type one '=', then the compiler can
complain about it, instead of silently assigning 3 to i.
I can only guess that is the sort of idea / habit that Jon got that style
from.
John
On Mon, 2009-11-02 at 20:46 +0100, John Kacur wrote:
> On Mon, Nov 2, 2009 at 4:43 PM, Thomas Gleixner <[email protected]> wrote:
> >> diff --git a/drivers/misc/hwlat_detector.c b/drivers/misc/hwlat_detector.c
> >> index be6553f..7191e17 100644
> >> --- a/drivers/misc/hwlat_detector.c
> >> +++ b/drivers/misc/hwlat_detector.c
> >> @@ -262,6 +262,8 @@ static int get_sample(void *unused)
> >> /* Keep a running maximum ever recorded hardware latency */
> >> if (sample > data.max_sample)
> >> data.max_sample = sample;
> >> +
> >> + wake_up(&data.wq); /* wake up reader(s) */
> >> }
> >>
> >> ret = 0;
> >> @@ -299,8 +301,6 @@ static int kthread_fn(void *unused)
> >> goto err_out;
> >> }
> >>
> >> - wake_up(&data.wq); /* wake up reader(s) */
> >> -
> >
> > This reverts -rt commit 5fc485f6341c177e9d5f641c6fb9e33ca465aa84.
> >
> > CONFIG_DEBUG_SPINLOCK_SLEEP resp. CONFIG_DEBUG_PREEMPT will tell you
> > why this is a bad idea on -rt :)
> >
> > Thanks,
> >
> > tglx
> > --
>
> Ugh. Thanks for catching that. To be honest not sure how it happened,
> so it must be my fault. :)
>
> Updated patch just adds the error checking from Jon Masters then,
> patched against -rt.
Actually, I think it might be my fault. I'll talk to you about the best
way to proceed with updates since we're still looking for a reproducer
on that lockup Clark was seeing a while back.
Jon.