2006-09-08 06:35:18

by Don Mullis

Subject: Re: Re: [patch 6/6] process filtering for fault-injection capabilities

Suggested changes, implemented by patch below:

1) Reorder kernel command line args alphabetically -- lets
output of `ls /debug/failslab` serve as a handy reminder
of the arg bindings.

2) Rename a variable to agree with the /debug file name.

---
lib/should_fail.c | 14 ++++++++------
1 file changed, 8 insertions(+), 6 deletions(-)

Index: linux-2.6.17/lib/should_fail.c
===================================================================
--- linux-2.6.17.orig/lib/should_fail.c
+++ linux-2.6.17/lib/should_fail.c
@@ -13,18 +13,20 @@ int setup_should_fail(struct should_fail
unsigned long interval;
int times;
int space;
- unsigned long filter;
+ unsigned long process_filter;

- /* "<probability>,<interval>,<times>,<space>,<process-filter>" */
- if (sscanf(str, "%lu,%lu,%d,%d,%lu", &probability, &interval, &times,
- &space, &filter) < 5)
- return 0;
+ /* "<interval>,<probability>,<process-filter>,<space>,<times>" */
+ if (sscanf(str, "%lu,%lu,%lu,%d,%d", &interval,
+ &probability, &process_filter, &space, &times) < 5) {
+ printk( "SHOULD_FAIL: failed to parse arguments\n");
+ return 0;
+ }

data->probability = probability;
data->interval = interval;
atomic_set(&data->times, times);
atomic_set(&data->space, space);
- data->process_filter = filter;
+ data->process_filter = process_filter;

return 1;
}

2006-09-08 07:06:18

by Akinobu Mita

Subject: Re: Re: [patch 6/6] process filtering for fault-injection capabilities

On Thu, Sep 07, 2006 at 11:29:57PM -0700, Don Mullis wrote:

> 1) Reorder kernel command line args alphabetically -- lets
> output of `ls /debug/failslab` serve as a handy reminder
> of the arg bindings.
>
> 2) Rename a variable to agree with the /debug file name.

Yes. this kernel command line was confusing.
This change makes it better.

Now I'm writing the filter which allow failing only for a specific
module by using unwinding kernel stacks API.

Then I'll send next version with your fixes.

2006-09-09 18:20:34

by Don Mullis

Subject: Re: Re: [patch 6/6] process filtering for fault-injection capabilities

On Fri, 2006-09-08 at 15:00 +0800, Akinobu Mita wrote:
> Now I'm writing the filter which allow failing only for a specific
> module by using unwinding kernel stacks API.

Are you planning to use STACK_UNWIND? I see that only i386
supports STACK_UNWIND.

Perhaps the much simpler STACKTRACE would suffice -- supported
by i386, x86_64, and s390?

2006-09-12 07:30:55

by Akinobu Mita

Subject: Re: Re: [patch 6/6] process filtering for fault-injection capabilities

On Sat, Sep 09, 2006 at 11:14:58AM -0700, Don Mullis wrote:
> On Fri, 2006-09-08 at 15:00 +0800, Akinobu Mita wrote:
> > Now I'm writing the filter which allow failing only for a specific
> > module by using unwinding kernel stacks API.
>
> Are you planning to use STACK_UNWIND? I see that only i386
> supports STACK_UNWIND.
>
> Perhaps the much simpler STACKTRACE would suffice -- supported
> by i386, x86_64, and s390?

Thanks, I didn't realize the include/linux/stacktrace.h