This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
Re: kprobes + nptl01 (test in ltp testsuite) problem
On Tue, Jul 24, 2007 at 06:09:44PM +0530, manjunath k wrote:
> Ive executed the same tests for the 2.6.16.y kernel before, but the tests
> used
> to pass.
> Ive used the put_user() in this testcase does it mean that the put_user()
> is a blocking call or goes into sleep in 2.6.22.1 kernel version.
No. IFAICS, put_user() could always sleep. If the tests passed for you,
it may be because the user address you dereferenced was in memory.
Ananth
>
> Manjunath
>
> On 7/24/07, Ananth N Mavinakayanahalli <ananth@in.ibm.com> wrote:
> >On Tue, Jul 24, 2007 at 05:25:57PM +0530, manjunath k wrote:
> >> Hi,
> >>
> >> Ive been using the kprobes on X86 m/c and found the following problem
> >with
> >> the following kprobe and the testsuite.
> >>
> >> 1. A probe is placed for the sys_gettimeofday with the pre/post/fault
> >> handlers
> >> defined (k-test.c is attached).
> >>
> >> 2. The nptl01 test from the ltp testuite is executed and the system gets
> >> into an
> >> infinite loop (nptl01.c is attached).
> >>
> >> Ive attached the log file that contains the output that was seen when
> >> the above test is performed (kprobe-nptl.log), the log file is
> >> stripped for the purpose of attaching.
> >>
> >> Please provide me some information regarding the above failure.
> >
> >You can't call functions that can either fault or sleep/block in your
> >kprobe handlers. get_user() can sleep.
> >
> >Ananth
> >