Re: [BUG] Null pointer deref with hrtimer_try_to_cancel()

[Oleg Nesterov <oleg@redhat.com>]

On 12/20, Oleg Nesterov wrote:
>
> On 12/19, Thomas Gleixner wrote:
> >
> > On Fri, 19 Dec 2008, Eric Sesterhenn wrote:
> >
> > > root@computer-desktop:~/testing/ltp-full-20081130/tools/strace_test#
> > > ./timer_create04
> > > timer_create04    1  FAIL  :  timer_create(2) failed to produce expected
> > > error; 22 , errno : EINVAL and got 0
> > > timer_create04    2  PASS  :  timer_create(2) expected failure; Got
> > > errno - EINVAL : Invalid parameter
> > > timer_create04    3  PASS  :  timer_create(2) expected failure; Got
> > > errno - EFAULT : Bad address
> > > timer_create04    4  PASS  :  timer_create(2) expected failure; Got
> > > errno - EFAULT : Bad address
> > > timer_create04    5  PASS  :  timer_create(2) expected failure; Got
> > > errno - EFAULT : Bad address
> > > timer_create04    6  PASS  :  timer_create(2) expected failure; Got
> > > errno - EFAULT : Bad address
>
> according to above, timer_create() always returns -EXXX ?

Aaah. I misread the first "FAIL" above. timer_create() succeeds!

hmm... it does timer_create(MAX_CLOCKS) and thus it should fail...

Can't find the original commit at
	http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git

but now we have CLOCK_MONOTONIC_RAW == 4, and MAX_CLOCKS == 4. So the
test should be fixed too, the first timer_create() should not fail on
2.6.28.

OK, sys_timer_create(CLOCK_MONOTONIC_RAW) calls
__hrtimer_init(CLOCK_MONOTONIC_RAW) and this looks just wrong:

	 timer->base = &cpu_base->clock_base[CLOCK_MONOTONIC_RAW];

while HRTIMER_MAX_CLOCK_BASES == 2. So time->base points to
"nowhere", this can explain the crash.

Thomas?

Oleg.