Subject: [PATCH] Use raw_smp_processor_id() in rcu_random()


With CONFIG_PREEMPT=y on 2.6.23-rc8-mm2, rcu_random() calls
smp_processor_id() from a preemptible context, triggering the
following kernel error.

BUG: using smp_processor_id() in preemptible [00000001] code: rcu_torture_fak/5155
caller is rcu_random+0x16/0x39 [rcutorture]
[<c010612f>] dump_trace+0x68/0x1d2
[<c01062b1>] show_trace_log_lvl+0x18/0x2c
[<c0106c23>] show_trace+0xf/0x11
[<c0106d33>] dump_stack+0x12/0x14
[<c02a0a25>] debug_smp_processor_id+0xa1/0xb4
[<f887064c>] rcu_random+0x16/0x39 [rcutorture]
[<f8870cdf>] <3>BUG: using smp_processor_id() in preemptible [00000001] code: rcu_torture_fak/5156
rcu_torture_fakewriter+0x4d/0xc5 [rcutorture]
[<c0139b32>] caller is rcu_random+0x16/0x39 [rcutorture]
kthread+0x38/0x5e
[<c0105ca7>] kernel_thread_helper+0x7/0x10
=======================
[<c010612f>] dump_trace+0x68/0x1d2
[<c01062b1>] show_trace_log_lvl+0x18/0x2c
[<c0106c23>] show_trace+0xf/0x11
[<c0106d33>] dump_stack+0x12/0x14
[<c02a0a25>] debug_smp_processor_id+0xa1/0xb4
[<f887064c>] rcu_random+0x16/0x39 [rcutorture]
[<f8870cdf>] rcu_torture_fakewriter+0x4d/0xc5 [rcutorture]

rcu_random() can do with raw_smp_processor_id() as a parameter to cpu_clock()
in this particular context.

Signed-off-by: Gautham R Shenoy <[email protected]>

---
kernel/rcutorture.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

Index: linux-2.6.23-rc8/kernel/rcutorture.c
===================================================================
--- linux-2.6.23-rc8.orig/kernel/rcutorture.c
+++ linux-2.6.23-rc8/kernel/rcutorture.c
@@ -170,7 +170,8 @@ static unsigned long
rcu_random(struct rcu_random_state *rrsp)
{
if (--rrsp->rrs_count < 0) {
- rrsp->rrs_state += (unsigned long)cpu_clock(smp_processor_id());
+ rrsp->rrs_state +=
+ (unsigned long)cpu_clock(raw_smp_processor_id());
rrsp->rrs_count = RCU_RANDOM_REFRESH;
}
rrsp->rrs_state = rrsp->rrs_state * RCU_RANDOM_MULT + RCU_RANDOM_ADD;
--
Gautham R Shenoy
Linux Technology Center
IBM India.
"Freedom comes with a price tag of responsibility, which is still a bargain,
because Freedom is priceless!"


2007-10-12 09:39:29

by Josh Triplett

[permalink] [raw]
Subject: Re: [PATCH] Use raw_smp_processor_id() in rcu_random()

Gautham R Shenoy wrote:
> With CONFIG_PREEMPT=y on 2.6.23-rc8-mm2, rcu_random() calls
> smp_processor_id() from a preemptible context, triggering the
> following kernel error.
>
> BUG: using smp_processor_id() in preemptible [00000001] code: rcu_torture_fak/5155
> caller is rcu_random+0x16/0x39 [rcutorture]
> [<c010612f>] dump_trace+0x68/0x1d2
> [<c01062b1>] show_trace_log_lvl+0x18/0x2c
> [<c0106c23>] show_trace+0xf/0x11
> [<c0106d33>] dump_stack+0x12/0x14
> [<c02a0a25>] debug_smp_processor_id+0xa1/0xb4
> [<f887064c>] rcu_random+0x16/0x39 [rcutorture]
> [<f8870cdf>] <3>BUG: using smp_processor_id() in preemptible [00000001] code: rcu_torture_fak/5156
> rcu_torture_fakewriter+0x4d/0xc5 [rcutorture]
> [<c0139b32>] caller is rcu_random+0x16/0x39 [rcutorture]
> kthread+0x38/0x5e
> [<c0105ca7>] kernel_thread_helper+0x7/0x10
> =======================
> [<c010612f>] dump_trace+0x68/0x1d2
> [<c01062b1>] show_trace_log_lvl+0x18/0x2c
> [<c0106c23>] show_trace+0xf/0x11
> [<c0106d33>] dump_stack+0x12/0x14
> [<c02a0a25>] debug_smp_processor_id+0xa1/0xb4
> [<f887064c>] rcu_random+0x16/0x39 [rcutorture]
> [<f8870cdf>] rcu_torture_fakewriter+0x4d/0xc5 [rcutorture]
>
> rcu_random() can do with raw_smp_processor_id() as a parameter to cpu_clock()
> in this particular context.
>
> Signed-off-by: Gautham R Shenoy <[email protected]>

Looks reasonable to me.

Acked-by: Josh Triplett <[email protected]>

2007-10-12 19:21:29

by Paul E. McKenney

[permalink] [raw]
Subject: Re: [PATCH] Use raw_smp_processor_id() in rcu_random()

On Fri, Oct 12, 2007 at 02:50:44PM +0530, Gautham R Shenoy wrote:
>
> With CONFIG_PREEMPT=y on 2.6.23-rc8-mm2, rcu_random() calls
> smp_processor_id() from a preemptible context, triggering the
> following kernel error.
>
> BUG: using smp_processor_id() in preemptible [00000001] code: rcu_torture_fak/5155
> caller is rcu_random+0x16/0x39 [rcutorture]
> [<c010612f>] dump_trace+0x68/0x1d2
> [<c01062b1>] show_trace_log_lvl+0x18/0x2c
> [<c0106c23>] show_trace+0xf/0x11
> [<c0106d33>] dump_stack+0x12/0x14
> [<c02a0a25>] debug_smp_processor_id+0xa1/0xb4
> [<f887064c>] rcu_random+0x16/0x39 [rcutorture]
> [<f8870cdf>] <3>BUG: using smp_processor_id() in preemptible [00000001] code: rcu_torture_fak/5156
> rcu_torture_fakewriter+0x4d/0xc5 [rcutorture]
> [<c0139b32>] caller is rcu_random+0x16/0x39 [rcutorture]
> kthread+0x38/0x5e
> [<c0105ca7>] kernel_thread_helper+0x7/0x10
> =======================
> [<c010612f>] dump_trace+0x68/0x1d2
> [<c01062b1>] show_trace_log_lvl+0x18/0x2c
> [<c0106c23>] show_trace+0xf/0x11
> [<c0106d33>] dump_stack+0x12/0x14
> [<c02a0a25>] debug_smp_processor_id+0xa1/0xb4
> [<f887064c>] rcu_random+0x16/0x39 [rcutorture]
> [<f8870cdf>] rcu_torture_fakewriter+0x4d/0xc5 [rcutorture]
>
> rcu_random() can do with raw_smp_processor_id() as a parameter to cpu_clock()
> in this particular context.

Good catch!!!

Acked-by: Paul E. McKenney <[email protected]>

> Signed-off-by: Gautham R Shenoy <[email protected]>
>
> ---
> kernel/rcutorture.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> Index: linux-2.6.23-rc8/kernel/rcutorture.c
> ===================================================================
> --- linux-2.6.23-rc8.orig/kernel/rcutorture.c
> +++ linux-2.6.23-rc8/kernel/rcutorture.c
> @@ -170,7 +170,8 @@ static unsigned long
> rcu_random(struct rcu_random_state *rrsp)
> {
> if (--rrsp->rrs_count < 0) {
> - rrsp->rrs_state += (unsigned long)cpu_clock(smp_processor_id());
> + rrsp->rrs_state +=
> + (unsigned long)cpu_clock(raw_smp_processor_id());
> rrsp->rrs_count = RCU_RANDOM_REFRESH;
> }
> rrsp->rrs_state = rrsp->rrs_state * RCU_RANDOM_MULT + RCU_RANDOM_ADD;
> --
> Gautham R Shenoy
> Linux Technology Center
> IBM India.
> "Freedom comes with a price tag of responsibility, which is still a bargain,
> because Freedom is priceless!"