Thomas,
I think I've gotten to the heart of the problem. Here's an excerpt
from the latest -rt patch: net/core/dev.c in the function dev_queue_xmit
@@ -1588,11 +1588,17 @@ gso:
Either shot noqueue qdisc, it is even simpler 8)
*/
if (dev->flags & IFF_UP) {
- int cpu = smp_processor_id(); /* ok because BHs are off */
+ /*
+ * No need to check for recursion with threaded interrupts:
+ */
+#ifdef CONFIG_PREEMPT_RT
+ if (1) {
+#else
+ int cpu = raw_smp_processor_id(); /* ok because BHs are off */
if (dev->xmit_lock_owner != cpu) {
-
- HARD_TX_LOCK(dev, cpu);
+#endif
+ HARD_TX_LOCK(dev);
if (!netif_queue_stopped(dev) &&
!netif_subqueue_stopped(dev, skb->queue_mapping)) {
I'm not sure why the check for recursion has been removed.
In the backtrace below, I think it would be caught by this check and
not recursively call the spinlock code.
> Call Trace:
>
> [D010BB40] [C01C7BA8] rt_spin_lock_slowlock+0x44/0x1f8 (unreliable)
>
> [D010BB90] [C0153464] dev_queue_xmit+0x298/0x2a0 Tunnel2
>
> [D010BBB0] [C0176398] ip_output+0x288/0x2dc
>
> [D010BBE0] [C01AC078] ipip_tunnel_xmit+0x508/0x698
>
> [D010BC60] [C0150DF4] dev_hard_start_xmit+0x1b4/0x2a4
>
> [D010BC80] [C0153430] dev_queue_xmit+0x264/0x2a0 Tunnel4
>
> [D010BCA0] [C0176398] ip_output+0x288/0x2dc
>
> [D010BCD0] [C01AC078] ipip_tunnel_xmit+0x508/0x698
>
> [D010BD50] [C0150DF4] dev_hard_start_xmit+0x1b4/0x2a4
>
> [D010BD70] [C0153430] dev_queue_xmit+0x264/0x2a0 Tunnel2
>
> [D010BD90] [C0176398] ip_output+0x288/0x2dc
>
> [D010BDC0] [C017685C] ip_queue_xmit+0x1ac/0x4e4
>
> [D010BE30] [C018762C] tcp_transmit_skb+0x390/0x810
>
> [D010BE70] [C018882C] tcp_retransmit_skb+0x160/0x638
>
> [D010BEA0] [C018BA5C] tcp_write_timer+0x274/0x6c0
>
> [D010BED0] [C0024314] run_timer_softirq+0x2d0/0xedc
>
> [D010BF80] [C001F1C4] ksoftirqd+0xf8/0x1b0
>
> [D010BFC0] [C0031588] kthread+0xc0/0xfc
>
> [D010BFF0] [C000471C] kernel_thread+0x44/0x60
>
I found one other place in the code which appears to do the same thing.
Although it is written to handle smp collisions, I think it should also
handle the error case above.
Index: linux-rt-rebase.q/net/sched/sch_generic.c
===================================================================
--- linux-rt-rebase.q.orig/net/sched/sch_generic.c
+++ linux-rt-rebase.q/net/sched/sch_generic.c
@@ -12,6 +12,7 @@
*/
#include <linux/bitops.h>
+#include <linux/kallsyms.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/kernel.h>
@@ -150,16 +151,28 @@ static inline int qdisc_restart(struct n
*/
lockless = (dev->features & NETIF_F_LLTX);
- if (!lockless && !netif_tx_trylock(dev)) {
- /* Another CPU grabbed the driver tx lock */
- return handle_dev_cpu_collision(skb, dev, q);
+ if (!lockless) {
+#ifdef CONFIG_PREEMPT_RT
+ netif_tx_lock(dev);
+#else
+ if (netif_tx_trylock(dev))
+ /* Another CPU grabbed the driver tx lock */
+ return handle_dev_cpu_collision(skb, dev, q);
+#endif
What do you think?
Thanks,
Mark
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
