per-cpu data preempt-safing

PATCH New Robert Love 1 participant 1 message Sep 24, 2002

Subsystems (4)

Networking Network drivers Memory management MM core

lore

Share & tools

lore

Robert Love

author

2002-09-24 20:47 UTC #0 lore raw

PATCH 1/1 per-cpu data preempt-safing

Linus,

Attached patch fixes unsafe access to per-CPU data via reordering of instructions or use of "get_cpu()".

The following files were affected:

 include/linux/brlock.h     |    5 ++++-
 include/linux/netdevice.h  |   12 ++++++++----
 include/linux/page-flags.h |    6 +++---
 mm/highmem.c               |    2 ++
 4 files changed, 17 insertions(+), 8 deletions(-)

Before anyone balks at the brlock.h fix, note this was in the alternative version of the code which is not used by default.

Patch is against current BK. Please, apply.

Robert Love

include/linux/brlock.h +4 -1

diff -urN linux-2.5.38/include/linux/brlock.h linux/include/linux/brlock.h
@@ -85,7 +85,8 @@
 	if (idx >= __BR_END)
 		__br_lock_usage_bug();
 
-	read_lock(&__brlock_array[smp_processor_id()][idx]);
+	preempt_disable();
+	_raw_read_lock(&__brlock_array[smp_processor_id()][idx]);
 }
 
 static inline void br_read_unlock (enum brlock_indices idx)

@@ -109,6 +110,7 @@
 	if (idx >= __BR_END)
 		__br_lock_usage_bug();
 
+	preempt_disable();
 	ctr = &__brlock_array[smp_processor_id()][idx];
 	lock = &__br_write_locks[idx].lock;
 again:

@@ -147,6 +149,7 @@
 
 	wmb();
 	(*ctr)--;
+	preempt_enable();
 }
 #endif /* __BRLOCK_USE_ATOMICS */

include/linux/netdevice.h +8 -4

diff -urN linux-2.5.38/include/linux/netdevice.h linux/include/linux/netdevice.h
@@ -514,9 +514,10 @@
 {
 	if (!test_and_set_bit(__LINK_STATE_SCHED, &dev->state)) {
 		unsigned long flags;
-		int cpu = smp_processor_id();
+		int cpu;
 
 		local_irq_save(flags);
+		cpu = smp_processor_id();
 		dev->next_sched = softnet_data[cpu].output_queue;
 		softnet_data[cpu].output_queue = dev;
 		cpu_raise_softirq(cpu, NET_TX_SOFTIRQ);

@@ -563,10 +564,11 @@
 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
 {
 	if (atomic_dec_and_test(&skb->users)) {
-		int cpu =smp_processor_id();
+		int cpu;
 		unsigned long flags;
 
 		local_irq_save(flags);
+		cpu = smp_processor_id();
 		skb->next = softnet_data[cpu].completion_queue;
 		softnet_data[cpu].completion_queue = skb;
 		cpu_raise_softirq(cpu, NET_TX_SOFTIRQ);

@@ -726,9 +728,10 @@
 static inline void __netif_rx_schedule(struct net_device *dev)
 {
 	unsigned long flags;
-	int cpu = smp_processor_id();
+	int cpu;
 
 	local_irq_save(flags);
+	cpu = smp_processor_id();
 	dev_hold(dev);
 	list_add_tail(&dev->poll_list, &softnet_data[cpu].poll_list);
 	if (dev->quota < 0)

@@ -754,11 +757,12 @@
 {
 	if (netif_rx_schedule_prep(dev)) {
 		unsigned long flags;
-		int cpu = smp_processor_id();
+		int cpu;
 
 		dev->quota += undo;
 
 		local_irq_save(flags);
+		cpu = smp_processor_id();
 		list_add_tail(&dev->poll_list, &softnet_data[cpu].poll_list);
 		__cpu_raise_softirq(cpu, NET_RX_SOFTIRQ);
 		local_irq_restore(flags);

include/linux/page-flags.h +3 -3

diff -urN linux-2.5.38/include/linux/page-flags.h linux/include/linux/page-flags.h
@@ -86,9 +86,9 @@
 
 #define mod_page_state(member, delta)					\
 	do {								\
-		preempt_disable();					\
-		page_states[smp_processor_id()].member += (delta);	\
-		preempt_enable();					\
+		int cpu = get_cpu();					\
+		page_states[cpu].member += (delta);			\
+		put_cpu();						\
 	} while (0)
 
 #define inc_page_state(member)	mod_page_state(member, 1UL)

mm/highmem.c +2 -0

diff -urN linux-2.5.38/mm/highmem.c linux/mm/highmem.c
@@ -472,6 +472,7 @@
 {
 	int idx, type;
 
+	preempt_disable();
 	for (type = 0; type < KM_TYPE_NR; type++) {
 		idx = type + KM_TYPE_NR*smp_processor_id();
 		if (!pte_none(*(kmap_pte-idx))) {

@@ -479,6 +480,7 @@
 			BUG();
 		}
 	}
+	preempt_enable();
 }
 #endif