[PATCH] Hot unplug CPU 3/4

From: Rusty Russell (rusty@rustcorp.com.au)
Date: Thu Sep 12 2002 - 20:55:09 EST


Name: set_cpus_allowed can fail
Author: Rusty Russell
Status: Tested on 2.5.30 SMP
Depends: Hotcpu/cpumask.patch.gz

D: Simply moves the "is this mask valid?" check inside set_cpus_allowed,
D: since it might be racy otherwise in the context of adding/removing
D: CPUs. This also introduces the CPU brlock.

diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32767-linux-2.5.33/arch/i386/kernel/smpboot.c .32767-linux-2.5.33.updated/arch/i386/kernel/smpboot.c
--- .32767-linux-2.5.33/arch/i386/kernel/smpboot.c 2002-09-04 16:04:52.000000000 +1000
+++ .32767-linux-2.5.33.updated/arch/i386/kernel/smpboot.c 2002-09-04 16:05:11.000000000 +1000
@@ -45,6 +45,7 @@
 
 #include <linux/delay.h>
 #include <linux/mc146818rtc.h>
+#include <linux/brlock.h>
 #include <asm/mtrr.h>
 #include <asm/pgalloc.h>
 #include <asm/tlbflush.h>
@@ -458,8 +459,9 @@ int __init start_secondary(void *unused)
          * the local TLBs too.
          */
         local_flush_tlb();
- set_bit(smp_processor_id(), cpu_online_map);
- wmb();
+ br_write_lock_irq(BR_CPU_LOCK);
+ set_bit(smp_processor_id(), cpu_online_map);
+ br_write_unlock_irq(BR_CPU_LOCK);
         return cpu_idle();
 }
 
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32767-linux-2.5.33/arch/ppc/kernel/smp.c .32767-linux-2.5.33.updated/arch/ppc/kernel/smp.c
--- .32767-linux-2.5.33/arch/ppc/kernel/smp.c 2002-09-04 16:04:52.000000000 +1000
+++ .32767-linux-2.5.33.updated/arch/ppc/kernel/smp.c 2002-09-04 16:05:11.000000000 +1000
@@ -432,7 +432,9 @@ int __cpu_up(unsigned int cpu)
         printk("Processor %d found.\n", cpu);
 
         smp_ops->give_timebase();
- set_bit(cpu, cpu_online_map);
+ br_write_lock_irq(BR_CPU_LOCK);
+ set_bit(cpu, &cpu_online_map);
+ br_write_unlock_irq(BR_CPU_LOCK);
         return 0;
 }
 
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32767-linux-2.5.33/include/linux/brlock.h .32767-linux-2.5.33.updated/include/linux/brlock.h
--- .32767-linux-2.5.33/include/linux/brlock.h 2002-08-28 09:29:52.000000000 +1000
+++ .32767-linux-2.5.33.updated/include/linux/brlock.h 2002-09-04 16:05:11.000000000 +1000
@@ -34,6 +34,7 @@
 /* Register bigreader lock indices here. */
 enum brlock_indices {
         BR_NETPROTO_LOCK,
+ BR_CPU_LOCK,
         __BR_END
 };
 
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32767-linux-2.5.33/include/linux/sched.h .32767-linux-2.5.33.updated/include/linux/sched.h
--- .32767-linux-2.5.33/include/linux/sched.h 2002-09-04 16:04:53.000000000 +1000
+++ .32767-linux-2.5.33.updated/include/linux/sched.h 2002-09-04 16:05:11.000000000 +1000
@@ -425,9 +425,14 @@ do { if (atomic_dec_and_test(&(tsk)->usa
 #define _STK_LIM (8*1024*1024)
 
 #if CONFIG_SMP
-extern void set_cpus_allowed(task_t *p, const unsigned long new_mask[]);
+extern int set_cpus_allowed(task_t *p, const unsigned long new_mask[]);
 #else
-# define set_cpus_allowed(p, new_mask) do { } while (0)
+static inline int set_cpus_allowed(struct task_struct *p,
+ const unsigned long *new_mask)
+{
+ if (!(*new_mask & 1)) return -EINVAL;
+ return 0;
+}
 #endif
 #define CPU_MASK_NONE { 0 }
 #define CPU_MASK_ALL { [0 ... BITS_TO_LONG(NR_CPUS)-1 ] = ~0UL }
@@ -435,9 +440,9 @@ extern void set_cpus_allowed(task_t *p,
 static inline void migrate_to_cpu(unsigned int cpu)
 {
         DECLARE_BITMAP(mask, NR_CPUS) = CPU_MASK_NONE;
- BUG_ON(!cpu_online(cpu));
         __set_bit(cpu, mask);
- set_cpus_allowed(current, mask);
+ if (set_cpus_allowed(current, mask) != 0)
+ BUG();
 }
 
 extern void set_user_nice(task_t *p, long nice);
diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .32767-linux-2.5.33/kernel/sched.c .32767-linux-2.5.33.updated/kernel/sched.c
--- .32767-linux-2.5.33/kernel/sched.c 2002-09-04 16:04:53.000000000 +1000
+++ .32767-linux-2.5.33.updated/kernel/sched.c 2002-09-04 16:05:32.000000000 +1000
@@ -30,6 +30,7 @@
 #include <linux/notifier.h>
 #include <linux/delay.h>
 #include <linux/bitops.h>
+#include <linux/brlock.h>
 
 /*
  * Convert user-nice values [ -20 ... 0 ... 19 ]
@@ -1553,9 +1554,6 @@ asmlinkage int sys_sched_setaffinity(pid
         if (copy_from_user(&new_mask, user_mask_ptr, sizeof(new_mask)))
                 return -EFAULT;
 
- if (any_online_cpu(new_mask) == NR_CPUS)
- return -EINVAL;
-
         read_lock(&tasklist_lock);
 
         p = find_process_by_pid(pid);
@@ -1577,8 +1575,7 @@ asmlinkage int sys_sched_setaffinity(pid
                         !capable(CAP_SYS_NICE))
                 goto out_unlock;
 
- retval = 0;
- set_cpus_allowed(p, new_mask);
+ retval = set_cpus_allowed(p, new_mask);
 
 out_unlock:
         put_task_struct(p);
@@ -1914,29 +1911,30 @@ typedef struct {
  * task must not exit() & deallocate itself prematurely. The
  * call is not atomic; no spinlocks may be held.
  */
-void set_cpus_allowed(task_t *p, const unsigned long new_mask[])
+int set_cpus_allowed(task_t *p, const unsigned long new_mask[])
 {
         unsigned long flags;
         migration_req_t req;
         runqueue_t *rq;
-
-#if 0 /* FIXME: Grab cpu_lock, return error on this case. --RR */
- new_mask &= cpu_online_map;
- if (!new_mask)
- BUG();
-#endif
+ int ret = 0;
 
         preempt_disable();
         rq = task_rq_lock(p, &flags);
+ /* Let's not deal with cpus leaving us */
+ br_read_lock(BR_CPU_LOCK);
+ if (any_online_cpu(new_mask) == NR_CPUS) {
+ ret = -EINVAL;
+ goto out_unlock;
+ }
+
         memcpy(p->cpus_allowed, new_mask, sizeof(p->cpus_allowed));
         /*
          * Can the task run on the task's current CPU? If not then
          * migrate the thread off to a proper CPU.
          */
- if (test_bit(task_cpu(p), new_mask)) {
- task_rq_unlock(rq, &flags);
- goto out;
- }
+ if (test_bit(task_cpu(p), new_mask))
+ goto out_unlock;
+
         /*
          * If the task is not on a runqueue (and not running), then
          * it is sufficient to simply update the task's cpu field.
@@ -1944,17 +1942,24 @@ void set_cpus_allowed(task_t *p, const u
         if (!p->array && !task_running(rq, p)) {
                 set_task_cpu(p, any_online_cpu(p->cpus_allowed));
                 task_rq_unlock(rq, &flags);
- goto out;
+ goto out_unlock;
         }
         init_completion(&req.done);
         req.task = p;
         list_add(&req.list, &rq->migration_queue);
+ br_read_unlock(BR_CPU_LOCK);
         task_rq_unlock(rq, &flags);
         wake_up_process(rq->migration_thread);
 
         wait_for_completion(&req.done);
-out:
         preempt_enable();
+ return 0;
+
+ out_unlock:
+ br_read_unlock(BR_CPU_LOCK);
+ task_rq_unlock(rq, &flags);
+ preempt_enable();
+ return ret;
 }
 
 /*

--
  Anyone who quotes me in their sig is an idiot. -- Rusty Russell.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Sun Sep 15 2002 - 22:00:31 EST