real-time enhanced page allocator and throttling

diff -urN linux-2.6.0-test2-mm4/include/linux/sched.h linux/include/linux/sched.h
@@ -282,7 +282,9 @@
 #define MAX_RT_PRIO		MAX_USER_RT_PRIO
 
 #define MAX_PRIO		(MAX_RT_PRIO + 40)
- 
+
+#define rt_task(p)		((p)->prio < MAX_RT_PRIO)
+
 /*
  * Some day this will be a full-fledged user tracking system..
  */

diff -urN linux-2.6.0-test2-mm4/kernel/sched.c linux/kernel/sched.c
@@ -199,7 +199,6 @@
 #define this_rq()		(cpu_rq(smp_processor_id())) /* not __get_cpu_var(runqueues)! */
 #define task_rq(p)		cpu_rq(task_cpu(p))
 #define cpu_curr(cpu)		(cpu_rq(cpu)->curr)
-#define rt_task(p)		((p)->prio < MAX_RT_PRIO)
 
 /*
  * Default context-switch locking:

diff -urN linux-2.6.0-test2-mm4/mm/page_alloc.c linux/mm/page_alloc.c
@@ -518,7 +518,8 @@
  *
  * Herein lies the mysterious "incremental min".  That's the
  *
- *	min += z->pages_low;
+ *	local_low = z->pages_low;
+ *	min += local_low;
  *
  * thing.  The intent here is to provide additional protection to low zones for
  * allocation requests which _could_ use higher zones.  So a GFP_HIGHMEM

@@ -536,10 +537,11 @@
 	unsigned long min;
 	struct zone **zones, *classzone;
 	struct page *page;
+	struct reclaim_state reclaim_state;
+	struct task_struct *p = current;
 	int i;
 	int cold;
 	int do_retry;
-	struct reclaim_state reclaim_state;
 
 	if (wait)
 		might_sleep();

@@ -557,8 +559,17 @@
 	min = 1UL << order;
 	for (i = 0; zones[i] != NULL; i++) {
 		struct zone *z = zones[i];
+		unsigned long local_low;
+
+		/*
+		 * This is the fabled 'incremental min'. We let real-time tasks
+		 * dip their real-time paws a little deeper into reserves.
+		 */
+		local_low = z->pages_low;
+		if (rt_task(p))
+			local_low >>= 1;
+		min += local_low;
 
-		min += z->pages_low;
 		if (z->free_pages >= min ||
 				(!wait && z->free_pages >= z->pages_high)) {
 			page = buffered_rmqueue(z, order, cold);

@@ -594,7 +605,7 @@
 	/* here we're in the low on memory slow path */
 
 rebalance:
-	if ((current->flags & (PF_MEMALLOC | PF_MEMDIE)) && !in_interrupt()) {
+	if ((p->flags & (PF_MEMALLOC | PF_MEMDIE)) && !in_interrupt()) {
 		/* go through the zonelist yet again, ignoring mins */
 		for (i = 0; zones[i] != NULL; i++) {
 			struct zone *z = zones[i];

@@ -610,14 +621,14 @@
 	if (!wait)
 		goto nopage;
 
-	current->flags |= PF_MEMALLOC;
+	p->flags |= PF_MEMALLOC;
 	reclaim_state.reclaimed_slab = 0;
-	current->reclaim_state = &reclaim_state;
+	p->reclaim_state = &reclaim_state;
 
 	try_to_free_pages(classzone, gfp_mask, order);
 
-	current->reclaim_state = NULL;
-	current->flags &= ~PF_MEMALLOC;
+	p->reclaim_state = NULL;
+	p->flags &= ~PF_MEMALLOC;
 
 	/* go through the zonelist yet one more time */
 	min = 1UL << order;

@@ -657,7 +668,7 @@
 	if (!(gfp_mask & __GFP_NOWARN)) {
 		printk("%s: page allocation failure."
 			" order:%d, mode:0x%x\n",
-			current->comm, order, gfp_mask);
+			p->comm, order, gfp_mask);
 	}
 	return NULL;
 got_pg:

diff -urN linux-2.6.0-test2-mm4/mm/page-writeback.c linux/mm/page-writeback.c
@@ -227,7 +227,11 @@
 	if (dirty_exceeded)
 		ratelimit = 8;
 
-	if (get_cpu_var(ratelimits)++ >= ratelimit) {
+	/*
+	 * Check the rate limiting. Also, we do not want to throttle real-time
+	 * tasks in balance_dirty_pages(). Period.
+	 */
+	if (get_cpu_var(ratelimits)++ >= ratelimit && !rt_task(current)) {
 		__get_cpu_var(ratelimits) = 0;
 		put_cpu_var(ratelimits);
 		return balance_dirty_pages(mapping);

diff -puN mm/page_alloc.c~rt-tasks-special-vm-treatment-2 mm/page_alloc.c
@@ -592,6 +592,8 @@ __alloc_pages(unsigned int gfp_mask, uns
 		local_min = z->pages_min;
 		if (gfp_mask & __GFP_HIGH)
 			local_min >>= 2;
+		if (rt_task(p))
+			local_min >>= 1;
 		min += local_min;
 		if (z->free_pages >= min ||
 				(!wait && z->free_pages >= z->pages_high)) {

diff -puN mm/page-writeback.c~rt-tasks-special-vm-treatment-2 mm/page-writeback.c
@@ -144,7 +144,7 @@ get_dirty_limits(struct page_state *ps, 
  * If we're over `background_thresh' then pdflush is woken to perform some
  * writeout.
  */
-void balance_dirty_pages(struct address_space *mapping)
+static void balance_dirty_pages(struct address_space *mapping)
 {
 	struct page_state ps;
 	long nr_reclaimable;

@@ -167,8 +167,9 @@ void balance_dirty_pages(struct address_
 		nr_reclaimable = ps.nr_dirty + ps.nr_unstable;
 		if (nr_reclaimable + ps.nr_writeback <= dirty_thresh)
 			break;
-
 		dirty_exceeded = 1;
+		if (rt_task(current))
+			break;
 
 		/* Note: nr_reclaimable denotes nr_dirty + nr_unstable.
 		 * Unstable writes are a feature of certain networked

@@ -223,7 +224,7 @@ void balance_dirty_pages_ratelimited(str
 	 * Check the rate limiting. Also, we do not want to throttle real-time
 	 * tasks in balance_dirty_pages(). Period.
 	 */
-	if (get_cpu_var(ratelimits)++ >= ratelimit && !rt_task(current)) {
+	if (get_cpu_var(ratelimits)++ >= ratelimit) {
 		__get_cpu_var(ratelimits) = 0;
 		put_cpu_var(ratelimits);
 		balance_dirty_pages(mapping);