Jan Beulich
2007-Feb-07 17:10 UTC
[Xen-devel] [PATCH 2/4] domain heap allocator changes - per-bit-width heap zones
Replace the 3-zone scheme of the heap allocator with one with one where
zones are distinguished by their bit widths.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Index: 2007-02-07/xen/common/page_alloc.c
==================================================================---
2007-02-07.orig/xen/common/page_alloc.c 2007-02-07 16:26:42.000000000 +0100
+++ 2007-02-07/xen/common/page_alloc.c 2007-02-07 16:26:45.000000000 +0100
@@ -53,16 +53,18 @@ unsigned long max_dma_mfn = (1UL << (CON
static void parse_dma_bits(char *s)
{
unsigned int v = simple_strtol(s, NULL, 0);
- if ( v >= (sizeof(long)*8 + PAGE_SHIFT) )
+ if ( v >= (BITS_PER_LONG + PAGE_SHIFT) )
{
- dma_bitsize = sizeof(long)*8 + PAGE_SHIFT;
+ dma_bitsize = BITS_PER_LONG + PAGE_SHIFT;
max_dma_mfn = ~0UL;
}
- else
+ else if ( v > PAGE_SHIFT )
{
dma_bitsize = v;
max_dma_mfn = (1UL << (dma_bitsize - PAGE_SHIFT)) - 1;
}
+ else
+ printk("Invalid dma_bits value of %u ignored.\n", v);
}
custom_param("dma_bits", parse_dma_bits);
@@ -279,12 +281,13 @@ unsigned long alloc_boot_pages(unsigned
*/
#define MEMZONE_XEN 0
-#define MEMZONE_DOM 1
-#define MEMZONE_DMADOM 2
-#define NR_ZONES 3
+#ifdef PADDR_BITS
+#define NR_ZONES (PADDR_BITS - PAGE_SHIFT)
+#else
+#define NR_ZONES (BITS_PER_LONG - PAGE_SHIFT)
+#endif
-#define pfn_dom_zone_type(_pfn) \
- (((_pfn) <= max_dma_mfn) ? MEMZONE_DMADOM : MEMZONE_DOM)
+#define pfn_dom_zone_type(_pfn) (fls(_pfn) - 1)
static struct list_head heap[NR_ZONES][MAX_NUMNODES][MAX_ORDER+1];
@@ -294,15 +297,17 @@ static DEFINE_SPINLOCK(heap_lock);
/* Allocate 2^@order contiguous pages. */
static struct page_info *alloc_heap_pages(
- unsigned int zone, unsigned int cpu, unsigned int order)
+ unsigned int zone_lo, unsigned zone_hi,
+ unsigned int cpu, unsigned int order)
{
unsigned int i, j, node = cpu_to_node(cpu), num_nodes = num_online_nodes();
- unsigned int request = (1UL << order);
+ unsigned int zone, request = (1UL << order);
struct page_info *pg;
ASSERT(node >= 0);
ASSERT(node < num_nodes);
- ASSERT(zone < NR_ZONES);
+ ASSERT(zone_lo <= zone_hi);
+ ASSERT(zone_hi < NR_ZONES);
if ( unlikely(order > MAX_ORDER) )
return NULL;
@@ -315,14 +320,17 @@ static struct page_info *alloc_heap_page
* needless computation on fast-path */
for ( i = 0; i < num_nodes; i++ )
{
- /* check if target node can support the allocation */
- if ( avail[zone][node] >= request )
+ for ( zone = zone_hi; zone >= zone_lo; --zone )
{
- /* Find smallest order which can satisfy the request. */
- for ( j = order; j <= MAX_ORDER; j++ )
+ /* check if target node can support the allocation */
+ if ( avail[zone][node] >= request )
{
- if ( !list_empty(&heap[zone][node][j]) )
- goto found;
+ /* Find smallest order which can satisfy the request. */
+ for ( j = order; j <= MAX_ORDER; j++ )
+ {
+ if ( !list_empty(&heap[zone][node][j]) )
+ goto found;
+ }
}
}
/* pick next node, wrapping around if needed */
@@ -447,16 +455,17 @@ void init_heap_pages(
}
static unsigned long avail_heap_pages(
- int zone, int node)
+ unsigned int zone_lo, unsigned int zone_hi, unsigned int node)
{
- unsigned int i, j, num_nodes = num_online_nodes();
+ unsigned int i, zone, num_nodes = num_online_nodes();
unsigned long free_pages = 0;
- for (i=0; i<NR_ZONES; i++)
- if ( (zone == -1) || (zone == i) )
- for (j=0; j < num_nodes; j++)
- if ( (node == -1) || (node == j) )
- free_pages += avail[i][j];
+ if ( zone_hi >= NR_ZONES )
+ zone_hi = NR_ZONES - 1;
+ for ( zone = zone_lo; zone <= zone_hi; zone++ )
+ for ( i = 0; i < num_nodes; i++ )
+ if ( (node == -1) || (node == i) )
+ free_pages += avail[zone][i];
return free_pages;
}
@@ -576,7 +585,7 @@ void *alloc_xenheap_pages(unsigned int o
int i;
local_irq_save(flags);
- pg = alloc_heap_pages(MEMZONE_XEN, smp_processor_id(), order);
+ pg = alloc_heap_pages(MEMZONE_XEN, MEMZONE_XEN, smp_processor_id(), order);
local_irq_restore(flags);
if ( unlikely(pg == NULL) )
@@ -621,22 +630,26 @@ void free_xenheap_pages(void *v, unsigne
void init_domheap_pages(paddr_t ps, paddr_t pe)
{
- unsigned long s_tot, e_tot, s_dma, e_dma, s_nrm, e_nrm;
+ unsigned long s_tot, e_tot;
+ unsigned int zone;
ASSERT(!in_irq());
s_tot = round_pgup(ps) >> PAGE_SHIFT;
e_tot = round_pgdown(pe) >> PAGE_SHIFT;
- s_dma = min(s_tot, max_dma_mfn + 1);
- e_dma = min(e_tot, max_dma_mfn + 1);
- if ( s_dma < e_dma )
- init_heap_pages(MEMZONE_DMADOM, mfn_to_page(s_dma), e_dma - s_dma);
-
- s_nrm = max(s_tot, max_dma_mfn + 1);
- e_nrm = max(e_tot, max_dma_mfn + 1);
- if ( s_nrm < e_nrm )
- init_heap_pages(MEMZONE_DOM, mfn_to_page(s_nrm), e_nrm - s_nrm);
+ zone = fls(s_tot);
+ BUG_ON(zone <= MEMZONE_XEN + 1);
+ for ( --zone; s_tot < e_tot; ++zone )
+ {
+ unsigned long end = e_tot;
+
+ BUILD_BUG_ON(NR_ZONES > BITS_PER_LONG);
+ if ( zone < BITS_PER_LONG - 1 && end > 1UL << (zone
+ 1) )
+ end = 1UL << (zone + 1);
+ init_heap_pages(zone, mfn_to_page(s_tot), end - s_tot);
+ s_tot = end;
+ }
}
@@ -703,17 +716,21 @@ struct page_info *__alloc_domheap_pages(
if ( !(memflags & MEMF_dma) )
{
- pg = alloc_heap_pages(MEMZONE_DOM, cpu, order);
+ pg = alloc_heap_pages(dma_bitsize - PAGE_SHIFT, NR_ZONES - 1, cpu,
order);
/* Failure? Then check if we can fall back to the DMA pool. */
if ( unlikely(pg == NULL) &&
((order > MAX_ORDER) ||
- (avail_heap_pages(MEMZONE_DMADOM,-1) <
+ (avail_heap_pages(MEMZONE_XEN + 1,
+ dma_bitsize - PAGE_SHIFT - 1,
+ -1) <
(dma_emergency_pool_pages + (1UL << order)))) )
return NULL;
}
if ( pg == NULL )
- if ( (pg = alloc_heap_pages(MEMZONE_DMADOM, cpu, order)) == NULL )
+ if ( (pg = alloc_heap_pages(MEMZONE_XEN + 1,
+ dma_bitsize - PAGE_SHIFT - 1,
+ cpu, order)) == NULL )
return NULL;
mask = pg->u.free.cpumask;
@@ -835,9 +852,14 @@ unsigned long avail_domheap_pages(void)
{
unsigned long avail_nrm, avail_dma;
- avail_nrm = avail_heap_pages(MEMZONE_DOM,-1);
+ avail_nrm = avail_heap_pages(dma_bitsize - PAGE_SHIFT,
+ NR_ZONES - 1,
+ -1);
+
+ avail_dma = avail_heap_pages(MEMZONE_XEN + 1,
+ dma_bitsize - PAGE_SHIFT - 1,
+ -1);
- avail_dma = avail_heap_pages(MEMZONE_DMADOM,-1);
if ( avail_dma > dma_emergency_pool_pages )
avail_dma -= dma_emergency_pool_pages;
else
@@ -848,18 +870,33 @@ unsigned long avail_domheap_pages(void)
unsigned long avail_nodeheap_pages(int node)
{
- return avail_heap_pages(-1, node);
+ return avail_heap_pages(0, NR_ZONES - 1, node);
}
static void pagealloc_keyhandler(unsigned char key)
{
+ unsigned int zone = MEMZONE_XEN;
+ unsigned long total = 0;
+
printk("Physical memory information:\n");
- printk(" Xen heap: %lukB free\n"
- " DMA heap: %lukB free\n"
- " Dom heap: %lukB free\n",
- avail_heap_pages(MEMZONE_XEN, -1) << (PAGE_SHIFT-10),
- avail_heap_pages(MEMZONE_DMADOM, -1) <<(PAGE_SHIFT-10),
- avail_heap_pages(MEMZONE_DOM, -1) <<(PAGE_SHIFT-10));
+ printk(" Xen heap: %lukB free\n",
+ avail_heap_pages(zone, zone, -1) << (PAGE_SHIFT-10));
+
+ while ( ++zone < NR_ZONES )
+ {
+ unsigned long n;
+
+ if ( zone == dma_bitsize - PAGE_SHIFT )
+ {
+ printk(" DMA heap: %lukB free\n", total <<
(PAGE_SHIFT-10));
+ total = 0;
+ }
+ n = avail_heap_pages(zone, zone, -1);
+ total += n;
+ if ( n )
+ printk(" heap[%02u]: %lukB free\n", zone, n <<
(PAGE_SHIFT-10));
+ }
+ printk(" Dom heap: %lukB free\n", total <<
(PAGE_SHIFT-10));
}
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel