aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86/mm/pat.c
diff options
context:
space:
mode:
authorVenkatesh Pallipadi <venkatesh.pallipadi@intel.com>2009-07-10 12:57:36 -0400
committerH. Peter Anvin <hpa@zytor.com>2009-08-26 18:41:19 -0400
commit335ef896d4c6639849d79367f0fef9abc06d121b (patch)
tree6567e3c9512a7c8bd31a324783d8c427baf2145d /arch/x86/mm/pat.c
parent9e36fda0b359d2a6ae039c3d7e71a04502a77898 (diff)
x86, pat: Add rbtree to do quick lookup in memtype tracking
PAT memtype tracking uses a linear link list to keep track of IO (non-RAM) regions and their memtypes. The code used a last_accessed pointer as a cache to speedup the lookup. As per discussions with H. Peter Anvin a while back, having a rbtree here will avoid bad performances in pathological cases where we may end up with huge linked list. This may not add any noticable performance speedup in normal case as the number of entires in PAT memtype list tend to be ~20-30 range. The patch removes the "cached_entry" logic as with rbtree we have more generic way of speeding up the lookup. With this patch, we use rbtree to do the quick lookup. We still use linked list as the memtype range tracked can be of different sizes and can overlap in different ways. We also keep track of usage counts with linked list. Example: Multiple ioremaps with different sizes uncached-minus @ 0xfffff00000-0xfffff04000 uncached-minus @ 0xfffff02000-0xfffff03000 And one userlevel mmap and the thread forks a new process uncached-minus @ 0xbf453000-0xbf454000 uncached-minus @ 0xbf453000-0xbf454000 Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Diffstat (limited to 'arch/x86/mm/pat.c')
-rw-r--r--arch/x86/mm/pat.c106
1 files changed, 86 insertions, 20 deletions
diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c
index 82d097ce3091..c90f2420f56c 100644
--- a/arch/x86/mm/pat.c
+++ b/arch/x86/mm/pat.c
@@ -15,6 +15,7 @@
15#include <linux/gfp.h> 15#include <linux/gfp.h>
16#include <linux/mm.h> 16#include <linux/mm.h>
17#include <linux/fs.h> 17#include <linux/fs.h>
18#include <linux/rbtree.h>
18 19
19#include <asm/cacheflush.h> 20#include <asm/cacheflush.h>
20#include <asm/processor.h> 21#include <asm/processor.h>
@@ -148,11 +149,10 @@ static char *cattr_name(unsigned long flags)
148 * areas). All the aliases have the same cache attributes of course. 149 * areas). All the aliases have the same cache attributes of course.
149 * Zero attributes are represented as holes. 150 * Zero attributes are represented as holes.
150 * 151 *
151 * Currently the data structure is a list because the number of mappings 152 * The data structure is a list that is also organized as an rbtree
152 * are expected to be relatively small. If this should be a problem 153 * sorted on the start address of memtype range.
153 * it could be changed to a rbtree or similar.
154 * 154 *
155 * memtype_lock protects the whole list. 155 * memtype_lock protects both the linear list and rbtree.
156 */ 156 */
157 157
158struct memtype { 158struct memtype {
@@ -160,11 +160,53 @@ struct memtype {
160 u64 end; 160 u64 end;
161 unsigned long type; 161 unsigned long type;
162 struct list_head nd; 162 struct list_head nd;
163 struct rb_node rb;
163}; 164};
164 165
166static struct rb_root memtype_rbroot = RB_ROOT;
165static LIST_HEAD(memtype_list); 167static LIST_HEAD(memtype_list);
166static DEFINE_SPINLOCK(memtype_lock); /* protects memtype list */ 168static DEFINE_SPINLOCK(memtype_lock); /* protects memtype list */
167 169
170static struct memtype *memtype_rb_search(struct rb_root *root, u64 start)
171{
172 struct rb_node *node = root->rb_node;
173 struct memtype *last_lower = NULL;
174
175 while (node) {
176 struct memtype *data = container_of(node, struct memtype, rb);
177
178 if (data->start < start) {
179 last_lower = data;
180 node = node->rb_right;
181 } else if (data->start > start) {
182 node = node->rb_left;
183 } else
184 return data;
185 }
186
187 /* Will return NULL if there is no entry with its start <= start */
188 return last_lower;
189}
190
191static void memtype_rb_insert(struct rb_root *root, struct memtype *data)
192{
193 struct rb_node **new = &(root->rb_node);
194 struct rb_node *parent = NULL;
195
196 while (*new) {
197 struct memtype *this = container_of(*new, struct memtype, rb);
198
199 parent = *new;
200 if (data->start <= this->start)
201 new = &((*new)->rb_left);
202 else if (data->start > this->start)
203 new = &((*new)->rb_right);
204 }
205
206 rb_link_node(&data->rb, parent, new);
207 rb_insert_color(&data->rb, root);
208}
209
168/* 210/*
169 * Does intersection of PAT memory type and MTRR memory type and returns 211 * Does intersection of PAT memory type and MTRR memory type and returns
170 * the resulting memory type as PAT understands it. 212 * the resulting memory type as PAT understands it.
@@ -218,9 +260,6 @@ chk_conflict(struct memtype *new, struct memtype *entry, unsigned long *type)
218 return -EBUSY; 260 return -EBUSY;
219} 261}
220 262
221static struct memtype *cached_entry;
222static u64 cached_start;
223
224static int pat_pagerange_is_ram(unsigned long start, unsigned long end) 263static int pat_pagerange_is_ram(unsigned long start, unsigned long end)
225{ 264{
226 int ram_page = 0, not_rampage = 0; 265 int ram_page = 0, not_rampage = 0;
@@ -382,17 +421,19 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type,
382 421
383 spin_lock(&memtype_lock); 422 spin_lock(&memtype_lock);
384 423
385 if (cached_entry && start >= cached_start) 424 entry = memtype_rb_search(&memtype_rbroot, new->start);
386 entry = cached_entry; 425 if (likely(entry != NULL)) {
387 else 426 /* To work correctly with list_for_each_entry_continue */
427 entry = list_entry(entry->nd.prev, struct memtype, nd);
428 } else {
388 entry = list_entry(&memtype_list, struct memtype, nd); 429 entry = list_entry(&memtype_list, struct memtype, nd);
430 }
389 431
390 /* Search for existing mapping that overlaps the current range */ 432 /* Search for existing mapping that overlaps the current range */
391 where = NULL; 433 where = NULL;
392 list_for_each_entry_continue(entry, &memtype_list, nd) { 434 list_for_each_entry_continue(entry, &memtype_list, nd) {
393 if (end <= entry->start) { 435 if (end <= entry->start) {
394 where = entry->nd.prev; 436 where = entry->nd.prev;
395 cached_entry = list_entry(where, struct memtype, nd);
396 break; 437 break;
397 } else if (start <= entry->start) { /* end > entry->start */ 438 } else if (start <= entry->start) { /* end > entry->start */
398 err = chk_conflict(new, entry, new_type); 439 err = chk_conflict(new, entry, new_type);
@@ -400,8 +441,6 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type,
400 dprintk("Overlap at 0x%Lx-0x%Lx\n", 441 dprintk("Overlap at 0x%Lx-0x%Lx\n",
401 entry->start, entry->end); 442 entry->start, entry->end);
402 where = entry->nd.prev; 443 where = entry->nd.prev;
403 cached_entry = list_entry(where,
404 struct memtype, nd);
405 } 444 }
406 break; 445 break;
407 } else if (start < entry->end) { /* start > entry->start */ 446 } else if (start < entry->end) { /* start > entry->start */
@@ -409,8 +448,6 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type,
409 if (!err) { 448 if (!err) {
410 dprintk("Overlap at 0x%Lx-0x%Lx\n", 449 dprintk("Overlap at 0x%Lx-0x%Lx\n",
411 entry->start, entry->end); 450 entry->start, entry->end);
412 cached_entry = list_entry(entry->nd.prev,
413 struct memtype, nd);
414 451
415 /* 452 /*
416 * Move to right position in the linked 453 * Move to right position in the linked
@@ -438,13 +475,13 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type,
438 return err; 475 return err;
439 } 476 }
440 477
441 cached_start = start;
442
443 if (where) 478 if (where)
444 list_add(&new->nd, where); 479 list_add(&new->nd, where);
445 else 480 else
446 list_add_tail(&new->nd, &memtype_list); 481 list_add_tail(&new->nd, &memtype_list);
447 482
483 memtype_rb_insert(&memtype_rbroot, new);
484
448 spin_unlock(&memtype_lock); 485 spin_unlock(&memtype_lock);
449 486
450 dprintk("reserve_memtype added 0x%Lx-0x%Lx, track %s, req %s, ret %s\n", 487 dprintk("reserve_memtype added 0x%Lx-0x%Lx, track %s, req %s, ret %s\n",
@@ -456,7 +493,7 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type,
456 493
457int free_memtype(u64 start, u64 end) 494int free_memtype(u64 start, u64 end)
458{ 495{
459 struct memtype *entry; 496 struct memtype *entry, *saved_entry;
460 int err = -EINVAL; 497 int err = -EINVAL;
461 int is_range_ram; 498 int is_range_ram;
462 499
@@ -474,17 +511,46 @@ int free_memtype(u64 start, u64 end)
474 return -EINVAL; 511 return -EINVAL;
475 512
476 spin_lock(&memtype_lock); 513 spin_lock(&memtype_lock);
514
515 entry = memtype_rb_search(&memtype_rbroot, start);
516 if (unlikely(entry == NULL))
517 goto unlock_ret;
518
519 /*
520 * Saved entry points to an entry with start same or less than what
521 * we searched for. Now go through the list in both directions to look
522 * for the entry that matches with both start and end, with list stored
523 * in sorted start address
524 */
525 saved_entry = entry;
477 list_for_each_entry(entry, &memtype_list, nd) { 526 list_for_each_entry(entry, &memtype_list, nd) {
478 if (entry->start == start && entry->end == end) { 527 if (entry->start == start && entry->end == end) {
479 if (cached_entry == entry || cached_start == start) 528 rb_erase(&entry->rb, &memtype_rbroot);
480 cached_entry = NULL; 529 list_del(&entry->nd);
530 kfree(entry);
531 err = 0;
532 break;
533 } else if (entry->start > start) {
534 break;
535 }
536 }
537
538 if (!err)
539 goto unlock_ret;
481 540
541 entry = saved_entry;
542 list_for_each_entry_reverse(entry, &memtype_list, nd) {
543 if (entry->start == start && entry->end == end) {
544 rb_erase(&entry->rb, &memtype_rbroot);
482 list_del(&entry->nd); 545 list_del(&entry->nd);
483 kfree(entry); 546 kfree(entry);
484 err = 0; 547 err = 0;
485 break; 548 break;
549 } else if (entry->start < start) {
550 break;
486 } 551 }
487 } 552 }
553unlock_ret:
488 spin_unlock(&memtype_lock); 554 spin_unlock(&memtype_lock);
489 555
490 if (err) { 556 if (err) {