As described, pages are charged to their first touchers.
The first toucher is determined using pages' _mapcount
manipulations in rmap calls.
A page is charged in two stages:
1. preparation, in which the resource availability is checked.
This stage may lead to page reclamation, thus it is performed
in a "might-sleep" places;
2. the container assignment to page. This is done in an atomic
code that handles races between multiple touchers.
The uncharge process is tricky. After the _mapcount is changed
and happens to hit zero we cannot trust the page->container
pointer any longer. So we get the pointer before the decrement.
Includes fixes from Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Pavel Emelianov <xemul@openvz.org>
---
diff -upr linux-2.6.22-rc2-mm1.orig/fs/exec.c linux-2.6.22-rc2-mm1-2/fs/exec.c
--- linux-2.6.22-rc2-mm1.orig/fs/exec.c 2007-05-30 12:32:34.000000000 +0400
+++ linux-2.6.22-rc2-mm1-2/fs/exec.c 2007-06-01 16:35:13.000000000 +0400
@@ -59,6 +59,8 @@
#include <linux/kmod.h>
#endif
+#include <linux/rss_container.h>
+
int core_uses_pid;
char core_pattern[CORENAME_MAX_SIZE] = "core";
int suid_dumpable = 0;
@@ -314,27 +316,34 @@ void install_arg_page(struct vm_area_str
struct mm_struct *mm = vma->vm_mm;
pte_t * pte;
spinlock_t *ptl;
+ struct page_container *pcont;
if (unlikely(anon_vma_prepare(vma)))
goto out;
+ if (container_rss_prepare(page, vma, &pcont))
+ goto out;
+
flush_dcache_page(page);
pte = get_locked_pte(mm, address, &ptl);
if (!pte)
- goto out;
+ goto out_release;
if (!pte_none(*pte)) {
pte_unmap_unlock(pte, ptl);
- goto out;
+ goto out_release;
}
inc_mm_counter(mm, anon_rss);
lru_cache_add_active(page);
set_pte_at(mm, address, pte, pte_mkdirty(pte_mkwrite(mk_pte(
page, vma->vm_page_prot))));
- page_add_new_anon_rmap(page, vma, address);
+ page_add_new_anon_rmap(page, vma, address, pcont);
pte_unmap_unlock(pte, ptl);
/* no need for flush_tlb */
return;
+
+out_release:
+ container_rss_release(pcont);
out:
__free_page(page);
force_sig(SIGKILL, current);
diff -upr linux-2.6.22-rc2-mm1.orig/include/linux/rmap.h linux-2.6.22-rc2-mm1-2/include/linux/rmap.h
--- linux-2.6.22-rc2-mm1.orig/include/linux/rmap.h 2007-05-29 18:01:57.000000000 +0400
+++ linux-2.6.22-rc2-mm1-2/include/linux/rmap.h 2007-06-01 16:35:13.000000000 +0400
@@ -69,9 +69,13 @@ void __anon_vma_link(struct vm_area_stru
/*
* rmap interfaces called when adding or removing pte of page
*/
-void page_add_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
-void page_add_new_anon_rmap(struct page *, struct vm_area_struct *, unsigned long);
-void page_add_file_rmap(struct page *);
+struct page_container;
+
+void page_add_anon_rmap(struct page *, struct vm_area_struct *,
+ unsigned long, struct page_container *);
+void page_add_new_anon_rmap(struct page *, struct vm_area_struct *,
+ unsigned long, struct page_container *);
+void page_add_file_rmap(struct page *, struct page_container *);
void page_remove_rmap(struct page *, struct vm_area_struct *);
#ifdef CONFIG_DEBUG_VM
diff -upr linux-2.6.22-rc2-mm1.orig/mm/memory.c linux-2.6.22-rc2-mm1-2/mm/memory.c
--- linux-2.6.22-rc2-mm1.orig/mm/memory.c 2007-05-30 12:32:36.000000000 +0400
+++ linux-2.6.22-rc2-mm1-2/mm/memory.c 2007-06-01 16:35:13.000000000 +0400
@@ -60,6 +60,8 @@
#include <linux/swapops.h>
#include <linux/elf.h>
+#include <linux/rss_container.h>
+
#ifndef CONFIG_NEED_MULTIPLE_NODES
/* use the per-pgdat data instead for discontigmem - mbligh */
unsigned long max_mapnr;
@@ -1155,7 +1157,7 @@ static int zeromap_pte_range(struct mm_s
break;
}
page_cache_get(page);
- page_add_file_rmap(page);
+ page_add_file_rmap(page, NULL);
inc_mm_counter(mm, file_rss);
set_pte_at(mm, addr, pte, zero_pte);
} while (pte++, addr += PAGE_SIZE, addr != end);
@@ -1263,7 +1265,7 @@ static int insert_page(struct mm_struct
/* Ok, finally just insert the thing.. */
get_page(page);
inc_mm_counter(mm, file_rss);
- page_add_file_rmap(page);
+ page_add_file_rmap(page, NULL);
set_pte_at(mm, addr, pte, mk_pte(page, prot));
retval = 0;
@@ -1668,6 +1670,7 @@ static int do_wp_page(struct mm_struct *
pte_t entry;
int reuse = 0, ret = VM_FAULT_MINOR;
struct page *dirty_page = NULL;
+ struct page_container *pcont;
old_page = vm_normal_page(vma, address, orig_pte);
if (!old_page)
@@ -1752,6 +1755,9 @@ gotten:
cow_user_page(new_page, old_page, address, vma);
}
+ if (container_rss_prepare(new_page, vma, &pcont))
+ goto oom;
+
/*
* Re-check the pte - we dropped the lock
*/
@@ -1779,12 +1785,14 @@ gotten:
set_pte_at(mm, address, page_table, entry);
update_mmu_cache(vma, address, entry);
lru_cache_add_active(new_page);
- page_add_new_anon_rmap(new_page, vma, address);
+ page_add_new_anon_rmap(new_page, vma, address, pcont);
/* Free the old page.. */
new_page = old_page;
ret |= VM_FAULT_WRITE;
- }
+ } else
+ container_rss_release(pcont);
+
if (new_page)
page_cache_release(new_page);
if (old_page)
@@ -2176,6 +2184,7 @@ static int do_swap_page(struct mm_struct
swp_entry_t entry;
pte_t pte;
int ret = VM_FAULT_MINOR;
+ struct page_container *pcont;
if (!pte_unmap_same(mm, pmd, page_table, orig_pte))
goto out;
@@ -2208,6 +2217,11 @@ static int do_swap_page(struct mm_struct
count_vm_event(PGMAJFAULT);
}
+ if (container_rss_prepare(page, vma, &pcont)) {
+ ret = VM_FAULT_OOM;
+ goto out;
+ }
+
delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
mark_page_accessed(page);
lock_page(page);
@@ -2221,6 +2235,7 @@ static int do_swap_page(struct mm_struct
if (unlikely(!PageUptodate(page))) {
ret = VM_FAULT_SIGBUS;
+ container_rss_release(pcont);
goto out_nomap;
}
@@ -2235,7 +2250,7 @@ static int do_swap_page(struct mm_struct
flush_icache_page(vma, page);
set_pte_at(mm, address, page_table, pte);
- page_add_anon_rmap(page, vma, address);
+ page_add_anon_rmap(page, vma, address, pcont);
swap_free(entry);
if (vm_swap_full())
@@ -2256,6 +2271,7 @@ unlock:
out:
return ret;
out_nomap:
+ container_rss_release(pcont);
pte_unmap_unlock(page_table, ptl);
unlock_page(page);
page_cache_release(page);
@@ -2274,6 +2290,7 @@ static int do_anonymous_page(struct mm_s
struct page *page;
spinlock_t *ptl;
pte_t entry;
+ struct page_container *pcont;
if (write_access) {
/* Allocate our own private page. */
@@ -2285,15 +2302,19 @@ static int do_anonymous_page(struct mm_s
if (!page)
goto oom;
+ if (container_rss_prepare(page, vma, &pcont))
+ goto oom_release;
+
entry = mk_pte(page, vma->vm_page_prot);
entry = maybe_mkwrite(pte_mkdirty(entry), vma);
page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
if (!pte_none(*page_table))
- goto release;
+ goto release_container;
+
inc_mm_counter(mm, anon_rss);
lru_cache_add_active(page);
- page_add_new_anon_rmap(page, vma, address);
+ page_add_new_anon_rmap(page, vma, address, pcont);
} else {
/* Map the ZERO_PAGE - vm_page_prot is readonly */
page = ZERO_PAGE(address);
@@ -2305,7 +2326,7 @@ static int do_anonymous_page(struct mm_s
if (!pte_none(*page_table))
goto release;
inc_mm_counter(mm, file_rss);
- page_add_file_rmap(page);
+ page_add_file_rmap(page, NULL);
}
set_pte_at(mm, address, page_table, entry);
@@ -2316,9 +2337,14 @@ static int do_anonymous_page(struct mm_s
unlock:
pte_unmap_unlock(page_table, ptl);
return VM_FAULT_MINOR;
+release_container:
+ container_rss_release(pcont);
release:
page_cache_release(page);
goto unlock;
+
+oom_release:
+ page_cache_release(page);
oom:
return VM_FAULT_OOM;
}
@@ -2346,6 +2372,7 @@ static int __do_fault(struct mm_struct *
int anon = 0;
struct page *dirty_page = NULL;
struct fault_data fdata;
+ struct page_container *pcont;
fdata.address = address & PAGE_MASK;
fdata.pgoff = pgoff;
@@ -2415,6 +2442,11 @@ static int __do_fault(struct mm_struct *
}
+ if (container_rss_prepare(page, vma, &pcont)) {
+ fdata.type = VM_FAULT_OOM;
+ goto out;
+ }
+
page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
/*
@@ -2437,10 +2469,10 @@ static int __do_fault(struct mm_struct *
if (anon) {
inc_mm_counter(mm, anon_rss);
lru_cache_add_active(page);
- page_add_new_anon_rmap(page, vma, address);
+ page_add_new_anon_rmap(page, vma, address, pcont);
} else {
inc_mm_counter(mm, file_rss);
- page_add_file_rmap(page);
+ page_add_file_rmap(page, pcont);
if (flags & FAULT_FLAG_WRITE) {
dirty_page = page;
get_page(dirty_page);
@@ -2451,6 +2483,7 @@ static int __do_fault(struct mm_struct *
update_mmu_cache(vma, address, entry);
lazy_mmu_prot_update(entry);
} else {
+ container_rss_release(pcont);
if (anon)
page_cache_release(page);
else
diff -upr linux-2.6.22-rc2-mm1.orig/mm/migrate.c linux-2.6.22-rc2-mm1-2/mm/migrate.c
--- linux-2.6.22-rc2-mm1.orig/mm/migrate.c 2007-05-30 12:32:36.000000000 +0400
+++ linux-2.6.22-rc2-mm1-2/mm/migrate.c 2007-06-01 16:35:13.000000000 +0400
@@ -28,6 +28,7 @@
#include <linux/mempolicy.h>
#include <linux/vmalloc.h>
#include <linux/security.h>
+#include <linux/rss_container.h>
#include "internal.h"
@@ -134,6 +135,7 @@ static void remove_migration_pte(struct
pte_t *ptep, pte;
spinlock_t *ptl;
unsigned long addr = page_address_in_vma(new, vma);
+ struct page_container *pcont;
if (addr == -EFAULT)
return;
@@ -157,6 +159,11 @@ static void remove_migration_pte(struct
return;
}
+ if (container_rss_prepare(new, vma, &pcont)) {
+ pte_unmap(ptep);
+ return;
+ }
+
ptl = pte_lockptr(mm, pmd);
spin_lock(ptl);
pte = *ptep;
@@ -175,16 +182,19 @@ static void remove_migration_pte(struct
set_pte_
...
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
