2013-09-03 10:37:51

by Wanpeng Li

[permalink] [raw]
Subject: [PATCH v5 1/4] mm/vmalloc: don't set area->caller twice

Changelog:
*v1 -> v2: rebase against mmotm tree

The caller address has already been set in set_vmalloc_vm(), there's no need
to set it again in __vmalloc_area_node.

Reviewed-by: Zhang Yanfei <[email protected]>
Signed-off-by: Wanpeng Li <[email protected]>
---
mm/vmalloc.c | 1 -
1 file changed, 1 deletion(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 1074543..d78d117 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -1566,7 +1566,6 @@ static void *__vmalloc_area_node(struct vm_struct *area, gfp_t gfp_mask,
pages = kmalloc_node(array_size, nested_gfp, node);
}
area->pages = pages;
- area->caller = caller;
if (!area->pages) {
remove_vm_area(area->addr);
kfree(area);
--
1.8.1.2


2013-09-03 10:37:44

by Wanpeng Li

[permalink] [raw]
Subject: [PATCH v5 2/4] mm/vmalloc: revert "mm/vmalloc.c: emit the failure message before return"

Changelog:
*v2 -> v3: revert commit 46c001a2 directly

Don't warning twice in __vmalloc_area_node and __vmalloc_node_range if
__vmalloc_area_node allocation failure. This patch revert commit 46c001a2
(mm/vmalloc.c: emit the failure message before return).

Reviewed-by: Zhang Yanfei <[email protected]>
Signed-off-by: Wanpeng Li <[email protected]>
---
mm/vmalloc.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index d78d117..e3ec8b4 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -1635,7 +1635,7 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,

addr = __vmalloc_area_node(area, gfp_mask, prot, node, caller);
if (!addr)
- goto fail;
+ return NULL;

/*
* In this function, newly allocated vm_struct has VM_UNINITIALIZED
--
1.8.1.2

2013-09-03 10:37:49

by Wanpeng Li

[permalink] [raw]
Subject: [PATCH v5 4/4] mm/vmalloc: fix show vmap_area information race with vmap_area tear down

Changelog:
*v4 -> v5: return directly for !VM_VM_AREA case and remove (VM_LAZY_FREE | VM_LAZY_FREEING) check

There is a race window between vmap_area tear down and show vmap_area information.

A B

remove_vm_area
spin_lock(&vmap_area_lock);
va->vm = NULL;
va->flags &= ~VM_VM_AREA;
spin_unlock(&vmap_area_lock);
spin_lock(&vmap_area_lock);
if (va->flags & (VM_LAZY_FREE | VM_LAZY_FREEZING))
return 0;
if (!(va->flags & VM_VM_AREA)) {
seq_printf(m, "0x%pK-0x%pK %7ld vm_map_ram\n",
(void *)va->va_start, (void *)va->va_end,
va->va_end - va->va_start);
return 0;
}
free_unmap_vmap_area(va);
flush_cache_vunmap
free_unmap_vmap_area_noflush
unmap_vmap_area
free_vmap_area_noflush
va->flags |= VM_LAZY_FREE

The assumption !VM_VM_AREA represents vm_map_ram allocation is introduced by
commit: d4033afd(mm, vmalloc: iterate vmap_area_list, instead of vmlist, in
vmallocinfo()). However, !VM_VM_AREA also represents vmap_area is being tear
down in race window mentioned above. This patch fix it by don't dump any
information for !VM_VM_AREA case and also remove (VM_LAZY_FREE | VM_LAZY_FREEING)
check since they are not possible for !VM_VM_AREA case.

Suggested-by: Joonsoo Kim <[email protected]>
Signed-off-by: Wanpeng Li <[email protected]>
---
mm/vmalloc.c | 15 ++++++---------
1 file changed, 6 insertions(+), 9 deletions(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 5368b17..9b75028 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -2582,16 +2582,13 @@ static int s_show(struct seq_file *m, void *p)
{
struct vmap_area *va = p;
struct vm_struct *v;
-
- if (va->flags & (VM_LAZY_FREE | VM_LAZY_FREEING))
- return 0;
-
- if (!(va->flags & VM_VM_AREA)) {
- seq_printf(m, "0x%pK-0x%pK %7ld vm_map_ram\n",
- (void *)va->va_start, (void *)va->va_end,
- va->va_end - va->va_start);
+
+ /*
+ * s_show can encounter race with remove_vm_area, !VM_VM_AREA on
+ * behalf of vmap area is being tear down or vm_map_ram allocation.
+ */
+ if (!(va->flags & VM_VM_AREA))
return 0;
- }

v = va->vm;

--
1.8.1.2

2013-09-03 10:38:12

by Wanpeng Li

[permalink] [raw]
Subject: [PATCH v5 3/4] mm/vmalloc: revert "mm/vmalloc.c: check VM_UNINITIALIZED flag in s_show instead of show_numa_info"

Changelog:
*v2 -> v3: revert commit d157a558 directly

The VM_UNINITIALIZED/VM_UNLIST flag introduced by commit f5252e00(mm: avoid
null pointer access in vm_struct via /proc/vmallocinfo) is used to avoid
accessing the pages field with unallocated page when show_numa_info() is
called. This patch move the check just before show_numa_info in order that
some messages still can be dumped via /proc/vmallocinfo. This patch revert
commit d157a558 (mm/vmalloc.c: check VM_UNINITIALIZED flag in s_show instead
of show_numa_info);

Reviewed-by: Zhang Yanfei <[email protected]>
Signed-off-by: Wanpeng Li <[email protected]>
---
mm/vmalloc.c | 10 +++++-----
1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index e3ec8b4..5368b17 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -2562,6 +2562,11 @@ static void show_numa_info(struct seq_file *m, struct vm_struct *v)
if (!counters)
return;

+ /* Pair with smp_wmb() in clear_vm_uninitialized_flag() */
+ smp_rmb();
+ if (v->flags & VM_UNINITIALIZED)
+ return;
+
memset(counters, 0, nr_node_ids * sizeof(unsigned int));

for (nr = 0; nr < v->nr_pages; nr++)
@@ -2590,11 +2595,6 @@ static int s_show(struct seq_file *m, void *p)

v = va->vm;

- /* Pair with smp_wmb() in clear_vm_uninitialized_flag() */
- smp_rmb();
- if (v->flags & VM_UNINITIALIZED)
- return 0;
-
seq_printf(m, "0x%pK-0x%pK %7ld",
v->addr, v->addr + v->size, v->size);

--
1.8.1.2