If we have to call get_block with the create argument set to 1, then
the filesystem almost certainly had to zero the block. which is an I/O,
which should be reported as a major fault.
Note that major faults on XIP files happen for different reasons than
major faults on non-XIP files. XIP files behave as if everything except
file holes is already cached. That's all the more reason to report
major faults when we do have to do I/O; it may be a valuable resource
for sysadmins trying to diagnose performance problems.
Signed-off-by: Matthew Wilcox <matthew.r.wilcox@xxxxxxxxx>
---
fs/xip.c | 14 ++++++++++++--
1 file changed, 12 insertions(+), 2 deletions(-)
diff --git a/fs/xip.c b/fs/xip.c
index d032838..4eb4596 100644
--- a/fs/xip.c
+++ b/fs/xip.c
@@ -19,10 +19,12 @@
#include <linux/fs.h>
#include <linux/genhd.h>
#include <linux/highmem.h>
+#include <linux/memcontrol.h>
#include <linux/mm.h>
#include <linux/mutex.h>
#include <linux/sched.h>
#include <linux/uio.h>
+#include <linux/vmstat.h>
int xip_clear_blocks(struct inode *inode, sector_t block, long size)
{
@@ -213,6 +215,7 @@ static int do_xip_fault(struct vm_area_struct *vma, struct vm_fault *vmf,
pgoff_t size;
unsigned long pfn;
int error;
+ int major = 0;
size = (i_size_read(inode) + PAGE_SIZE - 1) >> PAGE_SHIFT;
if (vmf->pgoff >= size)
@@ -226,6 +229,9 @@ static int do_xip_fault(struct vm_area_struct *vma, struct vm_fault *vmf,
/* Don't allocate backing store if we're going to COW a hole */
if (!error && !buffer_mapped(&bh) && !vmf->cow_page) {
error = get_block(inode, block, &bh, 1);
+ count_vm_event(PGMAJFAULT);
+ mem_cgroup_count_vm_event(vma->vm_mm, PGMAJFAULT);
+ major = VM_FAULT_MAJOR;
}
if (error)
@@ -255,7 +261,7 @@ static int do_xip_fault(struct vm_area_struct *vma, struct vm_fault *vmf,
/* -EBUSY is fine, somebody else faulted on the same PTE */
if (error != -EBUSY)
BUG_ON(error);
- return VM_FAULT_NOPAGE;
+ return VM_FAULT_NOPAGE | major;
}
int xip_fault(struct vm_area_struct *vma, struct vm_fault *vmf,
@@ -297,6 +303,7 @@ static int do_xip_pmd_fault(struct vm_area_struct *vma, unsigned long address,
pgoff_t size, pgoff;
sector_t block;
unsigned long pfn;
+ int major = 0;
/* Fall back to PTEs if we're going to COW */
if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED))
@@ -324,6 +331,9 @@ static int do_xip_pmd_fault(struct vm_area_struct *vma, unsigned long address,
/* Next, try to allocate the whole thing */
bh.b_size = PMD_SIZE;
length = get_block(inode, block, &bh, 1);
+ count_vm_event(PGMAJFAULT);
+ mem_cgroup_count_vm_event(vma->vm_mm, PGMAJFAULT);
+ major = VM_FAULT_MAJOR;
if (length)
return VM_FAULT_SIGBUS;
if (bh.b_size == PMD_SIZE)
@@ -356,7 +366,7 @@ static int do_xip_pmd_fault(struct vm_area_struct *vma, unsigned long address,
/* -EBUSY is fine, somebody else faulted on the same PMD */
if (length != -EBUSY)
BUG_ON(length);
- return VM_FAULT_NOPAGE;
+ return VM_FAULT_NOPAGE | major;
}
int xip_pmd_fault(struct vm_area_struct *vma, unsigned long address,
--
1.8.4.rc3
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
