[PATCH v6 22/22] dax: Add reporting of major faults

From: Matthew Wilcox
Date: Tue Feb 25 2014 - 09:21:22 EST


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 DAX files happen for different reasons than
major faults on non-DAX files. DAX 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/dax.c | 8 +++++++-
1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/fs/dax.c b/fs/dax.c
index cdc8012..79a67c5 100644
--- a/fs/dax.c
+++ b/fs/dax.c
@@ -20,10 +20,12 @@
#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 dax_clear_blocks(struct inode *inode, sector_t block, long size)
{
@@ -286,6 +288,7 @@ static int do_dax_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)
@@ -301,6 +304,9 @@ static int do_dax_fault(struct vm_area_struct *vma, struct vm_fault *vmf,
if (!buffer_written(&bh) && !vmf->cow_page) {
if (vmf->flags & FAULT_FLAG_WRITE) {
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 || bh.b_size < PAGE_SIZE)
return VM_FAULT_SIGBUS;
} else {
@@ -332,7 +338,7 @@ static int do_dax_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;
}

/**
--
1.8.5.3

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/