Skip to contentSkip to conversation

pmem, libnvdimm: complete REQ_FLUSH => REQ_PREFLUSH

PATCH In review Ross Zwisler 2 participants 8 messages Jun 6, 2018

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH v3 1/4 pmem, libnvdimm: complete REQ_FLUSH => REQ_PREFLUSH

Complete the move from REQ_FLUSH to REQ_PREFLUSH that apparently started way back in v4.8.

---
 drivers/nvdimm/pmem.c | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)
drivers/nvdimm/pmem.c +1 -6
index 9d714926ecf5..252adfab1e47 100644
@@ -164,11 +164,6 @@ static blk_status_t pmem_do_bvec(struct pmem_device *pmem, struct page *page,
 	return rc;
 }
 
-/* account for REQ_FLUSH rename, replace with REQ_PREFLUSH after v4.8-rc1 */
-#ifndef REQ_FLUSH
-#define REQ_FLUSH REQ_PREFLUSH
-#endif
-
 static blk_qc_t pmem_make_request(struct request_queue *q, struct bio *bio)
 {
 	blk_status_t rc = 0;
@@ -179,7 +174,7 @@ static blk_qc_t pmem_make_request(struct request_queue *q, struct bio *bio)
 	struct pmem_device *pmem = q->queuedata;
 	struct nd_region *nd_region = to_region(pmem);
 
-	if (bio->bi_opf & REQ_FLUSH)
+	if (bio->bi_opf & REQ_PREFLUSH)
 		nvdimm_flush(nd_region);
 
 	do_acct = nd_iostat_start(bio, &start);
Signature · 6 lines
2.14.4

_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm
#1 lore raw

PATCH v3 2/4 libnvdimm: unconditionally deep flush on *sync

Prior to this commit we would only do a "deep flush" (have nvdimm_flush() write to each of the flush hints for a region) in response to an msync/fsync/sync call if the nvdimm_has_cache() returned true at the time we were setting up the request queue. This happens due to the write cache value passed in to blk_queue_write_cache(), which then causes the block layer to send down BIOs with REQ_FUA and REQ_PREFLUSH set. We do have a "write_cache" sysfs entry for namespaces, i.e.:

/sys/bus/nd/devices/pfn0.1/block/pmem0/dax/write_cache

which can be used to control whether or not the kernel thinks a given namespace has a write cache, but this didn't modify the deep flush behavior that we set up when the driver was initialized. Instead, it only modified whether or not DAX would flush CPU caches via dax_flush() in response to *sync calls.

Simplify this by making the *sync deep flush always happen, regardless of the write cache setting of a namespace. The DAX CPU cache flushing will still be controlled the write_cache setting of the namespace.

---
 drivers/nvdimm/pmem.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
drivers/nvdimm/pmem.c +3 -4
index 252adfab1e47..97b4c39a9267 100644
@@ -294,7 +294,7 @@ static int pmem_attach_disk(struct device *dev,
 {
 	struct nd_namespace_io *nsio = to_nd_namespace_io(&ndns->dev);
 	struct nd_region *nd_region = to_nd_region(dev->parent);
-	int nid = dev_to_node(dev), fua, wbc;
+	int nid = dev_to_node(dev), fua;
 	struct resource *res = &nsio->res;
 	struct resource bb_res;
 	struct nd_pfn *nd_pfn = NULL;
@@ -330,7 +330,6 @@ static int pmem_attach_disk(struct device *dev,
 		dev_warn(dev, "unable to guarantee persistence of writes\n");
 		fua = 0;
 	}
-	wbc = nvdimm_has_cache(nd_region);
 
 	if (!devm_request_mem_region(dev, res->start, resource_size(res),
 				dev_name(&ndns->dev))) {
@@ -377,7 +376,7 @@ static int pmem_attach_disk(struct device *dev,
 		return PTR_ERR(addr);
 	pmem->virt_addr = addr;
 
-	blk_queue_write_cache(q, wbc, fua);
+	blk_queue_write_cache(q, true, fua);
 	blk_queue_make_request(q, pmem_make_request);
 	blk_queue_physical_block_size(q, PAGE_SIZE);
 	blk_queue_logical_block_size(q, pmem_sector_size(ndns));
@@ -408,7 +407,7 @@ static int pmem_attach_disk(struct device *dev,
 		put_disk(disk);
 		return -ENOMEM;
 	}
-	dax_write_cache(dax_dev, wbc);
+	dax_write_cache(dax_dev, nvdimm_has_cache(nd_region));
 	pmem->dax_dev = dax_dev;
 
 	gendev = disk_to_dev(disk);
Signature · 6 lines
2.14.4

_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm
#2 lore raw

Re: [PATCH v3 2/4] libnvdimm: unconditionally deep flush on *sync

Quoted 16 lines · Ross Zwisler

Prior to this commit we would only do a "deep flush" (have nvdimm_flush() write to each of the flush hints for a region) in response to an msync/fsync/sync call if the nvdimm_has_cache() returned true at the time we were setting up the request queue. This happens due to the write cache value passed in to blk_queue_write_cache(), which then causes the block layer to send down BIOs with REQ_FUA and REQ_PREFLUSH set. We do have a "write_cache" sysfs entry for namespaces, i.e.:

/sys/bus/nd/devices/pfn0.1/block/pmem0/dax/write_cache

which can be used to control whether or not the kernel thinks a given namespace has a write cache, but this didn't modify the deep flush behavior that we set up when the driver was initialized. Instead, it only modified whether or not DAX would flush CPU caches via dax_flush() in response to *sync calls.

Quoted 6 lines

Simplify this by making the *sync deep flush always happen, regardless of the write cache setting of a namespace. The DAX CPU cache flushing will still be controlled the write_cache setting of the namespace.

Looks, good. I believe we want this one and ["PATCH v3 4/4] libnvdimm:
don't flush power-fail protected CPU caches" marked for -stable and tagged with:

Fixes 5fdf8e5ba566 ("libnvdimm: re-enable deep flush for pmem devices

via fsync()")

...any concerns with that?
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm

#3 lore raw

Re: [PATCH v3 2/4] libnvdimm: unconditionally deep flush on *sync

Quoted 25 lines · Dan Williams
Quoted 16 lines · Ross Zwisler

Prior to this commit we would only do a "deep flush" (have nvdimm_flush() write to each of the flush hints for a region) in response to an msync/fsync/sync call if the nvdimm_has_cache() returned true at the time we were setting up the request queue. This happens due to the write cache value passed in to blk_queue_write_cache(), which then causes the block layer to send down BIOs with REQ_FUA and REQ_PREFLUSH set. We do have a "write_cache" sysfs entry for namespaces, i.e.:

/sys/bus/nd/devices/pfn0.1/block/pmem0/dax/write_cache

which can be used to control whether or not the kernel thinks a given namespace has a write cache, but this didn't modify the deep flush behavior that we set up when the driver was initialized. Instead, it only modified whether or not DAX would flush CPU caches via dax_flush() in response to *sync calls.

Quoted 6 lines

Simplify this by making the *sync deep flush always happen, regardless of the write cache setting of a namespace. The DAX CPU cache flushing will still be controlled the write_cache setting of the namespace.

Quoted 8 lines

Looks, good. I believe we want this one and ["PATCH v3 4/4] libnvdimm:
don't flush power-fail protected CPU caches" marked for -stable and tagged with:

Fixes 5fdf8e5ba566 ("libnvdimm: re-enable deep flush for pmem devices

via fsync()")

...any concerns with that?

Nope, sounds good. Can you fix that up when you apply, or would it be helpful for me to send another revision with those tags?
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm

#4 lore raw

Re: [PATCH v3 2/4] libnvdimm: unconditionally deep flush on *sync

Quoted 30 lines · Ross Zwisler
Quoted 29 lines · Dan Williams
Quoted 16 lines · Ross Zwisler

Prior to this commit we would only do a "deep flush" (have nvdimm_flush() write to each of the flush hints for a region) in response to an msync/fsync/sync call if the nvdimm_has_cache() returned true at the time we were setting up the request queue. This happens due to the write cache value passed in to blk_queue_write_cache(), which then causes the block layer to send down BIOs with REQ_FUA and REQ_PREFLUSH set. We do have a "write_cache" sysfs entry for namespaces, i.e.:

/sys/bus/nd/devices/pfn0.1/block/pmem0/dax/write_cache

which can be used to control whether or not the kernel thinks a given namespace has a write cache, but this didn't modify the deep flush behavior that we set up when the driver was initialized. Instead, it only modified whether or not DAX would flush CPU caches via dax_flush() in response to *sync calls.

Quoted 6 lines

Simplify this by making the *sync deep flush always happen, regardless of the write cache setting of a namespace. The DAX CPU cache flushing will still be controlled the write_cache setting of the namespace.

Looks, good. I believe we want this one and ["PATCH v3 4/4] libnvdimm:
don't flush power-fail protected CPU caches" marked for -stable and tagged with:

Quoted 7 lines
Quoted 4 lines
Fixes 5fdf8e5ba566 ("libnvdimm: re-enable deep flush for pmem devices

via fsync()")

...any concerns with that?

Nope, sounds good. Can you fix that up when you apply, or would it be helpful for me to send another revision with those tags?

I'll fix it up, thanks Ross.
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm

#5 lore raw

PATCH v3 3/4 libnvdimm: use dax_write_cache* helpers

In reply to Ross Zwisler

Use dax_write_cache() and dax_write_cache_enabled() instead of open coding the bit operations.

---
 drivers/dax/super.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)
drivers/dax/super.c +3 -6
index 2b2332b605e4..c2c46f96b18c 100644
@@ -182,8 +182,7 @@ static ssize_t write_cache_show(struct device *dev,
 	if (!dax_dev)
 		return -ENXIO;
 
-	rc = sprintf(buf, "%d\n", !!test_bit(DAXDEV_WRITE_CACHE,
-				&dax_dev->flags));
+	rc = sprintf(buf, "%d\n", !!dax_write_cache_enabled(dax_dev));
 	put_dax(dax_dev);
 	return rc;
 }
@@ -201,10 +200,8 @@ static ssize_t write_cache_store(struct device *dev,
 
 	if (rc)
 		len = rc;
-	else if (write_cache)
-		set_bit(DAXDEV_WRITE_CACHE, &dax_dev->flags);
 	else
-		clear_bit(DAXDEV_WRITE_CACHE, &dax_dev->flags);
+		dax_write_cache(dax_dev, write_cache);
 
 	put_dax(dax_dev);
 	return len;
@@ -286,7 +283,7 @@ EXPORT_SYMBOL_GPL(dax_copy_from_iter);
 void arch_wb_cache_pmem(void *addr, size_t size);
 void dax_flush(struct dax_device *dax_dev, void *addr, size_t size)
 {
-	if (unlikely(!test_bit(DAXDEV_WRITE_CACHE, &dax_dev->flags)))
+	if (unlikely(!dax_write_cache_enabled(dax_dev)))
 		return;
 
 	arch_wb_cache_pmem(addr, size);
Signature · 6 lines
2.14.4

_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm
#6 lore raw

PATCH v3 4/4 libnvdimm: don't flush power-fail protected CPU caches

In reply to Ross Zwisler

This commit:

5fdf8e5ba566 ("libnvdimm: re-enable deep flush for pmem devices via fsync()")

intended to make sure that deep flush was always available even on platforms which support a power-fail protected CPU cache. An unintended side effect of this change was that we also lost the ability to skip flushing CPU caches on those power-fail protected CPU cache.

Fix this by skipping the low level cache flushing in dax_flush() if we have CPU caches which are power-fail protected. The user can still override this behavior by manually setting the write_cache state of a namespace. See libndctl's ndctl_namespace_write_cache_is_enabled(), ndctl_namespace_enable_write_cache() and ndctl_namespace_disable_write_cache() functions.

Fixes 5fdf8e5ba566 ("libnvdimm: re-enable deep flush for pmem devices via fsync()")
---
 drivers/nvdimm/region_devs.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/nvdimm/region_devs.c +2 -1
index a612be6f019d..ec3543b83330 100644
@@ -1132,7 +1132,8 @@ EXPORT_SYMBOL_GPL(nvdimm_has_flush);
 
 int nvdimm_has_cache(struct nd_region *nd_region)
 {
-	return is_nd_pmem(&nd_region->dev);
+	return is_nd_pmem(&nd_region->dev) &&
+		!test_bit(ND_REGION_PERSIST_CACHE, &nd_region->flags);
 }
 EXPORT_SYMBOL_GPL(nvdimm_has_cache);
 
Signature · 6 lines
2.14.4

_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm