[2/3] block: Add badblock management for gendisks
diff mbox

Message ID 1448066960-20119-3-git-send-email-vishal.l.verma@intel.com
State New
Headers show

Commit Message

Verma, Vishal L Nov. 21, 2015, 12:49 a.m. UTC
NVDIMM devices, which can behave more like DRAM rather than block
devices, may develop bad cache lines, or 'poison'. A block device
exposed by the pmem driver can then consume poison via a read (or
write), and cause a machine check. On platforms without machine
check recovery features, this would mean a crash.

The block device maintaining a runtime list of all known sectors that
have poison can directly avoid this, and also provide a path forward
to enable proper handling/recovery for DAX faults on such a device.

Use the new badblock management interfaces to add a badblocks list to
gendisks.

Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
---
 block/genhd.c         | 64 +++++++++++++++++++++++++++++++++++++++++++++++++++
 include/linux/genhd.h |  6 +++++
 2 files changed, 70 insertions(+)

Comments

Jeff Moyer Nov. 24, 2015, 3:34 p.m. UTC | #1
Vishal Verma <vishal.l.verma@intel.com> writes:

> NVDIMM devices, which can behave more like DRAM rather than block
> devices, may develop bad cache lines, or 'poison'. A block device
> exposed by the pmem driver can then consume poison via a read (or
> write), and cause a machine check. On platforms without machine
> check recovery features, this would mean a crash.
>
> The block device maintaining a runtime list of all known sectors that
> have poison can directly avoid this, and also provide a path forward
> to enable proper handling/recovery for DAX faults on such a device.
>
> Use the new badblock management interfaces to add a badblocks list to
> gendisks.

Because disk_alloc_badblocks can fail, you need to check for a NULL
disk->bb in all of the utility functions you've defined.

Cheers,
Jeff

>
> Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
> ---
>  block/genhd.c         | 64 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  include/linux/genhd.h |  6 +++++
>  2 files changed, 70 insertions(+)
>
> diff --git a/block/genhd.c b/block/genhd.c
> index 0c706f3..4209c32 100644
> --- a/block/genhd.c
> +++ b/block/genhd.c
> @@ -20,6 +20,7 @@
>  #include <linux/idr.h>
>  #include <linux/log2.h>
>  #include <linux/pm_runtime.h>
> +#include <linux/badblocks.h>
>  
>  #include "blk.h"
>  
> @@ -505,6 +506,20 @@ static int exact_lock(dev_t devt, void *data)
>  	return 0;
>  }
>  
> +static void disk_alloc_badblocks(struct gendisk *disk)
> +{
> +	disk->bb = kzalloc(sizeof(disk->bb), GFP_KERNEL);
> +	if (!disk->bb) {
> +		pr_warn("%s: failed to allocate space for badblocks\n",
> +			disk->disk_name);
> +		return;
> +	}
> +
> +	if (badblocks_init(disk->bb, 1))
> +		pr_warn("%s: failed to initialize badblocks\n",
> +			disk->disk_name);
> +}
> +
>  static void register_disk(struct gendisk *disk)
>  {
>  	struct device *ddev = disk_to_dev(disk);
> @@ -609,6 +624,7 @@ void add_disk(struct gendisk *disk)
>  	disk->first_minor = MINOR(devt);
>  
>  	disk_alloc_events(disk);
> +	disk_alloc_badblocks(disk);
>  
>  	/* Register BDI before referencing it from bdev */
>  	bdi = &disk->queue->backing_dev_info;
> @@ -657,6 +673,9 @@ void del_gendisk(struct gendisk *disk)
>  	blk_unregister_queue(disk);
>  	blk_unregister_region(disk_devt(disk), disk->minors);
>  
> +	badblocks_free(disk->bb);
> +	kfree(disk->bb);
> +
>  	part_stat_set_all(&disk->part0, 0);
>  	disk->part0.stamp = 0;
>  
> @@ -670,6 +689,48 @@ void del_gendisk(struct gendisk *disk)
>  }
>  EXPORT_SYMBOL(del_gendisk);
>  
> +/*
> + * The gendisk usage of badblocks does not track acknowledgements for
> + * badblocks. We always assume they are acknowledged.
> + */
> +int disk_check_badblocks(struct gendisk *disk, sector_t s, int sectors,
> +		   sector_t *first_bad, int *bad_sectors)
> +{
> +	return badblocks_check(disk->bb, s, sectors, first_bad, bad_sectors);
> +}
> +EXPORT_SYMBOL(disk_check_badblocks);
> +
> +int disk_set_badblocks(struct gendisk *disk, sector_t s, int sectors)
> +{
> +	return badblocks_set(disk->bb, s, sectors, 1);
> +}
> +EXPORT_SYMBOL(disk_set_badblocks);
> +
> +int disk_clear_badblocks(struct gendisk *disk, sector_t s, int sectors)
> +{
> +	return badblocks_clear(disk->bb, s, sectors);
> +}
> +EXPORT_SYMBOL(disk_clear_badblocks);
> +
> +/* sysfs access to bad-blocks list. */
> +static ssize_t disk_badblocks_show(struct device *dev,
> +					struct device_attribute *attr,
> +					char *page)
> +{
> +	struct gendisk *disk = dev_to_disk(dev);
> +
> +	return badblocks_show(disk->bb, page, 0);
> +}
> +
> +static ssize_t disk_badblocks_store(struct device *dev,
> +					struct device_attribute *attr,
> +					const char *page, size_t len)
> +{
> +	struct gendisk *disk = dev_to_disk(dev);
> +
> +	return badblocks_store(disk->bb, page, len, 0);
> +}
> +
>  /**
>   * get_gendisk - get partitioning information for a given device
>   * @devt: device to get partitioning information for
> @@ -988,6 +1049,8 @@ static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
>  static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
>  static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
>  static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
> +static DEVICE_ATTR(badblocks, S_IRUGO | S_IWUSR, disk_badblocks_show,
> +		disk_badblocks_store);
>  #ifdef CONFIG_FAIL_MAKE_REQUEST
>  static struct device_attribute dev_attr_fail =
>  	__ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
> @@ -1009,6 +1072,7 @@ static struct attribute *disk_attrs[] = {
>  	&dev_attr_capability.attr,
>  	&dev_attr_stat.attr,
>  	&dev_attr_inflight.attr,
> +	&dev_attr_badblocks.attr,
>  #ifdef CONFIG_FAIL_MAKE_REQUEST
>  	&dev_attr_fail.attr,
>  #endif
> diff --git a/include/linux/genhd.h b/include/linux/genhd.h
> index 2adbfa6..5563bde 100644
> --- a/include/linux/genhd.h
> +++ b/include/linux/genhd.h
> @@ -162,6 +162,7 @@ struct disk_part_tbl {
>  };
>  
>  struct disk_events;
> +struct badblocks;
>  
>  struct gendisk {
>  	/* major, first_minor and minors are input parameters only,
> @@ -201,6 +202,7 @@ struct gendisk {
>  	struct blk_integrity *integrity;
>  #endif
>  	int node_id;
> +	struct badblocks *bb;
>  };
>  
>  static inline struct gendisk *part_to_disk(struct hd_struct *part)
> @@ -421,6 +423,10 @@ extern void add_disk(struct gendisk *disk);
>  extern void del_gendisk(struct gendisk *gp);
>  extern struct gendisk *get_gendisk(dev_t dev, int *partno);
>  extern struct block_device *bdget_disk(struct gendisk *disk, int partno);
> +extern int disk_check_badblocks(struct gendisk *disk, sector_t s, int sectors,
> +		   sector_t *first_bad, int *bad_sectors);
> +extern int disk_set_badblocks(struct gendisk *disk, sector_t s, int sectors);
> +extern int disk_clear_badblocks(struct gendisk *disk, sector_t s, int sectors);
>  
>  extern void set_device_ro(struct block_device *bdev, int flag);
>  extern void set_disk_ro(struct gendisk *disk, int flag);
--
To unsubscribe from this list: send the line "unsubscribe linux-block" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Jeff Moyer Nov. 24, 2015, 7:14 p.m. UTC | #2
"Verma, Vishal L" <vishal.l.verma@intel.com> writes:

> On Tue, 2015-11-24 at 10:34 -0500, Jeff Moyer wrote:
>> Vishal Verma <vishal.l.verma@intel.com> writes:
>> 
>> > NVDIMM devices, which can behave more like DRAM rather than block
>> > devices, may develop bad cache lines, or 'poison'. A block device
>> > exposed by the pmem driver can then consume poison via a read (or
>> > write), and cause a machine check. On platforms without machine
>> > check recovery features, this would mean a crash.
>> > 
>> > The block device maintaining a runtime list of all known sectors
>> > that
>> > have poison can directly avoid this, and also provide a path forward
>> > to enable proper handling/recovery for DAX faults on such a device.
>> > 
>> > Use the new badblock management interfaces to add a badblocks list
>> > to
>> > gendisks.
>> 
>> Because disk_alloc_badblocks can fail, you need to check for a NULL
>> disk->bb in all of the utility functions you've defined.
>> 
>
> Thanks, Jeff - I'll fix this. I have a handful of other fixes queued up
> too, will send out a v2 soon.

I'm not sure whether it makes sense to continue without badblock
management for the RAID code.  I was hoping Neil would comment on that.

-Jeff
--
To unsubscribe from this list: send the line "unsubscribe linux-block" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Verma, Vishal L Nov. 24, 2015, 8:10 p.m. UTC | #3
On Tue, 2015-11-24 at 14:14 -0500, Jeff Moyer wrote:
> 

> I'm not sure whether it makes sense to continue without badblock

> management for the RAID code.  I was hoping Neil would comment on

> that.

> 

> -Jeff


Not sure I follow? I believe I've kept all the badblocks functionality
RAID already had..


On a related note, something I observed when testing with md:

md's badblock list is per-device, and can be seen at:
	/sys/block/md0/md/dev-pmem0/bad_blocks

Now if we have badblocks in the gendisks too, there is also:
	/sys/block/pmem0/bad_blocks

The two are separate 'accounts' maintained by separate drivers (md for
the former, and pmem for the latter). This can potentially be
confusing..

Should we consolidate the two, i.e. make md (re)use the gendisk
badblocks for its purposes too?

	-Vishal
Williams, Dan J Nov. 24, 2015, 9:31 p.m. UTC | #4
On Tue, Nov 24, 2015 at 12:10 PM, Verma, Vishal L
<vishal.l.verma@intel.com> wrote:
> On Tue, 2015-11-24 at 14:14 -0500, Jeff Moyer wrote:
>>
>> I'm not sure whether it makes sense to continue without badblock
>> management for the RAID code.  I was hoping Neil would comment on
>> that.
>>
>> -Jeff
>
> Not sure I follow? I believe I've kept all the badblocks functionality
> RAID already had..
>
>
> On a related note, something I observed when testing with md:
>
> md's badblock list is per-device, and can be seen at:
>         /sys/block/md0/md/dev-pmem0/bad_blocks
>
> Now if we have badblocks in the gendisks too, there is also:
>         /sys/block/pmem0/bad_blocks
>
> The two are separate 'accounts' maintained by separate drivers (md for
> the former, and pmem for the latter). This can potentially be
> confusing..
>
> Should we consolidate the two, i.e. make md (re)use the gendisk
> badblocks for its purposes too?

If we get agreement that tracking bad blocks at the gendisk is useful
for more than just nvdimms then yes, I think it makes sense to move md
bad_blocks to the gendisk layer.  That is provided we can add a
symlink to make the move transparent to existing md userspace tooling.

The use cases I can envision being useful for other disks is:

1/ Bypassing / avoiding known bad blocks on disks / drivers that
inject long completion latency for error handling.

2/ Simulating bad blocks for disks that can't store them locally.
Similar to the md use case, if one encounters errors restoring a disk
image from a backup it's useful to have the option to record the error
in a bad block list and keep going.
--
To unsubscribe from this list: send the line "unsubscribe linux-block" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Jeff Moyer Nov. 25, 2015, 3:37 p.m. UTC | #5
"Verma, Vishal L" <vishal.l.verma@intel.com> writes:

> On Tue, 2015-11-24 at 14:14 -0500, Jeff Moyer wrote:
>> 
>> I'm not sure whether it makes sense to continue without badblock
>> management for the RAID code.  I was hoping Neil would comment on
>> that.
>> 
>> -Jeff
>
> Not sure I follow? I believe I've kept all the badblocks functionality
> RAID already had..

What I mean to say is that the RAID code had previously embedded the
badblocks structure in one of its other data structures.  As a result,
you would never get an allocation failure for it.

> On a related note, something I observed when testing with md:
>
> md's badblock list is per-device, and can be seen at:
> 	/sys/block/md0/md/dev-pmem0/bad_blocks
>
> Now if we have badblocks in the gendisks too, there is also:
> 	/sys/block/pmem0/bad_blocks
>
> The two are separate 'accounts' maintained by separate drivers (md for
> the former, and pmem for the latter). This can potentially be
> confusing..
>
> Should we consolidate the two, i.e. make md (re)use the gendisk
> badblocks for its purposes too?

I agree with what Dan said.

Cheers,
Jeff
--
To unsubscribe from this list: send the line "unsubscribe linux-block" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Verma, Vishal L Nov. 25, 2015, 5:55 p.m. UTC | #6
On Wed, 2015-11-25 at 10:37 -0500, Jeff Moyer wrote:
> "Verma, Vishal L" <vishal.l.verma@intel.com> writes:

> 

> > On Tue, 2015-11-24 at 14:14 -0500, Jeff Moyer wrote:

> > > 

> > > I'm not sure whether it makes sense to continue without badblock

> > > management for the RAID code.  I was hoping Neil would comment on

> > > that.

> > > 

> > > -Jeff

> > 

> > Not sure I follow? I believe I've kept all the badblocks

> > functionality

> > RAID already had..

> 

> What I mean to say is that the RAID code had previously embedded the

> badblocks structure in one of its other data structures.  As a result,

> you would never get an allocation failure for it.

> 

Ah I see - I don't think that has effectively changed. 'rdev' still
contains a statically allocated badblocks structure (as opposed to
gendisk, which just stores a pointer). md used to dynamically allocate
the storage space inside badblocks (bb->page), and that is still the
case using badblocks_init.

	-Vishal
Jeff Moyer Nov. 25, 2015, 6:07 p.m. UTC | #7
"Verma, Vishal L" <vishal.l.verma@intel.com> writes:

> On Wed, 2015-11-25 at 10:37 -0500, Jeff Moyer wrote:
>> "Verma, Vishal L" <vishal.l.verma@intel.com> writes:
>> 
>> > On Tue, 2015-11-24 at 14:14 -0500, Jeff Moyer wrote:
>> > > 
>> > > I'm not sure whether it makes sense to continue without badblock
>> > > management for the RAID code.  I was hoping Neil would comment on
>> > > that.
>> > > 
>> > > -Jeff
>> > 
>> > Not sure I follow? I believe I've kept all the badblocks
>> > functionality
>> > RAID already had..
>> 
>> What I mean to say is that the RAID code had previously embedded the
>> badblocks structure in one of its other data structures.  As a result,
>> you would never get an allocation failure for it.
>> 
> Ah I see - I don't think that has effectively changed. 'rdev' still
> contains a statically allocated badblocks structure (as opposed to
> gendisk, which just stores a pointer). md used to dynamically allocate
> the storage space inside badblocks (bb->page), and that is still the
> case using badblocks_init.

Ah, ok.  Sorry I didn't dig into that further.

Thanks,
Jeff
--
To unsubscribe from this list: send the line "unsubscribe linux-block" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch
diff mbox

diff --git a/block/genhd.c b/block/genhd.c
index 0c706f3..4209c32 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -20,6 +20,7 @@ 
 #include <linux/idr.h>
 #include <linux/log2.h>
 #include <linux/pm_runtime.h>
+#include <linux/badblocks.h>
 
 #include "blk.h"
 
@@ -505,6 +506,20 @@  static int exact_lock(dev_t devt, void *data)
 	return 0;
 }
 
+static void disk_alloc_badblocks(struct gendisk *disk)
+{
+	disk->bb = kzalloc(sizeof(disk->bb), GFP_KERNEL);
+	if (!disk->bb) {
+		pr_warn("%s: failed to allocate space for badblocks\n",
+			disk->disk_name);
+		return;
+	}
+
+	if (badblocks_init(disk->bb, 1))
+		pr_warn("%s: failed to initialize badblocks\n",
+			disk->disk_name);
+}
+
 static void register_disk(struct gendisk *disk)
 {
 	struct device *ddev = disk_to_dev(disk);
@@ -609,6 +624,7 @@  void add_disk(struct gendisk *disk)
 	disk->first_minor = MINOR(devt);
 
 	disk_alloc_events(disk);
+	disk_alloc_badblocks(disk);
 
 	/* Register BDI before referencing it from bdev */
 	bdi = &disk->queue->backing_dev_info;
@@ -657,6 +673,9 @@  void del_gendisk(struct gendisk *disk)
 	blk_unregister_queue(disk);
 	blk_unregister_region(disk_devt(disk), disk->minors);
 
+	badblocks_free(disk->bb);
+	kfree(disk->bb);
+
 	part_stat_set_all(&disk->part0, 0);
 	disk->part0.stamp = 0;
 
@@ -670,6 +689,48 @@  void del_gendisk(struct gendisk *disk)
 }
 EXPORT_SYMBOL(del_gendisk);
 
+/*
+ * The gendisk usage of badblocks does not track acknowledgements for
+ * badblocks. We always assume they are acknowledged.
+ */
+int disk_check_badblocks(struct gendisk *disk, sector_t s, int sectors,
+		   sector_t *first_bad, int *bad_sectors)
+{
+	return badblocks_check(disk->bb, s, sectors, first_bad, bad_sectors);
+}
+EXPORT_SYMBOL(disk_check_badblocks);
+
+int disk_set_badblocks(struct gendisk *disk, sector_t s, int sectors)
+{
+	return badblocks_set(disk->bb, s, sectors, 1);
+}
+EXPORT_SYMBOL(disk_set_badblocks);
+
+int disk_clear_badblocks(struct gendisk *disk, sector_t s, int sectors)
+{
+	return badblocks_clear(disk->bb, s, sectors);
+}
+EXPORT_SYMBOL(disk_clear_badblocks);
+
+/* sysfs access to bad-blocks list. */
+static ssize_t disk_badblocks_show(struct device *dev,
+					struct device_attribute *attr,
+					char *page)
+{
+	struct gendisk *disk = dev_to_disk(dev);
+
+	return badblocks_show(disk->bb, page, 0);
+}
+
+static ssize_t disk_badblocks_store(struct device *dev,
+					struct device_attribute *attr,
+					const char *page, size_t len)
+{
+	struct gendisk *disk = dev_to_disk(dev);
+
+	return badblocks_store(disk->bb, page, len, 0);
+}
+
 /**
  * get_gendisk - get partitioning information for a given device
  * @devt: device to get partitioning information for
@@ -988,6 +1049,8 @@  static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
 static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
+static DEVICE_ATTR(badblocks, S_IRUGO | S_IWUSR, disk_badblocks_show,
+		disk_badblocks_store);
 #ifdef CONFIG_FAIL_MAKE_REQUEST
 static struct device_attribute dev_attr_fail =
 	__ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
@@ -1009,6 +1072,7 @@  static struct attribute *disk_attrs[] = {
 	&dev_attr_capability.attr,
 	&dev_attr_stat.attr,
 	&dev_attr_inflight.attr,
+	&dev_attr_badblocks.attr,
 #ifdef CONFIG_FAIL_MAKE_REQUEST
 	&dev_attr_fail.attr,
 #endif
diff --git a/include/linux/genhd.h b/include/linux/genhd.h
index 2adbfa6..5563bde 100644
--- a/include/linux/genhd.h
+++ b/include/linux/genhd.h
@@ -162,6 +162,7 @@  struct disk_part_tbl {
 };
 
 struct disk_events;
+struct badblocks;
 
 struct gendisk {
 	/* major, first_minor and minors are input parameters only,
@@ -201,6 +202,7 @@  struct gendisk {
 	struct blk_integrity *integrity;
 #endif
 	int node_id;
+	struct badblocks *bb;
 };
 
 static inline struct gendisk *part_to_disk(struct hd_struct *part)
@@ -421,6 +423,10 @@  extern void add_disk(struct gendisk *disk);
 extern void del_gendisk(struct gendisk *gp);
 extern struct gendisk *get_gendisk(dev_t dev, int *partno);
 extern struct block_device *bdget_disk(struct gendisk *disk, int partno);
+extern int disk_check_badblocks(struct gendisk *disk, sector_t s, int sectors,
+		   sector_t *first_bad, int *bad_sectors);
+extern int disk_set_badblocks(struct gendisk *disk, sector_t s, int sectors);
+extern int disk_clear_badblocks(struct gendisk *disk, sector_t s, int sectors);
 
 extern void set_device_ro(struct block_device *bdev, int flag);
 extern void set_disk_ro(struct gendisk *disk, int flag);