2016-04-15 02:41:39

by Dan Williams

Subject: [PATCH] libnvdimm, pmem: clarify the write+clear_poison+write flow

The ACPI specification does not specify the state of data after a clear
poison operation. Potential future libnvdimm bus implementations for
other architectures also might not specify or disagree on the state of
data after clear poison. Clarify why we write twice.

Reported-by: Jeff Moyer <[email protected]>
Reported-by: Vishal Verma <[email protected]>
Signed-off-by: Dan Williams <[email protected]>
---
drivers/nvdimm/pmem.c | 14 ++++++++++++++
1 file changed, 14 insertions(+)

diff --git a/drivers/nvdimm/pmem.c b/drivers/nvdimm/pmem.c
index c6befaa9c708..d9a0dbc2d023 100644
--- a/drivers/nvdimm/pmem.c
+++ b/drivers/nvdimm/pmem.c
@@ -86,6 +86,20 @@ static int pmem_do_bvec(struct pmem_device *pmem, struct page *page,
flush_dcache_page(page);
}
} else {
+ /*
+ * Note that we write the data both before and after
+ * clearing poison. The write before clear poison
+ * handles situations where the latest written data is
+ * preserved and the clear poison operation simply marks
+ * the address range as valid without changing the data.
+ * In this case application software can assume that an
+ * interrupted write will either return the new good
+ * data or an error.
+ *
+ * However, if pmem_clear_poison() leaves the data in an
+ * indeterminate state we need to perform the write
+ * after clear poison.
+ */
flush_dcache_page(page);
memcpy_to_pmem(pmem_addr, mem + off, len);
if (unlikely(bad_pmem)) {

2016-04-15 07:11:42

by Johannes Thumshirn

Subject: Re: [PATCH] libnvdimm, pmem: clarify the write+clear_poison+write flow

On Donnerstag, 14. April 2016 19:40:47 CEST Dan Williams wrote:
> The ACPI specification does not specify the state of data after a clear
> poison operation. Potential future libnvdimm bus implementations for
> other architectures also might not specify or disagree on the state of
> data after clear poison. Clarify why we write twice.
>
> Reported-by: Jeff Moyer <[email protected]>
> Reported-by: Vishal Verma <[email protected]>
> Signed-off-by: Dan Williams <[email protected]>
> ---

Reviewed-by: Johannes Thumshirn <[email protected]>

--
Johannes Thumshirn Storage
[email protected] +49 911 74053 689
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 N?rnberg
GF: Felix Imend?rffer, Jane Smithard, Graham Norton
HRB 21284 (AG N?rnberg)
Key fingerprint = EC38 9CAB C2C4 F25D 8600 D0D0 0393 969D 2D76 0850

2016-04-15 12:32:18

by Jeff Moyer

Subject: Re: [PATCH] libnvdimm, pmem: clarify the write+clear_poison+write flow

Dan Williams <[email protected]> writes:

> The ACPI specification does not specify the state of data after a clear
> poison operation. Potential future libnvdimm bus implementations for
> other architectures also might not specify or disagree on the state of
> data after clear poison. Clarify why we write twice.

Thanks, Dan, looks good.

Reviewed-by: Jeff Moyer <[email protected]>

2016-04-15 16:58:23

by Verma, Vishal L

Subject: Re: [PATCH] libnvdimm, pmem: clarify the write+clear_poison+write flow

On Thu, 2016-04-14 at 19:40 -0700, Dan Williams wrote:
> The ACPI specification does not specify the state of data after a
> clear
> poison operation.  Potential future libnvdimm bus implementations for
> other architectures also might not specify or disagree on the state
> of
> data after clear poison.  Clarify why we write twice.
>
> Reported-by: Jeff Moyer <[email protected]>
> Reported-by: Vishal Verma <[email protected]>
> Signed-off-by: Dan Williams <[email protected]>
> ---
>  drivers/nvdimm/pmem.c |   14 ++++++++++++++
>  1 file changed, 14 insertions(+)

Looks good, thanks!

Reviewed-by: Vishal Verma <[email protected]>

>
> diff --git a/drivers/nvdimm/pmem.c b/drivers/nvdimm/pmem.c
> index c6befaa9c708..d9a0dbc2d023 100644
> --- a/drivers/nvdimm/pmem.c
> +++ b/drivers/nvdimm/pmem.c
> @@ -86,6 +86,20 @@ static int pmem_do_bvec(struct pmem_device *pmem,
> struct page *page,
>   flush_dcache_page(page);
>   }
>   } else {
> + /*
> +  * Note that we write the data both before and after
> +  * clearing poison.  The write before clear poison
> +  * handles situations where the latest written data
> is
> +  * preserved and the clear poison operation simply
> marks
> +  * the address range as valid without changing the
> data.
> +  * In this case application software can assume that
> an
> +  * interrupted write will either return the new good
> +  * data or an error.
> +  *
> +  * However, if pmem_clear_poison() leaves the data
> in an
> +  * indeterminate state we need to perform the write
> +  * after clear poison.
> +  */
>   flush_dcache_page(page);
>   memcpy_to_pmem(pmem_addr, mem + off, len);
>   if (unlikely(bad_pmem)) {
>