From: Dan Williams <dan.j.williams@intel.com>
To: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: linux-nvdimm <linux-nvdimm@lists.01.org>
Subject: Re: [RFC PATCH] libnvdimm: Update the meaning for persistence_domain values
Date: Wed, 15 Jan 2020 11:44:29 -0800 [thread overview]
Message-ID: <CAPcyv4hRhYUcCyCMET1i2bkUEjp_MWVj0hEr3mVBp24Oq4Yauw@mail.gmail.com> (raw)
In-Reply-To: <0f44df90-1f75-9d0a-10af-6e7f48158bc7@linux.ibm.com>
On Wed, Jan 15, 2020 at 9:31 AM Aneesh Kumar K.V
<aneesh.kumar@linux.ibm.com> wrote:
>
> On 1/15/20 10:57 PM, Aneesh Kumar K.V wrote:
> > On 1/15/20 10:25 PM, Jeff Moyer wrote:
> >> "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com> writes:
> >>
> >>> Currently, kernel shows the below values
> >>> "persistence_domain":"cpu_cache"
> >>> "persistence_domain":"memory_controller"
> >>> "persistence_domain":"unknown"
> >>>
> >>> This patch updates the meaning of these values such that
> >>>
> >>> "cpu_cache" indicates no extra instructions is needed to ensure the
> >>> persistence
> >>> of data in the pmem media on power failure.
> >>>
> >>> "memory_controller" indicates platform provided instructions need to
> >>> be issued
> >>> as per documented sequence to make sure data flushed is guaranteed to
> >>> be on pmem
> >>> media in case of system power loss.
> >>>
> >>> Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
> >>> ---
> >>> arch/powerpc/platforms/pseries/papr_scm.c | 7 ++++++-
> >>> include/linux/libnvdimm.h | 6 +++---
> >>> 2 files changed, 9 insertions(+), 4 deletions(-)
> >>>
> >>> diff --git a/arch/powerpc/platforms/pseries/papr_scm.c
> >>> b/arch/powerpc/platforms/pseries/papr_scm.c
> >>> index c2ef320ba1bf..26a5ef263758 100644
> >>> --- a/arch/powerpc/platforms/pseries/papr_scm.c
> >>> +++ b/arch/powerpc/platforms/pseries/papr_scm.c
> >>> @@ -360,8 +360,13 @@ static int papr_scm_nvdimm_init(struct
> >>> papr_scm_priv *p)
> >>> if (p->is_volatile)
> >>> p->region = nvdimm_volatile_region_create(p->bus, &ndr_desc);
> >>> - else
> >>> + else {
> >>> + /*
> >>> + * We need to flush things correctly to guarantee persistance
> >>> + */
> >>> + set_bit(ND_REGION_PERSIST_MEMCTRL, &ndr_desc.flags);
> >>> p->region = nvdimm_pmem_region_create(p->bus, &ndr_desc);
> >>> + }
> >>> if (!p->region) {
> >>> dev_err(dev, "Error registering region %pR from %pOF\n",
> >>> ndr_desc.res, p->dn);
> >>
> >> Would you also update of_pmem to indicate the persistence domain,
> >> please?
> >>
> >
> > sure.
> >
> >
> >>> diff --git a/include/linux/libnvdimm.h b/include/linux/libnvdimm.h
> >>> index f2a33f2e3ba8..9126737377e1 100644
> >>> --- a/include/linux/libnvdimm.h
> >>> +++ b/include/linux/libnvdimm.h
> >>> @@ -52,9 +52,9 @@ enum {
> >>> */
> >>> ND_REGION_PERSIST_CACHE = 1,
> >>> /*
> >>> - * Platform provides mechanisms to automatically flush outstanding
> >>> - * write data from memory controler to pmem on system power loss.
> >>> - * (ADR)
> >>> + * Platform provides instructions to flush data such that on
> >>> completion
> >>> + * of the instructions, data flushed is guaranteed to be on pmem
> >>> even
> >>> + * in case of a system power loss.
> >>
> >> I find the prior description easier to understand.
> >
> >
> > I was trying to avoid the term 'automatically, 'memory controler' and
> > ADR. Can I update the above as
> >
> > /*
> > * Platform provides mechanisms to flush outstanding write data
> > * to pmem on system power loss.
> > */
> >
>
> Wanted to add more details. So with the above interpretation, if the
> persistence_domain is found to be 'cpu_cache', application can expect a
> store instruction to guarantee persistence.
The expectation is globally visible stores are persisted
> If it is 'none' there is no
> persistence ( I am not sure how that is the difference from 'volatile'
> pmem region).
"None" means the "platform does not enumerate a persistence domain".
That's not necessarily "volatile" because the user may know that they
have battery backup, or some other private/out-of-band capability not
exported by the platform. In which case they would manually need to
manipulate the pmemX/write_cache property manually.
> If it is 'memory_controller' ( I am not sure whether that
> is the right term), application needs to follow the recommended
> mechanism to flush write data to pmem.
If it is memory_controller the expectation is that flushing data out
of the cpu caches and making those writebacks to memory globally
visible is sufficient for the platform to persist flushed data on a
power loss.
_______________________________________________
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-leave@lists.01.org
next prev parent reply other threads:[~2020-01-15 19:44 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-01-08 6:49 [RFC PATCH] libnvdimm: Update the meaning for persistence_domain values Aneesh Kumar K.V
2020-01-15 16:55 ` Jeff Moyer
2020-01-15 17:27 ` Aneesh Kumar K.V
2020-01-15 17:31 ` Aneesh Kumar K.V
2020-01-15 17:42 ` Jeff Moyer
2020-01-15 19:44 ` Dan Williams [this message]
2020-01-15 17:35 ` Jeff Moyer
2020-01-15 17:55 ` Aneesh Kumar K.V
2020-01-15 19:48 ` Dan Williams
2020-01-16 6:24 ` Aneesh Kumar K.V
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAPcyv4hRhYUcCyCMET1i2bkUEjp_MWVj0hEr3mVBp24Oq4Yauw@mail.gmail.com \
--to=dan.j.williams@intel.com \
--cc=aneesh.kumar@linux.ibm.com \
--cc=linux-nvdimm@lists.01.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).