See http://www.uefi.org/RFIC_LIST ("Virtual NVDIMM 0x1901"):
"Get Unsafe Shutdown Count (Function Index 2)".
Let's expose the info to the userspace (e.g. ntctl) via sysfs.
Signed-off-by: Dexuan Cui <[email protected]>
---
drivers/acpi/nfit/core.c | 51 ++++++++++++++++++++++++++++++++++++++
drivers/acpi/nfit/hyperv.h | 26 +++++++++++++++++++
2 files changed, 77 insertions(+)
create mode 100644 drivers/acpi/nfit/hyperv.h
diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
index d5a164b6833a..d504da34ce34 100644
--- a/drivers/acpi/nfit/core.c
+++ b/drivers/acpi/nfit/core.c
@@ -25,6 +25,7 @@
#include <asm/cacheflush.h>
#include <acpi/nfit.h>
#include "intel.h"
+#include "hyperv.h"
#include "nfit.h"
/*
@@ -1802,6 +1803,54 @@ __weak void nfit_intel_shutdown_status(struct nfit_mem *nfit_mem)
}
}
+__weak void nfit_hyperv_shutdown_status(struct nfit_mem *nfit_mem)
+{
+ struct device *dev = &nfit_mem->adev->dev;
+ struct nd_hyperv_shutdown_status status;
+ union acpi_object in_buf = {
+ .buffer.type = ACPI_TYPE_BUFFER,
+ .buffer.length = 0,
+ };
+ union acpi_object in_obj = {
+ .package.type = ACPI_TYPE_PACKAGE,
+ .package.count = 1,
+ .package.elements = &in_buf,
+ };
+ const u8 func = ND_HYPERV_GET_UNSAFE_SHUTDOWN_COUNT;
+ const guid_t *guid = to_nfit_uuid(nfit_mem->family);
+ u8 revid = nfit_dsm_revid(nfit_mem->family, func);
+ struct acpi_device *adev = nfit_mem->adev;
+ acpi_handle handle = adev->handle;
+ union acpi_object *out_obj;
+
+ if ((nfit_mem->dsm_mask & BIT(func)) == 0)
+ return;
+
+ out_obj = acpi_evaluate_dsm(handle, guid, revid, func, &in_obj);
+ if (!out_obj || out_obj->type != ACPI_TYPE_BUFFER
+ || out_obj->buffer.length < sizeof(status)) {
+ dev_dbg(dev->parent,
+ "%s: failed to get Unsafe Shutdown Count\n",
+ dev_name(dev));
+ ACPI_FREE(out_obj);
+ return;
+ }
+
+ memcpy(&status, out_obj->buffer.pointer, sizeof(status));
+ ACPI_FREE(out_obj);
+
+ if (status.general_err != 0) {
+ dev_dbg(dev->parent,
+ "%s: failed to get Unsafe Shutdown Count: err=0x%x\n",
+ dev_name(dev), status.status);
+ return;
+ }
+
+ set_bit(NFIT_MEM_DIRTY_COUNT, &nfit_mem->flags);
+ nfit_mem->dirty_shutdown = status.shutdown_count;
+}
+
+
static void populate_shutdown_status(struct nfit_mem *nfit_mem)
{
/*
@@ -1811,6 +1860,8 @@ static void populate_shutdown_status(struct nfit_mem *nfit_mem)
*/
if (nfit_mem->family == NVDIMM_FAMILY_INTEL)
nfit_intel_shutdown_status(nfit_mem);
+ else if (nfit_mem->family == NVDIMM_FAMILY_HYPERV)
+ nfit_hyperv_shutdown_status(nfit_mem);
}
static int acpi_nfit_add_dimm(struct acpi_nfit_desc *acpi_desc,
diff --git a/drivers/acpi/nfit/hyperv.h b/drivers/acpi/nfit/hyperv.h
new file mode 100644
index 000000000000..c4a25b8624cc
--- /dev/null
+++ b/drivers/acpi/nfit/hyperv.h
@@ -0,0 +1,26 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright(c) 2019 Microsoft Corporation. All rights reserved.
+ * Hyper-V specific definitions for _DSM of Hyper-V Virtual NVDIMM
+ *
+ * See http://www.uefi.org/RFIC_LIST (Virtual NVDIMM 0x1901)
+ */
+#ifndef _NFIT_HYPERV_H_
+#define _NFIT_HYPERV_H_
+
+#define ND_HYPERV_GET_UNSAFE_SHUTDOWN_COUNT 2
+
+struct nd_hyperv_shutdown_status {
+ union {
+ u32 status;
+ struct {
+ u16 general_err;
+ u8 func_err;
+ u8 vendor_err;
+ };
+ };
+ u32 shutdown_count;
+} __packed;
+
+
+#endif
--
2.19.1
On Wed, Jan 30, 2019 at 10:49 AM Dexuan Cui <[email protected]> wrote:
>
>
> See http://www.uefi.org/RFIC_LIST ("Virtual NVDIMM 0x1901"):
> "Get Unsafe Shutdown Count (Function Index 2)".
>
> Let's expose the info to the userspace (e.g. ntctl) via sysfs.
>
> Signed-off-by: Dexuan Cui <[email protected]>
> ---
> drivers/acpi/nfit/core.c | 51 ++++++++++++++++++++++++++++++++++++++
> drivers/acpi/nfit/hyperv.h | 26 +++++++++++++++++++
> 2 files changed, 77 insertions(+)
> create mode 100644 drivers/acpi/nfit/hyperv.h
Looks ok to me, but if we grow any more of these "shutdown count"
retrievals I'll want to take a look at being able to call
acpi_nfit_ctl() earlier in the init process so that this data
marshaling can be unified.
On Wed, Jan 30, 2019 at 06:48:40PM +0000, Dexuan Cui wrote:
>
> See http://www.uefi.org/RFIC_LIST ("Virtual NVDIMM 0x1901"):
> "Get Unsafe Shutdown Count (Function Index 2)".
>
> Let's expose the info to the userspace (e.g. ntctl) via sysfs.
If you add a new sysfs file, you need to add a new Documentation/ABI/
update as well :(
> From: Greg KH <[email protected]>
> Sent: Wednesday, January 30, 2019 11:38 AM
>
> On Wed, Jan 30, 2019 at 06:48:40PM +0000, Dexuan Cui wrote:
> >
> > Let's expose the info to the userspace (e.g. ntctl) via sysfs.
>
> If you add a new sysfs file, you need to add a new Documentation/ABI/
> update as well :(
It's an existing sysfs node, which was originally added by Dan in Sep 2018:
/sys/bus/nd/devices/nmem0/nfit/dirty_shutdown.
Before this patch, the node doesn't appear when Linux runs on Hyper-V,
and with this patch, the node can appear now.
However, indeed, the node and the other related nodes have not been
documented in Documentation/ABI/testing/sysfs-bus-nfit yet.
I suppose Dan would add that?
Thanks,
-- Dexuan
> From: Linux-nvdimm <[email protected]> On Behalf Of
> Dexuan Cui
> Sent: Wednesday, January 30, 2019 12:03 PM
> To: Greg KH <[email protected]>
> Cc: Josh Poulson <[email protected]>; [email protected];
> Haiyang Zhang <[email protected]>;
> [email protected]; Rafael J. Wysocki
> <[email protected]>; [email protected]; Michael Kelley
> <[email protected]>; Sasha Levin <[email protected]>;
> [email protected]; Ross Zwisler <[email protected]>; Stephen
> Hemminger <[email protected]>; Len Brown <[email protected]>
> Subject: RE: [PATCH] nfit: Collect shutdown status for
> NVDIMM_FAMILY_HYPERV
>
> > From: Greg KH <[email protected]>
> > Sent: Wednesday, January 30, 2019 11:38 AM
> >
> > On Wed, Jan 30, 2019 at 06:48:40PM +0000, Dexuan Cui wrote:
> > >
> > > Let's expose the info to the userspace (e.g. ntctl) via sysfs.
> >
> > If you add a new sysfs file, you need to add a new Documentation/ABI/
> > update as well :(
>
> It's an existing sysfs node, which was originally added by Dan in Sep 2018:
> /sys/bus/nd/devices/nmem0/nfit/dirty_shutdown.
>
> Before this patch, the node doesn't appear when Linux runs on Hyper-V,
> and with this patch, the node can appear now.
>
> However, indeed, the node and the other related nodes have not been
> documented in Documentation/ABI/testing/sysfs-bus-nfit yet.
> I suppose Dan would add that?
Actually the other nodes have been documented, and only the
"dirty_shutdown" is missed.
It should be straightfoward. Let me make a patch for this.
Thanks for the reminder, Greg!
Thanks,
-- Dexuan