On 3/6/19 12:28 PM, Oliver wrote:
On Wed, Mar 6, 2019 at 4:35 PM Aneesh Kumar K.V
<[email protected]> wrote:
Even if the kernel is built with THP or HUGEPAGE_PUD, the platform can decide
not to allow huge pages based on different parameters. The huge page support
checks are mostly arch specific and this patch provides arch specific
callbacks/abstraction for finding alignment values we should use when
configuring
pfn device.
Cc: Oliver O'Halloran <[email protected]>
Signed-off-by: Aneesh Kumar K.V <[email protected]>
---
arch/powerpc/include/asm/libnvdimm.h | 10 ++++++
arch/powerpc/mm/Makefile | 1 +
arch/powerpc/mm/nvdimm.c | 48 ++++++++++++++++++++++++++++
arch/x86/include/asm/libnvdimm.h | 19 +++++++++++
drivers/nvdimm/nd.h | 6 ----
drivers/nvdimm/pfn_devs.c | 31 ++++++++++++++++--
6 files changed, 107 insertions(+), 8 deletions(-)
create mode 100644 arch/powerpc/include/asm/libnvdimm.h
create mode 100644 arch/powerpc/mm/nvdimm.c
create mode 100644 arch/x86/include/asm/libnvdimm.h
diff --git a/arch/powerpc/include/asm/libnvdimm.h
b/arch/powerpc/include/asm/libnvdimm.h
new file mode 100644
index 000000000000..0928096eca90
--- /dev/null
+++ b/arch/powerpc/include/asm/libnvdimm.h
@@ -0,0 +1,10 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_POWERPC_LIBNVDIMM_H
+#define _ASM_POWERPC_LIBNVDIMM_H
+
+#define nd_pfn_supported_alignment nd_pfn_supported_alignment
+extern unsigned long *nd_pfn_supported_alignments(void);
+extern unsigned long nd_pfn_default_alignment(void);
+extern unsigned long nd_altmap_align_size(unsigned long nd_align);
+
+#endif
diff --git a/arch/powerpc/mm/Makefile b/arch/powerpc/mm/Makefile
index f965fc33a8b7..fed0d4b3b4bc 100644
--- a/arch/powerpc/mm/Makefile
+++ b/arch/powerpc/mm/Makefile
@@ -55,3 +55,4 @@ obj-$(CONFIG_PPC_BOOK3S_64) +=
dump_linuxpagetables-book3s64.o
endif
obj-$(CONFIG_PPC_HTDUMP) += dump_hashpagetable.o
obj-$(CONFIG_PPC_MEM_KEYS) += pkeys.o
+obj-$(CONFIG_NVDIMM_PFN) += nvdimm.o
diff --git a/arch/powerpc/mm/nvdimm.c b/arch/powerpc/mm/nvdimm.c
new file mode 100644
index 000000000000..ae47a65362a1
--- /dev/null
+++ b/arch/powerpc/mm/nvdimm.c
@@ -0,0 +1,48 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/mm.h>
+
+/*
+ * We support only pte and pmd mappings for now.
+ */
+const unsigned long *nd_pfn_supported_alignments(void)
+{
+ static unsigned long supported_alignments[3];
+
+#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
+#error "Update page size"
+#endif
+ supported_alignments[0] = PAGE_SIZE;
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+ if (mmu_psize_defs[MMU_PAGE_16M].shift ||
+ mmu_psize_defs[MMU_PAGE_2M].shift)
+ supported_alignments[1] = HPAGE_PMD_SIZE;
+#else
+ supported_alignments[1] = 0;
+#endif
+
+ supported_alignments[2] = 0;
+ return supported_alignments;
+}
+
+/*
+ * Use pmd mapping if supported as default alignment
+ */
+unsigned long nd_pfn_default_alignment(void)
+{
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+ if (mmu_psize_defs[MMU_PAGE_16M].shift ||
+ mmu_psize_defs[MMU_PAGE_2M].shift)
+ return HPAGE_PMD_SIZE;
+#endif
+ return PAGE_SIZE;
+}
+
+unsigned long nd_altmap_align_size(unsigned long nd_align)
+{
+ unsigned long vmemmap_psize;
+
+ vmemmap_psize = 1UL << mmu_psize_defs[mmu_vmemmap_psize].shift;
+ return max(nd_align, vmemmap_psize);
+}
It seems a bit heavyweight to have every arch define these functions
when we'll be the only ones doing anything interesting with them. I
think we'd be better off using the same method we use for PAGE_SIZE
where the arch #defines them to a variable that we fill out at
runtime. That way libnvdimm can keep using PFN_DEFAULT_ALIGNMENT or
whatever and the rest of the world doesn't need to care about the
weird stuff our firmware is doing today.
I can definitely do what we did with nd_pfn_supported_alginments. That
does a default implementation in case arch doesn't want to override. I
am not sure about doing this as #define, that will make it similar to
what we have with HPAGE_SHIFT which IMHO is not clean. But if your
concern is why bother other architecture when ppc is the only one
impacted, we can fix that by providing a default implementation of these
function?
diff --git a/arch/x86/include/asm/libnvdimm.h b/arch/x86/include/asm/libnvdimm.h
new file mode 100644
index 000000000000..3d5361db9164
--- /dev/null
+++ b/arch/x86/include/asm/libnvdimm.h
@@ -0,0 +1,19 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_X86_LIBNVDIMM_H
+#define _ASM_X86_LIBNVDIMM_H
+
+static inline unsigned long nd_pfn_default_alignment(void)
+{
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+ return HPAGE_PMD_SIZE;
+#else
+ return PAGE_SIZE;
+#endif
+}
+
+static inline unsigned long nd_altmap_align_size(unsigned long nd_align)
+{
+ return PMD_SIZE;
+}
+
+#endif
diff --git a/drivers/nvdimm/nd.h b/drivers/nvdimm/nd.h
index 379bf4305e61..e5a65868476f 100644
--- a/drivers/nvdimm/nd.h
+++ b/drivers/nvdimm/nd.h
@@ -292,12 +292,6 @@ static inline struct device *nd_btt_create(struct
nd_region *nd_region)
struct nd_pfn *to_nd_pfn(struct device *dev);
#if IS_ENABLED(CONFIG_NVDIMM_PFN)
-#ifdef CONFIG_TRANSPARENT_HUGEPAGE
-#define PFN_DEFAULT_ALIGNMENT HPAGE_PMD_SIZE
-#else
-#define PFN_DEFAULT_ALIGNMENT PAGE_SIZE
-#endif
-
int nd_pfn_probe(struct device *dev, struct nd_namespace_common *ndns);
bool is_nd_pfn(struct device *dev);
struct device *nd_pfn_create(struct nd_region *nd_region);
diff --git a/drivers/nvdimm/pfn_devs.c b/drivers/nvdimm/pfn_devs.c
index 6f22272e8d80..331e53bcb65e 100644
--- a/drivers/nvdimm/pfn_devs.c
+++ b/drivers/nvdimm/pfn_devs.c
@@ -18,6 +18,7 @@
#include <linux/slab.h>
#include <linux/fs.h>
#include <linux/mm.h>
+#include <asm/libnvdimm.h>
#include "nd-core.h"
#include "pfn.h"
#include "nd.h"
@@ -111,6 +112,8 @@ static ssize_t align_show(struct device *dev,
return sprintf(buf, "%ld\n", nd_pfn->align);
}
+#ifndef nd_pfn_supported_alignment
+#define nd_pfn_supported_alignment nd_pfn_supported_alignment
static const unsigned long *nd_pfn_supported_alignments(void)
{
/*
@@ -133,6 +136,7 @@ static const unsigned long
*nd_pfn_supported_alignments(void)
return data;
}
+#endif
static ssize_t align_store(struct device *dev,
struct device_attribute *attr, const char *buf, size_t len)
@@ -310,7 +314,7 @@ struct device *nd_pfn_devinit(struct nd_pfn *nd_pfn,
return NULL;
nd_pfn->mode = PFN_MODE_NONE;
- nd_pfn->align = PFN_DEFAULT_ALIGNMENT;
+ nd_pfn->align = nd_pfn_default_alignment();
dev = &nd_pfn->dev;
device_initialize(&nd_pfn->dev);
if (ndns && !__nd_attach_ndns(&nd_pfn->dev, ndns, &nd_pfn->ndns)) {
@@ -420,6 +424,20 @@ static int nd_pfn_clear_memmap_errors(struct nd_pfn
*nd_pfn)
return 0;
}
+static bool nd_supported_alignment(unsigned long align)
+{
+ int i;
+ const unsigned long *supported = nd_pfn_supported_alignments();
+
+ if (align == 0)
+ return false;
+
+ for (i = 0; supported[i]; i++)
+ if (align == supported[i])
+ return true;
+ return false;
+}
+
int nd_pfn_validate(struct nd_pfn *nd_pfn, const char *sig)
{
u64 checksum, offset;
@@ -474,6 +492,15 @@ int nd_pfn_validate(struct nd_pfn *nd_pfn, const char *sig)
align = 1UL << ilog2(offset);
mode = le32_to_cpu(pfn_sb->mode);
+ /*
+ * Check whether the we support the alignment
+ */
+ if (!nd_supported_alignment(align)) {
+ dev_err(&nd_pfn->dev, "init failed, settings mismatch\n");
+ dev_dbg(&nd_pfn->dev, "align: %lx:%lx\n", nd_pfn->align, align);
print both at pr_err(), context-less errors are mostly just annoying.
+ return -EINVAL;
+ }
That is the same style used in other part of the code. Was not sure why
that align details was not printed with dev_err. Do we want to switch
other usage
if (nd_pfn->align != align || nd_pfn->mode != mode) {
dev_err(&nd_pfn->dev,
"init failed, settings mismatch\n");
dev_dbg(&nd_pfn->dev, "align: %lx:%lx mode: %d:%d\n",
nd_pfn->align, align, nd_pfn->mode,
mode);
+
if (!nd_pfn->uuid) {
/*
* When probing a namepace via nd_pfn_probe() the uuid
@@ -743,7 +770,7 @@ static int nd_pfn_init(struct nd_pfn *nd_pfn)
* PMD_SIZE for most architectures.
*/
offset = ALIGN(start + SZ_8K + 64 * npfns + dax_label_reserve,
- max(nd_pfn->align, PMD_SIZE)) - start;
+ nd_altmap_align_size(nd_pfn->align)) - start;
} else if (nd_pfn->mode == PFN_MODE_RAM)
offset = ALIGN(start + SZ_8K + dax_label_reserve,
nd_pfn->align) - start;
--
2.20.1
-aneesh
_______________________________________________
Linux-nvdimm mailing list
[email protected]
https://lists.01.org/mailman/listinfo/linux-nvdimm
