tools/vm/page-types.c: support KPF_SOFTDIRTY bit

Soft dirty bit allows us to track which pages are written since the last
clear_ref (by "echo 4 > /proc/pid/clear_refs".) This is useful for
userspace applications to know their memory footprints.

Note that the kernel exposes this flag via bit[55] of /proc/pid/pagemap,
and the semantics is not a default one (scheduled to be the default in the
near future.) However, it shifts to the new semantics at the first
clear_ref, and the users of soft dirty bit always do it before utilizing
the bit, so that's not a big deal. Users must avoid relying on the bit in
page-types before the first clear_ref.

Signed-off-by: Naoya Horiguchi <[e-mail]>
Cc: Wu Fengguang <[e-mail]>
Cc: Pavel Emelyanov <[e-mail]>
Cc: Cyrill Gorcunov <[e-mail]>
Signed-off-by: Andrew Morton <[e-mail]>
Signed-off-by: Linus Torvalds <[e-mail]>

include/linux/kernel-page-flags.h

···
 #define KPF_OWNER_PRIVATE	37
 #define KPF_ARCH		38
 #define KPF_UNCACHED		39
+#define KPF_SOFTDIRTY		40
 
 #endif /* LINUX_KERNEL_PAGE_FLAGS_H */

tools/vm/page-types.c

···
 #define PM_PSHIFT_BITS      6
 #define PM_PSHIFT_OFFSET    (PM_STATUS_OFFSET - PM_PSHIFT_BITS)
 #define PM_PSHIFT_MASK      (((1LL << PM_PSHIFT_BITS) - 1) << PM_PSHIFT_OFFSET)
-#define PM_PSHIFT(x)        (((u64) (x) << PM_PSHIFT_OFFSET) & PM_PSHIFT_MASK)
+#define __PM_PSHIFT(x)      (((uint64_t) (x) << PM_PSHIFT_OFFSET) & PM_PSHIFT_MASK)
 #define PM_PFRAME_MASK      ((1LL << PM_PSHIFT_OFFSET) - 1)
 #define PM_PFRAME(x)        ((x) & PM_PFRAME_MASK)
 
+#define __PM_SOFT_DIRTY      (1LL)
 #define PM_PRESENT          PM_STATUS(4LL)
 #define PM_SWAP             PM_STATUS(2LL)
+#define PM_SOFT_DIRTY       __PM_PSHIFT(__PM_SOFT_DIRTY)
 
 
 /*
···
 #define KPF_OWNER_PRIVATE	37
 #define KPF_ARCH		38
 #define KPF_UNCACHED		39
+#define KPF_SOFTDIRTY		40
 
 /* [48-] take some arbitrary free slots for expanding overloaded flags
  * not part of kernel API
···
 	[KPF_OWNER_PRIVATE]	= "O:owner_private",
 	[KPF_ARCH]		= "h:arch",
 	[KPF_UNCACHED]		= "c:uncached",
+	[KPF_SOFTDIRTY]		= "f:softdirty",
 
 	[KPF_READAHEAD]		= "I:readahead",
 	[KPF_SLOB_FREE]		= "P:slob_free",
···
 	return 1;
 }
 
-static uint64_t expand_overloaded_flags(uint64_t flags)
+static uint64_t expand_overloaded_flags(uint64_t flags, uint64_t pme)
 {
 	/* SLOB/SLUB overload several page flags */
 	if (flags & BIT(SLAB)) {
···
 	/* PG_reclaim is overloaded as PG_readahead in the read path */
 	if ((flags & (BIT(RECLAIM) | BIT(WRITEBACK))) == BIT(RECLAIM))
 		flags ^= BIT(RECLAIM) | BIT(READAHEAD);
+
+	if (pme & PM_SOFT_DIRTY)
+		flags |= BIT(SOFTDIRTY);
 
 	return flags;
 }
···
 	return flags;
 }
 
-static uint64_t kpageflags_flags(uint64_t flags)
+static uint64_t kpageflags_flags(uint64_t flags, uint64_t pme)
 {
-	flags = expand_overloaded_flags(flags);
-
-	if (!opt_raw)
+	if (opt_raw)
+		flags = expand_overloaded_flags(flags, pme);
+	else
 		flags = well_known_flags(flags);
 
 	return flags;
···
 }
 
 static void add_page(unsigned long voffset,
-		     unsigned long offset, uint64_t flags)
+		     unsigned long offset, uint64_t flags, uint64_t pme)
 {
-	flags = kpageflags_flags(flags);
+	flags = kpageflags_flags(flags, pme);
 
 	if (!bit_mask_ok(flags))
 		return;
···
 #define KPAGEFLAGS_BATCH	(64 << 10)	/* 64k pages */
 static void walk_pfn(unsigned long voffset,
 		     unsigned long index,
-		     unsigned long count)
+		     unsigned long count,
+		     uint64_t pme)
 {
 	uint64_t buf[KPAGEFLAGS_BATCH];
 	unsigned long batch;
···
 			break;
 
 		for (i = 0; i < pages; i++)
-			add_page(voffset + i, index + i, buf[i]);
+			add_page(voffset + i, index + i, buf[i], pme);
 
 		index += pages;
 		count -= pages;
···
 		for (i = 0; i < pages; i++) {
 			pfn = pagemap_pfn(buf[i]);
 			if (pfn)
-				walk_pfn(index + i, pfn, 1);
+				walk_pfn(index + i, pfn, 1, buf[i]);
 		}
 
 		index += pages;
···
 
 	for (i = 0; i < nr_addr_ranges; i++)
 		if (!opt_pid)
-			walk_pfn(0, opt_offset[i], opt_size[i]);
+			walk_pfn(0, opt_offset[i], opt_size[i], 0);
 		else
 			walk_task(opt_offset[i], opt_size[i]);