Use ATTR_DBM for ATTR_SW_DBM instead of a software bit.

[markj]

Summary

On ARMv8.0 ATTR_DBM is reserved. But Linux uses it anyway, so I presume
it is safe to do so in FreeBSD as well.

I am torn on whether to keep using "ATTR_SW_DBM" or whether I should
just replace all references with ATTR_SW_DBM. Any preferences?

[alc]

Do we have access to a machine with hardware DBM support?

[markj]

In D21121#458634, @alc wrote:

Do we have access to a machine with hardware DBM support?

Not yet. It looks like the thunderx2 does not support hardware DBM or AF management: http://dmesgd.nycbug.org/index.cgi?do=view&id=4798

[alc]

I am torn on whether to keep using "ATTR_SW_DBM" or whether I should
just replace all references with ATTR_SW_DBM. Any preferences?

No, not really.

[andrew]

In D21121#458634, @alc wrote:

Do we have access to a machine with hardware DBM support?

As far as I know only in the form of a simulator, e.g. the Architecture Envelope Model from https://developer.arm.com/tools-and-software/simulation-models/fixed-virtual-platforms

head/sys/arm64/include/pte.h

 typedef uint64_t        pt_entry_t;             /* page table entry */
 #endif
 
 /* Block and Page attributes */
 #define ATTR_MASK_H     UINT64_C(0xfffc000000000000)
 #define ATTR_MASK_L     UINT64_C(0x0000000000000fff)
 #define ATTR_MASK       (ATTR_MASK_H | ATTR_MASK_L)
 /* Bits 58:55 are reserved for software */
-#define ATTR_SW_UNUSED  (1UL << 58)
-#define _ATTR_SW_DBM    (1UL << 57)
+#define ATTR_SW_UNUSED2 (1UL << 58)
+#define ATTR_SW_UNUSED1 (1UL << 57)
 #define ATTR_SW_MANAGED (1UL << 56)
 #define ATTR_SW_WIRED   (1UL << 55)
 #define ATTR_UXN        (1UL << 54)
 #define ATTR_PXN        (1UL << 53)
 #define ATTR_XN         (ATTR_PXN | ATTR_UXN)
 #define ATTR_CONTIGUOUS (1UL << 52)
 #define ATTR_DBM        (1UL << 51)
 #define ATTR_nG         (1 << 11)