rtld-elf: Fix VDSO Obj_Entry mapsize and report in utrace(2)

[jrtc27]

Summary

Subtracing addr is only appropriate for position-dependent objects,
where vaddrbase would also be the same value. For position-independent
objects, like the VDSO (which we already assume due to setting vaddrbase
to 0), the segments start at 0, not addr.

[jrtc27]

Before:

jrtc4@amoracia:~/freebsd-src$ LD_UTRACE=1 ktrace -t u /libexec/ld-elf.so.1 /usr/bin/true && kdump | grep loaded
  6613 ld-elf.so.1 USER  RTLD: loaded   0x47728f403408 @ 0x1dbb7f8e9000 - 0x1dbb7f8e8fff ([vdso])
  6613 ld-elf.so.1 USER  RTLD: loaded   0x47728f403808 @ 0x1dbb7ff71000 - 0x1dbb80273fff (/lib/libc.so.7)

After:

jrtc4@amoracia:~/freebsd-src$ LD_UTRACE=1 ktrace -t u ~/freebsd-obj/home/jrtc4/freebsd-src/amd64.amd64/libexec/rtld-elf/ld-elf.so.1 /usr/bin/true && kdump | grep loaded
  6607 ld-elf.so.1 USER  RTLD: loaded   0x55e2a0a03408 @ 0x2e106a0ad000 - 0x2e106a0adfff ([vdso])
  6607 ld-elf.so.1 USER  RTLD: loaded   0x55e2a0a03808 @ 0x2e106b725000 - 0x2e106ba27fff (/lib/libc.so.7)

[kib]

libexec/rtld-elf/rtld.c

3034

Why the round_page() is needed?

[jrtc27]

libexec/rtld-elf/rtld.c

3034

Because you can't have sub-page mappings? We round_page in digest_phdr. But maybe your point is that this a special sub-page-granularity object, unlike the ones we mmap from userspace?

[kib]

libexec/rtld-elf/rtld.c

3034

Exactly. The vdso object itself only occupies parts of the shared page(s), there is other stuff co-located with it, like timehands, or rndg data.

[jrtc27]

Don't round up

[kib]

In the description: s/eto/to/

[jrtc27]

Now:

jrtc4@amoracia:~/freebsd-src$ LD_UTRACE=1 ktrace -t u ~/freebsd-obj/home/jrtc4/freebsd-src/amd64.amd64/libexec/rtld-elf/ld-elf.so.1 /usr/bin/true && kdump | grep loaded
 80756 ld-elf.so.1 USER  RTLD: loaded   0x5a069ee03408 @ 0x3dae988c1000 - 0x3dae988c12e5 ([vdso])
 80756 ld-elf.so.1 USER  RTLD: loaded   0x5a069ee03808 @ 0x3dae98deb000 - 0x3dae990edfff (/lib/libc.so.7)

[jrtc27]

In D51924#1186691, @kib wrote:

In the description: s/eto/to/

Uh yes, and there's another typo in there too, thanks.