The big thing that needs to be dealt with for this to make sense is to have a way to specify the filename that isn't a table embedded in libc. A first cut could be a special index that means no-file and perhaps using -1 as the line, but without that, we can't scale beyond files in the base system. (This also poses a problem for out of tree kernel modules.) We might also want macros like the kernel, but I'm less sure of that.

I ended up pulling the implementations out of the commit I merged because their dependencies (kernel option for sys/conf/files and cheri headers aren't available yet.)

In D33964#1328991, @markj wrote:

In D33964#1328982, @brooks wrote:

In D33964#1328968, @markj wrote:

An old diff. Is ptraddr_t the right type here?

If you used ptraddr_t you'd have to restore the compat32 support since it is in practice size_t. This seems like a good use of kvaddr_t.

kib already pointed out that kvaddr_t seems like the wrong type since the sysctl is exporting user addresses.

In D33964#1328982, @brooks wrote:

In D33964#1328968, @markj wrote:

An old diff. Is ptraddr_t the right type here?

If you used ptraddr_t you'd have to restore the compat32 support since it is in practice size_t. This seems like a good use of kvaddr_t.

In D33964#1328968, @markj wrote:

An old diff. Is ptraddr_t the right type here?

Move _data implementation to libkern/memcpy_data. Move cherification of memmove to D57965.

I'm going to further decompose this review to pull the cherification of memmove into it's own commit.

Clarify which pointers provenance is preserved

Simplify C implementation of mem{cpy,move}.

Use __nodiscard

Fix typo

Use __nodiscard

Fix headers. Mention that one past the end is permitted.

Created by accident. I blame the heat

fix typo

Be clear that sooptcopyoutptr is missing because it makes no sense

Fix some manpage nits

Doc fixes. Add copyinptr_nofault

Address feedback, switch references to memory_model.7

Move contents to a new memory_model.7

Rework intro text.

In D57661#1325572, @ziaee wrote:

Could we also put some sbom tags?

Use title case, rebase on D57820

Clarify alignment requirement somewhat

Fix more typos

Address feedback from @emaste

Link manpages to the description of pointer provenance in arch(7)

Document and link manpages to the description of pointer provenacne in arch(7)

Link manpages to the description of pointer provenacne in arch(7)

Link manpages to the description of pointer provenacne in arch(7)

Sorry about missing the label in i386.

For some reason this failed to auto close. I've manually linked rG2ee2ef607f60.

Discourage use of bcopy

Drop gotos.

Fix sfbuf and transient map leaks

use memcpy_data

Drop bcopy_data

• Rename to uiomove_step to move to subr_uio.c
• Pass struct uio * as the last argument (similar to other functions) instead of separate segflg and rw arguments

In D57687#1322718, @kib wrote:

amd64 is fine, but there are tons of bcopy use in amd64/amd64/*.c. Why is it fine for them to stay with bcopy? Or put it differently, why amd64 is affected at all?

With CHERI we'll be adding to variants of UIO_READ/UIO_WRITE that preserved pointer provenance and centralizing the definitions will make adding that a lot cleaner.

Created D57680 to follow POSIX/C's lead and stop talking about strings in mem* functions.

Rebase after removing references to strings in base manpages