Mark Johnston

Change rangelock handling in FreeBSD's zfs_getpages()

2

Unconditionally hold the rangelock in zfs_getpages(). ### Motivation and Context This is reportedly required for direct I/O support on FreeBSD. ### Description This change modifies zfs_getpages() to uncondtionally acquire the...

Make jemalloc bound returned pointers to their usable size

9

This allows MRS to handle realloc() loops more efficiently, since it can avoid going back to the allocator every time the allocation size increases. See #2133. It is also consistent...

several caprevoke optimizations

1

The procstat-related patches should perhaps be upstreamed as well, though `procstat vm` doesn't have a verbose mode in vanilla FreeBSD.

Remove tests for experimental jemalloc ABIs

1

This is needed in order to build snmalloc's tests on recent FreeBSD main. I verified that all of the remaining tests pass there.

the job scheduler is slow when upgrading a desktop system

3

I have a desktop system that I'm upgrading from FreeBSD:15:amd64 to FreeBSD:16:amd64, about 1000 ports packages are to be upgraded. Once I confirm the upgrade, pkg burns the CPU for...

pkg solver is looking at too many packages

12

I tried upgrading some pkgbase systems from 15.0-PRERELEASE to 16.0-CURRENT. First the kernels: ``` # pkg -o ABI=FreeBSD:16:amd64 upgrade -r FreeBSD-base -g FreeBSD-kernel-* ... Installed packages to be UPGRADED: FreeBSD-kernel-generic:...

Import pf fixes for CHERI

2

These are sufficient to get all but 3 of the pf tests to pass; I'm still investigating the remaining failures.

caprevoke: Avoid crashing if the revoker encounters a kernel capability

9

This mitigates a bug in the `DTRACEIOC_EPROBE` ioctl handler (which I have since fixed locally) that caused kernel panics when running certain dtrace scripts.

The kstack allocator includes a single guard page at the bottom of each stack. In principle these are unnecessary in purecap kernels and we should try setting KSTACK_GUARD_PAGES to 0...