No objections, but why ? link_elf_obj also uses OBJT_DEFAULT.

I think you need to provide some explanation how your l/b bits functions are supposed to work.

I dislike the reliance on equal FreeBSD and Linux flags values. Please use LINUX_MFD_XXX constants in linuxolator. You may add _Static_asserts that they are equal and avoid translating the current flags, but I do not think that it is sustainable in long run.

In D20674#476867, @alc wrote:

In D20674#476856, @alc wrote:

In D20674#476783, @kib wrote:

That said, I suspect that we no longer need to avoid initial blocks. At least not for the label preservation.

I'm afraid that is not true. I found out the hard way that it is needed. I lost a disk label, swap area, and file system when I first enabled "trimonce". However, the configuration of that storage device is usual. The first partition is a swap area, not a file system.

I think the approach of relying on the object lock is fine.

In D20674#476657, @alc wrote:

I believe that there is a sound reason for not reporting the unmodified size of the device. Specifically, the swap pager can't use the first, if I recall correctly, 8KB of a swap device because it might contain a disk label. However, I can't explain why we use dmmax to calculate the modified size.

Why do you need a sysctl ? If bumping, simply increase to some larger but reasonable value, e.g. 2M. IOSIZE_MAX is definitely unreasonable.

Is there any public specs about NVDIMM enumeration using e820 ?

More notes by alc.

More alc' notes.

I stil has an opinion that the mp root vnode caching should be better integrated with existing code. Right now it feels as a side hack.

In D21703#475705, @ray wrote:

Return back #ifdef NEW_PCIB, @jhb not sure it is safe to omit.

I was told by bde (?) that after the block of code that is described by the multi-line comment, there should be a blank line. Since you carefully document other moments, this one should be described as well, I believe.

Alan, any more comments ? I want to close the review.

In D21767#475394, @markj wrote:

Here is a program that reproduces the panic: https://reviews.freebsd.org/P319

The program that syzkaller generated uses fewer system calls but is multi-threaded.

In D21767#475356, @alc wrote:

Do we actually allow truncation on a running executable file?

Would pmap_page_is_mapped() also a good or even better place to put such asserts ?

Note a use case which must not be broken: often the workstation where syscall files are regenerated is running some kind of stable or even a release.

Perhaps name it atomic_ops.h or atomic_algo.h.

In D21767#475000, @markj wrote:

In D21767#474999, @kib wrote:

What is the point of deactivating them ? Inactive scan frees invalid pages upon encountering them.

As opposed to doing nothing? vm_page_deactivate_noreuse() ensures that the pages skip LRU, so the page daemon (which handles races with concurrent wirings) will free them sooner.

What is the point of deactivating them ? Inactive scan frees invalid pages upon encountering them.

In D21762#474968, @rmacklem wrote:

So, do you think that just replacing n_mtx with n_solock makes more sense?

Yes.

Is there any way to enable SPARSE_MAPPING except manually adding #define ?