Handle overflow in calculating max kmem size.

[kib]

Summary

vm_kmem_size is u_long, and it might be not capable of holding page count times PAGE_SIZE, even when scaled down by VM_KMEM_SIZE_SCALE. As bde reported, 12G PAE config ends up with zero for kmem size.

Explicitly check for overflow and clamp kmem size at vm_kmem_size_max. If we end up at zero size because VM_KMEM_SIZE_MAX is not defined, panic with clear explanation rather then failing in weird way.

[markj]

sys/kern/kern_malloc.c

935 ↗

(On Diff #52618)

Why does vm_kmem_size * PAGE_SIZE < vm_kmem_size imply that the overflowed value is 0?

Also, if vm_kmem_size_max == 0 and vm_kmem_size_min != 0, the code above may just silently set vm_kmem_size = vm_kmem_size_min in the overflow case. Shouldn't we panic in that case too?

[kib]

sys/kern/kern_malloc.c

935 ↗

(On Diff #52618)

I only want to panic if the value for the vm_kmem_size is insane. If there are some defined values for VM_KMEM_SIZE_MIN/MAX, and they are used, this should be fine. Even if only _MIN is defined and used.

So no, the overflow does not imply that the overflown value is zero, it only means that we fall back to vm_kmem_size_max, whatever it is (except zero).

[markj]

sys/kern/kern_malloc.c

935 ↗

(On Diff #52618)

But if there is an overflow and vm_kmem_size_max == vm_kmem_size_min == 0, we may set vm_kmem_size to a non-zero insane value.

[kib]

Do not that vm_kmem_size_max > 0 when handling overflow.
We will panic later if it is 0.