Hello,
Jeff Rizzo<riz%tastylime.net@localhost> wrote:
I assume this is because of the rmind-uvmplock branch merge, but the
latest amd64 domU kernel downloaded from the releng cluster panics at
boot:
...
I have committed various Xen fixes. PR/45062 should also be fixed,
or at least XEN3_DOM0 boots and works for me. I am not sure yet why
"(pte& PG_G) == 0 || pm == pmap_kernel()" assert fails for Xen, but
see below.
Xen kernels print "pmap_kenter_pa: mapping already present" messages
during the boot and after. While not fatal, they indicate dirty page
table entries (PTEs), which we normally should not have. I would guess
that page tables are not zeroed somewhere. Perhaps early boot code,
xen_pmap_bootstrap() does zeroing, but not sure if correctly.
It is not a new problem, it happens with 5.99.52 (pre uvmplock merge)
Xen kernel as well, except now diagnostic message is printed.