Re: kern/60522: evbppc/RB800 hits MI pmap KASSERT failure

The following reply was made to PR kern/60522; it has been noted by GNATS.

From: Rin Okuyama <[email protected]>
To: Nick Hudson <[email protected]>
Cc: [email protected], [email protected],
 [email protected], [email protected]
Subject: Re: kern/60522: evbppc/RB800 hits MI pmap KASSERT failure
Date: Sun, 2 Aug 2026 11:53:25 +0900

 On 2026/08/01 23:48, Nick Hudson wrote:
 > 
 > 
 >> On 1 Aug 2026, at 15:44, Rin Okuyama <[email protected]> wrote:
 >>
 >> Hi,
 >>
 >> On 2026/08/01 21:42, Nick Hudson wrote:
 >>> Is this a regression? If so can you bisect please
 >>
 >> Sure. The first bad commit was:
 >>
 >> https://github.com/NetBSD/src/commit/6162b83d2b80429d87007f1f1c451e538e577581
 > 
 > thanks.
 > 
 >>
 >> KASSERT failure disappears if part of this commit is reverted:
 >>
 >> ```
 >> diff --git a/sys/uvm/pmap/pmap.c b/sys/uvm/pmap/pmap.c
 >> index cb0f6b3b065..eba0cbba4c7 100644
 >> --- a/sys/uvm/pmap/pmap.c
 >> +++ b/sys/uvm/pmap/pmap.c
 >> @@ -1326,7 +1326,7 @@ pmap_pte_protect(pmap_t pmap, vaddr_t sva, vaddr_t eva, pt_entry_t *ptep,
 >> struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg);
 >>
 >> pmap_page_set_attributes(mdpg, VM_PAGEMD_MODIFIED);
 >> - if (VM_PAGEMD_EXECPAGE_P(mdpg) || (prot & VM_PROT_EXECUTE)) {
 >> + if (VM_PAGEMD_EXECPAGE_P(mdpg)) {
 >> KASSERT(!VM_PAGEMD_PVLIST_EMPTY_P(mdpg));
 >> #ifdef PMAP_VIRTUAL_CACHE_ALIASES
 >> if (VM_PAGEMD_CACHED_P(mdpg)) {
 >> ```
 >>
 > 
 > This doesn’t make a lot of sense to me at first glance.
 > 
 > Could you get the UVMHIST dumps with it in place right after the panic, please?
 
 Here it is!
 
 https://www.netbsd.org/~rin/kern60522_uvmhist.tgz
 
 Thanks,
 rin
 
 >>
 >>> Otherwise, please enable UVMHIST and dump pmaphist and pmapexechist after the panic.
 >>
 >