Hi,

running as a SEV guest we now calculate the page frame mask from the CPUID
provided "physical address bit reduction".  In amd64 pmap code we use now
the variable pg_frame instead of the define PG_FRAME (0x000ffffffffff000).

There is one instance in pmap code, where pg_frame is applied to virtual
addresses, not physical addresses.   I found a machine, where the address
bit reduction is rather large (six bits).  So the calculated pg_frame
is 0x00003fffffe00000.  However, on amd64 VM_MAX_ADDRESS is defined
as 0x00007fbfdfeff000.  So masking such a large virtual address with
pg_frame on that machine causes havoc.  Therefore, when masking VAs we
still have to use PG_FRAME.

I've re-checked the other applications of pg_mask, those should be ok.

Thoughts?

Take care,
HJ.
---------------------------------------------------------------------------
commit bb02fd22fc19f6e8222d097794d775d226db84ec
Author: Hans-Joerg Hoexer <hshoexer@genua.de>
Date:   Thu Aug 8 19:27:30 2024 +0200

    kern: use PG_FRAME when masking virtual addresses

diff --git a/sys/arch/amd64/amd64/pmap.c b/sys/arch/amd64/amd64/pmap.c
index cf8e59287e4..a9155d239b7 100644
--- a/sys/arch/amd64/amd64/pmap.c
+++ b/sys/arch/amd64/amd64/pmap.c
@@ -2159,8 +2159,8 @@ pmap_write_protect(struct pmap *pmap, vaddr_t sva, vaddr_t eva, vm_prot_t prot)
 	shootself = (scr3 == 0);
 
 	/* should be ok, but just in case ... */
-	sva &= pg_frame;
-	eva &= pg_frame;
+	sva &= PG_FRAME;
+	eva &= PG_FRAME;
 
 	if (!(prot & PROT_READ))
 		set |= pg_xo;

• smime.p7s (5K)