Re: Current kernel with http://gnats.netbsd.org/39965 and pmap patches

[email protected] wrote:

> Thanks! I missed that one. I applied patch (but corrected place for minus sign
> for VM_MAX_KERNEL_ADDRESS just like in amiga/include/vmparam.h :))
> 
> Installer panics just like before with 030 cpu, but it works with 060 cpu, so it
> seems there's something funny happening with 030 that we need to sort out.

I don't understand what XXX calculations in amiga_init.c do and
how they map PA of Sysmap (Sysptmap_pa) into VM_MAX_KERNEL_ADDRESS
in mmu040_setup() case... (it's a bit different from other m68k ports)

But in mmu030_setup(), I don't think we can use the same way
to map Sysptmap_pa at the end of KVA with other m68k ports
which have 4KB page (i.e. PS/TIA/TIB == 12/10/10) because
amiga and atari have 8KB page (PS/TIA/TIB == 13/8/11) and
hardware segment table and software Sysptmap have different size.
(but I could be wrong since I'm not an m68k guru ;-)
---

--- atari/atari_init.c.0	2008-12-17 19:12:51.000000000 +0900
+++ atari/atari_init.c	2008-12-17 23:47:19.000000000 +0900
@@ -918,7 +918,7 @@
 	u_int		sysptmap_pa;	/* System page table		*/
 	u_int		kbase;
 {
-	st_entry_t	sg_proto, *sg;
+	st_entry_t	sg_proto, *sg, *esg;
 	pt_entry_t	pg_proto, *pg, *epg;
 
 	/*
@@ -938,14 +938,20 @@
 	}
 
 	/* 
-	 * invalidate the remainder of the tables
+	 * Invalidate the remainder of the tables.
 	 */
-	/* XXX PAGE_SIZE dependent constant: 256 or 1024 */
-	epg = (u_int *)(sysptmap_pa + (256 - 1) * sizeof(st_entry_t));
-	while(pg < epg) {
+#define TIA_SIZE	256
+	esg = (u_int *)(sysptmap_pa + (TIA_SIZE - 1) * sizeof(st_entry_t));
+	while (sg < esg)
 		*sg++ = SG_NV;
+
+	epg = (u_int *)(sysptmap_pa + (NPTEPG - 1) * sizeof(st_entry_t));
+	while (pg < epg)
 		*pg++ = PG_NV;
-	}
+
+	/*
+	 * Initialize the last one to point Sysptmap.
+	 */
 	*sg = sysptmap_pa | SG_RW | SG_V;
 	*pg = sysptmap_pa | PG_RW | PG_CI | PG_V;
 }

---
Izumi Tsutsui