Date: Sat, 30 Aug 2008 15:03:18 GMT From: Nathan Whitehorn <nwhitehorn@FreeBSD.org> To: Perforce Change Reviews <perforce@freebsd.org> Subject: PERFORCE change 148849 for review Message-ID: <200808301503.m7UF3IXs049735@repoman.freebsd.org>
next in thread | raw e-mail | index | archive | help
http://perforce.freebsd.org/chv.cgi?CH=148849 Change 148849 by nwhitehorn@nwhitehorn_trantor on 2008/08/30 15:02:20 Partially fix boot while WITNESS and INVARIANTS are enabled. It will no longer silently hang, but there are still panics from lock-order reversals that will need the standard treatment. Affected files ... .. //depot/projects/ppc-g5/sys/powerpc/aim/mmu_oea64.c#2 edit Differences ... ==== //depot/projects/ppc-g5/sys/powerpc/aim/mmu_oea64.c#2 (text+ko) ==== @@ -902,6 +902,8 @@ /* Enter the pages */ PMAP_LOCK(&ofw_pmap); + PMAP_LOCK(kernel_pmap); + DISABLE_TRANS(msr); for (off = 0; off < translations[i].om_len; off += PAGE_SIZE) { struct vm_page m; @@ -915,11 +917,12 @@ * so we need them if we want to keep executing. */ if (translations[i].om_va + off < SEGMENT_LENGTH) { - moea64_kenter(mmup,translations[i].om_va + off, - translations[i].om_pa_lo + off); + moea64_enter_locked(kernel_pmap, + translations[i].om_va + off, &m, + prot, 1); } - /* Map the frame buffer into the kernel pmap */ + /* Map the frame buffer into the kernel pmap instead */ if ((translations[i].om_va + off >= fbva) && (translations[i].om_va + off < fbva + fbsize)) { @@ -932,6 +935,8 @@ ofw_mappings++; } ENABLE_TRANS(msr); + + PMAP_UNLOCK(kernel_pmap); PMAP_UNLOCK(&ofw_pmap); }
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200808301503.m7UF3IXs049735>