So my previous diff only did W^X for the bootstrap page tables. Since
these page tables get replaced quickly afterwards, W^X wasn't really
enforced. The diff below hopefully does a better job. It removes
some commented out code that was clearly trying to do the same thing
as pmap_map_stolen(). It also kills a bunch of debug printfs that I
deem unnecessary at this point and weren't showing up on the console
anyway.
ok?
Index: arch/arm64/arm64/pmap.c
===================================================================
RCS file: /cvs/src/sys/arch/arm64/arm64/pmap.c,v
retrieving revision 1.28
diff -u -p -r1.28 pmap.c
--- arch/arm64/arm64/pmap.c 24 Mar 2017 19:48:01 -0000 1.28
+++ arch/arm64/arm64/pmap.c 27 Mar 2017 16:51:39 -0000
@@ -1132,6 +1132,9 @@ CTASSERT(sizeof(struct pmapvp0) == 8192)
int mappings_allocated = 0;
int pted_allocated = 0;
+extern char __text_start[], _etext[];
+extern char __rodata_start[], _erodata[];
+
vaddr_t
pmap_bootstrap(long kvo, paddr_t lpt1, long kernelstart, long kernelend,
long ram_start, long ram_end)
@@ -1241,8 +1244,6 @@ pmap_bootstrap(long kvo, paddr_t lpt1,
// enable mappings for existing 'allocated' mapping in the bootstrap
// page tables
extern uint64_t *pagetable;
- extern char __text_start[], _etext[];
- extern char __rodata_start[], _erodata[];
extern char _end[];
vp2 = (void *)((long)&pagetable + kvo);
struct mem_region *mp;
@@ -1273,36 +1274,15 @@ pmap_bootstrap(long kvo, paddr_t lpt1,
}
}
- // At this point we are still running on the bootstrap page tables
- // however all memory allocated for the final page tables is
- // 'allocated' and should now be mapped
- // This means we are able to use the virtual addressing to
- // enter the final mappings into the new mapping tables.
-
-#if 0
- vm_prot_t prot;
-
- for (mp = pmap_allocated; mp->size != 0; mp++) {
- vaddr_t va;
- paddr_t pa;
- vsize_t size;
- extern char *etext();
- //printf("mapping %08x sz %08x\n", mp->start, mp->size);
- for (pa = mp->start, va = pa + kvo, size = mp->size & ~0xfff;
- size > 0;
- va += PAGE_SIZE, pa+= PAGE_SIZE, size -= PAGE_SIZE)
- {
- pa = va - kvo;
- prot = PROT_READ|PROT_WRITE;
- if ((vaddr_t)va >= (vaddr_t)kernelstart &&
- (vaddr_t)va < (vaddr_t)etext)
- prot |= PROT_EXEC; // XXX remove WRITE?
- pmap_kenter_cache(va, pa, prot, PMAP_CACHE_WB);
- }
- }
-#endif
pmap_avail_fixup();
+ /*
+ * At this point we are still running on the bootstrap page
+ * tables however all memory for the final page tables is
+ * 'allocated' and should now be mapped. This means we are
+ * able to use the virtual addressing to enter the final
+ * mappings into the new mapping tables.
+ */
vstart = pmap_map_stolen(kernelstart);
void (switch_mmu_kernel)(long);
@@ -2117,42 +2097,33 @@ pmap_steal_avail(size_t size, int align,
vaddr_t
pmap_map_stolen(vaddr_t kernel_start)
{
- int prot;
struct mem_region *mp;
- uint64_t pa, va, e;
- extern char *etext();
-
+ paddr_t pa;
+ vaddr_t va;
+ uint64_t e;
- int oldprot = 0;
- printf("mapping self\n");
for (mp = pmap_allocated; mp->size; mp++) {
- printf("start %16llx end %16llx\n", mp->start, mp->start +
mp->size);
- printf("exe range %16llx %16llx\n", kernel_start,
- (uint64_t)&etext);
for (e = 0; e < mp->size; e += PAGE_SIZE) {
- /* XXX - is this a kernel text mapping? */
- /* XXX - Do we care about KDB ? */
+ int prot = PROT_READ | PROT_WRITE;
+
pa = mp->start + e;
va = pa - pmap_avail_kvo;
+
if (va < VM_MIN_KERNEL_ADDRESS ||
va >= VM_MAX_KERNEL_ADDRESS)
continue;
- if ((vaddr_t)va >= (vaddr_t)kernel_start &&
- (vaddr_t)va < (vaddr_t)&etext) {
- prot = PROT_READ|PROT_WRITE|
- PROT_EXEC;
- } else {
- prot = PROT_READ|PROT_WRITE;
- }
- if (prot != oldprot) {
- printf("mapping v %16llx p %16llx prot %x\n",
va,
- pa, prot);
- oldprot = prot;
- }
+
+ if (va >= (vaddr_t)__text_start &&
+ va < (vaddr_t)_etext)
+ prot = PROT_READ | PROT_EXEC;
+ else if (va >= (vaddr_t)__rodata_start &&
+ va < (vaddr_t)_erodata)
+ prot = PROT_READ;
+
pmap_kenter_cache(va, pa, prot, PMAP_CACHE_WB);
}
}
- printf("last mapping v %16llx p %16llx\n", va, pa);
+
return va + PAGE_SIZE;
}
