The KVM code has some repeated bugs in it around use of __pa() on
per-cpu data. Those data are not in an area on which __pa() is
valid. However, they are also called early enough in boot that
__vmalloc_start_set is not set, and thus the CONFIG_DEBUG_VIRTUAL
debugging does not catch them.
This adds a check to also verify them against max_low_pfn, which
we can use earler in boot than is_vmalloc_addr(). However, if
we are super-early in boot, max_low_pfn=0 and this will trip
on every call, so also make sure that max_low_pfn is set.
With this patch applied, CONFIG_DEBUG_VIRTUAL will actually
catch the bug I was chasing.
I'd love to find a generic way so that any __pa() call on percpu
areas could do a BUG_ON(), but there don't appear to be any nice
and easy ways to check if an address is a percpu one. Anybody
have ideas on a way to do this?
---
linux-2.6.git-dave/arch/x86/mm/numa.c | 2 +-
linux-2.6.git-dave/arch/x86/mm/pat.c | 4 ++--
linux-2.6.git-dave/arch/x86/mm/physaddr.c | 9 ++++++++-
3 files changed, 11 insertions(+), 4 deletions(-)
diff -puN arch/x86/mm/numa.c~make-DEBUG_VIRTUAL-work-earlier-in-boot arch/x86/mm/numa.c
--- linux-2.6.git/arch/x86/mm/numa.c~make-DEBUG_VIRTUAL-work-earlier-in-boot 2013-01-09 13:55:44.718676439 -0500
+++ linux-2.6.git-dave/arch/x86/mm/numa.c 2013-01-09 13:55:44.726676509 -0500
@@ -219,7 +219,7 @@ static void __init setup_node_data(int n
*/
nd = alloc_remap(nid, nd_size);
if (nd) {
- nd_pa = __pa(nd);
+ nd_pa = __phys_addr_nodebug(nd);
remapped = true;
} else {
nd_pa = memblock_alloc_nid(nd_size, SMP_CACHE_BYTES, nid);
diff -puN arch/x86/mm/pat.c~make-DEBUG_VIRTUAL-work-earlier-in-boot arch/x86/mm/pat.c
--- linux-2.6.git/arch/x86/mm/pat.c~make-DEBUG_VIRTUAL-work-earlier-in-boot 2013-01-09 13:55:44.722676474 -0500
+++ linux-2.6.git-dave/arch/x86/mm/pat.c 2013-01-09 13:55:44.726676509 -0500
@@ -560,10 +560,10 @@ int kernel_map_sync_memtype(u64 base, un
{
unsigned long id_sz;
- if (base >= __pa(high_memory))
+ if (base > __pa(high_memory-1))
return 0;
- id_sz = (__pa(high_memory) < base + size) ?
+ id_sz = (__pa(high_memory-1) <= base + size) ?
__pa(high_memory) - base :
size;
diff -puN arch/x86/mm/physaddr.c~make-DEBUG_VIRTUAL-work-earlier-in-boot arch/x86/mm/physaddr.c
--- linux-2.6.git/arch/x86/mm/physaddr.c~make-DEBUG_VIRTUAL-work-earlier-in-boot 2013-01-09 13:55:44.722676474 -0500
+++ linux-2.6.git-dave/arch/x86/mm/physaddr.c 2013-01-09 13:55:44.726676509 -0500
@@ -1,3 +1,4 @@
+#include <linux/bootmem.h>
#include <linux/mmdebug.h>
#include <linux/module.h>
#include <linux/mm.h>
@@ -47,10 +48,16 @@ EXPORT_SYMBOL(__virt_addr_valid);
#ifdef CONFIG_DEBUG_VIRTUAL
unsigned long __phys_addr(unsigned long x)
{
+ unsigned long phys_addr = x - PAGE_OFFSET;
/* VMALLOC_* aren't constants */
VIRTUAL_BUG_ON(x < PAGE_OFFSET);
VIRTUAL_BUG_ON(__vmalloc_start_set && is_vmalloc_addr((void *) x));
- return x - PAGE_OFFSET;
+ /* max_low_pfn is set early, but not _that_ early */
+ if (max_low_pfn) {
+ VIRTUAL_BUG_ON((phys_addr >> PAGE_SHIFT) > max_low_pfn);
+ BUG_ON(slow_virt_to_phys((void *)x) != phys_addr);
+ }
+ return phys_addr;
}
EXPORT_SYMBOL(__phys_addr);
#endif
_
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>