On x86 platform, if we use "/sys/devices/system/memory/soft_offline_page" to offline a
free page twice, the value of mce_bad_pages will be added twice. So this is an error,
since the page was already marked HWPoison, we should skip the page and don't add the
value of mce_bad_pages.
$ cat /proc/meminfo | grep HardwareCorrupted
soft_offline_page()
get_any_page()
atomic_long_add(1, &mce_bad_pages)
The free page which marked HWPoison is still managed by page buddy allocator. So when
offlining it again, get_any_page() always returns 0 with
"pr_info("%s: %#lx free buddy page\n", __func__, pfn);".
When page is allocated, the PageBuddy is removed in bad_page(), then get_any_page()
returns -EIO with pr_info("%s: %#lx: unknown zero refcount page type %lx\n", so
mce_bad_pages will not be added.
Signed-off-by: Xishi Qiu <qiuxishi@xxxxxxxxxx>
Signed-off-by: Jiang Liu <jiang.liu@xxxxxxxxxx>
---
mm/memory-failure.c | 5 +++++
1 files changed, 5 insertions(+), 0 deletions(-)
diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index 8b20278..02a522e 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -1375,6 +1375,11 @@ static int get_any_page(struct page *p, unsigned long pfn, int flags)
if (flags & MF_COUNT_INCREASED)
return 1;
+ if (PageHWPoison(p)) {
+ pr_info("%s: %#lx page already poisoned\n", __func__, pfn);
+ return -EBUSY;
+ }
+
/*
* The lock_memory_hotplug prevents a race with memory hotplug.
* This is a big hammer, a better would be nicer.
--
1.7.6.1
--
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>