[PATCH v4 1/2] mm/hwpoison: Avoid the impact of hwpoison_filter() return value on mce handler

[PATCH v4 1/2] mm/hwpoison: Avoid the impact of hwpoison_filter() return value on mce handler

[luofei]

When the hwpoison page meets the filter conditions, it should
not be regarded as successful memory_failure() processing for
mce handler, but should return a value(-EHWPOISON), otherwise
mce handler regards the error page has been identified and
isolated, which may lead to calling set_mce_nospec() to change
page attribute, etc.

Here a new MF_MCE_HANDLE flag is introduced to identify the
call from the mce handler and instruct hwpoison_filter() to
return -EHWPOISON, otherwise return 0 for compatibility with
the hwpoison injector.

Signed-off-by: luofei <luofei@unicloud.com>
---
 arch/x86/kernel/cpu/mce/core.c | 15 +++++++++------
 include/linux/mm.h             |  1 +
 mm/memory-failure.c            | 14 ++++++++++++--
 3 files changed, 22 insertions(+), 8 deletions(-)

diff --git a/arch/x86/kernel/cpu/mce/core.c b/arch/x86/kernel/cpu/mce/core.c
index 5818b837fd4d..de29165c65b6 100644
--- a/arch/x86/kernel/cpu/mce/core.c
+++ b/arch/x86/kernel/cpu/mce/core.c
@@ -612,7 +612,7 @@ static int uc_decode_notifier(struct notifier_block *nb, unsigned long val,
 		return NOTIFY_DONE;
 
 	pfn = mce->addr >> PAGE_SHIFT;
-	if (!memory_failure(pfn, 0)) {
+	if (!memory_failure(pfn, MF_MCE_HANDLE)) {
 		set_mce_nospec(pfn, whole_page(mce));
 		mce->kflags |= MCE_HANDLED_UC;
 	}
@@ -1286,7 +1286,7 @@ static void kill_me_now(struct callback_head *ch)
 static void kill_me_maybe(struct callback_head *cb)
 {
 	struct task_struct *p = container_of(cb, struct task_struct, mce_kill_me);
-	int flags = MF_ACTION_REQUIRED;
+	int flags = MF_ACTION_REQUIRED | MF_MCE_HANDLE;
 	int ret;
 
 	p->mce_count = 0;
@@ -1303,9 +1303,12 @@ static void kill_me_maybe(struct callback_head *cb)
 	}
 
 	/*
-	 * -EHWPOISON from memory_failure() means that it already sent SIGBUS
-	 * to the current process with the proper error info, so no need to
-	 * send SIGBUS here again.
+	 * -EHWPOISON from memory_failure() means that memory_failure() did
+	 * not handle the error event for the following reason:
+	 *  - SIGBUS has already been sent to the current process with the
+	 *    proper error info, or
+	 *  - hwpoison_filter() filtered the event,
+	 * so no need to deal with it more.
 	 */
 	if (ret == -EHWPOISON)
 		return;
@@ -1320,7 +1323,7 @@ static void kill_me_never(struct callback_head *cb)
 
 	p->mce_count = 0;
 	pr_err("Kernel accessed poison in user space at %llx\n", p->mce_addr);
-	if (!memory_failure(p->mce_addr >> PAGE_SHIFT, 0))
+	if (!memory_failure(p->mce_addr >> PAGE_SHIFT, MF_MCE_HANDLE))
 		set_mce_nospec(p->mce_addr >> PAGE_SHIFT, p->mce_whole_page);
 }
 
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 213cc569b192..f4703f948e9a 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -3188,6 +3188,7 @@ enum mf_flags {
 	MF_MUST_KILL = 1 << 2,
 	MF_SOFT_OFFLINE = 1 << 3,
 	MF_UNPOISON = 1 << 4,
+	MF_MCE_HANDLE = 1 << 5,
 };
 extern int memory_failure(unsigned long pfn, int flags);
 extern void memory_failure_queue(unsigned long pfn, int flags);
diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index 97a9ed8f87a9..59d6d939a752 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -1526,7 +1526,10 @@ static int memory_failure_hugetlb(unsigned long pfn, int flags)
 				if (TestClearPageHWPoison(head))
 					num_poisoned_pages_dec();
 				unlock_page(head);
-				return 0;
+				if (flags & MF_MCE_HANDLE)
+					return -EHWPOISON;
+				else
+					return 0;
 			}
 			unlock_page(head);
 			res = MF_FAILED;
@@ -1613,7 +1616,10 @@ static int memory_failure_dev_pagemap(unsigned long pfn, int flags,
 		goto out;
 
 	if (hwpoison_filter(page)) {
-		rc = 0;
+		if (flags & MF_MCE_HANDLE)
+			rc = -EHWPOISON;
+		else
+			rc = 0;
 		goto unlock;
 	}
 
@@ -1837,6 +1843,10 @@ int memory_failure(unsigned long pfn, int flags)
 			num_poisoned_pages_dec();
 		unlock_page(p);
 		put_page(p);
+		if (flags & MF_MCE_HANDLE)
+			res = -EHWPOISON;
+		else
+			res = 0;
 		goto unlock_mutex;
 	}
 
-- 
2.27.0