Skip to contentSkip to conversation

hw/poison: Avoid the impact of hwpoison_filter() return value on mce handler

PATCH Superseded luofei 2 participants 3 messages Feb 18, 2022 – Feb 21, 2022

Subsystems (6)

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore

luofei

author
#0 lore raw

PATCH v3 1/2 hw/poison: Avoid the impact of hwpoison_filter() return value on mce handler

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 introdulced to identify the call from the mce handler and instruct hwpoison_filter() to return -EHWPOISON, other return 0 for compatibility with the hwpoison injector.

---
 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(-)
arch/x86/kernel/cpu/mce/core.c +9 -6
index 5818b837fd4d..de29165c65b6 100644
@@ -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);
 }
 
include/linux/mm.h +1 -0
index 213cc569b192..f4703f948e9a 100644
@@ -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);
mm/memory-failure.c +12 -2
index 97a9ed8f87a9..59d6d939a752 100644
@@ -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;
 	}
 
Signature · 1 line
2.27.0
#1 lore raw

Conventionally, "hwpoison" or "HWPOISON" (instead of "hw/poison") is used in the subsystem tag in $SUBJECT, so could you align it (for grep friendliness) ?

# I should've mentioned this in previous review round, sorry.

Quoted 8 lines · 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 introdulced to identify the

^^ ^^^^^^^^^^^
single '_' ? introduced

Quoted 2 lines

call from the mce handler and instruct hwpoison_filter() to return -EHWPOISON, other return 0 for compatibility with the

                     ^^^^^
                     otherwise ?
Quoted 3 lines

hwpoison injector.

The diff looks ok to me, so with updates on the subject/description ...

Thanks,
Naoya Horiguchi

Quoted 106 lines
---
 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(-)
arch/x86/kernel/cpu/mce/core.c +9 -6
index 5818b837fd4d..de29165c65b6 100644
@@ -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);
 }
 
include/linux/mm.h +1 -0
index 213cc569b192..f4703f948e9a 100644
@@ -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);
mm/memory-failure.c +12 -2
index 97a9ed8f87a9..59d6d939a752 100644
@@ -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;
 	}
 
Signature · 1 line
2.27.0
author Acked-by Naoya Horiguchi
#2 lore raw
Quoted 30 lines

Conventionally, "hwpoison" or "HWPOISON" (instead of "hw/poison") is used in the subsystem tag in $SUBJECT, so could you align it (for grep friendliness) ?

# I should've mentioned this in previous review round, sorry.

Quoted 6 lines · 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

Quoted 7 lines

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 introdulced to identify the

^^ ^^^^^^^^^^^

single '_' ? introduced

Quoted 11 lines
Quoted 3 lines

call from the mce handler and instruct hwpoison_filter() to return

-EHWPOISON, other return 0 for compatibility with the

                    ^^^^^

                    otherwise ?
Quoted 1 line

hwpoison injector.

I am so sorry for my carelessness, I will resubmit the updated patches.

Thanks

________________________________
发件人: HORIGUCHI NAOYA(堀口 直也) <[email protected]>
发送时间: 2022年2月21日 9:32:16
收件人: 罗飞
抄送: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected] 主题: Re: [PATCH v3 1/2] hw/poison: Avoid the impact of hwpoison_filter() return value on mce handler

Conventionally, "hwpoison" or "HWPOISON" (instead of "hw/poison") is used in the subsystem tag in $SUBJECT, so could you align it (for grep friendliness) ?

# I should've mentioned this in previous review round, sorry.

Quoted 8 lines · 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 introdulced to identify the

^^ ^^^^^^^^^^^
single '_' ? introduced

Quoted 2 lines

call from the mce handler and instruct hwpoison_filter() to return -EHWPOISON, other return 0 for compatibility with the

                     ^^^^^
                     otherwise ?
Quoted 3 lines

hwpoison injector.

The diff looks ok to me, so with updates on the subject/description ...

Thanks,
Naoya Horiguchi

Quoted 106 lines
---
 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(-)
arch/x86/kernel/cpu/mce/core.c +9 -6
index 5818b837fd4d..de29165c65b6 100644
@@ -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);
 }
 
include/linux/mm.h +1 -0
index 213cc569b192..f4703f948e9a 100644
@@ -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);
mm/memory-failure.c +12 -3
index 97a9ed8f87a9..59d6d939a752 100644
@@ -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