在 2024/8/19 18:30, Jonathan Cameron 写道:

On Tue, 28 May 2024 16:59:11 +0800
Tong Tiangen [off-list ref] wrote:

quoted

For the arm64 kernel, when it processes hardware memory errors for
synchronize notifications(do_sea()), if the errors is consumed within the
kernel, the current processing is panic. However, it is not optimal.

Take copy_from/to_user for example, If ld* triggers a memory error, even in
kernel mode, only the associated process is affected. Killing the user
process and isolating the corrupt page is a better choice.

New fixup type EX_TYPE_KACCESS_ERR_ZERO_ME_SAFE is added to identify insn
that can recover from memory errors triggered by access to kernel memory.

Signed-off-by: Tong Tiangen <redacted>

Hi - this is going slow :(

A few comments inline in the meantime but this really needs ARM maintainers
to take a (hopefully final) look.

Jonathan

quoted

diff --git a/arch/arm64/include/asm/asm-extable.h b/arch/arm64/include/asm/asm-extable.h
index 980d1dd8e1a3..9c0664fe1eb1 100644
--- a/arch/arm64/include/asm/asm-extable.h
+++ b/arch/arm64/include/asm/asm-extable.h

@@ -5,11 +5,13 @@
  #include <linux/bits.h>
  #include <asm/gpr-num.h>
  
-#define EX_TYPE_NONE			0
-#define EX_TYPE_BPF			1
-#define EX_TYPE_UACCESS_ERR_ZERO	2
-#define EX_TYPE_KACCESS_ERR_ZERO	3
-#define EX_TYPE_LOAD_UNALIGNED_ZEROPAD	4
+#define EX_TYPE_NONE				0
+#define EX_TYPE_BPF				1
+#define EX_TYPE_UACCESS_ERR_ZERO		2
+#define EX_TYPE_KACCESS_ERR_ZERO		3
+#define EX_TYPE_LOAD_UNALIGNED_ZEROPAD		4
+/* kernel access memory error safe */
+#define EX_TYPE_KACCESS_ERR_ZERO_ME_SAFE	5

Does anyone care enough about the alignment to bother realigning for one
long line? I'd be tempted not to bother, but up to maintainers.

quoted

diff --git a/arch/arm64/lib/copy_to_user.S b/arch/arm64/lib/copy_to_user.S
index 802231772608..2ac716c0d6d8 100644
--- a/arch/arm64/lib/copy_to_user.S
+++ b/arch/arm64/lib/copy_to_user.S

@@ -20,7 +20,7 @@
   *	x0 - bytes not copied
   */
  	.macro ldrb1 reg, ptr, val
-	ldrb  \reg, [\ptr], \val
+	KERNEL_ME_SAFE(9998f, ldrb  \reg, [\ptr], \val)
  	.endm
  
  	.macro strb1 reg, ptr, val

@@ -28,7 +28,7 @@
  	.endm
  
  	.macro ldrh1 reg, ptr, val
-	ldrh  \reg, [\ptr], \val
+	KERNEL_ME_SAFE(9998f, ldrh  \reg, [\ptr], \val)
  	.endm
  
  	.macro strh1 reg, ptr, val

@@ -36,7 +36,7 @@
  	.endm
  
  	.macro ldr1 reg, ptr, val
-	ldr \reg, [\ptr], \val
+	KERNEL_ME_SAFE(9998f, ldr \reg, [\ptr], \val)
  	.endm
  
  	.macro str1 reg, ptr, val

@@ -44,7 +44,7 @@
  	.endm
  
  	.macro ldp1 reg1, reg2, ptr, val
-	ldp \reg1, \reg2, [\ptr], \val
+	KERNEL_ME_SAFE(9998f, ldp \reg1, \reg2, [\ptr], \val)
  	.endm
  
  	.macro stp1 reg1, reg2, ptr, val

@@ -64,7 +64,7 @@ SYM_FUNC_START(__arch_copy_to_user)
  9997:	cmp	dst, dstin
  	b.ne	9998f
  	// Before being absolutely sure we couldn't copy anything, try harder
-	ldrb	tmp1w, [srcin]
+KERNEL_ME_SAFE(9998f, ldrb	tmp1w, [srcin])

Alignment looks off?

Hi, Jonathan:

How about we change this in conjunction with mark's suggestion? :)

quoted

  USER(9998f, sttrb tmp1w, [dst])
  	add	dst, dst, #1
  9998:	sub	x0, end, dst			// bytes not copied

quoted

diff --git a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c
index 451ba7cbd5ad..2dc65f99d389 100644
--- a/arch/arm64/mm/fault.c
+++ b/arch/arm64/mm/fault.c

@@ -708,21 +708,32 @@ static int do_bad(unsigned long far, unsigned long esr, struct pt_regs *regs)
  	return 1; /* "fault" */
  }
  
+/*
+ * APEI claimed this as a firmware-first notification.
+ * Some processing deferred to task_work before ret_to_user().
+ */
+static bool do_apei_claim_sea(struct pt_regs *regs)
+{
+	if (user_mode(regs)) {
+		if (!apei_claim_sea(regs))

I'd keep to the the (apei_claim_sea(regs) == 0)
used in the original code. That hints to the reader that we are
interested here in an 'error' code rather than apei_claim_sea() returning
a bool.   I initially wondered why we return true when the code
fails to claim it.

Also, perhaps if you return 0 for success and an error code if not
you could just make this

	if (user_mode(regs))
		return apei_claim_sea(regs);

	if (IS_ENABLED(CONFIG_ARCH_HAS_COPY_MC)) {
		if (fixup_exception_me(regs)) {
			return apei_claim_sea(regs);
		}
	}

	return false;

or maybe even (I may have messed this up, but I think this logic
works).

	if (!user_mode(regs) && IS_ENABLED(CONFIG_ARCH_HAS_COPY_MC)) {
		if (!fixup_exception_me(regs))
			return false;
	}
	return apei_claim_sea(regs);

quoted

+			return true;
+	} else if (IS_ENABLED(CONFIG_ARCH_HAS_COPY_MC)) {
+		if (fixup_exception_me(regs) && !apei_claim_sea(regs))

Same here with using apei_claim_sea(regs) == 0 so it's obvious we
are checking for an error, not a boolean.

quoted

+			return true;
+	}
+
+	return false;
+}
+
  static int do_sea(unsigned long far, unsigned long esr, struct pt_regs *regs)
  {
  	const struct fault_info *inf;
  	unsigned long siaddr;
  
-	inf = esr_to_fault_info(esr);
-
-	if (user_mode(regs) && apei_claim_sea(regs) == 0) {
-		/*
-		 * APEI claimed this as a firmware-first notification.
-		 * Some processing deferred to task_work before ret_to_user().
-		 */
+	if (do_apei_claim_sea(regs))

It might be made sense to factor this out first, then could be reviewed
as a noop before the new stuff is added.  Still it's not much code, so doesn't
really matter.
Might be worth keeping to returning 0 for success, error code
otherwise as per apei_claim_sea(regs)

The bool returning functions in the nearby code tend to be is_xxxx
not things that succeed or not.

If you change it to return int make this
	if (do_apei_claim_sea(regs) == 0)
so it's obvious this is the no error case.

My fault, treating the return value of apei_claim_sea() as bool has
caused some confusion. Perhaps using "== 0" can reduce this confuse.

Here's the change:

	static int do_apei_claim_sea(struct pt_regs *regs)
	{
		if (!user_mode(regs) && IS_ENABLED(CONFIG_ARCH_HAS_COPY_MC)) {
			if (!fixup_exception_me(regs)))
				return -ENOENT;
		}
		
		return apei_claim_sea(regs);
	}
	
	static int do_sea(...)
	{
		[...]
		if (do_apei_claim_sea(regs) == 0)
			return 0;
		[...]
	}

I'll modify it later with the comments of mark.

Thanks,
Tong.

quoted

  		return 0;
-	}
  
+	inf = esr_to_fault_info(esr);
  	if (esr & ESR_ELx_FnV) {
  		siaddr = 0;
  	} else {

.