Skip to contentSkip to conversation

riscv: Fix debug instruction check and support trap-based WARN()

PATCH In review Vincent Chen 2 participants 6 messages Feb 28, 2019 – Mar 5, 2019 RISC-V

lore
Share & tools
lore
#0 lore raw

cover riscv: Fix debug instruction check and support trap-based WARN()

The handler for the debug exception will call is_valid_bugaddr(bugaddr) to check if the instruction in bugaddr is a real debug instruction. However, the expected instruction, ebreak, is possibly translated to c.ebreak by assmebler when C extension is supported. This patchset will add c.ebreak into the check mechanism. In addition, BUG() is currently unable to work in the kernel module due to an inappropriated condition in is_valid_bugaddr(). This issue will be fixed in this patchset. Finally, this patchset enables WARN() related functions to trap the code to help developers debug it.

Vincent Chen (3):
riscv: Add the support for c.ebreak check in is_valid_bugaddr()
riscv: Support BUG() in kernel module
riscv: Make WARN() related functions able to trigger a trap exception

_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv

#1 lore raw

PATCH 1/3 riscv: Add the support for c.ebreak check in is_valid_bugaddr()

The is_valid_bugaddr() function compares the instruction pointed to by $sepc with macro __BUG_INSN to check whether the current trap exception is caused by an "ebreak" instruction. Hence the macro __BUG_INSN is defined as "ebreak". However, this check flow is possibly erroneous because the expected trap instruction possibly be translated to "c.ebreak" by assembler if C extension is supported. Therefore, it requires a mechanism to distinguish the length of the instruction in $spec and compare it to the correct __BUG_INSN.

By the way, I make the kernel go to die() like BUG() when the trap type is BUG_TRAP_TYPE_WARN because currently the WARN() does not trigger any trap exception.

---
 arch/riscv/include/asm/bug.h |    7 ++++++-
 arch/riscv/kernel/traps.c    |    8 +++++---
 2 files changed, 11 insertions(+), 4 deletions(-)
arch/riscv/include/asm/bug.h +6 -1
index bfc7f09..1cab7f4 100644
@@ -21,7 +21,12 @@
 #include <asm/asm.h>
 
 #ifdef CONFIG_GENERIC_BUG
-#define __BUG_INSN	_AC(0x00100073, UL) /* ebreak */
+#define __INSN_LENGTH_MASK  _UL(0x3)
+#define __INSN_LENGTH_32    _UL(0x3)
+#define __COMPRESSED_INSN_MASK	_UL(0xffff)
+
+#define __BUG_INSN_32	_UL(0x00100073) /* ebreak */
+#define __BUG_INSN_16	_UL(0x9002) /* c.ebreak */
 
 #ifndef __ASSEMBLY__
 typedef u32 bug_insn_t;
arch/riscv/kernel/traps.c +5 -3
index 24a9333..deae0e5 100644
@@ -129,8 +129,7 @@ asmlinkage void do_trap_break(struct pt_regs *regs)
 		case BUG_TRAP_TYPE_NONE:
 			break;
 		case BUG_TRAP_TYPE_WARN:
-			regs->sepc += sizeof(bug_insn_t);
-			return;
+			die(regs, "Kernel BUG. Kernel got an unexpected WARN trapped by ebreak");
 		case BUG_TRAP_TYPE_BUG:
 			die(regs, "Kernel BUG");
 		}
@@ -149,7 +148,10 @@ int is_valid_bugaddr(unsigned long pc)
 		return 0;
 	if (probe_kernel_address((bug_insn_t *)pc, insn))
 		return 0;
-	return (insn == __BUG_INSN);
+	if ((insn & __INSN_LENGTH_MASK) == __INSN_LENGTH_32)
+		return insn == __BUG_INSN_32;
+	else
+		return (insn & __COMPRESSED_INSN_MASK) == __BUG_INSN_16;
 }
 #endif /* CONFIG_GENERIC_BUG */
 
Signature · 7 lines
1.7.1


_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
#2 lore raw

PATCH 2/3 riscv: Support BUG() in kernel module

In reply to Vincent Chen

The kernel module is loaded into vmalloc region which is located below to the PAGE_OFFSET. Hence the condition, pc < PAGE_OFFSET, in the is_valid_bugaddr() will filter out all trap exceptions triggered by kernel module. To support BUG() in kernel module, the condition is changed to pc < VMALLOC_START.

---
 arch/riscv/kernel/traps.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
arch/riscv/kernel/traps.c +1 -1
index deae0e5..dee0e5e 100644
@@ -144,7 +144,7 @@ int is_valid_bugaddr(unsigned long pc)
 {
 	bug_insn_t insn;
 
-	if (pc < PAGE_OFFSET)
+	if (pc < VMALLOC_START)
 		return 0;
 	if (probe_kernel_address((bug_insn_t *)pc, insn))
 		return 0;
Signature · 7 lines
1.7.1


_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
#3 lore raw

PATCH 3/3 riscv: Make WARN() related functions able to trigger a trap exception

In reply to Vincent Chen

This can help developers to analyze the cause of WARN() because the control will be transferred to debugging environment if the debugger is connected.

---
 arch/riscv/include/asm/bug.h |   27 ++++++++++++++++++---------
 arch/riscv/kernel/traps.c    |   19 ++++++++++++++++---
 2 files changed, 34 insertions(+), 12 deletions(-)
arch/riscv/include/asm/bug.h +18 -9
index 1cab7f4..b2aa88f 100644
@@ -43,38 +43,47 @@
 #define __BUG_ENTRY			\
 	__BUG_ENTRY_ADDR "\n\t"		\
 	__BUG_ENTRY_FILE "\n\t"		\
-	RISCV_SHORT " %1"
+	RISCV_SHORT " %1  \n\t"		\
+	RISCV_SHORT " %2"
 #else
 #define __BUG_ENTRY			\
-	__BUG_ENTRY_ADDR
+	__BUG_ENTRY_ADDR "\n\t"		\
+	RISCV_SHORT " %2"
 #endif
 
-#define BUG()							\
+#define __BUG_FLAGS(flags)					\
 do {								\
 	__asm__ __volatile__ (					\
 		"1:\n\t"					\
 			"ebreak\n"				\
-			".pushsection __bug_table,\"a\"\n\t"	\
+			".pushsection __bug_table,\"aw\"\n\t"	\
 		"2:\n\t"					\
 			__BUG_ENTRY "\n\t"			\
-			".org 2b + %2\n\t"			\
+			".org 2b + %3\n\t"                      \
 			".popsection"				\
 		:						\
 		: "i" (__FILE__), "i" (__LINE__),		\
-		  "i" (sizeof(struct bug_entry)));		\
-	unreachable();						\
+		  "i" (flags),					\
+		  "i" (sizeof(struct bug_entry)));              \
 } while (0)
+
 #endif /* !__ASSEMBLY__ */
 #else /* CONFIG_GENERIC_BUG */
 #ifndef __ASSEMBLY__
-#define BUG()							\
+#define __BUG_FLAGS(flags)					\
 do {								\
 	__asm__ __volatile__ ("ebreak\n");			\
-	unreachable();						\
 } while (0)
 #endif /* !__ASSEMBLY__ */
 #endif /* CONFIG_GENERIC_BUG */
 
+#define BUG() do {						\
+	__BUG_FLAGS(0);						\
+	unreachable();						\
+} while (0)
+
+#define __WARN_FLAGS(flags) __BUG_FLAGS(BUGFLAG_WARNING|(flags))
+
 #define HAVE_ARCH_BUG
 
 #include <asm-generic/bug.h>
arch/riscv/kernel/traps.c +16 -3
index dee0e5e..023208b 100644
@@ -118,6 +118,17 @@ asmlinkage void name(struct pt_regs *regs)				\
 DO_ERROR_INFO(do_trap_ecall_m,
 	SIGILL, ILL_ILLTRP, "environment call from M-mode");
 
+#ifdef CONFIG_GENERIC_BUG
+static inline unsigned long get_break_insn_length(unsigned long pc)
+{
+	bug_insn_t insn;
+
+	if (probe_kernel_address((bug_insn_t *)pc, insn))
+		return 0;
+	return ((insn & __INSN_LENGTH_MASK) == __INSN_LENGTH_32) ? 4UL : 2UL;
+}
+#endif /* CONFIG_GENERIC_BUG */
+
 asmlinkage void do_trap_break(struct pt_regs *regs)
 {
 #ifdef CONFIG_GENERIC_BUG
@@ -129,7 +140,8 @@ asmlinkage void do_trap_break(struct pt_regs *regs)
 		case BUG_TRAP_TYPE_NONE:
 			break;
 		case BUG_TRAP_TYPE_WARN:
-			die(regs, "Kernel BUG. Kernel got an unexpected WARN trapped by ebreak");
+			regs->sepc += get_break_insn_length(regs->sepc);
+			break;
 		case BUG_TRAP_TYPE_BUG:
 			die(regs, "Kernel BUG");
 		}
@@ -149,12 +161,13 @@ int is_valid_bugaddr(unsigned long pc)
 	if (probe_kernel_address((bug_insn_t *)pc, insn))
 		return 0;
 	if ((insn & __INSN_LENGTH_MASK) == __INSN_LENGTH_32)
-		return insn == __BUG_INSN_32;
+		return (insn == __BUG_INSN_32);
 	else
-		return (insn & __COMPRESSED_INSN_MASK) == __BUG_INSN_16;
+		return ((insn & __COMPRESSED_INSN_MASK) == __BUG_INSN_16);
 }
 #endif /* CONFIG_GENERIC_BUG */
 
+
 void __init trap_init(void)
 {
 	/*
Signature · 7 lines
1.7.1


_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
#4 lore raw
In reply to Vincent Chen
Quoted 13 lines · On Thu, 28 Feb 2019 02:31:28 PST (-0800), [email protected] wrote:

The handler for the debug exception will call is_valid_bugaddr(bugaddr) to check if the instruction in bugaddr is a real debug instruction. However, the expected instruction, ebreak, is possibly translated to c.ebreak by assmebler when C extension is supported. This patchset will add c.ebreak into the check mechanism. In addition, BUG() is currently unable to work in the kernel module due to an inappropriated condition in is_valid_bugaddr(). This issue will be fixed in this patchset. Finally, this patchset enables WARN() related functions to trap the code to help developers debug it.

Quoted 4 lines

Vincent Chen (3):
riscv: Add the support for c.ebreak check in is_valid_bugaddr()
riscv: Support BUG() in kernel module
riscv: Make WARN() related functions able to trigger a trap exception

I'm finding this patch set a bit hard to follow, and I think it has more diff than is necessary. For example, the first patch introduces a new die() only to have it removed by the third patch. There's also some unnecessary non-functional diff, like

@@ -149,12 +161,13 @@ int is_valid_bugaddr(unsigned long pc)

           if (probe_kernel_address((bug_insn_t *)pc, insn))
                   return 0;
           if ((insn & __INSN_LENGTH_MASK) == __INSN_LENGTH_32)
   -               return insn == __BUG_INSN_32;
   +               return (insn == __BUG_INSN_32);
           else
   -               return (insn & __COMPRESSED_INSN_MASK) == __BUG_INSN_16;
   +               return ((insn & __COMPRESSED_INSN_MASK) == __BUG_INSN_16);
    }
    #endif /* CONFIG_GENERIC_BUG */
   +
    void __init trap_init(void)
    {
           /*

I like the idea of the patch set, though. Do you have time to clean it up and submit a v2?

#5 lore raw
Quoted 42 lines · Palmer Dabbelt
Quoted 13 lines · On Thu, 28 Feb 2019 02:31:28 PST (-0800), [email protected] wrote:

The handler for the debug exception will call is_valid_bugaddr(bugaddr) to check if the instruction in bugaddr is a real debug instruction. However, the expected instruction, ebreak, is possibly translated to c.ebreak by assmebler when C extension is supported. This patchset will add c.ebreak into the check mechanism. In addition, BUG() is currently unable to work in the kernel module due to an inappropriated condition in is_valid_bugaddr(). This issue will be fixed in this patchset. Finally, this patchset enables WARN() related functions to trap the code to help developers debug it.

Quoted 4 lines

Vincent Chen (3):
riscv: Add the support for c.ebreak check in is_valid_bugaddr()
riscv: Support BUG() in kernel module
riscv: Make WARN() related functions able to trigger a trap exception

I'm finding this patch set a bit hard to follow, and I think it has more diff than is necessary. For example, the first patch introduces a new die() only to have it removed by the third patch. There's also some unnecessary non-functional diff, like

    @@ -149,12 +161,13 @@ int is_valid_bugaddr(unsigned long pc)
            if (probe_kernel_address((bug_insn_t *)pc, insn))
                    return 0;
            if ((insn & __INSN_LENGTH_MASK) == __INSN_LENGTH_32)
    -               return insn == __BUG_INSN_32;
    +               return (insn == __BUG_INSN_32);
            else
    -               return (insn & __COMPRESSED_INSN_MASK) == __BUG_INSN_16;
    +               return ((insn & __COMPRESSED_INSN_MASK) == __BUG_INSN_16);
     }
     #endif /* CONFIG_GENERIC_BUG */
    
    +
     void __init trap_init(void)
     {
            /*
Quoted 2 lines

I like the idea of the patch set, though. Do you have time to clean it up and submit a v2?

OK, I will improve it and submit a v2 patch.

_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv