Skip to contentSkip to conversation

kgdb: enable arch to handle more query packets

PATCH Superseded Vincent Chen 2 participants 3 messages Mar 3, 2020 – Mar 19, 2020

Subsystems (2)

v1 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH 3/5 kgdb: enable arch to handle more query packets

KGDB only supports parts of GDB query packets. Add kgdb_arch_cmd_query() hook function to enable arch to handle more query packets such as the qSupported packet.

---
 include/linux/kgdb.h   | 10 ++++++++++
 kernel/debug/gdbstub.c |  6 ++++++
 2 files changed, 16 insertions(+)
include/linux/kgdb.h +10 -0
index b072aeb1fd78..bbb35557f76d 100644
@@ -177,6 +177,16 @@ kgdb_arch_handle_exception(int vector, int signo, int err_code,
 			   struct pt_regs *regs);
 
 /**
+ *	kgdb_arch_handle_exception - Handle architecture specific query packet.
+ *	@remcom_in_buffer: The buffer of the packet we have read.
+ *	@remcom_out_buffer: The buffer of %BUFMAX bytes to write a packet into.
+ */
+
+extern void
+kgdb_arch_cmd_query(char *remcom_in_buffer,
+		    char *remcom_out_buffer);
+
+/**
  *	kgdb_call_nmi_hook - Call kgdb_nmicallback() on the current CPU
  *	@ignored: This parameter is only here to match the prototype.
  *
kernel/debug/gdbstub.c +6 -0
index 4b280fc7dd67..2b2e7b99edcc 100644
@@ -694,6 +694,9 @@ static int gdb_cmd_reboot(struct kgdb_state *ks)
 	return 0;
 }
 
+void __weak kgdb_arch_cmd_query(char *remcom_in_buffer,
+				char *remcom_out_buffer);
+
 /* Handle the 'q' query packets */
 static void gdb_cmd_query(struct kgdb_state *ks)
 {
@@ -792,6 +795,9 @@ static void gdb_cmd_query(struct kgdb_state *ks)
 		}
 		break;
 #endif
+	default:
+		kgdb_arch_cmd_query(remcom_in_buffer, remcom_out_buffer);
+
 	}
 }
 
Signature · 1 line
2.7.4
maintainer
#1 lore raw
Quoted 46 lines · On Tue, 03 Mar 2020 00:47:43 PST (-0800), [email protected] wrote:

KGDB only supports parts of GDB query packets. Add kgdb_arch_cmd_query() hook function to enable arch to handle more query packets such as the qSupported packet.

---
 include/linux/kgdb.h   | 10 ++++++++++
 kernel/debug/gdbstub.c |  6 ++++++
 2 files changed, 16 insertions(+)
include/linux/kgdb.h +10 -0
index b072aeb1fd78..bbb35557f76d 100644
@@ -177,6 +177,16 @@ kgdb_arch_handle_exception(int vector, int signo, int err_code,
 			   struct pt_regs *regs);
 
 /**
+ *	kgdb_arch_handle_exception - Handle architecture specific query packet.
+ *	@remcom_in_buffer: The buffer of the packet we have read.
+ *	@remcom_out_buffer: The buffer of %BUFMAX bytes to write a packet into.
+ */
+
+extern void
+kgdb_arch_cmd_query(char *remcom_in_buffer,
+		    char *remcom_out_buffer);
+
+/**
  *	kgdb_call_nmi_hook - Call kgdb_nmicallback() on the current CPU
  *	@ignored: This parameter is only here to match the prototype.
  *
kernel/debug/gdbstub.c +3 -0
index 4b280fc7dd67..2b2e7b99edcc 100644
@@ -694,6 +694,9 @@ static int gdb_cmd_reboot(struct kgdb_state *ks)
 	return 0;
 }
 
+void __weak kgdb_arch_cmd_query(char *remcom_in_buffer,
+				char *remcom_out_buffer);
+
 /* Handle the 'q' query packets */
 static void gdb_cmd_query(struct kgdb_state *ks)
 {

kernel/debug/gdbstub.c static void gdb_cmd_query(struct kgdb_state *ks) · line 802

Quoted 8 lines
kernel/debug/gdbstub.c +3 -0
@@ -792,6 +795,9 @@ static void gdb_cmd_query(struct kgdb_state *ks)
 		}
 		break;
 #endif
+	default:
+		kgdb_arch_cmd_query(remcom_in_buffer, remcom_out_buffer);
+
 	}
 }

Won't this blow up on architectures that don't implement kgdb_arch_cmd_query()? IIRC undefined weak functions have the absolute address 0. It's probably better to avoid the __weak entirely and introduce something along the lines of CONFIG_ARCH_HAS_CMD_QUERY.

One better: I don't think qSupported is a good example of an arch-specific packet -- it's probably cleaner to have the arch indicate which extra features it supports (presumably via Kconfig) rather than just punting on the whole packet. The XML stuff is a good example: rather than putting all that in the RISC-V port, we could just have something like CONFIG_ACRH_SUPPORTS_GDB_XML which indicates XML descriptions are supported and expects a function to fill out the XML response.

#2 lore raw
Quoted 51 lines · Palmer Dabbelt
Quoted 49 lines · On Tue, 03 Mar 2020 00:47:43 PST (-0800), [email protected] wrote:

KGDB only supports parts of GDB query packets. Add kgdb_arch_cmd_query() hook function to enable arch to handle more query packets such as the qSupported packet.

---
 include/linux/kgdb.h   | 10 ++++++++++
 kernel/debug/gdbstub.c |  6 ++++++
 2 files changed, 16 insertions(+)
include/linux/kgdb.h +10 -0
index b072aeb1fd78..bbb35557f76d 100644
@@ -177,6 +177,16 @@ kgdb_arch_handle_exception(int vector, int signo, int err_code,
                         struct pt_regs *regs);
 
 /**
+ *   kgdb_arch_handle_exception - Handle architecture specific query packet.
+ *   @remcom_in_buffer: The buffer of the packet we have read.
+ *   @remcom_out_buffer: The buffer of %BUFMAX bytes to write a packet into.
+ */
+
+extern void
+kgdb_arch_cmd_query(char *remcom_in_buffer,
+                 char *remcom_out_buffer);
+
+/**
  *   kgdb_call_nmi_hook - Call kgdb_nmicallback() on the current CPU
  *   @ignored: This parameter is only here to match the prototype.
  *
kernel/debug/gdbstub.c +3 -0
index 4b280fc7dd67..2b2e7b99edcc 100644
@@ -694,6 +694,9 @@ static int gdb_cmd_reboot(struct kgdb_state *ks)
      return 0;
 }
 
+void __weak kgdb_arch_cmd_query(char *remcom_in_buffer,
+                             char *remcom_out_buffer);
+
 /* Handle the 'q' query packets */
 static void gdb_cmd_query(struct kgdb_state *ks)
 {
@@ -792,6 +795,9 @@ static void gdb_cmd_query(struct kgdb_state *ks)
              }
              break;
 #endif
Quoted 8 lines

kernel/debug/gdbstub.c static void gdb_cmd_query(struct kgdb_state *ks) · line 802

Quoted 5 lines
kernel/debug/gdbstub.c +3 -0
+     default:
+             kgdb_arch_cmd_query(remcom_in_buffer, remcom_out_buffer);
+
      }
 }

Won't this blow up on architectures that don't implement kgdb_arch_cmd_query()? IIRC undefined weak functions have the absolute address 0. It's probably

Yes, you are right.
I planed to define an empty week function but I forgot to add braces after the function prototype.
This is my mistake.

Quoted 10 lines

better to avoid the __weak entirely and introduce something along the lines of CONFIG_ARCH_HAS_CMD_QUERY.

One better: I don't think qSupported is a good example of an arch-specific packet -- it's probably cleaner to have the arch indicate which extra features it supports (presumably via Kconfig) rather than just punting on the whole packet. The XML stuff is a good example: rather than putting all that in the RISC-V port, we could just have something like CONFIG_ACRH_SUPPORTS_GDB_XML which indicates XML descriptions are supported and expects a function to fill out the XML response.

I agree with you. I will follow your suggestion to modify it. Thank you.