Support stack traces for running threads on kdumps

Author: sdimitro

libdrgn/kdump.c

@@ -139,3 +139,61 @@ struct drgn_error *drgn_program_set_kdump(struct drgn_program *prog)
 	kdump_free(ctx);
 	return err;
 }
+
+struct drgn_error *drgn_program_cache_prstatus_kdump(struct drgn_program *prog)
+{
+	struct drgn_error *err;
+	kdump_num_t ncpus, i;
+	kdump_status ks;
+
+	ks = kdump_get_number_attr(prog->kdump_ctx, "cpu.number", &ncpus);
+	if (ks != KDUMP_OK) {
+		return drgn_error_format(DRGN_ERROR_OTHER,
+					 "kdump_get_number_attr(cpu.number): %s",
+					 kdump_get_err(prog->kdump_ctx));
+	}
+
+	/*
+	 * Note that in the following loop we never call kdump_attr_unref() on
+	 * prstatus_ref, nor kdump_blob_unpin() on the prstatus blob that we get
+	 * from libkdumpfile. Since drgn is completely read-only as a consumer
+	 * of that library, we "leak" both the attribute reference and blob pin
+	 * until kdump_free() is called which will clean up everything for us.
+	 */
+	for (i = 0; i < ncpus; i++) {
+		/* Enough for the longest possible PRSTATUS attribute name. */
+		char attr_name[64];
+		kdump_attr_ref_t prstatus_ref;
+		kdump_attr_t prstatus_attr;
+		void *prstatus_data;
+		size_t prstatus_size;
+
+		snprintf(attr_name, sizeof(attr_name),
+			 "cpu.%" PRIuFAST64 ".PRSTATUS", i);
+		ks = kdump_attr_ref(prog->kdump_ctx, attr_name, &prstatus_ref);
+		if (ks != KDUMP_OK) {
+			return drgn_error_format(DRGN_ERROR_OTHER,
+						 "kdump_attr_ref(%s): %s",
+						 attr_name,
+						 kdump_get_err(prog->kdump_ctx));
+		}
+
+		ks = kdump_attr_ref_get(prog->kdump_ctx, &prstatus_ref,
+					&prstatus_attr);
+		if (ks != KDUMP_OK) {
+			return drgn_error_format(DRGN_ERROR_OTHER,
+						 "kdump_attr_ref_get(%s): %s",
+						 attr_name,
+						 kdump_get_err(prog->kdump_ctx));
+		}
+
+		prstatus_data = kdump_blob_pin(prstatus_attr.val.blob);
+		prstatus_size = kdump_blob_size(prstatus_attr.val.blob);
+		err = drgn_program_cache_prstatus_entry(prog,
+							prstatus_data,
+							prstatus_size);
+		if (err)
+			return err;
+	}
+	return NULL;
+}

libdrgn/linux_kernel.h

@@ -34,13 +34,14 @@ linux_kernel_report_debug_info(struct drgn_program *prog,
 #define KDUMP_SIG_LEN (sizeof(KDUMP_SIGNATURE) - 1)
 
 #ifdef WITH_LIBKDUMPFILE
+struct drgn_error *drgn_program_cache_prstatus_kdump(struct drgn_program *prog);
 struct drgn_error *drgn_program_set_kdump(struct drgn_program *prog);
 #else
 static inline struct drgn_error *
 drgn_program_set_kdump(struct drgn_program *prog)
 {
         return drgn_error_create(DRGN_ERROR_INVALID_ARGUMENT,
-	                         "drgn was built without libkdumpfile support");
+				 "drgn was built without libkdumpfile support");
 }
 #endif
 

libdrgn/program.c

@@ -684,16 +684,45 @@ drgn_program_load_debug_info(struct drgn_program *prog, const char **paths,
 	return err;
 }
 
-static struct drgn_error *drgn_program_cache_prstatus(struct drgn_program *prog)
+struct drgn_error *drgn_program_cache_prstatus_entry(struct drgn_program *prog,
+						     char *data, size_t size)
 {
-	size_t phnum, i;
+	struct drgn_prstatus_map_entry entry;
 	size_t pr_pid_offset;
+	uint32_t pr_pid;
 	bool bswap;
 
 	pr_pid_offset = drgn_program_is_64_bit(prog) ? 32 : 24;
 	bswap = (drgn_program_is_little_endian(prog) !=
 		 (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__));
 
+	if (size < pr_pid_offset + sizeof(pr_pid))
+		return NULL;
+
+	memcpy(&pr_pid, data + pr_pid_offset, sizeof(pr_pid));
+	if (bswap)
+		pr_pid = bswap_32(pr_pid);
+	if (!pr_pid)
+		return NULL;
+
+	entry.key = pr_pid;
+	entry.value.str = data;
+	entry.value.len = size;
+	if (drgn_prstatus_map_insert(&prog->prstatus_cache, &entry,
+				     NULL) == -1) {
+		return &drgn_enomem;
+	}
+	return NULL;
+}
+
+static struct drgn_error *drgn_program_cache_prstatus(struct drgn_program *prog)
+{
+	size_t phnum, i;
+
+#ifdef WITH_LIBKDUMPFILE
+	if (prog->kdump_ctx)
+		return drgn_program_cache_prstatus_kdump(prog);
+#endif
 	if (elf_getphdrnum(prog->core, &phnum) != 0)
 		return drgn_error_libelf();
 	for (i = 0; i < phnum; i++) {
@@ -720,28 +749,18 @@ static struct drgn_error *drgn_program_cache_prstatus(struct drgn_program *prog)
 		       (offset = gelf_getnote(data, offset, &nhdr, &name_offset,
 					      &desc_offset))) {
 			const char *name;
-			uint32_t pr_pid;
-			struct drgn_prstatus_map_entry entry;
+			struct drgn_error *err;
 
 			name = (char *)data->d_buf + name_offset;
 			if (strncmp(name, "CORE", nhdr.n_namesz) != 0 ||
-			    nhdr.n_type != NT_PRSTATUS ||
-			    nhdr.n_descsz < pr_pid_offset + sizeof(pr_pid))
-				continue;
-			memcpy(&pr_pid,
-			       (char *)data->d_buf + desc_offset + pr_pid_offset,
-			       sizeof(pr_pid));
-			if (bswap)
-				pr_pid = bswap_32(pr_pid);
-			if (!pr_pid)
+			    nhdr.n_type != NT_PRSTATUS)
 				continue;
 
-			entry.key = pr_pid;
-			entry.value.str = (char *)data->d_buf + desc_offset;
-			entry.value.len = nhdr.n_descsz;
-			if (drgn_prstatus_map_insert(&prog->prstatus_cache,
-						     &entry, NULL) == -1)
-				return &drgn_enomem;
+			err = drgn_program_cache_prstatus_entry(prog,
+								(char *)data->d_buf + desc_offset,
+								nhdr.n_descsz);
+			if (err)
+				return err;
 		}
 	}
 	return NULL;

libdrgn/program.h

@@ -157,6 +157,15 @@ struct drgn_error *drgn_program_get_dwfl(struct drgn_program *prog, Dwfl **ret);
 struct drgn_error *drgn_program_find_prstatus(struct drgn_program *prog,
 					      uint32_t tid, struct string *ret);
 
+/**
+ * Cache the @c NT_PRSTATUS note provided by @p data in @p prog.
+ *
+ * @param[in] data The pointer to the note data.
+ * @param[in] size Size of data in note.
+ */
+struct drgn_error *drgn_program_cache_prstatus_entry(struct drgn_program *prog,
+                                                     char *data, size_t size);
+
 /*
  * Like @ref drgn_program_find_symbol_by_address(), but @p ret is already
  * allocated, we may already know the module, and doesn't return a @ref

libdrgn/stack_trace.c

@@ -398,8 +398,12 @@ static bool drgn_thread_set_initial_registers(Dwfl_Thread *thread,
 		}
 	}
 
-	/* Then, try the core dump. */
+	/* Then, try the core dump (and/or kdump if supported). */
+#ifdef WITH_LIBKDUMPFILE
+	if (prog->core || prog->kdump_ctx) {
+#else
 	if (prog->core) {
+#endif
 		uint32_t tid;
 		struct string prstatus;