kbuild: keep symbols for symbol_get() even with CONFIG_TRIM_UNUSED_KSYMS

Linus observed that the symbol_request(utf8_data_table) call fails when
CONFIG_UNICODE=y and CONFIG_TRIM_UNUSED_KSYMS=y.

symbol_get() relies on the symbol data being present in the ksymtab for
symbol lookups. However, EXPORT_SYMBOL_GPL(utf8_data_table) is dropped
due to CONFIG_TRIM_UNUSED_KSYMS, as no module references it in this case.

Probably, this has been broken since commit dbacb0ef670d ("kconfig option
for TRIM_UNUSED_KSYMS").

This commit addresses the issue by leveraging modpost. Symbol names
passed to symbol_get() are recorded in the special .no_trim_symbol
section, which is then parsed by modpost to forcibly keep such symbols.
The .no_trim_symbol section is discarded by the linker scripts, so there
is no impact on the size of the final vmlinux or modules.

This commit cannot resolve the issue for direct calls to __symbol_get()
because the symbol name is not known at compile-time.

Although symbol_get() may eventually be deprecated, this workaround
should be good enough meanwhile.

Reported-by: Linus Torvalds <[e-mail]>
Suggested-by: Linus Torvalds <[e-mail]>
Signed-off-by: Masahiro Yamada <[e-mail]>

include/asm-generic/vmlinux.lds.h

···
 	*(.discard)							\
 	*(.discard.*)							\
 	*(.export_symbol)						\
+	*(.no_trim_symbol)						\
 	*(.modinfo)							\
 	/* ld.bfd warns about .gnu.version* even when not emitted */	\
 	*(.gnu.version*)						\

include/linux/module.h

···
 /* Get/put a kernel symbol (calls must be symmetric) */
 void *__symbol_get(const char *symbol);
 void *__symbol_get_gpl(const char *symbol);
-#define symbol_get(x) ((typeof(&x))(__symbol_get(__stringify(x))))
+#define symbol_get(x)	({ \
+	static const char __notrim[] \
+		__used __section(".no_trim_symbol") = __stringify(x); \
+	(typeof(&x))(__symbol_get(__stringify(x))); })
 
 /* modules using other modules: kdb wants to see this. */
 struct module_use {

scripts/mod/modpost.c

···
 			info->modinfo_len = sechdrs[i].sh_size;
 		} else if (!strcmp(secname, ".export_symbol")) {
 			info->export_symbol_secndx = i;
+		} else if (!strcmp(secname, ".no_trim_symbol")) {
+			info->no_trim_symbol = (void *)hdr + sechdrs[i].sh_offset;
+			info->no_trim_symbol_len = sechdrs[i].sh_size;
 		}
 
 		if (sechdrs[i].sh_type == SHT_SYMTAB) {
···
 	/* strip trailing .o */
 	mod = new_module(modname, strlen(modname) - strlen(".o"));
 
+	/* save .no_trim_symbol section for later use */
+	if (info.no_trim_symbol_len) {
+		mod->no_trim_symbol = xmalloc(info.no_trim_symbol_len);
+		memcpy(mod->no_trim_symbol, info.no_trim_symbol,
+		       info.no_trim_symbol_len);
+		mod->no_trim_symbol_len = info.no_trim_symbol_len;
+	}
+
 	if (!mod->is_vmlinux) {
 		license = get_modinfo(&info, "license");
 		if (!license)
···
 	}
 
 	free(buf);
+}
+
+/*
+ * Keep symbols recorded in the .no_trim_symbol section. This is necessary to
+ * prevent CONFIG_TRIM_UNUSED_KSYMS from dropping EXPORT_SYMBOL because
+ * symbol_get() relies on the symbol being present in the ksymtab for lookups.
+ */
+static void keep_no_trim_symbols(struct module *mod)
+{
+	unsigned long size = mod->no_trim_symbol_len;
+
+	for (char *s = mod->no_trim_symbol; s; s = next_string(s , &size)) {
+		struct symbol *sym;
+
+		/*
+		 * If find_symbol() returns NULL, this symbol is not provided
+		 * by any module, and symbol_get() will fail.
+		 */
+		sym = find_symbol(s);
+		if (sym)
+			sym->used = true;
+	}
 }
 
 static void check_modname_len(struct module *mod)
···
 		read_symbols_from_files(files_source);
 
 	list_for_each_entry(mod, &modules, list) {
+		keep_no_trim_symbols(mod);
+
 		if (mod->dump_file || mod->is_vmlinux)
 			continue;
 

scripts/mod/modpost.h

···
  *
  * @dump_file: path to the .symvers file if loaded from a file
  * @aliases: list head for module_aliases
+ * @no_trim_symbol: .no_trim_symbol section data
+ * @no_trim_symbol_len: length of the .no_trim_symbol section
  */
 struct module {
 	struct list_head list;
···
 	// Actual imported namespaces
 	struct list_head imported_namespaces;
 	struct list_head aliases;
+	char *no_trim_symbol;
+	unsigned int no_trim_symbol_len;
 	char name[];
 };
 
···
 	char         *strtab;
 	char	     *modinfo;
 	unsigned int modinfo_len;
+	char         *no_trim_symbol;
+	unsigned int no_trim_symbol_len;
 
 	/* support for 32bit section numbers */
 

scripts/module.lds.S

···
 		*(.discard)
 		*(.discard.*)
 		*(.export_symbol)
+		*(.no_trim_symbol)
 	}
 
 	__ksymtab		0 : ALIGN(8) { *(SORT(___ksymtab+*)) }