Re: [PATCH 2/6] module: add support for symbol namespaces.

From: Jessica Yu
Date: Wed Jul 25 2018 - 11:55:18 EST


+++ Martijn Coenen [24/07/18 09:56 +0200]:
I did find an issue with my approach:

On Mon, Jul 16, 2018 at 2:21 PM, Martijn Coenen <maco@xxxxxxxxxxx> wrote:
The ELF symbols are renamed to include the namespace with an asm label;
for example, symbol 'usb_stor_suspend' in namespace USB_STORAGE becomes
'usb_stor_suspend.USB_STORAGE'. This allows modpost to do namespace
checking, without having to go through all the effort of parsing ELF and
reloction records just to get to the struct kernel_symbols.

depmod doesn't like this: if namespaced symbols are built-in to the
kernel, they will appear as 'symbol.NS' in the symbol table, whereas
modules using those symbols just depend on 'symbol'. This will cause
depmod to warn about unknown symbols. I didn't find this previously
because all the exports/imports I tested were done from modules
themselves.

One way to deal with it is to change depmod, but it looks like it
hasn't been changed in ages, and it would also introduce a dependency
on userspaces to update it to avoid these warnings. So, we'd have to
encode the symbol namespace information in another way for modpost to
use. I could just write a separate post-processing tool (much like
genksyms), or perhaps encode the information in a discardable section.
Any other suggestions welcome.

This seems to be because depmod uses System.map as a source for kernel
symbols and scans for only the ones prefixed with "__ksymtab" to find
the exported ones - and those happen to use the '.' to mark the
namespace it belongs to. It strips that prefix and uses the remainder
as the actual symbol name. To fix that it'd have to strip the
namespace suffix in the symbol name as well.

Just scanning the depmod source code, it seems really trivial to just
have depmod accommodate the new symbol name format. Adding Lucas (kmod
maintainer) to CC.

Thanks,

Jessica



On x86_64 I saw no difference in binary size (compression), but at
runtime this will require a word of memory per export to hold the
namespace. An alternative could be to store namespaced symbols in their
own section and use a separate 'struct namespaced_kernel_symbol' for
that section, at the cost of making the module loader more complex.

Signed-off-by: Martijn Coenen <maco@xxxxxxxxxxx>
---
include/asm-generic/export.h | 2 +-
include/linux/export.h | 83 +++++++++++++++++++++++++++---------
include/linux/module.h | 13 ++++++
kernel/module.c | 79 ++++++++++++++++++++++++++++++++++
4 files changed, 156 insertions(+), 21 deletions(-)

diff --git a/include/asm-generic/export.h b/include/asm-generic/export.h
index 68efb950a918..4c3d1afb702f 100644
--- a/include/asm-generic/export.h
+++ b/include/asm-generic/export.h
@@ -29,7 +29,7 @@
.section ___ksymtab\sec+\name,"a"
.balign KSYM_ALIGN
__ksymtab_\name:
- __put \val, __kstrtab_\name
+ __put \val, __kstrtab_\name, 0
.previous
.section __ksymtab_strings,"a"
__kstrtab_\name:
diff --git a/include/linux/export.h b/include/linux/export.h
index ad6b8e697b27..9f6e70eeb85f 100644
--- a/include/linux/export.h
+++ b/include/linux/export.h
@@ -22,6 +22,11 @@ struct kernel_symbol
{
unsigned long value;
const char *name;
+ const char *namespace;
+};
+
+struct namespace_import {
+ const char *namespace;
};

#ifdef MODULE
@@ -54,18 +59,28 @@ extern struct module __this_module;
#define __CRC_SYMBOL(sym, sec)
#endif

+#define NS_SEPARATOR "."
+
+#define MODULE_IMPORT_NS(ns) \
+ static const struct namespace_import __knsimport_##ns \
+ asm("__knsimport_" #ns) \
+ __attribute__((section("__knsimport"), used)) \
+ = { #ns }
+
/* For every exported symbol, place a struct in the __ksymtab section */
-#define ___EXPORT_SYMBOL(sym, sec) \
+#define ___EXPORT_SYMBOL(sym, sec, ns, nspost, nspost2) \
extern typeof(sym) sym; \
__CRC_SYMBOL(sym, sec) \
- static const char __kstrtab_##sym[] \
+ static const char __kstrtab_##sym##nspost[] \
__attribute__((section("__ksymtab_strings"), aligned(1))) \
= #sym; \
- static const struct kernel_symbol __ksymtab_##sym \
+ static const struct kernel_symbol __ksymtab_##sym##nspost \
+ asm("__ksymtab_" #sym nspost2) \
__used \
- __attribute__((section("___ksymtab" sec "+" #sym), used)) \
+ __attribute__((section("___ksymtab" sec "+" #sym #nspost))) \
+ __attribute__((used)) \
__attribute__((aligned(sizeof(void *)))) \
- = { (unsigned long)&sym, __kstrtab_##sym }
+ = { (unsigned long)&sym, __kstrtab_##sym##nspost, ns }

#if defined(__KSYM_DEPS__)

@@ -76,52 +91,80 @@ extern struct module __this_module;
* system filters out from the preprocessor output (see ksym_dep_filter
* in scripts/Kbuild.include).
*/
-#define __EXPORT_SYMBOL(sym, sec) === __KSYM_##sym ===
+#define __EXPORT_SYMBOL(sym, sec, ns, nspost, nspost2) === __KSYM_##sym ===

#elif defined(CONFIG_TRIM_UNUSED_KSYMS)

#include <generated/autoksyms.h>

-#define __EXPORT_SYMBOL(sym, sec) \
- __cond_export_sym(sym, sec, __is_defined(__KSYM_##sym))
-#define __cond_export_sym(sym, sec, conf) \
- ___cond_export_sym(sym, sec, conf)
-#define ___cond_export_sym(sym, sec, enabled) \
- __cond_export_sym_##enabled(sym, sec)
-#define __cond_export_sym_1(sym, sec) ___EXPORT_SYMBOL(sym, sec)
-#define __cond_export_sym_0(sym, sec) /* nothing */
+#define __EXPORT_SYMBOL(sym, sec, ns, nspost, nspost2) \
+ __cond_export_sym(sym, sec, ns, nspost, nspost2, \
+ __is_defined(__KSYM_##sym))
+#define __cond_export_sym(sym, sec, ns, nspost, nspost2, conf) \
+ ___cond_export_sym(sym, sec, ns, nspost, nspost2, conf)
+#define ___cond_export_sym(sym, sec, ns, nspost, nspost2, enabled) \
+ __cond_export_sym_##enabled(sym, sec, ns, nspost, nspost2)
+#define __cond_export_sym_1(sym, sec, ns, nspost, nspost2) \
+ ___EXPORT_SYMBOL(sym, sec, ns, nspost, nspost2)
+#define __cond_export_sym_0(sym, sec, ns, nspost, nspost2) /* nothing */

#else
#define __EXPORT_SYMBOL ___EXPORT_SYMBOL
#endif

#define EXPORT_SYMBOL(sym) \
- __EXPORT_SYMBOL(sym, "")
+ __EXPORT_SYMBOL(sym, "", NULL, ,)

#define EXPORT_SYMBOL_GPL(sym) \
- __EXPORT_SYMBOL(sym, "_gpl")
+ __EXPORT_SYMBOL(sym, "_gpl", NULL, ,)

#define EXPORT_SYMBOL_GPL_FUTURE(sym) \
- __EXPORT_SYMBOL(sym, "_gpl_future")
+ __EXPORT_SYMBOL(sym, "_gpl_future", NULL, ,)
+
+#define EXPORT_SYMBOL_NS(sym, ns) \
+ __EXPORT_SYMBOL(sym, "", #ns, __##ns, NS_SEPARATOR #ns)
+
+#define EXPORT_SYMBOL_NS_GPL(sym, ns) \
+ __EXPORT_SYMBOL(sym, "_gpl", #ns, __##ns, NS_SEPARATOR #ns)

#ifdef CONFIG_UNUSED_SYMBOLS
-#define EXPORT_UNUSED_SYMBOL(sym) __EXPORT_SYMBOL(sym, "_unused")
-#define EXPORT_UNUSED_SYMBOL_GPL(sym) __EXPORT_SYMBOL(sym, "_unused_gpl")
+#define EXPORT_UNUSED_SYMBOL(sym) __EXPORT_SYMBOL(sym, "_unused", , ,)
+#define EXPORT_UNUSED_SYMBOL_GPL(sym) __EXPORT_SYMBOL(sym, "_unused_gpl", , ,)
#else
#define EXPORT_UNUSED_SYMBOL(sym)
#define EXPORT_UNUSED_SYMBOL_GPL(sym)
#endif

-#endif /* __GENKSYMS__ */
+#endif /* __KERNEL__ && !__GENKSYMS__ */
+
+#if defined(__GENKSYMS__)
+/*
+ * When we're running genksyms, ignore the namespace and make the _NS
+ * variants look like the normal ones. There are two reasons for this:
+ * 1) In the normal definition of EXPORT_SYMBOL_NS, the 'ns' macro
+ * argument is itself not expanded because it's always tokenized or
+ * concatenated; but when running genksyms, a blank definition of the
+ * macro does allow the argument to be expanded; if a namespace
+ * happens to collide with a #define, this can cause issues.
+ * 2) There's no need to modify genksyms to deal with the _NS variants
+ */
+#define EXPORT_SYMBOL_NS(sym, ns) \
+ EXPORT_SYMBOL(sym)
+#define EXPORT_SYMBOL_NS_GPL(sym, ns) \
+ EXPORT_SYMBOL_GPL(sym)
+#endif

#else /* !CONFIG_MODULES... */

#define EXPORT_SYMBOL(sym)
+#define EXPORT_SYMBOL_NS(sym, ns)
+#define EXPORT_SYMBOL_NS_GPL(sym, ns)
#define EXPORT_SYMBOL_GPL(sym)
#define EXPORT_SYMBOL_GPL_FUTURE(sym)
#define EXPORT_UNUSED_SYMBOL(sym)
#define EXPORT_UNUSED_SYMBOL_GPL(sym)

+#define MODULE_IMPORT_NS(ns)
#endif /* CONFIG_MODULES */
#endif /* !__ASSEMBLY__ */

diff --git a/include/linux/module.h b/include/linux/module.h
index d44df9b2c131..afab4e8fa188 100644
--- a/include/linux/module.h
+++ b/include/linux/module.h
@@ -268,6 +268,12 @@ void *__symbol_get(const char *symbol);
void *__symbol_get_gpl(const char *symbol);
#define symbol_get(x) ((typeof(&x))(__symbol_get(VMLINUX_SYMBOL_STR(x))))

+/* namespace dependencies of the module */
+struct module_ns_dep {
+ struct list_head ns_dep;
+ const char *namespace;
+};
+
/* modules using other modules: kdb wants to see this. */
struct module_use {
struct list_head source_list;
@@ -359,6 +365,13 @@ struct module {
const struct kernel_symbol *gpl_syms;
const s32 *gpl_crcs;

+ /* Namespace imports */
+ unsigned int num_ns_imports;
+ const struct namespace_import *ns_imports;
+
+ /* Namespace dependencies */
+ struct list_head ns_dependencies;
+
#ifdef CONFIG_UNUSED_SYMBOLS
/* unused exported symbols. */
const struct kernel_symbol *unused_syms;
diff --git a/kernel/module.c b/kernel/module.c
index f475f30eed8c..63de0fe849f9 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -1166,6 +1166,51 @@ static inline int module_unload_init(struct module *mod)
}
#endif /* CONFIG_MODULE_UNLOAD */

+static bool module_has_ns_dependency(struct module *mod, const char *ns)
+{
+ struct module_ns_dep *ns_dep;
+
+ list_for_each_entry(ns_dep, &mod->ns_dependencies, ns_dep) {
+ if (strcmp(ns_dep->namespace, ns) == 0)
+ return true;
+ }
+
+ return false;
+}
+
+static int add_module_ns_dependency(struct module *mod, const char *ns)
+{
+ struct module_ns_dep *ns_dep;
+
+ if (module_has_ns_dependency(mod, ns))
+ return 0;
+
+ ns_dep = kmalloc(sizeof(*ns_dep), GFP_ATOMIC);
+ if (!ns_dep)
+ return -ENOMEM;
+
+ ns_dep->namespace = ns;
+
+ list_add(&ns_dep->ns_dep, &mod->ns_dependencies);
+
+ return 0;
+}
+
+static bool module_imports_ns(struct module *mod, const char *ns)
+{
+ size_t i;
+
+ if (!ns)
+ return true;
+
+ for (i = 0; i < mod->num_ns_imports; ++i) {
+ if (!strcmp(mod->ns_imports[i].namespace, ns))
+ return true;
+ }
+
+ return false;
+}
+
static size_t module_flags_taint(struct module *mod, char *buf)
{
size_t l = 0;
@@ -1415,6 +1460,18 @@ static const struct kernel_symbol *resolve_symbol(struct module *mod,
goto getname;
}

+ /*
+ * We can't yet verify that the module actually imports this
+ * namespace, because the imports themselves are only available
+ * after processing the symbol table and doing relocation; so
+ * instead just record the dependency and check later.
+ */
+ if (sym->namespace) {
+ err = add_module_ns_dependency(mod, sym->namespace);
+ if (err)
+ sym = ERR_PTR(err);
+ }
+
getname:
/* We must make copy under the lock if we failed to get ref. */
strncpy(ownername, module_name(owner), MODULE_NAME_LEN);
@@ -3061,6 +3118,11 @@ static int find_module_sections(struct module *mod, struct load_info *info)
sizeof(*mod->gpl_syms),
&mod->num_gpl_syms);
mod->gpl_crcs = section_addr(info, "__kcrctab_gpl");
+
+ mod->ns_imports = section_objs(info, "__knsimport",
+ sizeof(*mod->ns_imports),
+ &mod->num_ns_imports);
+
mod->gpl_future_syms = section_objs(info,
"__ksymtab_gpl_future",
sizeof(*mod->gpl_future_syms),
@@ -3381,6 +3443,19 @@ static int post_relocation(struct module *mod, const struct load_info *info)
return module_finalize(info->hdr, info->sechdrs, mod);
}

+static void verify_namespace_dependencies(struct module *mod)
+{
+ struct module_ns_dep *ns_dep;
+
+ list_for_each_entry(ns_dep, &mod->ns_dependencies, ns_dep) {
+ if (!module_imports_ns(mod, ns_dep->namespace)) {
+ pr_warn("%s: module uses symbols from namespace %s,"
+ " but does not import it.\n",
+ mod->name, ns_dep->namespace);
+ }
+ }
+}
+
/* Is this module of this name done loading? No locks held. */
static bool finished_loading(const char *name)
{
@@ -3682,6 +3757,8 @@ static int load_module(struct load_info *info, const char __user *uargs,
if (err)
goto free_module;

+ INIT_LIST_HEAD(&mod->ns_dependencies);
+
#ifdef CONFIG_MODULE_SIG
mod->sig_ok = info->sig_ok;
if (!mod->sig_ok) {
@@ -3730,6 +3807,8 @@ static int load_module(struct load_info *info, const char __user *uargs,
if (err < 0)
goto free_modinfo;

+ verify_namespace_dependencies(mod);
+
flush_module_icache(mod);

/* Now copy in args */
--
2.18.0.203.gfac676dfb9-goog