2015-08-07 06:55:50

by Takashi Iwai

Subject: [PATCH v2] modpost: abort if a module symbol is too long

Module symbols have a limited length, but currently the build system
allows the build finishing even if the driver code contains a too long
symbol name, which eventually overflows the modversion_info[] item.
The compiler may catch at compiling *.mod.c like
CC xxx.mod.o
xxx.mod.c:18:16: warning: initializer-string for array of chars is too long
but it's merely a warning.

This patch adds the check of the symbol length in modpost and stops
the build properly.

Currently MODULE_NAME_LEN is defined in modpost.c instead of referring
to the definition in kernel header because including linux/module.h is
messy and we must cover cross-compilation.

Signed-off-by: Takashi Iwai <[email protected]>
---
v1->v2: Fix MODULE_NAME_LEN to use Elf_Addr

scripts/mod/modpost.c | 11 +++++++++++
1 file changed, 11 insertions(+)

diff --git a/scripts/mod/modpost.c b/scripts/mod/modpost.c
index 12d3db3bd46b..d583c98fde31 100644
--- a/scripts/mod/modpost.c
+++ b/scripts/mod/modpost.c
@@ -2133,6 +2133,11 @@ static void add_staging_flag(struct buffer *b, const char *name)
buf_printf(b, "\nMODULE_INFO(staging, \"Y\");\n");
}

+/* In kernel, this size is defined in linux/module.h;
+ * here we use Elf_Addr instead of long for covering cross-compile
+ */
+#define MODULE_NAME_LEN (64 - sizeof(Elf_Addr))
+
/**
* Record CRCs for unresolved symbols
**/
@@ -2177,6 +2182,12 @@ static int add_versions(struct buffer *b, struct module *mod)
s->name, mod->name);
continue;
}
+ if (strlen(s->name) >= MODULE_NAME_LEN) {
+ merror("too long symbol \"%s\" [%s.ko]\n",
+ s->name, mod->name);
+ err = 1;
+ break;
+ }
buf_printf(b, "\t{ %#8x, __VMLINUX_SYMBOL_STR(%s) },\n",
s->crc, s->name);
}
--
2.5.0

2015-08-09 20:19:40

by Rusty Russell

Subject: Re: [PATCH v2] modpost: abort if a module symbol is too long

Takashi Iwai <[email protected]> writes:
> Module symbols have a limited length, but currently the build system
> allows the build finishing even if the driver code contains a too long
> symbol name, which eventually overflows the modversion_info[] item.
> The compiler may catch at compiling *.mod.c like
> CC xxx.mod.o
> xxx.mod.c:18:16: warning: initializer-string for array of chars is too long
> but it's merely a warning.

Thanks, applied!

Cheers,
Rusty.

>
> This patch adds the check of the symbol length in modpost and stops
> the build properly.
>
> Currently MODULE_NAME_LEN is defined in modpost.c instead of referring
> to the definition in kernel header because including linux/module.h is
> messy and we must cover cross-compilation.
>
> Signed-off-by: Takashi Iwai <[email protected]>
> ---
> v1->v2: Fix MODULE_NAME_LEN to use Elf_Addr
>
> scripts/mod/modpost.c | 11 +++++++++++
> 1 file changed, 11 insertions(+)
>
> diff --git a/scripts/mod/modpost.c b/scripts/mod/modpost.c
> index 12d3db3bd46b..d583c98fde31 100644
> --- a/scripts/mod/modpost.c
> +++ b/scripts/mod/modpost.c
> @@ -2133,6 +2133,11 @@ static void add_staging_flag(struct buffer *b, const char *name)
> buf_printf(b, "\nMODULE_INFO(staging, \"Y\");\n");
> }
>
> +/* In kernel, this size is defined in linux/module.h;
> + * here we use Elf_Addr instead of long for covering cross-compile
> + */
> +#define MODULE_NAME_LEN (64 - sizeof(Elf_Addr))
> +
> /**
> * Record CRCs for unresolved symbols
> **/
> @@ -2177,6 +2182,12 @@ static int add_versions(struct buffer *b, struct module *mod)
> s->name, mod->name);
> continue;
> }
> + if (strlen(s->name) >= MODULE_NAME_LEN) {
> + merror("too long symbol \"%s\" [%s.ko]\n",
> + s->name, mod->name);
> + err = 1;
> + break;
> + }
> buf_printf(b, "\t{ %#8x, __VMLINUX_SYMBOL_STR(%s) },\n",
> s->crc, s->name);
> }
> --
> 2.5.0