Re: [PATCH] modpost: move the namespace field in Module.symvers last

[Jessica Yu <jeyu@kernel.org>]

+++ Jessica Yu [04/03/20 18:03 +0100]:
>In order to preserve backwards compatability with kmod tools, we have to
>move the namespace field in Module.symvers last, as the depmod -e -E
>option looks at the first three fields in Module.symvers to check symbol
>versions (and it's expected they stay in the original order of crc,
>symbol, module).
>
>Fixes: cb9b55d21fe0 ("modpost: add support for symbol namespaces")
>Cc: stable@vger.kernel.org
>Signed-off-by: Jessica Yu <jeyu@kernel.org>

First, I apologize for not having caught this mistake earlier. I still
have questions about the Module.symvers format, please see below.

>---
> Documentation/kbuild/modules.rst |  4 ++--
> scripts/export_report.pl         |  2 +-
> scripts/mod/modpost.c            | 24 ++++++++++++------------
> 3 files changed, 15 insertions(+), 15 deletions(-)
>
>diff --git a/Documentation/kbuild/modules.rst b/Documentation/kbuild/modules.rst
>index 69fa48ee93d6..e0b45a257f21 100644
>--- a/Documentation/kbuild/modules.rst
>+++ b/Documentation/kbuild/modules.rst
>@@ -470,9 +470,9 @@ build.
>
> 	The syntax of the Module.symvers file is::
>
>-	<CRC>       <Symbol>          <Namespace>  <Module>                         <Export Type>
>+	<CRC>       <Symbol>         <Module>                         <Export Type>     <Namespace>
>
>-	0xe1cc2a05  usb_stor_suspend  USB_STORAGE  drivers/usb/storage/usb-storage  EXPORT_SYMBOL_GPL
>+	0xe1cc2a05  usb_stor_suspend drivers/usb/storage/usb-storage  EXPORT_SYMBOL_GPL USB_STORAGE
>
> 	The fields are separated by tabs and values may be empty (e.g.
> 	if no namespace is defined for an exported symbol).
>diff --git a/scripts/export_report.pl b/scripts/export_report.pl
>index 548330e8c4e7..feb3d5542a62 100755
>--- a/scripts/export_report.pl
>+++ b/scripts/export_report.pl
>@@ -94,7 +94,7 @@ if (defined $opt{'o'}) {
> #
> while ( <$module_symvers> ) {
> 	chomp;
>-	my (undef, $symbol, $namespace, $module, $gpl) = split('\t');
>+	my (undef, $symbol, $module, $gpl, $namespace) = split('\t');
> 	$SYMBOL { $symbol } =  [ $module , "0" , $symbol, $gpl];
> }
> close($module_symvers);
>diff --git a/scripts/mod/modpost.c b/scripts/mod/modpost.c
>index 7edfdb2f4497..6ab235354f36 100644
>--- a/scripts/mod/modpost.c
>+++ b/scripts/mod/modpost.c
>@@ -2427,7 +2427,7 @@ static void write_if_changed(struct buffer *b, const char *fname)
> }
>
> /* parse Module.symvers file. line format:
>- * 0x12345678<tab>symbol<tab>module[[<tab>export]<tab>something]
>+ * 0x12345678<tab>symbol<tab>module<tab>export<tab>namespace

So, this comment was a source of confusion for me and Matthias I
think. It suggests that the export field is optional, and that even
following the export field there may also be "something" else,
whatever that is.

I suspect that there were historical reasons behind that comment that
are no longer accurate. We have been unconditionally printing the
export type since 2.6.18 (commit bd5cbcedf44), which is over a decade
ago now. And let me explain the read_dump() changes...

>  **/
> static void read_dump(const char *fname, unsigned int kernel)
> {
>@@ -2440,7 +2440,7 @@ static void read_dump(const char *fname, unsigned int kernel)
> 		return;
>
> 	while ((line = get_next_line(&pos, file, size))) {
>-		char *symname, *namespace, *modname, *d, *export, *end;
>+		char *symname, *namespace, *modname, *d, *export;
> 		unsigned int crc;
> 		struct module *mod;
> 		struct symbol *s;
>@@ -2448,16 +2448,16 @@ static void read_dump(const char *fname, unsigned int kernel)
> 		if (!(symname = strchr(line, '\t')))
> 			goto fail;
> 		*symname++ = '\0';
>-		if (!(namespace = strchr(symname, '\t')))
>-			goto fail;
>-		*namespace++ = '\0';
>-		if (!(modname = strchr(namespace, '\t')))
>+		if (!(modname = strchr(symname, '\t')))
> 			goto fail;
> 		*modname++ = '\0';
>-		if ((export = strchr(modname, '\t')) != NULL)
>-			*export++ = '\0';
>-		if (export && ((end = strchr(export, '\t')) != NULL))
>-			*end = '\0';

I believe the original read_dump() code treated the export field here
as optional, to support pre <= 2.6.18 Module.symvers (which does not
have the export type field). But I don't believe we have to support
this case anymore, right? It's ages ago. So I cleaned up this area,
made each field non-optional (but empty string "" for namespace is
allowed), and updated the comment.

>+		if (!(export = strchr(modname, '\t')))
>+			goto fail;
>+		*export++ = '\0';
>+		if (!(namespace = strchr(export, '\t')))
>+			goto fail;
>+		*namespace++ = '\0';
>+
> 		crc = strtoul(line, &d, 16);
> 		if (*symname == '\0' || *modname == '\0' || *d != '\0')
> 			goto fail;
>@@ -2508,9 +2508,9 @@ static void write_dump(const char *fname)
> 				namespace = symbol->namespace;
> 				buf_printf(&buf, "0x%08x\t%s\t%s\t%s\t%s\n",
> 					   symbol->crc, symbol->name,
>-					   namespace ? namespace : "",
> 					   symbol->module->name,
>-					   export_str(symbol->export));
>+					   export_str(symbol->export),
>+					   namespace ? namespace : "");
> 			}
> 			symbol = symbol->next;
> 		}
>-- 
>2.16.4
>