Re: [PATCH 4/7] module: avoid code duplication in include/linux/export.h

[Jessica Yu <jeyu@kernel.org>]

+++ Rasmus Villemoes [27/09/19 13:07 +0200]:
>On 27/09/2019 11.36, Masahiro Yamada wrote:
>> include/linux/export.h has lots of code duplication between
>> EXPORT_SYMBOL and EXPORT_SYMBOL_NS.
>>
>> To improve the maintainability and readability, unify the
>> implementation.
>>
>> When the symbol has no namespace, pass the empty string "" to
>> the 'ns' parameter.
>>
>> The drawback of this change is, it grows the code size.
>> When the symbol has no namespace, sym->namespace was previously
>> NULL, but it is now am empty string "". So, it increases 1 byte
>> for every no namespace EXPORT_SYMBOL.
>>
>> A typical kernel configuration has 10K exported symbols, so it
>> increases 10KB in rough estimation.
>>
>> I did not come up with a good idea to refactor it without increasing
>> the code size.
>
>Can't we put the "aMS" flags on the __ksymtab_strings section? That
>would make the empty strings free, and would also deduplicate the
>USB_STORAGE string. And while almost per definition we don't have exact
>duplicates among the names of exported symbols, we might have both a foo
>and __foo, so that could save even more.
>
>I don't know if we have it already, but we'd need each arch to tell us
>what symbol to use for @ in @progbits (e.g. % for arm). It seems most
>are fine with @, so maybe a generic version could be
>
>#ifndef ARCH_SECTION_TYPE_CHAR
>#define ARCH_SECTION_TYPE_CHAR "@"
>#endif
>
>and then it would be
>section("__ksymtab_strings,\"aMS\","ARCH_SECTION_TYPE_CHAR"progbits,1")

FWIW, I've just tinkered with this, and unfortunately the strings
don't get deduplicated for kernel modules :-(

Apparently ld does not do the deduplication for SHF_MERGE|SHF_STRINGS
sections for relocatable files (ld -r), which kernel modules are. See:

    https://sourceware.org/ml/binutils/2009-07/msg00291.html

But, the strings do get deduplicated for vmlinux. Not sure if we can
find a workaround for modules or if the benefit is significant enough
if it only for vmlinux.

Thanks,

Jessica