[PATCH v3] kallsyms: add support for relative offsets in kallsyms address table

[Ard Biesheuvel]

On 22 January 2016 at 04:44, Michael Ellerman <mpe at ellerman.id.au> wrote:
> On Thu, 2016-01-21 at 14:55 -0800, Kees Cook wrote:
>> On Thu, Jan 21, 2016 at 2:50 PM, Andrew Morton
>> <akpm at linux-foundation.org> wrote:
>> > On Thu, 21 Jan 2016 18:19:43 +0100 Ard Biesheuvel <ard.biesheuvel at linaro.org> wrote:
>> >
>> > > Similar to how relative extables are implemented, it is possible to emit
>> > > the kallsyms table in such a way that it contains offsets relative to some
>> > > anchor point in the kernel image rather than absolute addresses. The benefit
>> > > is that such table entries are no longer subject to dynamic relocation when
>> > > the build time and runtime offsets of the kernel image are different. Also,
>> > > on 64-bit architectures, it essentially cuts the size of the address table
>> > > in half since offsets can typically be expressed in 32 bits.
>> > >
>> > > Since it is useful for some architectures (like x86) to retain the ability
>> > > to emit absolute values as well, this patch adds support for both, by
>> > > emitting absolute addresses as positive 32-bit values, and addresses
>> > > relative to the lowest encountered relative symbol as negative values, which
>> > > are subtracted from the runtime address of this base symbol to produce the
>> > > actual address.
>> > >
>> > > Support for the above is enabled by default for all architectures except
>> > > IA-64, whose symbols are too far apart to capture in this manner.
>> >
>> > I'm not really understanding the benefits of this.  A smaller address
>> > table is nice, but why is it desirable that "such table entries are no
>> > longer subject to dynamic relocation when the build time and runtime
>> > offsets of the kernel image are different"?
>>
>> IIUC, this means that the relocation work done after decompression now
>> doesn't have to do relocation updates for all these values, which
>> means a smaller relocation table as well.
>
> Yep. If I remember the figures rightly it saves ~250K of relocations for the
> powerpc build.
>

For ppc64_defconfig (which has CONFIG_RELOCATABLE=y, i.e., it has a
dynamic relocation section containing a 24-byte RELA entry per
relocated quantity), I got the following numbers

101740 kallsyms entries
397 KB saved in permanent .rodata
2.4 MB saved in __init rela.dyn section
~500 KB saved in compressed image

For arm64, we don't have a compressed image, which is the reason I
need this for my arm64 implementation of CONFIG_RELOCATABLE (for
KASLR), since the RELA overhead goes straight into the distributed
image.

Thanks,
Ard.