Re: [PATCH] Remove unnecessary IFUNC dispatch for __memset_chk.

[Zack Weinberg <zackw at panix dot com>]

On 08/09/2015 02:41 AM, Andreas Schwab wrote:
> Zack Weinberg <zackw@panix.com> writes:
> 
>> __memset_chk does a bounds check (which the C compiler will optimize
>> just fine) and then calls memset.  That call will itself go through
>> IFUNC dispatch,
> 
> Does it?

On further investigation it appears not to -- specifically, internal
calls using __GI_foo appear to go straight to the default implementation
of 'foo'.

If so, I am inclined to think that that is a bug -- there are a *lot* of
internal calls to memset and memcpy in libc, they should not miss out on
architectural tuning.  I don't particularly understand how IFUNC works,
but wouldn't it be sufficient to send internal calls to anything with an
IFUNC through the PLT?  (I suppose there would then be a question of
whether the architectural optimizations made up for the PLT overhead.)

I am also inclined to think that there is no good reason to gunk up the
fortify implementation with IFUNCed _chk functions, regardless.  If
losing out on architecturally tuned memory block functions is acceptable
for normal calls inside libc, it should also be acceptable for fortified
applications (which one *expects* to incur a cost).

Regardless, I don't want to block the other thing I'm working on on
this, and I can see that this patch needs to be more complicated than it
first appeared (in addition to all the above, there's even more
memset_chk goo in sysdeps/ I didn't delete, and several more _chk
functions with the same treatment) so I'll withdraw it for now and
figure out another way to solve the problem I had with the other thing.

cc:ing HJ who seems to have implemented this stuff in the first place.

zw

Attachment: signature.asc
Description: OpenPGP digital signature