Re: module: s390: keep mod_arch_specific for livepatch modules

On Thu, 4 Feb 2016, Josh Poimboeuf wrote:

On Wed, Feb 03, 2016 at 08:37:52PM -0500, Jessica Yu wrote:

quoted

+++ Jessica Yu [03/02/16 20:11 -0500]:

quoted

Livepatch needs to utilize the symbol information contained in the
mod_arch_specific struct in order to be able to call the s390
apply_relocate_add() function to apply relocations. Keep a reference to
syminfo if the module is a livepatch module. Remove the redundant vfree()
in module_finalize() since module_arch_freeing_init() (which also frees
those structures) is called in do_init_module(). If the module isn't a
livepatch module, we free the structures in module_arch_freeing_init() as
usual.

Signed-off-by: Jessica Yu <redacted>
---
arch/s390/kernel/module.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)

I must note that I have verified that the patchset boots on s390 and
that the sample livepatch module still works ...so that's good, but
not saying much since what we really want is to test the relocation
code. The kpatch build scripts however currently only support x86, so
the next step is for me to port the kpatch scripts to s390 before I
can really test this patchset. This in itself might take a while, so
in the meantime I'd like to just collect another round of comments and
feedback for v4.

I haven't reviewed the code yet, but otherwise I'm thinking it would
actually be fine to merge this patch set before testing with s390
relocations.  They aren't implemented on s390 today anyway, so there
can't be a regression if nobody is using it.

I asked Jessica to test it with s390 relocations. It would be great to 
know it works on an architecture which originally inspired this patch set. 
However it should not be a blocker. There are advantages apart from this 
and we can happily merge it before testing.

I am going to review the patch set early next week...

Miroslav