[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <57510765-be85-4f91-493b-f7213a7580f3@redhat.com>
Date: Tue, 16 Apr 2019 09:37:36 -0400
From: Joe Lawrence <joe.lawrence@...hat.com>
To: Balbir Singh <bsingharora@...il.com>
Cc: linux-kernel@...r.kernel.org, live-patching@...r.kernel.org,
linux-kbuild@...r.kernel.org, Jessica Yu <jeyu@...nel.org>,
Jiri Kosina <jikos@...nel.org>,
Joao Moreira <jmoreira@...e.de>,
Josh Poimboeuf <jpoimboe@...hat.com>,
Konstantin Khlebnikov <khlebnikov@...dex-team.ru>,
Masahiro Yamada <yamada.masahiro@...ionext.com>,
Michael Matz <matz@...e.de>, Miroslav Benes <mbenes@...e.cz>,
Nicolai Stange <nstange@...e.de>,
Petr Mladek <pmladek@...e.com>
Subject: Re: [PATCH v3 0/9] klp-convert livepatch build tooling
On 4/16/19 1:24 AM, Balbir Singh wrote:
>
> Could we get some details with examples or a sample, sorry I might be dense
> and missing simple information. The problem being solved is not clear to
> me from the changelog.
>
Hi Balbir,
I see that Miroslav has already pointed to documentation, samples and
self-tests for the patchset, but here is a summary in my own words:
kpatch-build constructs livepatch kernel modules by comparing a
reference build with a patched build, combing through ELF object
sections and extracting new and changed sections that include patched code.
An alternative approach is "source-based", in which a livepatch kernel
module is built (mostly entirely) using the kernel's build
infrastructure. Sources for a livepatch are gathered ahead of time and
then built like an ordinary kernel module.
In either approach, there lies the problem of symbol visibility: how can
a livepatch resolve symbols that a kernel module ordinarily can't. For
example, file local or simply unexported symbols in across the kernel
and other modules. Enter the concept of "livepatch symbols" described
in module-elf-format.txt.
kpatch-build already creates such "livepatch symbols" (see its
create_klp_relasecs_and_syms()) and the livepatching core kernel code
already knows how resolve such symbols at klp_object patch time (see
klp_write_object_relocations()).
The klp-convert tool and this supporting patchset would empower
source-based-constructed livepatch modules to do the same.
-- Joe
Powered by blists - more mailing lists