Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220901171252.2148348-1-song@kernel.org> <alpine.LSU.2.21.2212091352370.18933@pobox.suse.cz>
In-Reply-To: <alpine.LSU.2.21.2212091352370.18933@pobox.suse.cz>
From:   Song Liu <song@kernel.org>
Date:   Tue, 13 Dec 2022 00:28:34 -0800
Message-ID: <CAPhsuW5xb2T5FBXUqG2S+AXBvDYSkLVVvUyDamjrbLQwe-3kVQ@mail.gmail.com>
Subject: Re: [PATCH v6] livepatch: Clear relocation targets on a module removal
To:     Miroslav Benes <mbenes@suse.cz>
Cc:     live-patching@vger.kernel.org, linux-kernel@vger.kernel.org,
        jpoimboe@kernel.org, jikos@kernel.org, pmladek@suse.com,
        x86@kernel.org, joe.lawrence@redhat.com,
        linuxppc-dev@lists.ozlabs.org, Josh Poimboeuf <jpoimboe@redhat.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Dec 9, 2022 at 4:55 AM Miroslav Benes <mbenes@suse.cz> wrote:
>
> Hi,
>
> first thank you for taking over and I also appologize for not replying
> much sooner.
>
> On Thu, 1 Sep 2022, Song Liu wrote:
>
> > From: Miroslav Benes <mbenes@suse.cz>
> >
> > Josh reported a bug:
> >
> >   When the object to be patched is a module, and that module is
> >   rmmod'ed and reloaded, it fails to load with:
> >
> >   module: x86/modules: Skipping invalid relocation target, existing value is nonzero for type 2, loc 00000000ba0302e9, val ffffffffa03e293c
> >   livepatch: failed to initialize patch 'livepatch_nfsd' for module 'nfsd' (-8)
> >   livepatch: patch 'livepatch_nfsd' failed for module 'nfsd', refusing to load module 'nfsd'
> >
> >   The livepatch module has a relocation which references a symbol
> >   in the _previous_ loading of nfsd. When apply_relocate_add()
> >   tries to replace the old relocation with a new one, it sees that
> >   the previous one is nonzero and it errors out.
> >
> >   On ppc64le, we have a similar issue:
> >
> >   module_64: livepatch_nfsd: Expected nop after call, got e8410018 at e_show+0x60/0x548 [livepatch_nfsd]
> >   livepatch: failed to initialize patch 'livepatch_nfsd' for module 'nfsd' (-8)
> >   livepatch: patch 'livepatch_nfsd' failed for module 'nfsd', refusing to load module 'nfsd'
> >
> > He also proposed three different solutions. We could remove the error
> > check in apply_relocate_add() introduced by commit eda9cec4c9a1
> > ("x86/module: Detect and skip invalid relocations"). However the check
> > is useful for detecting corrupted modules.
> >
> > We could also deny the patched modules to be removed. If it proved to be
> > a major drawback for users, we could still implement a different
> > approach. The solution would also complicate the existing code a lot.
> >
> > We thus decided to reverse the relocation patching (clear all relocation
> > targets on x86_64). The solution is not
> > universal and is too much arch-specific, but it may prove to be simpler
> > in the end.
> >
> > Reported-by: Josh Poimboeuf <jpoimboe@redhat.com>
> > Signed-off-by: Miroslav Benes <mbenes@suse.cz>
> > Signed-off-by: Song Liu <song@kernel.org>
>
> Petr has commented on the code aspects. I will just add that s390x was not
> dealt with at the time because there was no live patching support for
> s390x back then if I remember correctly and my notes do not lie. The same
> applies to powerpc32. I think that both should be fixed as well with this
> patch. It might also help to clean up the ifdeffery in the patch a bit.

After reading the code (no testing), I think we don't need any logic for
ppc32 and s390.

We need clear_relocate_add() to handle module reload failure.
The failure happens when we

1) call apply_relocate_add() on klp load (or module first load,
   if klp was loaded first);
2) do nothing when the module is unloaded;
3) call apply_relocate_add() on module reload, which failed.

This failure happens in the sanity check in
apply_relocate_add().

For x86, the check is something like:
                                if (*(s32 *)loc != 0)
                                        goto invalid_relocation;

For ppc64, the check is in restore_r2():

        if (*instruction != PPC_RAW_NOP()) {
                pr_err("%s: Expected nop after call, got %08x at %pS\n",
                        me->name, *instruction, instruction);
                return 0;
        }

I don't think we have similar checks for ppc32 and s390, so
clear_relocate_add() is not needed for the two.

OTOH, we can argue that clear_relocate_add() should undo
everything apply_relocate_add() did. But I do think that
will be an overkill.

WDYT?

Thanks,
Song