Re: [Xen-devel] [PATCH] tools/livepatch: recognize functions in livepatch.ignore.functions

[Elena Ufimtseva <ufimtseva@xxxxxxxxx>]

Hi Ross

Thanks for the info.
I actually did look into the KPATCH_IGNORE_FUNCTION code. But.. I
somehow ended up having this:

#define KPATCH_IGNORE_FUNCTION(_fn) \
void *__kpatch_ignore_func_##_fn __section(.kpatch.ignore.functions) = (#_fn);

Which apparently caused the symbol not being a function type, but a string(?).

Ok, thanks, I will be now dealing with this "artifact", ugh.

Elena


On Tue, Dec 12, 2017 at 12:09 AM, Ross Lagerwall
<ross.lagerwall@xxxxxxxxxx> wrote:
> On 12/12/2017 05:54 AM, elena.ufimtseva@xxxxxxxxxx wrote:
>>
>> From: Elena Ufimtseva <elena.ufimtseva@xxxxxxxxxx>
>>
>> It is expected that the symbol has type STT_FUNC in
>> livpatch.ignore.functions
>> sections, but it is incorrect and results in functions not to be ignored.
>> To actually ignore functions in livepatch.ignore.functions section,
>> attempt to
>> find the symbol of type STT_FUNC by its name.
>>
>
> Hi Elena,
>
> I suspect you might have got the wrong idea about how it is meant to be used
> (or I misunderstood your patch). It is expected that the relocation points
> to a symbol that has type STT_FUNC. This is subtly different from having a
> symbol with type STT_FUNC in .livepatch.ignore.functions.
>
> The correct way to use it is to declare a pointer stored in
> .livepatch.ignore.functions that points to the function you want to ignore.
>
> For example, place this at the end of arch/x86/mm/p2m.c to ignore changes to
> map_domain_gfn:
> void *__lp_ignore_func_map_domain_gfn __section(.livepatch.ignore.functions)
> = map_domain_gfn;
>
> The code in livepatch-build-tools is based on kpatch and they have a macro
> to do this in a more friendly way [1]. If you want, it would be great if you
> could port this to Xen LivePatch.
>
> [1]
> https://github.com/dynup/kpatch/blob/f4c0f3209e8e856d93622344560c8794cd8d8a45/kmod/patch/kpatch-macros.h#L39
>
> As an aside, much of the code in livepatch-build-tools comes from kpatch, so
> patches are ideally sent there first. We do intend eventually to merge the
> tools to avoid code duplication and kpatch's build tool was recently
> refactored to make this easier.
>
> Cheers,
> --
> Ross Lagerwall
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxxx
> https://lists.xenproject.org/mailman/listinfo/xen-devel



-- 
Elena

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel