[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <5587B4FB.3040609@nod.at>
Date: Mon, 22 Jun 2015 09:10:51 +0200
From: Richard Weinberger <richard@....at>
To: Alexei Starovoitov <alexei.starovoitov@...il.com>,
Nicolai Stange <nicstange@...il.com>
CC: "David S. Miller" <davem@...emloft.net>, netdev@...r.kernel.org,
linux-kernel@...r.kernel.org, daniel@...earbox.net
Subject: Re: [PATCH] lib: test_bpf: purge CPP register redefinitions
Am 22.06.2015 um 08:52 schrieb Richard Weinberger:
> Am 22.06.2015 um 08:05 schrieb Alexei Starovoitov:
>> to get rid of warning you proposing to do 1k line renames?!
>> Just add:
>> +#undef R8
>> +#undef R9
>> +#undef R10
>> #define R0 BPF_REG_0
>
> This would be also just another hack.
>
>> Though I think the better fix woud be to clean up:
>> arch/x86/include/uapi/asm/ptrace-abi.h
>> What's the point of:
>> #define R8 72
>> from 'uapi' point of view?
>
> To query cpu registers using ptrace(2).
>
>> Look like kernel details that shouldn't be exposed in uapi.
>
> These are not kernel details.
>
> Actually the problem is the other way around.
> UML is Linux ported to it's own userspace ABI.
> Hence, the arch/um and arch/x86/um use uapi header files.
>
> Maybe we can rework UML's header files such that
> no uapi header pollutes the kernel namespace.
While riding the bus to my office I've materialized that idea.
Nicolai, can you please give the attached patch a try?
Thanks,
//richard
View attachment "ptrace_abi.diff" of type "text/x-patch" (1942 bytes)
Powered by blists - more mailing lists