Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 18 Aug 2008 12:11:27 +0200
From:      "Alexander Leidinger" <Alexander@Leidinger.net>
To:        "Adrian Penisoara" <ady@freebsd.ady.ro>
Cc:        freebsd-emulation@freebsd.org, Chagin Dmitry <dchagin@freebsd.org>
Subject:   Re: x86_64 linuxulator patches
Message-ID:  <20080818121127.745322h5l2ufrxs0@webmail.leidinger.net>
In-Reply-To: <78cb3d3f0808180138o6c2e7c3bhf6f49b4de547b218@mail.gmail.com>
References:  <20080810072013.GA15196@dchagin.dialup.corbina.ru> <20080810115406.GR97161@deviant.kiev.zoral.com.ua> <20080810120424.GA15768@dchagin.dialup.corbina.ru> <20080810122124.GS97161@deviant.kiev.zoral.com.ua> <20080817181757.GA2940@dchagin.dialup.corbina.ru> <20080818100541.20073jcxqbtyci80@webmail.leidinger.net> <78cb3d3f0808180138o6c2e7c3bhf6f49b4de547b218@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Quoting "Adrian Penisoara" <ady@freebsd.ady.ro> (from Mon, 18 Aug 2008 =20
10:38:02 +0200):

> Hi,
>
> On Mon, Aug 18, 2008 at 10:05 AM, Alexander Leidinger
> <Alexander@leidinger.net> wrote:
>>> in my opinion the best decision for amd64 looks so.
>>> we use two modules. linux.ko for x86_64 and linux32.ko for ia32,
>>> option COMPAT_LINUX for x86_64 and COMPAT_LINUX32 for ia32.
>>> and two  linux_base directories: /compat/linux for x86_64
>>> and /compat/linux32 for ia32.
>>>
>>> there are other opinions?
>>
>> I propose:
>>  - /compat/linux64 for 64bit stuff
>>  - /compat/linux32 a symlink to /compat/linux
>>  - /compat/linux for 32bit stuff (we can think about having)
>>
>
>  I agree with this later proposition, it's a bad thing to break the
> already established purpose for /compat/linux (32bit binaries).
>
>  Here is yet another variation:
>    /compat/linux64 -- 64bit
>    /compat/linux32 -- 32bit
>    /compat/linux -> linux32  -- symlink, would this break anything in
> the current ports/packages ?

Making linux a symlink to linux32 will not work. I made the proposal =20
the other way around on purpose. If an user does not update after the =20
changes are introduced, any new installation (without the update) will =20
mess-up the installation. If you don't have to worry about backwards =20
compatibility, it's ok (and I would prefer it), but unfortunately it =20
will create more hassle than it will help in the real world (there are =20
too much people which will not make the right thing in such a case).

I also thought about workarounds, by e.g. checking if all involved =20
paths are of the correct type and refuse to work or to adjust the =20
paths if they aren't, but this will make the LINUXBASE handling very =20
complicated, more than it is worth (there's already a lot of work to =20
do to get all infrastructure for this together).

>  I strongly oppose having linux.ko as the 64bit version, rather there
> should be linux64.ko (linux-x86_64.ko ?) and linux32.ko (linux-ia32.ko
> ?) and perhaps we should make linux.ko try loading both of these
> (since the user did not specify the intended platform).

I like this. linux.ko as a meta module which depends upon linux32 and =20
linux64 on amd64, and depends upon linux32 on i386. This way POLA is =20
preserved, the documentation is not wrong (to enable linux emulation =20
load the linux.ko), and people which only want one can load the =20
specific one.

Bye,
Alexander.

--=20
=09A domineering man married a mere wisp of a girl.  He came back from
his honeymoon a chastened man.  He'd become aware of the will of the wisp.

http://www.Leidinger.net    Alexander @ Leidinger.net: PGP ID =3D B0063FE7
http://www.FreeBSD.org       netchild @ FreeBSD.org  : PGP ID =3D 72077137



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20080818121127.745322h5l2ufrxs0>