netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Stephen Hemminger <shemminger@vyatta.com>
To: Andi Kleen <andi@firstfloor.org>
Cc: subbu <subbus@serverengines.com>,
	'Andi Kleen' <andi@firstfloor.org>,
	netdev@vger.kernel.org
Subject: Re: [PATCH 9/15] BE NIC driver - Linux system abstraction files
Date: Thu, 15 May 2008 13:22:18 -0700	[thread overview]
Message-ID: <20080515132218.2c27008f@extreme> (raw)
In-Reply-To: <20080515174927.GC30216@one.firstfloor.org>

On Thu, 15 May 2008 19:49:27 +0200
Andi Kleen <andi@firstfloor.org> wrote:

> On Thu, May 15, 2008 at 11:00:46PM +0530, subbu wrote:
> > The reason why we have these abstractions is 
> > because beclib functions are written as 
> > OS neutral code and that needs a helper to
> > do OS specific functions.  By eliminating
> > SA, we will need to make  beclib Linux
> > dependent - something that we would like avoid
> > for ease of maintenance.  Is this approach
> > not acceptable ?  What will is the recommended
> > way to have part so a driver OS independent ?
> 
> We do not recommend it,  but recommend to write Linux
> specific drivers. Operating systems are quite different and you
> get far better drivers if you really optimize them for specific OS.
> That's especially important for performance critical drivers like
> yours for high speed networking.
> 
> Also splitting drivers into generic and OS specific parts typically
> leads to poor and overcomplicated design.
> 
> Anyways if you really want abstraction elsewhere then use the Linux
> names for it. But it's really not recommended.
> 
> -Andi

The underlying reason for this is the Linux kernel developers hatred
of abstraction layers is long term maintenance. Once a driver is accepted
into the mainline, the kernel developers may have to do maintenance on
the driver to fix API's or if the original hardware vendor abandons it.
All the abstraction layer just makes this harder. Some drivers do have
some OS independent portions but they are rare, and usually abstract the
hardware not the device model.

      reply	other threads:[~2008-05-15 20:22 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-05-15  9:26 [PATCH 9/15] BE NIC driver - Linux system abstraction files Subbu Seetharaman
2008-05-15 11:33 ` Andi Kleen
2008-05-15 17:33   ` Jeff Garzik
     [not found]   ` <20080515173841.A0A48189004A@one.firstfloor.org>
2008-05-15 17:49     ` Andi Kleen
2008-05-15 20:22       ` Stephen Hemminger [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20080515132218.2c27008f@extreme \
    --to=shemminger@vyatta.com \
    --cc=andi@firstfloor.org \
    --cc=netdev@vger.kernel.org \
    --cc=subbus@serverengines.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).