From mboxrd@z Thu Jan  1 00:00:00 1970
From: Andi Kleen <andi@firstfloor.org>
Subject: Re: __bad_udelay in network driver breaks build
Date: Tue, 26 Oct 2010 00:10:42 +0200
Message-ID: <20101025221042.GE17622@basil.fritz.box>
References: <20101018115230.GA27565@basil.fritz.box>
 <20101025.130513.102558756.davem@davemloft.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: andi@firstfloor.org, netdev@vger.kernel.org,
	jeffrey.t.kirsher@intel.com, alexander.h.duyck@intel.com
To: David Miller <davem@davemloft.net>
Return-path: <netdev-owner@vger.kernel.org>
Received: from one.firstfloor.org ([213.235.205.2]:37653 "EHLO
	one.firstfloor.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1752048Ab0JYWKs (ORCPT
	<rfc822;netdev@vger.kernel.org>); Mon, 25 Oct 2010 18:10:48 -0400
Content-Disposition: inline
In-Reply-To: <20101025.130513.102558756.davem@davemloft.net>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

On Mon, Oct 25, 2010 at 01:05:13PM -0700, David Miller wrote:
> From: Andi Kleen <andi@firstfloor.org>
> Date: Mon, 18 Oct 2010 13:52:30 +0200
> 
> > 
> > Current Linus master x86-64 allyesconfig fails with
> > 
> > drivers/built-in.o: In function `tms380tr_chipset_init':
> > tms380tr.c:(.text+0x10f02de): undefined reference to `__bad_udelay'
> > tms380tr.c:(.text+0x10f03ab): undefined reference to `__bad_udelay'
> > tms380tr.c:(.text+0x10f0400): undefined reference to `__bad_udelay'
> > tms380tr.c:(.text+0x10f07b2): undefined reference to `__bad_udelay'
> > tms380tr.c:(.text+0x10f08ed): undefined reference to `__bad_udelay'
> > make[2]: *** [.tmp_vmlinux1] Error 1
> 
> Let me know if this fixes things:

Fixes that problem, but now I get (with Linus' latest again and a gcc 4.6
snapshot)  

In file included from /home/lsrc/git/linux-2.6/drivers/net/igbvf/ethtool.c:36:0:
/home/lsrc/git/linux-2.6/drivers/net/igbvf/igbvf.h:129:15: error: duplicate member 'page'
make[5]: *** [drivers/net/igbvf/ethtool.o] Error 1
make[4]: *** [drivers/net/igbvf] Error 2

struct igbvf_buffer {
        dma_addr_t dma;
        struct sk_buff *skb;
        union {
                /* Tx */
                struct {
                        unsigned long time_stamp;
                        u16 length;
                        u16 next_to_watch;
                        u16 mapped_as_page;
                };
                /* Rx */
                struct {
                        struct page *page; <--------------- No 1
                        u64 page_dma;
                        unsigned int page_offset;
                };
        };
        struct page *page;         <------------ No 2
};

Hmm conflict of a member with a transparent union.
Maybe older gccs didn't catch that. But it looks very broken

Alexander, can you sort out which "page" member should be used where?

And there's another problem in SCSI which I'll report separately.

-Andi