public inbox for linux-scsi@vger.kernel.org
 help / color / mirror / Atom feed
From: James Bottomley <James.Bottomley@HansenPartnership.com>
To: dl8bcu@dl8bcu.de
Cc: rth@twiddle.net, ink@jurassic.park.msu.ru,
	linux-scsi@vger.kernel.org, linux-alpha@vger.kernel.org
Subject: Re: Build error: `.exit.text' referenced in section `.rodata'
Date: Mon, 12 May 2008 09:01:44 -0500	[thread overview]
Message-ID: <1210600904.3078.1.camel@localhost.localdomain> (raw)
In-Reply-To: <20080512115255.GA16942@ds20.borg.net>

On Mon, 2008-05-12 at 11:52 +0000, Thorsten Kranzkowski wrote:
> On Sun, May 11, 2008 at 07:20:07PM +0000, Thorsten Kranzkowski wrote:
> > On Sun, May 11, 2008 at 11:55:07AM -0500, James Bottomley wrote:
> > > On Sun, 2008-05-11 at 15:19 +0000, Thorsten Kranzkowski wrote:
> > > > Hello,
> > > > 
> > > > compiling 2.6.26-rc1 (latest git) on an Alpha DS20 (ev6, everything built in,
> > > > gcc version 4.1.2) fails with the following link errors:
> > > > 
> > > >   CC      init/version.o
> > > >   LD      init/built-in.o
> > > >   LD      .tmp_vmlinux1
> > > > `.exit.text' referenced in section `.rodata' of drivers/built-in.o: defined in discarded section `.exit.text' of drivers/built-in.o
> > 
> > 
> > > > I believe the problem is static int sd_major(int major_idx) being inlined
> > > > into exit_sd(). IIUC the switch statement therein uses sort of a jump table.
> > > > And the targets of this jump table can not be resolved because the section
> > > > they live in is just discarded.
> > > 
> > > It's plausible.  However, my version of gcc (4.2.3) doesn't actually
> > > inline this; even when I force it to by adding the inline directive to
> > > sd_major() I don't see the problem.  I suspect it might be a gcc bug
> > > 
> > > > 
> > > > applying this patch makes the problem go away:
> > > > 
> > > > Signed-off-by: Thorsten Kranzkowski <dl8bcu@dl8bcu.de>
> > > > 
> > > > diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
> > > > index 01cefbb..adf79ad 100644
> > > > --- a/drivers/scsi/sd.c
> > > > +++ b/drivers/scsi/sd.c
> > > > @@ -1913,7 +1913,7 @@ err_out:
> > > >   *
> > > >   *	Note: this function unregisters this driver from the scsi mid-level.
> > > >   **/
> > > > -static void __exit exit_sd(void)
> > > > +static void exit_sd(void)
> > > 
> > > We can't really do this because that text genuinely does belong in the
> > > exit section.  We might be able to fix it up on the alpha by just adding
> > > the noinline directive to sd_major.  Does this work?
> > 
> > Yes, this works as well! Thanks!
> > 
> >    text    data     bss     dec     hex filename
> > 6834571  504320  397508 7736399  760c4f vmlinux
> > 6835267  504056  397508 7736831  760dff vmlinux-exit_sd-not-in-.exit.text
> > 
> > Tested-by: Thorsten Kranzkowski <dl8bcu@dl8bcu.de>
> >  
> > Now I'll update gcc and try again with gcc 4.2.3.
> 
> 
> With gcc 4.2.3 I get the same error. The noinline patch fixes it again.
> 
>    text    data     bss     dec     hex filename
> 6721720  624416  397508 7743644  76289c vmlinux-gcc-4.2.3

OK, so it's an unfixed bug in gcc, and it looks to be specific to the
alpha backend.  if the compiler chooses to inline a function, it has to
respect the sectional (and other constraints) of the outer function.
Can someone file a bug report with the relevant compiler people?

Jmaes



  reply	other threads:[~2008-05-12 14:01 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-05-11 15:19 Build error: `.exit.text' referenced in section `.rodata' Thorsten Kranzkowski
2008-05-11 16:55 ` James Bottomley
2008-05-11 19:20   ` Thorsten Kranzkowski
2008-05-12 11:52     ` Thorsten Kranzkowski
2008-05-12 14:01       ` James Bottomley [this message]
2008-05-13 14:31         ` Ivan Kokshaysky
2008-05-13 14:59           ` James Bottomley
2008-05-13 20:06             ` Ivan Kokshaysky
2008-05-13 20:51               ` James Bottomley

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=1210600904.3078.1.camel@localhost.localdomain \
    --to=james.bottomley@hansenpartnership.com \
    --cc=dl8bcu@dl8bcu.de \
    --cc=ink@jurassic.park.msu.ru \
    --cc=linux-alpha@vger.kernel.org \
    --cc=linux-scsi@vger.kernel.org \
    --cc=rth@twiddle.net \
    /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