xfs
[Top] [All Lists]

Re: Upgrading kernel, Compile error in module

To: Micah Yoder <yodermk@xxxxxxxx>
Subject: Re: Upgrading kernel, Compile error in module
From: Steve Lord <lord@xxxxxxx>
Date: Mon, 29 Oct 2001 09:26:41 -0600
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: Message from Micah Yoder <yodermk@xxxxxxxx> of "Sat, 27 Oct 2001 16:42:51 EDT." <20011027234239.VKNQ5793.femail48.sdc1.sfba.home.com@there>
Sender: owner-linux-xfs@xxxxxxxxxxx
> Hi again.  Decided it was finally time to upgrade my 2.4.6pre2-xfs, which has
>  
> been working great, for security and VM fixes.  Just grabbed latest CVS an 
> hour or two ago (2.4.14pre3).
> 
> Thing that makes me nervous is that there was a minor compile error in gvim 
> ./drivers/block/paride/pcd.c:
> 
> static struct block_device_operations pcd_bdops = {
>       owner:                  THIS_MODULE,
>       open:                   cdrom_open,
>       release:                cdrom_release,
>       ioctl:                  cdrom_ioctl,
>       check_media_change:     cdrom_media_changed,
> }   /* line 274 */
> 
> static struct cdrom_device_ops pcd_dops = {
>       pcd_open,
>       pcd_release,
> [...]
> 
> Note the lack of a semicolon on 274!!!  I added it and now it compiles.  
> Dunno if that booboo was inherited from Linus or from one of you, but you 
> might want to fix it. :-)
> 

It was inherited from Linus, most of the files (99%) in the tree are
unmodified by us. In general with this sort of problem if it is not fixed
in the latest pre-xxx kernel from Linus, submit a patch to Linux kernel
and Linus. Unless a fix affects mainline kernel functionality I try to keep
our tree identical to Linus's.

> So that's fixed, but the fact that it was there worries me some!  What other 
> nasties besides compile errors could be in there?
> 

There was a change went into the 2.4.13-pre3 which affects how block device
modules do reference counting to avoid module unload being performed on
running code. Most of this code change was probably a global cut and
paste, a few errors can easily creep into a change like this and they
usually get sorted out fairly quickly.


> 
> Think it's "safe" to put this thing on a production server?  Again, my 
> "problem" is that it's co-located, 500 miles away, so a mistake could be 
> costly...  but if it works for most people I'm willing to try it.

In general I would not put a pre-xxx kernel up in production unless you
yourself have done fairly extensive testing. The 2.4.13 kernel itself
seems fairly robust, there is a patch for this on the ftp site:

        ftp://oss.sgi.com/projects/xfs/download/patches


Steve


> 
> Thanks,
> Micah
> 
> -- 
> Like to travel?                        http://TravTalk.org
> Micah Yoder Internet Development       http://yoderdev.com



<Prev in Thread] Current Thread [Next in Thread>