Date:	Wed, 28 Nov 2007 11:45:43 -0800
From:	Greg KH <greg@...ah.com>
To:	darnok@....org
Cc:	linux-kernel@...r.kernel.org, pjones@...hat.com,
	konradr@...hat.com, konradr@...ux.vnet.ibm.com,
	randy.dunlap@...cle.com, hpa@...or.com, lenb@...nel.org,
	mike.anderson@...ibm.com, dwm@...tin.ibm.com
Subject: Re: [PATCH] Add iSCSI IBFT Support (v0.3)

On Wed, Nov 28, 2007 at 03:21:40PM -0400, darnok@....org wrote:
> On Tue, Nov 27, 2007 at 11:09:19AM -0800, Greg KH wrote:
> > On Tue, Nov 27, 2007 at 02:09:50PM -0400, darnok@....org wrote:
> > > On Mon, Nov 26, 2007 at 09:29:55PM -0800, Greg KH wrote:
> > > > On Mon, Nov 26, 2007 at 11:23:31PM -0500, Konrad Rzeszutek wrote:
> > > > > On Monday 26 November 2007 22:31:38 Greg KH wrote:
> > > > > > > +#if defined(CONFIG_ISCSI_IBFT) || defined(CONFIG_ISCSI_IBFT_MODULE)
> > > > > ..snip..
> > > > > > > +static ssize_t find_ibft(void)
> > > > > > > +{
> > > > > ..snip..
> > > > > > > +}
> > > > > >
> > > > > > What is a function (not even an inline one) doing in a .h file?
> > > > > 
> > > > > I was not sure where to put it. This function (find_ibft) is used by the 
> > > > > setup_[32|64].c and the iscsi_ibft.c code. Randy suggested I put in .c file, 
> > > > > but I am not sure exactly where? Should I make a new file in called 
> > > > > libs/iscsi_ibft_helper.c ?
> > > > 
> > > > Put it in your .c file and make it a global function to be called by
> > > > someone else if they need it.
> > > 
> > > If the kernel is built with CONFIG_ISCSI_IBFT=m, the
> > > setup_[32|64],c code would depend on the 'find_ibft' symbol which is
> > > in a module (in the iscsi_ibft.c), which is not available during
> > > the bootup phase and not linked to vmlinuz.
> > 
> > Ah, then don't allow that :)
> > 
> > > This isn't an issue if the module is built with CONFIG_ISCSI_IBFT=y of
> > > course.
> > > 
> > > Or did by 'your .c file' mean a new file in arch/x86/kernel directory?
> > 
> > I didn't realize an external file, outside of your changes, needed this
> > function.  If it does, then perhaps you need to just place it elsewhere.
> 
> The fundamental problem is that 'find_ibft' ought to be available
> from anywhere (or at least from the iscsi_ibft.c) so that the iscsi_ibft
> module can be loaded on any platform. But on x86, it needs to be called
>  from setup_[32|64].c because the IBFT can be located anywhere
> between 512KB and 1MB - and the E820 does not neccesarily have to
> exclude that region. Hence the patch I proposed implements a
> 'reserve_ibft_region' code which would reserve the region of memory
> found by 'find_ibft' so that it can be preserved when iscsi_ibft
> module is actually loaded.
> 
> It ends up that there are three consumers of 'find_ibft':
>  a) the module itself (iscsi_ibft.c)
>  b) setup_32.c
>  c) setup_64.c
> 
> The first choice, which looked the most flexible, was to have it
> in iscsi_ibft.h file.
> The second one, which would inhibit the user from making iscsi_ibft
> a module, would be to move it to iscsi_ibft.c and make it
> EXPORT_SYMBOL(), but that seems wasteful from a memory foot-print
> look.
> A third option was to put in /lib, but that doesn't seem right - this
> 'find_ibft' code is specific to this module.
> 
> Of all the options, the cleanest looks to be the first choice :-(
> (I am not trying to be obstinate here - I just can't think of any
> other reasonable place).

If you insist on putting it in a .h file, it needs to be marked "inline"
at the least.

But, why not just put it in a separate file, that is built in if the
user wants iscsi support?  That way the setup code can call it properly
if needed.

thanks,

greg k-h
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists