[PATCH 1/2] mem: multiple resource support allow exclude a resource

[Jean-Christophe PLAGNIOL-VILLARD]

On 09:24 Fri 19 Nov     , Sascha Hauer wrote:
> On Wed, Nov 17, 2010 at 02:59:24PM +0100, Jean-Christophe PLAGNIOL-VILLARD wrote:
> > introduce DEVFS_MEM_BAREBOX_ONLY for this purpose
> > 
> > Signed-off-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj at jcrosoft.com>
> > ---
> >  arch/arm/lib/armlinux.c |    4 ++++
> >  include/driver.h        |    1 +
> >  2 files changed, 5 insertions(+), 0 deletions(-)
> > 
> > diff --git a/arch/arm/lib/armlinux.c b/arch/arm/lib/armlinux.c
> > index b74c5e8..55d1401 100644
> > --- a/arch/arm/lib/armlinux.c
> > +++ b/arch/arm/lib/armlinux.c
> > @@ -79,6 +79,10 @@ static void setup_memory_tags(void)
> >  	list_for_each_entry(mem, &memory_list, list) {
> >  		for (i = 0; i < mem->dev->num_resources; i++) {
> >  			res = &mem->dev->resource[i];
> > +
> > +			if (res->flags & DEVFS_MEM_BAREBOX_ONLY)
> > +				break;
> 
> Shouldn't this be a 'continue'?
yeap
> 
> > +
> >  			params->hdr.tag = ATAG_MEM;
> >  			params->hdr.size = tag_size(tag_mem32);
> >  
> > diff --git a/include/driver.h b/include/driver.h
> > index c7dce1e..e98455c 100644
> > --- a/include/driver.h
> > +++ b/include/driver.h
> > @@ -354,6 +354,7 @@ ssize_t cdev_write(struct cdev *cdev, const void *buf, size_t count, ulong offse
> >  #define DEVFS_PARTITION_READONLY	(1 << 1)
> >  #define DEVFS_IS_PARTITION		(1 << 2)
> >  #define DEVFS_RDWR			(1 << 3)
> > +#define DEVFS_MEM_BAREBOX_ONLY		(1 << 4)
> 
> I realize this when looking at this patch: You should abuse the flags
> field in a resource pass custom bits. The only flags in a resource should
> be the ones defined in include/linux/ioport.h.
> With DEVFS_RDWR you are lucky, there is a IORESOURCE_MEM_WRITEABLE flag
> which can be used for this purpose. There is no correspondent flag for
> DEVFS_MEM_BAREBOX_ONLY though.
ok I'll update
> 
> Overall I'm not very happy with the multiple-resources-to-mem-driver
> approach. If you are concerned with the overhead of multiple statically
> allocated devices we could introduce a add_memory_device(char *name, void *start,
> size_t size, unsigned long flags) function which dynamically allocates a
> device.
> 
> I'm also not convinced that these few multiple statically allocated
> devices introduce an overhead at all, I mean most boards only have
> SDRAM (one device) and maybe an SRAM (second device).
no preference just want to avoid multiple device but if u prefer I let it as
is

Best Regards,
J.