Re: [PATCH 2/4] dmi: Add a DMI firmware node and handling

[Corey Minyard]

On 02/02/2016 12:25 PM, Andy Lutomirski wrote:
> On Feb 2, 2016 5:37 AM, "Corey Minyard" <minyard@xxxxxxx> wrote:
> > On 02/01/2016 03:25 AM, Jean Delvare wrote:
> > > Hi Corey,
> > >
> > > I won't comment on the IPMI side of this as this isn't my area. However
> > > I have a comment on the DMI part:
> > >
> > > Le Friday 29 January 2016 Ã 16:43 -0600, minyard@xxxxxxx a Ãcrit :
> > > > From: Corey Minyard <cminyard@xxxxxxxxxx>
> > > >
> > > > This is so that an IPMI platform device can be created from a
> > > > DMI firmware entry.
> > > >
> > > > Signed-off-by: Corey Minyard <cminyard@xxxxxxxxxx>
> > > > Cc: Jean Delvare <jdelvare@xxxxxxx>
> > > > Cc: Andy Lutomirski <luto@xxxxxxxxxx>
> > > > ---
> > > >    drivers/firmware/dmi_scan.c | 34 ++++++++++++++++++++++++----------
> > > >    include/linux/dmi.h         | 24 ++++++++++++++++++++++++
> > > >    include/linux/fwnode.h      |  1 +
> > > >    3 files changed, 49 insertions(+), 10 deletions(-)
> > > >
> > > > diff --git a/drivers/firmware/dmi_scan.c b/drivers/firmware/dmi_scan.c
> > > > index da471b2..13d9bca 100644
> > > > --- a/drivers/firmware/dmi_scan.c
> > > > +++ b/drivers/firmware/dmi_scan.c
> > > > @@ -41,6 +41,16 @@ static struct dmi_memdev_info {
> > > >    } *dmi_memdev;
> > > >    static int dmi_memdev_nr;
> > > >    +static void *dmi_zalloc(unsigned len)
> > > > +{
> > > > +       void *ret = dmi_alloc(len);
> > > > +
> > > > +       if (ret)
> > > > +               memset(ret, 0, len);
> > > > +
> > > > +       return ret;
> > > > +}
> > > > +
> > > >    static const char * __init dmi_string_nosave(const struct dmi_header *dm, u8 s)
> > > >    {
> > > >          const u8 *bp = ((u8 *) dm) + dm->length;
> > > > @@ -242,6 +252,12 @@ static void __init dmi_save_type(const struct dmi_header *dm, int slot,
> > > > (...)
> > > > @@ -250,15 +266,14 @@ static void __init dmi_save_one_device(int type, const char *name)
> > > >          if (dmi_find_device(type, name, NULL))
> > > >                  return;
> > > >    -     dev = dmi_alloc(sizeof(*dev) + strlen(name) + 1);
> > > > +       dev = dmi_zalloc(sizeof(*dev) + strlen(name) + 1);
> > > >          if (!dev)
> > > >                  return;
> > > >          dev->type = type;
> > > >          strcpy((char *)(dev + 1), name);
> > > >          dev->name = (char *)(dev + 1);
> > > > -       dev->device_data = NULL;
> > > This change seems rather unrelated, and I'm not sure what purpose it
> > > serves. On ia64 and arm64 it is clearly redundant as dmi_alloc calls
> > > kzalloc directly. On x86_64, extend_brk is called instead (don't ask me
> > > why, I have no clue) but looking at the code I see that it does
> > > memset(ret, 0, size) as well so memory is also zeroed there. Which makes
> > > dmi_alloc the same as dmi_zalloc on all 3 architectures.
> > >
> > > So please revert this change. This will make your patch easier to
> > > review, too.
> > Ok.  I had assumed extend_break wasn't zeroing since there were all the NULL assignments,
> > I should have looked.
> >
> > I was thinking about this, and the fwnode could just be added to the IPMI device.  I'm not
> > sure if you would prefer that over adding it to dmi_device.  The fwnode is in acpi_device,
> > and I was modelling these changes after that, but maybe that's not required here.
> I think dmi_device is right, especially if your patches result in a
> firmware_node sysfs link being created.  That way the link will point
> to the right place.

Yeah, that's the conclusion I had come to, I think.  It doesn't 
currently add the
firmware_node to sysfs, but that's easily added and probably a next logical
step.

I'll have a new set of patches out today after I compile test at each step.

Thanks,

-corey