Re: can a kmalloc be both GFP_ATOMIC and GFP_KERNEL at the same time?

[Alan Cox <alan@lxorguk.ukuu.org.uk>]

> > > > drivers/scsi/aic7xxx_old.c:  aic_dev = kmalloc(sizeof(struct aic_dev_data), GFP_ATOMIC | GFP_KERNEL);
> > > > drivers/message/i2o/device.c:   resblk = kmalloc(buflen + 8, GFP_KERNEL | GFP_ATOMIC);
> > > >
> > > >   clarification?
> > >
> > > GFP_ATOMIC implies that the memory comes from the zones which
> > > GFP_KERNEL also uses.  So the above usage of GFP_KERNEL is redundant
> > > and should be removed.
> > 
> > hang on ... based on an email i just got, is that reference to
> > GFP_KERNEL "redundant" or "conflicting"?  big difference there.  and
> > is the proper fix to remove "GFP_KERNEL" in both cases?
> > 
> 
> umm, yeah, oops.  GFP_KERNEL|GFP_ATOMIC is not a redundant combination. 
> It's GFP_KERNEL plus "is able to access emergency pools".  We'd normally
> represent that as GFP_KERNEL|__GFP_HIGH.

i2o/device.c should be GFP_KERNEL as far as I can tell. It was meant to
be that way and the callers appear to all be calling it in sleep capable
contexts.

aic7xxx_old.c should probably be GFP_KERNEL as ->slave_alloc methods
appear to be able to sleep (although some drivers use GFP_ATOMIC here and
some GFP_KERNEL).

Alan