On Sun, Oct 09, 2016 at 09:32:32PM +0300, Jarkko Sakkinen wrote: > On Sun, Oct 09, 2016 at 10:49:05AM -0600, Jason Gunthorpe wrote: > > On Sun, Oct 09, 2016 at 12:38:18PM +0300, Jarkko Sakkinen wrote: > > > On Sat, Oct 08, 2016 at 07:42:56PM -0600, Jason Gunthorpe wrote: > > > > On Sun, Oct 09, 2016 at 03:15:09AM +0300, Jarkko Sakkinen wrote: > > > > > + ctrl = crb_map_res(dev, priv, &io_res, buf->control_address, > > > > > + sizeof(struct crb_regs) - > > > > > + offsetof(struct crb_regs, ctrl_req)); > > > > > + if (IS_ERR(ctrl)) > > > > > + return PTR_ERR(ctrl); > > > > > + > > > > > + /* The control area always overrlaps IO memory mapped from the > > > > > ACPI > > > > > + * object with CRB start only devices. Thus, this is perfectly > > > > > safe. > > > > > + */ > > > > > + priv->regs = (void *)((unsigned long)ctrl - > > > > > + offsetof(struct crb_regs, ctrl_req)); > > > > > > > > Hum. No, this makes bad assumptions about the structure of iomapping. > > > > > > > > The map itself needs to be done with the adjustment: > > > > > > > > ctrl = crb_map_res(dev, priv, &io_res, buf->control_address - > > > > offsetof(struct crb_regs, ctrl_req), > > > > sizeof(struct crb_regs)); > > > > > > That would be wrong address for the control area as it does not start > > > from the beginning of CRB registers. > > > > Of course, I just pointed out what the map call should look like > > > > Something like this > > > > priv->regs = crb_map_res(dev, priv, &io_res, buf->control_address - > > offsetof(struct crb_regs, ctrl_req), > > sizeof(struct crb_regs)); > > ctrl = &priv->regs.ctrl_req; > > Sorry I missed this part. > > Here are the constraints for existing hardware: > > 1. All the existing CRB start only hardware has the iomem covering the > control area and registers for multiple localities. > 2. All the existing ACPI start hardware has only the control area. > > If you assume that SSDT does not have malicous behavior caused by either > a BIOS bug or maybe a rootkit, then the current patch works for all the > existing hardware. > > To counter-measure for unexpected behavior in non-existing hardware and > buggy or malicious firmware it probably make sense to use crb_map_res to > validate the part of the CRB registers that is not part of the control > area. > > Doing it in the way you proposed does not work for ACPI start devices. > > For them it should be done in the same way as I'm doing in the existing > patch as for ACPI start devices the address below the control area are > never accessed. Having a separate crb_map_res for CRB start only devices > is sane thing to do for validation.
Alternative is to do two structures crb_regs_head and crb_regs_tail, which might be cleaner. I'm fine with going either route. /Jarkko
