RE: [PATCH] tpm: tpm_crb: relinquish locality on error path.

["Winkler, Tomas" <tomas.winkler@intel.com>]

> > On Tue, 2018-04-10 at 09:00 +0000, Winkler, Tomas wrote:
> > > >
> > > > On Sat, 2018-04-07 at 19:12 +0300, Tomas Winkler wrote:
> > > > > In crb_map_io() function, __crb_request_locality() is called
> > > > > prior to crb_cmd_ready(), but if one of the consecutive function
> > > > > fails the flow bails out instead of trying to relinquish locality.
> > > > > This patch adds goto jump to __crb_relinquish_locality() on the
> > > > > error path.
> > > > >
> > > > > Fixes: 888d867df441 (tpm: cmd_ready command can be issued only
> > > > > after granting
> > > > > locality)
> > > > > Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
> > > > > ---
> > > > >  drivers/char/tpm/tpm_crb.c | 10 +++++++---
> > > > >  1 file changed, 7 insertions(+), 3 deletions(-)
> > > > >
> > > > > diff --git a/drivers/char/tpm/tpm_crb.c
> > > > > b/drivers/char/tpm/tpm_crb.c index 7f78482cd157..34fbc6cb097b
> > > > > 100644
> > > > > --- a/drivers/char/tpm/tpm_crb.c
> > > > > +++ b/drivers/char/tpm/tpm_crb.c
> > > > > @@ -511,8 +511,10 @@ static int crb_map_io(struct acpi_device
> > > > > *device, struct crb_priv *priv,
> > > > >
> > > > >  	priv->regs_t = crb_map_res(dev, priv, &io_res, buf-
> > > > > >control_address,
> > > > >  				   sizeof(struct crb_regs_tail));
> > > > > -	if (IS_ERR(priv->regs_t))
> > > > > -		return PTR_ERR(priv->regs_t);
> > > > > +	if (IS_ERR(priv->regs_t)) {
> > > > > +		ret = PTR_ERR(priv->regs_t);
> > > > > +		goto out_relinquish_locality;
> > > > > +	}
> > > > >
> > > > >  	/*
> > > > >  	 * PTT HW bug w/a: wake up the device to access @@ -520,7
> > > > > +522,7
> > > >
> > > > @@
> > > > > static int crb_map_io(struct acpi_device *device, struct
> > > > > crb_priv *priv,
> > > > >  	 */
> > > > >  	ret = crb_cmd_ready(dev, priv);
> > > > >  	if (ret)
> > > > > -		return ret;
> > > > > +		goto out_relinquish_locality;
> > > > >
> > > > >  	pa_high = ioread32(&priv->regs_t->ctrl_cmd_pa_high);
> > > > >  	pa_low  = ioread32(&priv->regs_t->ctrl_cmd_pa_low);
> > > > > @@ -565,6 +567,8 @@ static int crb_map_io(struct acpi_device
> > > > > *device, struct crb_priv *priv,
> > > > >
> > > > >  	crb_go_idle(dev, priv);
> > > > >
> > > > > +out_relinquish_locality:
> > > > > +
> > > > >  	__crb_relinquish_locality(dev, priv, 0);
> > > > >
> > > > >  	return ret;
> > > >
> > > > Thanks, please just call it before returning in the error path.
> > >
> > > Can you please elaborate why, isn't the centralized exiting of
> > > functions preferred kernel coding style?
> > > https://www.kernel.org/doc/html/v4.11/process/coding-style.html#cent
> > > ra
> > > lized-ex
> > > iting-of-functions
> >
> > You exit only from one location (not multiple) and not from a nested
> > context. Here you just add more complexity by doing this.
> 
> Where is the complexity ? I see it as a standard way of undoing on exit.
> Tomas

Jarkko, can you please respond.
Thanks
Tomas