[PATCH 2/2] phy: qcom-qmp: fix pipe-clock imbalance on power-on failure

[Vinod Koul]

On 02-05-22, 12:27, Johan Hovold wrote:
> On Mon, May 02, 2022 at 03:39:41PM +0530, Vinod Koul wrote:
> > On 20-04-22, 17:23, Johan Hovold wrote:
> > > Make sure to disable the pipe clock also if ufs-reset deassertion fails
> > > during power on.
> > > 
> > > Note that the ufs-reset is asserted in qcom_qmp_phy_com_exit().
> > > 
> > > Fixes: c9b589791fc1 ("phy: qcom: Utilize UFS reset controller")
> > > Cc: Evan Green <evgreen at chromium.org>
> > > Signed-off-by: Johan Hovold <johan+linaro at kernel.org>
> > > ---
> > >  drivers/phy/qualcomm/phy-qcom-qmp.c | 14 +++++++-------
> > >  1 file changed, 7 insertions(+), 7 deletions(-)
> > > 
> > > diff --git a/drivers/phy/qualcomm/phy-qcom-qmp.c b/drivers/phy/qualcomm/phy-qcom-qmp.c
> > > index 8c2300bfe489..7d2d1ab061f7 100644
> > > --- a/drivers/phy/qualcomm/phy-qcom-qmp.c
> > > +++ b/drivers/phy/qualcomm/phy-qcom-qmp.c
> > > @@ -5375,14 +5375,14 @@ static int qcom_qmp_phy_power_on(struct phy *phy)
> > >  		if (ret) {
> > >  			dev_err(qmp->dev, "lane%d reset deassert failed\n",
> > >  				qphy->index);
> > > -			goto err_lane_rst;
> > > +			return ret;
> > 
> > This can be skipped if we retain the err_lane_rst label
> > 
> > >  		}
> > >  	}
> > >  
> > >  	ret = clk_prepare_enable(qphy->pipe_clk);
> > >  	if (ret) {
> > >  		dev_err(qmp->dev, "pipe_clk enable failed err=%d\n", ret);
> > > -		goto err_clk_enable;
> > > +		goto err_reset_lane;
> > >  	}
> > >  
> > >  	/* Tx, Rx, and PCS configurations */
> > > @@ -5433,7 +5433,7 @@ static int qcom_qmp_phy_power_on(struct phy *phy)
> > >  
> > >  	ret = reset_control_deassert(qmp->ufs_reset);
> > >  	if (ret)
> > > -		goto err_lane_rst;
> > > +		goto err_disable_pipe_clk;
> > 
> > this is the actual fix...
> 
> Right, but with a one-line fix this would read
> 
> 	goto err_pcs_ready;
> 
> which makes no sense at all. Is that really what you prefer?

Yes I feel for a fix we should always have the minimal changes.. makes
porting easier

> 
> Renaming just this one and the below one, would leave the error labels
> named using two different schemes, which also isn't very nice, but
> perhaps ok.
> 
> > >  
> > >  	qcom_qmp_phy_configure(pcs_misc, cfg->regs, cfg->pcs_misc_tbl,
> > >  			       cfg->pcs_misc_tbl_num);
> > > @@ -5472,17 +5472,17 @@ static int qcom_qmp_phy_power_on(struct phy *phy)
> > >  					 PHY_INIT_COMPLETE_TIMEOUT);
> > >  		if (ret) {
> > >  			dev_err(qmp->dev, "phy initialization timed-out\n");
> > > -			goto err_pcs_ready;
> > > +			goto err_disable_pipe_clk;
> > 
> > same rename here
> > 
> > >  		}
> > >  	}
> > >  	return 0;
> > >  
> > > -err_pcs_ready:
> > > +err_disable_pipe_clk:
> > >  	clk_disable_unprepare(qphy->pipe_clk);
> > > -err_clk_enable:
> > > +err_reset_lane:
> > >  	if (cfg->has_lane_rst)
> > >  		reset_control_assert(qphy->lane_rst);
> > > -err_lane_rst:
> > > +
> > >  	return ret;
> > 
> > while I feel the names given by this patch are better, they should not
> > be in a fix patch. We should just add the one line fix here and do
> > renames later
> 
> I'll respin if you prefer, just want to double check that you really
> want a one line fix (i.e. goto err_pcs_ready).

Yes please


-- 
~Vinod