Re: [PATCH V1 3/3] scsi: ufs: qcom: Add support to disable UFS LPM Feature

[Rob Herring <robh@kernel.org>]

On Wed, Apr 23, 2025 at 01:14:27AM +0530, Nitin Rawat wrote:
> 
> 
> On 4/22/2025 6:15 PM, Rob Herring wrote:
> > On Thu, Apr 17, 2025 at 06:16:45PM +0530, Nitin Rawat wrote:
> > > There are emulation FPGA platforms or other platforms where UFS low
> > > power mode is either unsupported or power efficiency is not a critical
> > > requirement.
> > > 
> > > Disable all low power mode UFS feature based on the "disable-lpm" device
> > > tree property parsed in platform driver.
> > > 
> > > Signed-off-by: Nitin Rawat <quic_nitirawa@quicinc.com>
> > > ---
> > >   drivers/ufs/host/ufs-qcom.c | 15 ++++++++-------
> > >   1 file changed, 8 insertions(+), 7 deletions(-)
> > > 
> > > diff --git a/drivers/ufs/host/ufs-qcom.c b/drivers/ufs/host/ufs-qcom.c
> > > index 1b37449fbffc..1024edf36b68 100644
> > > --- a/drivers/ufs/host/ufs-qcom.c
> > > +++ b/drivers/ufs/host/ufs-qcom.c
> > > @@ -1014,13 +1014,14 @@ static void ufs_qcom_set_host_caps(struct ufs_hba *hba)
> > > 
> > >   static void ufs_qcom_set_caps(struct ufs_hba *hba)
> > >   {
> > > -	hba->caps |= UFSHCD_CAP_CLK_GATING | UFSHCD_CAP_HIBERN8_WITH_CLK_GATING;
> > > -	hba->caps |= UFSHCD_CAP_CLK_SCALING | UFSHCD_CAP_WB_WITH_CLK_SCALING;
> > > -	hba->caps |= UFSHCD_CAP_AUTO_BKOPS_SUSPEND;
> > > -	hba->caps |= UFSHCD_CAP_WB_EN;
> > > -	hba->caps |= UFSHCD_CAP_AGGR_POWER_COLLAPSE;
> > > -	hba->caps |= UFSHCD_CAP_RPM_AUTOSUSPEND;
> > > -
> > > +	if (!hba->disable_lpm) {
> > > +		hba->caps |= UFSHCD_CAP_CLK_GATING | UFSHCD_CAP_HIBERN8_WITH_CLK_GATING;
> > > +		hba->caps |= UFSHCD_CAP_CLK_SCALING | UFSHCD_CAP_WB_WITH_CLK_SCALING;
> > > +		hba->caps |= UFSHCD_CAP_AUTO_BKOPS_SUSPEND;
> > > +		hba->caps |= UFSHCD_CAP_WB_EN;
> > > +		hba->caps |= UFSHCD_CAP_AGGR_POWER_COLLAPSE;
> > > +		hba->caps |= UFSHCD_CAP_RPM_AUTOSUSPEND;
> > > +	}
> > 
> > Doesn't RuntimePM already have userspace controls? And that's a Linux
> > feature that shouldn't really be controlled by DT. I think this property
> > should still to things defined by the UFS spec.
> 
> Hi Rob,
> Yes userspace has runtime PM control but by the time UFS driver probes
> completes and userspace is up, there are chances runtime PM may get kicked
> in.

That sounds like a problem more than 1 device would have...

Rob