Date:   Wed, 26 May 2021 12:53:05 +0530
From:   nitirawa@...eaurora.org
To:     Bjorn Andersson <bjorn.andersson@...aro.org>
Cc:     asutoshd@...eaurora.org, cang@...eaurora.org,
        stummala@...eaurora.org, vbadigan@...eaurora.org,
        alim.akhtar@...sung.com, avri.altman@....com, jejb@...ux.ibm.com,
        martin.petersen@...cle.com, stanley.chu@...iatek.com,
        beanhuo@...ron.com, adrian.hunter@...el.com, bvanassche@....org,
        linux-scsi@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH V2 3/3] scsi: ufs-qcom: configure VCC voltage level in
 vendor file

On 2021-04-01 20:42, Bjorn Andersson wrote:
> On Thu 01 Apr 09:58 CDT 2021, nitirawa@...eaurora.org wrote:
> 
>> On 2021-03-31 23:49, Bjorn Andersson wrote:
>> > On Wed 24 Mar 16:55 CDT 2021, nitirawa@...eaurora.org wrote:
>> >
>> > > On 2021-03-23 20:58, Bjorn Andersson wrote:
>> > > > On Sun 21 Mar 16:57 CDT 2021, Nitin Rawat wrote:
>> > > >
>> > > > > As a part of vops handler, VCC voltage is updated
>> > > > > as per the ufs device probed after reading the device
>> > > > > descriptor. We follow below steps to configure voltage
>> > > > > level.
>> > > > >
>> > > > > 1. Set the device to SLEEP state.
>> > > > > 2. Disable the Vcc Regulator.
>> > > > > 3. Set the vcc voltage according to the device type and reenable
>> > > > >    the regulator.
>> > > > > 4. Set the device mode back to ACTIVE.
>> > > > >
>> > > >
>> > > > When we discussed this a while back this was described as a requirement
>> > > > from the device specification, you only operate on objects "owned" by
>> > > > ufshcd and you invoke ufshcd operations to perform the actions.
>> > > >
>> > > > So why is this a ufs-qcom patch and not something in ufshcd?
>> > > >
>> > > > Regards,
>> > > > Bjorn
>> > > >
>> > > > > Signed-off-by: Nitin Rawat <nitirawa@...eaurora.org>
>> > > > > Signed-off-by: Veerabhadrarao Badiganti <vbadigan@...eaurora.org>
>> > > > > ---
>> > > > >  drivers/scsi/ufs/ufs-qcom.c | 51
>> > > > > +++++++++++++++++++++++++++++++++++++++++++++
>> > > > >  1 file changed, 51 insertions(+)
>> > > > >
>> > > > > diff --git a/drivers/scsi/ufs/ufs-qcom.c b/drivers/scsi/ufs/ufs-qcom.c
>> > > > > index f97d7b0..ca35f5c 100644
>> > > > > --- a/drivers/scsi/ufs/ufs-qcom.c
>> > > > > +++ b/drivers/scsi/ufs/ufs-qcom.c
>> > > > > @@ -21,6 +21,17 @@
>> > > > >  #define UFS_QCOM_DEFAULT_DBG_PRINT_EN	\
>> > > > >  	(UFS_QCOM_DBG_PRINT_REGS_EN | UFS_QCOM_DBG_PRINT_TEST_BUS_EN)
>> > > > >
>> > > > > +#define	ANDROID_BOOT_DEV_MAX	30
>> > > > > +static char android_boot_dev[ANDROID_BOOT_DEV_MAX];
>> > > > > +
>> > > > > +/* Min and Max VCC voltage values for ufs 2.x and
>> > > > > + * ufs 3.x devices
>> > > > > + */
>> > > > > +#define UFS_3X_VREG_VCC_MIN_UV	2540000 /* uV */
>> > > > > +#define UFS_3X_VREG_VCC_MAX_UV	2700000 /* uV */
>> > > > > +#define UFS_2X_VREG_VCC_MIN_UV	2950000 /* uV */
>> > > > > +#define UFS_2X_VREG_VCC_MAX_UV	2960000 /* uV */
>> > > > > +
>> > > > >  enum {
>> > > > >  	TSTBUS_UAWM,
>> > > > >  	TSTBUS_UARM,
>> > > > > @@ -1293,6 +1304,45 @@ static void
>> > > > > ufs_qcom_print_hw_debug_reg_all(struct ufs_hba *hba,
>> > > > >  	print_fn(hba, reg, 9, "UFS_DBG_RD_REG_TMRLUT ", priv);
>> > > > >  }
>> > > > >
>> > > > > +  /**
>> > > > > +   * ufs_qcom_setup_vcc_regulators - Update VCC voltage
>> > > > > +   * @hba: host controller instance
>> > > > > +   * Update VCC voltage based on UFS device(ufs 2.x or
>> > > > > +   * ufs 3.x probed)
>> > > > > +   */
>> > > > > +static int ufs_qcom_setup_vcc_regulators(struct ufs_hba *hba)
>> > > > > +{
>> > > > > +	struct ufs_dev_info *dev_info = &hba->dev_info;
>> > > > > +	struct ufs_vreg *vreg = hba->vreg_info.vcc;
>> > > > > +	int ret;
>> > > > > +
>> > > > > +	/* Put the device in sleep before lowering VCC level */
>> > > > > +	ret = ufshcd_set_dev_pwr_mode(hba, UFS_SLEEP_PWR_MODE);
>> > > > > +
>> > > > > +	/* Switch off VCC before switching it ON at 2.5v or 2.96v */
>> > > > > +	ret = ufshcd_disable_vreg(hba->dev, vreg);
>> > > > > +
>> > > > > +	/* add ~2ms delay before renabling VCC at lower voltage */
>> > > > > +	usleep_range(2000, 2100);
>> > > > > +
>> > > > > +	/* set VCC min and max voltage according to ufs device type */
>> > > > > +	if (dev_info->wspecversion >= 0x300) {
>> > > > > +		vreg->min_uV = UFS_3X_VREG_VCC_MIN_UV;
>> > > > > +		vreg->max_uV = UFS_3X_VREG_VCC_MAX_UV;
>> > > > > +	}
>> > > > > +
>> > > > > +	else {
>> > > > > +		vreg->min_uV = UFS_2X_VREG_VCC_MIN_UV;
>> > > > > +		vreg->max_uV = UFS_2X_VREG_VCC_MAX_UV;
>> > > > > +	}
>> > > > > +
>> > > > > +	ret = ufshcd_enable_vreg(hba->dev, vreg);
>> > > > > +
>> > > > > +	/* Bring the device in active now */
>> > > > > +	ret = ufshcd_set_dev_pwr_mode(hba, UFS_ACTIVE_PWR_MODE);
>> > > > > +	return ret;
>> > > > > +}
>> > > > > +
>> > > > >  static void ufs_qcom_enable_test_bus(struct ufs_qcom_host *host)
>> > > > >  {
>> > > > >  	if (host->dbg_print_en & UFS_QCOM_DBG_PRINT_TEST_BUS_EN) {
>> > > > > @@ -1490,6 +1540,7 @@ static const struct ufs_hba_variant_ops
>> > > > > ufs_hba_qcom_vops = {
>> > > > >  	.device_reset		= ufs_qcom_device_reset,
>> > > > >  	.config_scaling_param = ufs_qcom_config_scaling_param,
>> > > > >  	.program_key		= ufs_qcom_ice_program_key,
>> > > > > +	.setup_vcc_regulators	= ufs_qcom_setup_vcc_regulators,
>> > > > >  };
>> > > > >
>> > > > >  /**
>> > > > > --
>> > > > > 2.7.4
>> > > > >
>> > >
>> > > Hi Bjorn,
>> > > Thanks for your review.
>> > > But As per the earlier discussion regarding handling of vcc voltage
>> > > for platform supporting both ufs 2.x and ufs 3.x , it was finally
>> > > concluded
>> > > to
>> > > use "vops and let vendors handle it, until specs or someone
>> > > has a better suggestion". Please correct me in case i am wrong.
>> > >
>> >
>> > I was under the impression that this would result in something custom
>> > per platform, but what I'm objecting to now that I see the code is that
>> > this is completely generic.
>> >
>> > And the concerns we discussed regarding these regulators being shared
>> > with other devices is not considered in this implementation. But in
>> > practice I don't see how you could support 2.x, 3.x and rail sharing at
>> > the same time.
>> >
>> > Regards,
>> > Bjorn
>> >
>> > > https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg2399116.html
>> > >
>> > > Regards,
>> > > Nitin
>> 
>> Hi Bjorn,
>> Thanks for your feedback.
>> Regarding your query for regulator being shared with other device,
>> Imho, the soc/pmic designer should share only those device
>> with ufs regulator which has the same voltage range (2.4-3.6v).
>> If that is not considered by the pmic designer,
>> wouldn't that would be a board design issue ???
>> 
> 
> It's not only that the rail needs to stay within 2.4-3.6V, depending on
> operating mode of this device it either need to be at 2.54-2.7V or
> 2.95-2.96V depending on wspecversion for this instance.
> 
> So either that other device need to be completely flexible in that 
> range
> and support the voltage jumping between them without notice, or such
> design isn't possible.
> 
> And as you say, that would be something that the hardware designers
> would need to handle for us.
> 
>> And I agree with you that - the code looks generic but
>> since the below steps is not part of the specs,
>> I had to keep it in vendor specific file for which I
>> had to export few api from ufshcd.c to use in vendor
>> specific files.
>> 
>> 1. Set the device to SLEEP state.
>> 2. Disable the Vcc Regulator.
>> 3. Set the vcc voltage according to the device type and reenable
>>    the regulator.
>> 4. Set the device mode back to ACTIVE.
>> 
>> Please correct me if my understanding is not correct.
>> 
> 
> Are you saying that steps 1 to 4 here are not defined in the
> specification and therefor Qualcomm specific? Do we expect other 
> vendors
> to not follow this sequence, or do they simply not have these voltage
> constraints?
> 
> And again, isn't this the voltage for the attached UFS device? (Rather
> than a Qualcomm thing)
> 
> Regards,
> Bjorn


Hi Bjorn,
Sorry for quite late reply.
Yes Bjorn above steps(1-4) are not mentioned in the specs. But 
definitely other
vendor can follow the same steps . If no vendor have any concerns,
I can put these steps as generic in ufshcd.c file.
Let me know what's you opinion on this ??

Thanks,
Nitin

Powered by blists - more mailing lists