| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <541c95a6-0b89-fdc1-4707-e0a1eef1bf91@codeaurora.org>
Date: Thu, 18 Jun 2020 12:00:05 +0530
From: Veerabhadrarao Badiganti <vbadigan@...eaurora.org>
To: Ulf Hansson <ulf.hansson@...aro.org>
Cc: Adrian Hunter <adrian.hunter@...el.com>,
Bjorn Andersson <bjorn.andersson@...aro.org>,
"linux-mmc@...r.kernel.org" <linux-mmc@...r.kernel.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
linux-arm-msm <linux-arm-msm@...r.kernel.org>,
Asutosh Das <asutoshd@...eaurora.org>,
Vijay Viswanath <vviswana@...eaurora.org>,
Andy Gross <agross@...nel.org>
Subject: Re: [PATCH V4 2/2] mmc: sdhci-msm: Use internal voltage control
On 6/17/2020 7:46 PM, Ulf Hansson wrote:
> On Wed, 17 Jun 2020 at 14:46, Veerabhadrarao Badiganti
> <vbadigan@...eaurora.org> wrote:
>> Thanks for comments Uffe.
>>
>> On 6/17/2020 3:04 PM, Ulf Hansson wrote:
>>> On Tue, 16 Jun 2020 at 17:38, Veerabhadrarao Badiganti
>>> <vbadigan@...eaurora.org> wrote:
>>>> On qcom SD host controllers voltage switching be done after the HW
>>>> is ready for it. The HW informs its readiness through power irq.
>>>> The voltage switching should happen only then.
>>>>
>>>> Use the internal voltage switching and then control the voltage
>>>> switching using power irq.
>>>>
>>>> IO-bus supply of eMMC would be kept always-on. So set the load
>>>> for this supply to configure it in LPM when eMMC is suspend state
>>>> and in HPM when eMMC is active.
>>>>
>>>> Co-developed-by: Asutosh Das <asutoshd@...eaurora.org>
>>>> Signed-off-by: Asutosh Das <asutoshd@...eaurora.org>
>>>> Co-developed-by: Vijay Viswanath <vviswana@...eaurora.org>
>>>> Signed-off-by: Vijay Viswanath <vviswana@...eaurora.org>
>>>> Co-developed-by: Veerabhadrarao Badiganti <vbadigan@...eaurora.org>
>>>> Signed-off-by: Veerabhadrarao Badiganti <vbadigan@...eaurora.org>
>>>> ---
>>>> drivers/mmc/host/sdhci-msm.c | 208 +++++++++++++++++++++++++++++++++++++++++--
>>>> 1 file changed, 199 insertions(+), 9 deletions(-)
>>>>
>>>> diff --git a/drivers/mmc/host/sdhci-msm.c b/drivers/mmc/host/sdhci-msm.c
>>>> index 15c42b059240..8297b2142748 100644
>>>> --- a/drivers/mmc/host/sdhci-msm.c
>>>> +++ b/drivers/mmc/host/sdhci-msm.c
>>>> @@ -37,7 +37,9 @@
>>>> #define CORE_PWRCTL_IO_LOW BIT(2)
>>>> #define CORE_PWRCTL_IO_HIGH BIT(3)
>>>> #define CORE_PWRCTL_BUS_SUCCESS BIT(0)
>>>> +#define CORE_PWRCTL_BUS_FAIL BIT(1)
>>>> #define CORE_PWRCTL_IO_SUCCESS BIT(2)
>>>> +#define CORE_PWRCTL_IO_FAIL BIT(3)
>>>> #define REQ_BUS_OFF BIT(0)
>>>> #define REQ_BUS_ON BIT(1)
>>>> #define REQ_IO_LOW BIT(2)
>>>> @@ -127,6 +129,9 @@
>>>> /* Timeout value to avoid infinite waiting for pwr_irq */
>>>> #define MSM_PWR_IRQ_TIMEOUT_MS 5000
>>>>
>>>> +/* Max load for eMMC Vdd-io supply */
>>>> +#define MMC_VQMMC_MAX_LOAD_UA 325000
>>>> +
>>>> #define msm_host_readl(msm_host, host, offset) \
>>>> msm_host->var_ops->msm_readl_relaxed(host, offset)
>>>>
>>>> @@ -278,6 +283,7 @@ struct sdhci_msm_host {
>>>> bool uses_tassadar_dll;
>>>> u32 dll_config;
>>>> u32 ddr_config;
>>>> + bool vqmmc_enabled;
>>>> };
>>>>
>>>> static const struct sdhci_msm_offset *sdhci_priv_msm_offset(struct sdhci_host *host)
>>>> @@ -1346,6 +1352,88 @@ static void sdhci_msm_set_uhs_signaling(struct sdhci_host *host,
>>>> sdhci_msm_hs400(host, &mmc->ios);
>>>> }
>>>>
>>>> +static int sdhci_msm_set_vmmc(struct mmc_host *mmc)
>>>> +{
>>>> + if (IS_ERR(mmc->supply.vmmc))
>>>> + return 0;
>>>> +
>>>> + return mmc_regulator_set_ocr(mmc, mmc->supply.vmmc, mmc->ios.vdd);
>>>> +}
>>>> +
>>>> +static int msm_toggle_vqmmc(struct sdhci_msm_host *msm_host,
>>>> + struct mmc_host *mmc, bool level)
>>>> +{
>>>> + int ret;
>>>> + struct mmc_ios ios;
>>>> +
>>>> + if (msm_host->vqmmc_enabled == level)
>>>> + return 0;
>>>> +
>>>> + if (level) {
>>>> + /* Set the IO voltage regulator to default voltage level */
>>>> + if (msm_host->caps_0 & CORE_3_0V_SUPPORT)
>>>> + ios.signal_voltage = MMC_SIGNAL_VOLTAGE_330;
>>>> + else if (msm_host->caps_0 & CORE_1_8V_SUPPORT)
>>>> + ios.signal_voltage = MMC_SIGNAL_VOLTAGE_180;
>>>> +
>>>> + if (msm_host->caps_0 & CORE_VOLT_SUPPORT) {
>>>> + ret = mmc_regulator_set_vqmmc(mmc, &ios);
>>>> + if (ret < 0) {
>>>> + dev_err(mmc_dev(mmc), "%s: vqmmc set volgate failed: %d\n",
>>>> + mmc_hostname(mmc), ret);
>>>> + goto out;
>>>> + }
>>>> + }
>>>> + ret = regulator_enable(mmc->supply.vqmmc);
>>>> + } else {
>>>> + ret = regulator_disable(mmc->supply.vqmmc);
>>>> + }
>>>> +
>>>> + if (ret)
>>>> + dev_err(mmc_dev(mmc), "%s: vqmm %sable failed: %d\n",
>>>> + mmc_hostname(mmc), level ? "en":"dis", ret);
>>>> + else
>>>> + msm_host->vqmmc_enabled = level;
>>>> +out:
>>>> + return ret;
>>>> +}
>>>> +
>>>> +static int msm_config_vqmmc_mode(struct sdhci_msm_host *msm_host,
>>>> + struct mmc_host *mmc, bool hpm)
>>>> +{
>>>> + int load, ret;
>>>> +
>>>> + load = hpm ? MMC_VQMMC_MAX_LOAD_UA : 0;
>>>> + ret = regulator_set_load(mmc->supply.vqmmc, load);
>>>> + if (ret)
>>>> + dev_err(mmc_dev(mmc), "%s: vqmmc set load failed: %d\n",
>>>> + mmc_hostname(mmc), ret);
>>>> + return ret;
>>>> +}
>>>> +
>>>> +static int sdhci_msm_set_vqmmc(struct sdhci_msm_host *msm_host,
>>>> + struct mmc_host *mmc, bool level)
>>>> +{
>>>> + int ret;
>>>> + bool always_on;
>>>> +
>>>> + if (IS_ERR(mmc->supply.vqmmc) ||
>>> White space.
>>>
>>>> + (mmc->ios.power_mode == MMC_POWER_UNDEFINED))
>>>> + return 0;
>>>> + /*
>>>> + * For eMMC don't turn off Vqmmc, Instead just configure it in LPM
>>>> + * and HPM modes by setting the right amonut of load.
>>> /s/right amonut of load/corresponding load
>>>
>>>> + */
>>>> + always_on = mmc->card && mmc_card_mmc(mmc->card);
>>>> +
>>>> + if (always_on)
>>>> + ret = msm_config_vqmmc_mode(msm_host, mmc, level);
>>>> + else
>>>> + ret = msm_toggle_vqmmc(msm_host, mmc, level);
>>> I am worried that this isn't really doing what you think it does.
>>> always_on may not always be set for an eMMC.
>>>
>>> This is because the mmc->card doesn't get assigned until the
>>> initialization of the eMMC has been completed. In other words, way
>>> after VCC and VCCQ have been turned on and changed voltage levels.
>> For the very first time, i have to enable the vqmmc.
>> And second time on-wards I have to set load instead of turning vqmmc off
>> / on.
>>
>> This condition helps me in doing exactly the same.
>>
>> This gets invoked first time when mmc_power_up() is called.
>> Yes, by that time card won't be initialized. So this check fails and
>> allows to turn on the regulator instead of setting load.
>>
>> After that, this gets invoked only next time mmc_power_up/off() is called
>> which is during mmc_suspend/resume.
> I see.
>
> That assumes the initialization always works at the first attempt, so
> no re-try or power cycle gets done in some error/reset path.
>
>> By this time card is initialized, it allows to set load.
> Okay, I get it. Thanks for clarifying.
>
> Perhaps you can add a small comment about this in the code, so we
> don't forget the reasons why the code looks like this.
Sure. Will update comment with this info.
>>
>>> Moreover, mmc_card_mmc() is also about legacy MMC cards, not solely for eMMCs.
>> Okay. But on qcom platform, we have support for only eMMC/SD/SDIO.
>> So this mmc_card_mmc() meets my requirements!!! Allows only eMMC not SD
>> & SDIO.
> Well, eMMC is fairly similar to MMC, from a basic functionality point
> of view. So, I am quite sure you can get it to work with your
> controller, even if it's not officially supported.
>>> If you want special treatment of an eMMC, I think it's better to use
>>> the DT bindings Documentation/devicetree/bindings/mmc/mmc-card.txt. If
>>> you have such a subnode, that indicates that there is an eMMC card
>>> attached.
> That said, if you don't like the pure DT based solution, as I
> suggested above - at least extend mmc_card_mmc() with checking the
> MMC_CAP_NONREMOVABLE flag.
sure. Will update it. Thank you.
>
> [...]
>
> Kind regards
> Uffe
Powered by blists - more mailing lists