Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH V2 1/1] mmc: sdhci: Fix O2 Host data read/write DLL lock
 Phase shift issue
To:     "Chevron Li (WH)" <chevron.li@bayhubtech.com>,
        Ulf Hansson <ulf.hansson@linaro.org>,
        "linux-mmc@vger.kernel.org" <linux-mmc@vger.kernel.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Cc:     "Max Huang (SC)" <max.huang@bayhubtech.com>,
        "Xiaoguang Yu (WH)" <xiaoguang.yu@bayhubtech.com>,
        "Shirley Her (SC)" <shirley.her@bayhubtech.com>,
        "Shaper Liu (WH)" <shaper.liu@bayhubtech.com>,
        "Louis Lu (TP)" <louis.lu@bayhubtech.com>
References: <1564166909-7437-1-git-send-email-shirley.her@bayhubtech.com>
 <0ce0b72d-46f6-ee06-092b-abb0d656c083@intel.com>
 <MW2PR16MB2249FF19BBE7890B0E7703C9EAD90@MW2PR16MB2249.namprd16.prod.outlook.com>
From:   Adrian Hunter <adrian.hunter@intel.com>
Organization: Intel Finland Oy, Registered Address: PL 281, 00181 Helsinki,
 Business Identity Code: 0357606 - 4, Domiciled in Helsinki
Message-ID: <48c25923-432f-25df-caa8-30b38f4ac9f7@intel.com>
Date:   Tue, 6 Aug 2019 14:54:29 +0300
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.8.0
MIME-Version: 1.0
In-Reply-To: <MW2PR16MB2249FF19BBE7890B0E7703C9EAD90@MW2PR16MB2249.namprd16.prod.outlook.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 2/08/19 1:53 PM, Chevron Li (WH) wrote:
> Hi, Adrian,
> 
> Firstly, really appreciate for your valuable suggestions.
> 
> For your question that " Tuning will be executed whenever the system resumes, which is not an error. Can you explain when error recovery is needed,
> and how this patch works."
> 
> We have blow questions need your help.
> 
> 1. In which cases card initialization module(mmc_init_card) will be called after first init completed exclude error happened? Whether the case is Resume from idle or sleep, any case else?
>   In which cases tuning module(execute_tuning) will be called after first init completed exclude error happened? Whether the case is Resume from idle or sleep, any case else?

Re-tuning can happen after CRC-error, resume, or re-tuning timer timeout.
Also as part of re-initializing a card after repeated errors in
mmc_blk_mq_rw_recovery.

> 
> Because in Bayhub driver, once tuning module is called exclude first initialization the tuning call will be thought as error recovery tuning.
> 
> 2. If the tuning and initialization are called cases are not clearly, can we get the mmc_blk_request error status(block request error) in tuning module?

It would be better for the Bayhub driver to save error status.

> 
> What method can be used to get the structure mmc_blk_request in tuning module(execute_tuning)
> 
> 3. If there is no method to get the block request error status in tuning module, we want to add a variable to save the error status during block request,
> and use it in tuning module as the condition to call Bayhub dll recovery.
> 
> So which structure the variable should be added, and how to acquire it in mmc_blk_mq_rw_recovery, how to acquire it in execute_tuning.
> 
> Looking forward to your reply!
> 
> Thanks,
> Chevron
> 
> 
> -----Original Message-----
> From: Adrian Hunter [mailto:adrian.hunter@intel.com] 
> Sent: 2019年7月29日 20:13
> To: Shirley Her (SC); Ulf Hansson; linux-mmc@vger.kernel.org; linux-kernel@vger.kernel.org
> Cc: Louis Lu (TP); Max Huang (SC); Chevron Li (WH)
> Subject: Re: [PATCH V2 1/1] mmc: sdhci: Fix O2 Host data read/write DLL lock Phase shift issue
> 
> On 26/07/19 9:48 PM, Shirley Her (SC) wrote:
>> Fix data read/write error in HS200 mode due to chip DLL lock phase 
>> shift
> 
> More explanation is needed.  See comment below where
> sdhci_o2_error_recovery() is called.
> 
>>
>> Signed-off-by:Shirley Her<shirley.her@bayhubtech.com>
>> ---
>> change in V2:
>>   1. Use usleep_range instead of udelay
>>   2. move dll_adjust_count to sdhci-pci-o2micro.c
>>
>> change in V1:
>>   1. Add error recovery function to relock the DLL with correct phase
>>   2. Retuning HS200 after DLL locked
>> ---
>> ---
>>  drivers/mmc/host/sdhci-pci-o2micro.c | 157 
>> +++++++++++++++++++++++++++++------
>>  1 file changed, 133 insertions(+), 24 deletions(-)
>>
>> diff --git a/drivers/mmc/host/sdhci-pci-o2micro.c 
>> b/drivers/mmc/host/sdhci-pci-o2micro.c
>> index 9dc4548..c43b2ce 100644
>> --- a/drivers/mmc/host/sdhci-pci-o2micro.c
>> +++ b/drivers/mmc/host/sdhci-pci-o2micro.c
>> @@ -51,13 +51,60 @@
>>  #define O2_SD_VENDOR_SETTING2	0x1C8
>>  #define O2_SD_HW_TUNING_DISABLE	BIT(4)
>>  
>> -#define O2_PLL_WDT_CONTROL1	0x1CC
>> +#define O2_PLL_DLL_WDT_CONTROL1	0x1CC
> 
> Renaming O2_PLL_WDT_CONTROL1 should be a separate patch.
> 
>>  #define  O2_PLL_FORCE_ACTIVE	BIT(18)
>>  #define  O2_PLL_LOCK_STATUS	BIT(14)
>>  #define  O2_PLL_SOFT_RESET	BIT(12)
>> +#define  O2_DLL_LOCK_STATUS	BIT(11)
>>  
>>  #define O2_SD_DETECT_SETTING 0x324
>>  
>> +static const u32 dmdn_table[5] = {0x25100000, 0x2B1C0000,
>> +	0x2C1A0000, 0x371B0000, 0x35100000}; static u8 dll_adjust_count;
> 
> This should be in a struct:
> 
> struct o2_host {
> 	u8 dll_adjust_count;
> }
> 
> 
>> +static int sdhci_o2_get_cd(struct mmc_host *mmc);
> 
> Please make a separate patch that reorders the functions so that forward declaration of sdhci_o2_get_cd() is not necessary.
> 
>> +
>> +static void o2_pci_set_baseclk(struct sdhci_pci_chip *chip, u32 
>> +value) {
>> +	u32 scratch_32;
>> +
>> +	pci_read_config_dword(chip->pdev,
>> +			      O2_SD_PLL_SETTING, &scratch_32);
>> +
>> +	scratch_32 &= 0x0000FFFF;
>> +	scratch_32 |= value;
>> +
>> +	pci_write_config_dword(chip->pdev,
>> +			       O2_SD_PLL_SETTING, scratch_32); }
>> +
>> +static int sdhci_o2_wait_dll_detect_lock(struct sdhci_host *host) {
>> +	ktime_t timeout;
>> +	u32 scratch32;
>> +
>> +	usleep_range(5000, 6000);
>> +	scratch32 = sdhci_readl(host, O2_PLL_DLL_WDT_CONTROL1);
>> +	if (!(scratch32 & O2_DLL_LOCK_STATUS)) {
>> +		pr_warn("%s: DLL is still unlocked after wait 5ms\n",
>> +			mmc_hostname(host->mmc));
>> +	}
>> +
>> +	/* Detect 500 ms */
>> +	timeout = ktime_add_ms(ktime_get(), 500);
>> +	while (1) {
>> +		bool timedout = ktime_after(ktime_get(), timeout);
>> +
>> +		scratch32 = sdhci_readl(host, O2_PLL_DLL_WDT_CONTROL1);
>> +		if (!(scratch32 & O2_DLL_LOCK_STATUS))
>> +			return 0;
>> +
>> +		if (timedout)
>> +			return 1;
>> +		usleep_range(10, 1000);
>> +	}
>> +}
>> +
>>  static void sdhci_o2_set_tuning_mode(struct sdhci_host *host)  {
>>  	u16 reg;
>> @@ -95,11 +142,85 @@ static void __sdhci_o2_execute_tuning(struct sdhci_host *host, u32 opcode)
>>  	sdhci_reset_tuning(host);
>>  }
>>  
>> +static int sdhci_o2_error_recovery(struct sdhci_pci_chip *chip,
>> +				 struct sdhci_host *host)
>> +{
>> +	int ret = 0;
>> +	u8 scratch_8 = 0;
>> +	u32 scratch_32 = 0;
> 
> Please add a blank line after local variable declaration
> 
>> +	/* Disable clock */
>> +	sdhci_writeb(host, 0, SDHCI_CLOCK_CONTROL);
>> +
>> +DLL_ADJUST:
> 
> Please use a 'do', 'for' or 'while' loop.
> 
>> +	if (dll_adjust_count >= 5) {
>> +		dll_adjust_count = 4;
>> +		pr_warn("%s: error recovery failed with the dll_adjust_count over max value.\n",
>> +			mmc_hostname(host->mmc));
> 
> Isn't it necessary to exit the loop here?
> 
>> +	}
>> +
>> +	/* UnLock WP */
>> +	ret = pci_read_config_byte(chip->pdev,
>> +			O2_SD_LOCK_WP, &scratch_8);
>> +	if (ret)
>> +		return ret;
>> +
>> +	scratch_8 &= 0x7f;
>> +	pci_write_config_byte(chip->pdev, O2_SD_LOCK_WP, scratch_8);
>> +
>> +	/* PLL software reset */
>> +	scratch_32 = sdhci_readl(host, O2_PLL_DLL_WDT_CONTROL1);
>> +	scratch_32 |= O2_PLL_SOFT_RESET;
>> +	sdhci_writel(host, scratch_32, O2_PLL_DLL_WDT_CONTROL1);
>> +
>> +	ret = pci_read_config_dword(chip->pdev,
>> +				    O2_SD_FUNC_REG4,
>> +				    &scratch_32);
>> +	/* Enable Base Clk setting change */
>> +	scratch_32 |= O2_SD_FREG4_ENABLE_CLK_SET;
>> +	pci_write_config_dword(chip->pdev, O2_SD_FUNC_REG4, scratch_32);
>> +	o2_pci_set_baseclk(chip, dmdn_table[dll_adjust_count]);
>> +
>> +
> 
> Please don't use multiple blank lines
> 
>> +	/* Enable internal clock */
>> +	scratch_8 = SDHCI_CLOCK_INT_EN;
>> +	sdhci_writeb(host, scratch_8, SDHCI_CLOCK_CONTROL);
>> +
>> +	if (sdhci_o2_get_cd(host->mmc)) {
>> +		if (sdhci_o2_wait_dll_detect_lock(host)) {
>> +			scratch_8 |= SDHCI_CLOCK_CARD_EN;
>> +			sdhci_writeb(host, scratch_8, SDHCI_CLOCK_CONTROL);
>> +			ret = 1;
>> +		} else {
>> +
> 
> Blank lines aren't necessary after an open brace '{'
> 
>> +			pr_warn("%s: DLL unlocked when dll_adjust_count is %d.\n",
>> +			mmc_hostname(host->mmc), dll_adjust_count);
>> +			dll_adjust_count++;
>> +			goto DLL_ADJUST;
>> +		}
>> +	} else
>> +		goto DLL_ADJUST;
> 
> braces {} should be used on all arms of this statement
> 
>> +
>> +		/* Lock WP */
>> +	ret = pci_read_config_byte(chip->pdev,
>> +				   O2_SD_LOCK_WP, &scratch_8);
>> +	if (ret)
>> +		return ret;
>> +	scratch_8 |= 0x80;
>> +	pci_write_config_byte(chip->pdev, O2_SD_LOCK_WP, scratch_8);
>> +	return ret;
>> +}
>> +
>>  static int sdhci_o2_execute_tuning(struct mmc_host *mmc, u32 opcode)  
>> {
>>  	struct sdhci_host *host = mmc_priv(mmc);
>> +	struct sdhci_pci_slot *slot = sdhci_priv(host);
> 
> This is how you get a reference to o2_host:
> 
> 	struct o2_host *o2_host = sdhci_pci_priv(slot);
> 
>>  	int current_bus_width = 0;
>>  
>> +
> 
> Please don't use multiple blank lines
> 
>> +	if (dll_adjust_count)
> 
> dll_adjust_count should be in struct o2_host
> 
> 	o2_host->dll_adjust_count
> 
>> +		sdhci_o2_error_recovery(slot->chip, host);
> 
> Tuning will be executed whenever the system resumes, which is not an error.
> Can you explain when error recovery is needed, and how this patch works.
> 
>> +
>> +	dll_adjust_count++;
>>  	/*
>>  	 * This handler only implements the eMMC tuning that is specific to
>>  	 * this controller.  Fall back to the standard method for other TIMING.
>> @@ -131,23 +252,11 @@ static int sdhci_o2_execute_tuning(struct mmc_host *mmc, u32 opcode)
>>  		mmc->ios.bus_width = MMC_BUS_WIDTH_8;
>>  		sdhci_set_bus_width(host, current_bus_width);
>>  	}
>> -
> 
> Please avoid unrelated white space changes.
> 
>>  	host->flags &= ~SDHCI_HS400_TUNING;
>>  	return 0;
>>  }
>>  
>> -static void o2_pci_set_baseclk(struct sdhci_pci_chip *chip, u32 
>> value) -{
>> -	u32 scratch_32;
>> -	pci_read_config_dword(chip->pdev,
>> -			      O2_SD_PLL_SETTING, &scratch_32);
>>  
>> -	scratch_32 &= 0x0000FFFF;
>> -	scratch_32 |= value;
>> -
>> -	pci_write_config_dword(chip->pdev,
>> -			       O2_SD_PLL_SETTING, scratch_32);
>> -}
>>  
>>  static void o2_pci_led_enable(struct sdhci_pci_chip *chip)  { @@ 
>> -316,23 +425,23 @@ static void sdhci_o2_enable_internal_clock(struct sdhci_host *host)
>>  	u32 scratch32;
>>  
>>  	/* PLL software reset */
>> -	scratch32 = sdhci_readl(host, O2_PLL_WDT_CONTROL1);
>> +	scratch32 = sdhci_readl(host, O2_PLL_DLL_WDT_CONTROL1);
>>  	scratch32 |= O2_PLL_SOFT_RESET;
>> -	sdhci_writel(host, scratch32, O2_PLL_WDT_CONTROL1);
>> +	sdhci_writel(host, scratch32, O2_PLL_DLL_WDT_CONTROL1);
>>  	udelay(1);
>>  	scratch32 &= ~(O2_PLL_SOFT_RESET);
>> -	sdhci_writel(host, scratch32, O2_PLL_WDT_CONTROL1);
>> +	sdhci_writel(host, scratch32, O2_PLL_DLL_WDT_CONTROL1);
>>  
>>  	/* PLL force active */
>>  	scratch32 |= O2_PLL_FORCE_ACTIVE;
>> -	sdhci_writel(host, scratch32, O2_PLL_WDT_CONTROL1);
>> +	sdhci_writel(host, scratch32, O2_PLL_DLL_WDT_CONTROL1);
>>  
>>  	/* Wait max 20 ms */
>>  	timeout = ktime_add_ms(ktime_get(), 20);
>>  	while (1) {
>>  		bool timedout = ktime_after(ktime_get(), timeout);
>>  
>> -		scratch = sdhci_readw(host, O2_PLL_WDT_CONTROL1);
>> +		scratch = sdhci_readw(host, O2_PLL_DLL_WDT_CONTROL1);
>>  		if (scratch & O2_PLL_LOCK_STATUS)
>>  			break;
>>  		if (timedout) {
>> @@ -350,16 +459,16 @@ static void 
>> sdhci_o2_enable_internal_clock(struct sdhci_host *host)
>>  
>>  out:
>>  	/* Cancel PLL force active */
>> -	scratch32 = sdhci_readl(host, O2_PLL_WDT_CONTROL1);
>> +	scratch32 = sdhci_readl(host, O2_PLL_DLL_WDT_CONTROL1);
>>  	scratch32 &= ~O2_PLL_FORCE_ACTIVE;
>> -	sdhci_writel(host, scratch32, O2_PLL_WDT_CONTROL1);
>> +	sdhci_writel(host, scratch32, O2_PLL_DLL_WDT_CONTROL1);
>>  }
>>  
>>  static int sdhci_o2_get_cd(struct mmc_host *mmc)  {
>>  	struct sdhci_host *host = mmc_priv(mmc);
>> -
> 
> Please keep the blank line after declarations.
> 
>> -	sdhci_o2_enable_internal_clock(host);
>> +	if (!(sdhci_readw(host, O2_PLL_DLL_WDT_CONTROL1) & O2_PLL_LOCK_STATUS))
>> +		sdhci_o2_enable_internal_clock(host);
>>  
>>  	return !!(sdhci_readl(host, SDHCI_PRESENT_STATE) & 
>> SDHCI_CARD_PRESENT);  } @@ -369,7 +478,7 @@ static void 
>> sdhci_o2_enable_clk(struct sdhci_host *host, u16 clk)
>>  	/* Enable internal clock */
>>  	clk |= SDHCI_CLOCK_INT_EN;
>>  	sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
>> -
>> +	sdhci_o2_enable_internal_clock(host);
>>  	if (sdhci_o2_get_cd(host->mmc)) {
>>  		clk |= SDHCI_CLOCK_CARD_EN;
>>  		sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL); @@ -400,7 +509,7 @@ 
>> int sdhci_pci_o2_probe_slot(struct sdhci_pci_slot *slot)
>>  
>>  	chip = slot->chip;
>>  	host = slot->host;
>> -
>> +	dll_adjust_count = 0;
>>  	caps = sdhci_readl(host, SDHCI_CAPABILITIES);
>>  
>>  	/*
>>
> 
> Set up o2_host like this:
> 
> diff --git a/drivers/mmc/host/sdhci-pci-o2micro.c
> b/drivers/mmc/host/sdhci-pci-o2micro.c
> index c43b2cecdf30..cfe57f23dcf1 100644
> --- a/drivers/mmc/host/sdhci-pci-o2micro.c
> +++ b/drivers/mmc/host/sdhci-pci-o2micro.c
> @@ -796,4 +796,5 @@ const struct sdhci_pci_fixes sdhci_o2 = {
>  	.resume = sdhci_pci_o2_resume,
>  #endif
>  	.ops = &sdhci_pci_o2_ops,
> +	.priv_size = sizeof(struct o2_host),
>  };
>