Re: [PATCH 2/2] mtd: cfi_cmdset_0002: fix delayed error detection on HyperFlash

[Vignesh Raghavendra <vigneshr@xxxxxx>]

Hi,

On 04/10/19 12:04 AM, Sergei Shtylyov wrote:
> The commit 4844ef80305d ("mtd: cfi_cmdset_0002: Add support for polling
> status register") added checking for the status register error bits into
> chip_good() to only return 1 if these bits are zero. Unfortunately, this
> means that polling using chip_good() always reaches a time-out condition
> when erase or program failure bits are set. I think the status register
> error checking should be fully delegated to cfi_check_err_status() that
> should return whether any error bits were set or not...
> 

Please reword last sentence to drop "I think". Something like:

Lets fully delegate the function of determining error condition to
cfi_check_err_status() and make chip_good() only look for Device
Ready/Busy condition.

> Fixes: 4844ef80305d ("mtd: cfi_cmdset_0002: Add support for polling status register")
> Signed-off-by: Sergei Shtylyov <sergei.shtylyov@xxxxxxxxxxxxxxxxxx>
> 
> ---
>  drivers/mtd/chips/cfi_cmdset_0002.c |   55 +++++++++++++++++++-----------------
>  1 file changed, 30 insertions(+), 25 deletions(-)
> 
> Index: linux/drivers/mtd/chips/cfi_cmdset_0002.c
> ===================================================================
> --- linux.orig/drivers/mtd/chips/cfi_cmdset_0002.c
> +++ linux/drivers/mtd/chips/cfi_cmdset_0002.c
> @@ -123,14 +123,14 @@ static int cfi_use_status_reg(struct cfi
>  		(extp->SoftwareFeatures & poll_mask) == CFI_POLL_STATUS_REG;
>  }
>  
> -static void cfi_check_err_status(struct map_info *map, struct flchip *chip,
> -				 unsigned long adr)
> +static int cfi_check_err_status(struct map_info *map, struct flchip *chip,
> +				unsigned long adr)
>  {
>  	struct cfi_private *cfi = map->fldrv_priv;
>  	map_word status;
>  
>  	if (!cfi_use_status_reg(cfi))
> -		return;
> +		return 0;
>  
>  	cfi_send_gen_cmd(0x70, cfi->addr_unlock1, chip->start, map, cfi,
>  			 cfi->device_type, NULL);
> @@ -138,7 +138,7 @@ static void cfi_check_err_status(struct
>  
>  	/* The error bits are invalid while the chip's busy */
>  	if (!map_word_bitsset(map, status, CMD(CFI_SR_DRB)))
> -		return;
> +		return 0;
>  
>  	if (map_word_bitsset(map, status, CMD(0x3a))) {
>  		unsigned long chipstatus = MERGESTATUS(status);
> @@ -155,7 +155,9 @@ static void cfi_check_err_status(struct
>  		if (chipstatus & CFI_SR_SLSB)
>  			pr_err("%s sector write protected, status %lx\n",
>  			       map->name, chipstatus);
> +		return 1;
>  	}
> +	return 0;
>  }
>  
>  /* #define DEBUG_CFI_FEATURES */
> @@ -852,20 +854,16 @@ static int __xipram chip_good(struct map
>  
>  	if (cfi_use_status_reg(cfi)) {
>  		map_word ready = CMD(CFI_SR_DRB);
> -		map_word err = CMD(CFI_SR_PSB | CFI_SR_ESB);
> +
>  		/*
>  		 * For chips that support status register, check device
> -		 * ready bit and Erase/Program status bit to know if
> -		 * operation succeeded.
> +		 * ready bit
>  		 */
>  		cfi_send_gen_cmd(0x70, cfi->addr_unlock1, chip->start, map, cfi,
>  				 cfi->device_type, NULL);
>  		curd = map_read(map, addr);
>  
> -		if (map_word_andequal(map, curd, ready, ready))
> -			return !map_word_bitsset(map, curd, err);
> -
> -		return 0;
> +		return map_word_andequal(map, curd, ready, ready);
>  	}
>  
>  	oldd = map_read(map, addr);
> @@ -1703,8 +1701,11 @@ static int __xipram do_write_oneword_onc

Nit: for some reason, your diff has function names truncated abruptly
which makes its slightly harder to locate the context. I use git
format-patch that produces better readable contexts.

>  			break;
>  		}
>  
> -		if (chip_good(map, chip, adr, datum))
> +		if (chip_good(map, chip, adr, datum)) {
> +			if (cfi_check_err_status(map, chip, adr))
> +				ret = -EIO;
>  			break;
> +		}
>  
>  		/* Latency issues. Drop the lock, wait a while and retry */
>  		UDELAY(map, chip, adr, 1);
> @@ -1777,7 +1778,6 @@ static int __xipram do_write_oneword_ret
>  	ret = do_write_oneword_once(map, chip, adr, datum, mode, cfi);
>  	if (ret) {
>  		/* reset on all failures. */
> -		cfi_check_err_status(map, chip, adr);
>  		map_write(map, CMD(0xF0), chip->start);
>  		/* FIXME - should have reset delay before continuing */
>  
> @@ -1974,12 +1974,17 @@ static int __xipram do_write_buffer_wait
>  		 */
>  		if (time_after(jiffies, timeo) &&
>  		    !chip_good(map, chip, adr, datum)) {
> +			pr_warn("MTD %s(): software timeout, address:0x%.8lx.\n",
> +				__func__, adr);

Since we are returning an error condition, this should be pr_err() (I
know that rest of the file does not follow this convention, but lets
make sure new code does)


Rest looks fine to me. Thanks for the patch!

-- 
Regards
Vignesh

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/