2022-06-11 11:04:44

by Cédric Le Goater

[permalink] [raw]
Subject: [PATCH] spi: aspeed: Fix division by zero

When using the normal read operation for data transfers, the dummy bus
width is zero. In that case, they are no dummy bytes to transfer and
setting the dummy field in the controller register becomes useless.

Issue was found on a custom "Bifrost" board with a AST2500 SoC and
using a MX25L51245GMI-08G SPI Flash.

Cc: Chin-Ting Kuo <[email protected]>
Reported-by: Ian Woloschin <[email protected]>
Fixes: 54613fc6659b ("spi: aspeed: Add support for direct mapping")
Signed-off-by: Cédric Le Goater <[email protected]>
---
drivers/spi/spi-aspeed-smc.c | 12 +++++++++++-
1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/drivers/spi/spi-aspeed-smc.c b/drivers/spi/spi-aspeed-smc.c
index 496f3e1e9079..3e891bf22470 100644
--- a/drivers/spi/spi-aspeed-smc.c
+++ b/drivers/spi/spi-aspeed-smc.c
@@ -558,6 +558,14 @@ static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
u32 ctl_val;
int ret = 0;

+ dev_dbg(aspi->dev,
+ "CE%d %s dirmap [ 0x%.8llx - 0x%.8llx ] OP %#x mode:%d.%d.%d.%d naddr:%#x ndummies:%#x\n",
+ chip->cs, op->data.dir == SPI_MEM_DATA_IN ? "read" : "write",
+ desc->info.offset, desc->info.offset + desc->info.length,
+ op->cmd.opcode, op->cmd.buswidth, op->addr.buswidth,
+ op->dummy.buswidth, op->data.buswidth,
+ op->addr.nbytes, op->dummy.nbytes);
+
chip->clk_freq = desc->mem->spi->max_speed_hz;

/* Only for reads */
@@ -574,9 +582,11 @@ static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
ctl_val = readl(chip->ctl) & ~CTRL_IO_CMD_MASK;
ctl_val |= aspeed_spi_get_io_mode(op) |
op->cmd.opcode << CTRL_COMMAND_SHIFT |
- CTRL_IO_DUMMY_SET(op->dummy.nbytes / op->dummy.buswidth) |
CTRL_IO_MODE_READ;

+ if (op->dummy.nbytes)
+ ctl_val |= CTRL_IO_DUMMY_SET(op->dummy.nbytes / op->dummy.buswidth);
+
/* Tune 4BYTE address mode */
if (op->addr.nbytes) {
u32 addr_mode = readl(aspi->regs + CE_CTRL_REG);
--
2.35.3


2022-06-13 09:04:11

by Pratyush Yadav

[permalink] [raw]
Subject: Re: [PATCH] spi: aspeed: Fix division by zero

On 11/06/22 12:39PM, C?dric Le Goater wrote:
> When using the normal read operation for data transfers, the dummy bus
> width is zero. In that case, they are no dummy bytes to transfer and
> setting the dummy field in the controller register becomes useless.
>
> Issue was found on a custom "Bifrost" board with a AST2500 SoC and
> using a MX25L51245GMI-08G SPI Flash.
>
> Cc: Chin-Ting Kuo <[email protected]>
> Reported-by: Ian Woloschin <[email protected]>
> Fixes: 54613fc6659b ("spi: aspeed: Add support for direct mapping")
> Signed-off-by: C?dric Le Goater <[email protected]>
> ---
> drivers/spi/spi-aspeed-smc.c | 12 +++++++++++-
> 1 file changed, 11 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/spi/spi-aspeed-smc.c b/drivers/spi/spi-aspeed-smc.c
> index 496f3e1e9079..3e891bf22470 100644
> --- a/drivers/spi/spi-aspeed-smc.c
> +++ b/drivers/spi/spi-aspeed-smc.c
> @@ -558,6 +558,14 @@ static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
> u32 ctl_val;
> int ret = 0;
>
> + dev_dbg(aspi->dev,
> + "CE%d %s dirmap [ 0x%.8llx - 0x%.8llx ] OP %#x mode:%d.%d.%d.%d naddr:%#x ndummies:%#x\n",
> + chip->cs, op->data.dir == SPI_MEM_DATA_IN ? "read" : "write",
> + desc->info.offset, desc->info.offset + desc->info.length,
> + op->cmd.opcode, op->cmd.buswidth, op->addr.buswidth,
> + op->dummy.buswidth, op->data.buswidth,
> + op->addr.nbytes, op->dummy.nbytes);
> +

Unrelated change. Please send as a separate patch.

> chip->clk_freq = desc->mem->spi->max_speed_hz;
>
> /* Only for reads */
> @@ -574,9 +582,11 @@ static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
> ctl_val = readl(chip->ctl) & ~CTRL_IO_CMD_MASK;
> ctl_val |= aspeed_spi_get_io_mode(op) |
> op->cmd.opcode << CTRL_COMMAND_SHIFT |
> - CTRL_IO_DUMMY_SET(op->dummy.nbytes / op->dummy.buswidth) |
> CTRL_IO_MODE_READ;
>
> + if (op->dummy.nbytes)
> + ctl_val |= CTRL_IO_DUMMY_SET(op->dummy.nbytes / op->dummy.buswidth);
> +

LGTM. With the above fixed,

Reviewed-by: Pratyush Yadav <[email protected]>

> /* Tune 4BYTE address mode */
> if (op->addr.nbytes) {
> u32 addr_mode = readl(aspi->regs + CE_CTRL_REG);
> --
> 2.35.3
>

--
Regards,
Pratyush Yadav
Texas Instruments Inc.

2022-06-13 10:04:37

by Cédric Le Goater

[permalink] [raw]
Subject: Re: [PATCH] spi: aspeed: Fix division by zero

On 6/13/22 10:39, Pratyush Yadav wrote:
> On 11/06/22 12:39PM, Cédric Le Goater wrote:
>> When using the normal read operation for data transfers, the dummy bus
>> width is zero. In that case, they are no dummy bytes to transfer and
>> setting the dummy field in the controller register becomes useless.
>>
>> Issue was found on a custom "Bifrost" board with a AST2500 SoC and
>> using a MX25L51245GMI-08G SPI Flash.
>>
>> Cc: Chin-Ting Kuo <[email protected]>
>> Reported-by: Ian Woloschin <[email protected]>
>> Fixes: 54613fc6659b ("spi: aspeed: Add support for direct mapping")
>> Signed-off-by: Cédric Le Goater <[email protected]>
>> ---
>> drivers/spi/spi-aspeed-smc.c | 12 +++++++++++-
>> 1 file changed, 11 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/spi/spi-aspeed-smc.c b/drivers/spi/spi-aspeed-smc.c
>> index 496f3e1e9079..3e891bf22470 100644
>> --- a/drivers/spi/spi-aspeed-smc.c
>> +++ b/drivers/spi/spi-aspeed-smc.c
>> @@ -558,6 +558,14 @@ static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
>> u32 ctl_val;
>> int ret = 0;
>>
>> + dev_dbg(aspi->dev,
>> + "CE%d %s dirmap [ 0x%.8llx - 0x%.8llx ] OP %#x mode:%d.%d.%d.%d naddr:%#x ndummies:%#x\n",
>> + chip->cs, op->data.dir == SPI_MEM_DATA_IN ? "read" : "write",
>> + desc->info.offset, desc->info.offset + desc->info.length,
>> + op->cmd.opcode, op->cmd.buswidth, op->addr.buswidth,
>> + op->dummy.buswidth, op->data.buswidth,
>> + op->addr.nbytes, op->dummy.nbytes);
>> +
>
> Unrelated change. Please send as a separate patch.

OK.

>> chip->clk_freq = desc->mem->spi->max_speed_hz;
>>
>> /* Only for reads */
>> @@ -574,9 +582,11 @@ static int aspeed_spi_dirmap_create(struct spi_mem_dirmap_desc *desc)
>> ctl_val = readl(chip->ctl) & ~CTRL_IO_CMD_MASK;
>> ctl_val |= aspeed_spi_get_io_mode(op) |
>> op->cmd.opcode << CTRL_COMMAND_SHIFT |
>> - CTRL_IO_DUMMY_SET(op->dummy.nbytes / op->dummy.buswidth) |
>> CTRL_IO_MODE_READ;
>>
>> + if (op->dummy.nbytes)
>> + ctl_val |= CTRL_IO_DUMMY_SET(op->dummy.nbytes / op->dummy.buswidth);
>> +
>
> LGTM. With the above fixed,
>
> Reviewed-by: Pratyush Yadav <[email protected]>

Thanks,

C.

>
>> /* Tune 4BYTE address mode */
>> if (op->addr.nbytes) {
>> u32 addr_mode = readl(aspi->regs + CE_CTRL_REG);
>> --
>> 2.35.3
>>
>