[PATCH v2] ARM: arria10: xload: retry FPGA configuration

[PATCH v2] ARM: arria10: xload: retry FPGA configuration

[Steffen Trumtrar]

From: Bruno Knittel <bruno.knittel@bruker.com>

The Intel documentation states that the FPGA configuration might fail.
This has been observed on the Enclustra AA1+ board where up to 4 retries
where required to configure the FPGA.

Debugging session showed that the data where correctly read from the
eMMC but yet the configuration failed.

This commit introduces a retry loop on the FPGA configuration.
Up to 10 retries (arbitrary) are attempted.
As the hardware can't be used anyway without the FPGA loaded, this
doesn't introduce any boot time problems. Taking longer is better than
just hang()ing.

Signed-off-by: Bruno Knittel <bruno.knittel@bruker.com>
Signed-off-by: Steffen Trumtrar <s.trumtrar@pengutronix.de>
---
Changes since v1:
    - split loop into inline function
---
 arch/arm/mach-socfpga/arria10-xload.c | 49 ++++++++++++++++++++++++++---------
 1 file changed, 37 insertions(+), 12 deletions(-)

diff --git a/arch/arm/mach-socfpga/arria10-xload.c b/arch/arm/mach-socfpga/arria10-xload.c
index 9d54a1de58..db3cc209ee 100644
--- a/arch/arm/mach-socfpga/arria10-xload.c
+++ b/arch/arm/mach-socfpga/arria10-xload.c
@@ -353,17 +353,9 @@ int arria10_prepare_mmc(int barebox_part, int rbf_part)
 	return 0;
 }
 
-int arria10_load_fpga(int offset, int bitstream_size)
+static inline int __arria10_load_fpga(void *buf, uint32_t count, uint32_t size)
 {
-	void *buf = (void *)0xffe00000 + SZ_256K - 256 - SZ_16K;
 	int ret;
-	uint32_t count;
-	uint32_t size = bitstream_size / SECTOR_SIZE;
-
-	if (offset)
-		offset = offset / SECTOR_SIZE;
-
-	count = offset;
 
 	arria10_read_blocks(buf, count + bitstream.first_sec, SZ_16K);
 
@@ -371,23 +363,56 @@ int arria10_load_fpga(int offset, int bitstream_size)
 
 	ret = a10_fpga_init(buf);
 	if (ret)
-		hang();
+		return -EAGAIN;
 
 	while (count <= size) {
 		ret = a10_fpga_write(buf, SZ_16K);
 		if (ret == -ENOSPC)
-			break;
+			return -EAGAIN;
+
 		count += SZ_16K / SECTOR_SIZE;
 		ret = arria10_read_blocks(buf, count, SZ_16K);
+		// Reading failed, consider this a failed attempt to configure the FPGA and retry
+		if (ret)
+			return -EAGAIN;
 	}
 
 	ret = a10_fpga_write_complete();
 	if (ret)
-		hang();
+		return -EAGAIN;
 
 	return 0;
 }
 
+int arria10_load_fpga(int offset, int bitstream_size)
+{
+	int ret;
+	void *buf = (void *)0xffe00000 + SZ_256K - 256 - SZ_16K;
+	uint32_t count;
+	uint32_t size = bitstream_size / SECTOR_SIZE;
+	uint32_t retryCount;
+
+	if (offset)
+		offset = offset / SECTOR_SIZE;
+
+	/* Up to 4 retries have been seen on the Enclustra Mercury AA1+ board, as
+	 * FPGA configuration is mandatory to be able to continue the boot, take
+	 * some margin and try up to 10 times
+	 */
+	for (retryCount = 0; retryCount < 10; ++retryCount) {
+		count = offset;
+
+		ret = __arria10_load_fpga(buf, count, size);
+		if (!ret)
+			return 0;
+		else if (ret == -EAGAIN)
+			continue;
+	}
+
+	hang();
+	return -EIO;
+}
+
 void arria10_start_image(int offset)
 {
 	void *buf = (void *)0x0;

---
base-commit: bc5e49335067d0c6c6e97c082a0368fe9d86b725
change-id: 20231115-arria10-fpga-reload-on-failure-ba64deafcc52

Best regards,
-- 
Steffen Trumtrar <s.trumtrar@pengutronix.de>

Re: [PATCH v2] ARM: arria10: xload: retry FPGA configuration

[Sascha Hauer]

On Wed, Nov 15, 2023 at 02:11:22PM +0100, Steffen Trumtrar wrote:
> From: Bruno Knittel <bruno.knittel@bruker.com>
> 
> The Intel documentation states that the FPGA configuration might fail.
> This has been observed on the Enclustra AA1+ board where up to 4 retries
> where required to configure the FPGA.
> 
> Debugging session showed that the data where correctly read from the
> eMMC but yet the configuration failed.
> 
> This commit introduces a retry loop on the FPGA configuration.
> Up to 10 retries (arbitrary) are attempted.
> As the hardware can't be used anyway without the FPGA loaded, this
> doesn't introduce any boot time problems. Taking longer is better than
> just hang()ing.
> 
> Signed-off-by: Bruno Knittel <bruno.knittel@bruker.com>
> Signed-off-by: Steffen Trumtrar <s.trumtrar@pengutronix.de>
> ---
> Changes since v1:
>     - split loop into inline function
> ---
>  arch/arm/mach-socfpga/arria10-xload.c | 49 ++++++++++++++++++++++++++---------
>  1 file changed, 37 insertions(+), 12 deletions(-)

Applied, thanks

Sascha

> 
> diff --git a/arch/arm/mach-socfpga/arria10-xload.c b/arch/arm/mach-socfpga/arria10-xload.c
> index 9d54a1de58..db3cc209ee 100644
> --- a/arch/arm/mach-socfpga/arria10-xload.c
> +++ b/arch/arm/mach-socfpga/arria10-xload.c
> @@ -353,17 +353,9 @@ int arria10_prepare_mmc(int barebox_part, int rbf_part)
>  	return 0;
>  }
>  
> -int arria10_load_fpga(int offset, int bitstream_size)
> +static inline int __arria10_load_fpga(void *buf, uint32_t count, uint32_t size)
>  {
> -	void *buf = (void *)0xffe00000 + SZ_256K - 256 - SZ_16K;
>  	int ret;
> -	uint32_t count;
> -	uint32_t size = bitstream_size / SECTOR_SIZE;
> -
> -	if (offset)
> -		offset = offset / SECTOR_SIZE;
> -
> -	count = offset;
>  
>  	arria10_read_blocks(buf, count + bitstream.first_sec, SZ_16K);
>  
> @@ -371,23 +363,56 @@ int arria10_load_fpga(int offset, int bitstream_size)
>  
>  	ret = a10_fpga_init(buf);
>  	if (ret)
> -		hang();
> +		return -EAGAIN;
>  
>  	while (count <= size) {
>  		ret = a10_fpga_write(buf, SZ_16K);
>  		if (ret == -ENOSPC)
> -			break;
> +			return -EAGAIN;
> +
>  		count += SZ_16K / SECTOR_SIZE;
>  		ret = arria10_read_blocks(buf, count, SZ_16K);
> +		// Reading failed, consider this a failed attempt to configure the FPGA and retry
> +		if (ret)
> +			return -EAGAIN;
>  	}
>  
>  	ret = a10_fpga_write_complete();
>  	if (ret)
> -		hang();
> +		return -EAGAIN;
>  
>  	return 0;
>  }
>  
> +int arria10_load_fpga(int offset, int bitstream_size)
> +{
> +	int ret;
> +	void *buf = (void *)0xffe00000 + SZ_256K - 256 - SZ_16K;
> +	uint32_t count;
> +	uint32_t size = bitstream_size / SECTOR_SIZE;
> +	uint32_t retryCount;
> +
> +	if (offset)
> +		offset = offset / SECTOR_SIZE;
> +
> +	/* Up to 4 retries have been seen on the Enclustra Mercury AA1+ board, as
> +	 * FPGA configuration is mandatory to be able to continue the boot, take
> +	 * some margin and try up to 10 times
> +	 */
> +	for (retryCount = 0; retryCount < 10; ++retryCount) {
> +		count = offset;
> +
> +		ret = __arria10_load_fpga(buf, count, size);
> +		if (!ret)
> +			return 0;
> +		else if (ret == -EAGAIN)
> +			continue;
> +	}
> +
> +	hang();
> +	return -EIO;
> +}
> +
>  void arria10_start_image(int offset)
>  {
>  	void *buf = (void *)0x0;
> 
> ---
> base-commit: bc5e49335067d0c6c6e97c082a0368fe9d86b725
> change-id: 20231115-arria10-fpga-reload-on-failure-ba64deafcc52
> 
> Best regards,
> -- 
> Steffen Trumtrar <s.trumtrar@pengutronix.de>
> 
> 
> 

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |