[PATCH net-next v2 2/2] net: ethtool: Add support for writing firmware blocks using EPL payload

Danielle Ratson posted 2 patches 2 months, 3 weeks ago
There is a newer version of this series
[PATCH net-next v2 2/2] net: ethtool: Add support for writing firmware blocks using EPL payload
Posted by Danielle Ratson 2 months, 3 weeks ago
In the CMIS specification for pluggable modules, LPL (Local Payload) and
EPL (Extended Payload) are two types of data payloads used for managing
various functions and features of the module.

EPL payloads are used for more complex and extensive management
functions that require a larger amount of data, so writing firmware
blocks using EPL is much more efficient.

Currently, only LPL payload is supported for writing firmware blocks to
the module.

Add support for writing firmware block using EPL payload, both to
support modules that supports only EPL write mechanism, and to optimize
the flashing process of modules that support LPL and EPL.

Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
---

Notes:
    v2:
    	* Initialize the variable 'bytes_written' before the first
    	  iteration.

 net/ethtool/cmis.h           |  4 ++
 net/ethtool/cmis_cdb.c       | 66 ++++++++++++++++++++++++--
 net/ethtool/cmis_fw_update.c | 91 ++++++++++++++++++++++++++++++++----
 3 files changed, 148 insertions(+), 13 deletions(-)

diff --git a/net/ethtool/cmis.h b/net/ethtool/cmis.h
index 73a5060d0f4c..1e790413db0e 100644
--- a/net/ethtool/cmis.h
+++ b/net/ethtool/cmis.h
@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: GPL-2.0-only */
 
 #define ETHTOOL_CMIS_CDB_LPL_MAX_PL_LENGTH		120
+#define ETHTOOL_CMIS_CDB_EPL_MAX_PL_LENGTH		2048
 #define ETHTOOL_CMIS_CDB_CMD_PAGE			0x9F
 #define ETHTOOL_CMIS_CDB_PAGE_I2C_ADDR			0x50
 
@@ -23,6 +24,7 @@ enum ethtool_cmis_cdb_cmd_id {
 	ETHTOOL_CMIS_CDB_CMD_FW_MANAGMENT_FEATURES	= 0x0041,
 	ETHTOOL_CMIS_CDB_CMD_START_FW_DOWNLOAD		= 0x0101,
 	ETHTOOL_CMIS_CDB_CMD_WRITE_FW_BLOCK_LPL		= 0x0103,
+	ETHTOOL_CMIS_CDB_CMD_WRITE_FW_BLOCK_EPL		= 0x0104,
 	ETHTOOL_CMIS_CDB_CMD_COMPLETE_FW_DOWNLOAD	= 0x0107,
 	ETHTOOL_CMIS_CDB_CMD_RUN_FW_IMAGE		= 0x0109,
 	ETHTOOL_CMIS_CDB_CMD_COMMIT_FW_IMAGE		= 0x010A,
@@ -38,6 +40,7 @@ enum ethtool_cmis_cdb_cmd_id {
  * @resv1: Added to match the CMIS standard request continuity.
  * @resv2: Added to match the CMIS standard request continuity.
  * @payload: Payload for the CDB commands.
+ * @epl: Extended payload for the CDB commands.
  */
 struct ethtool_cmis_cdb_request {
 	__be16 id;
@@ -49,6 +52,7 @@ struct ethtool_cmis_cdb_request {
 		u8 resv2;
 		u8 payload[ETHTOOL_CMIS_CDB_LPL_MAX_PL_LENGTH];
 	);
+	u8 *epl;	/* Everything above this field checksummed. */
 };
 
 #define CDB_F_COMPLETION_VALID		BIT(0)
diff --git a/net/ethtool/cmis_cdb.c b/net/ethtool/cmis_cdb.c
index f599b20479f6..e27cad17505e 100644
--- a/net/ethtool/cmis_cdb.c
+++ b/net/ethtool/cmis_cdb.c
@@ -33,12 +33,19 @@ void ethtool_cmis_cdb_compose_args(struct ethtool_cmis_cdb_cmd_args *args,
 {
 	args->req.id = cpu_to_be16(cmd);
 	args->req.lpl_len = lpl_len;
-	if (lpl)
+	if (lpl) {
 		memcpy(args->req.payload, lpl, args->req.lpl_len);
+		args->read_write_len_ext =
+			ethtool_cmis_get_max_lpl_size(read_write_len_ext);
+	}
+	if (epl) {
+		args->req.epl_len = cpu_to_be16(epl_len);
+		args->req.epl = epl;
+		args->read_write_len_ext =
+			ethtool_cmis_get_max_epl_size(read_write_len_ext);
+	}
 
 	args->max_duration = max_duration;
-	args->read_write_len_ext =
-		ethtool_cmis_get_max_lpl_size(read_write_len_ext);
 	args->msleep_pre_rpl = msleep_pre_rpl;
 	args->rpl_exp_len = rpl_exp_len;
 	args->flags = flags;
@@ -548,6 +555,49 @@ __ethtool_cmis_cdb_execute_cmd(struct net_device *dev,
 	return err;
 }
 
+#define CMIS_CDB_EPL_PAGE_START			0xA0
+#define CMIS_CDB_EPL_PAGE_END			0xAF
+#define CMIS_CDB_EPL_FW_BLOCK_OFFSET_START	128
+#define CMIS_CDB_EPL_FW_BLOCK_OFFSET_END	255
+
+static int
+ethtool_cmis_cdb_execute_epl_cmd(struct net_device *dev,
+				 struct ethtool_cmis_cdb_cmd_args *args,
+				 struct ethtool_module_eeprom *page_data)
+{
+	u16 epl_len = be16_to_cpu(args->req.epl_len);
+	u32 bytes_written = 0;
+	u8 page;
+	int err;
+
+	for (page = CMIS_CDB_EPL_PAGE_START;
+	     page <= CMIS_CDB_EPL_PAGE_END && bytes_written < epl_len; page++) {
+		u16 offset = CMIS_CDB_EPL_FW_BLOCK_OFFSET_START;
+
+		while (offset <= CMIS_CDB_EPL_FW_BLOCK_OFFSET_END &&
+		       bytes_written < epl_len) {
+			u32 bytes_left = epl_len - bytes_written;
+			u16 space_left, bytes_to_write;
+
+			space_left = CMIS_CDB_EPL_FW_BLOCK_OFFSET_END - offset + 1;
+			bytes_to_write = min_t(u16, bytes_left,
+					       min_t(u16, space_left,
+						     args->read_write_len_ext));
+
+			err = __ethtool_cmis_cdb_execute_cmd(dev, page_data,
+							     page, offset,
+							     bytes_to_write,
+							     args->req.epl + bytes_written);
+			if (err < 0)
+				return err;
+
+			offset += bytes_to_write;
+			bytes_written += bytes_to_write;
+		}
+	}
+	return 0;
+}
+
 static u8 cmis_cdb_calc_checksum(const void *data, size_t size)
 {
 	const u8 *bytes = (const u8 *)data;
@@ -569,7 +619,9 @@ int ethtool_cmis_cdb_execute_cmd(struct net_device *dev,
 	int err;
 
 	args->req.chk_code =
-		cmis_cdb_calc_checksum(&args->req, sizeof(args->req));
+		cmis_cdb_calc_checksum(&args->req,
+				       offsetof(struct ethtool_cmis_cdb_request,
+						epl));
 
 	if (args->req.lpl_len > args->read_write_len_ext) {
 		args->err_msg = "LPL length is longer than CDB read write length extension allows";
@@ -591,6 +643,12 @@ int ethtool_cmis_cdb_execute_cmd(struct net_device *dev,
 	if (err < 0)
 		return err;
 
+	if (args->req.epl_len) {
+		err = ethtool_cmis_cdb_execute_epl_cmd(dev, args, &page_data);
+		if (err < 0)
+			return err;
+	}
+
 	offset = CMIS_CDB_CMD_ID_OFFSET +
 		offsetof(struct ethtool_cmis_cdb_request, id);
 	err = __ethtool_cmis_cdb_execute_cmd(dev, &page_data,
diff --git a/net/ethtool/cmis_fw_update.c b/net/ethtool/cmis_fw_update.c
index a514127985d4..48aef6220f00 100644
--- a/net/ethtool/cmis_fw_update.c
+++ b/net/ethtool/cmis_fw_update.c
@@ -9,6 +9,7 @@
 
 struct cmis_fw_update_fw_mng_features {
 	u8	start_cmd_payload_size;
+	u8	write_mechanism;
 	u16	max_duration_start;
 	u16	max_duration_write;
 	u16	max_duration_complete;
@@ -36,7 +37,9 @@ struct cmis_cdb_fw_mng_features_rpl {
 };
 
 enum cmis_cdb_fw_write_mechanism {
+	CMIS_CDB_FW_WRITE_MECHANISM_NONE	= 0x00,
 	CMIS_CDB_FW_WRITE_MECHANISM_LPL		= 0x01,
+	CMIS_CDB_FW_WRITE_MECHANISM_EPL		= 0x10,
 	CMIS_CDB_FW_WRITE_MECHANISM_BOTH	= 0x11,
 };
 
@@ -68,10 +71,9 @@ cmis_fw_update_fw_mng_features_get(struct ethtool_cmis_cdb *cdb,
 	}
 
 	rpl = (struct cmis_cdb_fw_mng_features_rpl *)args.req.payload;
-	if (!(rpl->write_mechanism == CMIS_CDB_FW_WRITE_MECHANISM_LPL ||
-	      rpl->write_mechanism == CMIS_CDB_FW_WRITE_MECHANISM_BOTH)) {
+	if (rpl->write_mechanism == CMIS_CDB_FW_WRITE_MECHANISM_NONE) {
 		ethnl_module_fw_flash_ntf_err(dev, ntf_params,
-					      "Write LPL is not supported",
+					      "CDB write mechanism is not supported",
 					      NULL);
 		return  -EOPNOTSUPP;
 	}
@@ -83,6 +85,10 @@ cmis_fw_update_fw_mng_features_get(struct ethtool_cmis_cdb *cdb,
 	 */
 	cdb->read_write_len_ext = rpl->read_write_len_ext;
 	fw_mng->start_cmd_payload_size = rpl->start_cmd_payload_size;
+	fw_mng->write_mechanism =
+		rpl->write_mechanism == CMIS_CDB_FW_WRITE_MECHANISM_LPL ?
+		CMIS_CDB_FW_WRITE_MECHANISM_LPL :
+		CMIS_CDB_FW_WRITE_MECHANISM_EPL;
 	fw_mng->max_duration_start = be16_to_cpu(rpl->max_duration_start);
 	fw_mng->max_duration_write = be16_to_cpu(rpl->max_duration_write);
 	fw_mng->max_duration_complete = be16_to_cpu(rpl->max_duration_complete);
@@ -149,9 +155,9 @@ struct cmis_cdb_write_fw_block_lpl_pl {
 };
 
 static int
-cmis_fw_update_write_image(struct ethtool_cmis_cdb *cdb,
-			   struct ethtool_cmis_fw_update_params *fw_update,
-			   struct cmis_fw_update_fw_mng_features *fw_mng)
+cmis_fw_update_write_image_lpl(struct ethtool_cmis_cdb *cdb,
+			       struct ethtool_cmis_fw_update_params *fw_update,
+			       struct cmis_fw_update_fw_mng_features *fw_mng)
 {
 	u8 start = fw_mng->start_cmd_payload_size;
 	u32 offset, max_block_size, max_lpl_len;
@@ -202,6 +208,67 @@ cmis_fw_update_write_image(struct ethtool_cmis_cdb *cdb,
 	return 0;
 }
 
+struct cmis_cdb_write_fw_block_epl_pl {
+	u8 fw_block[ETHTOOL_CMIS_CDB_EPL_MAX_PL_LENGTH];
+};
+
+static int
+cmis_fw_update_write_image_epl(struct ethtool_cmis_cdb *cdb,
+			       struct ethtool_cmis_fw_update_params *fw_update,
+			       struct cmis_fw_update_fw_mng_features *fw_mng)
+{
+	u8 start = fw_mng->start_cmd_payload_size;
+	u32 image_size = fw_update->fw->size;
+	u32 offset, lpl_len;
+	int err;
+
+	lpl_len = sizeof_field(struct cmis_cdb_write_fw_block_lpl_pl,
+			       block_address);
+
+	for (offset = start; offset < image_size;
+	     offset += ETHTOOL_CMIS_CDB_EPL_MAX_PL_LENGTH) {
+		struct cmis_cdb_write_fw_block_lpl_pl lpl = {
+			.block_address = cpu_to_be32(offset - start),
+		};
+		struct cmis_cdb_write_fw_block_epl_pl *epl;
+		struct ethtool_cmis_cdb_cmd_args args = {};
+		u32 epl_len;
+
+		ethnl_module_fw_flash_ntf_in_progress(fw_update->dev,
+						      &fw_update->ntf_params,
+						      offset - start,
+						      image_size);
+
+		epl_len = min_t(u32, ETHTOOL_CMIS_CDB_EPL_MAX_PL_LENGTH,
+				image_size - offset);
+		epl = kmalloc_array(epl_len, sizeof(u8), GFP_KERNEL);
+		if (!epl)
+			return -ENOMEM;
+
+		memcpy(epl->fw_block, &fw_update->fw->data[offset], epl_len);
+
+		ethtool_cmis_cdb_compose_args(&args,
+					      ETHTOOL_CMIS_CDB_CMD_WRITE_FW_BLOCK_EPL,
+					      (u8 *)&lpl, lpl_len, (u8 *)epl,
+					      epl_len,
+					      fw_mng->max_duration_write,
+					      cdb->read_write_len_ext, 1, 0,
+					      CDB_F_COMPLETION_VALID | CDB_F_STATUS_VALID);
+
+		err = ethtool_cmis_cdb_execute_cmd(fw_update->dev, &args);
+		kfree(epl);
+		if (err < 0) {
+			ethnl_module_fw_flash_ntf_err(fw_update->dev,
+						      &fw_update->ntf_params,
+						      "Write FW block EPL command failed",
+						      args.err_msg);
+			return err;
+		}
+	}
+
+	return 0;
+}
+
 static int
 cmis_fw_update_complete_download(struct ethtool_cmis_cdb *cdb,
 				 struct net_device *dev,
@@ -238,9 +305,15 @@ cmis_fw_update_download_image(struct ethtool_cmis_cdb *cdb,
 	if (err < 0)
 		return err;
 
-	err = cmis_fw_update_write_image(cdb, fw_update, fw_mng);
-	if (err < 0)
-		return err;
+	if (fw_mng->write_mechanism == CMIS_CDB_FW_WRITE_MECHANISM_LPL) {
+		err = cmis_fw_update_write_image_lpl(cdb, fw_update, fw_mng);
+		if (err < 0)
+			return err;
+	} else {
+		err = cmis_fw_update_write_image_epl(cdb, fw_update, fw_mng);
+		if (err < 0)
+			return err;
+	}
 
 	err = cmis_fw_update_complete_download(cdb, fw_update->dev, fw_mng,
 					       &fw_update->ntf_params);
-- 
2.45.0
Re: [PATCH net-next v2 2/2] net: ethtool: Add support for writing firmware blocks using EPL payload
Posted by Simon Horman 2 months, 2 weeks ago
+ Andrew Lunn

On Tue, Sep 10, 2024 at 12:02:17PM +0300, Danielle Ratson wrote:
> In the CMIS specification for pluggable modules, LPL (Local Payload) and
> EPL (Extended Payload) are two types of data payloads used for managing
> various functions and features of the module.
> 
> EPL payloads are used for more complex and extensive management
> functions that require a larger amount of data, so writing firmware
> blocks using EPL is much more efficient.
> 
> Currently, only LPL payload is supported for writing firmware blocks to
> the module.
> 
> Add support for writing firmware block using EPL payload, both to
> support modules that supports only EPL write mechanism, and to optimize
> the flashing process of modules that support LPL and EPL.
> 
> Signed-off-by: Danielle Ratson <danieller@nvidia.com>
> Reviewed-by: Petr Machata <petrm@nvidia.com>
> ---
> 
> Notes:
>     v2:
>     	* Initialize the variable 'bytes_written' before the first
>     	  iteration.

Hi Danielle,

Thanks for the update. From a doing-what-it-says-on-the-wrapper
perspective, this looks good to me.

Reviewed-by: Simon Horman <horms@kernel.org>

I do note that there were some questions from Andrew Lunn (CCed) in v1
regarding the size of transfers over the bus. I see that you responded to
that. Thanks! But I do wonder if he has any further comments.

https://lore.kernel.org/all/DM6PR12MB45163C3543368036C03B6E05D8982@DM6PR12MB4516.namprd12.prod.outlook.com/

...
Re: [PATCH net-next v2 2/2] net: ethtool: Add support for writing firmware blocks using EPL payload
Posted by Andrew Lunn 2 months, 2 weeks ago
On Wed, Sep 11, 2024 at 08:32:34AM +0100, Simon Horman wrote:
> + Andrew Lunn
> 
> On Tue, Sep 10, 2024 at 12:02:17PM +0300, Danielle Ratson wrote:
> > In the CMIS specification for pluggable modules, LPL (Local Payload) and
> > EPL (Extended Payload) are two types of data payloads used for managing
> > various functions and features of the module.
> > 
> > EPL payloads are used for more complex and extensive management
> > functions that require a larger amount of data, so writing firmware
> > blocks using EPL is much more efficient.
> > 
> > Currently, only LPL payload is supported for writing firmware blocks to
> > the module.
> > 
> > Add support for writing firmware block using EPL payload, both to
> > support modules that supports only EPL write mechanism, and to optimize
> > the flashing process of modules that support LPL and EPL.
> > 
> > Signed-off-by: Danielle Ratson <danieller@nvidia.com>
> > Reviewed-by: Petr Machata <petrm@nvidia.com>
> > ---
> > 
> > Notes:
> >     v2:
> >     	* Initialize the variable 'bytes_written' before the first
> >     	  iteration.
> 
> Hi Danielle,
> 
> Thanks for the update. From a doing-what-it-says-on-the-wrapper
> perspective, this looks good to me.
> 
> Reviewed-by: Simon Horman <horms@kernel.org>
> 
> I do note that there were some questions from Andrew Lunn (CCed) in v1
> regarding the size of transfers over the bus. I see that you responded to
> that. Thanks! But I do wonder if he has any further comments.

I do wounder where the speedup comes from.

> The LPL contains ap to 128 bytes, and the EPL up to 2048

Are 128 consecutive 16 byte transfers per block really faster than 8
consecutive transfers per block? If you have an efficient I2C master,
both should be doing 400Kbp. If the master is inefficient, you end up
with the same amount of dead time on the wire.

And does the standard really say you can fragment the block at the I2C
layer?

I suspect in the end we will have an API between the core and the
driver to ask it what size of block it actually supports. But we can
probably add that when we need it.

	Andrew