Re: [RFC PATCH V4 net-next 1/5] ethtool: Allow network drivers to dump arbitrary EEPROM data

On 3/23/2021 2:54 AM, Andrew Lunn wrote:

quoted

+static int eeprom_prepare_data(const struct ethnl_req_info *req_base,
+                            struct ethnl_reply_data *reply_base,
+                            struct genl_info *info)
+{
+     struct eeprom_reply_data *reply = MODULE_EEPROM_REPDATA(reply_base);
+     struct eeprom_req_info *request = MODULE_EEPROM_REQINFO(req_base);
+     struct ethtool_module_eeprom page_data = {0};
+     struct net_device *dev = reply_base->dev;
+     int ret;
+
+     if (!dev->ethtool_ops->get_module_eeprom_by_page)
+             return -EOPNOTSUPP;
+
+     /* Allow dumps either of low or high page without crossing half page boundary */
+     if ((request->offset < ETH_MODULE_EEPROM_PAGE_LEN / 2 &&
+          request->offset + request->length > ETH_MODULE_EEPROM_PAGE_LEN / 2) ||
+         request->offset + request->length > ETH_MODULE_EEPROM_PAGE_LEN)
+             return -EINVAL;

Please keep all the parameter validation together, in
eeprom_parse_request(). At some point, we may extend
eeprom_parse_request() to make use of extack, to indicate which
parameter is invalid. Just getting an -EINVAL can be hard to debug,
where as NL_SET_ERR_MSG_ATTR() can help the user.

Sure, we can add that.

       Andrew