Re: [dpdk-dev] [PATCH V1 1/2] examples/ethtool: fix data type of MTU

["lihuisong (C)" <lihuisong@huawei.com>]

在 2021/11/18 1:49, David Marchand 写道:
> On Thu, Apr 29, 2021 at 12:53 PM Min Hu (Connor) <humin29@huawei.com> wrote:
>> From: Huisong Li <lihuisong@huawei.com>
>>
>> This patch changes the data type of 'mtu' in rte_ethtool_net_change_mtu()
>> from 'int' to 'uint16_t'.
> You did not describe why this change is needed.

Ensure that the input parameter of rte_ethtool_net_change_mtu() API in lib/rte_ethtool.c
re consistent with rte_eth_dev_set_mtu() in ethdev layer.
I will fix commit log.

>
>
>> Fixes: bda68ab9d1e7 ("examples/ethtool: add user-space ethtool sample application")
>> Cc: stable@dpdk.org
>>
>> Signed-off-by: Huisong Li <lihuisong@huawei.com>
>> ---
>>   examples/ethtool/ethtool-app/ethapp.c | 6 +-----
>>   examples/ethtool/lib/rte_ethtool.c    | 9 +++++----
>>   examples/ethtool/lib/rte_ethtool.h    | 2 +-
>>   3 files changed, 7 insertions(+), 10 deletions(-)
>>
>> diff --git a/examples/ethtool/ethtool-app/ethapp.c b/examples/ethtool/ethtool-app/ethapp.c
>> index 36a1c37..fc743ce 100644
>> --- a/examples/ethtool/ethtool-app/ethapp.c
>> +++ b/examples/ethtool/ethtool-app/ethapp.c
>> @@ -521,13 +521,9 @@ pcmd_mtu_callback(void *ptr_params,
>>   {
>>          struct pcmd_intstr_params *params = ptr_params;
>>          int stat;
>> -       int new_mtu;
>> +       uint16_t new_mtu;
> strtoul can overflow the stack, when storing to new_mtu some lines below.
> You should either change new_mtu to unsigned long int or switch
> strtoul to some other integer format helper.

You are right. I want add an mtu specific cmd parameter structure:

struct pcmd_mtu_params {
     cmdline_fixed_string_t cmd;
     uint16_t port;
     uint16_t mtu_value;
};

MTU size is limited when users enter the value.

Like:

cmdline_parse_token_num_t pcmd_mtu_token_port =
     TOKEN_NUM_INITIALIZER(struct pcmd_mtu_params, mtu_value, RTE_UINT16);

What do you think, David?

>
>>          char *ptr_parse_end;
>>
>> -       if (!rte_eth_dev_is_valid_port(params->port)) {
>> -               printf("Error: Invalid port number %i\n", params->port);
>> -               return;
>> -       }
> Adding the check in lib/rte_ethtool.c is ok, but why not keep the
> check in the application?
All right, let it stay here.
>
>
>>          new_mtu = strtoul(params->opt, &ptr_parse_end, 10);
>>          if (*ptr_parse_end != '\0' ||
>>                          new_mtu < RTE_ETHER_MIN_MTU ||
> With this patch, rte_ethtool_net_change_mtu() now takes a uint16_t.
> You should add a check on new_mtu < UINT16_MAX.
>
If we use above new method, this check can be removed.
>> diff --git a/examples/ethtool/lib/rte_ethtool.c b/examples/ethtool/lib/rte_ethtool.c
>> index 4132516..73193ed 100644
>> --- a/examples/ethtool/lib/rte_ethtool.c
>> +++ b/examples/ethtool/lib/rte_ethtool.c
>> @@ -345,12 +345,13 @@ rte_ethtool_net_validate_addr(uint16_t port_id __rte_unused,
>>          return rte_is_valid_assigned_ether_addr(addr);
>>   }
>>
>> +
>>   int
>> -rte_ethtool_net_change_mtu(uint16_t port_id, int mtu)
>> +rte_ethtool_net_change_mtu(uint16_t port_id, uint16_t mtu)
>>   {
>> -       if (mtu < 0 || mtu > UINT16_MAX)
>> -               return -EINVAL;
>> -       return rte_eth_dev_set_mtu(port_id, (uint16_t)mtu);
>> +       RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
>> +
>> +       return rte_eth_dev_set_mtu(port_id, mtu);
>>   }
>>
>>   int
>> diff --git a/examples/ethtool/lib/rte_ethtool.h b/examples/ethtool/lib/rte_ethtool.h
>> index f177096..fe3250e 100644
>> --- a/examples/ethtool/lib/rte_ethtool.h
>> +++ b/examples/ethtool/lib/rte_ethtool.h
>> @@ -309,7 +309,7 @@ int rte_ethtool_net_validate_addr(uint16_t port_id,
>>    *   - (-EINVAL) if parameters invalid.
>>    *   - others depends on the specific operations implementation.
>>    */
>> -int rte_ethtool_net_change_mtu(uint16_t port_id, int mtu);
>> +int rte_ethtool_net_change_mtu(uint16_t port_id, uint16_t mtu);
>>
>>   /**
>>    * Retrieve the Ethernet device traffic statistics
>> --
>> 2.8.1
>>
>