> -----Original Message----- > From: Thomas Monjalon [mailto:tho...@monjalon.net] > Sent: Monday, October 26, 2020 8:34 PM > To: Andrew Rybchenko <andrew.rybche...@oktetlabs.ru> > Cc: wangyunjian <wangyunj...@huawei.com>; dev@dpdk.org; > ferruh.yi...@intel.com; Lilijun (Jerry) <jerry.lili...@huawei.com>; xudingke > <xudin...@huawei.com>; ma...@nvidia.com > Subject: Re: [dpdk-dev] [PATCH] ethdev: fix data type for port id > > 26/10/2020 13:33, Andrew Rybchenko: > > On 10/26/20 3:29 PM, Thomas Monjalon wrote: > > > 26/10/2020 13:24, wangyunjian: > > >> From: Yunjian Wang <wangyunj...@huawei.com> > > >> > > >> The ethdev port id should be 16 bits now. This patch fixes the data > > >> type of the variable for 'pid', changing from uint32_t to uint16_t. > > >> > > >> Fixes: 5b7ba31148a8 ("ethdev: add port ownership") > > > > > > It was 32-bit on purpose, to avoid overflow in this loop: > > > for (pid = 0; pid < RTE_MAX_ETHPORTS; pid++) > > > > > > It is now replaced by RTE_ETH_FOREACH_VALID_DEV, but I wonder > > > whether we still have this theoritical overflow risk. > > > If yes, we should change more variables to 32-bit. > > > > Ah, it is too tricky. May be it is better to ensure that > > RTE_MAX_ETHPORTS is less or equal to UINT16_MAX? > > Yes could be another option. >
Add a check on RTE_MAX_ETHPORT in rte_eal_init()?
