On 10/31/19 7:33 PM, Pavan Nikhilesh Bhagavatula wrote: >> From: Pavan Nikhilesh Bhagavatula >>> Hi Matan, >>> >>>> Hi Pavan >>>> >>>> From: Pavan Nikhilesh <pbhagavat...@marvell.com> >>>>> Some PMDs cannot work when certain offloads are enable/disabled, as a >>>>> workaround PMDs auto enable/disable offloads internally and expose it >>>>> through dev->data->dev_conf.rxmode.offloads. >>>>> >>>>> After device specific dev_configure is called compare the requested >>>>> offloads >>>>> to the offloads exposed by the PMD and, if the PMD failed to enable a >>>>> given >>>>> offload then log it and return -EINVAL from rte_eth_dev_configure, else if >>>>> the PMD failed to disable a given offload log and continue with >>>>> rte_eth_dev_configure. >>>>> >>>> >>>> rte_eth_dev_configure can be called more than 1 time in the device life >>>> time, How can you know what is the minimum offload configurations >>>> required by the port after the first call? >>>> Maybe putting it in dev info is better, what do you think? >>>> >>> >>> We only return -EINVAL in the case where we enable an offload advertised >>> by dev_info and the port still fails to enable it. >> >> Are you sure it is ok that devices may disable\enable offloads under the >> hood without user notification? > > Some devices already do it. The above check adds validation for the same.
The problem is that some offloads cannot be disabled. If application does not request Rx checksum offload since it does use it, it is not a problem to report it. Of course, it could be a problem if the offload is used, but application wants to disable it, for example, for debugging purposes. In this case, the solution is to mask offloads on application level, which is not ideal as well. Anyway, the patch just tries to highlight difference of applied from requested. So, it is a step forward. Also, the patch will fail configure if an offload is requested, but not enabled. >> Can't it break applications? >> Why does the device expose unsupported offloads in dev info? >> Does it update the running offload usynchronically? Race? >> Can you explain also your specific use case? >> >> >>>> Matan
