On Tue, Aug 25, 2026 at 01:36:48PM +0530, Kumar, Rajesh wrote: > > On 17-08-2026 02:48 pm, Mandal, Anurag wrote: > > Hi All, > > > > Gentle reminder for review & ack. > > > > Thank you. > > > > Regards, > > Anurag M > > > -----Original Message----- > > > From: Mandal, Anurag <[email protected]> > > > Sent: 03 August 2026 15:00 > > > To: [email protected] > > > Cc: Richardson, Bruce <[email protected]>; Shetty, Praveen > > > <[email protected]>; Mandal, Anurag <[email protected]> > > > Subject: [PATCH v2] net/cpfl: add timesync support > > > > > > The cpfl PMD advertises RTE_ETH_RX_OFFLOAD_TIMESTAMP, so received > > > packets carry a 32-bit hardware timestamp that the common idpf Rx path > > > extends to 64-bit nanoseconds using the device PTP clock (PHC). > > > However, it never initialised the PHC, so the timestamp conversion had no > > > reference time. > > > Applications had no way to enable timestamping or to read the current > > > device > > > clock to correlate the mbuf timestamps. > > > > > > This patch adds PTP support by reusing the idpf common PTP helpers: > > > > > > - timesync_enable: allocate and initialise the adapter PTP state, > > > query the PTP capabilities, program the base increment value > > > and set the device clock to the current system time. > > > The PTP state is shared by all vports of an adapter, so it > > > is initialised only once. > > > > > > - timesync_read_time: return the current device clock as a timespec. > > > > > > - timesync_disable: release the PTP state. > > > > > > Once timesync is enabled, the common Rx path resolves the cached PHC time, > > > so the mbuf Rx timestamp dynfield reports valid nanosecond values and > > > applications can read the device clock with rte_eth_timesync_read_time(). > > > > > > Signed-off-by: Anurag Mandal <[email protected]> > Reviewed-by: Rajesh Kumar <[email protected]>
Applied to dpdk-next-net-intel. Thanks, /Bruce
