Re: [Intel-wired-lan] [PATCH iwl-next v4 5/6] ice: shorten ring stat names and add accessors
> -Original Message- > From: Intel-wired-lan On Behalf Of Jacob > Keller > Sent: 21 November 2025 01:51 > To: Loktionov, Aleksandr ; Lobakin, Aleksander > ; Nguyen, Anthony L > ; Kitszel, Przemyslaw > > Cc: Simon Horman ; [email protected]; > [email protected]; Keller, Jacob E ; > Loktionov, Aleksandr > Subject: [Intel-wired-lan] [PATCH iwl-next v4 5/6] ice: shorten ring stat > names and add accessors > > The ice Tx/Rx hotpath has a few statistics counters for tracking unexpected > events. These values are stored as u64 but are not accumulated using the > u64_stats API. This could result in load/tear stores on some architectures. > Even some 64-bit architectures could have issues since the fields are not > read or written using ACCESS_ONCE or READ_ONCE. > > A following change is going to refactor the stats accumulator code to use the > u64_stats API for all of these stats, and to use u64_stats_read and > u64_stats_inc properly to prevent load/store tears on all architectures. > > Using u64_stats_inc and the syncp pointer is slightly verbose and would be > duplicated in a number of places in the Tx and Rx hot path. Add accessor > macros for the cases where only a single stat value is touched at once. To > keep lines short, also shorten the stats names and convert ice_txq_stats and > ice_rxq_stats to struct_group. > > This will ease the transition to properly using the u64_stats API in the > following change. > > Reviewed-by: Aleksandr Loktionov > Signed-off-by: Jacob Keller > --- > drivers/net/ethernet/intel/ice/ice_txrx.h | 55 +++ > drivers/net/ethernet/intel/ice/ice_txrx_lib.h | 2 +- > drivers/net/ethernet/intel/ice/ice_main.c | 16 > drivers/net/ethernet/intel/ice/ice_txrx.c | 16 > drivers/net/ethernet/intel/ice/ice_txrx_lib.c | 2 +- > drivers/net/ethernet/intel/ice/ice_xsk.c | 4 +- > 6 files changed, 60 insertions(+), 35 deletions(-) > Tested-by: Rinitha S (A Contingent worker at Intel)
Re: [Intel-wired-lan] [PATCH iwl-next v4 5/6] ice: shorten ring stat names and add accessors
On 11/25/2025 2:17 AM, Simon Horman wrote: > On Thu, Nov 20, 2025 at 12:20:45PM -0800, Jacob Keller wrote: >> The ice Tx/Rx hotpath has a few statistics counters for tracking unexpected >> events. These values are stored as u64 but are not accumulated using the >> u64_stats API. This could result in load/tear stores on some architectures. >> Even some 64-bit architectures could have issues since the fields are not >> read or written using ACCESS_ONCE or READ_ONCE. >> >> A following change is going to refactor the stats accumulator code to use >> the u64_stats API for all of these stats, and to use u64_stats_read and >> u64_stats_inc properly to prevent load/store tears on all architectures. >> >> Using u64_stats_inc and the syncp pointer is slightly verbose and would be >> duplicated in a number of places in the Tx and Rx hot path. Add accessor >> macros for the cases where only a single stat value is touched at once. To >> keep lines short, also shorten the stats names and convert ice_txq_stats >> and ice_rxq_stats to struct_group. >> >> This will ease the transition to properly using the u64_stats API in the >> following change. >> >> Reviewed-by: Aleksandr Loktionov >> Signed-off-by: Jacob Keller > > I had to read this and the next patch a few times to understand what was > going on. In the end, the key for me understanding this patch is "... > accessor macros for the cases where only a single stat value is touched at > once.". Especially the "once" bit. > I'm a little unhappy with needing access macros like this because its yet another "driver specific wrapper".. but I couldn't come up with something better.. > In the context of the following patch I think this change makes sense. > And I appreciate that keeping lines short also makes sense. So no > objections to the direction you've taken here. But I might not have > thought to use struct_group for this myself. > Right. The main issue I had was all the places where we increment one stat we'd end up needing *at least* 3 lines. And trying to split or refactor all of those updates seemed like it would be problematic. I liked the use of struct_group since it lets us keep size information and logical separation without needing to add the extra layer of indirection in the accesses. > Reviewed-by: Simon Horman OpenPGP_signature.asc Description: OpenPGP digital signature
Re: [Intel-wired-lan] [PATCH iwl-next v4 5/6] ice: shorten ring stat names and add accessors
On Thu, Nov 20, 2025 at 12:20:45PM -0800, Jacob Keller wrote: > The ice Tx/Rx hotpath has a few statistics counters for tracking unexpected > events. These values are stored as u64 but are not accumulated using the > u64_stats API. This could result in load/tear stores on some architectures. > Even some 64-bit architectures could have issues since the fields are not > read or written using ACCESS_ONCE or READ_ONCE. > > A following change is going to refactor the stats accumulator code to use > the u64_stats API for all of these stats, and to use u64_stats_read and > u64_stats_inc properly to prevent load/store tears on all architectures. > > Using u64_stats_inc and the syncp pointer is slightly verbose and would be > duplicated in a number of places in the Tx and Rx hot path. Add accessor > macros for the cases where only a single stat value is touched at once. To > keep lines short, also shorten the stats names and convert ice_txq_stats > and ice_rxq_stats to struct_group. > > This will ease the transition to properly using the u64_stats API in the > following change. > > Reviewed-by: Aleksandr Loktionov > Signed-off-by: Jacob Keller I had to read this and the next patch a few times to understand what was going on. In the end, the key for me understanding this patch is "... accessor macros for the cases where only a single stat value is touched at once.". Especially the "once" bit. In the context of the following patch I think this change makes sense. And I appreciate that keeping lines short also makes sense. So no objections to the direction you've taken here. But I might not have thought to use struct_group for this myself. Reviewed-by: Simon Horman
