| Message ID | 20210325011200.145818-7-kuba@kernel.org |
|---|---|
| State | New |
| Headers | show |
| Series | [net-next,1/6] ethtool: fec: fix typo in kdoc | expand |
On 25/03/2021 01:12, Jakub Kicinski wrote: > Drivers should reject mixing %ETHTOOL_FEC_AUTO_BIT with other > + * FEC modes, because it's unclear whether in this case other modes constrain > + * AUTO or are independent choices. Does this mean you want me to spin a patch to sfc to reject this? Currently for us e.g. AUTO|RS means use RS if the cable and link partner both support it, otherwise let firmware choose (presumably between BASER and OFF) based on cable/module & link partner caps and/or parallel detect. We took this approach because our requirements writers believed that customers would have a need for this setting; they called it "prefer FEC", and I think the idea was to use FEC if possible (even on cables where the IEEE-recommended default is no FEC, such as CA-25G-N 3m DAC) but allow fallback to no FEC if e.g. link partner doesn't advertise FEC in AN. Similarly, AUTO|BASER ("prefer BASE-R FEC") might be desired by a user who wants to use BASE-R if possible to minimise latency, but fall back to RS FEC if the cable or link partner insists on it (eg CA-25G-L 5m DAC). Whether we were right and all this is actually useful, I couldn't say. -ed
On Mon, Mar 29, 2021 at 12:56:30PM +0100, Edward Cree wrote: > On 25/03/2021 01:12, Jakub Kicinski wrote: > > Drivers should reject mixing %ETHTOOL_FEC_AUTO_BIT with other > > + * FEC modes, because it's unclear whether in this case other modes constrain > > + * AUTO or are independent choices. > > Does this mean you want me to spin a patch to sfc to reject this? > Currently for us e.g. AUTO|RS means use RS if the cable and link partner > both support it, otherwise let firmware choose (presumably between BASER > and OFF) based on cable/module & link partner caps and/or parallel detect. > We took this approach because our requirements writers believed that > customers would have a need for this setting; they called it "prefer FEC", > and I think the idea was to use FEC if possible (even on cables where the > IEEE-recommended default is no FEC, such as CA-25G-N 3m DAC) but allow > fallback to no FEC if e.g. link partner doesn't advertise FEC in AN. > Similarly, AUTO|BASER ("prefer BASE-R FEC") might be desired by a user who > wants to use BASE-R if possible to minimise latency, but fall back to RS > FEC if the cable or link partner insists on it (eg CA-25G-L 5m DAC). > Whether we were right and all this is actually useful, I couldn't say. Jacub was talking about adding a netlink API as the next step. You should feed this in as a requirement for that. Being able to express preferences in the API in an explicitly documented way. It there any other existing ethtool setting which could be used as a model? EEE, master/slave? I would class pause as an anti model, that is frequently done wrong :-( Andrew
On Mon, 29 Mar 2021 12:56:30 +0100 Edward Cree wrote: > On 25/03/2021 01:12, Jakub Kicinski wrote: > > Drivers should reject mixing %ETHTOOL_FEC_AUTO_BIT with other > > + * FEC modes, because it's unclear whether in this case other modes constrain > > + * AUTO or are independent choices. > > Does this mean you want me to spin a patch to sfc to reject this? > Currently for us e.g. AUTO|RS means use RS if the cable and link partner > both support it, otherwise let firmware choose (presumably between BASER > and OFF) based on cable/module & link partner caps and/or parallel detect. > We took this approach because our requirements writers believed that > customers would have a need for this setting; they called it "prefer FEC", > and I think the idea was to use FEC if possible (even on cables where the > IEEE-recommended default is no FEC, such as CA-25G-N 3m DAC) but allow > fallback to no FEC if e.g. link partner doesn't advertise FEC in AN. > Similarly, AUTO|BASER ("prefer BASE-R FEC") might be desired by a user who > wants to use BASE-R if possible to minimise latency, but fall back to RS > FEC if the cable or link partner insists on it (eg CA-25G-L 5m DAC). > Whether we were right and all this is actually useful, I couldn't say. Interesting combo. Up to you, the API is quite unclear, I think users shouldn't expect anything beyond single bit set to work across implementations. IMHO supporting anything beyond that is just code complexity for little to no gain. But then again, as long as you don't confuse AUTO with autoneg there's no burning need to change :)
diff --git a/include/uapi/linux/ethtool.h b/include/uapi/linux/ethtool.h index 517b68c5fcec..f6ef7d42c7a1 100644 --- a/include/uapi/linux/ethtool.h +++ b/include/uapi/linux/ethtool.h @@ -1372,35 +1372,58 @@ struct ethtool_per_queue_op { __u32 cmd; __u32 sub_command; __u32 queue_mask[__KERNEL_DIV_ROUND_UP(MAX_NUM_QUEUE, 32)]; char data[]; }; /** - * struct ethtool_fecparam - Ethernet forward error correction(fec) parameters + * struct ethtool_fecparam - Ethernet Forward Error Correction parameters * @cmd: Command number = %ETHTOOL_GFECPARAM or %ETHTOOL_SFECPARAM - * @active_fec: FEC mode which is active on the port, GET only. - * @fec: Bitmask of supported/configured FEC modes + * @active_fec: FEC mode which is active on the port, single bit set, GET only. + * @fec: Bitmask of configured FEC modes. * @reserved: Reserved for future extensions, ignore on GET, write 0 for SET. + * + * FEC modes supported by the device can be read via %ETHTOOL_GLINKSETTINGS. + * FEC settings are configured by link autonegotiation whenever it's enabled. + * With autoneg on %ETHTOOL_GFECPARAM can be used to read the current mode. + * + * When autoneg is disabled %ETHTOOL_SFECPARAM controls the FEC settings. + * It is recommended that drivers only accept a single bit set in @fec. + * When multiple bits are set in @fec drivers may pick mode in an implementation + * dependent way. Drivers should reject mixing %ETHTOOL_FEC_AUTO_BIT with other + * FEC modes, because it's unclear whether in this case other modes constrain + * AUTO or are independent choices. + * Drivers must reject SET requests if they support none of the requested modes. + * + * If device does not support FEC drivers may use %ETHTOOL_FEC_NONE instead + * of returning %EOPNOTSUPP from %ETHTOOL_GFECPARAM. + * + * See enum ethtool_fec_config_bits for definition of valid bits for both + * @fec and @active_fec. */ struct ethtool_fecparam { __u32 cmd; /* bitmask of FEC modes */ __u32 active_fec; __u32 fec; __u32 reserved; }; /** * enum ethtool_fec_config_bits - flags definition of ethtool_fec_configuration - * @ETHTOOL_FEC_NONE: FEC mode configuration is not supported - * @ETHTOOL_FEC_AUTO: Default/Best FEC mode provided by driver + * @ETHTOOL_FEC_NONE: FEC mode configuration is not supported. Should not + * be used together with other bits. GET only. + * @ETHTOOL_FEC_AUTO: Select default/best FEC mode automatically, usually based + * link mode and SFP parameters read from module's EEPROM. + * This bit does _not_ mean autonegotiation. * @ETHTOOL_FEC_OFF: No FEC Mode - * @ETHTOOL_FEC_RS: Reed-Solomon Forward Error Detection mode - * @ETHTOOL_FEC_BASER: Base-R/Reed-Solomon Forward Error Detection mode + * @ETHTOOL_FEC_RS: Reed-Solomon FEC Mode + * @ETHTOOL_FEC_BASER: Base-R/Reed-Solomon FEC Mode + * @ETHTOOL_FEC_LLRS: Low Latency Reed Solomon FEC Mode (25G/50G Ethernet + * Consortium) */ enum ethtool_fec_config_bits { ETHTOOL_FEC_NONE_BIT, ETHTOOL_FEC_AUTO_BIT, ETHTOOL_FEC_OFF_BIT, ETHTOOL_FEC_RS_BIT, ETHTOOL_FEC_BASER_BIT,
The definition of the FEC driver interface is quite unclear. Improve the documentation. This is based on current driver and user space code, as well as the discussions about the interface: RFC v1 (24 Oct 2016): https://lore.kernel.org/netdev/1477363849-36517-1-git-send-email-vidya@cumulusnetworks.com/ - this version has the autoneg field - no active_fec field - none vs off confusion is already present RFC v2 (10 Feb 2017): https://lore.kernel.org/netdev/1486727004-11316-1-git-send-email-vidya@cumulusnetworks.com/ - autoneg removed - active_fec added v1 (10 Feb 2017): https://lore.kernel.org/netdev/1486751311-42019-1-git-send-email-vidya@cumulusnetworks.com/ - no changes in the code v1 (24 Jun 2017): https://lore.kernel.org/netdev/1498331985-8525-1-git-send-email-roopa@cumulusnetworks.com/ - include in tree user v2 (27 Jul 2017): https://lore.kernel.org/netdev/1501199248-24695-1-git-send-email-roopa@cumulusnetworks.com/ Signed-off-by: Jakub Kicinski <kuba@kernel.org> --- include/uapi/linux/ethtool.h | 37 +++++++++++++++++++++++++++++------- 1 file changed, 30 insertions(+), 7 deletions(-)