> On 9/25/24 11:04 AM, Michael Wu wrote: > > In commit 35eba185fd1a ("i2c: designware: Calculate SCL timing > > parameter for High Speed Mode") hs_hcnt and hs_hcnt are computed based > > on > > fixed tHIGH = 160 and tLOW = 320. However, this fixed values only applies > > to the set of conditions of IC_CAP_LOADING = 400pF and > > IC_FREQ_OPTIMIZATION = 1. Outside of this conditions set, if this fixed > > values are still used, the calculated HCNT and LCNT will make the SCL > > frequency unabled to reach 3.4 MHz. > > > > If hs_hcnt and hs_lcnt are calculated based on fixed tHIGH = 160 and > > tLOW = 320, SCL frequency may not reach 3.4 MHz when IC_CAP_LOADING is > > not > > 400pF or IC_FREQ_OPTIMIZATION is not 1. > > > > Section 3.15.4.5 in DesignWare DW_apb_i2c Databook v2.03 says when > > IC_CLK_FREQ_OPTIMIZATION = 0, > > > > MIN_SCL_HIGHtime = 60 ns for 3.4 Mbps, bus loading = 100pF > > = 120 ns for 3,4 Mbps, bus loading = 400pF > > MIN_SCL_LOWtime = 160 ns for 3.4 Mbps, bus loading = 100pF > > = 320 ns for 3.4 Mbps, bus loading = 400pF > > > > and section 3.15.4.6 says when IC_CLK_FREQ_OPTIMIZATION = 1, > > > > MIN_SCL_HIGHtime = 60 ns for 3.4 Mbps, bus loading = 100pF > > = 160 ns for 3.4 Mbps, bus loading = 400pF > > MIN_SCL_LOWtime = 120 ns for 3.4 Mbps, bus loading = 100pF > > = 320 ns for 3.4 Mbps, bus loading = 400pF > > > > In order to calculate more accurate hs_hcnt and hs_lcnt, two hardware > > parameters IC_CAP_LOADING and IC_CLK_FREQ_OPTIMIZATION must be > > considered together. > > > > Signed-off-by: Michael Wu <michael.wu@xxxxxxxxx> > > --- > > drivers/i2c/busses/i2c-designware-common.c | 16 ++++++++++++++ > > drivers/i2c/busses/i2c-designware-core.h | 8 +++++++ > > drivers/i2c/busses/i2c-designware-master.c | 24 > +++++++++++++++++++-- > > drivers/i2c/busses/i2c-designware-platdrv.c | 2 ++ > > 4 files changed, 48 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/i2c/busses/i2c-designware-common.c > b/drivers/i2c/busses/i2c-designware-common.c > > index e8a688d04aee..f0a7d0ce6fd6 100644 > > --- a/drivers/i2c/busses/i2c-designware-common.c > > +++ b/drivers/i2c/busses/i2c-designware-common.c > > @@ -332,6 +332,22 @@ void i2c_dw_adjust_bus_speed(struct dw_i2c_dev > *dev) > > } > > EXPORT_SYMBOL_GPL(i2c_dw_adjust_bus_speed); > > > > +void i2c_dw_parse_of(struct dw_i2c_dev *dev) > > +{ > > + int ret; > > + > > + ret = device_property_read_u32(dev->dev, "bus-loading", > > + &dev->bus_loading); > > Like Andy said better name would be bus_capacitance_pf. I'll rename it to bus_cap_pf due to the limit of not exceed 80 characters a line. > > Also i2c_dw_parse_of() sounds too generic and may lead to think all and > only device tree related parameters are parsed here. > > > + if (ret || dev->bus_loading < 400) > > + dev->bus_loading = 100; > > + else > > + dev->bus_loading = 400; > > + > > I think these are more understandable and robust if no parameter > adjustments are not done here but used straight in the if statements in > the i2c_dw_set_timings_master(). Less if statements that way and all > checked in one place. bus_loading is a hardware parameter, indicating the capacitance value. Literally, it has no direct relationship with time, so I may create a new function to parse it and other hardware parameters. Because the same reason, it is not suitable to be adjusted in i2c_dw_set_timings_master(). Like i2c_parse_timing() declared in drivers/i2c/i2c-core-base.c, I recommend changing its value after device_property_read_u32() if necessary. Thanks & Regards, Michael