On Fri, Sep 18, 2020 at 03:02:39PM +0000, Viorel Suman (OSS) wrote:
Please fix your mail client to word wrap within paragraphs at something substantially less than 80 columns. Doing this makes your messages much easier to read and reply to.
- regmap_read(regmap, FSL_XCVR_EXT_ISR, &isr);
- regmap_write(regmap, FSL_XCVR_EXT_ISR_CLR, isr);
This will unconditionally clear any interrupts, even those we don't understand - it might be better to only clear bits that are supported so the IRQ core can complain if there's something unexpected showing up.
The ARM core registers itself in "fsl_xcvr_prepare" (the code below) just for a subset of all supported interrupts:
ret = regmap_update_bits(xcvr->regmap, FSL_XCVR_EXT_IER0, FSL_XCVR_IRQ_EARC_ALL, FSL_XCVR_IRQ_EARC_ALL); ===== FSL_XCVR_IRQ_EARC_ALL - this mask represents all the interrupts we are interested in and we handle in interrupt handler, But this is just a subset of all interrupts the M0+ core is able to assert. Not very intuitive, I think I need to reword it somehow.
That's not the issue, the issue is that if we get into the ISR we just ack all the bits that are flagged by the hardware regardless of if we actually handled them. This won't work if there are ever systems that share the interrupt and it works against safety/debugging features that the interrupt has in case something goes wrong and we get spurious interrupts.
- if (isr & FSL_XCVR_IRQ_FIFO_UOFL_ERR)
dev_dbg(dev, "RX/TX FIFO full/empty\n");
Should this be dev_err()?
The interrupt may be asserted right before DMA starts to fill the TX FIFO if I recall correctly. I've added it just to debug the IP behavior, will check and change it to err it in next version if it is the case.
If it does come up normally then a comment or something to explain why this happens normally would probably be good.