On 2025-01-03, Petr Mladek <pmladek@xxxxxxxx> wrote: > My understanding is that the nested IER manipulation does not > harm. This statement implies that it is OK for UART_IER_RLSI|UART_IER_RDI bits to be set in UART_IER even though the console will write to UART_TX, because the _nesting_ contexts would set those bits rather than restoring the original value of 0x0. I ran some tests and leaving these bits set during Tx does not appear to cause an issue, but it is difficult to say because the context interrupted by a nesting context will only print at most 1 character. Also, it is writing under spin_lock_irqsave() so that might be masking any effects. Perhaps UART_IER is temporarly cleared because of other bits that would cause problems during Tx? I would need to create a specific test to investigate this further. Regardless, it still is a cosmetic ugliness that bits are not being properly restored, even if it turns out these particular bits are not problematic during Tx. As was with your LSR_DR comment, you are good at triggering fundamental investigations into the history of the 8250 driver. ;-) > All in all, the patch looks good to me. > > Reviewed-by: Petr Mladek <pmladek@xxxxxxxx> Thanks for the review. I interpret it to mean that I do not need to make any changes to this patch for v5. John
