On 08/07, Neeraj Upadhyay wrote: > Move the request_irq() call to the end of the msm_startup(), > so that we don't handle interrupts while msm_startup() is > running. This avoids potential races while initialization > is in progress. For example, consider below scenario > where rx handler reads the intermediate value of dma->chan, > set in msm_request_rx_dma(), and tries to do dma mapping, > which results in data abort. > > uart_port_startup() > msm_startup() > request_irq() > ... > msm_request_rx_dma() > ... > dma->chan = dma_request_slave_channel_reason(dev, "rx"); > <UART RX IRQ> > msm_uart_irq() > msm_handle_rx_dm() > msm_start_rx_dma() > dma->desc = dma_map_single() > <data abort> > > Signed-off-by: Neeraj Upadhyay <neeraju@xxxxxxxxxxxxxx> > --- Or we can write the IMR register with 0 to mask all irqs before requesting the irq handler be setup? We will unmask the interrupts we care about anyway when we setup termios. If not, this change is ok, but I would guess it doesn't fix spurious irqs from happening while we keep configuring the hardware. Feel free to take my reviewed-by though. Reviewed-by: Stephen Boyd <sboyd@xxxxxxxxxxxxxx> -- Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project -- To unsubscribe from this list: send the line "unsubscribe linux-soc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
