udelay is impelmented using a busy-wait loop and consumes CPU cycles
while usleep_range is implemented using interrupts.cls_flush_uart_write()
is called after a channel lock is acquired i.e. an atomic context.
Hence delay in this method should use udelay instead of usleep_range.
Signed-off-by: Aishwarya Pant <aishpant@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
writeb((UART_FCR_ENABLE_FIFO | UART_FCR_CLEAR_XMIT),
&ch->ch_cls_uart->isr_fcr);
- usleep_range(10, 20);
+
+ /* Must use *delay family functions in atomic context */
+ udelay(10);
ch->ch_flags |= (CH_TX_FIFO_EMPTY | CH_TX_FIFO_LWM);
}