Sergei Shtylyov wrote:
> Hello.
>
> Marc St-Jean wrote:
>
> > There are three different fixes:
> > 1. Fix for DesignWare APB THRE errata:
> > In brief, this is a non-standard 16550 in that the THRE interrupt
> > will not re-assert itself simply by disabling and re-enabling the
> > THRI bit in the IER, it is only re-enabled if a character is actually
> > sent out.
> > It appears that the "8250-uart-backup-timer.patch" in the "mm" tree also
> > fixes it so we have dropped our initial workaround.
> > This patch now needs to be applied on top of that "mm" patch.
>
> This patch has hit the mainline kernel already.
I see, I'll drop the reference to the "mm" patch.
> > 3. Workaround for interrupt/data concurrency issue:
> > The SoC needs to ensure that writes that can cause interrupts to be
> > cleared reach the UART before returning from the ISR. This fix reads
> > a non-destructive register on the UART so the read transaction
> > completion ensures the previously queued write transaction has
> > also completed.
>
> > The differences from the last attempt is dropping the call to
> > in_irq() and including more detailed description of the fixes.
>
> The difference part would better be under the "---" cutoff line, along
> with diffstat.
> This way it gets auto-removed by quilt/git when applying the patch.
I'll add to the next posting.
> > diff --git a/drivers/serial/8250.c b/drivers/serial/8250.c
> > index 3d91bfc..bfaacc5 100644
> > --- a/drivers/serial/8250.c
> > +++ b/drivers/serial/8250.c
> > @@ -308,6 +308,7 @@ static unsigned int serial_in(struct uar
> > return inb(up->port.iobase + 1);
> >
> > case UPIO_MEM:
> > + case UPIO_DWAPB:
> > return readb(up->port.membase + offset);
> >
> > case UPIO_MEM32:
> > @@ -333,6 +334,8 @@ #endif
> > static void
> > serial_out(struct uart_8250_port *up, int offset, int value)
> > {
> > + /* Save the offset before it's remapped */
> > + int save_offset = offset;
> > offset = map_8250_out_reg(up, offset) << up->port.regshift;
> >
> > switch (up->port.iotype) {
>
> I've just got an idea how to "beautify" this code -- rename the
> 'offset'
> arg to something like reg, and then declare/init 'offset' as local
> variable.
> Would certainly look better (and worth doing in all serial_* accessros).
I agree but am having enough of a hard time getting the bare minimum
accepted that I don't dare touch any unnecessary lines.
> > @@ -359,6 +362,18 @@ #endif
> > writeb(value, up->port.membase + offset);
> > break;
> >
> > + case UPIO_DWAPB:
> > + /* Save the LCR value so it can be re-written when a
> > + * Busy Detect interrupt occurs. */
> > + if (save_offset == UART_LCR)
> > + up->lcr = value;
> > + writeb(value, up->port.membase + offset);
> > + /* Read the IER to ensure any interrupt is cleared before
> > + * returning from ISR. */
> > + if (save_offset == UART_TX || save_offset == UART_IER)
> > + value = serial_in(up, UART_IER);
> > + break;
> > +
> > default:
> > outb(value, up->port.iobase + offset);
> > }
> > @@ -2454,8 +2485,8 @@ int __init serial8250_start_console(stru
> >
> > add_preferred_console("ttyS", line, options);
> > printk("Adding console on ttyS%d at %s 0x%lx (options '%s')\n",
> > - line, port->iotype == UPIO_MEM ? "MMIO" : "I/O port",
> > - port->iotype == UPIO_MEM ? (unsigned long) port->mapbase :
> > + line, port->iotype >= UPIO_MEM ? "MMIO" : "I/O port",
> > + port->iotype >= UPIO_MEM ? (unsigned long) port->mapbase :
> > (unsigned long) port->iobase, options);
> > if (!(serial8250_console.flags & CON_ENABLED)) {
> > serial8250_console.flags &= ~CON_PRINTBUFFER;
>
> I've remembered why I decided not to fix it: this code only gets called
> from 8250__eraly.c which can't handle anything beyond UPIO_MEM anyway.
We don't use 8250_early and I've tested it works without, so I'll drop this
change.
Thanks,
Marc
