Re: [PATCH] console: Use flush() before panic and reset

[Stefan Roese]

On 3/15/23 21:29, Tony Dinh wrote:

Hi Simon,

On Wed, Mar 15, 2023 at 12:43 PM Simon Glass <s...@chromium.org> wrote:

On Tue, 14 Mar 2023 at 18:24, Tony Dinh <mibo...@gmail.com> wrote:

To make sure the panic and the reset messages will go out, console flush() 
should be used.
Sleep periods do not work in early u-boot phase when timer driver is not 
initialized yet.

Reference: https://lists.denx.de/pipermail/u-boot/2023-March/512233.html

Signed-off-by: Tony Dinh <mibo...@gmail.com>
---

  arch/arm/lib/reset.c | 4 ++--
  lib/panic.c          | 4 +++-
  2 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/arch/arm/lib/reset.c b/arch/arm/lib/reset.c
index 95169bae1c..3e051e36f1 100644
--- a/arch/arm/lib/reset.c
+++ b/arch/arm/lib/reset.c
@@ -25,6 +25,7 @@
  #include <cpu_func.h>
  #include <irq_func.h>
  #include <linux/delay.h>
+#include <stdio.h>

Reviewed-by: Simon Glass <s...@chromium.org>

Is flush implemented widely?

Pali wrote that flush function and here was his comment on the
referenced thread:
"Moreover there is already new function named flush() which
does "wait until stdout message was sent" and can be used instead of
those sleeps. I have already did it on some places (see git history for
flush function) but seems that you find some more."

So I think the answer is not yet widely implemented.

I just checked and it seems it's supported for all platforms using
DM_SERIAL when the serial driver supports the pending() function:

drivers/serial/serial-uclass.c:
#ifdef CONFIG_CONSOLE_FLUSH_SUPPORT
static void _serial_flush(struct udevice *dev)
{
        struct dm_serial_ops *ops = serial_get_ops(dev);

        if (!ops->pending)
                return;
        while (ops->pending(dev, false) > 0)
                ;
}
#endif

So AFAICT it's implemented widely. Not sure, why
CONFIG_CONSOLE_FLUSH_SUPPORT was introduced. It's enabled by default
and should be available in all cases IMHO.

Thanks,
Stefan