[RESEND PATCH v2] netconsole: various improvements

[Tom Rini]

On Mon, Apr 24, 2023 at 07:06:08PM -0700, Tony Dinh wrote:
> Hi Simon,
> 
> On Mon, Apr 24, 2023 at 12:43 PM Simon Glass <sjg at chromium.org> wrote:
> >
> > Hi Tony,
> >
> > On Wed, 19 Apr 2023 at 21:55, Tony Dinh <mibodhi at gmail.com> wrote:
> > >
> > > On Wed, Apr 19, 2023 at 6:22 PM Tony Dinh <mibodhi at gmail.com> wrote:
> > > >
> > > > HI Simon,
> > > >
> > > > On Tue, Apr 18, 2023 at 6:46 PM Simon Glass <sjg at chromium.org> wrote:
> > > > >
> > > > > Hi Tony,
> > > > >
> > > > > On Mon, 3 Apr 2023 at 15:42, Tony Dinh <mibodhi at gmail.com> wrote:
> > > > > >
> > > > > > Use CONFIG_CONSOLE_MUX for netconsole. When netconsole is running,
> > > > > > stdin/stdout/stder must be set to some primary console, in addtion to nc.
> > > > > > For example, stdin=serial,nc. Some recent Linux kernels will not boot with
> > > > > > only nc on the stdout list, ie. stdout=nc. When netconsole exits, remove
> > > > > > nc from the list of devices in stdin/stdout/stderr.
> > > > > >
> > > > > > Signed-off-by: Tony Dinh <mibodhi at gmail.com>
> > > > > > ---
> > > > > >
> > > > > > Changes in v2:
> > > > > > - Select CONFIG_CONSOLE_MUX if CONFIG_NETCONSOLE is enabled
> > > > > > - Add new functions in netconsole driver to support CONSOLE_MUX
> > > > > > - Add new function to encapsulate the process of stopping netconsole
> > > > > > from bootm
> > > > > > - Remove unecessary net_timeout initial value = 1
> > > > > > - Resend to correct missing <stdio_dev.h> include in bootm.c
> > > > > >
> > > > > >  boot/bootm.c             | 14 +++++++---
> > > > > >  drivers/net/Kconfig      | 10 +++++++
> > > > > >  drivers/net/netconsole.c | 60 ++++++++++++++++++++++++++++++++++++++++
> > > > > >  include/stdio_dev.h      |  1 +
> > > > > >  4 files changed, 81 insertions(+), 4 deletions(-)
> > > > >
> > > > > This seems OK to me but for a few thoughts below.
> > > > >
> > > > > But can we use this opportunity to move this to driver model? The
> > > > > netconsole driver could be bound in eth_post_bind() and the code in
> > > > > netconsole.c converted to be a driver.
> > > > >
> > > > > I think that would be better than building on an out-of-date driver.
> > > >
> > > > I'd agree that converting to a driver model is the logical next step.
> > > > My motivation is to fix the booting problem for the Linux kernel first
> > > > (I'm having problems booting some kernels without this patch). And
> > > > then in the next go round, I'll convert netconsole to a driver. Most
> > > > of the code in this small patch will be needed whether it is an old
> > > > driver or DM anyway.
> >
> > This crosses the line to a feature, IMO...
> 
> My initial patch would be a pure bug fix. But it also breaks other
> boards like Nokia, as Pali has pointed out. So in the v2 patch,
> Console Mux is introduced to cover all boards. I don't think that we
> are adding any features here.
> 
> It's just that I'm using netconsole often so I noticed the booting
> problem with recent Linux kernels. It's kind of a catch-22, without a
> serial console I cannot debug Linux early booting problems, and with
> serial console active there is no booting problem.

Yes, we should fix functionality then work on clean-ups.

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20230424/9c34acf4/attachment.sig>