Re: [PATCH net-next v2 3/8] net: netconsole: move newline trimming to function

[Matthew Wood]

On Wed, Jan 31, 2024 at 8:45 PM Matthew Wood <thepacketgeek@xxxxxxxxx> wrote:
>
> On Tue, Jan 30, 2024 at 1:16 AM Breno Leitao <leitao@xxxxxxxxxx> wrote:
> >
> > On Fri, Jan 26, 2024 at 03:13:38PM -0800, Matthew Wood wrote:
> > > Move newline trimming logic from `dev_name_store()` to a new function
> > > (trim_newline()) for shared use in netconsole.c
> > >
> > > Signed-off-by: Matthew Wood <thepacketgeek@xxxxxxxxx>
> > > ---
> > >  drivers/net/netconsole.c | 17 +++++++++++------
> > >  1 file changed, 11 insertions(+), 6 deletions(-)
> > >
> > > diff --git a/drivers/net/netconsole.c b/drivers/net/netconsole.c
> > > index 085350beca87..b280d06bf152 100644
> > > --- a/drivers/net/netconsole.c
> > > +++ b/drivers/net/netconsole.c
> > > @@ -230,6 +230,16 @@ static struct netconsole_target *to_target(struct config_item *item)
> > >                           struct netconsole_target, group);
> > >  }
> > >
> > > +/* Get rid of possible trailing newline, returning the new length */
> > > +static void trim_newline(char *s, size_t maxlen)
> > > +{
> > > +     size_t len;
> > > +
> > > +     len = strnlen(s, maxlen);
> > > +     if (s[len - 1] == '\n')
> > > +             s[len - 1] = '\0';
> > > +}
> >
> > I am thinking about this one. Should we replace the first `\n` in the
> > file by `\0` no matter where it is? This will probably make it easier to
> > implement the netconsd, where we know it will be impossible to have `\n`
> > in the userdata.
> >
> > Maybe something as:
> >
> >         static inline void trim_newline(char *str)
> >         {
> >                 char *pos = strchr(str, '\n');
> >
> >                 if (pos)
> >                         *pos = '\0';
> >         }
> >
> >
> > All in all, this is a good clean up, which make the code easier to read.
> > Thanks!
>
> I like this idea, I agree that only accepting userdata values upto the
> first newline clears up the expectations for log output and parsing on
> the receiving side. I would prefer that to the case where multiple
> values (delimited by newlines) are somehow attempted with a single
> key, seems like just using additional key/value pairs would be
> cleaner.

In practice truncating at the first newline makes no difference as
printk, echo, and other methods seem to buffer and write per-line. So
in this example, the stored value will be "val2" with or without the
suggested change:

$ printf "val1\nval2" > userdata/testing/value
# This results in two calls to userdatum_value_store, the first with
"val1\n" and the second with "val2". "val2" remains as the latest
write.
$ cat userdata/testing/value
val2

I will add a warning about this possibly unexpected behavior in the
docs for v3 for the patch