Re: [PATCH v2 03/15] drivers: thermal: tsens: Add __func__ identifier to debug statements

[Daniel Thompson]

On Thu, Aug 29, 2019 at 09:44:04PM +0530, Amit Kucheria wrote:
> On Thu, Aug 29, 2019 at 8:49 PM Daniel Thompson
> <daniel.thompson@xxxxxxxxxx> wrote:
> >
> > On Thu, Aug 29, 2019 at 07:58:45PM +0530, Amit Kucheria wrote:
> > > On Thu, Aug 29, 2019 at 7:35 PM Daniel Thompson
> > > <daniel.thompson@xxxxxxxxxx> wrote:
> > > >
> > > > On Tue, Aug 27, 2019 at 05:43:59PM +0530, Amit Kucheria wrote:
> > > > > Printing the function name when enabling debugging makes logs easier to
> > > > > read.
> > > > >
> > > > > Signed-off-by: Amit Kucheria <amit.kucheria@xxxxxxxxxx>
> > > > > Reviewed-by: Stephen Boyd <swboyd@xxxxxxxxxxxx>
> > > > > Reviewed-by: Daniel Lezcano <daniel.lezcano@xxxxxxxxxx>
> > > >
> > > > This should need to be manually added at each call site; it is already
> > > > built into the logging system (the f flag for dynamic debug)?
> > >
> > > I assume you meant "shouldn't".
> >
> > Quite so. Sorry about that.
> >
> > > I haven't yet integrated dynamic debug into my daily workflow.
> > >
> > > Last time I looked at it, it was a bit bothersome to use because I
> > > needed to lookup exact line numbers to trigger useful information. And
> > > those line numbers constantly keep changing as I work on the driver,
> > > so it was a bit painful to script. Not to mention the syntax to frob
> > > the correct files in debugfs to enable this functionality.
> > >
> > > As opposed to this, adding the following to the makefile is so easy. :-)
> > >
> > > CFLAGS_tsens-common.o          := -DDEBUG
> > >
> > > Perhaps I am using it all wrong? How would I go about using dynamic
> > > debug instead of this patch?
> >
> > Throwing dyndbg="file <fname>.c +pf" onto the kernel command line is a
> > good start (+p enables debug level prints, +f causes messages to include
> > the function name).
> 
> That's useful to know.
> 
> $ git grep __func__ | wc -l
> 30914
> 
> Want to send some patches? :-)

I know. Sad isn't it?

To be fair plenty of patches already circulate tidying up this sort of
thing (along with the removal of inane messages such as informing us
that a function run).


> > When the C files map to module names (whether the modules are actually
> > built-in or not) then <module>.dyndbg=+pf is a bit cleaner and allows
> > you to debug the whole of a driver without how it is decomposed into
> > files.
> 
> And if changing the kernel cmdline options isn't possible or is inconvenient?

Architectures where this problem offer CONFIG_CMDLINE_FORCE meaning if
you are already building a custom kernel you can override whatever
cmdline the bootloader gives you.


> > There are (many) other controls to play with[1] but the above should be
> > sufficient to simulate -DDEBUG .
> 
> The "hard" bit is explicitly poking the line number in a file to
> activate a paricular pr_dbg statement. Even if I scripted it, those
> lines numbers keep changing in an actively developed driver.

Line numbers? Nothing I suggested contained a line number.


Daniel.

> 
> Somehow, I've always felt dyndbg was more useful to debug a production
> system where recompiling the kernel wasn't an option e.g. reporting an
> issue back to a distro-kernel vendor.
> 
> > Daniel.
> >
> > [1]
> > https://www.kernel.org/doc/html/latest/admin-guide/dynamic-debug-howto.html