Re: [PATCH][RESEND] fuse: add simple request tracepoints

[Josef Bacik <josef@toxicpanda.com>]

On Thu, Jul 04, 2024 at 07:20:16PM +0200, Bernd Schubert wrote:
> 
> 
> On 7/3/24 16:38, Josef Bacik wrote:
> > I've been timing various fuse operations and it's quite annoying to do
> > with kprobes.  Add two tracepoints for sending and ending fuse requests
> > to make it easier to debug and time various operations.
> 
> Thanks, this is super helpful.
> 
> [...]
> > 
> > +	EM( FUSE_STATX,			"FUSE_STATX")		\
> > +	EMe(CUSE_INIT,			"CUSE_INIT")
> > +
> > +/*
> > + * This will turn the above table into TRACE_DEFINE_ENUM() for each of the
> > + * entries.
> > + */
> > +#undef EM
> > +#undef EMe
> > +#define EM(a, b)	TRACE_DEFINE_ENUM(a);
> > +#define EMe(a, b)	TRACE_DEFINE_ENUM(a);
> 
> 
> I'm not super familiar with tracepoints and I'm a bit list why "EMe" is
> needed
> in addition to EM? CUSE_INIT is just another number?

This is just obnoxious preprocessor abuse, so you're right this first iteration
of EMe() is the same as EM(), but if you look right below that you have

/* Now we redfine it with the table that __print_symbolic needs. */
#undef EM
#undef EMe
#define EM(a, b)        {a, b},
#define EMe(a, b)       {a, b}

so later when we do

__print_symbolic(__entry->opcode, OPCODES)

OPCODES gets turned intoo

__print_symbolic(__entry->opcode,
		{FUSE_LOOKUP, "FUSE_LOOKUP"},{...},{CUSE_INIT, "CUSE_INIT"})

it's subtle and annoying, but the cleanest way to have these big opcode tables
that are easy to add/remove stuff from for clean output.  Thanks,

Josef