GSOC:Microproject

[Eugene Syromiatnikov]

On Thu, Mar 08, 2018 at 12:22:53AM +0530, Abhishek Tiwari wrote:
> On Mon, Mar 5, 2018 at 3:27 PM, Gabriel Laskar <gabriel at lse.epita.fr> wrote:
> 
> > On Mon, 5 Mar 2018 14:31:16 +0530
> > Abhishek Tiwari <erabhishektiwarics at gmail.com> wrote:
> >
> > > Please do review patch.
> >
> > It will be way simpler to review with the patch inlined, and not
> > attached to your mail.
> >
> > As a starter:
> >
> > * your commit message does not explain anything, a better message would
> >   be something like what you have in the NEWS file
> > * The NEWS file is not updated for the correct version. Also, most of
> >   the time, it is the maintainers that adds the digest of the version
> >   here.
> >
> > Made changes as suggested. Patch not attached.
> 
> * you did not answer to (message-id:
> >   <20180219142319.GJ4439 at asgard.redhat.com>)
> >
> 
> In this message, it is suggested to make class of different type of system
> calls and then user can specify which class to trace.

No, the question was the following:
---8<---
On Thu, Feb 15, 2018 at 07:07:59PM +0530, Abhishek Tiwari wrote:
> Hi,
> 
> I want to work on microproject
> 
> extend -e trace=class syntax with new classes
> 
> 
> I have gone through mailing list. Is following grouping ok
> 
> {,l,f}setxattr           TRACE_SETXATTR  TFX
Doesn't -e trace='/[lf]?setxattr' cover precisely this set of syscalls
on any architecture?
--->8---

The same question remains with the provided patch, as the syscall set
can be covered by -e trace=/xattr.  Note that that was the exact reason
for the commits v4.17~97 and v4.17~96.

> Last year I completed same microproject to trace statfs-like system calls
> and different hierarchy of system calls were define using % and %% i.e. %
> would define more specific system call and %% would define general group of
> system call.
> 
> * Did you started to take into account remarks from  Masatake YAMATO
> >   and Eugene Syromiatnikov (discussion starts at message-id:
> >   <20180223.174751.218891701122588475.yamato at redhat.com> ) about user
> >   defined groups instead of hardocoded ones ?
> >
> 
> How should we approach for user-defined group of system call?
> 
> For microproject should I submit patch for tracing xattr system calls or I
> should work on user-defined group of system call?

Well, at least to me, the *xattr syscall group looks redundant due to
the aforementioned reason.