[Bug tapsets/14325] New: for arm/s390x, the nd_syscall tapsets need improvement

[Bug tapsets/14325] New: for arm/s390x, the nd_syscall tapsets need improvement

[dsmith at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14325

             Bug #: 14325
           Summary: for arm/s390x, the nd_syscall tapsets need improvement
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: tapsets
        AssignedTo: systemtap@sourceware.org
        ReportedBy: dsmith@redhat.com
    Classification: Unclassified


The nd_syscall tapsets use the *_arg() functions to get syscall arguments. 
However, arm can only get the 1st 4 arguments and s390x can only get the 1st 6
arguments.  The nd_syscall tapset tries to get the 7th argument in several
probe aliases, which will fail.

Commit 65ddca0 is a hack for s390x that supplies a dummy value for 2 probe
aliases that are used heavily by the nd_syscall testsuite.

But, a better fix is needed for arm and s390x.

Possible solutions include:

- Have the *_arg() functions return a dummy value when asked for an argument it
can't handle.

- Have each architecture provide a function, something like syscall_max_args(),
that would provide the maximum argument that this architecture supports.  This
function could be tested in each probe alias, like this:

  argstr = sprintf("%d %d %d %d %d", int_arg(1), int_arg(2), int_arg(3),
           int_arg(4), (syscall_max_args() >= 5 ? int_arg(5) : -1))

- Continue like commit 65ddca0 and just do architecture checks everywhere
needed.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug tapsets/14325] for arm/s390x, the nd_syscall tapsets need improvement

[dsmith at redhat dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=14325

--- Comment #1 from David Smith <dsmith at redhat dot com> 2012-07-05 13:57:41 UTC ---
(In reply to comment #0)
> Possible solutions include:
> 
> - Have the *_arg() functions return a dummy value when asked for an argument it
> can't handle.
> 
> - Have each architecture provide a function, something like syscall_max_args(),
> that would provide the maximum argument that this architecture supports.  This
> function could be tested in each probe alias, like this:
> 
>   argstr = sprintf("%d %d %d %d %d", int_arg(1), int_arg(2), int_arg(3),
>            int_arg(4), (syscall_max_args() >= 5 ? int_arg(5) : -1))
> 
> - Continue like commit 65ddca0 and just do architecture checks everywhere
> needed.

Another solution would be to define variants of the *_arg() funtions that take
a default value (to be used when the argument isn't available).  It could look
something like this (where '0' is the default value for arg 4 and -1 is the
default value for arg 5):

  argstr = sprintf("%d %d %d %d %d", int_arg(1), int_arg(2), int_arg(3),
           int_arg_default(4, 0), int_arg(5, -1))

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

[Bug tapsets/14325] for arm, the nd_syscall tapsets need improvement

[dsmith at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=14325

David Smith <dsmith at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|for arm/s390x, the          |for arm, the nd_syscall
                   |nd_syscall tapsets need     |tapsets need improvement
                   |improvement                 |

--- Comment #2 from David Smith <dsmith at redhat dot com> ---
For s390x, bug #15913 fixed getting arguments off the stack for s390x. So, I
just reverted commit 65ddca0 (in commit 9241707).

This means this bug only applies to arm now.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug tapsets/14325] for arm, the nd_syscall tapsets need improvement

[dsmith at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=14325

David Smith <dsmith at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #3 from David Smith <dsmith at redhat dot com> ---
With the fix for bug #13481, we can now close this bug.

-- 
You are receiving this mail because:
You are the assignee for the bug.