This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
Re: [patch] For getrusage's argstr in tapset
Hello, Wenji.
Within the syscall getrusage, there are only 2 types of flag.
So in my opinion, all the other values are not getrusage's flag and should
be printed as UNKNOWN.
To output a clearer description, we need also print the value of flag in
the case of "UNKNOWN".
So I think it will be better if it can be revised in my way.
Regards
Zhaolei
----- Original Message -----
From: "Wenji Huang" <wenji.huang@oracle.com>
To: "Zhaolei" <zhaolei@cn.fujitsu.com>
Cc: <systemtap@sourceware.org>
Sent: Monday, September 10, 2007 2:18 PM
Subject: Re: [patch] For getrusage's argstr in tapset
> Hi zhaolei,
>
> In normal case, thing may be like that. But in abnormal case, maybe
> wrong arguments are passed to this function. So new entry will
> give a clearer description, instead of general UNKNOWN.
>
> Thanks,
> Wenji
>
> Zhaolei wrote:
>
> >Hello, Wenji.
> >
> >Thank you for your suggestion.
> >
> >I investigated RUSAGE_BOTH and got the following result:
> >1: _rusage_who_str is only used for sys_getrusage(and compat_sys_getrusage)
> > in current tapset, and there is no other syscall which needs to call
> > _rusage_who_str.
> >
> >2: RUSAGE_BOTH is only used in kernel's internal function getrusage, but
> > there is no syscall using this function with RUSAGE_BOTH.
> >
> >So I think RUSAGE_BOTH is not necessary in tracing syscall, but it may be
> > needed when tracing kernel function as getrusage.
> >
> >It is good idea disable RUSAGE_BOTH only in sys_getrusage
> >(and compat_sys_getrusage).
> >
> >Regards
> >Zhaolei
> >----- Original Message -----
> >From: "Wenji Huang" <wenji.huang@oracle.com>
> >To: "Zhaolei" <zhaolei@cn.fujitsu.com>
> >Cc: <systemtap@sourceware.org>
> >Sent: Monday, September 10, 2007 11:44 AM
> >Subject: Re: [patch] For getrusage's argstr in tapset
> >
> >
> >
> >
> >>Hi Zhaolei,
> >>
> >> probe getrusage set probe point at the entry of the function. So the
> >>arguments may be any value.
> >>
> >> So I think to keep the entry in _rusage_who_str, just give another
> >>mapping in case of some certain value.
> >>
> >> Of course, if you can ensure that "who" only be
> >>RUSAGE_SELF/RUSAGE_CHILDREN when the syscall called. Just remove.
> >>
> >>Thanks,
> >>Wenji
> >>
> >>
> >>Zhaolei wrote:
> >>
> >>
> >>
> >>>Hi, everyone
> >>>
> >>>In [tapset->probe getrusage], argstr's "who" argument is printed as
> >>>"RUSAGE_BOTH" when its value is -2.
> >>>But "RUSAGE_BOTH" is not used in getrusage syscall due to the following
> >>>kernel source:
> >>>asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
> >>>{
> >>> if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
> >>> return -EINVAL;
> >>> return getrusage(current, who, ru);
> >>>}
> >>>
> >>>So i want to remove RUSAGE_BOTH from _rusage_who_str.
> >>>Is there anyone knows why RUSAGE_BOTH is necessary, please tell me.
> >>>and if no objection, I will commit it.
> >>>
> >>>Signed-off-by: "Zhaolei" zhaolei@cn.fujitsu.com
> >>>
> >>>--- aux_syscalls.stp.old 2007-09-06 11:51:14.000000000 +0900
> >>>+++ aux_syscalls.stp 2007-09-06 11:52:22.000000000 +0900
> >>>@@ -1661,7 +1661,6 @@ function _rlimit_resource_str(r) {
> >>>function _rusage_who_str(w) {
> >>> if(w==0) return "RUSAGE_SELF"
> >>> if(w==-1) return "RUSAGE_CHILDREN"
> >>>- if(w==-2) return "RUSAGE_BOTH"
> >>> return sprintf("UNKNOWN VALUE: %d", w)
> >>>}
> >>>
> >>>Regards
> >>>Zhaolei
> >>>
> >>>
> >>>
> >>>
> >>>
> >>
> >>
> >
> >
> >
>
>