[Bug tapsets/15690] New: syscall.rt_sigprocmask probe alias broken on s390x/ppc64 for 32-bit apps

["dsmith at redhat dot com" <sourceware-bugzilla at sourceware dot org>]

http://sourceware.org/bugzilla/show_bug.cgi?id=15690

            Bug ID: 15690
           Summary: syscall.rt_sigprocmask probe alias broken on
                    s390x/ppc64 for 32-bit apps
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: tapsets
          Assignee: systemtap at sourceware dot org
          Reporter: dsmith at redhat dot com

On s390x/ppc64, the syscall.rt_sigprocmask probe alias is broken when probing
32-bit apps, always displaying '[EMPTY]' for the sigset value.

Here's the definition of syscall.rt_sigprocmask from
tapset/linux/syscalls2.stp:

====
# rt_sigprocmask _____________________________________________
# long sys32_rt_sigprocmask(u32 how, compat_sigset_t __user *set,
compat_sigset_t __user *oset, size_t sigsetsize)
# long compat_sys_rt_sigprocmask(int how, compat_sigset_t __user *set,
compat_sigset_t __user *oset, compat_size_t sigsetsize)
# long sys_rt_sigprocmask(int how, sigset_t __user *set, sigset_t __user *oset,
size_t sigsetsize)
#
probe syscall.rt_sigprocmask = kernel.function("sys32_rt_sigprocmask").call ?,
                          kernel.function("compat_sys_rt_sigprocmask").call ?,
                          kernel.function("sys_rt_sigprocmask").call ?
{
    name = "rt_sigprocmask"
    how  = $how
    how_str = _sigprocmask_how_str($how)
    set_uaddr    = @choose_defined($set, $nset)
    oldset_uaddr = $oset
    argstr = sprintf("%s, [%s], %p, %d", how_str, _stp_sigset_u(set_uaddr),
        $oset, $sigsetsize)
}
====

The problem is that the 'compat' entry points (compat_sys_rt_sigprocmask and
sys_rt_sigprocmask) don't get a 'sigset_t', they get a 'compat_sigset_t'. On
x86_64, a 'compat_sigset_t' is laid out similarly to a 'sigset_t'. On
big-endian platforms like ppc64 and s390x, a 'compat_sigset_t' is laid out
differently than a 'sigset_t'. Since _stp_sigset_u() isn't expecting a
'compat_sigset_t', it doesn't handle it correctly.

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