[Bug translator/7045] New: user-space probe x86 on x86-64 host

[Bug translator/7045] New: user-space probe x86 on x86-64 host

[fche at redhat dot com]

Current checks in tapsets.cxx:validate_module_elf do not consider
the valid possibility of probing an x86 binary on an x86-64 kernel,
and unfairly rejects debuginfo/elf processing for them.

(The same problem probably exists on ppc.)

-- 
           Summary: user-space probe x86 on x86-64 host
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: translator
        AssignedTo: systemtap at sources dot redhat dot com
        ReportedBy: fche at redhat dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=7045

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

[Bug translator/7045] user-space probe x86 on x86-64 host

[wenji dot huang at oracle dot com]

------- Additional Comments From wenji dot huang at oracle dot com  2008-12-03 08:47 -------
sorry, don't catch it well. I tried simple test on FC9-64 bits.
gcc -m32 -o /tmp/test /tmp/test.c
/tmp/test&
stap -vvve 'probe process("/tmp/test").syscall{}' 

Everything is OK. And validate_module_elf wasn't be called neither.

Could you elaborate it a little or provide some test cases? 


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=7045

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

[Bug translator/7045] user-space probe x86 on x86-64 host

[fche at redhat dot com]

------- Additional Comments From fche at redhat dot com  2008-12-05 17:06 -------
(In reply to comment #1)
> Could you elaborate it a little or provide some test cases? 

On my fc9 x86-64 machine, I get this.  Maybe your 'gcc -m32' doesn't work?

[12:03:46] % cat > hello.c
int main () {
}

[12:04:30] % gcc -g -m32 hello.c -o hello

[12:04:33] % file hello
hello: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically
linked (uses shared libs), for GNU/Linux 2.6.9, not stripped

[12:04:34] % sudo ./stap -e 'probe process("./hello").function("*").call {}' -vwt
Pass 1: parsed user script and 47 library script(s) in 160usr/20sys/178real ms.
semantic error: ELF machine i?86 (code 3) mismatch with target x86_64 in
'/home/fche/Private/DEVEL/DEVEL-systemtap/git/BUILD/hello'
semantic error: no match while resolving probe point
process("./hello").function("*").call
semantic error: no probes found
Pass 2: analyzed script: 0 probe(s), 0 function(s), 0 embed(s), 0 global(s) in
0usr/0sys/3real ms.
Pass 2: analysis failed.  Try again with another '--vp 01' option.


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=7045

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

[Bug translator/7045] user-space probe x86 on x86-64 host

[jkenisto at us dot ibm dot com]

------- Additional Comments From jkenisto at us dot ibm dot com  2008-12-05 20:39 -------
It's broken as described by FChE on my x86_64 system, too.  powerpc was
similarly broken (PR 6946), but I fixed it.  Unfortunately, that fix won't work
for the x86 problem.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=7045

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

[Bug translator/7045] user-space probe x86 on x86-64 host

[wenji dot huang at oracle dot com]

------- Additional Comments From wenji dot huang at oracle dot com  2008-12-08 03:24 -------
Yeah
stap -e 'probe process("./hello").function("*").call {}'  failed due to
mismatched elf.
But 
stap -e 'probe process("./hello").syscall {}'  gets rid of the checking  elf.



-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=7045

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

[Bug translator/7045] user-space probe x86 on x86-64 host

[dsmith at redhat dot com]

------- Additional Comments From dsmith at redhat dot com  2008-12-08 15:33 -------
(In reply to comment #4)
> Yeah
> stap -e 'probe process("./hello").function("*").call {}'  failed due to
> mismatched elf.
>
> But 
> stap -e 'probe process("./hello").syscall {}'  gets rid of the checking  elf.

Here's a small clarification of the last example above (the ".syscall" one). 
That example doesn't check the debuginfo because it doesn't need debuginfo
(internally that example uses utrace which doesn't require debuginfo).  So, the
fact that the .syscall probe doesn't check matching debuginfo is by design.


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=7045

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

[Bug translator/7045] user-space probe x86 on x86-64 host

[fche at redhat dot com]

------- Additional Comments From fche at redhat dot com  2009-03-27 17:04 -------
commit 756c946

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


http://sourceware.org/bugzilla/show_bug.cgi?id=7045

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