[Bug gdb/11842] New: compat_siginfo_from_siginfo and siginfo_from_compat_siginfo are wrong

[Bug gdb/11842] New: compat_siginfo_from_siginfo and siginfo_from_compat_siginfo are wrong

[jan dot kratochvil at redhat dot com]

Only biarch gdb is affected.

#include <unistd.h>
#include <sys/syscall.h>
#include <signal.h>
#define tgkill(tgid, tid, sig) syscall (__NR_tgkill, tgid, tid, sig)
int
main (void)
{
  return tgkill (getpid (), getpid (), SIGUSR1);
}

gcc -o 1 1.c -Wall -g -m32

GNU gdb (GDB) 7.2.50.20100726-cvs
This GDB was configured as "x86_64-unknown-linux-gnu".
gcc-4.4.4-10.fc13.x86_64
kernel-2.6.33.6-147.fc13.x86_64

./gdb -nx -ex start -ex 'p getpid ()' -ex c -ex 'p
$_siginfo._sifields._kill.si_pid' ./1
[...]
$1 = 23172
Continuing.

Program received signal SIGUSR1, User defined signal 1.
0x00110430 in __kernel_vsyscall ()
$2 = 0

$1 and $2 are the same for native x86_64 and for native i686.
$1 and $2 SHOULD be the same even for i686 on x86_64 debugger.

compat_siginfo_from_siginfo and siginfo_from_compat_siginfo are wrong in:
gdb/gdbserver/linux-x86-low.c
gdb/amd64-linux-nat.c

-- 
           Summary: compat_siginfo_from_siginfo and
                    siginfo_from_compat_siginfo are wrong
           Product: gdb
           Version: unknown
            Status: UNCONFIRMED
          Severity: minor
          Priority: P2
         Component: gdb
        AssignedTo: unassigned at sourceware dot org
        ReportedBy: jan dot kratochvil at redhat dot com
                CC: gdb-prs at sourceware dot org
GCC target triplet: x86_64-fedora13-linux-gnu


http://sourceware.org/bugzilla/show_bug.cgi?id=11842

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/11842] compat_siginfo_from_siginfo and siginfo_from_compat_siginfo are wrong

[jan dot kratochvil at redhat dot com]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   GCC host triplet|                            |x86_64-fedora13-linux-gnu


http://sourceware.org/bugzilla/show_bug.cgi?id=11842

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/11842] compat_siginfo_from_siginfo and siginfo_from_compat_siginfo are wrong

[pedro at codesourcery dot com]

------- Additional Comments From pedro at codesourcery dot com  2010-08-06 12:28 -------
This patch works for me.  Did you spot anything else wrong?

	gdb/
	* amd64-linux-nat.c (compat_siginfo_from_siginfo)
	(siginfo_from_compat_siginfo): Also copy si_pid and si_uid when
	si_code is < 0.

	gdb/gdbserver/
	* linux-x86-low.c (compat_siginfo_from_siginfo)
	(siginfo_from_compat_siginfo): Also copy si_pid and si_uid when
	si_code is < 0.

---
 gdb/amd64-linux-nat.c         |    4 ++++
 gdb/gdbserver/linux-x86-low.c |    4 ++++
 2 files changed, 8 insertions(+)

Index: src/gdb/amd64-linux-nat.c
===================================================================
--- src.orig/gdb/amd64-linux-nat.c	2010-08-06 13:19:55.000000000 +0100
+++ src/gdb/amd64-linux-nat.c	2010-08-06 13:21:05.000000000 +0100
@@ -576,6 +576,8 @@ compat_siginfo_from_siginfo (compat_sigi
 
   if (to->si_code < 0)
     {
+      to->cpt_si_pid = from->si_pid;
+      to->cpt_si_uid = from->si_uid;
       to->cpt_si_ptr = (intptr_t) from->si_ptr;
     }
   else if (to->si_code == SI_USER)
@@ -630,6 +632,8 @@ siginfo_from_compat_siginfo (siginfo_t *
 
   if (to->si_code < 0)
     {
+      to->si_pid = from->cpt_si_pid;
+      to->si_uid = from->cpt_si_uid;
       to->si_ptr = (void *) (intptr_t) from->cpt_si_ptr;
     }
   else if (to->si_code == SI_USER)
Index: src/gdb/gdbserver/linux-x86-low.c
===================================================================
--- src.orig/gdb/gdbserver/linux-x86-low.c	2010-08-06 13:19:55.000000000 +0100
+++ src/gdb/gdbserver/linux-x86-low.c	2010-08-06 13:21:05.000000000 +0100
@@ -749,6 +749,8 @@ compat_siginfo_from_siginfo (compat_sigi
 
   if (to->si_code < 0)
     {
+      to->cpt_si_pid = from->si_pid;
+      to->cpt_si_uid = from->si_uid;
       to->cpt_si_ptr = (intptr_t) from->si_ptr;
     }
   else if (to->si_code == SI_USER)
@@ -803,6 +805,8 @@ siginfo_from_compat_siginfo (siginfo_t *
 
   if (to->si_code < 0)
     {
+      to->si_pid = from->cpt_si_pid;
+      to->si_uid = from->cpt_si_uid;
       to->si_ptr = (void *) (intptr_t) from->cpt_si_ptr;
     }
   else if (to->si_code == SI_USER)


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1


http://sourceware.org/bugzilla/show_bug.cgi?id=11842

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/11842] compat_siginfo_from_siginfo and siginfo_from_compat_siginfo are wrong

[jan dot kratochvil at redhat dot com]

------- Additional Comments From jan dot kratochvil at redhat dot com  2010-08-06 12:41 -------
I find the conditional `to->si_code < 0' wrong there.

See for example:
drivers/usb/core/devio.c
                sinfo.si_code = SI_ASYNCIO;
                sinfo.si_addr = as->userurb;
include/asm-generic/siginfo.h
#define SI_ASYNCIO      -4              /* sent by AIO completion */

But you do not convert si_addr here.
Moreover when you start converting more fields you cannot - as they share the
same places in the union.

One has to probably investigate all the kernel points generating siginfo. :-/


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=11842

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/11842] compat_siginfo_from_siginfo and siginfo_from_compat_siginfo are wrong

[pedro at codesourcery dot com]

------- Additional Comments From pedro at codesourcery dot com  2010-08-06 13:31 -------
Yeah, though in practice si_addr ends up at the same address as si_pid to it is
already copied.  Note that that's driver code.  Who's to say what random drivers
do?  I'm thinking that we only need to be as thorough/good as 
arch/x86/ia32/ia32_signal.c:{copy_siginfo_to_user32,copy_siginfo_from_user32},
and that also has a catch-all si_code < 0 branch.

(I see that in our version, the "to->si_code == SI_TIMER" branch is
unreacheable, as SI_TIMER is < 0.  whoops.)


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=11842

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/11842] compat_siginfo_from_siginfo and siginfo_from_compat_siginfo are wrong

[jan dot kratochvil at redhat dot com]

------- Additional Comments From jan dot kratochvil at redhat dot com  2010-08-06 13:35 -------
(In reply to comment #3)
> I'm thinking that we only need to be as thorough/good as 
> arch/x86/ia32/ia32_signal.c:{copy_siginfo_to_user32,copy_siginfo_from_user32},

OK, yes, I agree, thanks.


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=11842

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/11842] compat_siginfo_from_siginfo and siginfo_from_compat_siginfo are wrong

[jan dot kratochvil at redhat dot com]

------- Additional Comments From jan dot kratochvil at redhat dot com  2010-08-30 08:47 -------
Post:
[patch 4/9]#2 Pedro's fix: compat_siginfo_from_siginfo (PR 11842)
http://sourceware.org/ml/gdb-patches/2010-08/msg00484.html


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=11842

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug gdb/11842] compat_siginfo_from_siginfo and siginfo_from_compat_siginfo are wrong

[pedro at codesourcery dot com]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2010-09-24 13:44 -------
Subject: Bug 11842

CVSROOT:	/cvs/src
Module name:	src
Changes by:	palves@sourceware.org	2010-09-24 13:44:02

Modified files:
	gdb            : ChangeLog 
	gdb/gdbserver  : ChangeLog 

Log message:
	PR gdb/11842
	
	gdb/
	* amd64-linux-nat.c (compat_siginfo_from_siginfo)
	(siginfo_from_compat_siginfo): Also copy si_pid and si_uid when
	si_code is < 0.  Check for si_code == SI_TIMER before checking for
	si_code < 0.
	
	gdb/gdbserver/
	* linux-x86-low.c (compat_siginfo_from_siginfo)
	(siginfo_from_compat_siginfo): Also copy si_pid and si_uid when
	si_code is < 0.  Check for si_code == SI_TIMER before checking for
	si_code < 0.

Patches:
http://sourceware.org/cgi-bin/cvsweb.cgi/src/gdb/ChangeLog.diff?cvsroot=src&r1=1.12200&r2=1.12201
http://sourceware.org/cgi-bin/cvsweb.cgi/src/gdb/gdbserver/ChangeLog.diff?cvsroot=src&r1=1.432&r2=1.433


------- Additional Comments From pedro at codesourcery dot com  2010-09-29 09:03 -------
Patch checked in.


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


http://sourceware.org/bugzilla/show_bug.cgi?id=11842

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.