Re: [LTP] [PATCH] syscalls: fix some failure on arch X86_64

[liubo <liubo2009@cn.fujitsu.com>]

[-- Attachment #1.1: Type: text/plain, Size: 12575 bytes --]

Hi, Garrett
On 11/26/2009 07:11 PM, Garrett Cooper wrote:
> On Mon, Nov 16, 2009 at 12:53 AM, liubo <liubo2009@cn.fujitsu.com> wrote:
>   
>> Hi, Subrata,
>> On 11/16/2009 04:13 PM, Subrata Modak wrote:
>>
>> Liubo,
>>
>> Are you sending an updated patch post this discussion ?
>>
>>
>>
>> Yes, I'll resend a updated patch about these rt_sigs.
>>
>> Regards--
>> Liubo
>>
>> Regards--
>> Subrata
>>
>> On Wed, 2009-11-11 at 13:03 +0800, Wei Yongjun wrote:
>>
>>
>> Mike Frysinger wrote:
>>
>>
>> On Tuesday 10 November 2009 04:38:30 liubo wrote:
>>
>>
>>
>> 1) rt_sigaction
>>     "sigaction" has the structure:
>>
>>  struct sigaction {
>>          __sighandler_t sa_handler;
>>          unsigned long sa_flags;
>>   #ifdef SA_RESTORER
>>           __sigrestore_t sa_restorer;
>>   #endif
>>           sigset_t sa_mask;               /* mask last for extensibility */
>>  };
>>
>>     However, on arch x86_64, if we directly get to call rt_sigaction,
>> the argument "sa_restorer" will not be fulfilled, and this will lead
>>  to segment fault.
>>     on arch x86_64, if sa_restorer is not set, kernel will lead to segment
>>  fault. In other arch, if sa_restorer is not set, kernel can do the correct
>>  work. To avoid this segment fault, we use glibc function
>> "int sigaction(...);" instead, which can fulfill the argument
>>  "sa_restorer".
>>
>>
>>
>> which defeats the purpose of the test.  there is no guarantee that the C
>> library sigaction function is implemented via the __NR_rt_sigaction syscall.
>>
>>
>>
>> In x86_64, it do this. And If we want to use __NR_rt_sigaction syscall
>> directly, we must fill the sa_restorer and set the RESTORER flag to
>> sa_mask. If we do not set the sa_restorer, kill will always cause
>> segment fault.
>>
>>
>>
>>
>>
>>
>> 2) rt_sigprocmask
>>     This failure contains two aspects,
>> the first is the segment fault as described in 1),
>> the second is that testcase uses a unknown signal 33 for test,
>> and this will lead sigaction cannot bind signal 33 to the action.
>>
>>     So, we attempt to use a known signal instead, such as 34.
>>
>>
>>
>> which is just as bogus and unportable.  if the test needs a real time
>> signal,
>> it should leverage the sigrtmin...sigrtmax defines.
>> -mike
>>     
>
> I see what you mean about this testcase segfaulting now.
>
> I've done some exploring, and while sigaction does map to rt_sigaction
> on ia64 and x86_64, it won't on arm, mips, s390*, sh, or x86.
>
> Furthermore, the problem is actually being caused by the fact that
> we're not even calling sigaddset properly!
>
> gcooper@orangebox
> /scratch/ltp-dev2/ltp/testcases/kernel/syscalls/rt_sigaction $ gdb
> ./rt_sigaction01
> GNU gdb 6.8
> Copyright (C) 2008 Free Software Foundation, Inc.
> License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
> This is free software: you are free to change and redistribute it.
> There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
> and "show warranty" for details.
> This GDB was configured as "x86_64-pc-linux-gnu"...
> (gdb) c
> The program is not being run.
> (gdb) r
> Starting program:
> /scratch/ltp-dev2/ltp/testcases/kernel/syscalls/rt_sigaction/rt_sigaction01
> thread:20
> [Thread debugging using libthread_db enabled]
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01    1  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01    2  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01    3  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01    4  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01    5  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01    6  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01    7  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01    8  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01    9  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   10  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   11  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   12  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   13  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   14  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   15  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   16  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   17  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   18  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   19  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   20  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   21  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   22  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   23  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   24  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   25  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   26  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   27  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   28  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   29  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
> rt_sigaction01    0  TINFO  :  sigaddset(..) failed
> rt_sigaction01   30  TFAIL  :  rt_sigaction01 failed:
> TEST_ERRNO=EINVAL(22): Invalid argument
>
> So, the test needs fixing, anyhow. After I fixed it, the test still
> segfaulted >:( --
>
> gcooper@orangebox
> /scratch/ltp-dev2/ltp/testcases/kernel/syscalls/rt_sigaction $ gdb
> ./rt_sigaction01
> GNU gdb 6.8
> Copyright (C) 2008 Free Software Foundation, Inc.
> License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
> This is free software: you are free to change and redistribute it.
> There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
> and "show warranty" for details.
> This GDB was configured as "x86_64-pc-linux-gnu"...
> (gdb) r
> Starting program:
> /scratch/ltp-dev2/ltp/testcases/kernel/syscalls/rt_sigaction/rt_sigaction01
> thread:20
> [Thread debugging using libthread_db enabled]
> rt_sigaction01    0  TINFO  :  signal: 34
> rt_sigaction01    0  TINFO  :  sa.sa_flags = SA_RESETHAND|SA_SIGINFO
> [New Thread 0x7fdd451d66f0 (LWP 31927)]
>
> Program received signal SIG34, Real-time event 34.
> [Switching to Thread 0x7fdd451d66f0 (LWP 31927)]
> 0x00007fdd448944f7 in kill () from /lib/libc.so.6
> (gdb) c
> Continuing.
>
> Program received signal SIGSEGV, Segmentation fault.
> 0x00007fdd448944f7 in kill () from /lib/libc.so.6
>
> So I tried to do some digging. glibc 2.8-20080602 does the following
> for alpha, i386, and sysv:
>
>       result = INLINE_SYSCALL (rt_sigaction, 4,
>                                sig, act ? __ptrvalue (&kact) : NULL,
>                                oact ? __ptrvalue (&koact) : NULL, _NSIG / 8);
>
> The kernel code for rt_sigaction in kernel/signal.c is as follows:
>
> #ifdef __ARCH_WANT_SYS_RT_SIGACTION
> SYSCALL_DEFINE4(rt_sigaction, int, sig,
>                 const struct sigaction __user *, act,
>                 struct sigaction __user *, oact,
>                 size_t, sigsetsize)
> {
>         struct k_sigaction new_sa, old_sa;
>         int ret = -EINVAL;
>
>         /* XXX: Don't preclude handling different sized sigset_t's.  */
>         if (sigsetsize != sizeof(sigset_t))
>                 goto out;
>
>         if (act) {
>                 if (copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
>                         return -EFAULT;
>         }
>
>         ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
>
>         if (!ret && oact) {
>                 if (copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
>                         return -EFAULT;
>         }
> out:
>         return ret;
> }
>
> This just blows my mind because rt_sigaction is just calling
> do_sigaction, which is the same code as sigaction AFAICT, apart from
> the fact that rt_sigaction completely disregards sa_restorer on some
> architectures because it's obsolete, and requires the end-user to
> specify a signal mask set (here's mips' version, which wasn't the same
> as i386's version -- i386 had the old sa_restorer code):
>
> SYSCALL_DEFINE3(sigaction, int, sig, const struct sigaction __user *, act,
>         struct sigaction __user *, oact)
> {
>         struct k_sigaction new_ka, old_ka;
>         int ret;
>         int err = 0;
>
>         if (act) {
>                 old_sigset_t mask;
>
>                 if (!access_ok(VERIFY_READ, act, sizeof(*act)))
>                         return -EFAULT;
>                 err |= __get_user(new_ka.sa.sa_handler, &act->sa_handler);
>                 err |= __get_user(new_ka.sa.sa_flags, &act->sa_flags);
>                 err |= __get_user(mask, &act->sa_mask.sig[0]);
>                 if (err)
>                         return -EFAULT;
>
>                 siginitset(&new_ka.sa.sa_mask, mask);
>         }
>
>         ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
>
>         if (!ret && oact) {
>                 if (!access_ok(VERIFY_WRITE, oact, sizeof(*oact)))
>                         return -EFAULT;
>                 err |= __put_user(old_ka.sa.sa_flags, &oact->sa_flags);
>                 err |= __put_user(old_ka.sa.sa_handler, &oact->sa_handler);
>                 err |= __put_user(old_ka.sa.sa_mask.sig[0], oact->sa_mask.sig);
>                 err |= __put_user(0, &oact->sa_mask.sig[1]);
>                 err |= __put_user(0, &oact->sa_mask.sig[2]);
>                 err |= __put_user(0, &oact->sa_mask.sig[3]);
>                 if (err)
>                         return -EFAULT;
>         }
>
>         return ret;
> }
>
> I dunno -- maybe someone else can read the kernel code better than me,
> but it looks like there's zero real value to the rt_sigaction
> testcases, because it can be covered by sigaction...
>
>   

    We just tell guys who use these testcases on arch x86_64 that
    "Syscall rt_sigaction cannot be called directly on arch x86_64.",
    can we?
   
> Thanks,
> -Garrett
>
>
>   


[-- Attachment #1.2: Type: text/html, Size: 12958 bytes --]

[-- Attachment #2: Type: text/plain, Size: 354 bytes --]

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
trial. Simplify your report design, integration and deployment - and focus on 
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july

[-- Attachment #3: Type: text/plain, Size: 155 bytes --]

_______________________________________________
Ltp-list mailing list
Ltp-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ltp-list