Re: [Xenomai] RTDM serial issue

[Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>]

On 01/01/2013 05:09 AM, jeff.webb@nta-inc.net wrote:

> When I try to open an RTDM serial device under Xenomai-2.6.2/Linux 3.4.6
> (Ubuntu 12.04), the call appears to succeed, but I get the following
> output in /var/log/kern.log:
> 
> kernel: [  109.250736] I-pipe: Detected stalled head domain, probably
> caused by a bug.
> kernel: [  109.250737]         A critical section may have been left
> unterminated.
> kernel: [  109.250743] Pid: 3585, comm: serial_tx Not tainted
> 3.4.6-xenomai-2.6.2 #1
> kernel: [  109.250745] Call Trace:
> kernel: [  109.250754]  [<ffffffff810c8a5e>] ipipe_root_only+0xbe/0x130
> kernel: [  109.250758]  [<ffffffff81104aa0>] ? irq_vfile_show+0x570/0x570
> kernel: [  109.250762]  [<ffffffff810c8f56>] ipipe_request_irq+0x36/0x100
> kernel: [  109.250765]  [<ffffffff810cb753>] ipipe_virtualize_irq+0x13/0x30
> kernel: [  109.250770]  [<ffffffff810ffe80>] rthal_irq_request+0x30/0x40
> kernel: [  109.250772]  [<ffffffff811061cb>] xnintr_attach+0x2fb/0x460
> kernel: [  109.250777]  [<ffffffff8119e572>] rtdm_irq_request+0x32/0x70
> kernel: [  109.250782]  [<ffffffff81594f01>] rt_16550_open+0x161/0x380
> kernel: [  109.250785]  [<ffffffff8119b681>] __rt_dev_open+0xb1/0x140
> kernel: [  109.250789]  [<ffffffff811a2738>] sys_rtdm_open+0x78/0xa0
> kernel: [  109.250792]  [<ffffffff8111ed00>] hisyscall_event+0x220/0x3b0
> kernel: [  109.250796]  [<ffffffff810cba4f>] ipipe_syscall_hook+0x5f/0xa0
> kernel: [  109.250799]  [<ffffffff810c9551>]
> __ipipe_notify_syscall+0x131/0x2e0
> kernel: [  109.250804]  [<ffffffff81020b46>] __ipipe_syscall_root+0x46/0x160
> kernel: [  109.250808]  [<ffffffff813a679a>]
> __ipipe_syscall_root_thunk+0x35/0x67
> kernel: [  109.250813]  [<ffffffff816c6cf7>] ?
> system_call_after_swapgs+0x54/0x6d
> kernel: [  109.250815] ------------[ cut here ]------------
> kernel: [  109.250818] WARNING: at include/linux/ipipe_debug.h:88
> xnintr_attach+0x41f/0x460()
> kernel: [  109.250820] Hardware name: Precision WorkStation T3400
> kernel: [  109.250822] Modules linked in: bnep rfcomm bluetooth
> binfmt_misc dm_crypt snd_hda_codec_analog snd_hda_intel snd_hda_codec
> snd_hwde
>  snd_seq_midi_event snd_seq snd_timer snd_seq_device snd soundcore
> usbhid x38_edac hid edac_core snd_page_alloc coretemp psmouse mac_hid dcdba
> t_pc lp parport reiserfs nouveau ttm drm_kms_helper drm i2c_algo_bit
> mxm_wmi video wmi tg3
> kernel: [  109.250874] Pid: 3585, comm: serial_tx Not tainted
> 3.4.6-xenomai-2.6.2 #1
> kernel: [  109.250876] Call Trace:
> kernel: [  109.250880]  [<ffffffff8103cc0f>] warn_slowpath_common+0x7f/0xc0
> kernel: [  109.250883]  [<ffffffff8103cc6a>] warn_slowpath_null+0x1a/0x20
> kernel: [  109.250886]  [<ffffffff811062ef>] xnintr_attach+0x41f/0x460
> kernel: [  109.250889]  [<ffffffff8119e572>] rtdm_irq_request+0x32/0x70
> kernel: [  109.250892]  [<ffffffff81594f01>] rt_16550_open+0x161/0x380
> kernel: [  109.250896]  [<ffffffff8119b681>] __rt_dev_open+0xb1/0x140
> kernel: [  109.250899]  [<ffffffff811a2738>] sys_rtdm_open+0x78/0xa0
> kernel: [  109.250902]  [<ffffffff8111ed00>] hisyscall_event+0x220/0x3b0
> kernel: [  109.250905]  [<ffffffff810cba4f>] ipipe_syscall_hook+0x5f/0xa0
> kernel: [  109.250908]  [<ffffffff810c9551>]
> __ipipe_notify_syscall+0x131/0x2e0
> kernel: [  109.250912]  [<ffffffff81020b46>] __ipipe_syscall_root+0x46/0x160
> kernel: [  109.250915]  [<ffffffff813a679a>]
> __ipipe_syscall_root_thunk+0x35/0x67
> kernel: [  109.250918]  [<ffffffff816c6cf7>] ?
> system_call_after_swapgs+0x54/0x6d
> kernel: [  109.250921] ---[ end trace b86b12c8ae154342 ]---
> 
> This only happens the first time I run my application.  Despite these
> messages, my application code seems to work for the most part.  I am having
> a few high latencies that may be associated with this, or may be due to
> something else entirely.  Regardless, the messages seem to indicate that
> something is not right.  I get similar behavior using a Xenomai build I made
> in mid October using the ipipe-core 3.4 branch and xenomai 2.6 git branch.
> I get the same results on several Ubuntu 12.04 machines.  Here is a simple
> test program that triggers the sort of output described above:
> 
> /* Standard includes */
> #include <stdio.h>
> #include <errno.h>
> #include <error.h>
> #include <stdlib.h>   /* Exit status macros and atexit */
> #include <sys/mman.h> /* Memory management (mlockall) */
> 
> #include <rtdm/rtserial.h>
> 
> /* Main program */
> int main(void)
> {
>   int err;
>   int fd;
> 
>   /* Disable paging for this program's memory */
>   err = mlockall(MCL_CURRENT | MCL_FUTURE);
>   if (err)
>   {
>       error(EXIT_FAILURE, errno,
>             "could not disable memory paging for this program");
>   }
> 
>   /* Open the serial port */
>   fd = rt_dev_open("rtser0", 0);
>   if (fd < 0)
>     {
>       error(EXIT_FAILURE, errno,
>             "could not open serial device");
>     }
> 
>   printf("open succeeded\n");
> 
>   /* Close the serial port */
>   rt_dev_close(fd);
> 
>   return 0;
> }
> 
> makefile flags:
> CFLAGS := $(shell xeno-config --skin=native --cflags)
> LDLIBS := -lrtdm $(shell xeno-config --skin=native --ldflags)
> 
> If I try to compile and run this same test program on an older machine
> running Xenomai 2.5.5.2 / Linux 2.6.32.20 (Ubuntu 10.04), I do not have
> any issues.  I also built a set of Xenomai 2.6.1 / Linux 3.2.21 packages
> for this older system and I don't seem to see the problem there either.
> I had a package build issue (too old of a version of some dpkg build tool)
> when trying to build the latest Xenomai/Linux combo under 10.04, so I
> didn't test that configuration on the older machine.  The newer machines
> have hardware that is not supported in older Xenomai/Linux configurations,
> so I haven't been able to do an apples-to-apples comparison to verify if
> the behavior is due to a change in the Xenomai version.  Any help in
> deciphering what's going on, or some next steps in debugging would be
> appreciated.


This is a known issue, for a proposed fix, see:
http://www.xenomai.org/pipermail/xenomai/2012-October/026566.html

Whether we are merging this patch is still not sure.

> 
> Thanks for all the fine work in putting together this latest release.


I am wondering whether you are ironical, a new release should happen
shortly to fix all the issues which were forgotten.

-- 
                                                                Gilles.