On 11/24/2017 03:38 PM, Leopold Palomo-Avellaneda wrote:
> On 24/11/17 15:21, Philippe Gerum wrote:
>> On 11/24/2017 01:24 PM, Leopold Palomo-Avellaneda wrote:
>>> On 24/11/17 13:22, Philippe Gerum wrote:
>>>> On 11/24/2017 01:10 PM, Leopold Palomo-Avellaneda wrote:
>>>>> On 24/11/17 13:04, Philippe Gerum wrote:
>>>>>> On 11/24/2017 01:01 PM, Leopold Palomo-Avellaneda wrote:
>>>>>>> On 24/11/17 11:55, Philippe Gerum wrote:
>>>>>>>> On 11/24/2017 11:34 AM, Leopold Palomo-Avellaneda wrote:
>>>>>>>>> Hi (again)
>>>>>>>>>
>>>>>>>>> running cross-link I got that:
>>>>>>>>>
>>>>>>>>>>> with my user
>>>>>>>>>>>
>>>>>>>>>>> ./cross-link
>>>>>>>>>>> mem=0x7f0d630488e0, alloc_size=8192
>>>>>>>>>>> Init memory pool returns 1808 bytes
>>>>>>>>>>> Running after ...
>>>>>>>>>>> sysregd:
>>>>>>>>>>> create_directory_recursive("/var/run/xenomai/leopold.palomo/anon@19479"):
>>>>>>>>>>> Permission denied
>>>>>>>>>
>>>>>>>>> ....
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> the permissions are:
>>>>>>>>> $ ls -la /var/run/xenomai/
>>>>>>>>> total 0
>>>>>>>>> drwxr-xr-x 3 root root 60 nov 23 16:03 .
>>>>>>>>> drwxr-xr-x 24 root root 860 nov 24 11:12 ..
>>>>>>>>> drwxr-xr-x 2 root root 40 nov 24 11:12 root
>>>>>>>>>
>>>>>>>>> so, IMHO this folder should belongs to xenomai group and permisions
>>>>>>>>> g+rw. Right?
>>>>>>>>
>>>>>>>> There is no such official "Xenomai" group, I mean this is only a
>>>>>>>> template in some rule file to illustrate the fact that one could
>>>>>>>> actually assign rt privileges to a user group, instead of running apps
>>>>>>>> as a sudoer. That could be any group you are willing to mention in the
>>>>>>>> xenomai.allowed_group kernel parameter on the kernel command line, or
>>>>>>>> no
>>>>>>>> group at all. This is a matter of local installation.
>>>>>>>
>>>>>>> Ok, the debian package creates it. I have xenomai.allowed_group=113 in
>>>>>>> the grub.
>>>>>>> So, the packaging (installation should solve it).
>>>>>>>
>>>>>>>>>
>>>>>>>>> Changing the permissions manually solved it. It's something about the
>>>>>>>>> packaging
>>>>>>>>> or the library?
>>>>>>>>>
>>>>>>>>
>>>>>>>> This is not a library issue.
>>>>>>>
>>>>>>> Ok, mine. Packaging problem.
>>>>>>>
>>>>>>>>>
>>>>>>>>>>> leopold.palomo@bmm3 ~/xenomai/xenomai-3.0.6/system/demo$ sudo
>>>>>>>>>>> ./cross-link
>>>>>>>>>>> mem=0x7fc9e73f88e0, alloc_size=8192
>>>>>>>>>>> Init memory pool returns 1808 bytes
>>>>>>>>>>> Running after ...
>>>>>>>>>>> mem=0x7f2e146318e0, alloc_size=8192
>>>>>>>>>>> Init memory pool returns 1808 bytes
>>>>>>>>>>> Running after ...
>>>>>>>>>>> main : can't open /dev/rtdm/rtser0 (write), No such file or
>>>>>>>>>>> directory
>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Not being able to open rtser0 is yet another kind of problem: the
>>>>>>>>>> uart
>>>>>>>>>> driver is not up and running, so the devices were not created.
>>>>>>>>>> crosslink
>>>>>>>>>> definitely needs such driver.
>>>>>>>>>
>>>>>>>>> Yes, I don't have rtser0 entry in /dev/rtdm/. Who does create it?
>>>>>>>>>
>>>>>>>>
>>>>>>>> Any RTDM driver from kernel/drivers/serial, depending on your platform.
>>>>>>>
>>>>>>> Well, I have /dev/rtdm/* entries but not that one. Is this normal?
>>>>>>>
>>>>>>
>>>>>> Do you have any RTDM-based serial driver enabled in Kconfig?
>>>>>>
>>>>>
>>>>> xeno_16550A
>>>>>
>>>>> I have all modules that I can.
>>>>>
>>>>> <M> 16550A UART driver
>>>>> Hardware access mode (Any access mode)
>>>>> ---> (X)
>>>>> Any access mode
>>>>>
>>>>> [*] PCI board support
>>>>> [*] Moxa PCI boards
>>>>>
>>>>> CONFIG_XENO_OPT_RTDM_COMPAT_DEVNODE=y
>>>>> CONFIG_XENO_DRIVERS_AUTOTUNE=m
>>>>>
>>>>> #
>>>>> # Serial drivers
>>>>> #
>>>>> CONFIG_XENO_DRIVERS_16550A=m
>>>>
>>>> missing modprobe?
>>>>
>>>
>>> $ lsmod | grep xeno
>>> xeno_16550A 28672 0
>>> xeno_autotune 24576 0
>>> xeno_analogy 57344 0
>>> xeno_rtdmtest 16384 0
>>> xeno_udd 16384 0
>>> xeno_switchtest 20480 0
>>> xeno_rtipc 49152 0
>>> xeno_can_peak_pci 16384 0
>>> xeno_can_sja1000 20480 1 xeno_can_peak_pci
>>> xeno_can 32768 2 xeno_can_peak_pci,xeno_can_sja1000
>>>
>>>
>>>
>>
>> It looks like the driver was not given the proper parameters at load up,
>> specifically the io= or mem= information.
>>
>> http://xenomai.org/serial-16550a-driver/ may help.
>>
>
> ok,
>
> I have followed this page and got:
>
> DMAR: DRHD: handling fault status reg 2
> [ 90.453076] DMAR: [INTR-REMAP] Request device [f0:1f.0] fault index 0
> [fault
> reason 37] Blocked a compatibility format interrupt request
It looks like there is a device on the PCI bus issuing interrupts with a
wrong format, specifically a compat format IRQ despite your kernel
config enables interrupt remapping. This is reported by a fault
interrupt. lspci should tell you which device is mentioned in the fault
report (f0:1f.0).
lspci -v may help finding the offender. passing iommu=debug on the
kernel command line should dump some useful information regarding the
DMA remapping hardware too.
I don't think this issue could be related to Xenomai though, since the
latter is not involved in the setup process of neither the IOMMU nor the
PCI devices. Enabling that specific PCI device in the RTDM serial driver
is likely to reveal an earlier misconfiguration though.
--
Philippe.
_______________________________________________
Xenomai mailing list
Xenomai@xenomai.org
https://xenomai.org/mailman/listinfo/xenomai
