hi, I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures: https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other... It will be great if someone tests them on other architectures too. cheers, stanislav
Millionth thanks for that. I have my test instance (1.6.0-4~64bit) on Debian Jesse running on Xen and I experience some jitter and voice deformation issues (vm is pretty powerful for single call to 3x00 number (made over pretty fast route)), I have also that nice gem in scaleway (on Jesse as well, but I probably have to try it on Centos) and it seems to be promising. Best regards, O. On 30 November 2015 at 16:27, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
with any public VM hosting, regardless of the hypervisor type, you get unpredictable CPU resource available for sending out RTP streams. That causes random jitter or packet loss, especially if you have many parallel calls. I moved my own PBX from a public Xen hosting to Scaleway, and I'm significantly more happy with the voice quality, even that the distance is higher (round-trip delay increased from few milliseconds to 25ms because of geographical distance). On Tue, Dec 8, 2015 at 12:07 PM, Olgierd Ziolko <lists.fs-users@uce.pw> wrote:
Millionth thanks for that. I have my test instance (1.6.0-4~64bit) on Debian Jesse running on Xen and I experience some jitter and voice deformation issues (vm is pretty powerful for single call to 3x00 number (made over pretty fast route)), I have also that nice gem in scaleway (on Jesse as well, but I probably have to try it on Centos) and it seems to be promising.
Best regards, O.
On 30 November 2015 at 16:27, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Your instructions worked great for me building on/for my Odroid-U3 (arm) running Debian Jessie. I know you are running on arm at scaleway. Do you ever run into any alignment issues? I'm getting alignment exception faults receiving faxes and I'm having trouble pulling any useful debug messages. Thanks, Jason On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Hi, I didn't see any errors, but I'm not using fax, and mostly it's bridging the calls and connecting users to a conference. Do you have more debugging information? Most probably the source of the problem is that x86 allows unaligned memory access: a word pointer can have any value. And other architectures don't allow this, and require the pointers to be aligned to word boundaries. Mongodb, for example, is known to fail on any non-x86 CPU because of that. So, probably mod_spandsp needs an audit. On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote:
Your instructions worked great for me building on/for my Odroid-U3 (arm) running Debian Jessie.
I know you are running on arm at scaleway. Do you ever run into any alignment issues? I'm getting alignment exception faults receiving faxes and I'm having trouble pulling any useful debug messages.
Thanks, Jason
On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
I get nothing in debug or sip trace. It always dies at mod_spandsp_fax so I figured it was in that module. At that point freeswitch just restarts without throwing any errors. Kernel log shows: Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not handling instruction ecb17a01 at [<b44fe32c>] Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: alignment exception (0x001) at 0xb4546e4a The only part of the exception that is the same every time it happens is "instruction ecb17a01." I would have just opened a JIRA, but I didn't think this was enough information. I have been Googling trying to figure out how to get more info. Jason On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
Hi, I didn't see any errors, but I'm not using fax, and mostly it's bridging the calls and connecting users to a conference. Do you have more debugging information?
Most probably the source of the problem is that x86 allows unaligned memory access: a word pointer can have any value. And other architectures don't allow this, and require the pointers to be aligned to word boundaries. Mongodb, for example, is known to fail on any non-x86 CPU because of that.
So, probably mod_spandsp needs an audit. On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote:
Your instructions worked great for me building on/for my Odroid-U3 (arm) running Debian Jessie.
I know you are running on arm at scaleway. Do you ever run into any alignment issues? I'm getting alignment exception faults receiving faxes and I'm having trouble pulling any useful debug messages.
Thanks, Jason
On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
This is on FreeSWITCH Version 1.6.5+git-20151209T192325Z~70b8c17763~32bit (git 70b8c17 2015-12-09 19:23:25Z 32bit) by the way. On Thu, Dec 10, 2015 at 2:49 PM, Komar, Jason <jkomar@jbox.ca> wrote:
I get nothing in debug or sip trace. It always dies at mod_spandsp_fax so I figured it was in that module. At that point freeswitch just restarts without throwing any errors.
Kernel log shows:
Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not handling instruction ecb17a01 at [<b44fe32c>] Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: alignment exception (0x001) at 0xb4546e4a
The only part of the exception that is the same every time it happens is "instruction ecb17a01."
I would have just opened a JIRA, but I didn't think this was enough information. I have been Googling trying to figure out how to get more info.
Jason
On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
Hi, I didn't see any errors, but I'm not using fax, and mostly it's bridging the calls and connecting users to a conference. Do you have more debugging information?
Most probably the source of the problem is that x86 allows unaligned memory access: a word pointer can have any value. And other architectures don't allow this, and require the pointers to be aligned to word boundaries. Mongodb, for example, is known to fail on any non-x86 CPU because of that.
So, probably mod_spandsp needs an audit. On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote:
Your instructions worked great for me building on/for my Odroid-U3 (arm) running Debian Jessie.
I know you are running on arm at scaleway. Do you ever run into any alignment issues? I'm getting alignment exception faults receiving faxes and I'm having trouble pulling any useful debug messages.
Thanks, Jason
On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin <ssinyagin@gmail.com
wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
if coredumps are not made, you need to enable them, and then install debug symbols (-dbg packages), and run gdb on the core fule. Then it shoul dindicate the line number in the sources where it crashed. Here's one suspect: mod_spandsp_fax.c, mod_spandsp_fax_process_fax: int16_t *buf = NULL; .... buf = switch_core_session_alloc(session, SWITCH_RECOMMENDED_BUFFER_SIZE); are we guaranteed that alloc returns a 16-bit aligned pointer? I think it makes sense to open a jira ticket anyway On Thu, Dec 10, 2015 at 10:51 PM, Komar, Jason <jkomar@jbox.ca> wrote:
This is on FreeSWITCH Version 1.6.5+git-20151209T192325Z~70b8c17763~32bit (git 70b8c17 2015-12-09 19:23:25Z 32bit) by the way.
On Thu, Dec 10, 2015 at 2:49 PM, Komar, Jason <jkomar@jbox.ca> wrote:
I get nothing in debug or sip trace. It always dies at mod_spandsp_fax so I figured it was in that module. At that point freeswitch just restarts without throwing any errors.
Kernel log shows:
Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not handling instruction ecb17a01 at [<b44fe32c>] Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: alignment exception (0x001) at 0xb4546e4a
The only part of the exception that is the same every time it happens is "instruction ecb17a01."
I would have just opened a JIRA, but I didn't think this was enough information. I have been Googling trying to figure out how to get more info.
Jason
On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
Hi, I didn't see any errors, but I'm not using fax, and mostly it's bridging the calls and connecting users to a conference. Do you have more debugging information?
Most probably the source of the problem is that x86 allows unaligned memory access: a word pointer can have any value. And other architectures don't allow this, and require the pointers to be aligned to word boundaries. Mongodb, for example, is known to fail on any non-x86 CPU because of that.
So, probably mod_spandsp needs an audit.
On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote:
Your instructions worked great for me building on/for my Odroid-U3 (arm) running Debian Jessie.
I know you are running on arm at scaleway. Do you ever run into any alignment issues? I'm getting alignment exception faults receiving faxes and I'm having trouble pulling any useful debug messages.
Thanks, Jason
On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Some of this is a bit of unfamiliar territory for me, but I am learning. I will try enabling core dumps, etc. and see if I can get some more information. I will open up a jira ticket in the next couple of days. Thanks for your help. Jason On Thu, Dec 10, 2015 at 4:19 PM, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
if coredumps are not made, you need to enable them, and then install debug symbols (-dbg packages), and run gdb on the core fule. Then it shoul dindicate the line number in the sources where it crashed.
Here's one suspect: mod_spandsp_fax.c, mod_spandsp_fax_process_fax:
int16_t *buf = NULL; .... buf = switch_core_session_alloc(session, SWITCH_RECOMMENDED_BUFFER_SIZE);
are we guaranteed that alloc returns a 16-bit aligned pointer?
I think it makes sense to open a jira ticket anyway
On Thu, Dec 10, 2015 at 10:51 PM, Komar, Jason <jkomar@jbox.ca> wrote:
This is on FreeSWITCH Version 1.6.5+git-20151209T192325Z~70b8c17763~32bit (git 70b8c17 2015-12-09 19:23:25Z 32bit) by the way.
On Thu, Dec 10, 2015 at 2:49 PM, Komar, Jason <jkomar@jbox.ca> wrote:
I get nothing in debug or sip trace. It always dies at mod_spandsp_fax
so
I figured it was in that module. At that point freeswitch just restarts without throwing any errors.
Kernel log shows:
Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not handling instruction ecb17a01 at [<b44fe32c>] Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: alignment exception (0x001) at 0xb4546e4a
The only part of the exception that is the same every time it happens is "instruction ecb17a01."
I would have just opened a JIRA, but I didn't think this was enough information. I have been Googling trying to figure out how to get more info.
Jason
On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin < ssinyagin@gmail.com> wrote:
Hi, I didn't see any errors, but I'm not using fax, and mostly it's bridging the calls and connecting users to a conference. Do you have
more
debugging information?
Most probably the source of the problem is that x86 allows unaligned memory access: a word pointer can have any value. And other architectures don't allow this, and require the pointers to be aligned to word boundaries. Mongodb, for example, is known to fail on any non-x86 CPU because of that.
So, probably mod_spandsp needs an audit.
On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote:
Your instructions worked great for me building on/for my Odroid-U3
(arm)
running Debian Jessie.
I know you are running on arm at scaleway. Do you ever run into any alignment issues? I'm getting alignment exception faults receiving faxes and I'm having trouble pulling any useful debug messages.
Thanks, Jason
On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
https://freeswitch.org/jira/browse/FS-8651 the source of the problem is quite deep in the sources of spandsp, so I can't offer a fix. On Thu, Dec 10, 2015 at 10:49 PM, Komar, Jason <jkomar@jbox.ca> wrote:
I get nothing in debug or sip trace. It always dies at mod_spandsp_fax so I figured it was in that module. At that point freeswitch just restarts without throwing any errors.
Kernel log shows:
Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not handling instruction ecb17a01 at [<b44fe32c>] Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: alignment exception (0x001) at 0xb4546e4a
The only part of the exception that is the same every time it happens is "instruction ecb17a01."
I would have just opened a JIRA, but I didn't think this was enough information. I have been Googling trying to figure out how to get more info.
Jason
On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
Hi, I didn't see any errors, but I'm not using fax, and mostly it's bridging the calls and connecting users to a conference. Do you have more debugging information?
Most probably the source of the problem is that x86 allows unaligned memory access: a word pointer can have any value. And other architectures don't allow this, and require the pointers to be aligned to word boundaries. Mongodb, for example, is known to fail on any non-x86 CPU because of that.
So, probably mod_spandsp needs an audit.
On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote:
Your instructions worked great for me building on/for my Odroid-U3 (arm) running Debian Jessie.
I know you are running on arm at scaleway. Do you ever run into any alignment issues? I'm getting alignment exception faults receiving faxes and I'm having trouble pulling any useful debug messages.
Thanks, Jason
On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin <ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
According to the ARM docs, the instruction in question decodes to "coprocessor data transfer (LDC, STC)". The way that instruction is encoded (with the first four bits == xE), the instruction will be executed every time it's encountered (i.e., not conditionally). Stanislav, you said you thought the trap was related to HD math and not alignment. Based on what that instruction apparently is, it looks to me like you were on the right track. Do all ARM variants have an FPU, and if not, could this problem stem from configure et al. determining there's an FPU on a platform where there isn't one? Hope this helps someone find and eradicate this bug. - Larry On Dec 12, 2015 7:04 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
https://freeswitch.org/jira/browse/FS-8651 the source of the problem is quite deep in the sources of spandsp, so I can't offer a fix.
On Thu, Dec 10, 2015 at 10:49 PM, Komar, Jason <jkomar@jbox.ca> wrote:
I get nothing in debug or sip trace. It always dies at mod_spandsp_fax so I figured it was in that module. At that point freeswitch just restarts without throwing any errors.
Kernel log shows:
Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not handling instruction ecb17a01 at [<b44fe32c>] Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: alignment exception (0x001) at 0xb4546e4a
The only part of the exception that is the same every time it happens is "instruction ecb17a01."
I would have just opened a JIRA, but I didn't think this was enough information. I have been Googling trying to figure out how to get more info.
Jason
On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin <ssinyagin@gmail.com
wrote:
Hi, I didn't see any errors, but I'm not using fax, and mostly it's bridging the calls and connecting users to a conference. Do you have
more
debugging information?
Most probably the source of the problem is that x86 allows unaligned memory access: a word pointer can have any value. And other architectures don't allow this, and require the pointers to be aligned to word boundaries. Mongodb, for example, is known to fail on any non-x86 CPU because of that.
So, probably mod_spandsp needs an audit.
On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote:
Your instructions worked great for me building on/for my Odroid-U3
(arm)
running Debian Jessie.
I know you are running on arm at scaleway. Do you ever run into any alignment issues? I'm getting alignment exception faults receiving faxes and I'm having trouble pulling any useful debug messages.
Thanks, Jason
On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin < ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
See the bottom of the bug description. There's a mismatch between a 16bit integer array and a function that treats it as array of 32bit floats. On 13 Dec 2015 3:30 am, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
According to the ARM docs, the instruction in question decodes to "coprocessor data transfer (LDC, STC)". The way that instruction is encoded (with the first four bits == xE), the instruction will be executed every time it's encountered (i.e., not conditionally).
Stanislav, you said you thought the trap was related to HD math and not alignment. Based on what that instruction apparently is, it looks to me like you were on the right track.
Do all ARM variants have an FPU, and if not, could this problem stem from configure et al. determining there's an FPU on a platform where there isn't one?
Hope this helps someone find and eradicate this bug.
- Larry On Dec 12, 2015 7:04 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
https://freeswitch.org/jira/browse/FS-8651 the source of the problem is quite deep in the sources of spandsp, so I can't offer a fix.
On Thu, Dec 10, 2015 at 10:49 PM, Komar, Jason <jkomar@jbox.ca> wrote:
I get nothing in debug or sip trace. It always dies at mod_spandsp_fax so I figured it was in that module. At that point freeswitch just restarts without throwing any errors.
Kernel log shows:
Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not handling instruction ecb17a01 at [<b44fe32c>] Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: alignment exception (0x001) at 0xb4546e4a
The only part of the exception that is the same every time it happens is "instruction ecb17a01."
I would have just opened a JIRA, but I didn't think this was enough information. I have been Googling trying to figure out how to get more info.
Jason
On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin < ssinyagin@gmail.com> wrote:
Hi, I didn't see any errors, but I'm not using fax, and mostly it's bridging the calls and connecting users to a conference. Do you have
more
debugging information?
Most probably the source of the problem is that x86 allows unaligned memory access: a word pointer can have any value. And other architectures don't allow this, and require the pointers to be aligned to word boundaries. Mongodb, for example, is known to fail on any non-x86 CPU because of that.
So, probably mod_spandsp needs an audit.
On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote:
Your instructions worked great for me building on/for my Odroid-U3
(arm)
running Debian Jessie.
I know you are running on arm at scaleway. Do you ever run into any alignment issues? I'm getting alignment exception faults receiving faxes and I'm having trouble pulling any useful debug messages.
Thanks, Jason
On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin < ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Stanislav, I saw that; what I was saying was it looked to me like the processor was interpreting data as code and it happened to work out that the bit pattern for that data and for LDC/STC are the same. I've run into problems before where misaligned inductions have resulted in an ARM processor either ignoring the improperly aligned portion of an instruction or treating it as data. If you disassemble the code around where it traps, what does the listing show the processor is trying to execute? One additional thought - if you're concerned that calloc() or malloc() is returning a pointer that's improperly aligned, you could always supply your own temporarily and see if the problem goes away. - Larry On Dec 13, 2015 4:19 AM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
See the bottom of the bug description. There's a mismatch between a 16bit integer array and a function that treats it as array of 32bit floats. On 13 Dec 2015 3:30 am, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
According to the ARM docs, the instruction in question decodes to "coprocessor data transfer (LDC, STC)". The way that instruction is encoded (with the first four bits == xE), the instruction will be executed every time it's encountered (i.e., not conditionally).
Stanislav, you said you thought the trap was related to HD math and not alignment. Based on what that instruction apparently is, it looks to me like you were on the right track.
Do all ARM variants have an FPU, and if not, could this problem stem from configure et al. determining there's an FPU on a platform where there isn't one?
Hope this helps someone find and eradicate this bug.
- Larry On Dec 12, 2015 7:04 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
https://freeswitch.org/jira/browse/FS-8651 the source of the problem is quite deep in the sources of spandsp, so I can't offer a fix.
On Thu, Dec 10, 2015 at 10:49 PM, Komar, Jason <jkomar@jbox.ca> wrote:
I get nothing in debug or sip trace. It always dies at mod_spandsp_fax so I figured it was in that module. At that point freeswitch just restarts without throwing any errors.
Kernel log shows:
Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not handling instruction ecb17a01 at [<b44fe32c>] Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: alignment exception (0x001) at 0xb4546e4a
The only part of the exception that is the same every time it happens is "instruction ecb17a01."
I would have just opened a JIRA, but I didn't think this was enough information. I have been Googling trying to figure out how to get more info.
Jason
On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin < ssinyagin@gmail.com> wrote:
Hi, I didn't see any errors, but I'm not using fax, and mostly it's bridging the calls and connecting users to a conference. Do you have
more
debugging information?
Most probably the source of the problem is that x86 allows unaligned memory access: a word pointer can have any value. And other architectures don't allow this, and require the pointers to be aligned to word boundaries. Mongodb, for example, is known to fail on any non-x86 CPU because of that.
So, probably mod_spandsp needs an audit.
On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote:
Your instructions worked great for me building on/for my Odroid-U3
(arm)
running Debian Jessie.
I know you are running on arm at scaleway. Do you ever run into any alignment issues? I'm getting alignment exception faults receiving faxes and I'm having trouble pulling any useful debug messages.
Thanks, Jason
On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin < ssinyagin@gmail.com> wrote: > > hi, > > I put together the build scenario that I tested for ARM (armhf at > scaleway.com), and it should also work for other architectures: > > > https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other... > > It will be great if someone tests them on other architectures too. > > > cheers, > stanislav > > >
> Professional FreeSWITCH Consulting Services: > consulting@freeswitch.org > http://www.freeswitchsolutions.com > > Official FreeSWITCH Sites > http://www.freeswitch.org > http://confluence.freeswitch.org > http://www.cluecon.com > > FreeSWITCH-users mailing list > FreeSWITCH-users@lists.freeswitch.org > http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users > http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Larry, here it's just a pointer that is 16bit aligned, and is given to a function that tries to treat it as an array of 32bit values. See my new comment in the ticket. We just need the module author to fix this, as it's difficult to figure out why some ifdefs have "x" at the end. On 14 Dec 2015 19:03, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
Stanislav,
I saw that; what I was saying was it looked to me like the processor was interpreting data as code and it happened to work out that the bit pattern for that data and for LDC/STC are the same. I've run into problems before where misaligned inductions have resulted in an ARM processor either ignoring the improperly aligned portion of an instruction or treating it as data.
If you disassemble the code around where it traps, what does the listing show the processor is trying to execute?
One additional thought - if you're concerned that calloc() or malloc() is returning a pointer that's improperly aligned, you could always supply your own temporarily and see if the problem goes away.
- Larry On Dec 13, 2015 4:19 AM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
See the bottom of the bug description. There's a mismatch between a 16bit integer array and a function that treats it as array of 32bit floats. On 13 Dec 2015 3:30 am, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
According to the ARM docs, the instruction in question decodes to "coprocessor data transfer (LDC, STC)". The way that instruction is encoded (with the first four bits == xE), the instruction will be executed every time it's encountered (i.e., not conditionally).
Stanislav, you said you thought the trap was related to HD math and not alignment. Based on what that instruction apparently is, it looks to me like you were on the right track.
Do all ARM variants have an FPU, and if not, could this problem stem from configure et al. determining there's an FPU on a platform where there isn't one?
Hope this helps someone find and eradicate this bug.
- Larry On Dec 12, 2015 7:04 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
https://freeswitch.org/jira/browse/FS-8651 the source of the problem is quite deep in the sources of spandsp, so I can't offer a fix.
On Thu, Dec 10, 2015 at 10:49 PM, Komar, Jason <jkomar@jbox.ca> wrote:
I get nothing in debug or sip trace. It always dies at mod_spandsp_fax so I figured it was in that module. At that point freeswitch just restarts without throwing any errors.
Kernel log shows:
Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not handling instruction ecb17a01 at [<b44fe32c>] Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: alignment exception (0x001) at 0xb4546e4a
The only part of the exception that is the same every time it happens is "instruction ecb17a01."
I would have just opened a JIRA, but I didn't think this was enough information. I have been Googling trying to figure out how to get more info.
Jason
On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin < ssinyagin@gmail.com> wrote:
Hi, I didn't see any errors, but I'm not using fax, and mostly it's bridging the calls and connecting users to a conference. Do you have
more
debugging information?
Most probably the source of the problem is that x86 allows unaligned memory access: a word pointer can have any value. And other architectures don't allow this, and require the pointers to be aligned to word boundaries. Mongodb, for example, is known to fail on any non-x86 CPU because of that.
So, probably mod_spandsp needs an audit.
On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote: > > Your instructions worked great for me building on/for my Odroid-U3 (arm) > running Debian Jessie. > > I know you are running on arm at scaleway. Do you ever run into any > alignment issues? I'm getting alignment exception faults receiving faxes and > I'm having trouble pulling any useful debug messages. > > Thanks, > Jason > > On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin < ssinyagin@gmail.com> > wrote: >> >> hi, >> >> I put together the build scenario that I tested for ARM (armhf at >> scaleway.com), and it should also work for other architectures: >> >> >> https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other... >> >> It will be great if someone tests them on other architectures too. >> >> >> cheers, >> stanislav >> >> >>
>> Professional FreeSWITCH Consulting Services: >> consulting@freeswitch.org >> http://www.freeswitchsolutions.com >> >> Official FreeSWITCH Sites >> http://www.freeswitch.org >> http://confluence.freeswitch.org >> http://www.cluecon.com >> >> FreeSWITCH-users mailing list >> FreeSWITCH-users@lists.freeswitch.org >> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >> http://www.freeswitch.org > > > >
> Professional FreeSWITCH Consulting Services: > consulting@freeswitch.org > http://www.freeswitchsolutions.com > > Official FreeSWITCH Sites > http://www.freeswitch.org > http://confluence.freeswitch.org > http://www.cluecon.com > > FreeSWITCH-users mailing list > FreeSWITCH-users@lists.freeswitch.org > http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users > http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Stanislav, Sometime in the next few days I need to spend some quality time with that code, out of sheer curiosity if for no other reason. There are a few things in that code that have me - uh, stumped , to put it mildly. Would you do me a favor, and if you do hear back from the author / maintainer, let me know what he says? Or when he fixes it? Thanks , Larry On Dec 14, 2015 3:16 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
Larry, here it's just a pointer that is 16bit aligned, and is given to a function that tries to treat it as an array of 32bit values. See my new comment in the ticket. We just need the module author to fix this, as it's difficult to figure out why some ifdefs have "x" at the end. On 14 Dec 2015 19:03, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
Stanislav,
I saw that; what I was saying was it looked to me like the processor was interpreting data as code and it happened to work out that the bit pattern for that data and for LDC/STC are the same. I've run into problems before where misaligned inductions have resulted in an ARM processor either ignoring the improperly aligned portion of an instruction or treating it as data.
If you disassemble the code around where it traps, what does the listing show the processor is trying to execute?
One additional thought - if you're concerned that calloc() or malloc() is returning a pointer that's improperly aligned, you could always supply your own temporarily and see if the problem goes away.
- Larry On Dec 13, 2015 4:19 AM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
See the bottom of the bug description. There's a mismatch between a 16bit integer array and a function that treats it as array of 32bit floats. On 13 Dec 2015 3:30 am, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
According to the ARM docs, the instruction in question decodes to "coprocessor data transfer (LDC, STC)". The way that instruction is encoded (with the first four bits == xE), the instruction will be executed every time it's encountered (i.e., not conditionally).
Stanislav, you said you thought the trap was related to HD math and not alignment. Based on what that instruction apparently is, it looks to me like you were on the right track.
Do all ARM variants have an FPU, and if not, could this problem stem from configure et al. determining there's an FPU on a platform where there isn't one?
Hope this helps someone find and eradicate this bug.
- Larry On Dec 12, 2015 7:04 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
https://freeswitch.org/jira/browse/FS-8651 the source of the problem is quite deep in the sources of spandsp, so I can't offer a fix.
On Thu, Dec 10, 2015 at 10:49 PM, Komar, Jason <jkomar@jbox.ca> wrote:
I get nothing in debug or sip trace. It always dies at mod_spandsp_fax so I figured it was in that module. At that point freeswitch just restarts without throwing any errors.
Kernel log shows:
Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not handling instruction ecb17a01 at [<b44fe32c>] Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: alignment exception (0x001) at 0xb4546e4a
The only part of the exception that is the same every time it happens is "instruction ecb17a01."
I would have just opened a JIRA, but I didn't think this was enough information. I have been Googling trying to figure out how to get more info.
Jason
On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin < ssinyagin@gmail.com> wrote: > > Hi, I didn't see any errors, but I'm not using fax, and mostly it's > bridging the calls and connecting users to a conference. Do you have more > debugging information? > > Most probably the source of the problem is that x86 allows unaligned > memory access: a word pointer can have any value. And other architectures > don't allow this, and require the pointers to be aligned to word boundaries. > Mongodb, for example, is known to fail on any non-x86 CPU because of that. > > So, probably mod_spandsp needs an audit. > > On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote: >> >> Your instructions worked great for me building on/for my Odroid-U3 (arm) >> running Debian Jessie. >> >> I know you are running on arm at scaleway. Do you ever run into any >> alignment issues? I'm getting alignment exception faults receiving faxes and >> I'm having trouble pulling any useful debug messages. >> >> Thanks, >> Jason >> >> On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin < ssinyagin@gmail.com> >> wrote: >>> >>> hi, >>> >>> I put together the build scenario that I tested for ARM (armhf at >>> scaleway.com), and it should also work for other architectures: >>> >>> >>> https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other... >>> >>> It will be great if someone tests them on other architectures too. >>> >>> >>> cheers, >>> stanislav >>> >>> >>>
>>> Professional FreeSWITCH Consulting Services: >>> consulting@freeswitch.org >>> http://www.freeswitchsolutions.com >>> >>> Official FreeSWITCH Sites >>> http://www.freeswitch.org >>> http://confluence.freeswitch.org >>> http://www.cluecon.com >>> >>> FreeSWITCH-users mailing list >>> FreeSWITCH-users@lists.freeswitch.org >>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >>> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >>> http://www.freeswitch.org >> >> >> >>
>> Professional FreeSWITCH Consulting Services: >> consulting@freeswitch.org >> http://www.freeswitchsolutions.com >> >> Official FreeSWITCH Sites >> http://www.freeswitch.org >> http://confluence.freeswitch.org >> http://www.cluecon.com >> >> FreeSWITCH-users mailing list >> FreeSWITCH-users@lists.freeswitch.org >> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >> http://www.freeswitch.org > > >
> Professional FreeSWITCH Consulting Services: > consulting@freeswitch.org > http://www.freeswitchsolutions.com > > Official FreeSWITCH Sites > http://www.freeswitch.org > http://confluence.freeswitch.org > http://www.cluecon.com > > FreeSWITCH-users mailing list > FreeSWITCH-users@lists.freeswitch.org > http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users > http://www.freeswitch.org
_________________________________________________________________________
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
I tried to reach Steve on Sunday, but no success so far. On Wed, Dec 16, 2015 at 12:25 AM, Larry Morley <lmorley@neny.cslimits.net> wrote:
Stanislav,
Sometime in the next few days I need to spend some quality time with that code, out of sheer curiosity if for no other reason. There are a few things in that code that have me - uh, stumped , to put it mildly.
Would you do me a favor, and if you do hear back from the author / maintainer, let me know what he says? Or when he fixes it?
Thanks , Larry
On Dec 14, 2015 3:16 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
Larry, here it's just a pointer that is 16bit aligned, and is given to a function that tries to treat it as an array of 32bit values. See my new comment in the ticket. We just need the module author to fix this, as it's difficult to figure out why some ifdefs have "x" at the end.
On 14 Dec 2015 19:03, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
Stanislav,
I saw that; what I was saying was it looked to me like the processor was interpreting data as code and it happened to work out that the bit pattern for that data and for LDC/STC are the same. I've run into problems before where misaligned inductions have resulted in an ARM processor either ignoring the improperly aligned portion of an instruction or treating it as data.
If you disassemble the code around where it traps, what does the listing show the processor is trying to execute?
One additional thought - if you're concerned that calloc() or malloc() is returning a pointer that's improperly aligned, you could always supply your own temporarily and see if the problem goes away.
- Larry
On Dec 13, 2015 4:19 AM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
See the bottom of the bug description. There's a mismatch between a 16bit integer array and a function that treats it as array of 32bit floats.
On 13 Dec 2015 3:30 am, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
According to the ARM docs, the instruction in question decodes to "coprocessor data transfer (LDC, STC)". The way that instruction is encoded (with the first four bits == xE), the instruction will be executed every time it's encountered (i.e., not conditionally).
Stanislav, you said you thought the trap was related to HD math and not alignment. Based on what that instruction apparently is, it looks to me like you were on the right track.
Do all ARM variants have an FPU, and if not, could this problem stem from configure et al. determining there's an FPU on a platform where there isn't one?
Hope this helps someone find and eradicate this bug.
- Larry
On Dec 12, 2015 7:04 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
https://freeswitch.org/jira/browse/FS-8651 the source of the problem is quite deep in the sources of spandsp, so I can't offer a fix.
On Thu, Dec 10, 2015 at 10:49 PM, Komar, Jason <jkomar@jbox.ca> wrote: > I get nothing in debug or sip trace. It always dies at > mod_spandsp_fax so I > figured it was in that module. At that point freeswitch just > restarts > without throwing any errors. > > Kernel log shows: > > Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not > handling > instruction ecb17a01 at [<b44fe32c>] > Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: > alignment > exception (0x001) at 0xb4546e4a > > The only part of the exception that is the same every time it > happens is > "instruction ecb17a01." > > I would have just opened a JIRA, but I didn't think this was enough > information. I have been Googling trying to figure out how to get > more info. > > Jason > > > On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin > <ssinyagin@gmail.com> > wrote: >> >> Hi, I didn't see any errors, but I'm not using fax, and mostly it's >> bridging the calls and connecting users to a conference. Do you >> have more >> debugging information? >> >> Most probably the source of the problem is that x86 allows >> unaligned >> memory access: a word pointer can have any value. And other >> architectures >> don't allow this, and require the pointers to be aligned to word >> boundaries. >> Mongodb, for example, is known to fail on any non-x86 CPU because >> of that. >> >> So, probably mod_spandsp needs an audit. >> >> On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote: >>> >>> Your instructions worked great for me building on/for my Odroid-U3 >>> (arm) >>> running Debian Jessie. >>> >>> I know you are running on arm at scaleway. Do you ever run into >>> any >>> alignment issues? I'm getting alignment exception faults receiving >>> faxes and >>> I'm having trouble pulling any useful debug messages. >>> >>> Thanks, >>> Jason >>> >>> On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin >>> <ssinyagin@gmail.com> >>> wrote: >>>> >>>> hi, >>>> >>>> I put together the build scenario that I tested for ARM (armhf at >>>> scaleway.com), and it should also work for other architectures: >>>> >>>> >>>> >>>> https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other... >>>> >>>> It will be great if someone tests them on other architectures >>>> too. >>>> >>>> >>>> cheers, >>>> stanislav >>>> >>>> >>>> >>>> _________________________________________________________________________ >>>> Professional FreeSWITCH Consulting Services: >>>> consulting@freeswitch.org >>>> http://www.freeswitchsolutions.com >>>> >>>> Official FreeSWITCH Sites >>>> http://www.freeswitch.org >>>> http://confluence.freeswitch.org >>>> http://www.cluecon.com >>>> >>>> FreeSWITCH-users mailing list >>>> FreeSWITCH-users@lists.freeswitch.org >>>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >>>> >>>> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users >>>> http://www.freeswitch.org >>> >>> >>> >>> >>> _________________________________________________________________________ >>> Professional FreeSWITCH Consulting Services: >>> consulting@freeswitch.org >>> http://www.freeswitchsolutions.com >>> >>> Official FreeSWITCH Sites >>> http://www.freeswitch.org >>> http://confluence.freeswitch.org >>> http://www.cluecon.com >>> >>> FreeSWITCH-users mailing list >>> FreeSWITCH-users@lists.freeswitch.org >>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >>> >>> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users >>> http://www.freeswitch.org >> >> >> >> _________________________________________________________________________ >> Professional FreeSWITCH Consulting Services: >> consulting@freeswitch.org >> http://www.freeswitchsolutions.com >> >> Official FreeSWITCH Sites >> http://www.freeswitch.org >> http://confluence.freeswitch.org >> http://www.cluecon.com >> >> FreeSWITCH-users mailing list >> FreeSWITCH-users@lists.freeswitch.org >> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >> >> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users >> http://www.freeswitch.org > > > > > _________________________________________________________________________ > Professional FreeSWITCH Consulting Services: > consulting@freeswitch.org > http://www.freeswitchsolutions.com > > Official FreeSWITCH Sites > http://www.freeswitch.org > http://confluence.freeswitch.org > http://www.cluecon.com > > FreeSWITCH-users mailing list > FreeSWITCH-users@lists.freeswitch.org > http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > > UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users > http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Hi, Stanislav, Okay, I'll take a stab at it some time between tonight and tomorrow, then. You still have access to the equipment at Scaleway, right? BTW it shouldn't matter, but, what version of spandsp, what Linux distro, what libc is on that host, etc.? Also, could you send me the output of "gcc -v" and "uname -i"? Thanks. - Larry On Dec 16, 2015 17:31, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
I tried to reach Steve on Sunday, but no success so far.
On Wed, Dec 16, 2015 at 12:25 AM, Larry Morley <lmorley@neny.cslimits.net> wrote:
Stanislav,
Sometime in the next few days I need to spend some quality time with that code, out of sheer curiosity if for no other reason. There are a few things in that code that have me - uh, stumped , to put it mildly.
Would you do me a favor, and if you do hear back from the author / maintainer, let me know what he says? Or when he fixes it?
Thanks , Larry
On Dec 14, 2015 3:16 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
Larry, here it's just a pointer that is 16bit aligned, and is given to a function that tries to treat it as an array of 32bit values. See my new comment in the ticket. We just need the module author to fix this, as
it's
difficult to figure out why some ifdefs have "x" at the end.
On 14 Dec 2015 19:03, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
Stanislav,
I saw that; what I was saying was it looked to me like the processor
was
interpreting data as code and it happened to work out that the bit pattern for that data and for LDC/STC are the same. I've run into problems before where misaligned inductions have resulted in an ARM processor either ignoring the improperly aligned portion of an instruction or treating it as data.
If you disassemble the code around where it traps, what does the listing show the processor is trying to execute?
One additional thought - if you're concerned that calloc() or malloc() is returning a pointer that's improperly aligned, you could always supply your own temporarily and see if the problem goes away.
- Larry
On Dec 13, 2015 4:19 AM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
See the bottom of the bug description. There's a mismatch between a 16bit integer array and a function that treats it as array of 32bit
floats.
On 13 Dec 2015 3:30 am, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
According to the ARM docs, the instruction in question decodes to "coprocessor data transfer (LDC, STC)". The way that instruction is
encoded
(with the first four bits == xE), the instruction will be executed every time it's encountered (i.e., not conditionally).
Stanislav, you said you thought the trap was related to HD math and not alignment. Based on what that instruction apparently is, it looks to me like you were on the right track.
Do all ARM variants have an FPU, and if not, could this problem stem from configure et al. determining there's an FPU on a platform where there isn't one?
Hope this helps someone find and eradicate this bug.
- Larry
On Dec 12, 2015 7:04 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote: > > https://freeswitch.org/jira/browse/FS-8651 > the source of the problem is quite deep in the sources of spandsp, so > I can't offer a fix. > > > On Thu, Dec 10, 2015 at 10:49 PM, Komar, Jason <jkomar@jbox.ca> wrote: > > I get nothing in debug or sip trace. It always dies at > > mod_spandsp_fax so I > > figured it was in that module. At that point freeswitch just > > restarts > > without throwing any errors. > > > > Kernel log shows: > > > > Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not > > handling > > instruction ecb17a01 at [<b44fe32c>] > > Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: > > alignment > > exception (0x001) at 0xb4546e4a > > > > The only part of the exception that is the same every time it > > happens is > > "instruction ecb17a01." > > > > I would have just opened a JIRA, but I didn't think this was enough > > information. I have been Googling trying to figure out how to get > > more info. > > > > Jason > > > > > > On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin > > <ssinyagin@gmail.com> > > wrote: > >> > >> Hi, I didn't see any errors, but I'm not using fax, and mostly it's > >> bridging the calls and connecting users to a conference. Do you > >> have more > >> debugging information? > >> > >> Most probably the source of the problem is that x86 allows > >> unaligned > >> memory access: a word pointer can have any value. And other > >> architectures > >> don't allow this, and require the pointers to be aligned to word > >> boundaries. > >> Mongodb, for example, is known to fail on any non-x86 CPU because > >> of that. > >> > >> So, probably mod_spandsp needs an audit. > >> > >> On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote: > >>> > >>> Your instructions worked great for me building on/for my Odroid-U3 > >>> (arm) > >>> running Debian Jessie. > >>> > >>> I know you are running on arm at scaleway. Do you ever run into > >>> any > >>> alignment issues? I'm getting alignment exception faults receiving > >>> faxes and > >>> I'm having trouble pulling any useful debug messages. > >>> > >>> Thanks, > >>> Jason > >>> > >>> On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin > >>> <ssinyagin@gmail.com> > >>> wrote: > >>>> > >>>> hi, > >>>> > >>>> I put together the build scenario that I tested for ARM (armhf at > >>>> scaleway.com), and it should also work for other architectures: > >>>> > >>>> > >>>> > >>>> https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other... > >>>> > >>>> It will be great if someone tests them on other architectures > >>>> too. > >>>> > >>>> > >>>> cheers, > >>>> stanislav > >>>> > >>>> > >>>> > >>>>
> >>>> Professional FreeSWITCH Consulting Services: > >>>> consulting@freeswitch.org > >>>> http://www.freeswitchsolutions.com > >>>> > >>>> Official FreeSWITCH Sites > >>>> http://www.freeswitch.org > >>>> http://confluence.freeswitch.org > >>>> http://www.cluecon.com > >>>> > >>>> FreeSWITCH-users mailing list > >>>> FreeSWITCH-users@lists.freeswitch.org > >>>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > >>>> > >>>> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users > >>>> http://www.freeswitch.org > >>> > >>> > >>> > >>> > >>>
> >>> Professional FreeSWITCH Consulting Services: > >>> consulting@freeswitch.org > >>> http://www.freeswitchsolutions.com > >>> > >>> Official FreeSWITCH Sites > >>> http://www.freeswitch.org > >>> http://confluence.freeswitch.org > >>> http://www.cluecon.com > >>> > >>> FreeSWITCH-users mailing list > >>> FreeSWITCH-users@lists.freeswitch.org > >>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > >>> > >>> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users > >>> http://www.freeswitch.org > >> > >> > >> > >>
> >> Professional FreeSWITCH Consulting Services: > >> consulting@freeswitch.org > >> http://www.freeswitchsolutions.com > >> > >> Official FreeSWITCH Sites > >> http://www.freeswitch.org > >> http://confluence.freeswitch.org > >> http://www.cluecon.com > >> > >> FreeSWITCH-users mailing list > >> FreeSWITCH-users@lists.freeswitch.org > >> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > >> > >> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users > >> http://www.freeswitch.org > > > > > > > > > >
> > Professional FreeSWITCH Consulting Services: > > consulting@freeswitch.org > > http://www.freeswitchsolutions.com > > > > Official FreeSWITCH Sites > > http://www.freeswitch.org > > http://confluence.freeswitch.org > > http://www.cluecon.com > > > > FreeSWITCH-users mailing list > > FreeSWITCH-users@lists.freeswitch.org > > http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > > > > UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users > > http://www.freeswitch.org > > >
> Professional FreeSWITCH Consulting Services: > consulting@freeswitch.org > http://www.freeswitchsolutions.com > > Official FreeSWITCH Sites > http://www.freeswitch.org > http://confluence.freeswitch.org > http://www.cluecon.com > > FreeSWITCH-users mailing list > FreeSWITCH-users@lists.freeswitch.org > http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > > UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users > http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
The spandsp is always the version in our tree, which is not a release version of spandsp. It is maintained by the maintainer of spandsp and is far ahead of any release version by multiple years. Its still not released due to changing api that is being worked on, but will at some point in the future, at which point we will shift to system packages for it.
On Dec 16, 2015, at 6:46 PM, Larry Morley <lmorley@neny.cslimits.net> wrote:
Hi, Stanislav,
Okay, I'll take a stab at it some time between tonight and tomorrow, then.
You still have access to the equipment at Scaleway, right?
BTW it shouldn't matter, but, what version of spandsp, what Linux distro, what libc is on that host, etc.? Also, could you send me the output of "gcc -v" and "uname -i"? Thanks.
- Larry
On Dec 16, 2015 17:31, "Stanislav Sinyagin" <ssinyagin@gmail.com <mailto:ssinyagin@gmail.com>> wrote: I tried to reach Steve on Sunday, but no success so far.
On Wed, Dec 16, 2015 at 12:25 AM, Larry Morley <lmorley@neny.cslimits.net <mailto:lmorley@neny.cslimits.net>> wrote:
Stanislav,
Sometime in the next few days I need to spend some quality time with that code, out of sheer curiosity if for no other reason. There are a few things in that code that have me - uh, stumped , to put it mildly.
Would you do me a favor, and if you do hear back from the author / maintainer, let me know what he says? Or when he fixes it?
Thanks , Larry
On Dec 14, 2015 3:16 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com <mailto:ssinyagin@gmail.com>> wrote:
Larry, here it's just a pointer that is 16bit aligned, and is given to a function that tries to treat it as an array of 32bit values. See my new comment in the ticket. We just need the module author to fix this, as it's difficult to figure out why some ifdefs have "x" at the end.
On 14 Dec 2015 19:03, "Larry Morley" <lmorley@neny.cslimits.net <mailto:lmorley@neny.cslimits.net>> wrote:
Stanislav,
I saw that; what I was saying was it looked to me like the processor was interpreting data as code and it happened to work out that the bit pattern for that data and for LDC/STC are the same. I've run into problems before where misaligned inductions have resulted in an ARM processor either ignoring the improperly aligned portion of an instruction or treating it as data.
If you disassemble the code around where it traps, what does the listing show the processor is trying to execute?
One additional thought - if you're concerned that calloc() or malloc() is returning a pointer that's improperly aligned, you could always supply your own temporarily and see if the problem goes away.
- Larry
On Dec 13, 2015 4:19 AM, "Stanislav Sinyagin" <ssinyagin@gmail.com <mailto:ssinyagin@gmail.com>> wrote:
See the bottom of the bug description. There's a mismatch between a 16bit integer array and a function that treats it as array of 32bit floats.
On 13 Dec 2015 3:30 am, "Larry Morley" <lmorley@neny.cslimits.net <mailto:lmorley@neny.cslimits.net>> wrote:
According to the ARM docs, the instruction in question decodes to "coprocessor data transfer (LDC, STC)". The way that instruction is encoded (with the first four bits == xE), the instruction will be executed every time it's encountered (i.e., not conditionally).
Stanislav, you said you thought the trap was related to HD math and not alignment. Based on what that instruction apparently is, it looks to me like you were on the right track.
Do all ARM variants have an FPU, and if not, could this problem stem from configure et al. determining there's an FPU on a platform where there isn't one?
Hope this helps someone find and eradicate this bug.
- Larry
On Dec 12, 2015 7:04 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com <mailto:ssinyagin@gmail.com>> wrote: > > https://freeswitch.org/jira/browse/FS-8651 <https://freeswitch.org/jira/browse/FS-8651> > the source of the problem is quite deep in the sources of spandsp, so > I can't offer a fix. > > > On Thu, Dec 10, 2015 at 10:49 PM, Komar, Jason <jkomar@jbox.ca <mailto:jkomar@jbox.ca>> wrote: > > I get nothing in debug or sip trace. It always dies at > > mod_spandsp_fax so I > > figured it was in that module. At that point freeswitch just > > restarts > > without throwing any errors. > > > > Kernel log shows: > > > > Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not > > handling > > instruction ecb17a01 at [<b44fe32c>] > > Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: > > alignment > > exception (0x001) at 0xb4546e4a > > > > The only part of the exception that is the same every time it > > happens is > > "instruction ecb17a01." > > > > I would have just opened a JIRA, but I didn't think this was enough > > information. I have been Googling trying to figure out how to get > > more info. > > > > Jason > > > > > > On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin > > <ssinyagin@gmail.com <mailto:ssinyagin@gmail.com>> > > wrote: > >> > >> Hi, I didn't see any errors, but I'm not using fax, and mostly it's > >> bridging the calls and connecting users to a conference. Do you > >> have more > >> debugging information? > >> > >> Most probably the source of the problem is that x86 allows > >> unaligned > >> memory access: a word pointer can have any value. And other > >> architectures > >> don't allow this, and require the pointers to be aligned to word > >> boundaries. > >> Mongodb, for example, is known to fail on any non-x86 CPU because > >> of that. > >> > >> So, probably mod_spandsp needs an audit. > >> > >> On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca <mailto:jkomar@jbox.ca>> wrote: > >>> > >>> Your instructions worked great for me building on/for my Odroid-U3 > >>> (arm) > >>> running Debian Jessie. > >>> > >>> I know you are running on arm at scaleway. Do you ever run into > >>> any > >>> alignment issues? I'm getting alignment exception faults receiving > >>> faxes and > >>> I'm having trouble pulling any useful debug messages. > >>> > >>> Thanks, > >>> Jason > >>> > >>> On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin > >>> <ssinyagin@gmail.com <mailto:ssinyagin@gmail.com>> > >>> wrote: > >>>> > >>>> hi, > >>>> > >>>> I put together the build scenario that I tested for ARM (armhf at > >>>> scaleway.com <http://scaleway.com/>), and it should also work for other architectures: > >>>> > >>>> > >>>> > >>>> https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other... <https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other+architectures> > >>>> > >>>> It will be great if someone tests them on other architectures > >>>> too. > >>>>
Hi Larry, yes, I'm actually running my production PBX at scaleway now. I'm just not using fax :) This is spandsp from FreeSWITCH master branch source tree, and the build procedure is described in confluence. It's a standard Debian 8. I have also daily debs for ARM in my private repo. If needed, I can create a new host for your tests. But so far, the source of the error is already clear, and we only need to understand the reasoning why there are "x" at the end of ifdefs in so many places. They seem to be placed there on purpose. On 17 Dec 2015 00:47, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
Hi, Stanislav,
Okay, I'll take a stab at it some time between tonight and tomorrow, then.
You still have access to the equipment at Scaleway, right?
BTW it shouldn't matter, but, what version of spandsp, what Linux distro, what libc is on that host, etc.? Also, could you send me the output of "gcc -v" and "uname -i"? Thanks.
- Larry On Dec 16, 2015 17:31, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
I tried to reach Steve on Sunday, but no success so far.
On Wed, Dec 16, 2015 at 12:25 AM, Larry Morley <lmorley@neny.cslimits.net> wrote:
Stanislav,
Sometime in the next few days I need to spend some quality time with that code, out of sheer curiosity if for no other reason. There are a few things in that code that have me - uh, stumped , to put it mildly.
Would you do me a favor, and if you do hear back from the author / maintainer, let me know what he says? Or when he fixes it?
Thanks , Larry
On Dec 14, 2015 3:16 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
Larry, here it's just a pointer that is 16bit aligned, and is given to
a
function that tries to treat it as an array of 32bit values. See my new comment in the ticket. We just need the module author to fix this, as it's difficult to figure out why some ifdefs have "x" at the end.
On 14 Dec 2015 19:03, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
Stanislav,
I saw that; what I was saying was it looked to me like the processor
was
interpreting data as code and it happened to work out that the bit pattern for that data and for LDC/STC are the same. I've run into problems before where misaligned inductions have resulted in an ARM processor either ignoring the improperly aligned portion of an instruction or treating it as data.
If you disassemble the code around where it traps, what does the listing show the processor is trying to execute?
One additional thought - if you're concerned that calloc() or malloc() is returning a pointer that's improperly aligned, you could always supply your own temporarily and see if the problem goes away.
- Larry
On Dec 13, 2015 4:19 AM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
See the bottom of the bug description. There's a mismatch between a 16bit integer array and a function that treats it as array of 32bit
floats.
On 13 Dec 2015 3:30 am, "Larry Morley" <lmorley@neny.cslimits.net> wrote: > > According to the ARM docs, the instruction in question decodes to > "coprocessor data transfer (LDC, STC)". The way that instruction is
encoded
> (with the first four bits == xE), the instruction will be executed every > time it's encountered (i.e., not conditionally). > > Stanislav, you said you thought the trap was related to HD math and not > alignment. Based on what that instruction apparently is, it looks to me like > you were on the right track. > > Do all ARM variants have an FPU, and if not, could this problem stem > from configure et al. determining there's an FPU on a platform where there > isn't one? > > Hope this helps someone find and eradicate this bug. > > - Larry > > On Dec 12, 2015 7:04 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> > wrote: >> >> https://freeswitch.org/jira/browse/FS-8651 >> the source of the problem is quite deep in the sources of spandsp, so >> I can't offer a fix. >> >> >> On Thu, Dec 10, 2015 at 10:49 PM, Komar, Jason <jkomar@jbox.ca> wrote: >> > I get nothing in debug or sip trace. It always dies at >> > mod_spandsp_fax so I >> > figured it was in that module. At that point freeswitch just >> > restarts >> > without throwing any errors. >> > >> > Kernel log shows: >> > >> > Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not >> > handling >> > instruction ecb17a01 at [<b44fe32c>] >> > Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: >> > alignment >> > exception (0x001) at 0xb4546e4a >> > >> > The only part of the exception that is the same every time it >> > happens is >> > "instruction ecb17a01." >> > >> > I would have just opened a JIRA, but I didn't think this was enough >> > information. I have been Googling trying to figure out how to get >> > more info. >> > >> > Jason >> > >> > >> > On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin >> > <ssinyagin@gmail.com> >> > wrote: >> >> >> >> Hi, I didn't see any errors, but I'm not using fax, and mostly it's >> >> bridging the calls and connecting users to a conference. Do you >> >> have more >> >> debugging information? >> >> >> >> Most probably the source of the problem is that x86 allows >> >> unaligned >> >> memory access: a word pointer can have any value. And other >> >> architectures >> >> don't allow this, and require the pointers to be aligned to word >> >> boundaries. >> >> Mongodb, for example, is known to fail on any non-x86 CPU because >> >> of that. >> >> >> >> So, probably mod_spandsp needs an audit. >> >> >> >> On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote: >> >>> >> >>> Your instructions worked great for me building on/for my Odroid-U3 >> >>> (arm) >> >>> running Debian Jessie. >> >>> >> >>> I know you are running on arm at scaleway. Do you ever run into >> >>> any >> >>> alignment issues? I'm getting alignment exception faults receiving >> >>> faxes and >> >>> I'm having trouble pulling any useful debug messages. >> >>> >> >>> Thanks, >> >>> Jason >> >>> >> >>> On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin >> >>> <ssinyagin@gmail.com> >> >>> wrote: >> >>>> >> >>>> hi, >> >>>> >> >>>> I put together the build scenario that I tested for ARM (armhf at >> >>>> scaleway.com), and it should also work for other architectures: >> >>>> >> >>>> >> >>>> >> >>>> https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other... >> >>>> >> >>>> It will be great if someone tests them on other architectures >> >>>> too. >> >>>> >> >>>> >> >>>> cheers, >> >>>> stanislav >> >>>> >> >>>> >> >>>> >> >>>>
>> >>>> Professional FreeSWITCH Consulting Services: >> >>>> consulting@freeswitch.org >> >>>> http://www.freeswitchsolutions.com >> >>>> >> >>>> Official FreeSWITCH Sites >> >>>> http://www.freeswitch.org >> >>>> http://confluence.freeswitch.org >> >>>> http://www.cluecon.com >> >>>> >> >>>> FreeSWITCH-users mailing list >> >>>> FreeSWITCH-users@lists.freeswitch.org >> >>>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >> >>>> >> >>>> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >> >>>> http://www.freeswitch.org >> >>> >> >>> >> >>> >> >>> >> >>>
>> >>> Professional FreeSWITCH Consulting Services: >> >>> consulting@freeswitch.org >> >>> http://www.freeswitchsolutions.com >> >>> >> >>> Official FreeSWITCH Sites >> >>> http://www.freeswitch.org >> >>> http://confluence.freeswitch.org >> >>> http://www.cluecon.com >> >>> >> >>> FreeSWITCH-users mailing list >> >>> FreeSWITCH-users@lists.freeswitch.org >> >>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >> >>> >> >>> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >> >>> http://www.freeswitch.org >> >> >> >> >> >> >> >>
>> >> Professional FreeSWITCH Consulting Services: >> >> consulting@freeswitch.org >> >> http://www.freeswitchsolutions.com >> >> >> >> Official FreeSWITCH Sites >> >> http://www.freeswitch.org >> >> http://confluence.freeswitch.org >> >> http://www.cluecon.com >> >> >> >> FreeSWITCH-users mailing list >> >> FreeSWITCH-users@lists.freeswitch.org >> >> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >> >> >> >> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >> >> http://www.freeswitch.org >> > >> > >> > >> > >> >
>> > Professional FreeSWITCH Consulting Services: >> > consulting@freeswitch.org >> > http://www.freeswitchsolutions.com >> > >> > Official FreeSWITCH Sites >> > http://www.freeswitch.org >> > http://confluence.freeswitch.org >> > http://www.cluecon.com >> > >> > FreeSWITCH-users mailing list >> > FreeSWITCH-users@lists.freeswitch.org >> > http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >> > >> > UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >> > http://www.freeswitch.org >> >> >>
>> Professional FreeSWITCH Consulting Services: >> consulting@freeswitch.org >> http://www.freeswitchsolutions.com >> >> Official FreeSWITCH Sites >> http://www.freeswitch.org >> http://confluence.freeswitch.org >> http://www.cluecon.com >> >> FreeSWITCH-users mailing list >> FreeSWITCH-users@lists.freeswitch.org >> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >> >> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >> http://www.freeswitch.org > > > >
> Professional FreeSWITCH Consulting Services: > consulting@freeswitch.org > http://www.freeswitchsolutions.com > > Official FreeSWITCH Sites > http://www.freeswitch.org > http://confluence.freeswitch.org > http://www.cluecon.com > > FreeSWITCH-users mailing list > FreeSWITCH-users@lists.freeswitch.org > http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > > UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users > http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Mike: thanks; that clears up a few things. Stanislav: if it wouldn't be too much trouble, could you spin up a host that I could use for a few days? If you could set it up so it's a mirror image of yours, that'd be great. Thanks Larry On Dec 16, 2015 19:37, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
Hi Larry, yes, I'm actually running my production PBX at scaleway now. I'm just not using fax :)
This is spandsp from FreeSWITCH master branch source tree, and the build procedure is described in confluence. It's a standard Debian 8. I have also daily debs for ARM in my private repo. If needed, I can create a new host for your tests.
But so far, the source of the error is already clear, and we only need to understand the reasoning why there are "x" at the end of ifdefs in so many places. They seem to be placed there on purpose.
On 17 Dec 2015 00:47, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
Hi, Stanislav,
Okay, I'll take a stab at it some time between tonight and tomorrow, then.
You still have access to the equipment at Scaleway, right?
BTW it shouldn't matter, but, what version of spandsp, what Linux distro, what libc is on that host, etc.? Also, could you send me the output of "gcc -v" and "uname -i"? Thanks.
- Larry On Dec 16, 2015 17:31, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
I tried to reach Steve on Sunday, but no success so far.
On Wed, Dec 16, 2015 at 12:25 AM, Larry Morley <lmorley@neny.cslimits.net> wrote:
Stanislav,
Sometime in the next few days I need to spend some quality time with that code, out of sheer curiosity if for no other reason. There are a few things in that code that have me - uh, stumped , to put it mildly.
Would you do me a favor, and if you do hear back from the author / maintainer, let me know what he says? Or when he fixes it?
Thanks , Larry
On Dec 14, 2015 3:16 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
Larry, here it's just a pointer that is 16bit aligned, and is given
to a
function that tries to treat it as an array of 32bit values. See my new comment in the ticket. We just need the module author to fix this, as it's difficult to figure out why some ifdefs have "x" at the end.
On 14 Dec 2015 19:03, "Larry Morley" <lmorley@neny.cslimits.net> wrote:
Stanislav,
I saw that; what I was saying was it looked to me like the
processor was
interpreting data as code and it happened to work out that the bit pattern for that data and for LDC/STC are the same. I've run into problems before where misaligned inductions have resulted in an ARM processor either ignoring the improperly aligned portion of an instruction or treating it as data.
If you disassemble the code around where it traps, what does the listing show the processor is trying to execute?
One additional thought - if you're concerned that calloc() or malloc() is returning a pointer that's improperly aligned, you could always supply your own temporarily and see if the problem goes away.
- Larry
On Dec 13, 2015 4:19 AM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote: > > See the bottom of the bug description. There's a mismatch between a > 16bit integer array and a function that treats it as array of 32bit floats. > > On 13 Dec 2015 3:30 am, "Larry Morley" <lmorley@neny.cslimits.net> > wrote: >> >> According to the ARM docs, the instruction in question decodes to >> "coprocessor data transfer (LDC, STC)". The way that instruction is encoded >> (with the first four bits == xE), the instruction will be executed every >> time it's encountered (i.e., not conditionally). >> >> Stanislav, you said you thought the trap was related to HD math and not >> alignment. Based on what that instruction apparently is, it looks to me like >> you were on the right track. >> >> Do all ARM variants have an FPU, and if not, could this problem stem >> from configure et al. determining there's an FPU on a platform where there >> isn't one? >> >> Hope this helps someone find and eradicate this bug. >> >> - Larry >> >> On Dec 12, 2015 7:04 PM, "Stanislav Sinyagin" <ssinyagin@gmail.com
>> wrote: >>> >>> https://freeswitch.org/jira/browse/FS-8651 >>> the source of the problem is quite deep in the sources of spandsp, so >>> I can't offer a fix. >>> >>> >>> On Thu, Dec 10, 2015 at 10:49 PM, Komar, Jason <jkomar@jbox.ca> wrote: >>> > I get nothing in debug or sip trace. It always dies at >>> > mod_spandsp_fax so I >>> > figured it was in that module. At that point freeswitch just >>> > restarts >>> > without throwing any errors. >>> > >>> > Kernel log shows: >>> > >>> > Dec 10 14:16:17 antman kernel: [ 5949.832802] Alignment trap: not >>> > handling >>> > instruction ecb17a01 at [<b44fe32c>] >>> > Dec 10 14:16:17 antman kernel: [ 5949.834290] Unhandled fault: >>> > alignment >>> > exception (0x001) at 0xb4546e4a >>> > >>> > The only part of the exception that is the same every time it >>> > happens is >>> > "instruction ecb17a01." >>> > >>> > I would have just opened a JIRA, but I didn't think this was enough >>> > information. I have been Googling trying to figure out how to get >>> > more info. >>> > >>> > Jason >>> > >>> > >>> > On Thu, Dec 10, 2015 at 2:37 PM, Stanislav Sinyagin >>> > <ssinyagin@gmail.com> >>> > wrote: >>> >> >>> >> Hi, I didn't see any errors, but I'm not using fax, and mostly it's >>> >> bridging the calls and connecting users to a conference. Do you >>> >> have more >>> >> debugging information? >>> >> >>> >> Most probably the source of the problem is that x86 allows >>> >> unaligned >>> >> memory access: a word pointer can have any value. And other >>> >> architectures >>> >> don't allow this, and require the pointers to be aligned to word >>> >> boundaries. >>> >> Mongodb, for example, is known to fail on any non-x86 CPU because >>> >> of that. >>> >> >>> >> So, probably mod_spandsp needs an audit. >>> >> >>> >> On 10 Dec 2015 21:11, "Komar, Jason" <jkomar@jbox.ca> wrote: >>> >>> >>> >>> Your instructions worked great for me building on/for my Odroid-U3 >>> >>> (arm) >>> >>> running Debian Jessie. >>> >>> >>> >>> I know you are running on arm at scaleway. Do you ever run into >>> >>> any >>> >>> alignment issues? I'm getting alignment exception faults receiving >>> >>> faxes and >>> >>> I'm having trouble pulling any useful debug messages. >>> >>> >>> >>> Thanks, >>> >>> Jason >>> >>> >>> >>> On Mon, Nov 30, 2015 at 9:27 AM, Stanislav Sinyagin >>> >>> <ssinyagin@gmail.com> >>> >>> wrote: >>> >>>> >>> >>>> hi, >>> >>>> >>> >>>> I put together the build scenario that I tested for ARM (armhf at >>> >>>> scaleway.com), and it should also work for other architectures: >>> >>>> >>> >>>> >>> >>>> >>> >>>> https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other... >>> >>>> >>> >>>> It will be great if someone tests them on other architectures >>> >>>> too. >>> >>>> >>> >>>> >>> >>>> cheers, >>> >>>> stanislav >>> >>>> >>> >>>> >>> >>>> >>> >>>>
>>> >>>> Professional FreeSWITCH Consulting Services: >>> >>>> consulting@freeswitch.org >>> >>>> http://www.freeswitchsolutions.com >>> >>>> >>> >>>> Official FreeSWITCH Sites >>> >>>> http://www.freeswitch.org >>> >>>> http://confluence.freeswitch.org >>> >>>> http://www.cluecon.com >>> >>>> >>> >>>> FreeSWITCH-users mailing list >>> >>>> FreeSWITCH-users@lists.freeswitch.org >>> >>>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >>> >>>> >>> >>>> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >>> >>>> http://www.freeswitch.org >>> >>> >>> >>> >>> >>> >>> >>> >>> >>>
>>> >>> Professional FreeSWITCH Consulting Services: >>> >>> consulting@freeswitch.org >>> >>> http://www.freeswitchsolutions.com >>> >>> >>> >>> Official FreeSWITCH Sites >>> >>> http://www.freeswitch.org >>> >>> http://confluence.freeswitch.org >>> >>> http://www.cluecon.com >>> >>> >>> >>> FreeSWITCH-users mailing list >>> >>> FreeSWITCH-users@lists.freeswitch.org >>> >>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >>> >>> >>> >>> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >>> >>> http://www.freeswitch.org >>> >> >>> >> >>> >> >>> >>
>>> >> Professional FreeSWITCH Consulting Services: >>> >> consulting@freeswitch.org >>> >> http://www.freeswitchsolutions.com >>> >> >>> >> Official FreeSWITCH Sites >>> >> http://www.freeswitch.org >>> >> http://confluence.freeswitch.org >>> >> http://www.cluecon.com >>> >> >>> >> FreeSWITCH-users mailing list >>> >> FreeSWITCH-users@lists.freeswitch.org >>> >> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >>> >> >>> >> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >>> >> http://www.freeswitch.org >>> > >>> > >>> > >>> > >>> >
>>> > Professional FreeSWITCH Consulting Services: >>> > consulting@freeswitch.org >>> > http://www.freeswitchsolutions.com >>> > >>> > Official FreeSWITCH Sites >>> > http://www.freeswitch.org >>> > http://confluence.freeswitch.org >>> > http://www.cluecon.com >>> > >>> > FreeSWITCH-users mailing list >>> > FreeSWITCH-users@lists.freeswitch.org >>> > http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >>> > >>> > UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >>> > http://www.freeswitch.org >>> >>> >>>
>>> Professional FreeSWITCH Consulting Services: >>> consulting@freeswitch.org >>> http://www.freeswitchsolutions.com >>> >>> Official FreeSWITCH Sites >>> http://www.freeswitch.org >>> http://confluence.freeswitch.org >>> http://www.cluecon.com >>> >>> FreeSWITCH-users mailing list >>> FreeSWITCH-users@lists.freeswitch.org >>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >>> >>> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >>> http://www.freeswitch.org >> >> >> >>
>> Professional FreeSWITCH Consulting Services: >> consulting@freeswitch.org >> http://www.freeswitchsolutions.com >> >> Official FreeSWITCH Sites >> http://www.freeswitch.org >> http://confluence.freeswitch.org >> http://www.cluecon.com >> >> FreeSWITCH-users mailing list >> FreeSWITCH-users@lists.freeswitch.org >> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users >> >> UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users >> http://www.freeswitch.org > > > >
> Professional FreeSWITCH Consulting Services: > consulting@freeswitch.org > http://www.freeswitchsolutions.com > > Official FreeSWITCH Sites > http://www.freeswitch.org > http://confluence.freeswitch.org > http://www.cluecon.com > > FreeSWITCH-users mailing list > FreeSWITCH-users@lists.freeswitch.org > http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users > http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________
Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Stanislav, A few months back - it was a few weeks after ClueCon; that's the closest I can come relying on my memory, and at the moment, that's the best I can do - I posted a similar announcement; the difference was that I had rebuilt everything for the PPC. I also posted a note regarding Tyan's release of a POWER 8 reference system; my goal was to see how many others in the FS community were interested in using either the current POWER platforms or the new one. I never did hear back from anyone, which surprised me; apparently, however, now that it's apparent we're both targeting non Intel platforms, perhaps, going forward, we can combine our efforts? Along, of course, with anyone else who wishes to toss their hat in the ring. We routinely build, test and deploy FS on numerous hardware and software platforms. Sincerely, Larry Morley ------------------------------------------------------------ Lawrence J. "Larry" Morley R&D Director and CTO Enlightened Computer Solutions, LLC 299 Hyspot Road, #9-D Greenfield Center , New York 12833 On Nov 30, 2015 11:28 AM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Larry, I guess POWER is a much less popular platform, so you won't get the testers easily. but ARM is widespread thanks to cheap boards like Raspberry Pi. Building FreeSWITCH 1.4 under Debian Wheezy was quite an easy task, there was only one small patch that used a newer library because the one that was used by default was not compatible with non-x86. The 1.6 under Jessie took a bit of an effort, as all prerequisites needed to be compiled separately. Now the build procedure is stable and it should work for any CPU architecture. Jason gave me access to the machine where he discovered the problem with fax receiving, so I'll open a jira ticket as soon as I test a fix. I thought it's a problem with memory alignment, but it looks like it's rather a problem with floating-point maths. On Sat, Dec 12, 2015 at 1:51 AM, Larry Morley <lmorley@neny.cslimits.net> wrote:
Stanislav,
A few months back - it was a few weeks after ClueCon; that's the closest I can come relying on my memory, and at the moment, that's the best I can do - I posted a similar announcement; the difference was that I had rebuilt everything for the PPC. I also posted a note regarding Tyan's release of a POWER 8 reference system; my goal was to see how many others in the FS community were interested in using either the current POWER platforms or the new one.
I never did hear back from anyone, which surprised me; apparently, however, now that it's apparent we're both targeting non Intel platforms, perhaps, going forward, we can combine our efforts? Along, of course, with anyone else who wishes to toss their hat in the ring.
We routinely build, test and deploy FS on numerous hardware and software platforms.
Sincerely,
Larry Morley
------------------------------------------------------------ Lawrence J. "Larry" Morley R&D Director and CTO Enlightened Computer Solutions, LLC 299 Hyspot Road, #9-D Greenfield Center , New York 12833
On Nov 30, 2015 11:28 AM, "Stanislav Sinyagin" <ssinyagin@gmail.com> wrote:
hi,
I put together the build scenario that I tested for ARM (armhf at scaleway.com), and it should also work for other architectures:
https://freeswitch.org/confluence/display/FREESWITCH/Building+debs+for+other...
It will be great if someone tests them on other architectures too.
cheers, stanislav
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
participants (5)
-
Komar, Jason
-
Larry Morley
-
Michael Jerris
-
Olgierd Ziolko
-
Stanislav Sinyagin