[OpenSIPS-Users] Cross Compile OpenSIPS ARMv7
Caio
caiortp at gmail.com
Wed Oct 21 15:37:36 CEST 2015
It's possible to disable fastlock.h and futex_lock.h in the Makefile.conf?
Thanks
Regards,
Caio Pereira
2015-10-20 18:49 GMT-02:00 Caio <caiortp at gmail.com>:
> Ovidiu,
>
> My Log
>
> echo $CC_EXTRA_OPTS
> -march=armv7-a -marm -mthumb-interwork -mfloat-abi=hard -mfpu=neon
> --sysroot=/home/caio/fsl-community-bsp/imx6/tmp/sysroots/imx6"
>
> CPU=armv7a
> LD=arm-poky-linux-gnueabi-ld
> --sysroot=/home/caio/fsl-community-bsp/imx6/tmp/sysroots/imx6
>
> CC=arm-poky-linux-gnueabi-gcc -march=armv7-a -marm -mthumb-interwork
> -mfloat-abi=hard -mfpu=neon
> --sysroot=/home/caio/fsl-community-bsp/imx6/tmp/sysroots/imx6
>
>
> cat Makefile.conf
>
> #DEFS+= -DPKG_MALLOC #Uses a faster malloc (exclusive w/ USE_SHM_MEM)
> DEFS+= -DSHM_MMAP #Use mmap instead of SYSV shared memory
> DEFS+= -DUSE_MCAST #Compile in support for IP Multicast
> DEFS+= -DDISABLE_NAGLE #Disabled the TCP NAgle Algorithm ( lower delay )
> DEFS+= -DSTATISTICS #Enables the statistics manager
> DEFS+= -DHAVE_RESOLV_RES #Support for changing some of the resolver
> parameters
> #DEFS+= -DHP_MALLOC #High performance allocator with fine-grained locking
> #DEFS+= -DF_MALLOC #An even faster allocator. Not recommended for debugging
> #DEFS+= -DF_MALLOC_OPTIMIZATIONS #Remove all internal checks in F_MALLOC
> #DEFS+= -DDBG_QM_MALLOC #Allocator used for debugging information
> #DEFS+= -DUSE_SHM_MEM #All PKG allocations are mapped to SHM ( exclusive
> w/ PKG_MALLOC )
> #DEFS+= -DDBG_F_MALLOC #TODO ?
> #DEFS+= -DNO_DEBUG #Turns off all debug messages
> #DEFS+= -DNO_LOG #Completely turns off all the logging
> #DEFS+= -DVQ_MALLOC #TODO ?
> #DEFS+= -DFAST_LOCK #Uses fast architecture specific locking
> #DEFS+= -DUSE_FUTEX #Uses linux futexs with fast architecture specific
> locking
> DEFS+= -DUSE_SYSV_SEM #Uses SYSV sems for locking ( slower & limited
> number of locks
> #DEFS+= -DUSE_PTHREAD_MUTEX #Uses pthread mutexes
> #DEFS+= -DBUSY_WAIT #Uses busy waiting on the lock
> #DEFS+= -DDBG_LOCK #TODO ?
> DEFS+= -DNOSMP #Do not use SMP sompliant locking. Faster but won't work on
> SMP machines
> #DEFS+= -DEXTRA_DEBUG #Compiles in some extra debugging code
>
>
> make all
> Compiling blacklists.c
> In file included from blacklists.h:30:0,
> from blacklists.c:38:
> ip_addr.h: In function 'sockaddr2ip_addr':
> ip_addr.h:224:25: warning: cast increases required alignment of target
> type [-Wcast-align]
> memcpy(ip->u.addr, &((struct sockaddr_in*)sa)->sin_addr, 4);
> ^
> ip_addr.h:229:25: warning: cast increases required alignment of target
> type [-Wcast-align]
> memcpy(ip->u.addr, &((struct sockaddr_in6*)sa)->sin6_addr, 16);
> ^
> In file included from locking.h:66:0,
> from blacklists.h:32,
> from blacklists.c:38:
> lock_alloc.h: In function 'lock_set_alloc':
> lock_alloc.h:70:13: warning: cast increases required alignment of target
> type [-Wcast-align]
> ls->locks=(gen_lock_t*)((char*)ls+sizeof(gen_lock_set_t));
> ^
> {standard input}: Assembler messages:
> {standard input}:105: Error: bad instruction `btsl $0,[r6]'
> {standard input}:106: Error: bad instruction `adcl $0,r3'
> {standard input}:208: Error: bad instruction `movb $0,(r3)'
> {standard input}:305: Error: bad instruction `btsl $0,[r9]'
> {standard input}:306: Error: bad instruction `adcl $0,r3'
> {standard input}:390: Error: bad instruction `movb $0,(r3)'
> {standard input}:436: Error: bad instruction `btsl $0,[r10]'
> {standard input}:437: Error: bad instruction `adcl $0,r3'
> {standard input}:480: Error: bad instruction `movb $0,(r3)'
> {standard input}:526: Error: bad instruction `btsl $0,[r10]'
> {standard input}:527: Error: bad instruction `adcl $0,r3'
> {standard input}:555: Error: bad instruction `movb $0,(r2)'
> {standard input}:619: Error: bad instruction `btsl $0,[r10]'
> {standard input}:620: Error: bad instruction `adcl $0,r3'
> {standard input}:718: Error: bad instruction `movb $0,(r3)'
> {standard input}:3194: Error: bad instruction `btsl $0,[r7]'
> {standard input}:3195: Error: bad instruction `adcl $0,r6'
> {standard input}:3249: Error: bad instruction `movb $0,(r3)'
> {standard input}:3300: Error: bad instruction `btsl $0,[r7]'
> {standard input}:3301: Error: bad instruction `adcl $0,r3'
> {standard input}:3340: Error: bad instruction `movb $0,(r0)'
> {standard input}:3366: Error: bad instruction `btsl $0,[r6]'
> {standard input}:3367: Error: bad instruction `adcl $0,r3'
> {standard input}:3411: Error: bad instruction `movb $0,(r1)'
> {standard input}:3497: Error: bad instruction `btsl $0,[r6]'
> {standard input}:3498: Error: bad instruction `adcl $0,r3'
> {standard input}:3542: Error: bad instruction `movb $0,(r1)'
> {standard input}:3795: Error: bad instruction `btsl $0,[r6]'
> {standard input}:3796: Error: bad instruction `adcl $0,r3'
> {standard input}:3839: Error: bad instruction `movb $0,(r3)'
> {standard input}:3882: Error: bad instruction `btsl $0,[fp]'
> {standard input}:3883: Error: bad instruction `adcl $0,r3'
> {standard input}:3981: Error: bad instruction `movb $0,(r3)'
> {standard input}:4020: Error: bad instruction `btsl $0,[r6]'
> {standard input}:4021: Error: bad instruction `adcl $0,r3'
> {standard input}:4081: Error: bad instruction `movb $0,(r3)'
> {standard input}:4129: Error: bad instruction `btsl $0,[r10]'
> {standard input}:4130: Error: bad instruction `adcl $0,r3'
> {standard input}:4381: Error: bad instruction `movb $0,(r3)'
> {standard input}:4837: Error: bad instruction `btsl $0,[fp]'
> {standard input}:4838: Error: bad instruction `adcl $0,r3'
> {standard input}:4951: Error: bad instruction `movb $0,(r3)'
> {standard input}:5312: Error: bad instruction `btsl $0,[r6]'
> {standard input}:5313: Error: bad instruction `adcl $0,r9'
> {standard input}:5383: Error: bad instruction `btsl $0,[r8]'
> {standard input}:5384: Error: bad instruction `adcl $0,r5'
> {standard input}:5471: Error: bad instruction `movb $0,(r3)'
> {standard input}:5510: Error: bad instruction `btsl $0,[r6]'
> {standard input}:5511: Error: bad instruction `adcl $0,r3'
> {standard input}:5557: Error: bad instruction `movb $0,(r3)'
> {standard input}:5596: Error: bad instruction `btsl $0,[r9]'
> {standard input}:5597: Error: bad instruction `adcl $0,r3'
> {standard input}:5628: Error: bad instruction `movb $0,(r3)'
> {standard input}:5889: Error: bad instruction `movb $0,(r3)'
> {standard input}:5928: Error: bad instruction `btsl $0,[r8]'
> {standard input}:5929: Error: bad instruction `adcl $0,r3'
> {standard input}:5975: Error: bad instruction `movb $0,(r3)'
> {standard input}:6014: Error: bad instruction `btsl $0,[r9]'
> {standard input}:6015: Error: bad instruction `adcl $0,r3'
> {standard input}:6045: Error: bad instruction `movb $0,(r3)'
> {standard input}:6103: Error: bad instruction `btsl $0,[r10]'
> {standard input}:6104: Error: bad instruction `adcl $0,r3'
> {standard input}:6203: Error: bad instruction `movb $0,(r3)'
> {standard input}:6323: Error: bad instruction `btsl $0,[r6]'
> {standard input}:6324: Error: bad instruction `adcl $0,r3'
> {standard input}:6421: Error: bad instruction `movb $0,(r3)'
> {standard input}:6481: Error: bad instruction `btsl $0,[r9]'
> {standard input}:6482: Error: bad instruction `adcl $0,r3'
> {standard input}:6582: Error: bad instruction `movb $0,(r3)'
> {standard input}:7101: Error: bad instruction `btsl $0,[r1]'
> {standard input}:7102: Error: bad instruction `adcl $0,r3'
> {standard input}:7232: Error: bad instruction `movb $0,(r3)'
> {standard input}:7394: Error: bad instruction `btsl $0,[r2]'
> {standard input}:7395: Error: bad instruction `adcl $0,r1'
> {standard input}:7527: Error: bad instruction `movb $0,(r3)'
> {standard input}:7803: Error: bad instruction `btsl $0,[r6]'
> {standard input}:7804: Error: bad instruction `adcl $0,r3'
> {standard input}:7912: Error: bad instruction `movb $0,(r3)'
> {standard input}:8809: Error: bad instruction `btsl $0,[r4]'
> {standard input}:8810: Error: bad instruction `adcl $0,r3'
> {standard input}:8852: Error: bad instruction `movb $0,(r1)'
> {standard input}:8883: Error: bad instruction `btsl $0,[r5]'
> {standard input}:8884: Error: bad instruction `adcl $0,r3'
> {standard input}:8930: Error: bad instruction `movb $0,(r3)'
> {standard input}:8980: Error: bad instruction `btsl $0,[r10]'
> {standard input}:8981: Error: bad instruction `adcl $0,r3'
> {standard input}:9020: Error: bad instruction `movb $0,(r0)'
> make: *** [blacklists.o] Error 1
>
>
>
>
>
> Thank you.
>
> Best Regards,
>
> Caio Pereira
>
>
> 2015-10-20 18:28 GMT-02:00 Caio <caiortp at gmail.com>:
>
>> Ovidiu Sas,
>>
>>
>> In the Makefile.defs there a code to get the ARCH
>>
>> 90 ARCH := $(shell $(GETARCH) |sed -e s/i.86/i386/ -e s/sun4u/sparc64/
>> \
>> 91 -e s/armv4l/arm/ -e "s/Power Macintosh/ppc/"
>> \
>> 92 -e s/sun4v/sparc64/ -e "s/cobalt/mips2/" \
>> 93 -e s/amd64/x86_64/ )
>>
>>
>> I tried export ARCH=armv7 but because of the code, the ARCH will be
>> always x86-64 (host machine) .
>>
>> I added the CFLAGS to the armv7 arch in the Makefile.defs
>>
>> 918 ifeq ($(ARCH), armv7)
>> 919 # if gcc
>> 920 ifeq ($(CC_NAME), gcc)
>> 921 #common stuff
>> 922 CFLAGS+=-O0 -mfloat-abi=hard
>> -march=armv7-a -mthumb-interwork -Wcast-align $(PROFILE) \
>> 923 -mfloat-abi=hard -mfpu=neon
>> -mtune=cortex-a9 -Wall
>> 924 endif
>> 925 endif
>>
>>
>> Which is the best way to pass the arch parameter to the Makefile?
>>
>> (sorry for the typo in the last email).
>>
>> Thank you very much.
>>
>> Best Regards,
>>
>> Caio Pereira
>>
>> 2015-10-20 17:17 GMT-02:00 Ovidiu Sas <osas at voipembedded.com>:
>>
>>> The optimizations are done via the compiler flags: CFLAGS.
>>> You need to replace -09 with -00.
>>> You need to check if you are using a native compiler or a cross one
>>> and adjust the proper flags inside Makefile.defs.
>>>
>>> Regards,
>>> Ovidiu Sas
>>>
>>> On Tue, Oct 20, 2015 at 1:53 PM, Caio <caiortp at gmail.com> wrote:
>>> > Hello, Oividiu!
>>> >
>>> > Sorry for the delay, I was on vacation.
>>> >
>>> > I'm writing a recipe to Yocto.
>>> > This ARM cpu is new (armv7-a), but I see that is not compatible with
>>> > opensips's optimization.
>>> > I'm new with opensips, I have only tested in my host machine x86-64.
>>> >
>>> >
>>> > Do you know with parameters I need to comment in the Makefile.conf to
>>> have a
>>> > basic compilation without any optimization in the opensips?
>>> >
>>> >
>>> > #DEFS+= -DPKG_MALLOC #Uses a faster malloc (exclusive w/ USE_SHM_MEM)
>>> > #DEFS+= -DSHM_MMAP #Use mmap instead of SYSV shared memory
>>> > DEFS+= -DUSE_MCAST #Compile in support for IP Multicast
>>> > DEFS+= -DDISABLE_NAGLE #Disabled the TCP NAgle Algorithm ( lower delay
>>> )
>>> > DEFS+= -DSTATISTICS #Enables the statistics manager
>>> > DEFS+= -DHAVE_RESOLV_RES #Support for changing some of the resolver
>>> > parameters
>>> > #DEFS+= -DHP_MALLOC #High performance allocator with fine-grained
>>> locking
>>> > DEFS+= -DF_MALLOC #An even faster allocator. Not recommended for
>>> debugging
>>> > #DEFS+= -DF_MALLOC_OPTIMIZATIONS #Remove all internal checks in
>>> F_MALLOC
>>> > #DEFS+= -DDBG_QM_MALLOC #Allocator used for debugging information
>>> > #DEFS+= -DUSE_SHM_MEM #All PKG allocations are mapped to SHM (
>>> exclusive w/
>>> > PKG_MALLOC )
>>> > #DEFS+= -DDBG_F_MALLOC #TODO ?
>>> > #DEFS+= -DNO_DEBUG #Turns off all debug messages
>>> > #DEFS+= -DNO_LOG #Completely turns off all the logging
>>> > #DEFS+= -DVQ_MALLOC #TODO ?
>>> > #DEFS+= -DFAST_LOCK #Uses fast architecture specific locking
>>> > #DEFS+= -DUSE_FUTEX #Uses linux futexs with fast architecture specific
>>> > locking
>>> > #DEFS+= -DUSE_SYSV_SEM #Uses SYSV sems for locking ( slower & limited
>>> number
>>> > of locks
>>> > #DEFS+= -DUSE_PTHREAD_MUTEX #Uses pthread mutexes
>>> > #DEFS+= -DBUSY_WAIT #Uses busy waiting on the lock
>>> > #DEFS+= -DDBG_LOCK #TODO ?
>>> >
>>> >
>>> >
>>> > Thank you!
>>> >
>>> > Best Regards,
>>> >
>>> > Caio Pereira
>>> >
>>> >
>>> > 2015-09-23 21:48 GMT-03:00 Ovidiu Sas <osas at voipembedded.com>:
>>> >>
>>> >> Hello Caio,
>>> >>
>>> >> Have you managed to cross compile opensips?
>>> >> The optimizations do not work for older type of arm CPUs.
>>> >> Did you experienced any issues running opensips on arm?
>>> >>
>>> >> Regards,
>>> >> Ovidiu Sas
>>> >>
>>> >> On Thu, Sep 17, 2015 at 2:51 PM, Caio <caiortp at gmail.com> wrote:
>>> >> > Hello,
>>> >> >
>>> >> > I would like to cross compile the opensips to ARMv7 architecture
>>> >> > (Freescale
>>> >> > IMX6).
>>> >> > I saw that there's some optimization in assembly to lock functions.
>>> >> > There's some configuration in the Makeconfig.conf options that I
>>> need
>>> >> > disable or enable to able the opensips run in my target
>>> architecutre?
>>> >> >
>>> >> > Thank you very much.
>>> >> >
>>> >> > Best Regards
>>> >> >
>>> >> >
>>> >> >
>>> >> > --
>>> >> > ----------------------------------------------
>>> >> > Caio Pereira
>>> >> >
>>> >> > _______________________________________________
>>> >> > Users mailing list
>>> >> > Users at lists.opensips.org
>>> >> > http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>>> >> >
>>> >>
>>> >>
>>> >>
>>> >> --
>>> >> VoIP Embedded, Inc.
>>> >> http://www.voipembedded.com
>>> >>
>>> >> _______________________________________________
>>> >> Users mailing list
>>> >> Users at lists.opensips.org
>>> >> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>>> >
>>> >
>>> >
>>> >
>>> > --
>>> > ----------------------------------------------
>>> > Caio Pereira
>>> >
>>> > _______________________________________________
>>> > Users mailing list
>>> > Users at lists.opensips.org
>>> > http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>>> >
>>>
>>>
>>>
>>> --
>>> VoIP Embedded, Inc.
>>> http://www.voipembedded.com
>>>
>>> _______________________________________________
>>> Users mailing list
>>> Users at lists.opensips.org
>>> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>>>
>>
>>
>>
>> --
>> ----------------------------------------------
>> Caio Pereira
>>
>
>
>
> --
> ----------------------------------------------
> Caio Pereira
>
--
----------------------------------------------
Caio Pereira
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensips.org/pipermail/users/attachments/20151021/9d746ecc/attachment-0001.htm>
More information about the Users
mailing list