btsync 1.1.69 ARM Alignment error

[THJ]

While version 1.1.48 workes without any problems on my ARM NAS device, 1.1.69 doesn't want to start and exists with Alignment error.

$ ./btsync
Alignment error

Here are last few lines of strace if it helps to debug the error:

getrlimit(RLIMIT_STACK, {rlim_cur=8192*1024, rlim_max=RLIM_INFINITY}) = 0
brk(0)								 = 0x349000
brk(0x36a000)						 = 0x36a000
gettimeofday({1376379137, 723324}, NULL) = 0
mkdir("/tmp", 0777)					 = -1 EEXIST (File exists)
mkdir("/tmp/btsync_dumps", 0777)	 = -1 EEXIST (File exists)
gettimeofday({1376379137, 727732}, NULL) = 0
futex(0x3361e8, FUTEX_WAKE, 2147483647) = 0
sigaltstack(NULL, {ss_sp=0, ss_flags=SS_DISABLE, ss_size=0}) = 0
sigaltstack({ss_sp=0x3499e0, ss_flags=0, ss_size=8192}, NULL) = 0
rt_sigaction(SIGSEGV, NULL, {SIG_DFL, [], 0}, 8) = 0
rt_sigaction(SIGABRT, NULL, {SIG_DFL, [], 0}, 8) = 0
rt_sigaction(SIGFPE, NULL, {SIG_DFL, [], 0}, 8) = 0
rt_sigaction(SIGILL, NULL, {SIG_DFL, [], 0}, 8) = 0
rt_sigaction(SIGBUS, NULL, {SIG_DFL, [], 0}, 8) = 0
rt_sigaction(SIGSEGV, {0x14af0, [ILL ABRT BUS FPE SEGV], SA_STACK|SA_SIGINFO|0x4000000}, NULL, 8) = 0
rt_sigaction(SIGABRT, {0x14af0, [ILL ABRT BUS FPE SEGV], SA_STACK|SA_SIGINFO|0x4000000}, NULL, 8) = 0
rt_sigaction(SIGFPE, {0x14af0, [ILL ABRT BUS FPE SEGV], SA_STACK|SA_SIGINFO|0x4000000}, NULL, 8) = 0
rt_sigaction(SIGILL, {0x14af0, [ILL ABRT BUS FPE SEGV], SA_STACK|SA_SIGINFO|0x4000000}, NULL, 8) = 0
rt_sigaction(SIGBUS, {0x14af0, [ILL ABRT BUS FPE SEGV], SA_STACK|SA_SIGINFO|0x4000000}, NULL, 8) = 0
fstat64(1, {st_mode=S_IFCHR|0620, st_rdev=makedev(136, 1), ...}) = 0
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x4001f000
write(1, "Alignment error\n", 16Alignment error
)	 = 16
exit_group(1)

LDD output looks ok:

$ ldd btsync
	    librt.so.1 => /lib/librt.so.1 (0x4002d000)
	    libdl.so.2 => /lib/libdl.so.2 (0x4003c000)
	    libm.so.6 => /lib/libm.so.6 (0x40047000)
	    libpthread.so.0 => /lib/libpthread.so.0 (0x400f2000)
	    libc.so.6 => /lib/libc.so.6 (0x40110000)
	    /lib/ld-linux.so.3 (0x40000000)

Any clue what changed that is preventing the new version to start?

Thanks.

[kos13]

There are some alignment on ARM that Sync is having troubles with. Basically you need to execute

echo 2 > /proc/cpu/alignment

So Sync will work.

[THJ]

There are some alignment on ARM that Sync is having troubles with. Basically you need to execute

echo 2 > /proc/cpu/alignment

So Sync will work.

Perfect solution.

$ echo 2 > /proc/cpu/alignment
$ ./btsync --help
BitTorrent Sync 1.1.69

Thank you.

[THJ]

Hello again.

This solution seems to be only valid for application to start. But after about 10 seconds when it starts the whole NAS will freeze and I have to reboot it. This is my btsync.conf:

{
  "device_name": "eTRAYz NAS",
  "listening_port" : 0,
  "storage_path" : "/home/epkg/share/btsync/.sync",
  "pid_file" : "/home/epkg/share/btsync/.sync/btsync.pid",
  "check_for_updates" : true,
  "use_upnp" : true,
  "download_limit" : 0,
  "upload_limit" : 0,
  "webui" :
  {
    "listen" : "0.0.0.0:7777",
    "login" : "admin",
    "password" : "sysadmin"
  },
  "shared_folders": []
}

And this is how I start the process:

/home/epkg/share/btsync/btsync --config /home/epkg/share/btsync/.sync/sync.conf

I've tried to change the value in /proc/cpu/alignment to 3, 4 or 5 and with all the NAS freezes as soon as btsync process starts.

Any other suggestion?

[dirkschwarz]

Hello,

same problem -> Zyxel NSA210

Any solution? Thanks

[THJ]

Same problem with 1.1.70.

[THJ]

Any other solution that wouldn't crash entire NAS?

[THJ]

Or at least explanation if this will ever get fixed?

[kos13]

The problem with alignment happens only on old ARM CPU types. The echo 2 command, force CPU to overcome alignment errors by means of emulation, so there will be significant performance degradation for the NAS.

Yes it will be fixed, but I can't give you a timeframe for the fix.

[demus]

Could you elaborate on what "old" means in ths case?

I am trying to run btsync on a SheevaPlug, where it dumps core after transfering for a short time.

[draggos]

So, on this same ARM CPU (old type), the 1.1.48 works perfectly fine - no alignment issues. The new version, the 1.1.70, needs a fix by means of emulation to overcome alignment errors that were not present in the previous version ?! Something doesn't sound right here..

[laurens]

Probably they optimized something, or added a new feature, which requires emulation to function properly on older ARM cpu's.

[hvtuananh]

Do you know how to downgrade to 1.1.48 version? I am using ArchLinux

[bittorrentomane]

I've had the same problem with the alignment, the posted solution worked for me.

No freezing yet on Pogoplug v2 Pink running 1.1.70 () on Debian Wheezy

[kbingham]

So, on this same ARM CPU (old type), the 1.1.48 works perfectly fine - no alignment issues.

I've got this issue too ... Where can we download the 1.1.48?

[laurens]

http://syncapp.bittorrent.com/1.1.48/

[THJ]

Just tested version 1.1.82 and this bug is still present. Any ETA when we can expect a solution?

 

Thank you.

[herzi]

I tested v1.2.68, it still yields alignment errors.

[GreatMarko]

I tested v1.2.68, it still yields alignment errors.

 

Even after you've executed as root "echo 2 > /proc/cpu/alignment" and run btsync?

[THJ]

Yes then it works. Device no longer freezes. But this could be solved from inside application and not by running a command prior to starting btsync.

[herzi]

Even after you've executed as root "echo 2 > /proc/cpu/alignment" and run btsync?

 

Are you familiar w/ GNU/Linux systems? Invoking that command should absolutely be _not_ necessary. Ever.

 

It also shouldn't be invoked from inside the application. The application should just work out of the box.

[GreatMarko]

Are you familiar w/ GNU/Linux systems? Invoking that command should absolutely be _not_ necessary. Ever.

 

It also shouldn't be invoked from inside the application. The application should just work out of the box.

 

I am merely quoting official advice from BitTorrent Inc on the matter!

[nwilks]

Can we get an update from the developers when this alignment issue will be fixed for those of us using the affected ARM systems? I really don't like how the performance is degraded on my machine after issuing the command, and it has been a really long time since you identified the issue.

[DePingus]

Is there a bugtracker where we can vote the issue up? I have 3 ARM devices I would like to reliably use this on.

[ikwyl6]

Yes, I would also like this fixed instead of doing the alignment fix.