File transfer (SMB or ftp) frozen during transfer

[GoogleCodeExporter]

What steps will reproduce the problem?
1. Simply upload or download to NAS files no too large around (700MB)
2. after certain number of files, the transfer breaks and application (SMB
or ftp) seems frozen
3. restart of the application will reset and restart the transfer

What is the expected output? What do you see instead?
* download and upload files via ftp and SMB.

What version of the product are you using? On what operating system?
SNAKE OS V1.0.0 (20091027)/2.6.16-gazineu. Windows XP prf edition.

Please provide any additional information below.
-No error logs are found in Nas logs.
-Sometimes the transfer recovers itself
-RAM usage is always over 90%

Original issue reported on code.google.com by Angel.M....@gmail.com on 13 Apr 2010 at 9:51

[GoogleCodeExporter]

Hello, I have the same problem, Windows 7 Pro installed, and...
I was experienced too it can't erase directories if they are not empty... Only 
can
erase them if first erase the archives containing one by one... Any solution?

Original comment by jasv6374...@gmail.com on 16 Apr 2010 at 8:51

[GoogleCodeExporter]

I have problems too with read and write on NTFS hard disc attached to NAS, it
contents "dissapears" and transfers and Explorer (Windows) was locked at few 
seconds
of start transfers.
I was tested this issues with 2 different routers with same results...
Anybody has the solution?

Original comment by jasv6374...@gmail.com on 16 Apr 2010 at 9:39

[GoogleCodeExporter]

Bad to know I am not the only one with the isse. If it was only me, I might be 
doing
a wrong setting.
In case it is of help to create a workaround/fix: I noticed that every time I 
find
the problem, the CPU of the nas is 100%. Normally, when every thing is fine, 
CPU is
around 50%~60%. 
When the CPU recovers normal values, and if the aplication did not go in 
timeout, the
session/transfer recovers itself. 
I hope we get it fixed. 

Original comment by Angel.M....@gmail.com on 17 Apr 2010 at 1:13

[GoogleCodeExporter]

Ok.. i had the same problem but i think i resolved it, or almost resolved it. 
By 
using WinSCP, accessing thru SSH with your root account, you will see in your 
root 
folder a folder named USB that contain the usb devices. You you notice the 
permissions of that folder you can see that the T option is active ( right 
click the 
folder and select Properties. You will notice that the STICKY BIT option is 
active. 
I don´t know nothing about linux, but for what i could read online on some 
forums, 
this options doesn´t allow user Y to delete a file owned by user X. As all of 
my 
files are owned by root, i could not delete some files and folders. After 
making USB 
and all files and folders inside 0777 instead of 1755 ( i think this was the 
initial 
configuration ) i can now delete files and folders with my user account, 
despite all 
files being marked as owned by root. sometimes i get a message that the file is 
being used by some process, but those errors telling me that i needed root 
permission to delete some file or folders, are gone for now. I have all my 
devices 
formated as EXT3.

ALL OUR FILES BELONG TO US !!! :P

Original comment by ulisses....@gmail.com on 18 Apr 2010 at 10:41

[GoogleCodeExporter]

Hi friends, I was made some tests and this problem persists with the new SnakeOS
1.1.0 firmware too...
File transfers freezes with disc formated in EXT3, then I was tested with FAT32 
with
same bad results, so I think the HD format isn't the question.
I was read anywhere this problem is inherent to Samba security (Windows 7 and 
Vista,
but with others too) but I don´t know how to solve this issue.
And can't erase directly any folder if there is any archive in... And some files
erased reappears again!!!
With FTP clients occurs identical fails in LAN, I was tested several like 
Cuteftp
(the best I think), Filezilla, Fireftp...
Is possible to correct the permissions in Samba to permits to any LAN user to
read/write OK?

Original comment by jasv6374...@gmail.com on 19 Apr 2010 at 10:44

[GoogleCodeExporter]

Try to use next settings for Samba daemon in /etc/smb.conf:
socket options = SO_KEEPALIVE TCP_NODELAY IPTOS_LOWDELAY SO_RCVBUF=65536 
SO_SNDBUF=65536 
log level = 1
read raw = yes
write raw = yes
oplocks = yes
max xmit = 65535
dead time = 15
getwd cache = yes

Original comment by pavel.ma...@gmail.com on 19 Apr 2010 at 11:36

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

Hello all, by chance, are all those with this problem using wireless? 
Encrypted? I
did, and to try I changed to wired. The problem (frozen transfers) disappeared. 
Now I am more convinced than ever it is ... windows... as usual.
I noticed that every minute there are short interruption in the transfers (and
always). The frozen cases seems to correspond with one of those narrow 
interruptions
no recovered. See attached pic, every interrupt corresponds to a minute. I 
found in
internet reference to wireless card with side effects because of those 
interruptions
and a comment about the Windows "Wireless Zero Configuration" (WZC) 
-unbelievable-. I
have stopped the service and *by now* is working. It seems that WZC is useless 
in any
case. 
I also changed the smb.conf file as Pavel indicates. Probably are good
recommendations improving smb sessions, but, Pavel, do you think they are the 
good
ones for wireless?
If someone has time, try: a-wired; b- wireless and stop WZC (admin 
tools->services->
Wzc and let me know your results.

Most of us uses wireless. Maybe Samba might come already with defaults values in
smb.conf optimized for wireless.

A bit long but I hope that helps. Angel

Original comment by Angel.M....@gmail.com on 24 Apr 2010 at 12:12

Attachments:

• Dibujo.JPG

[GoogleCodeExporter]

disregard the solution suggested in my previous mail. I still have the problem 
while
using wireless. However, it is not found using wired. Maybe it is still of help.
Angel

Original comment by Angel.M....@gmail.com on 24 Apr 2010 at 10:42

[GoogleCodeExporter]

The wireless packet losses may be the culprit.
There may be two ways of get these problems sorted out, to upgrade samba or 
change
kernel tcp settings for better handling of it.
I could not reproduce this issue, will let you know to help in testing.

Original comment by dgazi...@gmail.com on 25 Apr 2010 at 12:50

• Changed state: Started
• Added labels: Priority-High
• Removed labels: Priority-Medium

[GoogleCodeExporter]

Hi Duglas, I captured several traces with wireshark in case you want to have a 
look
(one of with 2 recovers and 3 interruptions). Last one is too heavy to attach to
anything. If needed I better filter it out removing the central transfer data. I
attache a small one by now. As you guessed, it looks related with lost frames 
not
retransmitted by the server.
I am available for testing as needed. 
Note: if needed, I may load V1.1 or other debugging one.
Angel   

Original comment by Angel.M....@gmail.com on 26 Apr 2010 at 9:38

Attachments:

• [Capture SMB with error 2400 on.pcap](https://storage.googleapis.com/google-code-attachments/snake-os/issue-11/comment-11/Capture SMB with error 2400 on.pcap)
• [Capture SMB with error 9600on.pcap](https://storage.googleapis.com/google-code-attachments/snake-os/issue-11/comment-11/Capture SMB with error 9600on.pcap)

[GoogleCodeExporter]

Hi,

I have the same issue as well.  Running Windows XP Pro SP3, hard wired to 
router.  Can 
provide further details if required; just let me know what you need

Thanks

Rob

Original comment by r.edwa...@hotmail.co.uk on 9 May 2010 at 9:58

[GoogleCodeExporter]

I Have the same problem. I turn off the "ntp client" on the "Snake web admin" 
and
never frozen again (with my latop via wireless).
Any comments?

NAhuel

Original comment by nahuel.r...@gmail.com on 14 May 2010 at 2:31

[GoogleCodeExporter]

I'm a Mac user and had these problems. I have managed to drastically reduce the 
number of errors by adding 
this to /etc/smb.conf. I don't know if this will reduce problems for Windows 
users.

in the [global] section

add: unix extensions = no

then restart samba by using "/etc/init.d/samba restart"

Credit - http://www.macosxhints.com/article.php?story=20100405023255445

-Dan

Original comment by dantu...@gmail.com on 21 May 2010 at 10:48

[GoogleCodeExporter]

Hello all, I am close to give up :-(

* I try all your smb.conf suggestions. The issue is still there. It makes 
sences it
is not a smb issue, it happens as well with ftp.  
* I set a log file via smb.conf (log file = /usb/500_root/download/log.%m) and 
the
following errors are shown:
[2010/05/21 23:24:37, 0] lib/util_sock.c:write_socket(566)
  write_socket: Error writing 61503 bytes to socket 7: ERRNO = Connection reset by peer
[2010/05/23 22:33:43, 0] lib/util_sock.c:write_socket_data(540)
  write_socket_data: write failure. Error = Connection timed out
[2010/05/23 22:33:48, 0] lib/util_sock.c:write_socket(566)
  write_socket: Error writing 32831 bytes to socket 7: ERRNO = Connection timed out

* btw, in the logs I found that entry "unix extensions" is not recognized.

I red in some samba old blogs that this might be caused by miss configuration 
in the
ethernet, and something might be related but maybe is the TCP implemenaton no 
enough
robust in the NAS. The reason i think so:
1- I tested 2 routers (wifi): same effect
2- with different PCs same behaviour.
3- removed all wifi encryption: same behaviour.
4- I connect with a fast ethernet (as mentioned before), and the problem 
disappears.

Wifi is the scenario presenting always the problem. But wifi plus other FTP at 
the
internet does not present the issue. Only with the NAS seems to be happening. 
Might
it be the TCP implementation in the NAS not able to recover the frequent frame 
lost
found in the Wifi?

Any hint to debug TCP errors at the NAS side? 
Thank you in advance and Cheers,
Angel
PS: By the way, with FTP, max bit rate I get is 2,60Mbps and using samba, 
9,60mbps. I
do not have a reason to explain that difference. I would expect similar 
throughput.

Original comment by Angel.M....@gmail.com on 30 May 2010 at 9:49

[GoogleCodeExporter]

Same problem here. I have tested this for one day with large files files and
different hdds, all with NTFS. The transfer was freezing regularly. I tested 
with 2
different access points (linksys WRT54G3G and tp-link TL-WA501G) - no 
difference.
With a wired connection I was not able to reproduce the problem.

I decided to check whether is is a packet-loss issue, so I tested with ping -t 
-l
65500 running at the same time. And while the ping was running I was not able to
reproduce the error. Then I found out, then if I'm able to catch the moment the
transfer freezes and send several ping commands before the error is reported 
then the
transfer was recovering. Seems like a concurrency bug in the software. Good 
that it
is reproducable.

Original comment by agsa...@gmail.com on 5 Jun 2010 at 1:26

[GoogleCodeExporter]

I've double checked my comment and one of the disks was FAT32. Same problem 
there.

Original comment by agsa...@gmail.com on 5 Jun 2010 at 1:34

[GoogleCodeExporter]

HI all,
Yesterday I made new tests transferring files from my computer, installed with 
Windows 7 32 bit and 64 respectively and within my own network and local 
network to my NAS WLX-652 with the latest firmware installed "-V1.2.0-snakeos 
20100621-from-original.zip "installed, and the results were as follows:
- Both ethernet and wireless connection, you can access the shared folder and 
download files from the NAS to my PC as an anonymous user equipment.
- Both ethernet and wireless connection is accessible but not send files from 
my computers to the NAS, nothing happens and you receive a message that is not 
the path to the NAS.
- Using an FTP client to send files to a user logged into my NAS with write 
access from my computers to the NAS, is immediately a report that says that 
writing rights are denied by the NAS.
The shared folder on my NAS is configured to access any user with write access, 
and configuration of Samba access is anonymous or registered. The FTP server 
configuration is active for the same shared folder and turned off anonymous 
access.
That is, in any case I can access my shared files on my NAS to view and 
download to my computer either by ethernet or wireless, but in no case Samba 
allows writing.
Will this have any solution this in the future?

Original comment by jasv6374...@gmail.com on 6 Jul 2010 at 6:45

[GoogleCodeExporter]

Hello, I do not see why does not allow you to modify them. I have never had 
such problem. I assume you have created your partitions with writable to "yes". 
You can check it as well under /etc/smb.conf. Each particion section has the 
writeable = yes. 
If it does not solve it, try "properties" once mounted. There is a folder 
"permissions". it should be also changed there. I will try now version and 
report any change, but I suspect that it is not prepared the fix of the 
downaload problesm.

Regards, 

Original comment by Angel.M....@gmail.com on 8 Jul 2010 at 7:24

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

I can confirm that it isn't a Windows only problem. It happens too with Linux 
(over wireless 802.11n). I've made a packet capture over windows and I can see 
that when the problem happens a duplicate frame is sent by the NAS, without any 
request from the client. After that, several duplicated packets are sent. That 
seems to happen in large delays. On short delays (less than 1 sec), what seems 
to happen is that is a delay between the smb "chunk" request and the response, 
so the request must be retransmited.
All these symptoms almost disapear using settings on comment #6 (Pavel). 
I'll keep testing and inform if I find out anything.

Regards,

Original comment by mdb...@gmail.com on 9 Jul 2010 at 2:09

[GoogleCodeExporter]

I use Windows 7 32 bit and can't copy LARGE files with wireless interface with 
FTP or SMB.  With wired interface works fine.  As workaround, I use robocopy  
to copy large files.  May be the problem is the tcp sliding window resize used 
for increase performance (Microsoft has a KBQ about it).

Original comment by bala...@gmail.com on 16 Jul 2010 at 12:09

[GoogleCodeExporter]

I can confirm same issue from Linux to NAS via SMB: file transfers fails using 
wireless.
NAS Model: 650 BT
I also tried workaround on comment #6 with no success.
Please consider my NAS available for any test.
Thanks.

Original comment by enrico.b...@gmail.com on 17 Aug 2010 at 10:12

[GoogleCodeExporter]

The problem happens to me as well. Only when I'm connected through WiFi.
Everything works great with Ethernet connection.
Ideas?

Original comment by magn...@gmail.com on 3 Oct 2010 at 3:33

[GoogleCodeExporter]

I have the freeze also, but in a different way.
I use the browse folder functionality from 
http://code.google.com/p/snake-os/issues/detail?id=50
So via the webserver on the NAS, I browse an external HD connected to the NAS 
via the USB port (directly).
And here also I have the problem that sometimes, while a directory listing is 
given, that it freezes for about half a minute.
The webserver is running on the NAS and via the NAS (unix ls command) a 
directory listing is made of the connected HD (formatted FAT 32) and then via a 
web page shown in the explorer of the connected computer. SAMBA or FTP are not 
used in this situation I guess.
So the problem seems to be really in the kernel of the NAS, is it not?
I have a second device and it has exactly the same problem.
The NAS device is a WLX-652

Can someone explain what is mend by accessing the NAS wireless (WiFi) or wired 
and why this would influence the problem?

Peter

Original comment by pen...@gmail.com on 6 Oct 2010 at 8:44

[GoogleCodeExporter]

Note that the change from comment 6 from 
http://code.google.com/p/snake-os/issues/detail?id=50 has as result that this 
problem does not occur that often anymore. So if you want to simulate this then 
use the older files before comment 6. It seems if the filesystem is accessed 
quickly that it does not occur... 

Original comment by pen...@gmail.com on 6 Oct 2010 at 8:53

[GoogleCodeExporter]

I have the same problem. 
I’ve tried all the snake OS firmware versions as well as the original and the 
updated original. 
I’ve tried EXT 2, 3, NTFS and FAT32.
I’ve tried it wirelessly on 2 different NETGEAR routers, one with N and the 
other through G with WEP and WPA2 security.
I’ve tried transferring files to a Linux and several windows 7 computers.
Nothing seems to change it. While watching a movie it will drop out but if left 
to play will come back (depending on the player).
Is there any more information I could possible give to help the problem.

Joel

Original comment by joelecl...@gmail.com on 12 Oct 2010 at 7:16

[GoogleCodeExporter]

Hi Peno64, simply a note, I can not assure how often it happens with the 
parameters as indicated in comment 6, but it still happens. Also, usability is 
very low if you have to access the filesystem quickly. I undertand that the 
large number people impacted seems to show is not a rare issue. I also 
understand that, if parameters suggested in comment 6 improve something the 
issue, they should be added to the default files within the new version. I have 
the experience that the problem is not solved with them, but if for instance, 
it improve behaviour for some people, they could be added. At least, it seems 
they do not cause more problems.
As Joel, I think some more help is needed here, wifi is largely used and 
software would really improve usability with this fixed.

Although, with some limitation on time now, I am available (and willing) to run 
tests to help. Unfortunately, my router and NAS are offline. Only accessible 
connecting directly (cable or wifi) to them. 
regards,
angel 

Original comment by Angel.M....@gmail.com on 14 Oct 2010 at 9:53

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

Hi there,

I was seeing these mysterious TCP-related freezes with my NS-K330, except that 
in my case they were also happening in the ssh shell. I noticed that sometimes 
it would freeze inexplicably when I was just typing something. During these 
events the NAS wasn't doing anything, virtually at rest.

It seems that the problem was solved after I disabled the NTP client as nahuel 
suggested. The simple fact that disabling that service affects the problem 
means that we are on to something interesting. The NTP client makes little 
adjustments to the clock once in a while to keep it in sync with a remote 
clock. These changes may be adversely affecting the timing in the TCP stack, 
which causes the problem.

Is there any kernel hacker out there?

At least it seems that a simple workaround is to just disable the NTP client.

Euler

Original comment by eulerdgr...@gmail.com on 26 Oct 2010 at 11:59

[GoogleCodeExporter]

Have the same problem on a cheap LAN-NAS device. using telnet I have to type 
slowly :-P .. otherwise it hangs

Original comment by tuxbox.g...@gmail.com on 27 Oct 2010 at 7:27

[GoogleCodeExporter]

Euler,
      It seems you're right. I disabled NTP and is working fine now. But, analyzing the configuration I've found that the NTP client isn't running as a deamon. Instead, is called by cron every hour (at xx:11). So, I still don't know why it's working now.

Regards,
Marcelo

Original comment by drno...@gmail.com on 27 Oct 2010 at 8:23

[GoogleCodeExporter]

[CODE]
This is the NTP script

#!/bin/sh
# SNAKE OS - 2009-06-27
# Called by rc.sysinit, udhcpc and crond (periodically)
# 
export PATH=/bin:/sbin:/usr/bin:/usr/sbin                                       

NTP_ENABLE=$(grep "ntp_enable=" /etc/default/config | cut -d = -f 2)
if [ $NTP_ENABLE -eq 1 ] ; then
    NTP_SERVER=$(grep "ntp_server=" /etc/default/config  | cut -d = -f 2)
    ntpclient -h $NTP_SERVER -s -i 2 -c 1 > /dev/null &
fi
[/CODE]

the script looks for ntp_enable flags and does what it needs to if found

Original comment by tuxbox.g...@gmail.com on 27 Oct 2010 at 9:07

[GoogleCodeExporter]

I tried disabling NTP with no success: the copy still hangs after few mbytes if 
connected via wireless. (both with and without comment 6 samba modification)

Original comment by enrico.b...@gmail.com on 27 Oct 2010 at 9:47

[GoogleCodeExporter]

It seems that messing with tcp settings helps a lot (but it isn't perfect)
Could you confirm that doing 

echo 1 > /proc/sys/net/ipv44/tcp_no_metrics_save 

really helps?. I don't know a way of getting objective data so I could be wrong.

Original comment by mdb...@gmail.com on 8 Nov 2010 at 12:29

[GoogleCodeExporter]

Hi mdbovo,
I have modified the file tcp_no_metrics_save as suggested.
No changes occured: file transfers still freeze.

Thanks for the support.

Original comment by enrico.b...@gmail.com on 8 Nov 2010 at 5:52

[GoogleCodeExporter]

Hello,

I always assumed that this was a hardware issue as the origonal non snake 
firmware had the same issue.

Is there any word from the devloper as to this being fixed in v1.3? (if there 
will be one)?

Apart from this really annoying issue I think the code is greag - well done and 
thanks :) :)

Original comment by osgood...@gmail.com on 12 Nov 2010 at 2:06

[GoogleCodeExporter]

I have the same problems of all IP connections stalling for 25 seconds 
randomly. I added USB->Ethernet .KO drivers and testing is continuing. 

Original comment by tuxbox.g...@gmail.com on 13 Nov 2010 at 10:19

[GoogleCodeExporter]

CONFIRMED:

USB->ETHERNET works 100% without stalling.... The STARNic driver in the kernel 
seems to be really really really buggy.

Now to figure out how to autoload the USB->Netowork drivers on boot.

Original comment by tuxbox.g...@gmail.com on 13 Nov 2010 at 10:31

[GoogleCodeExporter]

tuxbox.guru: Just to confirm. You're adding a USB to Ethernet adapter and using 
that as your network interface instead of the built in network port, or are you 
replacing the drivers for the built in network adapter?

Original comment by static.o...@gmail.com on 13 Nov 2010 at 11:25

[GoogleCodeExporter]

I dont replace them at all... I just dont use them. I boot the unit using the 
normal StarNIC interface.. then insmod the ASIX.KO and usbnet.ko .. bring up 
the interface and then telnet to the new interface .. then ifconfig eth0 down

I could build in the USBNET drivers in to the kernel and totally disable 
starNic (I think that is what you were thinking of) I think there is a kernel 
config for the StarNIC - but I was too scared to disable it (I didnt want to 
create a brick)

Original comment by tuxbox.g...@gmail.com on 13 Nov 2010 at 2:10

[GoogleCodeExporter]

Hi Tuxbox.guru

Can you provide simple step by step instruction how to change drivers to 
UBS->Ehternet.

Thanks!

Original comment by aky...@gmail.com on 18 Nov 2010 at 3:40

[GoogleCodeExporter]

If you have the snakeOS sdk, change the following lines in the SDK/config files 
(both) - then rebuild

This will build in the USB Ethernet.. (here are the configs I used)

#
# USB Network Adapters
#
# CONFIG_USB_CATC is not set
# CONFIG_USB_KAWETH is not set
# CONFIG_USB_PEGASUS is not set
CONFIG_USB_RTL8150=y
CONFIG_USB_USBNET=y
CONFIG_USB_NET_AX8817X=y
CONFIG_USB_NET_CDCETHER=y
# CONFIG_USB_NET_GL620A is not set
# CONFIG_USB_NET_NET1080 is not set
CONFIG_USB_NET_PLUSB=y
# CONFIG_USB_NET_RNDIS_HOST is not set
# CONFIG_USB_NET_CDC_SUBSET is not set
# CONFIG_USB_NET_ZAURUS is not set
# CONFIG_USB_MON is not set

#
# USB port drivers

Original comment by tuxbox.g...@gmail.com on 20 Nov 2010 at 8:55

Attachments:

• str8132_defconfig_ramimage
• str8132_defconfig_jffs2

[GoogleCodeExporter]

WARNING!!!!!

The above configs WILL disable the built in Ethernet Port on the device.... 
(Just use the section that is in the Issue... and the config files are there 
for reference ONLY)

Original comment by tuxbox.g...@gmail.com on 20 Nov 2010 at 8:56

[GoogleCodeExporter]

Could you attach the 2 new modules? 

Original comment by mdb...@gmail.com on 21 Nov 2010 at 11:40

[GoogleCodeExporter]

Here you go...

usbnet.ko is compulsory... the rest are for different devices.

Original comment by tuxbox.g...@gmail.com on 21 Nov 2010 at 12:52

Attachments:

• USBEther.tar.gz

[GoogleCodeExporter]

Hi all!!

@tuxbox.guru
First of all to say you're doing a great job, hacking a lot of things :-)
Thats why i ask you for a step-by-step commands to bring the new eth interface 
up, without any loss of connection to the NAS, e.g:

# insmod /usb/sda1/usbnet.ko  ???
# ifconfig eth1 192.168.0.2 netmask 255.255.255.0 up ???
# ... ???

Many many thanx in advance.

Regards.

David.

Original comment by DavidFL....@gmail.com on 30 Nov 2010 at 9:25

[GoogleCodeExporter]

Hi David,

you are almost there :D 

#insmod usbnet.ko
#insmod <your ethernet device driver>.ko
#ifconfig eth1 192.168.1.33 netmask 255.255.255.0 up

done...

the <your ethernet device driver> in my case was asix.ko. to verify device is 
loaded, type in dmesg and look at the last entries.

You might lose the connection to the device but just re-telnet in using the new 
number.

Original comment by tuxbox.g...@gmail.com on 30 Nov 2010 at 12:24

[GoogleCodeExporter]

What model of NAS are you using tuxbox.guru?
I can't see anything after insert module usbnet. I have a NSK-330

Thanks

Original comment by mdb...@gmail.com on 1 Dec 2010 at 12:47