[PC-BSD Testing] Dolphin bug

Kris Moore kris at pcbsd.com
Tue Jan 5 14:01:50 PST 2010


On 01/05/2010 11:39, doverosx at gmail.com wrote:
> Kris Moore wrote:
>> On 01/05/2010 11:06, doverosx at gmail.com wrote:
>>
>>> Kris Moore wrote:
>>>
>>>> On 01/05/2010 11:01, doverosx at gmail.com wrote:
>>>>
>>>>
>>>>> Kris Moore wrote:
>>>>>
>>>>>
>>>>>> On 01/05/2010 14:55, Kris Moore wrote:
>>>>>>
>>>>>>
>>>>>>
>>>>>>> On 01/05/2010 14:14, doverosx at gmail.com wrote:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>> ...continued from "Dictionary issues again?" which were caused by /tmp
>>>>>>>> being used improperly.
>>>>>>>>
>>>>>>>> Various dmesg, df -h and sysctl vm.vmtotal outputs:
>>>>>>>>
>>>>>>>> /********** DMESG *********/
>>>>>>>> kern.maxfiles limit exceeded by uid 1001, please see tuning(7).
>>>>>>>> kern.maxfiles limit exceeded by uid 1001, please see tuning(7).
>>>>>>>> kern.maxfiles limit exceeded by uid 1001, please see tuning(7).
>>>>>>>> kern.maxfiles limit exceeded by uid 1001, please see tuning(7).
>>>>>>>> kern.maxfiles limit exceeded by uid 1001, please see tuning(7).
>>>>>>>> pid 52607 (kdeinit4), uid 1001 inumber 70671 on /tmp: filesystem full
>>>>>>>> pid 44850 (firefox-bin), uid 1001 inumber 6 on /tmp: filesystem full
>>>>>>>> pid 44850 (firefox-bin), uid 1001 inumber 6 on /tmp: filesystem full
>>>>>>>> pid 28813 (npviewer.bin), uid 1001: exited on signal 11
>>>>>>>>
>>>>>>>> *This is from the dictionary issues dmesg output and not the SMB
>>>>>>>> specific usage of dmesg*
>>>>>>>>
>>>>>>>> *******************************************************************************
>>>>>>>> During 800MB transfer, Plasma space "could not read file"
>>>>>>>>
>>>>>>>> [brodey at pcbsd]/home/brodey(120)% ls -l /tmp
>>>>>>>> total 6
>>>>>>>> drwx------  2 brodey  wheel  1024 Jan  5 13:54 kde-brodey
>>>>>>>> drwx------  2 root    wheel   512 Jan  5 13:27 kde-root
>>>>>>>> drwx------  2 brodey  wheel   512 Jan  5 13:55 ksocket-brode
>>>>>>>> [brodey at pcbsd]/home/brodey(121)% ls -l /tmp/kde-brodey/
>>>>>>>> total 605538
>>>>>>>> -rw-------  1 brodey  wheel          0 Jan  5 13:42 dolphinAy8161.tmp
>>>>>>>> -rw-------  1 brodey  wheel          0 Jan  5 13:44 dolphinHc8161.tmp
>>>>>>>> -rw-------  1 brodey  wheel          0 Jan  5 13:42 dolphinNv8161.tmp
>>>>>>>> -rw-------  1 brodey  wheel          0 Jan  5 13:54 dolphinRS8161.tmp
>>>>>>>> -rw-r--r--  1 brodey  wheel  107618884 Jan  5 13:56 dolphinRS8161.tmp.part
>>>>>>>> -rw-------  1 brodey  wheel          0 Jan  5 13:47 dolphinVH8161.tmp
>>>>>>>> -rw-r--r--  1 brodey  wheel  134364212 Jan  5 13:52 dolphinVH8161.tmp.part
>>>>>>>> -rw-------  1 brodey  wheel          0 Jan  5 13:47 dolphinVg8161.tmp
>>>>>>>> -rw-r--r--  1 brodey  wheel  134020904 Jan  5 13:52 dolphinVg8161.tmp.part
>>>>>>>> -rw-------  1 brodey  wheel          0 Jan  5 13:44 dolphinYZ8161.tmp
>>>>>>>> -rw-------  1 brodey  wheel          0 Jan  5 13:54 dolphinZL8161.tmp
>>>>>>>> -rw-r--r--  1 brodey  wheel  109270032 Jan  5 13:56 dolphinZL8161.tmp.part
>>>>>>>> -rw-------  1 brodey  wheel          0 Jan  5 13:47 dolphiner8161.tmp
>>>>>>>> -rw-r--r--  1 brodey  wheel  134364212 Jan  5 13:52 dolphiner8161.tmp.part
>>>>>>>> -rw-------  1 brodey  wheel          0 Jan  5 13:44 dolphinfH8161.tmp
>>>>>>>> -rw-------  1 brodey  wheel       1406 Jan  5 13:37 systemsettingsff5492.tmp
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>> Ok, we may have found the issue here. Dolphin uses /tmp by default for
>>>>>>> file copy jobs, and when that fills up, you're sunk, as you've seen :P
>>>>>>>
>>>>>>> However, we've been limiting tmpfs to 800MB, and if we remove that, it
>>>>>>> should allow you to use much more space when doing large transfers. Give
>>>>>>> this a shot, in /etc/rc.conf look for these lines:
>>>>>>>
>>>>>>> # tmpmfs Flags
>>>>>>> tmpmfs="YES"
>>>>>>> tmpsize="800m"
>>>>>>> tmpmfs_flags="-S"
>>>>>>> clear_tmp="YES"
>>>>>>>
>>>>>>> Remove the tmpsize= and tmpmfs_flags= lines, and reboot. Then try to
>>>>>>> copy again, does /tmp fill up as quickly and die?
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>> Actually scratch that last, try this instead:
>>>>>>
>>>>>> comment out all tmpfs stuff in /etc/rc.conf
>>>>>>
>>>>>> Add this line to /etc/fstab
>>>>>> tmpfs                   /tmp                    tmpfs
>>>>>> rw,mode=1777    0       0
>>>>>>
>>>>>> Reboot, and run df -h, does /tmp show up with much more space available now?
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>     I will scrap tmpfs stuff from rc.conf and test now. Transferring files
>>>>> worked without the flags but 19MB is clearly too small ;). Do you mean
>>>>> tmpmfs in fstab?
>>>>>
>>>>> Brodey
>>>>>
>>>>>
>>>> Yea, I would go ahead and disable tmpfs lines in rc.conf entirely, and
>>>> add the /etc/fstab line I specified above. That seems to work fine on my
>>>> system here, shows /tmp has having all my RAM + Swap space size
>>>> available to work with.
>>>>
>>>>
>>>>
>>>>
>>> yeahh...tmpfs it is! 6.9GB available. Though, cleartemp option sounds
>>> like a good idea to keep around (?)
>>>
>>
>> Great!! I'm not sure the cleatmp option is required though, since tmpfs
>> already gets blown away / recreated when you restart anyway :)
>>
>>
>>
> Removing places, "preview this file.." service hasn't done anything to
> fix up the issue. I will try removing the "Information" feature right now.
>
> As an aside, when transferring files to/fro my FreeNAS box I reach
> 2.0+MiB/s but the info box is only a few hundred KB. That darn forum
> thread is exactly what I'm going through!
> _______________________________________________

Well, it sounds like you've been able to isolate a nice Dolphin bug :)

I would check bugs.kde.org and see if any of the guys from the forum 
posts reported it yet. If not, go ahead and open a new bug up there so 
we can get it resolved right away.


-- 

Kris Moore
PC-BSD Software
http://www.pcbsd.com


More information about the Testing mailing list