0.7.0 beta bug: Wiping memory -- segmentation fault (core dumped)

[Daniel Riedemann]

Arch Linux, KDE 4.3.2, Kernel 2.6.31.4, Python 2.6.3
###

Ok this is going to be a bit more complicated. I hope I got enough information for you, if not please tell me what I have to do.
I have done everything as root because a normal user shouldn't have the rights to clean the memory, right?

First I got a error and nothing was done because of that:

### BEFORE:
# cat /proc/swaps
Filename Type Size Used Priority
/dev/mapper/swap partition 2498096 0 -1

# free
             total used free shared buffers cached
Mem: 2057056 1363148 693908 0 107156 632284
-/+ buffers/cache: 623708 1433348
Swap: 2498096 0 2498096

### CLICK PREVIEW
debug: clean_operation('system'), options = '['memory']'
debug: deepscans= {}
debug: searches= {}

### CLICK ON DELETE
debug: clean_operation('system'), options = '['memory']'
debug: disabling swap
debug: running cmd ['swapoff', '-a', '-v']
Traceback (most recent call last):
  File "/usr/share/bleachbit/Worker.py", line 91, in execute
    for ret in cmd.execute(self.really_delete):
  File "/usr/share/bleachbit/Command.py", line 103, in execute
    func_ret = self.func()
  File "/usr/share/bleachbit/Memory.py", line 198, in wipe_memory
    devices = disable_swap_linux()
  File "/usr/share/bleachbit/Memory.py", line 67, in disable_swap_linux
    % { 'args' : str(args), 'stdout' : stdout, 'stderr' : stderr } )
RuntimeError: Unexpected output:
args='['swapoff', '-a', '-v']'
stdout='swapoff on /dev/mapper/swap
'
stderr=''
debug: deepscans= {}
debug: searches= {}

But the swap device was "umounted" correctly:
#cat /proc/swaps
Filename Type Size Used Priority

So why the error?

####################################
When you do it a second time, or do a manually swapoff before this happens:

### CLICK PREVIEW
debug: clean_operation('system'), options = '['memory']'
debug: deepscans= {}
debug: searches= {}

### CLICK ON DELETE
debug: clean_operation('system'), options = '['memory']'
debug: deepscans= {}
debug: searches= {}

debug: clean_operation('system'), options = '['memory']'
debug: wipe_memory(), devices= None
debug: physical free: 1359511552 B (1296 MB)
info: allocating 1291535974 bytes of memory (1231.00 MB)
debug: wiping the memory I just allocated
[1] 6441 segmentation fault (core dumped) bleachbit

The return code I get after the segmentation fault is always 139:
# echo $?
139

After the crash my memory looks like this:

# free
             total used free shared buffers cached
Mem: 2057056 1345984 711072 0 107300 632260
-/+ buffers/cache: 606424 1450632
Swap: 0 0 0

#cat /proc/swaps
Filename Type Size Used Priority

I think it crashes when it tries to delete my ram, and with my swap is nothing done, right?
Because it goes very fast and I don't see or hear any hdd activity.

Regards

[Daniel Riedemann (darie)]

Oh, and I see the following now in my post above:

debug: physical free: 1359511552 B (1296 MB)
info: allocating 1291535974 bytes of memory (1231.00 MB)

So there are 65 MB still free and not wiped. Is that intended?

[Daniel Riedemann (darie)]

I can't edit the bug report anymore -- I always get an error :-(
So just for clarification:

These three lines of output:
debug: clean_operation('system'), options = '['memory']'
debug: deepscans= {}
debug: searches= {}

Don't belong to "CLICK ON DELETE" I accidently copied them again. These are only part of "CLICK PREVIEW"

Regards

[Andrew Ziem (ahziem1)]

Thanks for the report

Issue #1: BleachBit expects the swap partition to look like /dev/sda5, so that explains the 'unexpected output'. That's easy to fix.

Issue#2: Segmentation fault. I've started on a way to mitigate this risk by using fork().

Issue#3: Only 95% is wiped. This is related to #3. I'm researching a way to get to 100% without a segmentation fault. In my tests going above 95% caused a seg fault, but apparently that does not work for you.

[Andrew Ziem (ahziem1)]

#2 using fork() is in SVN revision 1304, so after a seg fault, BleachBit will at least re-mount swap (not on your system yet since it doesn't know about /dev/mapper/swap) and continue running.

[Daniel Riedemann (darie)]

Update: With revision 1311 the swapoff doesn't work, the fork dies with 139 - but swap gets correctly remounted afterwards.

[Andrew Ziem (ahziem1)]

I checked a plain installation of Arch Linux, and it uses /dev/sda2 for swap. Do you use LVM for swap? Could you run these commands?

cat /proc/partitions
pvscan
lvscan
vgscan

[Andrew Ziem (ahziem1)]

SVN 1340 fixes everything except ' /dev/mapper/swap'. Regarding comment #6, do you use LVM? Could you please reply with the output of these commands?

cat /proc/swaps
cat /proc/partitions
cat /etc/fstab

[Andrew Ziem (ahziem1)]

I discovered /dev/mapper/swap is common name on many Linux distro's for encrypted swap partition. Fixed in SVN 1354

[Daniel Riedemann (darie)]

Hello Andrew,

I'm afraid to tell you that it doesn't work here :-(
Yes I have a fully encrypted hdd with lvm.

## First Click on "Delete"-button:

debug: clean_operation('system'), options = '['memory']'
debug: disabling swap
debug: running cmd ['swapoff', '-a', '-v']
Traceback (most recent call last):
  File "/usr/share/bleachbit/Worker.py", line 89, in execute
    for ret in cmd.execute(self.really_delete):
  File "/usr/share/bleachbit/Command.py", line 108, in execute
    for func_ret in self.func():
  File "/usr/share/bleachbit/Memory.py", line 241, in wipe_memory
    devices = disable_swap_linux()
  File "/usr/share/bleachbit/Memory.py", line 69, in disable_swap_linux
    % { 'args' : str(args), 'stdout' : stdout, 'stderr' : stderr } )
RuntimeError: Unexpected output:
args='['swapoff', '-a', '-v']'
stdout='swapoff für /dev/mapper/swap
'
stderr=''
debug: deepscans= {}
debug: searches= {}

But it get's umounted correctly !

## Second click on "Delete"-Button:
debug: clean_operation('system'), options = '['memory']'
debug: detected swap devices: None
debug: new nice value 19
Traceback (most recent call last):
  File "/usr/share/bleachbit/Memory.py", line 98, in fill_memory_linux
    login = os.getlogin()
OSError: [Errno 2] Datei oder Verzeichnis nicht gefunden
debug: physical free: 1498976256 B (1429 MB)
debug: wipe_memory() pid 6352 waiting for child pid 6386
info: allocating 1071.00 MB (1123946496 B) memory
debug: wiping 1071.00 MB I just allocated
warning: child process returned code 139
debug: re-enabling swap
debug: deepscans= {}
debug: searches= {}

re-enabling swap works. But the fork still dies with return code 139.

Maybe my many tmpfs mounts are the reason?
Please see my attachments for that.

[Daniel Riedemann (darie)]

Because I can't edit: I'm using svn revision 1369

[Andrew Ziem (ahziem1)]

BleachBit was confused on the German 'für'. Try SVN 1371. I also fixed the OSError 2.

[Daniel Riedemann (darie)]

Revision 1376:
First click on Delete I get: error: cannot find size of swap device '/dev/mapper/swap'

##Detailed output:
debug: clean_operation('system'), options = '['memory']'
debug: disabling swap
debug: running cmd ['swapoff', '-a', '-v']
debug: detected swap devices: ['/dev/mapper/swap']
info: wiping swap device '/dev/mapper/swap'
Traceback (most recent call last):
  File "/usr/share/bleachbit/Worker.py", line 89, in execute
    for ret in cmd.execute(self.really_delete):
  File "/usr/share/bleachbit/Command.py", line 108, in execute
    for func_ret in self.func():
  File "/usr/share/bleachbit/Memory.py", line 244, in wipe_memory
    wipe_swap_linux(devices)
  File "/usr/share/bleachbit/Memory.py", line 223, in wipe_swap_linux
    if get_swap_size_linux(device) > 8*1024**3:
  File "/usr/share/bleachbit/Memory.py", line 137, in get_swap_size_linux
    (device, open('/proc/partitions').read()))
RuntimeError: error: cannot find size of swap device '/dev/mapper/swap'
major minor #blocks name

   8 0 625131864 sda
   8 1 134217728 sda1
   8 2 128520 sda2
   8 3 2498107 sda3
   8 4 1 sda4
   8 5 64002928 sda5
   8 6 424280941 sda6
 254 0 64001900 dm-0
 254 1 424279913 dm-1
 254 2 2498107 dm-2

debug: deepscans= {}
debug: searches= {}

##Second click on Delete:
debug: clean_operation('system'), options = '['memory']'
debug: detected swap devices: None
debug: wipe_memory() pid 11021 waiting for child pid 11033
debug: new nice value 19
debug: dropping privileges of pid 11033 to uid 100
debug: physical free: 1320341504 B (1259 MB)
info: allocating 944.00 MB (990077952 B) memory
debug: wiping 944.00 MB I just allocated
warning: child process returned code 11
debug: re-enabling swap
debug: deepscans= {}
debug: searches= {}

Is return code 11 good or bad?
Do you try to get the swap size trough /proc/partitions? Wouldn't be /proc/swaps better and easier for that? (Because of LVM)

[Andrew Ziem (ahziem1)]

>Is return code 11 good or bad?

It means the forked process did not finish. Does /var/log/messages show the Linux 'out-of-memory killer' killed a process?

>Do you try to get the swap size trough /proc/partitions? Wouldn't be /proc/swaps better and easier for that? (Because of LVM)

Yes, try SVN 1382.

[Daniel Riedemann (darie)]

Revision 1383:

Same again on first click:

debug: clean_operation('system'), options = '['memory']'
debug: disabling swap
debug: running cmd ['swapoff', '-a', '-v']
debug: detected swap devices: ['/dev/mapper/swap']
info: wiping swap device '/dev/mapper/swap'
Traceback (most recent call last):
  File "/usr/share/bleachbit/Worker.py", line 89, in execute
    for ret in cmd.execute(self.really_delete):
  File "/usr/share/bleachbit/Command.py", line 108, in execute
    for func_ret in self.func():
  File "/usr/share/bleachbit/Memory.py", line 246, in wipe_memory
    wipe_swap_linux(devices)
  File "/usr/share/bleachbit/Memory.py", line 225, in wipe_swap_linux
    if get_swap_size_linux(device) > 8*1024**3:
  File "/usr/share/bleachbit/Memory.py", line 139, in get_swap_size_linux
    (device, open('/proc/partitions').read()))
RuntimeError: error: cannot find size of swap device '/dev/mapper/swap'
major minor #blocks name

   8 0 625131864 sda
   8 1 134217728 sda1
   8 2 128520 sda2
   8 3 2498107 sda3
   8 4 1 sda4
   8 5 64002928 sda5
   8 6 424280941 sda6
 254 0 64001900 dm-0
 254 1 424279913 dm-1
 254 2 2498107 dm-2

debug: deepscans= {}
debug: searches= {}

##########

Same again for second click:

debug: clean_operation('system'), options = '['memory']'
debug: detected swap devices: None
debug: wipe_memory() pid 6301 waiting for child pid 6313
debug: new nice value 19
debug: dropping privileges of pid 6313 to uid 100
debug: physical free: 1402400768 B (1337 MB)
info: allocating 1002.00 MB (1051610112 B) memory
debug: wiping 1002.00 MB I just allocated
warning: child process returned code 11
debug: re-enabling swap
debug: deepscans= {}
debug: searches= {}

###########

dmesg after that shows:

python[6313]: segfault at 1c393010 ip 00007f6d6a02090b sp 00007fff30e1d398 error 6 in libc-2.10.1.so[7f6d69fa5000+149000]
Adding 2498096k swap on /dev/mapper/swap. Priority:-1 extents:1 across:2498096k

Ouch! :-(

[Andrew Ziem (ahziem1)]

> RuntimeError: error: cannot find size of swap device '/dev/mapper/swap'

OK. Try SVN 1385.

> python[6313]: segfault at 1c393010 ip 00007f6d6a02090b sp 00007fff30e1d398 error 6 in libc-2.10.1.so[7f6d69fa5000+149000]

Once the swap problems are fixed, please file another bug ticket for the 'segfault' and 'returned code 11' to be fixed after 0.7.0 is released.

[Daniel Riedemann (darie)]

Okay, no problem... :-) Do you mean one ticket for segfault and return code or do you mean one each (so two in total)?
And sorry that I have to say that, but I found another bug (I'm going to create the ticket now). I think this new one should be fixed before the final release of 0.7.0 ...

With Rev. 1385 it's exactly the same.
Just in case I overlooked something, here the output again:

debug: clean_operation('system'), options = '['memory']'
debug: disabling swap
debug: running cmd ['swapoff', '-a', '-v']
debug: detected swap devices: ['/dev/mapper/swap']
info: wiping swap device '/dev/mapper/swap'
Traceback (most recent call last):
  File "/usr/share/bleachbit/Worker.py", line 89, in execute
    for ret in cmd.execute(self.really_delete):
  File "/usr/share/bleachbit/Command.py", line 108, in execute
    for func_ret in self.func():
  File "/usr/share/bleachbit/Memory.py", line 246, in wipe_memory
    wipe_swap_linux(devices)
  File "/usr/share/bleachbit/Memory.py", line 225, in wipe_swap_linux
    if get_swap_size_linux(device) > 8*1024**3:
  File "/usr/share/bleachbit/Memory.py", line 139, in get_swap_size_linux
    (device, open('/proc/partitions').read()))
RuntimeError: error: cannot find size of swap device '/dev/mapper/swap'
major minor #blocks name

   8 0 625131864 sda
   8 1 134217728 sda1
   8 2 128520 sda2
   8 3 2498107 sda3
   8 4 1 sda4
   8 5 64002928 sda5
   8 6 424280941 sda6
 254 0 64001900 dm-0
 254 1 424279913 dm-1
 254 2 2498107 dm-2

debug: deepscans= {}
debug: searches= {}

[Andrew Ziem (ahziem1)]

That output doesn't look like SVN 1385. The biggest clue is '/proc/partitions' which was removed in SVN 1382

http://bleachbit.svn.sourceforge.net/viewvc/bleachbit/trunk/bleachbit/Memory.py?r1=1382&r2=1384

Also, I don't 'install' SVN at all, but your output shows it is running from /usr/.

[Andrew Ziem (ahziem1)]

>Okay, no problem... :-) Do you mean one ticket for segfault and return code or do you mean one each (so two in total)?

Sorry, I forgot to answer this question. I mean one new ticket for these two. They are related.

[Daniel Riedemann (darie)]

>That output doesn't look like SVN 1385.

Oups, sorry... I build a package but I forgot to install it. I think I should go earlier to bed to today... :-(
Now it works and looks like this:

debug: clean_operation('system'), options = '['memory']'
debug: disabling swap
debug: running cmd ['swapoff', '-a', '-v']
debug: detected swap devices: ['/dev/mapper/swap']
info: wiping swap device '/dev/mapper/swap'
debug: running cmd ['blkid', '/dev/mapper/swap', '-s', 'UUID']
debug: uuid(/dev/mapper/swap)='c6ea33e5-0af6-4148-8a92-b0d8eddfdf1f'
debug: reinitializing swap device /dev/mapper/swap
debug: running cmd ['mkswap', '/dev/mapper/swap', '-U', 'c6ea33e5-0af6-4148-8a92-b0d8eddfdf1f']
debug: wipe_memory() pid 15519 waiting for child pid 15528
debug: new nice value 19
debug: dropping privileges of pid 15528 to uid 100
debug: physical free: 1087000576 B (1036 MB)
info: allocating 777.00 MB (814930944 B) memory
debug: wiping 777.00 MB I just allocated
warning: child process returned code 11
debug: re-enabling swap
debug: deepscans= {}
debug: searches= {}

What does bleachbit do when it wipes the swap? It goes so fast and has so less hdd activity it can't be something like a dd if=/dev/zero ...

[Andrew Ziem (ahziem1)]

>What does bleachbit do when it wipes the swap? It goes so fast and has
> so less hdd activity it can't be something like a dd if=/dev/zero ...

Yes, basically the same as 'dd if=/dev/zero'