Re: [coLinux-devel] Possible bug with unionfs and colinux

[Henry N. <hen...@ar...>]

  Hello Sergio,

At 05.03.2010 09:04, Sergio Paracuellos wrote:
> I am using the last stable version of colinux
> (stable-colinux-20100205.tar.gz) compiled in linux using an autoboot usb
> which uses unionfs for the filesystem. I am using windows 7 with Xming.
> When I tryed to launch firefox 3.5, colinux crash with an kernel oops. I
> attached the oops in the mail.
>
> When firefox starts sqlite tried to do a ftruncate in /tmp/var which is
> a directory of the unionfs and get always this oops. (In general I get
> this oops in every program which do the same operation ftruncate)
>
> My temporal solution is to make a link of this directory to a directory
> in another FS and the problem disappear.
>
> ln -fs tmp /media/data/tmp

> Mar  5 06:54:54 live kernel: WARNING: at /home/sergio/LIME-ACCESS/colinux/workspace/build/linux-2.6.22.18-source/fs/unionfs/subr.c:244 unionfs_copy_attr_times()
> Mar  5 06:54:54 live kernel:  [<c0103b5a>] show_trace_log_lvl+0x1a/0x30
> Mar  5 06:54:54 live kernel:  [<c0103ca2>] show_trace+0x12/0x20
> Mar  5 06:54:54 live kernel:  [<c0104b35>] dump_stack+0x15/0x20
> Mar  5 06:54:54 live kernel:  [<e08d7141>] unionfs_copy_attr_times+0x141/0x160 [unionfs]
> Mar  5 06:54:54 live kernel:  [<e08ea1f0>] unionfs_readpage+0x120/0x1e0 [unionfs]
> Mar  5 06:54:54 live kernel:  [<c013e66b>] do_generic_mapping_read+0x17b/0x570
> Mar  5 06:54:54 live kernel:  [<c0140891>] generic_file_aio_read+0xc1/0x1c0
> Mar  5 06:54:54 live kernel:  [<c015d57f>] do_sync_read+0xcf/0x110
> Mar  5 06:54:54 live kernel:  [<c015de3f>] vfs_read+0xaf/0x160
> Mar  5 06:54:54 live kernel:  [<c015e26d>] sys_read+0x3d/0x70
> Mar  5 06:54:54 live kernel:  [<c0102866>] sysenter_past_esp+0x5f/0x89
> Mar  5 06:54:54 live kernel:  =======================

This is the part of source:

/* copy a/m/ctime from the lower branch with the newest times */
void unionfs_copy_attr_times(struct inode *upper)
{
     int bindex;
     struct inode *lower;

     if (!upper)
         return;
     if (ibstart(upper) < 0) {
#ifdef CONFIG_UNION_FS_DEBUG
         WARN_ON(ibstart(upper) < 0); <=== line 244
#endif /* CONFIG_UNION_FS_DEBUG */
         return;
     }
...

coLinux developer can't help you for this warning.

Please try a newer version of unionfs.
You can get "unionfs-2.5.4_for_2.6.22.19" from 
http://www.filesystems.org/project-unionfs.html
The build is ok. But I can't test it. I have no filesystems ready for 
unionfs.

Run these steps inside your Linux kernel tree:
* patch -p0 -R < patches/unionfs-2.2.4_for_2.6.22.18.diff
* patch -p0 < patches/unionfs-2.5.4_for_2.6.22.19.diff
* Than rebuild your modules with "make modules && make modules_install".
* Reboot coLinux and check the version number 2.5.4 on module load.

Maybe, it is a known bug and new version helps? If not, then you should 
post this bug to unionfs, and give they and us simple shell steps or 
small C-Programs to repeat your problem. I think, there is not only a 
ftruncate. The warning comes from a file "read". maybe from read after 
ftruncate?

Exists a test-suite for this filesystem? So, everybody without knowing 
of all the usage of unionfs can run some automated tests?

-- 
Henry N.