Re: [fuse-devel] Problem with fuse on 2.6.25 kernel

[Miklos S. <mi...@sz...>]

> > 
> > OK, if you have MAGIC_SYSRQ config option enabled in the kernel, then
> > you could try pressing SysRq-t, when the mount gets stuck, or doing
> > 
> >  echo t > /proc/sysrq-trigger
> > 
> > and sending the resulting kernel output from 'dmesg' to me.
> > 
> 
> There's plenty of output, but you're probably interested in encfs only. I did
> the dump as root, so permissions should be of no concern. By the way, one
> umount is sufficient to clear the bad state off the target directory (don't
> know why I needed two previously).
> 
> Apr 28 10:34:07 mortug [ 2468.228845] encfs         S ffff81000100aac0     0 
> 8277   8270
> Apr 28 10:34:07 mortug [ 2468.228845]  ffff81003a8e9b08 0000000000000096
> 0000000000000000 0000000000000000
> Apr 28 10:34:07 mortug [ 2468.228845]  0000000200000001 ffffffff8073cac0
> ffffffff8073cac0 ffffffff8073cac0
> Apr 28 10:34:07 mortug [ 2468.228845]  ffffffff80738d20 ffffffff8073cac0
> ffff81003154a288 ffff81003a8e9ab8
> Apr 28 10:34:07 mortug [ 2468.228845] Call Trace:
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff8816d968>]
> :fuse:fuse_get_req+0x88/0x120
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff8024e180>]
> autoremove_wake_function+0x0/0x30
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff8039a916>]
> smack_d_instantiate+0x56/0x1b0
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff881702dd>]
> :fuse:fuse_getxattr+0x7d/0x180
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff8039a9b5>]
> smack_d_instantiate+0xf5/0x1b0
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff802c09a3>]
> d_alloc_root+0x63/0x70
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff88172930>]
> :fuse:fuse_fill_super+0x310/0x650
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff802c4762>]
> get_filesystem+0x12/0x40
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff88172620>]
> :fuse:fuse_fill_super+0x0/0x650
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff802aedd3>]
> get_sb_nodev+0x63/0xe0
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff802ae9b8>]
> vfs_kern_mount+0x78/0x160
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff802aeb03>]
> do_kern_mount+0x53/0x110
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff802c73ab>]
> do_new_mount+0x9b/0xe0
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff802c75de>]
> do_mount+0x1ee/0x240
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff80286680>]
> __get_free_pages+0x90/0xa0
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff802c76cb>]
> sys_mount+0x9b/0x100
> Apr 28 10:34:07 mortug [ 2468.228845]  [<ffffffff8020c1eb>]
> system_call_after_swapgs+0x7b/0x80
> Apr 28 10:34:07 mortug [ 2468.228845] 

Thanks.  Here's the problem:

:fuse:fuse_getxattr+0x7d/0x180
smack_d_instantiate+0xf5/0x1b0
d_alloc_root+0x63/0x70
:fuse:fuse_fill_super+0x310/0x650

Smack is doing something that fuse isn't prepared for: calling the
filesystem's ->getxattr() method before the super block is set up.  It
probably shouldn't be doing this.

Casey, what's up with this?

Miklos