lkml.org 
[lkml]   [2013]   [Oct]   [22]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    From
    SubjectRe: [uml-devel] fuzz tested 32 bit user mode linux image hangs at in histfs
    On 10/22/2013 06:12 PM, Richard Weinberger wrote:
    > On Tue, Oct 22, 2013 at 5:16 PM, Toralf Förster <toralf.foerster@gmx.de> wrote:
    >>
    >> When I fuzz testing a 32 bit UML at a 32 bit host (guest 3.12.-rc6-x, host 3.11.6) with trinity
    >> and use hostfs for the victom files for trinity. then trintiy often hangs while trying to finish.
    >>
    >> At the host I do have 1 process eating 100% CPU power of 1 core. A back trace of thet linux process at the hosts gives :
    >>
    >> tfoerste@n22 ~ $ sudo gdb /usr/local/bin/linux-v3.12-rc6-57-g69c88dc 16749 -n -batch -ex bt
    >> radix_tree_next_chunk (root=0x21, iter=0x47647c60, flags=12) at lib/radix-tree.c:769
    >> 769 while (++offset < RADIX_TREE_MAP_SIZE) {
    >> #0 radix_tree_next_chunk (root=0x21, iter=0x47647c60, flags=12) at lib/radix-tree.c:769
    >> #1 0x080cc13e in find_get_pages (mapping=0x483ed240, start=0, nr_pages=14, pages=0xc) at mm/filemap.c:844
    >> #2 0x080d5caa in pagevec_lookup (pvec=0x47647cc4, mapping=0x21, start=33, nr_pages=33) at mm/swap.c:914
    >> #3 0x080d609a in truncate_inode_pages_range (mapping=0x483ed240, lstart=0, lend=-1) at mm/truncate.c:241
    >> #4 0x080d643f in truncate_inode_pages (mapping=0x21, lstart=51539607585) at mm/truncate.c:358
    >> #5 0x08260838 in hostfs_evict_inode (inode=0x483ed188) at fs/hostfs/hostfs_kern.c:242
    >> #6 0x0811a8cf in evict (inode=0x483ed188) at fs/inode.c:549
    >> #7 0x0811b2ad in iput_final (inode=<optimized out>) at fs/inode.c:1391
    >> #8 iput (inode=0x483ed188) at fs/inode.c:1409
    >> #9 0x08117648 in dentry_iput (dentry=<optimized out>) at fs/dcache.c:331
    >> #10 d_kill (dentry=0x47d6d580, parent=0x47d95d10) at fs/dcache.c:477
    >> #11 0x08118068 in dentry_kill (dentry=<optimized out>, unlock_on_failure=<optimized out>) at fs/dcache.c:586
    >> #12 dput (dentry=0x47d6d580) at fs/dcache.c:641
    >> #13 0x08104903 in __fput (file=0x47471840) at fs/file_table.c:264
    >> #14 0x0810496b in ____fput (work=0x47471840) at fs/file_table.c:282
    >> #15 0x08094496 in task_work_run () at kernel/task_work.c:123
    >> #16 0x0807efd2 in exit_task_work (task=<optimized out>) at include/linux/task_work.h:21
    >> #17 do_exit (code=1196535808) at kernel/exit.c:787
    >> #18 0x0807f5dd in do_group_exit (exit_code=0) at kernel/exit.c:920
    >> #19 0x0807f649 in SYSC_exit_group (error_code=<optimized out>) at kernel/exit.c:931
    >> #20 SyS_exit_group (error_code=0) at kernel/exit.c:929
    >> #21 0x08062984 in handle_syscall (r=0x4763b1d4) at arch/um/kernel/skas/syscall.c:35
    >> #22 0x08074fb5 in handle_trap (local_using_sysemu=<optimized out>, regs=<optimized out>, pid=<optimized out>) at arch/um/os-Linux/skas/process.c:198
    >> #23 userspace (regs=0x4763b1d4) at arch/um/os-Linux/skas/process.c:431
    >> #24 0x0805f750 in fork_handler () at arch/um/kernel/process.c:160
    >> #25 0x00000000 in ?? ()
    >>
    >
    > That trace is identical to the one you reported yesterday.
    > But this time no nfs is in the game, right?
    >

    Right - I could narrow down it in the meanwhile to hostfs only. First I
    argued if NFS sometimes might force the issue to happen later but in the
    mean while I don't think so.


    >> Last message of trinity's watchdog are :
    >>
    >> ...
    >> [watchdog] exit_reason=2, but 2 children still running.
    >> Bailing main loop. Exit reason: Reached maximum syscall count.
    >> [watchdog] Reached limit 10001. Telling children to exit.
    >> [watchdog] [1516] Watchdog exiting
    >>
    >>
    >> I'm unsure if this is only UML specific, interesting for the fs people or mm or ... ?
    >>
    >>
    >> --
    >> MfG/Sincerely
    >> Toralf Förster
    >> pgp finger print: 7B1A 07F4 EC82 0F90 D4C2 8936 872A E508 7DB6 9DA3
    >>
    >> ------------------------------------------------------------------------------
    >> October Webinars: Code for Performance
    >> Free Intel webinars can help you accelerate application performance.
    >> Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from
    >> the latest Intel processors and coprocessors. See abstracts and register >
    >> http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk
    >> _______________________________________________
    >> User-mode-linux-devel mailing list
    >> User-mode-linux-devel@lists.sourceforge.net
    >> https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel
    >
    >
    >


    --
    MfG/Sincerely
    Toralf Förster
    pgp finger print: 7B1A 07F4 EC82 0F90 D4C2 8936 872A E508 7DB6 9DA3
    --
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/

    \
     
     \ /
      Last update: 2013-10-22 19:01    [W:2.256 / U:0.744 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site