+++ This bug was initially created as a clone of Bug #1374015 +++

Description of problem:

Ganesha crashes while removing files from clients.

Version-Release number of selected component (if applicable):

[root@dhcp43-116 ~]# rpm -qa|grep glusterfs
glusterfs-geo-replication-3.8.3-0.6.git7956718.el7.centos.x86_64
glusterfs-api-3.8.3-0.6.git7956718.el7.centos.x86_64
glusterfs-fuse-3.8.3-0.6.git7956718.el7.centos.x86_64
glusterfs-server-3.8.3-0.6.git7956718.el7.centos.x86_64
glusterfs-libs-3.8.3-0.6.git7956718.el7.centos.x86_64
glusterfs-client-xlators-3.8.3-0.6.git7956718.el7.centos.x86_64
glusterfs-ganesha-3.8.3-0.6.git7956718.el7.centos.x86_64
glusterfs-cli-3.8.3-0.6.git7956718.el7.centos.x86_64
glusterfs-debuginfo-3.8.3-0.6.git7956718.el7.centos.x86_64
glusterfs-3.8.3-0.6.git7956718.el7.centos.x86_64

[root@dhcp43-116 ~]# rpm -qa|grep ganesha
nfs-ganesha-gluster-next.20160827.7641daf-1.el7.centos.x86_64
glusterfs-ganesha-3.8.3-0.6.git7956718.el7.centos.x86_64
nfs-ganesha-debuginfo-next.20160827.7641daf-1.el7.centos.x86_64
nfs-ganesha-next.20160827.7641daf-1.el7.centos.x86_64


How reproducible:

Twice

Steps to Reproduce:
1.Create large number of files on a dist-rep volume via v4 ganesha mount from 2 different clients.
2.Start removing different set of files simultaneously from both the clients. 
3.Observe that while removal is in progress, ganesha crashes on the mounted node with below bt:

(gdb) bt
#0  0x00007f937af59c5f in __inode_ctx_free (inode=inode@entry=0x7f9356c9fe24)
    at inode.c:332
#1  0x00007f937af5ae42 in __inode_destroy (inode=0x7f9356c9fe24) at inode.c:353
#2  inode_table_prune (table=table@entry=0x7f9360103f30) at inode.c:1543
#3  0x00007f937af5b124 in inode_unref (inode=0x7f9356c9fe24) at inode.c:524
#4  0x00007f937b232216 in pub_glfs_h_close (object=0x7f925400f660)
    at glfs-handleops.c:1365
#5  0x00007f937b64a929 in handle_release (obj_hdl=0x7f92540308f8)
    at /usr/src/debug/nfs-ganesha/src/FSAL/FSAL_GLUSTER/handle.c:70
#6  0x00007f937fd1caf4 in mdcache_lru_clean (entry=0x7f924cd8f060)
    at /usr/src/debug/nfs-ganesha/src/FSAL/Stackable_FSALs/FSAL_MDCACHE/mdcache_lru.c:421
#7  mdcache_lru_get (entry=entry@entry=0x7f92e94e1a70)
    at /usr/src/debug/nfs-ganesha/src/FSAL/Stackable_FSALs/FSAL_MDCACHE/mdcache_lru.c:1229
#8  0x00007f937fd268b6 in mdcache_alloc_handle (fs=0x0, 
    sub_handle=0x7f924802e358, export=0x7f9380e0fab0)
    at /usr/src/debug/nfs-ganesha/src/FSAL/Stackable_FSALs/FSAL_MDCACHE/mdcache_helpers.c:117
#9  mdcache_new_entry (export=export@entry=0x7f9380e0fab0, 
---Type <return> to continue, or q <return> to quit---
    sub_handle=0x7f924802e358, attrs_in=attrs_in@entry=0x7f92e94e1bd0, 
    attrs_out=attrs_out@entry=0x0, new_directory=new_directory@entry=false, 
    entry=entry@entry=0x7f92e94e1b30, state=state@entry=0x0)
    at /usr/src/debug/nfs-ganesha/src/FSAL/Stackable_FSALs/FSAL_MDCACHE/mdcache_helpers.c:367
#10 0x00007f937fd208e4 in mdcache_alloc_and_check_handle (
    export=export@entry=0x7f9380e0fab0, sub_handle=<optimized out>, 
    new_obj=new_obj@entry=0x7f92e94e1bc8, 
    new_directory=new_directory@entry=false, 
    attrs_in=attrs_in@entry=0x7f92e94e1bd0, attrs_out=attrs_out@entry=0x0, 
    tag=tag@entry=0x7f937fd58b10 "lookup ", parent=parent@entry=0x7f9380e783c0, 
    name=name@entry=0x7f924e2d5e4c "def26337", invalidate=invalidate@entry=true, 
    state=state@entry=0x0)
    at /usr/src/debug/nfs-ganesha/src/FSAL/Stackable_FSALs/FSAL_MDCACHE/mdcache_handle.c:93
#11 0x00007f937fd27986 in mdc_lookup_uncached (
    mdc_parent=mdc_parent@entry=0x7f9380e783c0, 
    name=name@entry=0x7f924e2d5e4c "def26337", 
    new_entry=new_entry@entry=0x7f92e94e1d40, attrs_out=attrs_out@entry=0x0)
    at /usr/src/debug/nfs-ganesha/src/FSAL/Stackable_FSALs/FSAL_MDCACHE/mdcache_hel---Type <return> to continue, or q <return> to quit---
pers.c:981
#12 0x00007f937fd1f53f in mdcache_readdir (dir_hdl=0x7f9380e783f8, 
    whence=<optimized out>, dir_state=0x7f92e94e1dc0, 
    cb=0x7f937fc4bc00 <populate_dirent>, attrmask=<optimized out>, 
    eod_met=0x7f92e94e1e8b)
    at /usr/src/debug/nfs-ganesha/src/FSAL/Stackable_FSALs/FSAL_MDCACHE/mdcache_handle.c:626
#13 0x00007f937fc4d97d in fsal_readdir (directory=directory@entry=0x7f9380e783f8, 
    cookie=cookie@entry=4053976218433744092, 
    nbfound=nbfound@entry=0x7f92e94e1e8c, eod_met=eod_met@entry=0x7f92e94e1e8b, 
    attrmask=122830, cb=cb@entry=0x7f937fc88d50 <nfs4_readdir_callback>, 
    opaque=opaque@entry=0x7f92e94e1e90)
    at /usr/src/debug/nfs-ganesha/src/FSAL/fsal_helper.c:1457
#14 0x00007f937fc89d1b in nfs4_op_readdir (op=0x7f92a40219d0, 
    data=0x7f92e94e20b0, resp=0x7f9248025080)
    at /usr/src/debug/nfs-ganesha/src/Protocols/NFS/nfs4_op_readdir.c:631
#15 0x00007f937fc765bf in nfs4_Compound (arg=<optimized out>, 
    req=<optimized out>, res=0x7f924802cb50)
    at /usr/src/debug/nfs-ganesha/src/Protocols/NFS/nfs4_Compound.c:734
#16 0x00007f937fc65c0c in nfs_rpc_execute (reqdata=reqdata@entry=0x7f92a4013dd0)
---Type <return> to continue, or q <return> to quit---
    at /usr/src/debug/nfs-ganesha/src/MainNFSD/nfs_worker_thread.c:1281
#17 0x00007f937fc674bd in worker_run (ctx=0x7f9380ecf9e0)
    at /usr/src/debug/nfs-ganesha/src/MainNFSD/nfs_worker_thread.c:1548
#18 0x00007f937fcfb629 in fridgethr_start_routine (arg=0x7f9380ecf9e0)
    at /usr/src/debug/nfs-ganesha/src/support/fridgethr.c:550
#19 0x00007f937e1d2dc5 in start_thread () from /lib64/libpthread.so.0
#20 0x00007f937d8a01cd in clone () from /lib64/libc.so.6

Actual results:

Ganesha crashes on one of the nodes while files are being rmeoved from 2 clients.

Expected results:

There should not be any crash.

Additional info:

There is another bug filed for the ganesha crash seen during removal of files (https://bugzilla.redhat.com/show_bug.cgi?id=1373262), but in this case bt is different, so new bug is being filed.

--- Additional comment from Niels de Vos on 2016-09-12 07:39:03 CEST ---

All 3.8.x bugs are now reported against version 3.8 (without .x). For more information, see http://www.gluster.org/pipermail/gluster-devel/2016-September/050859.html