features/shard: Fix permission issues

This patch does the following:
* reverts commit b467af0e99b39ef708420d3f7f6696b0ca618512
* changes ownership on shards under /.shard to be root:root
* makes readv, writev, [f]truncate, rename, and unlink fops
  to perform operations on files under /.shard with
  frame->root->{uid,gid} as 0.

This would ensure that a [f]setattr on a sharded file
does not need to be called on all the shards associated with it.

Change-Id: Idcfb8c0dd354b0baab6b2356d2ab83ce51caa20e
BUG: 1251824
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: http://review.gluster.org/11992
Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
Tested-by: NetBSD Build System <jenkins@build.gluster.org>
Tested-by: Gluster Build System <jenkins@build.gluster.com>

1 files changed, 57 insertions, 11 deletions

diff --git a/tests/bugs/shard/bug-1251824.t b/tests/bugs/shard/bug-1251824.t
index 71bfdc7fda..9e6ceaa816 100644
--- a/tests/bugs/shard/bug-1251824.t
+++ b/tests/bugs/shard/bug-1251824.t
@@ -1,6 +1,7 @@
 #!/bin/bash
 
 . $(dirname $0)/../../include.rc
+. $(dirname $0)/../../common-utils.rc
 
 cleanup;
 
@@ -8,52 +9,97 @@ TEST glusterd
 TEST pidof glusterd
 TEST $CLI volume create $V0 $H0:$B0/${V0}{0,1}
 TEST $CLI volume set $V0 features.shard on
+TEST $CLI volume set $V0 performance.strict-write-ordering on
 TEST $CLI volume start $V0
 
-TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0
+TEST glusterfs --volfile-id=$V0 --volfile-server=$H0 $M0
 TEST useradd -M test_user 2>/dev/null
 
 # Create 3 files as root.
 TEST touch $M0/foo
 TEST touch $M0/bar
 TEST touch $M0/baz
+TEST touch $M0/qux
+TEST mkdir $M0/dir
 
 # Change ownership to non-root on foo and bar.
 TEST chown test_user:test_user $M0/foo
 TEST chown test_user:test_user $M0/bar
 
 # Write 6M of data on foo as non-root, 2M overflowing into block-1.
-su -m test_user -c "dd if=/dev/zero of=$M0/foo bs=1M count=6"
+TEST run_cmd_as_user test_user "dd if=/dev/zero of=$M0/foo bs=1M count=6"
 
-# Ensure owner and group are same on the shard as the main file.
+# Ensure owner and group are root on the block-1 shard.
 gfid_foo=`getfattr -n glusterfs.gfid.string $M0/foo 2>/dev/null \
           | grep glusterfs.gfid.string | cut -d '"' -f 2`
 
-EXPECT "test_user" echo `find $B0 -name $gfid_foo.1 | xargs stat -c %U`
-EXPECT "test_user" echo `find $B0 -name $gfid_foo.1 | xargs stat -c %G`
+EXPECT "root" echo `find $B0 -name $gfid_foo.1 | xargs stat -c %U`
+EXPECT "root" echo `find $B0 -name $gfid_foo.1 | xargs stat -c %G`
+
+#Ensure /.shard is owned by root.
+EXPECT "root" echo `find $B0/${V0}0 -name .shard | xargs stat -c %U`
+EXPECT "root" echo `find $B0/${V0}0 -name .shard | xargs stat -c %G`
+EXPECT "root" echo `find $B0/${V0}1 -name .shard | xargs stat -c %U`
+EXPECT "root" echo `find $B0/${V0}1 -name .shard | xargs stat -c %G`
 
 # Write 6M of data on bar as root.
 TEST dd if=/dev/zero of=$M0/bar bs=1M count=6
 
-# Ensure owner and group are same on the shard as the main file.
+# Ensure owner and group are root on the block-1 shard.
 gfid_bar=`getfattr -n glusterfs.gfid.string $M0/bar 2>/dev/null \
           | grep glusterfs.gfid.string | cut -d '"' -f 2`
 
-EXPECT "test_user" echo `find $B0 -name $gfid_bar.1 | xargs stat -c %U`
-EXPECT "test_user" echo `find $B0 -name $gfid_bar.1 | xargs stat -c %G`
+EXPECT "root" echo `find $B0 -name $gfid_bar.1 | xargs stat -c %U`
+EXPECT "root" echo `find $B0 -name $gfid_bar.1 | xargs stat -c %G`
 
 # Write 6M of data on baz as root.
 TEST dd if=/dev/zero of=$M0/baz bs=1M count=6
 
-# Ensure owner andgroup are same on the shard as the main file.
 gfid_baz=`getfattr -n glusterfs.gfid.string $M0/baz 2>/dev/null \
           | grep glusterfs.gfid.string | cut -d '"' -f 2`
 
+# Ensure owner and group are root on the block-1 shard.
 EXPECT "root" echo `find $B0 -name $gfid_baz.1 | xargs stat -c %U`
 EXPECT "root" echo `find $B0 -name $gfid_baz.1 | xargs stat -c %G`
-userdel test_user
 
-TEST umount $M0
+# Test to ensure unlink from an unauthorized user does not lead to only
+# the shards under /.shard getting unlinked while that on the base file fails
+# with EPERM/ACCES.
+
+TEST ! run_cmd_as_user test_user "unlink $M0/baz"
+TEST find $B0/*/.shard/$gfid_baz.1
+
+# Test to ensure rename of a file where the dest file exists and is sharded,
+# from an unauthorized user does not lead to only the shards under /.shard
+# getting unlinked while that on the base file fails with EPERM/ACCES.
+
+TEST ! run_cmd_as_user test_user "mv -f $M0/qux $M0/baz"
+TEST find $B0/*/.shard/$gfid_baz.1
+TEST stat $M0/qux
+
+# Shard translator executes steps in the following order while doing a truncate
+# to a lower size:
+# 1) unlinking shards under /.shard first with frame->root->{uid,gid} being 0,
+# 2) truncate the original file by the right amount.
+# The following two tests are towards ensuring that truncate attempt from an
+# unauthorised user doesn't result in only the shards under /.shard getting
+# removed (since they're being performed as root) while step 2) above fails,
+# leaving the file in an inconsistent state.
+
+TEST ! run_cmd_as_user test_user "truncate -s 1M $M0/baz"
+TEST find $B0/*/.shard/$gfid_baz.1
+
+# Perform a cp as non-root user. This should trigger readv() which will trigger
+# reads on first shard of "foo" under /.shard, and this must not fail if shard
+# translator correctly sets frame->root->uid,gid to 0 before reading off the
+# first shard, since it's owned by root.
+TEST chown test_user:test_user $M0/dir
+TEST run_cmd_as_user test_user "cp $M0/foo $M0/dir/quux"
+
+md5sum_foo=$(md5sum $M0/foo | awk '{print $1}')
+EXPECT "$md5sum_foo" echo `md5sum $M0/dir/quux | awk '{print $1}'`
+
+userdel test_user
 
 TEST $CLI volume stop $V0
 TEST $CLI volume delete $V0