From d5ce2300f77c25b38a076d4dd6a5521e82c56172 Mon Sep 17 00:00:00 2001 From: Kotresh HR Date: Mon, 29 Jul 2019 18:30:42 +0530 Subject: [PATCH 295/297] ctime/rebalance: Heal ctime xattr on directory during rebalance After add-brick and rebalance, the ctime xattr is not present on rebalanced directories on new brick. This patch fixes the same. Note that ctime still doesn't support consistent time across distribute sub-volume. This patch also fixes the in-memory inconsistency of time attributes when metadata is self healed. Backport of: > Patch: https://review.gluster.org/23127/ > Change-Id: Ia20506f1839021bf61d4753191e7dc34b31bb2df > fixes: bz#1734026 > Signed-off-by: Kotresh HR Change-Id: Ia20506f1839021bf61d4753191e7dc34b31bb2df BUG: 1728673 Signed-off-by: Kotresh HR Reviewed-on: https://code.engineering.redhat.com/gerrit/181105 Tested-by: RHGS Build Bot Reviewed-by: Atin Mukherjee --- tests/basic/afr/split-brain-healing-ctime.t | 253 +++++++++++++++++++++ tests/basic/afr/split-brain-healing.t | 1 + tests/basic/ctime/ctime-ec-heal.t | 71 ++++++ tests/basic/ctime/ctime-ec-rebalance.t | 44 ++++ tests/basic/ctime/ctime-rep-heal.t | 71 ++++++ tests/basic/ctime/ctime-rep-rebalance.t | 42 ++++ .../bug-1734370-entry-heal-restore-time.t | 84 +++++++ tests/volume.rc | 15 +- xlators/cluster/afr/src/afr-self-heal-common.c | 3 +- xlators/cluster/afr/src/afr-self-heal-entry.c | 2 + xlators/cluster/dht/src/dht-common.c | 1 + xlators/cluster/ec/src/ec-heal.c | 7 +- xlators/storage/posix/src/posix-entry-ops.c | 8 +- xlators/storage/posix/src/posix-helpers.c | 31 ++- xlators/storage/posix/src/posix-inode-fd-ops.c | 57 ++--- xlators/storage/posix/src/posix-metadata.c | 65 +++++- xlators/storage/posix/src/posix-metadata.h | 7 + xlators/storage/posix/src/posix.h | 5 +- 18 files changed, 714 insertions(+), 53 deletions(-) create mode 100644 tests/basic/afr/split-brain-healing-ctime.t create mode 100644 tests/basic/ctime/ctime-ec-heal.t create mode 100644 tests/basic/ctime/ctime-ec-rebalance.t create mode 100644 tests/basic/ctime/ctime-rep-heal.t create mode 100644 tests/basic/ctime/ctime-rep-rebalance.t create mode 100644 tests/bugs/replicate/bug-1734370-entry-heal-restore-time.t diff --git a/tests/basic/afr/split-brain-healing-ctime.t b/tests/basic/afr/split-brain-healing-ctime.t new file mode 100644 index 0000000..1ca18e3 --- /dev/null +++ b/tests/basic/afr/split-brain-healing-ctime.t @@ -0,0 +1,253 @@ +#!/bin/bash + +#Test the split-brain resolution CLI commands. +. $(dirname $0)/../../include.rc +. $(dirname $0)/../../volume.rc + +function get_replicate_subvol_number { + local filename=$1 + #get_backend_paths + if [ -f $B0/${V0}1/$filename ] + then + echo 0 + elif [ -f $B0/${V0}3/$filename ] + then echo 1 + else + echo -1 + fi +} + +cleanup; + +AREQUAL_PATH=$(dirname $0)/../../utils +GET_MDATA_PATH=$(dirname $0)/../../utils +CFLAGS="" +test "`uname -s`" != "Linux" && { + CFLAGS="$CFLAGS -lintl"; +} +build_tester $AREQUAL_PATH/arequal-checksum.c $CFLAGS +build_tester $GET_MDATA_PATH/get-mdata-xattr.c + +TEST glusterd +TEST pidof glusterd +TEST $CLI volume create $V0 replica 2 $H0:$B0/${V0}{1,2,3,4} +TEST $CLI volume set $V0 cluster.self-heal-daemon off +TEST $CLI volume set $V0 cluster.data-self-heal off +TEST $CLI volume set $V0 cluster.metadata-self-heal off +TEST $CLI volume set $V0 cluster.entry-self-heal off +TEST $CLI volume set $V0 ctime on +TEST $CLI volume start $V0 +TEST $GFS --volfile-id=/$V0 --volfile-server=$H0 $M0 + +cd $M0 +for i in {1..10} +do + echo "Initial content">>file$i +done + +replica_0_files_list=(`ls $B0/${V0}1|grep -v '^\.'`) +replica_1_files_list=(`ls $B0/${V0}3|grep -v '^\.'`) + +############ Create data split-brain in the files. ########################### +TEST kill_brick $V0 $H0 $B0/${V0}1 +for file in ${!replica_0_files_list[*]} +do + echo "B1 is down">>${replica_0_files_list[$file]} +done +TEST kill_brick $V0 $H0 $B0/${V0}3 +for file in ${!replica_1_files_list[*]} +do + echo "B3 is down">>${replica_1_files_list[$file]} +done + +SMALLER_FILE_SIZE=$(stat -c %s file1) + +TEST $CLI volume start $V0 force +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 0 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 2 + +TEST kill_brick $V0 $H0 $B0/${V0}2 +for file in ${!replica_0_files_list[*]} +do + echo "B2 is down">>${replica_0_files_list[$file]} + echo "appending more content to make it the bigger file">>${replica_0_files_list[$file]} +done +TEST kill_brick $V0 $H0 $B0/${V0}4 +for file in ${!replica_1_files_list[*]} +do + echo "B4 is down">>${replica_1_files_list[$file]} + echo "appending more content to make it the bigger file">>${replica_1_files_list[$file]} +done + +BIGGER_FILE_SIZE=$(stat -c %s file1) +TEST $CLI volume start $V0 force +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 3 + + +############### Acessing the files should now give EIO. ############################### +TEST ! cat file1 +TEST ! cat file2 +TEST ! cat file3 +TEST ! cat file4 +TEST ! cat file5 +TEST ! cat file6 +TEST ! cat file7 +TEST ! cat file8 +TEST ! cat file9 +TEST ! cat file10 +################### +TEST $CLI volume set $V0 cluster.self-heal-daemon on +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "Y" glustershd_up_status +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 0 +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 1 +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 2 +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 3 + +################ Heal file1 using the bigger-file option ############## +$CLI volume heal $V0 split-brain bigger-file /file1 +EXPECT "0" echo $? +EXPECT $BIGGER_FILE_SIZE stat -c %s file1 + +################ Heal file2 using the bigger-file option and its gfid ############## +subvolume=$(get_replicate_subvol_number file2) +if [ $subvolume == 0 ] +then + GFID=$(gf_get_gfid_xattr $B0/${V0}1/file2) +elif [ $subvolume == 1 ] +then + GFID=$(gf_get_gfid_xattr $B0/${V0}3/file2) +fi +GFIDSTR="gfid:$(gf_gfid_xattr_to_str $GFID)" +$CLI volume heal $V0 split-brain bigger-file $GFIDSTR +EXPECT "0" echo $? + +################ Heal file3 using the source-brick option ############## +################ Use the brick having smaller file size as source ####### +subvolume=$(get_replicate_subvol_number file3) +if [ $subvolume == 0 ] +then + $CLI volume heal $V0 split-brain source-brick $H0:$B0/${V0}2 /file3 +elif [ $subvolume == 1 ] +then + $CLI volume heal $V0 split-brain source-brick $H0:$B0/${V0}4 /file3 +fi +EXPECT "0" echo $? +EXPECT $SMALLER_FILE_SIZE stat -c %s file3 + +################ Heal file4 using the source-brick option and it's gfid ############## +################ Use the brick having smaller file size as source ####### +subvolume=$(get_replicate_subvol_number file4) +if [ $subvolume == 0 ] +then + GFID=$(gf_get_gfid_xattr $B0/${V0}1/file4) + GFIDSTR="gfid:$(gf_gfid_xattr_to_str $GFID)" + $CLI volume heal $V0 split-brain source-brick $H0:$B0/${V0}2 $GFIDSTR +elif [ $subvolume == 1 ] +then + GFID=$(gf_get_gfid_xattr $B0/${V0}3/file4) + GFIDSTR="gfid:$(gf_gfid_xattr_to_str $GFID)" + $CLI volume heal $V0 split-brain source-brick $H0:$B0/${V0}4 $GFIDSTR +fi +EXPECT "0" echo $? +EXPECT $SMALLER_FILE_SIZE stat -c %s file4 + +# With ctime enabled, the ctime xattr ("trusted.glusterfs.mdata") gets healed +# as part of metadata heal. So mtime would be same, hence it can't be healed +# using 'latest-mtime' policy, use 'source-brick' option instead. +################ Heal file5 using the source-brick option ############## +subvolume=$(get_replicate_subvol_number file5) +if [ $subvolume == 0 ] +then + $CLI volume heal $V0 split-brain source-brick $H0:$B0/${V0}1 /file5 +elif [ $subvolume == 1 ] +then + $CLI volume heal $V0 split-brain source-brick $H0:$B0/${V0}3 /file5 +fi +EXPECT "0" echo $? + +if [ $subvolume == 0 ] +then + mtime1_after_heal=$(get_mtime $B0/${V0}1/file5) + mtime2_after_heal=$(get_mtime $B0/${V0}2/file5) +elif [ $subvolume == 1 ] +then + mtime1_after_heal=$(get_mtime $B0/${V0}3/file5) + mtime2_after_heal=$(get_mtime $B0/${V0}4/file5) +fi + +#TODO: To below comparisons on full sub-second resolution + +TEST [ $mtime1_after_heal -eq $mtime2_after_heal ] + +mtime_mount_after_heal=$(stat -c %Y file5) + +TEST [ $mtime1_after_heal -eq $mtime_mount_after_heal ] + +################ Heal file6 using the source-brick option and its gfid ############## +subvolume=$(get_replicate_subvol_number file6) +if [ $subvolume == 0 ] +then + GFID=$(gf_get_gfid_xattr $B0/${V0}1/file6) + GFIDSTR="gfid:$(gf_gfid_xattr_to_str $GFID)" + $CLI volume heal $V0 split-brain source-brick $H0:$B0/${V0}1 $GFIDSTR +elif [ $subvolume == 1 ] +then + GFID=$(gf_get_gfid_xattr $B0/${V0}3/file6) + GFIDSTR="gfid:$(gf_gfid_xattr_to_str $GFID)" + $CLI volume heal $V0 split-brain source-brick $H0:$B0/${V0}3 $GFIDSTR +fi +EXPECT "0" echo $? + +if [ $subvolume == 0 ] +then + mtime1_after_heal=$(get_mtime $B0/${V0}1/file6) + mtime2_after_heal=$(get_mtime $B0/${V0}2/file6) +elif [ $subvolume == 1 ] +then + mtime1_after_heal=$(get_mtime $B0/${V0}3/file6) + mtime2_after_heal=$(get_mtime $B0/${V0}4/file6) +fi + +#TODO: To below comparisons on full sub-second resolution + +TEST [ $mtime1_after_heal -eq $mtime2_after_heal ] + +mtime_mount_after_heal=$(stat -c %Y file6) + +TEST [ $mtime1_after_heal -eq $mtime_mount_after_heal ] + +################ Heal remaining SB'ed files of replica_0 using B1 as source ############## +$CLI volume heal $V0 split-brain source-brick $H0:$B0/${V0}1 +EXPECT "0" echo $? + +################ Heal remaining SB'ed files of replica_1 using B3 as source ############## +$CLI volume heal $V0 split-brain source-brick $H0:$B0/${V0}3 +EXPECT "0" echo $? + +############### Reading the files should now succeed. ############################### +TEST cat file1 +TEST cat file2 +TEST cat file3 +TEST cat file4 +TEST cat file5 +TEST cat file6 +TEST cat file7 +TEST cat file8 +TEST cat file9 +TEST cat file10 + +################ File contents on the bricks must be same. ################################ +TEST diff <(arequal-checksum -p $B0/$V01 -i .glusterfs) <(arequal-checksum -p $B0/$V02 -i .glusterfs) +TEST diff <(arequal-checksum -p $B0/$V03 -i .glusterfs) <(arequal-checksum -p $B0/$V04 -i .glusterfs) + +############### Trying to heal files not in SB should fail. ############################### +$CLI volume heal $V0 split-brain bigger-file /file1 +EXPECT "1" echo $? +$CLI volume heal $V0 split-brain source-brick $H0:$B0/${V0}4 /file3 +EXPECT "1" echo $? + +cd - +TEST rm $AREQUAL_PATH/arequal-checksum +TEST rm $GET_MDATA_PATH/get-mdata-xattr +cleanup diff --git a/tests/basic/afr/split-brain-healing.t b/tests/basic/afr/split-brain-healing.t index 78553e6..315e815 100644 --- a/tests/basic/afr/split-brain-healing.t +++ b/tests/basic/afr/split-brain-healing.t @@ -35,6 +35,7 @@ TEST $CLI volume set $V0 cluster.self-heal-daemon off TEST $CLI volume set $V0 cluster.data-self-heal off TEST $CLI volume set $V0 cluster.metadata-self-heal off TEST $CLI volume set $V0 cluster.entry-self-heal off +TEST $CLI volume set $V0 ctime off TEST $CLI volume start $V0 TEST $GFS --volfile-id=/$V0 --volfile-server=$H0 $M0 diff --git a/tests/basic/ctime/ctime-ec-heal.t b/tests/basic/ctime/ctime-ec-heal.t new file mode 100644 index 0000000..1cb4516 --- /dev/null +++ b/tests/basic/ctime/ctime-ec-heal.t @@ -0,0 +1,71 @@ +#!/bin/bash +# +# This will test self healing of ctime xattr 'trusted.glusterfs.mdata' +# +### + +. $(dirname $0)/../../include.rc +. $(dirname $0)/../../volume.rc +. $(dirname $0)/../../afr.rc + +cleanup + +#cleate and start volume +TEST glusterd +TEST pidof glusterd +TEST $CLI volume create $V0 disperse 3 redundancy 1 $H0:$B0/${V0}{1..3} +TEST $CLI volume set $V0 ctime on +TEST $CLI volume start $V0 + +#Mount the volume +TEST $GFS --volfile-id=/$V0 --volfile-server=$H0 $M0; + +# Create files +mkdir $M0/dir1 +echo "Initial content" > $M0/file1 + +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/file1 + +# Kill brick +TEST kill_brick $V0 $H0 $B0/${V0}3 + +echo "B3 is down" >> $M0/file1 +echo "Change dir1 time attributes" > $M0/dir1/dir1_file1 +echo "Entry heal file" > $M0/entry_heal_file1 +mkdir $M0/entry_heal_dir1 + +# Check xattr +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '2' get_mdata_uniq_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '2' get_mdata_uniq_count $B0/${V0}{1..3}/file1 + +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '2' get_mdata_count $B0/${V0}{1..3}/dir1/dir1_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/dir1/dir1_file1 + +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '2' get_mdata_count $B0/${V0}{1..3}/entry_heal_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/entry_heal_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '2' get_mdata_count $B0/${V0}{1..3}/entry_heal_dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/entry_heal_dir1 + +TEST $CLI volume start $V0 force +$CLI volume heal $V0 + +# Check xattr +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/file1 + +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/dir1/dir1_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/dir1/dir1_file1 + +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/entry_heal_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/entry_heal_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/entry_heal_dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/entry_heal_dir1 + +cleanup; diff --git a/tests/basic/ctime/ctime-ec-rebalance.t b/tests/basic/ctime/ctime-ec-rebalance.t new file mode 100644 index 0000000..caccdc1 --- /dev/null +++ b/tests/basic/ctime/ctime-ec-rebalance.t @@ -0,0 +1,44 @@ +#!/bin/bash +# +# This will test healing of ctime xattr 'trusted.glusterfs.mdata' after add-brick and rebalance +# +### + +. $(dirname $0)/../../include.rc +. $(dirname $0)/../../volume.rc +. $(dirname $0)/../../fallocate.rc + +cleanup + +#cleate and start volume +TEST glusterd +TEST pidof glusterd +TEST $CLI volume create $V0 disperse 3 redundancy 1 $H0:$B0/${V0}{0..5} +TEST $CLI volume set $V0 ctime on +TEST $CLI volume start $V0 + +#Mount the volume +TEST $GFS --volfile-id=/$V0 --volfile-server=$H0 $M0; +EXPECT_WITHIN $CHILD_UP_TIMEOUT "3" ec_child_up_count $V0 0 + +# Create files +mkdir $M0/dir1 +echo "test data" > $M0/dir1/file1 + +# Add brick +TEST $CLI volume add-brick $V0 $H0:$B0/${V0}{6..8} + +#Trigger rebalance +TEST $CLI volume rebalance $V0 start force +EXPECT_WITHIN $REBALANCE_TIMEOUT "completed" rebalance_status_field $V0 + +#Verify ctime xattr heal on directory +EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}6/dir1" +EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}7/dir1" +EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}8/dir1" + +b6_mdata=$(get_mdata "$B0/${V0}6/dir1") +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "${b6_mdata}" get_mdata $B0/${V0}7/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "${b6_mdata}" get_mdata $B0/${V0}8/dir1 + +cleanup; diff --git a/tests/basic/ctime/ctime-rep-heal.t b/tests/basic/ctime/ctime-rep-heal.t new file mode 100644 index 0000000..ba8b08a --- /dev/null +++ b/tests/basic/ctime/ctime-rep-heal.t @@ -0,0 +1,71 @@ +#!/bin/bash +# +# This will test self healing of ctime xattr 'trusted.glusterfs.mdata' +# +### + +. $(dirname $0)/../../include.rc +. $(dirname $0)/../../volume.rc +. $(dirname $0)/../../afr.rc + +cleanup + +#cleate and start volume +TEST glusterd +TEST pidof glusterd +TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}{1..3} +TEST $CLI volume set $V0 ctime on +TEST $CLI volume start $V0 + +#Mount the volume +TEST $GFS --volfile-id=/$V0 --volfile-server=$H0 $M0; + +# Create files +mkdir $M0/dir1 +echo "Initial content" > $M0/file1 + +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/file1 + +# Kill brick +TEST kill_brick $V0 $H0 $B0/${V0}3 + +echo "B3 is down" >> $M0/file1 +echo "Change dir1 time attributes" > $M0/dir1/dir1_file1 +echo "Entry heal file" > $M0/entry_heal_file1 +mkdir $M0/entry_heal_dir1 + +# Check xattr +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '2' get_mdata_uniq_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '2' get_mdata_uniq_count $B0/${V0}{1..3}/file1 + +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '2' get_mdata_count $B0/${V0}{1..3}/dir1/dir1_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/dir1/dir1_file1 + +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '2' get_mdata_count $B0/${V0}{1..3}/entry_heal_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/entry_heal_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '2' get_mdata_count $B0/${V0}{1..3}/entry_heal_dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/entry_heal_dir1 + +TEST $CLI volume start $V0 force +$CLI volume heal $V0 + +# Check xattr +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/file1 + +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/dir1/dir1_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/dir1/dir1_file1 + +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/entry_heal_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/entry_heal_file1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '3' get_mdata_count $B0/${V0}{1..3}/entry_heal_dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT '1' get_mdata_uniq_count $B0/${V0}{1..3}/entry_heal_dir1 + +cleanup; diff --git a/tests/basic/ctime/ctime-rep-rebalance.t b/tests/basic/ctime/ctime-rep-rebalance.t new file mode 100644 index 0000000..dd9743e --- /dev/null +++ b/tests/basic/ctime/ctime-rep-rebalance.t @@ -0,0 +1,42 @@ +#!/bin/bash +# +# This will test healing of ctime xattr 'trusted.glusterfs.mdata' after add-brick and rebalance +# +### + +. $(dirname $0)/../../include.rc +. $(dirname $0)/../../volume.rc +. $(dirname $0)/../../afr.rc + +cleanup + +#cleate and start volume +TEST glusterd +TEST pidof glusterd +TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}{0..5} +TEST $CLI volume set $V0 ctime on +TEST $CLI volume start $V0 + +#Mount the volume +TEST $GFS --volfile-id=/$V0 --volfile-server=$H0 $M0; + +# Create files +mkdir $M0/dir1 + +# Add brick +TEST $CLI volume add-brick $V0 $H0:$B0/${V0}{6..8} + +#Trigger rebalance +TEST $CLI volume rebalance $V0 start force +EXPECT_WITHIN $REBALANCE_TIMEOUT "completed" rebalance_status_field $V0 + +#Verify ctime xattr heal on directory +EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}6/dir1" +EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}7/dir1" +EXPECT_WITHIN $PROCESS_UP_TIMEOUT 'trusted.glusterfs.mdata' check_for_xattr 'trusted.glusterfs.mdata' "$B0/${V0}8/dir1" + +b6_mdata=$(get_mdata "$B0/${V0}6/dir1") +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "${b6_mdata}" get_mdata $B0/${V0}7/dir1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "${b6_mdata}" get_mdata $B0/${V0}8/dir1 + +cleanup; diff --git a/tests/bugs/replicate/bug-1734370-entry-heal-restore-time.t b/tests/bugs/replicate/bug-1734370-entry-heal-restore-time.t new file mode 100644 index 0000000..298d6ed --- /dev/null +++ b/tests/bugs/replicate/bug-1734370-entry-heal-restore-time.t @@ -0,0 +1,84 @@ +#!/bin/bash + +. $(dirname $0)/../../include.rc +. $(dirname $0)/../../volume.rc +. $(dirname $0)/../../afr.rc + +cleanup; + +function time_stamps_match { + path=$1 + mtime_source_b0=$(get_mtime $B0/${V0}0/$path) + atime_source_b0=$(get_atime $B0/${V0}0/$path) + mtime_source_b2=$(get_mtime $B0/${V0}2/$path) + atime_source_b2=$(get_atime $B0/${V0}2/$path) + mtime_sink_b1=$(get_mtime $B0/${V0}1/$path) + atime_sink_b1=$(get_atime $B0/${V0}1/$path) + + #The same brick must be the source of heal for both atime and mtime. + if [[ ( $mtime_source_b0 -eq $mtime_sink_b1 && $atime_source_b0 -eq $atime_sink_b1 ) || \ + ( $mtime_source_b2 -eq $mtime_sink_b1 && $atime_source_b2 -eq $atime_sink_b1 ) ]] + then + echo "Y" + else + echo "N" + fi + +} + +# Test that the parent dir's timestamps are restored during entry-heal. +GET_MDATA_PATH=$(dirname $0)/../../utils +build_tester $GET_MDATA_PATH/get-mdata-xattr.c + +TEST glusterd; +TEST pidof glusterd; +TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}{0,1,2}; +TEST $CLI volume set $V0 ctime on +TEST $CLI volume start $V0; + +TEST $GFS --volfile-id=/$V0 --volfile-server=$H0 --attribute-timeout=0 --entry-timeout=0 $M0 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 0 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 2 + +############################################################################### +TEST mkdir $M0/DIR +TEST kill_brick $V0 $H0 $B0/${V0}1 +TEST touch $M0/DIR/FILE + +TEST $CLI volume start $V0 force +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 1 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "Y" glustershd_up_status +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 0 +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 1 +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 2 +TEST $CLI volume heal $V0 +EXPECT_WITHIN $HEAL_TIMEOUT "0" get_pending_heal_count $V0 + +EXPECT "Y" time_stamps_match DIR +ctime_source1=$(get_ctime $B0/${V0}0/$path) +ctime_source2=$(get_ctime $B0/${V0}2/$path) +ctime_sink=$(get_ctime $B0/${V0}1/$path) +TEST [ $ctime_source1 -eq $ctime_sink ] +TEST [ $ctime_source2 -eq $ctime_sink ] + +############################################################################### +# Repeat the test with ctime feature disabled. +TEST $CLI volume set $V0 features.ctime off +TEST mkdir $M0/DIR2 +TEST kill_brick $V0 $H0 $B0/${V0}1 +TEST touch $M0/DIR2/FILE + +TEST $CLI volume start $V0 force +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 0 +EXPECT_WITHIN $PROCESS_UP_TIMEOUT "Y" glustershd_up_status +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 0 +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 1 +EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 2 +TEST $CLI volume heal $V0 +EXPECT_WITHIN $HEAL_TIMEOUT "0" get_pending_heal_count $V0 + +EXPECT "Y" time_stamps_match DIR2 + +TEST rm $GET_MDATA_PATH/get-mdata-xattr +cleanup; diff --git a/tests/volume.rc b/tests/volume.rc index 76a8fd4..9a002d9 100644 --- a/tests/volume.rc +++ b/tests/volume.rc @@ -371,6 +371,19 @@ function get_gfid2path { getfattr -h --only-values -n glusterfs.gfidtopath $path 2>/dev/null } +function get_mdata { + local path=$1 + getfattr -h -e hex -n trusted.glusterfs.mdata $path 2>/dev/null | grep "trusted.glusterfs.mdata" | cut -f2 -d'=' +} + +function get_mdata_count { + getfattr -d -m . -e hex $@ 2>/dev/null | grep mdata | wc -l +} + +function get_mdata_uniq_count { + getfattr -d -m . -e hex $@ 2>/dev/null | grep mdata | uniq | wc -l +} + function get_xattr_key { local key=$1 local path=$2 @@ -925,7 +938,7 @@ function get_ctime { local time=$(get-mdata-xattr -c $1) if [ $time == "-1" ]; then - echo $(stat -c %Z $2) + echo $(stat -c %Z $1) else echo $time fi diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c index b38085a..81ef38a 100644 --- a/xlators/cluster/afr/src/afr-self-heal-common.c +++ b/xlators/cluster/afr/src/afr-self-heal-common.c @@ -513,7 +513,8 @@ afr_selfheal_restore_time(call_frame_t *frame, xlator_t *this, inode_t *inode, AFR_ONLIST(healed_sinks, frame, afr_sh_generic_fop_cbk, setattr, &loc, &replies[source].poststat, - (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME), NULL); + (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME | GF_SET_ATTR_CTIME), + NULL); loc_wipe(&loc); diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c index e07b521..35b600f 100644 --- a/xlators/cluster/afr/src/afr-self-heal-entry.c +++ b/xlators/cluster/afr/src/afr-self-heal-entry.c @@ -1032,6 +1032,8 @@ unlock: goto postop_unlock; } + afr_selfheal_restore_time(frame, this, fd->inode, source, healed_sinks, + locked_replies); ret = afr_selfheal_undo_pending( frame, this, fd->inode, sources, sinks, healed_sinks, undid_pending, AFR_ENTRY_TRANSACTION, locked_replies, postop_lock); diff --git a/xlators/cluster/dht/src/dht-common.c b/xlators/cluster/dht/src/dht-common.c index 219b072..99cccd6 100644 --- a/xlators/cluster/dht/src/dht-common.c +++ b/xlators/cluster/dht/src/dht-common.c @@ -115,6 +115,7 @@ char *xattrs_to_heal[] = {"user.", QUOTA_LIMIT_KEY, QUOTA_LIMIT_OBJECTS_KEY, GF_SELINUX_XATTR_KEY, + GF_XATTR_MDATA_KEY, NULL}; char *dht_dbg_vxattrs[] = {DHT_DBG_HASHED_SUBVOL_PATTERN, NULL}; diff --git a/xlators/cluster/ec/src/ec-heal.c b/xlators/cluster/ec/src/ec-heal.c index 0f0f398..06a7016 100644 --- a/xlators/cluster/ec/src/ec-heal.c +++ b/xlators/cluster/ec/src/ec-heal.c @@ -2301,9 +2301,10 @@ ec_restore_time_and_adjust_versions(call_frame_t *frame, ec_t *ec, fd_t *fd, loc.inode = inode_ref(fd->inode); gf_uuid_copy(loc.gfid, fd->inode->gfid); - ret = cluster_setattr(ec->xl_list, healed_sinks, ec->nodes, replies, - output, frame, ec->xl, &loc, &source_buf, - GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME, NULL); + ret = cluster_setattr( + ec->xl_list, healed_sinks, ec->nodes, replies, output, frame, + ec->xl, &loc, &source_buf, + GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME | GF_SET_ATTR_CTIME, NULL); EC_INTERSECT(healed_sinks, healed_sinks, output, ec->nodes); if (EC_COUNT(healed_sinks, ec->nodes) == 0) { ret = -ENOTCONN; diff --git a/xlators/storage/posix/src/posix-entry-ops.c b/xlators/storage/posix/src/posix-entry-ops.c index 34ee2b8..283b305 100644 --- a/xlators/storage/posix/src/posix-entry-ops.c +++ b/xlators/storage/posix/src/posix-entry-ops.c @@ -500,7 +500,7 @@ post_op: posix_set_gfid2path_xattr(this, real_path, loc->pargfid, loc->name); } - op_ret = posix_entry_create_xattr_set(this, real_path, xdata); + op_ret = posix_entry_create_xattr_set(this, loc, real_path, xdata); if (op_ret) { if (errno != EEXIST) gf_msg(this->name, GF_LOG_ERROR, errno, P_MSG_XATTR_FAILED, @@ -828,7 +828,7 @@ posix_mkdir(call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode, "setting ACLs on %s failed ", real_path); } - op_ret = posix_entry_create_xattr_set(this, real_path, xdata); + op_ret = posix_entry_create_xattr_set(this, loc, real_path, xdata); if (op_ret) { gf_msg(this->name, GF_LOG_ERROR, errno, P_MSG_XATTR_FAILED, "setting xattrs on %s failed", real_path); @@ -1529,7 +1529,7 @@ posix_symlink(call_frame_t *frame, xlator_t *this, const char *linkname, } ignore: - op_ret = posix_entry_create_xattr_set(this, real_path, xdata); + op_ret = posix_entry_create_xattr_set(this, loc, real_path, xdata); if (op_ret) { gf_msg(this->name, GF_LOG_ERROR, errno, P_MSG_XATTR_FAILED, "setting xattrs on %s failed ", real_path); @@ -2167,7 +2167,7 @@ posix_create(call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags, posix_set_gfid2path_xattr(this, real_path, loc->pargfid, loc->name); } ignore: - op_ret = posix_entry_create_xattr_set(this, real_path, xdata); + op_ret = posix_entry_create_xattr_set(this, loc, real_path, xdata); if (op_ret) { gf_msg(this->name, GF_LOG_ERROR, errno, P_MSG_XATTR_FAILED, "setting xattrs on %s failed ", real_path); diff --git a/xlators/storage/posix/src/posix-helpers.c b/xlators/storage/posix/src/posix-helpers.c index d143d4c..6a1a35c 100644 --- a/xlators/storage/posix/src/posix-helpers.c +++ b/xlators/storage/posix/src/posix-helpers.c @@ -1188,11 +1188,15 @@ posix_dump_buffer(xlator_t *this, const char *real_path, const char *key, #endif int -posix_handle_pair(xlator_t *this, const char *real_path, char *key, +posix_handle_pair(xlator_t *this, loc_t *loc, const char *real_path, char *key, data_t *value, int flags, struct iatt *stbuf) { int sys_ret = -1; int ret = 0; + int op_errno = 0; + struct mdata_iatt mdata_iatt = { + 0, + }; #ifdef GF_DARWIN_HOST_OS const int error_code = EINVAL; #else @@ -1216,6 +1220,23 @@ posix_handle_pair(xlator_t *this, const char *real_path, char *key, /* ignore this key value pair */ ret = 0; goto out; + } else if (!strncmp(key, GF_XATTR_MDATA_KEY, strlen(key))) { + /* This is either by rebalance or self heal. Create the xattr if it's + * not present. Compare and update the larger value if the xattr is + * already present. + */ + if (loc == NULL) { + ret = -EINVAL; + goto out; + } + posix_mdata_iatt_from_disk(&mdata_iatt, + (posix_mdata_disk_t *)value->data); + ret = posix_set_mdata_xattr_legacy_files(this, loc->inode, real_path, + &mdata_iatt, &op_errno); + if (ret != 0) { + ret = -op_errno; + } + goto out; } else { sys_ret = sys_lsetxattr(real_path, key, value->data, value->len, flags); #ifdef GF_DARWIN_HOST_OS @@ -1810,8 +1831,8 @@ _handle_entry_create_keyvalue_pair(dict_t *d, char *k, data_t *v, void *tmp) return 0; } - ret = posix_handle_pair(filler->this, filler->real_path, k, v, XATTR_CREATE, - filler->stbuf); + ret = posix_handle_pair(filler->this, filler->loc, filler->real_path, k, v, + XATTR_CREATE, filler->stbuf); if (ret < 0) { errno = -ret; return -1; @@ -1820,7 +1841,8 @@ _handle_entry_create_keyvalue_pair(dict_t *d, char *k, data_t *v, void *tmp) } int -posix_entry_create_xattr_set(xlator_t *this, const char *path, dict_t *dict) +posix_entry_create_xattr_set(xlator_t *this, loc_t *loc, const char *path, + dict_t *dict) { int ret = -1; @@ -1834,6 +1856,7 @@ posix_entry_create_xattr_set(xlator_t *this, const char *path, dict_t *dict) filler.this = this; filler.real_path = path; filler.stbuf = NULL; + filler.loc = loc; ret = dict_foreach(dict, _handle_entry_create_keyvalue_pair, &filler); diff --git a/xlators/storage/posix/src/posix-inode-fd-ops.c b/xlators/storage/posix/src/posix-inode-fd-ops.c index e0ea85b..a2a518f 100644 --- a/xlators/storage/posix/src/posix-inode-fd-ops.c +++ b/xlators/storage/posix/src/posix-inode-fd-ops.c @@ -429,22 +429,9 @@ posix_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, &frame->root->ctime, stbuf, valid); } - if (valid & GF_SET_ATTR_CTIME && !priv->ctime) { - /* - * If ctime is not enabled, we have no means to associate an - * arbitrary ctime with the file, so as a fallback, we ignore - * the ctime payload and update the file ctime to current time - * (which is possible directly with the POSIX API). - */ - op_ret = PATH_SET_TIMESPEC_OR_TIMEVAL(real_path, NULL); - if (op_ret == -1) { - op_errno = errno; - gf_msg(this->name, GF_LOG_ERROR, errno, P_MSG_UTIMES_FAILED, - "setattr (utimes) on %s " - "failed", - real_path); - goto out; - } + if ((valid & GF_SET_ATTR_CTIME) && priv->ctime) { + posix_update_ctime_in_mdata(this, real_path, -1, loc->inode, + &frame->root->ctime, stbuf, valid); } if (!valid) { @@ -469,14 +456,6 @@ posix_setattr(call_frame_t *frame, xlator_t *this, loc_t *loc, goto out; } - if (valid & GF_SET_ATTR_CTIME && priv->ctime) { - /* - * If we got ctime payload, we override - * the ctime of statpost with that. - */ - statpost.ia_ctime = stbuf->ia_ctime; - statpost.ia_ctime_nsec = stbuf->ia_ctime_nsec; - } posix_set_ctime(frame, this, real_path, -1, loc->inode, &statpost); if (xdata) @@ -592,6 +571,7 @@ posix_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt statpost = { 0, }; + struct posix_private *priv = NULL; struct posix_fd *pfd = NULL; dict_t *xattr_rsp = NULL; int32_t ret = -1; @@ -604,6 +584,9 @@ posix_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, VALIDATE_OR_GOTO(this, out); VALIDATE_OR_GOTO(fd, out); + priv = this->private; + VALIDATE_OR_GOTO(priv, out); + ret = posix_fd_ctx_get(fd, this, &pfd, &op_errno); if (ret < 0) { gf_msg_debug(this->name, 0, "pfd is NULL from fd=%p", fd); @@ -656,6 +639,11 @@ posix_fsetattr(call_frame_t *frame, xlator_t *this, fd_t *fd, &frame->root->ctime, stbuf, valid); } + if ((valid & GF_SET_ATTR_CTIME) && priv->ctime) { + posix_update_ctime_in_mdata(this, NULL, pfd->fd, fd->inode, + &frame->root->ctime, stbuf, valid); + } + if (!valid) { op_ret = sys_fchown(pfd->fd, -1, -1); if (op_ret == -1) { @@ -2578,7 +2566,7 @@ _handle_setxattr_keyvalue_pair(dict_t *d, char *k, data_t *v, void *tmp) filler = tmp; - return posix_handle_pair(filler->this, filler->real_path, k, v, + return posix_handle_pair(filler->this, filler->loc, filler->real_path, k, v, filler->flags, filler->stbuf); } @@ -2641,27 +2629,27 @@ posix_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict, priv = this->private; DISK_SPACE_CHECK_AND_GOTO(frame, priv, xdata, op_ret, op_errno, out); + MAKE_INODE_HANDLE(real_path, this, loc, NULL); + if (!real_path) { + op_ret = -1; + op_errno = ESTALE; + goto out; + } + ret = dict_get_mdata(dict, CTIME_MDATA_XDATA_KEY, &mdata_iatt); if (ret == 0) { /* This is initiated by lookup when ctime feature is enabled to create * "trusted.glusterfs.mdata" xattr if not present. These are the files * which were created when ctime feature is disabled. */ - ret = posix_set_mdata_xattr_legacy_files(this, loc->inode, &mdata_iatt, - &op_errno); + ret = posix_set_mdata_xattr_legacy_files(this, loc->inode, real_path, + &mdata_iatt, &op_errno); if (ret != 0) { op_ret = -1; } goto out; } - MAKE_INODE_HANDLE(real_path, this, loc, NULL); - if (!real_path) { - op_ret = -1; - op_errno = ESTALE; - goto out; - } - posix_pstat(this, loc->inode, loc->gfid, real_path, &preop, _gf_false); op_ret = -1; @@ -2796,6 +2784,7 @@ posix_setxattr(call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict, filler.real_path = real_path; filler.this = this; filler.stbuf = &preop; + filler.loc = loc; #ifdef GF_DARWIN_HOST_OS filler.flags = map_xattr_flags(flags); diff --git a/xlators/storage/posix/src/posix-metadata.c b/xlators/storage/posix/src/posix-metadata.c index 532daa2..9efaf99 100644 --- a/xlators/storage/posix/src/posix-metadata.c +++ b/xlators/storage/posix/src/posix-metadata.c @@ -56,6 +56,19 @@ posix_mdata_from_disk(posix_mdata_t *out, posix_mdata_disk_t *in) out->atime.tv_nsec = be64toh(in->atime.tv_nsec); } +void +posix_mdata_iatt_from_disk(struct mdata_iatt *out, posix_mdata_disk_t *in) +{ + out->ia_ctime = be64toh(in->ctime.tv_sec); + out->ia_ctime_nsec = be64toh(in->ctime.tv_nsec); + + out->ia_mtime = be64toh(in->mtime.tv_sec); + out->ia_mtime_nsec = be64toh(in->mtime.tv_nsec); + + out->ia_atime = be64toh(in->atime.tv_sec); + out->ia_atime_nsec = be64toh(in->atime.tv_nsec); +} + /* posix_fetch_mdata_xattr fetches the posix_mdata_t from disk */ static int posix_fetch_mdata_xattr(xlator_t *this, const char *real_path_arg, int _fd, @@ -341,6 +354,7 @@ posix_compare_timespec(struct timespec *first, struct timespec *second) int posix_set_mdata_xattr_legacy_files(xlator_t *this, inode_t *inode, + const char *realpath, struct mdata_iatt *mdata_iatt, int *op_errno) { posix_mdata_t *mdata = NULL; @@ -369,8 +383,8 @@ posix_set_mdata_xattr_legacy_files(xlator_t *this, inode_t *inode, goto unlock; } - ret = posix_fetch_mdata_xattr(this, NULL, -1, inode, (void *)mdata, - op_errno); + ret = posix_fetch_mdata_xattr(this, realpath, -1, inode, + (void *)mdata, op_errno); if (ret == 0) { /* Got mdata from disk. This is a race, another client * has healed the xattr during lookup. So set it in inode @@ -412,7 +426,7 @@ posix_set_mdata_xattr_legacy_files(xlator_t *this, inode_t *inode, } } - ret = posix_store_mdata_xattr(this, NULL, -1, inode, mdata); + ret = posix_store_mdata_xattr(this, realpath, -1, inode, mdata); if (ret) { gf_msg(this->name, GF_LOG_ERROR, errno, P_MSG_STOREMDATA_FAILED, "gfid: %s key:%s ", uuid_utoa(inode->gfid), @@ -445,7 +459,8 @@ posix_set_mdata_xattr(xlator_t *this, const char *real_path, int fd, GF_VALIDATE_OR_GOTO(this->name, inode, out); GF_VALIDATE_OR_GOTO(this->name, time, out); - if (update_utime && (!u_atime || !u_mtime)) { + if (update_utime && (flag->ctime && !time) && (flag->atime && !u_atime) && + (flag->mtime && !u_mtime)) { goto out; } @@ -652,6 +667,48 @@ posix_update_utime_in_mdata(xlator_t *this, const char *real_path, int fd, return; } +/* posix_update_ctime_in_mdata updates the posix_mdata_t when ctime needs + * to be modified + */ +void +posix_update_ctime_in_mdata(xlator_t *this, const char *real_path, int fd, + inode_t *inode, struct timespec *ctime, + struct iatt *stbuf, int valid) +{ + int32_t ret = 0; +#if defined(HAVE_UTIMENSAT) + struct timespec tv_ctime = { + 0, + }; +#else + struct timeval tv_ctime = { + 0, + }; +#endif + posix_mdata_flag_t flag = { + 0, + }; + + struct posix_private *priv = NULL; + priv = this->private; + + if (inode && priv->ctime) { + tv_ctime.tv_sec = stbuf->ia_ctime; + SET_TIMESPEC_NSEC_OR_TIMEVAL_USEC(tv_ctime, stbuf->ia_ctime_nsec); + flag.ctime = 1; + + ret = posix_set_mdata_xattr(this, real_path, -1, inode, &tv_ctime, NULL, + NULL, NULL, &flag, _gf_true); + if (ret) { + gf_msg(this->name, GF_LOG_WARNING, errno, P_MSG_SETMDATA_FAILED, + "posix set mdata atime failed on file:" + " %s gfid:%s", + real_path, uuid_utoa(inode->gfid)); + } + } + return; +} + static void posix_get_mdata_flag(uint64_t flags, posix_mdata_flag_t *flag) { diff --git a/xlators/storage/posix/src/posix-metadata.h b/xlators/storage/posix/src/posix-metadata.h index c176699..63e8771 100644 --- a/xlators/storage/posix/src/posix-metadata.h +++ b/xlators/storage/posix/src/posix-metadata.h @@ -43,6 +43,10 @@ posix_update_utime_in_mdata(xlator_t *this, const char *real_path, int fd, inode_t *inode, struct timespec *ctime, struct iatt *stbuf, int valid); void +posix_update_ctime_in_mdata(xlator_t *this, const char *real_path, int fd, + inode_t *inode, struct timespec *ctime, + struct iatt *stbuf, int valid); +void posix_set_ctime(call_frame_t *frame, xlator_t *this, const char *real_path, int fd, inode_t *inode, struct iatt *stbuf); void @@ -56,7 +60,10 @@ posix_set_ctime_cfr(call_frame_t *frame, xlator_t *this, int fd_out, inode_t *inode_out, struct iatt *stbuf_out); int posix_set_mdata_xattr_legacy_files(xlator_t *this, inode_t *inode, + const char *realpath, struct mdata_iatt *mdata_iatt, int *op_errno); +void +posix_mdata_iatt_from_disk(struct mdata_iatt *out, posix_mdata_disk_t *in); #endif /* _POSIX_METADATA_H */ diff --git a/xlators/storage/posix/src/posix.h b/xlators/storage/posix/src/posix.h index 64288a7..dd51062 100644 --- a/xlators/storage/posix/src/posix.h +++ b/xlators/storage/posix/src/posix.h @@ -339,7 +339,7 @@ dict_t * posix_xattr_fill(xlator_t *this, const char *path, loc_t *loc, fd_t *fd, int fdnum, dict_t *xattr, struct iatt *buf); int -posix_handle_pair(xlator_t *this, const char *real_path, char *key, +posix_handle_pair(xlator_t *this, loc_t *loc, const char *real_path, char *key, data_t *value, int flags, struct iatt *stbuf); int posix_fhandle_pair(call_frame_t *frame, xlator_t *this, int fd, char *key, @@ -352,7 +352,8 @@ int posix_gfid_heal(xlator_t *this, const char *path, loc_t *loc, dict_t *xattr_req); int -posix_entry_create_xattr_set(xlator_t *this, const char *path, dict_t *dict); +posix_entry_create_xattr_set(xlator_t *this, loc_t *loc, const char *path, + dict_t *dict); int posix_fd_ctx_get(fd_t *fd, xlator_t *this, struct posix_fd **pfd, -- 1.8.3.1