diff -Nru ceph-10.1.1/AUTHORS ceph-10.1.2/AUTHORS --- ceph-10.1.1/AUTHORS 2016-04-06 14:21:23.000000000 +0000 +++ ceph-10.1.2/AUTHORS 2016-04-12 17:54:23.000000000 +0000 @@ -2,6 +2,7 @@ Abhishek Dixit Abhishek Lekshmanan Abhishek Lekshmanan +Abhishek Lekshmanan Abhishek Varshney Accela Zhao Adam C. Emerson @@ -228,7 +229,7 @@ Jeff Epstein Jeffrey Lu Jeff Weber -Jenkins Build Slave User +Jenkins Build Slave User Jenkins Jens-Christian Fischer Jeremy Qian @@ -289,6 +290,7 @@ Kim Vandry Kiseleva Alyona Kongming Wu +Kris Jurka Krzysztof Kosiński Kuan Kai Chiu Kun Huang @@ -417,6 +419,7 @@ Ruben Kerkhof Ruifeng Yang runsisi +runsisi Rust Shen Rutger ter Borg Sage Weil @@ -430,7 +433,6 @@ Sandon Van Ness Sangdi Xu Sarthak Munshi -scienceluo Scott A. Brandt Scott Devoid Sean Channel @@ -516,6 +518,7 @@ Wei Luo Wei Luo Wei Qian +weiqiaomiao Wenjun Huang Wesley Spikes Wido den Hollander diff -Nru ceph-10.1.1/ceph.spec ceph-10.1.2/ceph.spec --- ceph-10.1.1/ceph.spec 2016-04-06 14:21:00.000000000 +0000 +++ ceph-10.1.2/ceph.spec 2016-04-12 17:53:58.000000000 +0000 @@ -74,7 +74,7 @@ # common ################################################################################# Name: ceph -Version: 10.1.1 +Version: 10.1.2 Release: 0%{?dist} Epoch: 1 Summary: User space components of the Ceph file system diff -Nru ceph-10.1.1/ChangeLog ceph-10.1.2/ChangeLog --- ceph-10.1.1/ChangeLog 2016-04-06 14:21:23.000000000 +0000 +++ ceph-10.1.2/ChangeLog 2016-04-12 17:54:24.000000000 +0000 @@ -1,5 +1,107 @@ -ce50389 (HEAD, tag: v10.1.1, origin/jewel) 10.1.1 +4a2a6f7 (HEAD, tag: v10.1.2, origin/jewel) 10.1.2 +8b98556 PG: set epoch_created and parent_split_bits for child pg +bd1c548 test: fix ut test failure caused by lfn change +45219e0 Fixed ceph-common install. +fd2f455 mds: take standby_for_fscid into account in FSMap::find_unused() +b6d8c32 librbd: Fixed bug in disabling non-primary image mirroring +14a66f6 ceph-disk: fix PrepareData.set_type() args should be class member. +4c203b3 ceph-disk: fix spelling mistake geattr to getattr. +5b098ca ceph-disk: fix lockbox set_or_create_partition() +b7708da radosgw-admin: fix name checking +770846b radosgw-admin: allow setting zone when there is no realm +49886d5 check-generated.sh: can't source bash from sh +8bc8085 tests: add Ubuntu 16.04 xenial dockerfile +0e4a92e crush: fix typo +0a622e6 doc: rgw admin uses "region list" not "regions list" +3c77292 journal: fix final result for JournalTrimmer::C_RemoveSet +3b54d5d test/rados/misc.cc: add long locator key and namespace tests +cb03d4d LFNIndex: use chain_getxattr_buf +c7db303 chain_xattr: add chain_getxattr_buf +755c685 LFNIndex::lfn_parse_object_name: return int rather than bool +e4916f6 LFNIndex::list_objects: lfn_translate does not set errno +25f937e FileStore::set_xattr_limits_via_conf: add warning if max xattr size smaller than max name +18b9f95 test/objectstore/chain_xattr.cc: add test for ensure_single_attr +73778f1 LFNIndex: ensure that lfn and directory attrs are written atomically +ac750ce chain_[f]getxattr: always use size, no reaon to consider CHAIN_XATTR_MAX_BLOCK_LEN +8dc0330 chain_xattr: s/onechunk/skip_chain_cleanup, add ensure_single_attr +21487fd os/,osd/: restructure the rados name length check +8770043 ceph-disk: fix set_data_partition() when data is partition. +a330078 rgw-admin: fix period delete error message +3320f8f rgw-admin: remove unused iterator +64a8a6a rbd-mirror: fixed bug that caused infinite loop when disabling image mirroring +a651598 mailmap: Luo Kexue name normalization +c36c5d4 mailmap: Ning Yao affiliation +eb536bb mailmap: Vitja Makarov affiliation +323276e release-notes: v10.1.1 release notes +18d06ef release-notes: v10.1.1 release notes (draft) +1be8dba mailmap: sunspot affiliation +c256a9c mailmap: Qinghua Jin affiliation +c312b52 mailmap: Richard W.M. Jones affiliation +b441722 mailmap: Ira Cooper affiliation +f4f9d91 mailmap: Star Guo affiliation +ceb2a72 mailmap: Xiaoxi Chen affiliation +885db95 mailmap: Jenkins name normalization +fb63721 releases: firefly was EOL december 2015 +fc29dc2 releases: update the v10.1.0 release link +d82c497 doc: add a page for CephFS' experimental features describing lockout rules +6974ed4 doc: update cephfs administration page for enable_multiple safety check +4b3802e mdsmonitor: unify experimental warnings +ea1b276 test: update tests for new flag requirements +4bc4e74 MDSMap: lock out multimds clusters and directory fragmentation by default +15d1fd4 MDSMap: switch from booleans to flags for feature enablement logging +cda1c1a FSMap: add output for enabled_multiple flags +3130132 MDSMonitor: make enabling multiple FSes a lot scarier +3b9371a FSMap: print out compat instead of erroneously repeating enable_multiple +2429463 ceph-dencoder: add FSMap +15cabdc rgw_ldap: make ldap.h inclusion conditional +d58e5fe rgw: fix problem deleting objects begining with double underscore +492a572 test: fix memory leaks in rbd-mirror test cases +acfc2b1 test: avoid leaking librados connections when creating pools +88e244c rbd: journal reset should disable/re-enable journaling feature +2fa4147 osd/ReplicatedPG: clean up temp object if copy-from fails +a0bb575 ceph_test_rados_api_misc: make CopyFrom omap test be a big object +134416a Revert "rados: Add new field flags for ceph_osd_op.copy_get." +017d830 rbd-mirror: workaround for intermingled lockdep singletons +98744fd logrotate.conf: poke ceph-fuse after log rotation +91e0be0 ceph-fuse: reopen log file on SIGHUP +60679fc librbd: restore out-of-band future callbacks to avoid lock cycles +e98d046 MDS: unregister command add in clean_up_admin_socket +2f4bc84 rgw: aws4 subdomain calling bugfix +dab0b55 rgw: the map 'headers' is assigned a wrong value +f01261f authtool: fix test output validation & minor whitespace. +408964e journal: fix context memory leak when shutting down live replay +f931066 journal: Future does not require metadata shared pointer +d3dbd85 mon: warn if 'sortbitwise' flag is not set and no legacy OSDs are present +d544e44 mds: validate file layouts during replay +9414bef debian/rules: include ceph-mds-*.conf upstart files in ceph-mds +45a0bc1 mds: add operator<< for file_layout_t +693f46a mds: skip size recovery on non-file inodes +0f09642 mds: tidy up backtrace pool handling +4ddcf41 mds: health metric for being read only +79b19a6 osd: cancel scrub if noscrub is set for pool or all +4d3aef7 osd: reset tp handle when search for boundary of chunky-scrub +25d8007 os/bluestore: use short, relative paths with bluefs +ce50389 (tag: v10.1.1) 10.1.1 +6c0ab75 librbd: disallow unsafe rbd_op_threads values +bb4c2ca librgw/rgw_file: correctly handle object permissions +6851822 rgw_file: print DIRS1 read parameters at verbose +d84f55f rgw_file: fix attributes for "special" test cases +1bd1ffd rgw_file unit tests: validate Unix owners in DIRS1 +8e2c804 authtool: update --help and manpage to match code. +dffd867 build: Respect TMPDIR for virtualenv. +048251b common/fs_types: dump pool_id signed +cd41ca2 mds: fix legacy layout decode with pool 0 +b8e0458 rgw: retry read_log_info() while master is down +05cafcf Drop any systemd imposed process/thread limits 02ab8a2 mrun: update path to cmake binaries +67f8f1f os/bluestore/BlueFS: add some perfcounters +75ddd73 os/bluestore/BlueFS: revamp bdev ids +a5564a6 os/ObjectStore: make device uuid probe output something friendly +4f6523d rgw: aws4 uri encoding bugfix +bc9607b mon/OSDMonitor: fix off-by-one for osd_map_message_max +81cc288 osd: improve full map requests +2e22f54 osd: create rerequest_full_maps() helper +961a46f client: fix pool permisson check d248128 config: fix setuser_match_path typo d5ec33f tests: Removing one ceph-dencoder call in check-generated.sh 4af1aa6 tests: Fixing python statement in ceph_objectstore_tool.py @@ -18,10 +120,24 @@ 8b6be11 tests: Adding parallelism to encoding/readable.sh db31cc6 tests: Adding parallelism helpers in ceph-helpers.sh 93ace63 cmake: fix the build of test_rados_api_list +524f8e6 mds: unregister newly added commands correctly +da27c33 mds: avoid key renew storm on clock skew +7dae094 mds: drop message reference on error exit +1f54f73 mds: fix message leak during handle_core_message() b7a5f8b test: TestMirroringWatcher test cases were not closing images +ae39517 rgw: fix a typo in error message 8231208 global/global_init: expand metavariables in setuser_match_path +1a6c686 mds: Add cmapv to ESessions default constructor initializer list dd167cf crush: fix error log +3562323 librados: use Objecter::prepare_*_op helpers to set c->tid safely +cd167c0 osdc/Objecter: create more prepare_foo_op() helpers f47e06b tests: Fixing broken test/cephtool-test-mon.sh test +8f9e7b0 rgw: Do not try to encode or decode time_t +778506e time: Change temporary variable types in time decode +4a88a7f librados: fix narrow race with tid return value assignment +b7eb86f osdc/Objecter: fix narrow race with tid assignment +1de73d7 qa/workunits/rest/test.py: fs flag set enable_multiple true +42e692a ceph-rest-api: do not include single-option CephChoices in prefix 9565a50 set 128MB tcmalloc cache size by bytes ff9843b Striper: reduce assemble_result log level f812199 qa/workunits/rbd: qemu tests need to wait for image to be created @@ -36,6 +152,7 @@ c432691 os/ObjectStore: add noexcept to ensure move ctor is used 1c2831a common/Cycles: Do not initialize Cycles globally. ec79b64 unittest_erasure_code_plugin: fix deadlock caused by locked mutex in cancelled thread +554d1b4 ReplicatedPG::_rollback_to: update the OMAP flag aedc529 test: Fix test to run with btrfs which has snap_### dirs 3dd5249 librbd: avoid throwing error if mirroring is unsupported 280b8a1 rgw: add exclusive flag to set_as_default() @@ -44,7 +161,9 @@ c4efef5 rgw: add a few missing cmdline switches in help 09b5356 cls_journal: fix -EEXIST checking 2c0f03a rgw_admin: remove unused parent_period arg +85229f6 mds: fix potential null pointer access a29b96a debian/rules: put init-ceph in /etc/init.d/ceph, not ceph-base +09c4195 MDSMonitor: introduce command 'fs set_default ' 602425a configure: Add -D_LARGEFILE64_SOURCE to Linux build. 639f158 mon: remove unnecessary comment for update_from_paxos f5ef4d4 cmake: add missing librbd/MirrorWatcher.cc and librd/ObjectWatcher.cc @@ -161,6 +280,7 @@ bc254c8 rbd: rbd-mirroring: Enabling image mirroring depends on pool mirroring mode 668c8f9 script: subscription-manager support (part 3) 97b74bd osd/ClassHandler: only dlclose() the classes not missing +dad3b84 client: pass 'newly issued caps' to Client::check_cap_issue() 349c81f ceph_test_rados_api_pool: fix command for readonly cache-mode ad2e6f4 ceph.in: update for cmake path changes 5da6ae8 vstart: update for cmake build path changes @@ -592,6 +712,8 @@ bdcff15 mds: fix FSMap upgrade on mixed mon versions 38fd3f1 rgw: LDAP pass-through authentication e52f7b4 mds: fix FSMap upgrade with daemons in the map +7b33156 common: thread: allow set_affinity() to return a error code +e10c6e4 common: buffer: put a guard for stat() syscall during read_file 1ea1735 osd: fix wrong counter for batch objects removal during remove_dir() 12d151f osd: initialize last_recalibrate field at construction f1a4490 ceph.spec.in: disable lttng and babeltrace explicitly diff -Nru ceph-10.1.1/configure ceph-10.1.2/configure --- ceph-10.1.1/configure 2016-04-06 14:20:21.000000000 +0000 +++ ceph-10.1.2/configure 2016-04-12 17:53:16.000000000 +0000 @@ -1,6 +1,6 @@ #! /bin/sh # Guess values for system-dependent variables and create Makefiles. -# Generated by GNU Autoconf 2.69 for ceph 10.1.1. +# Generated by GNU Autoconf 2.69 for ceph 10.1.2. # # Report bugs to . # @@ -590,8 +590,8 @@ # Identity of this package. PACKAGE_NAME='ceph' PACKAGE_TARNAME='ceph' -PACKAGE_VERSION='10.1.1' -PACKAGE_STRING='ceph 10.1.1' +PACKAGE_VERSION='10.1.2' +PACKAGE_STRING='ceph 10.1.2' PACKAGE_BUGREPORT='ceph-devel@vger.kernel.org' PACKAGE_URL='' @@ -1582,7 +1582,7 @@ # Omit some internal or obsolete options to make the list less imposing. # This message is too long to be a string in the A/UX 3.1 sh. cat <<_ACEOF -\`configure' configures ceph 10.1.1 to adapt to many kinds of systems. +\`configure' configures ceph 10.1.2 to adapt to many kinds of systems. Usage: $0 [OPTION]... [VAR=VALUE]... @@ -1653,7 +1653,7 @@ if test -n "$ac_init_help"; then case $ac_init_help in - short | recursive ) echo "Configuration of ceph 10.1.1:";; + short | recursive ) echo "Configuration of ceph 10.1.2:";; esac cat <<\_ACEOF @@ -1837,7 +1837,7 @@ test -n "$ac_init_help" && exit $ac_status if $ac_init_version; then cat <<\_ACEOF -ceph configure 10.1.1 +ceph configure 10.1.2 generated by GNU Autoconf 2.69 Copyright (C) 2012 Free Software Foundation, Inc. @@ -2913,7 +2913,7 @@ This file contains any messages produced by compilers while running configure, to aid debugging if configure makes a mistake. -It was created by ceph $as_me 10.1.1, which was +It was created by ceph $as_me 10.1.2, which was generated by GNU Autoconf 2.69. Invocation command line was $ $0 $@ @@ -16408,7 +16408,7 @@ # Define the identity of the package. PACKAGE='ceph' - VERSION='10.1.1' + VERSION='10.1.2' cat >>confdefs.h <<_ACEOF @@ -26100,7 +26100,7 @@ # report actual input values of CONFIG_FILES etc. instead of their # values after options handling. ac_log=" -This file was extended by ceph $as_me 10.1.1, which was +This file was extended by ceph $as_me 10.1.2, which was generated by GNU Autoconf 2.69. Invocation command line was CONFIG_FILES = $CONFIG_FILES @@ -26166,7 +26166,7 @@ cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1 ac_cs_config="`$as_echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`" ac_cs_version="\\ -ceph config.status 10.1.1 +ceph config.status 10.1.2 configured by $0, generated by GNU Autoconf 2.69, with options \\"\$ac_cs_config\\" diff -Nru ceph-10.1.1/configure.ac ceph-10.1.2/configure.ac --- ceph-10.1.1/configure.ac 2016-04-06 14:17:03.000000000 +0000 +++ ceph-10.1.2/configure.ac 2016-04-12 17:46:05.000000000 +0000 @@ -8,7 +8,7 @@ # VERSION define is not used by the code. It gets a version string # from 'git describe'; see src/ceph_ver.[ch] -AC_INIT([ceph], [10.1.1], [ceph-devel@vger.kernel.org]) +AC_INIT([ceph], [10.1.2], [ceph-devel@vger.kernel.org]) AX_CXX_COMPILE_STDCXX_11(, mandatory) diff -Nru ceph-10.1.1/debian/ceph-common.postinst ceph-10.1.2/debian/ceph-common.postinst --- ceph-10.1.1/debian/ceph-common.postinst 2016-03-30 08:21:40.000000000 +0000 +++ ceph-10.1.2/debian/ceph-common.postinst 2016-04-12 10:02:36.000000000 +0000 @@ -56,7 +56,7 @@ usermod -c "$SERVER_NAME" \ -d $SERVER_HOME \ -g $SERVER_GROUP \ - $SERVER_USER + $SERVER_USER 2>/dev/null # 5. adjust file and directory permissions if ! dpkg-statoverride --list $SERVER_HOME >/dev/null diff -Nru ceph-10.1.1/debian/ceph.logrotate ceph-10.1.2/debian/ceph.logrotate --- ceph-10.1.1/debian/ceph.logrotate 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/debian/ceph.logrotate 2016-04-12 17:46:05.000000000 +0000 @@ -4,7 +4,7 @@ compress sharedscripts postrotate - killall -q -1 ceph-mon ceph-mds ceph-osd radosgw || true + killall -q -1 ceph-mon ceph-mds ceph-osd ceph-fuse radosgw || true endscript missingok notifempty diff -Nru ceph-10.1.1/debian/changelog ceph-10.1.2/debian/changelog --- ceph-10.1.1/debian/changelog 2016-04-08 15:30:47.000000000 +0000 +++ ceph-10.1.2/debian/changelog 2016-04-14 13:46:59.000000000 +0000 @@ -1,3 +1,13 @@ +ceph (10.1.2-0ubuntu1) xenial; urgency=medium + + * New upstream release candidate for Ceph Jewel: + - FFe: http://pad.lv/1563714. + - d/p/32bit-compat.patch,tasksmax-infinity.patch: Dropped, + included upstream. + * d/ceph-common.postinst: Silence output of usermod call (LP: #1569249). + + -- James Page Thu, 14 Apr 2016 14:46:58 +0100 + ceph (10.1.1-0ubuntu1) xenial; urgency=medium * New upstream release candidate for Ceph Jewel: diff -Nru ceph-10.1.1/debian/patches/32bit-compat.patch ceph-10.1.2/debian/patches/32bit-compat.patch --- ceph-10.1.1/debian/patches/32bit-compat.patch 2016-04-08 11:10:04.000000000 +0000 +++ ceph-10.1.2/debian/patches/32bit-compat.patch 1970-01-01 00:00:00.000000000 +0000 @@ -1,55 +0,0 @@ -From 778506ed7d0efc9899d419cf31e0bf7e25cf9420 Mon Sep 17 00:00:00 2001 -From: "Adam C. Emerson" -Date: Thu, 31 Mar 2016 14:50:06 -0400 -Subject: [PATCH 1/2] time: Change temporary variable types in time decode - -Since we're decoding 32-bit integers, just use uint32_t and then cast them to -what utime_t expects. - -Fixes: http://tracker.ceph.com/issues/15330 -Signed-off-by: Adam C. Emerson ---- - src/common/ceph_time.h | 7 +++++-- - 1 file changed, 5 insertions(+), 2 deletions(-) - ---- a/src/common/ceph_time.h -+++ b/src/common/ceph_time.h -@@ -382,10 +382,13 @@ void encode(const std::chrono::time_poin - template - void decode(std::chrono::time_point& t, - bufferlist::iterator& p) { -- uint32_t s, ns; -+ uint32_t s; -+ uint32_t ns; - ::decode(s, p); - ::decode(ns, p); -- struct timespec ts = {s, ns}; -+ struct timespec ts = { -+ static_cast(s), -+ static_cast(ns)}; - - t = Clock::from_timespec(ts); - } ---- a/src/cls/rgw/cls_rgw_ops.h -+++ b/src/cls/rgw/cls_rgw_ops.h -@@ -180,7 +180,7 @@ struct rgw_cls_link_olh_op { - ::encode(olh_epoch, bl); - ::encode(log_op, bl); - ::encode(bilog_flags, bl); -- time_t t = ceph::real_clock::to_time_t(unmod_since); -+ uint64_t t = ceph::real_clock::to_time_t(unmod_since); - ::encode(t, bl); - ::encode(unmod_since, bl); - ::encode(high_precision_time, bl); -@@ -198,9 +198,9 @@ struct rgw_cls_link_olh_op { - ::decode(log_op, bl); - ::decode(bilog_flags, bl); - if (struct_v == 2) { -- time_t t; -+ uint64_t t; - ::decode(t, bl); -- unmod_since = ceph::real_clock::from_time_t(t); -+ unmod_since = ceph::real_clock::from_time_t(static_cast(t)); - } - if (struct_v >= 3) { - ::decode(unmod_since, bl); diff -Nru ceph-10.1.1/debian/patches/series ceph-10.1.2/debian/patches/series --- ceph-10.1.1/debian/patches/series 2016-04-05 16:27:02.000000000 +0000 +++ ceph-10.1.2/debian/patches/series 2016-04-14 09:06:28.000000000 +0000 @@ -17,6 +17,4 @@ rocksdb-flags.patch skip-setup.py-makefiles.patch disable-openssl-linking.patch -32bit-compat.patch drop-user-group-osd-prestart.patch -tasksmax-infinity.patch diff -Nru ceph-10.1.1/debian/patches/tasksmax-infinity.patch ceph-10.1.2/debian/patches/tasksmax-infinity.patch --- ceph-10.1.1/debian/patches/tasksmax-infinity.patch 2016-04-08 11:10:05.000000000 +0000 +++ ceph-10.1.2/debian/patches/tasksmax-infinity.patch 1970-01-01 00:00:00.000000000 +0000 @@ -1,71 +0,0 @@ -From 05cafcf19fb80dd2a76fdbbe5329bae53aa0def4 Mon Sep 17 00:00:00 2001 -From: James Page -Date: Tue, 5 Apr 2016 17:32:59 +0100 -Subject: [PATCH] Drop any systemd imposed process/thread limits - -If systemd has task accounting enabled, a default of 512 tasks -will be applied to all systemd units. - -For ceph, this is way to low even for a modest cluster, so stop -this restriction being applied and allow administrators to apply -limits using sysctl. - -Signed-off-by: James Page ---- - systemd/ceph-mds@.service | 1 + - systemd/ceph-mon@.service | 1 + - systemd/ceph-osd@.service | 1 + - systemd/ceph-radosgw@.service | 1 + - systemd/ceph-rbd-mirror@.service | 1 + - 5 files changed, 5 insertions(+) - ---- a/systemd/ceph-mds@.service -+++ b/systemd/ceph-mds@.service -@@ -15,6 +15,7 @@ PrivateDevices=yes - ProtectHome=true - ProtectSystem=full - PrivateTmp=true -+TasksMax=infinity - - [Install] - WantedBy=ceph-mds.target ---- a/systemd/ceph-mon@.service -+++ b/systemd/ceph-mon@.service -@@ -21,6 +21,7 @@ PrivateDevices=yes - ProtectHome=true - ProtectSystem=full - PrivateTmp=true -+TasksMax=infinity - - [Install] - WantedBy=ceph-mon.target ---- a/systemd/ceph-osd@.service -+++ b/systemd/ceph-osd@.service -@@ -15,6 +15,7 @@ ExecReload=/bin/kill -HUP $MAINPID - ProtectHome=true - ProtectSystem=full - PrivateTmp=true -+TasksMax=infinity - - [Install] - WantedBy=ceph-osd.target ---- a/systemd/ceph-radosgw@.service -+++ b/systemd/ceph-radosgw@.service -@@ -14,6 +14,7 @@ PrivateDevices=yes - ProtectHome=true - ProtectSystem=full - PrivateTmp=true -+TasksMax=infinity - - [Install] - WantedBy=ceph-radosgw.target ---- a/systemd/ceph-rbd-mirror@.service -+++ b/systemd/ceph-rbd-mirror@.service -@@ -17,6 +17,7 @@ PrivateTmp=true - Restart=on-failure - StartLimitInterval=30min - StartLimitBurst=3 -+TasksMax=infinity - - [Install] - WantedBy=ceph-rbd-mirror.target diff -Nru ceph-10.1.1/doc/man/8/ceph-authtool.rst ceph-10.1.2/doc/man/8/ceph-authtool.rst --- ceph-10.1.1/doc/man/8/ceph-authtool.rst 2016-04-06 14:17:03.000000000 +0000 +++ ceph-10.1.2/doc/man/8/ceph-authtool.rst 2016-04-12 17:46:05.000000000 +0000 @@ -9,9 +9,18 @@ Synopsis ======== -| **ceph-authtool** *keyringfile* [ -l | --list ] [ -C | --create-keyring - ] [ -p | --print ] [ -n | --name *entityname* ] [ --gen-key ] [ -a | - --add-key *base64_key* ] [ --caps *capfile* ] +| **ceph-authtool** *keyringfile* + [ -l | --list ] + [ -p | --print ] + [ -C | --create-keyring ] + [ -g | --gen-key ] + [ --gen-print-key ] + [ --import-keyring *otherkeyringfile* ] + [ -n | --name *entityname* ] + [ -u | --set-uid *auid* ] + [ -a | --add-key *base64_key* ] + [ --cap *subsystem* *capability* ] + [ --caps *capfile* ] Description @@ -45,19 +54,36 @@ will create a new keyring, overwriting any existing keyringfile -.. option:: --gen-key +.. option:: -g, --gen-key will generate a new secret key for the specified entityname -.. option:: --add-key +.. option:: --gen-print-key + + will generate a new secret key for the specified entityname, + without altering the keyringfile, printing the secret to stdout + +.. option:: --import-keyring *secondkeyringfile* + + will import the content of a given keyring to the keyringfile + +.. option:: -n, --name *name* + + specify entityname to operate on + +.. option:: -u, --set-uid *auid* + + sets the auid (authenticated user id) for the specified entityname + +.. option:: -a, --add-key *base64_key* will add an encoded key to the keyring -.. option:: --cap subsystem capability +.. option:: --cap *subsystem* *capability* will set the capability for given subsystem -.. option:: --caps capsfile +.. option:: --caps *capsfile* will set all of capabilities associated with a given key, for all subsystems diff -Nru ceph-10.1.1/doc/man/8/radosgw-admin.rst ceph-10.1.2/doc/man/8/radosgw-admin.rst --- ceph-10.1.1/doc/man/8/radosgw-admin.rst 2016-04-06 14:17:03.000000000 +0000 +++ ceph-10.1.2/doc/man/8/radosgw-admin.rst 2016-04-12 17:46:05.000000000 +0000 @@ -107,7 +107,7 @@ :command:`region get` Show region info. -:command:`regions list` +:command:`region list` List all regions set on this cluster. :command:`region set` diff -Nru ceph-10.1.1/man/ceph.8 ceph-10.1.2/man/ceph.8 --- ceph-10.1.1/man/ceph.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph \- ceph administration tool . diff -Nru ceph-10.1.1/man/ceph-authtool.8 ceph-10.1.2/man/ceph-authtool.8 --- ceph-10.1.1/man/ceph-authtool.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-authtool.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-AUTHTOOL" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-AUTHTOOL" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-authtool \- ceph keyring manipulation tool . @@ -32,9 +32,18 @@ .. .SH SYNOPSIS .nf -\fBceph\-authtool\fP \fIkeyringfile\fP [ \-l | \-\-list ] [ \-C | \-\-create\-keyring -] [ \-p | \-\-print ] [ \-n | \-\-name \fIentityname\fP ] [ \-\-gen\-key ] [ \-a | -\-\-add\-key \fIbase64_key\fP ] [ \-\-caps \fIcapfile\fP ] +\fBceph\-authtool\fP \fIkeyringfile\fP +[ \-l | \-\-list ] +[ \-p | \-\-print ] +[ \-C | \-\-create\-keyring ] +[ \-g | \-\-gen\-key ] +[ \-\-gen\-print\-key ] +[ \-\-import\-keyring \fIotherkeyringfile\fP ] +[ \-n | \-\-name \fIentityname\fP ] +[ \-u | \-\-set\-uid \fIauid\fP ] +[ \-a | \-\-add\-key \fIbase64_key\fP ] +[ \-\-cap \fIsubsystem\fP \fIcapability\fP ] +[ \-\-caps \fIcapfile\fP ] .fi .sp .SH DESCRIPTION @@ -69,22 +78,43 @@ .UNINDENT .INDENT 0.0 .TP -.B \-\-gen\-key +.B \-g, \-\-gen\-key will generate a new secret key for the specified entityname .UNINDENT .INDENT 0.0 .TP -.B \-\-add\-key +.B \-\-gen\-print\-key +will generate a new secret key for the specified entityname, +without altering the keyringfile, printing the secret to stdout +.UNINDENT +.INDENT 0.0 +.TP +.B \-\-import\-keyring *secondkeyringfile* +will import the content of a given keyring to the keyringfile +.UNINDENT +.INDENT 0.0 +.TP +.B \-n, \-\-name *name* +specify entityname to operate on +.UNINDENT +.INDENT 0.0 +.TP +.B \-u, \-\-set\-uid *auid* +sets the auid (authenticated user id) for the specified entityname +.UNINDENT +.INDENT 0.0 +.TP +.B \-a, \-\-add\-key *base64_key* will add an encoded key to the keyring .UNINDENT .INDENT 0.0 .TP -.B \-\-cap subsystem capability +.B \-\-cap *subsystem* *capability* will set the capability for given subsystem .UNINDENT .INDENT 0.0 .TP -.B \-\-caps capsfile +.B \-\-caps *capsfile* will set all of capabilities associated with a given key, for all subsystems .UNINDENT .SH CAPABILITIES diff -Nru ceph-10.1.1/man/ceph-clsinfo.8 ceph-10.1.2/man/ceph-clsinfo.8 --- ceph-10.1.1/man/ceph-clsinfo.8 2016-04-06 14:21:21.000000000 +0000 +++ ceph-10.1.2/man/ceph-clsinfo.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-CLSINFO" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-CLSINFO" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-clsinfo \- show class object information . diff -Nru ceph-10.1.1/man/ceph-conf.8 ceph-10.1.2/man/ceph-conf.8 --- ceph-10.1.1/man/ceph-conf.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-conf.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-CONF" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-CONF" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-conf \- ceph conf file tool . diff -Nru ceph-10.1.1/man/ceph-create-keys.8 ceph-10.1.2/man/ceph-create-keys.8 --- ceph-10.1.1/man/ceph-create-keys.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-create-keys.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-CREATE-KEYS" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-CREATE-KEYS" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-create-keys \- ceph keyring generate tool . diff -Nru ceph-10.1.1/man/ceph-debugpack.8 ceph-10.1.2/man/ceph-debugpack.8 --- ceph-10.1.1/man/ceph-debugpack.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-debugpack.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-DEBUGPACK" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-DEBUGPACK" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-debugpack \- ceph debug packer utility . diff -Nru ceph-10.1.1/man/ceph-dencoder.8 ceph-10.1.2/man/ceph-dencoder.8 --- ceph-10.1.1/man/ceph-dencoder.8 2016-04-06 14:21:21.000000000 +0000 +++ ceph-10.1.2/man/ceph-dencoder.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-DENCODER" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-DENCODER" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-dencoder \- ceph encoder/decoder utility . diff -Nru ceph-10.1.1/man/ceph-deploy.8 ceph-10.1.2/man/ceph-deploy.8 --- ceph-10.1.1/man/ceph-deploy.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-deploy.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-DEPLOY" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-DEPLOY" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-deploy \- Ceph deployment tool . diff -Nru ceph-10.1.1/man/ceph-detect-init.8 ceph-10.1.2/man/ceph-detect-init.8 --- ceph-10.1.1/man/ceph-detect-init.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-detect-init.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-DETECT-INIT" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-DETECT-INIT" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-detect-init \- display the init system Ceph should use . diff -Nru ceph-10.1.1/man/ceph-disk.8 ceph-10.1.2/man/ceph-disk.8 --- ceph-10.1.1/man/ceph-disk.8 2016-04-06 14:21:21.000000000 +0000 +++ ceph-10.1.2/man/ceph-disk.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-DISK" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-DISK" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-disk \- Ceph disk utility for OSD . diff -Nru ceph-10.1.1/man/cephfs.8 ceph-10.1.2/man/cephfs.8 --- ceph-10.1.1/man/cephfs.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/cephfs.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPHFS" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPHFS" "8" "April 12, 2016" "dev" "Ceph" .SH NAME cephfs \- ceph file system options utility . diff -Nru ceph-10.1.1/man/ceph-fuse.8 ceph-10.1.2/man/ceph-fuse.8 --- ceph-10.1.1/man/ceph-fuse.8 2016-04-06 14:21:21.000000000 +0000 +++ ceph-10.1.2/man/ceph-fuse.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-FUSE" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-FUSE" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-fuse \- FUSE-based client for ceph . diff -Nru ceph-10.1.1/man/ceph-mds.8 ceph-10.1.2/man/ceph-mds.8 --- ceph-10.1.1/man/ceph-mds.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-mds.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-MDS" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-MDS" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-mds \- ceph metadata server daemon . diff -Nru ceph-10.1.1/man/ceph-mon.8 ceph-10.1.2/man/ceph-mon.8 --- ceph-10.1.1/man/ceph-mon.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-mon.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-MON" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-MON" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-mon \- ceph monitor daemon . diff -Nru ceph-10.1.1/man/ceph-osd.8 ceph-10.1.2/man/ceph-osd.8 --- ceph-10.1.1/man/ceph-osd.8 2016-04-06 14:21:21.000000000 +0000 +++ ceph-10.1.2/man/ceph-osd.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-OSD" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-OSD" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-osd \- ceph object storage daemon . diff -Nru ceph-10.1.1/man/ceph-post-file.8 ceph-10.1.2/man/ceph-post-file.8 --- ceph-10.1.1/man/ceph-post-file.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-post-file.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-POST-FILE" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-POST-FILE" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-post-file \- post files for ceph developers . diff -Nru ceph-10.1.1/man/ceph-rbdnamer.8 ceph-10.1.2/man/ceph-rbdnamer.8 --- ceph-10.1.1/man/ceph-rbdnamer.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-rbdnamer.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-RBDNAMER" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-RBDNAMER" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-rbdnamer \- udev helper to name RBD devices . diff -Nru ceph-10.1.1/man/ceph-rest-api.8 ceph-10.1.2/man/ceph-rest-api.8 --- ceph-10.1.1/man/ceph-rest-api.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-rest-api.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-REST-API" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-REST-API" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-rest-api \- ceph RESTlike administration server . diff -Nru ceph-10.1.1/man/ceph-run.8 ceph-10.1.2/man/ceph-run.8 --- ceph-10.1.1/man/ceph-run.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-run.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-RUN" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-RUN" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-run \- restart daemon on core dump . diff -Nru ceph-10.1.1/man/ceph-syn.8 ceph-10.1.2/man/ceph-syn.8 --- ceph-10.1.1/man/ceph-syn.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/ceph-syn.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CEPH-SYN" "8" "April 06, 2016" "dev" "Ceph" +.TH "CEPH-SYN" "8" "April 12, 2016" "dev" "Ceph" .SH NAME ceph-syn \- ceph synthetic workload generator . diff -Nru ceph-10.1.1/man/crushtool.8 ceph-10.1.2/man/crushtool.8 --- ceph-10.1.1/man/crushtool.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/crushtool.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "CRUSHTOOL" "8" "April 06, 2016" "dev" "Ceph" +.TH "CRUSHTOOL" "8" "April 12, 2016" "dev" "Ceph" .SH NAME crushtool \- CRUSH map manipulation tool . diff -Nru ceph-10.1.1/man/librados-config.8 ceph-10.1.2/man/librados-config.8 --- ceph-10.1.1/man/librados-config.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/librados-config.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "LIBRADOS-CONFIG" "8" "April 06, 2016" "dev" "Ceph" +.TH "LIBRADOS-CONFIG" "8" "April 12, 2016" "dev" "Ceph" .SH NAME librados-config \- display information about librados . diff -Nru ceph-10.1.1/man/monmaptool.8 ceph-10.1.2/man/monmaptool.8 --- ceph-10.1.1/man/monmaptool.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/monmaptool.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "MONMAPTOOL" "8" "April 06, 2016" "dev" "Ceph" +.TH "MONMAPTOOL" "8" "April 12, 2016" "dev" "Ceph" .SH NAME monmaptool \- ceph monitor cluster map manipulation tool . diff -Nru ceph-10.1.1/man/mount.ceph.8 ceph-10.1.2/man/mount.ceph.8 --- ceph-10.1.1/man/mount.ceph.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/mount.ceph.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "MOUNT.CEPH" "8" "April 06, 2016" "dev" "Ceph" +.TH "MOUNT.CEPH" "8" "April 12, 2016" "dev" "Ceph" .SH NAME mount.ceph \- mount a ceph file system . diff -Nru ceph-10.1.1/man/osdmaptool.8 ceph-10.1.2/man/osdmaptool.8 --- ceph-10.1.1/man/osdmaptool.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/osdmaptool.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "OSDMAPTOOL" "8" "April 06, 2016" "dev" "Ceph" +.TH "OSDMAPTOOL" "8" "April 12, 2016" "dev" "Ceph" .SH NAME osdmaptool \- ceph osd cluster map manipulation tool . diff -Nru ceph-10.1.1/man/rados.8 ceph-10.1.2/man/rados.8 --- ceph-10.1.1/man/rados.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/rados.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "RADOS" "8" "April 06, 2016" "dev" "Ceph" +.TH "RADOS" "8" "April 12, 2016" "dev" "Ceph" .SH NAME rados \- rados object storage utility . diff -Nru ceph-10.1.1/man/radosgw.8 ceph-10.1.2/man/radosgw.8 --- ceph-10.1.1/man/radosgw.8 2016-04-06 14:21:21.000000000 +0000 +++ ceph-10.1.2/man/radosgw.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "RADOSGW" "8" "April 06, 2016" "dev" "Ceph" +.TH "RADOSGW" "8" "April 12, 2016" "dev" "Ceph" .SH NAME radosgw \- rados REST gateway . diff -Nru ceph-10.1.1/man/radosgw-admin.8 ceph-10.1.2/man/radosgw-admin.8 --- ceph-10.1.1/man/radosgw-admin.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/radosgw-admin.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "RADOSGW-ADMIN" "8" "April 06, 2016" "dev" "Ceph" +.TH "RADOSGW-ADMIN" "8" "April 12, 2016" "dev" "Ceph" .SH NAME radosgw-admin \- rados REST gateway user administration utility . @@ -127,7 +127,7 @@ .B \fBregion get\fP Show region info. .TP -.B \fBregions list\fP +.B \fBregion list\fP List all regions set on this cluster. .TP .B \fBregion set\fP diff -Nru ceph-10.1.1/man/rbd.8 ceph-10.1.2/man/rbd.8 --- ceph-10.1.1/man/rbd.8 2016-04-06 14:21:21.000000000 +0000 +++ ceph-10.1.2/man/rbd.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "RBD" "8" "April 06, 2016" "dev" "Ceph" +.TH "RBD" "8" "April 12, 2016" "dev" "Ceph" .SH NAME rbd \- manage rados block device (RBD) images . diff -Nru ceph-10.1.1/man/rbd-fuse.8 ceph-10.1.2/man/rbd-fuse.8 --- ceph-10.1.1/man/rbd-fuse.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/rbd-fuse.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "RBD-FUSE" "8" "April 06, 2016" "dev" "Ceph" +.TH "RBD-FUSE" "8" "April 12, 2016" "dev" "Ceph" .SH NAME rbd-fuse \- expose rbd images as files . diff -Nru ceph-10.1.1/man/rbdmap.8 ceph-10.1.2/man/rbdmap.8 --- ceph-10.1.1/man/rbdmap.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/rbdmap.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "RBDMAP" "8" "April 06, 2016" "dev" "Ceph" +.TH "RBDMAP" "8" "April 12, 2016" "dev" "Ceph" .SH NAME rbdmap \- map RBD devices at boot time . diff -Nru ceph-10.1.1/man/rbd-mirror.8 ceph-10.1.2/man/rbd-mirror.8 --- ceph-10.1.1/man/rbd-mirror.8 2016-04-06 14:21:21.000000000 +0000 +++ ceph-10.1.2/man/rbd-mirror.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "RBD-MIRROR" "8" "April 06, 2016" "dev" "Ceph" +.TH "RBD-MIRROR" "8" "April 12, 2016" "dev" "Ceph" .SH NAME rbd-mirror \- Ceph daemon for mirroring RBD images . diff -Nru ceph-10.1.1/man/rbd-nbd.8 ceph-10.1.2/man/rbd-nbd.8 --- ceph-10.1.1/man/rbd-nbd.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/rbd-nbd.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "RBD-NBD" "8" "April 06, 2016" "dev" "Ceph" +.TH "RBD-NBD" "8" "April 12, 2016" "dev" "Ceph" .SH NAME rbd-nbd \- map rbd images to nbd device . diff -Nru ceph-10.1.1/man/rbd-replay.8 ceph-10.1.2/man/rbd-replay.8 --- ceph-10.1.1/man/rbd-replay.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/rbd-replay.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "RBD-REPLAY" "8" "April 06, 2016" "dev" "Ceph" +.TH "RBD-REPLAY" "8" "April 12, 2016" "dev" "Ceph" .SH NAME rbd-replay \- replay rados block device (RBD) workloads . diff -Nru ceph-10.1.1/man/rbd-replay-many.8 ceph-10.1.2/man/rbd-replay-many.8 --- ceph-10.1.1/man/rbd-replay-many.8 2016-04-06 14:21:21.000000000 +0000 +++ ceph-10.1.2/man/rbd-replay-many.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "RBD-REPLAY-MANY" "8" "April 06, 2016" "dev" "Ceph" +.TH "RBD-REPLAY-MANY" "8" "April 12, 2016" "dev" "Ceph" .SH NAME rbd-replay-many \- replay a rados block device (RBD) workload on several clients . diff -Nru ceph-10.1.1/man/rbd-replay-prep.8 ceph-10.1.2/man/rbd-replay-prep.8 --- ceph-10.1.1/man/rbd-replay-prep.8 2016-04-06 14:21:22.000000000 +0000 +++ ceph-10.1.2/man/rbd-replay-prep.8 2016-04-12 17:54:22.000000000 +0000 @@ -1,6 +1,6 @@ .\" Man page generated from reStructuredText. . -.TH "RBD-REPLAY-PREP" "8" "April 06, 2016" "dev" "Ceph" +.TH "RBD-REPLAY-PREP" "8" "April 12, 2016" "dev" "Ceph" .SH NAME rbd-replay-prep \- prepare captured rados block device (RBD) workloads for replay . diff -Nru ceph-10.1.1/src/ceph-disk/ceph_disk/main.py ceph-10.1.2/src/ceph-disk/ceph_disk/main.py --- ceph-10.1.1/src/ceph-disk/ceph_disk/main.py 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/ceph-disk/ceph_disk/main.py 2016-04-12 17:46:05.000000000 +0000 @@ -122,7 +122,7 @@ @staticmethod def get_ready_by_name(name): - return [x[name]['ready'] for x in PTYPE.values()] + return [x[name]['ready'] for x in PTYPE.values() if name in x] @staticmethod def is_regular_space(ptype): @@ -1876,18 +1876,18 @@ if stat.S_ISBLK(mode): if getattr(args, name + '_file'): raise Error('%s is not a regular file' % name.capitalize, - geattr(args, name)) + getattr(args, name)) self.type = self.DEVICE return if stat.S_ISREG(mode): if getattr(args, name + '_dev'): raise Error('%s is not a block device' % name.capitalize, - geattr(args, name)) + getattr(args, name)) self.type = self.FILE raise Error('%s %s is neither a block device nor regular file' % - (name.capitalize, geattr(args, name))) + (name.capitalize, getattr(args, name))) def is_none(self): return self.type == self.NONE @@ -2241,8 +2241,8 @@ self.args.lockbox) self.partition = DevicePartition.factory( path=None, dev=self.args.lockbox, args=self.args) - ptype = partition.get_ptype() - ready = Ptype.get_ready_by_type('lockbox') + ptype = self.partition.get_ptype() + ready = Ptype.get_ready_by_name('lockbox') if ptype not in ready: LOG.warning('incorrect partition UUID: %s, expected %s' % (ptype, str(ready))) @@ -2384,7 +2384,7 @@ elif stat.S_ISBLK(dmode): self.type = self.DEVICE else: - raise Error('not a dir or block device', args.data) + raise Error('not a dir or block device', self.args.data) def is_file(self): return self.type == self.FILE @@ -2539,8 +2539,8 @@ self.args.data) self.partition = DevicePartition.factory( path=None, dev=self.args.data, args=self.args) - ptype = partition.get_ptype() - ready = Ptype.get_ready_by_type('osd') + ptype = self.partition.get_ptype() + ready = Ptype.get_ready_by_name('osd') if ptype not in ready: LOG.warning('incorrect partition UUID: %s, expected %s' % (ptype, str(ready))) diff -Nru ceph-10.1.1/src/ceph_fuse.cc ceph-10.1.2/src/ceph_fuse.cc --- ceph-10.1.1/src/ceph_fuse.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/ceph_fuse.cc 2016-04-12 17:46:05.000000000 +0000 @@ -34,6 +34,7 @@ #include "common/linux_version.h" #endif #include "global/global_init.h" +#include "global/signal_handler.h" #include "common/safe_io.h" #include @@ -223,6 +224,9 @@ goto out_messenger_start_failed; } + init_async_signal_handler(); + register_async_signal_handler(SIGHUP, sighup_handler); + // start client r = client->init(); if (r < 0) { @@ -268,6 +272,9 @@ out_shutdown: client->shutdown(); out_init_failed: + unregister_async_signal_handler(SIGHUP, sighup_handler); + shutdown_async_signal_handler(); + // wait for messenger to finish messenger->shutdown(); messenger->wait(); diff -Nru ceph-10.1.1/src/ceph_osd.cc ceph-10.1.2/src/ceph_osd.cc --- ceph-10.1.1/src/ceph_osd.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/ceph_osd.cc 2016-04-12 17:46:05.000000000 +0000 @@ -189,7 +189,8 @@ } if (get_device_fsid) { uuid_d uuid; - int r = ObjectStore::probe_block_device_fsid(device_path, &uuid); + int r = ObjectStore::probe_block_device_fsid(g_ceph_context, device_path, + &uuid); if (r < 0) { cerr << "failed to get device fsid for " << device_path << ": " << cpp_strerror(r) << std::endl; diff -Nru ceph-10.1.1/src/client/Client.cc ceph-10.1.2/src/client/Client.cc --- ceph-10.1.1/src/client/Client.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/client/Client.cc 2016-04-12 17:46:05.000000000 +0000 @@ -4807,7 +4807,7 @@ if (m->get_op() == CEPH_CAP_OP_IMPORT && m->get_wanted() != wanted) check = true; - check_cap_issue(in, cap, issued); + check_cap_issue(in, cap, new_caps); // update caps if (old_caps & ~new_caps) { @@ -12099,10 +12099,12 @@ if (!cct->_conf->client_check_pool_perm) return 0; - int64_t pool = in->layout.pool_id; + int64_t pool_id = in->layout.pool_id; + std::string pool_ns = in->layout.pool_ns; + std::pair perm_key(pool_id, pool_ns); int have = 0; while (true) { - std::map::iterator it = pool_perms.find(pool); + auto it = pool_perms.find(perm_key); if (it == pool_perms.end()) break; if (it->second == POOL_CHECKING) { @@ -12123,7 +12125,7 @@ return 0; } - pool_perms[pool] = POOL_CHECKING; + pool_perms[perm_key] = POOL_CHECKING; char oid_buf[32]; snprintf(oid_buf, sizeof(oid_buf), "%llx.00000000", (unsigned long long)in->ino); @@ -12155,7 +12157,7 @@ if (rd_ret == 0 || rd_ret == -ENOENT) have |= POOL_READ; else if (rd_ret != -EPERM) { - ldout(cct, 10) << "check_pool_perm on pool " << pool + ldout(cct, 10) << "check_pool_perm on pool " << pool_id << " ns " << pool_ns << " rd_err = " << rd_ret << " wr_err = " << wr_ret << dendl; errored = true; } @@ -12163,7 +12165,7 @@ if (wr_ret == 0 || wr_ret == -EEXIST) have |= POOL_WRITE; else if (wr_ret != -EPERM) { - ldout(cct, 10) << "check_pool_perm on pool " << pool + ldout(cct, 10) << "check_pool_perm on pool " << pool_id << " ns " << pool_ns << " rd_err = " << rd_ret << " wr_err = " << wr_ret << dendl; errored = true; } @@ -12172,22 +12174,22 @@ // Indeterminate: erase CHECKING state so that subsequent calls re-check. // Raise EIO because actual error code might be misleading for // userspace filesystem user. - pool_perms.erase(pool); + pool_perms.erase(perm_key); signal_cond_list(waiting_for_pool_perm); return -EIO; } - pool_perms[pool] = have | POOL_CHECKED; + pool_perms[perm_key] = have | POOL_CHECKED; signal_cond_list(waiting_for_pool_perm); } if ((need & CEPH_CAP_FILE_RD) && !(have & POOL_READ)) { - ldout(cct, 10) << "check_pool_perm on pool " << pool + ldout(cct, 10) << "check_pool_perm on pool " << pool_id << " ns " << pool_ns << " need " << ccap_string(need) << ", but no read perm" << dendl; return -EPERM; } if ((need & CEPH_CAP_FILE_WR) && !(have & POOL_WRITE)) { - ldout(cct, 10) << "check_pool_perm on pool " << pool + ldout(cct, 10) << "check_pool_perm on pool " << pool_id << " ns " << pool_ns << " need " << ccap_string(need) << ", but no write perm" << dendl; return -EPERM; } diff -Nru ceph-10.1.1/src/client/Client.h ceph-10.1.2/src/client/Client.h --- ceph-10.1.1/src/client/Client.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/client/Client.h 2016-04-12 17:46:05.000000000 +0000 @@ -541,7 +541,7 @@ bool is_quota_bytes_exceeded(Inode *in, int64_t new_bytes); bool is_quota_bytes_approaching(Inode *in); - std::map pool_perms; + std::map, int> pool_perms; list waiting_for_pool_perm; int check_pool_perm(Inode *in, int need); diff -Nru ceph-10.1.1/src/cls/rgw/cls_rgw_ops.h ceph-10.1.2/src/cls/rgw/cls_rgw_ops.h --- ceph-10.1.1/src/cls/rgw/cls_rgw_ops.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/cls/rgw/cls_rgw_ops.h 2016-04-12 17:46:05.000000000 +0000 @@ -180,7 +180,7 @@ ::encode(olh_epoch, bl); ::encode(log_op, bl); ::encode(bilog_flags, bl); - time_t t = ceph::real_clock::to_time_t(unmod_since); + uint64_t t = ceph::real_clock::to_time_t(unmod_since); ::encode(t, bl); ::encode(unmod_since, bl); ::encode(high_precision_time, bl); @@ -198,9 +198,9 @@ ::decode(log_op, bl); ::decode(bilog_flags, bl); if (struct_v == 2) { - time_t t; + uint64_t t; ::decode(t, bl); - unmod_since = ceph::real_clock::from_time_t(t); + unmod_since = ceph::real_clock::from_time_t(static_cast(t)); } if (struct_v >= 3) { ::decode(unmod_since, bl); diff -Nru ceph-10.1.1/src/common/buffer.cc ceph-10.1.2/src/common/buffer.cc --- ceph-10.1.1/src/common/buffer.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/common/buffer.cc 2016-04-12 17:46:05.000000000 +0000 @@ -1982,7 +1982,15 @@ struct stat st; memset(&st, 0, sizeof(st)); - ::fstat(fd, &st); + if (::fstat(fd, &st) < 0) { + int err = errno; + std::ostringstream oss; + oss << "bufferlist::read_file(" << fn << "): stat error: " + << cpp_strerror(err); + *error = oss.str(); + VOID_TEMP_FAILURE_RETRY(::close(fd)); + return -err; + } ssize_t ret = read_fd(fd, st.st_size); if (ret < 0) { diff -Nru ceph-10.1.1/src/common/ceph_time.h ceph-10.1.2/src/common/ceph_time.h --- ceph-10.1.1/src/common/ceph_time.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/common/ceph_time.h 2016-04-12 17:46:05.000000000 +0000 @@ -382,10 +382,13 @@ template void decode(std::chrono::time_point& t, bufferlist::iterator& p) { - uint32_t s, ns; + uint32_t s; + uint32_t ns; ::decode(s, p); ::decode(ns, p); - struct timespec ts = {s, ns}; + struct timespec ts = { + static_cast(s), + static_cast(ns)}; t = Clock::from_timespec(ts); } diff -Nru ceph-10.1.1/src/common/config_opts.h ceph-10.1.2/src/common/config_opts.h --- ceph-10.1.1/src/common/config_opts.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/common/config_opts.h 2016-04-12 17:46:05.000000000 +0000 @@ -271,6 +271,7 @@ OPTION(mon_warn_on_crush_straw_calc_version_zero, OPT_BOOL, true) // warn if crush straw_calc_version==0 OPTION(mon_warn_on_osd_down_out_interval_zero, OPT_BOOL, true) // warn if 'mon_osd_down_out_interval == 0' OPTION(mon_warn_on_cache_pools_without_hit_sets, OPT_BOOL, true) +OPTION(mon_warn_on_no_sortbitwise, OPT_BOOL, true) // warn when sortbitwise not set OPTION(mon_min_osdmap_epochs, OPT_INT, 500) OPTION(mon_max_pgmap_epochs, OPT_INT, 500) OPTION(mon_max_log_epochs, OPT_INT, 500) @@ -869,6 +870,7 @@ OPTION(osd_max_object_size, OPT_U64, 100*1024L*1024L*1024L) // OSD's maximum object size OPTION(osd_max_object_name_len, OPT_U32, 2048) // max rados object name len +OPTION(osd_max_object_namespace_len, OPT_U32, 256) // max rados object namespace len OPTION(osd_max_attr_name_len, OPT_U32, 100) // max rados attr name len; cannot go higher than 100 chars for file system backends OPTION(osd_max_attr_size, OPT_U64, 0) @@ -1026,6 +1028,18 @@ OPTION(filestore_max_inline_xattrs_btrfs, OPT_U32, 10) OPTION(filestore_max_inline_xattrs_other, OPT_U32, 2) +// max xattr value size +OPTION(filestore_max_xattr_value_size, OPT_U32, 0) //Override +OPTION(filestore_max_xattr_value_size_xfs, OPT_U32, 64<<10) +OPTION(filestore_max_xattr_value_size_btrfs, OPT_U32, 64<<10) +// ext4 allows 4k xattrs total including some smallish extra fields and the +// keys. We're allowing 2 512 inline attrs in addition some some filestore +// replay attrs. After accounting for those, we still need to fit up to +// two attrs of this value. That means we need this value to be around 1k +// to be safe. This is hacky, but it's not worth complicating the code +// to work around ext4's total xattr limit. +OPTION(filestore_max_xattr_value_size_other, OPT_U32, 1<<10) + OPTION(filestore_sloppy_crc, OPT_BOOL, false) // track sloppy crcs OPTION(filestore_sloppy_crc_block_size, OPT_INT, 65536) diff -Nru ceph-10.1.1/src/common/fs_types.cc ceph-10.1.2/src/common/fs_types.cc --- ceph-10.1.1/src/common/fs_types.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/common/fs_types.cc 2016-04-12 17:46:05.000000000 +0000 @@ -46,8 +46,9 @@ stripe_count = fl.fl_stripe_count; object_size = fl.fl_object_size; pool_id = (int32_t)fl.fl_pg_pool; - // in the legacy encoding, pool 0 was undefined. - if (pool_id == 0) + // in the legacy encoding, a zeroed structure was the default and + // would have pool 0 instead of -1. + if (pool_id == 0 && stripe_unit == 0 && stripe_count == 0 && object_size == 0) pool_id = -1; pool_ns.clear(); } @@ -108,7 +109,7 @@ f->dump_unsigned("stripe_unit", stripe_unit); f->dump_unsigned("stripe_count", stripe_count); f->dump_unsigned("object_size", object_size); - f->dump_unsigned("pool_id", pool_id); + f->dump_int("pool_id", pool_id); f->dump_string("pool_ns", pool_ns); } @@ -122,3 +123,12 @@ o.back()->pool_id = 3; o.back()->pool_ns = "myns"; } + +ostream& operator<<(ostream& out, const file_layout_t &layout) +{ + JSONFormatter f; + layout.dump(&f); + f.flush(out); + return out; +} + diff -Nru ceph-10.1.1/src/common/Thread.cc ceph-10.1.2/src/common/Thread.cc --- ceph-10.1.1/src/common/Thread.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/common/Thread.cc 2016-04-12 17:46:05.000000000 +0000 @@ -200,8 +200,9 @@ int Thread::set_affinity(int id) { + int r = 0; cpuid = id; if (pid && ceph_gettid() == pid) - _set_affinity(id); - return 0; + r = _set_affinity(id); + return r; } diff -Nru ceph-10.1.1/src/crush/CrushWrapper.cc ceph-10.1.2/src/crush/CrushWrapper.cc --- ceph-10.1.1/src/crush/CrushWrapper.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/crush/CrushWrapper.cc 2016-04-12 17:46:05.000000000 +0000 @@ -353,7 +353,7 @@ if (item < 0 && !unlink_only) { crush_bucket *t = get_bucket(item); if (t && t->size) { - ldout(cct, 1) << "remove_item_undef bucket " << item << " has " << t->size + ldout(cct, 1) << "remove_item_under bucket " << item << " has " << t->size << " items, not empty" << dendl; return -ENOTEMPTY; } diff -Nru ceph-10.1.1/src/.git_version ceph-10.1.2/src/.git_version --- ceph-10.1.1/src/.git_version 2016-04-06 14:21:14.000000000 +0000 +++ ceph-10.1.2/src/.git_version 2016-04-12 17:54:13.000000000 +0000 @@ -1,2 +1,2 @@ -ce50389b773fe7f72fca40a3dd69cfe6613eaeb1 -v10.1.1 +4a2a6f72640d6b74a3bbd92798bb913ed380dcd4 +v10.1.2 diff -Nru ceph-10.1.1/src/include/ceph_fs.h ceph-10.1.2/src/include/ceph_fs.h --- ceph-10.1.1/src/include/ceph_fs.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/include/ceph_fs.h 2016-04-12 17:46:05.000000000 +0000 @@ -234,6 +234,11 @@ */ #define CEPH_MDSMAP_DOWN (1<<0) /* cluster deliberately down */ #define CEPH_MDSMAP_ALLOW_SNAPS (1<<1) /* cluster allowed to create snapshots */ +#define CEPH_MDSMAP_ALLOW_MULTIMDS (1<<2) /* cluster allowed to have >1 active MDS */ +#define CEPH_MDSMAP_ALLOW_DIRFRAGS (1<<3) /* cluster allowed to fragment directories */ + +#define CEPH_MDSMAP_ALLOW_CLASSICS (CEPH_MDSMAP_ALLOW_SNAPS | CEPH_MDSMAP_ALLOW_MULTIMDS | \ + CEPH_MDSMAP_ALLOW_DIRFRAGS) /* * mds states diff -Nru ceph-10.1.1/src/include/fs_types.h ceph-10.1.2/src/include/fs_types.h --- ceph-10.1.1/src/include/fs_types.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/include/fs_types.h 2016-04-12 17:46:05.000000000 +0000 @@ -103,4 +103,6 @@ WRITE_EQ_OPERATORS_5(file_layout_t, stripe_unit, stripe_count, object_size, pool_id, pool_ns); +ostream& operator<<(ostream& out, const file_layout_t &layout); + #endif diff -Nru ceph-10.1.1/src/include/rados.h ceph-10.1.2/src/include/rados.h --- ceph-10.1.1/src/include/rados.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/include/rados.h 2016-04-12 17:46:05.000000000 +0000 @@ -446,10 +446,6 @@ }; enum { - CEPH_OSD_COPY_GET_FLAG_NOTSUPP_OMAP = 1, /* mean dest pool don't support omap*/ -}; - -enum { CEPH_OSD_TMAP2OMAP_NULLOK = 1, }; @@ -516,7 +512,6 @@ } __attribute__ ((packed)) clonerange; struct { __le64 max; /* max data in reply */ - __le32 flags; } __attribute__ ((packed)) copy_get; struct { __le64 snapid; diff -Nru ceph-10.1.1/src/journal/FutureImpl.cc ceph-10.1.2/src/journal/FutureImpl.cc --- ceph-10.1.1/src/journal/FutureImpl.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/journal/FutureImpl.cc 2016-04-12 17:46:05.000000000 +0000 @@ -2,15 +2,14 @@ // vim: ts=8 sw=2 smarttab #include "journal/FutureImpl.h" -#include "journal/JournalMetadata.h" #include "journal/Utils.h" namespace journal { -FutureImpl::FutureImpl(JournalMetadataPtr journal_metadata, uint64_t tag_tid, - uint64_t entry_tid, uint64_t commit_tid) - : RefCountedObject(NULL, 0), m_journal_metadata(journal_metadata), - m_tag_tid(tag_tid), m_entry_tid(entry_tid), m_commit_tid(commit_tid), +FutureImpl::FutureImpl(uint64_t tag_tid, uint64_t entry_tid, + uint64_t commit_tid) + : RefCountedObject(NULL, 0), m_tag_tid(tag_tid), m_entry_tid(entry_tid), + m_commit_tid(commit_tid), m_lock(utils::unique_lock_name("FutureImpl::m_lock", this)), m_safe(false), m_consistent(false), m_return_value(0), m_flush_state(FLUSH_STATE_NONE), m_consistent_ack(this) { @@ -51,7 +50,7 @@ } if (complete && on_safe != NULL) { - m_journal_metadata->queue(on_safe, m_return_value); + on_safe->complete(m_return_value); } else if (flush_handler) { // attached to journal object -- instruct it to flush all entries through // this one. possible to become detached while lock is released, so flush @@ -69,7 +68,8 @@ return; } } - m_journal_metadata->queue(on_safe, m_return_value); + + on_safe->complete(m_return_value); } bool FutureImpl::is_complete() const { diff -Nru ceph-10.1.1/src/journal/FutureImpl.h ceph-10.1.2/src/journal/FutureImpl.h --- ceph-10.1.1/src/journal/FutureImpl.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/journal/FutureImpl.h 2016-04-12 17:46:05.000000000 +0000 @@ -18,7 +18,6 @@ namespace journal { class FutureImpl; -class JournalMetadata; typedef boost::intrusive_ptr FutureImplPtr; class FutureImpl : public RefCountedObject, boost::noncopyable { @@ -29,11 +28,9 @@ virtual void get() = 0; virtual void put() = 0; }; - typedef boost::intrusive_ptr JournalMetadataPtr; typedef boost::intrusive_ptr FlushHandlerPtr; - FutureImpl(JournalMetadataPtr journal_metadata, uint64_t tag_tid, - uint64_t entry_tid, uint64_t commit_tid); + FutureImpl(uint64_t tag_tid, uint64_t entry_tid, uint64_t commit_tid); void init(const FutureImplPtr &prev_future); @@ -96,7 +93,6 @@ virtual void finish(int r) {} }; - JournalMetadataPtr m_journal_metadata; uint64_t m_tag_tid; uint64_t m_entry_tid; uint64_t m_commit_tid; diff -Nru ceph-10.1.1/src/journal/JournalRecorder.cc ceph-10.1.2/src/journal/JournalRecorder.cc --- ceph-10.1.1/src/journal/JournalRecorder.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/journal/JournalRecorder.cc 2016-04-12 17:46:05.000000000 +0000 @@ -80,8 +80,7 @@ ObjectRecorderPtr object_ptr = get_object(splay_offset); uint64_t commit_tid = m_journal_metadata->allocate_commit_tid( object_ptr->get_object_number(), tag_tid, entry_tid); - FutureImplPtr future(new FutureImpl(m_journal_metadata, tag_tid, entry_tid, - commit_tid)); + FutureImplPtr future(new FutureImpl(tag_tid, entry_tid, commit_tid)); future->init(m_prev_future); m_prev_future = future; diff -Nru ceph-10.1.1/src/journal/JournalTrimmer.cc ceph-10.1.2/src/journal/JournalTrimmer.cc --- ceph-10.1.1/src/journal/JournalTrimmer.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/journal/JournalTrimmer.cc 2016-04-12 17:46:05.000000000 +0000 @@ -194,7 +194,8 @@ void JournalTrimmer::C_RemoveSet::complete(int r) { lock.Lock(); - if (r < 0 && r != -ENOENT && return_value == -ENOENT) { + if (r < 0 && r != -ENOENT && + (return_value == -ENOENT || return_value == 0)) { return_value = r; } else if (r == 0 && return_value == -ENOENT) { return_value = 0; diff -Nru ceph-10.1.1/src/journal/ObjectPlayer.cc ceph-10.1.2/src/journal/ObjectPlayer.cc --- ceph-10.1.1/src/journal/ObjectPlayer.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/journal/ObjectPlayer.cc 2016-04-12 17:46:05.000000000 +0000 @@ -70,9 +70,15 @@ void ObjectPlayer::unwatch() { ldout(m_cct, 20) << __func__ << ": " << m_oid << " unwatch" << dendl; Mutex::Locker timer_locker(m_timer_lock); + cancel_watch(); - m_watch_ctx = NULL; + Context *watch_ctx = nullptr; + std::swap(watch_ctx, m_watch_ctx); + if (watch_ctx != nullptr) { + delete watch_ctx; + } + while (m_watch_in_progress) { m_watch_in_progress_cond.Wait(m_timer_lock); } @@ -202,18 +208,17 @@ ldout(m_cct, 10) << __func__ << ": " << m_oid << " poll complete, r=" << r << dendl; - Context *on_finish = NULL; + Context *on_finish = nullptr; { Mutex::Locker timer_locker(m_timer_lock); assert(m_watch_in_progress); if (r == -ENOENT) { r = 0; } - on_finish = m_watch_ctx; - m_watch_ctx = NULL; + std::swap(on_finish, m_watch_ctx); } - if (on_finish != NULL) { + if (on_finish != nullptr) { on_finish->complete(r); } diff -Nru ceph-10.1.1/src/librados/IoCtxImpl.cc ceph-10.1.2/src/librados/IoCtxImpl.cc --- ceph-10.1.1/src/librados/IoCtxImpl.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/librados/IoCtxImpl.cc 2016-04-12 17:46:05.000000000 +0000 @@ -678,7 +678,8 @@ Context *oncommit = new C_SafeCond(&mylock, &cond, &done, &r); int op = o->ops[0].op.op; - ldout(client->cct, 10) << ceph_osd_op_name(op) << " oid=" << oid << " nspace=" << oloc.nspace << dendl; + ldout(client->cct, 10) << ceph_osd_op_name(op) << " oid=" << oid + << " nspace=" << oloc.nspace << dendl; Objecter::Op *objecter_op = objecter->prepare_mutate_op(oid, oloc, *o, snapc, ut, flags, NULL, oncommit, &ver); @@ -745,7 +746,7 @@ Objecter::Op *objecter_op = objecter->prepare_read_op(oid, oloc, *o, snap_seq, pbl, flags, onack, &c->objver); - c->tid = objecter->op_submit(objecter_op); + objecter->op_submit(objecter_op, &c->tid); return 0; } @@ -764,8 +765,10 @@ c->io = this; queue_aio_write(c); - c->tid = objecter->mutate(oid, oloc, *o, snap_context, ut, flags, onack, - oncommit, &c->objver); + Objecter::Op *op = objecter->prepare_mutate_op( + oid, oloc, *o, snap_context, ut, flags, onack, + oncommit, &c->objver); + objecter->op_submit(op, &c->tid); return 0; } @@ -783,9 +786,11 @@ c->io = this; c->blp = pbl; - c->tid = objecter->read(oid, oloc, - off, len, snapid, pbl, 0, - onack, &c->objver); + Objecter::Op *o = objecter->prepare_read_op( + oid, oloc, + off, len, snapid, pbl, 0, + onack, &c->objver); + objecter->op_submit(o, &c->tid); return 0; } @@ -804,10 +809,11 @@ c->bl.push_back(buffer::create_static(len, buf)); c->blp = &c->bl; - c->tid = objecter->read(oid, oloc, - off, len, snapid, &c->bl, 0, - onack, &c->objver); - + Objecter::Op *o = objecter->prepare_read_op( + oid, oloc, + off, len, snapid, &c->bl, 0, + onack, &c->objver); + objecter->op_submit(o, &c->tid); return 0; } @@ -839,9 +845,11 @@ onack->m_ops.sparse_read(off, len, m, data_bl, NULL); - c->tid = objecter->read(oid, oloc, - onack->m_ops, snap_seq, NULL, 0, - onack, &c->objver); + Objecter::Op *o = objecter->prepare_read_op( + oid, oloc, + onack->m_ops, snap_seq, NULL, 0, + onack, &c->objver); + objecter->op_submit(o, &c->tid); return 0; } @@ -864,9 +872,11 @@ c->io = this; queue_aio_write(c); - c->tid = objecter->write(oid, oloc, - off, len, snapc, bl, ut, 0, - onack, onsafe, &c->objver); + Objecter::Op *o = objecter->prepare_write_op( + oid, oloc, + off, len, snapc, bl, ut, 0, + onack, onsafe, &c->objver); + objecter->op_submit(o, &c->tid); return 0; } @@ -888,9 +898,11 @@ c->io = this; queue_aio_write(c); - c->tid = objecter->append(oid, oloc, - len, snapc, bl, ut, 0, - onack, onsafe, &c->objver); + Objecter::Op *o = objecter->prepare_append_op( + oid, oloc, + len, snapc, bl, ut, 0, + onack, onsafe, &c->objver); + objecter->op_submit(o, &c->tid); return 0; } @@ -913,9 +925,11 @@ c->io = this; queue_aio_write(c); - c->tid = objecter->write_full(oid, oloc, - snapc, bl, ut, 0, - onack, onsafe, &c->objver); + Objecter::Op *o = objecter->prepare_write_full_op( + oid, oloc, + snapc, bl, ut, 0, + onack, onsafe, &c->objver); + objecter->op_submit(o, &c->tid); return 0; } @@ -934,9 +948,11 @@ c->io = this; queue_aio_write(c); - c->tid = objecter->remove(oid, oloc, - snapc, ut, 0, - onack, onsafe, &c->objver); + Objecter::Op *o = objecter->prepare_remove_op( + oid, oloc, + snapc, ut, 0, + onack, onsafe, &c->objver); + objecter->op_submit(o, &c->tid); return 0; } @@ -948,9 +964,11 @@ C_aio_stat_Ack *onack = new C_aio_stat_Ack(c, pmtime); c->io = this; - c->tid = objecter->stat(oid, oloc, - snap_seq, psize, &onack->mtime, 0, - onack, &c->objver); + Objecter::Op *o = objecter->prepare_stat_op( + oid, oloc, + snap_seq, psize, &onack->mtime, 0, + onack, &c->objver); + objecter->op_submit(o, &c->tid); return 0; } @@ -961,9 +979,11 @@ C_aio_stat2_Ack *onack = new C_aio_stat2_Ack(c, pts); c->io = this; - c->tid = objecter->stat(oid, oloc, - snap_seq, psize, &onack->mtime, 0, - onack, &c->objver); + Objecter::Op *o = objecter->prepare_stat_op( + oid, oloc, + snap_seq, psize, &onack->mtime, 0, + onack, &c->objver); + objecter->op_submit(o, &c->tid); return 0; } @@ -984,7 +1004,9 @@ ::ObjectOperation rd; rd.hit_set_ls(pls, NULL); object_locator_t oloc(poolid); - c->tid = objecter->pg_read(hash, oloc, rd, NULL, 0, onack, NULL, NULL); + Objecter::Op *o = objecter->prepare_pg_read_op( + hash, oloc, rd, NULL, 0, onack, NULL, NULL); + objecter->op_submit(o, &c->tid); return 0; } @@ -999,7 +1021,9 @@ ::ObjectOperation rd; rd.hit_set_get(ceph::real_clock::from_time_t(stamp), pbl, 0); object_locator_t oloc(poolid); - c->tid = objecter->pg_read(hash, oloc, rd, NULL, 0, onack, NULL, NULL); + Objecter::Op *o = objecter->prepare_pg_read_op( + hash, oloc, rd, NULL, 0, onack, NULL, NULL); + objecter->op_submit(o, &c->tid); return 0; } @@ -1041,8 +1065,10 @@ ::ObjectOperation op; op.scrub_ls(start_after, max_to_get, objects, interval, nullptr); object_locator_t oloc{poolid, pg.ps()}; - c->tid = objecter->pg_read(oloc.hash, oloc, op, nullptr, CEPH_OSD_FLAG_PGOP, onack, - nullptr, nullptr); + Objecter::Op *o = objecter->prepare_pg_read_op( + oloc.hash, oloc, op, nullptr, CEPH_OSD_FLAG_PGOP, onack, + nullptr, nullptr); + objecter->op_submit(o, &c->tid); return 0; } @@ -1060,8 +1086,10 @@ ::ObjectOperation op; op.scrub_ls(start_after, max_to_get, snapsets, interval, nullptr); object_locator_t oloc{poolid, pg.ps()}; - c->tid = objecter->pg_read(oloc.hash, oloc, op, nullptr, CEPH_OSD_FLAG_PGOP, onack, - nullptr, nullptr); + Objecter::Op *o = objecter->prepare_pg_read_op( + oloc.hash, oloc, op, nullptr, CEPH_OSD_FLAG_PGOP, onack, + nullptr, nullptr); + objecter->op_submit(o, &c->tid); return 0; } @@ -1119,8 +1147,9 @@ ::ObjectOperation rd; prepare_assert_ops(&rd); rd.call(cls, method, inbl); - c->tid = objecter->read(oid, oloc, rd, snap_seq, outbl, 0, onack, &c->objver); - + Objecter::Op *o = objecter->prepare_read_op( + oid, oloc, rd, snap_seq, outbl, 0, onack, &c->objver); + objecter->op_submit(o, &c->tid); return 0; } diff -Nru ceph-10.1.1/src/librbd/ImageCtx.cc ceph-10.1.2/src/librbd/ImageCtx.cc --- ceph-10.1.1/src/librbd/ImageCtx.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/librbd/ImageCtx.cc 2016-04-12 17:46:05.000000000 +0000 @@ -52,7 +52,7 @@ class ThreadPoolSingleton : public ThreadPool { public: explicit ThreadPoolSingleton(CephContext *cct) - : ThreadPool(cct, "librbd::thread_pool", "tp_librbd", cct->_conf->rbd_op_threads, + : ThreadPool(cct, "librbd::thread_pool", "tp_librbd", 1, "rbd_op_threads") { start(); } diff -Nru ceph-10.1.1/src/librbd/internal.cc ceph-10.1.2/src/librbd/internal.cc --- ceph-10.1.1/src/librbd/internal.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/librbd/internal.cc 2016-04-12 17:46:05.000000000 +0000 @@ -312,13 +312,10 @@ return r; } - if (!is_primary) { - if (!force) { - lderr(cct) << "Mirrored image is not the primary, add force option to" - " disable mirroring" << dendl; - return -EINVAL; - } - goto remove_mirroring_image; + if (!is_primary && !force) { + lderr(cct) << "Mirrored image is not the primary, add force option to" + " disable mirroring" << dendl; + return -EINVAL; } mirror_image_internal.state = cls::rbd::MIRROR_IMAGE_STATE_DISABLING; @@ -329,6 +326,10 @@ return r; } + if (!is_primary) { + goto remove_mirroring_image; + } + r = MirroringWatcher<>::notify_image_updated( ictx->md_ctx, cls::rbd::MIRROR_IMAGE_STATE_DISABLING, ictx->id, mirror_image_internal.global_image_id); @@ -341,6 +342,7 @@ header_oid = ::journal::Journaler::header_oid(ictx->id); while(true) { + clients.clear(); r = cls::journal::client::client_list(ictx->md_ctx, header_oid, &clients); if (r < 0) { lderr(cct) << "cannot disable mirroring: " << cpp_strerror(r) << dendl; diff -Nru ceph-10.1.1/src/librbd/Journal.cc ceph-10.1.2/src/librbd/Journal.cc --- ceph-10.1.1/src/librbd/Journal.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/librbd/Journal.cc 2016-04-12 17:46:05.000000000 +0000 @@ -827,7 +827,8 @@ << "length=" << length << ", " << "flush=" << flush_entry << ", tid=" << tid << dendl; - Context *on_safe = new C_IOEventSafe(this, tid); + Context *on_safe = create_async_context_callback( + m_image_ctx, new C_IOEventSafe(this, tid)); if (flush_entry) { future.flush(on_safe); } else { @@ -942,8 +943,9 @@ op_finish_future = m_journaler->append(m_tag_tid, bl); } - op_finish_future.flush(new C_OpEventSafe(this, op_tid, op_start_future, - op_finish_future)); + op_finish_future.flush(create_async_context_callback( + m_image_ctx, new C_OpEventSafe(this, op_tid, op_start_future, + op_finish_future))); } template @@ -971,7 +973,7 @@ } if (future.is_valid()) { - future.flush(NULL); + future.flush(nullptr); } } diff -Nru ceph-10.1.1/src/logrotate.conf ceph-10.1.2/src/logrotate.conf --- ceph-10.1.1/src/logrotate.conf 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/logrotate.conf 2016-04-12 17:46:05.000000000 +0000 @@ -4,7 +4,7 @@ compress sharedscripts postrotate - killall -q -1 ceph-mon ceph-mds ceph-osd radosgw || true + killall -q -1 ceph-mon ceph-mds ceph-osd ceph-fuse radosgw || true endscript missingok notifempty diff -Nru ceph-10.1.1/src/Makefile-env.am ceph-10.1.2/src/Makefile-env.am --- ceph-10.1.1/src/Makefile-env.am 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/Makefile-env.am 2016-04-12 17:46:05.000000000 +0000 @@ -299,6 +299,8 @@ # put virtualenvs in this directory # otherwise it may overflow #! 80 kernel limit -export CEPH_BUILD_VIRTUALENV = /tmp +# beware that some build environments might not be able to write to /tmp +export TMPDIR ?= /tmp +export CEPH_BUILD_VIRTUALENV = $(TMPDIR) radoslibdir = $(libdir)/rados-classes diff -Nru ceph-10.1.1/src/Makefile.in ceph-10.1.2/src/Makefile.in --- ceph-10.1.1/src/Makefile.in 2016-04-06 14:20:26.000000000 +0000 +++ ceph-10.1.2/src/Makefile.in 2016-04-12 17:53:21.000000000 +0000 @@ -5153,6 +5153,8 @@ rbd_OBJECTS = $(am_rbd_OBJECTS) @ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@rbd_DEPENDENCIES = libjournal.la \ @ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ libcls_journal_client.la \ +@ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ libcls_rbd_client.la \ +@ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ libcls_lock_client.la \ @ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ $(LIBRBD) \ @ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ $(LIBRBD_TYPES) \ @ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ $(LIBRADOS) \ @@ -12261,6 +12263,8 @@ @ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ $(am__append_267) @ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@rbd_LDADD = libjournal.la \ @ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ libcls_journal_client.la \ +@ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ libcls_rbd_client.la \ +@ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ libcls_lock_client.la \ @ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ $(LIBRBD) \ @ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ $(LIBRBD_TYPES) \ @ENABLE_CLIENT_TRUE@@WITH_RADOS_TRUE@@WITH_RBD_TRUE@ $(LIBRADOS) \ @@ -32388,7 +32392,9 @@ # put virtualenvs in this directory # otherwise it may overflow #! 80 kernel limit -export CEPH_BUILD_VIRTUALENV = /tmp +# beware that some build environments might not be able to write to /tmp +export TMPDIR ?= /tmp +export CEPH_BUILD_VIRTUALENV = $(TMPDIR) @NO_GIT_VERSION_TRUE@export NO_VERSION="yes" diff -Nru ceph-10.1.1/src/mds/Beacon.cc ceph-10.1.2/src/mds/Beacon.cc --- ceph-10.1.1/src/mds/Beacon.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/Beacon.cc 2016-04-12 17:46:05.000000000 +0000 @@ -443,6 +443,13 @@ large_completed_requests_metrics.clear(); } } + + // Report a health warning if we are readonly + if (mds->mdcache->is_readonly()) { + MDSHealthMetric m(MDS_HEALTH_READ_ONLY, HEALTH_WARN, + "MDS in read-only mode"); + health.metrics.push_back(m); + } } MDSMap::DaemonState Beacon::get_want_state() const diff -Nru ceph-10.1.1/src/mds/CInode.cc ceph-10.1.2/src/mds/CInode.cc --- ceph-10.1.1/src/mds/CInode.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/CInode.cc 2016-04-12 17:46:05.000000000 +0000 @@ -1143,13 +1143,7 @@ auth_pin(this); - int64_t pool; - if (is_dir()) { - pool = mdcache->mds->mdsmap->get_metadata_pool(); - } else { - pool = inode.layout.pool_id; - } - + const int64_t pool = get_backtrace_pool(); inode_backtrace_t bt; build_backtrace(pool, bt); bufferlist parent_bl; @@ -1212,8 +1206,10 @@ { if (r < 0) { dout(1) << "store backtrace error " << r << " v " << v << dendl; - mdcache->mds->clog->error() << "failed to store backtrace on dir ino " - << ino() << " object, errno " << r << "\n"; + mdcache->mds->clog->error() << "failed to store backtrace on ino " + << ino() << " object" + << ", pool " << get_backtrace_pool() + << ", errno " << r << "\n"; mdcache->mds->handle_write_error(r); return; } @@ -1229,13 +1225,7 @@ void CInode::fetch_backtrace(Context *fin, bufferlist *backtrace) { - int64_t pool; - if (is_dir()) - pool = mdcache->mds->mdsmap->get_metadata_pool(); - else - pool = inode.layout.pool_id; - - mdcache->fetch_backtrace(inode.ino, pool, *backtrace, fin); + mdcache->fetch_backtrace(inode.ino, get_backtrace_pool(), *backtrace, fin); } void CInode::_mark_dirty_parent(LogSegment *ls, bool dirty_pool) @@ -3725,12 +3715,7 @@ void fetch_backtrace_and_tag(CInode *in, std::string tag, Context *fin, int *bt_r, bufferlist *bt) { - int64_t pool; - if (in->is_dir()) - pool = in->mdcache->mds->mdsmap->get_metadata_pool(); - else - pool = in->inode.layout.pool_id; - + const int64_t pool = in->get_backtrace_pool(); object_t oid = CInode::get_object_name(in->ino(), frag_t(), ""); ObjectOperation fetch; @@ -3788,11 +3773,7 @@ results->performed_validation = true; // at least, some of it! results->backtrace.checked = true; - int64_t pool; - if (in->is_dir()) - pool = in->mdcache->mds->mdsmap->get_metadata_pool(); - else - pool = in->inode.layout.pool_id; + const int64_t pool = in->get_backtrace_pool(); inode_backtrace_t& memory_backtrace = results->backtrace.memory_value; in->build_backtrace(pool, memory_backtrace); bool equivalent, divergent; @@ -4302,3 +4283,15 @@ clog->info() << "scrub complete with tag '" << scrub_infop->header->tag << "'"; } } + +int64_t CInode::get_backtrace_pool() const +{ + if (is_dir()) { + return mdcache->mds->mdsmap->get_metadata_pool(); + } else { + // Files are required to have an explicit layout that specifies + // a pool + assert(inode.layout.pool_id != -1); + return inode.layout.pool_id; + } +} diff -Nru ceph-10.1.1/src/mds/CInode.h ceph-10.1.2/src/mds/CInode.h --- ceph-10.1.1/src/mds/CInode.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/CInode.h 2016-04-12 17:46:05.000000000 +0000 @@ -769,6 +769,15 @@ void store_backtrace(MDSInternalContextBase *fin, int op_prio=-1); void _stored_backtrace(int r, version_t v, Context *fin); void fetch_backtrace(Context *fin, bufferlist *backtrace); +protected: + /** + * Return the pool ID where we currently write backtraces for + * this inode (in addition to inode.old_pools) + * + * @returns a pool ID >=0 + */ + int64_t get_backtrace_pool() const; +public: void _mark_dirty_parent(LogSegment *ls, bool dirty_pool=false); void clear_dirty_parent(); void verify_diri_backtrace(bufferlist &bl, int err); diff -Nru ceph-10.1.1/src/mds/events/ESessions.h ceph-10.1.2/src/mds/events/ESessions.h --- ceph-10.1.1/src/mds/events/ESessions.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/events/ESessions.h 2016-04-12 17:46:05.000000000 +0000 @@ -28,7 +28,7 @@ map client_map; bool old_style_encode; - ESessions() : LogEvent(EVENT_SESSIONS), old_style_encode(false) { } + ESessions() : LogEvent(EVENT_SESSIONS), cmapv(0), old_style_encode(false) { } ESessions(version_t pv, map& cm) : LogEvent(EVENT_SESSIONS), cmapv(pv), diff -Nru ceph-10.1.1/src/mds/FSMap.cc ceph-10.1.2/src/mds/FSMap.cc --- ceph-10.1.1/src/mds/FSMap.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/FSMap.cc 2016-04-12 17:46:05.000000000 +0000 @@ -35,6 +35,11 @@ compat.dump(f); f->close_section(); + f->open_object_section("feature flags"); + f->dump_bool("enable_multiple", enable_multiple); + f->dump_bool("ever_enabled_multiple", ever_enabled_multiple); + f->close_section(); + f->open_array_section("standbys"); for (const auto &i : standby_daemons) { f->open_object_section("info"); @@ -76,8 +81,9 @@ void FSMap::print(ostream& out) const { out << "e" << epoch << std::endl; - out << "enable_multiple: " << enable_multiple << std::endl; - out << "compat: " << enable_multiple << std::endl; + out << "enable_multiple, ever_enabled_multiple: " << enable_multiple << "," + << ever_enabled_multiple << std::endl; + out << "compat: " << compat << std::endl; out << " " << std::endl; if (filesystems.empty()) { @@ -231,7 +237,7 @@ void FSMap::encode(bufferlist& bl, uint64_t features) const { if (features & CEPH_FEATURE_SERVER_JEWEL) { - ENCODE_START(6, 6, bl); + ENCODE_START(7, 6, bl); ::encode(epoch, bl); ::encode(next_filesystem_id, bl); ::encode(legacy_client_fscid, bl); @@ -245,6 +251,7 @@ ::encode(mds_roles, bl); ::encode(standby_daemons, bl, features); ::encode(standby_epochs, bl); + ::encode(ever_enabled_multiple, bl); ENCODE_FINISH(bl); } else { if (filesystems.empty()) { @@ -280,7 +287,7 @@ // MDSMonitor to store an FSMap instead of an MDSMap was // 5, so anything older than 6 is decoded as an MDSMap, // and anything newer is decoded as an FSMap. - DECODE_START_LEGACY_COMPAT_LEN_16(6, 4, 4, p); + DECODE_START_LEGACY_COMPAT_LEN_16(7, 4, 4, p); if (struct_v < 6) { // Decoding an MDSMap (upgrade) ::decode(epoch, p); @@ -334,11 +341,28 @@ if (ev >= 4) ::decode(legacy_mds_map.last_failure_osd_epoch, p); if (ev >= 6) { - ::decode(legacy_mds_map.ever_allowed_snaps, p); - ::decode(legacy_mds_map.explicitly_allowed_snaps, p); + if (ev < 10) { + // previously this was a bool about snaps, not a flag map + bool flag; + ::decode(flag, p); + legacy_mds_map.ever_allowed_features = flag ? + CEPH_MDSMAP_ALLOW_SNAPS : 0; + ::decode(flag, p); + legacy_mds_map.explicitly_allowed_features = flag ? + CEPH_MDSMAP_ALLOW_SNAPS : 0; + if (legacy_mds_map.max_mds > 1) { + legacy_mds_map.set_multimds_allowed(); + } + } else { + ::decode(legacy_mds_map.ever_allowed_features, p); + ::decode(legacy_mds_map.explicitly_allowed_features, p); + } } else { - legacy_mds_map.ever_allowed_snaps = true; - legacy_mds_map.explicitly_allowed_snaps = false; + legacy_mds_map.ever_allowed_features = CEPH_MDSMAP_ALLOW_CLASSICS; + legacy_mds_map.explicitly_allowed_features = 0; + if (legacy_mds_map.max_mds > 1) { + legacy_mds_map.set_multimds_allowed(); + } } if (ev >= 7) ::decode(legacy_mds_map.inline_data_enabled, p); @@ -416,6 +440,7 @@ ::decode(mds_roles, p); ::decode(standby_daemons, p); ::decode(standby_epochs, p); + ::decode(ever_enabled_multiple, p); } DECODE_FINISH(p); @@ -512,7 +537,8 @@ return result; } -mds_gid_t FSMap::find_unused(bool force_standby_active) const { +mds_gid_t FSMap::find_unused(fs_cluster_id_t fscid, + bool force_standby_active) const { for (const auto &i : standby_daemons) { const auto &gid = i.first; const auto &info = i.second; @@ -521,6 +547,10 @@ if (info.laggy() || info.rank >= 0) continue; + if (info.standby_for_fscid != FS_CLUSTER_ID_NONE && + info.standby_for_fscid != fscid) + continue; + if ((info.standby_for_rank == MDSMap::MDS_NO_STANDBY_PREF || info.standby_for_rank == MDSMap::MDS_MATCHED_ACTIVE || (info.standby_for_rank == MDSMap::MDS_STANDBY_ANY @@ -537,7 +567,7 @@ if (standby) return standby; else - return find_unused(force_standby_active); + return find_unused(role.fscid, force_standby_active); } void FSMap::sanity() const diff -Nru ceph-10.1.1/src/mds/FSMap.h ceph-10.1.2/src/mds/FSMap.h --- ceph-10.1.1/src/mds/FSMap.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/FSMap.h 2016-04-12 17:46:05.000000000 +0000 @@ -95,6 +95,7 @@ fs_cluster_id_t legacy_client_fscid; CompatSet compat; bool enable_multiple; + bool ever_enabled_multiple; // < the cluster had multiple MDSes enabled once std::map > filesystems; @@ -115,7 +116,7 @@ next_filesystem_id(FS_CLUSTER_ID_ANONYMOUS + 1), legacy_client_fscid(FS_CLUSTER_ID_NONE), compat(get_mdsmap_compat_set_default()), - enable_multiple(false) + enable_multiple(false), ever_enabled_multiple(false) { } FSMap(const FSMap &rhs) @@ -125,6 +126,7 @@ legacy_client_fscid(rhs.legacy_client_fscid), compat(rhs.compat), enable_multiple(rhs.enable_multiple), + ever_enabled_multiple(rhs.ever_enabled_multiple), mds_roles(rhs.mds_roles), standby_daemons(rhs.standby_daemons), standby_epochs(rhs.standby_epochs) @@ -159,6 +161,9 @@ void set_enable_multiple(const bool v) { enable_multiple = v; + if (true == v) { + ever_enabled_multiple = true; + } } bool get_enable_multiple() const @@ -411,7 +416,7 @@ mds_gid_t find_standby_for(mds_role_t mds, const std::string& name) const; - mds_gid_t find_unused(bool force_standby_active) const; + mds_gid_t find_unused(fs_cluster_id_t fscid, bool force_standby_active) const; mds_gid_t find_replacement_for(mds_role_t mds, const std::string& name, bool force_standby_active) const; diff -Nru ceph-10.1.1/src/mds/journal.cc ceph-10.1.2/src/mds/journal.cc --- ceph-10.1.1/src/mds/journal.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/journal.cc 2016-04-12 17:46:05.000000000 +0000 @@ -572,6 +572,25 @@ */ in->oldest_snap = oldest_snap; in->decode_snap_blob(snapbl); + + /* + * In case there was anything malformed in the journal that we are + * replaying, do sanity checks on the inodes we're replaying and + * go damaged instead of letting any trash into a live cache + */ + if (in->is_file()) { + // Files must have valid layouts with a pool set + if (in->inode.layout.pool_id == -1 || !in->inode.layout.is_valid()) { + dout(0) << "EMetaBlob.replay invalid layout on ino " << *in + << ": " << in->inode.layout << dendl; + std::ostringstream oss; + oss << "Invalid layout for inode 0x" << std::hex << in->inode.ino + << std::dec << " in journal"; + mds->clog->error() << oss.str(); + mds->damaged(); + assert(0); // Should be unreachable because damaged() calls respawn() + } + } } // EMetaBlob::remotebit diff -Nru ceph-10.1.1/src/mds/Locker.cc ceph-10.1.2/src/mds/Locker.cc --- ceph-10.1.1/src/mds/Locker.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/Locker.cc 2016-04-12 17:46:05.000000000 +0000 @@ -2187,6 +2187,7 @@ utime_t new_mtime) { assert(in->is_auth()); + assert(in->is_file()); inode_t *latest = in->get_projected_inode(); map new_ranges; diff -Nru ceph-10.1.1/src/mds/MDBalancer.cc ceph-10.1.2/src/mds/MDBalancer.cc --- ceph-10.1.1/src/mds/MDBalancer.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/MDBalancer.cc 2016-04-12 17:46:05.000000000 +0000 @@ -333,6 +333,7 @@ void MDBalancer::queue_split(CDir *dir) { + assert(mds->mdsmap->allows_dirfrags()); split_queue.insert(dir->dirfrag()); } @@ -984,6 +985,7 @@ // split if (g_conf->mds_bal_split_size > 0 && + mds->mdsmap->allows_dirfrags() && (dir->should_split() || (v > g_conf->mds_bal_split_rd && type == META_POP_IRD) || (v > g_conf->mds_bal_split_wr && type == META_POP_IWR)) && diff -Nru ceph-10.1.1/src/mds/MDCache.cc ceph-10.1.2/src/mds/MDCache.cc --- ceph-10.1.1/src/mds/MDCache.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/MDCache.cc 2016-04-12 17:46:05.000000000 +0000 @@ -6038,6 +6038,11 @@ CInode *in = p->second; if (!in->is_auth()) continue; + + // Only normal files need file size recovery + if (!in->is_file()) { + continue; + } bool recover = false; for (map::iterator p = in->inode.client_ranges.begin(); diff -Nru ceph-10.1.1/src/mds/MDSDaemon.cc ceph-10.1.2/src/mds/MDSDaemon.cc --- ceph-10.1.1/src/mds/MDSDaemon.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/MDSDaemon.cc 2016-04-12 17:46:05.000000000 +0000 @@ -337,11 +337,19 @@ admin_socket->unregister_command("dump_blocked_ops"); admin_socket->unregister_command("dump_historic_ops"); admin_socket->unregister_command("scrub_path"); + admin_socket->unregister_command("tag path"); admin_socket->unregister_command("flush_path"); + admin_socket->unregister_command("export dir"); + admin_socket->unregister_command("dump cache"); admin_socket->unregister_command("session evict"); + admin_socket->unregister_command("osdmap barrier"); admin_socket->unregister_command("session ls"); admin_socket->unregister_command("flush journal"); admin_socket->unregister_command("force_readonly"); + admin_socket->unregister_command("get subtrees"); + admin_socket->unregister_command("dirfrag split"); + admin_socket->unregister_command("dirfrag merge"); + admin_socket->unregister_command("dirfrag ls"); delete asok_hook; asok_hook = NULL; } @@ -474,8 +482,21 @@ mds_lock.Unlock(); return r; } + + int rotating_auth_attempts = 0; + const int max_rotating_auth_attempts = 10; + while (monc->wait_auth_rotating(30.0) < 0) { - derr << "unable to obtain rotating service keys; retrying" << dendl; + if (++rotating_auth_attempts <= max_rotating_auth_attempts) { + derr << "unable to obtain rotating service keys; retrying" << dendl; + continue; + } + derr << "ERROR: failed to refresh rotating keys, " + << "maximum retry time reached." << dendl; + mds_lock.Lock(); + suicide(); + mds_lock.Unlock(); + return -ETIMEDOUT; } objecter->start(); @@ -801,6 +822,7 @@ if (mds_rank == NULL) { r = -EINVAL; ss << "MDS not active"; + goto out; } // FIXME harmonize `session kill` with admin socket session evict int64_t session_id = 0; @@ -988,6 +1010,7 @@ // has taken our ID, we don't want to keep restarting and // fighting them for the ID. suicide(); + m->put(); return; } } @@ -1244,6 +1267,7 @@ if (mds_rank) { mds_rank->handle_osd_map(); } + m->put(); break; default: diff -Nru ceph-10.1.1/src/mds/MDSMap.cc ceph-10.1.2/src/mds/MDSMap.cc --- ceph-10.1.1/src/mds/MDSMap.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/MDSMap.cc 2016-04-12 17:46:05.000000000 +0000 @@ -134,6 +134,8 @@ { f->dump_int("epoch", epoch); f->dump_unsigned("flags", flags); + f->dump_unsigned("ever_allowed_features", ever_allowed_features); + f->dump_unsigned("explicitly_allowed_features", explicitly_allowed_features); f->dump_stream("created") << created; f->dump_stream("modified") << modified; f->dump_int("tableserver", tableserver); @@ -547,7 +549,7 @@ ::encode(cas_pool, bl); // kclient ignores everything from here - __u16 ev = 9; + __u16 ev = 10; ::encode(ev, bl); ::encode(compat, bl); ::encode(metadata_pool, bl); @@ -560,8 +562,8 @@ ::encode(failed, bl); ::encode(stopped, bl); ::encode(last_failure_osd_epoch, bl); - ::encode(ever_allowed_snaps, bl); - ::encode(explicitly_allowed_snaps, bl); + ::encode(ever_allowed_features, bl); + ::encode(explicitly_allowed_features, bl); ::encode(inline_data_enabled, bl); ::encode(enabled, bl); ::encode(fs_name, bl); @@ -624,11 +626,27 @@ if (ev >= 4) ::decode(last_failure_osd_epoch, p); if (ev >= 6) { - ::decode(ever_allowed_snaps, p); - ::decode(explicitly_allowed_snaps, p); + if (ev < 10) { + // previously this was a bool about snaps, not a flag map + bool flag; + ::decode(flag, p); + ever_allowed_features = flag ? CEPH_MDSMAP_ALLOW_SNAPS : 0; + ever_allowed_features |= CEPH_MDSMAP_ALLOW_MULTIMDS|CEPH_MDSMAP_ALLOW_DIRFRAGS; + ::decode(flag, p); + explicitly_allowed_features = flag ? CEPH_MDSMAP_ALLOW_SNAPS : 0; + if (max_mds > 1) { + set_multimds_allowed(); + } + } else { + ::decode(ever_allowed_features, p); + ::decode(explicitly_allowed_features, p); + } } else { - ever_allowed_snaps = true; - explicitly_allowed_snaps = false; + ever_allowed_features = CEPH_MDSMAP_ALLOW_CLASSICS; + explicitly_allowed_features = 0; + if (max_mds > 1) { + set_multimds_allowed(); + } } if (ev >= 7) ::decode(inline_data_enabled, p); diff -Nru ceph-10.1.1/src/mds/MDSMap.h ceph-10.1.2/src/mds/MDSMap.h --- ceph-10.1.1/src/mds/MDSMap.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/MDSMap.h 2016-04-12 17:46:05.000000000 +0000 @@ -209,8 +209,8 @@ std::map up; // who is in those roles std::map mds_info; - bool ever_allowed_snaps; //< the cluster has ever allowed snap creation - bool explicitly_allowed_snaps; //< the user has explicitly enabled snap creation + uint8_t ever_allowed_features; //< bitmap of features the cluster has allowed + uint8_t explicitly_allowed_features; //< bitmap of features explicitly enabled bool inline_data_enabled; @@ -235,8 +235,8 @@ cas_pool(-1), metadata_pool(0), max_mds(0), - ever_allowed_snaps(false), - explicitly_allowed_snaps(false), + ever_allowed_features(0), + explicitly_allowed_features(0), inline_data_enabled(false), cached_up_features(0) { } @@ -259,11 +259,27 @@ void set_snaps_allowed() { set_flag(CEPH_MDSMAP_ALLOW_SNAPS); - ever_allowed_snaps = true; - explicitly_allowed_snaps = true; + ever_allowed_features |= CEPH_MDSMAP_ALLOW_SNAPS; + explicitly_allowed_features |= CEPH_MDSMAP_ALLOW_SNAPS; } - bool allows_snaps() { return test_flag(CEPH_MDSMAP_ALLOW_SNAPS); } void clear_snaps_allowed() { clear_flag(CEPH_MDSMAP_ALLOW_SNAPS); } + bool allows_snaps() const { return test_flag(CEPH_MDSMAP_ALLOW_SNAPS); } + + void set_multimds_allowed() { + set_flag(CEPH_MDSMAP_ALLOW_MULTIMDS); + ever_allowed_features |= CEPH_MDSMAP_ALLOW_MULTIMDS; + explicitly_allowed_features |= CEPH_MDSMAP_ALLOW_MULTIMDS; + } + void clear_multimds_allowed() { clear_flag(CEPH_MDSMAP_ALLOW_MULTIMDS); } + bool allows_multimds() const { return test_flag(CEPH_MDSMAP_ALLOW_MULTIMDS); } + + void set_dirfrags_allowed() { + set_flag(CEPH_MDSMAP_ALLOW_DIRFRAGS); + ever_allowed_features |= CEPH_MDSMAP_ALLOW_DIRFRAGS; + explicitly_allowed_features |= CEPH_MDSMAP_ALLOW_DIRFRAGS; + } + void clear_dirfrags_allowed() { clear_flag(CEPH_MDSMAP_ALLOW_DIRFRAGS); } + bool allows_dirfrags() const { return test_flag(CEPH_MDSMAP_ALLOW_DIRFRAGS); } epoch_t get_epoch() const { return epoch; } void inc_epoch() { epoch++; } diff -Nru ceph-10.1.1/src/mds/MDSRank.cc ceph-10.1.2/src/mds/MDSRank.cc --- ceph-10.1.1/src/mds/MDSRank.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mds/MDSRank.cc 2016-04-12 17:46:05.000000000 +0000 @@ -502,7 +502,8 @@ if (!dir->get_parent_dir()) continue; // must be linked. if (!dir->is_auth()) continue; // must be auth. frag_t fg = dir->get_frag(); - if (fg == frag_t() || (rand() % (1 << fg.bits()) == 0)) + if (mdsmap->allows_dirfrags() && + (fg == frag_t() || (rand() % (1 << fg.bits()) == 0))) mdcache->split_dir(dir, 1); else balancer->queue_merge(dir); @@ -2166,6 +2167,11 @@ cmdmap_t cmdmap, std::ostream &ss) { + if (!mdsmap->allows_dirfrags()) { + ss << "dirfrags are disallowed by the mds map!"; + return false; + } + int64_t by = 0; if (!cmd_getval(g_ceph_context, cmdmap, "bits", by)) { ss << "missing bits argument"; @@ -2467,25 +2473,27 @@ dout(20) << "try_eval(" << inum << ", " << mask << ")" << dendl; } else dout(15) << "inode " << inum << " not in mdcache!" << dendl; } else if (args[0] == "fragment_dir") { - if (args.size() == 4) { - filepath fp(args[1].c_str()); - CInode *in = mdcache->cache_traverse(fp); - if (in) { - frag_t fg; - if (fg.parse(args[2].c_str())) { - CDir *dir = in->get_dirfrag(fg); - if (dir) { - if (dir->is_auth()) { - int by = atoi(args[3].c_str()); - if (by) - mdcache->split_dir(dir, by); - else - dout(0) << "need to split by >0 bits" << dendl; - } else dout(0) << "dir " << dir->dirfrag() << " not auth" << dendl; - } else dout(0) << "dir " << in->ino() << " " << fg << " dne" << dendl; - } else dout(0) << " frag " << args[2] << " does not parse" << dendl; - } else dout(0) << "path " << fp << " not found" << dendl; - } else dout(0) << "bad syntax" << dendl; + if (!mdsmap->allows_dirfrags()) { + if (args.size() == 4) { + filepath fp(args[1].c_str()); + CInode *in = mdcache->cache_traverse(fp); + if (in) { + frag_t fg; + if (fg.parse(args[2].c_str())) { + CDir *dir = in->get_dirfrag(fg); + if (dir) { + if (dir->is_auth()) { + int by = atoi(args[3].c_str()); + if (by) + mdcache->split_dir(dir, by); + else + dout(0) << "need to split by >0 bits" << dendl; + } else dout(0) << "dir " << dir->dirfrag() << " not auth" << dendl; + } else dout(0) << "dir " << in->ino() << " " << fg << " dne" << dendl; + } else dout(0) << " frag " << args[2] << " does not parse" << dendl; + } else dout(0) << "path " << fp << " not found" << dendl; + } else dout(0) << "bad syntax" << dendl; + } else dout(0) << "dirfrags are disallowed by the mds map!" << dendl; } else if (args[0] == "merge_dir") { if (args.size() == 3) { filepath fp(args[1].c_str()); diff -Nru ceph-10.1.1/src/messages/MMDSBeacon.h ceph-10.1.2/src/messages/MMDSBeacon.h --- ceph-10.1.1/src/messages/MMDSBeacon.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/messages/MMDSBeacon.h 2016-04-12 17:46:05.000000000 +0000 @@ -37,7 +37,8 @@ MDS_HEALTH_CLIENT_LATE_RELEASE_MANY, MDS_HEALTH_CLIENT_OLDEST_TID, MDS_HEALTH_CLIENT_OLDEST_TID_MANY, - MDS_HEALTH_DAMAGE + MDS_HEALTH_DAMAGE, + MDS_HEALTH_READ_ONLY }; /** diff -Nru ceph-10.1.1/src/mon/MDSMonitor.cc ceph-10.1.2/src/mon/MDSMonitor.cc --- ceph-10.1.1/src/mon/MDSMonitor.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mon/MDSMonitor.cc 2016-04-12 17:46:05.000000000 +0000 @@ -70,6 +70,11 @@ return cmd_getval(cct, cmdmap, k, (int64_t&)val); } +static const string EXPERIMENTAL_WARNING("Warning! This feature is experimental." +"It may cause problems up to and including data loss." +"Consult the documentation at ceph.com, and if unsure, do not proceed." +"Add --yes-i-really-mean-it if you are certain."); + static const string MDS_METADATA_PREFIX("mds_metadata"); @@ -1508,6 +1513,9 @@ string flag_val; cmd_getval(g_ceph_context, cmdmap, "val", flag_val); + string confirm; + cmd_getval(g_ceph_context, cmdmap, "confirm", confirm); + if (flag_name == "enable_multiple") { bool flag_bool = false; int r = parse_bool(flag_val, &flag_bool, ss); @@ -1521,7 +1529,9 @@ ss << "Multiple-filesystems are forbidden until all mons are updated"; return -EINVAL; } - + if (confirm != "--yes-i-really-mean-it") { + ss << EXPERIMENTAL_WARNING; + } fsmap.set_enable_multiple(flag_bool); return 0; } else { @@ -1740,6 +1750,17 @@ // Persist the new FSMap pending_fsmap.filesystems[new_fs->fscid] = new_fs; return 0; + } else if (prefix == "fs set_default") { + string fs_name; + cmd_getval(g_ceph_context, cmdmap, "fs_name", fs_name); + auto fs = pending_fsmap.get_filesystem(fs_name); + if (fs == nullptr) { + ss << "filesystem '" << fs_name << "' does not exist"; + return -ENOENT; + } + + pending_fsmap.legacy_client_fscid = fs->fscid; + return 0; } else { return -ENOSYS; } @@ -1812,6 +1833,11 @@ if (interr.length()) { return -EINVAL; } + if (!fs->mds_map.allows_multimds() && n > fs->mds_map.get_max_mds() && + n > 1) { + ss << "multi-MDS clusters are not enabled; set 'allow_multimds' to enable"; + return -EINVAL; + } if (n > MAX_MDS) { ss << "may not have more than " << MAX_MDS << " MDS ranks"; return -EINVAL; @@ -1833,7 +1859,7 @@ string confirm; if (!cmd_getval(g_ceph_context, cmdmap, "confirm", confirm) || confirm != "--yes-i-really-mean-it") { - ss << "inline data is new and experimental; you must specify --yes-i-really-mean-it"; + ss << EXPERIMENTAL_WARNING; return -EPERM; } ss << "inline data enabled"; @@ -1892,7 +1918,7 @@ string confirm; if (!cmd_getval(g_ceph_context, cmdmap, "confirm", confirm) || confirm != "--yes-i-really-mean-it") { - ss << "Snapshots are unstable and will probably break your FS! Set to --yes-i-really-mean-it if you are sure you want to enable them"; + ss << EXPERIMENTAL_WARNING; return -EPERM; } fsmap.modify_filesystem( @@ -1903,6 +1929,64 @@ }); ss << "enabled new snapshots"; } + } else if (var == "allow_multimds") { + bool enable_multimds = false; + int r = parse_bool(val, &enable_multimds, ss); + if (r != 0) { + return r; + } + + if (!enable_multimds) { + fsmap.modify_filesystem(fs->fscid, + [](std::shared_ptr fs) + { + fs->mds_map.clear_multimds_allowed(); + }); + ss << "disallowed increasing the cluster size past 1"; + } else { + string confirm; + if (!cmd_getval(g_ceph_context, cmdmap, "confirm", confirm) || + confirm != "--yes-i-really-mean-it") { + ss << EXPERIMENTAL_WARNING; + return -EPERM; + } + fsmap.modify_filesystem( + fs->fscid, + [](std::shared_ptr fs) + { + fs->mds_map.set_multimds_allowed(); + }); + ss << "enabled creation of more than 1 active MDS"; + } + } else if (var == "allow_dirfrags") { + bool enable_dirfrags = false; + int r = parse_bool(val, &enable_dirfrags, ss); + if (r != 0) { + return r; + } + + if (!enable_dirfrags) { + fsmap.modify_filesystem(fs->fscid, + [](std::shared_ptr fs) + { + fs->mds_map.clear_dirfrags_allowed(); + }); + ss << "disallowed new directory fragmentation"; + } else { + string confirm; + if (!cmd_getval(g_ceph_context, cmdmap, "confirm", confirm) || + confirm != "--yes-i-really-mean-it") { + ss << EXPERIMENTAL_WARNING; + return -EPERM; + } + fsmap.modify_filesystem( + fs->fscid, + [](std::shared_ptr fs) + { + fs->mds_map.set_dirfrags_allowed(); + }); + ss << "enabled directory fragmentation"; + } } else if (var == "cluster_down") { bool is_down = false; int r = parse_bool(val, &is_down, ss); @@ -2329,6 +2413,17 @@ if (!cmd_getval(g_ceph_context, cmdmap, "maxmds", maxmds) || maxmds < 0) { return -EINVAL; } + + const MDSMap& mdsmap = + pending_fsmap.filesystems.at(pending_fsmap.legacy_client_fscid)->mds_map; + + if (!mdsmap.allows_multimds() && + maxmds > mdsmap.get_max_mds() && + maxmds > 1) { + ss << "multi-MDS clusters are not enabled; set 'allow_multimds' to enable"; + return -EINVAL; + } + if (maxmds > MAX_MDS) { ss << "may not have more than " << MAX_MDS << " MDS ranks"; return -EINVAL; diff -Nru ceph-10.1.1/src/mon/MonCommands.h ceph-10.1.2/src/mon/MonCommands.h --- ceph-10.1.1/src/mon/MonCommands.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mon/MonCommands.h 2016-04-12 17:46:05.000000000 +0000 @@ -329,7 +329,8 @@ "name=maxmds,type=CephInt,range=0", \ "set max MDS index", "mds", "rw", "cli,rest") COMMAND("mds set " \ - "name=var,type=CephChoices,strings=max_mds|max_file_size|allow_new_snaps|inline_data " \ + "name=var,type=CephChoices,strings=max_mds|max_file_size" + "|allow_new_snaps|inline_data|allow_multimds|allow_dirfrags " \ "name=val,type=CephString " \ "name=confirm,type=CephString,req=false", \ "set mds parameter to ", "mds", "rw", "cli,rest") @@ -397,12 +398,13 @@ COMMAND("fs set " \ "name=fs_name,type=CephString " \ "name=var,type=CephChoices,strings=max_mds|max_file_size" - "|allow_new_snaps|inline_data|cluster_down " \ + "|allow_new_snaps|inline_data|cluster_down|allow_multimds|allow_dirfrags " \ "name=val,type=CephString " \ "name=confirm,type=CephString,req=false", \ "set mds parameter to ", "mds", "rw", "cli,rest") COMMAND("fs flag set name=flag_name,type=CephChoices,strings=enable_multiple " - "name=val,type=CephString", \ + "name=val,type=CephString " \ + "name=confirm,type=CephChoices,strings=--yes-i-really-mean-it,req=false", \ "Set a global CephFS flag", \ "fs", "rw", "cli,rest") COMMAND("fs add_data_pool name=fs_name,type=CephString " \ @@ -411,6 +413,9 @@ COMMAND("fs rm_data_pool name=fs_name,type=CephString " \ "name=pool,type=CephString", \ "remove data pool ", "mds", "rw", "cli,rest") +COMMAND("fs set_default name=fs_name,type=CephString", \ + "set the default to the named filesystem", \ + "fs", "rw", "cli,rest") /* * Monmap commands diff -Nru ceph-10.1.1/src/mon/OSDMonitor.cc ceph-10.1.2/src/mon/OSDMonitor.cc --- ceph-10.1.1/src/mon/OSDMonitor.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/mon/OSDMonitor.cc 2016-04-12 17:46:06.000000000 +0000 @@ -2564,7 +2564,8 @@ } while (first <= osdmap.get_epoch()) { - epoch_t last = MIN(first + g_conf->osd_map_message_max, osdmap.get_epoch()); + epoch_t last = MIN(first + g_conf->osd_map_message_max - 1, + osdmap.get_epoch()); MOSDMap *m = build_incremental(first, last); if (req) { @@ -2955,6 +2956,16 @@ } } + // Not using 'sortbitwise' and should be? + if (g_conf->mon_warn_on_no_sortbitwise && + !osdmap.test_flag(CEPH_OSDMAP_SORTBITWISE) && + (osdmap.get_features(CEPH_ENTITY_TYPE_OSD, NULL) & + CEPH_FEATURE_OSD_BITWISE_HOBJ_SORT)) { + ostringstream ss; + ss << "no legacy OSD present but 'sortbitwise' flag is not set"; + summary.push_back(make_pair(HEALTH_WARN, ss.str())); + } + // Warn if 'mon_osd_down_out_interval' is set to zero. // Having this option set to zero on the leader acts much like the // 'noout' flag. It's hard to figure out what's going wrong with clusters diff -Nru ceph-10.1.1/src/os/bluestore/BlueFS.cc ceph-10.1.2/src/os/bluestore/BlueFS.cc --- ceph-10.1.1/src/os/bluestore/BlueFS.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/bluestore/BlueFS.cc 2016-04-12 17:46:06.000000000 +0000 @@ -5,6 +5,7 @@ #include "common/debug.h" #include "common/errno.h" +#include "common/perf_counters.h" #include "BlockDevice.h" #include "Allocator.h" #include "StupidAllocator.h" @@ -15,23 +16,76 @@ #define dout_prefix *_dout << "bluefs " BlueFS::BlueFS() - : ino_last(0), + : logger(NULL), + ino_last(0), log_seq(0), - log_writer(NULL) + log_writer(NULL), + bdev(MAX_BDEV), + ioc(MAX_BDEV), + block_all(MAX_BDEV), + block_total(MAX_BDEV, 0) { } BlueFS::~BlueFS() { for (auto p : bdev) { - p->close(); - delete p; + if (p) { + p->close(); + delete p; + } } for (auto p : ioc) { delete p; } } +void BlueFS::_init_logger() +{ + PerfCountersBuilder b(g_ceph_context, "BlueFS", + l_bluefs_first, l_bluefs_last); + b.add_u64_counter(l_bluefs_gift_bytes, "gift_bytes", "Bytes gifted from BlueStore"); + b.add_u64_counter(l_bluefs_reclaim_bytes, "reclaim_bytes", "Bytes reclaimed by BlueStore"); + b.add_u64(l_bluefs_db_total_bytes, "db_total_bytes", "Total bytes (main db device)"); + b.add_u64(l_bluefs_db_free_bytes, "db_free_bytes", "Free bytes (main db device)"); + b.add_u64(l_bluefs_wal_total_bytes, "wal_total_bytes", "Total bytes (wal device)"); + b.add_u64(l_bluefs_wal_free_bytes, "wal_free_bytes", "Free bytes (wal device)"); + b.add_u64(l_bluefs_slow_total_bytes, "slow_total_bytes", "Total bytes (slow device)"); + b.add_u64(l_bluefs_slow_free_bytes, "slow_free_bytes", "Free bytes (slow device)"); + b.add_u64(l_bluefs_num_files, "num_files", "File count"); + b.add_u64(l_bluefs_log_bytes, "log_bytes", "Size of the metadata log"); + b.add_u64_counter(l_bluefs_log_compactions, "log_compactions", "Compactions of the metadata log"); + b.add_u64_counter(l_bluefs_logged_bytes, "logged_bytes", "Bytes written to the metadata log"); + logger = b.create_perf_counters(); + g_ceph_context->get_perfcounters_collection()->add(logger); +} + +void BlueFS::_shutdown_logger() +{ + g_ceph_context->get_perfcounters_collection()->remove(logger); + delete logger; +} + +void BlueFS::_update_logger_stats() +{ + // we must be holding the lock + logger->set(l_bluefs_num_files, file_map.size()); + logger->set(l_bluefs_log_bytes, log_writer->file->fnode.size); + + if (alloc[BDEV_WAL]) { + logger->set(l_bluefs_wal_total_bytes, block_total[BDEV_WAL]); + logger->set(l_bluefs_wal_free_bytes, alloc[BDEV_WAL]->get_free()); + } + if (alloc[BDEV_DB]) { + logger->set(l_bluefs_db_total_bytes, block_total[BDEV_DB]); + logger->set(l_bluefs_db_free_bytes, alloc[BDEV_DB]->get_free()); + } + if (alloc[BDEV_SLOW]) { + logger->set(l_bluefs_slow_total_bytes, block_total[BDEV_SLOW]); + logger->set(l_bluefs_slow_free_bytes, alloc[BDEV_SLOW]->get_free()); + } +} + /*static void aio_cb(void *priv, void *priv2) { BlueFS *fs = static_cast(priv); @@ -42,7 +96,8 @@ int BlueFS::add_block_device(unsigned id, string path) { dout(10) << __func__ << " bdev " << id << " path " << path << dendl; - assert(id == bdev.size()); + assert(id < bdev.size()); + assert(bdev[id] == NULL); BlockDevice *b = BlockDevice::create(path, NULL, NULL); //aio_cb, this); int r = b->open(path); if (r < 0) { @@ -51,15 +106,16 @@ } dout(1) << __func__ << " bdev " << id << " path " << path << " size " << pretty_si_t(b->get_size()) << "B" << dendl; - bdev.push_back(b); - ioc.push_back(new IOContext(NULL)); - block_all.resize(bdev.size()); + bdev[id] = b; + ioc[id] = new IOContext(NULL); return 0; } uint64_t BlueFS::get_block_device_size(unsigned id) { - return bdev[id]->get_size(); + if (bdev[id]) + return bdev[id]->get_size(); + return 0; } void BlueFS::add_block_extent(unsigned id, uint64_t offset, uint64_t length) @@ -68,8 +124,10 @@ dout(1) << __func__ << " bdev " << id << " " << offset << "~" << length << dendl; assert(id < bdev.size()); + assert(bdev[id]); assert(bdev[id]->get_size() >= offset + length); block_all[id].insert(offset, length); + block_total[id] += length; if (alloc.size()) { log_t.op_alloc_add(id, offset, length); @@ -77,6 +135,9 @@ assert(r == 0); alloc[id]->init_add_free(offset, length); } + + if (logger) + logger->inc(l_bluefs_gift_bytes, length); dout(10) << __func__ << " done" << dendl; } @@ -86,6 +147,7 @@ std::lock_guard l(lock); dout(1) << __func__ << " bdev " << id << " want " << want << dendl; assert(id < alloc.size()); + assert(alloc[id]); int r = alloc[id]->reserve(want); assert(r == 0); // caller shouldn't ask for more than they can get @@ -96,10 +158,13 @@ alloc[id]->unreserve(want - *length); block_all[id].erase(*offset, *length); + block_total[id] -= *length; log_t.op_alloc_rm(id, *offset, *length); r = _flush_log(); assert(r == 0); + if (logger) + logger->inc(l_bluefs_reclaim_bytes, *length); dout(1) << __func__ << " bdev " << id << " want " << want << " got " << *offset << "~" << *length << dendl; return 0; @@ -109,12 +174,7 @@ { std::lock_guard l(lock); assert(id < block_all.size()); - uint64_t r = 0; - interval_set& p = block_all[id]; - for (interval_set::iterator q = p.begin(); q != p.end(); ++q) { - r += q.get_len(); - } - return r; + return block_total[id]; } uint64_t BlueFS::get_free(unsigned id) @@ -129,14 +189,14 @@ std::lock_guard l(lock); usage->resize(bdev.size()); for (unsigned id = 0; id < bdev.size(); ++id) { - uint64_t total = 0; - interval_set& p = block_all[id]; - for (interval_set::iterator q = p.begin(); q != p.end(); ++q) { - total += q.get_len(); + if (!bdev[id]) { + (*usage)[id] = make_pair(0, 0); + continue; } (*usage)[id].first = alloc[id]->get_free(); - (*usage)[id].second = total; - uint64_t used = (total - (*usage)[id].first) * 100 / total; + (*usage)[id].second = block_total[id]; + uint64_t used = + (block_total[id] - (*usage)[id].first) * 100 / block_total[id]; dout(10) << __func__ << " bdev " << id << " free " << (*usage)[id].first << " (" << pretty_si_t((*usage)[id].first) << "B)" @@ -162,12 +222,12 @@ dout(1) << __func__ << " osd_uuid " << osd_uuid << dendl; - assert(bdev.size() >= 1); _init_alloc(); + _init_logger(); super.version = 1; - super.block_size = bdev[0]->get_block_size(); + super.block_size = bdev[BDEV_DB]->get_block_size(); super.osd_uuid = osd_uuid; super.uuid.generate_random(); dout(1) << __func__ << " uuid " << super.uuid << dendl; @@ -175,17 +235,20 @@ // init log FileRef log_file = new File; log_file->fnode.ino = 1; - log_file->fnode.prefer_bdev = bdev.size() - 1; - int r = _allocate(log_file->fnode.prefer_bdev, - g_conf->bluefs_max_log_runway, - &log_file->fnode.extents); + log_file->fnode.prefer_bdev = BDEV_WAL; + int r = _allocate( + log_file->fnode.prefer_bdev, + g_conf->bluefs_max_log_runway, + &log_file->fnode.extents); assert(r == 0); - log_writer = new FileWriter(log_file, bdev.size()); + log_writer = _create_writer(log_file); // initial txn log_t.op_init(); - for (unsigned bdev = 0; bdev < block_all.size(); ++bdev) { + for (unsigned bdev = 0; bdev < MAX_BDEV; ++bdev) { interval_set& p = block_all[bdev]; + if (p.empty()) + continue; for (interval_set::iterator q = p.begin(); q != p.end(); ++q) { dout(20) << __func__ << " op_alloc_add " << bdev << " " << q.get_start() << "~" << q.get_len() << dendl; @@ -204,7 +267,9 @@ _close_writer(log_writer); log_writer = NULL; block_all.clear(); + block_total.clear(); _stop_alloc(); + _shutdown_logger(); dout(10) << __func__ << " success" << dendl; return 0; @@ -213,8 +278,10 @@ void BlueFS::_init_alloc() { dout(20) << __func__ << dendl; - alloc.resize(bdev.size()); + alloc.resize(MAX_BDEV); for (unsigned id = 0; id < bdev.size(); ++id) { + if (!bdev[id]) + continue; alloc[id] = new StupidAllocator; interval_set& p = block_all[id]; for (interval_set::iterator q = p.begin(); q != p.end(); ++q) { @@ -235,7 +302,6 @@ int BlueFS::mount() { dout(1) << __func__ << dendl; - assert(!bdev.empty()); int r = _open_super(); if (r < 0) { @@ -244,7 +310,9 @@ } block_all.clear(); - block_all.resize(bdev.size()); + block_all.resize(MAX_BDEV); + block_total.clear(); + block_total.resize(MAX_BDEV, 0); _init_alloc(); r = _replay(); @@ -263,10 +331,12 @@ } // set up the log for future writes - log_writer = new FileWriter(_get_file(1), bdev.size()); + log_writer = _create_writer(_get_file(1)); assert(log_writer->file->fnode.ino == 1); log_writer->pos = log_writer->file->fnode.size; dout(10) << __func__ << " log write pos set to " << log_writer->pos << dendl; + + _init_logger(); return 0; out: @@ -283,12 +353,12 @@ _close_writer(log_writer); log_writer = NULL; - block_all.clear(); _stop_alloc(); file_map.clear(); dir_map.clear(); super = bluefs_super_t(); log_t.clear(); + _shutdown_logger(); } int BlueFS::fsck() @@ -311,8 +381,8 @@ bl.rebuild(); IOContext ioc(NULL); - bdev[0]->aio_write(get_super_offset(), bl, &ioc, false); - bdev[0]->aio_submit(&ioc); + bdev[BDEV_DB]->aio_write(get_super_offset(), bl, &ioc, false); + bdev[BDEV_DB]->aio_submit(&ioc); ioc.aio_wait(); dout(20) << __func__ << " v " << super.version << " crc " << crc << " offset " << get_super_offset() << dendl; @@ -328,8 +398,8 @@ int r; // always the second block - r = bdev[0]->read(get_super_offset(), get_super_length(), - &bl, ioc[0], false); + r = bdev[BDEV_DB]->read(get_super_offset(), get_super_length(), + &bl, ioc[BDEV_DB], false); if (r < 0) return r; @@ -458,6 +528,7 @@ dout(20) << __func__ << " " << pos << ": op_alloc_add " << " " << (int)id << ":" << offset << "~" << length << dendl; block_all[id].insert(offset, length); + block_total[id] += length; alloc[id]->init_add_free(offset, length); } break; @@ -472,6 +543,7 @@ dout(20) << __func__ << " " << pos << ": op_alloc_rm " << " " << (int)id << ":" << offset << "~" << length << dendl; block_all[id].erase(offset, length); + block_total[id] -= length; alloc[id]->init_rm_free(offset, length); } break; @@ -823,7 +895,7 @@ t.uuid = super.uuid; dout(20) << __func__ << " op_init" << dendl; t.op_init(); - for (unsigned bdev = 0; bdev < block_all.size(); ++bdev) { + for (unsigned bdev = 0; bdev < MAX_BDEV; ++bdev) { interval_set& p = block_all[bdev]; for (interval_set::iterator q = p.begin(); q != p.end(); ++q) { dout(20) << __func__ << " op_alloc_add " << bdev << " " << q.get_start() @@ -868,7 +940,7 @@ _close_writer(log_writer); log_file->fnode.size = bl.length(); - log_writer = new FileWriter(log_file, bdev.size()); + log_writer = _create_writer(log_file); log_writer->append(bl); int r = _flush(log_writer, true); assert(r == 0); @@ -884,6 +956,8 @@ for (auto& r : old_extents) { alloc[r.bdev]->release(r.offset, r.length); } + + logger->inc(l_bluefs_log_compactions); } void BlueFS::_pad_bl(bufferlist& bl) @@ -922,6 +996,8 @@ _pad_bl(bl); log_writer->append(bl); + logger->inc(l_bluefs_logged_bytes, bl.length()); + log_t.clear(); log_t.seq = 0; // just so debug output is less confusing @@ -941,6 +1017,8 @@ dirty_files.erase(p++); } + _update_logger_stats(); + return 0; } @@ -1004,7 +1082,9 @@ length += partial; dout(20) << __func__ << " waiting for previous aio to complete" << dendl; for (auto p : h->iocv) { - p->aio_wait(); + if (p) { + p->aio_wait(); + } } } if (length == partial + h->buffer.length()) { @@ -1044,8 +1124,9 @@ ++p; x_off = 0; } - for (unsigned i = 0; i < bdev.size(); ++i) { - if (h->iocv[i]->has_aios()) { + for (unsigned i = 0; i < MAX_BDEV; ++i) { + if (bdev[i] && h->iocv[i]->has_aios()) { + assert(h->iocv[i]); bdev[i]->aio_submit(h->iocv[i]); } } @@ -1058,7 +1139,9 @@ dout(10) << __func__ << " " << h << dendl; utime_t start = ceph_clock_now(NULL); for (auto p : h->iocv) { - p->aio_wait(); + if (p) { + p->aio_wait(); + } } utime_t end = ceph_clock_now(NULL); utime_t dur = end - start; @@ -1135,7 +1218,8 @@ { dout(20) << __func__ << dendl; for (auto p : bdev) { - p->flush(); + if (p) + p->flush(); } } @@ -1145,16 +1229,24 @@ assert(id < alloc.size()); uint64_t left = ROUND_UP_TO(len, g_conf->bluefs_alloc_size); - int r = alloc[id]->reserve(left); + int r = -ENOSPC; + if (alloc[id]) { + r = alloc[id]->reserve(left); + } if (r < 0) { - if (id) { - derr << __func__ << " failed to allocate " << left << " on bdev " << id - << ", free " << alloc[id]->get_free() - << "; fallback to bdev 0" << dendl; - return _allocate(0, len, ev); + if (id != BDEV_SLOW) { + if (bdev[id]) + derr << __func__ << " failed to allocate " << left << " on bdev " << id + << ", free " << alloc[id]->get_free() + << "; fallback to bdev " << id + 1 << dendl; + return _allocate(id + 1, len, ev); } - derr << __func__ << " failed to allocate " << left << " on bdev " << id - << ", free " << alloc[id]->get_free() << dendl; + if (bdev[id]) + derr << __func__ << " failed to allocate " << left << " on bdev " << id + << ", free " << alloc[id]->get_free() << dendl; + else + derr << __func__ << " failed to allocate " << left << " on bdev " << id + << ", dne" << dendl; return r; } @@ -1208,11 +1300,15 @@ dout(10) << __func__ << dendl; utime_t start = ceph_clock_now(NULL); for (auto p : alloc) { - p->commit_start(); + if (p) { + p->commit_start(); + } } _flush_log(); for (auto p : alloc) { - p->commit_finish(); + if (p) { + p->commit_finish(); + } } _maybe_compact_log(); utime_t end = ceph_clock_now(NULL); @@ -1276,41 +1372,53 @@ file->fnode.mtime = ceph_clock_now(NULL); } + file->fnode.prefer_bdev = BlueFS::BDEV_DB; if (dirname.length() > 5) { // the "db.slow" and "db.wal" directory names are hard-coded at // match up with bluestore. the slow device is always the second // one (when a dedicated block.db device is present and used at // bdev 0). the wal device is always last. if (strcmp(dirname.c_str() + dirname.length() - 5, ".slow") == 0) { - assert(bdev.size() > 1); - dout(20) << __func__ << " mapping " << dirname << "/" << filename - << " to bdev 1" << dendl; - file->fnode.prefer_bdev = 1; + file->fnode.prefer_bdev = BlueFS::BDEV_SLOW; } else if (strcmp(dirname.c_str() + dirname.length() - 4, ".wal") == 0) { - assert(bdev.size() > 1); - file->fnode.prefer_bdev = bdev.size() - 1; - dout(20) << __func__ << " mapping " << dirname << "/" << filename - << " to bdev " << (int)file->fnode.prefer_bdev << dendl; + file->fnode.prefer_bdev = BlueFS::BDEV_WAL; } } + dout(20) << __func__ << " mapping " << dirname << "/" << filename + << " to bdev " << (int)file->fnode.prefer_bdev << dendl; log_t.op_file_update(file->fnode); if (create) log_t.op_dir_link(dirname, filename, file->fnode.ino); - *h = new FileWriter(file, bdev.size()); + *h = _create_writer(file); dout(10) << __func__ << " h " << *h << " on " << file->fnode << dendl; return 0; } +BlueFS::FileWriter *BlueFS::_create_writer(FileRef f) +{ + FileWriter *w = new FileWriter(f); + for (unsigned i = 0; i < MAX_BDEV; ++i) { + if (bdev[i]) { + w->iocv[i] = new IOContext(NULL); + } else { + w->iocv[i] = NULL; + } + } + return w; +} + void BlueFS::_close_writer(FileWriter *h) { dout(10) << __func__ << " " << h << dendl; - for (unsigned i=0; iiocv[i]->aio_wait(); - bdev[i]->queue_reap_ioc(h->iocv[i]); + for (unsigned i=0; iiocv[i]); + h->iocv[i]->aio_wait(); + bdev[i]->queue_reap_ioc(h->iocv[i]); + } } - h->iocv.clear(); delete h; } diff -Nru ceph-10.1.1/src/os/bluestore/BlueFS.h ceph-10.1.2/src/os/bluestore/BlueFS.h --- ceph-10.1.1/src/os/bluestore/BlueFS.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/bluestore/BlueFS.h 2016-04-12 17:46:06.000000000 +0000 @@ -13,10 +13,34 @@ #include "boost/intrusive/list.hpp" #include +class PerfCounters; + class Allocator; +enum { + l_bluefs_first = 732600, + l_bluefs_gift_bytes, + l_bluefs_reclaim_bytes, + l_bluefs_db_total_bytes, + l_bluefs_db_free_bytes, + l_bluefs_wal_total_bytes, + l_bluefs_wal_free_bytes, + l_bluefs_slow_total_bytes, + l_bluefs_slow_free_bytes, + l_bluefs_num_files, + l_bluefs_log_bytes, + l_bluefs_log_compactions, + l_bluefs_logged_bytes, + l_bluefs_last, +}; + class BlueFS { public: + static constexpr unsigned MAX_BDEV = 3; + static constexpr unsigned BDEV_WAL = 0; + static constexpr unsigned BDEV_DB = 1; + static constexpr unsigned BDEV_SLOW = 2; + struct File : public RefCountedObject { bluefs_fnode_t fnode; int refs; @@ -80,22 +104,17 @@ bufferlist tail_block; ///< existing partial block at end of file, if any std::mutex lock; - vector iocv; ///< one for each bdev + std::array iocv; ///< for each bdev - FileWriter(FileRef f, unsigned num_bdev) + FileWriter(FileRef f) : file(f), pos(0) { ++file->num_writers; - iocv.resize(num_bdev); - for (unsigned i = 0; i < num_bdev; ++i) { - iocv[i] = new IOContext(NULL); - } } + // NOTE: caller must call BlueFS::close_writer() ~FileWriter() { --file->num_writers; - assert(iocv.empty()); // caller must call BlueFS::close_writer() } - void append(const char *buf, size_t len) { buffer.append(buf, len); } @@ -161,6 +180,8 @@ private: std::mutex lock; + PerfCounters *logger; + // cache map dir_map; ///< dirname -> Dir ceph::unordered_map file_map; ///< ino -> File @@ -173,25 +194,22 @@ bluefs_transaction_t log_t; ///< pending, unwritten log transaction /* - * - there can be from 1 to 3 block devices. + * There are up to 3 block devices: * - * - the first device always has the superblock. - * - * - if there is a dedicated db device, it is the first device, and the - * second device is shared with bluestore. the first device will be - * db/, and the second device will be db.slow/. - * - * - if there is no dedicated db device, then the first device is shared, and - * maps to the db/ directory. - * - * - a wal device, if present, it always the last device. it should be - * used for any files in the db.wal/ directory. + * BDEV_DB db/ - the primary db device + * BDEV_WAL db.wal/ - a small, fast device, specifically for the WAL + * BDEV_SLOW db.slow/ - a big, slow device, to spill over to as BDEV_DB fills */ vector bdev; ///< block devices we can use vector ioc; ///< IOContexts for bdevs vector > block_all; ///< extents in bdev we own + vector block_total; ///< sum of block_all vector alloc; ///< allocators for bdevs + void _init_logger(); + void _shutdown_logger(); + void _update_logger_stats(); + void _init_alloc(); void _stop_alloc(); @@ -237,6 +255,7 @@ int _write_super(); int _replay(); ///< replay journal + FileWriter *_create_writer(FileRef f); void _close_writer(FileWriter *h); // always put the super in the second 4k block. FIXME should this be diff -Nru ceph-10.1.1/src/os/bluestore/BlueStore.cc ceph-10.1.2/src/os/bluestore/BlueStore.cc --- ceph-10.1.1/src/os/bluestore/BlueStore.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/bluestore/BlueStore.cc 2016-04-12 17:46:06.000000000 +0000 @@ -1145,7 +1145,8 @@ } else if (s == "0") { do_bluefs = false; } else { - derr << __func__ << " bluefs = " << s << " : not 0 or 1, aborting" << dendl; + derr << __func__ << " bluefs = " << s << " : not 0 or 1, aborting" + << dendl; return -EIO; } } @@ -1162,36 +1163,42 @@ char bfn[PATH_MAX]; struct stat st; - int id = 0; snprintf(bfn, sizeof(bfn), "%s/block.db", path.c_str()); if (::stat(bfn, &st) == 0) { - r = bluefs->add_block_device(id, bfn); + r = bluefs->add_block_device(BlueFS::BDEV_DB, bfn); if (r < 0) { derr << __func__ << " add block device(" << bfn << ") returned: " << cpp_strerror(r) << dendl; goto free_bluefs; } - r = _check_or_set_bdev_label(bfn, bluefs->get_block_device_size(id), + r = _check_or_set_bdev_label( + bfn, + bluefs->get_block_device_size(BlueFS::BDEV_DB), "bluefs db", create); if (r < 0) { - derr << __func__ << " check block device(" << bfn << ") label returned: " + derr << __func__ + << " check block device(" << bfn << ") label returned: " << cpp_strerror(r) << dendl; goto free_bluefs; } if (create) { bluefs->add_block_extent( - id, BLUEFS_START, - bluefs->get_block_device_size(id) - BLUEFS_START); + BlueFS::BDEV_DB, + BLUEFS_START, + bluefs->get_block_device_size(BlueFS::BDEV_DB) - BLUEFS_START); } - ++id; + bluefs_shared_bdev = BlueFS::BDEV_SLOW; + } else { + bluefs_shared_bdev = BlueFS::BDEV_DB; } + // shared device snprintf(bfn, sizeof(bfn), "%s/block", path.c_str()); - r = bluefs->add_block_device(id, bfn); + r = bluefs->add_block_device(bluefs_shared_bdev, bfn); if (r < 0) { derr << __func__ << " add block device(" << bfn << ") returned: " - << cpp_strerror(r) << dendl; + << cpp_strerror(r) << dendl; goto free_bluefs; } if (create) { @@ -1204,21 +1211,23 @@ // align to bluefs's alloc_size initial = ROUND_UP_TO(initial, g_conf->bluefs_alloc_size); initial += g_conf->bluefs_alloc_size - BLUEFS_START; - bluefs->add_block_extent(id, BLUEFS_START, initial); + bluefs->add_block_extent(bluefs_shared_bdev, BLUEFS_START, initial); bluefs_extents.insert(BLUEFS_START, initial); } - bluefs_shared_bdev = id; - ++id; - if (id == 2) { + + // use a short, relative path, if it's bluefs. + strcpy(fn, "db"); + + if (bluefs_shared_bdev == BlueFS::BDEV_SLOW) { // we have both block.db and block; tell rocksdb! // note: the second (last) size value doesn't really matter char db_paths[PATH_MAX*3]; snprintf( - db_paths, sizeof(db_paths), "%s/db,%lld %s/db.slow,%lld", - path.c_str(), - (unsigned long long)bluefs->get_block_device_size(0) * 95 / 100, - path.c_str(), - (unsigned long long)bluefs->get_block_device_size(1) * 95 / 100); + db_paths, sizeof(db_paths), "db,%lld db.slow,%lld", + (unsigned long long)bluefs->get_block_device_size(BlueFS::BDEV_DB) * + 95 / 100, + (unsigned long long)bluefs->get_block_device_size(BlueFS::BDEV_SLOW) * + 95 / 100); g_conf->set_val("rocksdb_db_paths", db_paths, false, false); dout(10) << __func__ << " set rocksdb_db_paths to " << g_conf->rocksdb_db_paths << dendl; @@ -1226,23 +1235,26 @@ snprintf(bfn, sizeof(bfn), "%s/block.wal", path.c_str()); if (::stat(bfn, &st) == 0) { - r = bluefs->add_block_device(id, bfn); + r = bluefs->add_block_device(BlueFS::BDEV_WAL, bfn); if (r < 0) { derr << __func__ << " add block device(" << bfn << ") returned: " << cpp_strerror(r) << dendl; goto free_bluefs; } - r = _check_or_set_bdev_label(bfn, bluefs->get_block_device_size(id), + r = _check_or_set_bdev_label( + bfn, + bluefs->get_block_device_size(BlueFS::BDEV_WAL), "bluefs wal", create); if (r < 0) { - derr << __func__ << " check block device(" << bfn << ") label returned: " + derr << __func__ << " check block device(" << bfn << ") label returned: " << cpp_strerror(r) << dendl; goto free_bluefs; } if (create) { bluefs->add_block_extent( - id, BDEV_LABEL_BLOCK_SIZE, - bluefs->get_block_device_size(id) - BDEV_LABEL_BLOCK_SIZE); + BlueFS::BDEV_WAL, BDEV_LABEL_BLOCK_SIZE, + bluefs->get_block_device_size(BlueFS::BDEV_WAL) - + BDEV_LABEL_BLOCK_SIZE); } g_conf->set_val("rocksdb_separate_wal_dir", "true"); } else { @@ -1320,7 +1332,8 @@ delete bluefs; bluefs = NULL; } - // delete env manually here since we can't depend on db to do this under this case + // delete env manually here since we can't depend on db to do this + // under this case delete env; env = NULL; return -EIO; diff -Nru ceph-10.1.1/src/os/bluestore/BlueStore.h ceph-10.1.2/src/os/bluestore/BlueStore.h --- ceph-10.1.1/src/os/bluestore/BlueStore.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/bluestore/BlueStore.h 2016-04-12 17:46:06.000000000 +0000 @@ -667,8 +667,8 @@ int fsck() override; - unsigned get_max_object_name_length() override { - return 4096; + int validate_hobject_key(const hobject_t &obj) const override { + return 0; } unsigned get_max_attr_name_length() override { return 256; // arbitrary; there is no real limit internally diff -Nru ceph-10.1.1/src/os/filestore/chain_xattr.cc ceph-10.1.2/src/os/filestore/chain_xattr.cc --- ceph-10.1.1/src/os/filestore/chain_xattr.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/filestore/chain_xattr.cc 2016-04-12 17:46:06.000000000 +0000 @@ -37,7 +37,7 @@ * where marks the num of xattr in the chain. */ -static void get_raw_xattr_name(const char *name, int i, char *raw_name, int raw_len) +void get_raw_xattr_name(const char *name, int i, char *raw_name, int raw_len) { int pos = 0; @@ -135,7 +135,7 @@ return getxattr_len(fn, name); do { - chunk_size = (size < CHAIN_XATTR_MAX_BLOCK_LEN ? size : CHAIN_XATTR_MAX_BLOCK_LEN); + chunk_size = size; get_raw_xattr_name(name, i, raw_name, sizeof(raw_name)); r = sys_getxattr(fn, raw_name, (char *)val + pos, chunk_size); @@ -173,6 +173,35 @@ return ret; } +int chain_getxattr_buf(const char *fn, const char *name, bufferptr *bp) +{ + size_t size = 1024; // Initial + while (1) { + bufferptr buf(size); + int r = chain_getxattr( + fn, + name, + buf.c_str(), + size); + if (r > 0) { + buf.set_length(r); + if (bp) + bp->swap(buf); + return r; + } else if (r == 0) { + return 0; + } else { + if (r == -ERANGE) { + size *= 2; + } else { + return r; + } + } + } + assert(0 == "unreachable"); + return 0; +} + static int chain_fgetxattr_len(int fd, const char *name) { int i = 0, total = 0; @@ -206,7 +235,7 @@ return chain_fgetxattr_len(fd, name); do { - chunk_size = (size < CHAIN_XATTR_MAX_BLOCK_LEN ? size : CHAIN_XATTR_MAX_BLOCK_LEN); + chunk_size = size; get_raw_xattr_name(name, i, raw_name, sizeof(raw_name)); r = sys_fgetxattr(fd, raw_name, (char *)val + pos, chunk_size); @@ -247,7 +276,7 @@ // setxattr -static int get_xattr_block_size(size_t size) +int get_xattr_block_size(size_t size) { if (size <= CHAIN_XATTR_SHORT_LEN_THRESHOLD) // this may fit in the inode; stripe over short attrs so that XFS @@ -256,79 +285,6 @@ return CHAIN_XATTR_MAX_BLOCK_LEN; } -int chain_setxattr(const char *fn, const char *name, const void *val, size_t size, bool onechunk) -{ - int i = 0, pos = 0; - char raw_name[CHAIN_XATTR_MAX_NAME_LEN * 2 + 16]; - int ret = 0; - size_t max_chunk_size = get_xattr_block_size(size); - - do { - size_t chunk_size = (size < max_chunk_size ? size : max_chunk_size); - get_raw_xattr_name(name, i, raw_name, sizeof(raw_name)); - size -= chunk_size; - - int r = sys_setxattr(fn, raw_name, (char *)val + pos, chunk_size); - if (r < 0) { - ret = r; - break; - } - pos += chunk_size; - ret = pos; - i++; - } while (size); - - if (ret >= 0 && !onechunk) { - int r; - do { - get_raw_xattr_name(name, i, raw_name, sizeof(raw_name)); - r = sys_removexattr(fn, raw_name); - if (r < 0 && r != -ENODATA) - ret = r; - i++; - } while (r != -ENODATA); - } - - return ret; -} - -int chain_fsetxattr(int fd, const char *name, const void *val, size_t size, bool onechunk) -{ - int i = 0, pos = 0; - char raw_name[CHAIN_XATTR_MAX_NAME_LEN * 2 + 16]; - int ret = 0; - size_t max_chunk_size = get_xattr_block_size(size); - - do { - size_t chunk_size = (size < max_chunk_size ? size : max_chunk_size); - get_raw_xattr_name(name, i, raw_name, sizeof(raw_name)); - size -= chunk_size; - - int r = sys_fsetxattr(fd, raw_name, (char *)val + pos, chunk_size); - if (r < 0) { - ret = r; - break; - } - pos += chunk_size; - ret = pos; - i++; - } while (size); - - if (ret >= 0 && !onechunk) { - int r; - do { - get_raw_xattr_name(name, i, raw_name, sizeof(raw_name)); - r = sys_fremovexattr(fd, raw_name); - if (r < 0 && r != -ENODATA) - ret = r; - i++; - } while (r != -ENODATA); - } - - return ret; -} - - // removexattr int chain_removexattr(const char *fn, const char *name) diff -Nru ceph-10.1.1/src/os/filestore/chain_xattr.h ceph-10.1.2/src/os/filestore/chain_xattr.h --- ceph-10.1.1/src/os/filestore/chain_xattr.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/filestore/chain_xattr.h 2016-04-12 17:46:06.000000000 +0000 @@ -5,6 +5,10 @@ #define __CEPH_OSD_CHAIN_XATTR_H #include "common/xattr.h" +#include "include/assert.h" +#include "include/buffer.h" +#include +#include #include @@ -77,9 +81,100 @@ // wrappers to chain large values across multiple xattrs int chain_getxattr(const char *fn, const char *name, void *val, size_t size); +int chain_getxattr_buf(const char *fn, const char *name, bufferptr *bp); int chain_fgetxattr(int fd, const char *name, void *val, size_t size); -int chain_setxattr(const char *fn, const char *name, const void *val, size_t size, bool onechunk=false); -int chain_fsetxattr(int fd, const char *name, const void *val, size_t size, bool onechunk=false); + +int get_xattr_block_size(size_t size); +void get_raw_xattr_name(const char *name, int i, char *raw_name, int raw_len); + +template +int chain_setxattr( + const char *fn, const char *name, const void *val, size_t size) +{ + int i = 0, pos = 0; + char raw_name[CHAIN_XATTR_MAX_NAME_LEN * 2 + 16]; + int ret = 0; + size_t max_chunk_size = + ensure_single_attr ? size : get_xattr_block_size(size); + + static_assert( + !skip_chain_cleanup || ensure_single_attr, + "skip_chain_cleanup must imply ensure_single_attr"); + + do { + size_t chunk_size = (size < max_chunk_size ? size : max_chunk_size); + get_raw_xattr_name(name, i, raw_name, sizeof(raw_name)); + size -= chunk_size; + + int r = sys_setxattr(fn, raw_name, (char *)val + pos, chunk_size); + if (r < 0) { + ret = r; + break; + } + pos += chunk_size; + ret = pos; + i++; + assert(size == 0 || !ensure_single_attr); + } while (size); + + if (ret >= 0 && !skip_chain_cleanup) { + int r; + do { + get_raw_xattr_name(name, i, raw_name, sizeof(raw_name)); + r = sys_removexattr(fn, raw_name); + if (r < 0 && r != -ENODATA) + ret = r; + i++; + } while (r != -ENODATA); + } + + return ret; +} + +template +int chain_fsetxattr( + int fd, const char *name, const void *val, size_t size) +{ + int i = 0, pos = 0; + char raw_name[CHAIN_XATTR_MAX_NAME_LEN * 2 + 16]; + int ret = 0; + size_t max_chunk_size = + ensure_single_attr ? size : get_xattr_block_size(size); + + static_assert( + !skip_chain_cleanup || ensure_single_attr, + "skip_chain_cleanup must imply ensure_single_attr"); + + do { + size_t chunk_size = (size < max_chunk_size ? size : max_chunk_size); + get_raw_xattr_name(name, i, raw_name, sizeof(raw_name)); + size -= chunk_size; + + int r = sys_fsetxattr(fd, raw_name, (char *)val + pos, chunk_size); + if (r < 0) { + ret = r; + break; + } + pos += chunk_size; + ret = pos; + i++; + assert(size == 0 || !ensure_single_attr); + } while (size); + + if (ret >= 0 && !skip_chain_cleanup) { + int r; + do { + get_raw_xattr_name(name, i, raw_name, sizeof(raw_name)); + r = sys_fremovexattr(fd, raw_name); + if (r < 0 && r != -ENODATA) + ret = r; + i++; + } while (r != -ENODATA); + } + + return ret; +} + int chain_listxattr(const char *fn, char *names, size_t len); int chain_flistxattr(int fd, char *names, size_t len); int chain_removexattr(const char *fn, const char *name); diff -Nru ceph-10.1.1/src/os/filestore/FileStore.cc ceph-10.1.2/src/os/filestore/FileStore.cc --- ceph-10.1.1/src/os/filestore/FileStore.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/filestore/FileStore.cc 2016-04-12 17:46:06.000000000 +0000 @@ -123,6 +123,12 @@ return compat; } +int FileStore::validate_hobject_key(const hobject_t &obj) const +{ + unsigned len = LFNIndex::get_max_escaped_name_len(obj); + return len > m_filestore_max_xattr_value_size ? -ENAMETOOLONG : 0; +} + int FileStore::get_block_device_fsid(const string& path, uuid_d *fsid) { // make sure we don't try to use aio or direct_io (and get annoying @@ -300,8 +306,9 @@ << ") in index: " << cpp_strerror(-r) << dendl; goto fail; } - r = chain_fsetxattr(fd, XATTR_SPILL_OUT_NAME, - XATTR_NO_SPILL_OUT, sizeof(XATTR_NO_SPILL_OUT), true); + r = chain_fsetxattr( + fd, XATTR_SPILL_OUT_NAME, + XATTR_NO_SPILL_OUT, sizeof(XATTR_NO_SPILL_OUT)); if (r < 0) { VOID_TEMP_FAILURE_RETRY(::close(fd)); derr << "error setting spillout xattr for oid " << oid << " (" << (*path)->path() @@ -559,7 +566,8 @@ m_filestore_max_alloc_hint_size(g_conf->filestore_max_alloc_hint_size), m_fs_type(0), m_filestore_max_inline_xattr_size(0), - m_filestore_max_inline_xattrs(0) + m_filestore_max_inline_xattrs(0), + m_filestore_max_xattr_value_size(0) { m_filestore_kill_at.set(g_conf->filestore_kill_at); for (int i = 0; i < m_ondisk_finisher_num; ++i) { @@ -2146,7 +2154,8 @@ // then record that we did it bufferlist v; ::encode(spos, v); - int r = chain_fsetxattr(fd, GLOBAL_REPLAY_GUARD_XATTR, v.c_str(), v.length(), true); + int r = chain_fsetxattr( + fd, GLOBAL_REPLAY_GUARD_XATTR, v.c_str(), v.length()); if (r < 0) { derr << __func__ << ": fsetxattr " << GLOBAL_REPLAY_GUARD_XATTR << " got " << cpp_strerror(r) << dendl; @@ -2236,7 +2245,8 @@ bufferlist v(40); ::encode(spos, v); ::encode(in_progress, v); - int r = chain_fsetxattr(fd, REPLAY_GUARD_XATTR, v.c_str(), v.length(), true); + int r = chain_fsetxattr( + fd, REPLAY_GUARD_XATTR, v.c_str(), v.length()); if (r < 0) { derr << "fsetxattr " << REPLAY_GUARD_XATTR << " got " << cpp_strerror(r) << dendl; assert(0 == "fsetxattr failed"); @@ -2279,7 +2289,8 @@ ::encode(spos, v); bool in_progress = false; ::encode(in_progress, v); - int r = chain_fsetxattr(fd, REPLAY_GUARD_XATTR, v.c_str(), v.length(), true); + int r = chain_fsetxattr( + fd, REPLAY_GUARD_XATTR, v.c_str(), v.length()); if (r < 0) { derr << "fsetxattr " << REPLAY_GUARD_XATTR << " got " << cpp_strerror(r) << dendl; assert(0 == "fsetxattr failed"); @@ -3391,11 +3402,11 @@ r = chain_fgetxattr(**o, XATTR_SPILL_OUT_NAME, buf, sizeof(buf)); if (r >= 0 && !strncmp(buf, XATTR_NO_SPILL_OUT, sizeof(XATTR_NO_SPILL_OUT))) { - r = chain_fsetxattr(**n, XATTR_SPILL_OUT_NAME, XATTR_NO_SPILL_OUT, - sizeof(XATTR_NO_SPILL_OUT), true); + r = chain_fsetxattr(**n, XATTR_SPILL_OUT_NAME, XATTR_NO_SPILL_OUT, + sizeof(XATTR_NO_SPILL_OUT)); } else { - r = chain_fsetxattr(**n, XATTR_SPILL_OUT_NAME, XATTR_SPILL_OUT, - sizeof(XATTR_SPILL_OUT), true); + r = chain_fsetxattr(**n, XATTR_SPILL_OUT_NAME, XATTR_SPILL_OUT, + sizeof(XATTR_SPILL_OUT)); } if (r < 0) goto out3; @@ -5663,21 +5674,25 @@ { uint32_t fs_xattr_size; uint32_t fs_xattrs; + uint32_t fs_xattr_max_value_size; switch (m_fs_type) { #if defined(__linux__) case XFS_SUPER_MAGIC: fs_xattr_size = g_conf->filestore_max_inline_xattr_size_xfs; fs_xattrs = g_conf->filestore_max_inline_xattrs_xfs; + fs_xattr_max_value_size = g_conf->filestore_max_xattr_value_size_xfs; break; case BTRFS_SUPER_MAGIC: fs_xattr_size = g_conf->filestore_max_inline_xattr_size_btrfs; fs_xattrs = g_conf->filestore_max_inline_xattrs_btrfs; + fs_xattr_max_value_size = g_conf->filestore_max_xattr_value_size_btrfs; break; #endif default: fs_xattr_size = g_conf->filestore_max_inline_xattr_size_other; fs_xattrs = g_conf->filestore_max_inline_xattrs_other; + fs_xattr_max_value_size = g_conf->filestore_max_xattr_value_size_other; break; } @@ -5692,6 +5707,24 @@ m_filestore_max_inline_xattrs = g_conf->filestore_max_inline_xattrs; else m_filestore_max_inline_xattrs = fs_xattrs; + + // Use override value if set + if (g_conf->filestore_max_xattr_value_size) + m_filestore_max_xattr_value_size = g_conf->filestore_max_xattr_value_size; + else + m_filestore_max_xattr_value_size = fs_xattr_max_value_size; + + if (m_filestore_max_xattr_value_size < g_conf->osd_max_object_name_len) { + derr << "WARNING: max attr value size (" + << m_filestore_max_xattr_value_size + << ") is smaller than osd_max_object_name_len (" + << g_conf->osd_max_object_name_len + << "). Your backend filesystem appears to not support attrs large " + << "enough to handle the configured max rados name size. You may get " + << "unexpected ENAMETOOLONG errors on rados operations or buggy " + << "behavior" + << dendl; + } } // -- FSSuperblock -- diff -Nru ceph-10.1.1/src/os/filestore/FileStore.h ceph-10.1.2/src/os/filestore/FileStore.h --- ceph-10.1.1/src/os/filestore/FileStore.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/filestore/FileStore.h 2016-04-12 17:46:06.000000000 +0000 @@ -432,10 +432,9 @@ int write_op_seq(int, uint64_t seq); int mount(); int umount(); - unsigned get_max_object_name_length() { - // not safe for all file systems, btw! use the tunable to limit this. - return 4096; - } + + int validate_hobject_key(const hobject_t &obj) const override; + unsigned get_max_attr_name_length() { // xattr limit is 128; leave room for our prefixes (user.ceph._), // some margin, and cap at 100 @@ -739,6 +738,7 @@ void set_xattr_limits_via_conf(); uint32_t m_filestore_max_inline_xattr_size; uint32_t m_filestore_max_inline_xattrs; + uint32_t m_filestore_max_xattr_value_size; FSSuperblock superblock; diff -Nru ceph-10.1.1/src/os/filestore/IndexManager.cc ceph-10.1.2/src/os/filestore/IndexManager.cc --- ceph-10.1.1/src/os/filestore/IndexManager.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/filestore/IndexManager.cc 2016-04-12 17:46:06.000000000 +0000 @@ -36,8 +36,9 @@ static int set_version(const char *path, uint32_t version) { bufferlist bl; ::encode(version, bl); - return chain_setxattr(path, "user.cephos.collection_version", bl.c_str(), - bl.length(), true); + return chain_setxattr( + path, "user.cephos.collection_version", bl.c_str(), + bl.length()); } static int get_version(const char *path, uint32_t *version) { diff -Nru ceph-10.1.1/src/os/filestore/LFNIndex.cc ceph-10.1.2/src/os/filestore/LFNIndex.cc --- ceph-10.1.1/src/os/filestore/LFNIndex.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/filestore/LFNIndex.cc 2016-04-12 17:46:06.000000000 +0000 @@ -74,6 +74,14 @@ /* Public methods */ +uint64_t LFNIndex::get_max_escaped_name_len(const hobject_t &obj) +{ + ghobject_t ghobj(obj); + ghobj.shard_id = shard_id_t(0); + ghobj.generation = 0; + ghobj.hobj.snap = 0; + return lfn_generate_object_name_current(ghobj).size(); +} int LFNIndex::init() { @@ -375,14 +383,18 @@ ghobject_t *o) { char path[PATH_MAX]; - bufferptr bp(PATH_MAX); snprintf(path, sizeof(path), "%s/%s", dir, file); // Hack, user.ceph._ is the attribute used to store the object info - int r = chain_getxattr(path, "user.ceph._", bp.c_str(), bp.length()); + bufferptr bp; + int r = chain_getxattr_buf( + path, + "user.ceph._", + &bp); if (r < 0) return r; bufferlist bl; - bl.push_back(bp); + if (r > 0) + bl.push_back(bp); object_info_t oi(bl); *o = ghobject_t(oi.soid); return 0; @@ -421,10 +433,11 @@ ghobject_t obj; if (lfn_is_object(short_name)) { r = lfn_translate(to_list, short_name, &obj); - if (r < 0) { - r = -errno; + if (r == -EINVAL) { + continue; + } else if (r < 0) { goto cleanup; - } else if (r > 0) { + } else { string long_name = lfn_generate_object_name(obj); if (!lfn_must_hash(long_name)) { assert(long_name == short_name); @@ -434,8 +447,6 @@ out->insert(pair(short_name, obj)); ++listed; - } else { - continue; } } } @@ -521,9 +532,10 @@ { string full_path = get_full_path_subdir(path); maybe_inject_failure(); - return chain_setxattr(full_path.c_str(), mangle_attr_name(attr_name).c_str(), - reinterpret_cast(attr_value.c_str()), - attr_value.length()); + return chain_setxattr( + full_path.c_str(), mangle_attr_name(attr_name).c_str(), + reinterpret_cast(attr_value.c_str()), + attr_value.length()); } int LFNIndex::get_attr_path(const vector &path, @@ -531,26 +543,14 @@ bufferlist &attr_value) { string full_path = get_full_path_subdir(path); - size_t size = 1024; // Initial - while (1) { - bufferptr buf(size); - int r = chain_getxattr(full_path.c_str(), mangle_attr_name(attr_name).c_str(), - reinterpret_cast(buf.c_str()), - size); - if (r > 0) { - buf.set_length(r); - attr_value.push_back(buf); - break; - } else { - r = -errno; - if (r == -ERANGE) { - size *= 2; - } else { - return r; - } - } - } - return 0; + bufferptr bp; + int r = chain_getxattr_buf( + full_path.c_str(), + mangle_attr_name(attr_name).c_str(), + &bp); + if (r > 0) + attr_value.push_back(bp); + return r; } int LFNIndex::remove_attr_path(const vector &path, @@ -621,13 +621,8 @@ } } -string LFNIndex::lfn_generate_object_name(const ghobject_t &oid) +string LFNIndex::lfn_generate_object_name_current(const ghobject_t &oid) { - if (index_version == HASH_INDEX_TAG) - return lfn_generate_object_name_keyless(oid); - if (index_version == HASH_INDEX_TAG_2) - return lfn_generate_object_name_poolless(oid); - string full_name; string::const_iterator i = oid.hobj.oid.name.begin(); if (oid.hobj.oid.name.substr(0, 4) == "DIR_") { @@ -754,12 +749,14 @@ int i = 0; string candidate; string candidate_path; - char buf[FILENAME_MAX_LEN + 1]; for ( ; ; ++i) { candidate = lfn_get_short_name(oid, i); candidate_path = get_full_path(path, candidate); - r = chain_getxattr(candidate_path.c_str(), get_lfn_attr().c_str(), - buf, sizeof(buf)); + bufferptr bp; + r = chain_getxattr_buf( + candidate_path.c_str(), + get_lfn_attr().c_str(), + &bp); if (r < 0) { if (errno != ENODATA && errno != ENOENT) return -errno; @@ -780,8 +777,8 @@ return 0; } assert(r > 0); - buf[MIN((int)sizeof(buf) - 1, r)] = '\0'; - if (!strcmp(buf, full_name.c_str())) { + string lfn(bp.c_str(), bp.length()); + if (lfn == full_name) { if (mangled_name) *mangled_name = candidate; if (out_path) @@ -793,8 +790,11 @@ } return 0; } - r = chain_getxattr(candidate_path.c_str(), get_alt_lfn_attr().c_str(), - buf, sizeof(buf)); + bp = bufferptr(); + r = chain_getxattr_buf( + candidate_path.c_str(), + get_alt_lfn_attr().c_str(), + &bp); if (r > 0) { // only consider alt name if nlink > 1 struct stat st; @@ -805,7 +805,7 @@ // left over from incomplete unlink, remove maybe_inject_failure(); dout(20) << __func__ << " found extra alt attr for " << candidate_path - << ", long name " << string(buf, r) << dendl; + << ", long name " << string(bp.c_str(), bp.length()) << dendl; rc = chain_removexattr(candidate_path.c_str(), get_alt_lfn_attr().c_str()); maybe_inject_failure(); @@ -813,8 +813,8 @@ return rc; continue; } - buf[MIN((int)sizeof(buf) - 1, r)] = '\0'; - if (!strcmp(buf, full_name.c_str())) { + string lfn(bp.c_str(), bp.length()); + if (lfn == full_name) { dout(20) << __func__ << " used alt attr for " << full_name << dendl; if (mangled_name) *mangled_name = candidate; @@ -841,23 +841,29 @@ maybe_inject_failure(); // if the main attr exists and is different, move it to the alt attr. - char buf[FILENAME_MAX_LEN + 1]; - int r = chain_getxattr(full_path.c_str(), get_lfn_attr().c_str(), - buf, sizeof(buf)); - if (r >= 0 && (r != (int)full_name.length() || - memcmp(buf, full_name.c_str(), full_name.length()))) { - dout(20) << __func__ << " " << mangled_name - << " moving old name to alt attr " - << string(buf, r) - << ", new name is " << full_name << dendl; - r = chain_setxattr(full_path.c_str(), get_alt_lfn_attr().c_str(), - buf, r); - if (r < 0) - return r; + bufferptr bp; + int r = chain_getxattr_buf( + full_path.c_str(), + get_lfn_attr().c_str(), + &bp); + if (r > 0) { + string lfn(bp.c_str(), bp.length()); + if (lfn != full_name) { + dout(20) << __func__ << " " << mangled_name + << " moving old name to alt attr " + << lfn + << ", new name is " << full_name << dendl; + r = chain_setxattr( + full_path.c_str(), get_alt_lfn_attr().c_str(), + bp.c_str(), bp.length()); + if (r < 0) + return r; + } } - return chain_setxattr(full_path.c_str(), get_lfn_attr().c_str(), - full_name.c_str(), full_name.size()); + return chain_setxattr( + full_path.c_str(), get_lfn_attr().c_str(), + full_name.c_str(), full_name.size()); } int LFNIndex::lfn_unlink(const vector &path, @@ -936,31 +942,32 @@ return lfn_parse_object_name(short_name, out); } string full_path = get_full_path(path, short_name); - char attr[PATH_MAX]; // First, check alt attr - int r = chain_getxattr( + bufferptr bp; + int r = chain_getxattr_buf( full_path.c_str(), get_alt_lfn_attr().c_str(), - attr, - sizeof(attr) - 1); - if (r >= 0) { + &bp); + if (r > 0) { // There is an alt attr, does it match? - if (r < (int)sizeof(attr)) - attr[r] = '\0'; - if (short_name_matches(short_name.c_str(), attr)) { - string long_name(attr); - return lfn_parse_object_name(long_name, out); + string lfn(bp.c_str(), bp.length()); + if (short_name_matches(short_name.c_str(), lfn.c_str())) { + return lfn_parse_object_name(lfn, out); } } // Get lfn_attr - r = chain_getxattr(full_path.c_str(), get_lfn_attr().c_str(), attr, sizeof(attr) - 1); + bp = bufferptr(); + r = chain_getxattr_buf( + full_path.c_str(), + get_lfn_attr().c_str(), + &bp); if (r < 0) - return -errno; - if (r < (int)sizeof(attr)) - attr[r] = '\0'; + return r; + if (r == 0) + return -EINVAL; - string long_name(attr); + string long_name(bp.c_str(), bp.length()); return lfn_parse_object_name(long_name, out); } @@ -1032,7 +1039,7 @@ return 0; } -bool LFNIndex::lfn_parse_object_name_keyless(const string &long_name, ghobject_t *out) +int LFNIndex::lfn_parse_object_name_keyless(const string &long_name, ghobject_t *out) { bool r = parse_object(long_name.c_str(), *out); int64_t pool = -1; @@ -1042,7 +1049,7 @@ out->hobj.pool = pool; if (!r) return r; string temp = lfn_generate_object_name(*out); - return r; + return r ? 0 : -EINVAL; } static bool append_unescaped(string::const_iterator begin, @@ -1069,8 +1076,8 @@ return true; } -bool LFNIndex::lfn_parse_object_name_poolless(const string &long_name, - ghobject_t *out) +int LFNIndex::lfn_parse_object_name_poolless(const string &long_name, + ghobject_t *out) { string name; string key; @@ -1081,7 +1088,7 @@ if (*current == '\\') { ++current; if (current == long_name.end()) { - return false; + return -EINVAL; } else if (*current == 'd') { name.append("DIR_"); ++current; @@ -1096,27 +1103,27 @@ string::const_iterator end = current; for ( ; end != long_name.end() && *end != '_'; ++end) ; if (end == long_name.end()) - return false; + return -EINVAL; if (!append_unescaped(current, end, &name)) - return false; + return -EINVAL; current = ++end; for ( ; end != long_name.end() && *end != '_'; ++end) ; if (end == long_name.end()) - return false; + return -EINVAL; if (!append_unescaped(current, end, &key)) - return false; + return -EINVAL; current = ++end; for ( ; end != long_name.end() && *end != '_'; ++end) ; if (end == long_name.end()) - return false; + return -EINVAL; string snap_str(current, end); current = ++end; for ( ; end != long_name.end() && *end != '_'; ++end) ; if (end != long_name.end()) - return false; + return -EINVAL; string hash_str(current, end); if (snap_str == "head") @@ -1133,11 +1140,11 @@ if (coll().is_pg_prefix(&pg)) pool = (int64_t)pg.pgid.pool(); (*out) = ghobject_t(hobject_t(name, key, snap, hash, pool, "")); - return true; + return 0; } -bool LFNIndex::lfn_parse_object_name(const string &long_name, ghobject_t *out) +int LFNIndex::lfn_parse_object_name(const string &long_name, ghobject_t *out) { string name; string key; @@ -1157,7 +1164,7 @@ if (*current == '\\') { ++current; if (current == long_name.end()) { - return false; + return -EINVAL; } else if (*current == 'd') { name.append("DIR_"); ++current; @@ -1172,35 +1179,35 @@ string::const_iterator end = current; for ( ; end != long_name.end() && *end != '_'; ++end) ; if (end == long_name.end()) - return false; + return -EINVAL; if (!append_unescaped(current, end, &name)) - return false; + return -EINVAL; current = ++end; for ( ; end != long_name.end() && *end != '_'; ++end) ; if (end == long_name.end()) - return false; + return -EINVAL; if (!append_unescaped(current, end, &key)) - return false; + return -EINVAL; current = ++end; for ( ; end != long_name.end() && *end != '_'; ++end) ; if (end == long_name.end()) - return false; + return -EINVAL; string snap_str(current, end); current = ++end; for ( ; end != long_name.end() && *end != '_'; ++end) ; if (end == long_name.end()) - return false; + return -EINVAL; string hash_str(current, end); current = ++end; for ( ; end != long_name.end() && *end != '_'; ++end) ; if (end == long_name.end()) - return false; + return -EINVAL; if (!append_unescaped(current, end, &ns)) - return false; + return -EINVAL; current = ++end; for ( ; end != long_name.end() && *end != '_'; ++end) ; @@ -1212,7 +1219,7 @@ current = ++end; for ( ; end != long_name.end() && *end != '_'; ++end) ; if (end == long_name.end()) - return false; + return -EINVAL; genstring = string(current, end); generation = (gen_t)strtoull(genstring.c_str(), NULL, 16); @@ -1220,7 +1227,7 @@ current = ++end; for ( ; end != long_name.end() && *end != '_'; ++end) ; if (end != long_name.end()) - return false; + return -EINVAL; shardstring = string(current, end); shard_id = (shard_id_t)strtoul(shardstring.c_str(), NULL, 16); @@ -1240,7 +1247,7 @@ pool = strtoull(pstring.c_str(), NULL, 16); (*out) = ghobject_t(hobject_t(name, key, snap, hash, (int64_t)pool, ns), generation, shard_id); - return true; + return 0; } bool LFNIndex::lfn_is_hashed_filename(const string &name) diff -Nru ceph-10.1.1/src/os/filestore/LFNIndex.h ceph-10.1.2/src/os/filestore/LFNIndex.h --- ceph-10.1.1/src/os/filestore/LFNIndex.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/filestore/LFNIndex.h 2016-04-12 17:46:06.000000000 +0000 @@ -212,6 +212,11 @@ ); } + /** + * Returns the length of the longest escaped name which could result + * from any clone, shard, or rollback object of this object + */ + static uint64_t get_max_escaped_name_len(const hobject_t &obj); protected: virtual int _init() = 0; @@ -480,24 +485,36 @@ ); ///< @return Generated object name. /// Generate object name - string lfn_generate_object_name( + static string lfn_generate_object_name_current( const ghobject_t &oid ///< [in] Object for which to generate. ); ///< @return Generated object name. + /// Generate object name + string lfn_generate_object_name( + const ghobject_t &oid ///< [in] Object for which to generate. + ) { + if (index_version == HASH_INDEX_TAG) + return lfn_generate_object_name_keyless(oid); + if (index_version == HASH_INDEX_TAG_2) + return lfn_generate_object_name_poolless(oid); + else + return lfn_generate_object_name_current(oid); + } ///< @return Generated object name. + /// Parse object name - bool lfn_parse_object_name_keyless( + int lfn_parse_object_name_keyless( const string &long_name, ///< [in] Name to parse ghobject_t *out ///< [out] Resulting Object ); ///< @return True if successfull, False otherwise. /// Parse object name - bool lfn_parse_object_name_poolless( + int lfn_parse_object_name_poolless( const string &long_name, ///< [in] Name to parse ghobject_t *out ///< [out] Resulting Object ); ///< @return True if successfull, False otherwise. /// Parse object name - bool lfn_parse_object_name( + int lfn_parse_object_name( const string &long_name, ///< [in] Name to parse ghobject_t *out ///< [out] Resulting Object ); ///< @return True if successfull, False otherwise. diff -Nru ceph-10.1.1/src/os/kstore/KStore.h ceph-10.1.2/src/os/kstore/KStore.h --- ceph-10.1.1/src/os/kstore/KStore.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/kstore/KStore.h 2016-04-12 17:46:06.000000000 +0000 @@ -411,8 +411,9 @@ int fsck(); - unsigned get_max_object_name_length() { - return 4096; + + int validate_hobject_key(const hobject_t &obj) const override { + return 0; } unsigned get_max_attr_name_length() { return 256; // arbitrary; there is no real limit internally diff -Nru ceph-10.1.1/src/os/memstore/MemStore.h ceph-10.1.2/src/os/memstore/MemStore.h --- ceph-10.1.1/src/os/memstore/MemStore.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/memstore/MemStore.h 2016-04-12 17:46:06.000000000 +0000 @@ -365,8 +365,8 @@ int mount(); int umount(); - unsigned get_max_object_name_length() { - return 4096; + int validate_hobject_key(const hobject_t &obj) const override { + return 0; } unsigned get_max_attr_name_length() { return 256; // arbitrary; there is no real limit internally diff -Nru ceph-10.1.1/src/os/ObjectStore.cc ceph-10.1.2/src/os/ObjectStore.cc --- ceph-10.1.1/src/os/ObjectStore.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/ObjectStore.cc 2016-04-12 17:46:06.000000000 +0000 @@ -86,6 +86,7 @@ } int ObjectStore::probe_block_device_fsid( + CephContext *cct, const string& path, uuid_d *fsid) { @@ -95,14 +96,20 @@ // first try bluestore -- it has a crc on its header and will fail // reliably. r = BlueStore::get_block_device_fsid(path, fsid); - if (r == 0) + if (r == 0) { + lgeneric_dout(cct, 0) << __func__ << " " << path << " is bluestore, " + << *fsid << dendl; return r; + } #endif // okay, try FileStore (journal). r = FileStore::get_block_device_fsid(path, fsid); - if (r == 0) + if (r == 0) { + lgeneric_dout(cct, 0) << __func__ << " " << path << " is filestore, " + << *fsid << dendl; return r; + } return -EINVAL; } diff -Nru ceph-10.1.1/src/os/ObjectStore.h ceph-10.1.2/src/os/ObjectStore.h --- ceph-10.1.1/src/os/ObjectStore.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/os/ObjectStore.h 2016-04-12 17:46:06.000000000 +0000 @@ -119,8 +119,10 @@ * @param path path to device * @param fsid [out] osd uuid */ - static int probe_block_device_fsid(const string& path, - uuid_d *fsid); + static int probe_block_device_fsid( + CephContext *cct, + const string& path, + uuid_d *fsid); Logger *logger; @@ -1926,7 +1928,15 @@ virtual int fsck() { return -EOPNOTSUPP; } - virtual unsigned get_max_object_name_length() = 0; + + /** + * Returns 0 if the hobject is valid, -error otherwise + * + * Errors: + * -ENAMETOOLONG: locator/namespace/name too large + */ + virtual int validate_hobject_key(const hobject_t &obj) const = 0; + virtual unsigned get_max_attr_name_length() = 0; virtual int mkfs() = 0; // wipe virtual int mkjournal() = 0; // journal only diff -Nru ceph-10.1.1/src/osd/OSD.cc ceph-10.1.2/src/osd/OSD.cc --- ceph-10.1.1/src/osd/OSD.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/osd/OSD.cc 2016-04-12 17:46:06.000000000 +0000 @@ -4615,11 +4615,7 @@ // full map requests may happen while active or pre-boot if (requested_full_first) { - epoch_t first = requested_full_first; - epoch_t last = requested_full_last; - requested_full_first = 0; - requested_full_last = 0; - request_full_map(first, last); + rerequest_full_maps(); } } } @@ -4918,17 +4914,6 @@ monc->send_mon_message(req); } -void OSD::finish_full_map_request() -{ - if (requested_full_first == 0 && requested_full_last == 0) - return; - //Had requested some map but didn't receive in this message, - //This might because monitor capping the message to osd_map_message_max - dout(10) << __func__ << "still missing " << requested_full_first - << ".." << requested_full_last << ", but now give up." << dendl; - requested_full_first = requested_full_last = 0; -} - void OSD::got_full_map(epoch_t e) { assert(requested_full_first <= requested_full_last); @@ -6693,8 +6678,11 @@ // even if this map isn't from a mon, we may have satisfied our subscription monc->sub_got("osdmap", last); - if (!m->maps.empty()) - finish_full_map_request(); + if (!m->maps.empty() && requested_full_first) { + dout(10) << __func__ << " still missing full maps " << requested_full_first + << ".." << requested_full_last << dendl; + rerequest_full_maps(); + } if (last <= superblock.newest_map) { dout(10) << " no new maps here, dropping" << dendl; diff -Nru ceph-10.1.1/src/osd/OSD.h ceph-10.1.2/src/osd/OSD.h --- ceph-10.1.1/src/osd/OSD.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/osd/OSD.h 2016-04-12 17:46:06.000000000 +0000 @@ -2050,7 +2050,13 @@ epoch_t requested_full_first, requested_full_last; void request_full_map(epoch_t first, epoch_t last); - void finish_full_map_request(); + void rerequest_full_maps() { + epoch_t first = requested_full_first; + epoch_t last = requested_full_last; + requested_full_first = 0; + requested_full_last = 0; + request_full_map(first, last); + } void got_full_map(epoch_t e); // -- failures -- diff -Nru ceph-10.1.1/src/osd/PG.cc ceph-10.1.2/src/osd/PG.cc --- ceph-10.1.1/src/osd/PG.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/osd/PG.cc 2016-04-12 17:46:06.000000000 +0000 @@ -2298,6 +2298,7 @@ // Info child->info.history = info.history; + child->info.history.epoch_created = get_osdmap()->get_epoch(); child->info.purged_snaps = info.purged_snaps; if (info.last_backfill.is_max()) { @@ -2313,6 +2314,7 @@ } child->info.stats = info.stats; + child->info.stats.parent_split_bits = split_bits; info.stats.stats_invalid = true; child->info.stats.stats_invalid = true; child->info.last_epoch_started = info.last_epoch_started; @@ -3357,8 +3359,16 @@ //NOSCRUB so skip regular scrubs if ((osd->osd->get_osdmap()->test_flag(CEPH_OSDMAP_NOSCRUB) || - pool.info.has_flag(pg_pool_t::FLAG_NOSCRUB)) && !time_for_deep) + pool.info.has_flag(pg_pool_t::FLAG_NOSCRUB)) && !time_for_deep) { + if (scrubber.reserved) { + // cancel scrub if it is still in scheduling, + // so pgs from other pools where scrub are still legal + // have a chance to go ahead with scrubbing. + clear_scrub_reserved(); + scrub_unreserve_replicas(); + } return false; + } } if (cct->_conf->osd_scrub_auto_repair @@ -4108,6 +4118,7 @@ bool boundary_found = false; hobject_t start = scrubber.start; + unsigned loop = 0; while (!boundary_found) { vector objects; ret = get_pgbackend()->objects_list_partial( @@ -4137,6 +4148,12 @@ boundary_found = true; } } + + // reset handle once in a while, the search maybe takes long. + if (++loop >= g_conf->osd_loop_before_reset_tphandle) { + handle.reset_tp_timeout(); + loop = 0; + } } if (!_range_available_for_scrub(scrubber.start, candidate_end)) { diff -Nru ceph-10.1.1/src/osd/ReplicatedPG.cc ceph-10.1.2/src/osd/ReplicatedPG.cc --- ceph-10.1.1/src/osd/ReplicatedPG.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/osd/ReplicatedPG.cc 2016-04-12 17:46:06.000000000 +0000 @@ -1627,16 +1627,41 @@ return; } + hobject_t head(m->get_oid(), m->get_object_locator().key, + CEPH_NOSNAP, m->get_pg().ps(), + info.pgid.pool(), m->get_object_locator().nspace); + // object name too long? - unsigned max_name_len = MIN(g_conf->osd_max_object_name_len, - osd->osd->store->get_max_object_name_length()); - if (m->get_oid().name.size() > max_name_len) { - dout(4) << "do_op '" << m->get_oid().name << "' is longer than " - << max_name_len << " bytes" << dendl; + if (m->get_oid().name.size() > g_conf->osd_max_object_name_len) { + dout(4) << "do_op name is longer than " + << g_conf->osd_max_object_name_len + << " bytes" << dendl; + osd->reply_op_error(op, -ENAMETOOLONG); + return; + } + if (m->get_object_locator().key.size() > g_conf->osd_max_object_name_len) { + dout(4) << "do_op locator is longer than " + << g_conf->osd_max_object_name_len + << " bytes" << dendl; + osd->reply_op_error(op, -ENAMETOOLONG); + return; + } + if (m->get_object_locator().nspace.size() > + g_conf->osd_max_object_namespace_len) { + dout(4) << "do_op namespace is longer than " + << g_conf->osd_max_object_namespace_len + << " bytes" << dendl; osd->reply_op_error(op, -ENAMETOOLONG); return; } + if (int r = osd->store->validate_hobject_key(head)) { + dout(4) << "do_op object " << head << " invalid for backing store: " + << r << dendl; + osd->reply_op_error(op, r); + return; + } + // blacklisted? if (get_osdmap()->is_blacklisted(m->get_source_addr())) { dout(10) << "do_op " << m->get_source_addr() << " is blacklisted" << dendl; @@ -1702,11 +1727,6 @@ << " flags " << ceph_osd_flag_string(m->get_flags()) << dendl; - hobject_t head(m->get_oid(), m->get_object_locator().key, - CEPH_NOSNAP, m->get_pg().ps(), - info.pgid.pool(), m->get_object_locator().nspace); - - if (write_ordered && scrubber.write_blocked_by_scrub(head, get_sort_bitwise())) { dout(20) << __func__ << ": waiting for scrub" << dendl; @@ -6199,6 +6219,15 @@ obs.oi.set_omap_digest(rollback_to->obs.oi.omap_digest); else obs.oi.clear_omap_digest(); + + if (rollback_to->obs.oi.is_omap()) { + dout(10) << __func__ << " setting omap flag on " << obs.oi.soid << dendl; + obs.oi.set_flag(object_info_t::FLAG_OMAP); + } else { + dout(10) << __func__ << " clearing omap flag on " << obs.oi.soid << dendl; + obs.oi.clear_flag(object_info_t::FLAG_OMAP); + } + snapset.head_exists = true; } } @@ -6919,10 +6948,6 @@ return result; } - if ((osd_op.op.copy_get.flags & CEPH_OSD_COPY_GET_FLAG_NOTSUPP_OMAP) && - oi.is_omap()) - return -EOPNOTSUPP; - MOSDOp *op = reinterpret_cast(ctx->op->get_req()); uint64_t features = op->get_features(); @@ -7162,12 +7187,7 @@ // it already! assert(cop->cursor.is_initial()); } - - uint32_t copyget_flags = 0; - if (!pool.info.supports_omap()) - copyget_flags |= CEPH_OSD_COPY_GET_FLAG_NOTSUPP_OMAP; - - op.copy_get(&cop->cursor, get_copy_chunk_size(), copyget_flags, + op.copy_get(&cop->cursor, get_copy_chunk_size(), &cop->results.object_size, &cop->results.mtime, &cop->attrs, &cop->data, &cop->omap_header, &cop->omap_data, &cop->results.snaps, &cop->results.snap_seq, @@ -7330,6 +7350,16 @@ copy_ops.erase(cobc->obs.oi.soid); cobc->stop_block(); + if (r < 0 && cop->results.started_temp_obj) { + dout(10) << __func__ << " deleting partial temp object " + << cop->results.temp_oid << dendl; + ObjectContextRef tempobc = get_object_context(cop->results.temp_oid, true); + OpContextUPtr ctx = simple_opc_create(tempobc); + ctx->op_t->remove(cop->results.temp_oid); + ctx->discard_temp_oid = cop->results.temp_oid; + simple_opc_submit(std::move(ctx)); + } + // cancel and requeue proxy ops on this object if (!r) { for (map::iterator it = proxyread_ops.begin(); diff -Nru ceph-10.1.1/src/osdc/Objecter.cc ceph-10.1.2/src/osdc/Objecter.cc --- ceph-10.1.1/src/osdc/Objecter.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/osdc/Objecter.cc 2016-04-12 17:46:06.000000000 +0000 @@ -542,10 +542,10 @@ } sl.unlock(); - info->register_tid = _op_submit(o, sul); + _op_submit(o, sul, &info->register_tid); } else { // first send - info->register_tid = _op_submit_with_budget(o, sul); + _op_submit_with_budget(o, sul, &info->register_tid); } logger->inc(l_osdc_linger_send); @@ -2125,14 +2125,18 @@ // read | write --------------------------- -ceph_tid_t Objecter::op_submit(Op *op, int *ctx_budget) +void Objecter::op_submit(Op *op, ceph_tid_t *ptid, int *ctx_budget) { shunique_lock rl(rwlock, ceph::acquire_shared); - return _op_submit_with_budget(op, rl, ctx_budget); + ceph_tid_t tid = 0; + if (!ptid) + ptid = &tid; + _op_submit_with_budget(op, rl, ptid, ctx_budget); } -ceph_tid_t Objecter::_op_submit_with_budget(Op *op, shunique_lock& sul, - int *ctx_budget) +void Objecter::_op_submit_with_budget(Op *op, shunique_lock& sul, + ceph_tid_t *ptid, + int *ctx_budget) { assert(initialized.read()); @@ -2160,7 +2164,7 @@ op_cancel(tid, -ETIMEDOUT); }); } - return _op_submit(op, sul); + _op_submit(op, sul, ptid); } void Objecter::_send_op_account(Op *op) @@ -2242,7 +2246,7 @@ } } -ceph_tid_t Objecter::_op_submit(Op *op, shunique_lock& sul) +void Objecter::_op_submit(Op *op, shunique_lock& sul, ceph_tid_t *ptid) { // rwlock is locked @@ -2335,6 +2339,8 @@ if (check_for_latest_map) { _send_op_map_check(op); } + if (ptid) + *ptid = tid; op = NULL; sl.unlock(); @@ -2342,8 +2348,6 @@ ldout(cct, 5) << num_unacked.read() << " unacked, " << num_uncommitted.read() << " uncommitted" << dendl; - - return tid; } int Objecter::op_cancel(OSDSession *s, ceph_tid_t tid, int r) @@ -3219,7 +3223,7 @@ m->get_redirect().combine_with_locator(op->target.target_oloc, op->target.target_oid.name); op->target.flags |= CEPH_OSD_FLAG_REDIRECTED; - _op_submit(op, sul); + _op_submit(op, sul, NULL); m->put(); return; } diff -Nru ceph-10.1.1/src/osdc/Objecter.h ceph-10.1.2/src/osdc/Objecter.h --- ceph-10.1.1/src/osdc/Objecter.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/osdc/Objecter.h 2016-04-12 17:46:06.000000000 +0000 @@ -747,7 +747,6 @@ void copy_get(object_copy_cursor_t *cursor, uint64_t max, - uint32_t copyget_flags, uint64_t *out_size, ceph::real_time *out_mtime, std::map *out_attrs, @@ -765,7 +764,6 @@ int *prval) { OSDOp& osd_op = add_op(CEPH_OSD_OP_COPY_GET); osd_op.op.copy_get.max = max; - osd_op.op.copy_get.flags = copyget_flags; ::encode(*cursor, osd_op.indata); ::encode(max, osd_op.indata); unsigned p = ops.size() - 1; @@ -2074,14 +2072,15 @@ private: // low-level - ceph_tid_t _op_submit(Op *op, shunique_lock& lc); - ceph_tid_t _op_submit_with_budget(Op *op, shunique_lock& lc, - int *ctx_budget = NULL); + void _op_submit(Op *op, shunique_lock& lc, ceph_tid_t *ptid); + void _op_submit_with_budget(Op *op, shunique_lock& lc, + ceph_tid_t *ptid, + int *ctx_budget = NULL); inline void unregister_op(Op *op); // public interface public: - ceph_tid_t op_submit(Op *op, int *ctx_budget = NULL); + void op_submit(Op *op, ceph_tid_t *ptid = NULL, int *ctx_budget = NULL); bool is_active() { shared_lock l(rwlock); return !((!inflight_ops.read()) && linger_ops.empty() && @@ -2173,11 +2172,12 @@ } // mid-level helpers - Op *prepare_mutate_op(const object_t& oid, const object_locator_t& oloc, - ObjectOperation& op, const SnapContext& snapc, - ceph::real_time mtime, int flags, Context *onack, - Context *oncommit, version_t *objver = NULL, - osd_reqid_t reqid = osd_reqid_t()) { + Op *prepare_mutate_op( + const object_t& oid, const object_locator_t& oloc, + ObjectOperation& op, const SnapContext& snapc, + ceph::real_time mtime, int flags, Context *onack, + Context *oncommit, version_t *objver = NULL, + osd_reqid_t reqid = osd_reqid_t()) { Op *o = new Op(oid, oloc, op.ops, flags | global_op_flags.read() | CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->priority = op.priority; @@ -2187,20 +2187,25 @@ o->reqid = reqid; return o; } - ceph_tid_t mutate(const object_t& oid, const object_locator_t& oloc, - ObjectOperation& op, const SnapContext& snapc, - ceph::real_time mtime, int flags, Context *onack, - Context *oncommit, version_t *objver = NULL, - osd_reqid_t reqid = osd_reqid_t()) { + ceph_tid_t mutate( + const object_t& oid, const object_locator_t& oloc, + ObjectOperation& op, const SnapContext& snapc, + ceph::real_time mtime, int flags, Context *onack, + Context *oncommit, version_t *objver = NULL, + osd_reqid_t reqid = osd_reqid_t()) { Op *o = prepare_mutate_op(oid, oloc, op, snapc, mtime, flags, onack, oncommit, objver, reqid); - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } - Op *prepare_read_op(const object_t& oid, const object_locator_t& oloc, - ObjectOperation& op, - snapid_t snapid, bufferlist *pbl, int flags, - Context *onack, version_t *objver = NULL, - int *data_offset = NULL) { + Op *prepare_read_op( + const object_t& oid, const object_locator_t& oloc, + ObjectOperation& op, + snapid_t snapid, bufferlist *pbl, int flags, + Context *onack, version_t *objver = NULL, + int *data_offset = NULL, + uint64_t features = 0) { Op *o = new Op(oid, oloc, op.ops, flags | global_op_flags.read() | CEPH_OSD_FLAG_READ, onack, NULL, objver, data_offset); o->priority = op.priority; @@ -2213,22 +2218,26 @@ o->out_rval.swap(op.out_rval); return o; } - ceph_tid_t read(const object_t& oid, const object_locator_t& oloc, - ObjectOperation& op, - snapid_t snapid, bufferlist *pbl, int flags, - Context *onack, version_t *objver = NULL, - int *data_offset = NULL, - uint64_t features = 0) { + ceph_tid_t read( + const object_t& oid, const object_locator_t& oloc, + ObjectOperation& op, + snapid_t snapid, bufferlist *pbl, int flags, + Context *onack, version_t *objver = NULL, + int *data_offset = NULL, + uint64_t features = 0) { Op *o = prepare_read_op(oid, oloc, op, snapid, pbl, flags, onack, objver, data_offset); if (features) o->features = features; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } - ceph_tid_t pg_read(uint32_t hash, object_locator_t oloc, - ObjectOperation& op, bufferlist *pbl, int flags, - Context *onack, epoch_t *reply_epoch, - int *ctx_budget) { + Op *prepare_pg_read_op( + uint32_t hash, object_locator_t oloc, + ObjectOperation& op, bufferlist *pbl, int flags, + Context *onack, epoch_t *reply_epoch, + int *ctx_budget) { Op *o = new Op(object_t(), oloc, op.ops, flags | global_op_flags.read() | CEPH_OSD_FLAG_READ, onack, NULL, NULL); @@ -2245,7 +2254,18 @@ // budget is tracked by listing context o->ctx_budgeted = true; } - return op_submit(o, ctx_budget); + return o; + } + ceph_tid_t pg_read( + uint32_t hash, object_locator_t oloc, + ObjectOperation& op, bufferlist *pbl, int flags, + Context *onack, epoch_t *reply_epoch, + int *ctx_budget) { + Op *o = prepare_pg_read_op(hash, oloc, op, pbl, flags, + onack, reply_epoch, ctx_budget); + ceph_tid_t tid; + op_submit(o, &tid, ctx_budget); + return tid; } // caller owns a ref @@ -2297,10 +2317,11 @@ // high-level helpers - ceph_tid_t stat(const object_t& oid, const object_locator_t& oloc, - snapid_t snap, uint64_t *psize, ceph::real_time *pmtime, - int flags, Context *onfinish, version_t *objver = NULL, - ObjectOperation *extra_ops = NULL) { + Op *prepare_stat_op( + const object_t& oid, const object_locator_t& oloc, + snapid_t snap, uint64_t *psize, ceph::real_time *pmtime, + int flags, Context *onfinish, version_t *objver = NULL, + ObjectOperation *extra_ops = NULL) { vector ops; int i = init_ops(ops, 1, extra_ops); ops[i].op.op = CEPH_OSD_OP_STAT; @@ -2309,13 +2330,25 @@ CEPH_OSD_FLAG_READ, fin, 0, objver); o->snapid = snap; o->outbl = &fin->bl; - return op_submit(o); + return o; + } + ceph_tid_t stat( + const object_t& oid, const object_locator_t& oloc, + snapid_t snap, uint64_t *psize, ceph::real_time *pmtime, + int flags, Context *onfinish, version_t *objver = NULL, + ObjectOperation *extra_ops = NULL) { + Op *o = prepare_stat_op(oid, oloc, snap, psize, pmtime, flags, + onfinish, objver, extra_ops); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } - ceph_tid_t read(const object_t& oid, const object_locator_t& oloc, - uint64_t off, uint64_t len, snapid_t snap, bufferlist *pbl, - int flags, Context *onfinish, version_t *objver = NULL, - ObjectOperation *extra_ops = NULL, int op_flags = 0) { + Op *prepare_read_op( + const object_t& oid, const object_locator_t& oloc, + uint64_t off, uint64_t len, snapid_t snap, bufferlist *pbl, + int flags, Context *onfinish, version_t *objver = NULL, + ObjectOperation *extra_ops = NULL, int op_flags = 0) { vector ops; int i = init_ops(ops, 1, extra_ops); ops[i].op.op = CEPH_OSD_OP_READ; @@ -2328,7 +2361,18 @@ CEPH_OSD_FLAG_READ, onfinish, 0, objver); o->snapid = snap; o->outbl = pbl; - return op_submit(o); + return o; + } + ceph_tid_t read( + const object_t& oid, const object_locator_t& oloc, + uint64_t off, uint64_t len, snapid_t snap, bufferlist *pbl, + int flags, Context *onfinish, version_t *objver = NULL, + ObjectOperation *extra_ops = NULL, int op_flags = 0) { + Op *o = prepare_read_op(oid, oloc, off, len, snap, pbl, flags, + onfinish, objver, extra_ops, op_flags); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t read_trunc(const object_t& oid, const object_locator_t& oloc, @@ -2349,7 +2393,9 @@ CEPH_OSD_FLAG_READ, onfinish, 0, objver); o->snapid = snap; o->outbl = pbl; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t mapext(const object_t& oid, const object_locator_t& oloc, uint64_t off, uint64_t len, snapid_t snap, bufferlist *pbl, @@ -2366,7 +2412,9 @@ CEPH_OSD_FLAG_READ, onfinish, 0, objver); o->snapid = snap; o->outbl = pbl; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t getxattr(const object_t& oid, const object_locator_t& oloc, const char *name, snapid_t snap, bufferlist *pbl, int flags, @@ -2383,7 +2431,9 @@ CEPH_OSD_FLAG_READ, onfinish, 0, objver); o->snapid = snap; o->outbl = pbl; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t getxattrs(const object_t& oid, const object_locator_t& oloc, @@ -2398,7 +2448,9 @@ CEPH_OSD_FLAG_READ, fin, 0, objver); o->snapid = snap; o->outbl = &fin->bl; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t read_full(const object_t& oid, const object_locator_t& oloc, @@ -2420,13 +2472,16 @@ CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } - ceph_tid_t write(const object_t& oid, const object_locator_t& oloc, - uint64_t off, uint64_t len, const SnapContext& snapc, - const bufferlist &bl, ceph::real_time mtime, int flags, - Context *onack, Context *oncommit, version_t *objver = NULL, - ObjectOperation *extra_ops = NULL, int op_flags = 0) { + Op *prepare_write_op( + const object_t& oid, const object_locator_t& oloc, + uint64_t off, uint64_t len, const SnapContext& snapc, + const bufferlist &bl, ceph::real_time mtime, int flags, + Context *onack, Context *oncommit, version_t *objver = NULL, + ObjectOperation *extra_ops = NULL, int op_flags = 0) { vector ops; int i = init_ops(ops, 1, extra_ops); ops[i].op.op = CEPH_OSD_OP_WRITE; @@ -2440,14 +2495,27 @@ CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + return o; } - ceph_tid_t append(const object_t& oid, const object_locator_t& oloc, - uint64_t len, const SnapContext& snapc, - const bufferlist &bl, ceph::real_time mtime, int flags, - Context *onack, Context *oncommit, - version_t *objver = NULL, - ObjectOperation *extra_ops = NULL) { + ceph_tid_t write( + const object_t& oid, const object_locator_t& oloc, + uint64_t off, uint64_t len, const SnapContext& snapc, + const bufferlist &bl, ceph::real_time mtime, int flags, + Context *onack, Context *oncommit, version_t *objver = NULL, + ObjectOperation *extra_ops = NULL, int op_flags = 0) { + Op *o = prepare_write_op(oid, oloc, off, len, snapc, bl, mtime, flags, + onack, oncommit, objver, extra_ops, op_flags); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; + } + Op *prepare_append_op( + const object_t& oid, const object_locator_t& oloc, + uint64_t len, const SnapContext& snapc, + const bufferlist &bl, ceph::real_time mtime, int flags, + Context *onack, Context *oncommit, + version_t *objver = NULL, + ObjectOperation *extra_ops = NULL) { vector ops; int i = init_ops(ops, 1, extra_ops); ops[i].op.op = CEPH_OSD_OP_APPEND; @@ -2460,7 +2528,20 @@ CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + return o; + } + ceph_tid_t append( + const object_t& oid, const object_locator_t& oloc, + uint64_t len, const SnapContext& snapc, + const bufferlist &bl, ceph::real_time mtime, int flags, + Context *onack, Context *oncommit, + version_t *objver = NULL, + ObjectOperation *extra_ops = NULL) { + Op *o = prepare_append_op(oid, oloc, len, snapc, bl, mtime, flags, + onack, oncommit, objver, extra_ops); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t write_trunc(const object_t& oid, const object_locator_t& oloc, uint64_t off, uint64_t len, const SnapContext& snapc, @@ -2482,13 +2563,16 @@ CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } - ceph_tid_t write_full(const object_t& oid, const object_locator_t& oloc, - const SnapContext& snapc, const bufferlist &bl, - ceph::real_time mtime, int flags, Context *onack, - Context *oncommit, version_t *objver = NULL, - ObjectOperation *extra_ops = NULL, int op_flags = 0) { + Op *prepare_write_full_op( + const object_t& oid, const object_locator_t& oloc, + const SnapContext& snapc, const bufferlist &bl, + ceph::real_time mtime, int flags, Context *onack, + Context *oncommit, version_t *objver = NULL, + ObjectOperation *extra_ops = NULL, int op_flags = 0) { vector ops; int i = init_ops(ops, 1, extra_ops); ops[i].op.op = CEPH_OSD_OP_WRITEFULL; @@ -2500,7 +2584,19 @@ CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + return o; + } + ceph_tid_t write_full( + const object_t& oid, const object_locator_t& oloc, + const SnapContext& snapc, const bufferlist &bl, + ceph::real_time mtime, int flags, Context *onack, + Context *oncommit, version_t *objver = NULL, + ObjectOperation *extra_ops = NULL, int op_flags = 0) { + Op *o = prepare_write_full_op(oid, oloc, snapc, bl, mtime, flags, + onack, oncommit, objver, extra_ops, op_flags); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t trunc(const object_t& oid, const object_locator_t& oloc, const SnapContext& snapc, ceph::real_time mtime, int flags, @@ -2517,7 +2613,9 @@ CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t zero(const object_t& oid, const object_locator_t& oloc, uint64_t off, uint64_t len, const SnapContext& snapc, @@ -2532,7 +2630,9 @@ CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t rollback_object(const object_t& oid, const object_locator_t& oloc, const SnapContext& snapc, snapid_t snapid, @@ -2547,7 +2647,9 @@ objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t create(const object_t& oid, const object_locator_t& oloc, const SnapContext& snapc, ceph::real_time mtime, int global_flags, @@ -2562,12 +2664,15 @@ CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } - ceph_tid_t remove(const object_t& oid, const object_locator_t& oloc, - const SnapContext& snapc, ceph::real_time mtime, int flags, - Context *onack, Context *oncommit, - version_t *objver = NULL, ObjectOperation *extra_ops = NULL) { + Op *prepare_remove_op( + const object_t& oid, const object_locator_t& oloc, + const SnapContext& snapc, ceph::real_time mtime, int flags, + Context *onack, Context *oncommit, + version_t *objver = NULL, ObjectOperation *extra_ops = NULL) { vector ops; int i = init_ops(ops, 1, extra_ops); ops[i].op.op = CEPH_OSD_OP_DELETE; @@ -2575,7 +2680,18 @@ CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + return o; + } + ceph_tid_t remove( + const object_t& oid, const object_locator_t& oloc, + const SnapContext& snapc, ceph::real_time mtime, int flags, + Context *onack, Context *oncommit, + version_t *objver = NULL, ObjectOperation *extra_ops = NULL) { + Op *o = prepare_remove_op(oid, oloc, snapc, mtime, flags, + onack, oncommit, objver, extra_ops); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t setxattr(const object_t& oid, const object_locator_t& oloc, @@ -2595,7 +2711,9 @@ CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } ceph_tid_t removexattr(const object_t& oid, const object_locator_t& oloc, const char *name, const SnapContext& snapc, @@ -2613,7 +2731,9 @@ CEPH_OSD_FLAG_WRITE, onack, oncommit, objver); o->mtime = mtime; o->snapc = snapc; - return op_submit(o); + ceph_tid_t tid; + op_submit(o, &tid); + return tid; } void list_nobjects(NListContext *p, Context *onfinish); diff -Nru ceph-10.1.1/src/pybind/ceph_rest_api.py ceph-10.1.2/src/pybind/ceph_rest_api.py --- ceph-10.1.1/src/pybind/ceph_rest_api.py 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/pybind/ceph_rest_api.py 2016-04-12 17:46:06.000000000 +0000 @@ -224,15 +224,6 @@ # prefixes go in the URL path if desc.t == CephPrefix: url += '/' + desc.instance.prefix - # CephChoices with 1 required string (not --) do too, unless - # we've already started collecting params, in which case they - # too are params - elif (desc.t == CephChoices and - len(desc.instance.strings) == 1 and - desc.req and - not str(desc.instance).startswith('--') and - not params): - url += '/' + str(desc.instance) else: # tell/ is a weird case; the URL includes what # would everywhere else be a parameter diff -Nru ceph-10.1.1/src/rgw/librgw.cc ceph-10.1.2/src/rgw/librgw.cc --- ceph-10.1.1/src/rgw/librgw.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/rgw/librgw.cc 2016-04-12 17:46:06.000000000 +0000 @@ -545,16 +545,29 @@ } int RGWLibRequest::read_permissions(RGWOp* op) { + /* bucket and object ops */ int ret = rgw_build_bucket_policies(rgwlib.get_store(), get_state()); if (ret < 0) { - ldout(get_state()->cct, 10) << "read_permissions on " + ldout(get_state()->cct, 10) << "read_permissions (bucket policy) on " << get_state()->bucket << ":" << get_state()->object << " only_bucket=" << only_bucket() << " ret=" << ret << dendl; if (ret == -ENODATA) ret = -EACCES; + } else if (! only_bucket()) { + /* object ops */ + ret = rgw_build_object_policies(rgwlib.get_store(), get_state(), + op->prefetch_data()); + if (ret < 0) { + ldout(get_state()->cct, 10) << "read_permissions (object policy) on" + << get_state()->bucket << ":" + << get_state()->object + << " ret=" << ret << dendl; + if (ret == -ENODATA) + ret = -EACCES; + } } return ret; } /* RGWLibRequest::read_permissions */ diff -Nru ceph-10.1.1/src/rgw/rgw_admin.cc ceph-10.1.2/src/rgw/rgw_admin.cc --- ceph-10.1.1/src/rgw/rgw_admin.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/rgw/rgw_admin.cc 2016-04-12 17:46:06.000000000 +0000 @@ -2293,7 +2293,6 @@ } else if (ceph_argparse_witharg(args, i, &val, "--zone-new-name", (char*)NULL)) { zone_new_name = val; } else if (ceph_argparse_witharg(args, i, &val, "--endpoints", (char*)NULL)) { - list::iterator iter; get_str_list(val, endpoints); } else if (ceph_argparse_witharg(args, i, &val, "--source-zone", (char*)NULL)) { source_zone_name = val; @@ -2308,7 +2307,7 @@ tenant = user_id.tenant; } else { if (user_id.empty()) { - cerr << "ERROR: --tennant is set, but there's no user ID" << std::endl; + cerr << "ERROR: --tenant is set, but there's no user ID" << std::endl; return EINVAL; } user_id.tenant = tenant; @@ -2453,7 +2452,7 @@ case OPT_PERIOD_DELETE: { if (period_id.empty()) { - cerr << "missing realm name or id" << std::endl; + cerr << "missing period id" << std::endl; return -EINVAL; } RGWPeriod period(period_id); @@ -3315,13 +3314,18 @@ if(zone.realm_id.empty()) { RGWRealm realm(realm_id, realm_name); int ret = realm.init(g_ceph_context, store); - if (ret < 0) { + if (ret < 0 && ret != -ENOENT) { cerr << "failed to init realm: " << cpp_strerror(-ret) << std::endl; return -ret; } zone.realm_id = realm.get_id(); } + if( !zone_name.empty() && !zone.get_name().empty() && zone.get_name() != zone_name) { + cerr << "Error: zone name" << zone_name << " is different than the zone name " << zone.get_name() << " in the provided json " << std::endl; + return -EINVAL; + } + if (zone.get_name().empty()) { zone.set_name(zone_name); if (zone.get_name().empty()) { @@ -3330,11 +3334,6 @@ } } - if(zone.get_name() != zone_name) { - cerr << "Error: zone name" << zone_name << " is different than the zone name " << zone.get_name() << " in the provided json " << std::endl; - return -EINVAL; - } - zone_name = zone.get_name(); if (zone.get_id().empty()) { diff -Nru ceph-10.1.1/src/rgw/rgw_common.h ceph-10.1.2/src/rgw/rgw_common.h --- ceph-10.1.1/src/rgw/rgw_common.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/rgw/rgw_common.h 2016-04-12 17:46:06.000000000 +0000 @@ -1081,6 +1081,7 @@ const char *method; string script_uri; string request_uri; + string request_uri_aws4; string effective_uri; string request_params; string domain; diff -Nru ceph-10.1.1/src/rgw/rgw_ldap.h ceph-10.1.2/src/rgw/rgw_ldap.h --- ceph-10.1.1/src/rgw/rgw_ldap.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/rgw/rgw_ldap.h 2016-04-12 17:46:06.000000000 +0000 @@ -4,8 +4,10 @@ #ifndef RGW_LDAP_H #define RGW_LDAP_H +#if defined(HAVE_OPENLDAP) #define LDAP_DEPRECATED 1 #include "ldap.h" +#endif #include #include diff -Nru ceph-10.1.1/src/rgw/rgw_op.cc ceph-10.1.2/src/rgw/rgw_op.cc --- ceph-10.1.1/src/rgw/rgw_op.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/rgw/rgw_op.cc 2016-04-12 17:46:06.000000000 +0000 @@ -398,7 +398,8 @@ * only_bucket: If true, reads the bucket ACL rather than the object ACL. * Returns: 0 on success, -ERR# otherwise. */ -static int rgw_build_object_policies(RGWRados *store, struct req_state *s, bool prefetch_data) +int rgw_build_object_policies(RGWRados *store, struct req_state *s, + bool prefetch_data) { int ret = 0; @@ -2994,7 +2995,8 @@ return; } - rgw_obj obj(s->bucket, s->object); + rgw_obj obj(s->bucket, s->object.name); + obj.set_instance(s->object.instance); map attrs; diff -Nru ceph-10.1.1/src/rgw/rgw_op.h ceph-10.1.2/src/rgw/rgw_op.h --- ceph-10.1.1/src/rgw/rgw_op.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/rgw/rgw_op.h 2016-04-12 17:46:06.000000000 +0000 @@ -1338,6 +1338,8 @@ }; extern int rgw_build_bucket_policies(RGWRados* store, struct req_state* s); +extern int rgw_build_object_policies(RGWRados *store, struct req_state *s, + bool prefetch_data); static inline int put_data_and_throttle(RGWPutObjProcessor *processor, bufferlist& data, off_t ofs, diff -Nru ceph-10.1.1/src/rgw/rgw_rest.cc ceph-10.1.2/src/rgw/rgw_rest.cc --- ceph-10.1.1/src/rgw/rgw_rest.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/rgw/rgw_rest.cc 2016-04-12 17:46:06.000000000 +0000 @@ -1590,6 +1590,11 @@ { req_info& info = s->info; + /* save the request uri used to hash on the client side. request_uri may suffer + modifications as part of the bucket encoding in the subdomain calling format. + request_uri_aws4 will be used under aws4 auth */ + s->info.request_uri_aws4 = s->info.request_uri; + s->cio = cio; if (info.host.size()) { ldout(s->cct, 10) << "host=" << info.host << dendl; diff -Nru ceph-10.1.1/src/rgw/rgw_rest_conn.cc ceph-10.1.2/src/rgw/rgw_rest_conn.cc --- ceph-10.1.1/src/rgw/rgw_rest_conn.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/rgw/rgw_rest_conn.cc 2016-04-12 17:46:06.000000000 +0000 @@ -199,7 +199,7 @@ map headers; if (extra_headers) { - headers.insert(extra_params->begin(), extra_params->end()); + headers.insert(extra_headers->begin(), extra_headers->end()); } ret = req.get_resource(key, headers, resource, mgr); diff -Nru ceph-10.1.1/src/rgw/rgw_rest_s3.cc ceph-10.1.2/src/rgw/rgw_rest_s3.cc --- ceph-10.1.1/src/rgw/rgw_rest_s3.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/rgw/rgw_rest_s3.cc 2016-04-12 17:46:06.000000000 +0000 @@ -3409,7 +3409,7 @@ * that SigV4 typically does. this code follows the same approach that boto library * see auth.py:canonical_uri(...) */ - s->aws4_auth->canonical_uri = s->info.request_uri; + s->aws4_auth->canonical_uri = s->info.request_uri_aws4; if (s->aws4_auth->canonical_uri.empty()) { s->aws4_auth->canonical_uri = "/"; @@ -3437,8 +3437,20 @@ string encoded_key; string encoded_val; if (key != "X-Amz-Credential") { - aws4_uri_encode(key, encoded_key); - aws4_uri_encode(val, encoded_val); + string key_decoded; + url_decode(key, key_decoded); + if (key.length() != key_decoded.length()) { + encoded_key = key; + } else { + aws4_uri_encode(key, encoded_key); + } + string val_decoded; + url_decode(val, val_decoded); + if (val.length() != val_decoded.length()) { + encoded_val = val; + } else { + aws4_uri_encode(val, encoded_val); + } } else { encoded_key = key; encoded_val = val; diff -Nru ceph-10.1.1/src/rgw/rgw_sync.cc ceph-10.1.2/src/rgw/rgw_sync.cc --- ceph-10.1.1/src/rgw/rgw_sync.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/rgw/rgw_sync.cc 2016-04-12 17:46:06.000000000 +0000 @@ -1868,13 +1868,28 @@ } RGWObjectCtx obj_ctx(store, NULL); + int r = 0; // get shard count and oldest log period from master rgw_mdlog_info mdlog_info; - int r = read_log_info(&mdlog_info); - if (r < 0) { - lderr(store->ctx()) << "ERROR: fail to fetch master log info (r=" << r << ")" << dendl; - return r; + for (;;) { + if (going_down.read()) { + ldout(store->ctx(), 1) << __func__ << "(): going down" << dendl; + return 0; + } + r = read_log_info(&mdlog_info); + if (r == -EIO) { + // keep retrying if master isn't alive + ldout(store->ctx(), 10) << __func__ << "(): waiting for master.." << dendl; + backoff.backoff_sleep(); + continue; + } + backoff.reset(); + if (r < 0) { + lderr(store->ctx()) << "ERROR: fail to fetch master log info (r=" << r << ")" << dendl; + return r; + } + break; } do { @@ -1924,7 +1939,7 @@ auto num_shards = sync_status.sync_info.num_shards; if (num_shards != mdlog_info.num_shards) { lderr(store->ctx()) << "ERROR: can't sync, mismatch between num shards, master num_shards=" << mdlog_info.num_shards << " local num_shards=" << num_shards << dendl; - return r; + return -EINVAL; } RGWPeriodHistory::Cursor cursor; diff -Nru ceph-10.1.1/src/test/cli/ceph-authtool/help.t ceph-10.1.2/src/test/cli/ceph-authtool/help.t --- ceph-10.1.1/src/test/cli/ceph-authtool/help.t 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/cli/ceph-authtool/help.t 2016-04-12 17:46:06.000000000 +0000 @@ -14,12 +14,13 @@ specified entityname --gen-print-key will generate a new secret key without set it to the keyringfile, prints the secret to stdout - --import-keyring will import the content of a given keyring + --import-keyring FILE will import the content of a given keyring into the keyringfile - -u, --set-uid sets the auid (authenticated user id) for the + -n NAME, --name NAME specify entityname to operate on + -u AUID, --set-uid AUID sets the auid (authenticated user id) for the specified entityname - -a, --add-key will add an encoded key to the keyring - --cap subsystem capability will set the capability for given subsystem - --caps capsfile will set all of capabilities associated with a + -a BASE64, --add-key BASE64 will add an encoded key to the keyring + --cap SUBSYSTEM CAPABILITY will set the capability for given subsystem + --caps CAPSFILE will set all of capabilities associated with a given key, for all subsystems [1] diff -Nru ceph-10.1.1/src/test/cli/ceph-authtool/manpage.t ceph-10.1.2/src/test/cli/ceph-authtool/manpage.t --- ceph-10.1.1/src/test/cli/ceph-authtool/manpage.t 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/cli/ceph-authtool/manpage.t 2016-04-12 17:46:06.000000000 +0000 @@ -13,13 +13,14 @@ specified entityname --gen-print-key will generate a new secret key without set it to the keyringfile, prints the secret to stdout - --import-keyring will import the content of a given keyring + --import-keyring FILE will import the content of a given keyring into the keyringfile - -u, --set-uid sets the auid (authenticated user id) for the + -n NAME, --name NAME specify entityname to operate on + -u AUID, --set-uid AUID sets the auid (authenticated user id) for the specified entityname - -a, --add-key will add an encoded key to the keyring - --cap subsystem capability will set the capability for given subsystem - --caps capsfile will set all of capabilities associated with a + -a BASE64, --add-key BASE64 will add an encoded key to the keyring + --cap SUBSYSTEM CAPABILITY will set the capability for given subsystem + --caps CAPSFILE will set all of capabilities associated with a given key, for all subsystems [1] diff -Nru ceph-10.1.1/src/test/cli/ceph-authtool/simple.t ceph-10.1.2/src/test/cli/ceph-authtool/simple.t --- ceph-10.1.1/src/test/cli/ceph-authtool/simple.t 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/cli/ceph-authtool/simple.t 2016-04-12 17:46:06.000000000 +0000 @@ -13,12 +13,13 @@ specified entityname --gen-print-key will generate a new secret key without set it to the keyringfile, prints the secret to stdout - --import-keyring will import the content of a given keyring + --import-keyring FILE will import the content of a given keyring into the keyringfile - -u, --set-uid sets the auid (authenticated user id) for the + -n NAME, --name NAME specify entityname to operate on + -u AUID, --set-uid AUID sets the auid (authenticated user id) for the specified entityname - -a, --add-key will add an encoded key to the keyring - --cap subsystem capability will set the capability for given subsystem - --caps capsfile will set all of capabilities associated with a + -a BASE64, --add-key BASE64 will add an encoded key to the keyring + --cap SUBSYSTEM CAPABILITY will set the capability for given subsystem + --caps CAPSFILE will set all of capabilities associated with a given key, for all subsystems [1] diff -Nru ceph-10.1.1/src/test/encoding/check-generated.sh ceph-10.1.2/src/test/encoding/check-generated.sh --- ceph-10.1.1/src/test/encoding/check-generated.sh 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/encoding/check-generated.sh 2016-04-12 17:46:06.000000000 +0000 @@ -1,4 +1,4 @@ -#!/bin/sh -e +#!/bin/bash -e source ../qa/workunits/ceph-helpers.sh diff -Nru ceph-10.1.1/src/test/encoding/types.h ceph-10.1.2/src/test/encoding/types.h --- ceph-10.1.1/src/test/encoding/types.h 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/encoding/types.h 2016-04-12 17:46:06.000000000 +0000 @@ -192,6 +192,10 @@ TYPE_FEATUREFUL(MDSMap) TYPE_FEATUREFUL(MDSMap::mds_info_t) +#include "mds/FSMap.h" +//TYPE_FEATUREFUL(Filesystem) +TYPE_FEATUREFUL(FSMap) + #include "mds/Capability.h" TYPE_NOCOPY(Capability) diff -Nru ceph-10.1.1/src/test/journal/test_FutureImpl.cc ceph-10.1.2/src/test/journal/test_FutureImpl.cc --- ceph-10.1.1/src/test/journal/test_FutureImpl.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/journal/test_FutureImpl.cc 2016-04-12 17:46:06.000000000 +0000 @@ -25,12 +25,11 @@ } }; - journal::FutureImplPtr create_future(journal::JournalMetadataPtr metadata, - uint64_t tag_tid, uint64_t entry_tid, + journal::FutureImplPtr create_future(uint64_t tag_tid, uint64_t entry_tid, uint64_t commit_tid, const journal::FutureImplPtr &prev = journal::FutureImplPtr()) { - journal::FutureImplPtr future(new journal::FutureImpl(metadata, tag_tid, + journal::FutureImplPtr future(new journal::FutureImpl(tag_tid, entry_tid, commit_tid)); future->init(prev); @@ -50,7 +49,7 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future = create_future(metadata, 234, 123, 456); + journal::FutureImplPtr future = create_future(234, 123, 456); ASSERT_EQ(234U, future->get_tag_tid()); ASSERT_EQ(123U, future->get_entry_tid()); ASSERT_EQ(456U, future->get_commit_tid()); @@ -63,7 +62,7 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future = create_future(metadata, 234, 123, 456); + journal::FutureImplPtr future = create_future(234, 123, 456); ASSERT_FALSE(future->attach(&m_flush_handler)); ASSERT_EQ(1U, m_flush_handler.refs); } @@ -75,7 +74,7 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future = create_future(metadata, 234, 123, 456); + journal::FutureImplPtr future = create_future(234, 123, 456); future->flush(NULL); ASSERT_TRUE(future->attach(&m_flush_handler)); @@ -89,7 +88,7 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future = create_future(metadata, 234, 123, 456); + journal::FutureImplPtr future = create_future(234, 123, 456); ASSERT_FALSE(future->attach(&m_flush_handler)); future->detach(); ASSERT_EQ(0U, m_flush_handler.refs); @@ -102,7 +101,7 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future = create_future(metadata, 234, 123, 456); + journal::FutureImplPtr future = create_future(234, 123, 456); ASSERT_FALSE(future->attach(&m_flush_handler)); future.reset(); ASSERT_EQ(0U, m_flush_handler.refs); @@ -115,7 +114,7 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future = create_future(metadata, 234, 123, 456); + journal::FutureImplPtr future = create_future(234, 123, 456); ASSERT_FALSE(future->attach(&m_flush_handler)); C_SaferCond cond; @@ -133,7 +132,7 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future = create_future(metadata, 234, 123, 456); + journal::FutureImplPtr future = create_future(234, 123, 456); ASSERT_FALSE(future->attach(&m_flush_handler)); future->flush(NULL); @@ -150,10 +149,10 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future1 = create_future(metadata, 234, 123, 456); - journal::FutureImplPtr future2 = create_future(metadata, 234, 124, 457, + journal::FutureImplPtr future1 = create_future(234, 123, 456); + journal::FutureImplPtr future2 = create_future(234, 124, 457, future1); - journal::FutureImplPtr future3 = create_future(metadata, 235, 1, 458, + journal::FutureImplPtr future3 = create_future(235, 1, 458, future2); ASSERT_FALSE(future1->attach(&m_flush_handler)); ASSERT_FALSE(future2->attach(&m_flush_handler)); @@ -184,8 +183,8 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future1 = create_future(metadata, 234, 123, 456); - journal::FutureImplPtr future2 = create_future(metadata, 234, 124, 457, + journal::FutureImplPtr future1 = create_future(234, 123, 456); + journal::FutureImplPtr future2 = create_future(234, 124, 457, future1); ASSERT_FALSE(future1->attach(&m_flush_handler)); ASSERT_FALSE(future2->attach(&m_flush_handler)); @@ -206,7 +205,7 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future = create_future(metadata, 234, 123, 456); + journal::FutureImplPtr future = create_future(234, 123, 456); future->safe(-EIO); C_SaferCond cond; @@ -221,7 +220,7 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future = create_future(metadata, 234, 1, 456); + journal::FutureImplPtr future = create_future(234, 1, 456); C_SaferCond cond; future->wait(&cond); @@ -236,7 +235,7 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future = create_future(metadata, 234, 1, 456); + journal::FutureImplPtr future = create_future(234, 1, 456); future->safe(-EEXIST); C_SaferCond cond; @@ -251,8 +250,8 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future1 = create_future(metadata, 234, 123, 456); - journal::FutureImplPtr future2 = create_future(metadata, 234, 124, 457, + journal::FutureImplPtr future1 = create_future(234, 123, 456); + journal::FutureImplPtr future2 = create_future(234, 124, 457, future1); future1->safe(-EIO); @@ -268,8 +267,8 @@ journal::JournalMetadataPtr metadata = create_metadata(oid); ASSERT_EQ(0, init_metadata(metadata)); - journal::FutureImplPtr future1 = create_future(metadata, 234, 123, 456); - journal::FutureImplPtr future2 = create_future(metadata, 234, 124, 457, + journal::FutureImplPtr future1 = create_future(234, 123, 456); + journal::FutureImplPtr future2 = create_future(234, 124, 457, future1); future2->safe(-EEXIST); diff -Nru ceph-10.1.1/src/test/journal/test_ObjectRecorder.cc ceph-10.1.2/src/test/journal/test_ObjectRecorder.cc --- ceph-10.1.1/src/test/journal/test_ObjectRecorder.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/journal/test_ObjectRecorder.cc 2016-04-12 17:46:06.000000000 +0000 @@ -67,11 +67,10 @@ m_flush_age = i; } - journal::AppendBuffer create_append_buffer(journal::JournalMetadataPtr metadata, - uint64_t tag_tid, uint64_t entry_tid, + journal::AppendBuffer create_append_buffer(uint64_t tag_tid, uint64_t entry_tid, const std::string &payload) { - journal::FutureImplPtr future(new journal::FutureImpl(metadata, tag_tid, - entry_tid, 456)); + journal::FutureImplPtr future(new journal::FutureImpl(tag_tid, entry_tid, + 456)); future->init(journal::FutureImplPtr()); bufferlist bl; @@ -98,16 +97,14 @@ journal::ObjectRecorderPtr object = create_object(oid, 24); - journal::AppendBuffer append_buffer1 = create_append_buffer(metadata, - 234, 123, + journal::AppendBuffer append_buffer1 = create_append_buffer(234, 123, "payload"); journal::AppendBuffers append_buffers; append_buffers = {append_buffer1}; ASSERT_FALSE(object->append(append_buffers)); ASSERT_EQ(1U, object->get_pending_appends()); - journal::AppendBuffer append_buffer2 = create_append_buffer(metadata, - 234, 124, + journal::AppendBuffer append_buffer2 = create_append_buffer(234, 124, "payload"); append_buffers = {append_buffer2}; ASSERT_FALSE(object->append(append_buffers)); @@ -129,16 +126,14 @@ set_flush_interval(2); journal::ObjectRecorderPtr object = create_object(oid, 24); - journal::AppendBuffer append_buffer1 = create_append_buffer(metadata, - 234, 123, + journal::AppendBuffer append_buffer1 = create_append_buffer(234, 123, "payload"); journal::AppendBuffers append_buffers; append_buffers = {append_buffer1}; ASSERT_FALSE(object->append(append_buffers)); ASSERT_EQ(1U, object->get_pending_appends()); - journal::AppendBuffer append_buffer2 = create_append_buffer(metadata, - 234, 124, + journal::AppendBuffer append_buffer2 = create_append_buffer(234, 124, "payload"); append_buffers = {append_buffer2}; ASSERT_FALSE(object->append(append_buffers)); @@ -159,16 +154,14 @@ set_flush_bytes(10); journal::ObjectRecorderPtr object = create_object(oid, 24); - journal::AppendBuffer append_buffer1 = create_append_buffer(metadata, - 234, 123, + journal::AppendBuffer append_buffer1 = create_append_buffer(234, 123, "payload"); journal::AppendBuffers append_buffers; append_buffers = {append_buffer1}; ASSERT_FALSE(object->append(append_buffers)); ASSERT_EQ(1U, object->get_pending_appends()); - journal::AppendBuffer append_buffer2 = create_append_buffer(metadata, - 234, 124, + journal::AppendBuffer append_buffer2 = create_append_buffer(234, 124, "payload"); append_buffers = {append_buffer2}; ASSERT_FALSE(object->append(append_buffers)); @@ -189,15 +182,13 @@ set_flush_age(0.1); journal::ObjectRecorderPtr object = create_object(oid, 24); - journal::AppendBuffer append_buffer1 = create_append_buffer(metadata, - 234, 123, + journal::AppendBuffer append_buffer1 = create_append_buffer(234, 123, "payload"); journal::AppendBuffers append_buffers; append_buffers = {append_buffer1}; ASSERT_FALSE(object->append(append_buffers)); - journal::AppendBuffer append_buffer2 = create_append_buffer(metadata, - 234, 124, + journal::AppendBuffer append_buffer2 = create_append_buffer(234, 124, "payload"); append_buffers = {append_buffer2}; ASSERT_FALSE(object->append(append_buffers)); @@ -218,15 +209,13 @@ journal::ObjectRecorderPtr object = create_object(oid, 12); std::string payload(2048, '1'); - journal::AppendBuffer append_buffer1 = create_append_buffer(metadata, - 234, 123, + journal::AppendBuffer append_buffer1 = create_append_buffer(234, 123, payload); journal::AppendBuffers append_buffers; append_buffers = {append_buffer1}; ASSERT_FALSE(object->append(append_buffers)); - journal::AppendBuffer append_buffer2 = create_append_buffer(metadata, - 234, 124, + journal::AppendBuffer append_buffer2 = create_append_buffer(234, 124, payload); append_buffers = {append_buffer2}; ASSERT_TRUE(object->append(append_buffers)); @@ -246,8 +235,7 @@ journal::ObjectRecorderPtr object = create_object(oid, 24); - journal::AppendBuffer append_buffer1 = create_append_buffer(metadata, - 234, 123, + journal::AppendBuffer append_buffer1 = create_append_buffer(234, 123, "payload"); journal::AppendBuffers append_buffers; append_buffers = {append_buffer1}; @@ -273,8 +261,7 @@ journal::ObjectRecorderPtr object = create_object(oid, 24); - journal::AppendBuffer append_buffer = create_append_buffer(metadata, - 234, 123, + journal::AppendBuffer append_buffer = create_append_buffer(234, 123, "payload"); journal::AppendBuffers append_buffers; append_buffers = {append_buffer}; @@ -298,8 +285,7 @@ journal::ObjectRecorderPtr object = create_object(oid, 24); - journal::AppendBuffer append_buffer = create_append_buffer(metadata, - 234, 123, + journal::AppendBuffer append_buffer = create_append_buffer(234, 123, "payload"); journal::AppendBuffers append_buffers; @@ -326,11 +312,9 @@ journal::ObjectRecorderPtr object2 = create_object(oid, 12); std::string payload(2048, '1'); - journal::AppendBuffer append_buffer1 = create_append_buffer(metadata, - 234, 123, + journal::AppendBuffer append_buffer1 = create_append_buffer(234, 123, payload); - journal::AppendBuffer append_buffer2 = create_append_buffer(metadata, - 234, 124, + journal::AppendBuffer append_buffer2 = create_append_buffer(234, 124, payload); journal::AppendBuffers append_buffers; append_buffers = {append_buffer1, append_buffer2}; @@ -341,8 +325,7 @@ ASSERT_EQ(0, cond.wait()); ASSERT_EQ(0U, object1->get_pending_appends()); - journal::AppendBuffer append_buffer3 = create_append_buffer(metadata, - 456, 123, + journal::AppendBuffer append_buffer3 = create_append_buffer(456, 123, payload); append_buffers = {append_buffer3}; diff -Nru ceph-10.1.1/src/test/librados/misc.cc ceph-10.1.2/src/test/librados/misc.cc --- ceph-10.1.1/src/test/librados/misc.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/librados/misc.cc 2016-04-12 17:46:06.000000000 +0000 @@ -95,6 +95,88 @@ ASSERT_EQ(-ENAMETOOLONG, ioctx.write(string(maxlen*2, 'a').c_str(), bl, bl.length(), 0)); } +TEST_F(LibRadosMiscPP, LongLocatorPP) { + bufferlist bl; + bl.append("content"); + int maxlen = g_conf->osd_max_object_name_len; + ioctx.locator_set_key( + string((maxlen/2), 'a')); + ASSERT_EQ( + 0, + ioctx.write( + string("a").c_str(), + bl, bl.length(), 0)); + ioctx.locator_set_key( + string(maxlen - 1, 'a')); + ASSERT_EQ( + 0, + ioctx.write( + string("a").c_str(), + bl, bl.length(), 0)); + ioctx.locator_set_key( + string(maxlen, 'a')); + ASSERT_EQ( + 0, + ioctx.write( + string("a").c_str(), + bl, bl.length(), 0)); + ioctx.locator_set_key( + string(maxlen+1, 'a')); + ASSERT_EQ( + -ENAMETOOLONG, + ioctx.write( + string("a").c_str(), + bl, bl.length(), 0)); + ioctx.locator_set_key( + string((maxlen*2), 'a')); + ASSERT_EQ( + -ENAMETOOLONG, + ioctx.write( + string("a").c_str(), + bl, bl.length(), 0)); +} + +TEST_F(LibRadosMiscPP, LongNSpacePP) { + bufferlist bl; + bl.append("content"); + int maxlen = g_conf->osd_max_object_namespace_len; + ioctx.set_namespace( + string((maxlen/2), 'a')); + ASSERT_EQ( + 0, + ioctx.write( + string("a").c_str(), + bl, bl.length(), 0)); + ioctx.set_namespace( + string(maxlen - 1, 'a')); + ASSERT_EQ( + 0, + ioctx.write( + string("a").c_str(), + bl, bl.length(), 0)); + ioctx.set_namespace( + string(maxlen, 'a')); + ASSERT_EQ( + 0, + ioctx.write( + string("a").c_str(), + bl, bl.length(), 0)); + ioctx.set_namespace( + string(maxlen+1, 'a')); + ASSERT_EQ( + -ENAMETOOLONG, + ioctx.write( + string("a").c_str(), + bl, bl.length(), 0)); + ioctx.set_namespace( + string((maxlen*2), 'a')); + ASSERT_EQ( + -ENAMETOOLONG, + ioctx.write( + string("a").c_str(), + bl, bl.length(), 0)); +} + TEST_F(LibRadosMiscPP, LongAttrNamePP) { bufferlist bl; bl.append("content"); @@ -734,15 +816,29 @@ //copy_from between ecpool and no-ecpool. TEST_F(LibRadosTwoPoolsECPP, CopyFrom) { - //create object w/ omapheader + bufferlist z; + z.append_zero(4194304*2); bufferlist b; b.append("copyfrom"); - ASSERT_EQ(0, src_ioctx.omap_set_header("foo", b)); - version_t uv = src_ioctx.get_last_version(); - ObjectWriteOperation op; - op.copy_from("foo", src_ioctx, uv); - ASSERT_EQ(-EOPNOTSUPP, ioctx.operate("foo.copy", &op)); + // create big object w/ omapheader + { + ASSERT_EQ(0, src_ioctx.write_full("foo", z)); + ASSERT_EQ(0, src_ioctx.omap_set_header("foo", b)); + version_t uv = src_ioctx.get_last_version(); + ObjectWriteOperation op; + op.copy_from("foo", src_ioctx, uv); + ASSERT_EQ(-EOPNOTSUPP, ioctx.operate("foo.copy", &op)); + } + + // same with small object + { + ASSERT_EQ(0, src_ioctx.omap_set_header("bar", b)); + version_t uv = src_ioctx.get_last_version(); + ObjectWriteOperation op; + op.copy_from("bar", src_ioctx, uv); + ASSERT_EQ(-EOPNOTSUPP, ioctx.operate("bar.copy", &op)); + } } TEST_F(LibRadosMiscPP, CopyScrubPP) { diff -Nru ceph-10.1.1/src/test/librbd/test_mock_Journal.cc ceph-10.1.2/src/test/librbd/test_mock_Journal.cc --- ceph-10.1.1/src/test/librbd/test_mock_Journal.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/librbd/test_mock_Journal.cc 2016-04-12 17:46:06.000000000 +0000 @@ -944,6 +944,7 @@ // commit journal event followed by IO event (standard) on_journal_safe1->complete(0); + ictx->op_work_queue->drain(); expect_future_committed(mock_journaler); mock_journal.commit_io_event(1U, 0); @@ -954,6 +955,7 @@ C_SaferCond event_ctx; mock_journal.wait_event(2U, &event_ctx); on_journal_safe2->complete(0); + ictx->op_work_queue->drain(); ASSERT_EQ(0, event_ctx.wait()); } @@ -1054,6 +1056,7 @@ // failed IO remains uncommitted in journal on_journal_safe->complete(0); + ictx->op_work_queue->drain(); mock_journal.commit_io_event(1U, -EINVAL); } diff -Nru ceph-10.1.1/src/test/librgw_file_nfsns.cc ceph-10.1.2/src/test/librgw_file_nfsns.cc --- ceph-10.1.1/src/test/librgw_file_nfsns.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/librgw_file_nfsns.cc 2016-04-12 17:46:06.000000000 +0000 @@ -331,6 +331,14 @@ ASSERT_EQ(rc, 0); sf.sync(); ASSERT_TRUE(sf.rgw_fh->is_file()); + + /* because we made it the hard way, fixup attributes */ + struct stat st; + st.st_uid = owner_uid; + st.st_gid = owner_gid; + st.st_mode = 644; + sf.rgw_fh->create_stat(&st, create_mask); + /* open handle */ rc = rgw_open(fs, sf.fh, 0 /* flags */); ASSERT_EQ(rc, 0); @@ -523,6 +531,9 @@ ASSERT_TRUE(sobj.rgw_fh->is_dir()); ASSERT_TRUE(S_ISDIR(st.st_mode)); } + /* validate Unix owners */ + ASSERT_EQ(st.st_uid, owner_uid); + ASSERT_EQ(st.st_gid, owner_gid); if (verbose) { obj_rec_st rec_st{sobj, st}; std::cout << "\t" @@ -557,6 +568,10 @@ ASSERT_EQ(sobj.rgw_fh->get_size(), 16UL); // do it memset(buf, 0, 256); + if (verbose) { + std::cout << "reading 0,256 " << sobj.rgw_fh->relative_object_name() + << std::endl; + } rc = rgw_read(fs, sobj.fh, 0, 256, &nread, buf, RGW_READ_FLAG_NONE); ASSERT_EQ(rc, 0); if (verbose) { diff -Nru ceph-10.1.1/src/test/objectstore/chain_xattr.cc ceph-10.1.2/src/test/objectstore/chain_xattr.cc --- ceph-10.1.1/src/test/objectstore/chain_xattr.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/objectstore/chain_xattr.cc 2016-04-12 17:46:06.000000000 +0000 @@ -259,6 +259,120 @@ ::unlink(file); } +list get_xattrs(int fd) +{ + char _buf[1024]; + char *buf = _buf; + int len = sys_flistxattr(fd, _buf, sizeof(_buf)); + if (len < 0) + return list(); + list ret; + while (len > 0) { + size_t next_len = strlen(buf); + ret.push_back(string(buf, buf + next_len)); + assert(len >= (int)(next_len + 1)); + buf += (next_len + 1); + len -= (next_len + 1); + } + return ret; +} + +list get_xattrs(string fn) +{ + int fd = ::open(fn.c_str(), O_RDONLY); + if (fd < 0) + return list(); + auto ret = get_xattrs(fd); + ::close(fd); + return ret; +} + +TEST(chain_xattr, fskip_chain_cleanup_and_ensure_single_attr) +{ + const char *name = "user.foo"; + const char *file = FILENAME; + ::unlink(file); + int fd = ::open(file, O_CREAT|O_RDWR|O_TRUNC, 0700); + + char buf[800]; + memset(buf, sizeof(buf), 0x1F); + // set chunked without either + { + int r = chain_fsetxattr(fd, name, buf, sizeof(buf)); + ASSERT_EQ(r, sizeof(buf)); + ASSERT_GT(get_xattrs(fd).size(), 1); + } + + // verify + { + char buf2[sizeof(buf)*2]; + int r = chain_fgetxattr(fd, name, buf2, sizeof(buf2)); + ASSERT_EQ(r, sizeof(buf)); + ASSERT_EQ(memcmp(buf, buf2, sizeof(buf)), 0); + } + + // overwrite + { + int r = chain_fsetxattr(fd, name, buf, sizeof(buf)); + ASSERT_EQ(r, sizeof (buf)); + ASSERT_EQ(get_xattrs(fd).size(), 1); + } + + // verify + { + char buf2[sizeof(buf)*2]; + int r = chain_fgetxattr(fd, name, buf2, sizeof(buf2)); + ASSERT_EQ(r, sizeof(buf)); + ASSERT_EQ(memcmp(buf, buf2, sizeof(buf)), 0); + } + + ::close(fd); + ::unlink(file); +} + +TEST(chain_xattr, skip_chain_cleanup_and_ensure_single_attr) +{ + const char *name = "user.foo"; + const char *file = FILENAME; + ::unlink(file); + int fd = ::open(file, O_CREAT|O_RDWR|O_TRUNC, 0700); + ::close(fd); + + char buf[3000]; + memset(buf, sizeof(buf), 0x1F); + // set chunked without either + { + int r = chain_setxattr(file, name, buf, sizeof(buf)); + ASSERT_EQ(r, sizeof(buf)); + ASSERT_GT(get_xattrs(file).size(), 1); + } + + // verify + { + char buf2[sizeof(buf)*2]; + int r = chain_getxattr(file, name, buf2, sizeof(buf2)); + ASSERT_EQ(r, sizeof(buf)); + ASSERT_EQ(memcmp(buf, buf2, sizeof(buf)), 0); + } + + // overwrite + { + int r = chain_setxattr(file, name, buf, sizeof(buf)); + ASSERT_EQ(r, sizeof (buf)); + ASSERT_EQ(get_xattrs(file).size(), 1); + } + + // verify + { + char buf2[sizeof(buf)*2]; + int r = chain_getxattr(file, name, buf2, sizeof(buf2)); + ASSERT_EQ(r, sizeof(buf)); + ASSERT_EQ(memcmp(buf, buf2, sizeof(buf)), 0); + } + + ::unlink(file); +} + int main(int argc, char **argv) { vector args; argv_to_vec(argc, (const char **)argv, args); diff -Nru ceph-10.1.1/src/test/objectstore/test_bluefs.cc ceph-10.1.2/src/test/objectstore/test_bluefs.cc --- ceph-10.1.1/src/test/objectstore/test_bluefs.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/objectstore/test_bluefs.cc 2016-04-12 17:46:06.000000000 +0000 @@ -38,8 +38,8 @@ string fn = get_temp_bdev(size); uuid_d fsid; BlueFS fs; - fs.add_block_device(0, fn); - fs.add_block_extent(0, 1048576, size - 1048576); + fs.add_block_device(BlueFS::BDEV_DB, fn); + fs.add_block_extent(BlueFS::BDEV_DB, 1048576, size - 1048576); fs.mkfs(fsid); rm_temp_bdev(fn); } @@ -48,13 +48,13 @@ uint64_t size = 1048476 * 128; string fn = get_temp_bdev(size); BlueFS fs; - ASSERT_EQ(0, fs.add_block_device(0, fn)); - fs.add_block_extent(0, 1048576, size - 1048576); + ASSERT_EQ(0, fs.add_block_device(BlueFS::BDEV_DB, fn)); + fs.add_block_extent(BlueFS::BDEV_DB, 1048576, size - 1048576); uuid_d fsid; ASSERT_EQ(0, fs.mkfs(fsid)); ASSERT_EQ(0, fs.mount()); - ASSERT_EQ(fs.get_total(0), size - 1048576); - ASSERT_LT(fs.get_free(0), size - 1048576); + ASSERT_EQ(fs.get_total(BlueFS::BDEV_DB), size - 1048576); + ASSERT_LT(fs.get_free(BlueFS::BDEV_DB), size - 1048576); fs.umount(); rm_temp_bdev(fn); } @@ -63,8 +63,8 @@ uint64_t size = 1048476 * 128; string fn = get_temp_bdev(size); BlueFS fs; - ASSERT_EQ(0, fs.add_block_device(0, fn)); - fs.add_block_extent(0, 1048576, size - 1048576); + ASSERT_EQ(0, fs.add_block_device(BlueFS::BDEV_DB, fn)); + fs.add_block_extent(BlueFS::BDEV_DB, 1048576, size - 1048576); uuid_d fsid; ASSERT_EQ(0, fs.mkfs(fsid)); ASSERT_EQ(0, fs.mount()); @@ -99,8 +99,8 @@ uint64_t size = 1048476 * 128; string fn = get_temp_bdev(size); BlueFS fs; - ASSERT_EQ(0, fs.add_block_device(0, fn)); - fs.add_block_extent(0, 1048576, size - 1048576); + ASSERT_EQ(0, fs.add_block_device(BlueFS::BDEV_DB, fn)); + fs.add_block_extent(BlueFS::BDEV_DB, 1048576, size - 1048576); uuid_d fsid; ASSERT_EQ(0, fs.mkfs(fsid)); ASSERT_EQ(0, fs.mount()); diff -Nru ceph-10.1.1/src/test/os/TestLFNIndex.cc ceph-10.1.2/src/test/os/TestLFNIndex.cc --- ceph-10.1.1/src/test/os/TestLFNIndex.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/os/TestLFNIndex.cc 2016-04-12 17:46:06.000000000 +0000 @@ -49,7 +49,7 @@ const std::string mangled_name = lfn_generate_object_name(hoid); EXPECT_EQ(mangled_expected, mangled_name); ghobject_t hoid_parsed; - EXPECT_TRUE(lfn_parse_object_name(mangled_name, &hoid_parsed)); + EXPECT_EQ(0, lfn_parse_object_name(mangled_name, &hoid_parsed)); EXPECT_EQ(hoid, hoid_parsed); } diff -Nru ceph-10.1.1/src/test/pybind/test_ceph_argparse.py ceph-10.1.2/src/test/pybind/test_ceph_argparse.py --- ceph-10.1.1/src/test/pybind/test_ceph_argparse.py 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/pybind/test_ceph_argparse.py 2016-04-12 17:46:06.000000000 +0000 @@ -491,6 +491,11 @@ self.assert_valid_command(['fs', 'ls']) assert_equal({}, validate_command(sigdict, ['fs', 'ls', 'toomany'])) + def test_fs_set_default(self): + self.assert_valid_command(['fs', 'set_default', 'cephfs']) + assert_equal({}, validate_command(sigdict, ['fs', 'set_default'])) + assert_equal({}, validate_command(sigdict, ['fs', 'set_default', 'cephfs', 'toomany'])) + class TestMon(TestArgparse): def test_dump(self): diff -Nru ceph-10.1.1/src/test/rbd_mirror/test_ClusterWatcher.cc ceph-10.1.2/src/test/rbd_mirror/test_ClusterWatcher.cc --- ceph-10.1.1/src/test/rbd_mirror/test_ClusterWatcher.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/rbd_mirror/test_ClusterWatcher.cc 2016-04-12 17:46:06.000000000 +0000 @@ -45,7 +45,8 @@ void create_pool(bool enable_mirroring, const peer_t &peer, string *uuid = nullptr, string *name=nullptr) { string pool_name = get_temp_pool_name("test-rbd-mirror-"); - ASSERT_EQ("", create_one_pool_pp(pool_name, *m_cluster)); + ASSERT_EQ(0, m_cluster->pool_create(pool_name.c_str())); + int64_t pool_id = m_cluster->pool_lookup(pool_name.c_str()); ASSERT_GE(pool_id, 0); m_pools.insert(pool_name); @@ -84,7 +85,8 @@ void create_cache_pool(const string &base_pool, string *cache_pool_name) { bufferlist inbl; *cache_pool_name = get_temp_pool_name("test-rbd-mirror-"); - ASSERT_EQ("", create_one_pool_pp(*cache_pool_name, *m_cluster)); + ASSERT_EQ(0, m_cluster->pool_create(cache_pool_name->c_str())); + ASSERT_EQ(0, m_cluster->mon_command( "{\"prefix\": \"osd tier add\", \"pool\": \"" + base_pool + "\", \"tierpool\": \"" + *cache_pool_name + diff -Nru ceph-10.1.1/src/test/rbd_mirror/test_fixture.cc ceph-10.1.2/src/test/rbd_mirror/test_fixture.cc --- ceph-10.1.1/src/test/rbd_mirror/test_fixture.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/rbd_mirror/test_fixture.cc 2016-04-12 17:46:06.000000000 +0000 @@ -22,11 +22,13 @@ } void TestFixture::SetUpTestCase() { + ASSERT_EQ("", connect_cluster_pp(_rados)); + _local_pool_name = get_temp_pool_name("test-rbd-mirror-"); - ASSERT_EQ("", create_one_pool_pp(_local_pool_name, _rados)); + ASSERT_EQ(0, _rados.pool_create(_local_pool_name.c_str())); _remote_pool_name = get_temp_pool_name("test-rbd-mirror-"); - ASSERT_EQ("", create_one_pool_pp(_remote_pool_name, _rados)); + ASSERT_EQ(0, _rados.pool_create(_remote_pool_name.c_str())); } void TestFixture::TearDownTestCase() { diff -Nru ceph-10.1.1/src/test/rbd_mirror/test_ImageReplayer.cc ceph-10.1.2/src/test/rbd_mirror/test_ImageReplayer.cc --- ceph-10.1.1/src/test/rbd_mirror/test_ImageReplayer.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/rbd_mirror/test_ImageReplayer.cc 2016-04-12 17:46:06.000000000 +0000 @@ -76,14 +76,14 @@ EXPECT_EQ("", connect_cluster_pp(m_local_cluster)); m_local_pool_name = get_temp_pool_name(); - EXPECT_EQ("", create_one_pool_pp(m_local_pool_name, m_local_cluster)); + EXPECT_EQ(0, m_local_cluster.pool_create(m_local_pool_name.c_str())); EXPECT_EQ(0, m_local_cluster.ioctx_create(m_local_pool_name.c_str(), m_local_ioctx)); EXPECT_EQ("", connect_cluster_pp(m_remote_cluster)); m_remote_pool_name = get_temp_pool_name(); - EXPECT_EQ("", create_one_pool_pp(m_remote_pool_name, m_remote_cluster)); + EXPECT_EQ(0, m_remote_cluster.pool_create(m_remote_pool_name.c_str())); m_remote_pool_id = m_remote_cluster.pool_lookup(m_remote_pool_name.c_str()); EXPECT_GE(m_remote_pool_id, 0); @@ -104,6 +104,13 @@ ~TestImageReplayer() { + if (m_watch_handle != 0) { + m_remote_ioctx.unwatch2(m_watch_handle); + delete m_watch_ctx; + m_watch_ctx = nullptr; + m_watch_handle = 0; + } + delete m_replayer; delete m_threads; diff -Nru ceph-10.1.1/src/test/rbd_mirror/test_ImageSync.cc ceph-10.1.2/src/test/rbd_mirror/test_ImageSync.cc --- ceph-10.1.1/src/test/rbd_mirror/test_ImageSync.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/rbd_mirror/test_ImageSync.cc 2016-04-12 17:46:06.000000000 +0000 @@ -58,6 +58,11 @@ ASSERT_EQ(0, m_remote_journaler->register_client(client_data_bl)); } + virtual void TearDown() { + TestFixture::TearDown(); + delete m_remote_journaler; + } + void create_and_open(librados::IoCtx &io_ctx, librbd::ImageCtx **image_ctx) { librbd::RBD rbd; ASSERT_EQ(0, create_image(rbd, io_ctx, m_image_name, m_image_size)); diff -Nru ceph-10.1.1/src/test/rbd_mirror/test_PoolWatcher.cc ceph-10.1.2/src/test/rbd_mirror/test_PoolWatcher.cc --- ceph-10.1.1/src/test/rbd_mirror/test_PoolWatcher.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/test/rbd_mirror/test_PoolWatcher.cc 2016-04-12 17:46:06.000000000 +0000 @@ -54,7 +54,8 @@ void create_pool(bool enable_mirroring, const peer_t &peer, string *name=nullptr) { string pool_name = get_temp_pool_name("test-rbd-mirror-"); - ASSERT_EQ("", create_one_pool_pp(pool_name, *m_cluster)); + ASSERT_EQ(0, m_cluster->pool_create(pool_name.c_str())); + int64_t pool_id = m_cluster->pool_lookup(pool_name.c_str()); ASSERT_GE(pool_id, 0); m_pools.insert(pool_name); @@ -83,7 +84,8 @@ void create_cache_pool(const string &base_pool, string *cache_pool_name) { bufferlist inbl; *cache_pool_name = get_temp_pool_name("test-rbd-mirror-"); - ASSERT_EQ("", create_one_pool_pp(*cache_pool_name, *m_cluster)); + ASSERT_EQ(0, m_cluster->pool_create(cache_pool_name->c_str())); + ASSERT_EQ(0, m_cluster->mon_command( "{\"prefix\": \"osd tier add\", \"pool\": \"" + base_pool + "\", \"tierpool\": \"" + *cache_pool_name + diff -Nru ceph-10.1.1/src/tools/ceph_authtool.cc ceph-10.1.2/src/tools/ceph_authtool.cc --- ceph-10.1.1/src/tools/ceph_authtool.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/tools/ceph_authtool.cc 2016-04-12 17:46:06.000000000 +0000 @@ -37,13 +37,14 @@ << " specified entityname\n" << " --gen-print-key will generate a new secret key without set it\n" << " to the keyringfile, prints the secret to stdout\n" - << " --import-keyring will import the content of a given keyring\n" + << " --import-keyring FILE will import the content of a given keyring\n" << " into the keyringfile\n" - << " -u, --set-uid sets the auid (authenticated user id) for the\n" + << " -n NAME, --name NAME specify entityname to operate on\n" + << " -u AUID, --set-uid AUID sets the auid (authenticated user id) for the\n" << " specified entityname\n" - << " -a, --add-key will add an encoded key to the keyring\n" - << " --cap subsystem capability will set the capability for given subsystem\n" - << " --caps capsfile will set all of capabilities associated with a\n" + << " -a BASE64, --add-key BASE64 will add an encoded key to the keyring\n" + << " --cap SUBSYSTEM CAPABILITY will set the capability for given subsystem\n" + << " --caps CAPSFILE will set all of capabilities associated with a\n" << " given key, for all subsystems" << std::endl; exit(1); @@ -73,6 +74,9 @@ bool set_auid = false; std::vector::iterator i; + /* Handle options unique to ceph-authtool + * -n NAME, --name NAME is handled by global_init + * */ for (i = args.begin(); i != args.end(); ) { std::string val; if (ceph_argparse_double_dash(args, i)) { @@ -117,6 +121,7 @@ usage(); } } + if (fn.empty() && !gen_print_key) { cerr << argv[0] << ": must specify filename" << std::endl; usage(); diff -Nru ceph-10.1.1/src/tools/Makefile-client.am ceph-10.1.2/src/tools/Makefile-client.am --- ceph-10.1.1/src/tools/Makefile-client.am 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/tools/Makefile-client.am 2016-04-12 17:46:06.000000000 +0000 @@ -71,6 +71,7 @@ tools/rbd/Utils.h rbd_LDADD = \ libjournal.la libcls_journal_client.la \ + libcls_rbd_client.la libcls_lock_client.la \ $(LIBRBD) $(LIBRBD_TYPES) $(LIBRADOS) $(CEPH_GLOBAL) \ $(BOOST_REGEX_LIBS) $(BOOST_PROGRAM_OPTIONS_LIBS) if LINUX diff -Nru ceph-10.1.1/src/tools/rbd/action/Journal.cc ceph-10.1.2/src/tools/rbd/action/Journal.cc --- ceph-10.1.1/src/tools/rbd/action/Journal.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/tools/rbd/action/Journal.cc 2016-04-12 17:46:06.000000000 +0000 @@ -13,14 +13,13 @@ #include #include #include - +#include "cls/rbd/cls_rbd_client.h" #include "cls/journal/cls_journal_types.h" #include "cls/journal/cls_journal_client.h" #include "journal/Journaler.h" #include "journal/ReplayEntry.h" #include "journal/ReplayHandler.h" -//#include "librbd/Journal.h" // XXXMG: for librbd::Journal::reset() #include "librbd/journal/Types.h" namespace rbd { @@ -134,40 +133,35 @@ static int do_reset_journal(librados::IoCtx& io_ctx, const std::string& journal_id) { - // XXXMG: does not work due to a linking issue - //return librbd::Journal::reset(io_ctx, journal_id); - - ::journal::Journaler journaler(io_ctx, journal_id, "", 5); - - C_SaferCond cond; - journaler.init(&cond); - - int r = cond.wait(); + // disable/re-enable journaling to delete/re-create the journal + // to properly handle mirroring constraints + std::string image_name; + int r = librbd::cls_client::dir_get_name(&io_ctx, RBD_DIRECTORY, journal_id, + &image_name); if (r < 0) { - std::cerr << "failed to initialize journal: " << cpp_strerror(r) - << std::endl; + std::cerr << "failed to locate journal's image: " << cpp_strerror(r) + << std::endl; return r; } - uint8_t order, splay_width; - int64_t pool_id; - journaler.get_metadata(&order, &splay_width, &pool_id); - - r = journaler.remove(true); + librbd::Image image; + r = utils::open_image(io_ctx, image_name, false, &image); if (r < 0) { - std::cerr << "failed to reset journal: " << cpp_strerror(r) << std::endl; + std::cerr << "failed to open image: " << cpp_strerror(r) << std::endl; return r; } - r = journaler.create(order, splay_width, pool_id); + + r = image.update_features(RBD_FEATURE_JOURNALING, false); if (r < 0) { - std::cerr << "failed to create journal: " << cpp_strerror(r) << std::endl; + std::cerr << "failed to disable image journaling: " << cpp_strerror(r) + << std::endl; return r; } - // TODO register with librbd payload - r = journaler.register_client(bufferlist()); + r = image.update_features(RBD_FEATURE_JOURNALING, true); if (r < 0) { - std::cerr << "failed to register client: " << cpp_strerror(r) << std::endl; + std::cerr << "failed to re-enable image journaling: " << cpp_strerror(r) + << std::endl; return r; } return 0; diff -Nru ceph-10.1.1/src/tools/rbd_mirror/Replayer.cc ceph-10.1.2/src/tools/rbd_mirror/Replayer.cc --- ceph-10.1.1/src/tools/rbd_mirror/Replayer.cc 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/tools/rbd_mirror/Replayer.cc 2016-04-12 17:46:06.000000000 +0000 @@ -5,6 +5,9 @@ #include "common/Formatter.h" #include "common/admin_socket.h" +#include "common/ceph_argparse.h" +#include "common/code_environment.h" +#include "common/common_init.h" #include "common/debug.h" #include "common/errno.h" #include "include/stringify.h" @@ -143,37 +146,60 @@ { dout(20) << "replaying for " << m_peer << dendl; - int r = m_remote->init2(m_peer.client_name.c_str(), - m_peer.cluster_name.c_str(), 0); - if (r < 0) { - derr << "error initializing remote cluster handle for " << m_peer - << " : " << cpp_strerror(r) << dendl; - return r; - } + // NOTE: manually bootstrap a CephContext here instead of via + // the librados API to avoid mixing global singletons between + // the librados shared library and the daemon + // TODO: eliminate intermingling of global singletons within Ceph APIs + CephInitParameters iparams(CEPH_ENTITY_TYPE_CLIENT); + if (m_peer.client_name.empty() || + !iparams.name.from_str(m_peer.client_name)) { + derr << "error initializing remote cluster handle for " << m_peer << dendl; + return -EINVAL; + } + + CephContext *cct = common_preinit(iparams, CODE_ENVIRONMENT_LIBRARY, + CINIT_FLAG_UNPRIVILEGED_DAEMON_DEFAULTS); + cct->_conf->cluster = m_peer.cluster_name; - r = m_remote->conf_read_file(nullptr); + // librados::Rados::conf_read_file + int r = cct->_conf->parse_config_files(nullptr, nullptr, 0); if (r < 0) { - derr << "could not read ceph conf for " << m_peer - << " : " << cpp_strerror(r) << dendl; + derr << "could not read ceph conf for " << m_peer << ": " + << cpp_strerror(r) << dendl; + cct->put(); return r; } + cct->_conf->parse_env(); - r = m_remote->conf_parse_env(nullptr); + // librados::Rados::conf_parse_env + std::vector args; + env_to_vec(args, nullptr); + r = cct->_conf->parse_argv(args); if (r < 0) { - derr << "could not parse environment for " << m_peer - << " : " << cpp_strerror(r) << dendl; + derr << "could not parse environment for " << m_peer << ":" + << cpp_strerror(r) << dendl; + cct->put(); return r; } if (!m_args.empty()) { - r = m_remote->conf_parse_argv(m_args.size(), &m_args[0]); + // librados::Rados::conf_parse_argv + r = cct->_conf->parse_argv(m_args); if (r < 0) { - derr << "could not parse command line args for " << m_peer - << " : " << cpp_strerror(r) << dendl; + derr << "could not parse command line args for " << m_peer << ": " + << cpp_strerror(r) << dendl; + cct->put(); return r; } } + cct->_conf->apply_changes(nullptr); + cct->_conf->complain_about_parse_errors(cct); + + r = m_remote->init_with_context(cct); + assert(r == 0); + cct->put(); + r = m_remote->connect(); if (r < 0) { derr << "error connecting to remote cluster " << m_peer diff -Nru ceph-10.1.1/src/vstart.sh ceph-10.1.2/src/vstart.sh --- ceph-10.1.1/src/vstart.sh 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/src/vstart.sh 2016-04-12 17:46:06.000000000 +0000 @@ -625,7 +625,7 @@ if [ "$start_mds" -eq 1 -a "$CEPH_NUM_MDS" -gt 0 ]; then if [ "$CEPH_NUM_FS" -gt "1" ] ; then - $CEPH_ADM fs flag set enable_multiple true + $CEPH_ADM fs flag set enable_multiple true --yes-i-really-mean-it fi fs=0 diff -Nru ceph-10.1.1/systemd/ceph-mds@.service ceph-10.1.2/systemd/ceph-mds@.service --- ceph-10.1.1/systemd/ceph-mds@.service 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/systemd/ceph-mds@.service 2016-04-12 17:46:06.000000000 +0000 @@ -15,6 +15,7 @@ ProtectHome=true ProtectSystem=full PrivateTmp=true +TasksMax=infinity [Install] WantedBy=ceph-mds.target diff -Nru ceph-10.1.1/systemd/ceph-mon@.service ceph-10.1.2/systemd/ceph-mon@.service --- ceph-10.1.1/systemd/ceph-mon@.service 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/systemd/ceph-mon@.service 2016-04-12 17:46:06.000000000 +0000 @@ -21,6 +21,7 @@ ProtectHome=true ProtectSystem=full PrivateTmp=true +TasksMax=infinity [Install] WantedBy=ceph-mon.target diff -Nru ceph-10.1.1/systemd/ceph-osd@.service ceph-10.1.2/systemd/ceph-osd@.service --- ceph-10.1.1/systemd/ceph-osd@.service 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/systemd/ceph-osd@.service 2016-04-12 17:46:06.000000000 +0000 @@ -15,6 +15,7 @@ ProtectHome=true ProtectSystem=full PrivateTmp=true +TasksMax=infinity [Install] WantedBy=ceph-osd.target diff -Nru ceph-10.1.1/systemd/ceph-radosgw@.service ceph-10.1.2/systemd/ceph-radosgw@.service --- ceph-10.1.1/systemd/ceph-radosgw@.service 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/systemd/ceph-radosgw@.service 2016-04-12 17:46:06.000000000 +0000 @@ -14,6 +14,7 @@ ProtectHome=true ProtectSystem=full PrivateTmp=true +TasksMax=infinity [Install] WantedBy=ceph-radosgw.target diff -Nru ceph-10.1.1/systemd/ceph-rbd-mirror@.service ceph-10.1.2/systemd/ceph-rbd-mirror@.service --- ceph-10.1.1/systemd/ceph-rbd-mirror@.service 2016-04-06 14:17:04.000000000 +0000 +++ ceph-10.1.2/systemd/ceph-rbd-mirror@.service 2016-04-12 17:46:06.000000000 +0000 @@ -17,6 +17,7 @@ Restart=on-failure StartLimitInterval=30min StartLimitBurst=3 +TasksMax=infinity [Install] WantedBy=ceph-rbd-mirror.target