| Message ID | 20190326220730.3763-1-smayhew@redhat.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | restore nfsdcld | expand |
I've really only skimmed the changelogs and documentation and such, but that all looks good to me, and I'm still expecting to commit the kernel-side patches for 5.2, so I'm favor of these nfs-utils patches instead, as long as they look good to Steve. --b. On Tue, Mar 26, 2019 at 06:07:22PM -0400, Scott Mayhew wrote: > When nfsdcld was released, it was quickly deprecated in favor of the > nfsdcltrack usermodehelper, so as to not require another running daemon. > The nfsdcld code was removed from nfs-utils in 2012. That prevents > NFSv4 clients from reclaiming locks from nfsd's running in containers, > since neither nfsdcltrack nor the legacy client tracking code work in > containers. These patches restore the nfsdcld code. > > These patches are intended to go alongside some kernel patches that > introduce an enhancement that allows nfsd to "slurp" up the client > records during client tracking initialization and store them internally > in hash table. This enables nfsd to check whether an NFSv4 client is > allowed to reclaim without having to do an upcall to nfsdcld. It also > allows nfsd to decide to end the v4 grace period early if the number of > RECLAIM_COMPLETE operations it has received from "known" clients is > equal to the number of entries in the hash table. It also allows nfsd > to skip the v4 grace period altogether if it knows there are no clients > allowed to reclaim. > > The new nfsdcld code will work with older kernels, however in that case > there is no ability for nfsd to exit the grace period early or skip the > grace period altogether. > > v3: > - removed PathExists restriction from the nfsdcld.service file > - added the ability to do a one-time upgrade from older client tracking > methods (i.e. nfsdcltrack and/or the legacy v4recovery dir) > v2: > - Use sqlite3_column_int64() instead of sqlite3_column_int() to query > the current & recovery values from the grace table > - Uncommented the nfsdcld section header in nfs.conf > > Scott Mayhew (8): > Revert "nfsdcltrack: remove the nfsdcld daemon" > nfsdcld: move nfsdcld to its own directory > nfsdcld: a few enhancements > nfsdcld: remove some unused functions > nfsdcld: the -p option should specify the rpc_pipefs mountpoint > nfsdcld: add /etc/nfs.conf support > systemd: add a unit file for nfsdcld > nfsdcld: add a facility for migrating from older client tracking > methods > > .gitignore | 1 + > configure.ac | 23 + > nfs.conf | 4 + > support/include/cld.h | 1 + > systemd/nfs-server.service | 2 + > systemd/nfsdcld.service | 10 + > utils/Makefile.am | 4 + > utils/nfsdcld/Makefile.am | 19 + > utils/nfsdcld/cld-internal.h | 33 + > utils/nfsdcld/legacy.c | 180 ++++++ > utils/nfsdcld/legacy.h | 24 + > utils/nfsdcld/nfsdcld.c | 774 ++++++++++++++++++++++++ > utils/nfsdcld/nfsdcld.man | 199 ++++++ > utils/nfsdcld/sqlite.c | 1109 ++++++++++++++++++++++++++++++++++ > utils/nfsdcld/sqlite.h | 35 ++ > 15 files changed, 2418 insertions(+) > create mode 100644 systemd/nfsdcld.service > create mode 100644 utils/nfsdcld/Makefile.am > create mode 100644 utils/nfsdcld/cld-internal.h > create mode 100644 utils/nfsdcld/legacy.c > create mode 100644 utils/nfsdcld/legacy.h > create mode 100644 utils/nfsdcld/nfsdcld.c > create mode 100644 utils/nfsdcld/nfsdcld.man > create mode 100644 utils/nfsdcld/sqlite.c > create mode 100644 utils/nfsdcld/sqlite.h > > -- > 2.17.2
On 4/16/19 12:46 PM, J. Bruce Fields wrote: > I've really only skimmed the changelogs and documentation and such, but > that all looks good to me, and I'm still expecting to commit the > kernel-side patches for 5.2, so I'm favor of these nfs-utils patches > instead, as long as they look good to Steve. They look reasonable to me... Committed!! steved. > > --b. > > On Tue, Mar 26, 2019 at 06:07:22PM -0400, Scott Mayhew wrote: >> When nfsdcld was released, it was quickly deprecated in favor of the >> nfsdcltrack usermodehelper, so as to not require another running daemon. >> The nfsdcld code was removed from nfs-utils in 2012. That prevents >> NFSv4 clients from reclaiming locks from nfsd's running in containers, >> since neither nfsdcltrack nor the legacy client tracking code work in >> containers. These patches restore the nfsdcld code. >> >> These patches are intended to go alongside some kernel patches that >> introduce an enhancement that allows nfsd to "slurp" up the client >> records during client tracking initialization and store them internally >> in hash table. This enables nfsd to check whether an NFSv4 client is >> allowed to reclaim without having to do an upcall to nfsdcld. It also >> allows nfsd to decide to end the v4 grace period early if the number of >> RECLAIM_COMPLETE operations it has received from "known" clients is >> equal to the number of entries in the hash table. It also allows nfsd >> to skip the v4 grace period altogether if it knows there are no clients >> allowed to reclaim. >> >> The new nfsdcld code will work with older kernels, however in that case >> there is no ability for nfsd to exit the grace period early or skip the >> grace period altogether. >> >> v3: >> - removed PathExists restriction from the nfsdcld.service file >> - added the ability to do a one-time upgrade from older client tracking >> methods (i.e. nfsdcltrack and/or the legacy v4recovery dir) >> v2: >> - Use sqlite3_column_int64() instead of sqlite3_column_int() to query >> the current & recovery values from the grace table >> - Uncommented the nfsdcld section header in nfs.conf >> >> Scott Mayhew (8): >> Revert "nfsdcltrack: remove the nfsdcld daemon" >> nfsdcld: move nfsdcld to its own directory >> nfsdcld: a few enhancements >> nfsdcld: remove some unused functions >> nfsdcld: the -p option should specify the rpc_pipefs mountpoint >> nfsdcld: add /etc/nfs.conf support >> systemd: add a unit file for nfsdcld >> nfsdcld: add a facility for migrating from older client tracking >> methods >> >> .gitignore | 1 + >> configure.ac | 23 + >> nfs.conf | 4 + >> support/include/cld.h | 1 + >> systemd/nfs-server.service | 2 + >> systemd/nfsdcld.service | 10 + >> utils/Makefile.am | 4 + >> utils/nfsdcld/Makefile.am | 19 + >> utils/nfsdcld/cld-internal.h | 33 + >> utils/nfsdcld/legacy.c | 180 ++++++ >> utils/nfsdcld/legacy.h | 24 + >> utils/nfsdcld/nfsdcld.c | 774 ++++++++++++++++++++++++ >> utils/nfsdcld/nfsdcld.man | 199 ++++++ >> utils/nfsdcld/sqlite.c | 1109 ++++++++++++++++++++++++++++++++++ >> utils/nfsdcld/sqlite.h | 35 ++ >> 15 files changed, 2418 insertions(+) >> create mode 100644 systemd/nfsdcld.service >> create mode 100644 utils/nfsdcld/Makefile.am >> create mode 100644 utils/nfsdcld/cld-internal.h >> create mode 100644 utils/nfsdcld/legacy.c >> create mode 100644 utils/nfsdcld/legacy.h >> create mode 100644 utils/nfsdcld/nfsdcld.c >> create mode 100644 utils/nfsdcld/nfsdcld.man >> create mode 100644 utils/nfsdcld/sqlite.c >> create mode 100644 utils/nfsdcld/sqlite.h >> >> -- >> 2.17.2
When nfsdcld was released, it was quickly deprecated in favor of the nfsdcltrack usermodehelper, so as to not require another running daemon. The nfsdcld code was removed from nfs-utils in 2012. That prevents NFSv4 clients from reclaiming locks from nfsd's running in containers, since neither nfsdcltrack nor the legacy client tracking code work in containers. These patches restore the nfsdcld code. These patches are intended to go alongside some kernel patches that introduce an enhancement that allows nfsd to "slurp" up the client records during client tracking initialization and store them internally in hash table. This enables nfsd to check whether an NFSv4 client is allowed to reclaim without having to do an upcall to nfsdcld. It also allows nfsd to decide to end the v4 grace period early if the number of RECLAIM_COMPLETE operations it has received from "known" clients is equal to the number of entries in the hash table. It also allows nfsd to skip the v4 grace period altogether if it knows there are no clients allowed to reclaim. The new nfsdcld code will work with older kernels, however in that case there is no ability for nfsd to exit the grace period early or skip the grace period altogether. v3: - removed PathExists restriction from the nfsdcld.service file - added the ability to do a one-time upgrade from older client tracking methods (i.e. nfsdcltrack and/or the legacy v4recovery dir) v2: - Use sqlite3_column_int64() instead of sqlite3_column_int() to query the current & recovery values from the grace table - Uncommented the nfsdcld section header in nfs.conf Scott Mayhew (8): Revert "nfsdcltrack: remove the nfsdcld daemon" nfsdcld: move nfsdcld to its own directory nfsdcld: a few enhancements nfsdcld: remove some unused functions nfsdcld: the -p option should specify the rpc_pipefs mountpoint nfsdcld: add /etc/nfs.conf support systemd: add a unit file for nfsdcld nfsdcld: add a facility for migrating from older client tracking methods .gitignore | 1 + configure.ac | 23 + nfs.conf | 4 + support/include/cld.h | 1 + systemd/nfs-server.service | 2 + systemd/nfsdcld.service | 10 + utils/Makefile.am | 4 + utils/nfsdcld/Makefile.am | 19 + utils/nfsdcld/cld-internal.h | 33 + utils/nfsdcld/legacy.c | 180 ++++++ utils/nfsdcld/legacy.h | 24 + utils/nfsdcld/nfsdcld.c | 774 ++++++++++++++++++++++++ utils/nfsdcld/nfsdcld.man | 199 ++++++ utils/nfsdcld/sqlite.c | 1109 ++++++++++++++++++++++++++++++++++ utils/nfsdcld/sqlite.h | 35 ++ 15 files changed, 2418 insertions(+) create mode 100644 systemd/nfsdcld.service create mode 100644 utils/nfsdcld/Makefile.am create mode 100644 utils/nfsdcld/cld-internal.h create mode 100644 utils/nfsdcld/legacy.c create mode 100644 utils/nfsdcld/legacy.h create mode 100644 utils/nfsdcld/nfsdcld.c create mode 100644 utils/nfsdcld/nfsdcld.man create mode 100644 utils/nfsdcld/sqlite.c create mode 100644 utils/nfsdcld/sqlite.h