Ganesha with GPFS
by Alok Sinha
If we use Ganesha will GPFS , user api
will be NFS and we will lose most of GPFS functionalities,
like writing into single file parallelly.
Is my understanding correct?
If so, how can we get parallel write performance on a file
with NFS Ganesha?
Is my better option is to give up on Ganesha and use
Luster/Gluster directly? I need open source code.
-alok
4 years, 10 months
Change in ...nfs-ganesha[next]: Provision to get MDCACHE resource utilization
by Sachin Punadikar (GerritHub)
Sachin Punadikar has uploaded this change for review. ( https://review.gerrithub.io/c/ffilz/nfs-ganesha/+/485313 )
Change subject: Provision to get MDCACHE resource utilization
......................................................................
Provision to get MDCACHE resource utilization
It is helpful to know resource utilization at MDCACHE layer such as
chunks in use, entries used by LRU, open FD count etc.
Provision made to fetch this information via ganesha_stats utility.
Currently we have provision to get inode caching related stats via
"inode" option. Extended the same option, "inode", to cover both
inode cache stats and other resource utilization data.
Change-Id: Ic2ad3bc8d852134bdc139958e57230d9bb868bb9
Signed-off-by: Sachin Punadikar <psachin(a)in.ibm.com>
(cherry picked from commit ecf170abe66dc10a2a91a2799319d0a3da50b0fe)
(cherry picked from commit afc95c19242602cba0fa631549967095a9c76568)
---
M src/FSAL/Stackable_FSALs/FSAL_MDCACHE/mdcache_main.c
M src/include/server_stats_private.h
M src/scripts/ganeshactl/Ganesha/glib_dbus_stats.py
M src/scripts/ganeshactl/ganesha_stats.py
M src/support/export_mgr.c
5 files changed, 108 insertions(+), 31 deletions(-)
git pull ssh://review.gerrithub.io:29418/ffilz/nfs-ganesha refs/changes/13/485313/1
--
To view, visit https://review.gerrithub.io/c/ffilz/nfs-ganesha/+/485313
To unsubscribe, or for help writing mail filters, visit https://review.gerrithub.io/settings
Gerrit-Project: ffilz/nfs-ganesha
Gerrit-Branch: next
Gerrit-Change-Id: Ic2ad3bc8d852134bdc139958e57230d9bb868bb9
Gerrit-Change-Number: 485313
Gerrit-PatchSet: 1
Gerrit-Owner: Sachin Punadikar <psachin(a)in.ibm.com>
Gerrit-MessageType: newchange
4 years, 10 months
Change in ...nfs-ganesha[next]: Change to NULL.
by xijinyu (GerritHub)
xijinyu has uploaded this change for review. ( https://review.gerrithub.io/c/ffilz/nfs-ganesha/+/485257 )
Change subject: Change to NULL.
......................................................................
Change to NULL.
While p == NULL, should be NULL instead of non-NULL.
Signed-off-by: Xi Jinyu <xijinyu(a)cmss.chinamobile.com>
Change-Id: I1775a2936b367b40aba6a307368a7974bad1215b
---
M src/include/nfs_init.h
1 file changed, 2 insertions(+), 2 deletions(-)
git pull ssh://review.gerrithub.io:29418/ffilz/nfs-ganesha refs/changes/57/485257/1
--
To view, visit https://review.gerrithub.io/c/ffilz/nfs-ganesha/+/485257
To unsubscribe, or for help writing mail filters, visit https://review.gerrithub.io/settings
Gerrit-Project: ffilz/nfs-ganesha
Gerrit-Branch: next
Gerrit-Change-Id: I1775a2936b367b40aba6a307368a7974bad1215b
Gerrit-Change-Number: 485257
Gerrit-PatchSet: 1
Gerrit-Owner: xijinyu <xijinyu(a)cmss.chinamobile.com>
Gerrit-MessageType: newchange
4 years, 10 months
Re: Default KRB Principal name with new rpc.gssd
by Daniel Gryniewicz
I'm sorry, I'm very far from an expert in GSS/krb5. In particular, I've
never used NFS and gssd together, so I have no personal experience.
Maybe someone else on the list can help?
For questions like this, our default has been "what does knfsd do?" Do
you happen to know if knfsd accepts "shorthostname(a)REALM.COM" as a root
user?
Daniel
On 1/18/20 11:50 PM, Pushpesh Sharma wrote:
>
> Hi All,
>
> Any pointers are appreciated..
>
> -pushpesh
>
> *From:* Pushpesh Sharma <pushpeshs(a)vmware.com>
> *Sent:* Thursday, January 16, 2020 7:45 PM
> *To:* devel(a)lists.nfs-ganesha.org
> *Subject:* [NFS-Ganesha-Devel] Default KRB Principal name with new
> rpc.gssd
>
> Hi All,
>
> We are trying to use ganehsa with KRB. On NFS client centos7.5 we are
> joining Active Directory based using sssd(AD based KRB Realm). For
> root user we are getting krb ticket using kinit and valid AD user.
>
> Client principal while joining domain using sssd a default principal
> of shorthostname$(a)REALM.COM <mailto:shorthostname$@REALM.COM> is
> always generated. gss.rpcd by default send this principal as principal
> user name to ganesha server. Client do have other principal like
> nfs/client_fdqn(a)REALM.COM <mailto:nfs/client_fdqn@REALM.COM>. But as
> per rpc.gssd documentation <mailto:rpc.gssd%20documentation> as well
> first choice would be shorthostname$(a)REALM.COM
> <mailto:shorthostname$@REALM.COM>.
>
> Due to this server always recognize root user as someone else i.e.
> shorthostname$(a)REALM.COM <mailto:shorthostname$@REALM.COM>.
>
> We do see in src/idmapper/idmapper.c
> <https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub....>
> handling for mapping three principal patterns nfs/* , root/*, host/*
> to uid=0, gid=0. So this left shorthostname$(a)REALM.COM
> <mailto:horthostname$@REALM.COM> principal not being recognized as root.
>
> We are of the opinion that doing a small fix in above idmapper code to
> add this pattern as well can fix this issue.
>
> But we wanted to know any security concern around it? Or if the client
> behavior can be change in any way, so we don’t need this fix?
>
> If we try removing this shorthostname$(a)REALM.COM
> <mailto:shorthostname$@REALM.COM>principal after domain join, sssd
> cannot be re-loaded and complains about not finding this principal.
>
> Thanks
>
> -pushpesh
>
>
> _______________________________________________
> Devel mailing list -- devel(a)lists.nfs-ganesha.org
> To unsubscribe send an email to devel-leave(a)lists.nfs-ganesha.org
4 years, 10 months
Crash while trying to acquire the state_hdl lock
by Malahal Naineni
The code path is "open4_ex -> fsal_open2 -> mdcache_open2 ->
_mdcache_kill_entry -> cih_remove_checked -> _mdcache_lru_unref". In this
code path, open4_ex() acquires file_obj->state_hdl->state_lock
but entry->fsobj.hdl.no_cleanup is NOT set. I haven't looked at coredump
for more details, but based on the backtrace, this seems to a possible
issue. state_hdl lock and no_cleanup could co-ordinate, meaning set the
no_cleanup bit with every lock acquisition... Any other approach instead
of fixing everyone of code paths...
Regards, Malahal.
4 years, 10 months
Bug in svc_xprt_lookup() hash implemenation
by Malahal Naineni
It is quite possible that we could have zero refcount entries in the
xprt/rpc_dplx_rec hash table. The code tries to use some atomic flags to
avoid multiple threads freeing the xd/xprt but the code
in svc_xprt_lookup() could end up accessing freed memory. It places a REF
(SVC_REF) under the hash lock, but another process might be trying to
free/destroy the xprt. The code in svc_xprt_lookup() looks for a destroying
flag after releasing the hash lock and by this time there is no guaranty
that the xprt is still valid. The xprt is only valid as long as it is found
in the hash table and we hold the hash lock, correct?
We are able to create a crash consistently, but not sure if this code is
the culprit at this point.
Regards, Malahal.
4 years, 10 months
Announce Push of V4-dev.6
by Frank Filz
Branch next
Tag:V4-dev.6
Merge Highlights
* Don't convert posix acl to fsal acl when no effective acl
* Allow NFSv4 Reads when file is opened with write delegation
* selinux: add local policy for FSAL_VFS and FSAL_RGW
* MDCACHE: 4 fixes for readdir and dirent cache
* "ganesha_stats help" should return 0
* C++ - Don't load rpcent.h in C++ code, it conflicts with system headers
Signed-off-by: Frank S. Filz <ffilzlnx(a)mindspring.com>
Contents:
8e59f79 Frank S. Filz V4-dev.6
ffbf74b sepia-liu There will readdir loop if new_dir_entry in pre_chunk
bc5e0ea sepia-liu Fix all dirents to be skipped if reload_chunk keep work
851243d Daniel Gryniewicz C++ - Don't load rpcent.h in C++ code, it
conflicts with system headers
53ed177 Daniel Gryniewicz MDCACHE - Set next_ck on prev chunk
37dc26f Sachin Punadikar "ganesha_stats help" should return 0
eb98f5b Malahal Naineni Fix lru_reap_chunk_impl racing with
mdcache_lru_unref()
1b04741 Kaleb S. KEITHLEY selinux: add local policy for FSAL_VFS and
FSAL_RGW
35dcdfa Deepthi Shivaramu Allow NFSv4 Reads when file is opened with write
delegation
42338ec Zhu Shangzhong Don't convert posix acl to fsal acl when no effective
acl
4 years, 10 months
