- 21 Aug, 2016 40 commits
-
-
Hongchao Zhang authored
When RPC size or the max RPCs in flight is increased, the actual limit might be max_dirty_mb. This patch automatically increases the max_dirty_mb value at connection time and when the related values are tuned manually by proc file system. this patch also changes the unit of "cl_dirty" and "cl_dirty_max" in client_obd from byte to page. Signed-off-by:
Li Xi <lixi@ddn.com> Signed-off-by:
Hongchao Zhang <hongchao.zhang@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4933 Reviewed-on: http://review.whamcloud.com/10446Reviewed-by:
Jinshan Xiong <jinshan.xiong@intel.com> Reviewed-by:
Andreas Dilger <andreas.dilger@intel.com> Signed-off-by:
James Simmons <jsimmons@infradead.org> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Fan Yong authored
When creating a striped directory, the master object saves the slave objects (or shards) as internal sub-directories. The sub-directory's name is composed of ${shard_FID}:${shard_idx}. With the name, we can easily to know what the shard is and where it should be. On the other hand, we need to store some information related with the striped directory, such as magic, hash type, shards count, and so on. That is the LMV EA (header). We do NOT store the FID of each shard in the LMV EA. Instead, when we need the shards' FIDs (such as readdir() on client-side), we can build the entrie LMV EA on the MDT (in RAM) by iterating the sub-directory entries that are contained in the master object of the striped directroy. Above mechanism can simplify the striped directory create operation. For very large striped directory, logging the FIDs array in the LMV EA will be trouble. It also simplify the LFSCK for verifying striped directory, because it reduces the inconsistency sources. Another fixing is about the lmv_master_fid in master LMV EA header, it is redundant information, and may become one of the inconsistency sources. So replace it with two __u64 padding fields. Signed-off-by:Fan Yong <fan.yong@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5223 Reviewed-on: http://review.whamcloud.com/10751Reviewed-by:
Alex Zhuravlev <alexey.zhuravlev@intel.com> Reviewed-by:
Oleg Drokin <oleg.drokin@intel.com> Signed-off-by:
-
Andriy Skulysh authored
Exit condition from UNREGISTERING phase is releasing of both reply and bulk buffers. Call ptlrpc_unregister_bulk() if ptlrpc_unregister_reply() wasn't completed in async mode before switching to UNREGISTERING phase. Signed-off-by:
Andriy Skulysh <Andriy_Skulysh@xyratex.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5259 Xyratex-bug-id: MRP-1960 Reviewed-on: http://review.whamcloud.com/10846Reviewed-by:
Ann Koehler <amk@cray.com> Reviewed-by:
Liang Zhen <liang.zhen@intel.com> Reviewed-by:
-
wang di authored
For unknown hash type, LMV should try all stripes to locate the name entry. But it will only for lookup and unlink, i.e. we can only list and unlink entries under striped dir with unknown hash type. Signed-off-by:
wang di <di.wang@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4921 Reviewed-on: http://review.whamcloud.com/10041Reviewed-by:
John L. Hammond <john.hammond@intel.com> Reviewed-by:
-
Mikhail Pershin authored
The llog context id table cannot be shrunk easily because that will cause index shifting and incompatibility between old client and new server and vice versa. Patch moves llog_ctxt_id table to the lustre_idl.h because this is wire protocol data, these values are added to the wirecheck. Signed-off-by:
Mikhail Pershin <mike.pershin@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5218 Reviewed-on: http://review.whamcloud.com/10758Reviewed-by:
-
Lai Siyao authored
Add OBD_CONNECT_UNLINK_CLOSE flag for interop, once this is supported, client packs file handle in unlink RPC, and MDT will close file before unlink. Signed-off-by:
Lai Siyao <lai.siyao@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4367 Reviewed-on: http://review.whamcloud.com/10426Reviewed-by:
-
Lai Siyao authored
Currently client open-by-fid often packs name into the request, but the name may be invalid, eg. NFS export, and even if it's valid, it may cause inconsistency because this operation is done on this fid, which is globally unique, but name not. Since open-by-fid doesn't pack name, for striped dir we can't know parent stripe fid on client, so we set parent fid the same as child fid, and MDT has to find its parent fid from linkea (this is already supported by MDT). M_CHECK_STALE becomes obsolete. Unset MDS_OPEN_FL_INTERNAL from open syscall flags, because these flags are internally used, and should not be set from user space. It's not necessary to store parent fid in lli_pfid, because MDT can get it's parent fid from linkea, and now that DNE stripe directory stores master inode fid in lli_pfid, stop storing parent fid to avoid conflict. Signed-off-by:
Nathaniel Clark <nathaniel.l.clark@intel.com> Reviewed-by:
-
Brian Behlendorf authored
As the LU cache grows it can consume large enough chunks of memory that ends up preventing buffers for other objects, such as the OIs, from being cached and severely impacting the performance for FID lookups. Limit the lu_object cache to a maximum of lu_cache_nr objects. NOTES: * In order to be able to quickly determine the number of objects in the hash table the CFS_HASH_COUNTER flag is added. This adds an atomic_inc/dec to the hash insert/remove paths but is not expected to have any measurable impact of performance. Signed-off-by:
Brian Behlendorf <behlendorf1@llnl.gov> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5164 Reviewed-on: http://review.whamcloud.com/10237Reviewed-by:
Isaac Huang <he.huang@intel.com> Reviewed-by:
-
James Simmons authored
One of the versions of gcc I have refuses to build obd_mount.c due to index not be initialized in function lmd_make_exclusion before it is used. Signed-off-by:
James Simmons <uja.ornl@gmail.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4629 Reviewed-on: http://review.whamcloud.com/10705Reviewed-by:
Bob Glossman <bob.glossman@intel.com> Reviewed-by:
Dmitry Eremin <dmitry.eremin@intel.com> Reviewed-by:
-
Emoly Liu authored
ldlm_lock_create() and ldlm_resource_get() always return NULL as error reporting and "NULL" is interpretted as ENOMEM incorrectly sometimes. This patch fixes this problem by using ERR_PTR() rather than NULL. Signed-off-by:
Emoly Liu <emoly.liu@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4524 Reviewed-on: http://review.whamcloud.com/9004Reviewed-by:
Bobi Jam <bobijam@gmail.com> Reviewed-by:
-
Patrick Farrell authored
Add subsystem description from Di Wang to header file. Signed-off-by:
Patrick Farrell <paf@cray.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5153 Reviewed-on: http://review.whamcloud.com/10631Reviewed-by:
-
wang di authored
Also use D_INFO no matter what the error returned from mdc_put_rpc_lock. Signed-off-by:
-
Jinshan Xiong authored
A few changes are made in this patch for unstable pages tracking: 1. Remove kernel NFS unstable pages tracking because it killed performance 2. Track unstable pages as part of LRU cache. Otherwise Lustre can use much more memory than max_cached_mb 3. Remove obd_unstable_pages tracking to avoid using global atomic counter 4. Make unstable pages track optional. Tracking unstable pages is turned off by default, and can be controlled by llite.*.unstable_stats. Signed-off-by:
-
Jinshan Xiong authored
A page in clio used to occupy 584 bytes, which will use size-1024 slab cache. This patch reduces the per-page overhead to 512 bytes so it can use size-512 instead. Signed-off-by:
-
John L. Hammond authored
Rename each member of struct mdt_body, adding the prefix mbo_. Signed-off-by:
-
Hongchao Zhang authored
the LOV xattr of directory could be either lov_user_md_v1 (size is 32) or lov_user_md_v3 (size is 48), then the actual size of the LOV xattr should be return. Signed-off-by:
jacques-Charles Lafoucriere <jacques-charles.lafoucriere@cea.fr> Reviewed-by:
-
Niu Yawei authored
Simple variable rename. Signed-off-by:
Niu Yawei <yawei.niu@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4345 Reviewed-on: http://review.whamcloud.com/10223Reviewed-by:
-
wang di authored
Create striped directory with correct mode, which should be handling same as mkdir. Signed-off-by:
-
wang di authored
With DNE every object can have two locks in different namespaces: lookup lock in space of MDT storing direntry and update/open lock in space of MDT storing inode. In lmv_find_cbdata/lmv_lock_lock, it should try the MDT that the FID maps to first, since this can be easily found, and only try others if that fails. In the error handler of lmv_add_targets, it should check whether ld_tgt_count is being increased before ld_tgt_count is being -1. Signed-off-by:
-
Fan Yong authored
When the layout LFSCK repairs orphan OST-object, if the parent MDT-object was lost, then it will re-create the MDT-object and regenerate the LOV EA and fill the target LOV EA slot with the orphan information, and fill other slots with zero (LOV hole); if related LOV EA slot is invalid or hole, then it will refill the target LOV EA slot; if the target slot exceeds current LOV EA tail, then extend the LOV EA, and fill the gaps as zero. Some of the LOV EA holes may cannot be re-filled finally because of lost some OST-objects. And even if they can be re-filled, but there are still some possible race accessings from client before the re-filling. If the client access the LOV EA with hole(s), it may cause some strange behaviour, such as trigger LBUG()/LASSERT() on the client. So we will make the client to be aware of the LOV EA is incomplete. We introduce a new LOV EA pattern flag LOV_PATTERN_F_HOLE for that: any time when the LFSCK repairs the LOV EA with hole(s), the LOV EA will be marked as LOV_PATTERN_F_HOLE; when all the holes in the LOV EA are refilled, the LOV_PATTERN_F_HOLE will be dropped. For a new client, it recongizes the pattern flag LOV_PATTERN_F_HOLE, then it can permit/forbid some opertions on the file with LOV holes: 1) Normal read/write the file with LOV EA hole is permitted, but the application will get EIO error when read data from the dummy slot or write data to the dummy slot. 2) The users can dump the recovered data via some common read tools, such as "dd conv=sync,noerror". 3) Append data to the file which has LOV EA hole will get EIO failure. 4) Other operations will skip the LOV EA hole(s), and will not get failures, such as {s,g}etattr, {s,g}getxattr, stat, chown/chgrp, chmod, touch, unlink, and so on. For an old client, since it will not recognize the new pattern flag LOV_PATTERN_F_HOLE. So the LOV EA with hole will be dicarded with failure, but it will not cause the client to be crashed. Signed-off-by: -
wang di authored
In ll_lookup_it_finish, we need use the real parent(stripe) FID to validate the parent UPDATE lock. Signed-off-by:
-
wang di authored
Normally we know the value of op_mea1 when ll_readdir is called. In the case of '.' or '..' op_mea1 is unknown so for that case fetch the real parents FID. Signed-off-by:
Li Xi <pkuelelixi@gmail.com> Reviewed-by:
-
John L. Hammond authored
Move the definition of OBD_OCD_VERSION() and similar macros from lustre_idl.h to lustre_ver.h. These macros are primarily used in comparisons to LUSTRE_VERSION_CODE which is defined in lustre_ver.h and so should be defined there as well. Move a few definitions (related to FIDs, quota and striping) from lustre_idl.h to lustre_user.h. Signed-off-by:
-
Jinshan Xiong authored
To simplify the code, change the metrics from bytes to pages. Signed-off-by:
-
Jinshan Xiong authored
If the page is already dirtied in ll_write_end() and kernel tries to call balance_dirty_pages() to write back dirty pages in the same thread, this is deadlock case if the page is already held by clio. This can also fix the issue of LU-4873. Signed-off-by:
-
Christopher J. Morrone authored
It is not permitted in C++ to have a static declaration inside of an anonymous union. The g++ compiler will complaine with an error like this: error: struct ost_id::<anonymous union>::ostid invalid; an anonymous union can only have non-static data members [-fpermissive] This patch changes the code to use an unnamed struct in place of "struct ostid" inside of the anonymous union. That name declaration was completely unnecessary anyway, since it was not used anywhere else. Signed-off-by:
Christopher J. Morrone <morrone2@llnl.gov> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4987 Reviewed-on: http://review.whamcloud.com/10176Reviewed-by:
Robert Read <robert.read@intel.com> Reviewed-by:
-
Gregoire Pichon authored
Root squash exhibits inconsistent behaviour on a client when enabled. If a file is not cached on the client, then root will get a permission denied error when accessing the file. When the file has recently been accessed by a regular user and is still in cache, root will be able to access the file without error because the permission check is only done by the client that isn't aware of root squash. While the only real security benefit from root squash is to deny clients access to files owned by root itself, it also makes sense to treat file access on the client in a consistent manner regardless of whether the file is in cache or not. This patch adds root squash settings to llite so that client is able to apply root squashing when it is relevant. Configuration of MDT root squash settings will automatically be applied to llite config log as well. Update cfs_str2num_check() routine by removing any modification of the specified string parameter. Since string can come from ls_str field of a lstr structure, this avoids inconsistent ls_len field. Signed-off-by:
Gregoire Pichon <gregoire.pichon@bull.net> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-1778 Reviewed-on: http://review.whamcloud.com/5700Reviewed-by:
-
John L. Hammond authored
In ll_prep_md_op_data() validate names according to the same formula used in mdd_name_check(). Add mdc_pack_name() to validate the name actually packed in the request. Signed-off-by:
-
wang di authored
Separate master stripe with master object, so 1. stripeEA only exists on master object. 2. sub-stripe object will be inserted into master object as sub-directory, and it can get the master object by "..". By this, it will remove those specilities for stripe0 in LMV and LOD. And also simplify LFSCK, i.e. consistency check would be easier. When then master object becomes an orphan, we should mark all of its sub-stripes as dead object as well, otherwise client might still be able to create files under these stripes. A few fixes for striped directory layout lock: 1. stripe 0 should be locked as EX, same as other stripes. 2. Acquire the layout for directory, when it is being unliked. Signed-off-by:
-
Chris Horn authored
When determining whether an early reply can be sent the server will calculate the new deadline based on an offset from the request arrival time. However, when actually setting the new deadline the server offsets the current time. This can result in deadlines being extended more than at_max seconds past the request arrival time. Instead, the server should offset the arrival time when updating its request timeout. When a client receives an early reply it doesn't know the server side arrival time so we use the original sent time as an approximation. Signed-off-by:
Chris Horn <hornc@cray.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4578 Reviewed-on: http://review.whamcloud.com/9100Reviewed-by:
Christopher J. Morrone <chris.morrone.llnl@gmail.com> Reviewed-by:
-
wang di authored
Add error handler during inode inialization, so inode will become bad inode if something bad happens during inode prepare phase, otherwise the striped directory will not get its layout and being mis-regarded as normal directory. Signed-off-by:
-
John L. Hammond authored
Move ioctl definitions and related functions from lustre_dlm.h, lustre_lib.h, obd.h, to lustre_ioctl.h. Replace the definitions of retired ioctls with comment. Signed-off-by:
-
Vitaly Fertman authored
Move checks for FAILED, DESTROYED flags under ldlm spinlock, destroy flock atomically with the check it is not destroyed yet. Do not put the granted flock into the resource if this is UNLOCK, TEST, or DEADLOCK'ed flock. Later a regression for this patch was reported under LU-7626. The refcount nonzero (1) after lock cleanup errors was reported. The reason is that the case LCK_NL was not handled for obdecho. Patch 17791 resolved this issue which has been combined into this upstream patch. Signed-off-by:
Vitaly Fertman <vitaly_fertman@xyratex.com> Signed-off-by:
Andriy Skulysh <andriy.skulysh@seagate.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-2177Reviewed-by:
Alexey Lyashkov <alexey_lyashkov@xyratex.com> Reviewed-by:
Mirza Arshad Mirza Hussain <arshad.hussain@seagate.com> Reviewed-by:
Alexey Leonidovich Lyashkov <alexey.lyashkov@seagate.com> Reviewed-on: http://review.whamcloud.com/17791Reviewed-by:
-
Fan Yong authored
Since all the LOV_MAGIC_* definitions have the same postfix values break that value out into its own definition. With this we can check whether the magic's postfix match the LOV_MAGIC_MAGIC or not: if yes, then it is quite possible that the clients has encountered an newer LOV magic. This extra information can let us handle those cases more gracefully. Signed-off-by:
-
Fan Yong authored
Unify the flow control interfaces for MDC RPC and FLD RPC. We allow to adjust the maximum inflight RPCs count via /sys interface. Signed-off-by:
-
John L. Hammond authored
Rename request to preq. Signed-off-by:
-
John L. Hammond authored
In lmv_getattr_name() don't return a freed request in the error path. Signed-off-by:
-
Jinshan Xiong authored
Calculate the end of current stripe correctly when the stripe size is not power 2. Signed-off-by:
-
Jinshan Xiong authored
For a partial page write, it will have to issue a READ RPC firstly to get a full uptodate page. If another page is already locked by this thread it can easily cause deadlock. Signed-off-by:
-
Nathaniel Clark authored
Ensure there aren't invalid pointers hanging around after ptlrpc_req_finished is called. Signed-off-by:
-
