- 19 Sep, 2016 40 commits
-
-
Doug Oucharek authored
The previous patch, http://review.whamcloud.com/21304/, removed a check needed until LU-5718 is properly addressed. With the check, LU-5718 results in an error message and a lost RDMA operation. Without it, we have memory corruption and a crash (much harder to debug). Putting the check back in case LU-5718 is not fixed soon. Signed-off-by:
Doug Oucharek <doug.s.oucharek@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-7650 Reviewed-on: http://review.whamcloud.com/22281Reviewed-by:
James Simmons <uja.ornl@yahoo.com> Reviewed-by:
Olaf Weber <olaf@sgi.com> Reviewed-by:
Oleg Drokin <oleg.drokin@intel.com> Signed-off-by:
James Simmons <jsimmons@infradead.org> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Doug Oucharek authored
The code to allow peer_credits to be set per NI was originally "left inactive" because there were concerns about peer_credits interfering with the ability for IB nodes to connect to each other when peer_credits are not the same (peer_credits controls the queue depth for IB). With LU-3322, the values do not have to match so it is now safe to enable this code so peer_credits can be set per NI. This patch enables existing code for setting per NI peer_credits. Second this patch fixes a long standing bug in that the conf data was not being used to set variables in the lnet_ni structure until after lnd_startup() was called which meant LND drivers were ignoring struct lnet_ni tunable values being set. Now we change struct lnet_ni data fields based on conf data before calling lnd_startup(). Signed-off-by:
Dmitry Eremin <dmitry.eremin@intel.com> Reviewed-by:
-
Doug Oucharek authored
In lnet_rtrpools_enable(), a mistake was made and routing was not being turned on when the rtrpools are being allocated for the first time. This patch fixes that routine so we remember to turn on routing after allocating the rtrpools. Signed-off-by:
Amir Shehata <amir.shehata@intel.com> Reviewed-by:
-
Sebastien Buisson authored
Add new 'ni_net_ns' field to struct lnet_ni to hold a reference to original net namespace in which ni is created. In LNetDist(), check if ni was created in same net namespace as current's one. If not, assign order above 0xffff0000, to make this ni not a priority. Signed-off-by:
Sebastien Buisson <sbuisson@ddn.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-7845 Reviewed-on: http://review.whamcloud.com/21884Reviewed-by:
-
Quentin Bouget authored
Fixes potential deadlock in LNetMDAttach Signed-off-by:
Quentin Bouget <quentin.bouget.ocre@cea.fr> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-8249 Reviewed-on: http://review.whamcloud.com/20676Reviewed-by:
Henri Doreau <henri.doreau@cea.fr> Reviewed-by:
-
wang di authored
If the migrating directory is under striped directory, it needs to set right stripe FID for its parent. Signed-off-by:
wang di <di.wang@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-6263 Reviewed-on: http://review.whamcloud.com/13817Reviewed-by:
John L. Hammond <john.hammond@intel.com> Reviewed-by:
Andreas Dilger <andreas.dilger@intel.com> Reviewed-by:
Fan Yong <fan.yong@intel.com> Reviewed-by:
Lai Siyao <lai.siyao@intel.com> Signed-off-by:
-
Ned Bass authored
The client_obd::cl_default_mds_easize field should track the largest observed EA size advertised by the MDT, subject to a reasonable upper bound. The MDC uses cl_default_mds_easize to calculate the initial size of request buffers. The default value should be small enough to avoid wasted memory and excessive use of vmalloc(), yet large enough to accommodate the common use case. In the current code, the default value is only updated if client_obd::cl_max_mds_easize is strictly less than mdt_body::mbo_max_mdsize. This condition is almost never met, because client_obd::cl_max_mds_easize is computed at client mount-time based on the number of OSTs in the filesystem, so the MDT won't ever observe and advertise an EA size larger than that. As a result, client_obd::cl_default_mds_easize indefinitely retains its initial value, which is computed at client mount-time based on the filesystem's default stripe width. Any getattr() requests for widely striped files will consequently allocate a request buffer that is too small, forcing reallocations on both the client and server side. To avoid this, update client_obd::cl_default_mds_easize independently of the value of client_obd::cl_max_mds_easize. In addition, this patch includes these changes: - Add comments to the client_obd structure to clarify what the cl_{default,max}_mds_{cookie,ea}size values mean. - Prevent mdc_get_info() from storing uninitialized data in client_obd::cl_max_mds_cookiesize. - Use 4096 as an upper bound for the default values. The former bound of PAGE_CACHE_SIZE is too large on 64k-page platforms (i.e. PPC), so it fails to prevent the vmalloc() spinlock contention described in LU-3338. The new value was chosen to be large enough to accommodate common use cases while staying well below the 16k threshold at which allocations start using vmalloc(). Signed-off-by:Ned Bass <bass6@llnl.gov> Signed-off-by:
Kyle Blatter <kyleblatter@llnl.gov> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5549 Reviewed-on: http://review.whamcloud.com/11614Reviewed-by:
-
Ned Bass authored
Allow default_easize to be tuned via /sysfs. A system administrator might want this if a rare access to widely striped files drives up the value on a filesystem where narrowly striped files are the more common case. In practice, however, this is wanted primarily to facilitate a test case for LU-5549. - Plumb the necessary interfaces through the LMV and MDC layers to expose write access to this value by higher layers. - Add block comments to modified functions. Signed-off-by:
-
wang di authored
Add indexing option to default dirstripe EA. If MDT find out the client send the create req to the wrong MDT because of default stripeEA, it will return -EREMOTE, then client will retrieve default stripeEA through xattr cache, and re-create the object. Also merged patch for LU-6341 to resolve the following problem. Use ll_dir_getstripe to get default stripeEA in ll_new_node(), Because ll_getxattr_common requires admin rights for retrieving default LMVEA (because of trusted- prefix), which might cause mkdir (from normal user) failure. If parent does not have default stripeEA, then child should always be in the same MDT for mkdir. Otherwise MDT should return -EREMOTE, then client will refresh the default stripe index, and recreate the object. Signed-off-by:
-
Mikhail Pershin authored
Patch fixes the issue seen on the client with growing request timeout which occurred after the server side patch landed for LU-5079. While commit itself is correct, it reveals another issue. If request is being processed for a long time on server then client adaptive timeouts will adapt to that after receiving reply and new requests will have bigger timeout. Another problem is that server AT history is corrupted by recovery request processing time which not pure service time but includes also waiting time for clients to recover. Patch prevents the AT stats update from early replies on client and from recovering requests processing time on server. The ptlrpc_at_recv_early_reply() still updates the current request timeout as asked by server, but don't include this into AT stats. The real reply will bring that data from server after all. Signed-off-by:
Mikhail Pershin <mike.pershin@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-6084 Reviewed-on: http://review.whamcloud.com/13520Reviewed-by:
Jian Yu <jian.yu@intel.com> Reviewed-by:
-
James Simmons authored
When a parameter is permanently changed on the MGS the MGS send a changelog packet to the proper nodes that are affected by the change. Once the nodes receive the change they then call the userland utility lctl to change its local value. When calling a userland application from the kernel you specify a flag to control the interaction with the application. Originally by default the flag was set to 0 which is UMH_NO_WAIT which meant lctl was being called asynchronously. In older kernels this was fine since UHM_NO_WAIT and UHM_WAIT_PROC had nearly the same logic. This changed with newer kernels which broke updating our parameters. Plus doing a UHM_NO_WAIT doesn't report back a error if something goes wrong with lctl. The fix is to set the flag to UHM_WAIT_PROC so kernel space waits until lctl has finished and we get a proper error code if something does go wrong with lctl. Signed-off-by:
James Simmons <uja.ornl@gmail.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-6063 Reviewed-on: http://review.whamcloud.com/13677Reviewed-by:
Bob Glossman <bob.glossman@intel.com> Reviewed-by:
-
frank zago authored
A lot of symbols don't need to be exported at all because they are only used in the module they belong to. Signed-off-by:
frank zago <fzago@cray.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5829 Reviewed-on: http://review.whamcloud.com/12510Reviewed-by:
-
wang di authored
Because the inode and its connected dentries will be cleared out of the cache after migration, the inode needs to be locked during the migration. Signed-off-by:
-
frank zago authored
A lot of symbols don't need to be exported at all because they are only used in the module they belong to. Signed-off-by:
-
frank zago authored
A lot of symbols don't need to be exported at all because they are only used in the module they belong to. Signed-off-by:
-
Johann Lombardi authored
Use at_max in osc_enter_cache() to bound how long we wait for grant space before switching to synchronous I/Os. Do not print a message on the console when the timeout is hit since such long wait can be legitimate with flaky network (i.e. BRW is resent multiple times). Signed-off-by:
Johann Lombardi <johann.lombardi@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5521 Reviewed-on: http://review.whamcloud.com/12146Reviewed-by:
Niu Yawei <yawei.niu@intel.com> Reviewed-by:
Jinshan Xiong <jinshan.xiong@intel.com> Reviewed-by:
-
wang di authored
Do not revalidate slave stripes while holding master lock. Otherwise if the revalidating slaves are blocked, then the master lock can not be released in time. Remove some unnecesary merging in ll_revalidate_slave(), and the attributes will be stored in each stripe, only merging them if required. Signed-off-by:
-
Artem Blagodarenko authored
After DNE phase 2 has been added to client it sends create request to slave MDT. DNT1-only server doesn't expect request to slave MDT from client. It expects only cross-mdt request from master MDT. Thus if DNE2 client tries to "mkdir -i 1" on DNE1 server, then LBUG happened. This patch adds OBD_CONNECT_DIR_STRIPE connection flag check on client side. If striped directories are not supported by server, then create requrest is sent to master MDT. Signed-off-by:
Artem Blagodarenko <artem_blagodarenko@xyratex.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-6071 Xyratex-bug-id: MRP-2319 Reviewed-on: http://review.whamcloud.com/13189Reviewed-by:
-
Bobi Jam authored
Store inode's fid in cl_setattr_ost() and OSC packs this info on the wire (via lustre_set_wire_obdo) so that OST can use. NOTE: currently lu_fid::f_ver and obdo::o_parent_ver are not used on OFD device, and we use obdo::o_stripe_idx as filter_fid::ff_parent::f_ver and save it to the device. Signed-off-by:
Bobi Jam <bobijam.xu@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-1154 Reviewed-on: http://review.whamcloud.com/12902Reviewed-by:
-
Bobi Jam authored
coo_attr_set() is used to update object's attribute but its name makes confusion that people intuitively think that it is used to pass object's attribute down to server sides. Signed-off-by:
-
Fan Yong authored
The ll_lookup_it() may trigger IT_OPEN RPC to open a file by name. But at that time, the client does not know the target file's GID, so it cannot pack the necessary supplementary group ID in the RPC. Because of missing the supplementary group ID, the RPC maybe fail for open permission check on the MDS. Under such case, MDS should return the target file's GID, if the current thread on the client in the right group (according to the file's GID), the client will try the IT_OPEN RPC again with the right supplementary group ID. This patch is also helpful if some other(s) changed the file's GID after current RPC sent to the MDS with the suppgid as the original GID by race. Signed-off-by:
-
John L. Hammond authored
Merge the contents of lustre/include/linux/lvfs.h into lustre/include/lvfs.h. Merge lustre/include/linux/lustre_user.h into lustre/include/lustre/lustre_user.h. Move lustre_compat25.h and lustre_patchless_compat.h from lustre/include/linux/ to lustre/include/ and rename lustre_compat25.h to lustre_compat.h. Signed-off-by:
-
Liang Zhen authored
cpumask_of_node can return NULL if NUMA node is unavailable, in this case cfs_node_to_cpumask will try to copy from NULL and cause kernel panic. Signed-off-by:
Liang Zhen <liang.zhen@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5751 Reviewed-on: http://review.whamcloud.com/13207Reviewed-by:
Li Wei <wei.g.li@intel.com> Reviewed-by:
Bobi Jam <bobijam@hotmail.com> Reviewed-by:
-
Dmitry Eremin authored
Change type of cl_conn_count to size_t. Signed-off-by:
-
John L. Hammond authored
In ll_lov_setstripe_ea_info() release the inode size lock on all appropriate exit paths. Signed-off-by:
-
Andreas Dilger authored
Add comment blocks on lprocfs_stats_lock() and lprocfs_stats_unlock(). Move common NOPERCPU code out of the switch() statements to reduce code size and complexity, since it doesn't depend on the opc at all. Replace switch() in lprocfs_stats_unlock() with a simple if/else, since the lock opc was already checked in lprocfs_stats_lock(). Add an enum for the lprocfs_stats_lock() operations to make it clear what the valid values are and allow compiler checking. Signed-off-by:
-
Dmitry Eremin authored
Change the type accordant usage and remove warnings. Signed-off-by:
-
Bobi Jam authored
An OSC object could be destroyed with AGL locks waiting for granted, so we'd get rid of the osc_object_ast_clear() assertion that its dlm locks all getting granted. Signed-off-by:
-
Alexander Boyko authored
mgc_apply_recover_logs use only first nid from entry, this could be the problem for a cluster with several network address for a one node. Signed-off-by:
Alexander Boyko <alexander.boyko@seagate.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5950 Xyratex-bug-id: MRP-2255 Reviewed-on: http://review.whamcloud.com/12829Reviewed-by:
Ann Koehler <amk@cray.com> Signed-off-by:
-
Alexander Boyko authored
Buggy code at ptlrpc_connect_interpret() finish: rc = ptlrpc_import_recovery_state_machine(imp); ... Set import connection flags When import has FULL state ptlrpc_import_recovery_state_machine() wakeup all waiters on import and all delayed request, which was resented. And it could happened that request was send without updated flags and AT is disabled. If such request is in progress on the server, server drop the new instance, and could do early reply for it. But this early reply confuse client, cause it wait real reply(no AT for this request). Client try to touch buffer outside reply and got EPROTO error. The same bug existed for initital connect too. Import became FULL before import connection flags was set. Signed-off-by:Alexander Boyko <alexander_boyko@xyratex.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5528 Xyratex-bug-id: MRP-2034 Reviewed-on: http://review.whamcloud.com/11723Reviewed-by:
-
John L. Hammond authored
Flatten out the lsm_wire struct from the middle of struct lov_stripe_md and remove the member name macros. Signed-off-by:
Bobi Jam <bobijam@gmail.com> Reviewed-by:
-
Jinshan Xiong authored
lustre_idl.h only includes wire data; lustre_dlm.h is the right place for LDLM_GID_ANY. Signed-off-by:
-
Dmitry Eremin authored
Change return type and size argiments of lustre_msg_hdr_size(), lustre_msg_buf{len,count}() and req_capsule_*_size() to __u32. Change type of req_format->rf_idx and req_format->rf_fields.nr to size_t. Also return zero for incorrect message magic instead of -EINVAL. This will be more robust because of few of them after LASSERTF(0, "...") and will not be returned. In the rest places it return zero size instead of huge number after implicit unsigned conversion. Signed-off-by: -
Bobi Jam authored
Use cl_object_operations::coo_getstripe() to handle LL_IOC_LOV_GETSTRIPE ops. Signed-off-by:
-
Dmitry Eremin authored
Change the type accordant usage. Signed-off-by:
-
Dmitry Eremin authored
Pages count is unsigned. So, change the type accordant usage. Signed-off-by:
-
Nathaniel Clark authored
Instead of rechecking a static value, recalculate to see if pool stats need to be updated. Add newline so message will print instead of warning about missing newline. Signed-off-by:
Nathaniel Clark <nathaniel.l.clark@intel.com> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-4536 Reviewed-on: http://review.whamcloud.com/12547Reviewed-by:
-
Amir Shehata authored
There is a race condition on client reconnect when the import is being destroyed. Some outstanding client bound requests are being processed when the imp_sec has alread been freed. Ensure to suppress the error message in import_sec_validate_get() in that case Signed-off-by:
-
Bob Glossman authored
Always return an ERR_PTR() on errors, never return a NULL, in lu_object_find_slice(). Also clean up callers who no longer need special case handling of NULL returns. Signed-off-by:
-
Dmitry Eremin authored
Cleanup warnings about comparison between signed and unsigned. Signed-off-by:
-
