Branch: b1_4

Merging of changes from b1_2 into b1_4.
- don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
- lock /proc/sys/portals/routes internal state, avoiding oops (4827)
- debugging for client eviction looping (4908)

diff --git a/lustre/ChangeLog b/lustre/ChangeLog
index 04449b8..85f9bae 100644 (file)
--- a/lustre/ChangeLog
+++ b/lustre/ChangeLog
@@ -1,22 +1,27 @@
 tbd         Cluster File Systems, Inc. <info@clusterfs.com>
        * version 1.3.4
        * bug fixes
+        - changes from 1.2.9
        - flock/lockf fixes (but it's still disabled, pending 5135)
        - don't use EXT3 constants in llite code (5094)
-       - return async write errors to application if possible (2248)
-       - return last_committed value from OST to avoid OOM (4966)
        - memory shortage at startup could cause assertion (5176)
-       - the watchdog thread now runs as interruptible (5246)
        * miscellania
        - reorganization of lov code
        - single portals codebase
        - Infiniband NAL
-       - add pid to ldlm debugging output (4922)
-       - return last_committed value from OST to avoid OOM (4966)
        - add extents/mballoc support (5025)
        - direct I/O reads in the obdfilter (4048)
 
 tbd         Cluster File Systems, Inc. <info@clusterfs.com>
+       * version 1.2.9
+       - send OST transaction number in read/write reply to free req (4966)
+       - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
+       - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
+       - the watchdog thread now runs as interruptible (5246)
+       * miscellania
+       - add pid to ldlm debugging output (4922)
+
+2004-11-17  Cluster File Systems, Inc. <info@clusterfs.com>
        * version 1.2.8
        * bug fixes
        - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
@@ -39,6 +44,7 @@ tbd         Cluster File Systems, Inc. <info@clusterfs.com>
        - add software watchdogs to catch hung threads quickly (4941)
        - make lustrefs init script start after nfs is mounted
        - fix CWARN/ERROR duplication (4930)
+       - return async write errors to application if possible (2248)
        - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
        - print NAL number in %x format (4645)
 

diff --git a/lustre/autoMakefile.am b/lustre/autoMakefile.am
index 7830e54..81bde4b 100644 (file)
--- a/lustre/autoMakefile.am
+++ b/lustre/autoMakefile.am
@@ -56,7 +56,7 @@ lvfs-sources:
        $(MAKE) sources -C lvfs
 
 modules: lustre_build_version $(DEP) $(LDISKFS) lvfs-sources
-       $(MAKE) $(ARCH_UM) CC=$(CC) -C $(LINUX) -f $(PWD)/kernel-tests/Makefile LUSTRE_LINUX_CONFIG=$(LINUX_CONFIG) $(MODULE_TARGET)=$(PWD) -o tmp_include_depends -o scripts -o include/config/MARKER $@
+       $(MAKE) $(ARCH_UM) CC="$(CC)" -C $(LINUX) -f $(PWD)/kernel-tests/Makefile LUSTRE_LINUX_CONFIG=$(LINUX_CONFIG) $(MODULE_TARGET)=$(PWD) -o tmp_include_depends -o scripts -o include/config/MARKER $@
 
 endif # MODULES
 

diff --git a/lustre/ldlm/ldlm_flock.c b/lustre/ldlm/ldlm_flock.c
index 1dacf83..ac41bfe 100644 (file)
--- a/lustre/ldlm/ldlm_flock.c
+++ b/lustre/ldlm/ldlm_flock.c
@@ -469,7 +469,7 @@ ldlm_flock_completion_ast(struct ldlm_lock *lock, int flags, void *data)
         LDLM_DEBUG(lock, "client-side enqueue returned a blocked lock, "
                    "sleeping");
 
-        ldlm_lock_dump(D_DLMTRACE, lock, 0);
+        ldlm_lock_dump(D_OTHER, lock, 0);
 
         fwd.fwd_lock = lock;
         obd = class_exp2obd(lock->l_conn_export);

diff --git a/lustre/ldlm/ldlm_lockd.c b/lustre/ldlm/ldlm_lockd.c
index 4074048..39d24ac 100644 (file)
--- a/lustre/ldlm/ldlm_lockd.c
+++ b/lustre/ldlm/ldlm_lockd.c
@@ -170,7 +170,7 @@ static int expired_lock_main(void *arg)
 
 static void waiting_locks_callback(unsigned long unused)
 {
-        struct ldlm_lock *lock;
+        struct ldlm_lock *lock, *last = NULL;
         char str[PTL_NALFMT_SIZE];
 
         spin_lock_bh(&waiting_locks_spinlock);
@@ -187,6 +187,17 @@ static void waiting_locks_callback(unsigned long unused)
                            lock->l_export->exp_connection->c_remote_uuid.uuid,
                            ptlrpc_peernid2str(&lock->l_export->exp_connection->c_peer, str));
 
+                if (lock == last) {
+                        LDLM_ERROR(lock, "waiting on lock multiple times");
+                        CERROR("wll %p .prev %p, l_pending.next %p .prev %p\n",
+                               waiting_locks_list.next, waiting_locks_list.prev,
+                               lock->l_pending_chain.next,
+                               lock->l_pending_chain.prev);
+                        spin_unlock(&waiting_locks_spinlock);
+                        LBUG();
+                }
+                last = lock;
+
                 spin_lock_bh(&expired_lock_thread.elt_lock);
                 list_del(&lock->l_pending_chain);
                 list_add(&lock->l_pending_chain,

diff --git a/lustre/llite/llite_lib.c b/lustre/llite/llite_lib.c
index 33d93c2..2d09beb 100644 (file)
--- a/lustre/llite/llite_lib.c
+++ b/lustre/llite/llite_lib.c
@@ -742,7 +742,8 @@ static int null_if_equal(struct ldlm_lock *lock, void *data)
                 lock->l_ast_data = NULL;
 
                 if (lock->l_req_mode != lock->l_granted_mode)
-                        LDLM_ERROR(lock,"clearing inode with ungranted lock\n");        }
+                        LDLM_ERROR(lock,"clearing inode with ungranted lock\n");
+        }
 
         return LDLM_ITER_CONTINUE;
 }

diff --git a/lustre/llite/rw.c b/lustre/llite/rw.c
index 3d3c7ec..9798c3e 100644 (file)
--- a/lustre/llite/rw.c
+++ b/lustre/llite/rw.c
@@ -434,12 +434,12 @@ static int queue_or_sync_write(struct obd_export *exp, struct inode *inode,
                 GOTO(out, rc);
 
         /* make full-page requests if we are not at EOF (bug 4410) */
-        if (llap->llap_page->index < size_index) {
+        if (to != PAGE_SIZE && llap->llap_page->index < size_index) {
                 LL_CDEBUG_PAGE(D_PAGE, llap->llap_page,
                                "sync write before EOF: size_index %lu, to %d\n",
                                size_index, to);
                 to = PAGE_SIZE;
-        } else if (llap->llap_page->index == size_index) {
+        } else if (to != PAGE_SIZE && llap->llap_page->index == size_index) {
                 int size_to = inode->i_size & ~PAGE_MASK;
                 LL_CDEBUG_PAGE(D_PAGE, llap->llap_page,
                                "sync write at EOF: size_index %lu, to %d/%d\n",

diff --git a/lustre/lov/lov_internal.h b/lustre/lov/lov_internal.h
index e6ecc5e..752d074 100644 (file)
--- a/lustre/lov/lov_internal.h
+++ b/lustre/lov/lov_internal.h
@@ -56,10 +56,10 @@ struct lov_async_page {
         int                             lap_magic;
         int                             lap_stripe;
         obd_off                         lap_sub_offset;
+        obd_id                          lap_loi_id;
         void                            *lap_sub_cookie;
         struct obd_async_page_ops       *lap_caller_ops;
         void                            *lap_caller_data;
-        obd_id                          lap_loi_id;
 };
 
 #define LAP_FROM_COOKIE(c)                                                      \

diff --git a/lustre/obdfilter/filter.c b/lustre/obdfilter/filter.c
index 67df476..ddeba4d 100644 (file)
--- a/lustre/obdfilter/filter.c
+++ b/lustre/obdfilter/filter.c
@@ -258,7 +258,12 @@ static int filter_client_free(struct obd_export *exp, int flags)
         memset(&zero_fcd, 0, sizeof zero_fcd);
         push_ctxt(&saved, &obd->obd_ctxt, NULL);
         rc = fsfilt_write_record(obd, filter->fo_rcvd_filp, &zero_fcd,
-                                 sizeof(zero_fcd), &off, 1);
+                                 sizeof(zero_fcd), &off, 0);
+
+        if (rc == 0)
+                /* update server's transno */
+                filter_update_server_data(obd, filter->fo_rcvd_filp,
+                                          filter->fo_fsd, 1);
         pop_ctxt(&saved, &obd->obd_ctxt, NULL);
 
         CDEBUG(rc == 0 ? D_INFO : D_ERROR,
@@ -1033,8 +1038,8 @@ static int filter_destroy_internal(struct obd_device *obd, obd_id objid,
         ENTRY;
 
         if (inode->i_nlink != 1 || atomic_read(&inode->i_count) != 1) {
-                CERROR("destroying objid %.*s nlink = %lu, count = %d\n",
-                       dchild->d_name.len, dchild->d_name.name,
+                CERROR("destroying objid %.*s ino %lu nlink %lu count %d\n",
+                       dchild->d_name.len, dchild->d_name.name, inode->i_ino,
                        (unsigned long)inode->i_nlink,
                        atomic_read(&inode->i_count));
         }
@@ -1420,7 +1425,6 @@ static void filter_grant_sanity_check(struct obd_device *obd, const char *func)
         obd_size maxsize = obd->obd_osfs.os_blocks * obd->obd_osfs.os_bsize;
         obd_size tot_dirty = 0, tot_pending = 0, tot_granted = 0;
         obd_size fo_tot_dirty, fo_tot_pending, fo_tot_granted;
-        int level = D_CACHE;
 
         if (list_empty(&obd->obd_exports))
                 return;
@@ -1428,10 +1432,11 @@ static void filter_grant_sanity_check(struct obd_device *obd, const char *func)
         spin_lock(&obd->obd_osfs_lock);
         spin_lock(&obd->obd_dev_lock);
         list_for_each_entry(exp, &obd->obd_exports, exp_obd_chain) {
+                int error = 0;
                 fed = &exp->exp_filter_data;
                 if (fed->fed_grant < 0 || fed->fed_pending < 0 ||
                     fed->fed_dirty < 0)
-                        level = D_ERROR;
+                        error = 1;
                 if (maxsize > 0) { /* we may not have done a statfs yet */
                         LASSERTF(fed->fed_grant + fed->fed_pending <= maxsize,
                                  "%s: cli %s/%p %ld+%ld > "LPU64"\n", func,
@@ -1442,9 +1447,14 @@ static void filter_grant_sanity_check(struct obd_device *obd, const char *func)
                                  exp->exp_client_uuid.uuid, exp,
                                  fed->fed_dirty, maxsize);
                 }
-                CDEBUG(level, "%s: cli %s/%p dirty %ld pend %ld grant %ld\n",
-                       obd->obd_name, exp->exp_client_uuid.uuid, exp,
-                       fed->fed_dirty, fed->fed_pending, fed->fed_grant);
+                if (error)
+                        CERROR("%s: cli %s/%p dirty %ld pend %ld grant %ld\n",
+                               obd->obd_name, exp->exp_client_uuid.uuid, exp,
+                               fed->fed_dirty, fed->fed_pending,fed->fed_grant);
+                else
+                        CDEBUG(D_CACHE, "%s: cli %s/%p dirty %ld pend %ld grant %ld\n",
+                               obd->obd_name, exp->exp_client_uuid.uuid, exp,
+                               fed->fed_dirty, fed->fed_pending,fed->fed_grant);
                 tot_granted += fed->fed_grant + fed->fed_pending;
                 tot_pending += fed->fed_pending;
                 tot_dirty += fed->fed_dirty;

diff --git a/lustre/obdfilter/filter_io.c b/lustre/obdfilter/filter_io.c
index 9764996..599e9dc 100644 (file)
--- a/lustre/obdfilter/filter_io.c
+++ b/lustre/obdfilter/filter_io.c
@@ -187,18 +187,11 @@ restat:
         if (left >= tot_granted) {
                 left -= tot_granted;
         } else {
-                static unsigned long next;
-                if (left < tot_granted - obd->u.filter.fo_tot_pending &&
-                    time_after(jiffies, next)) {
-                        spin_unlock(&obd->obd_osfs_lock);
+                if (left < tot_granted - obd->u.filter.fo_tot_pending + 65536) {
                         CERROR("%s: cli %s/%p grant "LPU64" > available "
                                LPU64" and pending "LPU64"\n", obd->obd_name,
                                exp->exp_client_uuid.uuid, exp, tot_granted,
                                left, obd->u.filter.fo_tot_pending);
-                        if (next == 0)
-                                portals_debug_dumplog();
-                        next = jiffies + 20 * HZ;
-                        spin_lock(&obd->obd_osfs_lock);
                 }
                 left = 0;
         }

diff --git a/lustre/osc/osc_internal.h b/lustre/osc/osc_internal.h
index f7ac79a..f36d6ce 100644 (file)
--- a/lustre/osc/osc_internal.h
+++ b/lustre/osc/osc_internal.h
@@ -14,11 +14,10 @@
 
 struct osc_async_page {
         int                     oap_magic;
+        int                     oap_cmd;
         struct list_head        oap_pending_item;
         struct list_head        oap_urgent_item;
         struct list_head        oap_rpc_item;
-        struct page             *oap_page;
-        int                     oap_cmd;
 
         obd_off                 oap_obj_off;
         obd_off                 oap_page_off;
@@ -27,8 +26,9 @@ struct osc_async_page {
         enum async_flags        oap_async_flags;
 
         unsigned long           oap_interrupted:1;
-        struct obd_io_group     *oap_oig;
         struct oig_callback_context oap_occ;
+        struct page             *oap_page;
+        struct obd_io_group     *oap_oig;
         struct ptlrpc_request   *oap_request;
         struct client_obd       *oap_cli;
         struct lov_oinfo        *oap_loi;

diff --git a/lustre/portals/router/proc.c b/lustre/portals/router/proc.c
index a1397d2..9a3d4f2 100644 (file)
--- a/lustre/portals/router/proc.c
+++ b/lustre/portals/router/proc.c
@@ -31,6 +31,7 @@ struct proc_route_data {
         struct list_head *curr;
         unsigned int generation;
         off_t skip;
+        rwlock_t proc_route_rwlock;
 } kpr_read_routes_data;
 
 /* nal2name support re-used from utils/portals.c */
@@ -96,19 +97,22 @@ static int kpr_proc_router_write(struct file *file, const char *ubuffer,
 static int kpr_proc_routes_read(char *page, char **start, off_t off,
                                 int count, int *eof, void *data)
 {
-        struct proc_route_data *prd = data;
-        kpr_route_entry_t     *re;
-        kpr_gateway_entry_t *ge;
-        int                 chunk_len = 0;
-        int                 line_len = 0;
-        int                 user_len = 0;
+        struct proc_route_data  *prd = data;
+        kpr_route_entry_t       *re;
+        kpr_gateway_entry_t     *ge;
+        int                     chunk_len = 0;
+        int                     line_len = 0;
+        int                     user_len = 0;
+        int                     rc = 0;
 
         *eof = 1;
         *start = page;
 
+        write_lock(&(prd->proc_route_rwlock));
+
         if (prd->curr == NULL) {
                 if (off != 0)
-                        return 0;
+                        goto routes_read_exit;
 
                 /* First pass, initialize our private data */
                 prd->curr = kpr_routes.next;
@@ -118,13 +122,14 @@ static int kpr_proc_routes_read(char *page, char **start, off_t off,
                 /* Abort route list generation change */
                 if (prd->generation != kpr_routes_generation) {
                         prd->curr = NULL;
-                        return sprintf(page, "\nError: Routes Changed\n");
+                        rc = sprintf(page, "\nError: Routes Changed\n");
+                        goto routes_read_exit;
                 }
 
                 /* All the routes have been walked */
                 if (prd->curr == &kpr_routes) {
                         prd->curr = NULL;
-                        return 0;
+                        goto routes_read_exit;
                 }
         }
 
@@ -148,7 +153,8 @@ static int kpr_proc_routes_read(char *page, char **start, off_t off,
                 if (prd->curr->next == NULL) {
                         prd->curr = NULL;
                         read_unlock(&kpr_rwlock);
-                        return sprintf(page, "\nError: Routes Changed\n");
+                        rc = sprintf(page, "\nError: Routes Changed\n");
+                        goto routes_read_exit;
                 }
 
                 prd->curr = prd->curr->next;
@@ -169,13 +175,18 @@ static int kpr_proc_routes_read(char *page, char **start, off_t off,
                 prd->curr = prd->curr->prev;
                 prd->skip = line_len - (user_len - count);
                 read_unlock(&kpr_rwlock);
-                return count;
+                rc = count;
+                goto routes_read_exit;
         }
 
         /* Not enough data to entirely satify callers request */
         prd->skip = 0;
         read_unlock(&kpr_rwlock);
-        return user_len;
+        rc = user_len;
+
+routes_read_exit:
+        write_unlock(&(prd->proc_route_rwlock));
+        return rc;
 }
 
 static int kpr_proc_routes_write(struct file *file, const char *ubuffer,
@@ -215,6 +226,7 @@ void kpr_proc_init(void)
         kpr_read_routes_data.curr = NULL;
         kpr_read_routes_data.generation = 0;
         kpr_read_routes_data.skip = 0;
+        kpr_read_routes_data.proc_route_rwlock = RW_LOCK_UNLOCKED;
 
         routes_entry->data = &kpr_read_routes_data;
         routes_entry->read_proc = kpr_proc_routes_read;

diff --git a/lustre/ptlrpc/niobuf.c b/lustre/ptlrpc/niobuf.c
index 1171fb5..df2110d 100644 (file)
--- a/lustre/ptlrpc/niobuf.c
+++ b/lustre/ptlrpc/niobuf.c
@@ -447,6 +447,10 @@ int ptl_send_rpc(struct ptlrpc_request *request)
         if (rc != PTL_OK) {
                 CERROR("PtlMDAttach failed: %d\n", rc);
                 LASSERT (rc == PTL_NO_SPACE);
+                spin_lock_irqsave (&request->rq_lock, flags);
+                /* ...but the MD attach didn't succeed... */
+                request->rq_receiving_reply = 0;
+                spin_unlock_irqrestore (&request->rq_lock, flags);
                 GOTO(cleanup_me, rc -ENOMEM);
         }
 
@@ -456,7 +460,7 @@ int ptl_send_rpc(struct ptlrpc_request *request)
                request->rq_reply_portal,
                connection->c_peer.peer_ni->pni_name);
 
-        ptlrpc_request_addref(request);        /* +1 ref for the SENT callback */
+        ptlrpc_request_addref(request);       /* +1 ref for the SENT callback */
 
         request->rq_sent = CURRENT_SECONDS;
         ptlrpc_pinger_sending_on_import(request->rq_import);

diff --git a/lustre/ptlrpc/recover.c b/lustre/ptlrpc/recover.c
index 72924fc..b6bae88 100644 (file)
--- a/lustre/ptlrpc/recover.c
+++ b/lustre/ptlrpc/recover.c
@@ -224,8 +224,7 @@ int ptlrpc_replay_next(struct obd_import *imp, int *inflight)
 
 int ptlrpc_resend(struct obd_import *imp)
 {
-        struct list_head *tmp, *pos;
-        struct ptlrpc_request *req;
+        struct ptlrpc_request *req, *next;
         unsigned long flags;
 
         ENTRY;
@@ -243,8 +242,10 @@ int ptlrpc_resend(struct obd_import *imp)
         }
         spin_unlock_irqrestore(&imp->imp_lock, flags);
 
-        list_for_each_safe(tmp, pos, &imp->imp_sending_list) {
-                req = list_entry(tmp, struct ptlrpc_request, rq_list);
+        list_for_each_entry_safe(req, next, &imp->imp_sending_list, rq_list) {
+                LASSERTF((long)req > PAGE_SIZE && req != LP_POISON,
+                         "req %p bad\n", req);
+                LASSERTF(req->rq_type != LI_POISON, "req %p freed\n", req);
                 ptlrpc_resend_req(req);
         }
 

diff --git a/lustre/tests/recovery-small.sh b/lustre/tests/recovery-small.sh
index 6865e6c..157bca1 100755 (executable)
--- a/lustre/tests/recovery-small.sh
+++ b/lustre/tests/recovery-small.sh
@@ -148,7 +148,7 @@ test_10() {
 run_test 10 "finish request on server after client eviction (bug 1521)"
 
 #bug 2460
-# wake up a thead waiting for completion after eviction
+# wake up a thread waiting for completion after eviction
 test_11(){
     do_facet client multiop $MOUNT/$tfile Ow  || return 1
     do_facet client multiop $MOUNT/$tfile or  || return 2
@@ -161,7 +161,7 @@ test_11(){
 
     do_facet client munlink $MOUNT/$tfile  || return 4
 }
-run_test 11 "wake up a thead waiting for completion after eviction (b=2460)"
+run_test 11 "wake up a thread waiting for completion after eviction (b=2460)"
 
 #b=2494
 test_12(){

diff --git a/lustre/utils/lconf b/lustre/utils/lconf
index 96cf611..9fe97fa 100755 (executable)
--- a/lustre/utils/lconf
+++ b/lustre/utils/lconf
@@ -969,7 +969,7 @@ def sys_get_local_address(net_type, wildcard, cluster_id):
             iface, star = string.split(wildcard, ':')
             local = if2addr(iface)
             if not local:
-                panic ("unable to determine ip for:", wildcard)
+                panic("unable to determine ip for:", wildcard)
         else:
             host = socket.gethostname()
             local = socket.gethostbyname(host)