X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fosc%2Fosc_cache.c;h=99a18dc19998402d3629f6e8cd26b9424f7d1f27;hp=f2bad36ed7b807b2ee9f5ac098e84bb95d47eff3;hb=c2a0ab016383be46a8d04c00ff163eb6a4550f58;hpb=79c00733d0b4f8662b630fa2c6c1d475eb7556d2 diff --git a/lustre/osc/osc_cache.c b/lustre/osc/osc_cache.c index f2bad36..99a18dc 100644 --- a/lustre/osc/osc_cache.c +++ b/lustre/osc/osc_cache.c @@ -27,7 +27,7 @@ * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2012, 2013, Intel Corporation. + * Copyright (c) 2012, 2014, Intel Corporation. * */ /* @@ -48,7 +48,7 @@ static int extent_debug; /* set it to be true for more debug */ static void osc_update_pending(struct osc_object *obj, int cmd, int delta); static int osc_extent_wait(const struct lu_env *env, struct osc_extent *ext, - int state); + enum osc_extent_state state); static void osc_ap_completion(const struct lu_env *env, struct client_obd *cli, struct osc_async_page *oap, int sent, int rc); static int osc_make_ready(const struct lu_env *env, struct osc_async_page *oap, @@ -181,7 +181,7 @@ static int osc_extent_sanity_check0(struct osc_extent *ext, { struct osc_object *obj = ext->oe_obj; struct osc_async_page *oap; - int page_count; + size_t page_count; int rc = 0; if (!osc_object_is_locked(obj)) @@ -596,9 +596,9 @@ static inline int overlapped(struct osc_extent *ex1, struct osc_extent *ex2) * Find or create an extent which includes @index, core function to manage * extent tree. */ -struct osc_extent *osc_extent_find(const struct lu_env *env, - struct osc_object *obj, pgoff_t index, - int *grants) +static struct osc_extent *osc_extent_find(const struct lu_env *env, + struct osc_object *obj, pgoff_t index, + unsigned int *grants) { struct client_obd *cli = osc_cli(obj); struct osc_lock *olck; @@ -609,10 +609,10 @@ struct osc_extent *osc_extent_find(const struct lu_env *env, struct osc_extent *found = NULL; pgoff_t chunk; pgoff_t max_end; - int max_pages; /* max_pages_per_rpc */ - int chunksize; + unsigned int max_pages; /* max_pages_per_rpc */ + unsigned int chunksize; int ppc_bits; /* pages per chunk bits */ - int chunk_mask; + pgoff_t chunk_mask; int rc; ENTRY; @@ -667,8 +667,8 @@ restart: if (ext == NULL) ext = first_extent(obj); while (ext != NULL) { - loff_t ext_chk_start = ext->oe_start >> ppc_bits; - loff_t ext_chk_end = ext->oe_end >> ppc_bits; + pgoff_t ext_chk_start = ext->oe_start >> ppc_bits; + pgoff_t ext_chk_end = ext->oe_end >> ppc_bits; LASSERT(sanity_check_nolock(ext) == 0); if (chunk > ext_chk_end + 1) @@ -871,7 +871,7 @@ int osc_extent_finish(const struct lu_env *env, struct osc_extent *ext, RETURN(0); } -static int extent_wait_cb(struct osc_extent *ext, int state) +static int extent_wait_cb(struct osc_extent *ext, enum osc_extent_state state) { int ret; @@ -886,7 +886,7 @@ static int extent_wait_cb(struct osc_extent *ext, int state) * Wait for the extent's state to become @state. */ static int osc_extent_wait(const struct lu_env *env, struct osc_extent *ext, - int state) + enum osc_extent_state state) { struct osc_object *obj = ext->oe_obj; struct l_wait_info lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(600), NULL, @@ -915,7 +915,7 @@ static int osc_extent_wait(const struct lu_env *env, struct osc_extent *ext, rc = l_wait_event(ext->oe_waitq, extent_wait_cb(ext, state), &lwi); if (rc == -ETIMEDOUT) { OSC_EXTENT_DUMP(D_ERROR, ext, - "%s: wait ext to %d timedout, recovery in progress?\n", + "%s: wait ext to %u timedout, recovery in progress?\n", osc_export(obj)->exp_obd->obd_name, state); lwi = LWI_INTR(NULL, NULL); @@ -1056,7 +1056,7 @@ static int osc_extent_make_ready(const struct lu_env *env, struct osc_async_page *oap; struct osc_async_page *last = NULL; struct osc_object *obj = ext->oe_obj; - int page_count = 0; + unsigned int page_count = 0; int rc; ENTRY; @@ -1097,9 +1097,10 @@ static int osc_extent_make_ready(const struct lu_env *env, /* the last page is the only one we need to refresh its count by * the size of file. */ if (!(last->oap_async_flags & ASYNC_COUNT_STABLE)) { - last->oap_count = osc_refresh_count(env, last, OBD_BRW_WRITE); - LASSERT(last->oap_count > 0); - LASSERT(last->oap_page_off + last->oap_count <= PAGE_CACHE_SIZE); + int last_oap_count = osc_refresh_count(env, last, OBD_BRW_WRITE); + LASSERT(last_oap_count > 0); + LASSERT(last->oap_page_off + last_oap_count <= PAGE_CACHE_SIZE); + last->oap_count = last_oap_count; spin_lock(&last->oap_lock); last->oap_async_flags |= ASYNC_COUNT_STABLE; spin_unlock(&last->oap_lock); @@ -1130,7 +1131,8 @@ static int osc_extent_make_ready(const struct lu_env *env, * called to expand the extent for the same IO. To expand the extent, the * page index must be in the same or next chunk of ext->oe_end. */ -static int osc_extent_expand(struct osc_extent *ext, pgoff_t index, int *grants) +static int osc_extent_expand(struct osc_extent *ext, pgoff_t index, + unsigned int *grants) { struct osc_object *obj = ext->oe_obj; struct client_obd *cli = osc_cli(obj); @@ -1139,7 +1141,7 @@ static int osc_extent_expand(struct osc_extent *ext, pgoff_t index, int *grants) pgoff_t chunk = index >> ppc_bits; pgoff_t end_chunk; pgoff_t end_index; - int chunksize = 1 << cli->cl_chunkbits; + unsigned int chunksize = 1 << cli->cl_chunkbits; int rc = 0; ENTRY; @@ -1316,7 +1318,7 @@ static int osc_completion(const struct lu_env *env, struct osc_async_page *oap, if (rc == 0 && srvlock) { struct lu_device *ld = opg->ops_cl.cpl_obj->co_lu.lo_dev; struct osc_stats *stats = &lu2osc_dev(ld)->od_stats; - int bytes = oap->oap_count; + size_t bytes = oap->oap_count; if (crt == CRT_READ) stats->os_lockless_reads += bytes; @@ -1423,8 +1425,8 @@ static void __osc_unreserve_grant(struct client_obd *cli, } } -void osc_unreserve_grant(struct client_obd *cli, - unsigned int reserved, unsigned int unused) +static void osc_unreserve_grant(struct client_obd *cli, + unsigned int reserved, unsigned int unused) { spin_lock(&cli->cl_loi_list_lock); __osc_unreserve_grant(cli, reserved, unused); @@ -1449,7 +1451,7 @@ void osc_unreserve_grant(struct client_obd *cli, static void osc_free_grant(struct client_obd *cli, unsigned int nr_pages, unsigned int lost_grant) { - int grant = (1 << cli->cl_chunkbits) + cli->cl_extent_tax; + unsigned long grant = (1 << cli->cl_chunkbits) + cli->cl_extent_tax; spin_lock(&cli->cl_loi_list_lock); atomic_long_sub(nr_pages, &obd_dirty_pages); @@ -1489,7 +1491,7 @@ static int osc_enter_cache_try(struct client_obd *cli, { int rc; - OSC_DUMP_GRANT(D_CACHE, cli, "need:%d.\n", bytes); + OSC_DUMP_GRANT(D_CACHE, cli, "need:%d\n", bytes); rc = osc_reserve_grant(cli, bytes); if (rc < 0) @@ -1530,15 +1532,17 @@ static int ocw_granted(struct client_obd *cli, struct osc_cache_waiter *ocw) static int osc_enter_cache(const struct lu_env *env, struct client_obd *cli, struct osc_async_page *oap, int bytes) { - struct osc_object *osc = oap->oap_obj; - struct lov_oinfo *loi = osc->oo_oinfo; - struct osc_cache_waiter ocw; - struct l_wait_info lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(600), NULL, - LWI_ON_SIGNAL_NOOP, NULL); - int rc = -EDQUOT; + struct osc_object *osc = oap->oap_obj; + struct lov_oinfo *loi = osc->oo_oinfo; + struct osc_cache_waiter ocw; + struct l_wait_info lwi; + int rc = -EDQUOT; ENTRY; - OSC_DUMP_GRANT(D_CACHE, cli, "need:%d.\n", bytes); + lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(AT_OFF ? obd_timeout : at_max), + NULL, LWI_ON_SIGNAL_NOOP, NULL); + + OSC_DUMP_GRANT(D_CACHE, cli, "need:%d\n", bytes); spin_lock(&cli->cl_loi_list_lock); @@ -1546,12 +1550,16 @@ static int osc_enter_cache(const struct lu_env *env, struct client_obd *cli, * of queued writes and create a discontiguous rpc stream */ if (OBD_FAIL_CHECK(OBD_FAIL_OSC_NO_GRANT) || cli->cl_dirty_max_pages == 0 || - cli->cl_ar.ar_force_sync || loi->loi_ar.ar_force_sync) + cli->cl_ar.ar_force_sync || loi->loi_ar.ar_force_sync) { + OSC_DUMP_GRANT(D_CACHE, cli, "forced sync i/o\n"); GOTO(out, rc = -EDQUOT); + } /* Hopefully normal case - cache space and write credits available */ - if (osc_enter_cache_try(cli, oap, bytes, 0)) + if (osc_enter_cache_try(cli, oap, bytes, 0)) { + OSC_DUMP_GRANT(D_CACHE, cli, "granted from cache\n"); GOTO(out, rc = 0); + } /* We can get here for two reasons: too many dirty pages in cache, or * run out of grants. In both cases we should write dirty pages out. @@ -1576,42 +1584,51 @@ static int osc_enter_cache(const struct lu_env *env, struct client_obd *cli, spin_lock(&cli->cl_loi_list_lock); - /* l_wait_event is interrupted by signal, or timed out */ if (rc < 0) { - switch (rc) { - case -ETIMEDOUT: - OSC_DUMP_GRANT(D_ERROR, cli, - "try to reserve %d.\n", bytes); - osc_extent_tree_dump(D_ERROR, osc); - rc = -EDQUOT; - break; - case -EINTR: - /* Ensures restartability - LU-3581 */ - rc = -ERESTARTSYS; - break; - default: - CDEBUG(D_CACHE, "%s: event for cache space @" - " %p never arrived due to %d\n", - cli->cl_import->imp_obd->obd_name, - &ocw, rc); - break; - } + /* l_wait_event is interrupted by signal or timed out */ list_del_init(&ocw.ocw_entry); - GOTO(out, rc); + break; } - LASSERT(list_empty(&ocw.ocw_entry)); rc = ocw.ocw_rc; if (rc != -EDQUOT) - GOTO(out, rc); - if (osc_enter_cache_try(cli, oap, bytes, 0)) - GOTO(out, rc = 0); + break; + if (osc_enter_cache_try(cli, oap, bytes, 0)) { + rc = 0; + break; + } + } + + switch (rc) { + case 0: + OSC_DUMP_GRANT(D_CACHE, cli, "finally got grant space\n"); + break; + case -ETIMEDOUT: + OSC_DUMP_GRANT(D_CACHE, cli, + "timeout, fall back to sync i/o\n"); + osc_extent_tree_dump(D_CACHE, osc); + /* fall back to synchronous I/O */ + rc = -EDQUOT; + break; + case -EINTR: + /* Ensures restartability - LU-3581 */ + OSC_DUMP_GRANT(D_CACHE, cli, "interrupted\n"); + rc = -ERESTARTSYS; + break; + case -EDQUOT: + OSC_DUMP_GRANT(D_CACHE, cli, + "no grant space, fall back to sync i/o\n"); + break; + default: + CDEBUG(D_CACHE, "%s: event for cache space @ %p never arrived " + "due to %d, fall back to sync i/o\n", + cli->cl_import->imp_obd->obd_name, &ocw, rc); + break; } EXIT; out: spin_unlock(&cli->cl_loi_list_lock); - OSC_DUMP_GRANT(D_CACHE, cli, "returned %d.\n", rc); RETURN(rc); } @@ -1638,10 +1655,8 @@ void osc_wake_cache_waiters(struct client_obd *cli) goto wakeup; } - ocw->ocw_rc = 0; - if (!osc_enter_cache_try(cli, ocw->ocw_oap, ocw->ocw_grant, 0)) - ocw->ocw_rc = -EDQUOT; - + if (osc_enter_cache_try(cli, ocw->ocw_oap, ocw->ocw_grant, 0)) + ocw->ocw_rc = 0; wakeup: CDEBUG(D_CACHE, "wake up %p for oap %p, avail grant %ld, %d\n", ocw, ocw->ocw_oap, cli->cl_avail_grant, ocw->ocw_rc); @@ -1845,7 +1860,7 @@ static void osc_ap_completion(const struct lu_env *env, struct client_obd *cli, static int try_to_add_extent_for_io(struct client_obd *cli, struct osc_extent *ext, struct list_head *rpclist, - int *pc, unsigned int *max_pages) + unsigned int *pc, unsigned int *max_pages) { struct osc_extent *tmp; struct osc_async_page *oap = list_first_entry(&ext->oe_pages, @@ -1904,11 +1919,12 @@ static int try_to_add_extent_for_io(struct client_obd *cli, * 5. Traverse the extent tree from the 1st extent; * 6. Above steps exit if there is no space in this RPC. */ -static int get_write_extents(struct osc_object *obj, struct list_head *rpclist) +static unsigned int get_write_extents(struct osc_object *obj, + struct list_head *rpclist) { struct client_obd *cli = osc_cli(obj); struct osc_extent *ext; - int page_count = 0; + unsigned int page_count = 0; unsigned int max_pages = cli->cl_max_pages_per_rpc; LASSERT(osc_object_is_locked(obj)); @@ -1975,7 +1991,7 @@ __must_hold(osc) struct osc_extent *ext; struct osc_extent *tmp; struct osc_extent *first = NULL; - obd_count page_count = 0; + unsigned int page_count = 0; int srvlock = 0; int rc = 0; ENTRY; @@ -2048,7 +2064,7 @@ __must_hold(osc) struct osc_extent *ext; struct osc_extent *next; struct list_head rpclist = LIST_HEAD_INIT(rpclist); - int page_count = 0; + unsigned int page_count = 0; unsigned int max_pages = cli->cl_max_pages_per_rpc; int rc = 0; ENTRY; @@ -2266,7 +2282,8 @@ int osc_queue_async_io(const struct lu_env *env, struct cl_io *io, struct client_obd *cli = oap->oap_cli; struct osc_object *osc = oap->oap_obj; pgoff_t index; - int grants = 0; + unsigned int tmp; + unsigned int grants = 0; int brw_flags = OBD_BRW_ASYNC; int cmd = OBD_BRW_WRITE; int need_release = 0; @@ -2347,7 +2364,7 @@ int osc_queue_async_io(const struct lu_env *env, struct cl_io *io, grants = 0; need_release = 1; } else if (ext->oe_end < index) { - int tmp = grants; + tmp = grants; /* try to expand this extent */ rc = osc_extent_expand(ext, index, &tmp); if (rc < 0) { @@ -2372,7 +2389,7 @@ int osc_queue_async_io(const struct lu_env *env, struct cl_io *io, } if (ext == NULL) { - int tmp = (1 << cli->cl_chunkbits) + cli->cl_extent_tax; + tmp = (1 << cli->cl_chunkbits) + cli->cl_extent_tax; /* try to find new extent to cover this page */ LASSERT(oio->oi_active == NULL);