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,
{
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))
if (ext->oe_sync && ext->oe_grants > 0)
GOTO(out, rc = 90);
- if (ext->oe_dlmlock != NULL) {
+ if (ext->oe_dlmlock != NULL && !ldlm_is_failed(ext->oe_dlmlock)) {
struct ldlm_extent *extent;
extent = &ext->oe_dlmlock->l_policy_data.l_extent;
*/
static struct osc_extent *osc_extent_find(const struct lu_env *env,
struct osc_object *obj, pgoff_t index,
- int *grants)
-
+ unsigned int *grants)
{
struct client_obd *cli = osc_cli(obj);
struct osc_lock *olck;
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;
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)
int nr_pages = ext->oe_nr_pages;
int lost_grant = 0;
int blocksize = cli->cl_import->imp_obd->obd_osfs.os_bsize ? : 4096;
- __u64 last_off = 0;
+ loff_t last_off = 0;
int last_count = -1;
ENTRY;
/* For short writes we shouldn't count parts of pages that
* span a whole chunk on the OST side, or our accounting goes
* wrong. Should match the code in filter_grant_check. */
- int offset = last_off & ~CFS_PAGE_MASK;
+ int offset = last_off & ~PAGE_MASK;
int count = last_count + (offset & (blocksize - 1));
int end = (offset + last_count) & (blocksize - 1);
if (end)
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;
* 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,
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",
- osc_export(obj)->exp_obd->obd_name, state);
+ "%s: wait ext to %u timedout, recovery in progress?\n",
+ cli_name(osc_cli(obj)), state);
lwi = LWI_INTR(NULL, NULL);
rc = l_wait_event(ext->oe_waitq, extent_wait_cb(ext, state),
if (rc < 0)
GOTO(out, rc);
- /* discard all pages with index greater then trunc_index */
+ /* discard all pages with index greater than trunc_index */
list_for_each_entry_safe(oap, tmp, &ext->oe_pages,
oap_pending_item) {
pgoff_t index = osc_index(oap2osc(oap));
* 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);
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;
{
struct osc_page *opg = oap2osc_page(oap);
struct cl_page *page = oap2cl_page(oap);
- struct osc_object *obj = cl2osc(opg->ops_cl.cpl_obj);
enum cl_req_type crt;
int srvlock;
/* Clear opg->ops_transfer_pinned before VM lock is released. */
opg->ops_transfer_pinned = 0;
- spin_lock(&obj->oo_seatbelt);
- LASSERT(opg->ops_submitter != NULL);
- LASSERT(!list_empty(&opg->ops_inflight));
- list_del_init(&opg->ops_inflight);
- opg->ops_submitter = NULL;
- spin_unlock(&obj->oo_seatbelt);
-
opg->ops_submit_time = 0;
srvlock = oap->oap_brw_flags & OBD_BRW_SRVLOCK;
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;
#define OSC_DUMP_GRANT(lvl, cli, fmt, args...) do { \
struct client_obd *__tmp = (cli); \
- CDEBUG(lvl, "%s: grant { dirty: %ld/%ld dirty_pages: %ld/%lu " \
+ CDEBUG(lvl, "%s: grant { dirty: %lu/%lu dirty_pages: %ld/%lu " \
"dropped: %ld avail: %ld, reserved: %ld, flight: %d }" \
"lru {in list: %ld, left: %ld, waiters: %d }"fmt"\n", \
- __tmp->cl_import->imp_obd->obd_name, \
+ cli_name(__tmp), \
__tmp->cl_dirty_pages, __tmp->cl_dirty_max_pages, \
atomic_long_read(&obd_dirty_pages), obd_max_dirty_pages, \
__tmp->cl_lost_grant, __tmp->cl_avail_grant, \
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);
{
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)
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);
* 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.
osc_io_unplug_async(env, cli, NULL);
CDEBUG(D_CACHE, "%s: sleeping for cache space @ %p for %p\n",
- cli->cl_import->imp_obd->obd_name, &ocw, oap);
+ cli_name(cli), &ocw, oap);
rc = l_wait_event(ocw.ocw_waitq, ocw_granted(cli, &ocw), &lwi);
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_name(cli), &ocw, rc);
+ break;
}
EXIT;
out:
spin_unlock(&cli->cl_loi_list_lock);
- OSC_DUMP_GRANT(D_CACHE, cli, "returned %d.\n", rc);
RETURN(rc);
}
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);
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,
* 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));
static int
osc_send_write_rpc(const struct lu_env *env, struct client_obd *cli,
- struct osc_object *osc, pdl_policy_t pol)
+ struct osc_object *osc)
__must_hold(osc)
{
struct list_head rpclist = LIST_HEAD_INIT(rpclist);
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;
if (!list_empty(&rpclist)) {
LASSERT(page_count > 0);
- rc = osc_build_rpc(env, cli, &rpclist, OBD_BRW_WRITE, pol);
+ rc = osc_build_rpc(env, cli, &rpclist, OBD_BRW_WRITE);
LASSERT(list_empty(&rpclist));
}
*/
static int
osc_send_read_rpc(const struct lu_env *env, struct client_obd *cli,
- struct osc_object *osc, pdl_policy_t pol)
+ struct osc_object *osc)
__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;
osc_object_unlock(osc);
LASSERT(page_count > 0);
- rc = osc_build_rpc(env, cli, &rpclist, OBD_BRW_READ, pol);
+ rc = osc_build_rpc(env, cli, &rpclist, OBD_BRW_READ);
LASSERT(list_empty(&rpclist));
osc_object_lock(osc);
}
/* called with the loi list lock held */
-static void osc_check_rpcs(const struct lu_env *env, struct client_obd *cli,
- pdl_policy_t pol)
+static void osc_check_rpcs(const struct lu_env *env, struct client_obd *cli)
__must_hold(&cli->cl_loi_list_lock)
{
struct osc_object *osc;
* do io on writes while there are cache waiters */
osc_object_lock(osc);
if (osc_makes_rpc(cli, osc, OBD_BRW_WRITE)) {
- rc = osc_send_write_rpc(env, cli, osc, pol);
+ rc = osc_send_write_rpc(env, cli, osc);
if (rc < 0) {
CERROR("Write request failed with %d\n", rc);
}
}
if (osc_makes_rpc(cli, osc, OBD_BRW_READ)) {
- rc = osc_send_read_rpc(env, cli, osc, pol);
+ rc = osc_send_read_rpc(env, cli, osc);
if (rc < 0)
CERROR("Read request failed with %d\n", rc);
}
}
static int osc_io_unplug0(const struct lu_env *env, struct client_obd *cli,
- struct osc_object *osc, pdl_policy_t pol, int async)
+ struct osc_object *osc, int async)
{
int rc = 0;
return 0;
if (!async) {
- /* disable osc_lru_shrink() temporarily to avoid
- * potential stack overrun problem. LU-2859 */
- atomic_inc(&cli->cl_lru_shrinkers);
spin_lock(&cli->cl_loi_list_lock);
- osc_check_rpcs(env, cli, pol);
+ osc_check_rpcs(env, cli);
spin_unlock(&cli->cl_loi_list_lock);
- atomic_dec(&cli->cl_lru_shrinkers);
} else {
CDEBUG(D_CACHE, "Queue writeback work for client %p.\n", cli);
LASSERT(cli->cl_writeback_work != NULL);
static int osc_io_unplug_async(const struct lu_env *env,
struct client_obd *cli, struct osc_object *osc)
{
- /* XXX: policy is no use actually. */
- return osc_io_unplug0(env, cli, osc, PDL_POLICY_ROUND, 1);
+ return osc_io_unplug0(env, cli, osc, 1);
}
void osc_io_unplug(const struct lu_env *env, struct client_obd *cli,
- struct osc_object *osc, pdl_policy_t pol)
+ struct osc_object *osc)
{
- (void)osc_io_unplug0(env, cli, osc, pol, 0);
+ (void)osc_io_unplug0(env, cli, osc, 0);
}
int osc_prep_async_page(struct osc_object *osc, struct osc_page *ops,
oap->oap_page = page;
oap->oap_obj_off = offset;
- LASSERT(!(offset & ~CFS_PAGE_MASK));
+ LASSERT(!(offset & ~PAGE_MASK));
if (!client_is_remote(exp) && cfs_capable(CFS_CAP_SYS_RESOURCE))
oap->oap_brw_flags = OBD_BRW_NOQUOTA;
struct client_obd *cli = oap->oap_cli;
struct osc_object *osc = oap->oap_obj;
pgoff_t index;
- int grants = 0;
- int brw_flags = OBD_BRW_ASYNC;
+ unsigned int tmp;
+ unsigned int grants = 0;
+ u32 brw_flags = OBD_BRW_ASYNC;
int cmd = OBD_BRW_WRITE;
int need_release = 0;
int rc = 0;
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) {
}
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);
struct osc_object *obj, struct osc_page *ops)
{
struct osc_async_page *oap = &ops->ops_oap;
- struct osc_extent *ext = NULL;
int rc = 0;
ENTRY;
CDEBUG(D_INFO, "teardown oap %p page %p at index %lu.\n",
oap, ops, osc_index(oap2osc(oap)));
- osc_object_lock(obj);
if (!list_empty(&oap->oap_rpc_item)) {
CDEBUG(D_CACHE, "oap %p is not in cache.\n", oap);
rc = -EBUSY;
} else if (!list_empty(&oap->oap_pending_item)) {
+ struct osc_extent *ext = NULL;
+
+ osc_object_lock(obj);
ext = osc_extent_lookup(obj, osc_index(oap2osc(oap)));
+ osc_object_unlock(obj);
/* only truncated pages are allowed to be taken out.
* See osc_extent_truncate() and osc_cache_truncate_start()
* for details. */
osc_index(oap2osc(oap)));
rc = -EBUSY;
}
+ if (ext != NULL)
+ osc_extent_put(env, ext);
}
- osc_object_unlock(obj);
- if (ext != NULL)
- osc_extent_put(env, ext);
RETURN(rc);
}
ext = osc_extent_alloc(obj);
if (ext == NULL) {
- list_for_each_entry(oap, list, oap_pending_item) {
+ struct osc_async_page *tmp;
+
+ list_for_each_entry_safe(oap, tmp, list, oap_pending_item) {
list_del_init(&oap->oap_pending_item);
osc_ap_completion(env, cli, oap, 0, -ENOMEM);
}
/**
* Called by osc_io_setattr_start() to freeze and destroy covering extents.
*/
-int osc_cache_truncate_start(const struct lu_env *env, struct osc_io *oio,
- struct osc_object *obj, __u64 size)
+int osc_cache_truncate_start(const struct lu_env *env, struct osc_object *obj,
+ __u64 size, struct osc_extent **extp)
{
struct client_obd *cli = osc_cli(obj);
struct osc_extent *ext;
/* we need to hold this extent in OES_TRUNC state so
* that no writeback will happen. This is to avoid
- * BUG 17397. */
- LASSERT(oio->oi_trunc == NULL);
- oio->oi_trunc = osc_extent_get(ext);
+ * BUG 17397.
+ * Only partial truncate can reach here, if @size is
+ * not zero, the caller should provide a valid @extp. */
+ LASSERT(*extp == NULL);
+ *extp = osc_extent_get(ext);
OSC_EXTENT_DUMP(D_CACHE, ext,
"trunc at "LPU64"\n", size);
}
/**
* Called after osc_io_setattr_end to add oio->oi_trunc back to cache.
*/
-void osc_cache_truncate_end(const struct lu_env *env, struct osc_io *oio,
- struct osc_object *obj)
+void osc_cache_truncate_end(const struct lu_env *env, struct osc_extent *ext)
{
- struct osc_extent *ext = oio->oi_trunc;
-
- oio->oi_trunc = NULL;
if (ext != NULL) {
+ struct osc_object *obj = ext->oe_obj;
bool unplug = false;
EASSERT(ext->oe_nr_pages > 0, ext);
}
if (unplug)
- osc_io_unplug(env, osc_cli(obj), obj, PDL_POLICY_ROUND);
+ osc_io_unplug(env, osc_cli(obj), obj);
if (hp || discard) {
int rc;
struct cl_page *page = ops->ops_cl.cpl_page;
/* refresh non-overlapped index */
- tmp = osc_dlmlock_at_pgoff(env, osc, index, 0, 0);
+ tmp = osc_dlmlock_at_pgoff(env, osc, index,
+ OSC_DAP_FL_TEST_LOCK);
if (tmp != NULL) {
__u64 end = tmp->l_policy_data.l_extent.end;
/* Cache the first-non-overlapped index so as to skip
/* page is top page. */
info->oti_next_index = osc_index(ops) + 1;
if (cl_page_own(env, io, page) == 0) {
- KLASSERT(ergo(page->cp_type == CPT_CACHEABLE,
- !PageDirty(cl_page_vmpage(page))));
+ if (!ergo(page->cp_type == CPT_CACHEABLE,
+ !PageDirty(cl_page_vmpage(page))))
+ CL_PAGE_DEBUG(D_ERROR, env, page,
+ "discard dirty page?\n");
/* discard the page */
cl_page_discard(env, io, page);
git://git.whamcloud.com / fs / lustre-release.git / blobdiff