static void osc_lock_granted(const struct lu_env *env, struct osc_lock *oscl,
struct lustre_handle *lockh, bool lvb_update)
{
struct ldlm_lock *dlmlock;
dlmlock = ldlm_handle2lock_long(lockh, 0);
LASSERT(dlmlock);
/* lock reference taken by ldlm_handle2lock_long() is
* owned by osc_lock and released in osc_lock_detach()
*/
lu_ref_add(&dlmlock->l_reference, "osc_lock", oscl);
oscl->ols_has_ref = 1;
LASSERT(!oscl->ols_dlmlock);
oscl->ols_dlmlock = dlmlock;
/* This may be a matched lock for glimpse request, do not hold
* lock reference in that case.
*/
if (!oscl->ols_glimpse) {
/* hold a refc for non glimpse lock which will
* be released in osc_lock_cancel()
*/
lustre_handle_copy(&oscl->ols_handle, lockh);
ldlm_lock_addref(lockh, oscl->ols_einfo.ei_mode);
oscl->ols_hold = 1;
}
/* Lock must have been granted. */
lock_res_and_lock(dlmlock);
if (dlmlock->l_granted_mode == dlmlock->l_req_mode) {
struct ldlm_extent *ext = &dlmlock->l_policy_data.l_extent;
struct cl_lock_descr *descr = &oscl->ols_cl.cls_lock->cll_descr;
/* extend the lock extent, otherwise it will have problem when
* we decide whether to grant a lockless lock.
*/
descr->cld_mode = osc_ldlm2cl_lock(dlmlock->l_granted_mode);
descr->cld_start = cl_index(descr->cld_obj, ext->start);
descr->cld_end = cl_index(descr->cld_obj, ext->end);
descr->cld_gid = ext->gid;
/* no lvb update for matched lock */
if (lvb_update) {
LASSERT(oscl->ols_flags & LDLM_FL_LVB_READY);
osc_lock_lvb_update(env, cl2osc(oscl->ols_cl.cls_obj),
dlmlock, NULL);
}
LINVRNT(osc_lock_invariant(oscl));
}
unlock_res_and_lock(dlmlock);
LASSERT(oscl->ols_state != OLS_GRANTED);
oscl->ols_state = OLS_GRANTED;
}
static int osc_io_read_ahead(const struct lu_env *env,
const struct cl_io_slice *ios,
pgoff_t start, struct cl_read_ahead *ra)
{
struct osc_object *osc = cl2osc(ios->cis_obj);
struct ldlm_lock *dlmlock;
int result = -ENODATA;
ENTRY;
dlmlock = osc_dlmlock_at_pgoff(env, osc, start, 0);
if (dlmlock != NULL) {
if (dlmlock->l_req_mode != LCK_PR) {
struct lustre_handle lockh;
ldlm_lock2handle(dlmlock, &lockh);
ldlm_lock_addref(&lockh, LCK_PR);
ldlm_lock_decref(&lockh, dlmlock->l_req_mode);
}
ra->cra_end = cl_index(osc2cl(osc),
dlmlock->l_policy_data.l_extent.end);
ra->cra_release = osc_read_ahead_release;
ra->cra_cbdata = dlmlock;
result = 0;
}
RETURN(result);
}
/**
* Helper for osc_dlm_blocking_ast() handling discrepancies between cl_lock
* and ldlm_lock caches.
*/
static int mdc_dlm_blocking_ast0(const struct lu_env *env,
struct ldlm_lock *dlmlock,
void *data, int flag)
{
struct cl_object *obj = NULL;
int result = 0;
bool discard;
enum cl_lock_mode mode = CLM_READ;
ENTRY;
LASSERT(flag == LDLM_CB_CANCELING);
LASSERT(dlmlock != NULL);
lock_res_and_lock(dlmlock);
if (dlmlock->l_granted_mode != dlmlock->l_req_mode) {
dlmlock->l_ast_data = NULL;
unlock_res_and_lock(dlmlock);
RETURN(0);
}
discard = ldlm_is_discard_data(dlmlock);
if (dlmlock->l_granted_mode & (LCK_PW | LCK_GROUP))
mode = CLM_WRITE;
if (dlmlock->l_ast_data != NULL) {
obj = osc2cl(dlmlock->l_ast_data);
dlmlock->l_ast_data = NULL;
cl_object_get(obj);
}
ldlm_set_kms_ignore(dlmlock);
unlock_res_and_lock(dlmlock);
/* if l_ast_data is NULL, the dlmlock was enqueued by AGL or
* the object has been destroyed. */
if (obj != NULL) {
struct cl_attr *attr = &osc_env_info(env)->oti_attr;
/* Destroy pages covered by the extent of the DLM lock */
result = mdc_lock_flush(env, cl2osc(obj), cl_index(obj, 0),
CL_PAGE_EOF, mode, discard);
/* Losing a lock, set KMS to 0.
* NB: assumed that DOM lock covers whole data on MDT.
*/
/* losing a lock, update kms */
lock_res_and_lock(dlmlock);
cl_object_attr_lock(obj);
attr->cat_kms = 0;
cl_object_attr_update(env, obj, attr, CAT_KMS);
cl_object_attr_unlock(obj);
unlock_res_and_lock(dlmlock);
cl_object_put(env, obj);
}
RETURN(result);
}
/**
* Called when a lock is granted, from an upcall (when server returned a
* granted lock), or from completion AST, when server returned a blocked lock.
*
* Called under lock and resource spin-locks, that are released temporarily
* here.
*/
static void osc_lock_granted(const struct lu_env *env, struct osc_lock *olck,
struct ldlm_lock *dlmlock, int rc)
{
struct ldlm_extent *ext;
struct cl_lock *lock;
struct cl_lock_descr *descr;
LASSERT(dlmlock->l_granted_mode == dlmlock->l_req_mode);
ENTRY;
if (olck->ols_state < OLS_GRANTED) {
lock = olck->ols_cl.cls_lock;
ext = &dlmlock->l_policy_data.l_extent;
descr = &osc_env_info(env)->oti_descr;
descr->cld_obj = lock->cll_descr.cld_obj;
/* XXX check that ->l_granted_mode is valid. */
descr->cld_mode = osc_ldlm2cl_lock(dlmlock->l_granted_mode);
descr->cld_start = cl_index(descr->cld_obj, ext->start);
descr->cld_end = cl_index(descr->cld_obj, ext->end);
descr->cld_gid = ext->gid;
/*
* tell upper layers the extent of the lock that was actually
* granted
*/
olck->ols_state = OLS_GRANTED;
osc_lock_lvb_update(env, olck, rc);
/* release DLM spin-locks to allow cl_lock_{modify,signal}()
* to take a semaphore on a parent lock. This is safe, because
* spin-locks are needed to protect consistency of
* dlmlock->l_*_mode and LVB, and we have finished processing
* them. */
unlock_res_and_lock(dlmlock);
cl_lock_modify(env, lock, descr);
cl_lock_signal(env, lock);
LINVRNT(osc_lock_invariant(olck));
lock_res_and_lock(dlmlock);
}
EXIT;
}
static void osc_io_fsync_end(const struct lu_env *env,
const struct cl_io_slice *slice)
{
struct cl_fsync_io *fio = &slice->cis_io->u.ci_fsync;
struct cl_object *obj = slice->cis_obj;
pgoff_t start = cl_index(obj, fio->fi_start);
pgoff_t end = cl_index(obj, fio->fi_end);
int result = 0;
if (fio->fi_mode == CL_FSYNC_LOCAL) {
result = osc_cache_wait_range(env, cl2osc(obj), start, end);
} else if (fio->fi_mode == CL_FSYNC_ALL) {
struct osc_io *oio = cl2osc_io(env, slice);
struct osc_async_cbargs *cbargs = &oio->oi_cbarg;
wait_for_completion(&cbargs->opc_sync);
if (result == 0)
result = cbargs->opc_rc;
}
slice->cis_io->ci_result = result;
}
int lov_page_init_raid0(const struct lu_env *env, struct cl_object *obj,
struct cl_page *page, struct page *vmpage)
{
struct lov_object *loo = cl2lov(obj);
struct lov_layout_raid0 *r0 = lov_r0(loo);
struct lov_io *lio = lov_env_io(env);
struct cl_page *subpage;
struct cl_object *subobj;
struct lov_io_sub *sub;
struct lov_page *lpg = cl_object_page_slice(obj, page);
loff_t offset;
u64 suboff;
int stripe;
int rc;
offset = cl_offset(obj, page->cp_index);
stripe = lov_stripe_number(loo->lo_lsm, offset);
LASSERT(stripe < r0->lo_nr);
rc = lov_stripe_offset(loo->lo_lsm, offset, stripe,
&suboff);
LASSERT(rc == 0);
lpg->lps_invalid = 1;
cl_page_slice_add(page, &lpg->lps_cl, obj, &lov_page_ops);
sub = lov_sub_get(env, lio, stripe);
if (IS_ERR(sub)) {
rc = PTR_ERR(sub);
goto out;
}
subobj = lovsub2cl(r0->lo_sub[stripe]);
subpage = cl_page_find_sub(sub->sub_env, subobj,
cl_index(subobj, suboff), vmpage, page);
lov_sub_put(sub);
if (IS_ERR(subpage)) {
rc = PTR_ERR(subpage);
goto out;
}
if (likely(subpage->cp_parent == page)) {
lu_ref_add(&subpage->cp_reference, "lov", page);
lpg->lps_invalid = 0;
rc = 0;
} else {
CL_PAGE_DEBUG(D_ERROR, env, page, "parent page\n");
CL_PAGE_DEBUG(D_ERROR, env, subpage, "child page\n");
LASSERT(0);
}
out:
return rc;
}
static int osc_io_fsync_start(const struct lu_env *env,
const struct cl_io_slice *slice)
{
struct cl_io *io = slice->cis_io;
struct cl_fsync_io *fio = &io->u.ci_fsync;
struct cl_object *obj = slice->cis_obj;
struct osc_object *osc = cl2osc(obj);
pgoff_t start = cl_index(obj, fio->fi_start);
pgoff_t end = cl_index(obj, fio->fi_end);
int result = 0;
ENTRY;
if (fio->fi_end == OBD_OBJECT_EOF)
end = CL_PAGE_EOF;
result = osc_cache_writeback_range(env, osc, start, end, 0,
fio->fi_mode == CL_FSYNC_DISCARD);
if (result > 0) {
fio->fi_nr_written += result;
result = 0;
}
if (fio->fi_mode == CL_FSYNC_ALL) {
int rc;
/* we have to wait for writeback to finish before we can
* send OST_SYNC RPC. This is bad because it causes extents
* to be written osc by osc. However, we usually start
* writeback before CL_FSYNC_ALL so this won't have any real
* problem. */
rc = osc_cache_wait_range(env, osc, start, end);
if (result == 0)
result = rc;
rc = osc_fsync_ost(env, osc, fio);
if (result == 0)
result = rc;
}
RETURN(result);
}
static void osc_trunc_check(const struct lu_env *env, struct cl_io *io,
struct osc_io *oio, __u64 size)
{
struct cl_object *clob;
int partial;
pgoff_t start;
clob = oio->oi_cl.cis_obj;
start = cl_index(clob, size);
partial = cl_offset(clob, start) < size;
/*
* Complain if there are pages in the truncated region.
*/
cl_page_gang_lookup(env, clob, io, start + partial, CL_PAGE_EOF,
trunc_check_cb, (void *)&size);
}
ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
int rw, struct inode *inode,
struct ll_dio_pages *pv)
{
struct cl_page *clp;
struct cl_2queue *queue;
struct cl_object *obj = io->ci_obj;
int i;
ssize_t rc = 0;
loff_t file_offset = pv->ldp_start_offset;
size_t size = pv->ldp_size;
int page_count = pv->ldp_nr;
struct page **pages = pv->ldp_pages;
size_t page_size = cl_page_size(obj);
bool do_io;
int io_pages = 0;
ENTRY;
queue = &io->ci_queue;
cl_2queue_init(queue);
for (i = 0; i < page_count; i++) {
if (pv->ldp_offsets)
file_offset = pv->ldp_offsets[i];
LASSERT(!(file_offset & (page_size - 1)));
clp = cl_page_find(env, obj, cl_index(obj, file_offset),
pv->ldp_pages[i], CPT_TRANSIENT);
if (IS_ERR(clp)) {
rc = PTR_ERR(clp);
break;
}
rc = cl_page_own(env, io, clp);
if (rc) {
LASSERT(clp->cp_state == CPS_FREEING);
cl_page_put(env, clp);
break;
}
do_io = true;
/* check the page type: if the page is a host page, then do
* write directly */
if (clp->cp_type == CPT_CACHEABLE) {
struct page *vmpage = cl_page_vmpage(clp);
struct page *src_page;
struct page *dst_page;
void *src;
void *dst;
src_page = (rw == WRITE) ? pages[i] : vmpage;
dst_page = (rw == WRITE) ? vmpage : pages[i];
src = ll_kmap_atomic(src_page, KM_USER0);
dst = ll_kmap_atomic(dst_page, KM_USER1);
memcpy(dst, src, min(page_size, size));
ll_kunmap_atomic(dst, KM_USER1);
ll_kunmap_atomic(src, KM_USER0);
/* make sure page will be added to the transfer by
* cl_io_submit()->...->vvp_page_prep_write(). */
if (rw == WRITE)
set_page_dirty(vmpage);
if (rw == READ) {
/* do not issue the page for read, since it
* may reread a ra page which has NOT uptodate
* bit set. */
cl_page_disown(env, io, clp);
do_io = false;
}
}
if (likely(do_io)) {
cl_2queue_add(queue, clp);
/*
* Set page clip to tell transfer formation engine
* that page has to be sent even if it is beyond KMS.
*/
cl_page_clip(env, clp, 0, min(size, page_size));
++io_pages;
}
/* drop the reference count for cl_page_find */
cl_page_put(env, clp);
size -= page_size;
file_offset += page_size;
}
if (rc == 0 && io_pages) {
rc = cl_io_submit_sync(env, io,
rw == READ ? CRT_READ : CRT_WRITE,
queue, 0);
}
if (rc == 0)
rc = pv->ldp_size;
cl_2queue_discard(env, io, queue);
cl_2queue_disown(env, io, queue);
cl_2queue_fini(env, queue);
RETURN(rc);
}
static int llu_queue_pio(const struct lu_env *env, struct cl_io *io,
struct llu_io_group *group,
char *buf, size_t count, loff_t pos)
{
struct cl_object *obj = io->ci_obj;
struct inode *inode = ccc_object_inode(obj);
struct intnl_stat *st = llu_i2stat(inode);
struct obd_export *exp = llu_i2obdexp(inode);
struct page *page;
int rc = 0, ret_bytes = 0;
struct cl_page *clp;
struct cl_2queue *queue;
ENTRY;
if (!exp)
RETURN(-EINVAL);
queue = &io->ci_queue;
cl_2queue_init(queue);
/* prepare the pages array */
do {
unsigned long index, offset, bytes;
offset = (pos & ~CFS_PAGE_MASK);
index = pos >> PAGE_CACHE_SHIFT;
bytes = PAGE_CACHE_SIZE - offset;
if (bytes > count)
bytes = count;
/* prevent read beyond file range */
if (/* local_lock && */
io->ci_type == CIT_READ && pos + bytes >= st->st_size) {
if (pos >= st->st_size)
break;
bytes = st->st_size - pos;
}
/* prepare page for this index */
page = llu_get_user_page(index, buf - offset, offset, bytes);
if (!page) {
rc = -ENOMEM;
break;
}
clp = cl_page_find(env, obj,
cl_index(obj, pos),
page, CPT_TRANSIENT);
if (IS_ERR(clp)) {
rc = PTR_ERR(clp);
break;
}
rc = cl_page_own(env, io, clp);
if (rc) {
LASSERT(clp->cp_state == CPS_FREEING);
cl_page_put(env, clp);
break;
}
cl_2queue_add(queue, clp);
/* drop the reference count for cl_page_find, so that the page
* will be freed in cl_2queue_fini. */
cl_page_put(env, clp);
cl_page_clip(env, clp, offset, offset+bytes);
count -= bytes;
pos += bytes;
buf += bytes;
group->lig_rwcount += bytes;
ret_bytes += bytes;
page++;
} while (count);
if (rc == 0) {
enum cl_req_type iot;
iot = io->ci_type == CIT_READ ? CRT_READ : CRT_WRITE;
rc = cl_io_submit_sync(env, io, iot, queue, 0);
}
group->lig_rc = rc;
cl_2queue_discard(env, io, queue);
cl_2queue_disown(env, io, queue);
cl_2queue_fini(env, queue);
RETURN(ret_bytes);
}
static int vvp_io_fault_start(const struct lu_env *env,
const struct cl_io_slice *ios)
{
struct vvp_io *vio = cl2vvp_io(env, ios);
struct cl_io *io = ios->cis_io;
struct cl_object *obj = io->ci_obj;
struct inode *inode = ccc_object_inode(obj);
struct cl_fault_io *fio = &io->u.ci_fault;
struct vvp_fault_io *cfio = &vio->u.fault;
loff_t offset;
int result = 0;
struct page *vmpage = NULL;
struct cl_page *page;
loff_t size;
pgoff_t last; /* last page in a file data region */
if (fio->ft_executable &&
LTIME_S(inode->i_mtime) != vio->u.fault.ft_mtime)
CWARN("binary "DFID
" changed while waiting for the page fault lock\n",
PFID(lu_object_fid(&obj->co_lu)));
/* offset of the last byte on the page */
offset = cl_offset(obj, fio->ft_index + 1) - 1;
LASSERT(cl_index(obj, offset) == fio->ft_index);
result = ccc_prep_size(env, obj, io, 0, offset + 1, NULL);
if (result != 0)
return result;
/* must return locked page */
if (fio->ft_mkwrite) {
LASSERT(cfio->ft_vmpage != NULL);
lock_page(cfio->ft_vmpage);
} else {
result = vvp_io_kernel_fault(cfio);
if (result != 0)
return result;
}
vmpage = cfio->ft_vmpage;
LASSERT(PageLocked(vmpage));
if (OBD_FAIL_CHECK(OBD_FAIL_LLITE_FAULT_TRUNC_RACE))
ll_invalidate_page(vmpage);
size = i_size_read(inode);
/* Though we have already held a cl_lock upon this page, but
* it still can be truncated locally. */
if (unlikely((vmpage->mapping != inode->i_mapping) ||
(page_offset(vmpage) > size))) {
CDEBUG(D_PAGE, "llite: fault and truncate race happened!\n");
/* return +1 to stop cl_io_loop() and ll_fault() will catch
* and retry. */
GOTO(out, result = +1);
}
if (fio->ft_mkwrite ) {
pgoff_t last_index;
/*
* Capture the size while holding the lli_trunc_sem from above
* we want to make sure that we complete the mkwrite action
* while holding this lock. We need to make sure that we are
* not past the end of the file.
*/
last_index = cl_index(obj, size - 1);
if (last_index < fio->ft_index) {
CDEBUG(D_PAGE,
"llite: mkwrite and truncate race happened: "
"%p: 0x%lx 0x%lx\n",
vmpage->mapping,fio->ft_index,last_index);
/*
* We need to return if we are
* passed the end of the file. This will propagate
* up the call stack to ll_page_mkwrite where
* we will return VM_FAULT_NOPAGE. Any non-negative
* value returned here will be silently
* converted to 0. If the vmpage->mapping is null
* the error code would be converted back to ENODATA
* in ll_page_mkwrite0. Thus we return -ENODATA
* to handle both cases
*/
GOTO(out, result = -ENODATA);
}
}
page = cl_page_find(env, obj, fio->ft_index, vmpage, CPT_CACHEABLE);
if (IS_ERR(page))
GOTO(out, result = PTR_ERR(page));
/* if page is going to be written, we should add this page into cache
* earlier. */
if (fio->ft_mkwrite) {
wait_on_page_writeback(vmpage);
if (set_page_dirty(vmpage)) {
struct ccc_page *cp;
/* vvp_page_assume() calls wait_on_page_writeback(). */
cl_page_assume(env, io, page);
cp = cl2ccc_page(cl_page_at(page, &vvp_device_type));
vvp_write_pending(cl2ccc(obj), cp);
/* Do not set Dirty bit here so that in case IO is
* started before the page is really made dirty, we
* still have chance to detect it. */
result = cl_page_cache_add(env, io, page, CRT_WRITE);
LASSERT(cl_page_is_owned(page, io));
vmpage = NULL;
if (result < 0) {
cl_page_unmap(env, io, page);
cl_page_discard(env, io, page);
cl_page_disown(env, io, page);
cl_page_put(env, page);
/* we're in big trouble, what can we do now? */
if (result == -EDQUOT)
result = -ENOSPC;
GOTO(out, result);
} else
cl_page_disown(env, io, page);
}
}
last = cl_index(obj, size - 1);
/*
* The ft_index is only used in the case of
* a mkwrite action. We need to check
* our assertions are correct, since
* we should have caught this above
*/
LASSERT(!fio->ft_mkwrite || fio->ft_index <= last);
if (fio->ft_index == last)
/*
* Last page is mapped partially.
*/
fio->ft_nob = size - cl_offset(obj, fio->ft_index);
else
fio->ft_nob = cl_page_size(obj);
lu_ref_add(&page->cp_reference, "fault", io);
fio->ft_page = page;
EXIT;
out:
/* return unlocked vmpage to avoid deadlocking */
if (vmpage != NULL)
unlock_page(vmpage);
cfio->fault.ft_flags &= ~VM_FAULT_LOCKED;
return result;
}
static int vvp_io_read_start(const struct lu_env *env,
const struct cl_io_slice *ios)
{
struct vvp_io *vio = cl2vvp_io(env, ios);
struct ccc_io *cio = cl2ccc_io(env, ios);
struct cl_io *io = ios->cis_io;
struct cl_object *obj = io->ci_obj;
struct inode *inode = ccc_object_inode(obj);
struct ll_ra_read *bead = &vio->cui_bead;
struct file *file = cio->cui_fd->fd_file;
int result;
loff_t pos = io->u.ci_rd.rd.crw_pos;
long cnt = io->u.ci_rd.rd.crw_count;
long tot = cio->cui_tot_count;
int exceed = 0;
CLOBINVRNT(env, obj, ccc_object_invariant(obj));
CDEBUG(D_VFSTRACE, "read: -> [%lli, %lli)\n", pos, pos + cnt);
if (!can_populate_pages(env, io, inode))
return 0;
result = ccc_prep_size(env, obj, io, pos, tot, &exceed);
if (result != 0)
return result;
else if (exceed != 0)
goto out;
LU_OBJECT_HEADER(D_INODE, env, &obj->co_lu,
"Read ino %lu, %lu bytes, offset %lld, size %llu\n",
inode->i_ino, cnt, pos, i_size_read(inode));
/* turn off the kernel's read-ahead */
cio->cui_fd->fd_file->f_ra.ra_pages = 0;
/* initialize read-ahead window once per syscall */
if (!vio->cui_ra_window_set) {
vio->cui_ra_window_set = 1;
bead->lrr_start = cl_index(obj, pos);
bead->lrr_count = cl_index(obj, tot + PAGE_CACHE_SIZE - 1);
ll_ra_read_in(file, bead);
}
/* BUG: 5972 */
file_accessed(file);
switch (vio->cui_io_subtype) {
case IO_NORMAL:
result = lustre_generic_file_read(file, cio, &pos);
break;
case IO_SPLICE:
result = generic_file_splice_read(file, &pos,
vio->u.splice.cui_pipe, cnt,
vio->u.splice.cui_flags);
/* LU-1109: do splice read stripe by stripe otherwise if it
* may make nfsd stuck if this read occupied all internal pipe
* buffers. */
io->ci_continue = 0;
break;
default:
CERROR("Wrong IO type %u\n", vio->cui_io_subtype);
LBUG();
}
out:
if (result >= 0) {
if (result < cnt)
io->ci_continue = 0;
io->ci_nob += result;
ll_rw_stats_tally(ll_i2sbi(inode), current->pid, cio->cui_fd,
pos, result, READ);
result = 0;
}
return result;
}
static int vvp_mmap_locks(const struct lu_env *env,
struct ccc_io *vio, struct cl_io *io)
{
struct ccc_thread_info *cti = ccc_env_info(env);
struct mm_struct *mm = current->mm;
struct vm_area_struct *vma;
struct cl_lock_descr *descr = &cti->cti_descr;
ldlm_policy_data_t policy;
unsigned long addr;
unsigned long seg;
ssize_t count;
int result;
ENTRY;
LASSERT(io->ci_type == CIT_READ || io->ci_type == CIT_WRITE);
if (!cl_is_normalio(env, io))
RETURN(0);
if (vio->cui_iov == NULL) /* nfs or loop back device write */
RETURN(0);
/* No MM (e.g. NFS)? No vmas too. */
if (mm == NULL)
RETURN(0);
for (seg = 0; seg < vio->cui_nrsegs; seg++) {
const struct iovec *iv = &vio->cui_iov[seg];
addr = (unsigned long)iv->iov_base;
count = iv->iov_len;
if (count == 0)
continue;
count += addr & (~CFS_PAGE_MASK);
addr &= CFS_PAGE_MASK;
down_read(&mm->mmap_sem);
while((vma = our_vma(mm, addr, count)) != NULL) {
struct inode *inode = vma->vm_file->f_dentry->d_inode;
int flags = CEF_MUST;
if (ll_file_nolock(vma->vm_file)) {
/*
* For no lock case, a lockless lock will be
* generated.
*/
flags = CEF_NEVER;
}
/*
* XXX: Required lock mode can be weakened: CIT_WRITE
* io only ever reads user level buffer, and CIT_READ
* only writes on it.
*/
policy_from_vma(&policy, vma, addr, count);
descr->cld_mode = vvp_mode_from_vma(vma);
descr->cld_obj = ll_i2info(inode)->lli_clob;
descr->cld_start = cl_index(descr->cld_obj,
policy.l_extent.start);
descr->cld_end = cl_index(descr->cld_obj,
policy.l_extent.end);
descr->cld_enq_flags = flags;
result = cl_io_lock_alloc_add(env, io, descr);
CDEBUG(D_VFSTRACE, "lock: %d: [%lu, %lu]\n",
descr->cld_mode, descr->cld_start,
descr->cld_end);
if (result < 0)
RETURN(result);
if (vma->vm_end - addr >= count)
break;
count -= vma->vm_end - addr;
addr = vma->vm_end;
}
up_read(&mm->mmap_sem);
}
RETURN(0);
}
/**
* Creates sub-locks for a given lov_lock for the first time.
*
* Goes through all sub-objects of top-object, and creates sub-locks on every
* sub-object intersecting with top-lock extent. This is complicated by the
* fact that top-lock (that is being created) can be accessed concurrently
* through already created sub-locks (possibly shared with other top-locks).
*/
static struct lov_lock *lov_lock_sub_init(const struct lu_env *env,
const struct cl_object *obj,
struct cl_lock *lock)
{
int result = 0;
int i;
int nr;
u64 start;
u64 end;
u64 file_start;
u64 file_end;
struct lov_object *loo = cl2lov(obj);
struct lov_layout_raid0 *r0 = lov_r0(loo);
struct lov_lock *lovlck;
file_start = cl_offset(lov2cl(loo), lock->cll_descr.cld_start);
file_end = cl_offset(lov2cl(loo), lock->cll_descr.cld_end + 1) - 1;
for (i = 0, nr = 0; i < r0->lo_nr; i++) {
/*
* XXX for wide striping smarter algorithm is desirable,
* breaking out of the loop, early.
*/
if (likely(r0->lo_sub[i]) && /* spare layout */
lov_stripe_intersects(loo->lo_lsm, i,
file_start, file_end, &start, &end))
nr++;
}
LASSERT(nr > 0);
lovlck = libcfs_kvzalloc(offsetof(struct lov_lock, lls_sub[nr]),
GFP_NOFS);
if (!lovlck)
return ERR_PTR(-ENOMEM);
lovlck->lls_nr = nr;
for (i = 0, nr = 0; i < r0->lo_nr; ++i) {
if (likely(r0->lo_sub[i]) &&
lov_stripe_intersects(loo->lo_lsm, i,
file_start, file_end, &start, &end)) {
struct lov_lock_sub *lls = &lovlck->lls_sub[nr];
struct cl_lock_descr *descr;
descr = &lls->sub_lock.cll_descr;
LASSERT(!descr->cld_obj);
descr->cld_obj = lovsub2cl(r0->lo_sub[i]);
descr->cld_start = cl_index(descr->cld_obj, start);
descr->cld_end = cl_index(descr->cld_obj, end);
descr->cld_mode = lock->cll_descr.cld_mode;
descr->cld_gid = lock->cll_descr.cld_gid;
descr->cld_enq_flags = lock->cll_descr.cld_enq_flags;
lls->sub_stripe = i;
/* initialize sub lock */
result = lov_sublock_init(env, lock, lls);
if (result < 0)
break;
lls->sub_initialized = 1;
nr++;
}
}
LASSERT(ergo(result == 0, nr == lovlck->lls_nr));
if (result != 0) {
for (i = 0; i < nr; ++i) {
if (!lovlck->lls_sub[i].sub_initialized)
break;
cl_lock_fini(env, &lovlck->lls_sub[i].sub_lock);
}
kvfree(lovlck);
lovlck = ERR_PTR(result);
}
return lovlck;
}
/**
* Helper for osc_dlm_blocking_ast() handling discrepancies between cl_lock
* and ldlm_lock caches.
*/
static int osc_dlm_blocking_ast0(const struct lu_env *env,
struct ldlm_lock *dlmlock,
void *data, int flag)
{
struct cl_object *obj = NULL;
int result = 0;
int discard;
enum cl_lock_mode mode = CLM_READ;
LASSERT(flag == LDLM_CB_CANCELING);
lock_res_and_lock(dlmlock);
if (dlmlock->l_granted_mode != dlmlock->l_req_mode) {
dlmlock->l_ast_data = NULL;
unlock_res_and_lock(dlmlock);
return 0;
}
discard = ldlm_is_discard_data(dlmlock);
if (dlmlock->l_granted_mode & (LCK_PW | LCK_GROUP))
mode = CLM_WRITE;
if (dlmlock->l_ast_data) {
obj = osc2cl(dlmlock->l_ast_data);
dlmlock->l_ast_data = NULL;
cl_object_get(obj);
}
unlock_res_and_lock(dlmlock);
/* if l_ast_data is NULL, the dlmlock was enqueued by AGL or
* the object has been destroyed.
*/
if (obj) {
struct ldlm_extent *extent = &dlmlock->l_policy_data.l_extent;
struct cl_attr *attr = &osc_env_info(env)->oti_attr;
__u64 old_kms;
/* Destroy pages covered by the extent of the DLM lock */
result = osc_lock_flush(cl2osc(obj),
cl_index(obj, extent->start),
cl_index(obj, extent->end),
mode, discard);
/* losing a lock, update kms */
lock_res_and_lock(dlmlock);
cl_object_attr_lock(obj);
/* Must get the value under the lock to avoid race. */
old_kms = cl2osc(obj)->oo_oinfo->loi_kms;
/* Update the kms. Need to loop all granted locks.
* Not a problem for the client
*/
attr->cat_kms = ldlm_extent_shift_kms(dlmlock, old_kms);
cl_object_attr_set(env, obj, attr, CAT_KMS);
cl_object_attr_unlock(obj);
unlock_res_and_lock(dlmlock);
cl_object_put(env, obj);
}
return result;
}
