*
* You should have received a copy of the GNU General Public License
* version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
+ * http://www.gnu.org/licenses/gpl-2.0.html
*
* GPL HEADER END
*/
* Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2015, Intel Corporation.
+ * Copyright (c) 2011, 2017, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
#define DEBUG_SUBSYSTEM S_LLITE
-
#include <obd.h>
+#include <linux/pagevec.h>
+#include <linux/memcontrol.h>
+
#include "llite_internal.h"
#include "vvp_internal.h"
+#include <libcfs/linux/linux-misc.h>
static struct vvp_io *cl2vvp_io(const struct lu_env *env,
const struct cl_io_slice *slice)
* --bug 17336 */
loff_t size = i_size_read(inode);
unsigned long cur_index = start >>
- PAGE_CACHE_SHIFT;
+ PAGE_SHIFT;
if ((size == 0 && cur_index != 0) ||
- (((size - 1) >> PAGE_CACHE_SHIFT) <
+ (((size - 1) >> PAGE_SHIFT) <
cur_index))
*exceed = 1;
}
if (i_size_read(inode) < kms) {
i_size_write(inode, kms);
CDEBUG(D_VFSTRACE,
- DFID" updating i_size "LPU64"\n",
+ DFID" updating i_size %llu\n",
PFID(lu_object_fid(&obj->co_lu)),
(__u64)i_size_read(inode));
}
struct vvp_io *vio = cl2vvp_io(env, ios);
struct inode *inode = vvp_object_inode(ios->cis_obj);
- LASSERT(inode == vio->vui_fd->fd_file->f_path.dentry->d_inode);
- vio->u.fault.ft_mtime = LTIME_S(inode->i_mtime);
+ LASSERT(inode == file_inode(vio->vui_fd->fd_file));
return 0;
}
struct cl_object *obj = io->ci_obj;
struct vvp_io *vio = cl2vvp_io(env, ios);
struct inode *inode = vvp_object_inode(obj);
+ __u32 gen = 0;
+ int rc;
+ ENTRY;
CLOBINVRNT(env, obj, vvp_object_invariant(obj));
CDEBUG(D_VFSTRACE, DFID" ignore/verify layout %d/%d, layout version %d "
- "restore needed %d\n",
+ "need write layout %d, restore needed %d\n",
PFID(lu_object_fid(&obj->co_lu)),
io->ci_ignore_layout, io->ci_verify_layout,
- vio->vui_layout_gen, io->ci_restore_needed);
+ vio->vui_layout_gen, io->ci_need_write_intent,
+ io->ci_restore_needed);
if (io->ci_restore_needed) {
- int rc;
-
/* file was detected release, we need to restore it
* before finishing the io
*/
* block on layout lock held by the MDT
* as MDT will not send new layout in lvb (see LU-3124)
* we have to explicitly fetch it, all this will be done
- * by ll_layout_refresh()
+ * by ll_layout_refresh().
+ * Even if ll_layout_restore() returns zero, it doesn't mean
+ * that restore has been successful. Therefore it sets
+ * ci_verify_layout so that it will check layout at the end
+ * of this function.
*/
- if (rc == 0) {
- io->ci_restore_needed = 0;
- io->ci_need_restart = 1;
- io->ci_verify_layout = 1;
- } else {
+ if (rc) {
io->ci_restore_needed = 1;
io->ci_need_restart = 0;
io->ci_verify_layout = 0;
io->ci_result = rc;
+ GOTO(out, rc);
}
- }
- if (!io->ci_ignore_layout && io->ci_verify_layout) {
- __u32 gen = 0;
+ io->ci_restore_needed = 0;
- /* check layout version */
+ /* Even if ll_layout_restore() returns zero, it doesn't mean
+ * that restore has been successful. Therefore it should verify
+ * if there was layout change and restart I/O correspondingly.
+ */
ll_layout_refresh(inode, &gen);
io->ci_need_restart = vio->vui_layout_gen != gen;
if (io->ci_need_restart) {
ll_file_clear_flag(ll_i2info(vvp_object_inode(obj)),
LLIF_FILE_RESTORING);
}
+ GOTO(out, 0);
}
+
+ /**
+ * dynamic layout change needed, send layout intent
+ * RPC.
+ */
+ if (io->ci_need_write_intent) {
+ enum layout_intent_opc opc = LAYOUT_INTENT_WRITE;
+
+ io->ci_need_write_intent = 0;
+
+ LASSERT(io->ci_type == CIT_WRITE ||
+ cl_io_is_trunc(io) || cl_io_is_mkwrite(io));
+
+ CDEBUG(D_VFSTRACE, DFID" write layout, type %u "DEXT"\n",
+ PFID(lu_object_fid(&obj->co_lu)), io->ci_type,
+ PEXT(&io->ci_write_intent));
+
+ if (cl_io_is_trunc(io))
+ opc = LAYOUT_INTENT_TRUNC;
+
+ rc = ll_layout_write_intent(inode, opc, &io->ci_write_intent);
+ io->ci_result = rc;
+ if (!rc)
+ io->ci_need_restart = 1;
+ GOTO(out, rc);
+ }
+
+ if (!io->ci_need_restart &&
+ !io->ci_ignore_layout && io->ci_verify_layout) {
+ /* check layout version */
+ ll_layout_refresh(inode, &gen);
+ io->ci_need_restart = vio->vui_layout_gen != gen;
+ if (io->ci_need_restart) {
+ CDEBUG(D_VFSTRACE,
+ DFID" layout changed from %d to %d.\n",
+ PFID(lu_object_fid(&obj->co_lu)),
+ vio->vui_layout_gen, gen);
+ }
+ GOTO(out, 0);
+ }
+out:
+ EXIT;
}
static void vvp_io_fault_fini(const struct lu_env *env,
struct vm_area_struct *vma;
struct cl_lock_descr *descr = &vti->vti_descr;
union ldlm_policy_data policy;
-#ifdef HAVE_FILE_OPERATIONS_READ_WRITE_ITER
struct iovec iov;
struct iov_iter i;
-#else
- unsigned long seg;
-#endif
+ unsigned long addr;
+ ssize_t count;
int result = 0;
ENTRY;
if (mm == NULL)
RETURN(0);
-#ifdef HAVE_FILE_OPERATIONS_READ_WRITE_ITER
- iov_for_each(iov, i, *(vio->vui_iter)) {
-#else /* !HAVE_FILE_OPERATIONS_READ_WRITE_ITER */
- for (seg = 0; seg < vio->vui_iter->nr_segs; seg++) {
- const struct iovec iov = vio->vui_iter->iov[seg];
-#endif /* HAVE_FILE_OPERATIONS_READ_WRITE_ITER */
- unsigned long addr = (unsigned long)iov.iov_base;
- size_t count = iov.iov_len;
+ if (!iter_is_iovec(vio->vui_iter) && !iov_iter_is_kvec(vio->vui_iter))
+ RETURN(0);
+
+ for (i = *vio->vui_iter;
+ iov_iter_count(&i);
+ iov_iter_advance(&i, iov.iov_len)) {
+ iov = iov_iter_iovec(&i);
+ addr = (unsigned long)iov.iov_base;
+ count = iov.iov_len;
if (count == 0)
continue;
down_read(&mm->mmap_sem);
while((vma = our_vma(mm, addr, count)) != NULL) {
- struct dentry *de = vma->vm_file->f_path.dentry;
+ struct dentry *de = file_dentry(vma->vm_file);
struct inode *inode = de->d_inode;
int flags = CEF_MUST;
struct vvp_io *vio = cl2vvp_io(env, ios);
struct cl_io *io = ios->cis_io;
struct cl_object *obj = ios->cis_io->ci_obj;
-#ifndef HAVE_FILE_OPERATIONS_READ_WRITE_ITER
- struct iovec *iov;
-#endif
+
CLOBINVRNT(env, obj, vvp_object_invariant(obj));
if (!cl_is_normalio(env, io))
return;
-#ifdef HAVE_FILE_OPERATIONS_READ_WRITE_ITER
vio->vui_tot_count -= nob;
iov_iter_reexpand(vio->vui_iter, vio->vui_tot_count);
-#else /* !HAVE_FILE_OPERATIONS_READ_WRITE_ITER */
- LASSERT(vio->vui_tot_nrsegs >= vio->vui_iter->nr_segs);
- LASSERT(vio->vui_tot_count >= nob);
-
- /* Restore the iov changed in vvp_io_update_iov() */
- if (vio->vui_iov_olen > 0) {
- unsigned long idx = vio->vui_iter->nr_segs - 1;
-
- /* In the latest kernels iov is const so that
- * changes are done using iter helpers. In older
- * kernels those helpers don't exist we lustre
- * has to do some of the management of the iter
- * itself. */
- iov = (struct iovec *)&vio->vui_iter->iov[idx];
- iov->iov_len = vio->vui_iov_olen;
- vio->vui_iov_olen = 0;
- }
-
- /* In the latest kernels special helpers exist to help
- * advance the iov but we don't have that in older kernels
- * so we need to do the book keeping ourselves. */
- iov = (struct iovec *)vio->vui_iter->iov;
- while (nob > 0) {
- if (iov->iov_len > nob) {
- iov->iov_len -= nob;
- iov->iov_base += nob;
- break;
- }
-
- nob -= iov->iov_len;
- iov++;
- vio->vui_tot_nrsegs--;
- }
-
- vio->vui_iter->iov = iov;
- vio->vui_tot_count -= nob;
-#endif /* HAVE_FILE_OPERATIONS_READ_WRITE_ITER */
}
static void vvp_io_update_iov(const struct lu_env *env,
struct vvp_io *vio, struct cl_io *io)
{
size_t size = io->u.ci_rw.crw_count;
-#ifdef HAVE_FILE_OPERATIONS_READ_WRITE_ITER
+
if (!cl_is_normalio(env, io) || vio->vui_iter == NULL)
return;
iov_iter_truncate(vio->vui_iter, size);
-#else /* !HAVE_FILE_OPERATIONS_READ_WRITE_ITER */
- unsigned long i;
-
- vio->vui_iov_olen = 0;
- if (!cl_is_normalio(env, io) || vio->vui_tot_nrsegs == 0)
- return;
-
- for (i = 0; i < vio->vui_tot_nrsegs; i++) {
- struct iovec *iv = (struct iovec *) &vio->vui_iter->iov[i];
-
- if (iv->iov_len < size) {
- size -= iv->iov_len;
- } else {
- if (iv->iov_len > size) {
- vio->vui_iov_olen = iv->iov_len;
- iv->iov_len = size;
- }
- break;
- }
- }
-
- vio->vui_iter->nr_segs = i + 1;
- LASSERTF(vio->vui_tot_nrsegs >= vio->vui_iter->nr_segs,
- "tot_nrsegs: %lu, nrsegs: %lu\n",
- vio->vui_tot_nrsegs, vio->vui_iter->nr_segs);
-#endif /* !HAVE_FILE_OPERATIONS_READ_WRITE_ITER */
}
static int vvp_io_rw_lock(const struct lu_env *env, struct cl_io *io,
if (io->u.ci_rw.crw_nonblock)
ast_flags |= CEF_NONBLOCK;
+ if (io->ci_lock_no_expand)
+ ast_flags |= CEF_LOCK_NO_EXPAND;
+ if (vio->vui_fd) {
+ /* Group lock held means no lockless any more */
+ if (vio->vui_fd->fd_flags & LL_FILE_GROUP_LOCKED)
+ io->ci_ignore_lockless = 1;
+
+ if (ll_file_nolock(vio->vui_fd->fd_file) ||
+ (vio->vui_fd->fd_flags & LL_FILE_LOCKLESS_IO &&
+ !io->ci_ignore_lockless))
+ ast_flags |= CEF_NEVER;
+ }
result = vvp_mmap_locks(env, vio, io);
if (result == 0)
result = vvp_io_one_lock(env, io, ast_flags, mode, start, end);
- RETURN(result);
+ RETURN(result);
}
static int vvp_io_read_lock(const struct lu_env *env,
const struct cl_io_slice *ios)
{
- struct cl_io *io = ios->cis_io;
- struct cl_io_rw_common *rd = &io->u.ci_rd.rd;
+ struct cl_io *io = ios->cis_io;
+ struct cl_io_rw_common *rd = &io->u.ci_rd.rd;
int result;
ENTRY;
}
static int vvp_io_write_lock(const struct lu_env *env,
- const struct cl_io_slice *ios)
+ const struct cl_io_slice *ios)
{
- struct cl_io *io = ios->cis_io;
- loff_t start;
- loff_t end;
-
- if (io->u.ci_wr.wr_append) {
- start = 0;
- end = OBD_OBJECT_EOF;
- } else {
- start = io->u.ci_wr.wr.crw_pos;
- end = start + io->u.ci_wr.wr.crw_count - 1;
- }
- return vvp_io_rw_lock(env, io, CLM_WRITE, start, end);
+ struct cl_io *io = ios->cis_io;
+ loff_t start;
+ loff_t end;
+
+ if (io->u.ci_wr.wr_append) {
+ start = 0;
+ end = OBD_OBJECT_EOF;
+ } else {
+ start = io->u.ci_wr.wr.crw_pos;
+ end = start + io->u.ci_wr.wr.crw_count - 1;
+ }
+
+ RETURN(vvp_io_rw_lock(env, io, CLM_WRITE, start, end));
}
static int vvp_io_setattr_iter_init(const struct lu_env *env,
const struct cl_io_slice *ios)
+
{
return 0;
}
const struct cl_io_slice *ios)
{
struct cl_io *io = ios->cis_io;
- __u64 new_size;
+ __u64 lock_start = 0;
+ __u64 lock_end = OBD_OBJECT_EOF;
__u32 enqflags = 0;
- if (cl_io_is_trunc(io)) {
- new_size = io->u.ci_setattr.sa_attr.lvb_size;
- if (new_size == 0)
- enqflags = CEF_DISCARD_DATA;
- } else {
- unsigned int valid = io->u.ci_setattr.sa_valid;
+ if (cl_io_is_trunc(io)) {
+ if (io->u.ci_setattr.sa_attr.lvb_size == 0)
+ enqflags = CEF_DISCARD_DATA;
+ } else if (cl_io_is_fallocate(io)) {
+ lock_start = io->u.ci_setattr.sa_falloc_offset;
+ lock_end = lock_start + io->u.ci_setattr.sa_attr.lvb_size;
+ } else {
+ unsigned int valid = io->u.ci_setattr.sa_avalid;
if (!(valid & TIMES_SET_FLAGS))
return 0;
io->u.ci_setattr.sa_attr.lvb_atime >=
io->u.ci_setattr.sa_attr.lvb_ctime))
return 0;
-
- new_size = 0;
}
return vvp_io_one_lock(env, io, enqflags, CLM_WRITE,
- new_size, OBD_OBJECT_EOF);
+ lock_start, lock_end);
}
static int vvp_do_vmtruncate(struct inode *inode, size_t size)
int result;
unsigned valid = CAT_CTIME;
- cl_object_attr_lock(obj);
- attr->cat_ctime = io->u.ci_setattr.sa_attr.lvb_ctime;
- if (io->u.ci_setattr.sa_valid & ATTR_ATIME_SET) {
- attr->cat_atime = io->u.ci_setattr.sa_attr.lvb_atime;
- valid |= CAT_ATIME;
- }
- if (io->u.ci_setattr.sa_valid & ATTR_MTIME_SET) {
- attr->cat_mtime = io->u.ci_setattr.sa_attr.lvb_mtime;
- valid |= CAT_MTIME;
- }
+ cl_object_attr_lock(obj);
+ attr->cat_ctime = io->u.ci_setattr.sa_attr.lvb_ctime;
+ if (io->u.ci_setattr.sa_avalid & ATTR_ATIME_SET) {
+ attr->cat_atime = io->u.ci_setattr.sa_attr.lvb_atime;
+ valid |= CAT_ATIME;
+ }
+ if (io->u.ci_setattr.sa_avalid & ATTR_MTIME_SET) {
+ attr->cat_mtime = io->u.ci_setattr.sa_attr.lvb_mtime;
+ valid |= CAT_MTIME;
+ }
result = cl_object_attr_update(env, obj, attr, valid);
cl_object_attr_unlock(obj);
struct inode *inode = vvp_object_inode(io->ci_obj);
struct ll_inode_info *lli = ll_i2info(inode);
- mutex_lock(&inode->i_mutex);
if (cl_io_is_trunc(io)) {
- down_write(&lli->lli_trunc_sem);
+ trunc_sem_down_write(&lli->lli_trunc_sem);
+ inode_lock(inode);
inode_dio_wait(inode);
+ } else if (cl_io_is_fallocate(io)) {
+ inode_lock(inode);
+ inode_dio_wait(inode);
+ } else {
+ inode_lock(inode);
}
- if (io->u.ci_setattr.sa_valid & TIMES_SET_FLAGS)
+ if (io->u.ci_setattr.sa_avalid & TIMES_SET_FLAGS)
return vvp_io_setattr_time(env, ios);
return 0;
/* Truncate in memory pages - they must be clean pages
* because osc has already notified to destroy osc_extents. */
vvp_do_vmtruncate(inode, io->u.ci_setattr.sa_attr.lvb_size);
- inode_dio_write_done(inode);
- up_write(&lli->lli_trunc_sem);
+ inode_unlock(inode);
+ trunc_sem_up_write(&lli->lli_trunc_sem);
+ } else if (cl_io_is_fallocate(io)) {
+ inode_unlock(inode);
+ } else {
+ inode_unlock(inode);
}
- mutex_unlock(&inode->i_mutex);
}
static void vvp_io_setattr_fini(const struct lu_env *env,
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 cl_io *io = ios->cis_io;
- struct cl_object *obj = io->ci_obj;
- struct inode *inode = vvp_object_inode(obj);
- struct ll_inode_info *lli = ll_i2info(inode);
- struct file *file = vio->vui_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 = vio->vui_tot_count;
- int exceed = 0;
+ 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 = vvp_object_inode(obj);
+ struct ll_inode_info *lli = ll_i2info(inode);
+ struct file *file = vio->vui_fd->fd_file;
+ loff_t pos = io->u.ci_rd.rd.crw_pos;
+ size_t cnt = io->u.ci_rd.rd.crw_count;
+ size_t tot = vio->vui_tot_count;
+ int exceed = 0;
+ int result;
+ ENTRY;
CLOBINVRNT(env, obj, vvp_object_invariant(obj));
- CDEBUG(D_VFSTRACE, "read: -> [%lli, %lli)\n", pos, pos + cnt);
+ CDEBUG(D_VFSTRACE, "%s: read [%llu, %llu)\n",
+ file_dentry(file)->d_name.name,
+ pos, pos + cnt);
if (vio->vui_io_subtype == IO_NORMAL)
- down_read(&lli->lli_trunc_sem);
+ trunc_sem_down_read(&lli->lli_trunc_sem);
+
+ if (io->ci_async_readahead) {
+ file_accessed(file);
+ RETURN(0);
+ }
if (!can_populate_pages(env, io, inode))
- return 0;
+ RETURN(0);
- result = vvp_prep_size(env, obj, io, pos, tot, &exceed);
+ /* Unless this is reading a sparse file, otherwise the lock has already
+ * been acquired so vvp_prep_size() is an empty op. */
+ result = vvp_prep_size(env, obj, io, pos, cnt, &exceed);
if (result != 0)
- return result;
+ RETURN(result);
else if (exceed != 0)
- goto out;
+ GOTO(out, result);
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));
+ "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 */
vio->vui_fd->fd_file->f_ra.ra_pages = 0;
/* initialize read-ahead window once per syscall */
if (!vio->vui_ra_valid) {
vio->vui_ra_valid = true;
- vio->vui_ra_start = cl_index(obj, pos);
- vio->vui_ra_count = cl_index(obj, tot + PAGE_CACHE_SIZE - 1);
- ll_ras_enter(file);
+ vio->vui_ra_start_idx = cl_index(obj, pos);
+ vio->vui_ra_pages = cl_index(obj, tot + PAGE_SIZE - 1);
+ /* If both start and end are unaligned, we read one more page
+ * than the index math suggests. */
+ if ((pos & ~PAGE_MASK) != 0 && ((pos + tot) & ~PAGE_MASK) != 0)
+ vio->vui_ra_pages++;
+
+ CDEBUG(D_READA, "tot %zu, ra_start %lu, ra_count %lu\n",
+ tot, vio->vui_ra_start_idx, vio->vui_ra_pages);
}
/* BUG: 5972 */
switch (vio->vui_io_subtype) {
case IO_NORMAL:
LASSERT(vio->vui_iocb->ki_pos == pos);
-#ifdef HAVE_FILE_OPERATIONS_READ_WRITE_ITER
result = generic_file_read_iter(vio->vui_iocb, vio->vui_iter);
-#else
- result = generic_file_aio_read(vio->vui_iocb,
- vio->vui_iter->iov,
- vio->vui_iter->nr_segs,
- vio->vui_iocb->ki_pos);
-#endif
break;
case IO_SPLICE:
result = generic_file_splice_read(file, &pos,
CERROR("Wrong IO type %u\n", vio->vui_io_subtype);
LBUG();
}
+ GOTO(out, result);
out:
if (result >= 0) {
if (result < cnt)
io->ci_continue = 0;
io->ci_nob += result;
- ll_rw_stats_tally(ll_i2sbi(inode), current->pid, vio->vui_fd,
- pos, result, READ);
result = 0;
}
RETURN(bytes > 0 ? bytes : rc);
}
+/*
+ * Kernels 4.2 - 4.5 pass memcg argument to account_page_dirtied()
+ * Kernel v5.2-5678-gac1c3e4 no longer exports account_page_dirtied
+ */
+static inline void ll_account_page_dirtied(struct page *page,
+ struct address_space *mapping)
+{
+#ifdef HAVE_ACCOUNT_PAGE_DIRTIED_3ARGS
+ struct mem_cgroup *memcg = mem_cgroup_begin_page_stat(page);
+
+ account_page_dirtied(page, mapping, memcg);
+ mem_cgroup_end_page_stat(memcg);
+#elif defined(HAVE_ACCOUNT_PAGE_DIRTIED_EXPORT)
+ account_page_dirtied(page, mapping);
+#else
+ vvp_account_page_dirtied(page, mapping);
+#endif
+}
+
+/*
+ * From kernel v4.19-rc5-248-g9b89a0355144 use XArrary
+ * Prior kernels use radix_tree for tags
+ */
+static inline void ll_page_tag_dirty(struct page *page,
+ struct address_space *mapping)
+{
+#ifndef HAVE_RADIX_TREE_TAG_SET
+ __xa_set_mark(&mapping->i_pages, page_index(page), PAGECACHE_TAG_DIRTY);
+#else
+ radix_tree_tag_set(&mapping->page_tree, page_index(page),
+ PAGECACHE_TAG_DIRTY);
+#endif
+}
+
+/* Taken from kernel set_page_dirty, __set_page_dirty_nobuffers
+ * Last change to this area: b93b016313b3ba8003c3b8bb71f569af91f19fc7
+ *
+ * Current with Linus tip of tree (7/13/2019):
+ * v5.2-rc4-224-ge01e060fe0
+ *
+ * Backwards compat for 3.x, 5.x kernels relating to memcg handling
+ * & rename of radix tree to xarray.
+ */
+void vvp_set_pagevec_dirty(struct pagevec *pvec)
+{
+ struct page *page = pvec->pages[0];
+ struct address_space *mapping = page->mapping;
+ unsigned long flags;
+ int count = pagevec_count(pvec);
+ int dirtied = 0;
+ int i = 0;
+
+ ENTRY;
+
+ /* From set_page_dirty */
+ for (i = 0; i < count; i++)
+ ClearPageReclaim(pvec->pages[i]);
+
+ LASSERTF(page->mapping,
+ "mapping must be set. page %p, page->private (cl_page) %p\n",
+ page, (void *) page->private);
+
+ /* Rest of code derived from __set_page_dirty_nobuffers */
+ ll_xa_lock_irqsave(&mapping->i_pages, flags);
+
+ /* Notes on differences with __set_page_dirty_nobuffers:
+ * 1. We don't need to call page_mapping because we know this is a page
+ * cache page.
+ * 2. We have the pages locked, so there is no need for the careful
+ * mapping/mapping2 dance.
+ * 3. No mapping is impossible. (Race w/truncate mentioned in
+ * dirty_nobuffers should be impossible because we hold the page lock.)
+ * 4. All mappings are the same because i/o is only to one file.
+ * 5. We invert the lock order on lock_page_memcg(page) and the mapping
+ * xa_lock, but this is the only function that should use that pair of
+ * locks and it can't race because Lustre locks pages throughout i/o.
+ */
+ for (i = 0; i < count; i++) {
+ page = pvec->pages[i];
+ lock_page_memcg(page);
+ if (TestSetPageDirty(page)) {
+ unlock_page_memcg(page);
+ continue;
+ }
+ LASSERTF(page->mapping == mapping,
+ "all pages must have the same mapping. page %p, mapping %p, first mapping %p\n",
+ page, page->mapping, mapping);
+ WARN_ON_ONCE(!PagePrivate(page) && !PageUptodate(page));
+ ll_account_page_dirtied(page, mapping);
+ ll_page_tag_dirty(page, mapping);
+ dirtied++;
+ unlock_page_memcg(page);
+ }
+ ll_xa_unlock_irqrestore(&mapping->i_pages, flags);
+
+ CDEBUG(D_VFSTRACE, "mapping %p, count %d, dirtied %d\n", mapping,
+ count, dirtied);
+
+ if (mapping->host && dirtied) {
+ /* !PageAnon && !swapper_space */
+ __mark_inode_dirty(mapping->host, I_DIRTY_PAGES);
+ }
+
+ EXIT;
+}
+
static void write_commit_callback(const struct lu_env *env, struct cl_io *io,
- struct cl_page *page)
+ struct pagevec *pvec)
{
- struct page *vmpage = page->cp_vmpage;
+ int count = 0;
+ int i = 0;
- SetPageUptodate(vmpage);
- set_page_dirty(vmpage);
+ ENTRY;
- cl_page_disown(env, io, page);
+ count = pagevec_count(pvec);
+ LASSERT(count > 0);
- /* held in ll_cl_init() */
- lu_ref_del(&page->cp_reference, "cl_io", cl_io_top(io));
- cl_page_put(env, page);
+ for (i = 0; i < count; i++) {
+ struct page *vmpage = pvec->pages[i];
+ SetPageUptodate(vmpage);
+ }
+
+ vvp_set_pagevec_dirty(pvec);
+
+ for (i = 0; i < count; i++) {
+ struct page *vmpage = pvec->pages[i];
+ struct cl_page *page = (struct cl_page *) vmpage->private;
+ cl_page_disown(env, io, page);
+ lu_ref_del(&page->cp_reference, "cl_io", cl_io_top(io));
+ cl_page_put(env, page);
+ }
+
+ EXIT;
}
/* make sure the page list is contiguous */
struct cl_object *obj = io->ci_obj;
struct inode *inode = vvp_object_inode(obj);
struct ll_inode_info *lli = ll_i2info(inode);
+ struct file *file = vio->vui_fd->fd_file;
ssize_t result = 0;
loff_t pos = io->u.ci_wr.wr.crw_pos;
size_t cnt = io->u.ci_wr.wr.crw_count;
+ bool lock_inode = !IS_NOSEC(inode);
+ size_t nob = io->ci_nob;
+ struct iov_iter iter;
+ size_t written = 0;
ENTRY;
if (vio->vui_io_subtype == IO_NORMAL)
- down_read(&lli->lli_trunc_sem);
+ trunc_sem_down_read(&lli->lli_trunc_sem);
if (!can_populate_pages(env, io, inode))
RETURN(0);
pos = io->u.ci_wr.wr.crw_pos = i_size_read(inode);
vio->vui_iocb->ki_pos = pos;
} else {
- LASSERT(vio->vui_iocb->ki_pos == pos);
+ LASSERTF(vio->vui_iocb->ki_pos == pos,
+ "ki_pos %lld [%lld, %lld)\n",
+ vio->vui_iocb->ki_pos,
+ pos, pos + cnt);
}
- CDEBUG(D_VFSTRACE, "write: [%lli, %lli)\n", pos, pos + (long long)cnt);
+ CDEBUG(D_VFSTRACE, "%s: write [%llu, %llu)\n",
+ file_dentry(file)->d_name.name,
+ pos, pos + cnt);
/* The maximum Lustre file size is variable, based on the OST maximum
* object size and number of stripes. This needs another check in
* addition to the VFS checks earlier. */
if (pos + cnt > ll_file_maxbytes(inode)) {
CDEBUG(D_INODE,
- "%s: file "DFID" offset %llu > maxbytes "LPU64"\n",
- ll_get_fsname(inode->i_sb, NULL, 0),
+ "%s: file %s ("DFID") offset %llu > maxbytes %llu\n",
+ ll_i2sbi(inode)->ll_fsname,
+ file_dentry(file)->d_name.name,
PFID(ll_inode2fid(inode)), pos + cnt,
ll_file_maxbytes(inode));
RETURN(-EFBIG);
}
+ /* Tests to verify we take the i_mutex correctly */
+ if (OBD_FAIL_CHECK(OBD_FAIL_LLITE_IMUTEX_SEC) && !lock_inode)
+ RETURN(-EINVAL);
+
+ if (OBD_FAIL_CHECK(OBD_FAIL_LLITE_IMUTEX_NOSEC) && lock_inode)
+ RETURN(-EINVAL);
+
if (vio->vui_iter == NULL) {
/* from a temp io in ll_cl_init(). */
result = 0;
* consistency, proper locking to protect against writes,
* trucates, etc. is handled in the higher layers of lustre.
*/
-#ifdef HAVE_FILE_OPERATIONS_READ_WRITE_ITER
- result = generic_file_write_iter(vio->vui_iocb, vio->vui_iter);
+ lock_inode = !IS_NOSEC(inode);
+ iter = *vio->vui_iter;
+
+ if (unlikely(lock_inode))
+ inode_lock(inode);
+ result = __generic_file_write_iter(vio->vui_iocb,
+ vio->vui_iter);
+ if (unlikely(lock_inode))
+ inode_unlock(inode);
+
+ written = result;
+ if (result > 0)
+#ifdef HAVE_GENERIC_WRITE_SYNC_2ARGS
+ result = generic_write_sync(vio->vui_iocb, result);
#else
- result = __generic_file_aio_write(vio->vui_iocb,
- vio->vui_iter->iov,
- vio->vui_iter->nr_segs,
- &vio->vui_iocb->ki_pos);
-#endif
- if (result > 0 || result == -EIOCBQUEUED) {
+ {
ssize_t err;
- err = generic_write_sync(vio->vui_iocb->ki_filp,
- pos, result);
+ err = generic_write_sync(vio->vui_iocb->ki_filp, pos,
+ result);
if (err < 0 && result > 0)
result = err;
}
-
+#endif
}
+
if (result > 0) {
result = vvp_io_write_commit(env, io);
+ /* Simulate short commit */
+ if (CFS_FAULT_CHECK(OBD_FAIL_LLITE_SHORT_COMMIT)) {
+ vio->u.write.vui_written >>= 1;
+ if (vio->u.write.vui_written > 0)
+ io->ci_need_restart = 1;
+ }
if (vio->u.write.vui_written > 0) {
result = vio->u.write.vui_written;
- io->ci_nob += result;
-
- CDEBUG(D_VFSTRACE, "write: nob %zd, result: %zd\n",
+ CDEBUG(D_VFSTRACE, "%s: write nob %zd, result: %zd\n",
+ file_dentry(file)->d_name.name,
io->ci_nob, result);
+ io->ci_nob += result;
+ } else {
+ io->ci_continue = 0;
}
}
- if (result > 0) {
+ if (vio->vui_iocb->ki_pos != (pos + io->ci_nob - nob)) {
+ CDEBUG(D_VFSTRACE, "%s: write position mismatch: "
+ "ki_pos %lld vs. pos %lld, written %ld, commit %ld "
+ "rc %ld\n",
+ file_dentry(file)->d_name.name,
+ vio->vui_iocb->ki_pos, pos + io->ci_nob - nob,
+ written, io->ci_nob - nob, result);
+ /*
+ * Rewind ki_pos and vui_iter to where it has
+ * successfully committed.
+ */
+ vio->vui_iocb->ki_pos = pos + io->ci_nob - nob;
+ iov_iter_advance(&iter, io->ci_nob - nob);
+ vio->vui_iter->iov = iter.iov;
+ vio->vui_iter->nr_segs = iter.nr_segs;
+ vio->vui_iter->iov_offset = iter.iov_offset;
+ vio->vui_iter->count = iter.count;
+ }
+ if (result > 0 || result == -EIOCBQUEUED) {
ll_file_set_flag(ll_i2info(inode), LLIF_DATA_MODIFIED);
if (result < cnt)
io->ci_continue = 0;
- ll_rw_stats_tally(ll_i2sbi(inode), current->pid,
- vio->vui_fd, pos, result, WRITE);
- result = 0;
+ if (result > 0)
+ result = 0;
}
RETURN(result);
struct ll_inode_info *lli = ll_i2info(inode);
if (vio->vui_io_subtype == IO_NORMAL)
- up_read(&lli->lli_trunc_sem);
+ trunc_sem_up_read(&lli->lli_trunc_sem);
}
static int vvp_io_kernel_fault(struct vvp_fault_io *cfio)
{
struct vm_fault *vmf = cfio->ft_vmf;
- cfio->ft_flags = filemap_fault(cfio->ft_vma, vmf);
+ cfio->ft_flags = ll_filemap_fault(cfio->ft_vma, vmf);
cfio->ft_flags_valid = 1;
if (vmf->page) {
LL_CDEBUG_PAGE(D_PAGE, vmf->page, "got addr %p type NOPAGE\n",
- vmf->virtual_address);
+ get_vmf_address(vmf));
if (unlikely(!(cfio->ft_flags & VM_FAULT_LOCKED))) {
lock_page(vmf->page);
cfio->ft_flags |= VM_FAULT_LOCKED;
}
if (cfio->ft_flags & VM_FAULT_SIGBUS) {
- CDEBUG(D_PAGE, "got addr %p - SIGBUS\n", vmf->virtual_address);
+ CDEBUG(D_PAGE, "got addr %p - SIGBUS\n", get_vmf_address(vmf));
return -EFAULT;
}
if (cfio->ft_flags & VM_FAULT_OOM) {
- CDEBUG(D_PAGE, "got addr %p - OOM\n", vmf->virtual_address);
+ CDEBUG(D_PAGE, "got addr %p - OOM\n", get_vmf_address(vmf));
return -ENOMEM;
}
}
static void mkwrite_commit_callback(const struct lu_env *env, struct cl_io *io,
- struct cl_page *page)
+ struct pagevec *pvec)
{
- set_page_dirty(page->cp_vmpage);
+ vvp_set_pagevec_dirty(pvec);
}
static int vvp_io_fault_start(const struct lu_env *env,
pgoff_t last_index;
ENTRY;
- 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)));
-
- down_read(&lli->lli_trunc_sem);
+ trunc_sem_down_read_nowait(&lli->lli_trunc_sem);
/* offset of the last byte on the page */
offset = cl_offset(obj, fio->ft_index + 1) - 1;
LASSERT(PageLocked(vmpage));
if (OBD_FAIL_CHECK(OBD_FAIL_LLITE_FAULT_TRUNC_RACE))
- ll_invalidate_page(vmpage);
+ generic_error_remove_page(vmpage->mapping, vmpage);
size = i_size_read(inode);
/* Though we have already held a cl_lock upon this page, but
if (fio->ft_mkwrite) {
wait_on_page_writeback(vmpage);
if (!PageDirty(vmpage)) {
- struct cl_page_list *plist = &io->ci_queue.c2_qin;
+ struct cl_page_list *plist = &vio->u.fault.ft_queue;
struct vvp_page *vpg = cl_object_page_slice(obj, page);
int to = PAGE_SIZE;
/* size fixup */
if (last_index == vvp_index(vpg))
- to = size & ~PAGE_MASK;
+ to = ((size - 1) & ~PAGE_MASK) + 1;
/* 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_io_commit_async(env, io, plist, 0, to,
mkwrite_commit_callback);
+ /* Have overquota flag, trying sync write to check
+ * whether indeed out of quota */
+ if (result == -EDQUOT) {
+ cl_page_get(page);
+ result = vvp_io_commit_sync(env, io,
+ plist, 0, to);
+ if (result >= 0) {
+ io->ci_noquota = 1;
+ cl_page_own(env, io, page);
+ cl_page_list_add(plist, page);
+ lu_ref_add(&page->cp_reference,
+ "cl_io", io);
+ result = cl_io_commit_async(env, io,
+ plist, 0, to,
+ mkwrite_commit_callback);
+ io->ci_noquota = 0;
+ } else {
+ cl_page_put(env, page);
+ }
+ }
+
LASSERT(cl_page_is_owned(page, io));
cl_page_list_fini(env, plist);
if (result == -EDQUOT)
result = -ENOSPC;
GOTO(out, result);
- } else
+ } else {
cl_page_disown(env, io, page);
+ }
}
}
CLOBINVRNT(env, ios->cis_io->ci_obj,
vvp_object_invariant(ios->cis_io->ci_obj));
- up_read(&lli->lli_trunc_sem);
+ trunc_sem_up_read(&lli->lli_trunc_sem);
}
static int vvp_io_fsync_start(const struct lu_env *env,
struct vvp_io *vio = cl2vvp_io(env, ios);
if (unlikely(vio->vui_fd->fd_flags & LL_FILE_GROUP_LOCKED)) {
- ra->cra_end = CL_PAGE_EOF;
+ ra->cra_end_idx = CL_PAGE_EOF;
result = +1; /* no need to call down */
}
}
.cio_start = vvp_io_fsync_start,
.cio_fini = vvp_io_fini
},
+ [CIT_GLIMPSE] = {
+ .cio_fini = vvp_io_fini
+ },
[CIT_MISC] = {
.cio_fini = vvp_io_fini
},
size_t count;
struct ll_inode_info *lli = ll_i2info(inode);
- count = io->u.ci_rw.crw_count;
- /* "If nbyte is 0, read() will return 0 and have no other
- * results." -- Single Unix Spec */
- if (count == 0)
- result = 1;
- else {
+ count = io->u.ci_rw.crw_count;
+ /* "If nbyte is 0, read() will return 0 and have no other
+ * results." -- Single Unix Spec */
+ if (count == 0)
+ result = 1;
+ else
vio->vui_tot_count = count;
-#ifndef HAVE_FILE_OPERATIONS_READ_WRITE_ITER
- vio->vui_tot_nrsegs = 0;
-#endif
- }
/* for read/write, we store the jobid in the inode, and
* it'll be fetched by osc when building RPC.
* it's not accurate if the file is shared by different
* jobs.
*/
- lustre_get_jobid(lli->lli_jobid);
+ lustre_get_jobid(lli->lli_jobid, sizeof(lli->lli_jobid));
} else if (io->ci_type == CIT_SETATTR) {
if (!cl_io_is_trunc(io))
io->ci_lockreq = CILR_MANDATORY;
}
- /* ignore layout change for generic CIT_MISC but not for glimpse.
- * io context for glimpse must set ci_verify_layout to true,
- * see cl_glimpse_size0() for details. */
- if (io->ci_type == CIT_MISC && !io->ci_verify_layout)
- io->ci_ignore_layout = 1;
-
/* Enqueue layout lock and get layout version. We need to do this
* even for operations requiring to open file, such as read and write,
* because it might not grant layout lock in IT_OPEN. */
result = 0;
if (result < 0)
CERROR("%s: refresh file layout " DFID " error %d.\n",
- ll_get_fsname(inode->i_sb, NULL, 0),
- PFID(lu_object_fid(&obj->co_lu)), result);
+ ll_i2sbi(inode)->ll_fsname,
+ PFID(lu_object_fid(&obj->co_lu)), result);
}
+ io->ci_result = result < 0 ? result : 0;
RETURN(result);
}