Viewing: cl_io.c
// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
* Copyright (c) 2011, 2017, Intel Corporation.
*
*/
/*
* This file is part of Lustre, http://www.lustre.org/
*
* Client IO.
*
* Author: Nikita Danilov <nikita.danilov@sun.com>
* Author: Jinshan Xiong <jinshan.xiong@intel.com>
*
*/
#define DEBUG_SUBSYSTEM S_CLASS
#include <linux/sched.h>
#include <linux/list.h>
#include <linux/list_sort.h>
#include <linux/mmu_context.h>
#include <obd_class.h>
#include <obd_support.h>
#include <lustre_fid.h>
#include <cl_object.h>
#include "cl_internal.h"
/*
* cl_io interface.
*/
static inline int cl_io_type_is_valid(enum cl_io_type type)
{
return CIT_READ <= type && type < CIT_OP_NR;
}
static inline int cl_io_is_loopable(const struct cl_io *io)
{
return cl_io_type_is_valid(io->ci_type) && io->ci_type != CIT_MISC;
}
/*
* cl_io invariant that holds at all times when exported cl_io_*() functions
* are entered and left.
*/
static inline int cl_io_invariant(const struct cl_io *io)
{
/*
* io can own pages only when it is ongoing. Sub-io might
* still be in CIS_LOCKED state when top-io is in
* CIS_IO_GOING.
*/
return ergo(io->ci_owned_nr > 0, io->ci_state == CIS_IO_GOING ||
(io->ci_state == CIS_LOCKED && io->ci_parent != NULL));
}
/*
* Finalize @io, by calling cl_io_operations::cio_fini() bottom-to-top.
*/
void cl_io_fini(const struct lu_env *env, struct cl_io *io)
{
struct cl_io_slice *slice;
LINVRNT(cl_io_type_is_valid(io->ci_type));
LINVRNT(cl_io_invariant(io));
ENTRY;
while (!list_empty(&io->ci_layers)) {
slice = container_of(io->ci_layers.prev, struct cl_io_slice,
cis_linkage);
list_del_init(&slice->cis_linkage);
if (slice->cis_iop->op[io->ci_type].cio_fini != NULL)
slice->cis_iop->op[io->ci_type].cio_fini(env, slice);
/*
* Invalidate slice to catch use after free. This assumes that
* slices are allocated within session and can be touched
* after ->cio_fini() returns.
*/
slice->cis_io = NULL;
}
io->ci_state = CIS_FINI;
/* sanity check for layout change */
switch(io->ci_type) {
case CIT_READ:
case CIT_WRITE:
case CIT_DATA_VERSION:
case CIT_FAULT:
case CIT_SETATTR:
break;
case CIT_FSYNC:
LASSERT(!io->ci_need_restart);
break;
case CIT_MISC:
/* Check ignore layout change conf */
LASSERT(ergo(io->ci_ignore_layout || !io->ci_verify_layout,
!io->ci_need_restart));
break;
case CIT_GLIMPSE:
break;
case CIT_LADVISE:
case CIT_LSEEK:
break;
default:
LBUG();
}
EXIT;
}
EXPORT_SYMBOL(cl_io_fini);
static int __cl_io_init(const struct lu_env *env, struct cl_io *io,
enum cl_io_type iot, struct cl_object *obj)
{
struct cl_object *scan;
int result;
LINVRNT(io->ci_state == CIS_ZERO || io->ci_state == CIS_FINI);
LINVRNT(cl_io_type_is_valid(iot));
LINVRNT(cl_io_invariant(io));
ENTRY;
io->ci_type = iot;
INIT_LIST_HEAD(&io->ci_lockset.cls_todo);
INIT_LIST_HEAD(&io->ci_lockset.cls_done);
INIT_LIST_HEAD(&io->ci_layers);
result = 0;
cl_object_for_each(scan, obj) {
if (scan->co_ops->coo_io_init != NULL) {
result = scan->co_ops->coo_io_init(env, scan, io);
if (result != 0)
break;
}
}
if (result == 0)
io->ci_state = CIS_INIT;
RETURN(result);
}
/**
* cl_io_sub_init() - Initialize sub-IO (cl_io_operations::cio_init()
* top-to-bottom)
* @env: execution environment
* @io: pointer to a cl_io struct
* @iot: Requested transfer type (READ / WRITE)
* @obj: Sub object linked to IO operation (obj != cl_object_top(obj)).
*
* Return:
* * %0: Success
* * %-ERRNO: Failure
*/
int cl_io_sub_init(const struct lu_env *env, struct cl_io *io,
enum cl_io_type iot, struct cl_object *obj)
{
LASSERT(obj != cl_object_top(obj));
return __cl_io_init(env, io, iot, obj);
}
EXPORT_SYMBOL(cl_io_sub_init);
/**
* cl_io_init() - Initialize @io (cl_io_operations::cio_init() top-to-bottom)
* @env: execution environment
* @io: pointer to a cl_io struct
* @iot: Requested transfer type (READ / WRITE)
* @obj: Object linked to IO operation (obj == cl_object_top(obj))
*
* Caller has to call cl_io_fini() after a call to cl_io_init(), no matter
* what the latter returned.
*
* Return:
* * %0: Success
* * %-ERRNO: Failure
*/
int cl_io_init(const struct lu_env *env, struct cl_io *io,
enum cl_io_type iot, struct cl_object *obj)
{
LASSERT(obj == cl_object_top(obj));
/* clear I/O restart from previous instance */
io->ci_need_restart = 0;
return __cl_io_init(env, io, iot, obj);
}
EXPORT_SYMBOL(cl_io_init);
/**
* cl_io_rw_init() - Initialize read or write IO
* @env: execution environment
* @io: pointer to a cl_io struct
* @iot: Requested transfer type (READ / WRITE)
* @pos: IO start position
* @bytes: Number of bytes to Read/Write depending on @iot
*
* * Return:
* * %0: Success
* * %-ERRNO: Failure
*/
int cl_io_rw_init(const struct lu_env *env, struct cl_io *io,
enum cl_io_type iot, loff_t pos, size_t bytes)
{
LINVRNT(iot == CIT_READ || iot == CIT_WRITE);
LINVRNT(io->ci_obj != NULL);
ENTRY;
LU_OBJECT_HEADER(D_VFSTRACE, env, &io->ci_obj->co_lu,
"io range: %u [%llu, %llu) %u %u\n",
iot, (__u64)pos, (__u64)pos + bytes,
io->u.ci_rw.crw_nonblock, io->u.ci_wr.wr_append);
io->u.ci_rw.crw_pos = pos;
io->u.ci_rw.crw_bytes = bytes;
RETURN(cl_io_init(env, io, iot, io->ci_obj));
}
EXPORT_SYMBOL(cl_io_rw_init);
#ifdef HAVE_LIST_CMP_FUNC_T
static int cl_lock_descr_cmp(void *priv,
const struct list_head *a,
const struct list_head *b)
#else /* !HAVE_LIST_CMP_FUNC_T */
static int cl_lock_descr_cmp(void *priv,
struct list_head *a, struct list_head *b)
#endif /* HAVE_LIST_CMP_FUNC_T */
{
const struct cl_io_lock_link *l0 = list_entry(a, struct cl_io_lock_link,
cill_linkage);
const struct cl_io_lock_link *l1 = list_entry(b, struct cl_io_lock_link,
cill_linkage);
const struct cl_lock_descr *d0 = &l0->cill_descr;
const struct cl_lock_descr *d1 = &l1->cill_descr;
return lu_fid_cmp(lu_object_fid(&d0->cld_obj->co_lu),
lu_object_fid(&d1->cld_obj->co_lu));
}
static void cl_lock_descr_merge(struct cl_lock_descr *d0,
const struct cl_lock_descr *d1)
{
d0->cld_start = min(d0->cld_start, d1->cld_start);
d0->cld_end = max(d0->cld_end, d1->cld_end);
if (d1->cld_mode == CLM_WRITE && d0->cld_mode != CLM_WRITE)
d0->cld_mode = CLM_WRITE;
if (d1->cld_mode == CLM_GROUP && d0->cld_mode != CLM_GROUP)
d0->cld_mode = CLM_GROUP;
}
static int cl_lockset_merge(const struct cl_lockset *set,
const struct cl_lock_descr *need)
{
struct cl_io_lock_link *scan;
ENTRY;
list_for_each_entry(scan, &set->cls_todo, cill_linkage) {
if (!cl_object_same(scan->cill_descr.cld_obj, need->cld_obj))
continue;
/* Merge locks for the same object because ldlm lock server
* may expand the lock extent, otherwise there is a deadlock
* case if two conflicted locks are queueud for the same object
* and lock server expands one lock to overlap the another.
* The side effect is that it can generate a multi-stripe lock
* that may cause casacading problem */
cl_lock_descr_merge(&scan->cill_descr, need);
CDEBUG(D_VFSTRACE, "lock: %d: [%lu, %lu]\n",
scan->cill_descr.cld_mode, scan->cill_descr.cld_start,
scan->cill_descr.cld_end);
RETURN(+1);
}
RETURN(0);
}
static int cl_lockset_lock(const struct lu_env *env, struct cl_io *io,
struct cl_lockset *set)
{
struct cl_io_lock_link *link;
struct cl_io_lock_link *temp;
int result;
ENTRY;
result = 0;
list_for_each_entry_safe(link, temp, &set->cls_todo, cill_linkage) {
result = cl_lock_request(env, io, &link->cill_lock);
if (result < 0)
break;
list_move(&link->cill_linkage, &set->cls_done);
}
RETURN(result);
}
/*
* Takes locks necessary for the current iteration of IO.
*
* Calls cl_io_operations::cio_lock() top-to-bottom to collect locks required
* by layers for the current iteration. Then sort locks (to avoid dead-locks),
* and acquire them.
*/
int cl_io_lock(const struct lu_env *env, struct cl_io *io)
{
const struct cl_io_slice *scan;
int result = 0;
LINVRNT(cl_io_is_loopable(io));
LINVRNT(io->ci_state == CIS_IT_STARTED);
LINVRNT(cl_io_invariant(io));
ENTRY;
list_for_each_entry(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->op[io->ci_type].cio_lock == NULL)
continue;
result = scan->cis_iop->op[io->ci_type].cio_lock(env, scan);
if (result != 0)
break;
}
if (result == 0) {
/*
* Sort locks in lexicographical order of their (fid,
* start-offset) pairs to avoid deadlocks.
*/
list_sort(NULL, &io->ci_lockset.cls_todo, cl_lock_descr_cmp);
result = cl_lockset_lock(env, io, &io->ci_lockset);
}
if (result != 0)
cl_io_unlock(env, io);
else
io->ci_state = CIS_LOCKED;
RETURN(result);
}
EXPORT_SYMBOL(cl_io_lock);
/*
* Release locks takes by IO
*/
void cl_io_unlock(const struct lu_env *env, struct cl_io *io)
{
struct cl_lockset *set;
struct cl_io_lock_link *link;
struct cl_io_lock_link *temp;
const struct cl_io_slice *scan;
LASSERT(cl_io_is_loopable(io));
LASSERT(CIS_IT_STARTED <= io->ci_state && io->ci_state < CIS_UNLOCKED);
LINVRNT(cl_io_invariant(io));
ENTRY;
set = &io->ci_lockset;
list_for_each_entry_safe(link, temp, &set->cls_todo, cill_linkage) {
list_del_init(&link->cill_linkage);
if (link->cill_fini != NULL)
link->cill_fini(env, link);
}
list_for_each_entry_safe(link, temp, &set->cls_done, cill_linkage) {
list_del_init(&link->cill_linkage);
cl_lock_release(env, &link->cill_lock);
if (link->cill_fini != NULL)
link->cill_fini(env, link);
}
list_for_each_entry_reverse(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->op[io->ci_type].cio_unlock != NULL)
scan->cis_iop->op[io->ci_type].cio_unlock(env, scan);
}
io->ci_state = CIS_UNLOCKED;
EXIT;
}
EXPORT_SYMBOL(cl_io_unlock);
/**
* cl_io_iter_init() - Prepares next iteration of IO.
* @env: execution environment
* @io: pointer to a cl_io struct
*
* Calls cl_io_operations::cio_iter_init() top-to-bottom. This exists to give
* layers a chance to modify @io parameters, e.g., so that lov can restrict @io
* to a single stripe.
*
* * Return:
* * %0: successfully initialized
* * %-ERRNO: failure
*/
int cl_io_iter_init(const struct lu_env *env, struct cl_io *io)
{
const struct cl_io_slice *scan;
int result;
LINVRNT(cl_io_is_loopable(io));
LINVRNT(io->ci_state == CIS_INIT || io->ci_state == CIS_IT_ENDED);
LINVRNT(cl_io_invariant(io));
ENTRY;
result = 0;
list_for_each_entry(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->op[io->ci_type].cio_iter_init == NULL)
continue;
result = scan->cis_iop->op[io->ci_type].cio_iter_init(env,
scan);
if (result != 0)
break;
}
if (result == 0)
io->ci_state = CIS_IT_STARTED;
RETURN(result);
}
EXPORT_SYMBOL(cl_io_iter_init);
/*
* Finalizes IO iteration.
*
* Calls cl_io_operations::cio_iter_fini() bottom-to-top.
*/
void cl_io_iter_fini(const struct lu_env *env, struct cl_io *io)
{
const struct cl_io_slice *scan;
LINVRNT(cl_io_is_loopable(io));
LINVRNT(io->ci_state <= CIS_IT_STARTED ||
io->ci_state > CIS_IO_FINISHED);
LINVRNT(cl_io_invariant(io));
ENTRY;
list_for_each_entry_reverse(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->op[io->ci_type].cio_iter_fini != NULL)
scan->cis_iop->op[io->ci_type].cio_iter_fini(env, scan);
}
io->ci_state = CIS_IT_ENDED;
EXIT;
}
EXPORT_SYMBOL(cl_io_iter_fini);
/**
* cl_io_rw_advance() - Records read/write @io progressed @bytes forward
* @env: execution environment
* @io: pointer to a cl_io struct
* @bytes: Number of bytes IO should advance
*/
void cl_io_rw_advance(const struct lu_env *env, struct cl_io *io, size_t bytes)
{
const struct cl_io_slice *scan;
ENTRY;
LINVRNT(io->ci_type == CIT_READ || io->ci_type == CIT_WRITE ||
bytes == 0);
LINVRNT(cl_io_is_loopable(io));
LINVRNT(cl_io_invariant(io));
io->u.ci_rw.crw_pos += bytes;
io->u.ci_rw.crw_bytes -= bytes;
/* layers have to be notified. */
list_for_each_entry_reverse(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->op[io->ci_type].cio_advance != NULL)
scan->cis_iop->op[io->ci_type].cio_advance(env, scan,
bytes);
}
EXIT;
}
/*
* Adds a lock to a lockset.
*/
int cl_io_lock_add(const struct lu_env *env, struct cl_io *io,
struct cl_io_lock_link *link)
{
int result;
ENTRY;
if (cl_lockset_merge(&io->ci_lockset, &link->cill_descr))
result = +1;
else {
list_add(&link->cill_linkage, &io->ci_lockset.cls_todo);
result = 0;
}
RETURN(result);
}
EXPORT_SYMBOL(cl_io_lock_add);
static void cl_free_io_lock_link(const struct lu_env *env,
struct cl_io_lock_link *link)
{
OBD_FREE_PTR(link);
}
/*
* Allocates new lock link, and uses it to add a lock to a lockset.
*/
int cl_io_lock_alloc_add(const struct lu_env *env, struct cl_io *io,
struct cl_lock_descr *descr)
{
struct cl_io_lock_link *link;
int result;
ENTRY;
OBD_ALLOC_PTR(link);
if (link != NULL) {
link->cill_descr = *descr;
link->cill_fini = cl_free_io_lock_link;
result = cl_io_lock_add(env, io, link);
if (result) /* lock match */
link->cill_fini(env, link);
} else
result = -ENOMEM;
RETURN(result);
}
EXPORT_SYMBOL(cl_io_lock_alloc_add);
/*
* Starts IO by calling cl_io_operations::cio_start() top-to-bottom.
*/
int cl_io_start(const struct lu_env *env, struct cl_io *io)
{
const struct cl_io_slice *scan;
int result = 0;
LINVRNT(cl_io_is_loopable(io));
LINVRNT(io->ci_state == CIS_LOCKED);
LINVRNT(cl_io_invariant(io));
ENTRY;
io->ci_state = CIS_IO_GOING;
list_for_each_entry(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->op[io->ci_type].cio_start == NULL)
continue;
result = scan->cis_iop->op[io->ci_type].cio_start(env, scan);
if (result != 0)
break;
}
if (result >= 0)
result = 0;
RETURN(result);
}
EXPORT_SYMBOL(cl_io_start);
/*
* Wait until current IO iteration is finished by calling
* cl_io_operations::cio_end() bottom-to-top.
*/
void cl_io_end(const struct lu_env *env, struct cl_io *io)
{
const struct cl_io_slice *scan;
LINVRNT(cl_io_is_loopable(io));
LINVRNT(io->ci_state == CIS_IO_GOING);
LINVRNT(cl_io_invariant(io));
ENTRY;
list_for_each_entry_reverse(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->op[io->ci_type].cio_end != NULL)
scan->cis_iop->op[io->ci_type].cio_end(env, scan);
/* TODO: error handling. */
}
io->ci_state = CIS_IO_FINISHED;
EXIT;
}
EXPORT_SYMBOL(cl_io_end);
/*
* Called by read IO, to decide the readahead extent
*
* see cl_io_operations::cio_read_ahead_prep()
*/
int cl_io_read_ahead_prep(const struct lu_env *env, struct cl_io *io,
pgoff_t start, struct cl_read_ahead *ra)
{
const struct cl_io_slice *scan;
int result = 0;
LINVRNT(io->ci_type == CIT_READ ||
io->ci_type == CIT_FAULT ||
io->ci_type == CIT_WRITE);
LINVRNT(io->ci_state == CIS_IO_GOING || io->ci_state == CIS_LOCKED);
LINVRNT(cl_io_invariant(io));
ENTRY;
list_for_each_entry(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->cio_read_ahead_prep == NULL)
continue;
result = scan->cis_iop->cio_read_ahead_prep(env, scan,
start, ra);
if (result != 0)
break;
}
RETURN(result > 0 ? 0 : result);
}
EXPORT_SYMBOL(cl_io_read_ahead_prep);
/*
* Called before IO start, to reserve enough LRU slots to avoid
* deadlock.
*
* \see cl_io_operations::cio_lru_reserve()
*/
int cl_io_lru_reserve(const struct lu_env *env, struct cl_io *io,
loff_t pos, size_t bytes)
{
const struct cl_io_slice *scan;
int result = 0;
LINVRNT(io->ci_type == CIT_READ || io->ci_type == CIT_WRITE);
LINVRNT(cl_io_invariant(io));
ENTRY;
list_for_each_entry(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->cio_lru_reserve) {
result = scan->cis_iop->cio_lru_reserve(env, scan,
pos, bytes);
if (result)
break;
}
}
RETURN(result);
}
EXPORT_SYMBOL(cl_io_lru_reserve);
/**
* cl_io_commit_async() - Commit a list of contiguous pages into writeback cache
* @env: execution environment
* @io: pointer to a cl_io struct
* @queue: pointer to cl_page_list struct (pages that will be committed)
* @from: Starting position
* @to: Ending position
* @cb: callback function
* @prio: I/O priority
*
* Returns 0 if all pages committed, or errcode if error occurred.
* see cl_io_operations::cio_commit_async()
*/
int cl_io_commit_async(const struct lu_env *env, struct cl_io *io,
struct cl_page_list *queue, int from, int to,
cl_commit_cbt cb, enum cl_io_priority prio)
{
const struct cl_io_slice *scan;
int result = 0;
ENTRY;
list_for_each_entry(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->cio_commit_async == NULL)
continue;
result = scan->cis_iop->cio_commit_async(env, scan, queue,
from, to, cb, prio);
if (result != 0)
break;
}
RETURN(result);
}
EXPORT_SYMBOL(cl_io_commit_async);
void cl_io_extent_release(const struct lu_env *env, struct cl_io *io,
enum cl_io_priority prio)
{
const struct cl_io_slice *scan;
ENTRY;
list_for_each_entry(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->cio_extent_release == NULL)
continue;
scan->cis_iop->cio_extent_release(env, scan, prio);
}
EXIT;
}
EXPORT_SYMBOL(cl_io_extent_release);
int cl_dio_submit_rw(const struct lu_env *env, struct cl_io *io,
enum cl_req_type crt, struct cl_dio_pages *cdp)
{
const struct cl_io_slice *scan;
int result = 0;
ENTRY;
list_for_each_entry(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->cio_dio_submit == NULL)
continue;
result = scan->cis_iop->cio_dio_submit(env, io, scan, crt,
cdp);
if (result != 0)
break;
}
RETURN(result);
}
EXPORT_SYMBOL(cl_dio_submit_rw);
/**
* cl_io_submit_rw() - Submits a list of pages for immediate IO.
* @env: execution environment
* @io: pointer to a cl_io struct
* @crt: Requested transfer type (READ / WRITE)
* @queue: pointer to a cl_2queue struct (pages in IO operation)
*
* After the function gets returned, The submitted pages are moved to
* queue->c2_qout queue, and queue->c2_qin contain both the pages don't need
* to be submitted, and the pages are errant to submit.
*
* Returns 0 if at least one page was submitted, error code otherwise.
* (see cl_io_operations::cio_submit())
*/
int cl_io_submit_rw(const struct lu_env *env, struct cl_io *io,
enum cl_req_type crt, struct cl_2queue *queue)
{
const struct cl_io_slice *scan;
int result = 0;
ENTRY;
list_for_each_entry(scan, &io->ci_layers, cis_linkage) {
if (scan->cis_iop->cio_submit == NULL)
continue;
result = scan->cis_iop->cio_submit(env, io, scan, crt, queue);
if (result != 0)
break;
}
/*
* If ->cio_submit() failed, no pages were sent.
*/
LASSERT(ergo(result != 0, list_empty(&queue->c2_qout.pl_pages)));
RETURN(result);
}
EXPORT_SYMBOL(cl_io_submit_rw);
/**
* cl_io_submit_sync() - submit a sync_io and wait for the IO to be finished,
* or error happens. If @timeout is zero, it means to wait for the IO
* unconditionally.
* @env: execution environment
* @io: pointer to a cl_io struct
* @iot: Requested transfer type (READ / WRITE)
* @queue: pointer to a cl_2queue struct (pages in IO operation)
* @timeout: Time in seconds to wait for IO completion.
*
* This is used for synchronous submission of an async IO, so the waiting is
* done here in this function and the IO is done when this function returns.
*
* Return:
* * %0: Success
* * %-ERRNO: Failure
*/
int cl_io_submit_sync(const struct lu_env *env, struct cl_io *io,
enum cl_req_type iot, struct cl_2queue *queue,
long timeout)
{
struct cl_sync_io *anchor = &cl_env_info(env)->clt_anchor;
struct cl_page *pg;
int rc;
ENTRY;
cl_page_list_for_each(pg, &queue->c2_qin) {
LASSERT(pg->cp_sync_io == NULL);
/* this is for sync submission of async IO, IO that was always
* sync (like DIO) is handled differently
*/
LASSERT(pg->cp_type != CPT_TRANSIENT);
pg->cp_sync_io = anchor;
}
cl_sync_io_init(anchor, queue->c2_qin.pl_nr);
rc = cl_io_submit_rw(env, io, iot, queue);
if (rc == 0) {
/*
* If some pages weren't sent for any reason (e.g.,
* read found up-to-date pages in the cache, or write found
* clean pages), count them as completed to avoid infinite
* wait.
*/
cl_page_list_for_each(pg, &queue->c2_qin) {
pg->cp_sync_io = NULL;
cl_sync_io_note(env, anchor, 1);
}
/* wait for the IO to be finished. */
rc = cl_sync_io_wait(env, anchor, timeout);
cl_page_list_assume(env, io, &queue->c2_qout);
} else {
LASSERT(list_empty(&queue->c2_qout.pl_pages));
cl_page_list_for_each(pg, &queue->c2_qin)
pg->cp_sync_io = NULL;
}
RETURN(rc);
}
EXPORT_SYMBOL(cl_io_submit_sync);
/**
* cl_io_loop() - Main IO loop.
* @env: thread environment in which the I/O operation is executed
* @io: pointer to a cl_io struct
*
* Pumps IO through iterations calling
*
* - cl_io_iter_init()
*
* - cl_io_lock()
*
* - cl_io_start()
*
* - cl_io_end()
*
* - cl_io_unlock()
*
* - cl_io_iter_fini()
*
* repeatedly until there is no more @io to do.
*
* Return:
* * %0: IO was success
* * %-ERRNO: Failure
*/
int cl_io_loop(const struct lu_env *env, struct cl_io *io)
{
int result = 0;
int rc = 0;
LINVRNT(cl_io_is_loopable(io));
ENTRY;
do {
size_t bytes;
io->ci_continue = 0;
result = cl_io_iter_init(env, io);
if (result == 0) {
bytes = io->ci_bytes;
result = cl_io_lock(env, io);
if (result == 0) {
/*
* Notify layers that locks has been taken,
* and do actual i/o.
*
* - llite: kms, short read;
* - llite: generic_file_read();
*/
result = cl_io_start(env, io);
/*
* Send any remaining pending
* io, etc.
*
** - llite: ll_rw_stats_tally.
*/
cl_io_end(env, io);
cl_io_unlock(env, io);
cl_io_rw_advance(env, io, io->ci_bytes - bytes);
}
}
cl_io_iter_fini(env, io);
if (result)
rc = result;
} while ((result == 0 || result == -EIOCBQUEUED) &&
io->ci_continue);
if (rc && !result)
result = rc;
if (result == -EAGAIN && io->ci_ndelay && !io->ci_iocb_nowait) {
if (!io->ci_tried_all_mirrors) {
io->ci_need_restart = 1;
result = 0;
} else {
result = -EIO;
}
}
if (result == 0)
result = io->ci_result;
RETURN(result < 0 ? result : 0);
}
EXPORT_SYMBOL(cl_io_loop);
/**
* cl_io_slice_add() - Adds IO slice to the cl_io.
* @io: pointer to a cl_io struct
* @slice: pointer to the cl_io_slice struct (Part of IO operation)
* @obj: Object linked to IO operation
* @ops: operations/methods applicable on @slice
*
* This is called by cl_object_operations::coo_io_init() methods to add a
* per-layer state to the @io. New state is added at the end of
* cl_io::ci_layers list, that is, it is at the bottom of the stack.
*
* see cl_lock_slice_add(), cl_req_slice_add(), cl_page_slice_add()
*/
void cl_io_slice_add(struct cl_io *io, struct cl_io_slice *slice,
struct cl_object *obj,
const struct cl_io_operations *ops)
{
struct list_head *linkage = &slice->cis_linkage;
LASSERT((linkage->prev == NULL && linkage->next == NULL) ||
list_empty(linkage));
ENTRY;
list_add_tail(linkage, &io->ci_layers);
slice->cis_io = io;
slice->cis_obj = obj;
slice->cis_iop = ops;
EXIT;
}
EXPORT_SYMBOL(cl_io_slice_add);
/*
* Initializes page list.
*/
void cl_page_list_init(struct cl_page_list *plist)
{
ENTRY;
plist->pl_nr = 0;
INIT_LIST_HEAD(&plist->pl_pages);
EXIT;
}
EXPORT_SYMBOL(cl_page_list_init);
/*
* Adds a page to a page list.
*/
void cl_page_list_add(struct cl_page_list *plist, struct cl_page *page,
bool getref)
{
ENTRY;
/* it would be better to check that page is owned by "current" io, but
* it is not passed here. */
if (page->cp_type != CPT_TRANSIENT)
LASSERT(page->cp_owner != NULL);
LASSERT(list_empty(&page->cp_batch));
list_add_tail(&page->cp_batch, &plist->pl_pages);
++plist->pl_nr;
if (getref)
cl_page_get(page);
EXIT;
}
EXPORT_SYMBOL(cl_page_list_add);
/*
* Removes a page from a page list.
*/
void cl_page_list_del(const struct lu_env *env,
struct cl_page_list *plist, struct cl_page *page,
bool putref)
{
LASSERT(plist->pl_nr > 0);
ENTRY;
list_del_init(&page->cp_batch);
--plist->pl_nr;
if (putref)
cl_page_put(env, page);
EXIT;
}
EXPORT_SYMBOL(cl_page_list_del);
/*
* Moves a page from one page list to another.
*/
void cl_page_list_move(struct cl_page_list *dst, struct cl_page_list *src,
struct cl_page *page)
{
LASSERT(src->pl_nr > 0);
ENTRY;
list_move_tail(&page->cp_batch, &dst->pl_pages);
--src->pl_nr;
++dst->pl_nr;
EXIT;
}
EXPORT_SYMBOL(cl_page_list_move);
/*
* Moves a page from one page list to the head of another list.
*/
void cl_page_list_move_head(struct cl_page_list *dst, struct cl_page_list *src,
struct cl_page *page)
{
LASSERT(src->pl_nr > 0);
ENTRY;
list_move(&page->cp_batch, &dst->pl_pages);
--src->pl_nr;
++dst->pl_nr;
EXIT;
}
EXPORT_SYMBOL(cl_page_list_move_head);
/*
* splice the cl_page_list, just as list head does
*/
void cl_page_list_splice(struct cl_page_list *src, struct cl_page_list *dst)
{
ENTRY;
dst->pl_nr += src->pl_nr;
src->pl_nr = 0;
list_splice_tail_init(&src->pl_pages, &dst->pl_pages);
EXIT;
}
EXPORT_SYMBOL(cl_page_list_splice);
/*
* Disowns pages in a queue.
*/
void cl_page_list_disown(const struct lu_env *env, struct cl_page_list *plist)
{
struct cl_page *page;
struct cl_page *temp;
ENTRY;
cl_page_list_for_each_safe(page, temp, plist) {
LASSERT(plist->pl_nr > 0);
list_del_init(&page->cp_batch);
--plist->pl_nr;
/*
* __cl_page_disown rather than usual cl_page_disown() is used,
* because pages are possibly in CPS_FREEING state already due
* to the call to cl_page_list_discard().
*/
/*
* XXX __cl_page_disown() will fail if page is not locked.
*/
__cl_page_disown(env, page);
cl_page_put(env, page);
}
EXIT;
}
EXPORT_SYMBOL(cl_page_list_disown);
/*
* Releases pages from queue.
*/
void cl_page_list_fini(const struct lu_env *env, struct cl_page_list *plist)
{
struct cl_page *page;
struct cl_page *temp;
ENTRY;
cl_page_list_for_each_safe(page, temp, plist)
cl_page_list_del(env, plist, page, true);
LASSERT(plist->pl_nr == 0);
EXIT;
}
EXPORT_SYMBOL(cl_page_list_fini);
/*
* Assumes all pages in a queue.
*/
void cl_page_list_assume(const struct lu_env *env,
struct cl_io *io, struct cl_page_list *plist)
{
struct cl_page *page;
cl_page_list_for_each(page, plist)
cl_page_assume(env, io, page);
}
/*
* Discards all pages in a queue.
*/
void cl_page_list_discard(const struct lu_env *env, struct cl_io *io,
struct cl_page_list *plist)
{
struct cl_page *page;
ENTRY;
cl_page_list_for_each(page, plist)
cl_page_discard(env, io, page);
EXIT;
}
EXPORT_SYMBOL(cl_page_list_discard);
/*
* Initialize dual page queue.
*/
void cl_2queue_init(struct cl_2queue *queue)
{
ENTRY;
cl_page_list_init(&queue->c2_qin);
cl_page_list_init(&queue->c2_qout);
EXIT;
}
EXPORT_SYMBOL(cl_2queue_init);
/*
* Disown pages in both lists of a 2-queue.
*/
void cl_2queue_disown(const struct lu_env *env, struct cl_2queue *queue)
{
ENTRY;
cl_page_list_disown(env, &queue->c2_qin);
cl_page_list_disown(env, &queue->c2_qout);
EXIT;
}
EXPORT_SYMBOL(cl_2queue_disown);
/*
* Discard (truncate) pages in both lists of a 2-queue.
*/
void cl_2queue_discard(const struct lu_env *env,
struct cl_io *io, struct cl_2queue *queue)
{
ENTRY;
cl_page_list_discard(env, io, &queue->c2_qin);
cl_page_list_discard(env, io, &queue->c2_qout);
EXIT;
}
EXPORT_SYMBOL(cl_2queue_discard);
/*
* Assume to own the pages in cl_2queue
*/
void cl_2queue_assume(const struct lu_env *env,
struct cl_io *io, struct cl_2queue *queue)
{
cl_page_list_assume(env, io, &queue->c2_qin);
cl_page_list_assume(env, io, &queue->c2_qout);
}
/*
* Finalize both page lists of a 2-queue.
*/
void cl_2queue_fini(const struct lu_env *env, struct cl_2queue *queue)
{
ENTRY;
cl_page_list_fini(env, &queue->c2_qout);
cl_page_list_fini(env, &queue->c2_qin);
EXIT;
}
EXPORT_SYMBOL(cl_2queue_fini);
/*
* Initialize a 2-queue to contain @page in its incoming page list.
*/
void cl_2queue_init_page(struct cl_2queue *queue, struct cl_page *page)
{
ENTRY;
cl_2queue_init(queue);
/*
* Add a page to the incoming page list of 2-queue.
*/
cl_page_list_add(&queue->c2_qin, page, true);
EXIT;
}
EXPORT_SYMBOL(cl_2queue_init_page);
/* Returns top-level io. (See cl_object.c:cl_object_top()) */
struct cl_io *cl_io_top(struct cl_io *io)
{
ENTRY;
while (io->ci_parent != NULL)
io = io->ci_parent;
RETURN(io);
}
EXPORT_SYMBOL(cl_io_top);
/**
* cl_req_attr_set() - Set attr for IO request
* @env: execution environment
* @obj: Object linked to IO operation (obj == cl_object_top(obj))
* @attr: Per-transfer attributes
*
* Fills in attributes that are passed to server together with transfer. Only
* attributes from @obj->cra_flags may be touched. This can be called multiple
* times for the same request
*/
void cl_req_attr_set(const struct lu_env *env, struct cl_object *obj,
struct cl_req_attr *attr)
{
struct cl_object *scan;
ENTRY;
cl_object_for_each(scan, obj) {
if (scan->co_ops->coo_req_attr_set != NULL)
scan->co_ops->coo_req_attr_set(env, scan, attr);
}
EXIT;
}
EXPORT_SYMBOL(cl_req_attr_set);
/**
* cl_sync_io_init_notify() - Initialize synchronous IO wait @anchor for @nr
* pages with optional @end handler.
* @anchor: owned by caller, initialized here.
* @nr: number of pages initially pending in sync.
* @dio_aio: Is it a DIO or AIO
* @end: optional callback sync_io completion, can be used to trigger erasure
* coding, integrity, dedupe, or similar operation. @end is called with a
* spinlock on anchor->csi_waitq.lock
*/
void cl_sync_io_init_notify(struct cl_sync_io *anchor, int nr,
void *dio_aio, cl_sync_io_end_t *end)
{
ENTRY;
memset(anchor, 0, sizeof(*anchor));
init_waitqueue_head(&anchor->csi_waitq);
atomic_set(&anchor->csi_sync_nr, nr);
atomic_set(&anchor->csi_complete, 0);
anchor->csi_sync_rc = 0;
anchor->csi_end_io = end;
anchor->csi_dio_aio = dio_aio;
EXIT;
}
EXPORT_SYMBOL(cl_sync_io_init_notify);
/*
* Wait until all IO completes. Transfer completion routine has to call
* cl_sync_io_note() for every entity.
*/
int cl_sync_io_wait(const struct lu_env *env, struct cl_sync_io *anchor,
long timeout)
{
int rc = 0;
ENTRY;
LASSERT(timeout >= 0);
if (timeout > 0 &&
wait_event_idle_timeout(anchor->csi_waitq,
atomic_read(&anchor->csi_complete) == 1,
cfs_time_seconds(timeout)) == 0) {
rc = -ETIMEDOUT;
CERROR("IO failed: %d, still wait for %d remaining entries\n",
rc, atomic_read(&anchor->csi_complete));
}
wait_event_idle(anchor->csi_waitq,
atomic_read(&anchor->csi_complete) == 1);
if (!rc)
rc = anchor->csi_sync_rc;
/* We take the lock to ensure that cl_sync_io_note() has finished */
spin_lock(&anchor->csi_waitq.lock);
LASSERT(atomic_read(&anchor->csi_sync_nr) == 0);
LASSERT(atomic_read(&anchor->csi_complete) == 1);
spin_unlock(&anchor->csi_waitq.lock);
RETURN(rc);
}
EXPORT_SYMBOL(cl_sync_io_wait);
static inline void dio_aio_complete(struct kiocb *iocb, ssize_t res)
{
if (iocb->ki_complete)
#ifdef HAVE_KIOCB_COMPLETE_2ARGS
iocb->ki_complete(iocb, res);
#else
iocb->ki_complete(iocb, res, 0);
#endif
}
static void cl_dio_aio_end(const struct lu_env *env, struct cl_sync_io *anchor)
{
struct cl_dio_aio *aio = container_of(anchor, typeof(*aio), cda_sync);
ssize_t ret = anchor->csi_sync_rc;
ENTRY;
if (!aio->cda_no_aio_complete)
dio_aio_complete(aio->cda_iocb, ret ?: aio->cda_bytes);
EXIT;
}
static inline void csd_dup_free(struct cl_iter_dup *dup)
{
void *tmp = dup->id_vec;
dup->id_vec = NULL;
OBD_FREE(tmp, dup->id_vec_size);
}
static void cl_sub_dio_end(const struct lu_env *env, struct cl_sync_io *anchor)
{
struct cl_sub_dio *sdio = container_of(anchor, typeof(*sdio), csd_sync);
struct cl_dio_pages *cdp = &sdio->csd_dio_pages;
ssize_t ret = anchor->csi_sync_rc;
bool array_incomplete = false;
int i;
ENTRY;
if (cdp->cdp_cl_pages) {
for (i = 0; i < cdp->cdp_page_count; i++) {
struct cl_page *page = cdp->cdp_cl_pages[i];
/* if we failed allocating pages, the page array may be
* incomplete, so check the pointers
*
* FIXME: This extra tracking of array completeness is
* just a debug check and will be removed later in the
* series.
*/
if (page)
cl_page_put(env, page);
else if (array_incomplete)
LASSERT(!page);
else
array_incomplete = true;
}
OBD_FREE_PTR_ARRAY_LARGE(cdp->cdp_cl_pages,
cdp->cdp_page_count);
}
if (sdio->csd_unaligned) {
CDEBUG(D_VFSTRACE,
"finishing unaligned dio %s aio->cda_bytes %ld\n",
sdio->csd_write ? "write" : "read", sdio->csd_bytes);
/* read copies *from* the kernel buffer *to* userspace
* here at the end, write copies *to* the kernel
* buffer from userspace at the start
*/
if (!sdio->csd_write && sdio->csd_bytes > 0)
ret = ll_dio_user_copy(sdio);
ll_free_dio_buffer(cdp);
/* handle freeing here rather than in cl_sub_dio_free
* because we have the unmodified iovec pointer
*/
csd_dup_free(&sdio->csd_dup);
} else {
/* unaligned DIO does not get user pages, so it doesn't have to
* release them, but aligned I/O must
*/
ll_release_user_pages(cdp->cdp_pages, cdp->cdp_page_count);
}
cl_sync_io_note(env, &sdio->csd_ll_aio->cda_sync, ret);
EXIT;
}
struct cl_dio_aio *cl_dio_aio_alloc(struct kiocb *iocb, struct cl_object *obj,
bool is_aio)
{
struct cl_dio_aio *aio;
OBD_SLAB_ALLOC_PTR_GFP(aio, cl_dio_aio_kmem, GFP_NOFS);
if (aio != NULL) {
/*
* Hold one ref so that it won't be released until
* every pages is added.
*/
cl_sync_io_init_notify(&aio->cda_sync, 1, aio, cl_dio_aio_end);
aio->cda_iocb = iocb;
aio->cda_is_aio = is_aio;
aio->cda_no_aio_complete = !is_aio;
/* if this is true AIO, the memory is freed by the last call
* to cl_sync_io_note (when all the I/O is complete), because
* no one is waiting (in the kernel) for this to complete
*
* in other cases, the last user is cl_sync_io_wait, and in
* that case, the creator frees the struct after that call
*/
aio->cda_creator_free = !is_aio;
cl_object_get(obj);
aio->cda_obj = obj;
aio->cda_mm = get_task_mm(current);
}
return aio;
}
EXPORT_SYMBOL(cl_dio_aio_alloc);
struct cl_sub_dio *cl_sub_dio_alloc(struct cl_dio_aio *ll_aio,
struct iov_iter *iter, bool write,
bool unaligned, bool sync)
{
struct cl_sub_dio *sdio;
OBD_SLAB_ALLOC_PTR_GFP(sdio, cl_sub_dio_kmem, GFP_NOFS);
if (sdio != NULL) {
/*
* Hold one ref so that it won't be released until
* every pages is added.
*/
cl_sync_io_init_notify(&sdio->csd_sync, 1, sdio,
cl_sub_dio_end);
sdio->csd_ll_aio = ll_aio;
sdio->csd_creator_free = sync;
sdio->csd_write = write;
sdio->csd_unaligned = unaligned;
init_waitqueue_head(&sdio->csd_write_waitq);
spin_lock_init(&sdio->csd_write_lock);
atomic_add(1, &ll_aio->cda_sync.csi_sync_nr);
if (sdio->csd_unaligned) {
size_t v_sz = 0;
/* we need to make a copy of the user iovec at this
* point in time, in order to:
*
* A) have the correct state of the iovec for this
* chunk of I/O, ie, the main iovec is altered as we do
* I/O and this chunk needs the current state
* B) have a chunk-local copy; doing the IO later
* modifies the iovec, so to process each chunk from a
* separate thread requires a local copy of the iovec
*/
sdio->csd_iter = *iter;
if (iov_iter_is_bvec(iter))
v_sz = iter->nr_segs * sizeof(struct bio_vec);
else if (iov_iter_is_kvec(iter) || iter_is_iovec(iter))
v_sz = iter->nr_segs * sizeof(struct iovec);
/* xarray and discard do not need vec to be dup'd */
if (!v_sz)
goto out;
OBD_ALLOC(sdio->csd_dup.id_vec, v_sz);
if (!sdio->csd_dup.id_vec) {
cl_sub_dio_free(sdio);
sdio = NULL;
goto out;
}
memcpy(sdio->csd_dup.id_vec, iter->__iov, v_sz);
sdio->csd_dup.id_vec_size = v_sz;
sdio->csd_iter.__iov = sdio->csd_dup.id_vec;
}
}
out:
return sdio;
}
EXPORT_SYMBOL(cl_sub_dio_alloc);
void cl_dio_aio_free(const struct lu_env *env, struct cl_dio_aio *aio)
{
if (aio) {
if (aio->cda_mm)
mmput(aio->cda_mm);
cl_object_put(env, aio->cda_obj);
OBD_SLAB_FREE_PTR(aio, cl_dio_aio_kmem);
}
}
EXPORT_SYMBOL(cl_dio_aio_free);
void cl_sub_dio_free(struct cl_sub_dio *sdio)
{
if (sdio) {
if (sdio->csd_dup.id_vec) {
LASSERT(sdio->csd_unaligned);
csd_dup_free(&sdio->csd_dup);
sdio->csd_iter.__iov = NULL;
}
OBD_SLAB_FREE_PTR(sdio, cl_sub_dio_kmem);
}
}
EXPORT_SYMBOL(cl_sub_dio_free);
/*
* For unaligned DIO.
*
* Allocate the internal buffer from/to which we will perform DIO. This takes
* the user I/O parameters and allocates an internal buffer large enough to
* hold it. The pages in this buffer are aligned with pages in the file (ie,
* they have a 1-to-1 mapping with file pages).
*/
int ll_allocate_dio_buffer(struct cl_dio_pages *cdp, size_t io_size)
{
size_t pg_offset;
int result = 0;
ENTRY;
/* page level offset in the file where the I/O starts */
pg_offset = cdp->cdp_file_offset & ~PAGE_MASK;
/* this adds 1 for the first page and removes the bytes in it from the
* io_size, making the rest of the calculation aligned
*/
if (pg_offset) {
cdp->cdp_page_count++;
io_size -= min_t(size_t, PAGE_SIZE - pg_offset, io_size);
}
/* calculate pages for the rest of the buffer */
cdp->cdp_page_count += (io_size + PAGE_SIZE - 1) >> PAGE_SHIFT;
cdp->cdp_pages = kvzalloc(cdp->cdp_page_count * sizeof(struct page *),
GFP_NOFS);
if (cdp->cdp_pages == NULL)
GOTO(out, result = -ENOMEM);
if (CFS_FAIL_CHECK(OBD_FAIL_LLITE_DIO_BUFFER_ALLOC) ||
CFS_FAIL_CHECK(OBD_FAIL_LLITE_DIO_DRAIN_RETRY))
GOTO(out, result = -ENOMEM);
result = obd_pool_get_pages_array(cdp->cdp_pages, cdp->cdp_page_count);
if (result)
GOTO(out, result);
out:
if (result)
ll_free_dio_buffer(cdp);
if (result == 0)
result = cdp->cdp_page_count;
RETURN(result);
}
EXPORT_SYMBOL(ll_allocate_dio_buffer);
void ll_free_dio_buffer(struct cl_dio_pages *cdp)
{
if (!cdp->cdp_pages)
return;
obd_pool_put_pages_array(cdp->cdp_pages, cdp->cdp_page_count);
kvfree(cdp->cdp_pages);
cdp->cdp_pages = NULL;
}
EXPORT_SYMBOL(ll_free_dio_buffer);
/*
* ll_release_user_pages - tear down page struct array
* @pages: array of page struct pointers underlying target buffer
*/
void ll_release_user_pages(struct page **pages, int npages)
{
int i;
if (npages == 0) {
LASSERT(!pages);
return;
}
for (i = 0; i < npages; i++) {
if (!pages[i])
break;
put_page(pages[i]);
}
kvfree(pages);
}
EXPORT_SYMBOL(ll_release_user_pages);
#ifdef HAVE_FAULT_IN_IOV_ITER_READABLE
#define ll_iov_iter_fault_in_readable(iov, bytes) \
fault_in_iov_iter_readable(iov, bytes)
#else
#define ll_iov_iter_fault_in_readable(iov, bytes) \
iov_iter_fault_in_readable(iov, bytes)
#endif
#ifndef HAVE_KTHREAD_USE_MM
#define kthread_use_mm(mm) use_mm(mm)
#define kthread_unuse_mm(mm) unuse_mm(mm)
#endif
static inline size_t folio_from_iter(struct page *pg,
unsigned long offset, size_t bytes,
struct iov_iter *iter)
{
size_t copied; /* bytes successfully copied */
#if defined(HAVE_COPY_FOLIO_FROM_ITER_ATOMIC)
struct folio *folio = page_folio(pg);
int pgno = folio_page_idx(folio, pg);
offset += pgno * PAGE_SIZE;
copied = copy_folio_from_iter_atomic(folio, offset, bytes, iter);
flush_dcache_folio(folio);
#elif defined(HAVE_COPY_PAGE_FROM_ITER_ATOMIC)
copied = copy_page_from_iter_atomic(pg, offset, bytes, iter);
flush_dcache_page(pg);
#else
copied = iov_iter_copy_from_user_atomic(pg, iter, offset, bytes);
iov_iter_advance(iter, copied);
flush_dcache_page(pg);
#endif
return copied;
}
static inline size_t folio_to_iter(struct page *pg,
unsigned long offset, size_t bytes,
struct iov_iter *iter)
{
size_t copied; /* bytes successfully copied */
#ifdef HAVE___FILEMAP_GET_FOLIO
struct folio *folio = page_folio(pg);
int pgno = folio_page_idx(folio, pg);
offset += pgno * PAGE_SIZE;
copied = copy_folio_to_iter(folio, offset, bytes, iter);
#else
copied = copy_page_to_iter(pg, offset, bytes, iter);
#endif
return copied;
}
static inline size_t folio_iter(struct page *page,
unsigned long offset, size_t bytes,
struct iov_iter *iter, int rw)
{
if (rw == WRITE)
return folio_from_iter(page, offset, bytes, iter);
return folio_to_iter(page, offset, bytes, iter);
}
/* copy IO data to/from internal buffer and userspace iovec */
static ssize_t __ll_dio_user_copy(struct cl_sub_dio *sdio)
{
struct iov_iter *iter = &sdio->csd_iter;
struct cl_dio_pages *cdp = &sdio->csd_dio_pages;
struct mm_struct *mm = sdio->csd_ll_aio->cda_mm;
loff_t pos = cdp->cdp_file_offset;
size_t count = sdio->csd_bytes;
size_t original_count = count;
int short_copies = 0;
bool mm_used = false;
unsigned int i = 0;
int status = 0;
int rw;
ENTRY;
LASSERT(sdio->csd_unaligned);
if (sdio->csd_write)
rw = WRITE;
else
rw = READ;
/* read copying is protected by the reference count on the sdio, since
* it's done as part of getting rid of the sdio, but write copying is
* done at the start, where there may be multiple ptlrpcd threads
* using this sdio, so we must synchronize access
*/
if (rw == WRITE) {
unsigned long flags;
/* Use the wait queue's internal spinlock to protect state.
* We only hold it briefly to check/update flags, not during
* the actual copy operations which can sleep.
*/
spin_lock_irqsave(&sdio->csd_write_lock, flags);
/* Wait if another thread is currently copying */
while (sdio->csd_write_copying && !sdio->csd_write_copied) {
DEFINE_WAIT(wait);
prepare_to_wait(&sdio->csd_write_waitq, &wait,
TASK_UNINTERRUPTIBLE);
spin_unlock_irqrestore(&sdio->csd_write_lock, flags);
schedule();
spin_lock_irqsave(&sdio->csd_write_lock, flags);
finish_wait(&sdio->csd_write_waitq, &wait);
}
/* If copy is already done (or failed), return the status */
if (sdio->csd_write_copied) {
ssize_t ret = sdio->csd_write_status;
spin_unlock_irqrestore(&sdio->csd_write_lock, flags);
RETURN(ret);
}
/* We're the first thread here, claim the copy operation */
sdio->csd_write_copying = true;
spin_unlock_irqrestore(&sdio->csd_write_lock, flags);
}
/* if there's no mm, io is being done from a kernel thread, so there's
* no need to transition to its mm context anyway.
*
* Also, if mm == current->mm, that means this is being handled in the
* thread which created it, and not in a separate kthread - so it is
* unnecessary (and incorrect) to do a use_mm here
*
* assert that if we have an mm and it's not ours, we're doing this
* copying from a kernel thread - otherwise kthread_use_mm will happily
* trash memory and crash later
*/
if (mm && mm != current->mm) {
LASSERT(current->flags & PF_KTHREAD);
kthread_use_mm(mm);
mm_used = true;
}
/* fault in the entire userspace iovec */
if (rw == WRITE) {
if (unlikely(ll_iov_iter_fault_in_readable(iter, count)))
GOTO(out, status = -EFAULT);
}
if (CFS_FAIL_CHECK(OBD_FAIL_LLITE_DIO_COPY_ERR))
GOTO(out, status = -EFAULT);
/* modeled on kernel generic_file_buffered_read/write()
*
* note we only have one 'chunk' of i/o here, so we do not copy the
* whole iovec here (except when the chunk is the whole iovec) so we
* use the count of bytes in the chunk, csd_bytes, instead of looking
* at the iovec
*/
while (true) {
struct page *page = cdp->cdp_pages[i];
unsigned long offset; /* offset into kernel buffer page */
size_t copied; /* bytes successfully copied */
size_t bytes; /* bytes to copy for this page */
LASSERT(i < cdp->cdp_page_count);
offset = pos & ~PAGE_MASK;
bytes = min_t(unsigned long, PAGE_SIZE - offset, count);
CDEBUG(D_VFSTRACE,
"count %zd, offset %lu, pos %lld, cdp_page_count %u\n",
count, offset, pos, cdp->cdp_page_count);
if (fatal_signal_pending(current)) {
status = -EINTR;
break;
}
/* like btrfs, we do not have a mapping since this isn't
* a page cache page, so we must do this flush
* unconditionally
*
* NB: This is a noop on x86 but active on other
* architectures
*/
flush_dcache_page(page);
copied = folio_iter(page, offset, bytes, iter, rw);
pos += copied;
count -= copied;
if (unlikely(copied < bytes)) {
short_copies++;
CDEBUG(D_VFSTRACE,
"short copy - copied only %zd of %lu, short %d times\n",
copied, bytes, short_copies);
/* copies will very rarely be interrupted, but we
* should retry in those cases, since the other option
* is giving an IO error and this can occur in normal
* operation such as with racing unaligned AIOs
*
* but of course we should not retry indefinitely
*/
if (short_copies > 2) {
CERROR("Unaligned DIO copy repeatedly short, count %zd, offset %lu, bytes %lu, copied %zd, pos %lld\n",
count, offset, bytes, copied, pos);
status = -EFAULT;
break;
}
continue;
}
if (count == 0)
break;
i++;
}
/* if we complete successfully, we should reach all of the pages */
LASSERTF(ergo(status == 0, i == cdp->cdp_page_count - 1),
"status: %d, i: %d, cdp->cdp_page_count %u, count %zu\n",
status, i, cdp->cdp_page_count, count);
out:
if (mm_used)
kthread_unuse_mm(mm);
/* For write operations, update state and wake any waiting threads */
if (rw == WRITE) {
unsigned long flags;
ssize_t result = original_count - count ?
original_count - count : status;
spin_lock_irqsave(&sdio->csd_write_lock, flags);
/* Store result (bytes copied or error) for waiting threads */
sdio->csd_write_status = result;
/* Mark copy as complete (successfully or not) */
sdio->csd_write_copied = true;
sdio->csd_write_copying = false;
spin_unlock_irqrestore(&sdio->csd_write_lock, flags);
/* Wake up any threads waiting for the copy to complete */
wake_up_all(&sdio->csd_write_waitq);
}
/* the total bytes copied, or status */
RETURN(original_count - count ? original_count - count : status);
}
struct dio_user_copy_data {
struct cl_sub_dio *ducd_sdio;
struct completion ducd_completion;
ssize_t ducd_result;
};
static int ll_dio_user_copy_helper(void *data)
{
struct dio_user_copy_data *ducd = data;
struct cl_sub_dio *sdio = ducd->ducd_sdio;
ducd->ducd_result = __ll_dio_user_copy(sdio);
complete(&ducd->ducd_completion);
return 0;
}
ssize_t ll_dio_user_copy(struct cl_sub_dio *sdio)
{
struct dio_user_copy_data ducd;
struct task_struct *kthread;
/* normal case - copy is being done by ptlrpcd */
if (current->flags & PF_KTHREAD ||
/* for non-parallel DIO, the submitting thread does the copy */
sdio->csd_ll_aio->cda_mm == current->mm)
return __ll_dio_user_copy(sdio);
/* this is a slightly unfortunate workaround; when doing an fsync, a
* user thread may pick up a DIO extent which is about to be written
* out. we can't just ignore these, but we also can't handle them from
* the user thread, since user threads can't do data copying from
* another thread's memory.
*
* so we spawn a kthread to handle this case.
* this will be rare and is not a 'hot path', so the performance
* cost doesn't matter
*/
init_completion(&ducd.ducd_completion);
ducd.ducd_sdio = sdio;
kthread = kthread_run(ll_dio_user_copy_helper, &ducd,
"ll_ucp_%u", current->pid);
if (IS_ERR_OR_NULL(kthread))
return PTR_ERR(kthread);
wait_for_completion(&ducd.ducd_completion);
return ducd.ducd_result;
}
EXPORT_SYMBOL(ll_dio_user_copy);
/*
* Indicate that transfer of a single page completed.
*/
void __cl_sync_io_note(const struct lu_env *env, struct cl_sync_io *anchor,
int count, int ioret)
{
int sync_nr;
ENTRY;
if (anchor->csi_sync_rc == 0 && ioret < 0)
anchor->csi_sync_rc = ioret;
/* because there is no atomic_sub_and_lock, we have to do this slightly
* awkward subtraction when we have count > 1, handling all but 1 of
* our 'count' entries
*/
if (count > 1)
sync_nr = atomic_sub_return(count - 1, &anchor->csi_sync_nr);
else
sync_nr = atomic_read(&anchor->csi_sync_nr);
CDEBUG(D_VFSTRACE,
"Noting completion of %d items, %d items remaining.\n",
count, sync_nr - 1);
/*
* Synchronous IO done without releasing page lock (e.g., as a part of
* ->{prepare,commit}_write(). Completion is used to signal the end of
* IO.
*/
LASSERT(sync_nr > 0);
LASSERT(atomic_read(&anchor->csi_complete) == 0);
if (atomic_dec_and_lock(&anchor->csi_sync_nr,
&anchor->csi_waitq.lock)) {
struct cl_sub_dio *sub_dio_aio = NULL;
struct cl_dio_aio *dio_aio = NULL;
void *csi_dio_aio = NULL;
bool creator_free = true;
cl_sync_io_end_t *end_io = anchor->csi_end_io;
spin_unlock(&anchor->csi_waitq.lock);
/* we cannot do end_io while holding a spin lock, because
* end_io may sleep
*/
if (end_io)
end_io(env, anchor);
spin_lock(&anchor->csi_waitq.lock);
/* this tells the waiters we've completed, and can only be set
* after end_io() has been called and while we're holding the
* spinlock
*/
atomic_set(&anchor->csi_complete, 1);
/*
* Holding the lock across both the decrement and
* the wakeup ensures cl_sync_io_wait() doesn't complete
* before the wakeup completes and the contents of
* of anchor become unsafe to access as the owner is free
* to immediately reclaim anchor when cl_sync_io_wait()
* completes.
*/
wake_up_locked(&anchor->csi_waitq);
csi_dio_aio = anchor->csi_dio_aio;
sub_dio_aio = csi_dio_aio;
dio_aio = csi_dio_aio;
if (csi_dio_aio && end_io == cl_dio_aio_end)
creator_free = dio_aio->cda_creator_free;
else if (csi_dio_aio && end_io == cl_sub_dio_end)
creator_free = sub_dio_aio->csd_creator_free;
spin_unlock(&anchor->csi_waitq.lock);
if (csi_dio_aio && !creator_free) {
if (end_io == cl_dio_aio_end)
cl_dio_aio_free(env, dio_aio);
else if (end_io == cl_sub_dio_end)
cl_sub_dio_free(sub_dio_aio);
}
}
EXIT;
}
EXPORT_SYMBOL(__cl_sync_io_note);
void cl_sync_io_note(const struct lu_env *env, struct cl_sync_io *anchor,
int ioret)
{
__cl_sync_io_note(env, anchor, 1, ioret);
}
EXPORT_SYMBOL(cl_sync_io_note);
/* this function waits for completion of outstanding io and then re-initializes
* the anchor used to track it. This is used to wait to complete DIO before
* returning to userspace, and is never called for true AIO
*/
int cl_sync_io_wait_recycle(const struct lu_env *env, struct cl_sync_io *anchor,
long timeout, int ioret)
{
int rc = 0;
/*
* @anchor was inited as 1 to prevent end_io to be
* called before we add all pages for IO, so drop
* one extra reference to make sure we could wait
* count to be zero.
*/
cl_sync_io_note(env, anchor, ioret);
/* Wait for completion of outstanding dio before re-initializing for
* possible restart
*/
rc = cl_sync_io_wait(env, anchor, timeout);
/**
* One extra reference again, as if @anchor is
* reused we assume it as 1 before using.
*/
atomic_add(1, &anchor->csi_sync_nr);
/* we must also set this anchor as incomplete */
atomic_set(&anchor->csi_complete, 0);
return rc;
}
EXPORT_SYMBOL(cl_sync_io_wait_recycle);
