#define NFSDBG_FACILITY NFSDBG_PAGECACHE
static struct kmem_cache *nfs_page_cachep;
+static const struct rpc_call_ops nfs_pgio_common_ops;
+
+static void nfs_free_request(struct nfs_page *);
static bool nfs_pgarray_set(struct nfs_page_array *p, unsigned int pagecount)
{
{
if (atomic_dec_and_test(&c->io_count)) {
clear_bit(NFS_IO_INPROGRESS, &c->flags);
- smp_mb__after_clear_bit();
+ smp_mb__after_atomic();
wake_up_bit(&c->flags, NFS_IO_INPROGRESS);
}
}
return __nfs_iocounter_wait(c);
}
+static int nfs_wait_bit_uninterruptible(void *word)
+{
+ io_schedule();
+ return 0;
+}
+
+/*
+ * nfs_page_group_lock - lock the head of the page group
+ * @req - request in group that is to be locked
+ *
+ * this lock must be held if modifying the page group list
+ */
+void
+nfs_page_group_lock(struct nfs_page *req)
+{
+ struct nfs_page *head = req->wb_head;
+
+ WARN_ON_ONCE(head != head->wb_head);
+
+ wait_on_bit_lock(&head->wb_flags, PG_HEADLOCK,
+ nfs_wait_bit_uninterruptible,
+ TASK_UNINTERRUPTIBLE);
+}
+
+/*
+ * nfs_page_group_unlock - unlock the head of the page group
+ * @req - request in group that is to be unlocked
+ */
+void
+nfs_page_group_unlock(struct nfs_page *req)
+{
+ struct nfs_page *head = req->wb_head;
+
+ WARN_ON_ONCE(head != head->wb_head);
+
+ smp_mb__before_atomic();
+ clear_bit(PG_HEADLOCK, &head->wb_flags);
+ smp_mb__after_atomic();
+ wake_up_bit(&head->wb_flags, PG_HEADLOCK);
+}
+
+/*
+ * nfs_page_group_sync_on_bit_locked
+ *
+ * must be called with page group lock held
+ */
+static bool
+nfs_page_group_sync_on_bit_locked(struct nfs_page *req, unsigned int bit)
+{
+ struct nfs_page *head = req->wb_head;
+ struct nfs_page *tmp;
+
+ WARN_ON_ONCE(!test_bit(PG_HEADLOCK, &head->wb_flags));
+ WARN_ON_ONCE(test_and_set_bit(bit, &req->wb_flags));
+
+ tmp = req->wb_this_page;
+ while (tmp != req) {
+ if (!test_bit(bit, &tmp->wb_flags))
+ return false;
+ tmp = tmp->wb_this_page;
+ }
+
+ /* true! reset all bits */
+ tmp = req;
+ do {
+ clear_bit(bit, &tmp->wb_flags);
+ tmp = tmp->wb_this_page;
+ } while (tmp != req);
+
+ return true;
+}
+
+/*
+ * nfs_page_group_sync_on_bit - set bit on current request, but only
+ * return true if the bit is set for all requests in page group
+ * @req - request in page group
+ * @bit - PG_* bit that is used to sync page group
+ */
+bool nfs_page_group_sync_on_bit(struct nfs_page *req, unsigned int bit)
+{
+ bool ret;
+
+ nfs_page_group_lock(req);
+ ret = nfs_page_group_sync_on_bit_locked(req, bit);
+ nfs_page_group_unlock(req);
+
+ return ret;
+}
+
+/*
+ * nfs_page_group_init - Initialize the page group linkage for @req
+ * @req - a new nfs request
+ * @prev - the previous request in page group, or NULL if @req is the first
+ * or only request in the group (the head).
+ */
+static inline void
+nfs_page_group_init(struct nfs_page *req, struct nfs_page *prev)
+{
+ WARN_ON_ONCE(prev == req);
+
+ if (!prev) {
+ req->wb_head = req;
+ req->wb_this_page = req;
+ } else {
+ WARN_ON_ONCE(prev->wb_this_page != prev->wb_head);
+ WARN_ON_ONCE(!test_bit(PG_HEADLOCK, &prev->wb_head->wb_flags));
+ req->wb_head = prev->wb_head;
+ req->wb_this_page = prev->wb_this_page;
+ prev->wb_this_page = req;
+
+ /* grab extra ref if head request has extra ref from
+ * the write/commit path to handle handoff between write
+ * and commit lists */
+ if (test_bit(PG_INODE_REF, &prev->wb_head->wb_flags))
+ kref_get(&req->wb_kref);
+ }
+}
+
+/*
+ * nfs_page_group_destroy - sync the destruction of page groups
+ * @req - request that no longer needs the page group
+ *
+ * releases the page group reference from each member once all
+ * members have called this function.
+ */
+static void
+nfs_page_group_destroy(struct kref *kref)
+{
+ struct nfs_page *req = container_of(kref, struct nfs_page, wb_kref);
+ struct nfs_page *tmp, *next;
+
+ if (!nfs_page_group_sync_on_bit(req, PG_TEARDOWN))
+ return;
+
+ tmp = req;
+ do {
+ next = tmp->wb_this_page;
+ /* unlink and free */
+ tmp->wb_this_page = tmp;
+ tmp->wb_head = tmp;
+ nfs_free_request(tmp);
+ tmp = next;
+ } while (tmp != req);
+}
+
/**
* nfs_create_request - Create an NFS read/write request.
* @ctx: open context to use
- * @inode: inode to which the request is attached
* @page: page to write
+ * @last: last nfs request created for this page group or NULL if head
* @offset: starting offset within the page for the write
* @count: number of bytes to read/write
*
* User should ensure it is safe to sleep in this function.
*/
struct nfs_page *
-nfs_create_request(struct nfs_open_context *ctx, struct inode *inode,
- struct page *page,
- unsigned int offset, unsigned int count)
+nfs_create_request(struct nfs_open_context *ctx, struct page *page,
+ struct nfs_page *last, unsigned int offset,
+ unsigned int count)
{
struct nfs_page *req;
struct nfs_lock_context *l_ctx;
req->wb_bytes = count;
req->wb_context = get_nfs_open_context(ctx);
kref_init(&req->wb_kref);
+ nfs_page_group_init(req, last);
return req;
}
printk(KERN_ERR "NFS: Invalid unlock attempted\n");
BUG();
}
- smp_mb__before_clear_bit();
+ smp_mb__before_atomic();
clear_bit(PG_BUSY, &req->wb_flags);
- smp_mb__after_clear_bit();
+ smp_mb__after_atomic();
wake_up_bit(&req->wb_flags, PG_BUSY);
}
}
}
-
/**
* nfs_release_request - Release the count on an NFS read/write request
* @req: request to release
*
* Note: Should never be called with the spinlock held!
*/
-static void nfs_free_request(struct kref *kref)
+static void nfs_free_request(struct nfs_page *req)
{
- struct nfs_page *req = container_of(kref, struct nfs_page, wb_kref);
+ WARN_ON_ONCE(req->wb_this_page != req);
+
+ /* extra debug: make sure no sync bits are still set */
+ WARN_ON_ONCE(test_bit(PG_TEARDOWN, &req->wb_flags));
+ WARN_ON_ONCE(test_bit(PG_UNLOCKPAGE, &req->wb_flags));
+ WARN_ON_ONCE(test_bit(PG_UPTODATE, &req->wb_flags));
+ WARN_ON_ONCE(test_bit(PG_WB_END, &req->wb_flags));
+ WARN_ON_ONCE(test_bit(PG_REMOVE, &req->wb_flags));
/* Release struct file and open context */
nfs_clear_request(req);
void nfs_release_request(struct nfs_page *req)
{
- kref_put(&req->wb_kref, nfs_free_request);
-}
-
-static int nfs_wait_bit_uninterruptible(void *word)
-{
- io_schedule();
- return 0;
+ kref_put(&req->wb_kref, nfs_page_group_destroy);
}
/**
TASK_UNINTERRUPTIBLE);
}
-bool nfs_generic_pg_test(struct nfs_pageio_descriptor *desc, struct nfs_page *prev, struct nfs_page *req)
+/*
+ * nfs_generic_pg_test - determine if requests can be coalesced
+ * @desc: pointer to descriptor
+ * @prev: previous request in desc, or NULL
+ * @req: this request
+ *
+ * Returns zero if @req can be coalesced into @desc, otherwise it returns
+ * the size of the request.
+ */
+size_t nfs_generic_pg_test(struct nfs_pageio_descriptor *desc,
+ struct nfs_page *prev, struct nfs_page *req)
{
- /*
- * FIXME: ideally we should be able to coalesce all requests
- * that are not block boundary aligned, but currently this
- * is problematic for the case of bsize < PAGE_CACHE_SIZE,
- * since nfs_flush_multi and nfs_pagein_multi assume you
- * can have only one struct nfs_page.
- */
- if (desc->pg_bsize < PAGE_SIZE)
+ if (desc->pg_count > desc->pg_bsize) {
+ /* should never happen */
+ WARN_ON_ONCE(1);
return 0;
+ }
- return desc->pg_count + req->wb_bytes <= desc->pg_bsize;
+ return min(desc->pg_bsize - desc->pg_count, (size_t)req->wb_bytes);
}
EXPORT_SYMBOL_GPL(nfs_generic_pg_test);
struct nfs_pgio_header *hdr = &header->header;
INIT_LIST_HEAD(&hdr->pages);
- INIT_LIST_HEAD(&hdr->rpc_list);
spin_lock_init(&hdr->lock);
atomic_set(&hdr->refcnt, 0);
hdr->rw_ops = ops;
* @hdr: The header making a request
* @pagecount: Number of pages to create
*/
-struct nfs_pgio_data *nfs_pgio_data_alloc(struct nfs_pgio_header *hdr,
- unsigned int pagecount)
+static struct nfs_pgio_data *nfs_pgio_data_alloc(struct nfs_pgio_header *hdr,
+ unsigned int pagecount)
{
struct nfs_pgio_data *data, *prealloc;
}
EXPORT_SYMBOL_GPL(nfs_pgio_data_release);
+/**
+ * nfs_pgio_rpcsetup - Set up arguments for a pageio call
+ * @data: The pageio data
+ * @count: Number of bytes to read
+ * @offset: Initial offset
+ * @how: How to commit data (writes only)
+ * @cinfo: Commit information for the call (writes only)
+ */
+static void nfs_pgio_rpcsetup(struct nfs_pgio_data *data,
+ unsigned int count, unsigned int offset,
+ int how, struct nfs_commit_info *cinfo)
+{
+ struct nfs_page *req = data->header->req;
+
+ /* Set up the RPC argument and reply structs
+ * NB: take care not to mess about with data->commit et al. */
+
+ data->args.fh = NFS_FH(data->header->inode);
+ data->args.offset = req_offset(req) + offset;
+ /* pnfs_set_layoutcommit needs this */
+ data->mds_offset = data->args.offset;
+ data->args.pgbase = req->wb_pgbase + offset;
+ data->args.pages = data->pages.pagevec;
+ data->args.count = count;
+ data->args.context = get_nfs_open_context(req->wb_context);
+ data->args.lock_context = req->wb_lock_context;
+ data->args.stable = NFS_UNSTABLE;
+ switch (how & (FLUSH_STABLE | FLUSH_COND_STABLE)) {
+ case 0:
+ break;
+ case FLUSH_COND_STABLE:
+ if (nfs_reqs_to_commit(cinfo))
+ break;
+ default:
+ data->args.stable = NFS_FILE_SYNC;
+ }
+
+ data->res.fattr = &data->fattr;
+ data->res.count = count;
+ data->res.eof = 0;
+ data->res.verf = &data->verf;
+ nfs_fattr_init(&data->fattr);
+}
+
/**
* nfs_pgio_prepare - Prepare pageio data to go over the wire
* @task: The current task
rpc_exit(task, err);
}
+int nfs_initiate_pgio(struct rpc_clnt *clnt, struct nfs_pgio_data *data,
+ const struct rpc_call_ops *call_ops, int how, int flags)
+{
+ struct rpc_task *task;
+ struct rpc_message msg = {
+ .rpc_argp = &data->args,
+ .rpc_resp = &data->res,
+ .rpc_cred = data->header->cred,
+ };
+ struct rpc_task_setup task_setup_data = {
+ .rpc_client = clnt,
+ .task = &data->task,
+ .rpc_message = &msg,
+ .callback_ops = call_ops,
+ .callback_data = data,
+ .workqueue = nfsiod_workqueue,
+ .flags = RPC_TASK_ASYNC | flags,
+ };
+ int ret = 0;
+
+ data->header->rw_ops->rw_initiate(data, &msg, &task_setup_data, how);
+
+ dprintk("NFS: %5u initiated pgio call "
+ "(req %s/%llu, %u bytes @ offset %llu)\n",
+ data->task.tk_pid,
+ data->header->inode->i_sb->s_id,
+ (unsigned long long)NFS_FILEID(data->header->inode),
+ data->args.count,
+ (unsigned long long)data->args.offset);
+
+ task = rpc_run_task(&task_setup_data);
+ if (IS_ERR(task)) {
+ ret = PTR_ERR(task);
+ goto out;
+ }
+ if (how & FLUSH_SYNC) {
+ ret = rpc_wait_for_completion_task(task);
+ if (ret == 0)
+ ret = task->tk_status;
+ }
+ rpc_put_task(task);
+out:
+ return ret;
+}
+EXPORT_SYMBOL_GPL(nfs_initiate_pgio);
+
+/**
+ * nfs_pgio_error - Clean up from a pageio error
+ * @desc: IO descriptor
+ * @hdr: pageio header
+ */
+static int nfs_pgio_error(struct nfs_pageio_descriptor *desc,
+ struct nfs_pgio_header *hdr)
+{
+ set_bit(NFS_IOHDR_REDO, &hdr->flags);
+ nfs_pgio_data_release(hdr->data);
+ hdr->data = NULL;
+ desc->pg_completion_ops->error_cleanup(&desc->pg_list);
+ return -ENOMEM;
+}
+
/**
* nfs_pgio_release - Release pageio data
* @calldata: The pageio data to release
data->header->rw_ops->rw_result(task, data);
}
+/*
+ * Create an RPC task for the given read or write request and kick it.
+ * The page must have been locked by the caller.
+ *
+ * It may happen that the page we're passed is not marked dirty.
+ * This is the case if nfs_updatepage detects a conflicting request
+ * that has been written but not committed.
+ */
+int nfs_generic_pgio(struct nfs_pageio_descriptor *desc,
+ struct nfs_pgio_header *hdr)
+{
+ struct nfs_page *req;
+ struct page **pages;
+ struct nfs_pgio_data *data;
+ struct list_head *head = &desc->pg_list;
+ struct nfs_commit_info cinfo;
+
+ data = nfs_pgio_data_alloc(hdr, nfs_page_array_len(desc->pg_base,
+ desc->pg_count));
+ if (!data)
+ return nfs_pgio_error(desc, hdr);
+
+ nfs_init_cinfo(&cinfo, desc->pg_inode, desc->pg_dreq);
+ pages = data->pages.pagevec;
+ while (!list_empty(head)) {
+ req = nfs_list_entry(head->next);
+ nfs_list_remove_request(req);
+ nfs_list_add_request(req, &hdr->pages);
+ *pages++ = req->wb_page;
+ }
+
+ if ((desc->pg_ioflags & FLUSH_COND_STABLE) &&
+ (desc->pg_moreio || nfs_reqs_to_commit(&cinfo)))
+ desc->pg_ioflags &= ~FLUSH_COND_STABLE;
+
+ /* Set up the argument struct */
+ nfs_pgio_rpcsetup(data, desc->pg_count, 0, desc->pg_ioflags, &cinfo);
+ hdr->data = data;
+ desc->pg_rpc_callops = &nfs_pgio_common_ops;
+ return 0;
+}
+EXPORT_SYMBOL_GPL(nfs_generic_pgio);
+
+static int nfs_generic_pg_pgios(struct nfs_pageio_descriptor *desc)
+{
+ struct nfs_rw_header *rw_hdr;
+ struct nfs_pgio_header *hdr;
+ int ret;
+
+ rw_hdr = nfs_rw_header_alloc(desc->pg_rw_ops);
+ if (!rw_hdr) {
+ desc->pg_completion_ops->error_cleanup(&desc->pg_list);
+ return -ENOMEM;
+ }
+ hdr = &rw_hdr->header;
+ nfs_pgheader_init(desc, hdr, nfs_rw_header_free);
+ atomic_inc(&hdr->refcnt);
+ ret = nfs_generic_pgio(desc, hdr);
+ if (ret == 0)
+ ret = nfs_initiate_pgio(NFS_CLIENT(hdr->inode),
+ hdr->data, desc->pg_rpc_callops,
+ desc->pg_ioflags, 0);
+ if (atomic_dec_and_test(&hdr->refcnt))
+ hdr->completion_ops->completion(hdr);
+ return ret;
+}
+
static bool nfs_match_open_context(const struct nfs_open_context *ctx1,
const struct nfs_open_context *ctx2)
{
struct nfs_page *req,
struct nfs_pageio_descriptor *pgio)
{
- if (!nfs_match_open_context(req->wb_context, prev->wb_context))
- return false;
- if (req->wb_context->dentry->d_inode->i_flock != NULL &&
- !nfs_match_lock_context(req->wb_lock_context, prev->wb_lock_context))
- return false;
- if (req->wb_pgbase != 0)
- return false;
- if (prev->wb_pgbase + prev->wb_bytes != PAGE_CACHE_SIZE)
- return false;
- if (req_offset(req) != req_offset(prev) + prev->wb_bytes)
- return false;
- return pgio->pg_ops->pg_test(pgio, prev, req);
+ size_t size;
+
+ if (prev) {
+ if (!nfs_match_open_context(req->wb_context, prev->wb_context))
+ return false;
+ if (req->wb_context->dentry->d_inode->i_flock != NULL &&
+ !nfs_match_lock_context(req->wb_lock_context,
+ prev->wb_lock_context))
+ return false;
+ if (req_offset(req) != req_offset(prev) + prev->wb_bytes)
+ return false;
+ }
+ size = pgio->pg_ops->pg_test(pgio, prev, req);
+ WARN_ON_ONCE(size > req->wb_bytes);
+ if (size && size < req->wb_bytes)
+ req->wb_bytes = size;
+ return size > 0;
}
/**
static int nfs_pageio_do_add_request(struct nfs_pageio_descriptor *desc,
struct nfs_page *req)
{
+ struct nfs_page *prev = NULL;
if (desc->pg_count != 0) {
- struct nfs_page *prev;
-
prev = nfs_list_entry(desc->pg_list.prev);
- if (!nfs_can_coalesce_requests(prev, req, desc))
- return 0;
} else {
if (desc->pg_ops->pg_init)
desc->pg_ops->pg_init(desc, req);
desc->pg_base = req->wb_pgbase;
}
+ if (!nfs_can_coalesce_requests(prev, req, desc))
+ return 0;
nfs_list_remove_request(req);
nfs_list_add_request(req, &desc->pg_list);
desc->pg_count += req->wb_bytes;
* @desc: destination io descriptor
* @req: request
*
+ * This may split a request into subrequests which are all part of the
+ * same page group.
+ *
* Returns true if the request 'req' was successfully coalesced into the
* existing list of pages 'desc'.
*/
static int __nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
struct nfs_page *req)
{
- while (!nfs_pageio_do_add_request(desc, req)) {
- desc->pg_moreio = 1;
- nfs_pageio_doio(desc);
- if (desc->pg_error < 0)
- return 0;
- desc->pg_moreio = 0;
- if (desc->pg_recoalesce)
- return 0;
- }
+ struct nfs_page *subreq;
+ unsigned int bytes_left = 0;
+ unsigned int offset, pgbase;
+
+ nfs_page_group_lock(req);
+
+ subreq = req;
+ bytes_left = subreq->wb_bytes;
+ offset = subreq->wb_offset;
+ pgbase = subreq->wb_pgbase;
+
+ do {
+ if (!nfs_pageio_do_add_request(desc, subreq)) {
+ /* make sure pg_test call(s) did nothing */
+ WARN_ON_ONCE(subreq->wb_bytes != bytes_left);
+ WARN_ON_ONCE(subreq->wb_offset != offset);
+ WARN_ON_ONCE(subreq->wb_pgbase != pgbase);
+
+ nfs_page_group_unlock(req);
+ desc->pg_moreio = 1;
+ nfs_pageio_doio(desc);
+ if (desc->pg_error < 0)
+ return 0;
+ desc->pg_moreio = 0;
+ if (desc->pg_recoalesce)
+ return 0;
+ /* retry add_request for this subreq */
+ nfs_page_group_lock(req);
+ continue;
+ }
+
+ /* check for buggy pg_test call(s) */
+ WARN_ON_ONCE(subreq->wb_bytes + subreq->wb_pgbase > PAGE_SIZE);
+ WARN_ON_ONCE(subreq->wb_bytes > bytes_left);
+ WARN_ON_ONCE(subreq->wb_bytes == 0);
+
+ bytes_left -= subreq->wb_bytes;
+ offset += subreq->wb_bytes;
+ pgbase += subreq->wb_bytes;
+
+ if (bytes_left) {
+ subreq = nfs_create_request(req->wb_context,
+ req->wb_page,
+ subreq, pgbase, bytes_left);
+ if (IS_ERR(subreq))
+ goto err_ptr;
+ nfs_lock_request(subreq);
+ subreq->wb_offset = offset;
+ subreq->wb_index = req->wb_index;
+ }
+ } while (bytes_left > 0);
+
+ nfs_page_group_unlock(req);
return 1;
+err_ptr:
+ desc->pg_error = PTR_ERR(subreq);
+ nfs_page_group_unlock(req);
+ return 0;
}
static int nfs_do_recoalesce(struct nfs_pageio_descriptor *desc)
kmem_cache_destroy(nfs_page_cachep);
}
-const struct rpc_call_ops nfs_pgio_common_ops = {
+static const struct rpc_call_ops nfs_pgio_common_ops = {
.rpc_call_prepare = nfs_pgio_prepare,
.rpc_call_done = nfs_pgio_result,
.rpc_release = nfs_pgio_release,
};
+
+const struct nfs_pageio_ops nfs_pgio_rw_ops = {
+ .pg_test = nfs_generic_pg_test,
+ .pg_doio = nfs_generic_pg_pgios,
+};
projects / karo-tx-linux.git / blobdiff