diff options
Diffstat (limited to 'fs/nfs/read.c')
-rw-r--r-- | fs/nfs/read.c | 437 |
1 files changed, 211 insertions, 226 deletions
diff --git a/fs/nfs/read.c b/fs/nfs/read.c index 0a4be28c2ea3..86ced7836214 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -30,43 +30,73 @@ #define NFSDBG_FACILITY NFSDBG_PAGECACHE static const struct nfs_pageio_ops nfs_pageio_read_ops; -static const struct rpc_call_ops nfs_read_partial_ops; -static const struct rpc_call_ops nfs_read_full_ops; +static const struct rpc_call_ops nfs_read_common_ops; +static const struct nfs_pgio_completion_ops nfs_async_read_completion_ops; static struct kmem_cache *nfs_rdata_cachep; -struct nfs_read_data *nfs_readdata_alloc(unsigned int pagecount) +struct nfs_read_header *nfs_readhdr_alloc(void) { - struct nfs_read_data *p; - - p = kmem_cache_zalloc(nfs_rdata_cachep, GFP_KERNEL); - if (p) { - INIT_LIST_HEAD(&p->pages); - p->npages = pagecount; - if (pagecount <= ARRAY_SIZE(p->page_array)) - p->pagevec = p->page_array; - else { - p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_KERNEL); - if (!p->pagevec) { - kmem_cache_free(nfs_rdata_cachep, p); - p = NULL; - } - } + struct nfs_read_header *rhdr; + + rhdr = kmem_cache_zalloc(nfs_rdata_cachep, GFP_KERNEL); + if (rhdr) { + struct nfs_pgio_header *hdr = &rhdr->header; + + INIT_LIST_HEAD(&hdr->pages); + INIT_LIST_HEAD(&hdr->rpc_list); + spin_lock_init(&hdr->lock); + atomic_set(&hdr->refcnt, 0); + } + return rhdr; +} + +static struct nfs_read_data *nfs_readdata_alloc(struct nfs_pgio_header *hdr, + unsigned int pagecount) +{ + struct nfs_read_data *data, *prealloc; + + prealloc = &container_of(hdr, struct nfs_read_header, header)->rpc_data; + if (prealloc->header == NULL) + data = prealloc; + else + data = kzalloc(sizeof(*data), GFP_KERNEL); + if (!data) + goto out; + + if (nfs_pgarray_set(&data->pages, pagecount)) { + data->header = hdr; + atomic_inc(&hdr->refcnt); + } else { + if (data != prealloc) + kfree(data); + data = NULL; } - return p; +out: + return data; } -void nfs_readdata_free(struct nfs_read_data *p) +void nfs_readhdr_free(struct nfs_pgio_header *hdr) { - if (p && (p->pagevec != &p->page_array[0])) - kfree(p->pagevec); - kmem_cache_free(nfs_rdata_cachep, p); + struct nfs_read_header *rhdr = container_of(hdr, struct nfs_read_header, header); + + kmem_cache_free(nfs_rdata_cachep, rhdr); } void nfs_readdata_release(struct nfs_read_data *rdata) { + struct nfs_pgio_header *hdr = rdata->header; + struct nfs_read_header *read_header = container_of(hdr, struct nfs_read_header, header); + put_nfs_open_context(rdata->args.context); - nfs_readdata_free(rdata); + if (rdata->pages.pagevec != rdata->pages.page_array) + kfree(rdata->pages.pagevec); + if (rdata != &read_header->rpc_data) + kfree(rdata); + else + rdata->header = NULL; + if (atomic_dec_and_test(&hdr->refcnt)) + hdr->completion_ops->completion(hdr); } static @@ -78,39 +108,11 @@ int nfs_return_empty_page(struct page *page) return 0; } -static void nfs_readpage_truncate_uninitialised_page(struct nfs_read_data *data) -{ - unsigned int remainder = data->args.count - data->res.count; - unsigned int base = data->args.pgbase + data->res.count; - unsigned int pglen; - struct page **pages; - - if (data->res.eof == 0 || remainder == 0) - return; - /* - * Note: "remainder" can never be negative, since we check for - * this in the XDR code. - */ - pages = &data->args.pages[base >> PAGE_CACHE_SHIFT]; - base &= ~PAGE_CACHE_MASK; - pglen = PAGE_CACHE_SIZE - base; - for (;;) { - if (remainder <= pglen) { - zero_user(*pages, base, remainder); - break; - } - zero_user(*pages, base, pglen); - pages++; - remainder -= pglen; - pglen = PAGE_CACHE_SIZE; - base = 0; - } -} - void nfs_pageio_init_read_mds(struct nfs_pageio_descriptor *pgio, - struct inode *inode) + struct inode *inode, + const struct nfs_pgio_completion_ops *compl_ops) { - nfs_pageio_init(pgio, inode, &nfs_pageio_read_ops, + nfs_pageio_init(pgio, inode, &nfs_pageio_read_ops, compl_ops, NFS_SERVER(inode)->rsize, 0); } @@ -121,11 +123,12 @@ void nfs_pageio_reset_read_mds(struct nfs_pageio_descriptor *pgio) } EXPORT_SYMBOL_GPL(nfs_pageio_reset_read_mds); -static void nfs_pageio_init_read(struct nfs_pageio_descriptor *pgio, - struct inode *inode) +void nfs_pageio_init_read(struct nfs_pageio_descriptor *pgio, + struct inode *inode, + const struct nfs_pgio_completion_ops *compl_ops) { - if (!pnfs_pageio_init_read(pgio, inode)) - nfs_pageio_init_read_mds(pgio, inode); + if (!pnfs_pageio_init_read(pgio, inode, compl_ops)) + nfs_pageio_init_read_mds(pgio, inode, compl_ops); } int nfs_readpage_async(struct nfs_open_context *ctx, struct inode *inode, @@ -146,9 +149,10 @@ int nfs_readpage_async(struct nfs_open_context *ctx, struct inode *inode, if (len < PAGE_CACHE_SIZE) zero_user_segment(page, len, PAGE_CACHE_SIZE); - nfs_pageio_init_read(&pgio, inode); + nfs_pageio_init_read(&pgio, inode, &nfs_async_read_completion_ops); nfs_pageio_add_request(&pgio, new); nfs_pageio_complete(&pgio); + NFS_I(inode)->read_io += pgio.pg_bytes_written; return 0; } @@ -169,16 +173,49 @@ static void nfs_readpage_release(struct nfs_page *req) nfs_release_request(req); } -int nfs_initiate_read(struct nfs_read_data *data, struct rpc_clnt *clnt, - const struct rpc_call_ops *call_ops) +/* Note io was page aligned */ +static void nfs_read_completion(struct nfs_pgio_header *hdr) +{ + unsigned long bytes = 0; + + if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) + goto out; + while (!list_empty(&hdr->pages)) { + struct nfs_page *req = nfs_list_entry(hdr->pages.next); + struct page *page = req->wb_page; + + if (test_bit(NFS_IOHDR_EOF, &hdr->flags)) { + if (bytes > hdr->good_bytes) + zero_user(page, 0, PAGE_SIZE); + else if (hdr->good_bytes - bytes < PAGE_SIZE) + zero_user_segment(page, + hdr->good_bytes & ~PAGE_MASK, + PAGE_SIZE); + } + bytes += req->wb_bytes; + if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) { + if (bytes <= hdr->good_bytes) + SetPageUptodate(page); + } else + SetPageUptodate(page); + nfs_list_remove_request(req); + nfs_readpage_release(req); + } +out: + hdr->release(hdr); +} + +int nfs_initiate_read(struct rpc_clnt *clnt, + struct nfs_read_data *data, + const struct rpc_call_ops *call_ops, int flags) { - struct inode *inode = data->inode; + struct inode *inode = data->header->inode; int swap_flags = IS_SWAPFILE(inode) ? NFS_RPC_SWAPFLAGS : 0; struct rpc_task *task; struct rpc_message msg = { .rpc_argp = &data->args, .rpc_resp = &data->res, - .rpc_cred = data->cred, + .rpc_cred = data->header->cred, }; struct rpc_task_setup task_setup_data = { .task = &data->task, @@ -187,7 +224,7 @@ int nfs_initiate_read(struct nfs_read_data *data, struct rpc_clnt *clnt, .callback_ops = call_ops, .callback_data = data, .workqueue = nfsiod_workqueue, - .flags = RPC_TASK_ASYNC | swap_flags, + .flags = RPC_TASK_ASYNC | swap_flags | flags, }; /* Set up the initial task struct. */ @@ -212,19 +249,15 @@ EXPORT_SYMBOL_GPL(nfs_initiate_read); /* * Set up the NFS read request struct */ -static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data, +static void nfs_read_rpcsetup(struct nfs_read_data *data, unsigned int count, unsigned int offset) { - struct inode *inode = req->wb_context->dentry->d_inode; - - data->req = req; - data->inode = inode; - data->cred = req->wb_context->cred; + struct nfs_page *req = data->header->req; - data->args.fh = NFS_FH(inode); + data->args.fh = NFS_FH(data->header->inode); data->args.offset = req_offset(req) + offset; data->args.pgbase = req->wb_pgbase + offset; - data->args.pages = data->pagevec; + 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; @@ -238,9 +271,9 @@ static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data, static int nfs_do_read(struct nfs_read_data *data, const struct rpc_call_ops *call_ops) { - struct inode *inode = data->args.context->dentry->d_inode; + struct inode *inode = data->header->inode; - return nfs_initiate_read(data, NFS_CLIENT(inode), call_ops); + return nfs_initiate_read(NFS_CLIENT(inode), data, call_ops, 0); } static int @@ -253,7 +286,7 @@ nfs_do_multiple_reads(struct list_head *head, while (!list_empty(head)) { int ret2; - data = list_entry(head->next, struct nfs_read_data, list); + data = list_first_entry(head, struct nfs_read_data, list); list_del_init(&data->list); ret2 = nfs_do_read(data, call_ops); @@ -275,6 +308,24 @@ nfs_async_read_error(struct list_head *head) } } +static const struct nfs_pgio_completion_ops nfs_async_read_completion_ops = { + .error_cleanup = nfs_async_read_error, + .completion = nfs_read_completion, +}; + +static void nfs_pagein_error(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) +{ + set_bit(NFS_IOHDR_REDO, &hdr->flags); + while (!list_empty(&hdr->rpc_list)) { + struct nfs_read_data *data = list_first_entry(&hdr->rpc_list, + struct nfs_read_data, list); + list_del(&data->list); + nfs_readdata_release(data); + } + desc->pg_completion_ops->error_cleanup(&desc->pg_list); +} + /* * Generate multiple requests to fill a single page. * @@ -288,93 +339,95 @@ nfs_async_read_error(struct list_head *head) * won't see the new data until our attribute cache is updated. This is more * or less conventional NFS client behavior. */ -static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc, struct list_head *res) +static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) { - struct nfs_page *req = nfs_list_entry(desc->pg_list.next); + struct nfs_page *req = hdr->req; struct page *page = req->wb_page; struct nfs_read_data *data; size_t rsize = desc->pg_bsize, nbytes; unsigned int offset; - int requests = 0; - int ret = 0; - - nfs_list_remove_request(req); offset = 0; nbytes = desc->pg_count; do { size_t len = min(nbytes,rsize); - data = nfs_readdata_alloc(1); - if (!data) - goto out_bad; - data->pagevec[0] = page; - nfs_read_rpcsetup(req, data, len, offset); - list_add(&data->list, res); - requests++; + data = nfs_readdata_alloc(hdr, 1); + if (!data) { + nfs_pagein_error(desc, hdr); + return -ENOMEM; + } + data->pages.pagevec[0] = page; + nfs_read_rpcsetup(data, len, offset); + list_add(&data->list, &hdr->rpc_list); nbytes -= len; offset += len; - } while(nbytes != 0); - atomic_set(&req->wb_complete, requests); - desc->pg_rpc_callops = &nfs_read_partial_ops; - return ret; -out_bad: - while (!list_empty(res)) { - data = list_entry(res->next, struct nfs_read_data, list); - list_del(&data->list); - nfs_readdata_release(data); - } - nfs_readpage_release(req); - return -ENOMEM; + } while (nbytes != 0); + + nfs_list_remove_request(req); + nfs_list_add_request(req, &hdr->pages); + desc->pg_rpc_callops = &nfs_read_common_ops; + return 0; } -static int nfs_pagein_one(struct nfs_pageio_descriptor *desc, struct list_head *res) +static int nfs_pagein_one(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) { struct nfs_page *req; struct page **pages; - struct nfs_read_data *data; + struct nfs_read_data *data; struct list_head *head = &desc->pg_list; - int ret = 0; - data = nfs_readdata_alloc(nfs_page_array_len(desc->pg_base, - desc->pg_count)); + data = nfs_readdata_alloc(hdr, nfs_page_array_len(desc->pg_base, + desc->pg_count)); if (!data) { - nfs_async_read_error(head); - ret = -ENOMEM; - goto out; + nfs_pagein_error(desc, hdr); + return -ENOMEM; } - pages = data->pagevec; + pages = data->pages.pagevec; while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); - nfs_list_add_request(req, &data->pages); + nfs_list_add_request(req, &hdr->pages); *pages++ = req->wb_page; } - req = nfs_list_entry(data->pages.next); - nfs_read_rpcsetup(req, data, desc->pg_count, 0); - list_add(&data->list, res); - desc->pg_rpc_callops = &nfs_read_full_ops; -out: - return ret; + nfs_read_rpcsetup(data, desc->pg_count, 0); + list_add(&data->list, &hdr->rpc_list); + desc->pg_rpc_callops = &nfs_read_common_ops; + return 0; } -int nfs_generic_pagein(struct nfs_pageio_descriptor *desc, struct list_head *head) +int nfs_generic_pagein(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) { if (desc->pg_bsize < PAGE_CACHE_SIZE) - return nfs_pagein_multi(desc, head); - return nfs_pagein_one(desc, head); + return nfs_pagein_multi(desc, hdr); + return nfs_pagein_one(desc, hdr); } static int nfs_generic_pg_readpages(struct nfs_pageio_descriptor *desc) { - LIST_HEAD(head); + struct nfs_read_header *rhdr; + struct nfs_pgio_header *hdr; int ret; - ret = nfs_generic_pagein(desc, &head); + rhdr = nfs_readhdr_alloc(); + if (!rhdr) { + desc->pg_completion_ops->error_cleanup(&desc->pg_list); + return -ENOMEM; + } + hdr = &rhdr->header; + nfs_pgheader_init(desc, hdr, nfs_readhdr_free); + atomic_inc(&hdr->refcnt); + ret = nfs_generic_pagein(desc, hdr); if (ret == 0) - ret = nfs_do_multiple_reads(&head, desc->pg_rpc_callops); + ret = nfs_do_multiple_reads(&hdr->rpc_list, + desc->pg_rpc_callops); + if (atomic_dec_and_test(&hdr->refcnt)) + hdr->completion_ops->completion(hdr); return ret; } @@ -389,20 +442,21 @@ static const struct nfs_pageio_ops nfs_pageio_read_ops = { */ int nfs_readpage_result(struct rpc_task *task, struct nfs_read_data *data) { + struct inode *inode = data->header->inode; int status; dprintk("NFS: %s: %5u, (status %d)\n", __func__, task->tk_pid, task->tk_status); - status = NFS_PROTO(data->inode)->read_done(task, data); + status = NFS_PROTO(inode)->read_done(task, data); if (status != 0) return status; - nfs_add_stats(data->inode, NFSIOS_SERVERREADBYTES, data->res.count); + nfs_add_stats(inode, NFSIOS_SERVERREADBYTES, data->res.count); if (task->tk_status == -ESTALE) { - set_bit(NFS_INO_STALE, &NFS_I(data->inode)->flags); - nfs_mark_for_revalidate(data->inode); + set_bit(NFS_INO_STALE, &NFS_I(inode)->flags); + nfs_mark_for_revalidate(inode); } return 0; } @@ -412,15 +466,13 @@ static void nfs_readpage_retry(struct rpc_task *task, struct nfs_read_data *data struct nfs_readargs *argp = &data->args; struct nfs_readres *resp = &data->res; - if (resp->eof || resp->count == argp->count) - return; - /* This is a short read! */ - nfs_inc_stats(data->inode, NFSIOS_SHORTREAD); + nfs_inc_stats(data->header->inode, NFSIOS_SHORTREAD); /* Has the server at least made some progress? */ - if (resp->count == 0) + if (resp->count == 0) { + nfs_set_pgio_error(data->header, -EIO, argp->offset); return; - + } /* Yes, so retry the read at the end of the data */ data->mds_offset += resp->count; argp->offset += resp->count; @@ -429,114 +481,46 @@ static void nfs_readpage_retry(struct rpc_task *task, struct nfs_read_data *data rpc_restart_call_prepare(task); } -/* - * Handle a read reply that fills part of a page. - */ -static void nfs_readpage_result_partial(struct rpc_task *task, void *calldata) +static void nfs_readpage_result_common(struct rpc_task *task, void *calldata) { struct nfs_read_data *data = calldata; - + struct nfs_pgio_header *hdr = data->header; + + /* Note the only returns of nfs_readpage_result are 0 and -EAGAIN */ if (nfs_readpage_result(task, data) != 0) return; if (task->tk_status < 0) - return; - - nfs_readpage_truncate_uninitialised_page(data); - nfs_readpage_retry(task, data); + nfs_set_pgio_error(hdr, task->tk_status, data->args.offset); + else if (data->res.eof) { + loff_t bound; + + bound = data->args.offset + data->res.count; + spin_lock(&hdr->lock); + if (bound < hdr->io_start + hdr->good_bytes) { + set_bit(NFS_IOHDR_EOF, &hdr->flags); + clear_bit(NFS_IOHDR_ERROR, &hdr->flags); + hdr->good_bytes = bound - hdr->io_start; + } + spin_unlock(&hdr->lock); + } else if (data->res.count != data->args.count) + nfs_readpage_retry(task, data); } -static void nfs_readpage_release_partial(void *calldata) +static void nfs_readpage_release_common(void *calldata) { - struct nfs_read_data *data = calldata; - struct nfs_page *req = data->req; - struct page *page = req->wb_page; - int status = data->task.tk_status; - - if (status < 0) - set_bit(PG_PARTIAL_READ_FAILED, &req->wb_flags); - - if (atomic_dec_and_test(&req->wb_complete)) { - if (!test_bit(PG_PARTIAL_READ_FAILED, &req->wb_flags)) - SetPageUptodate(page); - nfs_readpage_release(req); - } nfs_readdata_release(calldata); } void nfs_read_prepare(struct rpc_task *task, void *calldata) { struct nfs_read_data *data = calldata; - NFS_PROTO(data->inode)->read_rpc_prepare(task, data); -} - -static const struct rpc_call_ops nfs_read_partial_ops = { - .rpc_call_prepare = nfs_read_prepare, - .rpc_call_done = nfs_readpage_result_partial, - .rpc_release = nfs_readpage_release_partial, -}; - -static void nfs_readpage_set_pages_uptodate(struct nfs_read_data *data) -{ - unsigned int count = data->res.count; - unsigned int base = data->args.pgbase; - struct page **pages; - - if (data->res.eof) - count = data->args.count; - if (unlikely(count == 0)) - return; - pages = &data->args.pages[base >> PAGE_CACHE_SHIFT]; - base &= ~PAGE_CACHE_MASK; - count += base; - for (;count >= PAGE_CACHE_SIZE; count -= PAGE_CACHE_SIZE, pages++) - SetPageUptodate(*pages); - if (count == 0) - return; - /* Was this a short read? */ - if (data->res.eof || data->res.count == data->args.count) - SetPageUptodate(*pages); -} - -/* - * This is the callback from RPC telling us whether a reply was - * received or some error occurred (timeout or socket shutdown). - */ -static void nfs_readpage_result_full(struct rpc_task *task, void *calldata) -{ - struct nfs_read_data *data = calldata; - - if (nfs_readpage_result(task, data) != 0) - return; - if (task->tk_status < 0) - return; - /* - * Note: nfs_readpage_retry may change the values of - * data->args. In the multi-page case, we therefore need - * to ensure that we call nfs_readpage_set_pages_uptodate() - * first. - */ - nfs_readpage_truncate_uninitialised_page(data); - nfs_readpage_set_pages_uptodate(data); - nfs_readpage_retry(task, data); -} - -static void nfs_readpage_release_full(void *calldata) -{ - struct nfs_read_data *data = calldata; - - while (!list_empty(&data->pages)) { - struct nfs_page *req = nfs_list_entry(data->pages.next); - - nfs_list_remove_request(req); - nfs_readpage_release(req); - } - nfs_readdata_release(calldata); + NFS_PROTO(data->header->inode)->read_rpc_prepare(task, data); } -static const struct rpc_call_ops nfs_read_full_ops = { +static const struct rpc_call_ops nfs_read_common_ops = { .rpc_call_prepare = nfs_read_prepare, - .rpc_call_done = nfs_readpage_result_full, - .rpc_release = nfs_readpage_release_full, + .rpc_call_done = nfs_readpage_result_common, + .rpc_release = nfs_readpage_release_common, }; /* @@ -668,11 +652,12 @@ int nfs_readpages(struct file *filp, struct address_space *mapping, if (ret == 0) goto read_complete; /* all pages were read */ - nfs_pageio_init_read(&pgio, inode); + nfs_pageio_init_read(&pgio, inode, &nfs_async_read_completion_ops); ret = read_cache_pages(mapping, pages, readpage_async_filler, &desc); nfs_pageio_complete(&pgio); + NFS_I(inode)->read_io += pgio.pg_bytes_written; npages = (pgio.pg_bytes_written + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; nfs_add_stats(inode, NFSIOS_READPAGES, npages); read_complete: @@ -684,7 +669,7 @@ out: int __init nfs_init_readpagecache(void) { nfs_rdata_cachep = kmem_cache_create("nfs_read_data", - sizeof(struct nfs_read_data), + sizeof(struct nfs_read_header), 0, SLAB_HWCACHE_ALIGN, NULL); if (nfs_rdata_cachep == NULL) |