--- zzzz-none-000/linux-3.10.107/fs/nfs/direct.c 2017-06-27 09:49:32.000000000 +0000 +++ scorpion-7490-727/linux-3.10.107/fs/nfs/direct.c 2021-02-04 17:41:59.000000000 +0000 @@ -66,6 +66,10 @@ /* * This represents a set of asynchronous requests that we're waiting on */ +struct nfs_direct_mirror { + ssize_t count; +}; + struct nfs_direct_req { struct kref kref; /* release manager */ @@ -78,8 +82,13 @@ /* completion state */ atomic_t io_count; /* i/os we're waiting for */ spinlock_t lock; /* protect completion state */ + + struct nfs_direct_mirror mirrors[NFS_PAGEIO_DESCRIPTOR_MIRROR_MAX]; + int mirror_count; + ssize_t count, /* bytes actually processed */ bytes_left, /* bytes left to be sent */ + io_start, /* start of IO */ error; /* any reported error */ struct completion completion; /* wait for i/o completion */ @@ -108,9 +117,135 @@ return atomic_dec_and_test(&dreq->io_count); } +void nfs_direct_set_resched_writes(struct nfs_direct_req *dreq) +{ + dreq->flags = NFS_ODIRECT_RESCHED_WRITES; +} +EXPORT_SYMBOL_GPL(nfs_direct_set_resched_writes); + +static void +nfs_direct_good_bytes(struct nfs_direct_req *dreq, struct nfs_pgio_header *hdr) +{ + int i; + ssize_t count; + + if (dreq->mirror_count == 1) { + dreq->mirrors[hdr->pgio_mirror_idx].count += hdr->good_bytes; + dreq->count += hdr->good_bytes; + } else { + /* mirrored writes */ + count = dreq->mirrors[hdr->pgio_mirror_idx].count; + if (count + dreq->io_start < hdr->io_start + hdr->good_bytes) { + count = hdr->io_start + hdr->good_bytes - dreq->io_start; + dreq->mirrors[hdr->pgio_mirror_idx].count = count; + } + /* update the dreq->count by finding the minimum agreed count from all + * mirrors */ + count = dreq->mirrors[0].count; + + for (i = 1; i < dreq->mirror_count; i++) + count = min(count, dreq->mirrors[i].count); + + dreq->count = count; + } +} + +/* + * nfs_direct_select_verf - select the right verifier + * @dreq - direct request possibly spanning multiple servers + * @ds_clp - nfs_client of data server or NULL if MDS / non-pnfs + * @commit_idx - commit bucket index for the DS + * + * returns the correct verifier to use given the role of the server + */ +static struct nfs_writeverf * +nfs_direct_select_verf(struct nfs_direct_req *dreq, + struct nfs_client *ds_clp, + int commit_idx) +{ + struct nfs_writeverf *verfp = &dreq->verf; + +#ifdef CONFIG_NFS_V4_1 + /* + * pNFS is in use, use the DS verf except commit_through_mds is set + * for layout segment where nbuckets is zero. + */ + if (ds_clp && dreq->ds_cinfo.nbuckets > 0) { + if (commit_idx >= 0 && commit_idx < dreq->ds_cinfo.nbuckets) + verfp = &dreq->ds_cinfo.buckets[commit_idx].direct_verf; + else + WARN_ON_ONCE(1); + } +#endif + return verfp; +} + + +/* + * nfs_direct_set_hdr_verf - set the write/commit verifier + * @dreq - direct request possibly spanning multiple servers + * @hdr - pageio header to validate against previously seen verfs + * + * Set the server's (MDS or DS) "seen" verifier + */ +static void nfs_direct_set_hdr_verf(struct nfs_direct_req *dreq, + struct nfs_pgio_header *hdr) +{ + struct nfs_writeverf *verfp; + + verfp = nfs_direct_select_verf(dreq, hdr->ds_clp, hdr->ds_commit_idx); + WARN_ON_ONCE(verfp->committed >= 0); + memcpy(verfp, &hdr->verf, sizeof(struct nfs_writeverf)); + WARN_ON_ONCE(verfp->committed < 0); +} + +/* + * nfs_direct_cmp_hdr_verf - compare verifier for pgio header + * @dreq - direct request possibly spanning multiple servers + * @hdr - pageio header to validate against previously seen verf + * + * set the server's "seen" verf if not initialized. + * returns result of comparison between @hdr->verf and the "seen" + * verf of the server used by @hdr (DS or MDS) + */ +static int nfs_direct_set_or_cmp_hdr_verf(struct nfs_direct_req *dreq, + struct nfs_pgio_header *hdr) +{ + struct nfs_writeverf *verfp; + + verfp = nfs_direct_select_verf(dreq, hdr->ds_clp, hdr->ds_commit_idx); + if (verfp->committed < 0) { + nfs_direct_set_hdr_verf(dreq, hdr); + return 0; + } + return memcmp(verfp, &hdr->verf, sizeof(struct nfs_writeverf)); +} + +/* + * nfs_direct_cmp_commit_data_verf - compare verifier for commit data + * @dreq - direct request possibly spanning multiple servers + * @data - commit data to validate against previously seen verf + * + * returns result of comparison between @data->verf and the verf of + * the server used by @data (DS or MDS) + */ +static int nfs_direct_cmp_commit_data_verf(struct nfs_direct_req *dreq, + struct nfs_commit_data *data) +{ + struct nfs_writeverf *verfp; + + verfp = nfs_direct_select_verf(dreq, data->ds_clp, + data->ds_commit_index); + + /* verifier not set so always fail */ + if (verfp->committed < 0) + return 1; + + return memcmp(verfp, &data->verf, sizeof(struct nfs_writeverf)); +} + /** * nfs_direct_IO - NFS address space operation for direct I/O - * @rw: direction (read or write) * @iocb: target I/O control block * @iov: array of vectors that define I/O buffer * @pos: offset in file to begin the operation @@ -121,7 +256,7 @@ * shunt off direct read and write requests before the VFS gets them, * so this method is only ever called for swap. */ -ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t pos, unsigned long nr_segs) +ssize_t nfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter, loff_t pos) { struct inode *inode = iocb->ki_filp->f_mapping->host; @@ -129,22 +264,11 @@ if (!IS_SWAPFILE(inode)) return 0; -#ifndef CONFIG_NFS_SWAP - dprintk("NFS: nfs_direct_IO (%s) off/no(%Ld/%lu) EINVAL\n", - iocb->ki_filp->f_path.dentry->d_name.name, - (long long) pos, nr_segs); - - return -EINVAL; -#else - VM_BUG_ON(iocb->ki_left != PAGE_SIZE); - VM_BUG_ON(iocb->ki_nbytes != PAGE_SIZE); - - if (rw == READ || rw == KERNEL_READ) - return nfs_file_direct_read(iocb, iov, nr_segs, pos, - rw == READ ? true : false); - return nfs_file_direct_write(iocb, iov, nr_segs, pos, - rw == WRITE ? true : false); -#endif /* CONFIG_NFS_SWAP */ + VM_BUG_ON(iov_iter_count(iter) != PAGE_SIZE); + + if (iov_iter_rw(iter) == READ) + return nfs_file_direct_read(iocb, iter, pos); + return nfs_file_direct_write(iocb, iter); } static void nfs_direct_release_pages(struct page **pages, unsigned int npages) @@ -157,13 +281,25 @@ void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo, struct nfs_direct_req *dreq) { - cinfo->lock = &dreq->lock; + cinfo->lock = &dreq->inode->i_lock; cinfo->mds = &dreq->mds_cinfo; cinfo->ds = &dreq->ds_cinfo; cinfo->dreq = dreq; cinfo->completion_ops = &nfs_direct_commit_completion_ops; } +static inline void nfs_direct_setup_mirroring(struct nfs_direct_req *dreq, + struct nfs_pageio_descriptor *pgio, + struct nfs_page *req) +{ + int mirror_count = 1; + + if (pgio->pg_ops->pg_get_mirror_count) + mirror_count = pgio->pg_ops->pg_get_mirror_count(pgio, req); + + dreq->mirror_count = mirror_count; +} + static inline struct nfs_direct_req *nfs_direct_req_alloc(void) { struct nfs_direct_req *dreq; @@ -176,7 +312,9 @@ kref_get(&dreq->kref); init_completion(&dreq->completion); INIT_LIST_HEAD(&dreq->mds_cinfo.list); + dreq->verf.committed = NFS_INVALID_STABLE_HOW; /* not set yet */ INIT_WORK(&dreq->work, nfs_direct_write_schedule_work); + dreq->mirror_count = 1; spin_lock_init(&dreq->lock); return dreq; @@ -231,14 +369,31 @@ * Synchronous I/O uses a stack-allocated iocb. Thus we can't trust * the iocb is still valid here if this is a synchronous request. */ -static void nfs_direct_complete(struct nfs_direct_req *dreq) +static void nfs_direct_complete(struct nfs_direct_req *dreq, bool write) { + struct inode *inode = dreq->inode; + + if (dreq->iocb && write) { + loff_t pos = dreq->iocb->ki_pos + dreq->count; + + spin_lock(&inode->i_lock); + if (i_size_read(inode) < pos) + i_size_write(inode, pos); + spin_unlock(&inode->i_lock); + } + + if (write) + nfs_zap_mapping(inode, inode->i_mapping); + + inode_dio_end(inode); + if (dreq->iocb) { long res = (long) dreq->error; if (!res) res = (long) dreq->count; - aio_complete(dreq->iocb, res, 0); + dreq->iocb->ki_complete(dreq->iocb, res, 0); } + complete_all(&dreq->completion); nfs_direct_req_release(dreq); @@ -246,9 +401,9 @@ static void nfs_direct_readpage_release(struct nfs_page *req) { - dprintk("NFS: direct read done (%s/%lld %d@%lld)\n", - req->wb_context->dentry->d_inode->i_sb->s_id, - (long long)NFS_FILEID(req->wb_context->dentry->d_inode), + dprintk("NFS: direct read done (%s/%llu %d@%lld)\n", + d_inode(req->wb_context->dentry)->i_sb->s_id, + (unsigned long long)NFS_FILEID(d_inode(req->wb_context->dentry)), req->wb_bytes, (long long)req_offset(req)); nfs_release_request(req); @@ -266,7 +421,8 @@ if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && (hdr->good_bytes == 0)) dreq->error = hdr->error; else - dreq->count += hdr->good_bytes; + nfs_direct_good_bytes(dreq, hdr); + spin_unlock(&dreq->lock); while (!list_empty(&hdr->pages)) { @@ -281,7 +437,7 @@ } out_put: if (put_dreq(dreq)) - nfs_direct_complete(dreq); + nfs_direct_complete(dreq, false); hdr->release(hdr); } @@ -314,66 +470,42 @@ * handled automatically by nfs_direct_read_result(). Otherwise, if * no requests have been sent, just return an error. */ -static ssize_t nfs_direct_read_schedule_segment(struct nfs_pageio_descriptor *desc, - const struct iovec *iov, - loff_t pos, bool uio) -{ - struct nfs_direct_req *dreq = desc->pg_dreq; - struct nfs_open_context *ctx = dreq->ctx; - struct inode *inode = ctx->dentry->d_inode; - unsigned long user_addr = (unsigned long)iov->iov_base; - size_t count = iov->iov_len; - size_t rsize = NFS_SERVER(inode)->rsize; - unsigned int pgbase; - int result; - ssize_t started = 0; - struct page **pagevec = NULL; - unsigned int npages; - - do { - size_t bytes; - int i; - pgbase = user_addr & ~PAGE_MASK; - bytes = min(max_t(size_t, rsize, PAGE_SIZE), count); +static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq, + struct iov_iter *iter, + loff_t pos) +{ + struct nfs_pageio_descriptor desc; + struct inode *inode = dreq->inode; + ssize_t result = -EINVAL; + size_t requested_bytes = 0; + size_t rsize = max_t(size_t, NFS_SERVER(inode)->rsize, PAGE_SIZE); - result = -ENOMEM; - npages = nfs_page_array_len(pgbase, bytes); - if (!pagevec) - pagevec = kmalloc(npages * sizeof(struct page *), - GFP_KERNEL); - if (!pagevec) - break; - if (uio) { - down_read(¤t->mm->mmap_sem); - result = get_user_pages(current, current->mm, user_addr, - npages, 1, 0, pagevec, NULL); - up_read(¤t->mm->mmap_sem); - if (result < 0) - break; - } else { - WARN_ON(npages != 1); - result = get_kernel_page(user_addr, 1, pagevec); - if (WARN_ON(result != 1)) - break; - } + nfs_pageio_init_read(&desc, dreq->inode, false, + &nfs_direct_read_completion_ops); + get_dreq(dreq); + desc.pg_dreq = dreq; + inode_dio_begin(inode); - if ((unsigned)result < npages) { - bytes = result * PAGE_SIZE; - if (bytes <= pgbase) { - nfs_direct_release_pages(pagevec, result); - break; - } - bytes -= pgbase; - npages = result; - } + while (iov_iter_count(iter)) { + struct page **pagevec; + size_t bytes; + size_t pgbase; + unsigned npages, i; + result = iov_iter_get_pages_alloc(iter, &pagevec, + rsize, &pgbase); + if (result < 0) + break; + + bytes = result; + iov_iter_advance(iter, bytes); + npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE; for (i = 0; i < npages; i++) { struct nfs_page *req; unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase); /* XXX do we need to do the eof zeroing found in async_filler? */ - req = nfs_create_request(dreq->ctx, dreq->inode, - pagevec[i], + req = nfs_create_request(dreq->ctx, pagevec[i], NULL, pgbase, req_len); if (IS_ERR(req)) { result = PTR_ERR(req); @@ -381,54 +513,21 @@ } req->wb_index = pos >> PAGE_SHIFT; req->wb_offset = pos & ~PAGE_MASK; - if (!nfs_pageio_add_request(desc, req)) { - result = desc->pg_error; + if (!nfs_pageio_add_request(&desc, req)) { + result = desc.pg_error; nfs_release_request(req); break; } pgbase = 0; bytes -= req_len; - started += req_len; - user_addr += req_len; + requested_bytes += req_len; pos += req_len; - count -= req_len; dreq->bytes_left -= req_len; } - /* The nfs_page now hold references to these pages */ nfs_direct_release_pages(pagevec, npages); - } while (count != 0 && result >= 0); - - kfree(pagevec); - - if (started) - return started; - return result < 0 ? (ssize_t) result : -EFAULT; -} - -static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq, - const struct iovec *iov, - unsigned long nr_segs, - loff_t pos, bool uio) -{ - struct nfs_pageio_descriptor desc; - ssize_t result = -EINVAL; - size_t requested_bytes = 0; - unsigned long seg; - - NFS_PROTO(dreq->inode)->read_pageio_init(&desc, dreq->inode, - &nfs_direct_read_completion_ops); - get_dreq(dreq); - desc.pg_dreq = dreq; - - for (seg = 0; seg < nr_segs; seg++) { - const struct iovec *vec = &iov[seg]; - result = nfs_direct_read_schedule_segment(&desc, vec, pos, uio); + kvfree(pagevec); if (result < 0) break; - requested_bytes += result; - if ((size_t)result < vec->iov_len) - break; - pos += vec->iov_len; } nfs_pageio_complete(&desc); @@ -438,29 +537,70 @@ * generic layer handle the completion. */ if (requested_bytes == 0) { + inode_dio_end(inode); nfs_direct_req_release(dreq); return result < 0 ? result : -EIO; } if (put_dreq(dreq)) - nfs_direct_complete(dreq); + nfs_direct_complete(dreq, false); return 0; } -static ssize_t nfs_direct_read(struct kiocb *iocb, const struct iovec *iov, - unsigned long nr_segs, loff_t pos, bool uio) +/** + * nfs_file_direct_read - file direct read operation for NFS files + * @iocb: target I/O control block + * @iter: vector of user buffers into which to read data + * @pos: byte offset in file where reading starts + * + * We use this function for direct reads instead of calling + * generic_file_aio_read() in order to avoid gfar's check to see if + * the request starts before the end of the file. For that check + * to work, we must generate a GETATTR before each direct read, and + * even then there is a window between the GETATTR and the subsequent + * READ where the file size could change. Our preference is simply + * to do all reads the application wants, and the server will take + * care of managing the end of file boundary. + * + * This function also eliminates unnecessarily updating the file's + * atime locally, as the NFS server sets the file's atime, and this + * client must read the updated atime from the server back into its + * cache. + */ +ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter, + loff_t pos) { - ssize_t result = -ENOMEM; - struct inode *inode = iocb->ki_filp->f_mapping->host; + struct file *file = iocb->ki_filp; + struct address_space *mapping = file->f_mapping; + struct inode *inode = mapping->host; struct nfs_direct_req *dreq; struct nfs_lock_context *l_ctx; + ssize_t result = -EINVAL; + size_t count = iov_iter_count(iter); + nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count); + + dfprintk(FILE, "NFS: direct read(%pD2, %zd@%Ld)\n", + file, count, (long long) pos); + result = 0; + if (!count) + goto out; + + mutex_lock(&inode->i_mutex); + result = nfs_sync_mapping(mapping); + if (result) + goto out_unlock; + + task_io_account_read(count); + + result = -ENOMEM; dreq = nfs_direct_req_alloc(); if (dreq == NULL) - goto out; + goto out_unlock; dreq->inode = inode; - dreq->bytes_left = iov_length(iov, nr_segs); + dreq->bytes_left = count; + dreq->io_start = pos; dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); l_ctx = nfs_get_lock_context(dreq->ctx); if (IS_ERR(l_ctx)) { @@ -471,23 +611,42 @@ if (!is_sync_kiocb(iocb)) dreq->iocb = iocb; - NFS_I(inode)->read_io += iov_length(iov, nr_segs); - result = nfs_direct_read_schedule_iovec(dreq, iov, nr_segs, pos, uio); - if (!result) + NFS_I(inode)->read_io += count; + result = nfs_direct_read_schedule_iovec(dreq, iter, pos); + + mutex_unlock(&inode->i_mutex); + + if (!result) { result = nfs_direct_wait(dreq); + if (result > 0) + iocb->ki_pos = pos + result; + } + + nfs_direct_req_release(dreq); + return result; + out_release: nfs_direct_req_release(dreq); +out_unlock: + mutex_unlock(&inode->i_mutex); out: return result; } -static void nfs_inode_dio_write_done(struct inode *inode) -{ - nfs_zap_mapping(inode, inode->i_mapping); - inode_dio_done(inode); +static void +nfs_direct_write_scan_commit_list(struct inode *inode, + struct list_head *list, + struct nfs_commit_info *cinfo) +{ + spin_lock(cinfo->lock); +#ifdef CONFIG_NFS_V4_1 + if (cinfo->ds != NULL && cinfo->ds->nwritten != 0) + NFS_SERVER(inode)->pnfs_curr_ld->recover_commit_reqs(list, cinfo); +#endif + nfs_scan_commit_list(&cinfo->mds->list, list, cinfo, 0); + spin_unlock(cinfo->lock); } -#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4) static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) { struct nfs_pageio_descriptor desc; @@ -495,20 +654,23 @@ LIST_HEAD(reqs); struct nfs_commit_info cinfo; LIST_HEAD(failed); + int i; nfs_init_cinfo_from_dreq(&cinfo, dreq); - pnfs_recover_commit_reqs(dreq->inode, &reqs, &cinfo); - spin_lock(cinfo.lock); - nfs_scan_commit_list(&cinfo.mds->list, &reqs, &cinfo, 0); - spin_unlock(cinfo.lock); + nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo); dreq->count = 0; + for (i = 0; i < dreq->mirror_count; i++) + dreq->mirrors[i].count = 0; get_dreq(dreq); - NFS_PROTO(dreq->inode)->write_pageio_init(&desc, dreq->inode, FLUSH_STABLE, + nfs_pageio_init_write(&desc, dreq->inode, FLUSH_STABLE, false, &nfs_direct_write_completion_ops); desc.pg_dreq = dreq; + req = nfs_list_entry(reqs.next); + nfs_direct_setup_mirroring(dreq, &desc, req); + list_for_each_entry_safe(req, tmp, &reqs, wb_list) { if (!nfs_pageio_add_request(&desc, req)) { nfs_list_remove_request(req); @@ -544,7 +706,7 @@ dprintk("NFS: %5u commit failed with error %d.\n", data->task.tk_pid, status); dreq->flags = NFS_ODIRECT_RESCHED_WRITES; - } else if (memcmp(&dreq->verf, &data->verf, sizeof(data->verf))) { + } else if (nfs_direct_cmp_commit_data_verf(dreq, data)) { dprintk("NFS: %5u commit verify failed\n", data->task.tk_pid); dreq->flags = NFS_ODIRECT_RESCHED_WRITES; } @@ -555,7 +717,7 @@ nfs_list_remove_request(req); if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) { /* Note the rewrite will go through mds */ - nfs_mark_request_commit(req, NULL, &cinfo); + nfs_mark_request_commit(req, NULL, &cinfo, 0); } else nfs_release_request(req); nfs_unlock_and_release_request(req); @@ -602,8 +764,7 @@ nfs_direct_write_reschedule(dreq); break; default: - nfs_inode_dio_write_done(dreq->inode); - nfs_direct_complete(dreq); + nfs_direct_complete(dreq, true); } } @@ -612,126 +773,11 @@ schedule_work(&dreq->work); /* Calls nfs_direct_write_schedule_work */ } -#else -static void nfs_direct_write_schedule_work(struct work_struct *work) -{ -} - -static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode) -{ - nfs_inode_dio_write_done(inode); - nfs_direct_complete(dreq); -} -#endif - -/* - * NB: Return the value of the first error return code. Subsequent - * errors after the first one are ignored. - */ -/* - * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE - * operation. If nfs_writedata_alloc() or get_user_pages() fails, - * bail and stop sending more writes. Write length accounting is - * handled automatically by nfs_direct_write_result(). Otherwise, if - * no requests have been sent, just return an error. - */ -static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *desc, - const struct iovec *iov, - loff_t pos, bool uio) -{ - struct nfs_direct_req *dreq = desc->pg_dreq; - struct nfs_open_context *ctx = dreq->ctx; - struct inode *inode = ctx->dentry->d_inode; - unsigned long user_addr = (unsigned long)iov->iov_base; - size_t count = iov->iov_len; - size_t wsize = NFS_SERVER(inode)->wsize; - unsigned int pgbase; - int result; - ssize_t started = 0; - struct page **pagevec = NULL; - unsigned int npages; - - do { - size_t bytes; - int i; - - pgbase = user_addr & ~PAGE_MASK; - bytes = min(max_t(size_t, wsize, PAGE_SIZE), count); - - result = -ENOMEM; - npages = nfs_page_array_len(pgbase, bytes); - if (!pagevec) - pagevec = kmalloc(npages * sizeof(struct page *), GFP_KERNEL); - if (!pagevec) - break; - - if (uio) { - down_read(¤t->mm->mmap_sem); - result = get_user_pages(current, current->mm, user_addr, - npages, 0, 0, pagevec, NULL); - up_read(¤t->mm->mmap_sem); - if (result < 0) - break; - } else { - WARN_ON(npages != 1); - result = get_kernel_page(user_addr, 0, pagevec); - if (WARN_ON(result != 1)) - break; - } - - if ((unsigned)result < npages) { - bytes = result * PAGE_SIZE; - if (bytes <= pgbase) { - nfs_direct_release_pages(pagevec, result); - break; - } - bytes -= pgbase; - npages = result; - } - - for (i = 0; i < npages; i++) { - struct nfs_page *req; - unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase); - - req = nfs_create_request(dreq->ctx, dreq->inode, - pagevec[i], - pgbase, req_len); - if (IS_ERR(req)) { - result = PTR_ERR(req); - break; - } - nfs_lock_request(req); - req->wb_index = pos >> PAGE_SHIFT; - req->wb_offset = pos & ~PAGE_MASK; - if (!nfs_pageio_add_request(desc, req)) { - result = desc->pg_error; - nfs_unlock_and_release_request(req); - break; - } - pgbase = 0; - bytes -= req_len; - started += req_len; - user_addr += req_len; - pos += req_len; - count -= req_len; - dreq->bytes_left -= req_len; - } - /* The nfs_page now hold references to these pages */ - nfs_direct_release_pages(pagevec, npages); - } while (count != 0 && result >= 0); - - kfree(pagevec); - - if (started) - return started; - return result < 0 ? (ssize_t) result : -EFAULT; -} - static void nfs_direct_write_completion(struct nfs_pgio_header *hdr) { struct nfs_direct_req *dreq = hdr->dreq; struct nfs_commit_info cinfo; - int bit = -1; + bool request_commit = false; struct nfs_page *req = nfs_list_entry(hdr->pages.next); if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) @@ -745,40 +791,33 @@ dreq->flags = 0; dreq->error = hdr->error; } - if (dreq->error != 0) - bit = NFS_IOHDR_ERROR; - else { - dreq->count += hdr->good_bytes; - if (test_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags)) { - dreq->flags = NFS_ODIRECT_RESCHED_WRITES; - bit = NFS_IOHDR_NEED_RESCHED; - } else if (test_bit(NFS_IOHDR_NEED_COMMIT, &hdr->flags)) { + if (dreq->error == 0) { + nfs_direct_good_bytes(dreq, hdr); + if (nfs_write_need_commit(hdr)) { if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) - bit = NFS_IOHDR_NEED_RESCHED; + request_commit = true; else if (dreq->flags == 0) { - memcpy(&dreq->verf, hdr->verf, - sizeof(dreq->verf)); - bit = NFS_IOHDR_NEED_COMMIT; + nfs_direct_set_hdr_verf(dreq, hdr); + request_commit = true; dreq->flags = NFS_ODIRECT_DO_COMMIT; } else if (dreq->flags == NFS_ODIRECT_DO_COMMIT) { - if (memcmp(&dreq->verf, hdr->verf, sizeof(dreq->verf))) { - dreq->flags = NFS_ODIRECT_RESCHED_WRITES; - bit = NFS_IOHDR_NEED_RESCHED; - } else - bit = NFS_IOHDR_NEED_COMMIT; + request_commit = true; + if (nfs_direct_set_or_cmp_hdr_verf(dreq, hdr)) + dreq->flags = + NFS_ODIRECT_RESCHED_WRITES; } } } spin_unlock(&dreq->lock); while (!list_empty(&hdr->pages)) { + req = nfs_list_entry(hdr->pages.next); nfs_list_remove_request(req); - switch (bit) { - case NFS_IOHDR_NEED_RESCHED: - case NFS_IOHDR_NEED_COMMIT: + if (request_commit) { kref_get(&req->wb_kref); - nfs_mark_request_commit(req, hdr->lseg, &cinfo); + nfs_mark_request_commit(req, hdr->lseg, &cinfo, + hdr->ds_commit_idx); } nfs_unlock_and_release_request(req); } @@ -806,33 +845,80 @@ .completion = nfs_direct_write_completion, }; + +/* + * NB: Return the value of the first error return code. Subsequent + * errors after the first one are ignored. + */ +/* + * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE + * operation. If nfs_writedata_alloc() or get_user_pages() fails, + * bail and stop sending more writes. Write length accounting is + * handled automatically by nfs_direct_write_result(). Otherwise, if + * no requests have been sent, just return an error. + */ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, - const struct iovec *iov, - unsigned long nr_segs, - loff_t pos, bool uio) + struct iov_iter *iter, + loff_t pos) { struct nfs_pageio_descriptor desc; struct inode *inode = dreq->inode; ssize_t result = 0; size_t requested_bytes = 0; - unsigned long seg; + size_t wsize = max_t(size_t, NFS_SERVER(inode)->wsize, PAGE_SIZE); - NFS_PROTO(inode)->write_pageio_init(&desc, inode, FLUSH_COND_STABLE, + nfs_pageio_init_write(&desc, inode, FLUSH_COND_STABLE, false, &nfs_direct_write_completion_ops); desc.pg_dreq = dreq; get_dreq(dreq); - atomic_inc(&inode->i_dio_count); + inode_dio_begin(inode); + + NFS_I(inode)->write_io += iov_iter_count(iter); + while (iov_iter_count(iter)) { + struct page **pagevec; + size_t bytes; + size_t pgbase; + unsigned npages, i; - NFS_I(dreq->inode)->write_io += iov_length(iov, nr_segs); - for (seg = 0; seg < nr_segs; seg++) { - const struct iovec *vec = &iov[seg]; - result = nfs_direct_write_schedule_segment(&desc, vec, pos, uio); + result = iov_iter_get_pages_alloc(iter, &pagevec, + wsize, &pgbase); if (result < 0) break; - requested_bytes += result; - if ((size_t)result < vec->iov_len) + + bytes = result; + iov_iter_advance(iter, bytes); + npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE; + for (i = 0; i < npages; i++) { + struct nfs_page *req; + unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase); + + req = nfs_create_request(dreq->ctx, pagevec[i], NULL, + pgbase, req_len); + if (IS_ERR(req)) { + result = PTR_ERR(req); + break; + } + + nfs_direct_setup_mirroring(dreq, &desc, req); + + nfs_lock_request(req); + req->wb_index = pos >> PAGE_SHIFT; + req->wb_offset = pos & ~PAGE_MASK; + if (!nfs_pageio_add_request(&desc, req)) { + result = desc.pg_error; + nfs_unlock_and_release_request(req); + break; + } + pgbase = 0; + bytes -= req_len; + requested_bytes += req_len; + pos += req_len; + dreq->bytes_left -= req_len; + } + nfs_direct_release_pages(pagevec, npages); + kvfree(pagevec); + if (result < 0) break; - pos += vec->iov_len; } nfs_pageio_complete(&desc); @@ -841,7 +927,7 @@ * generic layer handle the completion. */ if (requested_bytes == 0) { - inode_dio_done(inode); + inode_dio_end(inode); nfs_direct_req_release(dreq); return result < 0 ? result : -EIO; } @@ -851,100 +937,10 @@ return 0; } -static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov, - unsigned long nr_segs, loff_t pos, - size_t count, bool uio) -{ - ssize_t result = -ENOMEM; - struct inode *inode = iocb->ki_filp->f_mapping->host; - struct nfs_direct_req *dreq; - struct nfs_lock_context *l_ctx; - - dreq = nfs_direct_req_alloc(); - if (!dreq) - goto out; - - dreq->inode = inode; - dreq->bytes_left = count; - dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); - l_ctx = nfs_get_lock_context(dreq->ctx); - if (IS_ERR(l_ctx)) { - result = PTR_ERR(l_ctx); - goto out_release; - } - dreq->l_ctx = l_ctx; - if (!is_sync_kiocb(iocb)) - dreq->iocb = iocb; - - result = nfs_direct_write_schedule_iovec(dreq, iov, nr_segs, pos, uio); - if (!result) - result = nfs_direct_wait(dreq); -out_release: - nfs_direct_req_release(dreq); -out: - return result; -} - -/** - * nfs_file_direct_read - file direct read operation for NFS files - * @iocb: target I/O control block - * @iov: vector of user buffers into which to read data - * @nr_segs: size of iov vector - * @pos: byte offset in file where reading starts - * - * We use this function for direct reads instead of calling - * generic_file_aio_read() in order to avoid gfar's check to see if - * the request starts before the end of the file. For that check - * to work, we must generate a GETATTR before each direct read, and - * even then there is a window between the GETATTR and the subsequent - * READ where the file size could change. Our preference is simply - * to do all reads the application wants, and the server will take - * care of managing the end of file boundary. - * - * This function also eliminates unnecessarily updating the file's - * atime locally, as the NFS server sets the file's atime, and this - * client must read the updated atime from the server back into its - * cache. - */ -ssize_t nfs_file_direct_read(struct kiocb *iocb, const struct iovec *iov, - unsigned long nr_segs, loff_t pos, bool uio) -{ - ssize_t retval = -EINVAL; - struct file *file = iocb->ki_filp; - struct address_space *mapping = file->f_mapping; - size_t count; - - count = iov_length(iov, nr_segs); - nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count); - - dfprintk(FILE, "NFS: direct read(%s/%s, %zd@%Ld)\n", - file->f_path.dentry->d_parent->d_name.name, - file->f_path.dentry->d_name.name, - count, (long long) pos); - - retval = 0; - if (!count) - goto out; - - retval = nfs_sync_mapping(mapping); - if (retval) - goto out; - - task_io_account_read(count); - - retval = nfs_direct_read(iocb, iov, nr_segs, pos, uio); - if (retval > 0) - iocb->ki_pos = pos + retval; - -out: - return retval; -} - /** * nfs_file_direct_write - file direct write operation for NFS files * @iocb: target I/O control block - * @iov: vector of user buffers from which to write data - * @nr_segs: size of iov vector + * @iter: vector of user buffers from which to write data * @pos: byte offset in file where writing starts * * We use this function for direct writes instead of calling @@ -962,51 +958,88 @@ * Note that O_APPEND is not supported for NFS direct writes, as there * is no atomic O_APPEND write facility in the NFS protocol. */ -ssize_t nfs_file_direct_write(struct kiocb *iocb, const struct iovec *iov, - unsigned long nr_segs, loff_t pos, bool uio) +ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter) { - ssize_t retval = -EINVAL; + ssize_t result = -EINVAL; struct file *file = iocb->ki_filp; struct address_space *mapping = file->f_mapping; - size_t count; + struct inode *inode = mapping->host; + struct nfs_direct_req *dreq; + struct nfs_lock_context *l_ctx; + loff_t pos, end; - count = iov_length(iov, nr_segs); - nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count); + dfprintk(FILE, "NFS: direct write(%pD2, %zd@%Ld)\n", + file, iov_iter_count(iter), (long long) iocb->ki_pos); - dfprintk(FILE, "NFS: direct write(%s/%s, %zd@%Ld)\n", - file->f_path.dentry->d_parent->d_name.name, - file->f_path.dentry->d_name.name, - count, (long long) pos); + nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, + iov_iter_count(iter)); - retval = generic_write_checks(file, &pos, &count, 0); - if (retval) - goto out; + pos = iocb->ki_pos; + end = (pos + iov_iter_count(iter) - 1) >> PAGE_CACHE_SHIFT; - retval = -EINVAL; - if ((ssize_t) count < 0) - goto out; - retval = 0; - if (!count) - goto out; + mutex_lock(&inode->i_mutex); - retval = nfs_sync_mapping(mapping); - if (retval) - goto out; + result = nfs_sync_mapping(mapping); + if (result) + goto out_unlock; - task_io_account_write(count); + if (mapping->nrpages) { + result = invalidate_inode_pages2_range(mapping, + pos >> PAGE_CACHE_SHIFT, end); + if (result) + goto out_unlock; + } - retval = nfs_direct_write(iocb, iov, nr_segs, pos, count, uio); - if (retval > 0) { - struct inode *inode = mapping->host; + task_io_account_write(iov_iter_count(iter)); - iocb->ki_pos = pos + retval; - spin_lock(&inode->i_lock); - if (i_size_read(inode) < iocb->ki_pos) - i_size_write(inode, iocb->ki_pos); - spin_unlock(&inode->i_lock); + result = -ENOMEM; + dreq = nfs_direct_req_alloc(); + if (!dreq) + goto out_unlock; + + dreq->inode = inode; + dreq->bytes_left = iov_iter_count(iter); + dreq->io_start = pos; + dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); + l_ctx = nfs_get_lock_context(dreq->ctx); + if (IS_ERR(l_ctx)) { + result = PTR_ERR(l_ctx); + goto out_release; } -out: - return retval; + dreq->l_ctx = l_ctx; + if (!is_sync_kiocb(iocb)) + dreq->iocb = iocb; + + result = nfs_direct_write_schedule_iovec(dreq, iter, pos); + + if (mapping->nrpages) { + invalidate_inode_pages2_range(mapping, + pos >> PAGE_CACHE_SHIFT, end); + } + + mutex_unlock(&inode->i_mutex); + + if (!result) { + result = nfs_direct_wait(dreq); + if (result > 0) { + struct inode *inode = mapping->host; + + iocb->ki_pos = pos + result; + spin_lock(&inode->i_lock); + if (i_size_read(inode) < iocb->ki_pos) + i_size_write(inode, iocb->ki_pos); + spin_unlock(&inode->i_lock); + generic_write_sync(file, pos, result); + } + } + nfs_direct_req_release(dreq); + return result; + +out_release: + nfs_direct_req_release(dreq); +out_unlock: + mutex_unlock(&inode->i_mutex); + return result; } /**