new helper: iov_iter_get_pages_alloc()

same as iov_iter_get_pages(), except that pages array is allocated
(kmalloc if possible, vmalloc if that fails) and left for caller to
free.  Lustre and NFS ->direct_IO() switched to it.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

4 files changed, 167 insertions, 257 deletions

diff --git a/drivers/staging/lustre/lustre/llite/rw26.c b/drivers/staging/lustre/lustre/llite/rw26.c
index f718585..6b59945 100644
--- a/drivers/staging/lustre/lustre/llite/rw26.c
+++ b/drivers/staging/lustre/lustre/llite/rw26.c
@@ -218,14 +218,11 @@ static void ll_free_user_pages(struct page **pages, int npages, int do_dirty)
 	int i;
 
 	for (i = 0; i < npages; i++) {
-		if (pages[i] == NULL)
-			break;
 		if (do_dirty)
 			set_page_dirty_lock(pages[i]);
 		page_cache_release(pages[i]);
 	}
-
-	OBD_FREE_LARGE(pages, npages * sizeof(*pages));
+	kvfree(pages);
 }
 
 ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
@@ -370,10 +367,9 @@ static ssize_t ll_direct_IO_26(int rw, struct kiocb *iocb,
 	struct file *file = iocb->ki_filp;
 	struct inode *inode = file->f_mapping->host;
 	struct ccc_object *obj = cl_inode2ccc(inode);
-	long count = iov_iter_count(iter);
-	long tot_bytes = 0, result = 0;
+	ssize_t count = iov_iter_count(iter);
+	ssize_t tot_bytes = 0, result = 0;
 	struct ll_inode_info *lli = ll_i2info(inode);
-	unsigned long seg = 0;
 	long size = MAX_DIO_SIZE;
 	int refcheck;
 
@@ -407,63 +403,49 @@ static ssize_t ll_direct_IO_26(int rw, struct kiocb *iocb,
 		mutex_lock(&inode->i_mutex);
 
 	LASSERT(obj->cob_transient_pages == 0);
-	for (seg = 0; seg < iter->nr_segs; seg++) {
-		long iov_left = iter->iov[seg].iov_len;
-		unsigned long user_addr = (unsigned long)iter->iov[seg].iov_base;
+	while (iov_iter_count(iter)) {
+		struct page **pages;
+		size_t offs;
 
+		count = min_t(size_t, iov_iter_count(iter), size);
 		if (rw == READ) {
 			if (file_offset >= i_size_read(inode))
 				break;
-			if (file_offset + iov_left > i_size_read(inode))
-				iov_left = i_size_read(inode) - file_offset;
+			if (file_offset + count > i_size_read(inode))
+				count = i_size_read(inode) - file_offset;
 		}
 
-		while (iov_left > 0) {
-			struct page **pages;
-			int page_count, max_pages = 0;
-			long bytes;
-
-			bytes = min(size, iov_left);
-			page_count = ll_get_user_pages(rw, user_addr, bytes,
-						       &pages, &max_pages);
-			if (likely(page_count > 0)) {
-				if (unlikely(page_count <  max_pages))
-					bytes = page_count << PAGE_CACHE_SHIFT;
-				result = ll_direct_IO_26_seg(env, io, rw, inode,
-							     file->f_mapping,
-							     bytes, file_offset,
-							     pages, page_count);
-				ll_free_user_pages(pages, max_pages, rw==READ);
-			} else if (page_count == 0) {
-				GOTO(out, result = -EFAULT);
-			} else {
-				result = page_count;
-			}
-			if (unlikely(result <= 0)) {
-				/* If we can't allocate a large enough buffer
-				 * for the request, shrink it to a smaller
-				 * PAGE_SIZE multiple and try again.
-				 * We should always be able to kmalloc for a
-				 * page worth of page pointers = 4MB on i386. */
-				if (result == -ENOMEM &&
-				    size > (PAGE_CACHE_SIZE / sizeof(*pages)) *
-					   PAGE_CACHE_SIZE) {
-					size = ((((size / 2) - 1) |
-						 ~CFS_PAGE_MASK) + 1) &
-						CFS_PAGE_MASK;
-					CDEBUG(D_VFSTRACE,"DIO size now %lu\n",
-					       size);
-					continue;
-				}
-
-				GOTO(out, result);
+		result = iov_iter_get_pages_alloc(iter, &pages, count, &offs);
+		if (likely(result > 0)) {
+			int n = (result + offs + PAGE_SIZE - 1) / PAGE_SIZE;
+			result = ll_direct_IO_26_seg(env, io, rw, inode,
+						     file->f_mapping,
+						     result, file_offset,
+						     pages, n);
+			ll_free_user_pages(pages, n, rw==READ);
+		}
+		if (unlikely(result <= 0)) {
+			/* If we can't allocate a large enough buffer
+			 * for the request, shrink it to a smaller
+			 * PAGE_SIZE multiple and try again.
+			 * We should always be able to kmalloc for a
+			 * page worth of page pointers = 4MB on i386. */
+			if (result == -ENOMEM &&
+			    size > (PAGE_CACHE_SIZE / sizeof(*pages)) *
+				   PAGE_CACHE_SIZE) {
+				size = ((((size / 2) - 1) |
+					 ~CFS_PAGE_MASK) + 1) &
+					CFS_PAGE_MASK;
+				CDEBUG(D_VFSTRACE,"DIO size now %lu\n",
+				       size);
+				continue;
 			}
 
-			tot_bytes += result;
-			file_offset += result;
-			iov_left -= result;
-			user_addr += result;
+			GOTO(out, result);
 		}
+		iov_iter_advance(iter, result);
+		tot_bytes += result;
+		file_offset += result;
 	}
 out:
 	LASSERT(obj->cob_transient_pages == 0);
diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c
index 1d34f45..b122fe2 100644
--- a/fs/nfs/direct.c
+++ b/fs/nfs/direct.c
@@ -322,60 +322,37 @@ static const struct nfs_pgio_completion_ops nfs_direct_read_completion_ops = {
  * 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(&current->mm->mmap_sem);
-			result = get_user_pages(current, current->mm, user_addr,
-					npages, 1, 0, pagevec, NULL);
-			up_read(&current->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_PROTO(dreq->inode)->read_pageio_init(&desc, dreq->inode,
+			     &nfs_direct_read_completion_ops);
+	get_dreq(dreq);
+	desc.pg_dreq = dreq;
+	atomic_inc(&inode->i_dio_count);
 
-		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);
@@ -389,55 +366,21 @@ static ssize_t nfs_direct_read_schedule_segment(struct nfs_pageio_descriptor *de
 			}
 			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,
-					      struct iov_iter *iter,
-					      loff_t pos, bool uio)
-{
-	struct nfs_pageio_descriptor desc;
-	struct inode *inode = dreq->inode;
-	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;
-	atomic_inc(&inode->i_dio_count);
-
-	for (seg = 0; seg < iter->nr_segs; seg++) {
-		const struct iovec *vec = &iter->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);
@@ -521,7 +464,7 @@ ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter,
 		dreq->iocb = iocb;
 
 	NFS_I(inode)->read_io += count;
-	result = nfs_direct_read_schedule_iovec(dreq, iter, pos, uio);
+	result = nfs_direct_read_schedule_iovec(dreq, iter, pos);
 
 	mutex_unlock(&inode->i_mutex);
 
@@ -677,109 +620,6 @@ static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode
 }
 #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(&current->mm->mmap_sem);
-			result = get_user_pages(current, current->mm, user_addr,
-						npages, 0, 0, pagevec, NULL);
-			up_read(&current->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;
@@ -859,15 +699,27 @@ static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = {
 	.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,
 					       struct iov_iter *iter,
-					       loff_t pos, bool uio)
+					       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_direct_write_completion_ops);
@@ -875,16 +727,50 @@ static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
 	get_dreq(dreq);
 	atomic_inc(&inode->i_dio_count);
 
-	NFS_I(dreq->inode)->write_io += iov_iter_count(iter);
-	for (seg = 0; seg < iter->nr_segs; seg++) {
-		const struct iovec *vec = &iter->iov[seg];
-		result = nfs_direct_write_schedule_segment(&desc, vec, pos, uio);
+	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;
+
+		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, 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;
+			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);
 
@@ -985,7 +871,7 @@ ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter,
 	if (!is_sync_kiocb(iocb))
 		dreq->iocb = iocb;
 
-	result = nfs_direct_write_schedule_iovec(dreq, iter, pos, uio);
+	result = nfs_direct_write_schedule_iovec(dreq, iter, pos);
 
 	if (mapping->nrpages) {
 		invalidate_inode_pages2_range(mapping,
diff --git a/include/linux/uio.h b/include/linux/uio.h
index 2f8825b..4876e9f 100644
--- a/include/linux/uio.h
+++ b/include/linux/uio.h
@@ -73,6 +73,8 @@ void iov_iter_init(struct iov_iter *i, int direction, const struct iovec *iov,
 			unsigned long nr_segs, size_t count);
 ssize_t iov_iter_get_pages(struct iov_iter *i, struct page **pages,
 			size_t maxsize, size_t *start);
+ssize_t iov_iter_get_pages_alloc(struct iov_iter *i, struct page ***pages,
+			size_t maxsize, size_t *start);
 int iov_iter_npages(const struct iov_iter *i, int maxpages);
 
 static inline size_t iov_iter_count(struct iov_iter *i)
diff --git a/mm/iov_iter.c b/mm/iov_iter.c
index 0b677f8..a5c691c 100644
--- a/mm/iov_iter.c
+++ b/mm/iov_iter.c
@@ -1,6 +1,8 @@
 #include <linux/export.h>
 #include <linux/uio.h>
 #include <linux/pagemap.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
 
 size_t copy_page_to_iter(struct page *page, size_t offset, size_t bytes,
 			 struct iov_iter *i)
@@ -263,6 +265,44 @@ ssize_t iov_iter_get_pages(struct iov_iter *i,
 }
 EXPORT_SYMBOL(iov_iter_get_pages);
 
+ssize_t iov_iter_get_pages_alloc(struct iov_iter *i,
+		   struct page ***pages, size_t maxsize,
+		   size_t *start)
+{
+	size_t offset = i->iov_offset;
+	const struct iovec *iov = i->iov;
+	size_t len;
+	unsigned long addr;
+	void *p;
+	int n;
+	int res;
+
+	len = iov->iov_len - offset;
+	if (len > i->count)
+		len = i->count;
+	if (len > maxsize)
+		len = maxsize;
+	addr = (unsigned long)iov->iov_base + offset;
+	len += *start = addr & (PAGE_SIZE - 1);
+	addr &= ~(PAGE_SIZE - 1);
+	n = (len + PAGE_SIZE - 1) / PAGE_SIZE;
+	
+	p = kmalloc(n * sizeof(struct page *), GFP_KERNEL);
+	if (!p)
+		p = vmalloc(n * sizeof(struct page *));
+	if (!p)
+		return -ENOMEM;
+
+	res = get_user_pages_fast(addr, n, (i->type & WRITE) != WRITE, p);
+	if (unlikely(res < 0)) {
+		kvfree(p);
+		return res;
+	}
+	*pages = p;
+	return (res == n ? len : res * PAGE_SIZE) - *start;
+}
+EXPORT_SYMBOL(iov_iter_get_pages_alloc);
+
 int iov_iter_npages(const struct iov_iter *i, int maxpages)
 {
 	size_t offset = i->iov_offset;