struct cl_page *page;
struct cl_object *obj;
- int refcheck;
+ __u16 refcheck;
LASSERT(PageLocked(vmpage));
LASSERT(!PageWriteback(vmpage));
#define MAX_DIRECTIO_SIZE 2*1024*1024*1024UL
-static inline int ll_get_user_pages(int rw, unsigned long user_addr,
- size_t size, struct page ***pages,
- int *max_pages)
-{
- int result = -ENOMEM;
-
- /* set an arbitrary limit to prevent arithmetic overflow */
- if (size > MAX_DIRECTIO_SIZE) {
- *pages = NULL;
- return -EFBIG;
- }
-
- *max_pages = (user_addr + size + PAGE_CACHE_SIZE - 1) >>
- PAGE_CACHE_SHIFT;
- *max_pages -= user_addr >> PAGE_CACHE_SHIFT;
-
- OBD_ALLOC_LARGE(*pages, *max_pages * sizeof(**pages));
- if (*pages) {
- down_read(¤t->mm->mmap_sem);
- result = get_user_pages(current, current->mm, user_addr,
- *max_pages, (rw == READ), 0, *pages,
- NULL);
- up_read(¤t->mm->mmap_sem);
- if (unlikely(result <= 0))
- OBD_FREE_LARGE(*pages, *max_pages * sizeof(**pages));
- }
-
- return result;
-}
-
-/* ll_free_user_pages - tear down page struct array
- * @pages: array of page struct pointers underlying target buffer */
-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));
-}
-
ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
int rw, struct inode *inode,
struct ll_dio_pages *pv)
}
EXPORT_SYMBOL(ll_direct_rw_pages);
-static ssize_t ll_direct_IO_26_seg(const struct lu_env *env, struct cl_io *io,
- int rw, struct inode *inode,
- struct address_space *mapping,
- size_t size, loff_t file_offset,
- struct page **pages, int page_count)
+static ssize_t
+ll_direct_IO_seg(const struct lu_env *env, struct cl_io *io, int rw,
+ struct inode *inode, size_t size, loff_t file_offset,
+ struct page **pages, int page_count)
{
- struct ll_dio_pages pvec = { .ldp_pages = pages,
- .ldp_nr = page_count,
- .ldp_size = size,
- .ldp_offsets = NULL,
- .ldp_start_offset = file_offset
- };
-
- return ll_direct_rw_pages(env, io, rw, inode, &pvec);
+ struct ll_dio_pages pvec = { .ldp_pages = pages,
+ .ldp_nr = page_count,
+ .ldp_size = size,
+ .ldp_offsets = NULL,
+ .ldp_start_offset = file_offset
+ };
+
+ return ll_direct_rw_pages(env, io, rw, inode, &pvec);
+}
+
+/* ll_free_user_pages - tear down page struct array
+ * @pages: array of page struct pointers underlying target buffer */
+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]);
+ }
+
+#if defined(HAVE_DIRECTIO_ITER) || defined(HAVE_IOV_ITER_RW)
+ kvfree(pages);
+#else
+ OBD_FREE_LARGE(pages, npages * sizeof(*pages));
+#endif
}
#ifdef KMALLOC_MAX_SIZE
* up to 22MB for 128kB kmalloc and up to 682MB for 4MB kmalloc. */
#define MAX_DIO_SIZE ((MAX_MALLOC / sizeof(struct brw_page) * PAGE_CACHE_SIZE) & \
~(DT_MAX_BRW_SIZE - 1))
-static ssize_t ll_direct_IO_26(int rw, struct kiocb *iocb,
- const struct iovec *iov, loff_t file_offset,
- unsigned long nr_segs)
+
+#ifndef HAVE_IOV_ITER_RW
+# define iov_iter_rw(iter) rw
+#endif
+
+#if defined(HAVE_DIRECTIO_ITER) || defined(HAVE_IOV_ITER_RW)
+static ssize_t
+ll_direct_IO(
+# ifndef HAVE_IOV_ITER_RW
+ int rw,
+# endif
+ struct kiocb *iocb, struct iov_iter *iter,
+ loff_t file_offset)
+{
+ struct lu_env *env;
+ struct cl_io *io;
+ struct file *file = iocb->ki_filp;
+ struct inode *inode = file->f_mapping->host;
+ ssize_t count = iov_iter_count(iter);
+ ssize_t tot_bytes = 0, result = 0;
+ size_t size = MAX_DIO_SIZE;
+ __u16 refcheck;
+
+ /* FIXME: io smaller than PAGE_SIZE is broken on ia64 ??? */
+ if ((file_offset & ~PAGE_MASK) || (count & ~PAGE_MASK))
+ return -EINVAL;
+
+ CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p), size=%zd (max %lu), "
+ "offset=%lld=%llx, pages %zd (max %lu)\n",
+ PFID(ll_inode2fid(inode)), inode, count, MAX_DIO_SIZE,
+ file_offset, file_offset, count >> PAGE_CACHE_SHIFT,
+ MAX_DIO_SIZE >> PAGE_CACHE_SHIFT);
+
+ /* Check that all user buffers are aligned as well */
+ if (iov_iter_alignment(iter) & ~PAGE_MASK)
+ return -EINVAL;
+
+ env = cl_env_get(&refcheck);
+ LASSERT(!IS_ERR(env));
+ io = vvp_env_io(env)->vui_cl.cis_io;
+ LASSERT(io != NULL);
+
+ /* 0. Need locking between buffered and direct access. and race with
+ * size changing by concurrent truncates and writes.
+ * 1. Need inode mutex to operate transient pages.
+ */
+ if (iov_iter_rw(iter) == READ)
+ mutex_lock(&inode->i_mutex);
+
+ while (iov_iter_count(iter)) {
+ struct page **pages;
+ size_t offs;
+
+ count = min_t(size_t, iov_iter_count(iter), size);
+ if (iov_iter_rw(iter) == READ) {
+ if (file_offset >= i_size_read(inode))
+ break;
+
+ if (file_offset + count > i_size_read(inode))
+ count = i_size_read(inode) - file_offset;
+ }
+
+ result = iov_iter_get_pages_alloc(iter, &pages, count, &offs);
+ if (likely(result > 0)) {
+ int n = DIV_ROUND_UP(result + offs, PAGE_SIZE);
+
+ result = ll_direct_IO_seg(env, io, iov_iter_rw(iter),
+ inode, result, file_offset,
+ pages, n);
+ ll_free_user_pages(pages, n,
+ iov_iter_rw(iter) == 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) |
+ ~PAGE_MASK) + 1) & PAGE_MASK;
+ CDEBUG(D_VFSTRACE, "DIO size now %zu\n",
+ size);
+ continue;
+ }
+
+ GOTO(out, result);
+ }
+
+ iov_iter_advance(iter, result);
+ tot_bytes += result;
+ file_offset += result;
+ }
+out:
+ if (iov_iter_rw(iter) == READ)
+ mutex_unlock(&inode->i_mutex);
+
+ if (tot_bytes > 0) {
+ struct vvp_io *vio = vvp_env_io(env);
+
+ /* no commit async for direct IO */
+ vio->u.write.vui_written += tot_bytes;
+ }
+
+ cl_env_put(env, &refcheck);
+ return tot_bytes ? : result;
+}
+#else /* !HAVE_DIRECTIO_ITER && !HAVE_IOV_ITER_RW */
+
+static inline int ll_get_user_pages(int rw, unsigned long user_addr,
+ size_t size, struct page ***pages,
+ int *max_pages)
{
- struct lu_env *env;
- struct cl_io *io;
- struct file *file = iocb->ki_filp;
- struct inode *inode = file->f_mapping->host;
- long count = iov_length(iov, nr_segs);
- long tot_bytes = 0, result = 0;
- struct ll_inode_info *lli = ll_i2info(inode);
- unsigned long seg = 0;
- long size = MAX_DIO_SIZE;
- int refcheck;
- ENTRY;
-
- if (!lli->lli_has_smd)
- RETURN(-EBADF);
+ int result = -ENOMEM;
+
+ /* set an arbitrary limit to prevent arithmetic overflow */
+ if (size > MAX_DIRECTIO_SIZE) {
+ *pages = NULL;
+ return -EFBIG;
+ }
+
+ *max_pages = (user_addr + size + PAGE_CACHE_SIZE - 1) >>
+ PAGE_CACHE_SHIFT;
+ *max_pages -= user_addr >> PAGE_CACHE_SHIFT;
+
+ OBD_ALLOC_LARGE(*pages, *max_pages * sizeof(**pages));
+ if (*pages) {
+ down_read(¤t->mm->mmap_sem);
+ result = get_user_pages(current, current->mm, user_addr,
+ *max_pages, (rw == READ), 0, *pages,
+ NULL);
+ up_read(¤t->mm->mmap_sem);
+ if (unlikely(result <= 0))
+ OBD_FREE_LARGE(*pages, *max_pages * sizeof(**pages));
+ }
+
+ return result;
+}
+
+static ssize_t
+ll_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
+ loff_t file_offset, unsigned long nr_segs)
+{
+ struct lu_env *env;
+ struct cl_io *io;
+ struct file *file = iocb->ki_filp;
+ struct inode *inode = file->f_mapping->host;
+ ssize_t count = iov_length(iov, nr_segs);
+ ssize_t tot_bytes = 0, result = 0;
+ unsigned long seg = 0;
+ size_t size = MAX_DIO_SIZE;
+ __u16 refcheck;
+ ENTRY;
/* FIXME: io smaller than PAGE_SIZE is broken on ia64 ??? */
- if ((file_offset & ~CFS_PAGE_MASK) || (count & ~CFS_PAGE_MASK))
+ if ((file_offset & ~PAGE_MASK) || (count & ~PAGE_MASK))
RETURN(-EINVAL);
- CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p), size=%lu (max %lu), "
- "offset=%lld=%llx, pages %lu (max %lu)\n",
+ CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p), size=%zd (max %lu), "
+ "offset=%lld=%llx, pages %zd (max %lu)\n",
PFID(ll_inode2fid(inode)), inode, count, MAX_DIO_SIZE,
file_offset, file_offset, count >> PAGE_CACHE_SHIFT,
MAX_DIO_SIZE >> PAGE_CACHE_SHIFT);
/* Check that all user buffers are aligned as well */
for (seg = 0; seg < nr_segs; seg++) {
- if (((unsigned long)iov[seg].iov_base & ~CFS_PAGE_MASK) ||
- (iov[seg].iov_len & ~CFS_PAGE_MASK))
+ if (((unsigned long)iov[seg].iov_base & ~PAGE_MASK) ||
+ (iov[seg].iov_len & ~PAGE_MASK))
RETURN(-EINVAL);
}
LASSERT(io != NULL);
for (seg = 0; seg < nr_segs; seg++) {
- long iov_left = iov[seg].iov_len;
+ size_t iov_left = iov[seg].iov_len;
unsigned long user_addr = (unsigned long)iov[seg].iov_base;
if (rw == READ) {
while (iov_left > 0) {
struct page **pages;
int page_count, max_pages = 0;
- long bytes;
+ size_t bytes;
bytes = min(size, iov_left);
page_count = ll_get_user_pages(rw, user_addr, bytes,
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);
+ result = ll_direct_IO_seg(env, io, rw, inode,
+ bytes, file_offset,
+ pages, page_count);
ll_free_user_pages(pages, max_pages, rw==READ);
} else if (page_count == 0) {
GOTO(out, result = -EFAULT);
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",
+ ~PAGE_CACHE_MASK) + 1) &
+ PAGE_CACHE_MASK;
+ CDEBUG(D_VFSTRACE, "DIO size now %zu\n",
size);
continue;
}
cl_env_put(env, &refcheck);
RETURN(tot_bytes ? tot_bytes : result);
}
+#endif /* HAVE_DIRECTIO_ITER || HAVE_IOV_ITER_RW */
/**
* Prepare partially written-to page for a write.
static int ll_prepare_partial_page(const struct lu_env *env, struct cl_io *io,
struct cl_page *pg)
{
- struct cl_attr *attr = ccc_env_thread_attr(env);
+ struct cl_attr *attr = vvp_env_thread_attr(env);
struct cl_object *obj = io->ci_obj;
struct vvp_page *vpg = cl_object_page_slice(obj, pg);
loff_t offset = cl_offset(obj, vvp_index(vpg));
CDEBUG(D_VFSTRACE, "Writing %lu of %d to %d bytes\n", index, from, len);
lcc = ll_cl_find(file);
- if (lcc == NULL)
+ if (lcc == NULL) {
+ io = NULL;
GOTO(out, result = -EIO);
+ }
env = lcc->lcc_env;
io = lcc->lcc_io;
unlock_page(vmpage);
page_cache_release(vmpage);
}
+ if (io)
+ io->ci_result = result;
} else {
*pagep = vmpage;
*fsdata = lcc;
unplug = true;
}
if (unplug ||
- file->f_flags & O_SYNC || IS_SYNC(file->f_dentry->d_inode))
+ file->f_flags & O_SYNC || IS_SYNC(file->f_path.dentry->d_inode))
result = vvp_io_write_commit(env, io);
+ if (result < 0)
+ io->ci_result = result;
RETURN(result >= 0 ? copied : result);
}
}
#endif
-#ifndef MS_HAS_NEW_AOPS
const struct address_space_operations ll_aops = {
.readpage = ll_readpage,
- .direct_IO = ll_direct_IO_26,
- .writepage = ll_writepage,
- .writepages = ll_writepages,
- .set_page_dirty = __set_page_dirty_nobuffers,
- .write_begin = ll_write_begin,
- .write_end = ll_write_end,
- .invalidatepage = ll_invalidatepage,
- .releasepage = (void *)ll_releasepage,
+ .direct_IO = ll_direct_IO,
+ .writepage = ll_writepage,
+ .writepages = ll_writepages,
+ .set_page_dirty = __set_page_dirty_nobuffers,
+ .write_begin = ll_write_begin,
+ .write_end = ll_write_end,
+ .invalidatepage = ll_invalidatepage,
+ .releasepage = (void *)ll_releasepage,
#ifdef CONFIG_MIGRATION
- .migratepage = ll_migratepage,
+ .migratepage = ll_migratepage,
#endif
};
-#else
-const struct address_space_operations_ext ll_aops = {
- .orig_aops.readpage = ll_readpage,
- .orig_aops.direct_IO = ll_direct_IO_26,
- .orig_aops.writepage = ll_writepage,
- .orig_aops.writepages = ll_writepages,
- .orig_aops.set_page_dirty = __set_page_dirty_nobuffers,
- .orig_aops.invalidatepage = ll_invalidatepage,
- .orig_aops.releasepage = ll_releasepage,
-#ifdef CONFIG_MIGRATION
- .orig_aops.migratepage = ll_migratepage,
-#endif
- .write_begin = ll_write_begin,
- .write_end = ll_write_end
-};
-#endif
git://git.whamcloud.com / fs / lustre-release.git / blobdiff