ceph: sync read inline data

we can't use getattr to fetch inline data while holding Fr cap,
because it can cause deadlock. If we need to sync read inline data,
drop cap refs first, then use getattr to fetch inline data.

Signed-off-by: Yan, Zheng <zyan@redhat.com>

fs/ceph/addr.c

@@ -192,17 +192,30 @@ static int readpage_nounlock(struct file *filp, struct page *page)
 	struct ceph_osd_client *osdc =
 		&ceph_inode_to_client(inode)->client->osdc;
 	int err = 0;
+	u64 off = page_offset(page);
 	u64 len = PAGE_CACHE_SIZE;
 
-	err = ceph_readpage_from_fscache(inode, page);
+	if (off >= i_size_read(inode)) {
+		zero_user_segment(page, err, PAGE_CACHE_SIZE);
+		SetPageUptodate(page);
+		return 0;
+	}
 
+	/*
+	 * Uptodate inline data should have been added into page cache
+	 * while getting Fcr caps.
+	 */
+	if (ci->i_inline_version != CEPH_INLINE_NONE)
+		return -EINVAL;
+
+	err = ceph_readpage_from_fscache(inode, page);
 	if (err == 0)
 		goto out;
 
 	dout("readpage inode %p file %p page %p index %lu\n",
 	     inode, filp, page, page->index);
 	err = ceph_osdc_readpages(osdc, ceph_vino(inode), &ci->i_layout,
-				  (u64) page_offset(page), &len,
+				  off, &len,
 				  ci->i_truncate_seq, ci->i_truncate_size,
 				  &page, 1, 0);
 	if (err == -ENOENT)
@@ -384,6 +397,9 @@ static int ceph_readpages(struct file *file, struct address_space *mapping,
 	int rc = 0;
 	int max = 0;
 
+	if (ceph_inode(inode)->i_inline_version != CEPH_INLINE_NONE)
+		return -EINVAL;
+
 	rc = ceph_readpages_from_fscache(mapping->host, mapping, page_list,
 					 &nr_pages);
 
@@ -1219,8 +1235,8 @@ static int ceph_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 		want = CEPH_CAP_FILE_CACHE;
 	while (1) {
 		got = 0;
-		ret = ceph_get_caps(ci, CEPH_CAP_FILE_RD, want, -1,
-				    &got, &pinned_page);
+		ret = ceph_get_caps(ci, CEPH_CAP_FILE_RD, want,
+				    -1, &got, &pinned_page);
 		if (ret == 0)
 			break;
 		if (ret != -ERESTARTSYS) {
@@ -1231,7 +1247,11 @@ static int ceph_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 	dout("filemap_fault %p %llu~%zd got cap refs on %s\n",
 	     inode, off, (size_t)PAGE_CACHE_SIZE, ceph_cap_string(got));
 
-	ret = filemap_fault(vma, vmf);
+	if ((got & (CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_LAZYIO)) ||
+	    ci->i_inline_version == CEPH_INLINE_NONE)
+		ret = filemap_fault(vma, vmf);
+	else
+		ret = -EAGAIN;
 
 	dout("filemap_fault %p %llu~%zd dropping cap refs on %s ret %d\n",
 	     inode, off, (size_t)PAGE_CACHE_SIZE, ceph_cap_string(got), ret);
@@ -1239,6 +1259,42 @@ static int ceph_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 		page_cache_release(pinned_page);
 	ceph_put_cap_refs(ci, got);
 
+	if (ret != -EAGAIN)
+		return ret;
+
+	/* read inline data */
+	if (off >= PAGE_CACHE_SIZE) {
+		/* does not support inline data > PAGE_SIZE */
+		ret = VM_FAULT_SIGBUS;
+	} else {
+		int ret1;
+		struct address_space *mapping = inode->i_mapping;
+		struct page *page = find_or_create_page(mapping, 0,
+						mapping_gfp_mask(mapping) &
+						~__GFP_FS);
+		if (!page) {
+			ret = VM_FAULT_OOM;
+			goto out;
+		}
+		ret1 = __ceph_do_getattr(inode, page,
+					 CEPH_STAT_CAP_INLINE_DATA, true);
+		if (ret1 < 0 || off >= i_size_read(inode)) {
+			unlock_page(page);
+			page_cache_release(page);
+			ret = VM_FAULT_SIGBUS;
+			goto out;
+		}
+		if (ret1 < PAGE_CACHE_SIZE)
+			zero_user_segment(page, ret1, PAGE_CACHE_SIZE);
+		else
+			flush_dcache_page(page);
+		SetPageUptodate(page);
+		vmf->page = page;
+		ret = VM_FAULT_MAJOR | VM_FAULT_LOCKED;
+	}
+out:
+	dout("filemap_fault %p %llu~%zd read inline data ret %d\n",
+	     inode, off, (size_t)PAGE_CACHE_SIZE, ret);
 	return ret;
 }
 

fs/ceph/file.c

@@ -333,6 +333,11 @@ int ceph_release(struct inode *inode, struct file *file)
 	return 0;
 }
 
+enum {
+	CHECK_EOF = 1,
+	READ_INLINE = 2,
+};
+
 /*
  * Read a range of bytes striped over one or more objects.  Iterate over
  * objects we stripe over.  (That's not atomic, but good enough for now.)
@@ -412,7 +417,7 @@ more:
 		ret = read;
 		/* did we bounce off eof? */
 		if (pos + left > inode->i_size)
-			*checkeof = 1;
+			*checkeof = CHECK_EOF;
 	}
 
 	dout("striped_read returns %d\n", ret);
@@ -808,7 +813,7 @@ static ssize_t ceph_read_iter(struct kiocb *iocb, struct iov_iter *to)
 	struct page *pinned_page = NULL;
 	ssize_t ret;
 	int want, got = 0;
-	int checkeof = 0, read = 0;
+	int retry_op = 0, read = 0;
 
 again:
 	dout("aio_read %p %llx.%llx %llu~%u trying to get caps on %p\n",
@@ -830,8 +835,12 @@ again:
 		     inode, ceph_vinop(inode), iocb->ki_pos, (unsigned)len,
 		     ceph_cap_string(got));
 
-		/* hmm, this isn't really async... */
-		ret = ceph_sync_read(iocb, to, &checkeof);
+		if (ci->i_inline_version == CEPH_INLINE_NONE) {
+			/* hmm, this isn't really async... */
+			ret = ceph_sync_read(iocb, to, &retry_op);
+		} else {
+			retry_op = READ_INLINE;
+		}
 	} else {
 		dout("aio_read %p %llx.%llx %llu~%u got cap refs on %s\n",
 		     inode, ceph_vinop(inode), iocb->ki_pos, (unsigned)len,
@@ -846,12 +855,50 @@ again:
 		pinned_page = NULL;
 	}
 	ceph_put_cap_refs(ci, got);
+	if (retry_op && ret >= 0) {
+		int statret;
+		struct page *page = NULL;
+		loff_t i_size;
+		if (retry_op == READ_INLINE) {
+			page = __page_cache_alloc(GFP_NOFS);
+			if (!page)
+				return -ENOMEM;
+		}
+
+		statret = __ceph_do_getattr(inode, page,
+					    CEPH_STAT_CAP_INLINE_DATA, !!page);
+		if (statret < 0) {
+			 __free_page(page);
+			if (statret == -ENODATA) {
+				BUG_ON(retry_op != READ_INLINE);
+				goto again;
+			}
+			return statret;
+		}
 
-	if (checkeof && ret >= 0) {
-		int statret = ceph_do_getattr(inode, CEPH_STAT_CAP_SIZE, false);
+		i_size = i_size_read(inode);
+		if (retry_op == READ_INLINE) {
+			/* does not support inline data > PAGE_SIZE */
+			if (i_size > PAGE_CACHE_SIZE) {
+				ret = -EIO;
+			} else if (iocb->ki_pos < i_size) {
+				loff_t end = min_t(loff_t, i_size,
+						   iocb->ki_pos + len);
+				if (statret < end)
+					zero_user_segment(page, statret, end);
+				ret = copy_page_to_iter(page,
+						iocb->ki_pos & ~PAGE_MASK,
+						end - iocb->ki_pos, to);
+				iocb->ki_pos += ret;
+			} else {
+				ret = 0;
+			}
+			__free_pages(page, 0);
+			return ret;
+		}
 
 		/* hit EOF or hole? */
-		if (statret == 0 && iocb->ki_pos < inode->i_size &&
+		if (retry_op == CHECK_EOF && iocb->ki_pos < i_size &&
 			ret < len) {
 			dout("sync_read hit hole, ppos %lld < size %lld"
 			     ", reading more\n", iocb->ki_pos,
@@ -859,7 +906,7 @@ again:
 
 			read += ret;
 			len -= ret;
-			checkeof = 0;
+			retry_op = 0;
 			goto again;
 		}
 	}