1 // SPDX-License-Identifier: GPL-2.0
6  * See COPYING in top-level directory.
16 #include "orangefs-kernel.h"
17 #include "orangefs-bufmap.h"
22 	struct inode *inode = page->mapping->host;  in orangefs_writepage_locked()
23 	struct orangefs_write_range *wr = NULL;  in orangefs_writepage_locked()  local
34 		wr = (struct orangefs_write_range *)page_private(page);  in orangefs_writepage_locked()
35 		WARN_ON(wr->pos >= len);  in orangefs_writepage_locked()
36 		off = wr->pos;  in orangefs_writepage_locked()
37 		if (off + wr->len > len)  in orangefs_writepage_locked()
38 			wlen = len - off;  in orangefs_writepage_locked()
40 			wlen = wr->len;  in orangefs_writepage_locked()
45 			wlen = len - off;  in orangefs_writepage_locked()
57 	    len, wr, NULL, NULL);  in orangefs_writepage_locked()
59 		mapping_set_error(page->mapping, ret);  in orangefs_writepage_locked()
90 	struct inode *inode = ow->pages[0]->mapping->host;  in orangefs_writepages_work()
91 	struct orangefs_write_range *wrp, wr;  in orangefs_writepages_work()  local
100 	for (i = 0; i < ow->npages; i++) {  in orangefs_writepages_work()
101 		set_page_writeback(ow->pages[i]);  in orangefs_writepages_work()
102 		bvec_set_page(&ow->bv[i], ow->pages[i],  in orangefs_writepages_work()
103 			      min(page_offset(ow->pages[i]) + PAGE_SIZE,  in orangefs_writepages_work()
104 			          ow->off + ow->len) -  in orangefs_writepages_work()
105 			      max(ow->off, page_offset(ow->pages[i])),  in orangefs_writepages_work()
106 			      i == 0 ? ow->off - page_offset(ow->pages[i]) : 0);  in orangefs_writepages_work()
108 	iov_iter_bvec(&iter, ITER_SOURCE, ow->bv, ow->npages, ow->len);  in orangefs_writepages_work()
110 	WARN_ON(ow->off >= len);  in orangefs_writepages_work()
111 	if (ow->off + ow->len > len)  in orangefs_writepages_work()
112 		ow->len = len - ow->off;  in orangefs_writepages_work()
114 	off = ow->off;  in orangefs_writepages_work()
115 	wr.uid = ow->uid;  in orangefs_writepages_work()
116 	wr.gid = ow->gid;  in orangefs_writepages_work()
117 	ret = wait_for_direct_io(ORANGEFS_IO_WRITE, inode, &off, &iter, ow->len,  in orangefs_writepages_work()
118 	    0, &wr, NULL, NULL);  in orangefs_writepages_work()
120 		for (i = 0; i < ow->npages; i++) {  in orangefs_writepages_work()
121 			mapping_set_error(ow->pages[i]->mapping, ret);  in orangefs_writepages_work()
122 			if (PagePrivate(ow->pages[i])) {  in orangefs_writepages_work()
124 				    page_private(ow->pages[i]);  in orangefs_writepages_work()
125 				ClearPagePrivate(ow->pages[i]);  in orangefs_writepages_work()
126 				put_page(ow->pages[i]);  in orangefs_writepages_work()
129 			end_page_writeback(ow->pages[i]);  in orangefs_writepages_work()
130 			unlock_page(ow->pages[i]);  in orangefs_writepages_work()
134 		for (i = 0; i < ow->npages; i++) {  in orangefs_writepages_work()
135 			if (PagePrivate(ow->pages[i])) {  in orangefs_writepages_work()
137 				    page_private(ow->pages[i]);  in orangefs_writepages_work()
138 				ClearPagePrivate(ow->pages[i]);  in orangefs_writepages_work()
139 				put_page(ow->pages[i]);  in orangefs_writepages_work()
142 			end_page_writeback(ow->pages[i]);  in orangefs_writepages_work()
143 			unlock_page(ow->pages[i]);  in orangefs_writepages_work()
153 	struct orangefs_write_range *wr = folio->private;  in orangefs_writepages_callback()  local
156 	if (!wr) {  in orangefs_writepages_callback()
164 	ret = -1;  in orangefs_writepages_callback()
165 	if (ow->npages == 0) {  in orangefs_writepages_callback()
166 		ow->off = wr->pos;  in orangefs_writepages_callback()
167 		ow->len = wr->len;  in orangefs_writepages_callback()
168 		ow->uid = wr->uid;  in orangefs_writepages_callback()
169 		ow->gid = wr->gid;  in orangefs_writepages_callback()
170 		ow->pages[ow->npages++] = &folio->page;  in orangefs_writepages_callback()
174 	if (!uid_eq(ow->uid, wr->uid) || !gid_eq(ow->gid, wr->gid)) {  in orangefs_writepages_callback()
176 		ow->npages = 0;  in orangefs_writepages_callback()
177 		ret = -1;  in orangefs_writepages_callback()
180 	if (ow->off + ow->len == wr->pos) {  in orangefs_writepages_callback()
181 		ow->len += wr->len;  in orangefs_writepages_callback()
182 		ow->pages[ow->npages++] = &folio->page;  in orangefs_writepages_callback()
187 	if (ret == -1) {  in orangefs_writepages_callback()
188 		if (ow->npages) {  in orangefs_writepages_callback()
190 			ow->npages = 0;  in orangefs_writepages_callback()
192 		ret = orangefs_writepage_locked(&folio->page, wbc);  in orangefs_writepages_callback()
193 		mapping_set_error(folio->mapping, ret);  in orangefs_writepages_callback()
197 		if (ow->npages == ow->maxpages) {  in orangefs_writepages_callback()
199 			ow->npages = 0;  in orangefs_writepages_callback()
213 		return -ENOMEM;  in orangefs_writepages()
214 	ow->maxpages = orangefs_bufmap_size_query()/PAGE_SIZE;  in orangefs_writepages()
215 	ow->pages = kcalloc(ow->maxpages, sizeof(struct page *), GFP_KERNEL);  in orangefs_writepages()
216 	if (!ow->pages) {  in orangefs_writepages()
218 		return -ENOMEM;  in orangefs_writepages()
220 	ow->bv = kcalloc(ow->maxpages, sizeof(struct bio_vec), GFP_KERNEL);  in orangefs_writepages()
221 	if (!ow->bv) {  in orangefs_writepages()
222 		kfree(ow->pages);  in orangefs_writepages()
224 		return -ENOMEM;  in orangefs_writepages()
228 	if (ow->npages)  in orangefs_writepages()
231 	kfree(ow->pages);  in orangefs_writepages()
232 	kfree(ow->bv);  in orangefs_writepages()
243 	struct inode *inode = rac->mapping->host;  in orangefs_readahead()
250 	loff_t bytes_remaining = inode->i_size - readahead_pos(rac);  in orangefs_readahead()
262 	i_pages = &rac->mapping->i_pages;  in orangefs_readahead()
269 			inode->i_size, NULL, NULL, rac->file)) < 0)  in orangefs_readahead()
285 	struct inode *inode = folio->mapping->host;  in orangefs_read_folio()
299 			folio_size(folio), inode->i_size, NULL, NULL, file);  in orangefs_read_folio()
314 	struct orangefs_write_range *wr;  in orangefs_write_begin()  local
336 		struct orangefs_write_range *wr;  in orangefs_write_begin()  local
337 		wr = folio_get_private(folio);  in orangefs_write_begin()
338 		if (wr->pos + wr->len == pos &&  in orangefs_write_begin()
339 		    uid_eq(wr->uid, current_fsuid()) &&  in orangefs_write_begin()
340 		    gid_eq(wr->gid, current_fsgid())) {  in orangefs_write_begin()
341 			wr->len += len;  in orangefs_write_begin()
350 	wr = kmalloc(sizeof *wr, GFP_KERNEL);  in orangefs_write_begin()
351 	if (!wr)  in orangefs_write_begin()
352 		return -ENOMEM;  in orangefs_write_begin()
354 	wr->pos = pos;  in orangefs_write_begin()
355 	wr->len = len;  in orangefs_write_begin()
356 	wr->uid = current_fsuid();  in orangefs_write_begin()
357 	wr->gid = current_fsgid();  in orangefs_write_begin()
358 	folio_attach_private(folio, wr);  in orangefs_write_begin()
367 	struct inode *inode = folio->mapping->host;  in orangefs_write_end()
372 	 * cannot change under us because we hold the i_mutex.  in orangefs_write_end()
374 	if (last_pos > inode->i_size)  in orangefs_write_end()
379 		unsigned from = pos & (PAGE_SIZE - 1);  in orangefs_write_end()
381 			folio_zero_range(folio, from + copied, len - copied);  in orangefs_write_end()
385 		    (len == PAGE_SIZE || pos + len == inode->i_size)) {  in orangefs_write_end()
402 	struct orangefs_write_range *wr = folio_get_private(folio);  in orangefs_invalidate_folio()  local
408 	} else if (folio_pos(folio) + offset <= wr->pos &&  in orangefs_invalidate_folio()
409 	    wr->pos + wr->len <= folio_pos(folio) + offset + length) {  in orangefs_invalidate_folio()
415 	} else if (wr->pos < folio_pos(folio) + offset &&  in orangefs_invalidate_folio()
416 	    wr->pos + wr->len <= folio_pos(folio) + offset + length &&  in orangefs_invalidate_folio()
417 	     folio_pos(folio) + offset < wr->pos + wr->len) {  in orangefs_invalidate_folio()
419 		x = wr->pos + wr->len - (folio_pos(folio) + offset);  in orangefs_invalidate_folio()
420 		WARN_ON(x > wr->len);  in orangefs_invalidate_folio()
421 		wr->len -= x;  in orangefs_invalidate_folio()
422 		wr->uid = current_fsuid();  in orangefs_invalidate_folio()
423 		wr->gid = current_fsgid();  in orangefs_invalidate_folio()
425 	} else if (folio_pos(folio) + offset <= wr->pos &&  in orangefs_invalidate_folio()
426 	    folio_pos(folio) + offset + length < wr->pos + wr->len &&  in orangefs_invalidate_folio()
427 	    wr->pos < folio_pos(folio) + offset + length) {  in orangefs_invalidate_folio()
429 		x = folio_pos(folio) + offset + length - wr->pos;  in orangefs_invalidate_folio()
430 		WARN_ON(x > wr->len);  in orangefs_invalidate_folio()
431 		wr->pos += x;  in orangefs_invalidate_folio()
432 		wr->len -= x;  in orangefs_invalidate_folio()
433 		wr->uid = current_fsuid();  in orangefs_invalidate_folio()
434 		wr->gid = current_fsgid();  in orangefs_invalidate_folio()
436 	} else if (wr->pos < folio_pos(folio) + offset &&  in orangefs_invalidate_folio()
437 	    folio_pos(folio) + offset + length < wr->pos + wr->len) {  in orangefs_invalidate_folio()
446 	/* non-overlapping ranges */  in orangefs_invalidate_folio()
449 		if (!((folio_pos(folio) + offset + length <= wr->pos) ^  in orangefs_invalidate_folio()
450 		    (wr->pos + wr->len <= folio_pos(folio) + offset))) {  in orangefs_invalidate_folio()
455 			    wr->pos, wr->len);  in orangefs_invalidate_folio()
461 	 * Above there are returns where wr is freed or where we WARN.  in orangefs_invalidate_folio()
462 	 * Thus the following runs if wr was modified above.  in orangefs_invalidate_folio()
487 		r = orangefs_writepage_locked(&folio->page, &wbc);  in orangefs_launder_folio()
504 	struct file *file = iocb->ki_filp;  in orangefs_direct_IO()
505 	loff_t pos = iocb->ki_pos;  in orangefs_direct_IO()
509 	struct inode *inode = file->f_mapping->host;  in orangefs_direct_IO()
511 	struct orangefs_khandle *handle = &orangefs_inode->refn.khandle;  in orangefs_direct_IO()
514 	ssize_t ret = -EINVAL;  in orangefs_direct_IO()
517 		"%s-BEGIN(%pU): count(%d) after estimate_max_iovecs.\n",  in orangefs_direct_IO()
625 	struct folio *folio = page_folio(vmf->page);  in orangefs_page_mkwrite()
626 	struct inode *inode = file_inode(vmf->vma->vm_file);  in orangefs_page_mkwrite()
628 	unsigned long *bitlock = &orangefs_inode->bitlock;  in orangefs_page_mkwrite()
630 	struct orangefs_write_range *wr;  in orangefs_page_mkwrite()  local
632 	sb_start_pagefault(inode->i_sb);  in orangefs_page_mkwrite()
652 		wr = folio_get_private(folio);  in orangefs_page_mkwrite()
653 		if (uid_eq(wr->uid, current_fsuid()) &&  in orangefs_page_mkwrite()
654 		    gid_eq(wr->gid, current_fsgid())) {  in orangefs_page_mkwrite()
655 			wr->pos = page_offset(vmf->page);  in orangefs_page_mkwrite()
656 			wr->len = PAGE_SIZE;  in orangefs_page_mkwrite()
665 	wr = kmalloc(sizeof *wr, GFP_KERNEL);  in orangefs_page_mkwrite()
666 	if (!wr) {  in orangefs_page_mkwrite()
670 	wr->pos = page_offset(vmf->page);  in orangefs_page_mkwrite()
671 	wr->len = PAGE_SIZE;  in orangefs_page_mkwrite()
672 	wr->uid = current_fsuid();  in orangefs_page_mkwrite()
673 	wr->gid = current_fsgid();  in orangefs_page_mkwrite()
674 	folio_attach_private(folio, wr);  in orangefs_page_mkwrite()
677 	file_update_time(vmf->vma->vm_file);  in orangefs_page_mkwrite()
678 	if (folio->mapping != inode->i_mapping) {  in orangefs_page_mkwrite()
693 	sb_end_pagefault(inode->i_sb);  in orangefs_page_mkwrite()
702 	int ret = -EINVAL;  in orangefs_setattr_size()
708 		     &orangefs_inode->refn.khandle,  in orangefs_setattr_size()
709 		     orangefs_inode->refn.fs_id,  in orangefs_setattr_size()
710 		     iattr->ia_size);  in orangefs_setattr_size()
714 	if (ret == -ESTALE)  in orangefs_setattr_size()
715 		ret = -EIO;  in orangefs_setattr_size()
724 	truncate_pagecache(inode, iattr->ia_size);  in orangefs_setattr_size()
725 	i_size_write(inode, iattr->ia_size);  in orangefs_setattr_size()
726 	if (iattr->ia_size > orig_size)  in orangefs_setattr_size()
727 		pagecache_isize_extended(inode, orig_size, iattr->ia_size);  in orangefs_setattr_size()
731 		return -ENOMEM;  in orangefs_setattr_size()
733 	new_op->upcall.req.truncate.refn = orangefs_inode->refn;  in orangefs_setattr_size()
734 	new_op->upcall.req.truncate.size = (__s64) iattr->ia_size;  in orangefs_setattr_size()
752 		iattr->ia_valid |= ATTR_CTIME | ATTR_MTIME;  in orangefs_setattr_size()
761 	if (iattr->ia_valid & ATTR_MODE) {  in __orangefs_setattr()
762 		if (iattr->ia_mode & (S_ISVTX)) {  in __orangefs_setattr()
769 				iattr->ia_mode -= S_ISVTX;  in __orangefs_setattr()
772 					     "User attempted to set sticky bit on non-root directory; returning EINVAL.\n");  in __orangefs_setattr()
773 				ret = -EINVAL;  in __orangefs_setattr()
777 		if (iattr->ia_mode & (S_ISUID)) {  in __orangefs_setattr()
780 			ret = -EINVAL;  in __orangefs_setattr()
785 	if (iattr->ia_valid & ATTR_SIZE) {  in __orangefs_setattr()
792 	spin_lock(&inode->i_lock);  in __orangefs_setattr()
793 	if (ORANGEFS_I(inode)->attr_valid) {  in __orangefs_setattr()
794 		if (uid_eq(ORANGEFS_I(inode)->attr_uid, current_fsuid()) &&  in __orangefs_setattr()
795 		    gid_eq(ORANGEFS_I(inode)->attr_gid, current_fsgid())) {  in __orangefs_setattr()
796 			ORANGEFS_I(inode)->attr_valid = iattr->ia_valid;  in __orangefs_setattr()
798 			spin_unlock(&inode->i_lock);  in __orangefs_setattr()
803 		ORANGEFS_I(inode)->attr_valid = iattr->ia_valid;  in __orangefs_setattr()
804 		ORANGEFS_I(inode)->attr_uid = current_fsuid();  in __orangefs_setattr()
805 		ORANGEFS_I(inode)->attr_gid = current_fsgid();  in __orangefs_setattr()
808 	spin_unlock(&inode->i_lock);  in __orangefs_setattr()
823 	if (!ret && (iattr->ia_valid & ATTR_MODE))  in __orangefs_setattr_mode()
824 		ret = posix_acl_chmod(&nop_mnt_idmap, dentry, inode->i_mode);  in __orangefs_setattr_mode()
855 	struct inode *inode = path->dentry->d_inode;  in orangefs_getattr()
859 		     path->dentry, request_mask);  in orangefs_getattr()
868 			stat->result_mask &= ~STATX_SIZE;  in orangefs_getattr()
881 		return -ECHILD;  in orangefs_permission()
921 	if (ret < 0 && ret != -ENODATA)  in orangefs_fileattr_get()
945 	    (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | FS_NOATIME_FL | ORANGEFS_MIRROR_FL))) {  in orangefs_fileattr_set()
948 		return -EOPNOTSUPP;  in orangefs_fileattr_set()
950 	val = fa->flags;  in orangefs_fileattr_set()
972 	inode->i_mapping->a_ops = &orangefs_address_operations;  in orangefs_init_iops()
974 	switch (inode->i_mode & S_IFMT) {  in orangefs_init_iops()
976 		inode->i_op = &orangefs_file_inode_operations;  in orangefs_init_iops()
977 		inode->i_fop = &orangefs_file_operations;  in orangefs_init_iops()
980 		inode->i_op = &orangefs_symlink_inode_operations;  in orangefs_init_iops()
983 		inode->i_op = &orangefs_dir_inode_operations;  in orangefs_init_iops()
984 		inode->i_fop = &orangefs_dir_operations;  in orangefs_init_iops()
990 		return -EINVAL;  in orangefs_init_iops()
998  * a ino_t type that will be used as a hash-index from where the handle will
1005 	return orangefs_khandle_to_ino(&(ref->khandle));  in orangefs_handle_hash()
1014 	ORANGEFS_I(inode)->refn.fs_id = ref->fs_id;  in orangefs_set_inode()
1015 	ORANGEFS_I(inode)->refn.khandle = ref->khandle;  in orangefs_set_inode()
1016 	ORANGEFS_I(inode)->attr_valid = 0;  in orangefs_set_inode()
1017 	hash_init(ORANGEFS_I(inode)->xattr_cache);  in orangefs_set_inode()
1018 	ORANGEFS_I(inode)->mapping_time = jiffies - 1;  in orangefs_set_inode()
1019 	ORANGEFS_I(inode)->bitlock = 0;  in orangefs_set_inode()
1033 	return (!ORANGEFS_khandle_cmp(&(orangefs_inode->refn.khandle),  in orangefs_test_inode()
1034 				&(ref->khandle)) &&  in orangefs_test_inode()
1035 			orangefs_inode->refn.fs_id == ref->fs_id);  in orangefs_test_inode()
1039  * Front-end to lookup the inode-cache maintained by the VFS using the ORANGEFS
1060 		return ERR_PTR(-ENOMEM);  in orangefs_iget()
1062 	if (!(inode->i_state & I_NEW))  in orangefs_iget()
1071 	inode->i_ino = hash;	/* needed for stat etc */  in orangefs_iget()
1077 		     &ref->khandle,  in orangefs_iget()
1078 		     ref->fs_id,  in orangefs_iget()
1080 		     inode->i_ino);  in orangefs_iget()
1106 		return ERR_PTR(-ENOMEM);  in orangefs_new_inode()
1113 	inode->i_ino = hash;	/* needed for stat etc */  in orangefs_new_inode()
1120 	inode->i_rdev = dev;  in orangefs_new_inode()
1142 	if (mode != inode->i_mode) {  in orangefs_new_inode()
1147 		inode->i_mode = mode;  in orangefs_new_inode()
1149 		__posix_acl_chmod(&acl, GFP_KERNEL, inode->i_mode);  in orangefs_new_inode()