diff options
author | Mark Fasheh <mark.fasheh@oracle.com> | 2007-09-07 17:46:51 -0400 |
---|---|---|
committer | Mark Fasheh <mark.fasheh@oracle.com> | 2007-10-12 14:54:40 -0400 |
commit | 1afc32b952335f665327a1a9001ba1b44bb76fd9 (patch) | |
tree | c914afd0ef5d32b426c3cf65820de7599e570656 /fs/ocfs2/aops.c | |
parent | 6798d35a31c413bbb3f83bbaa844bd2598168ccc (diff) |
ocfs2: Write support for inline data
This fixes up write, truncate, mmap, and RESVSP/UNRESVP to understand inline
inode data.
For the most part, the changes to the core write code can be relied on to do
the heavy lifting. Any code calling ocfs2_write_begin (including shared
writeable mmap) can count on it doing the right thing with respect to
growing inline data to an extent tree.
Size reducing truncates, including UNRESVP can simply zero that portion of
the inode block being removed. Size increasing truncatesm, including RESVP
have to be a little bit smarter and grow the inode to an extent tree if
necessary.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Reviewed-by: Joel Becker <joel.becker@oracle.com>
Diffstat (limited to 'fs/ocfs2/aops.c')
-rw-r--r-- | fs/ocfs2/aops.c | 173 |
1 files changed, 171 insertions, 2 deletions
diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index fef0186a91cd..34d10452c56d 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c | |||
@@ -206,8 +206,8 @@ bail: | |||
206 | return err; | 206 | return err; |
207 | } | 207 | } |
208 | 208 | ||
209 | static int ocfs2_read_inline_data(struct inode *inode, struct page *page, | 209 | int ocfs2_read_inline_data(struct inode *inode, struct page *page, |
210 | struct buffer_head *di_bh) | 210 | struct buffer_head *di_bh) |
211 | { | 211 | { |
212 | void *kaddr; | 212 | void *kaddr; |
213 | unsigned int size; | 213 | unsigned int size; |
@@ -1432,6 +1432,130 @@ out: | |||
1432 | return ret; | 1432 | return ret; |
1433 | } | 1433 | } |
1434 | 1434 | ||
1435 | static int ocfs2_write_begin_inline(struct address_space *mapping, | ||
1436 | struct inode *inode, | ||
1437 | struct ocfs2_write_ctxt *wc) | ||
1438 | { | ||
1439 | int ret; | ||
1440 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | ||
1441 | struct page *page; | ||
1442 | handle_t *handle; | ||
1443 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)wc->w_di_bh->b_data; | ||
1444 | |||
1445 | page = find_or_create_page(mapping, 0, GFP_NOFS); | ||
1446 | if (!page) { | ||
1447 | ret = -ENOMEM; | ||
1448 | mlog_errno(ret); | ||
1449 | goto out; | ||
1450 | } | ||
1451 | /* | ||
1452 | * If we don't set w_num_pages then this page won't get unlocked | ||
1453 | * and freed on cleanup of the write context. | ||
1454 | */ | ||
1455 | wc->w_pages[0] = wc->w_target_page = page; | ||
1456 | wc->w_num_pages = 1; | ||
1457 | |||
1458 | handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); | ||
1459 | if (IS_ERR(handle)) { | ||
1460 | ret = PTR_ERR(handle); | ||
1461 | mlog_errno(ret); | ||
1462 | goto out; | ||
1463 | } | ||
1464 | |||
1465 | ret = ocfs2_journal_access(handle, inode, wc->w_di_bh, | ||
1466 | OCFS2_JOURNAL_ACCESS_WRITE); | ||
1467 | if (ret) { | ||
1468 | ocfs2_commit_trans(osb, handle); | ||
1469 | |||
1470 | mlog_errno(ret); | ||
1471 | goto out; | ||
1472 | } | ||
1473 | |||
1474 | if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL)) | ||
1475 | ocfs2_set_inode_data_inline(inode, di); | ||
1476 | |||
1477 | if (!PageUptodate(page)) { | ||
1478 | ret = ocfs2_read_inline_data(inode, page, wc->w_di_bh); | ||
1479 | if (ret) { | ||
1480 | ocfs2_commit_trans(osb, handle); | ||
1481 | |||
1482 | goto out; | ||
1483 | } | ||
1484 | } | ||
1485 | |||
1486 | wc->w_handle = handle; | ||
1487 | out: | ||
1488 | return ret; | ||
1489 | } | ||
1490 | |||
1491 | int ocfs2_size_fits_inline_data(struct buffer_head *di_bh, u64 new_size) | ||
1492 | { | ||
1493 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; | ||
1494 | |||
1495 | if (new_size < le16_to_cpu(di->id2.i_data.id_count)) | ||
1496 | return 1; | ||
1497 | return 0; | ||
1498 | } | ||
1499 | |||
1500 | static int ocfs2_try_to_write_inline_data(struct address_space *mapping, | ||
1501 | struct inode *inode, loff_t pos, | ||
1502 | unsigned len, struct page *mmap_page, | ||
1503 | struct ocfs2_write_ctxt *wc) | ||
1504 | { | ||
1505 | int ret, written = 0; | ||
1506 | loff_t end = pos + len; | ||
1507 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | ||
1508 | |||
1509 | mlog(0, "Inode %llu, write of %u bytes at off %llu. features: 0x%x\n", | ||
1510 | (unsigned long long)oi->ip_blkno, len, (unsigned long long)pos, | ||
1511 | oi->ip_dyn_features); | ||
1512 | |||
1513 | /* | ||
1514 | * Handle inodes which already have inline data 1st. | ||
1515 | */ | ||
1516 | if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { | ||
1517 | if (mmap_page == NULL && | ||
1518 | ocfs2_size_fits_inline_data(wc->w_di_bh, end)) | ||
1519 | goto do_inline_write; | ||
1520 | |||
1521 | /* | ||
1522 | * The write won't fit - we have to give this inode an | ||
1523 | * inline extent list now. | ||
1524 | */ | ||
1525 | ret = ocfs2_convert_inline_data_to_extents(inode, wc->w_di_bh); | ||
1526 | if (ret) | ||
1527 | mlog_errno(ret); | ||
1528 | goto out; | ||
1529 | } | ||
1530 | |||
1531 | /* | ||
1532 | * Check whether the inode can accept inline data. | ||
1533 | */ | ||
1534 | if (oi->ip_clusters != 0 || i_size_read(inode) != 0) | ||
1535 | return 0; | ||
1536 | |||
1537 | /* | ||
1538 | * Check whether the write can fit. | ||
1539 | */ | ||
1540 | if (mmap_page || end > ocfs2_max_inline_data(inode->i_sb)) | ||
1541 | return 0; | ||
1542 | |||
1543 | do_inline_write: | ||
1544 | ret = ocfs2_write_begin_inline(mapping, inode, wc); | ||
1545 | if (ret) { | ||
1546 | mlog_errno(ret); | ||
1547 | goto out; | ||
1548 | } | ||
1549 | |||
1550 | /* | ||
1551 | * This signals to the caller that the data can be written | ||
1552 | * inline. | ||
1553 | */ | ||
1554 | written = 1; | ||
1555 | out: | ||
1556 | return written ? written : ret; | ||
1557 | } | ||
1558 | |||
1435 | /* | 1559 | /* |
1436 | * This function only does anything for file systems which can't | 1560 | * This function only does anything for file systems which can't |
1437 | * handle sparse files. | 1561 | * handle sparse files. |
@@ -1483,6 +1607,19 @@ int ocfs2_write_begin_nolock(struct address_space *mapping, | |||
1483 | return ret; | 1607 | return ret; |
1484 | } | 1608 | } |
1485 | 1609 | ||
1610 | if (ocfs2_supports_inline_data(osb)) { | ||
1611 | ret = ocfs2_try_to_write_inline_data(mapping, inode, pos, len, | ||
1612 | mmap_page, wc); | ||
1613 | if (ret == 1) { | ||
1614 | ret = 0; | ||
1615 | goto success; | ||
1616 | } | ||
1617 | if (ret < 0) { | ||
1618 | mlog_errno(ret); | ||
1619 | goto out; | ||
1620 | } | ||
1621 | } | ||
1622 | |||
1486 | ret = ocfs2_expand_nonsparse_inode(inode, pos, len, wc); | 1623 | ret = ocfs2_expand_nonsparse_inode(inode, pos, len, wc); |
1487 | if (ret) { | 1624 | if (ret) { |
1488 | mlog_errno(ret); | 1625 | mlog_errno(ret); |
@@ -1570,6 +1707,7 @@ int ocfs2_write_begin_nolock(struct address_space *mapping, | |||
1570 | if (meta_ac) | 1707 | if (meta_ac) |
1571 | ocfs2_free_alloc_context(meta_ac); | 1708 | ocfs2_free_alloc_context(meta_ac); |
1572 | 1709 | ||
1710 | success: | ||
1573 | *pagep = wc->w_target_page; | 1711 | *pagep = wc->w_target_page; |
1574 | *fsdata = wc; | 1712 | *fsdata = wc; |
1575 | return 0; | 1713 | return 0; |
@@ -1637,6 +1775,31 @@ out_fail: | |||
1637 | return ret; | 1775 | return ret; |
1638 | } | 1776 | } |
1639 | 1777 | ||
1778 | static void ocfs2_write_end_inline(struct inode *inode, loff_t pos, | ||
1779 | unsigned len, unsigned *copied, | ||
1780 | struct ocfs2_dinode *di, | ||
1781 | struct ocfs2_write_ctxt *wc) | ||
1782 | { | ||
1783 | void *kaddr; | ||
1784 | |||
1785 | if (unlikely(*copied < len)) { | ||
1786 | if (!PageUptodate(wc->w_target_page)) { | ||
1787 | *copied = 0; | ||
1788 | return; | ||
1789 | } | ||
1790 | } | ||
1791 | |||
1792 | kaddr = kmap_atomic(wc->w_target_page, KM_USER0); | ||
1793 | memcpy(di->id2.i_data.id_data + pos, kaddr + pos, *copied); | ||
1794 | kunmap_atomic(kaddr, KM_USER0); | ||
1795 | |||
1796 | mlog(0, "Data written to inode at offset %llu. " | ||
1797 | "id_count = %u, copied = %u, i_dyn_features = 0x%x\n", | ||
1798 | (unsigned long long)pos, *copied, | ||
1799 | le16_to_cpu(di->id2.i_data.id_count), | ||
1800 | le16_to_cpu(di->i_dyn_features)); | ||
1801 | } | ||
1802 | |||
1640 | int ocfs2_write_end_nolock(struct address_space *mapping, | 1803 | int ocfs2_write_end_nolock(struct address_space *mapping, |
1641 | loff_t pos, unsigned len, unsigned copied, | 1804 | loff_t pos, unsigned len, unsigned copied, |
1642 | struct page *page, void *fsdata) | 1805 | struct page *page, void *fsdata) |
@@ -1650,6 +1813,11 @@ int ocfs2_write_end_nolock(struct address_space *mapping, | |||
1650 | handle_t *handle = wc->w_handle; | 1813 | handle_t *handle = wc->w_handle; |
1651 | struct page *tmppage; | 1814 | struct page *tmppage; |
1652 | 1815 | ||
1816 | if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { | ||
1817 | ocfs2_write_end_inline(inode, pos, len, &copied, di, wc); | ||
1818 | goto out_write_size; | ||
1819 | } | ||
1820 | |||
1653 | if (unlikely(copied < len)) { | 1821 | if (unlikely(copied < len)) { |
1654 | if (!PageUptodate(wc->w_target_page)) | 1822 | if (!PageUptodate(wc->w_target_page)) |
1655 | copied = 0; | 1823 | copied = 0; |
@@ -1687,6 +1855,7 @@ int ocfs2_write_end_nolock(struct address_space *mapping, | |||
1687 | block_commit_write(tmppage, from, to); | 1855 | block_commit_write(tmppage, from, to); |
1688 | } | 1856 | } |
1689 | 1857 | ||
1858 | out_write_size: | ||
1690 | pos += copied; | 1859 | pos += copied; |
1691 | if (pos > inode->i_size) { | 1860 | if (pos > inode->i_size) { |
1692 | i_size_write(inode, pos); | 1861 | i_size_write(inode, pos); |