diff options
| author | H. Peter Anvin <hpa@linux.intel.com> | 2010-05-13 14:38:04 -0700 |
|---|---|---|
| committer | H. Peter Anvin <hpa@linux.intel.com> | 2010-05-13 14:38:04 -0700 |
| commit | 4ade7735e6607f3db51df50d4e79e636ba6a0121 (patch) | |
| tree | 77b487df25ca96007c92fd49c99943e03c977039 | |
| parent | ae197a98b905a7da7872eef29dcbb0f822fa81de (diff) | |
| download | syslinux-4ade7735e6607f3db51df50d4e79e636ba6a0121.tar.gz syslinux-4ade7735e6607f3db51df50d4e79e636ba6a0121.tar.xz syslinux-4ade7735e6607f3db51df50d4e79e636ba6a0121.zip | |
fat: fix confusion between byte and sector counts
clust_shift is in units of sectors, not in bytes; this was mixed up in
the extent finder.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
| -rw-r--r-- | core/fs/fat/fat.c | 21 |
1 files changed, 11 insertions, 10 deletions
diff --git a/core/fs/fat/fat.c b/core/fs/fat/fat.c index 029c2d4a..e2fcc14c 100644 --- a/core/fs/fat/fat.c +++ b/core/fs/fat/fat.c @@ -93,13 +93,17 @@ static int fat_next_extent(struct inode *inode, uint32_t lstart) uint32_t lcluster; uint32_t pcluster; uint32_t tcluster; - uint32_t cluster_size = UINT32_C(1) << sbi->clust_shift; + const uint32_t cluster_bytes = UINT32_C(1) << sbi->clust_byte_shift; + const uint32_t cluster_secs = UINT32_C(1) << sbi->clust_shift; sector_t data_area = sbi->data; - tcluster = (inode->size + cluster_size - 1) >> sbi->clust_shift; + tcluster = (inode->size + cluster_bytes - 1) >> sbi->clust_byte_shift; if (mcluster >= tcluster) goto err; /* Requested cluster beyond end of file */ + dprintf("Getting next cluster, inode = %p, lstart = %u, mcluster = %u, tcluster = %u\n", + inode, lstart, mcluster, tcluster); + if (inode->next_extent.len) { if (inode->next_extent.pstart < data_area) goto err; /* Root directory has only one extent */ @@ -132,7 +136,7 @@ static int fat_next_extent(struct inode *inode, uint32_t lstart) inode->next_extent.pstart = ((sector_t)(pcluster-2) << sbi->clust_shift) + data_area; - inode->next_extent.len = cluster_size; + inode->next_extent.len = cluster_secs; lcluster++; while (lcluster < tcluster) { @@ -141,7 +145,7 @@ static int fat_next_extent(struct inode *inode, uint32_t lstart) if (xcluster != pcluster+1) break; /* Not contiguous */ pcluster = xcluster; - inode->next_extent.len += cluster_size; + inode->next_extent.len += cluster_secs; lcluster++; } return 0; @@ -183,12 +187,9 @@ static sector_t get_next_sector(struct fs_info* fs, uint32_t sector) } /* - * Here comes the place I don't like VFAT fs most; if we need seek - * the file to the right place, we need get the right sector address - * from begining everytime! Since it's a kind a signle link list, we - * need to traver from the head-node to find the right node in that list. - * - * What a waste of time! + * The FAT is a single-linked list. We remember the last place we + * were, so for a forward seek we can move forward from there, but + * for a reverse seek we have to start over... */ static sector_t get_the_right_sector(struct file *file) { |