File and directory collision bug #17
Comments
|
I've written some tests (calls the int21/16 function directly) for Dosemu2's test suite that show that at least some part of the problem is in the kernel. FreeDOS (git and 1.2) kernel on FAT does: With MS-DOS 6.22 and DR-DOS 7.01 the existing file or directory is left intact and the volume label is set in a duplicate file, but with the VOL attr set(effectively hidden). These DOSs also set the volume name in the BPB. Edit: updated some incorrect statements. |
|
Probably connected or the same issue: https://gitlab.com/FreeDOS/issue-reporting/-/issues/31 |
|
Thanks Robert, I have seen this already and I do agree that it's probably connected. Once I figure out what exactly is required in both Label and Kernel I'll try to submit patches. |
Part of a fix for [FDOS/label#17]
Part of a fix for [FDOS/label#17]
If the BPB is either v4.1 or v7 long, then its volume label field should be written. Note: This site https://jdebp.uk/FGA/bios-parameter-block.html suggests that it is perfectly valid to have a v7 long BPB with a FAT12 or FAT16 filesystem, although more usually it's used for FAT32. However I can't see any confirmation of this elsewhere, haven't seen an example of this in the wild, and have no means of generating a test article. More importantly since we are writing to the filesystem, it's important to not have any false positives or we could cause corruption. So for now this combination, should it exist, will not be updated. See the discussion here dosemu2/fdpp#202. Part of a fix for [FDOS/label#17]
Part of a fix for [FDOS/label#17]
If the BPB is either v4.1 or v7 long, then its volume label field should be written. Note: This site https://jdebp.uk/FGA/bios-parameter-block.html suggests that it is perfectly valid to have a v7 long BPB with a FAT12 or FAT16 filesystem, although more usually it's used for FAT32. However I can't see any confirmation of this elsewhere, haven't seen an example of this in the wild, and have no means of generating a test article. More importantly since we are writing to the filesystem, it's important to not have any false positives or we could cause corruption. So for now this combination, should it exist, will not be updated. See the discussion here dosemu2/fdpp#202. Part of a fix for [FDOS/label#17]
Part of a fix for [FDOS/label#17]
If the BPB is either v4.1 or v7 long, then its volume label field should be written. Note: This site https://jdebp.uk/FGA/bios-parameter-block.html suggests that it is perfectly valid to have a v7 long BPB with a FAT12 or FAT16 filesystem, although more usually it's used for FAT32. However I can't see any confirmation of this elsewhere, haven't seen an example of this in the wild, and have no means of generating a test article. More importantly since we are writing to the filesystem, it's important to not have any false positives or we could cause corruption. So for now this combination, should it exist, will not be updated. See the discussion here dosemu2/fdpp#202. Part of a fix for [FDOS/label#17]
Reported on SourceForge as Issue 365
According to the bug report... Basically, LABEL is overwriting root directory entries for files and directories without checking for conflicts. Without testing myself, I would assume it is probably also corrupting the filesystem when it overwrites an entry.
The text was updated successfully, but these errors were encountered: