freebsd-skq/sys/ufs/lfs/TODO

117 lines
5.0 KiB
Plaintext
Raw Normal View History

1994-05-24 10:09:53 +00:00
# @(#)TODO 8.1 (Berkeley) 6/11/93
NOTE: Changed the lookup on a page of inodes to search from the back
in case the same inode gets written twice on the same page.
Make sure that if you are writing a file, but not all the blocks
make it into a single segment, that you do not write the inode in
that segment.
Keith:
Why not delete the lfs_bmapv call, just mark everything dirty
that isn't deleted/truncated? Get some numbers about
what percentage of the stuff that the cleaner thinks
might be live is live. If it's high, get rid of lfs_bmapv.
There is a nasty problem in that it may take *more* room to write
the data to clean a segment than is returned by the new segment
because of indirect blocks in segment 2 being dirtied by the data
being copied into the log from segment 1. The suggested solution
at this point is to detect it when we have no space left on the
filesystem, write the extra data into the last segment (leaving
no clean ones), make it a checkpoint and shut down the file system
for fixing by a utility reading the raw partition. Argument is
that this should never happen and is practically impossible to fix
since the cleaner would have to theoretically build a model of the
entire filesystem in memory to detect the condition occurring.
A file coalescing cleaner will help avoid the problem, and one
that reads/writes from the raw disk could fix it.
DONE Currently, inodes are being flushed to disk synchronously upon
creation -- see ufs_makeinode. However, only the inode
is flushed, the directory "name" is written using VOP_BWRITE,
so it's not synchronous. Possible solutions: 1: get some
ordering in the writes so that inode/directory entries get
stuffed into the same segment. 2: do both synchronously
3: add Mendel's information into the stream so we log
creation/deletion of inodes. 4: do some form of partial
segment when changing the inode (creation/deletion/rename).
DONE Fix i_block increment for indirect blocks.
If the file system is tar'd, extracted on top of another LFS, the
IFILE ain't worth diddly. Is the cleaner writing the IFILE?
If not, let's make it read-only.
DONE Delete unnecessary source from utils in main-line source tree.
DONE Make sure that we're counting meta blocks in the inode i_block count.
Overlap the version and nextfree fields in the IFILE
DONE Vinvalbuf (Kirk):
Why writing blocks that are no longer useful?
Are the semantics of close such that blocks have to be flushed?
How specify in the buf chain the blocks that don't need
to be written? (Different numbering of indirect blocks.)
Margo:
Change so that only search one sector of inode block file for the
inode by using sector addresses in the ifile instead of
logical disk addresses.
Fix the use of the ifile version field to use the generation
number instead.
DONE Unmount; not doing a bgetvp (VHOLD) in lfs_newbuf call.
DONE Document in the README file where the checkpoint information is
on disk.
Variable block sizes (Margo/Keith).
Switch the byte accounting to sector accounting.
DONE Check lfs.h and make sure that the #defines/structures are all
actually needed.
DONE Add a check in lfs_segment.c so that if the segment is empty,
we don't write it.
Need to keep vnode v_numoutput up to date for pending writes?
DONE USENIX paper (Carl/Margo).
Evelyn:
lfsck: If delete a file that's being executed, the version number
isn't updated, and lfsck has to figure this out; case is the same as if have an inode that no directory references,
so the file should be reattached into lost+found.
Recovery/fsck.
Carl:
Investigate: clustering of reads (if blocks in the segment are ordered,
should read them all) and writes (McVoy paper).
Investigate: should the access time be part of the IFILE:
pro: theoretically, saves disk writes
con: cacheing inodes should obviate this advantage
the IFILE is already humongous
Cleaner.
Port to OSF/1 (Carl/Keith).
Currently there's no notion of write error checking.
+ Failed data/inode writes should be rescheduled (kernel level
bad blocking).
+ Failed superblock writes should cause selection of new
superblock for checkpointing.
FUTURE FANTASIES: ============
+ unrm, versioning
+ transactions
+ extended cleaner policies (hot/cold data, data placement)
==============================
Problem with the concept of multiple buffer headers referencing the segment:
Positives:
Don't lock down 1 segment per file system of physical memory.
Don't copy from buffers to segment memory.
Don't tie down the bus to transfer 1M.
Works on controllers supporting less than large transfers.
Disk can start writing immediately instead of waiting 1/2 rotation
and the full transfer.
Negatives:
Have to do segment write then segment summary write, since the latter
is what verifies that the segment is okay. (Is there another way
to do this?)
==============================
The algorithm for selecting the disk addresses of the super-blocks
has to be available to the user program which checks the file system.
(Currently in newfs, becomes a common subroutine.)