CVS log for src/sbin/fsck_lfs/segwrite.c

Up to [cvs.NetBSD.org] / src / sbin / fsck_lfs

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

Implementation of exFAT filesystem, with compilation conditional on MKEXFATFS
make variable.

Remove extra semicolon.

Merge changes from current as of 20200406

Avoid common symbols for fsck_lfs.

Pull up following revision(s) (requested by maya in ticket #330):
	sbin/fsck_lfs/inode.c: 1.69
	sbin/fsck_lfs/lfs.c: 1.73
	sbin/fsck_lfs/pass6.c: 1.50
	sbin/fsck_lfs/segwrite.c: 1.46
	sys/ufs/lfs/lfs.h: 1.202-1.203
	sys/ufs/lfs/lfs_accessors.h: 1.48
	sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
	sys/ufs/lfs/lfs_balloc.c: 1.94
	sys/ufs/lfs/lfs_bio.c: 1.141
	sys/ufs/lfs/lfs_extern.h: 1.113
	sys/ufs/lfs/lfs_inode.c: 1.156-1.157
	sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
	sys/ufs/lfs/lfs_itimes.c: 1.20
	sys/ufs/lfs/lfs_pages.c: 1.13-1.15
	sys/ufs/lfs/lfs_rename.c: 1.22
	sys/ufs/lfs/lfs_segment.c: 1.270-1.275
	sys/ufs/lfs/lfs_subr.c: 1.94-1.97
	sys/ufs/lfs/lfs_syscalls.c: 1.175
	sys/ufs/lfs/lfs_vfsops.c: 1.360
	sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
	sys/ufs/lfs/ulfs_inode.c: 1.20
	sys/ufs/lfs/ulfs_inode.h: 1.24
	sys/ufs/lfs/ulfs_lookup.c: 1.41
	sys/ufs/lfs/ulfs_quota2.c: 1.31
	sys/ufs/lfs/ulfs_readwrite.c: 1.24
	sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment

Update inode member i_flag --> i_state to keep up with kernel changes

Use IINFO in lfs_writeinode().
(both the kernel and the userland copies)

Use daddr_t, not ulfs_daddr_t, as the latter's 32 bits wide.
Don't use either for on-disk items.
Declare external data in header files.
Part 3 of 3.

Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)

The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.

Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.

Part two of dinodes; use the same union everywhere.
(previously the ufs-derived code had things set up slightly different)

Remove a bunch of associated mess.

Hack up dinode usage to be 64 vs. 32 as needed. Part 1.

(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)

Provide 32-bit and 64-bit versions of FINFO.

This also entailed sorting out part of struct segment, as that
contains a pointer into the current FINFO data.

Make 32-bit and 64-bit versions of SEGSUM.
Also fix some of the FINFO handling as it's closely entangled.

Add IFILE32 and IFILE64 structures for the on-disk ifile entries.
Add and use accessors. There are also a bunch of places that cast and
I hope I've found them all...

Make 32-bit and 64-bit versions of CLEANERINFO.

XXX: while this is written to disk, it seems like much of it would
XXX: be better set up as a commpage shared with the cleaner.

Add a (draft) 64-bit superblock. Make things build again.

Add pieces of support for using both superblock types where
convenient, and specifically to the superblock accessors, but don't
actually enable it anywhere.

First substantive step on PR 50000.

Use accessor functions for the version field of the lfs superblock.
I thought at first maybe the cases that test the version should be
rolled into the accessors, but on the whole I think the conclusion on
that is no.

Add a new lfs header file: lfs_accessors.h.

This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.

More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)

The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.

Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)

It also gets rid of cpp abuse in the form of fake structure member
macros.

Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)

XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.

add missing ;

fix error messages containing \n

Use VFS_PROTOS() for lfs.
Rename conflicting struct lfs field "lfs_start" to "lfs_s0addr".

No functional change.

- Fix fallout for recent bread() change (removing cred arg).

Rebase to HEAD as of a few days ago.

sync with head.

for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.

this commit was splitted into small chunks to avoid
a limitation of cvs.  ("Protocol error: too many arguments")

fix unused variable warnings.

resync from head

Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.

Tidy up the LFS userland build hacks.
Don't use -I${NETBSDSRCDIR}/sys; don't include files other than the
exported LFS headers, which are lfs.h, lfs_inode.h, and (for now)
lfs_extern.h.

Move the dinode (on-disk inode) structures to lfs.h, since they are
and will be obviously required by userland tools that need to read
the on-disk structures.

Also, DINODE{1,2}_SIZE -> LFS_DINODE{1,2}_SIZE.

Cleanups and hacks to make lfs userland stuff build:
   - lfs_cksum.c doesn't actually need ulfs_inode.h any more.
   - neither does lfs_itimes.c.
   - add hacks to fsck_lfs to make it compile.
   - add hacks to newfs_lfs to make it compile.
   - fix warning in ulfs_quota.c when quotas are fully disabled
     (as I guess is happening with the rumpity version)

XXX: This commit adds -I${NETBSDSRCDIR}/sys to the Makefiles for
XXX: fsck_lfs, newfs_lfs, and lfs_cleanerd. This needs to be cleaned
XXX: up ASAP; but I consider this less problematic in the short term
XXX: than spewing ulfs_*.h into /usr/include.

ufs -> ulfs for fsck_lfs.

resync with head

sync with head

Stuff UFS_ in front of a few of ufs's symbols to reduce namespace
pollution. Specifically:
   ROOTINO -> UFS_ROOTINO
   WINO -> UFS_WINO
   NXADDR -> UFS_NXADDR
   NDADDR -> UFS_NDADDR
   NIADDR -> UFS_NIADDR
   MAXSYMLINKLEN -> UFS_MAXSYMLINKLEN
   MAXSYMLINKLEN_UFS[12] -> UFS[12]_MAXSYMLINKLEN (for consistency)

Sort out ext2fs's misuse of NDADDR and NIADDR; fortunately, these have
the same values in ext2fs and ffs.

No functional change intended.

Three changes in a single commit.

- drop the notion of frags (LFS fragments) vs fsb (FFS fragments)
  The code uses a complicated unity function that just makes the
  code difficult to understand.

- support larger sector sizes. Fix disk address computations
  to use DEV_BSIZE in the kernel as required by device drivers
  and to use sector sizes in userland.

- Fix several locking bugs in lfs_bio.c and lfs_subr.c.

Sync w/ -current. 34 merge conflicts to follow.

Sync with HEAD.

sync with head.

Make sure all cached buffers with valid, not yet written data have been
run through copy-on-write.  Call fscow_run() with valid data where possible.

The LP_UFSCOW hack is no longer needed to protect ffs_copyonwrite() against
endless recursion.

- Add a flag B_MODIFY to bread(), breada() and breadn().  If set the caller
  intends to modify the buffer returned.

- Always run copy-on-write on buffers returned from ffs_balloc().

- Add new function ffs_getblk() that gets a buffer, assigns a new blkno,
  may clear the buffer and runs copy-on-write.  Process possible errors
  from getblk() or fscow_run().  Part of PR kern/38664.

Welcome to 4.99.63

Reviewed by: YAMAMOTO Takashi <yamt@netbsd.org>

Remove clause 3 and 4 from TNF licenses

sync with HEAD

Merge from vmlocking:

- Split vnode::v_flag into three fields, depending on field locking.
- simple_lock -> kmutex in a few places.
- Fix some simple locking problems.

Give brelse() a second argument so that it matches the kernel.
fsck_lfs now compiles again.

Fix malloc/realloc/calloc issues: always check and exit, use EEXIT instead
of 8.

Several fixes to improve the reliability of the roll-forward agent.
Also, note "properly orphaned" files as distinct from corrupted files.

Various improvements to fsck_lfs, to wit:

* Add lfs_balloc capability to the lfs library.
* Extend the Ifile if we run out of free inodes when creating lost+found.
* Don't roll forward if we have allocated a lost+found, to avoid
  conflicts when adding new files in roll-forward.
* Make some messages slightly more verbose (e.g. include inode number,
  and use pwarn() instead of printf() so the messages include the device
  name when preening).
* Change superblock detection/avoidance to use the offset table in the
  primary superblock, rather than looking at the contents.
* Be more verbose about various operations when passed the -d flag,
  especially roll-forward.
* Be more careful about dirops during roll forward, since the cleaner can
  sometimes write blocks from dirop vnodes.  Detect and avoid this problem.
* Always check the free list, even if given -i; if we're going to write
  it we have to check it first.
* Mark inodes dirty when blocks are found during roll forward, so the
  inodes are written with the new block locations.
* Update size of inodes if blocks beyond EOF are found during roll
  forward.
* Fix segment accounting for blocks and inodes found during roll
  forward.
* Report statistics on roll forward: how many new/deleted/moved files
  and how many updated blocks (or "nothing new").
* Don't care if the device being checked is really a device, if we have
  been passed the -f flag (to facilitate automated testing).
* When writing to the disk, use the current time in the segment headers
  rathern than time 0.
* When passed the -i flag, locate the partial segment containing the
  Ifile inode and use that to calculate lfs_offset, lfs_curseg,
  lfs_nextseg.  (Again for automated testing.)

Coverity CID 3447: Add extraneous checks to shut up Coverity.

Check for allocation failures in malloc, calloc, realloc, asprintf, and
vasprintf and try to handle them.

rename lfs.h to lfs_user.h so that it does not conflict.

Pull up the following revisions (requested by perseant in ticket #1281):

        1.8 sys/ufs/lfs/TODO
        1.75 sys/ufs/lfs/lfs.h                  (via patch)
        1.74 sys/ufs/lfs/lfs_alloc.c            (via patch)
        1.49, 1.51 sys/ufs/lfs/lfs_balloc.c     (1.51 via patch)
        1.78 sys/ufs/lfs/lfs_bio.c
        1.62 sys/ufs/lfs/lfs_extern.h           (via patch)
        1.156 sys/ufs/lfs/lfs_segment.c         (via patch)
        1.48 sys/ufs/lfs/lfs_subr.c
        1.101 sys/ufs/lfs/lfs_syscalls.c
        1.163 sys/ufs/lfs/lfs_vfsops.c          (via patch)
        1.134 sys/ufs/lfs/lfs_vnops.c           (via patch)
        1.61 sys/ufs/ufs/ufs_readwrite.c        (via patch)

        1.20 libexec/lfs_cleanerd/clean.h       (via patch)
        1.52 libexec/lfs_cleanerd/cleanerd.c    (via patch)
        1.41 libexec/lfs_cleanerd/library.c     (via patch)

        1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
        1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
        1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
        1.3 sbin/fsck_lfs/bufcache.c
        1.3 sbin/fsck_lfs/bufcache.h
        1.3 sbin/fsck_lfs/lfs.h
        1.8 sbin/fsck_lfs/lfs.c                 (via patch)
        1.8 sbin/fsck_lfs/pass3.c               (via patch)
        1.18 sbin/fsck_lfs/pass0.c              (via patch)
        1.18 sbin/fsck_lfs/utilities.c          (via patch)
        1.7 sbin/fsck_lfs/segwrite.c
        1.19 sbin/fsck_lfs/setup.c              (via patch)
        1.3 sbin/newfs_lfs/Makefile
        0 sbin/newfs_lfs/lfs.c                  (yes, remove it)
        1.1 sbin/newfs_lfs/make_lfs.c
        1.15 sbin/newfs_lfs/newfs.c             (via patch)

Various minor LFS improvements.

Kernel:

* Note when lfs_putpages(9) thinks it is not going to be writing any
   pages before calling genfs_putpages(9).  This prevents a situation in
   which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
   overestimate in most cases.  Note that if NRESERVE() is too high, it
   may be impossible to create files on the filesystem.  We catch this
   case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
   entries in indirect blocks again, triggering a failed assertion "daddr
   <= LFS_MAX_DADDR".  Explicitly convert to and from int32_t to correct
   this.  Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
   hold before triggering a flush.  This is settable by sysctl, but off
   (zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
   shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
   even though their v_size == 0.  Don't panic when we see this.
   Fixes PR #26680.
* Change lfs_bfree to a signed quantity.  The manner in which it is
   processed before being passed to the cleaner means that sometimes it
   may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
   lfs_statfs(9).  This prevents df(1) from ever telling us that our full
   filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
   associated buffer headers, so that the pagedaemon doesn't run us out
   of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
   unmounted.  Because vfs_busy() is a shared lock, and
   lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
   holding the lock that umount() is blocking on, then try to vfs_busy()
   again in getnewvnode().

cleaner:

* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
   so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
   segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.

newfs_lfs, fsck_lfs, and regression:

* Extend the lfs library from fsck_lfs(8) so that it can be used with a
   not-yet-existent LFS.  Make newfs_lfs(8) use this library, so it can
   create LFSs whose Ifile is larger than one segment.  Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
   -n flag.
* Remove the Ifile from the filesystem namespace.  The cleaner now uses
   a fcntl call on the root inode to find the Ifile filehandle.  (As a
   side-effect, addresses PR #29144.)

Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
  roll-forward
* Keep better track of the free list during roll-forward, avoiding
  a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
  truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
  in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
  write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.

Move the definition of simple_lock() and simple_unlock() to a common
header, since more of the LFS macros now use these functions.  Since
we're outside of the kernel, these are defined to be empty.

"#define lfs_devvp lfs_unlockvp" for readability, since that's what we
use it for in fsck_lfs/newfs_lfs.

Various minor LFS improvements:

* Extend the lfs library from fsck_lfs(8) so that it can be used with a
  not-yet-existent LFS.  Make newfs_lfs(8) use this library, so it can
  create LFSs whose Ifile is larger than one segment.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
  -n flag.

Check result of malloc().
reviewed by Martin.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22308, verified by myself.

don't use uninitialized variables.
PR/22123 (Izumi Tsutsui)

Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.

This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.

Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.

Check inode free list tail pointer as well as head pointer, and write both
into the CLEANERINFO block of the Ifile as well as into the superblock.
Make preen update both superblocks.

Add working writing ability to fsck_lfs, including roll-forward, based on
a partial-segment writer ported from the kernel.