piecemeal, middle-of-file writes for NFS. These hacks have caused no
end of trouble, especially when combined with mmap(). I've removed
them. Instead, NFS will issue a read-before-write to fully
instantiate the struct buf containing the write. NFS does, however,
optimize piecemeal appends to files. For most common file operations,
you will not notice the difference. The sole remaining fragment in
the VFS/BIO system is b_dirtyoff/end, which NFS uses to avoid cache
coherency issues with read-merge-write style operations. NFS also
optimizes the write-covers-entire-buffer case by avoiding the
read-before-write. There is quite a bit of room for further
optimization in these areas.
The VM system marks pages fully-valid (AKA vm_page_t->valid =
VM_PAGE_BITS_ALL) in several places, most noteably in vm_fault. This
is not correct operation. The vm_pager_get_pages() code is now
responsible for marking VM pages all-valid. A number of VM helper
routines have been added to aid in zeroing-out the invalid portions of
a VM page prior to the page being marked all-valid. This operation is
necessary to properly support mmap(). The zeroing occurs most often
when dealing with file-EOF situations. Several bugs have been fixed
in the NFS subsystem, including bits handling file and directory EOF
situations and buf->b_flags consistancy issues relating to clearing
B_ERROR & B_INVAL, and handling B_DONE.
getblk() and allocbuf() have been rewritten. B_CACHE operation is now
formally defined in comments and more straightforward in
implementation. B_CACHE for VMIO buffers is based on the validity of
the backing store. B_CACHE for non-VMIO buffers is based simply on
whether the buffer is B_INVAL or not (B_CACHE set if B_INVAL clear,
and vise-versa). biodone() is now responsible for setting B_CACHE
when a successful read completes. B_CACHE is also set when a bdwrite()
is initiated and when a bwrite() is initiated. VFS VOP_BWRITE
routines (there are only two - nfs_bwrite() and bwrite()) are now
expected to set B_CACHE. This means that bowrite() and bawrite() also
set B_CACHE indirectly.
There are a number of places in the code which were previously using
buf->b_bufsize (which is DEV_BSIZE aligned) when they should have
been using buf->b_bcount. These have been fixed. getblk() now clears
B_DONE on return because the rest of the system is so bad about
dealing with B_DONE.
Major fixes to NFS/TCP have been made. A server-side bug could cause
requests to be lost by the server due to nfs_realign() overwriting
other rpc's in the same TCP mbuf chain. The server's kernel must be
recompiled to get the benefit of the fixes.
Submitted by: Matthew Dillon <dillon@apollo.backplane.com>
This is a seriously beefed up chroot kind of thing. The process
is jailed along the same lines as a chroot does it, but with
additional tough restrictions imposed on what the superuser can do.
For all I know, it is safe to hand over the root bit inside a
prison to the customer living in that prison, this is what
it was developed for in fact: "real virtual servers".
Each prison has an ip number associated with it, which all IP
communications will be coerced to use and each prison has its own
hostname.
Needless to say, you need more RAM this way, but the advantage is
that each customer can run their own particular version of apache
and not stomp on the toes of their neighbors.
It generally does what one would expect, but setting up a jail
still takes a little knowledge.
A few notes:
I have no scripts for setting up a jail, don't ask me for them.
The IP number should be an alias on one of the interfaces.
mount a /proc in each jail, it will make ps more useable.
/proc/<pid>/status tells the hostname of the prison for
jailed processes.
Quotas are only sensible if you have a mountpoint per prison.
There are no privisions for stopping resource-hogging.
Some "#ifdef INET" and similar may be missing (send patches!)
If somebody wants to take it from here and develop it into
more of a "virtual machine" they should be most welcome!
Tools, comments, patches & documentation most welcome.
Have fun...
Sponsored by: http://www.rndassociates.com/
Run for almost a year by: http://www.servetheweb.com/
1:
s/suser/suser_xxx/
2:
Add new function: suser(struct proc *), prototyped in <sys/proc.h>.
3:
s/suser_xxx(\([a-zA-Z0-9_]*\)->p_ucred, \&\1->p_acflag)/suser(\1)/
The remaining suser_xxx() calls will be scrutinized and dealt with
later.
There may be some unneeded #include <sys/cred.h>, but they are left
as an exercise for Bruce.
More changes to the suser() API will come along with the "jail" code.
unallocated parts of the last page when the file ended on a frag
but not a page boundary.
Delimitted by tags PRE_MATT_MMAP_EOF and POST_MATT_MMAP_EOF,
in files alpha/alpha/pmap.c i386/i386/pmap.c nfs/nfs_bio.c vm/pmap.h
vm/vm_page.c vm/vm_page.h vm/vnode_pager.c miscfs/specfs/spec_vnops.c
ufs/ufs/ufs_readwrite.c kern/vfs_bio.c
Submitted by: Matt Dillon <dillon@freebsd.org>
Reviewed by: Alan Cox <alc@freebsd.org>
in my tree for 12+ months, and I just noticed that NetBSD have (I think,
I've just seen the commit, not the change) just zapped it there.
It wasn't in the options files or LINT either.
include of <sys/queue.h> in the !KERNEL case. The prerequisites
for <ufs/ufs/quota.h> were broken in Lite2 by converting some of
the kernel declarations to use queue macros without including
<sys/queue.h>. <sys/queue.h> was included in applications in
/usr/src instead. We polluted this file instead of merging the
changes in the applications.
Include <sys/queue.h> in the KERNEL case, and forward-declare all
structs that are used in prototypes, so that this file is almost
self-sufficient even in the kernel.
Obtained from: mostly from NetBSD
so that non-sloppy applications can call it without using disgusting
casts to avoid warnings. The 4th arg is sort of varargs -- it must
sometimes represent a filename, sometimes a struct pointer, and is
sometimes unused. The arg type is still caddr_t in the kernel.
Obtained from: mostly from NetBSD
lives in ext2_vnops.c for ext2fs. Also remove cast from comparision.
Bruce pointed out that it was bogus since we'd force a signed
comparision when we really wanted an unsigned comparison.
to write all the dirty blocks. If some of those blocks have dependencies,
they will be remarked dirty when the I/O completes. On systems with
really fast I/O systems, it is possible to get in an infinite loop trying
to flush the buffers, because the I/O finishes before we can get all the
dirty buffers off the v_dirtyblkhd list and into the I/O queue. (The
previous algorithm looped over the v_dirtyblkhd list writing out buffers
until the list emptied.) So, now we mark each buffer that we try to
write so that we can distinguish the ones that are being remarked dirty
from those that we have not yet tried to flush. Once we have tried to
push every buffer once, we then push any associated metadata that is
causing the remaining buffers to be redirtied.
Submitted by: Matthew Dillon <dillon@apollo.backplane.com>
Specifically, the test was in the wrong place, lacked a cast, didn't
unlock the node, and exited to bad rather than abortit. Now we don't
allow renaming of a file with LINK_MAX references. Move the test to
earlier in the code as it is closer to where ip is obtained, as that
is the style of the rest of the function.
Didn't fix the problems bruce pointed out in the rename man page to
include EMLINK, nor address his complaints about how the whole idea of
incrementing the link count during a rename is potentially asking for
trouble.
Also didn't try to correct potential problem Terry pointed out with
decrements not being similarly protected against underflow.
turns out to not be useful to unwind the dependencies and continue in
the face of a fatal error.
Also changed the log() to a printf() in softdep_error() so that it will
be output in the case of a impending panic.
Submitted by: Kirk McKusick <mckusick@mckusick.com>
changes to the VM system to support the new swapper, VM bug
fixes, several VM optimizations, and some additional revamping of the
VM code. The specific bug fixes will be documented with additional
forced commits. This commit is somewhat rough in regards to code
cleanup issues.
Reviewed by: "John S. Dyson" <root@dyson.iquest.net>, "David Greenman" <dg@root.com>
MNT_WAIT when we mean boolean `true' or check for that value not being
passed. There was no problem in practice because MNT_WAIT had the
magic value of 1.
I/O requests must be marked P_SYSTEM because if it isn't and the system
decides to swap it or (god forbid) kill it, the system stands a good
chance of locking up.
may be revoked, so vnop routines must be careful about accessing
the vnode if they may have blocked.
Fixed marking for update after successfully reading or writing 0
bytes. In this case, POSIX.1 specifies marking if and only if the
requested count is nonzero, but rev.1.86 never marked.
basically do a on-the-fly defragmentation of the FFS filesystem, changing
file block allocations to make them contiguous. Thanks to Kirk McKusick
for providing hints on what needed to be done to get this working.
They checked for the magic major number for the "device" behind mfs
mount points. Use a more obvious check for this device.
Debugged by: Andrew Gallatin <gallatin@cs.duke.edu>
when bdevsw[] became sparse. We still depend on magic to avoid having to
check that (v_rdev) device numbers in vnodes are not NODEV.
Removed redundant `major(dev) < nblkdev' tests instead of updating them.
partition that the label ioctl is being done on just because it has
offset 0, since there is no guarantee that such a partition is large
enough to contain the label. Don't use the wrong raw partition (0
instead of RAW_PART).
This fixes problems rewriting bizarre labels (with a nonzero offset
for the 'a' partition) in newfs(8). Such labels shouldn't normally
be used, but creating them was allowed if the ioctl was done on the
raw partition, and sysinstall creates them if the root partition isn't
allocated first.
Note that allowing write access to a partition other than the one that
has been checked for write access doesn't increase security holes
significantly, since write access to any partition already allows
changing the in-core label.
This fix should be in 3.0R. Rev.1.26 of newfs/newfs.c shouldn't be
in 3.0R.
but when i_effnlink was added to support soft updates, there was only
room for 4 spares. The number of spares was not reduced, so the inode
size became 260 (on i386's), or 512 after rounding up by malloc().
Use one spare field in `struct dinode' instead of the 5th spare field
in the inode and reduced to 4 spares in the inode so that the size is
256 again.
Changed the types of the spares in the inode from int to u_int32_t
so that the inode size has more chance of being <= 256 under other
arches, and downdated ext2fs to match (it was broken to use ints
before rev.1.1).
As a side effect, a few wakeup() calls are added, which might fix some of the
missing vm_page wakeups people have been seeing.
Reviewed by: Doug Rabson <dfr@nlsystems.com>
The problem is caused when a directory block is compacted. When this
occurs, softdep_change_directoryentry_offset() is called to relocate each
directory entry and adjust its matching diradd structure, if any, to match
the new location of the entry. The bug is that while
softdep_change_directoryentry_offset() correctly adjusts the offsets of
the diradd structures on the pd_diraddhd[] lists (which are not yet ready
to be committed to disk), it fails to adjust the offsets of the diradd
structures on the pd_pendinghd list (which are ready to be committed to
disk). This causes the dependency structures to be inconsistent with
the buf contents. Now, if the compaction has moved a directory entry to
the same offset as one of the diradd structures on the pd_pendinghd list
*and* a syscall is done that tries to remove this directory entry before
this directory block has been written to disk (which would empty
pd_pendinghd), a sanity check in newdirrem() will call panic() when it
notices that the inode number in the entry that it is to be removed doesn't
match the inode number in the diradd structure with that offset of that
entry.
Reviewed by: Kirk McKusick <mckusick@McKusick.COM>
Submitted by: Don Lewis <Don.Lewis@tsc.tdk.com>
happen when an NFS exported filesystem tries to remove a locally
mounted on directory.
PR: kern/7272
Submitted by: Andre Albsmeier <andre.albsmeier@mchp.siemens.de>
- don't set the clean flag on unmount of an unclean filesystem that was
(forcibly) mounted rw.
- set the clean flag on rw -> ro update of a mounted initially-clean
filesystem.
- fixed some style bugs (mostly long lines).
This uses the fs_flags field and FS_UNCLEAN state bit which were
introduced in the softdep changes. NetBSD uses extra state bits in
fs_clean.
Reviewed by: luoqui
Submitted by: Kirk McKusick <mckusick@McKusick.COM>
Two minor changes are also included,
1. Remove gratuitious checks for error return from vn_lock with LK_RETRY set,
vn_lock should always succeed in these cases.
2. Back out change rev. 1.36->1.37, which unnecessarily makes async mount
a little more unstable. It also keeps us in sync with other BSDs.
Suggested by: Bruce Evans <bde@zeta.org.au>
references to them.
The change a couple of days ago to ignore these numbers in statically
configured vfsconf structs was slightly premature because the cd9660,
cfs, devfs, ext2fs, nfs vfs's still used MOUNT_* instead of the number
in their vfsconf struct.
device drivers about sectors no longer in use.
Device-drivers receive the call through d_strategy, if they have
D_CANFREE in d_flags.
This allows flash based devices to erase the sectors and avoid
pointlessly carrying them around in compactions.
Reviewed by: Kirk Mckusick, bde
Sponsored by: M-Systems (www.m-sys.com)
clustering is obsolescent technology so hardly anyone noticed. On
a DORS 32160 SCSI drive with 4 tags, read clustering makes very
little difference even for huge sequential reads. However, on a
ZIP SCSI drive with 0 tags, the minimum overhead per block is about
40 msec, so very large clusters must be used to get anywhere near
the maximum transfer rate. Using clusters consisting of 1 8K block
reduces the transfer rate to about 250K/sec. Under msdosfs, missing
read clustering is normal and a cluster size of 1 512 byte block
reduces the transfer rate to about 25K/sec.
Broken in: rev.1.18
formats and args to match. Fixed old printf format errors (all related;
most were hidden by calling printf indirectly).
This change somehow avoids compiler bugs for 64-bit longs on i386's,
although it increases the number of 64-bit calculations.
respectively. Most of the longs should probably have been
u_longs, but this changes is just to prevent warnings about
casts between pointers and integers of different sizes, not
to fix poorly chosen types.
as the value in b_vp is often not really what you want.
(and needs to be frobbed). more cleanups will follow this.
Reviewed by: Bruce Evans <bde@freebsd.org>
as possible (when the inode is reclaimed). Temporarily only do
this if option UFS_LAZYMOD configured and softupdates aren't enabled.
UFS_LAZYMOD is intentionally left out of /sys/conf/options.
This is mainly to avoid almost useless disk i/o on battery powered
machines. It's silly to write to disk (on the next sync or when the
inode becomes inactive) just because someone hit a key or something
wrote to the screen or /dev/null.
PR: 5577
Previous version reviewed by: phk
in ufs_setattr() so that there is no need to pass timestamps to
UFS_UPDATE() (everything else just needs the current time). Ignore
the passed-in timestamps in UFS_UPDATE() and always call ufs_itimes()
(was: itimes()) to do the update. The timestamps are still passed
so that all the callers don't need to be changed yet.
that had an inode that has not yet been written to disk, when the inode of the
new file is also not yet written to disk, and your old directory entry is not
yet on disk but you need to remove it and the new name exists in memory
but has been deleted but the transaction to write the deleted name to disk
exists and has not yet been cancelled by the request to delete the non
existant name. I don't know how kirk could have missed such a glaring
problem for so long. :-) Especially since the inconsitency survived on
the disk for a whole 4 second on average before being fixed by other code.
This was not a crashing bug but just led to filesystem inconsitencies
if you crashed.
Submitted by: Kirk McKusick (mckusick@mckusick.com)
(doingdirectory && !newparent) case of ufs_rename().
rename("D1/X/", "D2/Y/") gives a wrong link count for D2.
Submitted by: Bruce Evans <bde@zeta.org.au>
Reviewed by: Kirk McKusick <mckusick@McKusick.COM>
is now ignored for special files, so that mounting root with option
noatime doesn't break reporting of idle times in programs like `w'.
The problem of execessive disk updates just to stamp atimes will be
handled for special files by only writing atimes to disk when inodes
become active. This works well because special files are relatively
uncommon and their atimes are even more disposable at panic time than
regular files' atimes.
1. mark atimes and mtimes of special files and fifos for update upon
successful completion of non-null i/o, not at the beginning of the
syscall.
2. never update file times for readonly filesystems. They were updated
for stats and closes but not for syncs. The updates were of course
only in-core and were thrown away when the inode was uncached, so
the times sometimes appeared to go backwards.
Improved comments in code related to (1) (mostly by removing them).
Unmacroized ITIMES(). The test in (2) bloated it even more. Don't
call getmicrotime() in the function version of it when we only need
the time in seconds.
---------
Make callers of namei() responsible for releasing references or locks
instead of having the underlying filesystems do it. This eliminates
redundancy in all terminal filesystems and makes it possible for stacked
transport layers such as umapfs or nullfs to operate correctly.
Quality testing was done with testvn, and lat_fs from the lmbench suite.
Some NFS client testing courtesy of Patrik Kudo.
vop_mknod and vop_symlink still release the returned vpp. vop_rename
still releases 4 vnode arguments before it returns. These remaining cases
will be corrected in the next set of patches.
---------
Submitted by: Michael Hancock <michaelh@cet.co.jp>
Reverse the VFS_VRELE patch. Reference counting of vnodes does not need
to be done per-fs. I noticed this while fixing vfs layering violations.
Doing reference counting in generic code is also the preference cited by
John Heidemann in recent discussions with him.
The implementation of alternative vnode management per-fs is still a valid
requirement for some filesystems but will be revisited sometime later,
most likely using a different framework.
Submitted by: Michael Hancock <michaelh@cet.co.jp>
This code will be turned on with the TWO options
DEVFS and SLICE. (see LINT)
Two labels PRE_DEVFS_SLICE and POST_DEVFS_SLICE will deliniate these changes.
/dev will be automatically mounted by init (thanks phk)
on bootup. See /sys/dev/slice/slice.4 for more info.
All code should act the same without these options enabled.
Mike Smith, Poul Henning Kamp, Soeren, and a few dozen others
This code does not support the following:
bad144 handling.
Persistance. (My head is still hurting from the last time we discussed this)
ATAPI flopies are not handled by the SLICE code yet.
When this code is running, all major numbers are arbitrary and COULD
be dynamically assigned. (this is not done, for POLA only)
Minor numbers for disk slices ARE arbitray and dynamically assigned.
- restored async mount support. The first entry in a block is still
always written synchronously, although it probably shouldn't be in
the async case.
- restored use of BWRITE() instead of bowrite() for the DOWHITEOUT
case, although bowrite() is probably better.
Broken by: merge of softdep changes (rev.1.22).
Found by: lmbench2 delete-file benchmarks.
(because we can :-). This means you can open a link file (or pseudo-file
in the case of short links where the data is stored in the inode rather
than disk blocks) and read the contents.
However, trap any writes from the user as it's difficult to do the right
thing in all cases. A link may be short and the user may be trying to
extend it beyond the limit and so on. Although.. being able to re-target
a symlink without deleting it first might have been nice.
This stuff is a bit perverse since symlink() and readlink() calls can
end up actually being implemented as read/write vnode ops.
Reviewed by: phk
* Figure out UTC relative to boottime. Four new functions provide
time relative to boottime.
* move "runtime" into struct proc. This helps fix the calcru()
problem in SMP.
* kill mono_time.
* add timespec{add|sub|cmp} macros to time.h. (XXX: These may change!)
* nanosleep, select & poll takes long sleeps one day at a time
Reviewed by: bde
Tested by: ache and others
"time" wasn't a atomic variable, so splfoo() protection were needed
around any access to it, unless you just wanted the seconds part.
Most uses of time.tv_sec now uses the new variable time_second instead.
gettime() changed to getmicrotime(0.
Remove a couple of unneeded splfoo() protections, the new getmicrotime()
is atomic, (until Bruce sets a breakpoint in it).
A couple of places needed random data, so use read_random() instead
of mucking about with time which isn't random.
Add a new nfs_curusec() function.
Mark a couple of bogosities involving the now disappeard time variable.
Update ffs_update() to avoid the weird "== &time" checks, by fixing the
one remaining call that passwd &time as args.
Change profiling in ncr.c to use ticks instead of time. Resolution is
the same.
Add new function "tvtohz()" to avoid the bogus "splfoo(), add time, call
hzto() which subtracts time" sequences.
Reviewed by: bde
softdep mode could only be activated on the initial mount of a filesystem
and then only if it was a read-write mount. A 'mount -r' (as done in the
rootfs mount) followed by a 'mount -u' to convert to read-write didn't
start softdep mode.
They are atomic, but return in essence what is in the "time" variable.
gettime() is now a macro front for getmicrotime().
Various patches to use the two new functions instead of the various
hacks used in their absence.
Some puntuation and grammer patches from Bruce.
A couple of XXX comments.
problems. Tor Egge and others have helped with various VM bugs
lately, but don't blame him -- blame me!!!
pmap.c:
1) Create an object for kernel page table allocations. This
fixes a bogus allocation method previously used for such, by
grabbing pages from the kernel object, using bogus pindexes.
(This was a code cleanup, and perhaps a minor system stability
issue.)
pmap.c:
2) Pre-set the modify and accessed bits when prudent. This will
decrease bus traffic under certain circumstances.
vfs_bio.c, vfs_cluster.c:
3) Rather than calculating the beginning virtual byte offset
multiple times, stick the offset into the buffer header, so
that the calculated offset can be reused. (Long long multiplies
are often expensive, and this is a probably unmeasurable performance
improvement, and code cleanup.)
vfs_bio.c:
4) Handle write recursion more intelligently (but not perfectly) so
that it is less likely to cause a system panic, and is also
much more robust.
vfs_bio.c:
5) getblk incorrectly wrote out blocks that are incorrectly sized.
The problem is fixed, and writes blocks out ONLY when B_DELWRI
is true.
vfs_bio.c:
6) Check that already constituted buffers have fully valid pages. If
not, then make sure that the B_CACHE bit is not set. (This was
a major source of Sig-11 type problems.)
vfs_bio.c:
7) Fix a potential system deadlock due to an incorrectly specified
sleep priority while waiting for a buffer write operation. The
change that I made opens the system up to serious problems, and
we need to examine the issue of process sleep priorities.
vfs_cluster.c, vfs_bio.c:
8) Make clustered reads work more correctly (and more completely)
when buffers are already constituted, but not fully valid.
(This was another system reliability issue.)
vfs_subr.c, ffs_inode.c:
9) Create a vtruncbuf function, which is used by filesystems that
can truncate files. The vinvalbuf forced a file sync type operation,
while vtruncbuf only invalidates the buffers past the new end of file,
and also invalidates the appropriate pages. (This was a system reliabiliy
and performance issue.)
10) Modify FFS to use vtruncbuf.
vm_object.c:
11) Make the object rundown mechanism for OBJT_VNODE type objects work
more correctly. Included in that fix, create pager entries for
the OBJT_DEAD pager type, so that paging requests that might slip
in during race conditions are properly handled. (This was a system
reliability issue.)
vm_page.c:
12) Make some of the page validation routines be a little less picky
about arguments passed to them. Also, support page invalidation
change the object generation count so that we handle generation
counts a little more robustly.
vm_pageout.c:
13) Further reduce pageout daemon activity when the system doesn't
need help from it. There should be no additional performance
decrease even when the pageout daemon is running. (This was
a significant performance issue.)
vnode_pager.c:
14) Teach the vnode pager to handle race conditions during vnode
deallocations.
