blocks of zeros could wind up in a file written to over NFS by a client.
The problem only occurs a few times per several gigabytes of data. This
problem turned out to be bug #3 below.
bug #1:
B_CLUSTEROK must be cleared when an NFS buffer is reverted from
stage 2 (ready for commit rpc) to stage 1 (ready for write).
Reversions can occur when a dirty NFS buffer is redirtied with new
data.
Otherwise the VFS/BIO system may end up thinking that a stage 1
NFS buffer is clusterable. Stage 1 NFS buffers are not clusterable.
bug #2:
B_CLUSTEROK was inappropriately set for a 'short' NFS buffer (short
buffers only occur near the EOF of the file). Change to only set
when the buffer is a full biosize (usually 8K). This bug has no
effect but should be fixed in -current anyway. It need not be
backported.
bug #3:
B_NEEDCOMMIT was inappropriately set in nfs_flush() (which is
typically only called by the update daemon). nfs_flush()
does a multi-pass loop but due to the lack of vnode locking it
is possible for new buffers to be added to the dirtyblkhd list
while a flush operation is going on. This may result in nfs_flush()
setting B_NEEDCOMMIT on a buffer which has *NOT* yet gone through its
stage 1 write, causing only the commit rpc to be made and thus
causing the contents of the buffer to be thrown away (never sent to
the server).
The patch also contains some cleanup, which only applies to the commit
into -current.
Reviewed by: dg, julian
Originally Reported by: Dan Nelson <dnelson@emsphone.com>
Merge the contents (less some trivial bordering the silly comments)
of <vm/vm_prot.h> and <vm/vm_inherit.h> into <vm/vm.h>. This puts
the #defines for the vm_inherit_t and vm_prot_t types next to their
typedefs.
This paves the road for the commit to follow shortly: change
useracc() to use VM_PROT_{READ|WRITE} rather than B_{READ|WRITE}
as argument.
previously issued synchronously even if async daemons (nfsiod's) were
available. The commit has been moved from the strategy code to the doio
code in order to asynchronize it.
Removed use of lastr in preparation for removal of vnode->v_lastr. It
has been replaced with seqcount, which is already supported by the system
and, in fact, gives us a better heuristic for sequential detection then
lastr ever did.
Made major performance improvements to the server side commit. The
server previously fsync'd the entire file for each commit rpc. The
server now bawrite()s only those buffers related to the offset/size
specified in the commit rpc.
Note that we do not commit the meta-data yet. This works still needs
to be done.
Note that a further optimization can be done (and has not yet been done)
on the client: we can merge multiple potential commit rpc's into a
single rpc with a greater file offset/size range and greatly reduce
rpc traffic.
Reviewed by: Alan Cox <alc@cs.rice.edu>, David Greenman <dg@root.com>
lockmgr locks. This commit should be functionally equivalent to the old
semantics. That is, all buffer locking is done with LK_EXCLUSIVE
requests. Changes to take advantage of LK_SHARED and LK_RECURSIVE will
be done in future commits.
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>
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>
including alan, john, me, luoqi, and kirk
Submitted by: Matt Dillon <dillon@frebsd.org>
This change implements a relatively sophisticated fix to getnewbuf().
There were two problems with getnewbuf(). First, the writerecursion
can lead to a system stack overflow when you have NFS and/or VN
devices in the system. Second, the free/dirty buffer accounting was
completely broken. Not only did the nfs routines blow it trying to
manually account for the buffer state, but the accounting that was
done did not work well with the purpose of their existance: figuring
out when getnewbuf() needs to sleep.
The meat of the change is to kern/vfs_bio.c. The remaining diffs are
all minor except for NFS, which includes both the fixes for bp
interaction AND fixes for a 'biodone(): buffer already done' lockup.
Sys/buf.h also contains a chaining structure which is not used by
this patchset but is used by other patches that are coming soon.
This patch deliniated by tags PRE_MAT_GETBUF and POST_MAT_GETBUF.
(sorry for the missing T matt)
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>
second argument. np_size is a 64 bit int, so is the second arg. This
might have caused needless 2G/4G file size problems.
I believe it was Bruce who queried this.
Pre-2.8 versions of gcc generate a call to __divdi3() for all 64-bit
signed divisions, but egcs optimizes them to a shift and fixup when
the divisor is a constant power of 2. Unfortunately, it generates
a call to __cmpdi2() for the fixup, although all except possibly
ancient versions of gcc and egcs do ordinary 64-bit comparisons
inline.
readrpc/writerpc, since they assume it's already been done. This could
break if the first read/write access to a nfs filesystem was an exec() or
mmap() instead of a read(), write() syscall. (or statfs()).
nfs_getpages() could return an errno (EOPNOTSUPP) instead of a VM_PAGER_*
return code. Some layout tweaks for the get/putpages code.
rather than assuming 2^64. It may not like files that big. :-)
On the nfs server, calculate and report the max file size as the point
that the block numbers in the cache would turn negative.
(ie: 1099511627775 bytes (1TB)).
One of the things I'm worried about however, is that directory offsets
are really cookies on a NFSv3 server and can be rather large, especially
when/if the server generates the opaque directory cookies by using a local
filesystem offset in what comes out as the upper 32 bits of the 64 bit
cookie. (a server is free to do this, it could save byte swapping
depending on the native 64 bit byte order)
Obtained from: NetBSD
We had run out of bits in the nfs mount flags, I have moved the internal
state flags into a seperate variable. These are no longer visible via
statfs(), but I don't know of anything that looks at them.
has been some bitrot and incorrect assumptions in the vfs_bio code. These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances. Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code. This code might have been committed seperately, but
almost everything is interrelated.
1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse. The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite. When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me. Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE. The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur. (use p->busy more for pageouts instead
of PG_BUSY.) Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy. Make the
page allocation code handle that case correctly. (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly. I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep. It didn't handle
consistancy problems very well, so make the design a little less
lofty. After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.
Make vfs_bio buffer mgmt work better.
Buffers were being used after brelse.
Make nfs_getpages work independently of other NFS
interfaces. This eliminates some difficult
recursion problems and decreases pagefault
overhead.
Remove an erroneous vfs_unbusy_pages.
Fix a reentrancy problem, with nfs_vinvalbuf when
vnode is already being rundown.
Reassignbuf wasn't being called when needed under
certain circumstances.
(Thanks to Bill Paul for help.)
original BSD code. The association between the vnode and the vm_object
no longer includes reference counts. The major difference is that
vm_object's are no longer freed gratuitiously from the vnode, and so
once an object is created for the vnode, it will last as long as the
vnode does.
When a vnode object reference count is incremented, then the underlying
vnode reference count is incremented also. The two "objects" are now
more intimately related, and so the interactions are now much less
complex.
When vnodes are now normally placed onto the free queue with an object still
attached. The rundown of the object happens at vnode rundown time, and
happens with exactly the same filesystem semantics of the original VFS
code. There is absolutely no need for vnode_pager_uncache and other
travesties like that anymore.
A side-effect of these changes is that SMP locking should be much simpler,
the I/O copyin/copyout optimizations work, NFS should be more ponderable,
and further work on layered filesystems should be less frustrating, because
of the totally coherent management of the vnode objects and vnodes.
Please be careful with your system while running this code, but I would
greatly appreciate feedback as soon a reasonably possible.
nfs_bio.c code worked better than the 1.44. This commit reverts
the important parts of 1.44 to 1.41, and we will fix it when we
can get a handle on the problem.
and b_validend. The changes to vfs_bio.c are a bit ugly but hopefully
can be tidied up later by a slight redesign.
PR: kern/2573, kern/2754, kern/3046 (possibly)
Reviewed by: dyson
wind up writing zeros instead of real data when the file is on an NFSv2
mounted directory.
While tracking this bug down, I noticed that nfs_asyncio was waking *all*
the iods when a block was written instead of just one per block. Fixing this
gives a 25% performance improvment for writes on v2 (less for v3).
Both are 2.2 candidates.
PR: kern/2774
Zero the b_dirty{off,end} after cluster-comitting a group of buffers.
With these fixes, I was able to complete a 'make world' with remote src
and obj directories.