clean and without writable mappings. if we try to flush dirty pages past
EOF to the server when NMODIFIED is clear, we'll update the attrcache before
doing the write, which will try to free the pages past EOF and deadlock.
to deal with this, we write-protect pages before we send them to the server,
and restrict ourselves to creating read-only mappings if NMODIFIED isn't set.
score another one for enami.
frank's scheme, with one new twist: don't wait until we've totally run
out of free pages before committing, but instead notice when we've built
up a largish range of uncommitted pages and commit only the older half of
the range, which is likely to already be on disk on the server.
- remove special treatment of pager_map mappings in pmaps. this is
required now, since I've removed the globals that expose the address range.
pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's
no longer any need to special-case it.
- eliminate struct uvm_vnode by moving its fields into struct vnode.
- rewrite the pageout path. the pager is now responsible for handling the
high-level requests instead of only getting control after a bunch of work
has already been done on its behalf. this will allow us to UBCify LFS,
which needs tighter control over its pages than other filesystems do.
writing a page to disk no longer requires making it read-only, which
allows us to write wired pages without causing all kinds of havoc.
- use a new PG_PAGEOUT flag to indicate that a page should be freed
on behalf of the pagedaemon when it's unlocked. this flag is very similar
to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the
pageout fails due to eg. an indirect-block buffer being locked.
this allows us to remove the "version" field from struct vm_page,
and together with shrinking "loan_count" from 32 bits to 16,
struct vm_page is now 4 bytes smaller.
- no longer use PG_RELEASED for swap-backed pages. if the page is busy
because it's being paged out, we can't release the swap slot to be
reallocated until that write is complete, but unlike with vnodes we
don't keep a count of in-progress writes so there's no good way to
know when the write is done. instead, when we need to free a busy
swap-backed page, just sleep until we can get it busy ourselves.
- implement a fast-path for extending writes which allows us to avoid
zeroing new pages. this substantially reduces cpu usage.
- encapsulate the data used by the genfs code in a struct genfs_node,
which must be the first element of the filesystem-specific vnode data
for filesystems which use genfs_{get,put}pages().
- eliminate many of the UVM pagerops, since they aren't needed anymore
now that the pager "put" operation is a higher-level operation.
- enhance the genfs code to allow NFS to use the genfs_{get,put}pages
instead of a modified copy.
- clean up struct vnode by removing all the fields that used to be used by
the vfs_cluster.c code (which we don't use anymore with UBC).
- remove kmem_object and mb_object since they were useless.
instead of allocating pages to these objects, we now just allocate
pages with no object. such pages are mapped in the kernel until they
are freed, so we can use the mapping to find the page to free it.
this allows us to remove splvm() protection in several places.
The sum of all these changes improves write throughput on my
decstation 5000/200 to within 1% of the rate of NetBSD 1.5
and reduces the elapsed time for "make release" of a NetBSD 1.5
source tree on my 128MB pc to 10% less than a 1.5 kernel took.
the mapping is:
VM_PAGER_OK 0
VM_PAGER_BAD <unused>
VM_PAGER_FAIL <unused>
VM_PAGER_PEND 0 (see below)
VM_PAGER_ERROR EIO
VM_PAGER_AGAIN EAGAIN
VM_PAGER_UNLOCK EBUSY
VM_PAGER_REFAULT ERESTART
for async i/o requests, it used to be possible for the request to
be convert to sync, and the pager would return VM_PAGER_OK or VM_PAGER_PEND
to indicate whether the caller should perform post-i/o cleanup.
this is no longer allowed; pagers must now return 0 to indicate that
the async i/o was successfully started, and the caller never needs to
worry about doing the post-i/o cleanup.
- in the cases where we skip over the i/o loop, increment npages by ridx
so that when the cleanup code starts processing the pgs array at index 0
it'll actually process all of the pages.
- process the PG_RELEASED flag when unbusying pages.
- add some missing MP locking.
- use MIN() and MAX() instead of min() and max() since the latter are
functions which take arguments of type "int" but we call them with
values of type "off_t", so the values could be truncated.
problem reported by msaitoh@netbsd.org. NOTE: These are marked XXXUBC
since the code that allocates the bufs is new with UBC, but it may be
the case that bp->b_proc needs to be intialized to curproc (it's used
in a call to nfs_sigintr()).
- fix math when skipping writing pages that just need a commit.
- clear the needcommit stuff and PG_RDONLY flags on pages returned for
overwrite requests as well as for normal write faults.
- bail out of nfs_write() if we get an error.
- remove a bogus attempt to clean up after failed uiomove()s.
- bring over a workaround for a lock-ordering problem from the genfs code.
- add some missing MP locking.
for them are actually done asynchronously. Idea taken from FreeBSD.
Do away with nfs_writebp completely, it's not needed anymore.
Keep an eye on the range of a file that needs to be committed, and
do it in heaps.
that required to support NFSv2 mounts. Not finished yet, but already
provides some 44k of saving in code size on arm26. More savings, and some
documentation, are still to come.
default, as the copyright on the main file (ffs_softdep.c) is such
that is has been put into gnusrc. options SOFTDEP will pull this
in. This code also contains the trickle syncer.
Bump version number to 1.4O
UVM was written by chuck cranor <chuck@maria.wustl.edu>, with some
minor portions derived from the old Mach code. i provided some help
getting swap and paging working, and other bug fixes/ideas. chuck
silvers <chuq@chuq.com> also provided some other fixes.
this is the rest of the MI portion changes.
this will be KNF'd shortly. :-)
whether we get it off the wire. An nfsiod might have been busy with
it, and finished while we were waiting for it in nfs_getcacheblk, so
we need to check for EOF again no matter what.
the directory cache as translation table. See nfs_subs.c for comments.
Makes the code a bit more complex to look at than I would have liked,
but doesn't affect the speed of the default behavior.
* Optimize caching behavior a bit when buffers are invalidated.
* Save some RPCs in readdir operations by not bothering if there is
a small amount left to do to fill the buffer. It'll be done in the
next RPC with a larger chunk anyway. Wastes a bit of buffer space
but is faster.
* Make n_vattr an allocated vattr struct. This avoids nfsnode bloat,
and is friendlier to the malloc routines.
directory cookie that may be thrown back at us from userspace, up
to a size limit. Fixes double entry problem.
* Split flags for internal and external use in the NFS mount structure.
* Fix some buffer structure fields that weren're being used correctly.
* Fix missing directory cache inval call in nfs_open.
* Limit on NFS_DIRBLKSIZ no longer needed, bumped to the more reasonable
value of 8k.
* Various other things that I forget, all related to the dir caching
somehow, though.
From Olaf Seibert <rhialto@polder.ubc.kun.nl> (PR 3687)
* Make an attempt to check the maximum filesize before attempting
a write to the server, as write RPCs will typically happen
asynchronously, and the process will not see the error.
Fixes problems with unexpectly truncated files at 4G
* Pass up errors in nfs_writerpc correctly
architectures), truncate them intelligently instead.
The truncation is done centralized in vnode_pager.c.
This prevents from wrap-over effects when parts of large (>2^32 byte) files
are mmapped.
Don't allow to mmap above the numerical range of vm_offset_t.
This is considered a temporary solution until the vm system handles the
object sizes/offsets more cleanly.
Improve the queuing algorithms used by NFS' asynchronous i/o. The
existing mechanism uses a global queue for some buffers and the
vp->b_dirtyblkhd queue for others. This turns sequential writes into
randomly ordered writes to the server, affecting both read and write
performance. The existing mechanism also copes badly with hung
servers, tending to block accesses to other servers when all the iods
are waiting for a hung server.
The new mechanism uses a queue for each mount point. All asynchronous
i/o goes through this queue which preserves the ordering of requests.
A simple mechanism ensures that the iods are shared out fairly between
active mount points.
Reviewed/integrated/approved by Frank van der Linden <fvdl@netbsd.org>
chs