proc or any VM system structure will have to be rebuilt!!!
Much needed overhaul of the VM system. Included in this first round of
changes:
1) Improved pager interfaces: init, alloc, dealloc, getpages, putpages,
haspage, and sync operations are supported. The haspage interface now
provides information about clusterability. All pager routines now take
struct vm_object's instead of "pagers".
2) Improved data structures. In the previous paradigm, there is constant
confusion caused by pagers being both a data structure ("allocate a
pager") and a collection of routines. The idea of a pager structure has
escentially been eliminated. Objects now have types, and this type is
used to index the appropriate pager. In most cases, items in the pager
structure were duplicated in the object data structure and thus were
unnecessary. In the few cases that remained, a un_pager structure union
was created in the object to contain these items.
3) Because of the cleanup of #1 & #2, a lot of unnecessary layering can now
be removed. For instance, vm_object_enter(), vm_object_lookup(),
vm_object_remove(), and the associated object hash list were some of the
things that were removed.
4) simple_lock's removed. Discussion with several people reveals that the
SMP locking primitives used in the VM system aren't likely the mechanism
that we'll be adopting. Even if it were, the locking that was in the code
was very inadequate and would have to be mostly re-done anyway. The
locking in a uni-processor kernel was a no-op but went a long way toward
making the code difficult to read and debug.
5) Places that attempted to kludge-up the fact that we don't have kernel
thread support have been fixed to reflect the reality that we are really
dealing with processes, not threads. The VM system didn't have complete
thread support, so the comments and mis-named routines were just wrong.
We now use tsleep and wakeup directly in the lock routines, for instance.
6) Where appropriate, the pagers have been improved, especially in the
pager_alloc routines. Most of the pager_allocs have been rewritten and
are now faster and easier to maintain.
7) The pagedaemon pageout clustering algorithm has been rewritten and
now tries harder to output an even number of pages before and after
the requested page. This is sort of the reverse of the ideal pagein
algorithm and should provide better overall performance.
8) Unnecessary (incorrect) casts to caddr_t in calls to tsleep & wakeup
have been removed. Some other unnecessary casts have also been removed.
9) Some almost useless debugging code removed.
10) Terminology of shadow objects vs. backing objects straightened out.
The fact that the vm_object data structure escentially had this
backwards really confused things. The use of "shadow" and "backing
object" throughout the code is now internally consistent and correct
in the Mach terminology.
11) Several minor bug fixes, including one in the vm daemon that caused
0 RSS objects to not get purged as intended.
12) A "default pager" has now been created which cleans up the transition
of objects to the "swap" type. The previous checks throughout the code
for swp->pg_data != NULL were really ugly. This change also provides
the rudiments for future backing of "anonymous" memory by something
other than the swap pager (via the vnode pager, for example), and it
allows the decision about which of these pagers to use to be made
dynamically (although will need some additional decision code to do
this, of course).
13) (dyson) MAP_COPY has been deprecated and the corresponding "copy
object" code has been removed. MAP_COPY was undocumented and non-
standard. It was furthermore broken in several ways which caused its
behavior to degrade to MAP_PRIVATE. Binaries that use MAP_COPY will
continue to work correctly, but via the slightly different semantics
of MAP_PRIVATE.
14) (dyson) Sharing maps have been removed. It's marginal usefulness in a
threads design can be worked around in other ways. Both #12 and #13
were done to simplify the code and improve readability and maintain-
ability. (As were most all of these changes)
TODO:
1) Rewrite most of the vnode pager to use VOP_GETPAGES/PUTPAGES. Doing
this will reduce the vnode pager to a mere fraction of its current size.
2) Rewrite vm_fault and the swap/vnode pagers to use the clustering
information provided by the new haspage pager interface. This will
substantially reduce the overhead by eliminating a large number of
VOP_BMAP() calls. The VOP_BMAP() filesystem interface should be
improved to provide both a "behind" and "ahead" indication of
contiguousness.
3) Implement the extended features of pager_haspage in swap_pager_haspage().
It currently just says 0 pages ahead/behind.
4) Re-implement the swap device (swstrategy) in a more elegant way, perhaps
via a much more general mechanism that could also be used for disk
striping of regular filesystems.
5) Do something to improve the architecture of vm_object_collapse(). The
fact that it makes calls into the swap pager and knows too much about
how the swap pager operates really bothers me. It also doesn't allow
for collapsing of non-swap pager objects ("unnamed" objects backed by
other pagers).
1) Files weren't properly synced on filesystems other than UFS. In some
cases, this lead to lost data. Most likely would be noticed on NFS.
The fix is to make the VM page sync/object_clean general rather than
in each filesystem.
2) Mixing regular and mmaped file I/O on NFS was very broken. It caused
chunks of files to end up as zeroes rather than the intended contents.
The fix was to fix several race conditions and to kludge up the
"b_dirtyoff" and "b_dirtyend" that NFS relies upon - paying attention
to page modifications that occurred via the mmapping.
Reviewed by: David Greenman
Submitted by: John Dyson
space for the hash list buckets and is a little faster. The features
of tailq aren't needed. Increased the size of the object hash table
to improve performance. In the future, this will be changed so that
the table is sized dynamically.
pages that are in FS buffers. This fixes the (believed to already have been
fixed) problem with msync() not doing it's job...in other words, the
stuff that Andrew has continuously been complaining about.
Submitted by: John Dyson, w/minor changes by me.
Fixed remaining known bugs in the buffer IO and VM system.
vfs_bio.c:
Fixed some race conditions and locking bugs. Improved performance
by removing some (now) unnecessary code and fixing some broken
logic.
Fixed process accounting of # of FS outputs.
Properly handle NFS interrupts (B_EINTR).
(various)
Replaced calls to clrbuf() with calls to an optimized routine
called vfs_bio_clrbuf().
(various FS sync)
Sync out modified vnode_pager backed pages.
ffs_vnops.c:
Do two passes: Sync out file data first, then indirect blocks.
vm_fault.c:
Fixed deadly embrace caused by acquiring locks in the wrong order.
vnode_pager.c:
Changed to use buffer I/O system for writing out modified pages. This
should fix the problem with the modification date previous not getting
updated. Also dramatically simplifies the code. Note that this is
going to change in the future and be implemented via VOP_PUTPAGES().
vm_object.c:
Fixed a pile of bugs related to cleaning (vnode) objects. The performance
of vm_object_page_clean() is terrible when dealing with huge objects,
but this will change when we implement a binary tree to keep the object
pages sorted.
vm_pageout.c:
Fixed broken clustering of pageouts. Fixed race conditions and other
lockup style bugs in the scanning of pages. Improved performance.
to accurately track this. It isn't an indicator of resource consumption
anyway.
Removed cnt.v_kernel_pages: We don't implement this and doing so accurately
would be very difficult (and ambiguous - since process pages are often
double mapped in the kernel and the process address spaces).
VTEXT not always getting cleared when it is supposed to. Added check to
make sure that vm_object_remove() isn't called with a NULL pager or for
a pager for an OBJ_INTERNAL object (neither of which will be on the hash
list). Clear OBJ_CANPERSIST if we decide to terminate it because of no
resident pages.
New functions create - vm_object_pip_wakeup and pagedaemon_wakeup that
are used to reduce the actual number of wakeups.
New function vm_page_protect which is used in conjuction with some new
page flags to reduce the number of calls to pmap_page_protect.
Minor changes to reduce unnecessary spl nesting.
Rewrote vm_page_alloc() to improve readability.
Various other mostly cosmetic changes.
pager(). Almost completely rewrote vm_mmap(); when John gets done with
the bottom half, it will be a complete rewrite. Deprecated most use of
vm_object_setpager(). Removed side effect of setting object persist
in vm_object_enter and moved this into the pager(s). A few other
cosmetic changes.
Slight change to reverse collapsing so that vm_object_deallocate doesn't
have to be called recursively.
Removed half of a previous fix - the renamed page during a collapse doesn't
need to be marked dirty because the pager backing store pointers are copied
- thus preserving the page's data. This assumes that pages without backing
store are always dirty (except perhaps for when they are first zeroed, but
this doesn't matter).
Switch order of two lines of code so that the correct pager is removed
from the hash list. The previous code bogusly passed a NULL pointer to
vm_object_remove(). The call to vm_object_remove() should be unnecessary
if named anonymous objects were being dealt with correctly. They are
currently marked as OBJ_INTERNAL, which really screws up things (such as
this).
2) bump reference counts by 2 instead of 1 so that an object deallocate
doesn't try to recursively collapse the object.
3) mark pages renamed during the collapse as dirty so that their contents
are preserved.
Submitted by: John and me.
Fixed long standing bug in freeing swap space during object collapses.
Fixed 'out of space' messages from printing out too often.
Modified to use new kmem_malloc() calling convention.
Implemented an additional stat in the swap pager struct to count the
amount of space allocated to that pager. This may be removed at some
point in the future.
Minimized unnecessary wakeups.
vm_fault.c:
Don't try to collect fault stats on 'swapped' processes - there aren't
any upages to store the stats in.
Changed read-ahead policy (again!).
vm_glue.c:
Be sure to gain a reference to the process's map before swapping.
Be sure to lose it when done.
kern_malloc.c:
Added the ability to specify if allocations are at interrupt time or
are 'safe'; this affects what types of pages can be allocated.
vm_map.c:
Fixed a variety of map lock problems; there's still a lurking bug that
will eventually bite.
vm_object.c:
Explicitly initialize the object fields rather than bzeroing the struct.
Eliminated the 'rcollapse' code and folded it's functionality into the
"real" collapse routine.
Moved an object_unlock() so that the backing_object is protected in
the qcollapse routine.
Make sure nobody fools with the backing_object when we're destroying it.
Added some diagnostic code which can be called from the debugger that
looks through all the internal objects and makes certain that they
all belong to someone.
vm_page.c:
Fixed a rather serious logic bug that would result in random system
crashes. Changed pagedaemon wakeup policy (again!).
vm_pageout.c:
Removed unnecessary page rotations on the inactive queue.
Changed the number of pages to explicitly free to just free_reserved
level.
Submitted by: John Dyson
There is similar bogusness in the pageout daemon that will be fixed soon.
This fixes a panic pointed out to me by Bruce Evans that occurs when
/dev/mem is used to map managed memory.
Added hook for pmap_prefault() and use symbolic constant for new third
argument to vm_page_alloc() (vm_fault.c, various)
Changed the way that upages and page tables are held. (vm_glue.c)
Fixed architectural flaw in allocating pages at interrupt time that was
introduced with the merged cache changes. (vm_page.c, various)
Adjusted some algorithms to acheive better paging performance and to
accomodate the fix for the architectural flaw mentioned above. (vm_pageout.c)
Fixed pbuf handling problem, changed policy on handling read-behind page.
(vnode_pager.c)
Submitted by: John Dyson
need to be moved into the qcollapse and rcollapse routines, but I don't
have time at the moment to make all the required changes...this will do
for now.
being cleared in some cases for vnode backed objects; we now do this in
vnode_pager_alloc proper to guarantee it. Also be more careful in the
rcollapse code about messing with busy/bmapped pages.
much higher filesystem I/O performance, and much better paging performance. It
represents the culmination of over 6 months of R&D.
The majority of the merged VM/cache work is by John Dyson.
The following highlights the most significant changes. Additionally, there are
(mostly minor) changes to the various filesystem modules (nfs, msdosfs, etc) to
support the new VM/buffer scheme.
vfs_bio.c:
Significant rewrite of most of vfs_bio to support the merged VM buffer cache
scheme. The scheme is almost fully compatible with the old filesystem
interface. Significant improvement in the number of opportunities for write
clustering.
vfs_cluster.c, vfs_subr.c
Upgrade and performance enhancements in vfs layer code to support merged
VM/buffer cache. Fixup of vfs_cluster to eliminate the bogus pagemove stuff.
vm_object.c:
Yet more improvements in the collapse code. Elimination of some windows that
can cause list corruption.
vm_pageout.c:
Fixed it, it really works better now. Somehow in 2.0, some "enhancements"
broke the code. This code has been reworked from the ground-up.
vm_fault.c, vm_page.c, pmap.c, vm_object.c
Support for small-block filesystems with merged VM/buffer cache scheme.
pmap.c vm_map.c
Dynamic kernel VM size, now we dont have to pre-allocate excessive numbers of
kernel PTs.
vm_glue.c
Much simpler and more effective swapping code. No more gratuitous swapping.
proc.h
Fixed the problem that the p_lock flag was not being cleared on a fork.
swap_pager.c, vnode_pager.c
Removal of old vfs_bio cruft to support the past pseudo-coherency. Now the
code doesn't need it anymore.
machdep.c
Changes to better support the parameter values for the merged VM/buffer cache
scheme.
machdep.c, kern_exec.c, vm_glue.c
Implemented a seperate submap for temporary exec string space and another one
to contain process upages. This eliminates all map fragmentation problems
that previously existed.
ffs_inode.c, ufs_inode.c, ufs_readwrite.c
Changes for merged VM/buffer cache. Add "bypass" support for sneaking in on
busy buffers.
Submitted by: John Dyson and David Greenman
1. The pageout daemon used to block under certain
circumstances, and we needed to add new functionality
that would cause the pageout daemon to block more often.
Now, the pageout daemon mostly just gets rid of pages
and kills processes when the system is out of swap.
The swapping, rss limiting and object cache trimming
have been folded into a new daemon called "vmdaemon".
This new daemon does things that need to be done for
the VM system, but can block. For example, if the
vmdaemon blocks for memory, the pageout daemon
can take care of it. If the pageout daemon had
blocked for memory, it was difficult to handle
the situation correctly (and in some cases, was
impossible).
2. The collapse problem has now been entirely fixed.
It now appears to be impossible to accumulate unnecessary
vm objects. The object collapsing now occurs when ref counts
drop to one (where it is more likely to be more simple anyway
because less pages would be out on disk.) The original
fixes were incomplete in that pathological circumstances
could still be contrived to cause uncontrolled growth
of swap. Also, the old code still, under steady state
conditions, used more swap space than necessary. When
using the new code, users will generally notice a
significant decrease in swap space usage, and theoretically,
the system should be leaving fewer unused pages around
competing for memory.
Submitted by: John Dyson
Enabled via REL2_1.
Added support for doing object collapses "on the fly". Enabled via REL2_1a.
Improved object collapses so that they can happen in more cases. Improved
sensing of modified pages to fix an apparant race condition and improve
clustered pageout opportunities. Fixed an "oops" with not restarting page
scan after a potential block in vm_pageout_clean() (not doing this can result
in strange behavior in some cases).
Submitted by: John Dyson & David Greenman
