and uncontrolled growth.
The key fix is from Dan Carasone, who noticed that buf_canfree() was
counting in _bytes_ but freeing in _buffers_, which caused the instant
drop to lowater observed by some users.
We now control the rate of growth; the probability of getting a new
allocation is inversely proportional to the current size of the
cache. This idea is from a long-ago conversation with Kirk McKusick
and, if memory serves, was used for the file-system cache in some
other BSD variant at some point in history.
With growth and shrinkage more or less dealt with, we return the
default maximum cache size to 15%. The default _minimum_ cache size
is raised from 1/16 of the maximum cache size to 1/8, since 1/16 was
chosen when the maximum size was 30% of memory.
Finally, after observing the behaviour of the pagedaemon and the
buffer cache drainer under pathological workloads (e.g. a benchmark
that steps through 75% of available memory backwards) I have moved
the call to buf_drain() to the beginning of the pagedaemon from the
end; if the pagedaemon bogs down, it still won't get run as often
as it should, but at least this way it will see the state of the
free count and free target _before_ the scan step does its thing.
- for in-kernel maps, disable map entry merging so that
unmap operations won't block. (workaround for PR/24039)
- for in-kernel maps, allocate kva for vm_map_entry from
the map itsself and eliminate MAX_KMAPENT and
uvm_map_entry_kmem_pool.
VOP_STRATEGY(bp) is replaced by one of two new functions:
- VOP_STRATEGY(vp, bp) Call the strategy routine of vp for bp.
- DEV_STRATEGY(bp) Call the d_strategy routine of bp->b_dev for bp.
DEV_STRATEGY(bp) is used only for block-to-block device situations.
This avoids problems with the kernel grovelling vmstat -u/-U when
using LOCKDEBUG, which changes the size of struct simplelock.
Replaced the original location of the simplelock with "int unused"
so that binary compatibility will be retained with old vmstat.
process context ('reaper').
From within the exiting process context:
* deactivate pmap and free vmspace while we can still block
* introduce MD cpu_lwp_free() - this cleans all MD-specific context (such
as FPU state), and is the last potentially blocking operation;
all of cpu_wait(), and most of cpu_exit(), is now folded into cpu_lwp_free()
* process is now immediatelly marked as zombie and made available for pickup
by parent; the remaining last lwp continues the exit as fully detached
* MI (rather than MD) code bumps uvmexp.swtch, cpu_exit() is now same
for both 'process' and 'lwp' exit
uvm_lwp_exit() is modified to never block; the u-area memory is now
always just linked to the list of available u-areas. Introduce (blocking)
uvm_uarea_drain(), which is called to release the excessive u-area memory;
this is called by parent within wait4(), or by pagedaemon on memory shortage.
uvm_uarea_free() is now private function within uvm_glue.c.
MD process/lwp exit code now always calls lwp_exit2() immediatelly after
switching away from the exiting lwp.
g/c now unneeded routines and variables, including the reaper kernel thread
virtual memory reservation and a private pool of memory pages -- by a scheme
based on memory pools.
This allows better utilization of memory because buffers can now be allocated
with a granularity finer than the system's native page size (useful for
filesystems with e.g. 1k or 2k fragment sizes). It also avoids fragmentation
of virtual to physical memory mappings (due to the former fixed virtual
address reservation) resulting in better utilization of MMU resources on some
platforms. Finally, the scheme is more flexible by allowing run-time decisions
on the amount of memory to be used for buffers.
On the other hand, the effectiveness of the LRU queue for buffer recycling
may be somewhat reduced compared to the traditional method since, due to the
nature of the pool based memory allocation, the actual least recently used
buffer may release its memory to a pool different from the one needed by a
newly allocated buffer. However, this effect will kick in only if the
system is under memory pressure.
uvm_km_valloc1(), and use it to express all of
uvm_km_valloc()
uvm_km_valloc_wait()
uvm_km_valloc_prefer()
uvm_km_valloc_prefer_wait()
uvm_km_valloc_align()
in terms of it by macro expansion.
Gone are the old kern_sysctl(), cpu_sysctl(), hw_sysctl(),
vfs_sysctl(), etc, routines, along with sysctl_int() et al. Now all
nodes are registered with the tree, and nodes can be added (or
removed) easily, and I/O to and from the tree is handled generically.
Since the nodes are registered with the tree, the mapping from name to
number (and back again) can now be discovered, instead of having to be
hard coded. Adding new nodes to the tree is likewise much simpler --
the new infrastructure handles almost all the work for simple types,
and just about anything else can be done with a small helper function.
All existing nodes are where they were before (numerically speaking),
so all existing consumers of sysctl information should notice no
difference.
PS - I'm sorry, but there's a distinct lack of documentation at the
moment. I'm working on sysctl(3/8/9) right now, and I promise to
watch out for buses.
copyin() or copyout().
uvm_useracc() tells us whether the mapping permissions allow access to
the desired part of an address space, and many callers assume that
this is the same as knowing whether an attempt to access that part of
the address space will succeed. however, access to user space can
fail for reasons other than insufficient permission, most notably that
paging in any non-resident data can fail due to i/o errors. most of
the callers of uvm_useracc() make the above incorrect assumption. the
rest are all misguided optimizations, which optimize for the case
where an operation will fail. we'd rather optimize for operations
succeeding, in which case we should just attempt the access and handle
failures due to insufficient permissions the same way we handle i/o
errors. since there appear to be no good uses of uvm_useracc(), we'll
just remove it.
(1) split the single list of pages allocated to a pool into three lists:
completely full, partially full, and completely empty.
there is no longer any need to traverse any list looking for a
certain type of page.
(2) replace the 8-element hash table for out-of-page page headers
with a splay tree.
these two changes (together with the recent enhancements to the wait code)
give us linear scaling for a fork+exit microbenchmark.
to improve scalability of operations on the map.
originally done by Niels Provos for OpenBSD.
tweaked for NetBSD by me with some advices from enami tsugutomo.
discussed on tech-kern@ and tech-perform@.
- after sleeping for memory, re-check if we have a page.
- put the allocated page to pageq to appease UVM_PAGE_TRKOWN.
- dequeue the page when doing ->K loan.
The case where l_stat == LSONPROC and l_cpu == curcpu cannot happen
because the pagedaemon is the LWP on curcpu and the pagedaemon is a
kernel thread and the code is only used by the pagedaemon.
See also updated patch in PR kern/23095, which I ment to checkin
originally.
uvm_swapout_threads will swapout LWPs which are running on another CPU:
- uvm_swapout_threads considers LWPs running on another CPU for swapout
if their l_swtime is high
- uvm_swapout_threads considers LWPs on the runqueue for swapout if their
l_swtime is high but these LWPs might be running by the time uvm_swapout
is called
symptoms of failure: panic in setrunqueue
fixes PR kern/23095
with the KVA of the newly-wired uarea.
This is useful on some architectures (e.g. xscale) where the uarea mapping
can be tweaked to use the mini-data cache instead of the main cache.
