ways:
(1) Cached pages are no longer kept in the object's resident page
splay tree and memq. Instead, they are kept in a separate per-object
splay tree of cached pages. However, access to this new per-object
splay tree is synchronized by the _free_ page queues lock, not to be
confused with the heavily contended page queues lock. Consequently, a
cached page can be reclaimed by vm_page_alloc(9) without acquiring the
object's lock or the page queues lock.
This solves a problem independently reported by tegge@ and Isilon.
Specifically, they observed the page daemon consuming a great deal of
CPU time because of pages bouncing back and forth between the cache
queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of
this problem turned out to be a deadlock avoidance strategy employed
when selecting a cached page to reclaim in vm_page_select_cache().
However, the root cause was really that reclaiming a cached page
required the acquisition of an object lock while the page queues lock
was already held. Thus, this change addresses the problem at its
root, by eliminating the need to acquire the object's lock.
Moreover, keeping cached pages in the object's primary splay tree and
memq was, in effect, optimizing for the uncommon case. Cached pages
are reclaimed far, far more often than they are reactivated. Instead,
this change makes reclamation cheaper, especially in terms of
synchronization overhead, and reactivation more expensive, because
reactivated pages will have to be reentered into the object's primary
splay tree and memq.
(2) Cached pages are now stored alongside free pages in the physical
memory allocator's buddy queues, increasing the likelihood that large
allocations of contiguous physical memory (i.e., superpages) will
succeed.
Finally, as a result of this change long-standing restrictions on when
and where a cached page can be reclaimed and returned by
vm_page_alloc(9) are eliminated. Specifically, calls to
vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and
return a formerly cached page. Consequently, a call to malloc(9)
specifying M_NOWAIT is less likely to fail.
Discussed with: many over the course of the summer, including jeff@,
Justin Husted @ Isilon, peter@, tegge@
Tested by: an earlier version by kris@
Approved by: re (kensmith)
This patch was part of ACPI-CA 20070508 release and the
following is excerpt from its change log:
Fixed a problem where the Global Lock handle was not properly
updated if a thread that acquired the Global Lock via executing
AML code then attempted to acquire the lock via the
AcpiAcquireGlobalLock interface. Reported by Joe Liu.
Approved by: re (kensmith)
Tested by: ambrisko
Obtained from: Intel
polling/interrupt-driven fallback and instead use polling only during
boot and pure interrupt-driven mode after boot. Polled mode could be
relegated completely to a legacy role if we could enable interrupts
during boot. Polled mode can be forced after boot by setting
debug.acpi.ec.polled="1", i.e. if there are timeouts.
- Use polling only during boot, shutdown, or if requested by the user.
Otherwise, use a generation count of GPEs, incremented atomically. This
prevents an old status value from being used if the EC is really slow
and the same condition (i.e. multiple IBEs for a write transaction) is
being checked.
- Check for and run the query handler directly if the SCI bit is set in
the status register during boot. Previously, the query handler wouldn't
run until interrupts were finally enabled late in boot.
- During boot and after starting a command, check if the event appears
to already have occurred before we even start waiting. If so, it's
possible the EC is very slow and we might accept an old status value.
Print a warning in this case. Once we've booted, interrupt-driven mode
should work just fine but polled mode could be unreliable. There's not
much more we can do about this until interrupts are enabled during boot.
- In the above case, we also do one final check if the interrupt-driven
mode gets a timeout. If the status is complete, it will force the
system back into polled mode since interrupt mode doesn't work. For
polled mode during boot, if the status appears to be already complete
before beginning the check loop, it waits 10 us before actually checking
the status, just in case the EC is really slow and hasn't gotten to work
on the new request yet.
- Use upper-case hex for the _Qxx method
- Use device_printf for errors, don't hide them under verbose
- Increase default total timeout to 750 ms and decrease polling interval
to 5 us.
- Don't pass the status value via the softc. Just read it directly.
- Remove the mutex. We use the sx lock for transaction serialization
with the query handler.
- Remove the Intel copyright notice as no code of theirs was ever
present in this file (verified against rev 1.1)
- Allow KTR module-only builds for ease of testing
Thanks to jkim and Alexey Starikovskiy for helpful discussions and testing.
Approved by: re
MFC after: 2 weeks
because of differences between how tinderbox compile LINT - after this
change I only compiled LINT on i386 and the bug shows up on 64bit platforms.
Requested by: des
Approved by: re (implicitly)
- Reintegrate the ANSI C function declaration change
from tcp_timer.c rev 1.92
- Reorganize the tcpcb structure so that it has a single
pointer to the "tcp_timer" structure which contains all
of the tcp timer callouts. This change means that when
the single tcp timer change is reintegrated, tcpcb will
not change in size, and therefore the ABI between
netstat and the kernel will not change.
Neither of these changes should have any functional
impact.
Reviewed by: bmah, rrs
Approved by: re (bmah)
route and once they are done with it, call rtfree(). rtfree() should
only be used when we are certain we hold the last reference to the
route. This bug results in console messages like the following:
rtfree: 0xc40f7000 has 1 refs
This patch switches the rtfree() to use RTFREE_LOCKED() instead,
which should handle the reference counting on the route better.
Approved by: re@ (gnn)
Reviewed by: bms
Reported by: many via net@ and current@
Tested by: many
$ ipfw -n add 1 allow layer2 not mac-type ip
00001 allow ip from any to any layer2 not not mac-type 0x0800
PR: bin/115372
Submitted by: Andrey V. Elsukov
Approved by: re (hrs)
MFC after: 3 weeks
All active fields in fsi are advisory/optional, so we shouldn't do
extra work to make them valid at all times, but instead we write to
the fsi too often (we still do), and we searched for a free cluster
for fsinxtfree too often.
This commit just removes the whole search and its results, so that we
write out our in-core copy of fsinxtfree instead of writing a "fixed"
copy and clobbering our in-core copy. This saves fixing 3 bugs:
- off-by-1 error for the end of the search, resulting in fsinxtfree
not actually being adjusted iff only the last cluster is free.
- missing adjustment when no clusters are free.
- off-by-many error for the start of the search. Starting the search
at 0 instead of at (the in-core copy of) fsinxtfree did more than
defeat the reasons for existence of fsinxtfree. fsinxtfree exists
mainly to avoid having to start at 0 for just the first search per
mount, but has the side effect of reducing bias towards allocating
near cluster 0. The bias would normally only be generated by the
first search per mount (if fsinxtfree is not supported), but since
we also adjusted the in-core copy of fsinxtfree here, we were doing
extra work to maximize the bias.
Approved by: re (kensmith)
providers with limited physical storage and add physical storage as
needed.
Submitted by: Ivan Voras
Sponsored by: Google Summer of Code 2006
Approved by: re (kensmith)
to an int to remove the warning from using a size_t variable on 64-bit
platforms.
Submitted by: Xin LI <delphij@FreeBSD.org>
Approved by: wes
Approved by: re (kensmith)
- Improve load long-term load balancer by always IPIing exactly once.
Previously the delay after rebalancing could cause problems with
uneven workloads.
- Allow nice to have a linear effect on the interactivity score. This
allows negatively niced programs to stay interactive longer. It may be
useful with very expensive Xorg servers under high loads. In general
it should not be necessary to alter the nice level to improve interactive
response. We may also want to consider never allowing positively niced
processes to become interactive at all.
- Initialize ccpu to 0 rather than 0.0. The decimal point was leftover
from when the code was copied from 4bsd. ccpu is 0 in ULE because ULE
only exports weighted cpu values.
Reported by: Steve Kargl (Load balancing problem)
Approved by: re
Eliminates panics due to locking issues.
Idea taken from src/sys/gnu/fs/xfs/FreeBSD/xfs_super.c.
PR: 89966, 92000, 104393
Reported by: H. Matsuo <hiroshi50000 yahoo co jp>,
Chris <m2chrischou gmail.com>,
Andrey V. Elsukov <bu7cher yandex ru>,
Jan Henrik Sylvester <me janh de>
Approved by: re (kensmith)
as they would have been translated from partitions of type "GPT".
This fixes sysinstall, now that geom_part has taken over from
geom_gpt.
Approved by: re (kensmith)
This allows to use numbers in human-readable form in many geom(8)
utilities. Such a simple change and makes live so much nicer.
Some examples:
gstripe label -s 16k
gmirror label -s 4k
gnop create -o 1g -s 128m -S 2k
gjournal label -s 2g
geli label -i 128k -s 4k
Approved by: re (kensmith)
simplifies code and should speedup pppoe_findsession() function which is
called for every incoming packet.
Approved by: re (kensmith), glebius (mentor)
changes the units from seconds to the value of 'ticks' when swapped
in/out. ULE does not have a periodic timer that scans all threads in
the system and as such maintaining a per-second counter is difficult.
- Change computations requiring the unit in seconds to subtract ticks
and divide by hz. This does make the wraparound condition hz times
more frequent but this is still in the range of several months to
years and the adverse effects are minimal.
Approved by: re
number of bytes written, even when used to write files to
disk. Extend the test suite to verify the correct return
values for archive_write_data() and archive_write_data_block().
Thanks to: Bruce Mah, for stepping in promptly to back out the
earlier broken version of this fix
Thanks to: Colin Percival, for pointing out the correct fix
MFC after: 5 days
Approved by: re (ksmith)
Pointy hat: \me
last interface should own the address, but the current code
fumbles the handoff. This fixes that.
- move address related debugs to PCB4 and add additional ones to
help in debugging address problems.
Approved by: re@freebsd.org (K Smith)
changes the units from seconds to the value of 'ticks' when swapped
in/out. ULE does not have a periodic timer that scans all threads in
the system and as such maintaining a per-second counter is difficult.
- Change computations requiring the unit in seconds to subtract ticks
and divide by hz. This does make the wraparound condition hz times
more frequent but this is still in the range of several months to
years and the adverse effects are minimal.
Approved by: re
Some ports will install with compressed manpages. man handles
this by looking for the .gz version of a man source file.
It is also common to include other files with the .so
directive where commands or functions share a man page.
Traditionally ports have had to handle this by either not
compressing the manpages, or using the _MLINKS macro in the
port makefile to create symlinks to the actual source file,
rather than using .so versions. Notably, the current version
of Xorg port breaks. See ports/113096 and ports/115845.
PR: bin/115850
Submitted by: Callum Gibson <callumgibson@optusnet.com.au>
Approved by: re@ (ken smith), grog@ (mentor)
MFC after: 1 week
In particular:
- smp_tlb_mtx is no longer used, so it is axed.
- smp rendezvous lock isn't really a leaf spin-mutex. Its bad placement in
the table, however, has been the source of a false positive LOR reporting
with the dt_lock. However, smp rendezvous lock would have had sched_lock
there for older lock, so it wasn't still a leaf lock.
- allpmaps is only used in ia32 architecture, so it is inserted in the
appropriate stub.
Addictionally:
- kse_zombie_lock is no longer present, so its definition is axed out.
- zombie_lock doesn't need to have an exported symbol, so just let's it be
declared as static.
Tested by: kris
Approved by: jeff (mentor)
Approved by: re
