This allocator uses a binary buddy system with a twist. First and
foremost, this allocator is required to support the implementation of
superpages. As a side effect, it enables a more robust implementation
of contigmalloc(9). Moreover, this reimplementation of
contigmalloc(9) eliminates the acquisition of Giant by
contigmalloc(..., M_NOWAIT, ...).
The twist is that this allocator tries to reduce the number of TLB
misses incurred by accesses through a direct map to small, UMA-managed
objects and page table pages. Roughly speaking, the physical pages
that are allocated for such purposes are clustered together in the
physical address space. The performance benefits vary. In the most
extreme case, a uniprocessor kernel running on an Opteron, I measured
an 18% reduction in system time during a buildworld.
This allocator does not implement page coloring. The reason is that
superpages have much the same effect. The contiguous physical memory
allocation necessary for a superpage is inherently colored.
Finally, the one caveat is that this allocator does not effectively
support prezeroed pages. I hope this is temporary. On i386, this is
a slight pessimization. However, on amd64, the beneficial effects of
the direct-map optimization outweigh the ill effects. I speculate
that this is true in general of machines with a direct map.
Approved by: re
In particular:
- Add an explicative table for locking of struct vmmeter members
- Apply new rules for some of those members
- Remove some unuseful comments
Heavily reviewed by: alc, bde, jeff
Approved by: jeff (mentor)
to the build.
This allocator uses a binary buddy system with a twist. First and
foremost, this allocator is required to support the implementation of
superpages. As a side effect, it enables a more robust implementation
of contigmalloc(9). Moreover, this reimplementation of
contigmalloc(9) eliminates the acquisition of Giant by
contigmalloc(..., M_NOWAIT, ...).
The twist is that this allocator tries to reduce the number of TLB
misses incurred by accesses through a direct map to small, UMA-managed
objects and page table pages. Roughly speaking, the physical pages
that are allocated for such purposes are clustered together in the
physical address space. The performance benefits vary. In the most
extreme case, a uniprocessor kernel running on an Opteron, I measured
an 18% reduction in system time during a buildworld.
This allocator does not implement page coloring. The reason is that
superpages have much the same effect. The contiguous physical memory
allocation necessary for a superpage is inherently colored.
Finally, the one caveat is that this allocator does not effectively
support prezeroed pages. I hope this is temporary. On i386, this is
a slight pessimization. However, on amd64, the beneficial effects of
the direct-map optimization outweigh the ill effects. I speculate
that this is true in general of machines with a direct map.
Approved by: re
- Use thread_lock() rather than sched_lock for per-thread scheduling
sychronization.
- Use the per-process spinlock rather than the sched_lock for per-process
scheduling synchronization.
Tested by: kris, current@
Tested on: i386, amd64, ULE, 4BSD, libthr, libkse, PREEMPTION, etc.
Discussed with: kris, attilio, kmacy, jhb, julian, bde (small parts each)
Now, we assume no more sched_lock protection for some of them and use the
distribuited loads method for vmmeter (distribuited through CPUs).
Reviewed by: alc, bde
Approved by: jeff (mentor)
- Rename PCPU_LAZY_INC into PCPU_INC
- Add the PCPU_ADD interface which just does an add on the pcpu member
given a specific value.
Note that for most architectures PCPU_INC and PCPU_ADD are not safe.
This is a point that needs some discussions/work in the next days.
Reviewed by: alc, bde
Approved by: jeff (mentor)
td_ru. This removes the requirement for per-process synchronization in
statclock() and mi_switch(). This was previously supported by
sched_lock which is going away. All modifications to rusage are now
done in the context of the owning thread. reads proceed without locks.
- Aggregate exiting threads rusage in thread_exit() such that the exiting
thread's rusage is not lost.
- Provide a new routine, rufetch() to fetch an aggregate of all rusage
structures from all threads in a process. This routine must be used
in any place requiring a rusage from a process prior to it's exit. The
exited process's rusage is still available via p_ru.
- Aggregate tick statistics only on demand via rufetch() or when a thread
exits. Tick statistics are kept in the thread and protected by sched_lock
until it exits.
Initial patch by: attilio
Reviewed by: attilio, bde (some objections), arch (mostly silent)
Probabilly, a general approach is not the better solution here, so we should
solve the sched_lock protection problems separately.
Requested by: alc
Approved by: jeff (mentor)
Change the VOP_OPEN(), vn_open() vnode operation and d_fdopen() cdev operation
argument from being file descriptor index into the pointer to struct file.
Proposed and reviewed by: jhb
Reviewed by: daichi (unionfs)
Approved by: re (kensmith)
These functions are intended to do the same actions of sx_xlock() and
sx_slock() but with the difference to perform an interruptible sleep, so
that sleep can be interrupted by external events.
In order to support these new featueres, some code renstruction is needed,
but external API won't be affected at all.
Note: use "void" cast for "int" returning functions in order to avoid tools
like Coverity prevents to whine.
Requested by: rwatson
Tested by: rwatson
Reviewed by: jhb
Approved by: jeff (mentor)
vm_map_pmap_enter() unless the caller is madvise(MADV_WILLNEED). With
the exception of calls to vm_map_pmap_enter() from
madvise(MADV_WILLNEED), vm_fault_prefault() and vm_map_pmap_enter()
are both used to create speculative mappings. Thus, always
reactivating cached pages is a mistake. In principle, cached pages
should only be reactivated by an actual access. Otherwise, the
following misbehavior can occur. On a hard fault for a text page the
clustering algorithm fetches not only the required page but also
several of the adjacent pages. Now, suppose that one or more of the
adjacent pages are never accessed. Ultimately, these unused pages
become cached pages through the efforts of the page daemon. However,
the next activation of the executable reactivates and maps these
unused pages. Consequently, they are never replaced. In effect, they
become pinned in memory.
vmcnts. This can be used to abstract away pcpu details but also changes
to use atomics for all counters now. This means sched lock is no longer
responsible for protecting counts in the switch routines.
Contributed by: Attilio Rao <attilio@FreeBSD.org>
VM_PHYSSEG_SPARSE depending on whether the physical address space is
densely or sparsely populated with memory. The effect of this
definition is to determine which of two implementations of
vm_page_array and PHYS_TO_VM_PAGE() is used. The legacy
implementation is obtained by defining VM_PHYSSEG_DENSE, and a new
implementation that trades off time for space is obtained by defining
VM_PHYSSEG_SPARSE. For now, all architectures except for ia64 and
sparc64 define VM_PHYSSEG_DENSE. Defining VM_PHYSSEG_SPARSE on ia64
allows the entirety of my Itanium 2's memory to be used. Previously,
only the first 1 GB could be used. Defining VM_PHYSSEG_SPARSE on
sparc64 allows USIIIi-based systems to boot without crashing.
This change is a combination of Nathan Whitehorn's patch and my own
work in perforce.
Discussed with: kmacy, marius, Nathan Whitehorn
PR: 112194
contigmalloc2() was always testing the first physical page for PG_ZERO,
not the current page of interest.
Submitted by: Michael Plass
PR: 81301
MFC after: 1 week
The problem is this: vm_fault_additional_pages() calls vm_pager_has_page(),
which calls vnode_pager_haspage(). Now when VOP_BMAP() returns an error (eg.
EOPNOTSUPP), vnode_pager_haspage() returns TRUE without initializing 'before'
and 'after' arguments, so we have some accidental values there. This bascially
was causing this condition to be meet:
if ((rahead + rbehind) >
((cnt.v_free_count + cnt.v_cache_count) - cnt.v_free_reserved)) {
pagedaemon_wakeup();
[...]
}
(we have some random values in rahead and rbehind variables)
I'm not entirely sure this is the right fix, maybe we should just return FALSE
in vnode_pager_haspage() when VOP_BMAP() fails?
alc@ knows about this problem, maybe he will be able to come up with a better
fix if this is not the right one.
causing a crash.
Suppose that we have two objects, obj and backing_obj, where
backing_obj is obj's backing object. Further, suppose that
backing_obj has a reference count of two. One being the reference
held by obj and the other by a map entry. Now, suppose that the map
entry is deallocated and its reference removed by
vm_object_deallocate(). vm_object_deallocate() recognizes that the
only remaining reference is from a shadow object, obj, and calls
vm_object_collapse() on obj. vm_object_collapse() executes
if (backing_object->ref_count == 1) {
/*
* If there is exactly one reference to the backing
* object, we can collapse it into the parent.
*/
vm_object_backing_scan(object, OBSC_COLLAPSE_WAIT);
vm_object_backing_scan(OBSC_COLLAPSE_WAIT) executes
if (op & OBSC_COLLAPSE_WAIT) {
vm_object_set_flag(backing_object, OBJ_DEAD);
}
Finally, suppose that either vm_object_backing_scan() or
vm_object_collapse() sleeps releasing its locks. At this instant,
another thread executes vm_object_split(). It crashes in
vm_object_reference_locked() on the assertion that the object is not
dead. If, however, assertions are not enabled, it crashes much later,
after the object has been recycled, in vm_object_deallocate() because
the shadow count and shadow list are inconsistent.
Reviewed by: tegge
Reported by: jhb
MFC after: 1 week
1) Eliminate an unnecessary check for fictitious pages. Specifically,
only device-backed objects contain fictitious pages and the object is
not device-backed.
2) Change the types of "psize" and "tmpidx" to vm_pindex_t in order to
prevent possible wrap around with extremely large maps and objects,
respectively. Observed by: tegge (last summer)
immediately flag any page that is allocated to a OBJT_PHYS object as
unmanaged in vm_page_alloc() rather than waiting for a later call to
vm_page_unmanage(). This allows for the elimination of some uses of
the page queues lock.
Change the type of the kernel and kmem objects from OBJT_DEFAULT to
OBJT_PHYS. This allows us to take advantage of the above change to
simplify the allocation of unmanaged pages in kmem_alloc() and
kmem_malloc().
Remove vm_page_unmanage(). It is no longer used.
vm_page_free_toq() to account for recent changes that allow
vm_page_free_toq() to be called on some pages without the page queues lock
being held, specifically, pages that are not contained in a vm object and
not a member of a page queue. (Examples of such pages include page table
pages, pv entry pages, and uma small alloc pages.)
is actually being added to the hold queue, not the free queue. At the same
time, avoid unnecessary tests to wake up threads waiting for free memory
and the idle thread that zeroes free pages. (These tests will be performed
later when the page finally moves from the hold queue to the free queue.)
boot by MD code to indicated detected alignment preference. Rather than
cache alignment being encoded in UMA consumers by defining a global
alignment value of (16 - 1) in UMA_ALIGN_CACHE, UMA_ALIGN_CACHE is now
a special value (-1) that causes UMA to look at registered alignment. If
no preferred alignment has been selected by MD code, a default alignment
of (16 - 1) will be used.
Currently, no hardware platforms specify alignment; architecture
maintainers will need to modify MD startup code to specify an alignment
if desired. This must occur before initialization of UMA so that all UMA
zones pick up the requested alignment.
Reviewed by: jeff, alc
Submitted by: attilio
- Restore support for fetching swap information from crash dumps via
kvm_get_swapinfo(3) to fix pstat -T/-s on crash dumps.
Reviewed by: arch@, phk
MFC after: 1 week
zone. Cluster allocations fail when this happens. Also processes that may have
blocked on cluster allocations will never be woken up. Thanks to rwatson for
an overview of the issue and pointers to the mbuma paper and his tool to dump
out UMA zones.
Reviewed by: andre@
maxpages on a zone is woken up, with the rest never being woken up as
a result of the ZFLAG_FULL flag being cleared. Wakeup all such blocked
procsses instead. This change introduces a thundering herd, but since
this should be relatively infrequent, optimizing this (by introducing
a count of blocked processes, for example) may be premature.
Reviewd by: ups@
setrunqueue() was mostly empty. The few asserts and thread state
setting were moved to the individual schedulers. sched_add() was
chosen to displace it for naming consistency reasons.
- Remove adjustrunqueue, it was 4 lines of code that was ifdef'd to be
different on all three schedulers where it was only called in one place
each.
- Remove the long ifdef'd out remrunqueue code.
- Remove the now redundant ts_state. Inspect the thread state directly.
- Don't set TSF_* flags from kern_switch.c, we were only doing this to
support a feature in one scheduler.
- Change sched_choose() to return a thread rather than a td_sched. Also,
rely on the schedulers to return the idlethread. This simplifies the
logic in choosethread(). Aside from the run queue links kern_switch.c
mostly does not care about the contents of td_sched.
Discussed with: julian
- Move the idle thread loop into the per scheduler area. ULE wants to
do something different from the other schedulers.
Suggested by: jhb
Tested on: x86/amd64 sched_{4BSD, ULE, CORE}.
allocations were made using improper flags in interrupt context.
Replace with a simple WITNESS warning call. This restores the
invariant that M_WAITOK allocations will always succeed or die
horribly trying, which is relied on by many UMA consumers.
MFC after: 3 weeks
Discussed with: jhb
- Add a printf in swp_pager_meta_build() to warn if the swapzone becomes
exhausted so that there's at least a warning before a box that runs out
of swapzone space before running out of swap space deadlocks.
MFC after: 1 week
Reviwed by: alc
Make part of John Birrell's KSE patch permanent..
Specifically, remove:
Any reference of the ksegrp structure. This feature was
never fully utilised and made things overly complicated.
All code in the scheduler that tried to make threaded programs
fair to unthreaded programs. Libpthread processes will already
do this to some extent and libthr processes already disable it.
Also:
Since this makes such a big change to the scheduler(s), take the opportunity
to rename some structures and elements that had to be moved anyhow.
This makes the code a lot more readable.
The ULE scheduler compiles again but I have no idea if it works.
The 4bsd scheduler still reqires a little cleaning and some functions that now do
ALMOST nothing will go away, but I thought I'd do that as a separate commit.
Tested by David Xu, and Dan Eischen using libthr and libpthread.
specific privilege names to a broad range of privileges. These may
require some future tweaking.
Sponsored by: nCircle Network Security, Inc.
Obtained from: TrustedBSD Project
Discussed on: arch@
Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri,
Alex Lyashkov <umka at sevcity dot net>,
Skip Ford <skip dot ford at verizon dot net>,
Antoine Brodin <antoine dot brodin at laposte dot net>
begun with a repo-copy of mac.h to mac_framework.h. sys/mac.h now
contains the userspace and user<->kernel API and definitions, with all
in-kernel interfaces moved to mac_framework.h, which is now included
across most of the kernel instead.
This change is the first step in a larger cleanup and sweep of MAC
Framework interfaces in the kernel, and will not be MFC'd.
Obtained from: TrustedBSD Project
Sponsored by: SPARTA
it introduced a check after the call to file system's get pages method
that assumes that the get pages method does not change the array of pages
that is passed to it. In the case of vnode_pager_generic_getpages(),
this assumption has been incorrect. The contents of the array of pages
may be shifted by vnode_pager_generic_getpages(). Likely, the problem
has been hidden by vnode_pager_haspage() limiting the set of pages that
are passed to vnode_pager_generic_getpages() such that a shift never
occurs.
The fix implemented herein is to adjust the pointer to the array of pages
rather than shifting the pages within the array.
MFC after: 3 weeks
Fix suggested by: tegge
an error returned by VOP_BMAP() and a hole in the file.
Change the callers to vnode_pager_addr() such that they return
VM_PAGER_ERROR when VOP_BMAP fails instead of a zero-filled page.
Reviewed by: tegge
MFC after: 3 weeks
size aligned requiring heavy usage of vm_page_alloc_contig
This change makes vm_page_alloc_contig SMP safe
Approved by: scottl (acting as backup for mentor rwatson)
vnode_pager_generic_getpages(): (1) that VOP_BMAP() is unsupported by the
underlying file system and (2) an error in performing the VOP_BMAP().
Previously, vnode_pager_generic_getpages() assumed that all errors were
of the first type. If, in fact, the error was of the second type, the
likely outcome was for the process to become permanently blocked on a busy
page.
MFC after: 3 weeks
Reviewed by: tegge
inlined and a procedure call is made in the rare case, i.e., when it is
necessary to sleep. In this case, inlining the test actually makes the
kernel smaller.
page queues-synchronized flag. Reduce the scope of the page queues lock in
vm_fault() accordingly.
Move vm_fault()'s call to vm_object_set_writeable_dirty() outside of the
scope of the page queues lock. Reviewed by: tegge
Additionally, eliminate an unnecessary dereference in computing the
argument that is passed to vm_object_set_writeable_dirty().
synchronized by the lock on the object containing the page.
Transition PG_WANTED and PG_SWAPINPROG to use the new field,
eliminating the need for holding the page queues lock when setting
or clearing these flags. Rename PG_WANTED and PG_SWAPINPROG to
VPO_WANTED and VPO_SWAPINPROG, respectively.
Eliminate the assertion that the page queues lock is held in
vm_page_io_finish().
Eliminate the acquisition and release of the page queues lock
around calls to vm_page_io_finish() in kern_sendfile() and
vfs_unbusy_pages().
Originally, I had adopted sparc64's name, pmap_clear_write(), for the
function that is now pmap_remove_write(). However, this function is more
like pmap_remove_all() than like pmap_clear_modify() or
pmap_clear_reference(), hence, the name change.
The higher-level rationale behind this change is described in
src/sys/amd64/amd64/pmap.c revision 1.567. The short version is that I'm
trying to clean up and fix our support for execute access.
Reviewed by: marcel@ (ia64)
system's machine-dependent and machine-independent layers. Once
pmap_clear_write() is implemented on all of our supported
architectures, I intend to replace all calls to pmap_page_protect() by
calls to pmap_clear_write(). Why? Both the use and implementation of
pmap_page_protect() in our virtual memory system has subtle errors,
specifically, the management of execute permission is broken on some
architectures. The "prot" argument to pmap_page_protect() should
behave differently from the "prot" argument to other pmap functions.
Instead of meaning, "give the specified access rights to all of the
physical page's mappings," it means "don't take away the specified
access rights from all of the physical page's mappings, but do take
away the ones that aren't specified." However, owing to our i386
legacy, i.e., no support for no-execute rights, all but one invocation
of pmap_page_protect() specifies VM_PROT_READ only, when the intent
is, in fact, to remove only write permission. Consequently, a
faithful implementation of pmap_page_protect(), e.g., ia64, would
remove execute permission as well as write permission. On the other
hand, some architectures that support execute permission have
basically ignored whether or not VM_PROT_EXECUTE is passed to
pmap_page_protect(), e.g., amd64 and sparc64. This change represents
the first step in replacing pmap_page_protect() by the less subtle
pmap_clear_write() that is already implemented on amd64, i386, and
sparc64.
Discussed with: grehan@ and marcel@
pointer: When vm_object_deallocate() sleeps because of a non-zero
paging in progress count on either object or object's shadow,
vm_object_deallocate() must ensure that object is still the shadow's
backing object when it reawakens. In fact, object may have been
deallocated while vm_object_deallocate() slept. If so, reacquiring
the lock on object can lead to a deadlock.
Submitted by: ups@
MFC after: 3 weeks
libmemstat(3) is used by vmstat (and friends) to produce more accurate
and more detailed statistics information in a machine-readable way,
and vmstat continues to provide the same text-based front-end.
This change should not be MFC'd.
vm_page_startup(). As a result, we now only lookup the tunable once
instead of looking it up once for every physical page of memory in the
system. This cuts out about a 1 second or so delay in boot on x86
systems. The delay is much larger and more noticable on sun4v apparently.
Reported by: kmacy
MFC after: 1 week
There is a race with the current locking scheme and removing
it should have no measurable performance impact.
This fixes page faults leading to panics in pmap_enter_quick_locked()
on amd64/i386.
Reviewed by: alc,jhb,peter,ps
Found mapped cache page. Specifically, if cnt.v_free_count dips below
cnt.v_free_reserved after p_start has been set to a non-NULL value,
then vm_map_pmap_enter() would break out of the loop and incorrectly
call pmap_enter_object() for the remaining address range. To correct
this error, this revision truncates the address range so that
pmap_enter_object() will not map any cache pages.
In collaboration with: tegge@
Reported by: kris@
These pages are allocated from the direct map, and were not previous
tracked. This included the vm_page_array and the early UMA bootstrap
pages.
Reviewed by: peter
vmspace_exitfree() and vmspace_free() which could result in the same
vmspace being freed twice.
Factor out part of exit1() into new function vmspace_exit(). Attach
to vmspace0 to allow old vmspace to be freed earlier.
Add new function, vmspace_acquire_ref(), for obtaining a vmspace
reference for a vmspace belonging to another process. Avoid changing
vmspace refcount from 0 to 1 since that could also lead to the same
vmspace being freed twice.
Change vmtotal() and swapout_procs() to use vmspace_acquire_ref().
Reviewed by: alc
report this as an allocation failure for the item type. The failure
will be separately recorded with the bucket type. This my eliminate
high mbuf allocation failure counts under some circumstances, which
can be alarming in appearance, but not actually a problem in
practice.
MFC after: 2 weeks
Reported by: ps, Peter J. Blok <pblok at bsd4all dot org>,
OxY <oxy at field dot hu>,
Gabor MICSKO <gmicskoa at szintezis dot hu>
via the debug.minidump sysctl and tunable.
Traditional dumps store all physical memory. This was once a good thing
when machines had a maximum of 64M of ram and 1GB of kvm. These days,
machines often have many gigabytes of ram and a smaller amount of kvm.
libkvm+kgdb don't have a way to access physical ram that is not mapped
into kvm at the time of the crash dump, so the extra ram being dumped
is mostly wasted.
Minidumps invert the process. Instead of dumping physical memory in
in order to guarantee that all of kvm's backing is dumped, minidumps
instead dump only memory that is actively mapped into kvm.
amd64 has a direct map region that things like UMA use. Obviously we
cannot dump all of the direct map region because that is effectively
an old style all-physical-memory dump. Instead, introduce a bitmap
and two helper routines (dump_add_page(pa) and dump_drop_page(pa)) that
allow certain critical direct map pages to be included in the dump.
uma_machdep.c's allocator is the intended consumer.
Dumps are a custom format. At the very beginning of the file is a header,
then a copy of the message buffer, then the bitmap of pages present in
the dump, then the final level of the kvm page table trees (2MB mappings
are expanded into a 4K page mappings), then the sparse physical pages
according to the bitmap. libkvm can now conveniently access the kvm
page table entries.
Booting my test 8GB machine, forcing it into ddb and forcing a dump
leads to a 48MB minidump. While this is a best case, I expect minidumps
to be in the 100MB-500MB range. Obviously, never larger than physical
memory of course.
minidumps are on by default. It would want be necessary to turn them off
if it was necessary to debug corrupt kernel page table management as that
would mess up minidumps as well.
Both minidumps and regular dumps are supported on the same machine.
if the specified priority is zero. This avoids a race where the calling
thread could read a snapshot of it's current priority, then a different
thread could change the first thread's priority, then the original thread
would call sched_prio() inside msleep() undoing the change made by the
second thread. I used a priority of zero as no thread that calls msleep()
or tsleep() should be specifying a priority of zero anyway.
The various places that passed 'curthread->td_priority' or some variant
as the priority now pass 0.
Kernel changes:
Inform hwpmc of executable objects brought into the system by
kldload() and mmap(), and of their removal by kldunload() and
munmap(). A helper function linker_hwpmc_list_objects() has been
added to "sys/kern/kern_linker.c" and is used by hwpmc to retrieve
the list of currently loaded kernel modules.
The unused `MAPPINGCHANGE' event has been deprecated in favour
of separate `MAP_IN' and `MAP_OUT' events; this change reduces
space wastage in the log.
Bump the hwpmc's ABI version to "2.0.00". Teach hwpmc(4) to
handle the map change callbacks.
Change the default per-cpu sample buffer size to hold
32 samples (up from 16).
Increment __FreeBSD_version.
libpmc(3) changes:
Update libpmc(3) to deal with the new events in the log file; bring
the pmclog(3) manual page in sync with the code.
pmcstat(8) changes:
Introduce new options to pmcstat(8): "-r" (root fs path), "-M"
(mapfile name), "-q"/"-v" (verbosity control). Option "-k" now
takes a kernel directory as its argument but will also work with
the older invocation syntax.
Rework string handling in pmcstat(8) to use an opaque type for
interned strings. Clean up ELF parsing code and add support for
tracking dynamic object mappings reported by a v2.0.00 hwpmc(4).
Report statistics at the end of a log conversion run depending
on the requested verbosity level.
Reviewed by: jhb, dds (kernel parts of an earlier patch)
Tested by: gallatin (earlier patch)
be called without any vnode locks held. Remove calls to vn_start_write() and
vn_finished_write() in vnode_pager_putpages() and add these calls before the
vnode lock is obtained to most of the callers that don't already have them.
object that requires Giant in vm_object_deallocate(). This is somewhat
hairy in that if we can't obtain Giant directly, we have to drop the
object lock, then lock Giant, then relock the object lock and verify that
we still need Giant. If we don't (because the object changed to OBJT_DEAD
for example), then we drop Giant before continuing.
Reviewed by: alc
Tested by: kris
and it has not plenty of free pages it tries to free pages in the cache queue.
Unfortunately freeing a cached page requires the locking of the object that
owns the page. However in the context of allocating pages we may not be able
to lock the object and thus can only TRY to lock the object. If the locking try
fails the cache page can not be freed and is activated to move it out of the way
so that we may try to free other cache pages.
If all pages in the cache belong to objects that are currently locked the
cache queue can be emptied without freeing a single page. This scenario caused
two problems:
1) vm_page_alloc always failed allocation when it tried freeing pages from
the cache queue and failed to do so. However if there are more than
cnt.v_interrupt_free_min pages on the free list it should return pages
when requested with priority VM_ALLOC_SYSTEM. Failure to do so can cause
resource exhaustion deadlocks.
2) Threads than need to allocate pages spend a lot of time cleaning up the
page queue without really getting anything done while the pagedaemon
needs to work overtime to refill the cache.
This change fixes the first problem. (1)
Reviewed by: tegge@
lock also protects this flag so it is not necessary.
- Don't rely on v_mount to detect whether or not we've been recycled, use
the more appropriate VI_DOOMED instead.
Sponsored by: Isilon Systems, Inc.
MFC After: 1 week
The former type, size_t, was causing truncation to 32 bits on i386,
which immediately led to undersizing of VM objects backed by
files >4GB. In particular, sendfile(2) was broken for such files.
PR: kern/92243
MFC after: 5 days
It detects both: buffer underflows and buffer overflows bugs at runtime
(on free(9) and realloc(9)) and prints backtraces from where memory was
allocated and from where it was freed.
Tested by: kris
of physical RAM instead of the bottom was a sound idea, but the implementation
left a lot to be desired. Scans would spend considerable time looking at
pages that are above of the address range given by the caller, and multiple
calls (like what happens in busdma) would spend more time on top of that
rescanning the same pages over and over.
Solve this, at least for now, with two simple optimizations. The first is
to not bother scanning high ordered pages that are outside of the provided
address range. Second is to cache the page index from the last successful
operation so that subsequent scans don't have to restart from the top. This
is conditional on the numpages argument being the same or greater between
calls.
MFC After: 2 weeks
The difference between WITNESS_CHECK() and WITNESS_WARN() is that
WITNESS_CHECK() should be used in the places that the return value of
witness_warn() is checked, whereas WITNESS_WARN() should be used in places
where the return value is ignored. Specifically, in a kernel without
WITNESS enabled, WITNESS_WARN() evaluates to an empty string where as
WITNESS_CHECK evaluates to 0. I also updated the one place that was
checking the return value of WITNESS_WARN() to use WITNESS_CHECK.
they are. They should be NULL at this point, except if we're coming from
swapdev_strategy().
It should only affect the case where we're swapping directly on a file over
NFS.
the resident page count matches the object size. We know it fully backs
its parent in this case.
Reviewed by: acl, tegge
Sponsored by: Isilon Systems, Inc.
statement. Specifically, a break statement that previously broke out of
the enclosing switch was not changed. Consequently, the enclosing loop
terminated prematurely.
This could result in "vm_page_insert: page already inserted" panics.
Submitted by: tegge
- provide an interface (macros) to the page coloring part of the VM system,
this allows to try different coloring algorithms without the need to
touch every file [1]
- make the page queue tuning values readable: sysctl vm.stats.pagequeue
- autotuning of the page coloring values based upon the cache size instead
of options in the kernel config (disabling of the page coloring as a
kernel option is still possible)
MD changes:
- detection of the cache size: only IA32 and AMD64 (untested) contains
cache size detection code, every other arch just comes with a dummy
function (this results in the use of default values like it was the
case without the autotuning of the page coloring)
- print some more info on Intel CPU's (like we do on AMD and Transmeta
CPU's)
Note to AMD owners (IA32 and AMD64): please run "sysctl vm.stats.pagequeue"
and report if the cache* values are zero (= bug in the cache detection code)
or not.
Based upon work by: Chad David <davidc@acns.ab.ca> [1]
Reviewed by: alc, arch (in 2004)
Discussed with: alc, Chad David, arch (in 2004)
- Provide tunable vm.memguard.desc, so one can specify memory type without
changing the code and recompiling the kernel.
- Allow to use memguard for kernel modules by providing sysctl
vm.memguard.desc, which can be changed to short description of memory
type before module is loaded.
- Move as much memguard code as possible to memguard.c.
- Add sysctl node vm.memguard. and move memguard-specific sysctl there.
- Add malloc_desc2type() function for finding memory type based on its
short description (ks_shortdesc field).
- Memory type can be changed (via vm.memguard.desc sysctl) only if it
doesn't exist (will be loaded later) or when no memory is allocated yet.
If there is allocated memory for the given memory type, return EBUSY.
- Implement two ways of memory types comparsion and make safer/slower the
default.
ephemeral mappings that are used as the source for three copy
operations from kernel space to user space. There are two reasons for
making this change: (1) Under heavy load exec_map can fill up causing
vm_map_find() to fail. When it fails, the nascent process is aborted
(SIGABRT). Whereas, this reimplementation using sf_buf_alloc()
sleeps. (2) Although it is possible to sleep on vm_map_find()'s
failure until address space becomes available (see kmem_alloc_wait()),
using sf_buf_alloc() is faster. Furthermore, the reimplementation
uses a CPU private mapping, avoiding a TLB shootdown on
multiprocessors.
Problem uncovered by: kris@
Reviewed by: tegge@
MFC after: 3 weeks
reclamation synchronously from get_pv_entry() instead of
asynchronously as part of the page daemon. Additionally, limit the
reclamation to inactive pages unless allocation from the PV entry zone
or reclamation from the inactive queue fails. Previously, reclamation
destroyed mappings to both inactive and active pages. get_pv_entry()
still, however, wakes up the page daemon when reclamation occurs. The
reason being that the page daemon may move some pages from the active
queue to the inactive queue, making some new pages available to future
reclamations.
Print the "reclaiming PV entries" message at most once per minute, but
don't stop printing it after the fifth time. This way, we do not give
the impression that the problem has gone away.
Reviewed by: tegge
by the zero-copy sockets method, and written to before the transmission
completes, we need to destroy all of the existing mappings to the page,
not just the one that we fault on. Otherwise, the mappings will no longer
be to the same page and changes made through one of the mappings will not
be visible through the others.
Observed by: tegge
If a copy-on-write fault occurs on the page, the new copy should inherit
a part of the original page's wire count.
Submitted by: tegge
MFC after: 1 week
- Prefer '_' to ' ', as it results in more easily parsed results in
memory monitoring tools such as vmstat.
- Remove punctuation that is incompatible with using memory type names
as file names, such as '/' characters.
- Disambiguate some collisions by adding subsystem prefixes to some
memory types.
- Generally prefer lower case to upper case.
- If the same type is defined in multiple architecture directories,
attempt to use the same name in additional cases.
Not all instances were caught in this change, so more work is required to
finish this conversion. Similar changes are required for UMA zone names.
vm_object_backing_scan() was not written to handle. Specifically, a wired
page within a backing object that is shadowed by a page within the shadow
object. Handle this state by removing the wired page from the backing
object. The wired page will be freed by socow_iodone().
Stop masking errors: If a page is being freed by vm_object_backing_scan(),
assert that it is no longer mapped rather than quietly destroying any
mappings.
Tested by: Harald Schmalzbauer
Before this change a copy operation with cp(1) would not update the
file access times.
According to the POSIX mmap(2) documentation: the st_atime field
of the mapped file may be marked for update at any time between the
mmap() call and the corresponding munmap() call. The initial read
or write reference to a mapped region shall cause the file's st_atime
field to be marked for update if it has not already been marked for
update.
I'm not sure this is the right thing to do, but at least I don't panic
anymore when swapping on a NFS file without using md(4).
X-MFC after: proper review
UMA boot pages.
Disable recursion on the general UMA lock now that startup_alloc() no
longer uses it.
Eliminate the variable uma_boot_free. It serves no purpose.
Note: This change eliminates a lock-order reversal between a system
map mutex and the UMA lock. See
http://sources.zabbadoz.net/freebsd/lor.html#109 for details.
MFC after: 3 days
vm_pager_init() is run before required nswbuf variable has been set
to correct value. This caused system to run with single pbuf available
for vnode_pager. Handle both cluster_pbuf_freecnt and vnode_pbuf_freecnt
variable in the same way.
Reported by: ade
Obtained from: alc
MFC after: 2 days
due to the vm object being locked.
When a process writes large amounts of data to a file, the vm object associated
with that file can contain most of the physical pages on the machine. If the
process is preempted while holding the lock on the vm object, pagedaemon would
be able to move very few pages from PQ_INACTIVE to PQ_CACHE or from PQ_ACTIVE
to PQ_INACTIVE, resulting in unlimited cleaning of dirty pages belonging to
other vm objects.
Temporarily unlock the page queues lock while locking vm objects to avoid lock
order violation. Detect and handle relevant page queue changes.
This change depends on both the lock portion of struct vm_object and normal
struct vm_page being type stable.
Reviewed by: alc
as opt_vmpage.h will not be available to user space library builds. A
similar existing check is present for KLD_MODULE for similar reasons.
MFC after: 3 days
monitoring API, which might or might not be the same as the internal
maximum (currently none).
Export flag information on UMA zones -- in particular, whether or
not this is a secondary zone, and so the keg free count should be
considered in that light.
MFC after: 1 day
make the b_iodone callback responsible for setting it if it is needed.
Previously, it was set unconditionally by bufdone() without holding
whichever lock is shared by the b_iodone callback and the corresponding
top-half function. Consequently, in a race, the top-half function could
conclude that operation was done before the b_iodone callback finished.
See, for example, aio_physwakeup() and aio_fphysio().
Note: I don't believe that the other, more widely-used b_iodone callbacks
are affected.
Discussed with: jeff
Reviewed by: phk
MFC after: 2 weeks
- Add a new uma_zfree_internal() flag, ZFREE_STATFREE, which causes it to
to update the zone's uz_frees statistic. Previously, the statistic was
updated unconditionally.
- Use the flag in situations where a "real" free occurs: i.e., one where
the caller is freeing an allocated item, to be differentiated from
situations where uma_zfree_internal() is used to tear down the item
during slab teardown in order to invoke its fini() method. Also use
the flag when UMA is freeing its internal objects.
- When exchanging a bucket with the zone from the per-CPU cache when
freeing an item, flush cache statistics back to the zone (since the
zone lock and critical section are both held) to match the allocation
case.
MFC after: 3 days
per-CPU cache statistics. UMA sizes the cache array based on the
number of CPUs at boot (mp_maxid + 1), and iterating based on MAXCPU
could read off the end of the array (into the next zone).
Reported by: yongari
MFC after: 1 week
it covers the following of the uc_alloc/freebucket cache pointers.
Originally, I felt that the race wasn't helped by holding the mutex,
hence a comment in the code and not holding it across the cache access.
However, it does improve consistency, as while it doesn't prevent
bucket exchange, it does prevent bucket pointer invalidation. So a
race in gathering cache free space statistics still can occur, but not
one that follows an invalid bucket pointer, if the mutex is held.
Submitted by: yongari
MFC after: 1 week
fs.first_object->flags & OBJ_NEEDGIANT test that was missed in an earlier
revision. This fixes mutex assertion failures in the debug.mpsafevm=0
case.
Reported by: ps
MFC after: 3 days
statistics via a binary structure stream:
- Add structure 'uma_stream_header', which defines a stream version,
definition of MAXCPUs used in the stream, and the number of zone
records in the stream.
- Add structure 'uma_type_header', which defines the name, alignment,
size, resource allocation limits, current pages allocated, preferred
bucket size, and central zone + keg statistics.
- Add structure 'uma_percpu_stat', which, for each per-CPU cache,
includes the number of allocations and frees, as well as the number
of free items in the cache.
- When the sysctl is queried, return a stream header, followed by a
series of type descriptions, each consisting of a type header
followed by a series of MAXCPUs uma_percpu_stat structures holding
per-CPU allocation information. Typical values of MAXCPU will be
1 (UP compiled kernel) and 16 (SMP compiled kernel).
This query mechanism allows user space monitoring tools to extract
memory allocation statistics in a machine-readable form, and to do so
at a per-CPU granularity, allowing monitoring of allocation patterns
across CPUs in order to better understand the distribution of work and
memory flow over multiple CPUs.
While here, also export the number of UMA zones as a sysctl
vm.uma_count, in order to assist in sizing user swpace buffers to
receive the stream.
A follow-up commit of libmemstat(3), a library to monitor kernel memory
allocation, will occur in the next few days. This change directly
supports converting netstat(1)'s "-mb" mode to using UMA-sourced stats
rather than separately maintained mbuf allocator statistics.
MFC after: 1 week
zone whenever it was moving buckets between the zone and the cache,
or when coalescing statistics across the CPU. Remove flushing of
statistics to the zone when coalescing statistics as part of sysctl,
as we won't be running on the right CPU to write to the cache
statistics.
Add a missed gathering of statistics: when uma_zalloc_internal()
does a special case allocation of a single item, make sure to update
the zone statistics to represent this. Previously this case wasn't
accounted for in user-visible statistics.
MFC after: 1 week
many regions checked again and again despite knowing the pages
contained were not usable and only satisfied the alignment constraints
This case was compounded, especially for large allocations, by the
practice of looping from the top of memory so as to keep out of the
important low-memory regions. While the old contigmalloc(9) has the
same problem, it is not as noticeable due to looping from the low
memory to high.
This degenerate case is fixed, as well as reversing the sense of the
rest of the loops within it, to provide a tremendous speed increase.
This makes the best case O(n * VM overhead) much more likely than the
worst case O(4 * VM overhead). For comparison, the worst case for old
contigmalloc would be O(5 * VM overhead) in addition to its strategy
of turning used memory into free being highly pessimal.
Also, fix a bug that in practice most likely couldn't have been triggered,
int the new contigmalloc(9): it walked backwards from the end of memory
without accounting for how many pages it needed. Potentially, nonexistant
pages could have been mapped. This hasn't occurred because the kernel
generally requests as its first contigmalloc(9) a single page.
Reported by: Nicolas Dehaine <nicko@stbernard.com>, wes
MFC After: 1 month
More testing by: Nicolas Dehaine <nicko@stbernard.com>, wes
vm_page's machine-dependent fields. Use this function in
vm_pageq_add_new_page() so that the vm_page's machine-dependent and
machine-independent fields are initialized at the same time.
Remove code from pmap_init() for initializing the vm_page's
machine-dependent fields.
Remove stale comments from pmap_init().
Eliminate the Boolean variable pmap_initialized from the alpha, amd64,
i386, and ia64 pmap implementations. Its use is no longer required
because of the above changes and earlier changes that result in physical
memory that is being mapped at initialization time being mapped without
pv entries.
Tested by: cognet, kensmith, marcel
underlying vnode requires Giant.
- In vm_fault only acquire Giant if the underlying object has NEEDSGIANT
set.
- In vm_object_shadow inherit the NEEDSGIANT flag from the backing object.
mutexes, which offers lower overhead on both UP and SMP. When allocating
from or freeing to the per-cpu cache, without INVARIANTS enabled, we now
no longer perform any mutex operations, which offers a 1%-3% performance
improvement in a variety of micro-benchmarks. We rely on critical
sections to prevent (a) preemption resulting in reentrant access to UMA on
a single CPU, and (b) migration of the thread during access. In the event
we need to go back to the zone for a new bucket, we release the critical
section to acquire the global zone mutex, and must re-acquire the critical
section and re-evaluate which cache we are accessing in case migration has
occured, or circumstances have changed in the current cache.
Per-CPU cache statistics are now gathered lock-free by the sysctl, which
can result in small races in statistics reporting for caches.
Reviewed by: bmilekic, jeff (somewhat)
Tested by: rwatson, kris, gnn, scottl, mike at sentex dot net, others
the number of entries in exec_map (maximum number of simultaneous execs
that can be handled by the kernel). The default value of 16 is
insufficient on heavily loaded machines (particularly SMP machines), and
if it is exceeded then executing further processes will generate a SIGABRT.
This is a workaround until a better solution can be implemented.
Reviewed by: alc
MFC after: 3 days
of physical addresses. The pages containing these physical addresses will
not be added to the free list and thus will effectively be ignored by the
VM system. This is mostly useful for the case when one knows of specific
physical addresses that have bit errors (such as from a memtest run) so
that one can blacklist the bad pages while waiting for the new sticks of
RAM to arrive. The physical addresses of any ignored pages are listed in
the message buffer as well.
MAP_SHARED so that the entry point gets executed un-conditionally.
This may be useful for security policies which want to perform access
control checks around run-time linking.
-add the mmap(2) flags argument to the check_vnode_mmap entry point
so that we can make access control decisions based on the type of
mapped object.
-update any dependent API around this parameter addition such as
function prototype modifications, entry point parameter additions
and the inclusion of sys/mman.h header file.
-Change the MLS, BIBA and LOMAC security policies so that subject
domination routines are not executed unless the type of mapping is
shared. This is done to maintain compatibility between the old
vm_mmap_vnode(9) and these policies.
Reviewed by: rwatson
MFC after: 1 month
down. If we have dirty pages, the putpages routine will need to know
what the vnode's object is so that it may write out dirty pages.
Pointy hat: phk
Found by: obrien
the type of object represented by the handle argument.
- Allow vm_mmap() to map device memory via cdev objects in addition to
vnodes and anonymous memory. Note that mmaping a cdev directly does not
currently perform any MAC checks like mapping a vnode does.
- Unbreak the DRM getbufs ioctl by having it call vm_mmap() directly on the
cdev the ioctl is acting on rather than trying to find a suitable vnode
to map from.
Reviewed by: alc, arch@
queues lock in vm_object_backing_scan(). Updates to the page's PG_BUSY
flag and busy field are synchronized by the containing object's lock.
Testing the page's hold_count and wire_count in vm_object_backing_scan()'s
OBSC_COLLAPSE_NOWAIT case is unnecessary. There is no reason why the held
or wired pages cannot be migrated to the shadow object.
Reviewed by: tegge
is inserted.
- In vm_page_remove() drop the backing vnode when the last page
is removed.
- Don't check the vnode to see if it must be reclaimed on every
call to vm_page_free_toq() as we only check it now when it is
actually required. This saves us two lock operations per call.
Sponsored by: Isilon Systems, Inc.
patch from kan@).
Pull bufobj_invalbuf() out of vinvalbuf() and make g_vfs call it on
close. This is not yet a generally safe function, but for this very
specific use it is safe. This solves the problem with buffers not
being flushed by unmount or after failed mount attempts.
statement from some files, so re-add it for the moment, until the
related legalese is sorted out. This change affects:
sys/kern/kern_mbuf.c
sys/vm/memguard.c
sys/vm/memguard.h
sys/vm/uma.h
sys/vm/uma_core.c
sys/vm/uma_dbg.c
sys/vm/uma_dbg.h
sys/vm/uma_int.h
UMA_ZONE_REFCNT and UMA_ZONE_MALLOC zones, as the page(s) undoubtedly
came from kmem_map for those two. Previously it would set it back
to NULL for UMA_ZONE_REFCNT zones and although this was probably not
fatal, it added MORE code for no reason.
an unused pageq queue reference in the page structure to stash a pointer
to the MemGuard FIFO. Using the page->object field caused problems
because when vm_map_protect() was called the second time to set
VM_PROT_DEFAULT back onto a set of pages in memguard_map, the protection
in the VM would be changed but the PMAP code would lazily not restore
the PG_RW bit on the underlying pages right away (see pmap_protect()).
So when a page fault finally occured and the VM noticed the faulting
address corresponds to a page that _does_ have write access now, it
would then call into PMAP to set back PG_RW (i386 case being discussed
here). However, before it got to do that, an assertion on the object
lock not being owned would get triggered, as the object of the faulting
page would need to be locked but was overloaded by MemGuard. This is
precisely why MemGuard cannot overload page->object.
Submitted by: Alan Cox (alc@)
a little bit of complexity but performance requirements lacking (this is
a debugging allocator after all), it's really not too bad (still
only 317 lines).
Also add an additional check to help catch really weird 3-threads-involved
races: make memguard_free() write to the first page handed back, always,
before it does anything else.
Note that there is still a problem in VM+PMAP (specifically with
vm_map_protect) w.r.t. MemGuard uses it, but this will be fixed shortly
and this change stands on its own.
