do not map the b_pages pages into buffer_map KVA. The use of the
unmapped buffers eliminate the need to perform TLB shootdown for
mapping on the buffer creation and reuse, greatly reducing the amount
of IPIs for shootdown on big-SMP machines and eliminating up to 25-30%
of the system time on i/o intensive workloads.
The unmapped buffer should be explicitely requested by the GB_UNMAPPED
flag by the consumer. For unmapped buffer, no KVA reservation is
performed at all. The consumer might request unmapped buffer which
does have a KVA reserve, to manually map it without recursing into
buffer cache and blocking, with the GB_KVAALLOC flag.
When the mapped buffer is requested and unmapped buffer already
exists, the cache performs an upgrade, possibly reusing the KVA
reservation.
Unmapped buffer is translated into unmapped bio in g_vfs_strategy().
Unmapped bio carry a pointer to the vm_page_t array, offset and length
instead of the data pointer. The provider which processes the bio
should explicitely specify a readiness to accept unmapped bio,
otherwise g_down geom thread performs the transient upgrade of the bio
request by mapping the pages into the new bio_transient_map KVA
submap.
The bio_transient_map submap claims up to 10% of the buffer map, and
the total buffer_map + bio_transient_map KVA usage stays the
same. Still, it could be manually tuned by kern.bio_transient_maxcnt
tunable, in the units of the transient mappings. Eventually, the
bio_transient_map could be removed after all geom classes and drivers
can accept unmapped i/o requests.
Unmapped support can be turned off by the vfs.unmapped_buf_allowed
tunable, disabling which makes the buffer (or cluster) creation
requests to ignore GB_UNMAPPED and GB_KVAALLOC flags. Unmapped
buffers are only enabled by default on the architectures where
pmap_copy_page() was implemented and tested.
In the rework, filesystem metadata is not the subject to maxbufspace
limit anymore. Since the metadata buffers are always mapped, the
buffers still have to fit into the buffer map, which provides a
reasonable (but practically unreachable) upper bound on it. The
non-metadata buffer allocations, both mapped and unmapped, is
accounted against maxbufspace, as before. Effectively, this means that
the maxbufspace is forced on mapped and unmapped buffers separately.
The pre-patch bufspace limiting code did not worked, because
buffer_map fragmentation does not allow the limit to be reached.
By Jeff Roberson request, the getnewbuf() function was split into
smaller single-purpose functions.
Sponsored by: The FreeBSD Foundation
Discussed with: jeff (previous version)
Tested by: pho, scottl (previous version), jhb, bf
MFC after: 2 weeks
Replace the per-object resident and cached pages splay tree with a
path-compressed multi-digit radix trie.
Along with this, switch also the x86-specific handling of idle page
tables to using the radix trie.
This change is supposed to do the following:
- Allowing the acquisition of read locking for lookup operations of the
resident/cached pages collections as the per-vm_page_t splay iterators
are now removed.
- Increase the scalability of the operations on the page collections.
The radix trie does rely on the consumers locking to ensure atomicity of
its operations. In order to avoid deadlocks the bisection nodes are
pre-allocated in the UMA zone. This can be done safely because the
algorithm needs at maximum one new node per insert which means the
maximum number of the desired nodes is the number of available physical
frames themselves. However, not all the times a new bisection node is
really needed.
The radix trie implements path-compression because UFS indirect blocks
can lead to several objects with a very sparse trie, increasing the number
of levels to usually scan. It also helps in the nodes pre-fetching by
introducing the single node per-insert property.
This code is not generalized (yet) because of the possible loss of
performance by having much of the sizes in play configurable.
However, efforts to make this code more general and then reusable in
further different consumers might be really done.
The only KPI change is the removal of the function vm_page_splay() which
is now reaped.
The only KBI change, instead, is the removal of the left/right iterators
from struct vm_page, which are now reaped.
Further technical notes broken into mealpieces can be retrieved from the
svn branch:
http://svn.freebsd.org/base/user/attilio/vmcontention/
Sponsored by: EMC / Isilon storage division
In collaboration with: alc, jeff
Tested by: flo, pho, jhb, davide
Tested by: ian (arm)
Tested by: andreast (powerpc)
pages around, taking array of vm_page_t both for source and
destination. Starting offsets and total transfer size are specified.
The function implements optimal algorithm for copying using the
platform-specific optimizations. For instance, on the architectures
were the direct map is available, no transient mappings are created,
for i386 the per-cpu ephemeral page frame is used. The code was
typically borrowed from the pmap_copy_page() for the same
architecture.
Only i386/amd64, powerpc aim and arm/arm-v6 implementations were
tested at the time of commit. High-level code, not committed yet to
the tree, ensures that the use of the function is only allowed after
explicit enablement.
For sparc64, the existing code has known issues and a stab is added
instead, to allow the kernel linking.
Sponsored by: The FreeBSD Foundation
Tested by: pho (i386, amd64), scottl (amd64), ian (arm and arm-v6)
MFC after: 2 weeks
future further optimizations where the vm_object lock will be held
in read mode most of the time the page cache resident pool of pages
are accessed for reading purposes.
The change is mostly mechanical but few notes are reported:
* The KPI changes as follow:
- VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK()
- VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK()
- VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK()
- VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED()
(in order to avoid visibility of implementation details)
- The read-mode operations are added:
VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(),
VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED()
* The vm/vm_pager.h namespace pollution avoidance (forcing requiring
sys/mutex.h in consumers directly to cater its inlining functions
using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h
consumers now must include also sys/rwlock.h.
* zfs requires a quite convoluted fix to include FreeBSD rwlocks into
the compat layer because the name clash between FreeBSD and solaris
versions must be avoided.
At this purpose zfs redefines the vm_object locking functions
directly, isolating the FreeBSD components in specific compat stubs.
The KPI results heavilly broken by this commit. Thirdy part ports must
be updated accordingly (I can think off-hand of VirtualBox, for example).
Sponsored by: EMC / Isilon storage division
Reviewed by: jeff
Reviewed by: pjd (ZFS specific review)
Discussed with: alc
Tested by: pho
Introduce a new KPI that verifies if the page cache is empty for a
specified vm_object. This KPI does not make assumptions about the
locking in order to be used also for building assertions at init and
destroy time.
It is mostly used to hide implementation details of the page cache.
Sponsored by: EMC / Isilon storage division
Reviewed by: jeff
Reviewed by: alc (vm_radix based version)
Tested by: flo, pho, jhb, davide
from being indirectly called via cpu_startup()+vm_ksubmap_init().
The boot order position remains the same at SI_SUB_CPU.
Allocation of the callout array is changed to stardard kernel malloc
from a slightly obscure direct kernel_map allocation.
kern_timeout_callwheel_alloc() is renamed to callout_callwheel_init()
to better describe its purpose.
kern_timeout_callwheel_init() is removed simplifying the per-cpu
initialization.
Reviewed by: davide
As vm objects are type-stable there is no need to initialize the
resident splay tree pointer and the cache splay tree pointer in
_vm_object_allocate() but this could be done in the init UMA zone
handler.
The destructor UMA zone handler, will further check if the condition is
retained at every destruction and catch for bugs.
Sponsored by: EMC / Isilon storage division
Submitted by: alc
valid if the flag OBJ_COLORED is set. Since _vm_object_allocate()
doesn't set this flag, it needn't initialize pg_color.
Sponsored by: EMC / Isilon Storage Division
- Capability is no longer separate descriptor type. Now every descriptor
has set of its own capability rights.
- The cap_new(2) system call is left, but it is no longer documented and
should not be used in new code.
- The new syscall cap_rights_limit(2) should be used instead of
cap_new(2), which limits capability rights of the given descriptor
without creating a new one.
- The cap_getrights(2) syscall is renamed to cap_rights_get(2).
- If CAP_IOCTL capability right is present we can further reduce allowed
ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed
ioctls can be retrived with cap_ioctls_get(2) syscall.
- If CAP_FCNTL capability right is present we can further reduce fcntls
that can be used with the new cap_fcntls_limit(2) syscall and retrive
them with cap_fcntls_get(2).
- To support ioctl and fcntl white-listing the filedesc structure was
heavly modified.
- The audit subsystem, kdump and procstat tools were updated to
recognize new syscalls.
- Capability rights were revised and eventhough I tried hard to provide
backward API and ABI compatibility there are some incompatible changes
that are described in detail below:
CAP_CREATE old behaviour:
- Allow for openat(2)+O_CREAT.
- Allow for linkat(2).
- Allow for symlinkat(2).
CAP_CREATE new behaviour:
- Allow for openat(2)+O_CREAT.
Added CAP_LINKAT:
- Allow for linkat(2). ABI: Reuses CAP_RMDIR bit.
- Allow to be target for renameat(2).
Added CAP_SYMLINKAT:
- Allow for symlinkat(2).
Removed CAP_DELETE. Old behaviour:
- Allow for unlinkat(2) when removing non-directory object.
- Allow to be source for renameat(2).
Removed CAP_RMDIR. Old behaviour:
- Allow for unlinkat(2) when removing directory.
Added CAP_RENAMEAT:
- Required for source directory for the renameat(2) syscall.
Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR):
- Allow for unlinkat(2) on any object.
- Required if target of renameat(2) exists and will be removed by this
call.
Removed CAP_MAPEXEC.
CAP_MMAP old behaviour:
- Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and
PROT_WRITE.
CAP_MMAP new behaviour:
- Allow for mmap(2)+PROT_NONE.
Added CAP_MMAP_R:
- Allow for mmap(PROT_READ).
Added CAP_MMAP_W:
- Allow for mmap(PROT_WRITE).
Added CAP_MMAP_X:
- Allow for mmap(PROT_EXEC).
Added CAP_MMAP_RW:
- Allow for mmap(PROT_READ | PROT_WRITE).
Added CAP_MMAP_RX:
- Allow for mmap(PROT_READ | PROT_EXEC).
Added CAP_MMAP_WX:
- Allow for mmap(PROT_WRITE | PROT_EXEC).
Added CAP_MMAP_RWX:
- Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC).
Renamed CAP_MKDIR to CAP_MKDIRAT.
Renamed CAP_MKFIFO to CAP_MKFIFOAT.
Renamed CAP_MKNODE to CAP_MKNODEAT.
CAP_READ old behaviour:
- Allow pread(2).
- Disallow read(2), readv(2) (if there is no CAP_SEEK).
CAP_READ new behaviour:
- Allow read(2), readv(2).
- Disallow pread(2) (CAP_SEEK was also required).
CAP_WRITE old behaviour:
- Allow pwrite(2).
- Disallow write(2), writev(2) (if there is no CAP_SEEK).
CAP_WRITE new behaviour:
- Allow write(2), writev(2).
- Disallow pwrite(2) (CAP_SEEK was also required).
Added convinient defines:
#define CAP_PREAD (CAP_SEEK | CAP_READ)
#define CAP_PWRITE (CAP_SEEK | CAP_WRITE)
#define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ)
#define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE)
#define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL)
#define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W)
#define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X)
#define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X)
#define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X)
#define CAP_RECV CAP_READ
#define CAP_SEND CAP_WRITE
#define CAP_SOCK_CLIENT \
(CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \
CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN)
#define CAP_SOCK_SERVER \
(CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \
CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \
CAP_SETSOCKOPT | CAP_SHUTDOWN)
Added defines for backward API compatibility:
#define CAP_MAPEXEC CAP_MMAP_X
#define CAP_DELETE CAP_UNLINKAT
#define CAP_MKDIR CAP_MKDIRAT
#define CAP_RMDIR CAP_UNLINKAT
#define CAP_MKFIFO CAP_MKFIFOAT
#define CAP_MKNOD CAP_MKNODAT
#define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER)
Sponsored by: The FreeBSD Foundation
Reviewed by: Christoph Mallon <christoph.mallon@gmx.de>
Many aspects discussed with: rwatson, benl, jonathan
ABI compatibility discussed with: kib
VM_OBJECT_LOCKED() macro is only used to implement a custom version
of lock assertions right now (which likely spread out thanks to
copy and paste).
Remove it and implement actual assertions.
Sponsored by: EMC / Isilon storage division
Reviewed by: alc
Tested by: pho
Replace the sub-optimal uma_zone_set_obj() primitive with more modern
uma_zone_reserve_kva(). The new primitive reserves before hand
the necessary KVA space to cater the zone allocations and allocates pages
with ALLOC_NOOBJ. More specifically:
- uma_zone_reserve_kva() does not need an object to cater the backend
allocator.
- uma_zone_reserve_kva() can cater M_WAITOK requests, in order to
serve zones which need to do uma_prealloc() too.
- When possible, uma_zone_reserve_kva() uses directly the direct-mapping
by uma_small_alloc() rather than relying on the KVA / offset
combination.
The removal of the object attribute allows 2 further changes:
1) _vm_object_allocate() becomes static within vm_object.c
2) VM_OBJECT_LOCK_INIT() is removed. This function is replaced by
direct calls to mtx_init() as there is no need to export it anymore
and the calls aren't either homogeneous anymore: there are now small
differences between arguments passed to mtx_init().
Sponsored by: EMC / Isilon storage division
Reviewed by: alc (which also offered almost all the comments)
Tested by: pho, jhb, davide
macro VM_OBJECT_SLEEP().
This hides some implementation details like the usage of the msleep()
primitive and the necessity to access to the lock address directly.
For this reason VM_OBJECT_MTX() macro is now retired.
Sponsored by: EMC / Isilon storage division
Reviewed by: alc
Tested by: pho
machine to another. Therefore, VM_MAX_KERNEL_ADDRESS can't be a constant.
Instead, #define it to be a variable, vm_max_kernel_address, just like we
do on sparc64.
Reviewed by: kib
Tested by: ian
in devfs if a particular race condition is hit in the device pager
code.
This was a side effect of change 227530 which changed the device
pager interface to call a new destructor routine for the cdev.
That destructor routine, old_dev_pager_dtor(), takes a VM object
handle.
The object handle is cast to a struct cdev *, and passed into
dev_rel().
That works in most cases, except the case in cdev_pager_allocate()
where there is a race condition between two threads allocating an
object backed by the same device. The loser of the race
deallocates its object at the end of the function.
The problem is that before inserting the object into the
dev_pager_object_list, the object's handle is changed from the
struct cdev pointer to the object's own address. This is to avoid
conflicts with the winner of the race, which already inserted an
object in the list with a handle that is a pointer to the same cdev
structure.
The object is then passed to vm_object_deallocate(), and eventually
makes its way down to old_dev_pager_dtor(). That function passes
the handle pointer (which is actually a VM object, not a struct
cdev as usual) into dev_rel(). dev_rel() decrements the reference
count in the assumed struct cdev (which happens to be 0), and
that triggers the assertion in dev_rel() that the reference count
is greater than or equal to 0.
The fix is to add a cdev pointer to the VM object, and use that
pointer when calling the cdev_pg_dtor() routine.
vm_object.h: Add a struct cdev pointer to the VM object
structure.
device_pager.c: In cdev_pager_allocate(), populate the new cdev
pointer.
In dev_pager_dealloc(), use the new cdev pointer
when calling the object's cdev_pg_dtor() routine.
Reviewed by: kib
Sponsored by: Spectra Logic Corporation
MFC after: 1 week
similar changes had to be made in various places throughout the machine-
independent virtual memory layer to support the new vm object type.
However, in most of these places, it's actually not the type of the vm
object that matters to us but instead certain attributes of its pages.
For example, OBJT_DEVICE, OBJT_MGTDEVICE, and OBJT_SG objects contain
fictitious pages. In other words, in most of these places, we were
testing the vm object's type to determine if it contained fictitious (or
unmanaged) pages.
To both simplify the code in these places and make the addition of future
vm object types easier, this change introduces two new vm object flags
that describe attributes of the vm object's pages, specifically, whether
they are fictitious or unmanaged.
Reviewed and tested by: kib
will be printed once the given zone becomes full and cannot allocate an
item. The warning will not be printed more often than every five minutes.
All UMA warnings can be globally turned off by setting sysctl/tunable
vm.zone_warnings to 0.
Discussed on: arch
Obtained from: WHEEL Systems
MFC after: 2 weeks
Add detail to the comment describing this function. In particular,
describe what MAP_PREFAULT_PARTIAL does.
Eliminate the abrupt change in behavior when the specified address range
grows from MAX_INIT_PT pages to MAX_INIT_PT plus one pages. Instead of
doing nothing, i.e., preloading no mappings whatsoever, map any resident
pages that fall within the start of the specified address range, i.e.,
[addr, addr + ulmin(size, ptoa(MAX_INIT_PT))).
Long ago, the vm object's list of resident pages was not ordered, so
this function had to choose between probing the global hash table of
all resident pages and iterating over the vm object's unordered list of
resident pages. Now, the list is ordered, so there is no reason for
MAP_PREFAULT_PARTIAL to be concerned with the vm object's count of
resident changes.
MFC after: 14 days
sleep, and perform the page allocations with VM_ALLOC_SYSTEM
class. Previously, the allocation was also allowed to completely drain
the reserve of the free pages, being translated to VM_ALLOC_INTERRUPT
request class for vm_page_alloc() and similar functions.
Allow the caller of malloc* to request the 'deep drain' semantic by
providing M_USE_RESERVE flag, now translated to VM_ALLOC_INTERRUPT
class. Previously, it resulted in less aggressive VM_ALLOC_SYSTEM
allocation class.
Centralize the translation of the M_* malloc(9) flags in the single
inline function malloc2vm_flags().
Discussion started by: "Sears, Steven" <Steven.Sears@netapp.com>
Reviewed by: alc, mdf (previous version)
Tested by: pho (previous version)
MFC after: 2 weeks
- Check that an argument is always available, otherwise current map
printing before to recurse is garbage.
- Spit out a message if an argument is not provided.
- Remove unread nlines variable.
- Use an explicit recursive function, disassociated from the
DB_SHOW_COMMAND() body, in order to make clear prototype and recursion
of the above mentioned function. The code results now much less
obscure.
Submitted by: gianni
was still possible to open for write from the lower filesystem. There
is a symmetric situation where the binary could already has file
descriptors opened for write, but it can be executed from the nullfs
overlay.
Handle the issue by passing one v_writecount reference to the lower
vnode if nullfs vnode has non-zero v_writecount. Note that only one
write reference can be donated, since nullfs only keeps one use
reference on the lower vnode. Always use the lower vnode v_writecount
for the checks.
Introduce the VOP_GET_WRITECOUNT to read v_writecount, which is
currently always bypassed to the lower vnode, and VOP_ADD_WRITECOUNT
to manipulate the v_writecount value, which manages a single bypass
reference to the lower vnode. Caling the VOPs instead of directly
accessing v_writecount provide the fix described in the previous
paragraph.
Tested by: pho
MFC after: 3 weeks
the call to pmap_remove_all() within vm_page_cache() is usually redundant.
This change eliminates that call to pmap_remove_all() and introduces a
call to pmap_remove_all() before vm_page_cache() in the one place where
it didn't already exist.
When iterating over a paging queue, if the object containing the current
page has a zero reference count, then the page can't have any managed
mappings. So, a call to pmap_remove_all() is pointless.
Change a panic() call in vm_page_cache() to a KASSERT().
MFC after: 6 weeks
cache line in order to avoid manual frobbing but using
struct mtx_padalign.
The sole exception being nvme and sxfge drivers, where the author
redefined CACHE_LINE_SIZE manually, so they need to be analyzed and
dealt with separately.
Reviwed by: jimharris, alc
