user-mode lock manager, build a kernel with the NFSLOCKD option and
add '-k' to 'rpc_lockd_flags' in rc.conf.
Highlights include:
* Thread-safe kernel RPC client - many threads can use the same RPC
client handle safely with replies being de-multiplexed at the socket
upcall (typically driven directly by the NIC interrupt) and handed
off to whichever thread matches the reply. For UDP sockets, many RPC
clients can share the same socket. This allows the use of a single
privileged UDP port number to talk to an arbitrary number of remote
hosts.
* Single-threaded kernel RPC server. Adding support for multi-threaded
server would be relatively straightforward and would follow
approximately the Solaris KPI. A single thread should be sufficient
for the NLM since it should rarely block in normal operation.
* Kernel mode NLM server supporting cancel requests and granted
callbacks. I've tested the NLM server reasonably extensively - it
passes both my own tests and the NFS Connectathon locking tests
running on Solaris, Mac OS X and Ubuntu Linux.
* Userland NLM client supported. While the NLM server doesn't have
support for the local NFS client's locking needs, it does have to
field async replies and granted callbacks from remote NLMs that the
local client has contacted. We relay these replies to the userland
rpc.lockd over a local domain RPC socket.
* Robust deadlock detection for the local lock manager. In particular
it will detect deadlocks caused by a lock request that covers more
than one blocking request. As required by the NLM protocol, all
deadlock detection happens synchronously - a user is guaranteed that
if a lock request isn't rejected immediately, the lock will
eventually be granted. The old system allowed for a 'deferred
deadlock' condition where a blocked lock request could wake up and
find that some other deadlock-causing lock owner had beaten them to
the lock.
* Since both local and remote locks are managed by the same kernel
locking code, local and remote processes can safely use file locks
for mutual exclusion. Local processes have no fairness advantage
compared to remote processes when contending to lock a region that
has just been unlocked - the local lock manager enforces a strict
first-come first-served model for both local and remote lockers.
Sponsored by: Isilon Systems
PR: 95247 107555 115524 116679
MFC after: 2 weeks
after each SYSINIT() macro invocation. This makes a number of
lightweight C parsers much happier with the FreeBSD kernel
source, including cflow's prcc and lxr.
MFC after: 1 month
Discussed with: imp, rink
always curthread.
As KPI gets broken by this patch, manpages and __FreeBSD_version will be
updated by further commits.
Tested by: Andrea Barberio <insomniac at slackware dot it>
namespace in order to handle lockmgr fields in a controlled way instead
than spreading all around bogus stubs:
- VN_LOCK_AREC() allows lock recursion for a specified vnode
- VN_LOCK_ASHARE() allows lock sharing for a specified vnode
In FFS land:
- BUF_AREC() allows lock recursion for a specified buffer lock
- BUF_NOREC() disallows recursion for a specified buffer lock
Side note: union_subr.c::unionfs_node_update() is the only other function
directly handling lockmgr fields. As this is not simple to fix, it has
been left behind as "sole" exception.
doesn't overflow in arc.c in this check:
if (kmem_used() > (kmem_size() * 4) / 5)
return (1);
With this bug ZFS almost doesn't cache.
Only 32bit machines are affected that have vm.kmem_size set to values >=1GB.
Reported by: David Taylor <davidt@yadt.co.uk>
conjuction with 'thread' argument passing which is always curthread.
Remove the unuseful extra-argument and pass explicitly curthread to lower
layer functions, when necessary.
KPI results broken by this change, which should affect several ports, so
version bumping and manpage update will be further committed.
Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
Remove this argument and pass curthread directly to underlying
VOP_LOCK1() VFS method. This modify makes the code cleaner and in
particular remove an annoying dependence helping next lockmgr() cleanup.
KPI results, obviously, changed.
Manpage and FreeBSD_version will be updated through further commits.
As a side note, would be valuable to say that next commits will address
a similar cleanup about VFS methods, in particular vop_lock1 and
vop_unlock.
Tested by: Diego Sardina <siarodx at gmail dot com>,
Andrea Di Pasquale <whyx dot it at gmail dot com>
characters (mostly "&"). Because top(1) shows only first six characters of
wait channel, without this change we saw only one meaningful character.
Requested by: kris & others
MFC after: 1 week
(BIO_WRITE and BIO_FLUSH) as it is done is Solaris. The difference is
that Solaris calls it only for sync requests, but we can't say in GEOM
is the request is sync or async, so we do it for every request.
MFC after: 1 week
to kproc_xxx as they actually make whole processes.
Thos makes way for us to add REAL kthread_create() and friends
that actually make theads. it turns out that most of these
calls actually end up being moved back to the thread version
when it's added. but we need to make this cosmetic change first.
I'd LOVE to do this rename in 7.0 so that we can eventually MFC the
new kthread_xxx() calls.
value, so we don't run out of KVA. The default vnodes limit fits better for
UFS, but ZFS allocated more file system specific memory for a vnode than UFS.
Don't touch vnodes limit if we detect it was tuned by system administrator
and restore original value when ZFS is unloaded.
This isn't final fix, but before we implement something better, this will
help to stabilize ZFS under heavy load on i386.
Approved by: re (bmah)
some cases, move to priv_check() if it was an operation on a thread and
no other flags were present.
Eliminate caller-side jail exception checking (also now-unused); jail
privilege exception code now goes solely in kern_jail.c.
We can't yet eliminate suser() due to some cases in the KAME code where
a privilege check is performed and then used in many different deferred
paths. Do, however, move those prototypes to priv.h.
Reviewed by: csjp
Obtained from: TrustedBSD Project
implementing some of them using existing ones.
- Allow to compile ZFS on all archs and use atomic operations surrounded
by global mutex on archs we don't have or can't have all atomic
operations needed by ZFS.
sysctl_handle_int is not sizeof the int type you want to export.
The type must always be an int or an unsigned int.
Remove the instances where a sizeof(variable) is passed to stop
people accidently cut and pasting these examples.
In a few places this was sysctl_handle_int was being used on 64 bit
types, which would truncate the value to be exported. In these
cases use sysctl_handle_quad to export them and change the format
to Q so that sysctl(1) can still print them.
1. Pass locking flags to VFS_ROOT().
2. Check v_mountedhere while the vnode is locked.
3. Always return locked vnode on success.
Change 1 fixes problem reported by Stephen M. Rumble - after
zfs_vfsops.c,1.9 change, zfs_root() no longer locks the vnode
unconditionally and traverse() didn't pass right lock type to
VFS_ROOT(). The result was that kernel paniced when .zfs/ directory
was accessed via NFS.
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)
based on individual fields beeing set. This doesn't work for setattr replay,
because va_type is set there, so we add AT_TYPE flag to va_mask, which won't
be accepted by zfs_setattr().
Reported by: kris
debug is turned off, initialize locks with NOWITNESS flag.
At some point I'll get back to them, we would probably need BLESSING
functionality, which is currently turned off by default.
initialization is complete. This fixes some root-on-ZFS
configurations.
Reported by: Bruno Damour <freebsd.ruomad@free.fr>
Tested by: Bruno Damour <freebsd.ruomad@free.fr>
and show up with different names: first try to open provider using
remembered name and compare its ident, if equal, this is our provider,
if not equal or there is no provider with such name, find provider with
remembered ident and don't care about the name.
non-mapped data as possible at once and not page-by-page. Which this change we
combain I/Os, but also saves many VM_OBJECT_UNLOCK()/VM_OBJECT_LOCK()
operations.
Simple 'fsx -l 33554432 -o 524288 -N 10000 /tank/fsx' test shows ~23%
performance increase.
- vm_page_undirty() is enough (instead of vm_page_set_validclean()), but it has
to be called before we write the data in case someone makes page dirty after
our write, but before our vm_page_undirty() call.
- Always dmu_write, not matter if uiomove() succeeded, because it could
partially be ok and we would lose some changes.
All good ideas from: ups
in all other file system on FreeBSD (instead from inactive() method).
A nice side-effect of this change, except that it speedups file system
when mmaped file are often open/closed, is that it makes FreeBSD's
namecache work:)
This fixes slow operations on mmaped files, because without this fix,
pages were written to disk multiple times.
If one is looking for even greater speed up for such operation, he should
disable ZIL (by setting vfs.zfs.zil_disable to 1 in /boot/loader.conf).
Disabling ZIL makes fsx run ~9 times faster.
o Fix linewrap issues.
o Fix two typos (s/Recomended/Recommended/ and s/tunning/tuning/)
o Remove a couple of extra instances of the word "of".
o Update names of kmem_size variables.
Approved by: pjd
- Move FreeBSD-specific code to zfs_freebsd_*() functions in zfs_vnops.c
and keep original functions as similar to vendor's code as possible.
- Add various includes back, now that we have them.
macro, as za_first_integer field also contains type. This should be fixed in
ZFS itself, but this bug is not visible on Solaris, because there, type is
not stored in za_first_integer. On the other hand it will be visible on
MacOS X.
Reported by: Barry Pederson <bp@barryp.org>
@118370 Correct typo.
@118371 Integrate changes from vendor.
@118491 Show backtrace on unexpected code paths.
@118494 Integrate changes from vendor.
@118504 Fix sendfile(2). I had two ways of fixing it:
1. Fixing sendfile(2) itself to use VOP_GETPAGES() instead of
hacking around with vn_rdwr(UIO_NOCOPY), which was suggested
by ups.
2. Modify ZFS behaviour to handle this special case.
Although 1 is more correct, I've choosen 2, because hack from 1
have a side-effect of beeing faster - it reads ahead MAXBSIZE
bytes instead of reading page by page. This is not easy to implement
with VOP_GETPAGES(), at least not for me in this very moment.
Reported by: Andrey V. Elsukov <bu7cher@yandex.ru>
@118525 Reorganize the code to reduce diff.
@118526 This code path is expected. It is simply when file is opened with
O_FSYNC flag.
Reported by: kris
Reported by: Michal Suszko <dry@dry.pl>
on a snapshot directory:
- Remove PRIV_VFS_MOUNT check - regular users can mount snapshots
via lookups on snapshot directory.
- Reset mount credential to kcred, so user won't be able to unmount
the snapshot.
- Reset owner uid.
- Unlock vnode in case of a failure.
Reported by: simokawa
This fixes stange panics when listing .zfs/snapshot/ directory for me.
Reported by: simokawa
Reported by: Johan Hendriks <Johan@double-l.nl>
- Hide cache_purge() under FREEBSD_NAMECACHE like in other files.
- Protect mnt_flag with mount interlock.
unload instead of returning EBUSY. This check tells if there are mounted
ZFS file systems or not. We can't unload if there are mounted file systems.
Reported by: Andrey V. Elsukov <bu7cher@yandex.ru>
- Reduce default number of spa_zio_* threads to N*spa_zio_issue
plus N*spa_zio_intr threads per ZIO type, where N is the number
of CPUs.
- Put ZIO type number in thread's name.
than 2GB of RAM. This was because our physmem is long and 'physmem*PAGESIZE'
can be negative for more than 2GB of memory.
Reported by: Andrey V. Elsukov <bu7cher@yandex.ru>
It is not yet tested by Andrey, so there can be other problems, but this
was definiately a bug, so I'm committing a fix now.
- Tune number of namecache entires better (based on desiredvnodes).
- Handle vfs_lowvnodes event by releasing requested number of name cache
entries, but no less than 5%.
Reported by: simokawa
popular names. Hence:
- comment current index() and rindex() functions, as these serve the same
functionality as, respectively, strchr() and strrchr() from userland;
- add inlined version of strchr() and strrchr(), as we tend to use them more
often;
- remove str[r]chr() definitions from ZFS code;
Reviewed by: pjd
Approved by: cognet (mentor)
- Allow to shrink ARC down to 16MB (instead of 64MB).
- Set arc_max to 1/2 of kmem_map by default.
- Start freeing things earlier when low memory situation is detected.
- Serialize execution of arc_lowmem().
I decided to setup minimum ZFS memory requirements to 512MB of RAM and 256MB of
kmem_map size. If there is less RAM or kmem_map, a warning will be printed.
World is cruel, be no better. In other words: modern file system requires
modern hardware:)
From ZFS administration guide:
"Currently the minimum amount of memory recommended to install a Solaris
system is 512 Mbytes. However, for good ZFS performance, at least one
Gbyte or more of memory is recommended."
ZFS file system was ported from OpenSolaris operating system. The code in under
CDDL license.
I'd like to thank all SUN developers that created this great piece of software.
Supported by: Wheel LTD (http://www.wheel.pl/)
Supported by: The FreeBSD Foundation (http://www.freebsdfoundation.org/)
Supported by: Sentex (http://www.sentex.net/)
