critical_enter() and critical_exit() are now solely a mechanism for
deferring kernel preemptions. They no longer have any affect on
interrupts. This means that standalone critical sections are now very
cheap as they are simply unlocked integer increments and decrements for the
common case.
Spin mutexes now use a separate KPI implemented in MD code: spinlock_enter()
and spinlock_exit(). This KPI is responsible for providing whatever MD
guarantees are needed to ensure that a thread holding a spin lock won't
be preempted by any other code that will try to lock the same lock. For
now all archs continue to block interrupts in a "spinlock section" as they
did formerly in all critical sections. Note that I've also taken this
opportunity to push a few things into MD code rather than MI. For example,
critical_fork_exit() no longer exists. Instead, MD code ensures that new
threads have the correct state when they are created. Also, we no longer
try to fixup the idlethreads for APs in MI code. Instead, each arch sets
the initial curthread and adjusts the state of the idle thread it borrows
in order to perform the initial context switch.
This change is largely a big NOP, but the cleaner separation it provides
will allow for more efficient alternative locking schemes in other parts
of the kernel (bare critical sections rather than per-CPU spin mutexes
for per-CPU data for example).
Reviewed by: grehan, cognet, arch@, others
Tested on: i386, alpha, sparc64, powerpc, arm, possibly more
to see what features they may support before calling identify/probe/attach.
This is necessary because the ACPI 3.0 spec requires driver support be
advertised before running any methods. For now, the flags are as specified
in for the _PDC and _OSC methods but we can support private flags as needed.
Add an implementation of this for acpi_cpu. It checks all its children
(notably cpufreq drivers) and calls the _PDC method to report the results.
instances in a given devclass. This is useful for systems that want to
call code in driver static methods, similar to device_identify().
Reviewed by: dfr
MFC after: 2 weeks
one to become available for one second and then return ENFILE. We
can run out of vnodes, and there must be a hard limit because without
one we can quickly run out of KVA on x86. Presently the system can
deadlock if there are maxvnodes directories in the namecache. The
original 4.x BSD behavior was to return ENFILE if we reached the max,
but 4.x BSD did not have the vnlru proc so it was less profitable to
wait.
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
in a devclass. All the other uses of maxunit are correct and this one was
safe since it checks the return value of devclass_get_device(), which would
always say that the highest unit device doesn't exist.
Reviewed by: dfr
MFC after: 3 days
completed I/O requests here.
- First allocate all needed bios, so if any of allocations fail, we can
free memory before sending any I/O requests down.
Reported by: Pawel Malachowski
MFC after: 3 days
- Don't intermingle direct calls to lockmgr and indirect calls through
VOPs. This will be important in the future.
- Dont lock the devvp's interlock just to release it on the next line by
passing LK_INTERLOCK to lockmgr.
- Restructure ffs_snapshot_unmount so we don't call free() with the
devvp's interlock locked.
because it may change identities while we're sleeping on the lock.
Otherwise we may bail out of ffs_sync() early due to an error from
deadfs.
- Collapse a VOP_UNLOCK, vrele into a single vput().
two bugs.
- ffs_disk_prewrite was pulling the vp from the buf and checking for
COPYONWRITE, when really it wanted the vp from the bufobj that we're
writing to, which is the devvp. This lead to us skipping the copy on
write to all file data, which significantly broke snapshots for the
last few months.
- When the SOFTUPDATES option was not included in the kernel config we
would also skip the copy on write check, which would effectively disable
snapshots.
- Remove an invalid mp_fixme().
Debugging tips from: mckusick
Reported by: iedowse, others
Discussed with: phk
generate dirty bufs even with a locked vnode, 100 retries is not that
many. This should probably change from a retry count to an abort when
we are no longer cleaning any buffers.
- Don't call vprint() while we still hold the vnode locked. Move the call
to later in the function.
- Clean up a comment.
implementations inspired by the ones in DragonFly. Unlike the
DragonFly versions, these have a small data cache footprint, and my
tests show that they're never slower than the old code except when the
charset or the span is 0 or 1 characters. This implementation is
generally faster than DragonFly until either the charset or the span
gets in the ballpark of 32 to 64 characters.
compiler features tests. This is ok, since machine/ieeefp.h is an internal
interface. But floatingpoint.h is a public interface and some ports use it,
so include sys/cdefs.h in the amd64 and i386 version of floatingpoint.h.
Note: some architectures don't provide recursive inclusion protection in
floatingpoint.h, namely alpha and ia64. Except for this part and now the
include of sys/cdefs.h, all those files are equal (from a compiler POV),
so they could be moved to only one version in src/include/.
Approved by: joerg
these at the moment, but applications that test for them will now
have a better chance of compiling.
I have intentionally omitted errnos that are only good for STREAMS,
since apps that use STREAMS won't compile anyway. The exception is
EPROTO, which was apparently intended for STREAMS, but worth having
anyway because Linux (mis)uses it for other things.
specific code will migrate to these files to augment or replace the
version in i386/include and/or i386/linux. This should, in the
fullness of time, allow many of the #ifdef PC98 in the tree.
# These files are in the public domain because there is insufficient
# creative content in them. When you customize them, please add a
# copyright notice and license.
OK'd in principle by: nyan@
creating the /dev/dpti%d entry that the software expects. This is just
a band-aid until either someone (hopefully) rewrites the utilities, or all
asr/dpt cards in existance get blasted into the sun.
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@
pc98 machines because (a) it is PCIe or PCI-X (b) there's a BIOS that
must run at boot which assumes IBM-AT compatible boot environment.
Noticed by: scottl
There are too many questions in freebsd-amd64@ about how to enable Linux
support that it seems a required piece of functionality. Thus we should
just have it on by default.
series of controllers. Areca provides a CLI and HTTP management tool for
FreeBSD/i386 and FreeBSD/amd64 on their website. Many thanks to Areca for
their support of FreeBSD. Thanks also to Mike Tansca and Sentex Communications
for donating hardware.
Obtained from: Erich Chen <erich at areca com tw>
ndis_timercall() in NdisMInitializeTimer(), we can't use the raw
function pointer. This is because ntoskrnl_run_dpc() expects to
invoke a function with Microsoft calling conventions. On i386,
this works because ndis_timercall() is declared with the __stdcall
attribute, but this is a no-op on amd64. To do it correctly, we
have to generate a wrapper for ndis_timercall() and us the wrapper
instead of of the raw function pointer.
Fix this by adding ndis_timercall() to the funcptr table in subr_ndis.c,
and create ndis_findwrap() to extract the wrapped function from the
table in NdisMInitializeTimer() instead of just passing ndis_timercall()
to KeInitializeDpc() directly.
checks, including cpuid_is_k7(), will catch CPUs that really don't support
this method.
Submitted by: Bruno Ducrot
Tested by: Jari Kirma (kirma cs.hut.fi)
on filesystems which safely support them. It appears that many
network filesystems specifically are not shared lock safe.
Sponsored by: Isilon Systems, Inc.
since simply unlocking a mutex does not ensure that one of the waiters
will run and acquire it. We're more likely to reacquire the mutex
before anyone else has a chance. It has also bit me three times now, as
it's not safe to drop the interlock before sleeping in many cases.
Sponsored by: Isilon Systems, Inc.
objdump --disassemble when disassembling itself in userland. I've added
the cmovCC instruction group and tweaked a bunch of size sensitive array
indexes to either fix my mistakes and/or force it to work by any means
necessary.
I'm committing this because it is usable enough to see what is going on
when single stepping via ddb.
It might still tell lies, but its lies will be far more subtle now. I'm
not sure that this is a good thing or not.
instructions as it was when I dropped it back in May 31, 2003. I'm
committing this as an intermediate stage because back then I thought I
understood what I was doing with this file.
an ap in 11g with protection enabled
o correct rate selection when operating in 11g with protection when no
packets have been sent yet (from John Bicket)
o track api change to get first descriptor and use it to collect the frame
length for calculating the state bin
o add more debugging and shuffle some existing debugging to give more info
o bump version to distinguish bug fixes
to the rate control module for tx complete processing; this enables
rate control algorithms to extract the packet length for xmits that
require multiple descriptors
o ATA is now fully newbus'd and split into modules.
This means that on a modern system you just load "atapci and ata"
to get the base support, and then one or more of the device
subdrivers "atadisk atapicd atapifd atapist ataraid".
All can be loaded/unloaded anytime, but for obvious reasons you
dont want to unload atadisk when you have mounted filesystems.
o The device identify part of the probe has been rewritten to fix
the problems with odd devices the old had, and to try to remove
so of the long delays some HW could provoke. Also probing is done
without the need for interrupts, making earlier probing possible.
o SATA devices can be hot inserted/removed and devices will be created/
removed in /dev accordingly.
NOTE: only supported on controllers that has this feature:
Promise and Silicon Image for now.
On other controllers the usual atacontrol detach/attach dance is
still needed.
o Support for "atomic" composite ATA requests used for RAID.
o ATA RAID support has been rewritten and and now supports these
metadata formats:
"Adaptec HostRAID"
"Highpoint V2 RocketRAID"
"Highpoint V3 RocketRAID"
"Intel MatrixRAID"
"Integrated Technology Express"
"LSILogic V2 MegaRAID"
"LSILogic V3 MegaRAID"
"Promise FastTrak"
"Silicon Image Medley"
"FreeBSD PseudoRAID"
o Update the ioctl API to match new RAID levels etc.
o Update atacontrol to know about the new RAID levels etc
NOTE: you need to recompile atacontrol with the new sys/ata.h,
make world will take care of that.
NOTE2: that rebuild is done differently from the old system as
the rebuild is now done piggybacked on read requests to the
array, so atacontrol simply starts a background "dd" to rebuild
the array.
o The reinit code has been worked over to be much more robust.
o The timeout code has been overhauled for races.
o Support of new chipsets.
o Lots of fixes for bugs found while doing the modulerization and
reviewing the old code.
Missing or changed features from current ATA:
o atapi-cd no longer has support for ATAPI changers. Todays its
much cheaper and alot faster to copy those CD images to disk
and serve them from there. Besides they dont seem to be made
anymore, maybe for that exact reason.
o ATA RAID can only read metadata from all the above metadata formats,
not write all of them (Promise and Highpoint V2 so far). This means
that arrays can be picked up from the BIOS, but they cannot be
created from FreeBSD. There is more to it than just the missing
write metadata support, those formats are not unique to a given
controller like Promise and Highpoint formats, instead they exist
for several types, and even worse, some controllers can have
different formats and its impossible to tell which one.
The outcome is that we cannot reliably create the metadata of those
formats and be sure the controller BIOS will understand it.
However write support is needed to update/fail/rebuild the arrays
properly so it sits fairly high on the TODO list.
o So far atapicam is not supported with these changes. When/if this
will change is up to the maintainer of atapi-cam so go there for
questions.
HW donated by: Webveveriet AS
HW donated by: Frode Nordahl
HW donated by: Yahoo!
HW donated by: Sentex
Patience by: Vife and my boys (and even the cats)
carp_carpdev_state_locked() is called every time carp interface is attached.
The first call backs up flags of the first interface, and the second
call backs up them again, erasing correct values.
To solve this, a carp_sc_state_locked() function is introduced. It is
called when interface is attached to parent, instead of calling
carp_carpdev_state_locked. carp_carpdev_state_locked() calls
carp_sc_state_locked() for each sc in chain.
Reported by: Yuriy N. Shkandybin, sem
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
filesystem modules must be recompiled. (Since struct vnode has
already changed in 6-CURRENT, there's little advantage to leaving
the unused fields around.)
vnodes whose names it caches, so we no longer need a `generation
number' to tell us if a referenced vnode is invalid. Replace the use
of the parent's v_id in the hash function with the address of the
parent vnode.
Tested by: Peter Holm
Glanced at by: jeff, phk
except for places where people forget to update one of them. We now
collect only one set of stats for both of these routines. Other
changes in this commit include:
- Start acquiring Giant again in vn_fullpath(), since it is required
when crossing a mount point.
- Expand the scope of the cache lock to avoid dropping it and
picking it up again for every pathname component. This also
makes it trivial to avoid races in stats collection.
- Assert that nc_dvp == v_dd for directories instead of returning
an error to userland when this is not true. AFAIK, it should
always be true when v_dd is non-null.
- For vn_fullpath(), handle the first (non-directory) vnode
separately.
Glanced at by: jeff, phk
returns error. In this case mbuf has already been freed. [1]
- Remove redundant declaration.
PR: kern/78893 [1]
Submitted by: Liang Yi [1]
Reviewed by: sam
MFC after: 1 day
to cache_lookup(). This allows us to acquire the vnode interlock before
dropping the cache lock. This protects the vnodes identity until we
have locked it.
Sponsored by: Isilon Systems, Inc.
We don't need a mknod(2) call
No tricky install documentation
Kernel leave them dev_t alone
Hey Kernel leave them cdevsw alone
All in all it's just another struct in src/sys
All in all you're just another struct in src/sys
Don't remove the now unused element from cdev yet, wait until
we have a better reason to bump the version.
There is now no longer any upper limit on how many device drivers
a FreeBSD kernel can have.
only allow proper values. ENTROPYSOURCE is a maxval+1, not an
allowable number.
Suggested loose protons in the solution: phk
Prefers to keep the pH close to seven: markm
acquire shared locks on intermediate directories.
- For the LASTCN, we may have to LK_UPGRADE the parent directory before
we lookup the last component.
- Acquire VFS_ROOT and dp locks based on the cn_lkflag.
Sponsored by: Isilon Systems, Inc.
vhold()s us.
- Avoid an extra mutex acquire and release in the common case of vgonel()
by checking for OWEINACT at the start of the function.
- Fix the case where we set OWEINACT in vput(). LK_EXCLUPGRADE drops our
shared lock if it fails.
Sponsored by: Isilon Systems, Inc.
ExAllocatePoolWithTag(), not malloc(), so it should be released
with ExFreePool(), not free(). Fix a couple if instances of
free(fh, ...) that got overlooked.
- On amd64, InterlockedPushEntrySList() and InterlockedPopEntrySList()
are mapped to ExpInterlockedPushEntrySList and
ExpInterlockedPopEntrySList() via macros (which do the same thing).
Add IMPORT_FUNC_MAP()s for these.
- Implement ExQueryDepthSList().
alloc and free routine pointers in the lookaside list with pointers
to ExAllocatePoolWithTag() and ExFreePool() (in the case where the
driver does not provide its own alloc and free routines). For amd64,
this is wrong: we have to use pointers to the wrapped versions of these
functions, not the originals.
zero'ing their length (copied from m_adj where this code came from
after the equivalent change there has had time to soak)
Noticed by: Coverity Prevent analysis tool
This adds support for the SiS intergrated NIC on some Athlon64 motherboards.
The MAC address is stored in the APC CMOS RAM and this fixes the
sis driver ending up with a 00:00:00:00:00:00 MAC address.
Submitted by: Stasys Smailys <ssmailys@komvista.lt>
nll_obsoletelock field in the lookaside list structure is only defined
for the i386 arch. For amd64, the field is gone, and different list
update routines are used which do their locking internally. Apparently
the Inprocomm amd64 driver uses lookaside lists. I'm not positive this
will make it work yet since I don't have an Inprocomm NIC to test, but
this needs to be fixed anyway.
- Assert that REMOVE, CREATE, and RENAME callers have WANTPARENT
or LOCKPARENT set. You can't complete any of these operations without
at least a reference to the parent. Many filesystems check for this case
even though it isn't possible in the current system.
- Only unlock the directory if this is a DOTDOT lookup. Previously this
code could have deadlocked if there was a DOTDOT lookup with LOCKPARENT
set and another thread was locking the other way up the tree.
Sponsored by: Isilon Systems, Inc.
handled in vfs_lookup.c. This code was missing PDIRUNLOCK use prior
to the removal of PDIRUNLOCK in rev 1.73 of vfs_lookup.c.
Sponsored by: Isilon Systems, Inc.
handled in vfs_lookup.c. This code was missing PDIRUNLOCK use prior
to the removal of PDIRUNLOCK in rev 1.73 of vfs_lookup.c.
Sponsored by: Isilon Systems, Inc.
handled in vfs_lookup.c. This code was missing PDIRUNLOCK use prior
to the removal of PDIRUNLOCK in rev 1.73 of vfs_lookup.c.
Sponsored by: Isilon Systems, Inc.
rely on ufs to always leave the parent locked except in the ISDOTDOT
case. Adjust asserts to deal with these changes.
Sponsored by: Isilon Systems, Inc.
- In the ISDOTDOT case we have to unlock the dvp before locking the child,
if this fails we must relock dvp before returning an error. This was
missing before.
Sponsored by: Isilon Systems, Inc.
- Network filesystems are written with a special idiom that checks the
cache first, and may even unlock dvp before discovering that a network
round-trip is required to resolve the name. I believe dvp is prevented
from being recycled even in the forced unmount case by the shared lock
on the mount point. If not, this code should grow checks for VI_DOOMED
after it relocks dvp or it will access NULL v_data fields.
Sponsored by: Isilon Systems, Inc.
calling VOP_LOOKUP(). Rather than having each filesystem check the
LOCKPARENT flag, we simply check it once here and unlock as required.
The only unusual case is ISDOTDOT, where we require an unlocked vnode
on return. Relocking this vnode with the child locked is allowed since
the child is actually its parent.
- Add a few asserts for some unusual conditions that I do not believe can
happen. These will later go away and turn into implementations for these
conditions.
Sponsored by: Isilon Systems, Inc.
case where filesystems legitimately need to unlock the directory vp is
in the DOTDOT case, which we can explicitly check for in lookup().
Furthermore, allowing filesystems to unlock dvp can lead to lock order
reversals in lookup() when we vrele the dvp while the child is still
locked.
Sponsored by: Isilon Systems, Inc.
