have some negative effect on interactivity but it yields great perf. gains.
This also brings the conditions under which ULE context switches inline
with SCHED_4BSD.
- Define some new kseq_* functions for manipulating the run queue.
- Add a new kseq member ksq_rslices and ksq_bload. rslices is the sum of
the slices of runnable kses. This will be used for push load balance
decisions. bload is the number of threads blocked waiting on IO.
I'm not convinced there is anything major wrong with the patch but
them's the rules..
I am using my "David's mentor" hat to revert this as he's
offline for a while.
than having change_dir() release the vnode lock on success, hold the
lock so that we can use it later when invoking MAC checks and
VOP_ACCESS() in the chroot() code. Update the comment to reflect
this calling convention. Update callers to unlock the vnode
lock. Correct a typo regarding vnode naming in the MAC case that
crept in via the previous patch applied.
cases: we might multiply vrele() a vnode when certain classes of
failures occur. This appears to stem from earlier Giant/file
descriptor lock pushdown and restructuring.
Submitted by: maxim
This implicitly removes the need for major numbers, but a number of
drivers still know things they shouldn't need to, and we need to
consider if there are applications which cache major(+minor) gleaned
from stat(2) and rely on it being constant over reboots before we
start assigning random majors.
sched_runnable() et all.
- Remove some dead code in sched_clock().
- Define two macros KSEQ_SELF() and KSEQ_CPU() for getting the kseq of the
current cpu or some alternate cpu.
- Start introducing kseq_() functions, such as kseq_choose() and kseq_setup().
run queue for each cpu.
- Introduce kse stealing into the sched_choose() code. This helps balance
cpus better in cases where process turnover is high. This implementation
is fairly trivial and will likely be only a temporary measure until
something more sophisticated has been written.
data structure called kse_upcall to manage UPCALL. All KSE binding
and loaning code are gone.
A thread owns an upcall can collect all completed syscall contexts in
its ksegrp, turn itself into UPCALL mode, and takes those contexts back
to userland. Any thread without upcall structure has to export their
contexts and exit at user boundary.
Any thread running in user mode owns an upcall structure, when it enters
kernel, if the kse mailbox's current thread pointer is not NULL, then
when the thread is blocked in kernel, a new UPCALL thread is created and
the upcall structure is transfered to the new UPCALL thread. if the kse
mailbox's current thread pointer is NULL, then when a thread is blocked
in kernel, no UPCALL thread will be created.
Each upcall always has an owner thread. Userland can remove an upcall by
calling kse_exit, when all upcalls in ksegrp are removed, the group is
atomatically shutdown. An upcall owner thread also exits when process is
in exiting state. when an owner thread exits, the upcall it owns is also
removed.
KSE is a pure scheduler entity. it represents a virtual cpu. when a thread
is running, it always has a KSE associated with it. scheduler is free to
assign a KSE to thread according thread priority, if thread priority is changed,
KSE can be moved from one thread to another.
When a ksegrp is created, there is always N KSEs created in the group. the
N is the number of physical cpu in the current system. This makes it is
possible that even an userland UTS is single CPU safe, threads in kernel still
can execute on different cpu in parallel. Userland calls kse_create to add more
upcall structures into ksegrp to increase concurrent in userland itself, kernel
is not restricted by number of upcalls userland provides.
The code hasn't been tested under SMP by author due to lack of hardware.
Reviewed by: julian
potential discontinuities in our UTC timescale.
Applications can monitor this variable if they want to be informed
about steps in the timescale. Slews (ntp and adjtime(2)) and
frequency adjustments (ntp) will not increment this counter, only
operations which set the clock. No attempt is made to classify
size or direction of the step.
correctly against PF_LOCAL. It seems that the test always fails then
sockaddr was not filled. So, I added else clause for workaround.
I doubt if it is right fix. However, it is better than nothing. I
found that NetBSD has same potential problem. But, fortunately,
NetBSD has equivalent else clause.
MFC after: 1 week
1. eliminate unnecessary loop which frees and re-allocates
the just allocated array
2. eliminate the newsize recomputation
3. eliminate unnecessary unlock and relock around free
4. correctly match the free with the malloc into M_KQUEUE instead of M_TEMP
5. eliminate conditional assignment of oldlist, which is equivalent to a
simple assignment
6. eliminate the oldlist temporary variable completely
Reviewed by: jhb
metadata. This fixes module dependency resolution by the kernel linker on
sparc64, where the relocations for the metadata are different than on other
architectures; the relative offset is in the addend of an Elf_Rela record
instead of the original value of the location being patched.
Also fix printf formats in debug code.
Submitted by: Hartmut Brandt <brandt@fokus.gmd.de>
PR: 46732
Tested on: alpha (obrien), i386, sparc64
was used to control code which were conditional on DEVFS' precense
since this avoided the need for large-scale source pollution with
#include "opt_geom.h"
Now that we approach making DEVFS standard, replace these tests
with an #ifdef to facilitate mechanical removal once DEVFS becomes
non-optional.
No functional change by this commit.
more than return ENXIO from its open routine, so most of this file
is unneeded.
A straight #ifdef'ing would look quite messy, and make the file
quite unreadable, so instead I have simply added the DEVFS version
of the file at the top, protected by #ifndef NODEVFS.
Once we have removed NODEVFS option, we can retain 86 the 86 lines at
the top and drop the other 287 lines.
SS_ISCONNECTING state, returning EINVAL (which is what POSIX mandates
in this case).
listen() on connected or connecting sockets would cause them to enter
a bad state; in the TCP case, this could cause sockets to go
catatonic or panics, depending on how the socket was connected.
Reviewed by: -net
MFC after: 2 weeks
vm_pageout_deficit:
1. Update vm_pageout_deficit before VM_WAIT. There is no sense in
delaying the update; the sooner the pageout daemon receives this
information the better. Reviewed by: tegge
2. Update vm_pageout_deficit according to the number of pages still
needed to complete the allocation, not the original size of the
allocation. Submitted by: tegge
(These errors have existed since the introduction of vm_pageout_deficit
in revision 1.144.)
portable copy. Note that pmap_extract() must be used instead of
pmap_kextract().
This is precursor work to a reorganization of vmapbuf() to close remaining
user/kernel races (which can lead to a panic).
in addition to secure level 1. The mask supports up to a secure level of 8
but only add defines through CTLFLAG_SECURE3 for now.
As per the missif in the log entry for 1.11 of ip_fw2.c which added the
secure flag to the IPFW sysctl's in the first place, change the secure
level requirement from 1 to 3 now that we have support for it.
Reviewed by: imp
With Design Suggestions by: imp
dereferenced when a process exits due to the vmspace ref-count being
bumped. Change shmexit() and shmexit_myhook() to take a vmspace instead
of a process and call it in vmspace_dofree(). This way if it is missed
in exit1()'s early-resource-free it will still be caught when the zombie is
reaped.
Also fix a potential race in shmexit_myhook() by NULLing out
vmspace->vm_shm prior to calling shm_delete_mapping() and free().
MFC after: 7 days
to access the pctcpu. This will have to be sorted out more later as the
new scheduler requires a procedural interface for this data. A more
complete solution will follow.
pointer types, and remove a huge number of casts from code using it.
Change struct xfile xf_data to xun_data (ABI is still compatible).
If we need to add a #define for f_data and xf_data we can, but I don't
think it will be necessary. There are no operational changes in this
commit.
that crept in recently. GCC will optimize the divides and multiplies for us.
Submitted by: David Schultz <dschultz@uclink.Berkeley.EDU>
MFC after: 1 day
so that entities that want to use the post_sync hook to write stuff
to devices and other tidy-up can do so before the device tree is
shot down. eg: da doing a SYNC_CACHE etc. This should get crashdumps
working on mpt devices again, and stops the ia64 boxes locking up
on regular shutdown when da tries to issue the scsi commands to mpt.
Obtained from: njl, gibbs
which expects it to be NULL unless the return value was 0 will work.
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
so it's value is not sign extended when assigned to the uintmax_t variable
used internally by printf. For example, if bit 31 is set in the cpuid
feature word, then %b would print out the initial value as a 16 character
hexadecimal value. Now it only prints out an 8 character value.
Reviewed by: bde
The leak in lseek was introduced in vfs_syscalls.c revision 1.218.
The leak in do_dup was introduced in kern_descrip.c revision 1.158.
Submitted by: iedowse
Add support for GCC's --test-coverage --profile-arcs options.
Add code to call the functions listed in the .ctors section, these are
used to string the per .o file counter blocks into a linked list.
Add empty __bb_fork_func() to cope with GCC magic gandling of exec*()
named functions.
To add support for other platforms should be trivial, but involves
determining the exact data-types gcc uses on that platform.
called. Otherwise (depending on a non-deterministic sort), the timecounter
code can be initialized before the clock rate has been set (on ia64) and it
assumes hz = 100, rather than the real value of 1024. I'm not sure how much
gets upset by this.
Glanced at by: phk
then call do_setopt_accept_filter(so, NULL) which will free the filter
instead of duplicating the code in do_setopt_accept_filter().
Pointed out by: Hiten Pandya <hiten@angelica.unixdaemons.com>
to sort out disk-io from file-io in the vm/buffer/filesystem space.
The intent is to sort VOP_STRATEGY calls into those which operate
on "real" vnodes and those which operate on VCHR vnodes. For
the latter kind, the call will be changed to VOP_SPECSTRATEGY,
possibly conditionally for those places where dual-use happens.
Add a default VOP_SPECSTRATEGY method which will call the normal
VOP_STRATEGY. First time it is called it will print debugging
information. This will only happen if a normal vnode is passed
to VOP_SPECSTRATEGY by mistake.
Add a real VOP_SPECSTRATEGY in specfs, which does what VOP_STRATEGY
does on a VCHR vnode today.
Add a new VOP_STRATEGY method in specfs to catch instances where
the conversion to VOP_SPECSTRATEGY has not yet happened. Handle
the request just like we always did, but first time called print
debugging information.
Apart up to two instances of console messages per boot, this amounts
to a glorified no-op commit.
If you get any of the messages on your console I would very much
like a copy of them mailed to phk@freebsd.org
included in the kernel. Include imgact_elf.c in conf/files, instead of
both imgact_elf32.c and imgact_elf64.c, which will use the default word
size for an architecture as defined in machine/elf.h. Architectures that
wish to build an additional image activator for an alternate word size can
include either imgact_elf32.c or imgact_elf64.c in files.${ARCH}, which
allows it to be dependent on MD options instead of solely on architecture.
Glanced at by: peter
void backtrace(void);
function which will print a backtrace if DDB is in the kernel and an
explanation if not.
This is useful for recording backtraces in non-fatal circumstances and
does not require pollution with DDB #includes in the files where it
is used.
It would of course be nice to have a non-DDB dependent version too,
but since the meat of a backtrace is MD it is probably not worth it.
On architectures with a non-executable stack, eg sparc64, this is used by
libgcc to determine at runtime if its necessary to enable execute permissions
on a region of the stack which will be used to execute code, allowing the
call to mprotect to be avoided if the kernel is configured to map the stack
executable.
o Allow callers of m_extadd() to allocate their own reference
m_ext.ref_cnt pointer, rather than having the mbuf system allocate it
with a malloc() in the critical path. This speeds m_extadd() up, and
also simplifies locking (malloc() may need Giant).
A driver or subsystem wishing to take use its own ref counter must
initialize m_ext.ref_cnt to point to its ref counter prior to
calling m_extadd(), and it must use EXT_EXTREF as its external type.
Eg:
m->m_ext.ref_cnt = my_ref_cnt_ptr;
m_extadd(.....,EXT_EXTREF);
Reviewed by: bosko
this was causing filedesc work to be very painful.
In order to make this work split out sigio definitions to thier own header
(sigio.h) which is included from proc.h for the time being.
take pointers to filedesc structures instead of threads. This makes
it more clear that they do not do any voodoo with the thread/proc
or anything other than the filedesc passed in or returned.
Remove some XXX KSE's as this resolves the issue.
calling getmicrouptime (but maintain the struct timeval-based calling
convention for compatibility)
o eliminate the use of timersub in ratecheck
Note that flood ping tests indicate ppsratecheck is inaccurate (but on the
conservative side) with this revised implementation. If more accuracy is
needed we'll have to introduce an alternate interface or increase the
overhead.
Reviewed by: silby, dillon, bde
were sometimes propagated using M_COPY_PKTHDR which actually did
something between a "move" and a "copy" operation. This is replaced
by M_MOVE_PKTHDR (which copies the pkthdr contents and "removes" it
from the source mbuf) and m_dup_pkthdr which copies the packet
header contents including any m_tag chain. This corrects numerous
problems whereby mbuf tags could be lost during packet manipulations.
These changes also introduce arguments to m_tag_copy and m_tag_copy_chain
to specify if the tag copy work should potentially block. This
introduces an incompatibility with openbsd which we may want to revisit.
Note that move/dup of packet headers does not handle target mbufs
that have a cluster bound to them. We may want to support this;
for now we watch for it with an assert.
Finally, M_COPYFLAGS was updated to include M_FIRSTFRAG|M_LASTFRAG.
Supported by: Vernier Networks
Reviewed by: Robert Watson <rwatson@FreeBSD.org>
__acl_get_link(), __acl_set_link(), acl_delete_link(), and
__acl_aclcheck_link(), with almost identical implementations to
the existing __acl_*_file() variants on these calls. Update
copyright.
Obtained from: TrustedBSD Project
__acl_get_link() Retrieve an ACL by name without following
symbolic links.
__acl_set_link() Set an ACL by name without following
symbolic links.
__acl_delete_link() Delete an ACL by name without following
symbolic links.
__acl_aclcheck_link() Check an ACL against a file by name without
following symbolic links.
These calls are similar in spirit to lstat(), lchown(), lchmod(), etc,
and will be used under similar circumstances.
Obtained from: TrustedBSD Project
call is in progress on the vnode. When vput() or vrele() sees a
1->0 reference count transition, it now return without any further
action if this flag is set. This flag is necessary to avoid recursion
into VOP_INACTIVE if the filesystem inactive routine causes the
reference count to increase and then drop back to zero. It is also
used to guarantee that an unlocked vnode will not be recycled while
blocked in VOP_INACTIVE().
There are at least two cases where the recursion can occur: one is
that the softupdates code called by ufs_inactive() via ffs_truncate()
can call vput() on the vnode. This has been reported by many people
as "lockmgr: draining against myself" panics. The other case is
that nfs_inactive() can call vget() and then vrele() on the vnode
to clean up a sillyrename file.
Reviewed by: mckusick (an older version of the patch)
to treat desiredvnodes much more like a limit than as a vague concept.
On a 2GB RAM machine where desired vnodes is 130k, we run out of
kmem_map space when we hit about 190k vnodes.
If we wake up the vnode washer in getnewvnode(), sleep until it is done,
so that it has a chance to offer us a washed vnode. If we don't sleep
here we'll just race ahead and allocate yet a vnode which will never
get freed.
In the vnodewasher, instead of doing 10 vnodes per mountpoint per
rotation, do 10% of the vnodes distributed evenly across the
mountpoints.
(show thread {address})
Remove the IDLE kse state and replace it with a change in
the way threads sahre KSEs. Every KSE now has a thread, which is
considered its "owner" however a KSE may also be lent to other
threads in the same group to allow completion of in-kernel work.
n this case the owner remains the same and the KSE will revert to the
owner when the other work has been completed.
All creations of upcalls etc. is now done from
kse_reassign() which in turn is called from mi_switch or
thread_exit(). This means that special code can be removed from
msleep() and cv_wait().
kse_release() does not leave a KSE with no thread any more but
converts the existing thread into teh KSE's owner, and sets it up
for doing an upcall. It is just inhibitted from being scheduled until
there is some reason to do an upcall.
Remove all trace of the kse_idle queue since it is no-longer needed.
"Idle" KSEs are now on the loanable queue.
The duplication is caused by the fact that imgact_elf.c is included
by both imgact_elf32.c and imgact_elf64.c and both are compiled by
default on ia64. Consequently, we have two seperate copies of the
elf_legacy_coredump variable due to them being declared static, and
two entries for the same sysctl in the linker set, both referencing
the unique copy of the elf_legacy_coredump variable. Since the second
sysctl cannot be registered, one of the elf_legacy_coredump variables
can not be tuned (if ordering still holds, it's the ELF64 related one).
The only solution is to create two different sysctl variables, just
like the elf<32|64>_trace sysctl variables. This unfortunately is an
(user) interface change, but unavoidable. Thus, on ELF32 platforms
the sysctl variable is called elf32_legacy_coredump and on ELF64
platforms it is called elf64_legacy_coredump. Platforms that have
both ELF formats have both sysctl variables.
These variables should probably be retired sooner rather than later.
skipping read-only pages, which can result in valuable non-text-related
data not getting dumped, the ELF loader and the dynamic loader now mark
read-only text pages NOCORE and the coredump code only checks (primarily) for
complete inaccessibility of the page or NOCORE being set.
Certain applications which map large amounts of read-only data will
produce much larger cores. A new sysctl has been added,
debug.elf_legacy_coredump, which will revert to the old behavior.
This commit represents collaborative work by all parties involved.
The PR contains a program demonstrating the problem.
PR: kern/45994
Submitted by: "Peter Edwards" <pmedwards@eircom.net>, Archie Cobbs <archie@dellroad.org>
Reviewed by: jdp, dillon
MFC after: 7 days
_KERNEL scope from "src/sys/sys/mchain.h".
Replace each occurrence of the above in _KERNEL scope with the
appropriate macro from the set of hto(be|le)(16|32|64) and
(be|le)toh(16|32|64) from "src/sys/sys/endian.h".
Tested by: tjr
Requested by: comment marked with XXX
resource starvation we clean-up as much of the vmspace structure as we
can when the last process using it exits. The rest of the structure
is cleaned up when it is reaped. But since exit1() decrements the ref
count it is possible for a double-free to occur if someone else, such as
the process swapout code, references and then dereferences the structure.
Additionally, the final cleanup of the structure should not occur until
the last process referencing it is reaped.
This commit solves the problem by introducing a secondary reference count,
calling 'vm_exitingcnt'. The normal reference count is decremented on exit
and vm_exitingcnt is incremented. vm_exitingcnt is decremented when the
process is reaped. When both vm_exitingcnt and vm_refcnt are 0, the
structure is freed for real.
MFC after: 3 weeks
they may be the only viable ones to flush. Thus it will now wait for
an inode lock if the other alternatives will result in rollbacks (and
immediate redirtying of the buffer). If only buffers with rollbacks
are available, one will be flushed, but then the buffer daemon will
wait briefly before proceeding. Failing to wait briefly effectively
deadlocks a uniprocessor since every other process writing to that
filesystem will wait for the buffer daemon to clean up which takes
close enough to forever to feel like a deadlock.
Reported by: Archie Cobbs <archie@dellroad.org>
Sponsored by: DARPA & NAI Labs.
Approved by: re
These call uma_large_malloc() and uma_large_free() which require Giant.
Fixes panic when descriptor table is larger than KMEM_ZMAX bytes
noticed by kkenn.
Reviewed by: jhb
unused. Replace it with a dm_mount back-pointer to the struct mount
that the devfs_mount is associated with. Export that pointer to MAC
Framework entry points, where all current policies don't use the
pointer. This permits the SEBSD port of SELinux's FLASK/TE to compile
out-of-the-box on 5.0-CURRENT with full file system labeling support.
Approved by: re (murray)
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
converting from individual vnode locks to the snapshot
lock, be sure to pass any waiting processes along to the
new lock as well. This transfer is done by a new function
in the lock manager, transferlockers(from_lock, to_lock);
Thanks to Lamont Granquist <lamont@scriptkiddie.org> for
his help in pounding on snapshots beyond all reason and
finding this deadlock.
Sponsored by: DARPA & NAI Labs.
1) Record all device events when devctl is enabled, rather than just when
devd has devctl open. This is necessary to prevent races between when
a device arrives, and when devd starts.
2) Add hw.bus.devctl_disable to disable devctl, this can also be set as a
tunable.
3) Fix async support. Reset nonblocking and async_td in open. remove
async flags.
4) Free all memory when devctl is disabled.
Approved by: re (blanket)
on this.
o Update the `cur' pointer in the cluster loop in m_getm() to avoid
incorrect truncation and leaked mbufs.
Reviewed by: bmilekic
Approved by: re
create an ABI that encodes offsets and sizes of structures into client
drivers. The functions isolate the ABI from changes to the resource
structure. Since these are used very rarely (once at startup), the
speed penalty will be down in the noise.
Also, add r_rid to the structure so that clients can save the 'rid' of
the resource in the struct resource, plus accessor functions. Future
additions to newbus will make use of this to present a simplified
interface for resource specification.
Approved by: re (jhb)
Reviewed by: jhb, jake
problem was a locked directory vnode), do not give the process a chance
to sleep in state "stopevent" (depends on the S_EXEC bit being set in
p_stops) until most resources have been released again.
Approved by: re
instead of panicing. Also, perform some of the simpler sanity checks on
the fds before acquiring the filedesc lock.
Approved by: re
Reported by: Dan Nelson <dan@emsphone.com> and others
by policy modules making use of downgrades in the MAC AST event. This
is required by the mac_lomac port of LOMAC to the MAC Framework.
Approved by: re
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories
