This change refactors the existing create_thread() function to be more
generic. It replaces almost all of its arguments by a callback that can
be used to extract the thread ID and copy it out to the right place, but
also to perform additional initialization steps, such as setting the
trapframe. This also makes the difference between thr_new() and
thr_create() more clear in my opinion.
This function is going to be used by the CloudABI compatibility layer.
Reviewed by: kib
MFC after: 1 month
This is based on work done by jeff@ and jhb@, as well as the numa.diff
patch that has been circulating when someone asks for first-touch NUMA
on -10 or -11.
* Introduce a simple set of VM policy and iterator types.
* tie the policy types into the vm_phys path for now, mirroring how
the initial first-touch allocation work was enabled.
* add syscalls to control changing thread and process defaults.
* add a global NUMA VM domain policy.
* implement a simple cascade policy order - if a thread policy exists, use it;
if a process policy exists, use it; use the default policy.
* processes inherit policies from their parent processes, threads inherit
policies from their parent threads.
* add a simple tool (numactl) to query and modify default thread/process
policities.
* add documentation for the new syscalls, for numa and for numactl.
* re-enable first touch NUMA again by default, as now policies can be
set in a variety of methods.
This is only relevant for very specific workloads.
This doesn't pretend to be a final NUMA solution.
The previous defaults in -HEAD (with MAXMEMDOM set) can be achieved by
'sysctl vm.default_policy=rr'.
This is only relevant if MAXMEMDOM is set to something other than 1.
Ie, if you're using GENERIC or a modified kernel with non-NUMA, then
this is a glorified no-op for you.
Thank you to Norse Corp for giving me access to rather large
(for FreeBSD!) NUMA machines in order to develop and verify this.
Thank you to Dell for providing me with dual socket sandybridge
and westmere v3 hardware to do NUMA development with.
Thank you to Scott Long at Netflix for providing me with access
to the two-socket, four-domain haswell v3 hardware.
Thank you to Peter Holm for running the stress testing suite
against the NUMA branch during various stages of development!
Tested:
* MIPS (regression testing; non-NUMA)
* i386 (regression testing; non-NUMA GENERIC)
* amd64 (regression testing; non-NUMA GENERIC)
* westmere, 2 socket (thankyou norse!)
* sandy bridge, 2 socket (thankyou dell!)
* ivy bridge, 2 socket (thankyou norse!)
* westmere-EX, 4 socket / 1TB RAM (thankyou norse!)
* haswell, 2 socket (thankyou norse!)
* haswell v3, 2 socket (thankyou dell)
* haswell v3, 2x18 core (thankyou scott long / netflix!)
* Peter Holm ran a stress test suite on this work and found one
issue, but has not been able to verify it (it doesn't look NUMA
related, and he only saw it once over many testing runs.)
* I've tested bhyve instances running in fixed NUMA domains and cpusets;
all seems to work correctly.
Verified:
* intel-pcm - pcm-numa.x and pcm-memory.x, whilst selecting different
NUMA policies for processes under test.
Review:
This was reviewed through phabricator (https://reviews.freebsd.org/D2559)
as well as privately and via emails to freebsd-arch@. The git history
with specific attributes is available at https://github.com/erikarn/freebsd/
in the NUMA branch (https://github.com/erikarn/freebsd/compare/local/adrian_numa_policy).
This has been reviewed by a number of people (stas, rpaulo, kib, ngie,
wblock) but not achieved a clear consensus. My hope is that with further
exposure and testing more functionality can be implemented and evaluated.
Notes:
* The VM doesn't handle unbalanced domains very well, and if you have an overly
unbalanced memory setup whilst under high memory pressure, VM page allocation
may fail leading to a kernel panic. This was a problem in the past, but it's
much more easily triggered now with these tools.
* This work only controls the path through vm_phys; it doesn't yet strongly/predictably
affect contigmalloc, KVA placement, UMA, etc. So, driver placement of memory
isn't really guaranteed in any way. That's next on my plate.
Sponsored by: Norse Corp, Inc.; Dell
Thread credentials are maintained as follows: each thread has a pointer to
creds and a reference on them. The pointer is compared with proc's creds on
userspace<->kernel boundary and updated if needed.
This patch introduces a counter which can be compared instead, so that more
structures can use this scheme without adding more comparisons on the boundary.
threads introduce kern_thr_alloc() which will be used later in the
linux_clone().
Differential Revision: https://reviews.freebsd.org/D1029
Reviewed by: trasz
threads split sys_thr_exit() up into sys_thr_exit() and kern_thr_exit().
Move
Where the second will be used in linux_exit() system call later.
Differential Revision: https://reviews.freebsd.org/D1028
Reviewed by: trasz
The point of this is to be able to add RACCT (with RACCT_DISABLED)
to GENERIC, to avoid having to rebuild the kernel to use rctl(8).
Differential Revision: https://reviews.freebsd.org/D2369
Reviewed by: kib@
MFC after: 1 month
Relnotes: yes
Sponsored by: The FreeBSD Foundation
currently a spin lock. Apparently, the only reason for this is that
umtx_thread_exit() is called under the process spinlock, which put the
requirement on the umtx_lock. Note that the witness static order list
is wrong for the umtx_lock, umtx_lock is explicitely before any thread
lock, so it is also before sleepq locks.
Change umtx_lock to be the sleepable mutex. For the reason above, the
calls to umtx_thread_exit() are moved from thread_exit() earlier in
each caller, when the process spin lock is not yet taken.
Discussed with: jhb
Tested by: pho (previous version)
Sponsored by: The FreeBSD Foundation
MFC after: 3 weeks
talked about. Explain where the mentioned trampoline located
(usermode), and the fact that attempt to exit last thread is denied in
kernel (by delegating the work to usermode).
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
the cached name used for KTR_SCHED traces when a thread's name changes.
This way KTR_SCHED traces (and thus schedgraph) will notice when a thread's
name changes, most commonly via execve().
MFC after: 2 weeks
thread_free(newtd). This to avoid a possible page fault in
cpu_thread_clean() as seen on amd64 with syscall fuzzing.
Reviewed by: kib
MFC after: 1 week
The SYSCTL_NODE macro defines a list that stores all child-elements of
that node. If there's no SYSCTL_DECL macro anywhere else, there's no
reason why it shouldn't be static.
patch modifies makesyscalls.sh to prefix all of the non-compatibility
calls (e.g. not linux_, freebsd32_) with sys_ and updates the kernel
entry points and all places in the code that use them. It also
fixes an additional name space collision between the kernel function
psignal and the libc function of the same name by renaming the kernel
psignal kern_psignal(). By introducing this change now we will ease future
MFCs that change syscalls.
Reviewed by: rwatson
Approved by: re (bz)
won't happen before 9.0. This commit adds "#ifdef RACCT" around all the
"PROC_LOCK(p); racct_whatever(p, ...); PROC_UNLOCK(p)" instances, in order
to avoid useless locking/unlocking in kernels built without "options RACCT".
file where they are used. Declare the kern.threads sysctl node at the
same location. Since no external use for the variables exists, make them
static.
Discussed with: dchagin
MFC after: 1 week
rwlock to protect the table. In old code, thread lookup is done with
process lock held, to find a thread, kernel has to iterate through
process and thread list, this is quite inefficient.
With this change, test shows in extreme case performance is
dramatically improved.
Earlier patch was reviewed by: jhb, julian
standard kill(). On other systems, SI_LWP is generated by lwp_kill().
This will allow conforming applications to differentiate between
signals generated by standard events and those generated by other
implementation events in a manner compatible with existing practice.
- Bump __FreeBSD_version
- Rename tdsignal() to tdsendsignal() and make it private to kern_sig.c.
- Add tdsignal() and tdksignal() routines that mirror psignal() and
pksignal() except that they accept a thread as an argument instead of
a process. They send a signal to a specific thread rather than to an
individual process.
Reviewed by: kib
for upcoming 64-bit PowerPC and MIPS support. This renames the COMPAT_IA32
option to COMPAT_FREEBSD32, removes some IA32-specific code from MI parts
of the kernel and enhances the freebsd32 compatibility code to support
big-endian platforms.
Reviewed by: kib, jhb
not blocking the signal, signal is placed on the thread sigqueue. If
the selected thread is in kernel executing thr_exit() or sigprocmask()
syscalls, then signal might be not delivered to usermode for arbitrary
amount of time, and for exiting thread it is lost.
Put process-directed signals to the process queue unconditionally,
selecting the thread to deliver the signal only by the thread returning
to usermode, since only then the thread can handle delivery of signal
reliably. For exiting thread or thread that has blocked some signals,
check whether the newly blocked signal is queued for the process, and
try to find a thread to wakeup for delivery, in reschedule_signal(). For
exiting thread, assume that all signals are blocked.
Change cursig() and postsig() to look both into the thread and process
signal queues. When there is a signal that thread returning to usermode
could consume, TDF_NEEDSIGCHK flag is not neccessary set now. Do
unlocked read of p_siglist and p_pendingcnt to check for queued signals.
Note that thread that has a signal unblocked might get spurious wakeup
and EINTR from the interruptible system call now, due to the possibility
of being selected by reschedule_signals(), while other thread returned
to usermode earlier and removed the signal from process queue. This
should not cause compliance issues, since the thread has not blocked a
signal and thus should be ready to receive it anyway.
Reported by: Justin Teller <justin.teller gmail com>
Reviewed by: davidxu, jilles
MFC after: 1 month
Remove the altkstacks, instead instantiate threads with kernel stack
allocated with the right size from the start. For the thread that has
kernel stack cached, verify that requested stack size is equial to the
actual, and reallocate the stack if sizes differ [1].
This fixes the bug introduced by r173361 that was committed several days
after r173004 and consisted of kthread_add(9) ignoring the non-default
kernel stack size.
Also, r173361 removed the caching of the kernel stacks for a non-first
thread in the process. Introduce separate kernel stack cache that keeps
some limited amount of preallocated kernel stacks to lower the latency
of thread allocation. Add vm_lowmem handler to prune the cache on
low memory condition. This way, system with reasonable amount of the
threads get lower latency of thread creation, while still not exhausting
significant portion of KVA for unused kstacks.
Submitted by: peter [1]
Discussed with: jhb, julian, peter
Reviewed by: jhb
Tested by: pho (and retested according to new test scenarious)
MFC after: 1 week
allocated with the right size from the start. For the thread that has
kernel stack cached, verify that requested stack size is equial to the
actual, and reallocate the stack if sizes differ [1].
This fixes the bug introduced by r173361 that was committed several days
after r173004 and consisted of kthread_add(9) ignoring the non-default
kernel stack size.
Also, r173361 removed the caching of the kernel stacks for a non-first
thread in the process. Introduce separate kernel stack cache that keeps
some limited amount of preallocated kernel stacks to lower the latency
of thread allocation. Add vm_lowmem handler to prune the cache on
low memory condition. This way, system with reasonable amount of the
threads get lower latency of thread creation, while still not exhausting
significant portion of KVA for unused kstacks.
Submitted by: peter [1]
Discussed with: jhb, julian, peter
Reviewed by: jhb
Tested by: pho
MFC after: 1 week
specific macros for each audit argument type. This makes it easier to
follow call-graphs, especially for automated analysis tools (such as
fxr).
In MFC, we should leave the existing AUDIT_ARG() macros as they may be
used by third-party kernel modules.
Suggested by: brooks
Approved by: re (kib)
Obtained from: TrustedBSD Project
MFC after: 1 week
kern_time.c:
- Unused variable `p'.
kern_thr.c:
- Variable `error' is always caught immediately, so no reason to
initialize it. There is no way that error != 0 at the end of
create_thread().
kern_sig.c:
- Unused variable `code'.
kern_synch.c:
- `rval' is always assigned in all different cases.
kern_rwlock.c:
- `v' is always overwritten with RW_UNLOCKED further on.
kern_malloc.c:
- `size' is always initialized with the proper value before being used.
kern_exit.c:
- `error' is always caught and returned immediately. abort2() never
returns a non-zero value.
kern_exec.c:
- `len' is always assigned inside the if-statement right below it.
tty_info.c:
- `td' is always overwritten by FOREACH_THREAD_IN_PROC().
Found by: LLVM's scan-build
of the ABI of the currently executing image. Change some places to test
the flags instead of explicit comparing with address of known sysentvec
structures to determine ABI features.
Discussed with: dchagin, imp, jhb, peter
unnecessary, the normal process lock and thread lock are enough. The
spin lock is still needed for process and thread exiting to mimic
single sched_lock.
to enter thread_suspend_check().
- Set TDF_ASTPENDING along with TDF_NEEDSUSPCHK so we can move the
thread_suspend_check() to ast() rather than userret().
- Check TDF_NEEDSUSPCHK in the sleepq_catch_signals() optimization so
that we don't miss a suspend request. If this is set use the
expensive signal path.
- Set NEEDSUSPCHK when creating a new thread in thr in case the
creating thread is due to be suspended as well but has not yet.
Reviewed by: davidxu (Authored original patch)
requiring the per-process spinlock to only requiring the process lock.
- Reflect these changes in the proc.h documentation and consumers throughout
the kernel. This is a substantial reduction in locking cost for these
fields and was made possible by recent changes to threading support.
silent NULL pointer dereference in the i386 and sparc64 pmap_pinit()
when the kmem_alloc_nofault() failed to allocate address space. Both
functions now return error instead of panicing or dereferencing NULL.
As consequence, vmspace_exec() and vmspace_unshare() returns the errno
int. struct vmspace arg was added to vm_forkproc() to avoid dealing
with failed allocation when most of the fork1() job is already done.
The kernel stack for the thread is now set up in the thread_alloc(),
that itself may return NULL. Also, allocation of the first process
thread is performed in the fork1() to properly deal with stack
allocation failure. proc_linkup() is separated into proc_linkup()
called from fork1(), and proc_linkup0(), that is used to set up the
kernel process (was known as swapper).
In collaboration with: Peter Holm
Reviewed by: jhb