1
0
mirror of https://git.FreeBSD.org/src.git synced 2024-12-16 10:20:30 +00:00
freebsd/lib/libc
Adrian Chadd 6520495abc Add an initial NUMA affinity/policy configuration for threads and processes.
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
2015-07-11 15:21:37 +00:00
..
aarch64 Add support for makecontext. This supports up to 8 arguments as this 2015-07-10 08:36:22 +00:00
amd64 Remove reboot.S (part of libc). It's not needed and was actually 2015-07-07 09:25:51 +00:00
arm Export the ARM __aeabi_mem* functions from libc, they are needed by the gcc 2015-06-16 16:40:25 +00:00
capability use .Mt to mark up email addresses consistently (part4) 2014-06-23 08:25:03 +00:00
compat-43 Reduce the size of the interposing table and amount of 2015-01-11 22:16:31 +00:00
db Drop some unnecessary casts. 2015-05-18 16:28:13 +00:00
gdtoa Replace use of ${.CURDIR} by ${LIBC_SRCTOP} and define ${LIBC_SRCTOP} 2014-03-04 02:19:39 +00:00
gen Since sh(1) now supports mulitbyte (only UTF-8) clarify the related BUGS 2015-07-11 13:07:50 +00:00
gmon Replace use of ${.CURDIR} by ${LIBC_SRCTOP} and define ${LIBC_SRCTOP} 2014-03-04 02:19:39 +00:00
i386 Remove reboot.S (part of libc). It's not needed and was actually 2015-07-07 09:25:51 +00:00
iconv Fix improbable memory leak in _citrus_prop_read_str(). 2015-04-20 22:09:50 +00:00
include Reassign copyright statements on several files from Advanced 2015-04-23 14:22:20 +00:00
inet Update our stub resolver to final version of libbind. 2014-08-12 12:36:06 +00:00
isc Update our stub resolver to final version of libbind. 2014-08-12 12:36:06 +00:00
locale __xlocale_C_ctype should not be const. It contains a reference count that is modified by newlocale / duplocale / freelocale. 2015-04-24 10:21:20 +00:00
md Bring the md5 functions into libc for internal use only. 2014-08-12 12:25:56 +00:00
mips Removed unused special fork() implementations. 2015-05-29 19:42:55 +00:00
nameser Add missing break. 2014-08-20 14:58:25 +00:00
net Fix name of a constant. 2015-06-14 20:16:36 +00:00
nls Use mdoc(7) macros to handle parenthesis 2015-04-26 10:54:52 +00:00
posix1e libc: plug memory leaks in edge cases for the posix1e code. 2015-03-13 18:42:43 +00:00
powerpc The lseek(2), mmap(2), truncate(2), ftruncate(2), pread(2), and 2015-04-18 21:50:13 +00:00
powerpc64 The lseek(2), mmap(2), truncate(2), ftruncate(2), pread(2), and 2015-04-18 21:50:13 +00:00
quad Remove MK_ARM_EABI, the armeb issues have been fixed. The code to support 2014-10-01 08:26:51 +00:00
regex mdoc: rendering fixes 2015-04-26 10:55:39 +00:00
resolv Add support for '_' occurring at the beginning or end of a name 2015-06-10 16:15:22 +00:00
rpc mdoc: fix functions declarations 2015-04-26 10:53:50 +00:00
softfloat Replace use of ${.CURDIR} by ${LIBC_SRCTOP} and define ${LIBC_SRCTOP} 2014-03-04 02:19:39 +00:00
sparc64 The lseek(2), mmap(2), truncate(2), ftruncate(2), pread(2), and 2015-04-18 21:50:13 +00:00
stdio Add fdclose(3) function. 2015-07-04 16:42:14 +00:00
stdlib Fix major copy/paste and other style errors. 2015-05-05 10:44:17 +00:00
stdtime 1) Fix the case we have less arguments for format string than we expected. 2014-10-07 06:34:05 +00:00
string strchr(3): Mdoc fixes. 2015-06-04 21:54:46 +00:00
sys Add an initial NUMA affinity/policy configuration for threads and processes. 2015-07-11 15:21:37 +00:00
tests Add support for makecontext. This supports up to 8 arguments as this 2015-07-10 08:36:22 +00:00
uuid Replace use of ${.CURDIR} by ${LIBC_SRCTOP} and define ${LIBC_SRCTOP} 2014-03-04 02:19:39 +00:00
xdr Use proper prototype for harmless(). 2015-05-31 19:09:24 +00:00
yp Replace use of ${.CURDIR} by ${LIBC_SRCTOP} and define ${LIBC_SRCTOP} 2014-03-04 02:19:39 +00:00
libc.ldscript Attempt to move the POSIX iconv* symbols out of runtime linker space. 2013-11-17 22:52:17 +00:00
Makefile Revert r284417 it is not necessary anymore 2015-06-15 19:28:07 +00:00
Makefile.depend new depends 2015-06-16 23:37:19 +00:00
Versions.def Fix the inheritance of the FBSDprivate_1.0 namespace. 2014-04-03 17:31:38 +00:00