1
1
Граф коммитов

26409 Коммитов

Автор SHA1 Сообщение Дата
Ralph Castain
568b58af75 Merge pull request #2738 from rhc54/topic/cancel
Cancel the waitpid callback once the waitpid on a process has fired to avoid multiple notifications
2017-01-16 15:53:53 -08:00
Ralph Castain
74a285be83 Cancel the waitpid callback once the waitpid on a process has fired to avoid multiple notifications
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-16 14:32:02 -08:00
Mike Dubman
bd6da46821 Merge pull request #2736 from alex-mikheev/topic/memheap_init_fix
oshmem: memheap: refactor component selection code
2017-01-16 16:37:52 +02:00
Alex Mikheev
83c2ab76a5
oshmem: memheap: refactor component selection code
Do not call component's init function until the component has been
selected.

Use mca_base_select() instead of the custom component selection code.

Signed-off-by: Alex Mikheev <alexm@mellanox.com>
2017-01-16 13:48:58 +02:00
Ralph Castain
9e8c7d6295 Silence Coverity warning
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-15 07:51:37 -08:00
Ralph Castain
6b34cc67d6 Correct typo
Fixes #2691

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-15 07:48:31 -08:00
Ralph Castain
d9b30e429f Merge pull request #2735 from rhc54/topic/sigh
Plug fd leaks
2017-01-14 18:28:32 -08:00
Ralph Castain
3a157f0496 One more time - we "push" IOF for stdout, stderr, and stddiag with separate calls. However, we were creating the sinks for all three of them each time, which caused them to leak. Create the sinks only once for each channel.
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-14 17:40:36 -08:00
Ralph Castain
d9fc88c2c7 Merge pull request #2733 from rhc54/topic/dvm
Cleanup DVM leaks
2017-01-12 21:29:54 -08:00
Ralph Castain
5c87fc10dc Update the mpirun man page to accurately reflect the use of -host
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-12 20:28:53 -08:00
Ralph Castain
b55c03255a Strange - I had created a new IOF API "complete" for cleaning up at the end of jobs, but somehow the implementation is missing. It also appears that the orted's never actually cleaned up their job-related information. These things are fine for normal mpirun-based operations, but cause significant resource leaks for the DVM.
Complete the implementation and seal the leaks

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-12 19:54:18 -08:00
Ralph Castain
6a092a41c2 Merge pull request #2732 from rhc54/topic/fudge
Missed one spot - plug fd leaks in orteds
2017-01-12 16:28:33 -08:00
Ralph Castain
0e2df3be3e Missed one spot - plug fd leaks in orteds
Fixes #2691

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-12 13:45:46 -08:00
Ralph Castain
9ad02b5d13 Merge pull request #2718 from rhc54/topic/leaks
Don't remove the IOF framework's tracking info for a proc until the state machine tells it to do so.
2017-01-12 09:57:17 -08:00
Nathan Hjelm
110840fc87 ess/hnp: add support for forwarding additional signals (#2712)
* ess/hnp: add support for forwarding additional signals

This commit adds support to the hnp ess module to forward additional
signals beyond the default SIGUSR1, SIGUSR2, SIGSTP, and SIGCONT.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>

* Generalize this a bit to allow a broader range of signals to be forwarded. Turns out that SIGURG is now a "standard" signal, though the value differs across systems. So setup to forward it (and some friends) if they are defined. Allow users to provide the signal name (instead of the integer value) as the value of even the more common signals does vary across systems. Don't limit the number that can be supported.

Signed-off-by: Ralph Castain <rhc@open-mpi.org>

* ess/hnp: fix some bugs in the signal forwarding code

This commit fixes two bugs:

 - signals_set needs to be set even if no signals are being
   forwarded. If it is not set we will SEGV in libevent if
   ess_hnp_forward_signals == none.

 - SIGTERM and SIGHUP are handled with a different type of handler. Do
   not allow the user to specify these to be forwarded.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>

* We are sure to get "dinged" if error messages aren't nicely output via show_help, so do so here

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-12 10:09:41 -07:00
Nathan Hjelm
91c34c8df6 Merge pull request #2703 from hjelmn/rcache_fix
rcache/base: do not release vma stuctures in vma_tree_delete
2017-01-12 09:53:34 -07:00
Ralph Castain
fa419d3c0d Don't remove the IOF framework's tracking info for a proc until the
state machine tells it to do so. This plugs leaked file descriptors as
we were losing track prior to destructing the resources.

Fixes #2691

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-12 08:34:29 -08:00
Nysal Jan K A
16ca8c18c6 Merge pull request #2706 from nysal/ppc_atomic_master
asm/ppc: Fix a regression in powerpc atomics
2017-01-12 19:43:33 +05:30
Jeff Squyres
938ab01ad6 Merge pull request #2714 from hjelmn/timer_rollover
timer/linux: prevent 64-bit overflow
2017-01-12 06:40:52 -05:00
Nathan Hjelm
45c05880aa timer/linux: prevent 64-bit overflow
The linux timer code was multiplying the result of the x86 time stamp
counter by 1000000 before dividing by the cpu frequency. This can
cause us to overflow 64 bits if the time stamp counter grows larger
than ~ 1.8e13 (about 8400 seconds after boot). To fix the issue the
units of opal_timer_linux_freq have been changed to MHz.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2017-01-11 20:03:10 -07:00
Gilles Gouaillardet
aeee48357a btl/sm: correctly handle nodes with zero NUMA hwloc object
the hwloc topology might not contain a NUMA object with hwloc < v2
if the node is not NUMA, so force the NUMA object count to one
in order to correctly allocate mca_btl_sm_component.sm_mpools.

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-12 11:45:29 +09:00
Gilles Gouaillardet
4932391002 ompi/proc: fix ompi_proc_finalize()
revert bits of open-mpi/ompi@cf534d0c95
we cannot del_procs here since the pml framework has already been closed

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-12 11:41:35 +09:00
Ralph Castain
269e5e3e87 Merge pull request #2711 from rhc54/topic/numa
Protect default mapping/binding options for cases where no NUMA or SOCKET objects exist - like VMs
2017-01-11 11:29:15 -08:00
Ralph Castain
aff3a00059 Protect default mapping/binding options for cases where no NUMA or
SOCKET objects exist - like VMs

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-11 09:44:44 -08:00
Nysal Jan K.A
97801028ba asm/ppc: Fix a regression in powerpc atomics
Add a missing constraint to the input operand list.
This fixes a regression caused by d4be138a7b.
Thanks to Orion Poplawski for reporting the issue.

Refs #2610

Signed-off-by: Nysal Jan K.A <jnysal@in.ibm.com>
2017-01-11 11:00:11 -05:00
Ralph Castain
31a8476223 Merge pull request #2702 from rhc54/topic/cov
Silence Coverity CID 1398541
2017-01-10 17:50:23 -08:00
Jeff Squyres
5b1d4fddd2 Merge pull request #2692 from jsquyres/pr/master/usnic-queue-fixes
master: usnic queue fixes
2017-01-10 19:38:40 -05:00
Nathan Hjelm
79cabc92fd rcache/base: do not release vma stuctures in vma_tree_delete
This commit fixes a deadlock that can occur when the libc version
holds a lock when calling munmap. In this case we could end up calling
free() from vma_tree_delete which would in turn try to obtain the lock
in libc. To avoid the issue put any deleted vma's in a new list on the
vma module and release them on the next call to vma_tree_insert. This
should be safe as this function is not called from the memory hooks.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2017-01-10 16:58:07 -07:00
Ralph Castain
e568b211e4 Silence Coverity CID 1398541
Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-10 15:30:50 -08:00
Ralph Castain
d8ed7b55a3 Merge pull request #2701 from rhc54/topic/dvmfix
Be a tad more cautious before releasing objects when running in DVM mode
2017-01-10 15:01:41 -08:00
Ralph Castain
93e4935902 Be a tad more cautious before releasing objects when running in DVM mode
Fixes #2700

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
2017-01-10 14:04:27 -08:00
Jeff Squyres
b980e334dc usnic: add completion stats
This should probably not go to the v2.x branch, since it changes the
output format of the usnic stats.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:06:54 -08:00
Jeff Squyres
706f53bb01 usnic: ensure that stats string is always truncated
Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:06:54 -08:00
Jeff Squyres
1fdd0fe228 usnic: add missing params to show_help() call
Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:06:54 -08:00
Jeff Squyres
7048adec04 usnic: add some assert()s
Add some run-time assert checks for debug builds.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:06:32 -08:00
Jeff Squyres
2d28ccb5fd usnic: add verbose output of queue lengths
Show the actual RX/TX and CQ length returned by libfabric in verbose
output.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:06:32 -08:00
Jeff Squyres
bd5b8ed754 usnic: ensure that queues are long enough
Double check the queue lengths that we get back from libfabric to
ensure that they are at least as long as we need.  They *should* never
be shorter than we need, but let's just check to be sure.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:06:32 -08:00
Jeff Squyres
53dc75a89c usnic: ensure to reset flags on returned frags
Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:06:31 -08:00
Jeff Squyres
c4d7876ca0 usnic: check send credits on data channel for data frags
Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:06:31 -08:00
Jeff Squyres
879d25e5df usnic: ensure to check send credits for ACKs
Don't just blindly send ACKs; ensure that we have send credits before
doing so.  If we don't have any send credits, just don't send the ACK
(it'll come again soon enough; it's not a tragedy if we don't send it
now).

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:06:31 -08:00
Jeff Squyres
7787dad4db usnic: ensure CQs are long enough
The libfabric usnic provider may give you back TX/RX queues that are
longer than you asked for.  So just use the TX/RQ/CQ lengths that we
asked for, regardless of what length comes back.

Additionally, keep the length of the priority channel CQ separate from
the length of the data CQ.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:03:53 -08:00
Jeff Squyres
b02d8c48f5 usnic: make the releasing safer
Since the usnic BTL is single-threaded in this area, there really is
no danger, but don't use one of the pointers hanging off the frag
after we return it to the freelist.  Instead, save the endpoint
pointer before returning the frag.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:03:53 -08:00
Jeff Squyres
e25b860627 usnic: clarify types
The types are technically typedef equivalent, but it's less confusing
to use the types that agree with the name of the constructor.

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 12:03:53 -08:00
Jeff Squyres
40fe575132 usnic: trivial updates (no code/logic changes)
- Add more explanatory comments
- Trivial whitespace / style updates
- Rename opal_btl_usnic_force_retrans() -> opal_btl_usnic_fast_retrans()

Signed-off-by: Jeff Squyres <jsquyres@cisco.com>
2017-01-10 10:40:02 -08:00
Nathan Hjelm
3593fad4d2 Merge pull request #2679 from hjelmn/cpuid_fix
master: amd64: save/restore all 64 bits of rbx around cpuid
2017-01-10 09:12:28 -07:00
Gilles Gouaillardet
6d59b476de Merge pull request #2686 from ggouaillardet/topic/pmix2x_ptl_base_sendrecv
pmix2x: ptl/base: send header and message data together via writev()
2017-01-10 16:26:10 +09:00
Gilles Gouaillardet
44c1ff60f1 Merge pull request #2672 from ggouaillardet/topic/misc_memory_leaks
Plug misc memory leaks
2017-01-10 13:16:04 +09:00
Gilles Gouaillardet
a01960bee5 pmix2x: ptl/base: send header and message data together via writev()
on Linux, sending the header and then the message data does severely
impact performances of ptl/tcp :
on the receiver, reading the data can often result in an PMIX_ERR_RESOURCE_BUSY
or PMIX_ERR_WOULD_BLOCK, which ends up degrading performances)
this commit send both header and message data at the same time via writev()
and makes ptl/tcp virtually as efficient as ptl/usock.

Short writev generally occur when the kernel buffer is full, so there is no
point for retrying in this case.

fwiw, no such degradation was observed on OSX.

Refs open-mpi/ompi#2657

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2017-01-10 13:07:39 +09:00
Nathan Hjelm
b320882932 Merge pull request #2688 from hjelmn/cvar_fix
mca/base: account for NULL string_value in verbose set
2017-01-09 14:27:31 -07:00
Nathan Hjelm
d6bd69dc93 mca/base: account for NULL string_value in verbose set
The MCA variable code calls the string from value function with a NULL
string to verify values. The verbosity enumerator was not correctly
checking for a non-NULL value before trying to set the string.

Signed-off-by: Nathan Hjelm <hjelmn@lanl.gov>
2017-01-09 11:52:31 -07:00