| Age | Commit message (Collapse) | Author |
|---|
|
The removal set now uses a red-black tree instead of an array on
large inputs, decreasing runtime complexity from `n*n` to
`n*log(n)`. It will also exit early when there are no more items
left in the removal set, drastically improving performance and
memory use when the items to be removed are present near the head
of the list.
This got a lot more complicated than before as the overhead of
always using a red-black tree was unacceptable when either of the
inputs were small, but this compromise has okay-to-decent
performance regardless of input size.
Co-authored-by: Dmytro Lytovchenko <[email protected]>
|
|
|
|
* sverker/erts/trap_exit-race/OTP-13452:
erts: Fix race for process_flag(trap_exit,true)
|
|
* rickard/proc-free-fix/OTP-13446:
Fix bad refc management of process struct
# Conflicts:
# erts/emulator/beam/erl_process.c
|
|
* rickard/port-sig-dropped-fix/OTP-13424:
Fix implementation of dropped signal to port
|
|
* rickard/ethr-event-futex-wait-timeout/OTP-13420:
Fix premature timeouts for ethread events on Linux
|
|
* rickard/last_calls/OTP-13418:
Unbreak process_info(Pid,last_calls)
|
|
and a concurrent exit signal.
We now actually guarantee that the process will not die
from exit signal *after* the call to process_flag(trap_exit,true)
has returned.
The race is narrow and probably quite hard to observe even if you
manage to provoke it. Has only been confirmed with the help of
return trace and a sleep in send_exit_signal().
Solution:
Seize status lock to prevent send_exit_signal() from reading
an old status (without TRAP_EXIT) and then writing PENDING_EXIT
after TRAP_EXIT has been set by process_flag_2().
|
|
|
|
|
|
|
|
|
|
This was accidentally removed in commit cd6903be0740db
|
|
|
|
|
|
|
|
No point in checking tmp_alloc instance 0
as any non-scheduler thread could race us.
|
|
One little (unsigned long) left behind.
|
|
|
|
* ia/msantos/unsupported-protocols:
Skips any sockets with unsupported protocols.
|
|
Patch from end of PR 864
Fixes problem reported in the PR by jeckersb
In this error case, a negative file descriptor is now saved at
listensock[i], which leads to a buffer overflow later on at
https://github.com/msantos/otp/blob/epmd-IPv6-node-reg2/erts/epmd/src/epmd_srv.c#L498
when calling FD_ISSET(listensock[i],&read_mask).
To reproduce this:
Completely disable IPv6 with ipv6.disable=1 on the kernel command line
Build with fortified code (RHEL/CentOS default): CFLAGS='-O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -m64 -mtune=generic -fno-strict-aliasing'
|
|
OTP-13251
* sverk/halt-INT_MIN:
erts: Make erlang:halt() accept bignums as Status
erts: Change erl_exit into erts_exit
kernel: Remove calls to erl_exit
|
|
* msantos/epmd-IPv6-node-reg2/PR-864/OTP-13364:
epmd: support IPv6 node registration
|
|
|
|
Just mask away the high bits to get a more tolerant erlang:halt
that behaves the same on 32 and 64 bit architectures.
|
|
This is mostly a pure refactoring.
Except for the buggy cases when calling erlang:halt() with a positive
integer in the range -(INT_MIN+2) to -INT_MIN that got confused with
ERTS_ABORT_EXIT, ERTS_DUMP_EXIT and ERTS_INTR_EXIT.
Outcome OLD erl_exit(n, ) NEW erts_exit(n, )
------- ------------------- -------------------------------------------
exit(Status) n = -Status <= 0 n = Status >= 0
crashdump+abort n > 0, ignore n n = ERTS_ERROR_EXIT < 0
The outcome of the old ERTS_ABORT_EXIT, ERTS_INTR_EXIT and
ERTS_DUMP_EXIT are the same as before (even though their values have
changed).
|
|
* rickard/proc-not-running/OTP-13365:
Do not wait for main lock when looking up process not running
|
|
|
|
|
|
|
|
* rickard/dw-cmpxchg-clang/OTP-13336:
Improve cmpxchg8b/cmpxchg16b inline asm
|
|
* rickard/gcc5-pic-cmpxchg8b/OTP-13326:
Improve cmpxchg8b inline asm configure test
|
|
* sverk/proc-lock-check-fix:
erts: Fix lock checker for process locks
|
|
|
|
* rickard/rq-state-bug/OTP-13298:
Fix bug causing run-queue mask to become inconsistent
|
|
Do lock order check *before* trying to seize lock... duh!
|
|
Clang didn't like that ecx/rcx was mapped to input and output
variables of different types.
|
|
* sverk/nif-test-cuddle:
erts: Fix wobbling test failure in nif_SUITE
|
|
* msantos/ms/typo-windows-build-script:
Fix typo in Windows build scripts
OTP-13337
|
|
|
|
|
|
|
|
This commit implements erts_internal:system_check(schedulers) with the
intent of a basic responsiveness test check of the schedulers.
|
|
ResA may have been GC'd after its last use.
|
|
Allow IPv6 nodes to register with and query epmd. On systems with
IPv6 support:
* epmd listens on both the IPv4 and IPv6 ANY or loopback sockets
* the epmd cli client connects to epmd over the IPv6 loopback
* distributed nodes started with "-proto_dist inet6_tcp" will register
with epmd over IPv6
To work on IPv6 capable systems that have IPv6 support disabled,
epmd ignores errors opening the socket if the protocol is not
supported. Similarly, the epmd client will fall back to IPv4 if the IPv6
socket is not available.
Update the minimum supported version of Windows to Windows Vista to
support IPv6.
|
|
* sverk/hipe-line-table-bug/OTP-13282:
erts: Fix bug concerning line information for hipe modules
|
|
* sverk/proc-exiting-timer-race/OTP-13245:
erts: Fix race between receive timeout and exit signal
|
|
|
|
* binarin/epmd-race:
Prevent down nodes going undetected in epmd
OTP-13301
|
|
* lukas/erts/enomem_no_abort/OTP-13292:
erts: When erts_alloc fails, the emulator no longer aborts
|
