thread_isolate() and thread_isolate_full() were relying on a set of thread
masks for all threads in different states (rdv, harmless, idle). This cannot
work anymore when the number of threads increases beyond LONGBITS so we need
to change the mechanism.
What is done here is to have a counter of requesters and the number of the
current isolated thread. Threads which want to isolate themselves increment
the request counter and wait for all threads to be marked harmless (or idle)
by scanning all groups and watching the respective masks. This is possible
because threads cannot escape once they discover this counter, unless they
also want to isolate and possibly pass first. Once all threads are harmless,
the requesting thread tries to self-assign the isolated thread number, and
if it fails it loops back to checking all threads. If it wins it's guaranted
to be alone, and can drop its harmless bit, so that other competing threads
go back to the loop waiting for all threads to be harmless. The benefit of
proceeding this way is that there's very little write contention on the
thread number (none during work), hence no cache line moves between caches,
thus frozen threads do not slow down the isolated one.
Once it's done, the isolated thread resets the thread number (hence lets
another thread take the place) and decrements the requester count, thus
possibly releasing all harmless threads.
With this change there's no more need for any global mask to synchronize
any thread, and we only need to loop over a number of groups to check
64 threads at a time per iteration. As such, tinfo's threads_want_rdv
could be dropped.
This was tested with 64 threads spread into 2 groups, running 64 tasks
(from the debug dev command), 20 "show sess" (thread_isolate()), 20
"add server blah/blah" (thread_isolate()), and 20 "del server blah/blah"
(thread_isolate_full()). The load remained very low (limited by external
socat forks) and no stuck nor starved thread was found.
The thread group info is not sufficient to represent a thread group's
current state as it's read-only. We also need something comparable to
the thread context to represent the aggregate state of the threads in
that group. This patch introduces ha_tgroup_ctx[] and tg_ctx for this.
It's indexed on the group id and must be cache-line aligned. The thread
masks that were global and that do not need to remain global were moved
there (want_rdv, harmless, idle).
Given that all the masks placed there now become group-specific, the
associated thread mask (tid_bit) now switches to the thread's local
bit (ltid_bit). Both are the same for nbtgroups 1 but will differ for
other values.
There's also a tg_ctx pointer in the thread so that it can be reached
from other threads.
This function was added in 2.0 when reworking the thread isolation
mechanism to make it more reliable. However it if fundamentally
incompatible with the full isolation mechanism provided by
thread_isolate_full() since that one will wait for all threads to
become idle while the former will wait for all threads to finish
waiting, causing a deadlock.
Given that it's not used, let's just drop it entirely before it gets
used by accident.
In order to kill all_threads_mask we'll need to have an equivalent for
the thread groups. The all_tgroups_mask does just this, it keeps one bit
set per enabled group.
At several places we're dereferencing the thread group just to catch
the group number, and this will become even more required once we start
to use per-group contexts. Let's just add the tgid in the thread_info
struct to make this easier.
For now we still set tid_bit to (1UL << tid) because FDs will not
work with more than one group without this, but once FDs start to
adopt local masks this must change to thr->ltid_bit.
Each thread has its own local thread id and its own global thread id,
in addition to the masks corresponding to each. Once the global thread
ID can go beyond 64 it will not be possible to have a global thread Id
bit anymore, so better start to remove it and use only the local one
from the struct thread_info.
In the configuration sometimes we'll omit a thread group number to designate
a global thread number range, and sometimes we'll mention the group and
designate IDs within that group. The operation is more complex than it
seems due to the need to check for ranges spanning between multiple groups
and determining groups from threads from bit masks and remapping bit masks
between local/global.
This patch adds a function to perform this operation, it takes a group and
mask on input and updates them on output. It's designed to be used by "bind"
lines but will likely be usable at other places if needed.
For situations where specified threads do not exist in the group, we have
the choice in the code between silently fixing the thread set or failing
with a message. For now the better option seems to return an error, but if
it turns out to be an issue we can easily change that in the future. Note
that it should only happen with "x/even" when group x only has one thread.
This is the equivalent of "tid" for ease of access. In the future if we
make th_cfg a pure thread-local array (not a pointer), it may make sense
to move it there.
ha_set_tid() was randomly used either to explicitly set thread 0 or to
set any possibly incomplete thread during boot. Let's replace it with
a pointer to a valid thread or NULL for any thread. This allows us to
check that the designated threads are always valid, and to ignore the
thread 0's mapping when setting it to NULL, and always use group 0 with
it during boot.
The initialization code is also cleaner, as we don't pass ugly casts
of a thread ID to a pointer anymore.
This takes care of unassigned threads groups and places unassigned
threads there, in a more or less balanced way. Too sparse allocations
may still fail though. For now with a maximum group number fixed to 1
nothing can really fail.
A the "tg" thread-local variable now always points to the current
thread group. It's pre-initializd to the first one during boot and is
set to point to the thread's one by ha_set_tid(). This last one takes
care of checking whether the thread group was assigned or not because
it may be called during boot before threads are initialized.
The scheduler contains a lot of stuff that is thread-local and not
exclusively tied to the scheduler. Other parts (namely thread_info)
contain similar thread-local context that ought to be merged with
it but that is even less related to the scheduler. However moving
more data into this structure isn't possible since task.h is high
level and cannot be included everywhere (e.g. activity) without
causing include loops.
In the end, it appears that the task_per_thread represents most of
the per-thread context defined with generic types and should simply
move to tinfo.h so that everyone can use them.
The struct was renamed to thread_ctx and the variable "sched" was
renamed to "th_ctx". "sched" used to be initialized manually from
run_thread_poll_loop(), now it's initialized by ha_set_tid() just
like ti, tid, tid_bit.
The memset() in init_task() was removed in favor of a bss initialization
of the array, so that other subsystems can put their stuff in this array.
Since the tasklet array has TL_CLASSES elements, the TL_* definitions
was moved there as well, but it's not a problem.
The vast majority of the change in this patch is caused by the
renaming of the structures.
These ones are rarely used or only to waste CPU cycles waiting, and are
the last ones requiring system includes in thread.h. Let's uninline them
and move them to thread.c.
haproxy.c still has to deal with pthread-specific low-level stuff that
is OS-dependent. We should not have to deal with this there, and we do
not need to access pthread anywhere else.
Let's move these 3 functions to thread.c and keep empty inline ones for
when threads are disabled.
The lock-debugging code in thread.h has no reason to be inlined. the
functions are quite fat and perform a lot of operations so there's no
saving keeping them inlined. Worse, most of them are in fact not
inlined, resulting in a significantly bigger executable.
This patch moves all this part from thread.h to thread.c. The functions
are still exported in thread.h of course. This results in ~166kB less
code:
text data bss dec hex filename
3165938 99424 897376 4162738 3f84b2 haproxy-before
2991987 99424 897376 3988787 3cdd33 haproxy-after
In addition the build time with thread debugging enabled has shrunk
from 19.2 to 17.7s thanks to much less code to be parsed in thread.h
that is included virtually everywhere.
We must protect from concurrent the tree which stores the QUIC packets received
by the dgram I/O handler, these packets being also parsed by the xprt task.
"f(void)" is the correct and preferred form for a function taking no
argument, while some places use the older "f()". These were reported
by clang's -Wmissing-prototypes, for example:
src/cpuset.c:111:5: warning: no previous prototype for function 'ha_cpuset_size' [-Wmissing-prototypes]
int ha_cpuset_size()
include/haproxy/cpuset.h:42:5: note: this declaration is not a prototype; add 'void' to make it a prototype for a zero-parameter function
int ha_cpuset_size();
^
void
This aggregate patch fixes this for the following functions:
ha_backtrace_to_stderr(), ha_cpuset_size(), ha_panic(), ha_random64(),
ha_thread_dump_all_to_trash(), get_exec_path(), check_config_validity(),
mworker_child_nb(), mworker_cli_proxy_(create|stop)(),
mworker_cleantasks(), mworker_cleanlisteners(), mworker_ext_launch_all(),
mworker_reload(), mworker_(env|proc_list)_to_(proc_list|env)(),
mworker_(un|)block_signals(), proxy_adjust_all_maxconn(),
proxy_destroy_all_defaults(), get_tainted(),
pool_total_(allocated|used)(), thread_isolate(_full|)(),
thread(_sync|)_release(), thread_harmless_till_end(),
thread_cpu_mask_forced(), dequeue_all_listeners(), next_timer_expiry(),
wake_expired_tasks(), process_runnable_tasks(), init_acl(),
init_buffer(), (de|)init_log_buffers(), (de|)init_pollers(),
fork_poller(), pool_destroy_all(), pool_evict_from_local_caches(),
pool_total_failures(), dump_pools_to_trash(), cfg_run_diagnostics(),
tv_init_(process|thread)_date(), __signal_process_queue(),
deinit_signals(), haproxy_unblock_signals()
Building with an old musl-based toolchain reported this warning:
include/haproxy/thread.h: In function 'ha_thread_relax':
include/haproxy/thread.h:256:5: warning: "_POSIX_PRIORITY_SCHEDULING" is not defined [-Wundef]
#if _POSIX_PRIORITY_SCHEDULING
^
There were indeed two "#if" insteadd of #ifdef" for this macro, let's
fix them.
The current principle of running under isolation was made to access
sensitive data while being certain that no other thread was using them
in parallel, without necessarily having to place locks everywhere. The
main use case are "show sess" and "show fd" which run over long chains
of pointers.
The thread_isolate() call relies on the "harmless" bit that indicates
for a given thread that it's not currently doing such sensitive things,
which is advertised using thread_harmless_now() and which ends usings
thread_harmless_end(), which also waits for possibly concurrent threads
to complete their work if they took this opportunity for starting
something tricky.
As some system calls were notoriously slow (e.g. mmap()), a bunch of
thread_harmless_now() / thread_harmless_end() were placed around them
to let waiting threads do their work while such other threads were not
able to modify memory contents.
But this is not sufficient for performing memory modifications. One such
example is the server deletion code. By modifying memory, it not only
requires that other threads are not playing with it, but are not either
in the process of touching it. The fact that a pool_alloc() or pool_free()
on some structure may call thread_harmless_now() and let another thread
start to release the same object's memory is not acceptable.
This patch introduces the concept of "idle threads". Threads entering
the polling loop are idle, as well as those that are waiting for all
others to become idle via the new function thread_isolate_full(). Once
thread_isolate_full() is granted, the thread is not idle anymore, and
it is released using thread_release() just like regular isolation. Its
users have to keep in mind that across this call nothing is granted as
another thread might have performed shared memory modifications. But
such users are extremely rare and are actually expecting this from their
peers as well.
Note that that in case of backport, this patch depends on previous patch:
MINOR: threads: make thread_release() not wait for other ones to complete
Till now whenever a server or proxy's queue was touched, this server
or proxy's lock was taken. Not only this requires distinct code paths,
but it also causes unnecessary contention with other uses of these locks.
This patch adds a lock inside the "queue" structure that will be used
the same way by the server and the proxy queuing code. The server used
to use a spinlock and the proxy an rwlock, though the queue only used
it for locked writes. This new version uses a spinlock since we don't
need the read lock part here. Tests have not shown any benefit nor cost
in using this one versus the rwlock so we could change later if needed.
The lower contention on the locks increases the performance from 362k
to 374k req/s on 16 threads with 20 servers and leastconn. The gain
with roundrobin even increases by 9%.
This is tagged medium because the lock is changed, but no other part of
the code touches the queues, with nor without locking, so this should
remain invisible.
This reverts commit fcb8bf8650.
The recent changes since 5304669e1 MEDIUM: queue: make
pendconn_process_next_strm() only return the pendconn opened a tiny race
condition between stream_free() and process_srv_queue(), as the pendconn
is accessed outside of the lock, possibly while it's being freed. A
different approach is required.
Till now whenever a server or proxy's queue was touched, this server
or proxy's lock was taken. Not only this requires distinct code paths,
but it also causes unnecessary contention with other uses of these locks.
This patch adds a lock inside the "queue" structure that will be used
the same way by the server and the proxy queuing code. The server used
to use a spinlock and the proxy an rwlock, though the queue only used
it for locked writes. This new version uses a spinlock since we don't
need the read lock part here. Tests have not shown any benefit nor cost
in using this one versus the rwlock so we could change later if needed.
The lower contention on the locks increases the performance from 491k
to 507k req/s on 16 threads with 20 servers and leastconn. The gain
with roundrobin even increases by 6%.
The performance profile changes from this:
13.03% haproxy [.] fwlc_srv_reposition
8.08% haproxy [.] fwlc_get_next_server
3.62% haproxy [.] process_srv_queue
1.78% haproxy [.] pendconn_dequeue
1.74% haproxy [.] pendconn_add
to this:
11.95% haproxy [.] fwlc_srv_reposition
7.57% haproxy [.] fwlc_get_next_server
3.51% haproxy [.] process_srv_queue
1.74% haproxy [.] pendconn_dequeue
1.70% haproxy [.] pendconn_add
At this point the differences are mostly measurement noise.
This is tagged medium because the lock is changed, but no other part of
the code touches the queues, with nor without locking, so this should
remain invisible.
Since threads were introduced in 1.8, the USE_PRIVATE_CACHE mode of the
shctx was not updated to use locks. Originally it was meant to disable
sharing between processes, so it removes the lock/unlock instructions.
But with threads enabled, it's not possible to work like this anymore.
It's easy to see that once built with private cache and threads enabled,
sending violent SSL traffic to the the process instantly makes it die.
The HTTP cache is very likely affected as well.
This patch addresses this by falling back to our native spinlocks when
USE_PRIVATE_CACHE is used. In practice we could use them also for other
modes and remove all older implementations, but this patch aims at keeping
the changes very low and easy to backport. A new SHCTX_LOCK label was
added to help with debugging, but OTHER_LOCK might be usable as well
for backports.
An even lighter approach for backports may consist in always declaring
the lock (or reusing "waiters"), and calling pl_take_s() for the lock()
and pl_drop_s() for the unlock() operation. This could even be used in
all modes (process and threads), even when thread support is disabled.
Subsequent patches will further clean up this area.
This patch must be backported to all supported versions since 1.8.
Now that the modified lockless variant does not need a DWCAS anymore,
there's no reason to keep the much slower locked version, so let's
just get rid of it.
Create a function thread_cpu_mask_forced. Its purpose is to report if a
restrictive cpu mask is active for the current proces, for example due
to a taskset invocation. It is only implemented for the linux platform
currently.
When a thread ends its harmeless period, we must only consider running
threads when testing threads_want_rdv_mask mask. To do so, we reintroduce
all_threads_mask mask in the bitwise operation (It was removed to fix a
deadlock).
Note that for now it is useless because there is no way to stop threads or
to have threads reserved for another task. But it is safer this way to avoid
bugs in the future.
A previous patch was pushed to fix a deadlock when an isolated thread ends
its harmless period (a9a9e9aac ["BUG/MEDIUM: thread: Fix a deadlock if an
isolated thread is marked as harmless"]). But, unfortunately, the fix is
incomplete. The same must be done in the outer loop, in
thread_harmless_end() function. The current thread must be ignored when
threads_want_rdv_mask mask is tested.
This patch must also be backported as far as 2.0.
This patch replaces roughly all occurrences of an HA_ATOMIC_ADD(&foo, 1)
or HA_ATOMIC_SUB(&foo, 1) with the equivalent HA_ATOMIC_INC(&foo) and
HA_ATOMIC_DEC(&foo) respectively. These are 507 changes over 45 files.
The lock was still used exclusively to deal with the concurrency between
the "show sess" release handler and a stream_new() or stream_free() on
another thread. All other accesses made by "show sess" are already done
under thread isolation. The release handler only requires to unlink its
node when stopping in the middle of a dump (error, timeout etc). Let's
just isolate the thread to deal with this case so that it's compatible
with the dump conditions, and remove all remaining locking on the streams.
This effectively kills the streams lock. The measured gain here is around
1.6% with 4 threads (374krps -> 380k).
Since OTHER_LOCK is commonly used it's become much more difficult to
profile lock contention by temporarily changing a lock label. Let's
add DEBUG1..5 to serve only for debugging. These ones must not be
used in committed code. We could decide to only define them when
DEBUG_THREAD is set but that would complicate attempts at measuring
performance with debugging turned off.
This is a preparation work for connection reuse with sni/proxy
protocol/specific src-dst addresses.
Protect every access to idle conn lists with a lock. This is currently
strictly not needed because the access to the list are made with atomic
operations. However, to be able to reuse connection with specific
parameters, the list storage will be converted to eb-trees. As this
structure does not have atomic operation, it is mandatory to protect it
with a lock.
For this, the takeover lock is reused. Its role was to protect during
connection takeover. As it is now extended to general idle conns usage,
it is renamed to idle_conns_lock. A new lock section is also
instantiated named IDLE_CONNS_LOCK to isolate its impact on performance.
Amaury reported that the commit 3ce6eed ("MEDIUM: ssl: add a rwlock for
SSL server session cache") introduced some warning during compilation:
include/haproxy/thread.h|411 col 2| warning: enumeration value 'SSL_SERVER_LOCK' not handled in switch [-Wswitch]
This patch fix the issue by adding the right entry in the switch block.
Must be backported where 3ce6eed is backported. (2.4 only for now)
When adding the server side support for certificate update over the CLI
we encountered a design problem with the SSL session cache which was not
locked.
Indeed, once a certificate is updated we need to flush the cache, but we
also need to ensure that the cache is not used during the update.
To prevent the use of the cache during an update, this patch introduce a
rwlock for the SSL server session cache.
In the SSL session part this patch only lock in read, even if it writes.
The reason behind this, is that in the session part, there is one cache
storage per thread so it is not a problem to write in the cache from
several threads. The problem is only when trying to write in the cache
from the CLI (which could be on any thread) when a session is trying to
access the cache. So there is a write lock in the CLI part to prevent
simultaneous access by a session and the CLI.
This patch also remove the thread_isolate attempt which is eating too
much CPU time and was not protecting from the use of a free ptr in the
session.
In addition to the previous simplification, most locks don't use the
seek or read lock (e.g. spinlocks etc) so let's split the dump into
distinct operations (write/seek/read) and only report those which
were used. Now the output size is roughly divided by 5 compared
to previous ones.
The lock stats are very verbose and more than half of them are used in
a typical test, making it hard to spot the sought values. Let's simply
report "not used" for those which have not been called at all.
Since our locks are based on progressive locks, we support the upgradable
seek lock that is compatible with readers and upgradable to a write lock.
The main purpose is to take it while seeking down a tree for modification
while other threads may seek the same tree for an input (e.g. compute the
next event date).
The newly supported operations are:
HA_RWLOCK_SKLOCK(lbl,l) pl_take_s(l) /* N --> S */
HA_RWLOCK_SKTOWR(lbl,l) pl_stow(l) /* S --> W */
HA_RWLOCK_WRTOSK(lbl,l) pl_wtos(l) /* W --> S */
HA_RWLOCK_SKTORD(lbl,l) pl_stor(l) /* S --> R */
HA_RWLOCK_WRTORD(lbl,l) pl_wtor(l) /* W --> R */
HA_RWLOCK_SKUNLOCK(lbl,l) pl_drop_s(l) /* S --> N */
HA_RWLOCK_TRYSKLOCK(lbl,l) (!pl_try_s(l)) /* N -?> S */
HA_RWLOCK_TRYRDTOSK(lbl,l) (!pl_try_rtos(l)) /* R -?> S */
Existing code paths are left unaffected so this patch doesn't affect
any running code.
We currently use only read and write lock operations with rwlocks, but
ours also support upgradable seek locks for which we do not report any
stats. Let's add them now when DEBUG_THREAD is enabled.
The thread_info struct is convenient to store various per-thread info
without having to resort to a painful thread_local storage which is
slow and painful to initialize.
The problem is, by having this one in thread.h it's very difficult to
add more entries there because everyone already includes thread.h so
conversely thread.h cannot reference certain types.
There's no point in having this there, instead let's create a new pair
of files, tinfo{,-t}.h, which declare the structure. This way it will
become possible to extend them with other includes and have certain
files store their own types there.
clang just failed on fd.c with this error:
src/fd.c:491:9: error: logical not is only applied to the left hand side of this comparison [-Werror,-Wlogical-not-parentheses]
while (HA_SPIN_TRYLOCK(OTHER_LOCK, &log_lock) != 0) {
^ ~~
That's because this expands to this:
while (!pl_try_s(&log_lock) != 0) {
Let's just add parenthesis in the TRYLOCK macros to avoid this.
This may need to be backported if commit df187875d ("BUG/MEDIUM: log:
don't hold the log lock during writev() on a file descriptor") is
backported as well as it seems to be the first one to trigger it.
This splits the hathreads.h file into types+macros and functions. Given
that most users of this file used to include it only to get the definition
of THREAD_LOCAL and MAXTHREADS, the bare minimum was placed into thread-t.h
(i.e. types and macros).
All the thread management was left to haproxy/thread.h. It's worth noting
the drop of the trailing "s" in the name, to remove the permanent confusion
that arises between this one and the system implementation (no "s") and the
makefile's option (no "s").
For consistency, src/hathreads.c was also renamed thread.c.
A number of files were updated to only include thread-t which is the one
they really needed.
Some future improvements are possible like replacing empty inlined
functions with macros for the thread-less case, as building at -O0 disables
inlining and causes these ones to be emitted. But this really is cosmetic.
