The last 3 fields were 3 list heads that are per-thread, and which are:
- the pool's LRU head
- the buffer_wq
- the streams list head
Moving them into thread_ctx completes the removal of dynamic elements
from the struct thread_info. Now all these dynamic elements are packed
together at a single place for a thread.
The "show pools" command provides some "allocated" and "used" estimates
on the pools objects, but this applies to the shared pool and the "used"
includes what is currently assigned to thread-local caches. It's possible
to know how much each thread uses, so let's dump the total size allocated
by thread caches as an estimate. It's only done when pools are enabled,
which explains why the patch adds quite a lot of ifdefs.
pool-os.h relies on a number of includes solely because the
pool_alloc_area() function was inlined, and this only because we want
the normal version to be inlined so that we can track the calling
places for the memory profiler. It's worth noting that it already
does not work at -O0, and that when UAF is enabled we don't care a
dime about profiling.
This patch does two things at once:
- force-inline the functions so that pool_alloc_area() is still
inlined at -O0 to help track malloc() users ;
- uninline the UAF version of these (that rely on mmap/munmap)
and move them to pools.c so that we can remove all unneeded
includes.
Doing so reduces by ~270kB or 0.15% the total build size.
A number of files currently access activity counters but rely on their
definitions to be inherited from other files (task.c, backend.c hlua.c,
sock.c, pool.c, stats.c, fd.c).
Ilya reported in issue #1391 a build warning on Fedora about mallinfo()
being deprecated in favor of mallinfo2() since glibc-2.33. Let's add
support for it. This should be backported where the following commit is
also backported: 157e39303 ("MINOR: pools: automatically disable
malloc_trim() with external allocators").
Pierre Cheynier reported some occasional crashes in malloc_trim() on a
recent glibc when running with jemalloc(). While in theory there should
not be any link between the two, it remains plausible that something
allocated early with one is tentatively freed with the other and that
attempts to trim end up badly. There's no point calling the glibc specific
malloc_trim() with external allocators anyway. However these ones are often
enabled at link time or even at run time with LD_PRELOAD, so we cannot rely
on build options for this.
This patch implements runtime detection for the allocator in use by checking
with mallinfo() that a malloc() call is properly accounted for in glibc's
malloc. It only enables malloc_trim() in this case, and ignores it for
other cases. It's fine to proceed like this because mallinfo() is provided
by a wider range of glibcs than malloc_trim().
This could be backported to 2.4 and 2.3. If so, it will also need previous
patch "CLEANUP: pools: factor all malloc_trim() calls into trim_all_pools()".
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.
In GH issue #1275, Fabiano Nunes Parente provided a nicely detailed
report showing reproducible crashes under musl. Musl is one of the libs
coming with a simple allocator for which we prefer to keep the shared
cache. On x86 we have a DWCAS so the lockless implementation is enabled
for such libraries.
And this implementation has had a small race since day one: the allocator
will need to read the first object's <next> pointer to place it into the
free list's head. If another thread picks the same element and immediately
releases it, while both the local and the shared pools are too crowded, it
will be freed to the OS. If the libc's allocator immediately releases it,
the memory area is unmapped and we can have a crash while trying to read
that pointer. However there is no problem as long as the item remains
mapped in memory because whatever value found there will not be placed
into the head since the counter will have changed.
The probability for this to happen is extremely low, but as analyzed by
Fabiano, it increases with the buffer size. On 16 threads it's relatively
easy to reproduce with 2MB buffers above 200k req/s, where it should
happen within the first 20 seconds of traffic usually.
This is a structural issue for which there are two non-trivial solutions:
- place a read lock in the alloc call and a barrier made of lock/unlock
in the free() call to force to serialize operations; this will have
a big performance impact since free() is already one of the contention
points;
- change the allocator to use a self-locked head, similar to what is
done in the MT_LISTS. This requires two memory writes to the head
instead of a single one, thus the overhead is exactly one memory
write during alloc and one during free;
This patch implements the second option. A new POOL_DUMMY pointer was
defined for the locked pointer value, allowing to both read and lock it
with a single xchg call. The code was carefully optimized so that the
locked period remains the shortest possible and that bus writes are
avoided as much as possible whenever the lock is held.
Tests show that while a bit slower than the original lockless
implementation on large buffers (2MB), it's 2.6 times faster than both
the no-cache and the locked implementation on such large buffers, and
remains as fast or faster than the all implementations when buffers are
48k or higher. Tests were also run on arm64 with similar results.
Note that this code is not used on modern libcs featuring a fast allocator.
A nice benefit of this change is that since it removes a dependency on
the DWCAS, it will be possible to remove the locked implementation and
replace it with this one, that is then usable on all systems, thus
significantly increasing their performance with large buffers.
Given that lockless pools were introduced in 1.9 (not supported anymore),
this patch will have to be backported as far as 2.0. The code changed
several times in this area and is subject to many ifdefs which will
complicate the backport. What is important is to remove all the DWCAS
code from the shared cache alloc/free lockless code and replace it with
this one. The pool_flush() code is basically the same code as the
allocator, retrieving the whole list at once. If in doubt regarding what
barriers to use in older versions, it's safe to use the generic ones.
This patch depends on the following previous commits:
- MINOR: pools: do not maintain the lock during pool_flush()
- MINOR: pools: call malloc_trim() under thread isolation
- MEDIUM: pools: use a single pool_gc() function for locked and lockless
The last one also removes one occurrence of an unneeded DWCAS in the
code that was incompatible with this fix. The removal of the now unused
seq field will happen in a future patch.
Many thanks to Fabiano for his detailed report, and to Olivier for
his help on this issue.
Locked and lockless shared pools don't need to use a different pool_gc()
function because this function isolates itself during the operation, so
we do not need to rely on DWCAS nor any atomic operation in fact. Let's
just get rid of the lockless one in favor of the simple one. This should
even result in a faster execution.
The ifdefs were slightly moved so that we can have pool_gc() defined
as soon as there are global pools, this avoids duplicating the function.
pool_gc() was adjusted to run under thread isolation by commit c0e2ff202
("MEDIUM: memory: make pool_gc() run under thread isolation") so that the
underlying malloc() and free() don't compete between threads during these
potentially aggressive moments (especially when mmap/munmap are involved).
Commit 88366c292 ("MEDIUM: pools: call malloc_trim() from pool_gc()")
later added a call to malloc_trim() but made it outside of the thread
isolation, which is contrary to the principle explained above. Also it
missed it in the locked version, meaning that those without a lockless
implementation cannot benefit from trimming.
This patch fixes that by calling it before thread_release() in both
places.
The locked version of pool_flush() is absurd, it locks the pool for each
and every element to be released till the end. Not only this is extremely
inefficient, but it may even never finish if other threads spend their
time refilling the pool. The only case where this can happen is during
soft-stop so the risk remains limited, but it should be addressed.
Since the code was reorganized, DEBUG_UAF was still tested in the locked
pool code despite pools being disabled when DEBUG_UAF is used. Let's move
the test to pool_put_to_os() which is the one that is always called in
this condition.
The impact is only a possible misleading analysis during a troubleshooting
session due to a missing double-frees or free of const area test that is
normally already dealt with by the underlying code anyway. In practice it's
unlikely anyone will ever notice.
This should only be backported to 2.4.
The lockless version of pool_flush() had a leftover of the original
version causing the pool's first entry to be set to NULL at the end.
The problem is that it does this outside of any lock and in a non-
atomic way, so that any concurrent alloc+free would result in a lost
object.
The risk is low and the consequence even lower, given that pool_flush()
is only used in pool_destroy() (hence single-threaded) or by stream_free()
during a soft-stop (not the place where most allocations happen), so in
the worst case it could result in valgrind complaining on soft-stop.
The bug was introduced with the first version of the code, in 1.9, so
the fix can be backported to all stable versions.
There were 102 CLI commands whose help were zig-zagging all along the dump
making them unreadable. This patch realigns all these messages so that the
command now uses up to 40 characters before the delimiting colon. About a
third of the commands did not correctly list their arguments which were
added after the first version, so they were all updated. Some abuses of
the term "id" were fixed to use a more explanatory term. The
"set ssl ocsp-response" command was not listed because it lacked a help
message, this was fixed as well. The deprecated enable/disable commands
for agent/health/server were prominently written as deprecated. Whenever
possible, clearer explanations were provided.
The purpose of this debugging option was to prevent certain pools from
masking other ones when they were shared. For example, task, http_txn,
h2s, h1s, h1c, session, fcgi_strm, and connection are all 192 bytes and
would normally be mergedi, but not with this option. The problem is that
certain pools are declared multiple times with various parameters, which
are often very close, and due to the way the option works, they're not
shared either. Good examples of this are captures and stick tables. Some
configurations have large numbers of stick-tables of pretty similar types
and it's very common to end up with the following when the option is
enabled:
$ socat - /tmp/sock1 <<< "show pools" | grep stick
- Pool sticktables (160 bytes) : 0 allocated (0 bytes), 0 used, needed_avg 0, 0 failures, 1 users, @0x753800=56
- Pool sticktables (160 bytes) : 0 allocated (0 bytes), 0 used, needed_avg 0, 0 failures, 1 users, @0x753880=57
- Pool sticktables (160 bytes) : 0 allocated (0 bytes), 0 used, needed_avg 0, 0 failures, 1 users, @0x753900=58
- Pool sticktables (160 bytes) : 0 allocated (0 bytes), 0 used, needed_avg 0, 0 failures, 1 users, @0x753980=59
- Pool sticktables (160 bytes) : 0 allocated (0 bytes), 0 used, needed_avg 0, 0 failures, 1 users, @0x753a00=60
- Pool sticktables (160 bytes) : 0 allocated (0 bytes), 0 used, needed_avg 0, 0 failures, 1 users, @0x753a80=61
- Pool sticktables (160 bytes) : 0 allocated (0 bytes), 0 used, needed_avg 0, 0 failures, 1 users, @0x753b00=62
- Pool sticktables (224 bytes) : 0 allocated (0 bytes), 0 used, needed_avg 0, 0 failures, 1 users, @0x753780=55
In addition to not being convenient, it can have important effects on the
memory usage because these pools will not share their entries, so one stick
table cannot allocate from another one's pool.
This patch solves this by going back to the initial goal which was not to
have different pools in the same list. Instead of masking the MAP_F_SHARED
flag, it simply adds a test on the pool's name, and disables pool sharing
if the names differ. This way pools are not shared unless they're of the
same name and size, which doesn't hinder debugging. The same test above
now returns this:
$ socat - /tmp/sock1 <<< "show pools" | grep stick
- Pool sticktables (160 bytes) : 0 allocated (0 bytes), 0 used, needed_avg 0, 0 failures, 7 users, @0x3fadb30 [SHARED]
- Pool sticktables (224 bytes) : 0 allocated (0 bytes), 0 used, needed_avg 0, 0 failures, 1 users, @0x3facaa0 [SHARED]
This is much better. This should probably be backported, in order to limit
the side effects of DEBUG_DONT_SHARE_POOLS being enabled in production.
The current "ADD" vs "ADDQ" is confusing because when thinking in terms
of appending at the end of a list, "ADD" naturally comes to mind, but
here it does the opposite, it inserts. Several times already it's been
incorrectly used where ADDQ was expected, the latest of which was a
fortunate accident explained in 6fa922562 ("CLEANUP: stream: explain
why we queue the stream at the head of the server list").
Let's use more explicit (but slightly longer) names now:
LIST_ADD -> LIST_INSERT
LIST_ADDQ -> LIST_APPEND
LIST_ADDED -> LIST_INLIST
LIST_DEL -> LIST_DELETE
The same is true for MT_LISTs, including their "TRY" variant.
LIST_DEL_INIT keeps its short name to encourage to use it instead of the
lazier LIST_DELETE which is often less safe.
The change is large (~674 non-comment entries) but is mechanical enough
to remain safe. No permutation was performed, so any out-of-tree code
can easily map older names to new ones.
The list doc was updated.
Amaury noticed that I managed to break the build of DEBUG_FAIL_ALLOC
for the second time with 207c09509 ("MINOR: pools: move the fault
injector to __pool_alloc()"). The joy of endlessly reworking patch
sets... No backport is needed, that was in the just merged cleanup
series.
This function has become too big (251 bytes) and is now hurting
performance a lot, with up to 4% request rate being lost over the last
pool changes. Let's move it to pool.c as a regular function. Other
attempts were made to cut it in half but it's still inefficient. Doing
this results in saving ~90kB of object code, and even 112kB since the
pool changes, with code that is even slightly faster!
Conversely, pool_get_from_cache(), which remains half of this size, is
still faster inlined, likely in part due to the immediate use of the
returned pointer afterwards.
Till now it used to call it only if there were not too many objects into
the local cache otherwise would send the latest one directly into the
shared cache. Now it always sends to the local cache and it's up to the
local cache to free its oldest objects. From a cache freshness perspective
it's better this way since we always evict cold objects instead of hot
ones. From an API perspective it's better because it will help make the
shared cache invisible to the public API.
Till now we could only evict oldest objects from all local caches using
pool_evict_from_local_caches() until the cache size was satisfying again,
but there was no way to evict excess objects from a single cache, which
is the reason why pool_put_to_cache() used to refrain from putting into
the local cache and would directly write to the shared cache, resulting
in massive writes when caches were full.
Let's add this new function now. It will stop once the number of objects
in the local cache is no higher than 16+total/8 or the cache size is no
more than 75% full, just like before.
For now the function is not used.
Continuing the unification of local and shared pools, now the usage of
pools is governed by CONFIG_HAP_POOLS without which allocations and
releases are performed directly from the OS using pool_alloc_nocache()
and pool_free_nocache().
There are two levels of freeing to the OS:
- code that wants to keep the pool's usage counters updated uses
pool_free_area() and handles the counters itself. That's what
pool_put_to_shared_cache() does in the no-global-pools case.
- code that does not want to update the counters because they were
already updated only calls pool_free_area().
Let's extract these calls to establish the symmetry with pool_get_from_os()
and pool_alloc_nocache(), resulting in pool_put_to_os() (which only updates
the allocated counter) and pool_free_nocache() (which also updates the used
counter). This will later allow to simplify the generic code.
A part of the code cannot be factored out because it still uses non-atomic
inc/dec for pool->used and pool->allocated as these are located under the
pool's lock. While it can make sense in terms of bus cycles, it does not
make sense in terms of code normalization. Further, some operations were
still performed under a lock that could be totally removed via the use of
atomic ops.
There is still one occurrence in pool_put_to_shared_cache() in the locked
code where pool_free_area() is called under the lock, which must absolutely
be fixed.
Now there's one part dealing with the allocation itself and keeping
counters up to date, and another one on top of it to return such an
allocated pointer to the user and update the use count and stats.
This is in anticipation for being able to group cache-related parts.
The release code is still done at once.
Till now it was limited to objects allocated from the OS which means
it had little use as soon as pools were enabled. Let's move it upper
in the layers so that any code can benefit from fault injection. In
addition this allows to pass a new flag POOL_F_NO_FAIL to disable it
if some callers prefer a no-failure approach.
ha_random() is quite heavy and uses atomic ops or even a lock on some
architectures. Here we don't seek good randoms, just statistical ones,
so let's use the statistical prng instead.
Now the multi-level cache becomes more visible:
pool_get_from_local_cache()
pool_put_to_local_cache()
pool_get_from_shared_cache()
pool_put_to_shared_cache()
The functions were rightfully called from/to_cache when the thread-local
cache was considered as the only cache, but this is getting terribly
confusing. Let's call them from/to local_cache to make it clear that
it is not related with the shared cache.
As a side note, since pool_evict_from_cache() used not to work for a
particular pool but for all of them at once, it was renamed to
pool_evict_from_local_caches() (plural form).
They were strictly equivalent, let's remerge them and rename them to
pool_alloc_nocache() as it's the call which performs a real allocation
which does not check nor update the cache. The only difference in the
past was the former taking the lock and not the second but now the lock
is not needed anymore at this stage since the pool's list is not touched.
In addition, given that the "avail" argument is no longer used by the
function nor by its callers, let's drop it.
Now we don't loop anymore trying to refill multiple items at once, and
an allocated object is directly returned to the requester instead of
being stored into the shared pool. This has multiple benefits. The
first one is that no locking is needed anymore on the allocation path
and the second one is that the loop will no longer cause latency spikes.
This is a first step towards unifying all the fallback code. Right now
these two functions are the only ones which do not update the needed_avg
rate counter since there's currently no shared pool kept when using them.
But their code is similar to what could be used everywhere except for
this one, so let's make them capable of maintaining usage statistics.
As a side effect the needed field in "show pools" will now be populated.
The mem_should_fail() call enabled by DEBUG_FAIL_ALLOC used to be placed
only in the no-cache version of the allocator. Now we can generalize it
to all modes and remove the exclusive test on CONFIG_HAP_NO_GLOBAL_POOLS.
We're going to make the local pool always present unless pools are
completely disabled. This means that pools are always enabled by
default, regardless of the use of threads. Let's drop this notion
of "local" pools and make it just "pool". The equivalent debug
option becomes DEBUG_NO_POOLS instead of DEBUG_NO_LOCAL_POOLS.
For now this changes nothing except the option and dropping the
dependency on USE_THREAD.
Initially per-thread pool caches were stored into a fixed-size array.
But this was a bit ugly because the last allocated pools were not able
to benefit from the cache at all. As a work around to preserve
performance, a size of 64 cacheable pools was set by default (there
are 51 pools at the moment, excluding any addon and debugging code),
so all in-tree pools were covered, at the expense of higher memory
usage.
In addition an index had to be calculated for each pool, and was used
to acces the pool cache head into that array. The pool index was not
even stored into the pools so it was required to determine it to access
the cache when the pool was already known.
This patch changes this by moving the pool cache head into the pool
head itself. This way it is certain that each pool will have its own
cache. This removes the need for index calculation.
The pool cache head is 32 bytes long so it was aligned to 64B to avoid
false sharing between threads. The extra cost is not huge (~2kB more
per pool than before), and we'll make better use of that space soon.
The pool cache head contains the size, which should probably be removed
since it's already in the pool's head.
When building with DEBUG_FAIL_ALLOC we call a random generator to decide
whether the pool alloc should succeed or fail, and there was a preliminary
debugging mechanism to keep sort of a history of the previous decisions. But
it was never used, enforces a lock during the allocation, and forces to use
static variables, all of which are limiting the ability to pursue the pools
cleanups with no real benefit. Let's get rid of them now.
When running with CONFIG_HAP_NO_GLOBAL_POOLS, it's theoritically possible
to keep an incorrect count of allocated entries in a pool because the
allocated counter was used as a cumulated counter of alloc calls instead
of a number of currently allocated items (it's possible the meaning has
changed over time). The only impact in this mode essentially is that
"show pools" will report incorrect values. But this would only happen on
limited pools, which is not even certain still exist.
This was added by recent commit 0bae07592 ("MEDIUM: pools: add
CONFIG_HAP_NO_GLOBAL_POOLS and CONFIG_HAP_GLOBAL_POOLS") so no backport
is needed.
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.
This fixes a gcc warning about a missing const on defproxy for
mem_parse_global_fail_alloc.
This is needed since the commit :
018251667e
CLEANUP: config: make the cfg_keyword parsers take a const for the
defproxy
We've reached a point where the global pools represent a significant
bottleneck with threads. On a 64-core machine, the performance was
divided by 8 between 32 and 64 H2 connections only because there were
not enough entries in the local caches to avoid picking from the global
pools, and the contention on the list there was very high. It becomes
obvious that we need to have an array of lists, but that will require
more changes.
In parallel, standard memory allocators have improved, with tcmalloc
and jemalloc finding their ways through mainstream systems, and glibc
having upgraded to a thread-aware ptmalloc variant, keeping this level
of contention here isn't justified anymore when we have both the local
per-thread pool caches and a fast process-wide allocator.
For these reasons, this patch introduces a new compile time setting
CONFIG_HAP_NO_GLOBAL_POOLS which is set by default when threads are
enabled with thread local pool caches, and we know we have a fast
thread-aware memory allocator (currently set for glibc>=2.26). In this
case we entirely bypass the global pool and directly use the standard
memory allocator when missing objects from the local pools. It is also
possible to force it at compile time when a good allocator is used with
another setup.
It is still possible to re-enable the global pools using
CONFIG_HAP_GLOBAL_POOLS, if a corner case is discovered regarding the
operating system's default allocator, or when building with a recent
libc but a different allocator which provides other benefits but does
not scale well with threads.
The LRU cache head was an array of list, which causes false sharing
between 4 to 8 threads in the same cache line. Let's move it to the
thread_info structure instead. There's no need to do the same for the
pool_cache[] array since it's already quite large (32 pointers each).
By doing this the request rate increased by 1% on a 16-thread machine.
This patch fixes all the leftovers from the include cleanup campaign. There
were not that many (~400 entries in ~150 files) but it was definitely worth
doing it as it revealed a few duplicates.
Most of the files dealing with error reports have to include log.h in order
to access ha_alert(), ha_warning() etc. But while these functions don't
depend on anything, log.h depends on a lot of stuff because it deals with
log-formats and samples. As a result it's impossible not to embark long
dependencies when using ha_warning() or qfprintf().
This patch moves these low-level functions to errors.h, which already
defines the error codes used at the same places. About half of the users
of log.h could be adjusted, sometimes revealing other issues such as
missing tools.h. Interestingly the total preprocessed size shrunk by
4%.
There's no point splitting the file in two since only cfgparse uses the
types defined there. A few call places were updated and cleaned up. All
of them were in C files which register keywords.
There is nothing left in common/ now so this directory must not be used
anymore.
The current state of the logging is a real mess. The main problem is
that almost all files include log.h just in order to have access to
the alert/warning functions like ha_alert() etc, and don't care about
logs. But log.h also deals with real logging as well as log-format and
depends on stream.h and various other things. As such it forces a few
heavy files like stream.h to be loaded early and to hide missing
dependencies depending where it's loaded. Among the missing ones is
syslog.h which was often automatically included resulting in no less
than 3 users missing it.
Among 76 users, only 5 could be removed, and probably 70 don't need the
full set of dependencies.
A good approach would consist in splitting that file in 3 parts:
- one for error output ("errors" ?).
- one for log_format processing
- and one for actual logging.
