When the current check is overloaded (more running checks than the
configured limit), we'll try more aggressively to find another thread.
Instead of just opportunistically looking for one half as loaded, now if
the current thread has more than 1% more active checks than another one,
or has more than a configured limit of concurrent running checks, it will
search for a more suitable thread among 3 other random ones in order to
migrate the check there. The number of migrations remains very low (~1%)
and the checks load very fair across all threads (~1% as well). The new
parameter is called tune.max-checks-per-thread.
Let's keep two check counters per thread:
- one for "active" checks, i.e. checks that are no more sleeping
and are assigned to the thread. These include sleeping and
running checks ;
- one for "running" checks, i.e. those which are currently
executing on the thread.
By doing so, we'll be able to spread the health checks load a bit better
and refrain from sending too many at once per thread. The counters are
atomic since a migration increments the target thread's active counter.
These numbers are reported in "show activity", which allows to check
per thread and globally how many checks are currently pending and running
on the system.
Ideally, we should only consider checks in the process of establishing
a connection since that's really the expensive part (particularly with
OpenSSL 3.0). But the inner layers are really not suitable to doing
this. However knowing the number of active checks is already a good
enough hint.
We now count the number of times a check was started on each thread
and the number of times a check was adopted. This helps understand
better what is observed regarding checks.
The goal here is to explicitly mark that a check was migrated so that
we don't do it again. This will allow us to perform other actions on
the target thread while still knowing that we don't want to be migrated
again. The new READY bit combine with SLEEPING to form 4 possible states:
SLP RDY State Description
0 0 - (reserved)
0 1 RUNNING Check is bound to current thread and running
1 0 SLEEPING Check is sleeping, not bound to a thread
1 1 MIGRATING Check is migrating to another thread
Thus we set READY upon migration, and check for it before migrating, this
is sufficient to prevent a second migration. To make things a bit clearer,
the SLEEPING bit was switched with FASTINTER so that SLEEPING and READY are
adjacent.
The CHK_ST_SLEEPING state was introduced by commit d114f4a68 ("MEDIUM:
checks: spread the checks load over random threads") to indicate that
a check was not currently bound to a thread and that it could easily
be migrated to any other thread. However it did not start the checks
in this state, meaning that they were not redispatchable on startup.
Sometimes under heavy load (e.g. when using SSL checks with OpenSSL 3.0)
the cost of setting up new connections is so high that some threads may
experience connection timeouts on startup. In this case it's better if
they can transfer their excess load to other idle threads. By just
marking the check as sleeping upon startup, we can do this and
significantly reduce the number of failed initial checks.
When a thread creates a new session for a server, if none was known yet,
we assign the thread id (hence the reused_sess index) to a shared variable
so that other threads will later be able to find it when they don't have
one yet. For now we only set and clear the pointer upon session creation,
we do not yet pick it.
Note that we could have done it per thread-group, so as to avoid any
cross-thread exchanges, but it's anticipated that this is essentially
used during startup, at a moment where the cost of inter-thread contention
is very low compared to the ability to restart at full speed, which
explains why instead we store a single entry.
The goal will be to permit a thread to update its session while having
it shared with other threads. For now we only place the lock and arrange
the code around it so that this is quite light. For now only the owner
thread uses this lock so there is no contention.
Note that there is a subtlety in the openssl API regarding
i2s_SSL_SESSION() in that it fills the area pointed to by its argument
with a dump of the session and returns a size that's equal to the
previously allocated one. As such, it does modify the shared area even
if that's not obvious at first glance.
We already have a call that can retreive an TLV with any value.
Therefore, the fetch logic is redundant and can be simplified
by simply calling the generic fetch with the correct TLV ID
set as an argument.
In order to be able to implement fetches in the future that allow
retrieval of any TLVs, a new generic data structure for TLVs is introduced.
Existing TLV fetches for PP2_TYPE_AUTHORITY and PP2_TYPE_UNIQUE_ID are
migrated to use this new data structure. TLV related pools are updated
to not rely on type, but only on size. Pools accomodate the TLV list
element with their associated value. For now, two pools for 128 B and
256 B values are introduced. More fine-grained solutions are possible
in the future, if necessary.
This patch improves readability by scoping HA proxy related PPv2 constants
with a 'HA" prefix. Besides, a new constant for the length of a CRC32C
TLV is introduced. The length is derived from the PPv2 spec, so 32 Bit.
For a while there has been the constraint of having to run as root for
transparent proxying, and we're starting to see some cases where QUIC is
not running in socket-per-connection mode due to the missing capability
that would be needed to bind a privileged port. It's not realistic to
ask all QUIC users on port 443 to run as root, so instead let's provide
a basic support for capabilities at least on linux. The ones currently
supported are cap_net_raw, cap_net_admin and cap_net_bind_service. The
mechanism was made OS-specific with a dedicated file because it really
is. It can be easily refined later for other OSes if needed.
A new keyword "setcaps" is added to the global section, to enumerate the
capabilities that must be kept when switching from root to non-root. This
is ignored in other situations though. HAProxy has to be built with
USE_LINUX_CAP=1 for this to be supported, which is enabled by default
for linux-glibc, linux-glibc-legacy and linux-musl.
A good way to test this is to start haproxy with such a config:
global
uid 1000
setcap cap_net_bind_service
frontend test
mode http
timeout client 3s
bind quic4@:443 ssl crt rsa+dh2048.pem allow-0rtt
and run it under "sudo strace -e trace=bind,setuid", then connecting
there from an H3 client. The bind() syscall must succeed despite the
user id having been switched.
Bug was introduced by commit 26654 ("MINOR: ssl: add "crt" in the
cert_exts array").
When looking for a .crt directly in the cert_exts array, the
ssl_sock_load_pem_into_ckch() function will be called with a argument
which does not have its ".crt" extensions anymore.
If "ssl-load-extra-del-ext" is used this is not a problem since we try
to add the ".crt" when doing the lookup in the tree.
However when using directly a ".crt" without this option it will failed
looking for the file in the tree.
The fix removes the "crt" entry from the array since it does not seem to
be really useful without a rework of all the lookups.
Should fix issue #2265
Must be backported as far as 2.6.
Surprisingly there's no include guard in plock.h though there is one in
atomic-ops.h. Let's add one, or we cannot risk including the file multiple
times.
Backend idle connections are purged on a recurring occurence during the
process lifetime. An estimated number of needed connections is
calculated and the excess is removed periodically.
Before this patch, purge was done directly using the idle then the safe
connection tree of a server instance. This has a major drawback to take
no account of a specific ordre and it may removed functional connections
while leaving ones which will fail on the next reuse.
The problem can be worse when using criteria to differentiate idle
connections such as the SSL SNI. In this case, purge may remove
connections with a high rate of reusing while leaving connections with
criteria never matched once, thus reducing drastically the reuse rate.
To improve this, introduce an alternative storage for idle connection
used in parallel of the idle/safe trees. Now, each connection inserted
in one of this tree is also inserted in the new list at
`srv_per_thread.idle_conn_list`. This guarantees that recently used
connection is present at the end of the list.
During the purge, use this list instead of idle/safe trees. Remove first
connection in front of the list which were not reused recently. This
will ensure that connection that are frequently reused are not purged
and should increase the reuse rate, particularily if distinct idle
connection criterias are in used.
Define a new function _srv_add_idle(). This is a simple wrapper to
insert a connection in the server idle tree. This is reserved for simple
usage and require to idle_conns lock. In most cases,
srv_add_to_idle_list() should be used.
This patch does not have any functional change. However, it will help
with the next patch as idle connection will be always inserted in a list
as secondary storage along with idle/safe trees.
Small change of API for conn_delete_from_tree(). Now the connection
instance is taken as argument instead of its inner node.
No functional change introduced with this commit. This simplifies
slightly invocation of conn_delete_from_tree(). The most useful changes
is that this function will be extended in the next patch to be able to
remove the connection from its new idle list at the same time as in its
idle tree.
Replace ->lock type of pat_ref struct by HA_RWLOCK_T.
Replace all calls to HA_SPIN_LOCK() (resp. HA_SPIN_UNLOCK()) by HA_RWLOCK_WRLOCK()
(resp. HA_RWLOCK_WRUNLOCK()) when a write access is required.
There is only one read access which is needed. This is in the "show map" command
callback, cli_io_handler_map_lookup() where a HA_SPIN_LOCK() call is replaced
by HA_RWLOCK_RDLOCK() (resp. HA_SPIN_UNLOCK() by HA_RWLOCK_RDUNLOCK).
Replace HA_SPIN_INIT() calls by HA_RWLOCK_INIT() calls.
Store a pointer to the expression (struct pattern_expr) into the data structure
used to chain/store the map element references (struct pat_ref_elt) , e.g. the
struct pattern_tree when stored into an ebtree or struct pattern_list when
chained to a list.
Modify pat_ref_set_elt() to stop inspecting all the expressions attached to a map
and to look for the <elt> element passed as parameter to retrieve the sample data
to be parsed. Indeed, thanks to the pointer added above to each pattern tree nodes
or list elements, they all can be inspected directly from the <elt> passed as
parameter and its ->tree_head and ->list_head member: the pattern tree nodes are
stored into elt->tree_head, and the pattern list elements are chained to
elt->list_head list. This inspection was also the job of pattern_find_smp() which
is no more useful. This patch removes the code of this function.
Organize reference to pattern element of map (struct pat_ref_elt) into an ebtree:
- add an eb_root member to the map (pat_ref struct) and an ebpt_node to its
element (pat_ref_elt struct),
- modify the code to insert these nodes into their ebtrees each time they are
allocated. This is done in pat_ref_append().
Note that ->head member (struct list) of map (struct pat_ref) is not removed
could have been removed. This is not the case because still necessary to dump
the map contents from the CLI in the order the map elememnts have been inserted.
This patch also modifies http_action_set_map() which is the callback at least
used by "set-map" action. The pat_ref_elt element returned by pat_ref_find_elt()
is no more ignored, but reused if not NULL by pat_ref_set() as first element to
lookup from. This latter is also modified to use the ebtree attached to the map
in place of the ->head list attached to each map element (pat_ref_elt struct).
Also modify pat_ref_find_elt() to makes it use ->eb_root map ebtree added to the
map by this patch in place of inspecting all the elements with a strcmp() call.
HTTP/2 demux must be handled with care for active reverse connection.
Until accept has been completed, it should be forbidden to handle
HEADERS frame as session is not yet ready to handle streams.
To implement this, use the flag H2_CF_DEM_TOOMANY which blocks demux
process. This flag is automatically set just after conn_reverse()
invocation. The flag is removed on rev_accept_conn() callback via a new
H2 ctl enum. H2 tasklet is woken up to restart demux process.
As a side-effect, reporting in H2 mux may be blocked as demux functions
are used to convert error status at the connection level with
CO_FL_ERROR. To ensure error is reported for a reverse connection, check
h2c_is_dead() specifically for this case in h2_wake(). This change also
has its own side-effect : h2c_is_dead() conditions have been adjusted to
always exclude !h2c->conn->owner condition which is always true for
reverse connection or else H2 mux may kill them unexpectedly.
Implement active reverse connection initialization. This is done through
a new task stored in the receiver structure. This task is instantiated
via bind callback and first woken up via enable callback.
Task handler is separated into two halves. On the first step, a new
connection is allocated and stored in <pend_conn> member of the
receiver. This new client connection will proceed to connect using the
server instance referenced in the bind_conf.
When connect has successfully been executed and HTTP/2 connection is
ready for exchange after SETTINGS, reverse_connect task is woken up. As
<pend_conn> is still set, the second halve is executed which only
execute listener_accept(). This will in turn execute accept_conn
callback which is defined to return the pending connection.
The task is automatically requeued inside accept_conn callback if bind
maxconn is not yet reached. This allows to specify how many connection
should be opened. Each connection is instantiated and reversed serially
one by one until maxconn is reached.
conn_free() has been modified to handle failure if a reverse connection
fails before being accepted. In this case, no session exists to notify
about the failure. Instead, reverse_connect task is requeud with a 1
second delay, giving time to fix a possible network issue. This will
allow to attempt a new connection reverse.
Note that for the moment connection rebinding after accept is disabled
for simplicity. Extra operations are required to migrate an existing
connection and its stack to a new thread which will be implemented
later.
Implement parsing for "rev@" addresses on bind line. On config parsing,
server name is stored on the bind_conf.
Several new callbacks are defined on reverse_connect protocol to
complete parsing. listen callback is used to retrieve the server
instance from the bind_conf server name. If found, the server instance
is stored on the receiver. Checks are implemented to ensure HTTP/2
protocol only is used by the server.
Implement active reverse support inside conn_reverse(). This is used to
transfer the connection from the backend to the frontend side.
A new flag is defined CO_FL_REVERSED which is set just after this
transition. This will be used to identify connections which were
reversed but not yet accepted.
A new protocol named "reverse_connect" is created. This will be used to
instantiate connections that are opened by a reverse bind.
For the moment, only a minimal set of callbacks are defined with no real
work. This will be extended along the next patches.
On connection passive reverse from frontend to backend, its hash node is
calculated to be able to select it from the idle server pool. If
attach-srv rule defined an associated name, reuse it as the value for
SNI prehash.
This change allows a client to select a reverse connection by its name
by configuring its server line with a SNI to permit this.
Add an optional argument 'name' for attach-srv rule. This contains an
expression which will be used as an identifier inside the server idle
pool after reversal. To match this connection for a future transfer
through the server, the SNI server parameter must match this name. If no
name is defined, match will only occur with an empty SNI value.
For the moment, only the parsing step is implemented. An extra check is
added to ensure that the reverse server uses SSL with a SNI. Indeed, if
name is defined but server does not uses a SNI, connections will never
be selected on reused after reversal due to a hash mismatch.
Create a new tcp-request session rule 'attach-srv'.
The parsing handler is used to extract the server targetted with the
notation 'backend/server'. The server instance is stored in the act_rule
instance under the new union variant 'attach_srv'.
Extra checks are implemented in parsing to ensure attach-srv is only
used for proxy in HTTP mode and with listeners/server with no explicit
protocol reference or HTTP/2 only.
The action handler itself is really simple. It assigns the stored server
instance to the 'reverse' member of the connection instance. It will be
used in a future patch to implement passive reverse-connect.
A reverse server relies solely on its pool of idle connection to
transfer requests which will be populated through a new tcp-request rule
'attach-srv'.
Several changes are required on connect_server() to implement this.
First, reuse mode is forced to always for this type of server. Then, if
no idle connection is found, the request will be aborted. This results
with a 503 HTTP error code, similarly to when no server is available.
Implement reverse-connect server. This server type cannot instantiate
its own connection on transfer. Instead, it can only reuse connection
from its idle pool. These connections will be populated using the future
'tcp-request session attach-srv' rule.
A reverse-connect has no address. Instead, it uses a new custom server
notation with '@' character prefix. For the moment, only '@reverse' is
defined. An extra check is implemented to ensure server is used in a
HTTP proxy.
Reverse connection after SETTINGS reception if it was set as reversable.
This operation is done in a new function h2_conn_reverse(). It regroups
common changes which are needed for both reversal direction :
H2_CF_IS_BACK is set or unset and timeouts are inverted.
For the moment, only passive reverse is fully implemented. Once done,
the connection instance is directly inserted in its targetted server
pool. It can then be used immediately for future transfers using this
server.
Define a new method conn_reverse(). This method is used to reverse a
connection from frontend to backend or vice-versa depending on its
initial status.
For the moment, passive reverse only is implemented. This covers the
transition from frontend to backend side. The connection is detached
from its owner session which can then be freed. Then the connection is
linked to the server instance.
only for passive connection on
frontend to transfer them on the backend side. This requires to free the
connection session after detaching it from.
Several CLI handlers use a server argument specified with the format
'<backend>/<server>'. The parsing of this arguement is done in two
steps, first splitting the string with '/' delimiter and then use
get_backend_server() to retrieve the server instance.
Refactor this code sections with the following changes :
* splitting is reimplented using ist API
* get_backend_server() is removed. Instead use the already existing
proxy_be_by_name() then server_find_by_name() which contains
duplicated code with the now removed function.
No functional change occurs with this commit. However, it will be useful
to add new configuration options reusing the same '<backend>/<server>'
for reverse connect.
Peter Varkoly reported a build issue on ppc64le in xxhash.h. Our version
(0.8.1) was the last one 9 months ago, and since then this specific issue
was addressed in 0.8.2, so let's apply the maintenance update.
This should be backported to 2.8 and 2.7.
Compilation is broken due to missing __pl_wait_unlock_long() definition
when building with PLOCK_DISABLE_EBO=1. This has been introduced since
the following commit which activates the inlining version of
pl_wait_unlock_long() :
commit 071d689a51
MINOR: threads: inline the wait function for pthread_rwlock emulation
Add an extra check on PLOCK_DISABLE_EBO before choosing the inline or
default version of pl_wait_unlock_long() to fix this.
There were a few places left where we forgot to call __ha_cpu_relax()
after a failed CAS, in the HA_ATOMIC_UPDATE_{MIN,MAX} macros, and in
a few sync_* API macros (the same as above plus HA_ATOMIC_CAS and
HA_ATOMIC_XCHG). Let's add them now.
This could have been a cause of contention, particularly with
process_stream() calling stream_update_time_stats() which uses 8 of them
in a call (4 for the server, 4 for the proxy). This may be a possible
explanation for the high CPU consumption reported in GH issue #2251.
This should be backported at least to 2.6 as it's harmless.
When using pthread_rwlock emulation, contention is reported on
pl_wait_unlock_long(). This is really not convenient to analyse what is
happening. Now plock supports inlining the wait call for just the lorw
functions by enabling PLOCK_LORW_INLINE_WAIT. Let's do this so that now
the wait time will be precisely reported as either pthread_rwlock_rdlock()
or pthread_rwlock_wrlock() depending on the contended function, but no
more on pl_wait_unlock_long(), which will still be reported for all
other locks.
Now when PLOCK_LORW_INLINE_WAIT is defined, the pl_wait_unlock_long()
calls in pl_lorw_rdlock() and pl_lorw_wrlock() will be inlined so that
all the CPU time is accounted for in the calling function.
This is plock upstream commit c993f81d581732a6eb8fe3033f21970420d21e5e.
Doing so will allow to expose the time spent in certain highly
contended functions, which can be desirable for more accurate CPU
profiling. For example this could be done in locking functions that
are already not inlined so that they are the ones being reported as
those consuming the CPU instead of just pl_wait_unlock_long().
This is plock upstream commit 7505c2e2c8c4aa0ab8f52a2288e1334ae6412be4.
Commit 9db830b ("plock: support inlining exponential backoff code")
added an option to support inlining of the wait code for longs but
forgot to do it for ints. Let's do it now.
This is plock upstream commit b1f9f0d252fa40577d11cfb2bc0a809d6960a297.
This aims at further reducing the contention on the free_list when using
global pools. The free_list pointer now appears for each bucket, and both
the alloc and the release code skip to a next bucket when ending on a
contended entry. The default entry used for allocations and releases
depend on the thread ID so that locality is preserved as much as possible
under low contention.
It would be nice to improve the situation to make sure that releases to
the shared pools doesn't consider the first entry's pointer but only an
argument that would be passed and that would correspond to the bucket in
the thread's cache. This would reduce computations and make sure that the
shared cache only contains items whose pointers match the same bucket.
This was not yet done. One possibility could be to keep the same splitting
in the local cache.
With this change, an h2load test with 5 * 160 conns & 40 streams on 80
threads that was limited to 368k RPS with the shared cache jumped to
3.5M RPS for 8 buckets, 4M RPS for 16 buckets, 4.7M RPS for 32 buckets
and 5.5M RPS for 64 buckets.
The failed allocation counter cannot depend on a pointer, but since it's
a perpetually increasing counter and not a gauge, we don't care where
it's incremented. Thus instead we're hashing on the TID. There's no
contention there anyway, but it's better not to waste the room in
the pool's heads and to move that with the other counters.
That's the same principle as for ->allocated and ->used. Here we return
the summ of the raw values, so the result still needs to be fed to
swrate_avg(). It also means that we now use the local ->used instead
of the global one for the calculations and do not need to call pool_used()
anymore on fast paths. The number of samples should likely be divided by
the number of buckets, but that's not done yet (better observe first).
A function pool_needed_avg() was added to report aggregated values for
the "show pools" command.
With this change, an h2load made of 5 * 160 conn * 40 streams on 80
threads raised from 1.5M RPS to 6.7M RPS.
That's the same principle as for ->allocated. The small difference here
is that it's no longer possible to decrement ->used in batches when
releasing clusters from the cache to the shared cache, so the counter
has to be decremented for each of them. But as it provides less
contention and it's done only during forced eviction, it shouldn't be
a problem.
A function "pool_used()" was added to return the sum of the entries.
It's used by pool_alloc_nocache() and pool_free_nocache() which need
to count the number of used entries. It's not a problem since such
operations are done when picking/releasing objects to/from the OS,
but it is a reminder that the number of buckets should remain small.
With this change, an h2load test made of 5 * 160 conn * 40 streams on
80 threads raised from 812k RPS to 1.5M RPS.
The ->used counter is one of the most stressed, and it heavily
depends on the ->allocated one, so let's first move ->allocated
to a few buckets.
A function "pool_allocated()" was added to return the sum of the entries.
It's important not to abuse it as it does iterate, so everywhere it's
possible to avoid it by keeping a local counter, it's better. Currently
it's used for limited pools which need to make sure they do not allocate
too many objects. That's an acceptable tradeoff to save CPU on large
machines at the expense of spending a little bit more on small ones which
normally are not under load.
On many threads and without the shared cache, there can be extreme
contention on the ->allocated counter, the ->free_list pointer, and
the ->used counter. It's possible to limit this contention by spreading
the counters a little bit over multiple entries, that are summed up when
a consultation is needed. The criterion used to spread the values cannot
be related to the thread ID due to migrations, since we need to keep
consistent stats (allocated vs used).
Instead we'll just hash the pointer, it provides an index that does the
job and that is consistent for the object. When having just a few entries
(16 here as it showed almost identical performance between global and
non-global pools) even iterations should be short enough during
measurements to not be a problem.
A pair of functions designed to ease pointer hash bucket calculation were
added, with one of them doing it for thread IDs because allocation failures
will be associated with a thread and not a pointer.
For now this patch only brings in the relevant parts of the infrastructure,
the CONFIG_HAP_POOL_BUCKETS_BITS macro that defaults to 6 bits when 512
threads or more are supported, 5 bits when 128 or more are supported, 4
bits when 16 or more are supported, otherwise 3 bits for small setups.
The array in the pool_head and the two utility functions are already
added. It should have no measurable impact beyond inflating the pool_head
structure.
The pool's allocation counter doesn't strictly require to be updated
from these functions, it may more efficiently be done in the caller
(even out of a loop for pool_flush() and pool_gc()), and doing so will
also help us spread the counters over an array later. The functions
were renamed _noinc and _nodec to make sure we catch any possible
user in an external patch. If needed, the original functions may easily
be reimplemented in an inline function.
During tests it was noticed that the current hash is not that good
on 4- and 5- bit hashes. About 7.5% of all the 32-bit primes were tested
as candidates for the hash function, by submitting them 128 arrangements
of N pointers among 40k extracted from haproxy's pools, and the average
fill rates for 1- to 12- bit hashes were measured and compared. It was
clear that some values do not provide great hashes and other ones are
way more resistant.
The current value is not bad at all but delivers 42.6% unique 2-bit
outputs, 41.6% 3-bit, 38.0% 4-bit, 38.2% 5-bit and 37.1% 10-bit. Some
values did perform significantly better, among which 0xacd1be85 which
does 43.2% 2-bit, 42.5% 3-bit, 42.2% 4-bit, 39.2% 5-bit and 37.3% 10-bit.
The reverse value used in the ptr2_hash() was really underperforming and
was replaced with 0x9d28e4e9 which does 49.6%, 40.4%, 42.6%, 39.1%, and
37.2% respectvely.
This should slightly improve the accuracy of the task and memory
profiling, and will be useful for pools.
When testing the pointer hash on 64-bit real pointers (map entries),
it appeared that the shift by 33 bits that hoped to compensate for the
3 nul LSB degrades the hash, and the centering is more optimal on
31-(bits+1)/2. This makes sense since the topmost bit of the
multiplicator is 31, so for an input of 1 bit and 1 bit of output we
would always get zero. With the formula adjusted this way, we can get
up to ~15% more unique entries at 10 bits and ~24% more at 11 bits.
When dealing with macro-based size definitions, it is useful to be able
to hash pointers on zero bits so that the macro automatically returns a
constant 0. For now it only supports 1-32. Let's just add this special
case. It's automatically optimized out by the compiler since the function
is inlined.
LONGBITS was defined long ago with old compilers that didn't provide the
word size. It's still present as being referenced in various places in the
code, but we must not use it to define other macros that may be evaluated
at pre-processing time since it contains sizeof() and casts that are not
compatible with preprocessor conditions. Let's switch MAX_THREADS_PER_GROUP
to __WORDSIZE so that we can condition blocks of code on it if needed.
LONGBITS should really be removed by now, given that we don't support
compilers not providing __WORDSIZE anymore (gcc < 4.2).
By moving the config-time stuff after the updt_lock, we can plug some
holes without interfering with it. This allows us to get back to the
768-bytes struct. The performance was not affected at all.
The read-lock contention observed on the update lock while turning it
into an upgradable lock were due to false sharing with the nearby
updates. Simply moving the lock alone into its own cache line is
sufficient to almost double the performance again, raising from 2355
to 4480k RPS with very low contention:
Samples: 1M of event 'cycles', 4000 Hz, Event count (approx.): 743422995452 lost
Overhead Shared Object Symbol
15.88% haproxy [.] stktable_lookup_key
5.94% haproxy [.] ebmb_lookup
5.69% haproxy [.] http_wait_for_request
3.66% haproxy [.] stktable_touch_with_exp
2.62% [kernel] [k] _raw_spin_unlock_irqrestore
1.86% haproxy [.] http_action_return
1.79% haproxy [.] stream_process_counters
1.78% [kernel] [k] skb_release_data
1.77% haproxy [.] process_stream
Unfortunately, trying to move the line anywhere else didn't work,
despite the remaining holes, because this structure is not quite
clean. This adds 64 bytes to a struct that was already 768 long,
so it's now 832. It's possible to repack it a little bit and regain
these bytes by removing the THREAD_ALIGN before "keys" because we
rarely use the config stuff, but that's a bit unsafe.
Updating an entry in the updates tree is currently performed under the
table's write lock, which causes huge contention with other accesses
such as lookups and free. Aside the updates tree, the update,
localupdate and commitupdate variables, nothing is manipulated, so
let's create a distinct lock (updt_lock) to protect these together
to remove this contention. It required to add an extra lock in the
few places where we delete the update (though only if we're really
going to delete it) to protect the tree. This is very convenient
because now peer_send_teachmsgs() only needs to take this read lock,
and there is very little contention left on the stick-table.
With this alone, the performance jumped from 614k to 1140k/s on a
80-thread machine with a peers section! Stick-table updates with
no peers however now has to stand two locks and slightly regressed
from 4.0-4.1M/s to 3.9-4.0. This is fairly minimal compared to the
significant unlocking of the peers updates and considered totally
acceptable.
The structure currently mixes R/O and R/W fields, let's organize them
by access type, focusing mainly on splitting the updates from the rest
so that peers activity does not affect the rest. For now it doesn't
bring any benefit but it paves the way for splitting the lock.
Due to the ts->ref_cnt being manipulated and checked inside wrlocks,
we continue to have it updated under plenty of read locks, which have
an important cost on many-thread machines.
This patch turns them all to atomic ops and carefully moves them outside
of locks every time this is possible:
- the ref_cnt is incremented before write-unlocking on creation otherwise
the element could vanish before we can do it
- the ref_cnt is decremented after write-locking on release
- for all other cases it's updated out of locks since it's guaranteed by
the sequence that it cannot vanish
- checks are done before locking every time it's used to decide
whether we're going to release the element (saves several write locks)
- expiration tests are just done using atomic loads, since there's no
particular ordering constraint there, we just want consistent values.
For Lua, the loop that is used to dump stick-tables could switch to read
locks only, but this was not done.
For peers, the loop that builds updates in peer_send_teachmsgs is extremely
expensive in write locks and it doesn't seem this is really needed since
the only updated variables are last_pushed and commitupdate, the first
one being on the shared table (thus not used by other threads) and the
commitupdate could likely be changed using a CAS. Thus all of this could
theoretically move under a read lock, but that was not done here.
On a 80-thread machine with a peers section enabled, the request rate
increased from 415 to 520k rps.
This updates the local copy of the plock library to benefit from finer
memory ordering, EBO on more operations such as when take_w() and stow()
wait for readers to leave and refined EBO, especially on common operation
such as attempts to upgade R to S, and avoids a counter-productive prior
read in rtos() and take_r().
These changes have shown a 5% increase on regular operations on ARM,
a 33% performance increase on ARM on stick-tables and 2% on x86, and
a 14% and 4% improvements on peers updates respectively on ARM and x86.
The availability of relaxed operations will probably be useful for stats
counters which are still extremely expensive to update.
The following plock commits were included in this update:
9db830b plock: support inlining exponential backoff code
008d3c2 plock: make the rtos upgrade faster
2f76dde atomic: clean up the generic xchg()
3c6919b atomic: make sure that the no-return macros do not return a value
97c2bb7 atomic: make the fallback bts use the pointed type for the shift
f4c1880 atomic: also implement the missing pl_btr()
8329b82 atomic: guard all generic definitions to make it easier to provide specific ones
7c5cb62 atomic: use C11 atomics when available
96afaf9 atomic: prefer the C11 definitions in general
f3ec7a6 atomic: implement load/store/atomic barriers
8bdbd1e atomic: add atomic load/stores
0f604c0 atomic: add more _noret operations
3fe35db atomic: remove the (void) cast from the C11 operations
3b08a7c atomic: allow to define the fallback _noret variants
28deb22 atomic: make x86 arithmetic operations the _noret variants
8061fe2 atomic: handle modern compilers that support returning flags
b8b91b7 atomic: add the fetch-and-<op> operations (pl_ld<op>)
59817ca atomic: add memory order variants for most operations
a40774f plock: explicitly make use of the pl_*_noret operations
6f1861b plock: switch to pl_sub_noret_lax() for cancellation
c013980 plock: use pl_ldadd{_lax,_acq,} instead of pl_xadd()
382eea3 plock: use a release ordering when dropping the lock
60d750d plock: use EBO when waiting for readers to leave in take_w() and stow()
fc01c4f plock: improve EBO a little bit
1ef6390 plock: switch to CAS + XADD for pl_take_r()
->xprt_ctx (struct ssl_sock_ctx) and ->conn (struct connection) must be kept
by the remaining QUIC connection object (struct quic_cc_conn) after having
release the previous one (struct quic_conn) to allow "show fd/sess" commands
to be functional without causing haproxy crashes.
No need to backport.
We're adding a new argument "relaxed" to h2_make_htx_request() so that
we can control its level of acceptance of certain invalid requests at
the proxy level with "option accept-invalid-http-request". The goal
will be to add deactivable checks that are still desirable to have by
default. For now no test is subject to it.
As its name implies, this function checks if a path component has any
forbidden headers starting at the designated location. The goal is to
seek from the result of a successful ist_find_range() for more precise
chars. Here we're focusing on 0x00-0x1F, 0x20 and 0x23 to make sure
we're not too strict at this point.
This looks up the character range <min>..<max> in the input string and
returns a pointer to the first one found. It's essentially the equivalent
of ist_find_ctl() in that it searches by 32 or 64 bits at once, but deals
with a range.
This function is not H2 specific but rather generic to HTTP. We'll
need it in H3 soon, so let's move it to HTTP and rename it to
http_header_has_forbidden_char().
When the connection enters the "connection closing" state after having sent
a datagram with CONNECTION_CLOSE frames inside its packets, a lot of memory
may be freed from quic_conn objects (QUIC connection). This is done allocating
a reduced sized object which keeps enough information to handle the remaining
incoming packets for the connection in "connection closing" state, and to
continue to send again the previous datagram with CONNECTION_CLOSE frames inside
which has already been sent.
Define a new quic_cc_conn struct which represents the connection object after
entering the "connection close" state and after having release the quic_conn
connection object.
Define <pool_head_quic_cc_conn> new pool for these quic_cc_conn struct objects.
Define QUIC_CONN_COMMON structure which is shared between quic_conn struct object
(the connection before entering "connection close" state), and new quic_cc_conn
struct object (the connection after entering "connection close"). So, all the
members inside QUIC_CONN_COMMON may be indifferently dereferenced from a
quic_conn struct or a quic_cc_conn struct pointer.
Implement qc_new_cc_conn() function to allocate such connections in
"connection close" state. This function is responsible of copying the
required information from the original connection (quic_conn) to the remaining
connection (quic_cc_conn). Among others initialization, it redefined the
QUIC packet handler task to quic_cc_conn_io_cb() and the idle timer task
to qc_cc_idle_timer_task(). quic_cc_conn_io_cb() drains the received and
resend the datagram which CONNECTION_CLOSE frame which has already been sent
when entering "connection close" state. qc_cc_idle_timer_task() only releases
the remaining quic_cc_conn struct object.
Modify quic_conn_release() to allocate quic_cc_conn struct objects from the
original connection passed as argument. It does nothing if this original
connection is not in closing state, or if the idle timer has already expired.
Implement quic_release_cc_conn() to release a "connection close" connection.
It is called when its timer expires or if an error occured when sending
a packet from this connection when the peer is no more reachable.
Add "quic_cids" new pool to allocate the ->cids trees of quic_conn objects.
Replace ->cids member of quic_conn objects by pointer to "quic_cids" and
adapt the code consequently. Nothing special.
Add a new pool <pool_head_quic_cc_buf> for buffer used when building datagram
wich CONNECTION_CLOSE frames inside with QUIC_MIN_CC_PKTSIZE(128) as minimum
size.
Add ->cc_buf_area to quic_conn struct to store such buffers.
Add ->cc_dgram_len to store the size of the "connection close" datagrams
and ->cc_buf a buffer struct to be used with ->cc_buf_area as ->area member
value.
Implement qc_get_txb() to be called in place of qc_txb_alloc() to allocate
a struct "quic_cc_buf" buffer when the connection needs an immediate close
or a buffer struct if not.
Modify qc_prep_hptks() and qc_prep_app_pkts() to allow them to use such
"quic_cc_buf" buffer when an immediate close is required.
Move rx.bytes, tx.bytes and tx.prep_bytes quic_conn struct member to
bytes anonymous struct (bytes.rx, bytes.tx and bytes.prep member respectively).
They are moved before being defined into a bytes anonoymous struct common to
a future struct to be defined.
Consequently adapt the code.
Add a BUG_ON() to quic_peer_validated_addr() to check the amplification limit
is respected when it return false(0), i.e. when the connection is not validated.
Implement quic_may_send_bytes() which returns the number of bytes which may be
sent when the connection has not already been validated and call this functions
at several places when this is the case (after having called
quic_peer_validated_addr()).
Furthermore, this patch improves the code maintainability. Some patches to
come will have to rename ->[rt]x.bytes quic_conn struct members.
A HTTP server may provide a complete response even prior receiving the
full request. In this case, RFC 9114 allows the server to abort read
with a STOP_SENDING with error code H3_NO_ERROR.
This scenario was notably reproduced with haproxy and an inactive
server. If the client send a POST request, haproxy may provide a full
HTTP 503 response before the end of the full request.
Fields of sedesc structure were documented in the comment about the
structure itself. It was not really convenient, hard to read, hard to
update. So comments about the fields are moved on the corresponding field
line, as usual.
There is a mechanisme in the H1 and H2 multiplexer to skip the payload when
a response is returned to the client when it must not contain any payload
(response to a HEAD request or a 204/304 response). However, this does not
work when the splicing is used. The H2 multiplexer does not support the
splicing, so there is no issue. But with the mux-h1, when data are sent
using the kernel splicing, the mux on the server side is not aware the
client side should skip the payload. And once the data are put in a pipe,
there is no way to stop the sending.
It is a defect of the current design. This will be easier to deal with this
case when the mux-to-mux forwarding will be implemented. But for now, to fix
the issue, we should add an HTX flag on the start-line to pass the info from
the client side to the server side and be able to disable the splicing in
necessary.
The associated reg-test was improved to be sure it does not fail when the
splicing is configured.
This patch should be backported as far as 2.4..
Move the TX part of the code to quic_tx.c.
Add quic_tx-t.h and quic_tx.h headers for this TX part code.
The definition of quic_tx_packet struct has been move from quic_conn-t.h to
quic_tx-t.h.
Same thing for the TX part:
Move the RX part of the code to quic_rx.c.
Add quic_rx-t.h and quic_rx.h headers for this TX part code.
The definition of quic_rx_packet struct has been move from quic_conn-t.h to
quic_rx-t.h.
Move the code which directly calls the functions of the OpenSSL QUIC API into
quic_ssl.c new C file.
Some code have been extracted from qc_conn_finalize() to implement only
the QUIC TLS part (see quic_tls_finalize()) into quic_tls.c.
qc_conn_finalize() has also been exported to be used from this new quic_ssl.c
C module.
To accelerate the compilation of quic_conn.c file, export the code in relation
with the traces from quic_conn.c to quic_trace.c.
Also add some headers (quic_trace-t.h and quic_trace.h).
This setting which may be used into a "global" section, enables the QUIC listener
bindings when haproxy is compiled with the OpenSSL wrapper. It has no effect
when haproxy is compiled against a TLS stack with QUIC support, typically quictls.
This wrapper needs to have an access to an encoded version of the local transport
parameter (to be sent to the peer). They are provided to the TLS stack thanks to
qc_ssl_compat_add_tps_cb() callback.
These encoded transport parameters were attached to the QUIC connection but
removed by this commit to save memory:
MINOR: quic: Stop storing the TX encoded transport parameters
This patch restores these transport parameters and attaches them again
to the QUIC connection (quic_conn struct), but only when the QUIC OpenSSL wrapper
is compiled.
Implement qc_set_quic_transport_params() to encode the transport parameters
for a connection and to set them into the stack and make this function work
for both the OpenSSL wrapper or any other TLS stack with QUIC support. Its uses
the encoded version of the transport parameters attached to the connection
when compiled for the OpenSSL wrapper, or local parameters when compiled
with TLS stack with QUIC support. These parameters are passed to
quic_transport_params_encode() and SSL_set_quic_transport_params() as before
this patch.
Include haproxy/quic_openssl_compat.h from haproxy/openssl-compat.h when the
compilation of the QUIC openssl wrapper for TLS stacks is enabled with
USE_QUIC_OPENSSLCOMPAT.
Highly inspired from nginx openssl wrapper code.
This wrapper implement this list of functions:
SSL_set_quic_method(),
SSL_quic_read_level(),
SSL_quic_write_level(),
SSL_set_quic_transport_params(),
SSL_provide_quic_data(),
SSL_process_quic_post_handshake()
and SSL_QUIC_METHOD QUIC specific bio method which are also implemented by quictls
to support QUIC from OpenSSL. So, its aims is to support QUIC from a standard OpenSSL
stack without QUIC support. It relies on the OpenSSL keylog feature to retreive
the secrets derived by the OpenSSL stack during a handshake and to pass them to
the ->set_encryption_secrets() callback as this is done by quictls. It makes
usage of a callback (quic_tls_compat_msg_callback()) to handle some TLS messages
only on the receipt path. Some of them must be passed to the ->add_handshake_data()
callback as this is done with quictls to be sent to the peer as CRYPTO data.
quic_tls_compat_msg_callback() callback also sends the received TLS alert with
->send_alert() callback.
AES 128-bits with CCM mode is not supported at this time. It is often disabled by
the OpenSSL stack, but as it can be enabled by "ssl-default-bind-ciphersuites",
the wrapper will send a TLS alerts (Handhshake failure) if this algorithm is
negotiated between the client and the server.
0rtt is also not supported by this wrapper.
The aim of this patch is to allow the building of QUIC datagrams with
as much as packets with different encryption levels inside during handshake.
At this time, this is possible only for at most two encryption levels.
That said, most of the time, a server only needs to use two encryption levels
by datagram, except during retransmissions.
Modify qc_prep_pkts(), the function responsible of building datagrams, to pass
a list of encryption levels as parameter in place of two encryption levels. This
function is also used when retransmitting datagrams. In this case this is a
customized/flexible list of encryption level which is passed to this function.
Add ->retrans new member to quic_enc_level struct, to be used as attach point
to list of encryption level used only during retransmission, and ->retrans_frms
new member which is a pointer to a list of frames to be retransmitted.
This context may be released at the same time as the Initial TLS context.
This is done calling quic_tls_ctx_secs_free() and pool_free() in two code locations.
Implement quic_nictx_free() to do that.
Shorten ->negotiated_ictx quic_conn struct member (->nictx).
This variable is used during version negotiation. Indeed, a connection
may have to support support several QUIC versions of paquets during
the handshake. ->nictx is the QUIC TLS cipher context used for the negotiated
QUIC version.
This patch allows a connection to dynamically allocate this TLS cipher context.
Add a new pool (pool_head_quic_tls_ctx) for such QUIC TLS cipher context object.
Modify qc_new_conn() to initialize ->nictx to NULL value.
quic_tls_ctx_secs_free() frees all the secrets attached to a QUIC TLS cipher context.
Modify it to do nothing if it is called with a NULL TLS cipher context.
Modify to allocate ->nictx from qc_conn_finalize() just before initializing
its secrets. qc_conn_finalize() allocates -nictx only if needed (if a new QUIC
version was negotiated).
Modify qc_conn_release() which release a QUIC connection (quic_conn struct) to
release ->nictx TLS cipher context.
There is no need to keep an encoded version of the QUIC listener transport
parameters attache to the connection.
Remove ->enc_params and ->enc_params_len member of quic_conn struct.
Use variables to build the encoded transport parameter local to
ha_quic_set_encryption_secrets() before they are passed to
SSL_set_quic_transport_params().
Modify qc_ssl_sess_init() prototype. It was expected to be used with
the encoded transport parameters as passed parameter, but they were not
used. Cleanup this function.
parse_cpu_set() stopped returning the undocumented -1 which was a
leftover from an earlier attempt, changed from ulong to int since
it only returns a success/failure and no more a mask. Thus it must
not return -1 and its callers must only test for != 0, as is
documented.
This field used to store the cpumap of the first thread in a group, and
was used till 2.4 to hold some default settings, after which it was no
longer used. Let's just drop it.
We're currently having a problem with the porting from cpu_map from
processes to thread-groups as it happened in 2.7 with commit 5b09341c0
("MEDIUM: cpu-map: replace the process number with the thread group
number"), though it seems that it has deeper roots even in 2.0 and
that it was progressively made worng over time.
The issue stems in the way the per-process and per-thread cpu-sets were
employed over time. Originally only processes were supported. Then
threads were added after an optional "/" and it was documented that
"cpu-map 1" is exactly equivalent to "cpu-map 1/all" (this was clarified
in 2.5 by commit 317804d28 ("DOC: update references to process numbers
in cpu-map and bind-process").
The reality is different: when processes were still supported, setting
"cpu-map 1" would apply the mask to the process itself (and only when
run in the background, which is not documented either and is also a
bug for another fix), and would be combined with any possible per-thread
mask when calculating the threads' affinity, possibly resulting in empty
sets. However, "cpu-map 1/all" would only set the mask for the threads
and not the process. As such the following:
cpu-map 1 odd
cpu-map 1/1-8 even
would leave no CPU while doing:
cpu-map 1/all odd
cpu-map 1/1-8 even
would allow all CPUs.
While such configs are very unlikely to ever be met (which is why this
bug is tagged minor), this is becoming quite more visible while testing
automatic CPU binding during 2.9 development because due to this bug
it's much more common to end up with incorrect bindings.
This patch fixes it by simply removing the .proc entry from cpu_map and
always setting all threads' maps. The process is no longer arbitrarily
bound to the group 1's mask, but in case threads are disabled, we'll
use thread 1's mask since it contains the configured CPUs.
This fix should be backported at least to 2.6, but no need to insist if
it resists as it's easier to break cpu-map than to fix an unlikely issue.
Since we'll soon want to adjust the "thread-groups" degree of freedom
based on the presence of cpu-map, we first need to be able to detect
if cpu-map was used. This function scans all cpu-map sets to detect if
any is present, and returns true accordingly.
lua_yieldk ctx argument is of type lua_KContext which is typedefed to
intptr_t when available so it can be used to store pointers.
But the wrapper function hlua_yieldk() passes it as a regular it so it
breaks that promise.
Changing hlua_yieldk() prototype so that ctx argument is of type
lua_KContext.
This bug had no functional impact because ctx argument is not being
actively used so far. This may be backported to all stable versions
anyway.
