This reverts commit 072e774939.
Doing h2load with h3 tests we notice this behavior:
Client ---- INIT no token SCID = a , DCID = A ---> Server (1)
Client <--- RETRY+TOKEN DCID = a, SCID = B ---- Server (2)
Client ---- INIT+TOKEN SCID = a , DCID = B ---> Server (3)
Client <--- INIT DCID = a, SCID = C ---- Server (4)
Client ---- INIT+TOKEN SCID = a, DCID = C ---> Server (5)
With (5) dropped by haproxy due to token validation.
Indeed the previous patch adds SCID of retry packet sent to the aad
of the token ciphering aad. It was useful to validate the next INIT
packets including the token are sent by the client using the new
provided SCID for DCID as mantionned into the RFC 9000.
But this stateless information is lost on received INIT packets
following the first outgoing INIT packet from the server because
the client is also supposed to re-use a second time the lastest
received SCID for its new DCID. This will break the token validation
on those last packets and they will be dropped by haproxy.
It was discussed there:
https://mailarchive.ietf.org/arch/msg/quic/7kXVvzhNCpgPk6FwtyPuIC6tRk0/
To resume: this is not the role of the server to verify the re-use of
retry's SCID for DCID in further client's INIT packets.
The previous patch must be reverted in all versions where it was
backported (supposed until 2.6)
When a stream is caught looping, we produce some output to help figure
its internal state explaining why it's looping. The problem is that this
debug output is quite old and the info it provides are quite insufficient
to debug a modern process, and since such bugs happen only once or twice
a year the situation doesn't improve.
On the other hand the output of "show sess all" is extremely detailed
and kept up to date with code evolutions since it's a heavily used
debugging tool.
This commit replaces the call to the totally outdated stream_dump() with
a call to strm_dump_to_buffer(), and removes the filters dump since they
are already emitted there, and it now produces much more exploitable
output:
[ALERT] (5936) : A bogus STREAM [0x7fa8dc02f660] is spinning at 5653514 calls per second and refuses to die, aborting now! Please report this error to developers:
0x7fa8dc02f660: [28/Sep/2023:09:53:08.811818] id=2 proto=tcpv4 source=127.0.0.1:58306
flags=0xc4a, conn_retries=0, conn_exp=<NEVER> conn_et=0x000 srv_conn=0x133f220, pend_pos=(nil) waiting=0 epoch=0x1
frontend=public (id=2 mode=http), listener=? (id=1) addr=127.0.0.1:4080
backend=public (id=2 mode=http) addr=127.0.0.1:61932
server=s1 (id=1) addr=127.0.0.1:7443
task=0x7fa8dc02fa40 (state=0x01 nice=0 calls=5749559 rate=5653514 exp=3s tid=1(1/1) age=1s)
txn=0x7fa8dc02fbf0 flags=0x3000 meth=1 status=-1 req.st=MSG_DONE rsp.st=MSG_RPBEFORE req.f=0x4c rsp.f=0x00
scf=0x7fa8dc02f5f0 flags=0x00000482 state=EST endp=CONN,0x7fa8dc02b4b0,0x05004001 sub=1 rex=58s wex=<NEVER>
h1s=0x7fa8dc02b4b0 h1s.flg=0x100010 .sd.flg=0x5004001 .req.state=MSG_DONE .res.state=MSG_RPBEFORE
.meth=GET status=0 .sd.flg=0x05004001 .sc.flg=0x00000482 .sc.app=0x7fa8dc02f660
.subs=0x7fa8dc02f608(ev=1 tl=0x7fa8dc02fae0 tl.calls=0 tl.ctx=0x7fa8dc02f5f0 tl.fct=sc_conn_io_cb)
h1c=0x7fa8dc0272d0 h1c.flg=0x0 .sub=0 .ibuf=0@(nil)+0/0 .obuf=0@(nil)+0/0 .task=0x7fa8dc0273f0 .exp=<NEVER>
co0=0x7fa8dc027040 ctrl=tcpv4 xprt=RAW mux=H1 data=STRM target=LISTENER:0x12840c0
flags=0x00000300 fd=32 fd.state=20 updt=0 fd.tmask=0x2
scb=0x7fa8dc02fb30 flags=0x00001411 state=EST endp=CONN,0x7fa8dc0300c0,0x05000001 sub=1 rex=58s wex=<NEVER>
h1s=0x7fa8dc0300c0 h1s.flg=0x4010 .sd.flg=0x5000001 .req.state=MSG_DONE .res.state=MSG_RPBEFORE
.meth=GET status=0 .sd.flg=0x05000001 .sc.flg=0x00001411 .sc.app=0x7fa8dc02f660
.subs=0x7fa8dc02fb48(ev=1 tl=0x7fa8dc02feb0 tl.calls=2 tl.ctx=0x7fa8dc02fb30 tl.fct=sc_conn_io_cb)
h1c=0x7fa8dc02ff00 h1c.flg=0x80000000 .sub=1 .ibuf=0@(nil)+0/0 .obuf=0@(nil)+0/0 .task=0x7fa8dc030020 .exp=<NEVER>
co1=0x7fa8dc02fcd0 ctrl=tcpv4 xprt=RAW mux=H1 data=STRM target=SERVER:0x133f220
flags=0x10000300 fd=33 fd.state=10421 updt=0 fd.tmask=0x2
req=0x7fa8dc02f680 (f=0x1840000 an=0x8000 pipe=0 tofwd=0 total=79)
an_exp=<NEVER> buf=0x7fa8dc02f688 data=(nil) o=0 p=0 i=0 size=0
htx=0xc18f60 flags=0x0 size=0 data=0 used=0 wrap=NO extra=0
res=0x7fa8dc02f6d0 (f=0x80000000 an=0x1400000 pipe=0 tofwd=0 total=0)
an_exp=<NEVER> buf=0x7fa8dc02f6d8 data=(nil) o=0 p=0 i=0 size=0
htx=0xc18f60 flags=0x0 size=0 data=0 used=0 wrap=NO extra=0
call trace(10):
| 0x59f2b7 [0f 0b 0f 1f 80 00 00 00]: stream_dump_and_crash+0x1f7/0x2bf
| 0x5a0d71 [e9 af e6 ff ff ba 40 00]: process_stream+0x19f1/0x3a56
| 0x68d7bb [49 89 c7 4d 85 ff 74 77]: run_tasks_from_lists+0x3ab/0x924
| 0x68e0b4 [29 44 24 14 8b 4c 24 14]: process_runnable_tasks+0x374/0x6d6
| 0x656f67 [83 3d f2 75 84 00 01 0f]: run_poll_loop+0x127/0x5a8
| 0x6575d7 [48 8b 1d 42 50 5c 00 48]: main+0x1b22f7
| 0x7fa8e0f35e45 [64 48 89 04 25 30 06 00]: libpthread:+0x7e45
| 0x7fa8e0e5a4af [48 89 c7 b8 3c 00 00 00]: libc:clone+0x3f/0x5a
Note that the output is subject to the global anon key so that IPs and
object names can be anonymized if required. It could make sense to
backport this and the few related previous patches next time such an
issue is reported.
There used to be two working modes for this function, a single-line one
and a multi-line one, the difference being made on the "eol" argument
which could contain either a space or an LF (and with the prefix being
adjusted accordingly). Let's get rid of the single-line mode as it's
what limits the output contents because it's difficult to produce
exploitable structured data this way. It was only used in the rare case
of spinning streams and applets and these are the ones lacking info. Now
a spinning stream produces:
[ALERT] (3511) : A bogus STREAM [0x227e7b0] is spinning at 5581202 calls per second and refuses to die, aborting now! Please report this error to developers:
strm=0x227e7b0,c4a src=127.0.0.1 fe=public be=public dst=s1
txn=0x2041650,3000 txn.req=MSG_DONE,4c txn.rsp=MSG_RPBEFORE,0
rqf=1840000 rqa=8000 rpf=80000000 rpa=1400000
scf=0x24af280,EST,482 scb=0x24af430,EST,1411
af=(nil),0 sab=(nil),0
cof=0x7fdb28026630,300:H1(0x24a6f60)/RAW((nil))/tcpv4(33)
cob=0x23199f0,10000300:H1(0x24af630)/RAW((nil))/tcpv4(32)
filters={}
call trace(11):
(...)
Since 2.4-dev18 with commit b4476c6a8 ("CLEANUP: freq_ctr: make
arguments of freq_ctr_total() const"), most of the freq_ctr readers
should be fine with a const, except that they were not updated to
reflect this and they continue to force variable on some functions
that call them. Let's update this. This could even be backported if
needed.
Added set-timeout for frontend side of session, so it can be used to set
custom per-client timeouts if needed. Added cur_client_timeout to fetch
client timeout samples.
Add reporting using send_log() for preconnect operation. This is minimal
to ensure we understand the current status of listener in active reverse
connect.
To limit logging quantity, only important transition are considered.
This requires to implement a minimal state machine as a new field in
receiver structure.
Here are the logs produced :
* Initiating : first time preconnect is enabled on a listener
* Error : last preconnect attempt interrupted on a connection error
* Reaching maxconn : all necessary connections were reversed and are
operational on a listener
When a connection is freed during preconnect before reversal, the error
must be notified to the listener to remove any connection reference and
rearm a new preconnect attempt. Currently, this can occur through 2 code
paths :
* conn_free() called directly by H2 mux
* error during conn_create_mux(). For this case, connection is flagged
with CO_FL_ERROR and reverse_connect task is woken up. The process
task handler is then responsible to call conn_free() for such
connection.
Duplicated steps where done both in conn_free() and process task
handler. These are now removed. To facilitate code maintenance,
dedicated operation have been centralized in a new function
rev_notify_preconn_err() which is called by conn_free().
This patch adds the ability to externalize and customize the code
of the computation of extra CIDs after the first one was derived from
the ODCID.
This is to prepare interoperability with extra components such as
different QUIC proxies or routers for instance.
To process the patch defines two function callbacks:
- the first one to compute a hash 64bits from the first generated CID
(itself continues to be derived from ODCID). Resulting hash is stored
into the 'quic_conn' and 64bits is chosen large enought to be able to
store an entire haproxy's CID.
- the second callback re-uses the previoulsy computed hash to derive
an extra CID using the custom algorithm. If not set haproxy will
continue to choose a randomized CID value.
Those two functions have also the 'cluster_secret' passed as an argument:
this way, it is usable for obfuscation or ciphering.
Function comments were outdated, probably because they have not been
updated during the previous refactors.
Fixing comments to better reflect the current behavior.
This may be backported up to 2.2, or even 2.0 by slightly adapting the
patch (in 2.0, such functions are documented in proto/pattern.h)
The ring lock was initially mostly used for the logs and used to inherit
its name in lock stats. Now that it's exclusively used by rings, let's
rename it accordingly.
The log server lock is pretty visible in perf top when using log samples
because it's taken for each server in turn while trying to validate and
update the log server's index. Let's change this for a CAS, since we have
the index and the range at hand now. This allow us to remove the logsrv
lock.
The test on 4 servers now shows a 3.7 times improvement thanks to much
lower contention. Without log sampling a test producing 4.4M logs/s
delivers 4.4M logs/s at 21 CPUs used, everything spent in the kernel.
After enabling 4 samples (1:4, 2:4, 3:4 and 4:4), the throughput would
previously drop to 1.13M log/s with 37 CPUs used and 75% spent in
process_send_log(). Now with this change, 4.25M logs/s are emitted,
using 26 CPUs and 22% in process_send_log(). That's a 3.7x throughput
improvement for a 30% global CPU usage reduction, but in practice it
mostly shows that the performance drop caused by having samples is much
less noticeable (each of the 4 servers has its index updated for each
log).
Note that in order to even avoid incrementing an index for each log srv
that is consulted, it would be more convenient to have a single index
per frontend and apply the modulus on each log server in turn to see if
the range has to be updated. It would then only perform one write per
range switch. However the place where this is done doesn't have access
to a frontend, so some changes would need to be performed for this, and
it would require to update the current range independently in each
logsrv, which is not necessarily easier since we don't know yet if we
can commit it.
By using a single long long to store both the current range and the
next index, we'll make it possible to perform atomic operations instead
of locking. Let's only regroup them for now under a new "curr_rg_idx".
The upper word is the range, the lower is the index.
This index is useless because it only serves to know when the global
index reached the end, while the global one already knows it. Let's
just drop it and perform the test on the global range.
It was verified with the following config that the first server continues
to take 1/10 of the traffic, the 2nd one 2/10, the 3rd one 3/10 and the
4th one 4/10:
log 127.0.0.1:10001 sample 1:10 local0
log 127.0.0.1:10002 sample 2,5:10 local0
log 127.0.0.1:10003 sample 3,7,9:10 local0
log 127.0.0.1:10004 sample 4,6,8,10:10 local0
The test of the log range is not very clear, in part due to the
reuse of the "curr_idx" name that happens at two levels. The call
to in_smp_log_range() applies to the smp_info's index to which 1 is
added: it verifies that the next index is still within the current
range.
Let's just have a local variable "next_index" in process_send_log()
that gets assigned the next index (current+1) and compare it to the
current range's boundaries. This makes the test much clearer. We can
then simply remove in_smp_log_range() that's no longer needed.
This reverts commit c618ed5ff4.
The list iterator is broken. As found by Fred, running QUIC single-
threaded shows that only the first connection is accepted because the
accepter relies on the element being initialized once detached (which
is expected and matches what MT_LIST_DELETE_SAFE() used to do before).
However while doing this in the quic_sock code seems to work, doing it
inside the macro show total breakage and the unit test doesn't work
anymore (random crashes). Thus it looks like the fix is not trivial,
let's roll this back for the time it will take to fix the loop.
In 1.9 with commit 627505d36 ("MINOR: freq_ctr: add swrate_add_scaled()
to work with large samples") we got the ability to indicate when adding
some values that they represent a number of samples. However there is an
issue in the calculation which is that the number of samples that is
added to the sum before the division in order to avoid fading away too
fast, is multiplied by the scale. The problem it causes is that this is
done in the negative part of the expression, and that as soon if the sum
of old_sum and v*s is too small (e.g. zero), we end up with a negative
value of -s.
This is visible in "show pools" which occasionally report a very large
value on "needed_avg" since 2.9, though the bug has been there for longer.
Indeed in 2.9 since they're hashed in buckets, it suffices that any
thread got one such error once for the sum to be wrong. One possible
impact is memory usage not shrinking after a short burst due to pools
refraining from releasing objects, believing they don't have enough.
This must be backported to all versions. Note that the opportunistic
version can be dropped before 2.8.
The new mt_list code supports exponential back-off on conflict, which
is important for use cases where there is contention on a large number
of threads. The API evolved a little bit and required some updates:
- mt_list_for_each_entry_safe() is now in upper case to explicitly
show that it is a macro, and only uses the back element, doesn't
require a secondary pointer for deletes anymore.
- MT_LIST_DELETE_SAFE() doesn't exist anymore, instead one just has
to set the list iterator to NULL so that it is not re-inserted
into the list and the list is spliced there. One must be careful
because it was usually performed before freeing the element. Now
instead the element must be nulled before the continue/break.
- MT_LIST_LOCK_ELT() and MT_LIST_UNLOCK_ELT() have always been
unclear. They were replaced by mt_list_cut_around() and
mt_list_connect_elem() which more explicitly detach the element
and reconnect it into the list.
- MT_LIST_APPEND_LOCKED() was only in haproxy so it was left as-is
in list.h. It may however possibly benefit from being upstreamed.
This required tiny adaptations to event_hdl.c and quic_sock.c. The
test case was updated and the API doc added. Note that in order to
keep include files small, the struct mt_list definition remains in
list-t.h (par of the internal API) and was ifdef'd out in mt_list.h.
A test on QUIC with both quictls 1.1.1 and wolfssl 5.6.3 on ARM64 with
80 threads shows a drastic reduction of CPU usage thanks to this and
the refined memory barriers. Please note that the CPU usage on OpenSSL
3.0.9 is significantly higher due to the excessive use of atomic ops
by openssl, but 3.1 is only slightly above 1.1.1 though:
- before: 35 Gbps, 3.5 Mpps, 7800% CPU
- after: 41 Gbps, 4.2 Mpps, 2900% CPU
This bug arrived with this commit in 2.9-dev3:
MEDIUM: quic: Allow the quic_conn memory to be asap released.
When sending packets from quic_cc_conn_io_cb(), e.g. when the quic_conn
object has been released and replaced by a lighter one (quic_cc_conn),
some counters may have to be incremented. But they were not reachable
because not shared between quic_conn and quic_cc_conn struct.
To fix this, one has only to move the ->cntrs counters from quic_conn
to QUIC_CONN_COMMON struct which is shared between both quic_cc_conn
Thank you to Tristan for having reported this in GH #2247.
No need to backport.
It's a bit frustrating sometimes to see pool checks catch a bug but not
provide exploitable information without a core.
Here we're adding a function "pool_inspect_item()" which is called just
before aborting in pool_check_pattern() and POOL_DEBUG_CHECK_MARK() and
which will display the error type, the pool's pointer and name, and will
try to check if the item's tag matches the pool, and if not, will iterate
over all pools to see if one would be a better candidate, then will try
to figure the last known caller and possibly other likely candidates if
the pool's tag is not sufficiently trusted. This typically helps better
diagnose corruption in use-after-free scenarios, or freeing to a pool
that differs from the one the object was allocated from, and will also
indicate calling points that may help figure where an object was last
released or allocated. The info is printed on stderr just before the
backtrace.
For example, the recent off-by-one test in the PPv2 changes would have
produced the following output in vtest logs:
*** h1 debug|FATAL: pool inconsistency detected in thread 1: tag mismatch on free().
*** h1 debug| caller: 0x62bb87 (conn_free+0x147/0x3c5)
*** h1 debug| pool: 0x2211ec0 ('pp_tlv_256', size 304, real 320, users 1)
*** h1 debug|Tag does not match. Possible origin pool(s):
*** h1 debug| tag: @0x2565530 = 0x2216740 (pp_tlv_128, size 176, real 192, users 1)
*** h1 debug|Recorded caller if pool 'pp_tlv_128':
*** h1 debug| @0x2565538 (+0184) = 0x62c76d (conn_recv_proxy+0x4cd/0xa24)
A mismatch in the allocated/released pool is already visible, and the
callers confirm it once resolved, where the allocator indeed allocates
from pp_tlv_128 and conn_free() releases to pp_tlv_256:
$ addr2line -spafe ./haproxy <<< $'0x62bb87\n0x62c76d'
0x000000000062bb87: conn_free at connection.c:568
0x000000000062c76d: conn_recv_proxy at connection.c:1177
In preparation for more detailed pool error reports, let's pass the
caller pointers to the check functions. This will be useful to produce
messages indicating where the issue happened.
When recording the caller of a pool_alloc(), we currently store it only
when the object comes from the cache and never when it comes from the
heap. There's no valid reason for this except that the caller's pointer
was not passed to pool_alloc_nocache(), so it used to set NULL there.
Let's just pass it down the chain.
These ones were still in cfgparse.c but they're not specific to the
config at all and may actually be used even when parsing cpu list
entries in /sys. Better move them where they can be reused.
cpu_map is 8.2kB/entry and there's one such entry per group, that's
~520kB total. In addition, the init code is still in haproxy.c enclosed
in ifdefs. Let's make this a dynamically allocated array in the cpuset
code and remove that init code.
Later we may even consider reallocating it once the number of threads
and groups is known, in order to shrink it a little bit, as the typical
setup with a single group will only need 8.2kB, thus saving half a MB
of RAM. This would require that the upper bound is placed in a variable
though.
This function takes on input a printf format for the file name, making
it particularly suitable for /proc or /sys entries which take a lot of
numbers. It also automatically trims the trailing CR and/or LF chars.
The function generate_random_cluster_secret() which initializes the cluster secret
when not supplied by configuration is buggy. There 1/256 that the cluster secret
string is empty.
To fix this, one stores the cluster as a reduced size first 128 bits of its own
SHA1 (160 bits) digest, if defined by configuration. If this is not the case, it
is initialized with a 128 bits random value. Furthermore, thus the cluster secret
is always initialized.
As the cluster secret is always initialized, there are several tests which
are for now on useless. This patch removes such tests (if(global.cluster_secret))
in the QUIC code part and at parsing time: no need to check that a cluster
secret was initialized with "quic-force-retry" option.
Must be backported as far as 2.6.
This patch implements the 'curves' keyword on server lines as well as
the 'ssl-default-server-curves' keyword in the global section.
It also add the keyword on the server line in the ssl_curves reg-test.
These keywords allow the configuration of the curves list for a server.
We currently know the number of tasks in the run queue that are niced,
and we don't expose it. It's too bad because it can give a hint about
what share of the load is relevant. For example if one runs a Lua
script that was purposely reniced, or if a stats page or the CLI is
hammered with slow operations, seeing them appear there can help
identify what part of the load is not caused by the traffic, and
improve monitoring systems or autoscalers.
When building the secondary signature for cache entries when vary is
enabled, the referer part of the signature was a simple crc32 of the
first referer header.
This patch changes it to a 64bits hash based of xxhash algorithm with a
random seed built during init. This will prevent "malicious" hash
collisions between entries of the cache.
We previously had postparsing logic but only for logsrv sinks, but now we
need to make this operation on logsrv directly instead of sinks to prepare
for additional postparsing logic that is not sink-specific.
To do this, we migrated post_sink_resolve() and sink_postresolve_logsrvs()
to their postresolve_logsrvs() and postresolve_logsrv_list() equivalents.
Then, we split postresolve_logsrv_list() so that the sink-only logic stays
in sink.c (sink_resolve_logsrv_buffer() function), and the "generic"
target part stays in log.c as resolve_logsrv().
Error messages formatting was preserved as far as possible but some slight
variations are to be expected.
As for the functional aspect, no change should be expected.
All applets only check the -1 error value (need room) for applet_put*
functions while the underlying functions may also return -2 if the input is
closed or -3 if the data length is invalid. It means applets already handle
other cases by their own.
The API should be fixed but for now, to ease backports, we only fix
applet_put* functions to always return -1 on error. This way, at least for
the applets point of view, the API is consistent.
This patch should be backported to 2.8. Probably not further. Except if we
suspect it could fix a bug.
Due to the fact that several variable values (rtt_var, srtt) were stored as multiple
of their real values, some calculations were less accurate as expected.
Stop storing 4*rtt_var values, and 8*srtt values.
Adjust all the impacted statements.
Must be backported as far as 2.6.
This detects when there are more threads bound via cpu-map than CPUs
enabled in cpu-map, or when there are more total threads than the total
number of CPUs available at boot (for unbound threads) and configured
for bound threads. In this case, a warning is emitted to explain the
problems it will cause, and explaining how to address the situation.
Note that some configurations will not be detected as faulty because
the algorithmic complexity to resolve all arrangements grows in O(N!).
This means that having 3 threads on 2 CPUs and one thread on 2 CPUs
will not be detected as it's 4 threads for 4 CPUs. But at least configs
such as T0:(1,4) T1:(1,4) T2:(2,4) T3:(3,4) will not trigger a warning
since they're valid.
It's very easy to mess up with some cpu-map directives and to leave
some thread unbound. Let's add a test that checks that either all
threads are bound or none are bound, but that we do not face the
intermediary situation where some are pinned and others are left
wandering around, possibly on the same CPUs as bound ones.
Note that this should not be backported, or maybe turned into a
notice only, as it appears that it will easily catch invalid
configs and that may break updates for some users.
Till now the CPUs that were bound were only retrieved in
thread_cpus_enabled() in order to count the number of CPUs allowed,
and it relied on arch-specific code.
Let's slightly arrange this into ha_cpuset_detect_bound() that
reuses the ha_cpuset struct and the accompanying code. This makes
the code much clearer without having to carry along some arch-specific
stuff out of this area.
Note that the macos-specific code used in thread.c to only count
online CPUs but not retrieve a mask, so for now we can't infer
anything from it and can't implement it.
In addition and more importantly, this function is reliable in that
it will only return a value when the detection is accurate, and will
not return incomplete sets on operating systems where we don't have
an exact list, such as online CPUs.
When building without threads, the recently introduced BUG_ON(tid != 0)
turns to a constant expression that evaluates to 0 and that is not used,
resulting in this warning:
src/connection.c: In function 'conn_free':
src/connection.c:584:3: warning: statement with no effect [-Wunused-value]
This is because the whole thing is declared as an expression for clarity.
Make it return void to avoid this. No backport is needed.
This adds a new option for the Makefile USE_OPENSSL_AWSLC, and
update the documentation with instructions to use HAProxy with
AWS-LC.
Update the type of the OCSP callback retrieved with
SSL_CTX_get_tlsext_status_cb with the actual type for
libcrypto versions greater than 1.0.2. This doesn't affect
OpenSSL which casts the callback to void* in SSL_CTX_ctrl.
For the same reason than the previous patch, we must not block the sends
when there is a pending shutdown. In other words, we must consider the sends
are allowed when there is a pending shutdown.
This patch must slowly be backported as far as 2.2. It should partially fix
issue #2249.
The progressive adoption of OpenSSL 3 and its abysmal handshake
performance has started to reveal situations where it simply isn't
possible anymore to succesfully run health checks on many servers,
because between the moment all the checks are started and the moment
the handshake finally completes, the timeout has expired!
This also has consequences on production traffic which gets
significantly delayed as well, all that for lots of checks. While it's
possible to increase the check delays, it doesn't solve everything as
checks still take a huge amount of time to converge in such conditions.
Here we take a different approach by permitting to enforce the maximum
concurrent checks per thread limitation and implementing an ordered
queue. Thanks to this, if a thread about to start a check has reached
its limit, it will add the check at the end of a queue and it will be
processed once another check is finished. This proves to be extremely
efficient, with all checks completing in a reasonable amount of time
and not being disturbed by the rest of the traffic from other checks.
They're just cycling slower, but at the speed the machine can handle.
One must understand however that if some complex checks perform multiple
exchanges, they will take a check slot for all the required duration.
This is why the limit is not enforced by default.
Tests on SSL show that a limit of 5-50 checks per thread on local
servers gives excellent results already, so that could be a good starting
point.
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.
The internal tick clock was used to export the timestamp int the token
on retry packets. Doing this in cluster mode the nodes don't
understand the timestamp from tokens generated by others.
This patch re-work this using the the real current date (wall-clock time).
Timestamp are also now considered in secondes instead of milleseconds.
This patch should be backported until v2.6
sink_write() currently relies on sink->maxlen to know when to stop
writing a given payload.
But it could be useful to pass a smaller, explicit value to sink_write()
to stop before the ring maxlen, for instance if the ring is shared between
multiple feeders.
sink_write() now takes an optional maxlen parameter:
if maxlen is > 0, then sink_write will stop writing at maxlen if maxlen
is smaller than ring->maxlen, else only ring->maxlen will be considered.
[for haproxy <= 2.7, patch must be applied by hand: that is:
__sink_write() and sink_write() should be patched to take maxlen into
account and function calls to sink_write() should use 0 as second argument
to keep original behavior]
Thierry Fournier reported an annoying side-effect when using the debug()
converter.
Consider the following examples:
[1] http-request set-var(txn.test) bool(true),ipmask(24)
[2] http-request redirect location /match if { bool(true),ipmask(32) }
When starting haproxy with [1] example we get:
config : parsing [test.conf:XX] : error detected in frontend 'fe' while parsing 'http-request set-var(txn.test)' rule : converter 'ipmask' cannot be applied.
With [2], we get:
config : parsing [test.conf:XX] : error detected in frontend 'fe' while parsing 'http-request redirect' rule : error in condition: converter 'ipmask' cannot be applied in ACL expression 'bool(true),ipmask(32)'.
Now consider the following examples which are based on [1] and [2]
but with the debug() sample conv inserted in-between those incompatible
sample types:
[1*] http-request set-var(txn.test) bool(true),debug,ipmask(24)
[2*] http-request redirect location /match if { bool(true),debug,ipmask(32) }
According to the documentation, "it is safe to insert the debug converter
anywhere in a chain, even with non-printable sample types".
Thus we don't expect any side-effect from using it within a chain.
However in current implementation, because of debug() returning SMP_T_ANY
type which is a pseudo type (only resolved at runtime), the sample
compatibility checks performed at parsing time are completely uneffective.
(haproxy will start and no warning will be emitted)
The indesirable effect of this is that debug() prevents haproxy from
warning you about impossible type conversions, hiding potential errors
in the configuration that could result to unexpected evaluation or logic
while serving live traffic. We better let haproxy warn you about this kind
of errors when it has the chance.
With our previous examples, this could cause some inconveniences. Let's
say for example that you are testing a configuration prior to deploying
it. When testing the config you might want to use debug() converter from
time to time to check how the conversion chain behaves.
Now after deploying the exact same conf, you might want to remove those
testing debug() statements since they are no longer relevant.. but
removing them could "break" the config and suddenly prevent haproxy from
starting upon reload/restart. (this is the expected behavior, but it
comes a bit too late because of debug() hiding errors during tests..)
To fix this, we introduce a new output type for sample expressions:
SMP_T_SAME - may only be used as "expected" out_type (parsing time)
for sample converters.
As it name implies, it is a way for the developpers to indicate that the
resulting converter's output type is guaranteed to match the type of the
sample that is presented on the converter's input side.
(converter may alter data content, but data type must not be changed)
What it does is that it tells haproxy that if switching to the converter
(by looking at the converter's input only, since outype is SAME) is
conversion-free, then the converter type can safely be ignored for type
compatibility checks within the chain.
debug()'s out_type is thus set to SMP_T_SAME instead of ANY, which allows
it to fully comply with the doc in the sense that it does not impact the
conversion chain when inserted between sample items.
Co-authored-by: Thierry Fournier <thierry.f.78@gmail.com>
SMP_T_ADDR support was added in b805f71 ("MEDIUM: sample: let the cast
functions set their output type").
According to the above commit, it is made clear that the ADDR type is
a pseudo/generic type that may be used for compatibility checks but
that cannot be emitted from a fetch or converter.
With that in mind, all conversions from ADDR to other types were
explicitly disabled in the compatibility matrix.
But later, when map_*_ip functions were updated in b2f8f08 ("MINOR: map:
The map can return IPv4 and IPv6"), we started using ADDR as "expected"
output type for converters. This still complies with the original
description from b805f71, because it is used as the theoric output
type, and is never emitted from the converters themselves (only "real"
types such as IPV4 or IPV6 are actually being emitted at runtime).
But this introduced an ambiguity as well as a few bugs, because some
compatibility checks are being performed at config parse time, and thus
rely on the expected output type to check if the conversion from current
element to the next element in the chain is theorically supported.
However, because the compatibility matrix doesn't support ADDR to other
types it is currently impossible to use map_*_ip converters in the middle
of a chain (the only supported usage is when map_*_ip converters are
at the very end of the chain).
To illustrate this, consider the following examples:
acl test str(ok),map_str_ip(test.map) -m found # this will work
acl test str(ok),map_str_ip(test.map),ipmask(24) -m found # this will raise an error
Likewise, stktable_compatible_sample() check for stick tables also relies
on out_type[table_type] compatibility check, so map_*_ip cannot be used
with sticktables at the moment:
backend st_test
stick-table type string size 1m expire 10m store http_req_rate(10m)
frontend fe
bind localhost:8080
mode http
http-request track-sc0 str(test),map_str_ip(test.map) table st_test # raises an error
To fix this, and prevent future usage of ADDR as expected output type
(for either fetches or converters) from introducing new bugs, the
ADDR=>? part of the matrix should follow the ANY type logic.
That is, ADDR, which is a pseudo-type, must be compatible with itself,
and where IPV4 and IPV6 both support a backward conversion to a given
type, ADDR must support it as well. It is done by setting the relevant
ADDR entries to c_pseudo() in the compatibility matrix to indicate that
the operation is theorically supported (c_pseudo() will never be executed
because ADDR should not be emitted: this only serves as a hint for
compatibility checks at parse time).
This is what's being done in this commit, thanks to this the broken
examples documented above should work as expected now, and future
usage of ADDR as out_type should not cause any issue.
This function is used for ANY=>!ANY conversions in the compatibility
matrix to help differentiate between real NOOP (c_none) and pseudo
conversions that are theorically supported at config parse time but can
never occur at runtime,. That is, to explicit the fact that actual related
runtime operations (e.g.: ANY->IPV4) are not NOOP since they might require
some conversion to be performed depending on the input type.
When checking the conf we don't know the effective out types so
cast[pseudo type][pseudo type] is allowed in the compatibility matrix,
but at runtime we only expect cast[real type][(real type || pseudo type)]
because fetches and converters may not emit pseudo types, thus using
c_none() everywhere was too ambiguous.
The process will crash if c_pseudo() is invoked to help catch bugs:
crashing here means that a pseudo type has been encountered on a
converter's input at runtime (because it was emitted earlier in the
chain), which is not supported and results from a broken sample fetch
or converter implementation. (pseudo types may only be used as out_type
in sample definitions for compatibility checks at parsing time)
Both sample_parse_expr() and parse_acl_expr() implement some code
logic to parse sample conv list after respective fetch or acl keyword.
(Seems like the acl one was inspired by the sample one historically)
But there is clearly code duplication between the two functions, making
them hard to maintain.
Hopefully, the parsing logic between them has stayed pretty much the
same, thus the sample conv parsing part may be moved in a dedicated
helper parsing function.
This is what's being done in this commit, we're adding the new function
sample_parse_expr_cnv() which does a single thing: parse the converters
that are listed right after a sample fetch keyword and inject them into
an already existing sample expression.
Both sample_parse_expr() and parse_acl_expr() were adapted to now make
use of this specific parsing function and duplicated code parts were
cleaned up.
Although sample_parse_expr() remains quite complicated (numerous function
arguments due to contextual parsing data) the first goal was to get rid of
code duplication without impacting the current behavior, with the added
benefit that it may allow further code cleanups / simplification in the
future.
This patch allows the low level packet parser to drop packets with type for discarded
packet number spaces. Furthermore, this prevents it from reallocating new encryption
levels and packet number spaces already released/discarded. When a packet number space
is discarded, it MUST NOT be reallocated.
As the packet number space discarding is done asap the type of packet received is
known, some packet number space discarding check may be safely removed from qc_try_rm_hp()
and qc_qel_may_rm_hp() which are called after having parse the packet header, and
is type.
As the packet number spaces and encryption level are dynamically allocated,
the information about the packet number space discarded status must be kept
somewhere else than in these objects.
quic_tls_discard_keys() is no more useful.
Modify quic_pktns_discard() to do the same job: flag the quic_conn object
has having discarded packet number space.
Implement quic_tls_pktns_is_disarded() to check if a packet number space is
discarded. Note the Application data packet number space is never discarded.
The maximum length of the secrets derived by the TLS stack is 384 bits.
This reduces the size of the objects provided by the "quic_tls_secret" pool by
16 bytes.
Should be backported as far as 2.6
Replace ->els static array of encryption levels by 4 pointers into the QUIC
connection object, quic_conn struct.
->iel denotes the Initial encryption level,
->eel the Early-Data encryption level,
->hel the Handshaske encryption level and
->ael the Application Data encryption level.
Add ->qel_list to this structure to list the encryption levels after having been
allocated. Modify consequently the encryption level object itself (quic_enc_level
struct) so that it might be added to ->qel_list QUIC connection list of
encryption levels.
Implement qc_enc_level_alloc() to initialize the value of a pointer to an encryption
level object. It is used to initialized the pointer newly added to the quic_conn
structure. It also takes a packet number space pointer address as argument to
initialize it if not already initialized.
Modify quic_tls_ctx_reset() to call it from quic_conn_enc_level_init() which is
called by qc_enc_level_alloc() to allocate an encryption level object.
Implement 2 new helper functions:
- ssl_to_qel_addr() to match and pointer address to a quic_encryption level
attached to a quic_conn object with a TLS encryption level enum value;
- qc_quic_enc_level() to match a pointer to a quic_encryption level attached
to a quic_conn object with an internal encryption level enum value.
This functions are useful to be called from ->set_encryption_secrets() and
->add_handshake_data() TLS stack called which takes a TLS encryption enum
as argument (enum ssl_encryption_level_t).
Replace all the use of the qc->els[] array element values by one of the newly
added ->[ieha]el quic_conn struct member values.
Very simple patch to define and declare a pool for the QUIC TLS encryptions levels.
It will be used to dynamically allocate such objects to be attached to the
QUIC connection object (quic_conn struct) and to remove from quic_conn struct the
static array of encryption levels (see ->els[]).
Add a pool to dynamically handle the memory used for the QUIC TLS packet number spaces.
Remove the static array of packet number spaces at QUIC connection level (struct
quic_conn) and add three new members to quic_conn struc as pointers to quic_pktns
struct, one by packet number space as follows:
->ipktns for Initial packet number space,
->hpktns for Handshake packet number space and
->apktns for Application packet number space.
Also add a ->pktns_list new member (struct list) to quic_conn struct to attach
the list of the packet number spaces allocated for the QUIC connection.
Implement ssl_to_quic_pktns() to map and retrieve the addresses of these pointers
from TLS stack encryption levels.
Modify quic_pktns_init() to initialize these members.
Modify ha_quic_set_encryption_secrets() and ha_quic_add_handshake_data() to
allocate the packet numbers and initialize the encryption level.
Implement quic_pktns_release() which takes pointers to pointers to packet number
space objects to release the memory allocated for a packet number space attached
to a QUIC connection and reset their address values.
Modify qc_new_conn() to allocation only the Initial packet number space and
Initial encryption level.
Modify QUIC loss detection API (quic_loss.c) to use the new ->pktns_list
list attached to a QUIC connection in place of a static array of packet number
spaces.
Replace at several locations the use of elements of an array of packet number
spaces by one of the three pointers to packet number spaces
haproxy/quic_tls-t.h is the correct place to quic_enc_level structure
definition.
Should be backported as far as 2.6 to ease any further backport to come.
In some cases it may be desirable for latency reasons to forcefully
flush the queue even if it results in suboptimal compression. In our
case the queue might contain up to almost 4 bytes, which need an EOB
and a switch to literal mode, followed by 4 bytes to encode an empty
message. This means that each call can add 5 extra bytes in the ouput
stream. And the flush may also result in the header being produced for
the first time, which can amount to 2 or 10 bytes (zlib or gzip). In
the worst case, a total of 19 bytes may be emitted at once upon a flush
with 31 pending bits and a gzip header.
This is libslz upstream commit cf8c4668e4b4216e930b56338847d8d46a6bfda9.
This patch introduces the "client-sigalgs" keyword for the server line,
which allows to configure the list of server signature algorithms
negociated during the handshake. Also available as
"ssl-default-server-client-sigalgs" in the global section.
This patch introduces the "sigalgs" keyword for the server line, which
allows to configure the list of server signature algorithms negociated
during the handshake. Also available as "ssl-default-server-sigalgs" in
the global section.
This bug arrived with this commit:
MINOR: quic: Remove pool_zalloc() from qc_new_conn()
Missing initialization of largest packet number received during a keyupdate phase.
This prevented the keyupdate feature from working and made the keyupdate interop
tests to fail for all the clients.
Furthermore, ->flags from quic_tls_ctx was also not initialized. This could
also impact the keyupdate feature at least.
No backport needed.
qc_new_conn() is ued to initialize QUIC connections with quic_conn struct objects.
This function calls quic_conn_release() when it fails to initialize a connection.
quic_conn_release() is also called to release the memory allocated by a QUIC
connection.
Replace pool_zalloc() by pool_alloc() in this function and initialize
all quic_conn struct members which are referenced by quic_conn_release() to
prevent use of non initialized variables in this fonction.
The ebtrees, the lists attached to quic_conn struct must be initialized.
The tasks must be reset to their NULL default values to be safely destroyed
by task_destroy(). This is all the case for all the TLS cipher contexts
of the encryption levels (struct quic_enc_level) and those for the keyupdate.
The packet number spaces (struct quic_pktns) must also be initialized.
->prx_counters pointer must be initialized to prevent quic_conn_prx_cntrs_update()
from dereferencing this pointer.
->latest_rtt member of quic_loss struct must also be initialized. This is done
by quic_loss_init() called by quic_path_init().
->tx.pto_probe member of quic_pktns struct was not initialized by quic_pktns_init().
This bug never occured because all quic_pktns structs are attached to quic_conn
structs which are always pool_zalloc()'ed.
Must be backported as far as 2.6.
proxy default-server is a specific type of server that is not allocated
using new_server(): it is directly stored within the parent proxy
structure. However, since it may contain some default config options that
may be inherited by regular servers, it is also subject to dynamic members
(strings, structures..) that needs to be deallocated when the parent proxy
is cleaned up.
Unfortunately, srv_drop() may not be used directly from p->defsrv since
this function is meant to be used on regular servers only (those created
using new_server()).
To circumvent this, we're splitting srv_drop() to make a new function
called srv_free_params() that takes care of the member cleaning which
originally takes place in srv_drop(). This function is exposed through
server.h, so it may be called from outside server.c.
Thanks to this, calling srv_free_params(&p->defsrv) from free_proxy()
prevents any memory leaks due to dynamic parameters allocated when
parsing a default-server line from a proxy section.
This partially fixes GH #2173 and may be backported to 2.8.
[While it could also be relevant for other stable versions, the patch
won't apply due to architectural changes / name changes between 2.4 => 2.6
and then 2.6 => 2.8. Considering this is a minor fix that only makes
memory analyzers happy during deinit paths (at least for <= 2.8), it might
not be worth the trouble to backport them any further?]
Historically the client-fin and server-fin timeouts were made to allow
a connection closure to be effective quickly if the last data were sent
down a socket and the client didn't close, something that can happen
when the peer's FIN is lost and retransmits are blocked by a firewall
for example. This made complete sense in 1.5 for TCP and HTTP in close
mode. But nowadays with muxes, it's not done at the right layer anymore
and even the description doesn't match what is being done, because what
happens is that the stream will abort the whole transfer after it's done
sending to the mux and this timeout expires.
We've seen in GH issue 2095 that this can happen with very short timeout
values, and while this didn't trigger often before, now that the muxes
(h2 & quic) properly report an end of stream before even the first
sc_conn_sync_recv(), it seems that it can happen more often, and have
two undesirable effects:
- logging a timeout when that's not the case
- aborting the request channel, hence the server-side conn, possibly
before it had a chance to be put back to the idle list, causing
this connection to be closed and not reusable.
Unfortunately for TCP (mux_pt) this remains necessary because the mux
doesn't have a timeout task. So here we're adding tests to only do
this through an HTX mux. But to be really clean we should in fact
completely drop all of this and implement these timeouts in the mux
itself.
This needs to be backported to 2.8 where the issue was discovered,
and maybe carefully to older versions, though that is not sure at
all. In any case, using a higher timeout or removing client-fin in
HTTP proxies is sufficient to make the issue disappear.
As seen in commits 33a4461fa ("BUG/MINOR: stats: Fix Lua's `get_stats`
function") and a46b142e8 ("BUG/MINOR: Missing stat_field_names (since
f21d17bb)") it seems frequent to omit to update stats_fields[] when
adding a new ST_F_xxx entry. This breaks Lua's get_stats() and shows
a "(null)" in the header of "show stat", but that one is not detectable
to the naked eye anymore.
Let's add a reminder above the enum declaration about this, and a small
reg tests checking for the absence of "(null)". It was verified to fail
before the last patch above.
This patch is similar to the previous one but for QUIC mux functions
used inside the mux code itself or application layer. Replace all
occurences of qc_* prefix by qcc_* or qcs_*. This should help to better
differentiate code between quic_conn and MUX.
This should be backported up to 2.7.
There was a misnaming in stats counter for *_BLOCKED frames in regard to
QUIC rfc convention. This patch fixes it to prevent future ambiguity :
- STREAMS_BLOCKED -> STREAM_DATA_BLOCKED
- STREAMS_DATA_BLOCKED_BIDI -> STREAMS_BLOCKED_BIDI
- STREAMS_DATA_BLOCKED_UNI -> STREAMS_BLOCKED_UNI
This should be backported up to 2.7.
Remove nb_streams field from qcc. It was not used outside of a BUG_ON()
statement to ensure we never have a negative count of streams. However
this is already checked with other fields.
This should be backported up to 2.7.
The Linux kernel maintains data structures to track a processes' open file
descriptors, and it expands these structures as necessary when FD usage grows
(at every FD=2^X starting at 64). However when threading is in use, during
expansion the kernel will pause (observed up to 47ms) while it waits for thread
synchronization (see https://bugzilla.kernel.org/show_bug.cgi?id=217366).
This change addresses the issue and avoids the random pauses by opening the
maximum file descriptor during initialization, so that expansion will not occur
while processing traffic.
TCC silently ignores the weak and section attributes, which ruins the
initcalls. Technically we're exactly in the same situation as with an
obsolete linker. Let's just automatically set the flag if TCC is
detected, this avoids surprises where the program compiles but does
not start.
No backport is needed.
TCC doesn't knoow about __attribute__((weak)), it silently ignores it.
We could add a "static" modifier there in this case but we already have
an alternate portable mode that is based on a slightly larger literal
for obsolete linkers (and non-ELF systems) which choke on weak. Let's
just add the test for tcc there and use it in this case.
No backport is needed.
TCC is upset by the declaration looking like:
const unsigned char ist_lc[256] __attribute__((weak)) = ((const unsigned char[256]){ ... });
It was written like this because it's expanded from the _IST_LC macro
but it's never used as-is, it's only used from ist_lc, which should be
the one containing the cast so that the macro only contains the list of
bytes that can be used in both places. And this assigns more consistent
roles to the lower and upper case macro/variable now, one is typed and
the other one not. No backport is needed.
Add ->sent_pkt counter to quic_conn struct to count the packet at QUIC connection
level. Then, when the connection is released, the ->sent_pkt counter value
is added to the one for the listener.
Must be backported to 2.7.
Add some statistical counters to quic_conn struct from quic_counters struct which
are used at listener level to handle them at QUIC connection level. This avoid
calling atomic functions. Furthermore this will be useful soon when a counter will
be added for the total number of packets which have been sent which will be very
often incremented.
Some counters were not added, espcially those which count the number of QUIC errors
by QUIC error types. Indeed such counters would be incremented most of the time
only one time at QUIC connection level.
Implement quic_conn_prx_cntrs_update() which accumulates the QUIC connection level
statistical counters to the listener level statistical counters.
Must be backported to 2.7.
Thierry Fournier reported a build breakage with the ubiquitous make
3.81, LDFLAGS were ignored. This is caused by the declaration of the
collect_opt_flags macro that is defined with an "=" sign, something
that only appeared in 3.82 and that is not necessary. With it removed,
the build now works fine at least from 3.80 to 4.3.
No backport is needed since this makefile cleanup appeared in 2.8.
During a code audit of the various situations that promote ERR_PENDING to
ERROR, it appeared that:
- all muxes use se_fl_set_error() to set it, which chooses either based
on EOI/EOS presence ;
- EOI/EOS that arrive late after ERR_PENDING were not systematically
upgraded to ERROR
This results in confusion about how such ERROR or ERR_PENDING ought to
be handled, which is not quite desirable.
This patch adds a test to se_fl_set() to detect if we're setting EOI or
EOS while ERR_PENDING is present, or the other way around so that any
sequence of EOI/EOS <-> ERR_PENDING results in ERROR being set. This
way there will no longer be possible situations where ERROR is missing
while the other ones are set.
quic_aead_iv_build() should never fail unless we call it with buffers of
different size. This never happens in the code as every input buffers
are of size QUIC_TLS_IV_LEN.
Remove the return value and add a BUG_ON() to prevent future misusage.
This is especially useful to remove one error handling on the sending
patch via quic_packet_encrypt().
This should be backported up to 2.7.
Right now there's no way to enforce a specific value of now_ms upon
startup in order to compensate for the time it takes to load a config,
specifically when dealing with the health check startup. For this we'd
need to force the now_offset value to compensate for the last known
value of the current date. This patch exposes a function to do exactly
this.
Just like we have the uptime in "show info", let's add the boot time.
It's trivial to collect as it's just the difference between the ready
date and the start date, and will allow users to monitor this element
in order to take action before it starts becoming problematic. Here
the boot time is reported in milliseconds, so this allows to even
observe sub-second anomalies in startup delays.
Some huge configs take a significant amount of time to start and this
can cause some trouble (e.g. health checks getting delayed and grouped,
process not responding to the CLI etc). For example, some configs might
start fast in certain environments and slowly in other ones just due to
the use of a wrong DNS server that delays all libc's resolutions. Let's
first start by measuring it by keeping a copy of the most recently known
ready date, once before calling check_config_validity() and then refine
it when leaving this function. A last call is finally performed just
before deciding to split between master and worker processes, and it covers
the whole boot. It's trivial to collect and even allows to get rid of a
call to clock_update_date() in function check_config_validity() that was
used in hope to better schedule future events.
Uninline and move qc_attach_sc() function to implementation source file.
This will be useful for next commit to add traces in it.
This should be backported up to 2.7.
MUX is responsible to put EOS on stream when read channel is closed.
This happens if underlying connection is closed or a RESET_STREAM is
received. FIN STREAM is ignored in this case.
For connection closure, simply check for CO_FL_SOCK_RD_SH.
For RESET_STREAM reception, a new flag QC_CF_RECV_RESET has been
introduced. It is set when RESET_STREAM is received, unless we already
received all data. This is conform to QUIC RFC which allows to ignore a
RESET_STREAM in this case. During RESET_STREAM processing, input buffer
is emptied so EOS can be reported right away on recv_buf operation.
This should be backported up to 2.7.
Fix the openssl build with older openssl version by disabling the new
ssl_c_r_dn fetch.
This also disable the ssl_client_samples.vtc file for OpenSSL version
older than 1.1.1
This patch addresses #1514, adds the ability to fetch DN of the root
ca that was in the chain when client certificate was verified during SSL
handshake.
