Emeric suggested that it's sometimes convenient to instantly know if a
client has advertised support for window scaling or timestamps for
example. While the info is present in the TCP options output, it's hard
to extract since it respects the options order.
So here we're extending the 56-bit fingerprint with 8 extra bits that
indicate the presence of options 2..8, and any option above 9 for the
last bit. In practice this is sufficient since higher options are not
commonly used. Also TCP option 5 is normally not sent on the SYN (SACK,
only SACK_perm is sent), and echo options 6 & 7 are no longer used
(replaced with timestamps). These fields might be repurposed in the
future if some more meaningful options are to be mapped (e.g. MPTCP,
TFO cookie, auth).
When a backend is created at runtime, the new proxy instance is inserted
at the end of proxies_list. This operation is buggy if this list is
empty : the code causes a null dereference which will lead to a crash.
This causes the following compilation error :
CC src/proxy.o
src/proxy.c: In function 'cli_parse_add_backend':
src/proxy.c:4933:36: warning: null pointer dereference [-Wnull-dereference]
4933 | proxies_list->next = px;
| ~~~~~~~~~~~~~~~~~~~^~~~
This patch fixes this issue. Note that in reality it cannot occur at
this moment as proxies_list cannot be empty (haproxy requires at least
one frontend to start, and the list also always contains internal
proxies).
No need to backport.
This patch fixes the following compilation error :
src/proxy.c:4954:12: error: format string is not a string literal
(potentially insecure) [-Werror,-Wformat-security]
4954 | ha_notice(msg);
| ^~~
No need to backport.
Implement proxy ID generation for dynamic backends. This is performed
through the already function existing proxy_get_next_id().
As an optimization, lookup will performed starting from a global
variable <dynpx_next_id>. It is initialized to the greatest ID assigned
after parsing, and updated each time a backend instance is created. When
backend deletion will be implemented, it could be lowered to the newly
available slot.
Implement the required operations for "add backend" handler. This
requires a new proxy allocation, settings copy from the specified
default instance and proxy config finalization. All handlers registered
via REGISTER_POST_PROXY_CHECK() are also called on the newly created
instance.
If no error were encountered, the newly created proxy is finally
attached in the proxies list.
This commits completes "add backend" handler with some checks performed
on the specified default proxy instance. These are additional checks
outside of the already existing inheritance rules, specific to dynamic
backends.
For now, a default proxy is considered not compatible if it is not in
mode TCP/HTTP. Also, a default proxy is rejected if it references HTTP
errors. This limitation may be lifted in the future, when HTTP errors
are partiallay reworked.
Add an optional "mode" argument to "add backend" CLI command. This
argument allows to specify if the backend is in TCP or HTTP mode.
By default, it is mandatory, unless the inherited default proxy already
explicitely specifies the mode. To differentiate if TCP mode is implicit
or explicit, a new proxy flag PR_FL_DEF_EXPLICIT_MODE is defined. It is
set for every defaults instances which explicitely defined their mode.
Define a basic CLI handler for "add backend".
For now, this handler only performs a parsing of the name argument and
return an error if a duplicate already exists. It runs under thread
isolation, to guarantee thread safety during the proxy creation.
This feature is considered in development. CLI command requires to set
experimental-mode.
Move backend compatibility checks performed during 'add server' in a
dedicated function be_supports_dynamic_srv(). This should simplify
addition of future restriction.
This function will be reused when implementing backend creation at
runtime.
Define a new utility function str_to_proxy_mode() which is able to
convert a string into the corresponding proxy mode if possible. This new
function is used for the parsing of "mode" configuration proxy keyword.
This patch will be reused for dynamic backend implementation, in order
to parse a similar "mode" argument via a CLI handler.
If a proxy is referencing a defaults instance, some checks must be
performed to ensure that inheritance will be compatible. Refcount of the
defaults instance may also be incremented if some settings cannot be
copied. This operation is performed when parsing a new proxy of defaults
section which references a defaults, either implicitely or explicitely.
This patch extracts this code into a dedicated function named
proxy_ref_defaults(). This in turn may call defaults_px_ref()
(previously called proxy_ref_defaults()) to increment its refcount.
The objective of this patch is to be able to reuse defaults inheritance
validation for dynamic backends created at runtime, outside of the
parsing code.
A lot of proxies initialization code is delayed on post-parsing stage,
as it depends on the configuration fully parsed. This is performed via a
loop on proxies_list.
Extract this code in a dedicated function proxy_finalize(). This patch
will be useful for dynamic backends creation.
Note that for the moment the code has been extracted as-is. With each
new features, some init code was added there. This has become a giant
loop with no real ordering. A future patch may provide some cleanup in
order to reorganize this.
Default proxies validation occurs during post-parsing. The objective is
to report any tcp/http-rules which could not behave as expected.
Previously, this was performed while looping over standard proxies list,
when such proxy is referencing a default instance. This was enough as
only named referenced proxies were kept after parsing. However, this is
not the case anymore in the context of dynamic backends creation at
runtime.
As such, this patch now performs validation on every named defaults
outside of the standard proxies list loop. This should not cause any
behavior difference, as defaults are validated without using the proxy
which relies on it.
Along with this change, PR_FL_READY proxy flag is now removed. Its usage
was only really needed for defaults, to avoid validating a same instance
multiple times. With the validation of defaults in their own loop, it is
now redundant.
Fix unhandled strdup() failure when initializing global.log_tag.
Bug was introduced with the fix UAF for global progname pointer from
351ae5dbe. So it must be backported as far as 3.1.
Initially when init_early was introduced the progname string was a local
used for temporary storage of log_tag. Now it's global and detached from
log_tag enough. Thus, in the past we could inform that log_tag
allocation has been failed but not now.
Must be backported since the progname string became global, that is
v3.1-dev9-96-g49772c55e
Differ checking the max threads per group number until we're done
parsing the configuration file, as it may be set after a "thread-group-
directive. Otherwise the default value of 64 will be used, even if there
is a max-threads-per-group directive.
This should be backported to 3.3.
Give global.maxthrpertgroup its default value at global creation,
instead of later when we're trying to detect the thread count.
It is used when verifying the configuration file validity, and if it was
not set in the config file, in a few corner cases, the value of 0 would
be used, which would then reject perfectly fine configuration files.
This should be backported to 3.3.
This patch fixes the bug presented in issue #3254
(https://github.com/haproxy/haproxy/issues/3254), which
occured on FreeBSD when using a stream socket for in
nameserver section. This bug occured due to an incorrect
reset of the alt_proto for a stream socket when the default
socket is created as a datagram socket. This patch fixes
this bug by doing a late assignment to alt_proto when
a datagram socket is requested, leaving only the modification
of alt_proto done by mptcp. Additional documentation
for the use of alt_proto has also been added to
clarify the use of the alt_proto variable.
As reported by Damien Claisse and Cdric Paillet, the "random" LB
algorithm can become particularly unfair with large numbers of servers
having few connections. It's indeed fairly common to see many servers
with zero connection in a thousand-server large farm, and in this case
the P2C algo consisting in checking the servers' loads doesn't help at
all and is basically similar to random(1). In this case, we only rely
on the distribution of server IDs in the random space to pick the best
server, but it's possible to observe huge discrepancies.
An attempt to model the problem clearly shows that with 1600 servers
with weight 10, for 1 million requests, the lowest loaded ones will
take 300 req while the most loaded ones will get 780, with most of
the values between 520 and 700.
In addition, only the first 28 lower bits of server IDs are used for
the key calculation, which means that node keys are more determinist.
Setting random keys in the lowest 28 bits only better packs values
with min around 530 and max around 710, with values mostly between
550 and 680.
This can only be compensated by increasing weights and draws without
being a perfect fix either. At 4 draws, the min is around 560 and the
max around 670, with most values bteween 590 and 650.
This patch takes another approach to this problem: when servers are on
tie regarding their loads, instead of arbitrarily taking the second one,
we now compare their current request rates, which is updated all the
time and smoothed over one second, and we pick the server with the
lowest request rate. Now with 2 draws, the curve is mostly flat, with
the min at 580 and the max at 628, and almost all values between 611
and 625. And 4 draws exclusively gives values from 614 to 624.
Other points will need to be addressed separately (bits of server ID,
maybe refine the hash algorithm), but these ones would affect how
caches are selected, and cannot be changed without an extra option.
For random however we can perform a change without impacting anyone.
This should be backported, probably only to 3.3 since it's where the
"random" algo became the default.
The per-cpu capacity of a cluster was taken into account since 3.2 with
commit 6c88e27cf4 ("MEDIUM: cpu-topo: change "performance" to consider
per-core capacity").
In cpu_policy_performance() and cpu_policy_efficiency(), we're trying
to figure which cores have more capacity than others by comparing their
cluster's average capacity. However, contrary to what the comment says,
we're not averaging per core but per cpu, which makes a difference for
CPUs mixing SMT with non-SMT cores on the same SoC, such as intel's 14th
gen CPUs. Indeed, on a machine where cpufreq is not enabled, all CPUs
can be reported with a capacity of 1024, resulting in a big cluster of
16*1024, and 4 small clusters of 4*1024 each, giving an average of 1024
per CPU, making it impossible to distinguish one from the other. In this
situation, both "cpu-policy performance" and "cpu-policy efficiency"
enable all cores.
But this is wrong, what needs to be taken into account in the divide is
the number of *cores*, not *cpus*, that allows to distinguish big from
little clusters. This was not noticeable on the ARM machines the commit
above aimed at fixing because there, the number of CPUs equals the number
of cores. And on an x86 machine with cpu_freq enabled, the frequencies
continue to help spotting which ones are big/little.
By using nb_cores instead of nb_cpus in the comparison and in the avg_capa
compare function, it properly works again on x86 without affecting other
machines with 1 CPU per core.
This can be backported to 3.2.
When we're about to enter polling, atomically set TH_FL_SLEEPING and
remove TH_FL_NOTIFIED, instead of doing it in sequence. Otherwise,
another thread may sett that both the TH_FL_SLEEPING and the
TH_FL_NOTIFIED bits are set, and don't wake up the thread then it should
be doing that.
This prevents a bug where a thread is sleeping while it should be
handling a new connection, which can happen if there are very few
incoming connection. This is easy to reproduce when using only two
threads, and injecting with only one connection, the connection may then
never be handled.
This should be backported up to 2.8.
This patch adds a new 'post-80' option that sets the
CLIENT_PLUGIN_AUTH (0x00080000) capability flag
and explicitly specifies mysql_native_password as
the authentication plugin in the handshake response.
This patch also addes documentation content for post-80 option
support in MySQL 8.x version. Which handles new default auth
plugin caching_sha2_password.
MySQL 8.0 changed the default authentication plugin from
mysql_native_password to caching_sha2_password.
The current mysql-check implementation only supports pre-41
and post-41 client auth protocols, which lack the CLIENT_PLUGIN_AUTH
capability flag. When HAProxy sends a post-41 authentication
packet to a MySQL 8.x server, the server responds with error 1251:
"Client does not support authentication protocol requested by server".
The new client capabilities for post-80 are:
- CLIENT_PROTOCOL_41 (0x00000200)
- CLIENT_SECURE_CONNECTION (0x00008000)
- CLIENT_PLUGIN_AUTH (0x00080000)
Usage example:
backend mysql_servers
option mysql-check user haproxy post-80
server db1 192.168.1.10:3306 check
The health check user must be created with mysql_native_password:
CREATE USER 'haproxy'@'%' IDENTIFIED WITH mysql_native_password BY '';
This addresses https://github.com/haproxy/haproxy/issues/2934.
Commit fa094d0b61 changed the msg callback
args, but forgot to fix quic_tls_msg_callback() accordingly, so do that,
and remove the unused struct connection paramter.
A regression was introduced in the commit 0ea601127 ("BUG/MAJOR: applet: Don't
call I/O handler if the applet was shut"). The test on shut flags for legacy
applets is inverted.
It should be harmeless on 3.4 and 3.3 because all applets were converted. But
this fix is mandatory for 3.2 and older.
The patch must be backported as far as 3.0 with the commit above.
OpenSSL 4.0 is deprecating X509_STORE_get0_objects().
Every occurence of X509_STORE_get0_objects() was first replaced by
X509_STORE_get1_objects().
This changes the ref count of the STACK_OF(X509_OBJECT) everywhere, and
need it to be sk_X509_OBJECT_pop_free(objs, X509_OBJECT_free) each time.
X509_STORE_get1_objects() is not available in AWS-LC, OpenSSL < 3.2,
LibreSSL and WolfSSL, so we need to still be compatible with get0.
To achieve this, 2 macros were added X509_STORE_getX_objects() and
sk_X509_OBJECT_popX_free(), these macros will use either the get0 or the
get1 macro depending on their availability. In the case of get0,
sk_X509_OBJECT_popX_free() will just do nothing instead of trying to
free.
Don't backport that unless really needed if we want to be compatible
with OpenSSL 4.0. It changes all the refcounts.
SSL msg callbacks are used for notification about sent/received SSL
messages. Such callbacks are registered via
ssl_sock_register_msg_callback().
Prior to this patch, connection was passed as first argument of these
callbacks. However, most of them do not use it. Worst, this may lead to
confusion as connection can be NULL in QUIC context.
This patch cleans this by removing connection argument. As an
alternative, connection can be retrieved in callbacks if needed using
ssl_sock_get_conn() but the code must be ready to deal with potential
NULL instances. As an example, heartbeat parsing callback has been
adjusted in this manner.
With QUIC backend implementation, SSL code has been adjusted in several
place when accessing connection instance. Indeed, with QUIC usage, SSL
context is tied up to quic_conn, and code may be executed prior/after
connection instantiation. For example, on frontend side, connection is
only created after QUIC handshake completion.
The following patch tried to fix unsafe accesses to connection. In
particular, msg callbacks are not called anymore if connection is NULL.
fab7da0fd0
BUG/MEDIUM: quic-be/ssl_sock: TLS callback called without connection
However, most msg callbacks do not need to use the connection instance.
The only occurence where it is accessed is for heartbeat message
parsing, which is the only case of crash solved. The above fix is too
restrictive as it completely prevents execution of these callbacks when
connection is unset. This breaks several features with QUIC, such as SSL
key logging or samples based on ClientHello capture.
The current patch reverts the above one. Thus, this restores invokation
of msg callbacks for QUIC during the whole low-level connection
lifetime. This requires a small adjustment in heartbeat parsing callback
to prevent access on a NULL connection.
The issue on ClientHello capture was mentionned in github issue #2495.
This must be backported up to 3.3.
The help message for "ktls" mentions "expose-experimental-directive"
without the final 's', which is particularly annoying when copy-pasting
the directive from the error message directly into the config.
This should be backported to 3.3.
As reported by Ben Kallus in the following thread:
https://www.mail-archive.com/haproxy@formilux.org/msg46471.html
there exist some agents which mistakenly accept CRLF inside quoted
chunk extensions, making it possible to fool them by injecting one
extra chunk they won't see for example, or making them miss the end
of the body depending on how it's done. Haproxy, like most other
agents nowadays, doesn't care at all about chunk extensions and just
drops them, in agreement with the spec.
However, as discussed, since chunk extensions are basically never used
except for attacks, and that the cost of just matching quote pairs and
checking backslashed quotes is escape consistency remains relatively
low, it can make sense to add such a check to abort the message parsing
when this situation is encountered. Note that it has to be done at two
places, because there is a fast path and a slow path for chunk parsing.
Also note that it *will* cause transfers using improperly formatted chunk
extensions to fail, but since these are really not used, and that the
likelihood of them being used but improperly quoted certainly is much
lower than the risk of crossing a broken parser on the client's request
path or on the server's response path, we consider the risk as
acceptable. The test is not subject to the configurable parser exceptions
and it's very unlikely that it will ever be needed.
Since this is done in 3.4 which will be LTS, this patch will have to be
backported to 3.3 so that any unlikely trouble gets a chance to be
detected before users upgrade to 3.4.
Thanks to Ben for the discussion, and to Rajat Raghav for sparking it
in the first place even though the original report was mistaken.
Cc: Ben Kallus <benjamin.p.kallus.gr@dartmouth.edu>
Cc: Rajat Raghav <xclow3n@gmail.com>
Cc: Christopher Faulet <cfaulet@haproxy.com>
The code in srv_add_to_idle_list() has its roots in 2.0 with commit
9ea5d361ae ("MEDIUM: servers: Reorganize the way idle connections are
cleaned."). At this era we didn't yet have the current set of atomic
load/store operations and we used to perform loads using volatile casts
after a barrier. It turns out that this function has kept this schema
over the years, resulting in a big mfence stalling all the pipeline
in the function:
| static __inline void
| __ha_barrier_full(void)
| {
| __asm __volatile("mfence" ::: "memory");
27.08 | mfence
| if ((volatile void *)srv->idle_node.node.leaf_p == NULL) {
0.84 | cmpq $0x0,0x158(%r15)
0.74 | je 35f
| return 1;
Switching these for a pair of atomic loads got rid of this and brought
0.5 to 3% extra performance depending on the tests due to variations
elsewhere, but it has never been below 0.5%. Note that the second load
doesn't need to be atomic since it's protected by the lock, but it's
cleaner from an API and code review perspective. That's also why it's
relaxed.
This was the last user of _ha_barrier_full(), let's try not to
reintroduce it now!
This option allows to disable the certificate compression (RFC 8879)
using OpenSSL >= 3.2.0.
This feature is known to permit some denial of services by causing extra
memory allocations of approximately 22MiB and extra CPU work per
connection with OpenSSL versions affected by CVE-2025-66199.
( https://openssl-library.org/news/vulnerabilities/index.html#CVE-2025-66199 )
Setting this to "off" permits to mitigate the problem.
Must be backported to every stable branches.
In 3.0, it was stated an applet could not be woken up after it was shutdown.
So the corresponding test in the applets I/O handler was removed. However,
it seems it may happen, especially when outgoing data are blocked on the
opposite side. But it is really unexpected because the "release" callback
function was already called and the appctx context was most probably
released.
Strangely, it was never detected by any applet till now. But the Prometheus
exporter was never updated and was still testing the shutdown. But when it
was refactored to use the new applet API in 3.3, the test was removed. And
this introduced a regression leading a crash because a server object could
be corrupted. Conditions to hit the bug are not really clear however.
So, now, to avoid any issue with all other applets, the test is performed in
task_process_applet(). The I/O handler is no longer called if the applet is
already shut.
The same is performed for applets still relying on the old API.
An amazing thanks to @idl0r for his invaluable help on this issue !
This patch should fix the issue #3244. It should first be backported to 3.3
and then slowly as far as 3.0.
The SSL passphrase callback function was only called when loading
private keys from a dedicated file (separate from the corresponding
certificate) but not when both the certificate and the key were in the
same file.
We can now load them properly, regardless of how they are provided.
A flas had to be added in the 'passphrase_cb_data' structure because in
the 'ssl_sock_load_pem_into_ckch' function, when calling
'PEM_read_bio_PrivateKey' there might be no private key in the PEM file
which would mean that the callback never gets called (and cannot set the
'passphrase_idx' to -1).
This patch can be backported to 3.3.
Some error paths in 'ssl_sock_passwd_cb' (allocation failures) did not
set the 'passphrase_idx' to -1 which is the way for the caller to know
not to call the callback again so in some memory contention contexts we
could end up calling the callback 'infinitely' (or until memory is
finally available).
This patch must be backported to 3.3.
Certain object sizes cannot be controlled at declaration time because
the resulting object size may be slightly extended (tag, caller),
aligned and rounded up, or even doubled depending on pool settings
(e.g. if backup is used).
This patch addresses this by enlarging the type in the pool registration
to 64-bit so that no info is lost from the declaration, and extra checks
for overflows can be performed during registration after various rounding
steps. This allows to catch issues such as these ones and to report a
suitable error:
global
tune.http.logurilen 2147483647
frontend
capture request header name len 2147483647
http-request capture src len 2147483647
tcp-request content capture src len 2147483647
Since 3.3 with commit 945aa0ea82 ("MINOR: initcalls: Add a new initcall
stage, STG_INIT_2"), stkt_late_init() calls stkt_create_stk_ctr_pool()
but doesn't check its return value, so if the pool creation fails, the
process still starts, which is not correct. This patch adds a check for
the return value to make sure we fail to start in this case. This was
not an issue before 3.3 because the function was called as a post-check
handler which did check for errors in the returned values.
A few capture pools can fail in case of too large values for example.
These include the req_uri, capture, and caphdr pools, and may be triggered
with "tune.http.logurilen 2147483647" in the global section, or one of
these in a frontend:
capture request header name len 2147483647
http-request capture src len 2147483647
tcp-request content capture src len 2147483647
These seem to be the only occurrences where create_pool()'s return value
is assigned without being checked, so let's add the proper check for
errors there. This can be backported as a hardening measure though the
risks and impacts are extremely low.
Condition to report the support for HAVE_TCP_MD5SIG feature was inverted. It
is only an issue for the reg-test related to this feature.
This patch must be backported to 3.3.
UNUSED blocks were not properly handled when the H1 multiplexer was
formatting the start line of a request or a response. UNUSED was ignored but
not removed from HTX message. So the mux can loop infinitly on such block.
It could be seen a a major issue but in fact it happens only if a very
specific case on the reponse processing (at least I think so): the server
must send an interim message (a 100-continue for intance) with the final
response. HAProxy must receive both in same time and the final reponse must
be intercepted (via a http-response return action for instance), In that
case, the interim message is fowarded and the server final reponse is
removed and replaced by a proxy error message.
Now UNUSED htx blocks are properly skipped and removed.
This patch must be backported as far as 3.0.
We frequently use lua_pcall() to provide safe alternative functions
(encapsulated helpers) that prevent the process from crashing in case
of Lua error when Lua is executed from an unsafe environment.
However, some of those safe helpers don't handle errors properly. In case
of error, the Lua API will always put an error object on top of the stack
as stated in the documentation. This error object can be used to retrieve
more info about the error. But in some cases when we ignore it, we should
still consume it to prevent the stack from being altered with an extra
object when returning from the helper function.
It should be backported to all stable versions. If the patch doesn't apply
automatically, all that's needed is to check for lua_pcall() in hlua.c
and for other cases than 'LUA_OK', make sure that the error object is popped
from the stack before the function returns.
Since commit 365ee28 ("BUG/MINOR: hlua: prevent LJMP in hlua_traceback()")
we now use lua_pcall() to protect sensitive parts of hlua_traceback()
function, and this to prevent Lua from crashing the process in case of
unexpected Lua error.
This is still relevant, but an error was made, as lua_pcall() was given
the nresult argument '1' when _hlua_traceback() internal function
doesn't push any argument on the stack. Because of this, it seems Lua
API still tries to push garbage object on top of the stack before
returning. This may cause functions that leverage hlua_traceback() in
the middle of stack manipulation to end up having a corrupted stack when
continuing after the hlua_traceback().
There doesn't seem to be many places where this could be a problem, as
this was discovered using the reproducer documented in f535d3e
("BUG/MEDIUM: debug: only dump Lua state when panicking"). Indeed, when
hlua_traceback() was used from the signal handler while the thread was
previously executing Lua, when returning to Lua after the handler the
Lua stack would be corrupted.
To fix the issue, we emphasize on the fact that the _hlua_traceback()
function doesn't push anything on the stack, returns 0, thus lua_pcall()
is given 0 'nresult' argument to prevent anything from being pushed after
the execution, preserving the original stack state.
This should be backported to all stable versions (because 365ee28 was
backported there)
This patch changes the handling of named defaults sections. Prior to
this patch, every unreferenced defaults proxies were removed on post
parsing. Now by default, these sections are kept after postparsing and
only purged on deinit. The objective is to allow reusing them as base
configuration for dynamic backends.
To implement this, refcount of every still addressable named sections is
incremented by one after parsing. This ensures that they won't be
removed even if referencing proxies are removed at runtime. This is done
via the new function proxy_ref_all_defaults().
To ensure defaults instances are still properly removed on deinit, the
inverse operation is performed : refcount is decremented by one on every
defaults sections via proxy_unref_all_defaults().
The original behavior can still be used by using the new global keyword
tune.defaults.purge. This is useful for users using configuration with
large number of defaults and not interested in dynamic backends
creation.
Defaults section are indexed by their name in defproxy_by_name tree. For
named sections, there is no duplicate : if two instances have the same
name, the older one is removed from the tree. However, this was not the
case for unnamed defaults which are all stored inconditionnally in
defproxy_by_name.
This commit introduces a new approach for unnamed defaults. Now, these
instances are never inserted in the defproxy_by_name tree. Indeed, this
is not needed as no tree lookup is performed with empty names. This may
optimize slightly config parsing with a huge number of named and unnamed
defaults sections, as the first ones won't fill up the tree needlessly.
However, defproxy_by_name tree is also used to purge unreferenced
defaults instances, both on postparsing and deinit. Thus, a new approach
is needed for unnamed sections cleanup. Now, each time a new defaults is
parsed, if the previous instance is unnamed, it is freed unless if
referenced by a proxy. When config parsing is ended, a similar operation
is performed to ensure the last unnamed defaults section won't stay in
memory. To implement this, last_defproxy static variable is now set to
global. Unnamed sections which cannot be removed due to proxies
referencing proxies will still be removed when such proxies are freed
themselves, at runtime or on deinit.
Defaults proxies instance are stored in a global name tree. When there
is a name conflict and the older entry cannot be simply discarded as it
is already referenced, the older entry is instead removed from the name
tree and inserted into the orphaned list.
The purpose of the orphaned list was to guarantee that any remaining
unreferenced defaults are purged either on postparsing or deinit.
However, this is in fact completely useless. Indeed on postparsing,
orphaned entries are always referenced. On deinit instead, defaults are
already freed along the cleanup of all frontend/backend instances clean
up, thanks to their refcounting.
This patch streamlines this by removing orphaned list. Instead, a
defaults section is inserted into a new global defaults_list during
their whole lifetime. This is not strictly necessary but it ensures that
defaults instances can still be accessed easily in the future if needed
even if not present in the name tree. On deinit, a BUG_ON() is added to
ensure that defaults_list is indeed emptied.
Another benefit from this patch is to simplify the defaults deletion
procedure. Orphaned simple list is replaced by a proper double linked
list implementation, so a single LIST_DELETE() is now performed. This
will be notably useful as defaults may be removed at runtime in the
future if backends deletion at runtime is implemented.
This patch renames functions which deal with defaults section. A common
"defaults_px_" prefix is defined. This serves as a marker to identify
functions which can only be used with proxies defaults capability. New
BUG_ON() are enforced to ensure this is valid.
Also, older proxy_unref_or_destroy_defaults() is renamed
defaults_px_detach().
Function proxy_preset_defaults() purpose has evolved over time.
Originally, it was only used to initialize defaults proxies instances.
Until today, it was extended so that all proxies use it. Its objective
is to initialize settings to common default values.
To remove the confusion, this function is now removed. Its content is
integrated directly into init_new_proxy().
For a long time, we've tried to show the Lua state and backtrace when
dumping threads so as to be able to figure is (and which) Lua code was
misbehaving, e.g. by performing expensive library calls. Since 3.1 with
commit 365ee28510 ("BUG/MINOR: hlua: prevent LJMP in hlua_traceback()"),
it appears that the approach is more fragile (though that fix addressed
a real issue about out-of-memory), and it's possible to occasionally
observe crashes or CPU loops with "show threads" while running Lua
heavily. While users of "show threads" are rare, the watchdog warnings,
which were also enabled on 3.1, also trigger these issues, which is
even more of a concern.
This patch goes the simple way to address this for now: since the purpose
of the Lua backtrace was to help locate Lua call places upon a panic,
let's only call the backtrace on panic but not in other situations. After
a panic we obviously don't care that the Lua stack might be corrupted
since it's never going to be resumed anyway. This may be relaxed in the
future if a solution is found to reliably produce harmless Lua backtraces.
The commit above was backported to all stable branches, so this patch
will be needed everywhere. However, TAINTED_PANIC only appeared in 2.8,
and given the rarety of this bug before 3.1, it's probably not needed
to make any extra effort to go beyond 2.8.
It's easy enough to test a version for being subject to this issue,
by running the following Lua code:
local function stress(txn)
for _, backend in pairs(core.backends) do
for _, server in pairs(backend.servers) do
local stats = server:get_stats()
end
end
end
core.register_fetches("stress", stress)
in the following config file:
global
stats socket /tmp/haproxy.stat level admin mode 666
tune.lua.bool-sample-conversion normal
lua-load-per-thread "stress.lua"
listen stress
bind :8001
mode http
timeout client 5s
timeout server 5s
timeout connect 5s
http-request return status 200 content-type text/plain lf-string %[lua.stress()]
server s1 127.0.0.1:8000
and stressing port 8001 with 100+ connections requesting / in loop, then
issuing "show threads" on the CLI using socat in loops as well. Normally
it instantly segfaults (sometimes during the first "show").
