BUG/MEDIUM: htx: Fix htx_xfer() to consume more data than expected

When an htx DATA block is partially transfer, we must take care to remove
exactly the copied size. To do so, we must save the size of the last block
value copied and not rely on the last data block after the copy. Indeed,
data can be merged with an existing DATA block, so the last block size can
be larger than the last part copied.

Because of this issue, it is possible to remove more data than
expected. Worse, this could lead to a crash by performing an integer
overflow on the block size.

No backport needed.

doc/configuration.txt

@@ -1886,10 +1886,13 @@ The following keywords are supported in the "global" section :
    - tune.h2.be.glitches-threshold
    - tune.h2.be.initial-window-size
    - tune.h2.be.max-concurrent-streams
+   - tune.h2.be.max-frames-at-once
    - tune.h2.be.rxbuf
    - tune.h2.fe.glitches-threshold
    - tune.h2.fe.initial-window-size
    - tune.h2.fe.max-concurrent-streams
+   - tune.h2.fe.max-frames-at-once
+   - tune.h2.fe.max-rst-at-once
    - tune.h2.fe.max-total-streams
    - tune.h2.fe.rxbuf
    - tune.h2.header-table-size
@@ -4162,8 +4165,11 @@ tune.bufsize.small <size>
   If however a small buffer is not sufficient, a reallocation is automatically
   done to switch to a standard size buffer.
 
-  For the moment, it is used only by HTTP/3 protocol to emit the response
+  For the moment, it is automatically used only by HTTP/3 protocol to emit the
-  headers.
+  response headers. Otherwise, small buffers support can be enabled for
+  specific proxies via the "use-small-buffers" option.
+
+  See also: option use-small-buffers
 
 tune.comp.maxlevel <number>
   Sets the maximum compression level. The compression level affects CPU
@@ -4368,6 +4374,13 @@ tune.h2.be.max-concurrent-streams <number>
   case). It is highly recommended not to increase this value; some might find
   it optimal to run at low values (1..5 typically).
 
+tune.h2.be.max-frames-at-once <number>
+  Sets the maximum number of HTTP/2 incoming frames that will be processed at
+  once on a backend connection. It can be useful to set this to a low value
+  (a few tens to a few hundreds) when dealing with very large buffers in order
+  to maintain a low latency and a better fairness between multiple connections.
+  The default value is zero, which means that no limitation is enforced.
+
 tune.h2.be.rxbuf <size>
   Sets the HTTP/2 receive buffer size for outgoing connections, in bytes. This
   size will be rounded up to the next multiple of tune.bufsize and will be
@@ -4458,6 +4471,25 @@ tune.h2.fe.max-concurrent-streams <number> [args...]
 
      tune.h2.fe.max-concurrent-streams 100 rq-load auto min 15
 
+tune.h2.fe.max-frames-at-once <number>
+  Sets the maximum number of HTTP/2 incoming frames that will be processed at
+  once on a frontend connection. It can be useful to set this to a low value
+  (a few tens to a few hundreds) when dealing with very large buffers in order
+  to maintain a low latency and a better fairness between multiple connections.
+  The default value is zero, which means that no limitation is enforced.
+
+tune.h2.fe.max-rst-at-once <number>
+  Sets the maximum number of HTTP/2 incoming RST_STREAM that will be processed
+  at once on a frontend connection. Once the specified number of RST_STREAM
+  frames are received, the connection handler will be placed in a low priority
+  queue and be processed after all other tasks. It can be useful to set this to
+  a very low value (1 or a few units) to significantly reduce the impacts of
+  RST_STREAM floods. RST_STREAM do happen when a user clicks on the Stop button
+  in their browser, but the few extra milliseconds caused by this requeuing are
+  generally unnoticeable, however they are generally effective at significantly
+  lowering the load caused from such floods. The default value is zero, which
+  means that no limitation is enforced.
+
 tune.h2.fe.max-total-streams <number>
   Sets the HTTP/2 maximum number of total streams processed per incoming
   connection. Once this limit is reached, HAProxy will send a graceful GOAWAY
@@ -4604,22 +4636,22 @@ tune.http.maxhdr <number>
   protocols. This limit is large enough but not documented on purpose. The same
   limit is applied on the first steps of the decoding for the same reason.
 
-tune.idle-pool.shared { on | off }
+tune.idle-pool.shared { auto | on | off }
-  Enables ('on') or disables ('off') sharing of idle connection pools between
+  Controls sharing idle connection pools between threads for a same server.
-  threads for a same server. The default is to share them between threads in
+  It can be enabled for all threads in a same thread group ('on'), enabled for
-  order to minimize the number of persistent connections to a server, and to
+  all threads ('full') or disabled ('off'). The default is to share them
-  optimize the connection reuse rate. But to help with debugging or when
+  between threads in the same thread group ('on'), in order to minimize the
+  number of persistent connections to a server, and to optimize the connection
+  reuse rate. Sharing with threads from other thread groups can have a
+  performance impact, and is not enabled by default, but can be useful if
+  maximizing connection reuse is a priority. To help with debugging or when
   suspecting a bug in HAProxy around connection reuse, it can be convenient to
-  forcefully disable this idle pool sharing between multiple threads, and force
+  forcefully disable this idle pool sharing between multiple threads,
-  this option to "off". The default is on. It is strongly recommended against
+  and force this option to "off". It is strongly recommended against disabling
-  disabling this option without setting a conservative value on "pool-low-conn"
+  this option without setting a conservative value on "pool-low-conn" for all
-  for all servers relying on connection reuse to achieve a high performance
+  servers relying on connection reuse to achieve a high performance level,
-  level, otherwise connections might be closed very often as the thread count
+  otherwise connections might be closed very often as the thread count
-  increases. Note that in any case, connections are only shared between threads
+  increases.
-  of the same thread group. This means that systems with many NUMA nodes may
-  show slightly more persistent connections while machines with unified caches
-  and many CPU cores per node may experience higher CPU usage. In the latter
-  case, the "max-thread-per-group" tunable may be used to improve the behavior.
 
 tune.idletimer <timeout>
   Sets the duration after which HAProxy will consider that an empty buffer is
@@ -5555,6 +5587,9 @@ tune.takeover-other-tg-connections <value>
   connections.
   Note that using connections from other thread groups can occur performance
   penalties, so it should not be used unless really needed.
+  Note that this behavior is now controlled by tune.idle-pool.shared, and
+  this keyword is just there for compatibility with older configurations, and
+  will be deprecated.
 
 tune.vars.global-max-size <size>
 tune.vars.proc-max-size <size>
@@ -5942,6 +5977,8 @@ errorloc302                               X          X         X         X
 -- keyword -------------------------- defaults - frontend - listen -- backend -
 errorloc303                               X          X         X         X
 error-log-format                          X          X         X         -
+external-check command                    X          -         X         X
+external-check path                       X          -         X         X
 force-persist                             -          -         X         X
 force-be-switch                           -          X         X         -
 filter                                    -          X         X         X
@@ -5987,6 +6024,7 @@ option disable-h2-upgrade            (*)  X          X         X         -
 option dontlog-normal                (*)  X          X         X         -
 option dontlognull                   (*)  X          X         X         -
 -- keyword -------------------------- defaults - frontend - listen -- backend -
+option external-check                     X          -         X         X
 option forwardfor                         X          X         X         X
 option forwarded                     (*)  X          -         X         X
 option h1-case-adjust-bogus-client   (*)  X          X         X         -
@@ -6005,9 +6043,9 @@ option httpchk                            X          -         X         X
 option httpclose                     (*)  X          X         X         X
 option httplog                            X          X         X         -
 option httpslog                           X          X         X         -
+option idle-close-on-response        (*)  X          X         X         -
 option independent-streams           (*)  X          X         X         X
 option ldap-check                         X          -         X         X
-option external-check                     X          -         X         X
 option log-health-checks             (*)  X          -         X         X
 option log-separate-errors           (*)  X          X         X         -
 option logasap                       (*)  X          X         X         -
@@ -6034,9 +6072,7 @@ option tcp-smart-connect             (*)  X          -         X         X
 option tcpka                              X          X         X         X
 option tcplog                             X          X         X         -
 option transparent      (deprecated) (*)  X          -         X         X
-option idle-close-on-response        (*)  X          X         X         -
+option use-small-buffers             (*)  X          -         X         X
-external-check command                    X          -         X         X
-external-check path                       X          -         X         X
 persist rdp-cookie                        X          -         X         X
 quic-initial                              X (!)      X         X         -
 rate-limit sessions                       X          X         X         -
@@ -7699,6 +7735,96 @@ force-persist { if | unless } <condition>
              and section 7 about ACL usage.
 
 
+external-check command <command>
+  Executable to run when performing an external-check
+
+  May be used in the following contexts: tcp, http, log
+
+  May be used in sections :   defaults | frontend | listen | backend
+                                 yes   |    no    |   yes  |   yes
+
+  Arguments :
+    <command> is the external command to run
+
+  The arguments passed to the to the command are:
+
+  <proxy_address> <proxy_port> <server_address> <server_port>
+
+  The <proxy_address> and <proxy_port> are derived from the first listener
+  that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
+  listener the proxy_address will be the path of the socket and the
+  <proxy_port> will be the string "NOT_USED". In a backend section, it's not
+  possible to determine a listener, and both <proxy_address> and <proxy_port>
+  will have the string value "NOT_USED".
+
+  Some values are also provided through environment variables.
+
+  Environment variables :
+    HAPROXY_PROXY_ADDR      The first bind address if available (or empty if not
+                            applicable, for example in a "backend" section).
+
+    HAPROXY_PROXY_ID        The backend id.
+
+    HAPROXY_PROXY_NAME      The backend name.
+
+    HAPROXY_PROXY_PORT      The first bind port if available (or empty if not
+                            applicable, for example in a "backend" section or
+                            for a UNIX socket).
+
+    HAPROXY_SERVER_ADDR     The server address.
+
+    HAPROXY_SERVER_CURCONN  The current number of connections on the server.
+
+    HAPROXY_SERVER_ID       The server id.
+
+    HAPROXY_SERVER_MAXCONN  The server max connections.
+
+    HAPROXY_SERVER_NAME     The server name.
+
+    HAPROXY_SERVER_PORT     The server port if available (or empty for a UNIX
+                            socket).
+
+    HAPROXY_SERVER_SSL      "0" when SSL is not used, "1" when it is used
+
+    HAPROXY_SERVER_PROTO    The protocol used by this server, which can be one
+                            of "cli" (the haproxy CLI), "syslog" (syslog TCP
+                            server), "peers" (peers TCP server), "h1" (HTTP/1.x
+                            server), "h2" (HTTP/2 server), or "tcp" (any other
+                            TCP server).
+
+    PATH                    The PATH environment variable used when executing
+                            the command may be set using "external-check path".
+
+  If the command executed and exits with a zero status then the check is
+  considered to have passed, otherwise the check is considered to have
+  failed.
+
+  Example :
+        external-check command /bin/true
+
+  See also : "external-check", "option external-check", "external-check path"
+
+
+external-check path <path>
+  The value of the PATH environment variable used when running an external-check
+
+  May be used in the following contexts: tcp, http, log
+
+  May be used in sections :   defaults | frontend | listen | backend
+                                 yes   |    no    |   yes  |   yes
+
+  Arguments :
+    <path> is the path used when executing external command to run
+
+  The default path is "".
+
+  Example :
+        external-check path "/usr/bin:/bin"
+
+  See also : "external-check", "option external-check",
+             "external-check command"
+
+
 force-be-switch { if | unless } <condition>
   Allow content switching to select a backend instance even if it is disabled
   or unpublished. This rule can be used by admins to test traffic to services
@@ -8203,6 +8329,11 @@ http-check expect [min-recv <int>] [comment <msg>]
                        occurred during the expect rule evaluation. <fmt> is a
                        Custom log format string (see section 8.2.6).
 
+    status-code <expr> is optional and can be used to set the check status code
+                       reported in logs, on success or on error. <expr> is a
+                       standard HAProxy expression formed by a sample-fetch
+                       followed by some converters.
+
     <match>   is a keyword indicating how to look for a specific pattern in the
               response. The keyword may be one of "status", "rstatus", "hdr",
               "fhdr", "string", or "rstring". The keyword may be preceded by an
@@ -9920,6 +10051,24 @@ no option dontlognull
   See also : "log", "http-ignore-probes", "monitor-uri", and
              section 8 about logging.
 
+
+option external-check
+  Use external processes for server health checks
+
+  May be used in the following contexts: tcp, http, log
+
+  May be used in sections :   defaults | frontend | listen | backend
+                                 yes   |    no    |   yes  |   yes
+
+  It is possible to test the health of a server using an external command.
+  This is achieved by running the executable set using "external-check
+  command".
+
+  Requires the "external-check" global to be set.
+
+  See also : "external-check", "external-check command", "external-check path"
+
+
 option forwarded [ proto ]
                  [ host | host-expr <host_expr> ]
                  [ by | by-expr <by_expr> ] [ by_port | by_port-expr <by_port_expr>]
@@ -10710,6 +10859,39 @@ option httpslog
   See also :  section 8 about logging.
 
 
+option idle-close-on-response
+no option idle-close-on-response
+  Avoid closing idle frontend connections if a soft stop is in progress
+
+  May be used in the following contexts: http
+
+  May be used in sections :   defaults | frontend | listen | backend
+                                 yes   |    yes   |   yes  |   no
+
+  Arguments : none
+
+  By default, idle connections will be closed during a soft stop. In some
+  environments, a client talking to the proxy may have prepared some idle
+  connections in order to send requests later. If there is no proper retry on
+  write errors, this can result in errors while haproxy is reloading. Even
+  though a proper implementation should retry on connection/write errors, this
+  option was introduced to support backwards compatibility with haproxy prior
+  to version 2.4. Indeed before v2.4, haproxy used to wait for a last request
+  and response to add a "connection: close" header before closing, thus
+  notifying the client that the connection would not be reusable.
+
+  In a real life example, this behavior was seen in AWS using the ALB in front
+  of a haproxy. The end result was ALB sending 502 during haproxy reloads.
+
+  Users are warned that using this option may increase the number of old
+  processes if connections remain idle for too long. Adjusting the client
+  timeouts and/or the "hard-stop-after" parameter accordingly might be
+  needed in case of frequent reloads.
+
+  See also: "timeout client", "timeout client-fin", "timeout http-request",
+            "hard-stop-after"
+
+
 option independent-streams
 no option independent-streams
   Enable or disable independent timeout processing for both directions
@@ -10774,56 +10956,6 @@ option ldap-check
   See also : "option httpchk"
 
 
-option external-check
-  Use external processes for server health checks
-
-  May be used in the following contexts: tcp, http, log
-
-  May be used in sections :   defaults | frontend | listen | backend
-                                 yes   |    no    |   yes  |   yes
-
-  It is possible to test the health of a server using an external command.
-  This is achieved by running the executable set using "external-check
-  command".
-
-  Requires the "external-check" global to be set.
-
-  See also : "external-check", "external-check command", "external-check path"
-
-
-option idle-close-on-response
-no option idle-close-on-response
-  Avoid closing idle frontend connections if a soft stop is in progress
-
-  May be used in the following contexts: http
-
-  May be used in sections :   defaults | frontend | listen | backend
-                                 yes   |    yes   |   yes  |   no
-
-  Arguments : none
-
-  By default, idle connections will be closed during a soft stop. In some
-  environments, a client talking to the proxy may have prepared some idle
-  connections in order to send requests later. If there is no proper retry on
-  write errors, this can result in errors while haproxy is reloading. Even
-  though a proper implementation should retry on connection/write errors, this
-  option was introduced to support backwards compatibility with haproxy prior
-  to version 2.4. Indeed before v2.4, haproxy used to wait for a last request
-  and response to add a "connection: close" header before closing, thus
-  notifying the client that the connection would not be reusable.
-
-  In a real life example, this behavior was seen in AWS using the ALB in front
-  of a haproxy. The end result was ALB sending 502 during haproxy reloads.
-
-  Users are warned that using this option may increase the number of old
-  processes if connections remain idle for too long. Adjusting the client
-  timeouts and/or the "hard-stop-after" parameter accordingly might be
-  needed in case of frequent reloads.
-
-  See also: "timeout client", "timeout client-fin", "timeout http-request",
-            "hard-stop-after"
-
-
 option log-health-checks
 no option log-health-checks
   Enable or disable logging of health checks status updates
@@ -11785,95 +11917,35 @@ no option transparent     (deprecated)
             "transparent" option of the "bind" keyword.
 
 
-external-check command <command>
+option use-small-buffers [ queue | l7-retries | check ]*
-  Executable to run when performing an external-check
 
-  May be used in the following contexts: tcp, http, log
+  Enable support for small buffers for the given categories.
+
+  May be used in the following contexts: tcp, http
 
   May be used in sections :   defaults | frontend | listen | backend
                                  yes   |    no    |   yes  |   yes
 
-  Arguments :
+  This option can be used to enable the small buffers support at diffent places
-    <command> is the external command to run
+  to save memory. By default, with no parameter, small buffers are used as far
+  as possible at all possible places. Otherwise, it is possible to limit it to
+  following the places:
 
-  The arguments passed to the to the command are:
+   - queue:      When set, small buffers will be used to store the requests, if
+                 small enough, when the connection is queued.
+   - l7-retries: When set, small buffers will be used to save the requests
+                 when L7 retries are enabled.
+   - check:      When set, small buffers will be used for the health-checks
+                 requests.
 
-  <proxy_address> <proxy_port> <server_address> <server_port>
+  When enabled, small buffers are used, but only if it is possible. Otherwise,
+  when data are too large, a regular buffer is automtically used. The size of
+  small buffers is configurable via the "tune.bufsize.small" global setting.
 
-  The <proxy_address> and <proxy_port> are derived from the first listener
+  If this option has been enabled in a "defaults" section, it can be disabled
-  that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
+  in a specific instance by prepending the "no" keyword before it.
-  listener the proxy_address will be the path of the socket and the
-  <proxy_port> will be the string "NOT_USED". In a backend section, it's not
-  possible to determine a listener, and both <proxy_address> and <proxy_port>
-  will have the string value "NOT_USED".
-
-  Some values are also provided through environment variables.
-
-  Environment variables :
-    HAPROXY_PROXY_ADDR      The first bind address if available (or empty if not
-                            applicable, for example in a "backend" section).
-
-    HAPROXY_PROXY_ID        The backend id.
-
-    HAPROXY_PROXY_NAME      The backend name.
-
-    HAPROXY_PROXY_PORT      The first bind port if available (or empty if not
-                            applicable, for example in a "backend" section or
-                            for a UNIX socket).
-
-    HAPROXY_SERVER_ADDR     The server address.
-
-    HAPROXY_SERVER_CURCONN  The current number of connections on the server.
-
-    HAPROXY_SERVER_ID       The server id.
-
-    HAPROXY_SERVER_MAXCONN  The server max connections.
-
-    HAPROXY_SERVER_NAME     The server name.
-
-    HAPROXY_SERVER_PORT     The server port if available (or empty for a UNIX
-                            socket).
-
-    HAPROXY_SERVER_SSL      "0" when SSL is not used, "1" when it is used
-
-    HAPROXY_SERVER_PROTO    The protocol used by this server, which can be one
-                            of "cli" (the haproxy CLI), "syslog" (syslog TCP
-                            server), "peers" (peers TCP server), "h1" (HTTP/1.x
-                            server), "h2" (HTTP/2 server), or "tcp" (any other
-                            TCP server).
-
-    PATH                    The PATH environment variable used when executing
-                            the command may be set using "external-check path".
-
-  If the command executed and exits with a zero status then the check is
-  considered to have passed, otherwise the check is considered to have
-  failed.
-
-  Example :
-        external-check command /bin/true
-
-  See also : "external-check", "option external-check", "external-check path"
-
-
-external-check path <path>
-  The value of the PATH environment variable used when running an external-check
-
-  May be used in the following contexts: tcp, http, log
-
-  May be used in sections :   defaults | frontend | listen | backend
-                                 yes   |    no    |   yes  |   yes
-
-  Arguments :
-    <path> is the path used when executing external command to run
-
-  The default path is "".
-
-  Example :
-        external-check path "/usr/bin:/bin"
-
-  See also : "external-check", "option external-check",
-             "external-check command"
 
+  See also: tune.bufsize.small
 
 persist rdp-cookie
 persist rdp-cookie(<name>)
@@ -13632,13 +13704,6 @@ tcp-check expect [min-recv <int>] [comment <msg>]
               does not match, the check will wait for more data. If set to 0,
               the evaluation result is always conclusive.
 
-    <match>   is a keyword indicating how to look for a specific pattern in the
-              response. The keyword may be one of "string", "rstring", "binary" or
-              "rbinary".
-              The keyword may be preceded by an exclamation mark ("!") to negate
-              the match. Spaces are allowed between the exclamation mark and the
-              keyword. See below for more details on the supported keywords.
-
     ok-status <st>     is optional and can be used to set the check status if
                        the expect rule is successfully evaluated and if it is
                        the last rule in the tcp-check ruleset. "L7OK", "L7OKC",
@@ -13686,6 +13751,13 @@ tcp-check expect [min-recv <int>] [comment <msg>]
                        standard HAProxy expression formed by a sample-fetch
                        followed by some converters.
 
+    <match>   is a keyword indicating how to look for a specific pattern in the
+              response. The keyword may be one of "string", "rstring", "binary" or
+              "rbinary".
+              The keyword may be preceded by an exclamation mark ("!") to negate
+              the match. Spaces are allowed between the exclamation mark and the
+              keyword. See below for more details on the supported keywords.
+
     <pattern> is the pattern to look for. It may be a string or a regular
               expression. If the pattern contains spaces, they must be escaped
               with the usual backslash ('\').
@@ -15382,15 +15454,14 @@ disable-l7-retry
   reason than a connection failure. This can be useful for example to make
   sure POST requests aren't retried on failure.
 
-do-log
+do-log [profile <log_profile>]
   Usable in:  QUIC Ini|    TCP RqCon| RqSes| RqCnt| RsCnt|    HTTP Req| Res| Aft
                     X |          X  |   X  |   X  |   X  |          X |  X |  X
 
   This action manually triggers a log emission on the proxy. This means
   log options on the proxy will be considered (including formatting options
   such as "log-format"), but it will not interfere with the logs automatically
-  generated by the proxy during transaction handling. It currently doesn't
+  generated by the proxy during transaction handling.
-  support any argument, though extensions may appear in future versions.
 
   Using "log-profile", it is possible to precisely describe how the log should
   be emitted for each of the available contexts where the action may be used.
@@ -15400,15 +15471,28 @@ do-log
 
   Also, they will be properly reported when using "%OG" logformat alias.
 
+  Optional "profile" argument may be used to specify the name of a log-profile
+  section that should be used for this do-log action specifically instead of
+  the one associated to the current logger that applies by default.
+
   Example:
-    log-profile myprof
+    log-profile my-dft-prof
       on tcp-req-conn format "Connect: %ci"
 
+    log-profile my-local-prof
+      on tcp-req-conn format "Local Connect: %ci"
+
     frontend myfront
-      log stdout format rfc5424 profile myprof local0
+      log stdout format rfc5424 profile my-dft-prof local0
       log-format "log generated using proxy logformat, from '%OG'"
-      tcp-request connection do-log             #uses special log-profile format
+      acl local src 127.0.0.1
-      tcp-request content do-log                #uses proxy logformat
+      # on connection use either log-profile from the logger (my-dft-prof) or
+      # explicit my-local-prof if source ip is localhost
+      tcp-request connection do-log if !local
+      tcp-request connection do-log profile my-local-prof if local
+      # on content use proxy logformat, since no override was specified
+      # in my-dft-prof
+      tcp-request content do-log
 
 do-resolve(<var>,<resolvers>[,ipv4|ipv6]) <expr>
   Usable in:  QUIC Ini|    TCP RqCon| RqSes| RqCnt| RsCnt|    HTTP Req| Res| Aft

doc/internals/api/htx-api.txt

@@ -539,10 +539,22 @@ message. These functions are used by HTX analyzers or by multiplexers.
       with the first block not removed, or NULL if everything was removed, and
       the amount of data drained.
 
-    - htx_xfer_blks() transfers HTX blocks from an HTX message to another,
+    - htx_xfer() transfers HTX blocks from an HTX message to another, stopping
-      stopping after the first block of a specified type is transferred or when
+      when a specific amount of bytes, including meta-data, was copied. If the
-      a specific amount of bytes, including meta-data, was moved. If the tail
+      tail block is a DATA block, it may be partially copied. All other block
-      block is a DATA block, it may be partially moved. All other block are
+      are transferred at once. By default, copied blocks are removed from the
+      original HTX message and headers and trailers parts cannot be partially
+      copied. But flags can be set to change the default behavior:
+
+          - HTX_XFER_KEEP_SRC_BLKS: source blocks are not removed
+          - HTX_XFER_PARTIAL_HDRS_COPY: partial headers and trailers
+            part can be xferred
+          - HTX_XFER_HDRS_ONLY: Only the headers part is xferred
+
+    - htx_xfer_blks() [DEPRECATED] transfers HTX blocks from an HTX message to
+      another, stopping after the first block of a specified type is transferred
+      or when a specific amount of bytes, including meta-data, was moved. If the
+      tail block is a DATA block, it may be partially moved. All other block are
       transferred at once or kept. This function returns a mixed value, with the
       last block moved, or NULL if nothing was moved, and the amount of data
       transferred. When HEADERS or TRAILERS blocks must be transferred, this

doc/management.txt

@@ -1740,10 +1740,7 @@ add backend <name> from <defproxy> [mode <mode>] [guid <guid>] [ EXPERIMENTAL ]
 
   All named default proxies can be used, given that they validate the same
   inheritance rules applied during configuration parsing. There is some
-  exceptions though, for example when the mode is neither TCP nor HTTP. Another
+  exceptions though, for example when the mode is neither TCP nor HTTP.
-  exception is that it is not yet possible to use a default proxies which
-  reference custom HTTP errors, for example via the errorfiles or http-rules
-  keywords.
 
   This command is restricted and can only be issued on sockets configured for
   level "admin". Moreover, this feature is still considered in development so it

include/haproxy/acme-t.h

@@ -58,6 +58,7 @@ struct acme_auth {
        struct ist auth;   /* auth URI */
        struct ist chall;  /* challenge URI */
        struct ist token;  /* token */
+       int validated;     /* already validated */
        int ready;         /* is the challenge ready ? */
        void *next;
 };

include/haproxy/action-t.h

@@ -198,6 +198,11 @@ struct act_rule {
 			struct server *srv; /* target server to attach the connection */
 			struct sample_expr *name; /* used to differentiate idle connections */
 		} attach_srv; /* 'attach-srv' rule */
+		struct {
+			enum log_orig_id orig;
+			char *profile_name;
+			struct log_profile *profile;
+		} do_log; /* 'do-log' action */
 		struct {
 			int value;
 			struct sample_expr *expr;

include/haproxy/check-t.h

@@ -59,6 +59,7 @@ enum chk_result {
 #define CHK_ST_FASTINTER        0x0400  /* force fastinter check */
 #define CHK_ST_READY            0x0800  /* check ready to migrate or run, see below */
 #define CHK_ST_SLEEPING         0x1000  /* check was sleeping, i.e. not currently bound to a thread, see below */
+#define CHK_ST_USE_SMALL_BUFF   0x2000  /* Use small buffers if possible for the request */
 
 /* 4 possible states for CHK_ST_SLEEPING and CHK_ST_READY:
  *   SLP  RDY   State      Description
@@ -188,6 +189,7 @@ struct check {
 	char **envp;				/* the environment to use if running a process-based check */
 	struct pid_list *curpid;		/* entry in pid_list used for current process-based test, or -1 if not in test */
 	struct sockaddr_storage addr;   	/* the address to check */
+	struct protocol *proto;                 /* protocol used for check, may be different from the server's one */
 	char *pool_conn_name;                   /* conn name used on reuse */
 	char *sni;				/* Server name */
 	char *alpn_str;                         /* ALPN to use for checks */

include/haproxy/check.h

@@ -78,7 +78,7 @@ struct task *process_chk(struct task *t, void *context, unsigned int state);
 struct task *srv_chk_io_cb(struct task *t, void *ctx, unsigned int state);
 
 int check_buf_available(void *target);
-struct buffer *check_get_buf(struct check *check, struct buffer *bptr);
+struct buffer *check_get_buf(struct check *check, struct buffer *bptr, unsigned int small_buffer);
 void check_release_buf(struct check *check, struct buffer *bptr);
 const char *init_check(struct check *check, int type);
 void free_check(struct check *check);

include/haproxy/chunk.h

@@ -33,6 +33,7 @@
 
 extern struct pool_head *pool_head_trash;
 extern struct pool_head *pool_head_large_trash;
+extern struct pool_head *pool_head_small_trash;
 
 /* function prototypes */
 
@@ -48,6 +49,7 @@ int chunk_strcmp(const struct buffer *chk, const char *str);
 int chunk_strcasecmp(const struct buffer *chk, const char *str);
 struct buffer *get_trash_chunk(void);
 struct buffer *get_large_trash_chunk(void);
+struct buffer *get_small_trash_chunk(void);
 struct buffer *get_trash_chunk_sz(size_t size);
 struct buffer *get_larger_trash_chunk(struct buffer *chunk);
 int init_trash_buffers(int first);
@@ -133,6 +135,29 @@ static forceinline struct buffer *alloc_large_trash_chunk(void)
 	return chunk;
 }
 
+/*
+ * Allocate a small trash chunk from the reentrant pool. The buffer starts at
+ * the end of the chunk. This chunk must be freed using free_trash_chunk(). This
+ * call may fail and the caller is responsible for checking that the returned
+ * pointer is not NULL.
+ */
+static forceinline struct buffer *alloc_small_trash_chunk(void)
+{
+	struct buffer *chunk;
+
+	if (!pool_head_small_trash)
+		return NULL;
+
+	chunk = pool_alloc(pool_head_small_trash);
+	if (chunk) {
+		char *buf = (char *)chunk + sizeof(struct buffer);
+		*buf = 0;
+		chunk_init(chunk, buf,
+			   pool_head_small_trash->size - sizeof(struct buffer));
+	}
+	return chunk;
+}
+
 /*
  * Allocate a trash chunk accordingly to the requested size. This chunk must be
  * freed using free_trash_chunk(). This call may fail and the caller is
@@ -140,7 +165,9 @@ static forceinline struct buffer *alloc_large_trash_chunk(void)
  */
 static forceinline struct buffer *alloc_trash_chunk_sz(size_t size)
 {
-	if (likely(size <= pool_head_trash->size))
+	if (pool_head_small_trash && size <= pool_head_small_trash->size)
+		return alloc_small_trash_chunk();
+	else if (size <= pool_head_trash->size)
 		return alloc_trash_chunk();
 	else if (pool_head_large_trash && size <= pool_head_large_trash->size)
 		return alloc_large_trash_chunk();
@@ -153,10 +180,12 @@ static forceinline struct buffer *alloc_trash_chunk_sz(size_t size)
  */
 static forceinline void free_trash_chunk(struct buffer *chunk)
 {
-	if (likely(chunk && chunk->size == pool_head_trash->size - sizeof(struct buffer)))
+	if (pool_head_small_trash && chunk && chunk->size == pool_head_small_trash->size - sizeof(struct buffer))
-		pool_free(pool_head_trash, chunk);
+		pool_free(pool_head_small_trash, chunk);
-	else
+	else if (pool_head_large_trash && chunk && chunk->size == pool_head_large_trash->size - sizeof(struct buffer))
 		pool_free(pool_head_large_trash, chunk);
+	else
+		pool_free(pool_head_trash, chunk);
 }
 
 /* copies chunk <src> into <chk>. Returns 0 in case of failure. */

include/haproxy/connection.h

@@ -34,6 +34,7 @@
 #include <haproxy/listener-t.h>
 #include <haproxy/obj_type.h>
 #include <haproxy/pool-t.h>
+#include <haproxy/protocol.h>
 #include <haproxy/server.h>
 #include <haproxy/session-t.h>
 #include <haproxy/task-t.h>
@@ -609,13 +610,13 @@ void list_mux_proto(FILE *out);
  */
 static inline const struct mux_proto_list *conn_get_best_mux_entry(
         const struct ist mux_proto,
-        int proto_side, int proto_mode)
+        int proto_side, int proto_is_quic, int proto_mode)
 {
 	struct mux_proto_list *item;
 	struct mux_proto_list *fallback = NULL;
 
 	list_for_each_entry(item, &mux_proto_list.list, list) {
-		if (!(item->side & proto_side) || !(item->mode & proto_mode))
+		if (!(item->side & proto_side) || !(item->mode & proto_mode) || (proto_is_quic && !(item->mux->flags & MX_FL_FRAMED)))
 			continue;
 		if (istlen(mux_proto) && isteq(mux_proto, item->token))
 			return item;
@@ -640,7 +641,7 @@ static inline const struct mux_ops *conn_get_best_mux(struct connection *conn,
 {
 	const struct mux_proto_list *item;
 
-	item = conn_get_best_mux_entry(mux_proto, proto_side, proto_mode);
+	item = conn_get_best_mux_entry(mux_proto, proto_side, proto_is_quic(conn->ctrl), proto_mode);
 
 	return item ? item->mux : NULL;
 }

include/haproxy/dynbuf.h

@@ -37,6 +37,7 @@
 
 extern struct pool_head *pool_head_buffer;
 extern struct pool_head *pool_head_large_buffer;
+extern struct pool_head *pool_head_small_buffer;
 
 int init_buffer(void);
 void buffer_dump(FILE *o, struct buffer *b, int from, int to);
@@ -66,6 +67,12 @@ static inline int b_is_large_sz(size_t sz)
 	return (pool_head_large_buffer && sz == pool_head_large_buffer->size);
 }
 
+/* Return 1 if <sz> is the size of a small buffer */
+static inline int b_is_small_sz(size_t sz)
+{
+	return (pool_head_small_buffer && sz == pool_head_small_buffer->size);
+}
+
 /* Return 1 if <bug> is a  default buffer */
 static inline int b_is_default(struct buffer *buf)
 {
@@ -78,6 +85,12 @@ static inline int b_is_large(struct buffer *buf)
 	return b_is_large_sz(b_size(buf));
 }
 
+/* Return 1 if <buf> is a small buffer */
+static inline int b_is_small(struct buffer *buf)
+{
+	return b_is_small_sz(b_size(buf));
+}
+
 /**************************************************/
 /* Functions below are used for buffer allocation */
 /**************************************************/
@@ -172,6 +185,8 @@ static inline char *__b_get_emergency_buf(void)
 		 * than the default buffers */				\
 		if (unlikely(b_is_large_sz(sz)))			\
 			pool_free(pool_head_large_buffer, area);	\
+		else if (unlikely(b_is_small_sz(sz)))			\
+			pool_free(pool_head_small_buffer, area);	\
 		else if (th_ctx->emergency_bufs_left < global.tune.reserved_bufs) \
 			th_ctx->emergency_bufs[th_ctx->emergency_bufs_left++] = area; \
 		else							\
@@ -185,6 +200,35 @@ static inline char *__b_get_emergency_buf(void)
 			__b_free((_buf));	\
 	} while (0)
 
+
+static inline struct buffer *b_alloc_small(struct buffer *buf)
+{
+	char *area = NULL;
+
+	if (!buf->size) {
+		area = pool_alloc(pool_head_small_buffer);
+		if (!area)
+			return NULL;
+		buf->area = area;
+		buf->size = global.tune.bufsize_small;
+	}
+	return buf;
+}
+
+static inline struct buffer *b_alloc_large(struct buffer *buf)
+{
+	char *area = NULL;
+
+	if (!buf->size) {
+		area = pool_alloc(pool_head_large_buffer);
+		if (!area)
+			return NULL;
+		buf->area = area;
+		buf->size = global.tune.bufsize_large;
+	}
+	return buf;
+}
+
 /* Offer one or multiple buffer currently belonging to target <from> to whoever
  * needs one. Any pointer is valid for <from>, including NULL. Its purpose is
  * to avoid passing a buffer to oneself in case of failed allocations (e.g.

include/haproxy/http_htx-t.h

@@ -93,4 +93,22 @@ struct http_errors {
 	struct list list;                     /* http-errors list */
 };
 
+/* Indicates the keyword origin of an http-error definition. This is used in
+ * <conf_errors> type to indicate which part of the internal union should be
+ * manipulated.
+ */
+enum http_err_directive {
+	HTTP_ERR_DIRECTIVE_SECTION = 0, /* "errorfiles" keyword referencing a http-errors section */
+	HTTP_ERR_DIRECTIVE_INLINE,      /* "errorfile" keyword with inline error definition */
+};
+
+/* Used with "errorfiles" directives. It indicates for each known HTTP error
+ * status codes if they are defined in the target http-errors section.
+ */
+enum http_err_import {
+	HTTP_ERR_IMPORT_NO = 0,
+	HTTP_ERR_IMPORT_IMPLICIT, /* import every errcode defined in a section */
+	HTTP_ERR_IMPORT_EXPLICIT, /* import a specific errcode from a section */
+};
+
 #endif /* _HAPROXY_HTTP_HTX_T_H */

include/haproxy/http_htx.h

@@ -78,6 +78,7 @@ struct buffer *http_load_errorfile(const char *file, char **errmsg);
 struct buffer *http_load_errormsg(const char *key, const struct ist msg, char **errmsg);
 struct buffer *http_parse_errorfile(int status, const char *file, char **errmsg);
 struct buffer *http_parse_errorloc(int errloc, int status, const char *url, char **errmsg);
+int proxy_check_http_errors(struct proxy *px);
 int proxy_dup_default_conf_errors(struct proxy *curpx, const struct proxy *defpx, char **errmsg);
 void proxy_release_conf_errors(struct proxy *px);
 

include/haproxy/htx.h

@@ -57,6 +57,16 @@ size_t htx_add_data(struct htx *htx, const struct ist data);
 struct htx_blk *htx_add_last_data(struct htx *htx, struct ist data);
 void htx_move_blk_before(struct htx *htx, struct htx_blk **blk, struct htx_blk **ref);
 int htx_append_msg(struct htx *dst, const struct htx *src);
+struct buffer *htx_move_to_small_buffer(struct buffer *dst, struct buffer *src);
+struct buffer *htx_move_to_large_buffer(struct buffer *dst, struct buffer *src);
+struct buffer *htx_copy_to_small_buffer(struct buffer *dst, struct buffer *src);
+struct buffer *htx_copy_to_large_buffer(struct buffer *dst, struct buffer *src);
+
+#define HTX_XFER_DEFAULT           0x00000000 /* Default XFER: no partial xfer / remove blocks from source */
+#define HTX_XFER_KEEP_SRC_BLKS     0x00000001 /* Don't remove xfer blocks from source messages during xfer */
+#define HTX_XFER_PARTIAL_HDRS_COPY 0x00000002 /* Allow partial copy of headers and trailers part */
+#define HTX_XFER_HDRS_ONLY         0x00000003 /* Only Transfert header blocks (start-line, header and EOH) */
+size_t htx_xfer(struct htx *dst, struct htx *src, size_t count, unsigned int flags);
 
 /* Functions and macros to get parts of the start-line or length of these
  * parts. Request and response start-lines are both composed of 3 parts.

include/haproxy/protocol.h

@@ -124,6 +124,12 @@ static inline int real_family(int ss_family)
 	return fam ? fam->real_family : AF_UNSPEC;
 }
 
+static inline int proto_is_quic(const struct protocol *proto)
+{
+	return (proto->proto_type == PROTO_TYPE_DGRAM &&
+		proto->xprt_type == PROTO_TYPE_STREAM);
+}
+
 #endif /* _HAPROXY_PROTOCOL_H */
 
 /*

include/haproxy/proxy-t.h

@@ -156,14 +156,17 @@ enum PR_SRV_STATE_FILE {
 #define PR_O2_RSTRICT_REQ_HDR_NAMES_NOOP 0x01000000 /* preserve request header names containing chars outside of [0-9a-zA-Z-] charset */
 #define PR_O2_RSTRICT_REQ_HDR_NAMES_MASK 0x01c00000 /* mask for restrict-http-header-names option */
 
-/* unused : 0x02000000 ... 0x08000000 */
-
 /* server health checks */
-#define PR_O2_CHK_NONE  0x00000000      /* no L7 health checks configured (TCP by default) */
+#define PR_O2_CHK_NONE   0x00000000    /* no L7 health checks configured (TCP by default) */
-#define PR_O2_TCPCHK_CHK 0x90000000     /* use TCPCHK check for server health */
+#define PR_O2_TCPCHK_CHK 0x02000000    /* use TCPCHK check for server health */
-#define PR_O2_EXT_CHK   0xA0000000      /* use external command for server health */
+#define PR_O2_EXT_CHK    0x04000000    /* use external command for server health */
-/* unused: 0xB0000000 to 0xF000000, reserved for health checks */
+#define PR_O2_CHK_ANY    0x06000000    /* Mask to cover any check */
-#define PR_O2_CHK_ANY   0xF0000000      /* Mask to cover any check */
+
+#define PR_O2_USE_SBUF_QUEUE    0x08000000 /* use small buffer for request when stream are queued*/
+#define PR_O2_USE_SBUF_L7_RETRY 0x10000000 /* use small buffer for request when L7 retires are enabled */
+#define PR_O2_USE_SBUF_CHECK    0x20000000 /* use small buffer for request's healthchecks */
+#define PR_O2_USE_SBUF_ALL      0x38000000 /* all flags for use-large-buffer option */
+/* unused : 0x40000000 ... 0x80000000 */
 /* end of proxy->options2 */
 
 /* bits for proxy->options3 */

include/haproxy/task-t.h

@@ -130,20 +130,22 @@ struct notification {
  * on return.
  */
 #define TASK_COMMON							\
-	struct {							\
+	unsigned int state; /* task state : bitfield of TASK_	*/ \
-		unsigned int state; /* task state : bitfield of TASK_	*/ \
+	int tid;            /* tid of task/tasklet. <0 = local for tasklet, unbound for task */ \
-		int tid;            /* tid of task/tasklet. <0 = local for tasklet, unbound for task */ \
+	struct task *(*process)(struct task *t, void *ctx, unsigned int state); /* the function which processes the task */ \
-		struct task *(*process)(struct task *t, void *ctx, unsigned int state); /* the function which processes the task */ \
+	void *context; /* the task's context */			\
-		void *context; /* the task's context */			\
+	const struct ha_caller *caller;	 /* call place of last wakeup(); 0 on init, -1 on free */ \
-		const struct ha_caller *caller;	 /* call place of last wakeup(); 0 on init, -1 on free */ \
+	uint32_t wake_date;              /* date of the last task wakeup */ \
-		uint32_t wake_date;              /* date of the last task wakeup */ \
+	unsigned int calls;              /* number of times process was called */ \
-		unsigned int calls;              /* number of times process was called */ \
+	TASK_DEBUG_STORAGE;					\
-		TASK_DEBUG_STORAGE;					\
+	short last_run;                  /* 16-bit now_ms of last run */
-	}
+	/* a 16- or 48-bit hole remains here and is used by task */
 
 /* The base for all tasks */
 struct task {
 	TASK_COMMON;			/* must be at the beginning! */
+	short nice;                     /* task prio from -1024 to +1024 */
+	int expire;			/* next expiration date for this task, in ticks */
 	struct eb32_node rq;		/* ebtree node used to hold the task in the run queue */
 	/* WARNING: the struct task is often aliased as a struct tasklet when
 	 * it is NOT in the run queue. The tasklet has its struct list here
@@ -151,14 +153,12 @@ struct task {
 	 * ever reorder these fields without taking this into account!
 	 */
 	struct eb32_node wq;		/* ebtree node used to hold the task in the wait queue */
-	int expire;			/* next expiration date for this task, in ticks */
-	short nice;                     /* task prio from -1024 to +1024 */
-	/* 16-bit hole here */
 };
 
 /* lightweight tasks, without priority, mainly used for I/Os */
 struct tasklet {
 	TASK_COMMON;			/* must be at the beginning! */
+	/* 48-bit hole here */
 	struct list list;
 	/* WARNING: the struct task is often aliased as a struct tasklet when
 	 * it is not in the run queue. The task has its struct rq here where

include/haproxy/tcpcheck-t.h

@@ -121,6 +121,7 @@ enum tcpcheck_rule_type {
 /* Unused 0x000000A0..0x00000FF0 (reserved for future proto) */
 #define TCPCHK_RULES_TCP_CHK     0x00000FF0
 #define TCPCHK_RULES_PROTO_CHK   0x00000FF0 /* Mask to cover protocol check */
+#define TCPCHK_RULES_MAY_USE_SBUF 0x00001000 /* checks may try to use small buffers if possible for the request */
 
 struct check;
 struct tcpcheck_connect {

src/acme.c

@@ -15,6 +15,7 @@
 #include <haproxy/acme-t.h>
 
 #include <haproxy/base64.h>
+#include <haproxy/intops.h>
 #include <haproxy/cfgparse.h>
 #include <haproxy/cli.h>
 #include <haproxy/errors.h>
@@ -266,7 +267,6 @@ static int cfg_parse_acme(const char *file, int linenum, char **args, int kwm)
 	mark_tainted(TAINTED_CONFIG_EXP_KW_DECLARED);
 
 	if (strcmp(args[0], "acme") == 0) {
-		struct acme_cfg *tmp_acme = acme_cfgs;
 
 		if (alertif_too_many_args(1, file, linenum, args, &err_code))
 			goto out;
@@ -292,7 +292,7 @@ static int cfg_parse_acme(const char *file, int linenum, char **args, int kwm)
 			 * name */
 			err_code |= ERR_ALERT | ERR_FATAL;
 			ha_alert("parsing [%s:%d]: acme section '%s' already exists (%s:%d).\n",
-					file, linenum, args[1], tmp_acme->filename, tmp_acme->linenum);
+			         file, linenum, args[1], cur_acme->filename, cur_acme->linenum);
 			goto out;
 		}
 
@@ -1188,7 +1188,7 @@ int acme_res_certificate(struct task *task, struct acme_ctx *ctx, char **errmsg)
 		}
 		/* get the next retry timing */
 		if (isteqi(hdr->n, ist("Retry-After"))) {
-			ctx->retryafter = atol(hdr->v.ptr);
+			ctx->retryafter = __strl2uic(hdr->v.ptr, hdr->v.len);
 		}
 	}
 
@@ -1261,7 +1261,7 @@ int acme_res_chkorder(struct task *task, struct acme_ctx *ctx, char **errmsg)
 		}
 		/* get the next retry timing */
 		if (isteqi(hdr->n, ist("Retry-After"))) {
-			ctx->retryafter = atol(hdr->v.ptr);
+			ctx->retryafter = __strl2uic(hdr->v.ptr, hdr->v.len);
 		}
 	}
 
@@ -1344,7 +1344,6 @@ int acme_req_finalize(struct task *task, struct acme_ctx *ctx, char **errmsg)
 	csr->data = ret;
 
 	chunk_printf(req_in, "{ \"csr\": \"%.*s\" }", (int)csr->data, csr->area);
-	OPENSSL_free(data);
 
 
 	if (acme_jws_payload(req_in, ctx->nonce, ctx->finalize, ctx->cfg->account.pkey, ctx->kid, req_out, errmsg) != 0)
@@ -1358,6 +1357,7 @@ int acme_req_finalize(struct task *task, struct acme_ctx *ctx, char **errmsg)
 error:
 	memprintf(errmsg, "couldn't request the finalize URL");
 out:
+	OPENSSL_free(data);
 	free_trash_chunk(req_in);
 	free_trash_chunk(req_out);
 	free_trash_chunk(csr);
@@ -1391,7 +1391,7 @@ int acme_res_finalize(struct task *task, struct acme_ctx *ctx, char **errmsg)
 		}
 		/* get the next retry timing */
 		if (isteqi(hdr->n, ist("Retry-After"))) {
-			ctx->retryafter = atol(hdr->v.ptr);
+			ctx->retryafter = __strl2uic(hdr->v.ptr, hdr->v.len);
 		}
 	}
 
@@ -1492,7 +1492,7 @@ enum acme_ret acme_res_challenge(struct task *task, struct acme_ctx *ctx, struct
 		}
 		/* get the next retry timing */
 		if (isteqi(hdr->n, ist("Retry-After"))) {
-			ctx->retryafter = atol(hdr->v.ptr);
+			ctx->retryafter = __strl2uic(hdr->v.ptr, hdr->v.len);
 		}
 	}
 
@@ -1618,7 +1618,7 @@ int acme_res_auth(struct task *task, struct acme_ctx *ctx, struct acme_auth *aut
 		}
 		/* get the next retry timing */
 		if (isteqi(hdr->n, ist("Retry-After"))) {
-			ctx->retryafter = atol(hdr->v.ptr);
+			ctx->retryafter = __strl2uic(hdr->v.ptr, hdr->v.len);
 		}
 
 	}
@@ -1654,6 +1654,19 @@ int acme_res_auth(struct task *task, struct acme_ctx *ctx, struct acme_auth *aut
 
 	auth->dns = istdup(ist2(t2->area, t2->data));
 
+	ret = mjson_get_string(hc->res.buf.area, hc->res.buf.data, "$.status", trash.area, trash.size);
+	if (ret == -1) {
+		memprintf(errmsg, "couldn't get a \"status\" from Authorization URL \"%s\"", auth->auth.ptr);
+		goto error;
+	}
+	trash.data = ret;
+
+	/* if auth is already valid we need to skip solving challenges */
+	if (strncasecmp("valid", trash.area, trash.data) == 0) {
+		auth->validated = 1;
+		goto out;
+	}
+
 	/* get the multiple challenges and select the one from the configuration */
 	for (i = 0; ; i++) {
 		int ret;
@@ -1761,6 +1774,7 @@ int acme_res_auth(struct task *task, struct acme_ctx *ctx, struct acme_auth *aut
 		break;
 	}
 
+out:
 	ret = 0;
 
 error:
@@ -1849,7 +1863,7 @@ int acme_res_neworder(struct task *task, struct acme_ctx *ctx, char **errmsg)
 		}
 		/* get the next retry timing */
 		if (isteqi(hdr->n, ist("Retry-After"))) {
-			ctx->retryafter = atol(hdr->v.ptr);
+			ctx->retryafter = __strl2uic(hdr->v.ptr, hdr->v.len);
 		}
 		/* get the order URL */
 		if (isteqi(hdr->n, ist("Location"))) {
@@ -2009,7 +2023,7 @@ int acme_res_account(struct task *task, struct acme_ctx *ctx, int newaccount, ch
 		}
 		/* get the next retry timing */
 		if (isteqi(hdr->n, ist("Retry-After"))) {
-			ctx->retryafter = atol(hdr->v.ptr);
+			ctx->retryafter = __strl2uic(hdr->v.ptr, hdr->v.len);
 		}
 		if (isteqi(hdr->n, ist("Replay-Nonce"))) {
 			istfree(&ctx->nonce);
@@ -2263,6 +2277,14 @@ re:
 		break;
 		case ACME_CHALLENGE:
 			if (http_st == ACME_HTTP_REQ) {
+				/* if challenge is already validated we skip this stage */
+				if (ctx->next_auth->validated) {
+					if ((ctx->next_auth = ctx->next_auth->next) == NULL) {
+						st = ACME_CHKCHALLENGE;
+						ctx->next_auth = ctx->auths;
+					}
+					goto nextreq;
+				}
 
 				/* if the challenge is not ready, wait to be wakeup */
 				if (!ctx->next_auth->ready)
@@ -2292,6 +2314,14 @@ re:
 		break;
 		case ACME_CHKCHALLENGE:
 			if (http_st == ACME_HTTP_REQ) {
+				/* if challenge is already validated we skip this stage */
+				if (ctx->next_auth->validated) {
+					if ((ctx->next_auth = ctx->next_auth->next) == NULL)
+						st = ACME_FINALIZE;
+
+					goto nextreq;
+				}
+
 				if (acme_post_as_get(task, ctx, ctx->next_auth->chall, &errmsg) != 0)
 					goto retry;
 			}
@@ -2526,9 +2556,9 @@ X509_REQ *acme_x509_req(EVP_PKEY *pkey, char **san)
 {
 	struct buffer *san_trash = NULL;
 	X509_REQ *x = NULL;
-	X509_NAME *nm;
+	X509_NAME *nm = NULL;
 	STACK_OF(X509_EXTENSION) *exts = NULL;
-	X509_EXTENSION *ext_san;
+	X509_EXTENSION *ext_san = NULL;
 	char *str_san = NULL;
 	int i = 0;
 
@@ -2559,26 +2589,36 @@ X509_REQ *acme_x509_req(EVP_PKEY *pkey, char **san)
 	for (i = 0; san[i]; i++) {
 		chunk_appendf(san_trash, "%sDNS:%s", i ? "," : "", san[i]);
 	}
-	str_san = my_strndup(san_trash->area, san_trash->data);
+	if ((str_san = my_strndup(san_trash->area, san_trash->data)) == NULL)
+		goto error;
 
 	if ((ext_san = X509V3_EXT_conf_nid(NULL, NULL, NID_subject_alt_name, str_san)) == NULL)
 		goto error;
 
 	if (!sk_X509_EXTENSION_push(exts, ext_san))
 		goto error;
+
+	ext_san = NULL; /* handle double-free upon error */
+
 	if (!X509_REQ_add_extensions(x, exts))
 		goto error;
 
-	sk_X509_EXTENSION_pop_free(exts, X509_EXTENSION_free);
-
 	if (!X509_REQ_sign(x, pkey, EVP_sha256()))
 		goto error;
 
+	sk_X509_EXTENSION_pop_free(exts, X509_EXTENSION_free);
+	X509_NAME_free(nm);
+	free(str_san);
 	free_trash_chunk(san_trash);
 
 	return x;
 
 error:
+	X509_EXTENSION_free(ext_san);
+	sk_X509_EXTENSION_pop_free(exts, X509_EXTENSION_free);
+	X509_REQ_free(x);
+	X509_NAME_free(nm);
+	free(str_san);
 	free_trash_chunk(san_trash);
 	return NULL;
 
@@ -2627,7 +2667,7 @@ EVP_PKEY *acme_gen_tmp_pkey()
 /* start an ACME task */
 static int acme_start_task(struct ckch_store *store, char **errmsg)
 {
-	struct task *task;
+	struct task *task = NULL;
 	struct acme_ctx *ctx = NULL;
 	struct acme_cfg *cfg;
 	struct ckch_store *newstore = NULL;
@@ -2712,6 +2752,8 @@ err:
 		HA_RWLOCK_WRUNLOCK(OTHER_LOCK, &acme_lock);
 		acme_ctx_destroy(ctx);
 	}
+	if (task)
+		task_destroy(task);
 	memprintf(errmsg, "%sCan't start the ACME client.", *errmsg ? *errmsg : "");
 	return 1;
 }
@@ -2721,7 +2763,10 @@ static int cli_acme_renew_parse(char **args, char *payload, struct appctx *appct
 	struct ckch_store *store = NULL;
 	char *errmsg = NULL;
 
-	if (!*args[1]) {
+	if (!cli_has_level(appctx, ACCESS_LVL_ADMIN))
+		return 1;
+
+	if (!*args[2]) {
 		memprintf(&errmsg, ": not enough parameters\n");
 		goto err;
 	}
@@ -2760,8 +2805,11 @@ static int cli_acme_chall_ready_parse(char **args, char *payload, struct appctx
 	int remain = 0;
 	struct ebmb_node *node = NULL;
 
-	if (!*args[2] && !*args[3] && !*args[4]) {
+	if (!cli_has_level(appctx, ACCESS_LVL_ADMIN))
-		memprintf(&msg, ": not enough parameters\n");
+		return 1;
+
+	if (!*args[2] || !*args[3] || !*args[4]) {
+		memprintf(&msg, "Not enough parameters: \"acme challenge_ready <certfile> domain <domain>\"\n");
 		goto err;
 	}
 
@@ -2882,8 +2930,12 @@ end:
 	return 1;
 }
 
-static int cli_acme_ps(char **args, char *payload, struct appctx *appctx, void *private)
+static int cli_acme_parse_status(char **args, char *payload, struct appctx *appctx, void *private)
 {
+
+	if (!cli_has_level(appctx, ACCESS_LVL_ADMIN))
+		return 1;
+
 	return 0;
 }
 
@@ -2891,7 +2943,7 @@ static int cli_acme_ps(char **args, char *payload, struct appctx *appctx, void *
 
 static struct cli_kw_list cli_kws = {{ },{
 	{ { "acme", "renew", NULL },           "acme renew <certfile>                   : renew a certificate using the ACME protocol", cli_acme_renew_parse, NULL, NULL, NULL, 0 },
-	{ { "acme", "status", NULL },          "acme status                             : show status of certificates configured with ACME", cli_acme_ps, cli_acme_status_io_handler, NULL, NULL, 0 },
+	{ { "acme", "status", NULL },          "acme status                             : show status of certificates configured with ACME", cli_acme_parse_status, cli_acme_status_io_handler, NULL, NULL, 0 },
 	{ { "acme", "challenge_ready", NULL }, "acme challenge_ready <certfile> domain <domain> : notify HAProxy that the ACME challenge is ready", cli_acme_chall_ready_parse, NULL, NULL, NULL, 0 },
 	{ { NULL }, NULL, NULL, NULL }
 }};

src/applet.c

@@ -511,7 +511,7 @@ size_t appctx_htx_rcv_buf(struct appctx *appctx, struct buffer *buf, size_t coun
 		goto out;
 	}
 
-	htx_xfer_blks(buf_htx, appctx_htx, count, HTX_BLK_UNUSED);
+	htx_xfer(buf_htx, appctx_htx, count, HTX_XFER_DEFAULT);
 	buf_htx->flags |= (appctx_htx->flags & (HTX_FL_PARSING_ERROR|HTX_FL_PROCESSING_ERROR));
 	if (htx_is_empty(appctx_htx)) {
 		buf_htx->flags |= (appctx_htx->flags & HTX_FL_EOM);
@@ -608,7 +608,7 @@ size_t appctx_htx_snd_buf(struct appctx *appctx, struct buffer *buf, size_t coun
 		goto end;
 	}
 
-	htx_xfer_blks(appctx_htx, buf_htx, count, HTX_BLK_UNUSED);
+	htx_xfer(appctx_htx, buf_htx, count, HTX_XFER_DEFAULT);
 	if (htx_is_empty(buf_htx)) {
 		appctx_htx->flags |= (buf_htx->flags & HTX_FL_EOM);
 	}

src/cfgparse-listen.c

@@ -1358,14 +1358,15 @@ int cfg_parse_listen(const char *file, int linenum, char **args, int kwm)
 			goto out;
 		}
 
-		err_code |= warnif_misplaced_http_req(curproxy, file, linenum, args[0], NULL);
+		if (warnif_misplaced_http_req(curproxy, file, linenum, args[0], NULL))
+			err_code |= ERR_WARN;
 
 		if (curproxy->cap & PR_CAP_FE)
 			where |= SMP_VAL_FE_HRQ_HDR;
 		if (curproxy->cap & PR_CAP_BE)
 			where |= SMP_VAL_BE_HRQ_HDR;
 		err_code |= warnif_cond_conflicts(rule->cond, where, &errmsg);
-		if (err_code)
+		if (errmsg && *errmsg)
 			ha_warning("parsing [%s:%d] : '%s.\n'", file, linenum, errmsg);
 
 		LIST_APPEND(&curproxy->http_req_rules, &rule->list);
@@ -1400,7 +1401,7 @@ int cfg_parse_listen(const char *file, int linenum, char **args, int kwm)
 		if (curproxy->cap & PR_CAP_BE)
 			where |= SMP_VAL_BE_HRS_HDR;
 		err_code |= warnif_cond_conflicts(rule->cond, where, &errmsg);
-		if (err_code)
+		if (errmsg && *errmsg)
 			ha_warning("parsing [%s:%d] : '%s.\n'", file, linenum, errmsg);
 
 		LIST_APPEND(&curproxy->http_res_rules, &rule->list);
@@ -1434,7 +1435,7 @@ int cfg_parse_listen(const char *file, int linenum, char **args, int kwm)
 		if (curproxy->cap & PR_CAP_BE)
 			where |= SMP_VAL_BE_HRS_HDR;
 		err_code |= warnif_cond_conflicts(rule->cond, where, &errmsg);
-		if (err_code)
+		if (errmsg && *errmsg)
 			ha_warning("parsing [%s:%d] : '%s.\n'", file, linenum, errmsg);
 
 		LIST_APPEND(&curproxy->http_after_res_rules, &rule->list);
@@ -1491,14 +1492,15 @@ int cfg_parse_listen(const char *file, int linenum, char **args, int kwm)
 		}
 
 		LIST_APPEND(&curproxy->redirect_rules, &rule->list);
-		err_code |= warnif_misplaced_redirect(curproxy, file, linenum, args[0], NULL);
+		if (warnif_misplaced_redirect(curproxy, file, linenum, args[0], NULL))
+			err_code |= ERR_WARN;
 
 		if (curproxy->cap & PR_CAP_FE)
 			where |= SMP_VAL_FE_HRQ_HDR;
 		if (curproxy->cap & PR_CAP_BE)
 			where |= SMP_VAL_BE_HRQ_HDR;
 		err_code |= warnif_cond_conflicts(rule->cond, where, &errmsg);
-		if (err_code)
+		if (errmsg && *errmsg)
 			ha_warning("parsing [%s:%d] : '%s.\n'", file, linenum, errmsg);
 	}
 	else if (strcmp(args[0], "use_backend") == 0) {
@@ -1528,7 +1530,7 @@ int cfg_parse_listen(const char *file, int linenum, char **args, int kwm)
 			}
 
 			err_code |= warnif_cond_conflicts(cond, SMP_VAL_FE_SET_BCK, &errmsg);
-			if (err_code)
+			if (errmsg && *errmsg)
 				ha_warning("parsing [%s:%d] : '%s.\n'", file, linenum, errmsg);
 		}
 		else if (*args[2]) {
@@ -1591,7 +1593,7 @@ int cfg_parse_listen(const char *file, int linenum, char **args, int kwm)
 		}
 
 		err_code |= warnif_cond_conflicts(cond, SMP_VAL_BE_SET_SRV, &errmsg);
-		if (err_code)
+		if (errmsg && *errmsg)
 			ha_warning("parsing [%s:%d] : '%s.\n'", file, linenum, errmsg);
 
 		rule = calloc(1, sizeof(*rule));
@@ -1646,7 +1648,7 @@ int cfg_parse_listen(const char *file, int linenum, char **args, int kwm)
 		 * where force-persist is applied.
 		 */
 		err_code |= warnif_cond_conflicts(cond, SMP_VAL_BE_REQ_CNT, &errmsg);
-		if (err_code)
+		if (errmsg && *errmsg)
 			ha_warning("parsing [%s:%d] : '%s.\n'", file, linenum, errmsg);
 
 		rule = calloc(1, sizeof(*rule));
@@ -1814,7 +1816,7 @@ int cfg_parse_listen(const char *file, int linenum, char **args, int kwm)
 			err_code |= warnif_cond_conflicts(cond, SMP_VAL_BE_STO_RUL, &errmsg);
 		else
 			err_code |= warnif_cond_conflicts(cond, SMP_VAL_BE_SET_SRV, &errmsg);
-		if (err_code)
+		if (errmsg && *errmsg)
 			ha_warning("parsing [%s:%d] : '%s.\n'", file, linenum, errmsg);
 
 		rule = calloc(1, sizeof(*rule));
@@ -1872,7 +1874,7 @@ int cfg_parse_listen(const char *file, int linenum, char **args, int kwm)
 			if (curproxy->cap & PR_CAP_BE)
 				where |= SMP_VAL_BE_HRQ_HDR;
 			err_code |= warnif_cond_conflicts(cond, where, &errmsg);
-			if (err_code)
+			if (errmsg && *errmsg)
 				ha_warning("parsing [%s:%d] : '%s.\n'", file, linenum, errmsg);
 
 			rule = calloc(1, sizeof(*rule));
@@ -1952,7 +1954,7 @@ int cfg_parse_listen(const char *file, int linenum, char **args, int kwm)
 			if (curproxy->cap & PR_CAP_BE)
 				where |= SMP_VAL_BE_HRQ_HDR;
 			err_code |= warnif_cond_conflicts(rule->cond, where, &errmsg);
-			if (err_code)
+			if (errmsg && *errmsg)
 				ha_warning("parsing [%s:%d] : '%s.\n'", file, linenum, errmsg);
 			LIST_APPEND(&curproxy->uri_auth->http_req_rules, &rule->list);
 
@@ -2200,6 +2202,42 @@ stats_error_parsing:
 			err_code |= ERR_ALERT | ERR_FATAL;
 			goto out;
 		}
+		else if (strcmp(args[1], "use-small-buffers") == 0) {
+			unsigned int flags = PR_O2_USE_SBUF_ALL;
+
+			if (warnifnotcap(curproxy, PR_CAP_BE, file, linenum, args[1], NULL)) {
+				err_code |= ERR_WARN;
+				goto out;
+			}
+
+			if (*(args[2])) {
+				int cur_arg;
+
+				flags = 0;
+				for (cur_arg = 2; *(args[cur_arg]); cur_arg++) {
+					if (strcmp(args[cur_arg], "queue") == 0)
+						flags |= PR_O2_USE_SBUF_QUEUE;
+					else if (strcmp(args[cur_arg], "l7-retries") == 0)
+						flags |= PR_O2_USE_SBUF_L7_RETRY;
+					else if (strcmp(args[cur_arg], "check") == 0)
+						flags |= PR_O2_USE_SBUF_CHECK;
+					else {
+						ha_alert("parsing [%s:%d] : invalid parameter '%s'. option '%s' expects 'queue', 'l7-retries' or 'check' value.\n",
+							 file, linenum, args[cur_arg], args[1]);
+						err_code |= ERR_ALERT | ERR_FATAL;
+						goto out;
+					}
+				}
+			}
+			if (kwm == KWM_STD) {
+				curproxy->options2 &= ~PR_O2_USE_SBUF_ALL;
+				curproxy->options2 |= flags;
+			}
+			else if (kwm == KWM_NO) {
+				curproxy->options2 &= ~flags;
+			}
+			goto out;
+		}
 
 		if (kwm != KWM_STD) {
 			ha_alert("parsing [%s:%d]: negation/default is not supported for option '%s'.\n",
@@ -2557,7 +2595,8 @@ stats_error_parsing:
 				goto out;
 			}
 
-			err_code |= warnif_misplaced_monitor(curproxy, file, linenum, args[0], args[1]);
+			if (warnif_misplaced_monitor(curproxy, file, linenum, args[0], args[1]))
+				err_code |= ERR_WARN;
 			if ((cond = build_acl_cond(file, linenum, &curproxy->acl, curproxy, (const char **)args + 2, &errmsg)) == NULL) {
 				ha_alert("parsing [%s:%d] : error detected while parsing a '%s %s' condition : %s.\n",
 					 file, linenum, args[0], args[1], errmsg);

src/cfgparse.c

@@ -63,6 +63,7 @@
 #include <haproxy/global.h>
 #include <haproxy/http_ana.h>
 #include <haproxy/http_rules.h>
+#include <haproxy/http_htx.h>
 #include <haproxy/lb_chash.h>
 #include <haproxy/lb_fas.h>
 #include <haproxy/lb_fwlc.h>
@@ -2318,6 +2319,18 @@ int check_config_validity()
 				   "Please fix either value to remove this warning.\n",
 				   global.tune.bufsize_large, global.tune.bufsize);
 			global.tune.bufsize_large = 0;
+			err_code |= ERR_WARN;
+		}
+	}
+
+	if (global.tune.bufsize_small > 0) {
+		if (global.tune.bufsize_small == global.tune.bufsize)
+			global.tune.bufsize_small = 0;
+		else if (global.tune.bufsize_small > global.tune.bufsize) {
+			ha_warning("invalid small buffer size %d bytes which is greater to default bufsize %d bytes.\n",
+				   global.tune.bufsize_small, global.tune.bufsize);
+			global.tune.bufsize_small = 0;
+			err_code |= ERR_WARN;
 		}
 	}
 
@@ -2392,6 +2405,8 @@ int check_config_validity()
 		else {
 			cfgerr += acl_find_targets(defpx);
 		}
+
+		err_code |= proxy_check_http_errors(defpx);
 	}
 
 	/* starting to initialize the main proxies list */

src/check.c

@@ -1515,13 +1515,15 @@ int check_buf_available(void *target)
 /*
  * Allocate a buffer. If it fails, it adds the check in buffer wait queue.
  */
-struct buffer *check_get_buf(struct check *check, struct buffer *bptr)
+struct buffer *check_get_buf(struct check *check, struct buffer *bptr, unsigned int small_buffer)
 {
 	struct buffer *buf = NULL;
 
-	if (likely(!LIST_INLIST(&check->buf_wait.list)) &&
+	if (small_buffer == 0 || (buf = b_alloc_small(bptr)) == NULL) {
-	    unlikely((buf = b_alloc(bptr, DB_CHANNEL)) == NULL)) {
+		if (likely(!LIST_INLIST(&check->buf_wait.list)) &&
-		b_queue(DB_CHANNEL, &check->buf_wait, check, check_buf_available);
+		    unlikely((buf = b_alloc(bptr, DB_CHANNEL)) == NULL)) {
+			b_queue(DB_CHANNEL, &check->buf_wait, check, check_buf_available);
+		}
 	}
 	return buf;
 }
@@ -1533,8 +1535,11 @@ struct buffer *check_get_buf(struct check *check, struct buffer *bptr)
 void check_release_buf(struct check *check, struct buffer *bptr)
 {
 	if (bptr->size) {
+		int defbuf = b_is_default(bptr);
+
 		b_free(bptr);
-		offer_buffers(check->buf_wait.target, 1);
+		if (defbuf)
+			offer_buffers(check->buf_wait.target, 1);
 	}
 }
 
@@ -1654,7 +1659,6 @@ int start_check_task(struct check *check, int mininter,
  */
 static int start_checks()
 {
-
 	struct proxy *px;
 	struct server *s;
 	char *errmsg = NULL;
@@ -1681,6 +1685,10 @@ static int start_checks()
 	 */
 	for (px = proxies_list; px; px = px->next) {
 		for (s = px->srv; s; s = s->next) {
+			if ((px->options2 & PR_O2_USE_SBUF_CHECK) &&
+			    (s->check.tcpcheck_rules->flags & TCPCHK_RULES_MAY_USE_SBUF))
+				s->check.state |= CHK_ST_USE_SMALL_BUFF;
+
 			if (s->check.state & CHK_ST_CONFIGURED) {
 				nbcheck++;
 				if ((srv_getinter(&s->check) >= SRV_CHK_INTER_THRES) &&
@@ -1805,7 +1813,15 @@ int init_srv_check(struct server *srv)
 		 * specified.
 		 */
 		if (!srv->check.port && !is_addr(&srv->check.addr)) {
-			if (!srv->check.use_ssl && srv->use_ssl != -1)
+			/*
+			 * If any setting is set for the check, then we can't
+			 * assume we'll use the same XPRT as the server, the
+			 * server may be QUIC, but we want a TCP check.
+			 */
+			if (!srv->check.use_ssl && srv->use_ssl != -1 &&
+			    !srv->check.via_socks4 && !srv->check.send_proxy &&
+			    (!srv->check.alpn_len || (srv->check.alpn_len == srv->ssl_ctx.alpn_len && !strncmp(srv->check.alpn_str, srv->ssl_ctx.alpn_str, srv->check.alpn_len))) &&
+			    (!srv->check.mux_proto || srv->check.mux_proto != srv->mux_proto))
 				srv->check.xprt = srv->xprt;
 			else if (srv->check.use_ssl == 1)
 				srv->check.xprt = xprt_get(XPRT_SSL);
@@ -2056,6 +2072,7 @@ static int srv_parse_addr(char **args, int *cur_arg, struct proxy *curpx, struct
 			  char **errmsg)
 {
 	struct sockaddr_storage *sk;
+	struct protocol *proto;
 	int port1, port2, err_code = 0;
 
 
@@ -2064,7 +2081,7 @@ static int srv_parse_addr(char **args, int *cur_arg, struct proxy *curpx, struct
 		goto error;
 	}
 
-	sk = str2sa_range(args[*cur_arg+1], NULL, &port1, &port2, NULL, NULL, NULL, errmsg, NULL, NULL, NULL,
+	sk = str2sa_range(args[*cur_arg+1], NULL, &port1, &port2, NULL, &proto, NULL, errmsg, NULL, NULL, NULL,
 	                  PA_O_RESOLVE | PA_O_PORT_OK | PA_O_STREAM | PA_O_CONNECT);
 	if (!sk) {
 		memprintf(errmsg, "'%s' : %s", args[*cur_arg], *errmsg);
@@ -2072,6 +2089,7 @@ static int srv_parse_addr(char **args, int *cur_arg, struct proxy *curpx, struct
 	}
 
 	srv->check.addr = *sk;
+	srv->check.proto = proto;
 	/* if agentaddr was never set, we can use addr */
 	if (!(srv->flags & SRV_F_AGENTADDR))
 		srv->agent.addr = *sk;
@@ -2101,7 +2119,11 @@ static int srv_parse_agent_addr(char **args, int *cur_arg, struct proxy *curpx,
 		goto error;
 	}
 	set_srv_agent_addr(srv, &sk);
-
+	/* Agent currently only uses TCP */
+	if (sk.ss_family == AF_INET)
+		srv->agent.proto = &proto_tcpv4;
+	else
+		srv->agent.proto = &proto_tcpv6;
   out:
 	return err_code;
 

src/chunk.c

@@ -53,6 +53,22 @@ struct pool_head *pool_head_large_trash __read_mostly = NULL;
 /* this is used to drain data, and as a temporary large buffer */
 THREAD_LOCAL struct buffer trash_large = { };
 
+/* small trash chunks used for various conversions */
+static THREAD_LOCAL struct buffer *small_trash_chunk;
+static THREAD_LOCAL struct buffer small_trash_chunk1;
+static THREAD_LOCAL struct buffer small_trash_chunk2;
+
+/* small trash buffers used for various conversions */
+static int small_trash_size __read_mostly = 0;
+static THREAD_LOCAL char *small_trash_buf1 = NULL;
+static THREAD_LOCAL char *small_trash_buf2 = NULL;
+
+/* the trash pool for reentrant allocations */
+struct pool_head *pool_head_small_trash __read_mostly = NULL;
+
+/* this is used to drain data, and as a temporary small buffer */
+THREAD_LOCAL struct buffer trash_small = { };
+
 /*
 * Returns a pre-allocated and initialized trash chunk that can be used for any
 * type of conversion. Two chunks and their respective buffers are alternatively
@@ -103,14 +119,40 @@ struct buffer *get_large_trash_chunk(void)
 	return large_trash_chunk;
 }
 
+/* Similar to get_trash_chunk() but return a pre-allocated small chunk
+ * instead. Becasuse small buffers are not enabled by default, this function may
+ * return NULL.
+ */
+struct buffer *get_small_trash_chunk(void)
+{
+	char *small_trash_buf;
+
+	if (!small_trash_size)
+		return NULL;
+
+	if (small_trash_chunk == &small_trash_chunk1) {
+		small_trash_chunk = &small_trash_chunk2;
+		small_trash_buf = small_trash_buf2;
+	}
+	else {
+		small_trash_chunk = &small_trash_chunk1;
+		small_trash_buf = small_trash_buf1;
+	}
+	*small_trash_buf = 0;
+	chunk_init(small_trash_chunk, small_trash_buf, small_trash_size);
+	return small_trash_chunk;
+}
+
 /* Returns a trash chunk accordingly to the requested size. This function may
  * fail if the requested size is too big or if the large chubks are not
  * configured.
  */
 struct buffer *get_trash_chunk_sz(size_t size)
 {
-	if (likely(size <= trash_size))
+	if (likely(size > small_trash_size && size <= trash_size))
-	    return get_trash_chunk();
+		return get_trash_chunk();
+	else if (small_trash_size && size <= small_trash_size)
+		return get_small_trash_chunk();
 	else if (large_trash_size && size <= large_trash_size)
 		return get_large_trash_chunk();
 	else
@@ -122,17 +164,20 @@ struct buffer *get_trash_chunk_sz(size_t size)
  */
 struct buffer *get_larger_trash_chunk(struct buffer *chk)
 {
-	struct buffer *chunk;
+	struct buffer *chunk = NULL;
 
-	if (!chk)
+	if (!chk || chk->size == small_trash_size) {
-		return get_trash_chunk();
+		/* no chunk or a small one, use a regular buffer */
+		chunk = get_trash_chunk();
+	}
+	else if (large_trash_size && chk->size <= large_trash_size) {
+		/* a regular byffer, use a large buffer if possible */
+		chunk = get_large_trash_chunk();
+	}
 
-	/* No large buffers or current chunk is alread a large trash chunk */
+	if (chk && chunk)
-	if (!large_trash_size || chk->size == large_trash_size)
+		b_xfer(chunk, chk, b_data(chk));
-		return NULL;
 
-	chunk = get_large_trash_chunk();
-	b_xfer(chunk, chk, b_data(chk));
 	return chunk;
 }
 
@@ -166,9 +211,29 @@ static int alloc_large_trash_buffers(int bufsize)
 	return trash_large.area && large_trash_buf1 && large_trash_buf2;
 }
 
+/* allocates the trash small buffers if necessary. Returns 0 in case of
+ * failure. Unlike alloc_trash_buffers(), It is unexpected to call this function
+ * multiple times. Small buffers are not used during configuration parsing.
+ */
+static int alloc_small_trash_buffers(int bufsize)
+{
+	small_trash_size = bufsize;
+	if (!small_trash_size)
+		return 1;
+
+	BUG_ON(trash_small.area && small_trash_buf1 && small_trash_buf2);
+
+	chunk_init(&trash_small, my_realloc2(trash_small.area, bufsize), bufsize);
+	small_trash_buf1 = (char *)my_realloc2(small_trash_buf1, bufsize);
+	small_trash_buf2 = (char *)my_realloc2(small_trash_buf2, bufsize);
+	return trash_small.area && small_trash_buf1 && small_trash_buf2;
+}
+
 static int alloc_trash_buffers_per_thread()
 {
-	return alloc_trash_buffers(global.tune.bufsize) && alloc_large_trash_buffers(global.tune.bufsize_large);
+	return (alloc_trash_buffers(global.tune.bufsize) &&
+		alloc_large_trash_buffers(global.tune.bufsize_large) &&
+		alloc_small_trash_buffers(global.tune.bufsize_large));
 }
 
 static void free_trash_buffers_per_thread()
@@ -180,6 +245,10 @@ static void free_trash_buffers_per_thread()
 	chunk_destroy(&trash_large);
 	ha_free(&large_trash_buf2);
 	ha_free(&large_trash_buf1);
+
+	chunk_destroy(&trash_small);
+	ha_free(&small_trash_buf2);
+	ha_free(&small_trash_buf1);
 }
 
 /* Initialize the trash buffers. It returns 0 if an error occurred. */
@@ -207,6 +276,14 @@ int init_trash_buffers(int first)
 		if (!pool_head_large_trash)
 			return 0;
 	}
+
+	if (!first && global.tune.bufsize_small) {
+		pool_head_small_trash = create_pool("small_trash",
+						    sizeof(struct buffer) + global.tune.bufsize_small,
+						    MEM_F_EXACT);
+		if (!pool_head_small_trash)
+			return 0;
+	}
 	return 1;
 }
 

src/dynbuf.c

@@ -24,6 +24,7 @@
 
 struct pool_head *pool_head_buffer __read_mostly;
 struct pool_head *pool_head_large_buffer __read_mostly = NULL;
+struct pool_head *pool_head_small_buffer __read_mostly;
 
 /* perform minimal initializations, report 0 in case of error, 1 if OK. */
 int init_buffer()
@@ -43,6 +44,12 @@ int init_buffer()
 			return 0;
 	}
 
+	if (global.tune.bufsize_small) {
+		pool_head_small_buffer = create_aligned_pool("small_buffer", global.tune.bufsize_small, 64, MEM_F_SHARED|MEM_F_EXACT);
+		if (!pool_head_small_buffer)
+			return 0;
+	}
+
 	/* make sure any change to the queues assignment isn't overlooked */
 	BUG_ON(DB_PERMANENT - DB_UNLIKELY - 1 != DYNBUF_NBQ);
 	BUG_ON(DB_MUX_RX_Q  < DB_SE_RX_Q   || DB_MUX_RX_Q  >= DYNBUF_NBQ);

src/ech.c

@@ -136,6 +136,10 @@ static int cli_parse_show_ech(char **args, char *payload,
 {
 	struct show_ech_ctx *ctx = applet_reserve_svcctx(appctx, sizeof(*ctx));
 
+	if (!cli_has_level(appctx, ACCESS_LVL_ADMIN))
+		return 1;
+
+
 	/* no parameter, shows only file list */
 	if (*args[3]) {
 		SSL_CTX *sctx = NULL;
@@ -297,6 +301,9 @@ static int cli_parse_add_ech(char **args, char *payload, struct appctx *appctx,
 	OSSL_ECHSTORE *es = NULL;
 	BIO *es_in = NULL;
 
+	if (!cli_has_level(appctx, ACCESS_LVL_ADMIN))
+		return 1;
+
 	if (!*args[3] || !payload)
 		return cli_err(appctx, "syntax: add ssl ech <name> <PEM file content>");
 	if (cli_find_ech_specific_ctx(args[3], &sctx) != 1)
@@ -324,6 +331,9 @@ static int cli_parse_set_ech(char **args, char *payload, struct appctx *appctx,
 	OSSL_ECHSTORE *es = NULL;
 	BIO *es_in = NULL;
 
+	if (!cli_has_level(appctx, ACCESS_LVL_ADMIN))
+		return 1;
+
 	if (!*args[3] || !payload)
 		return cli_err(appctx, "syntax: set ssl ech <name> <PEM file content>");
 	if (cli_find_ech_specific_ctx(args[3], &sctx) != 1)
@@ -351,6 +361,9 @@ static int cli_parse_del_ech(char **args, char *payload, struct appctx *appctx,
 	char success_message[ECH_SUCCESS_MSG_MAX];
 	OSSL_ECHSTORE *es = NULL;
 
+	if (!cli_has_level(appctx, ACCESS_LVL_ADMIN))
+		return 1;
+
 	if (!*args[3])
 		return cli_err(appctx, "syntax: del ssl ech <name>");
 	if (*args[4])

src/flt_spoe.c

@@ -1114,7 +1114,7 @@ static int spoe_process_event(struct stream *s, struct spoe_context *ctx,
 	}
 	else if (ret == 0) {
 		if ((s->scf->flags & SC_FL_ERROR) ||
-		    ((s->scf->flags & (SC_FL_EOS|SC_FL_ABRT_DONE)) && proxy_abrt_close_def(s->be, 1))) {
+		    ((s->scf->flags & SC_FL_EOS) && proxy_abrt_close_def(s->be, 1))) {
 			ctx->status_code = SPOE_CTX_ERR_INTERRUPT;
 			spoe_stop_processing(agent, ctx);
 			spoe_handle_processing_error(s, agent, ctx, dir);

src/http_ana.c

@@ -2828,8 +2828,7 @@ static enum rule_result http_req_get_intercept_rule(struct proxy *px, struct lis
 	int act_opts = 0;
 
 	if ((s->scf->flags & SC_FL_ERROR) ||
-	    ((s->scf->flags & (SC_FL_EOS|SC_FL_ABRT_DONE)) &&
+	    ((s->scf->flags & SC_FL_EOS) && proxy_abrt_close_def(px, 1)))
-	     proxy_abrt_close_def(px, 1)))
 		act_opts |= ACT_OPT_FINAL | ACT_OPT_FINAL_EARLY;
 
 	/* If "the current_rule_list" match the executed rule list, we are in
@@ -3020,8 +3019,7 @@ static enum rule_result http_res_get_intercept_rule(struct proxy *px, struct lis
 	if (final)
 		act_opts |= ACT_OPT_FINAL;
 	if ((s->scf->flags & SC_FL_ERROR) ||
-	    ((s->scf->flags & (SC_FL_EOS|SC_FL_ABRT_DONE)) &&
+	    ((s->scf->flags & SC_FL_EOS) && proxy_abrt_close_def(px, 1)))
-	     proxy_abrt_close_def(px, 1)))
 		act_opts |= ACT_OPT_FINAL | ACT_OPT_FINAL_EARLY;
 
 	/* If "the current_rule_list" match the executed rule list, we are in
@@ -4337,20 +4335,10 @@ enum rule_result http_wait_for_msg_body(struct stream *s, struct channel *chn,
 	}
 
 	if (channel_htx_full(chn, htx, global.tune.maxrewrite) || sc_waiting_room(chn_prod(chn))) {
-		struct buffer lbuf;
+		struct buffer lbuf = BUF_NULL;
-		char *area;
 
-		if (large_buffer == 0 || b_is_large(&chn->buf))
+		if (large_buffer == 0 || b_is_large(&chn->buf) || !htx_move_to_large_buffer(&lbuf, &chn->buf))
-			goto end; /* don't use large buffer or large buffer is full */
+			goto end; /* don't use large buffer or already a large buffer */
-
-		/* normal buffer is full, allocate a large one
-		 */
-		area = pool_alloc(pool_head_large_buffer);
-		if (!area)
-			goto end; /* Allocation failure: TODO must be improved to use buffer_wait */
-		lbuf = b_make(area, global.tune.bufsize_large, 0, 0);
-		htx_xfer_blks(htx_from_buf(&lbuf), htx, htx_used_space(htx), HTX_BLK_UNUSED);
-		htx_to_buf(htx, &chn->buf);
 		b_free(&chn->buf);
 		offer_buffers(s, 1);
 		chn->buf = lbuf;
@@ -4366,8 +4354,7 @@ enum rule_result http_wait_for_msg_body(struct stream *s, struct channel *chn,
 	/* we get here if we need to wait for more data */
 
 	if ((s->scf->flags & SC_FL_ERROR) ||
-	    ((s->scf->flags & (SC_FL_EOS|SC_FL_ABRT_DONE)) &&
+	    ((s->scf->flags & SC_FL_EOS) && proxy_abrt_close_def(s->be, 1)))
-	     proxy_abrt_close_def(s->be, 1)))
 		ret = HTTP_RULE_RES_CONT;
 	else if (!(chn_prod(chn)->flags & (SC_FL_ERROR|SC_FL_EOS|SC_FL_ABRT_DONE))) {
 		if (!tick_isset(chn->analyse_exp))

src/http_client.c

@@ -604,10 +604,7 @@ void httpclient_applet_io_handler(struct appctx *appctx)
 						htx_to_buf(htx, outbuf);
 						b_xfer(outbuf, &hc->req.buf, b_data(&hc->req.buf));
 					} else {
-						struct htx_ret ret;
+						if (!htx_xfer(htx, hc_htx, htx_used_space(hc_htx), HTX_XFER_DEFAULT)) {
-
-						ret = htx_xfer_blks(htx, hc_htx, htx_used_space(hc_htx), HTX_BLK_UNUSED);
-						if (!ret.ret) {
 							applet_have_more_data(appctx);
 							goto out;
 						}
@@ -711,7 +708,6 @@ void httpclient_applet_io_handler(struct appctx *appctx)
 				if (hc->options & HTTPCLIENT_O_RES_HTX) {
 					/* HTX mode transfers the header to the hc buffer */
 					struct htx *hc_htx;
-					struct htx_ret ret;
 
 					if (!b_alloc(&hc->res.buf, DB_MUX_TX)) {
 						applet_wont_consume(appctx);
@@ -720,8 +716,7 @@ void httpclient_applet_io_handler(struct appctx *appctx)
 					hc_htx = htxbuf(&hc->res.buf);
 
 					/* xfer the headers */
-					ret = htx_xfer_blks(hc_htx, htx, htx_used_space(htx), HTX_BLK_EOH);
+					if (!htx_xfer(hc_htx, htx, htx_used_space(htx), HTX_XFER_HDRS_ONLY)) {
-					if (!ret.ret) {
 						applet_need_more_data(appctx);
 						goto out;
 					}
@@ -811,12 +806,10 @@ void httpclient_applet_io_handler(struct appctx *appctx)
 				if (hc->options & HTTPCLIENT_O_RES_HTX) {
 					/* HTX mode transfers the header to the hc buffer */
 					struct htx *hc_htx;
-					struct htx_ret ret;
 
 					hc_htx = htxbuf(&hc->res.buf);
 
-					ret = htx_xfer_blks(hc_htx, htx, htx_used_space(htx), HTX_BLK_UNUSED);
+					if (!htx_xfer(hc_htx, htx, htx_used_space(htx), HTX_XFER_DEFAULT))
-					if (!ret.ret)
 						applet_wont_consume(appctx);
 					else
 						applet_fl_clr(appctx, APPCTX_FL_INBLK_FULL);

src/http_htx.c

@@ -41,17 +41,18 @@ struct list http_replies_list = LIST_HEAD_INIT(http_replies_list);
 /* The declaration of an errorfiles/errorfile directives. Used during config
  * parsing only. */
 struct conf_errors {
-	char type;                                  /* directive type (0: errorfiles, 1: errorfile) */
+	enum http_err_directive directive;          /* directive type: inline (errorfile <code> <file>) / section (errorfiles <section>) */
 	union {
 		struct {
 			int status;                 /* the status code associated to this error */
 			struct http_reply *reply;   /* the http reply for the errorfile */
-		} errorfile;                        /* describe an "errorfile" directive */
+		} inl;                              /* for HTTP_ERR_DIRECTIVE_INLINE only */
 		struct {
 			char *name;                 /* the http-errors section name */
-			char status[HTTP_ERR_SIZE]; /* list of status to import (0: ignore, 1: implicit import, 2: explicit import) */
+			struct http_errors *resolved; /* resolved section pointer set via proxy_check_http_errors() */
-		} errorfiles;                       /* describe an "errorfiles" directive */
+			enum http_err_import status[HTTP_ERR_SIZE]; /* list of status to import */
-	} info;
+		} section;                          /* for HTTP_ERR_DIRECTIVE_SECTION only */
+	} type;
 
 	char *file;                                 /* file where the directive appears */
 	int line;                                   /* line where the directive appears */
@@ -2034,9 +2035,9 @@ static int proxy_parse_errorloc(char **args, int section, struct proxy *curpx,
 		ret = -1;
 		goto out;
 	}
-	conf_err->type = 1;
+	conf_err->directive = HTTP_ERR_DIRECTIVE_INLINE;
-	conf_err->info.errorfile.status = status;
+	conf_err->type.inl.status = status;
-	conf_err->info.errorfile.reply = reply;
+	conf_err->type.inl.reply = reply;
 
 	conf_err->file = strdup(file);
 	conf_err->line = line;
@@ -2105,9 +2106,9 @@ static int proxy_parse_errorfile(char **args, int section, struct proxy *curpx,
 		ret = -1;
 		goto out;
 	}
-	conf_err->type = 1;
+	conf_err->directive = HTTP_ERR_DIRECTIVE_INLINE;
-	conf_err->info.errorfile.status = status;
+	conf_err->type.inl.status = status;
-	conf_err->info.errorfile.reply = reply;
+	conf_err->type.inl.reply = reply;
 	conf_err->file = strdup(file);
 	conf_err->line = line;
 	LIST_APPEND(&curpx->conf.errors, &conf_err->list);
@@ -2146,12 +2147,12 @@ static int proxy_parse_errorfiles(char **args, int section, struct proxy *curpx,
 		memprintf(err, "%s : out of memory.", args[0]);
 		goto error;
 	}
-	conf_err->type = 0;
 
-	conf_err->info.errorfiles.name = name;
+	conf_err->directive = HTTP_ERR_DIRECTIVE_SECTION;
+	conf_err->type.section.name = name;
 	if (!*(args[2])) {
 		for (rc = 0; rc < HTTP_ERR_SIZE; rc++)
-			conf_err->info.errorfiles.status[rc] = 1;
+			conf_err->type.section.status[rc] = HTTP_ERR_IMPORT_IMPLICIT;
 	}
 	else {
 		int cur_arg, status;
@@ -2160,7 +2161,7 @@ static int proxy_parse_errorfiles(char **args, int section, struct proxy *curpx,
 
 			for (rc = 0; rc < HTTP_ERR_SIZE; rc++) {
 				if (http_err_codes[rc] == status) {
-					conf_err->info.errorfiles.status[rc] = 2;
+					conf_err->type.section.status[rc] = HTTP_ERR_IMPORT_EXPLICIT;
 					break;
 				}
 			}
@@ -2231,16 +2232,16 @@ static int proxy_parse_http_error(char **args, int section, struct proxy *curpx,
 	if (reply->type == HTTP_REPLY_ERRFILES) {
 		int rc = http_get_status_idx(reply->status);
 
-		conf_err->type = 2;
+		conf_err->directive = HTTP_ERR_DIRECTIVE_SECTION;
-		conf_err->info.errorfiles.name = reply->body.http_errors;
+		conf_err->type.section.name = reply->body.http_errors;
-		conf_err->info.errorfiles.status[rc] = 2;
+		conf_err->type.section.status[rc] = HTTP_ERR_IMPORT_EXPLICIT;
 		reply->body.http_errors = NULL;
 		release_http_reply(reply);
 	}
 	else {
-		conf_err->type = 1;
+		conf_err->directive = HTTP_ERR_DIRECTIVE_INLINE;
-		conf_err->info.errorfile.status = reply->status;
+		conf_err->type.inl.status = reply->status;
-		conf_err->info.errorfile.reply = reply;
+		conf_err->type.inl.reply = reply;
 		LIST_APPEND(&http_replies_list, &reply->list);
 	}
 	conf_err->file = strdup(file);
@@ -2260,60 +2261,46 @@ static int proxy_parse_http_error(char **args, int section, struct proxy *curpx,
 
 }
 
-/* Check "errorfiles" proxy keyword */
+/* Converts <conf_errors> initialized during config parsing for <px> proxy.
-static int proxy_check_errors(struct proxy *px)
+ * Each one of them is transfromed in a http_reply type, stored in proxy
+ * replies array member. The original <conf_errors> becomes unneeded and is
+ * thus removed and freed.
+ */
+static int proxy_finalize_http_errors(struct proxy *px)
 {
 	struct conf_errors *conf_err, *conf_err_back;
 	struct http_errors *http_errs;
-	int rc, err = ERR_NONE;
+	int rc;
 
 	list_for_each_entry_safe(conf_err, conf_err_back, &px->conf.errors, list) {
-		if (conf_err->type == 1) {
+		switch (conf_err->directive) {
-			/* errorfile */
+		case HTTP_ERR_DIRECTIVE_INLINE:
-			rc = http_get_status_idx(conf_err->info.errorfile.status);
+			rc = http_get_status_idx(conf_err->type.inl.status);
-			px->replies[rc] = conf_err->info.errorfile.reply;
+			px->replies[rc] = conf_err->type.inl.reply;
 
 			/* For proxy, to rely on default replies, just don't reference a reply */
 			if (px->replies[rc]->type == HTTP_REPLY_ERRMSG && !px->replies[rc]->body.errmsg)
 				px->replies[rc] = NULL;
-		}
+			break;
-		else {
-			/* errorfiles */
-			list_for_each_entry(http_errs, &http_errors_list, list) {
-				if (strcmp(http_errs->id, conf_err->info.errorfiles.name) == 0)
-					break;
-			}
 
-			/* unknown http-errors section */
+		case HTTP_ERR_DIRECTIVE_SECTION:
-			if (&http_errs->list == &http_errors_list) {
+			http_errs = conf_err->type.section.resolved;
-				ha_alert("proxy '%s': unknown http-errors section '%s' (at %s:%d).\n",
+			if (http_errs) {
-					 px->id, conf_err->info.errorfiles.name, conf_err->file, conf_err->line);
+				for (rc = 0; rc < HTTP_ERR_SIZE; rc++) {
-				err |= ERR_ALERT | ERR_FATAL;
+					if (conf_err->type.section.status[rc] == HTTP_ERR_IMPORT_NO)
-				free(conf_err->info.errorfiles.name);
+						continue;
-				goto next;
-			}
-
-			free(conf_err->info.errorfiles.name);
-			for (rc = 0; rc < HTTP_ERR_SIZE; rc++) {
-				if (conf_err->info.errorfiles.status[rc] > 0) {
 					if (http_errs->replies[rc])
 						px->replies[rc] = http_errs->replies[rc];
-					else if (conf_err->info.errorfiles.status[rc] == 2)
-						ha_warning("config: proxy '%s' : status '%d' not declared in"
-							   " http-errors section '%s' (at %s:%d).\n",
-							   px->id, http_err_codes[rc], http_errs->id,
-							   conf_err->file, conf_err->line);
 				}
 			}
 		}
-	  next:
+
 		LIST_DELETE(&conf_err->list);
 		free(conf_err->file);
 		free(conf_err);
 	}
 
-  out:
+	return ERR_NONE;
-	return err;
 }
 
 static int post_check_errors()
@@ -2343,6 +2330,55 @@ static int post_check_errors()
 	return err_code;
 }
 
+/* Checks the validity of conf_errors stored in <px> proxy after the
+ * configuration is completely parsed.
+ *
+ * Returns ERR_NONE on success and a combination of ERR_CODE on failure.
+ */
+int proxy_check_http_errors(struct proxy *px)
+{
+	struct http_errors *http_errs;
+	struct conf_errors *conf_err;
+	int section_found;
+	int rc, err = ERR_NONE;
+
+	list_for_each_entry(conf_err, &px->conf.errors, list) {
+		if (conf_err->directive == HTTP_ERR_DIRECTIVE_SECTION) {
+			section_found = 0;
+			list_for_each_entry(http_errs, &http_errors_list, list) {
+				if (strcmp(http_errs->id, conf_err->type.section.name) == 0) {
+					section_found = 1;
+					break;
+				}
+			}
+
+			if (!section_found) {
+				ha_alert("proxy '%s': unknown http-errors section '%s' (at %s:%d).\n",
+				         px->id, conf_err->type.section.name, conf_err->file, conf_err->line);
+				ha_free(&conf_err->type.section.name);
+				err |= ERR_ALERT | ERR_FATAL;
+				continue;
+			}
+
+			conf_err->type.section.resolved = http_errs;
+			ha_free(&conf_err->type.section.name);
+
+			for (rc = 0; rc < HTTP_ERR_SIZE; rc++) {
+				if (conf_err->type.section.status[rc] == HTTP_ERR_IMPORT_EXPLICIT &&
+				    !http_errs->replies[rc]) {
+					ha_warning("config: proxy '%s' : status '%d' not declared in"
+					           " http-errors section '%s' (at %s:%d).\n",
+					           px->id, http_err_codes[rc], http_errs->id,
+					           conf_err->file, conf_err->line);
+					err |= ERR_WARN;
+				}
+			}
+		}
+	}
+
+	return err;
+}
+
 int proxy_dup_default_conf_errors(struct proxy *curpx, const struct proxy *defpx, char **errmsg)
 {
 	struct conf_errors *conf_err, *new_conf_err = NULL;
@@ -2354,19 +2390,22 @@ int proxy_dup_default_conf_errors(struct proxy *curpx, const struct proxy *defpx
 			memprintf(errmsg, "unable to duplicate default errors (out of memory).");
 			goto out;
 		}
-		new_conf_err->type = conf_err->type;
+		new_conf_err->directive = conf_err->directive;
-		if (conf_err->type == 1) {
+		switch (conf_err->directive) {
-			new_conf_err->info.errorfile.status = conf_err->info.errorfile.status;
+		case HTTP_ERR_DIRECTIVE_INLINE:
-			new_conf_err->info.errorfile.reply  = conf_err->info.errorfile.reply;
+			new_conf_err->type.inl.status = conf_err->type.inl.status;
-		}
+			new_conf_err->type.inl.reply  = conf_err->type.inl.reply;
-		else {
+			break;
-			new_conf_err->info.errorfiles.name = strdup(conf_err->info.errorfiles.name);
+
-			if (!new_conf_err->info.errorfiles.name) {
+		case HTTP_ERR_DIRECTIVE_SECTION:
+			new_conf_err->type.section.name = strdup(conf_err->type.section.name);
+			if (!new_conf_err->type.section.name) {
 				memprintf(errmsg, "unable to duplicate default errors (out of memory).");
 				goto out;
 			}
-			memcpy(&new_conf_err->info.errorfiles.status, &conf_err->info.errorfiles.status,
+			memcpy(&new_conf_err->type.section.status, &conf_err->type.section.status,
-			       sizeof(conf_err->info.errorfiles.status));
+			       sizeof(conf_err->type.section.status));
+			break;
 		}
 		new_conf_err->file = strdup(conf_err->file);
 		new_conf_err->line = conf_err->line;
@@ -2385,8 +2424,8 @@ void proxy_release_conf_errors(struct proxy *px)
 	struct conf_errors *conf_err, *conf_err_back;
 
 	list_for_each_entry_safe(conf_err, conf_err_back, &px->conf.errors, list) {
-		if (conf_err->type == 0)
+		if (conf_err->directive == HTTP_ERR_DIRECTIVE_SECTION)
-			free(conf_err->info.errorfiles.name);
+			free(conf_err->type.section.name);
 		LIST_DELETE(&conf_err->list);
 		free(conf_err->file);
 		free(conf_err);
@@ -2505,7 +2544,7 @@ static struct cfg_kw_list cfg_kws = {ILH, {
 }};
 
 INITCALL1(STG_REGISTER, cfg_register_keywords, &cfg_kws);
-REGISTER_POST_PROXY_CHECK(proxy_check_errors);
+REGISTER_POST_PROXY_CHECK(proxy_finalize_http_errors);
 REGISTER_POST_CHECK(post_check_errors);
 
 REGISTER_CONFIG_SECTION("http-errors", cfg_parse_http_errors, NULL);

src/htx.c

@@ -11,6 +11,7 @@
  */
 
 #include <haproxy/chunk.h>
+#include <haproxy/dynbuf.h>
 #include <haproxy/global.h>
 #include <haproxy/htx.h>
 #include <haproxy/net_helper.h>
@@ -719,10 +720,163 @@ struct htx_blk *htx_replace_blk_value(struct htx *htx, struct htx_blk *blk,
 	return blk;
 }
 
+/* Transfer HTX blocks from <src> to <dst>, stopping if <count> bytes were
+ * transferred (including payload and meta-data). It returns the number of bytes
+ * copied. By default, copied blocks are removed from <src> and only full
+ * headers and trailers part can be moved. <flags> can be set to change the
+ * default behavior:
+ *  - HTX_XFER_KEEP_SRC_BLKS: source blocks are not removed
+ *  - HTX_XFER_PARTIAL_HDRS_COPY: partial headers and trailers part can be xferred
+ *  - HTX_XFER_HDRS_ONLY: Only the headers part is xferred
+ */
+size_t htx_xfer(struct htx *dst, struct htx *src, size_t count, unsigned int flags)
+{
+	struct htx_blk *blk, *last_dstblk;
+	size_t ret = 0;
+	uint32_t max, last_dstblk_sz;
+	int dst_full = 0;
+
+	last_dstblk = NULL;
+	last_dstblk_sz = 0;
+	for (blk = htx_get_head_blk(src); blk && count; blk = htx_get_next_blk(src, blk)) {
+		struct ist v;
+		enum htx_blk_type type;
+		uint32_t sz;
+
+		/* Ignore unused block */
+		type = htx_get_blk_type(blk);
+		if (type == HTX_BLK_UNUSED)
+			continue;
+
+		if ((flags & HTX_XFER_HDRS_ONLY) &&
+		    type != HTX_BLK_REQ_SL && type != HTX_BLK_RES_SL &&
+		    type != HTX_BLK_HDR && type != HTX_BLK_EOH)
+			break;
+
+		max = htx_get_max_blksz(dst, count);
+		if (!max)
+			break;
+
+		sz = htx_get_blksz(blk);
+		switch (type) {
+		case HTX_BLK_DATA:
+			v = htx_get_blk_value(src, blk);
+			if (v.len > max)
+				v.len = max;
+			v.len = htx_add_data(dst, v);
+			if (!v.len) {
+				dst_full = 1;
+				goto stop;
+			}
+			last_dstblk = htx_get_tail_blk(dst);
+			last_dstblk_sz = v.len;
+			count -= sizeof(*blk) + v.len;
+			ret += sizeof(*blk) + v.len;
+			if (v.len != sz) {
+				dst_full = 1;
+				goto stop;
+			}
+			break;
+
+		default:
+			if (sz > max) {
+				dst_full = 1;
+				goto stop;
+			}
+
+			last_dstblk = htx_add_blk(dst, type, sz);
+			if (!last_dstblk) {
+				dst_full = 1;
+				goto stop;
+			}
+			last_dstblk->info = blk->info;
+			htx_memcpy(htx_get_blk_ptr(dst, last_dstblk), htx_get_blk_ptr(src, blk), sz);
+			last_dstblk_sz = sz;
+			count -= sizeof(*blk) + sz;
+			ret += sizeof(*blk) + sz;
+			break;
+		}
+
+		last_dstblk = NULL; /* Reset last_dstblk because it was fully copied */
+		last_dstblk_sz = 0;
+	}
+  stop:
+	/* Here, if not NULL, <blk> point on the first not fully copied block in
+	 * <src>. And <last_dstblk>, if defined, is the last not fully copied
+	 * block in <dst>. So have:
+	 *   - <blk> == NULL: everything was copied. <last_dstblk> must be NULL
+	 *   - <blk> != NULL && <last_dstblk> == NULL: partial copy but the last block was fully copied
+	 *   - <blk> != NULL && <last_dstblk> != NULL: partial copy and the last block was patially copied (DATA block only)
+	 */
+	if (!(flags & HTX_XFER_PARTIAL_HDRS_COPY)) {
+		/* Partial headers/trailers copy is not supported */
+		struct htx_blk *dstblk;
+		enum htx_blk_type type = HTX_BLK_UNUSED;
+
+		dstblk = htx_get_tail_blk(dst);
+		if (dstblk)
+			type = htx_get_blk_type(dstblk);
+
+		/* the last copied block is a start-line, a header or a trailer */
+		if (type == HTX_BLK_REQ_SL || type == HTX_BLK_RES_SL || type == HTX_BLK_HDR || type == HTX_BLK_TLR) {
+			/* <src > cannot have partial headers or trailers part */
+			BUG_ON(blk == NULL);
+
+			/* Remove partial headers/trailers from <dst> and rollback on <str> to not remove them later */
+			while (type == HTX_BLK_REQ_SL || type == HTX_BLK_RES_SL || type == HTX_BLK_HDR || type == HTX_BLK_TLR) {
+				BUG_ON(type != htx_get_blk_type(blk));
+				ret -= sizeof(*blk) + htx_get_blksz(blk);
+				htx_remove_blk(dst, dstblk);
+				dstblk = htx_get_tail_blk(dst);
+				blk = htx_get_prev_blk(src, blk);
+				if (!dstblk)
+					break;
+				type = htx_get_blk_type(dstblk);
+			}
+
+			/* Report if the xfer was interrupted because <dst> was
+			 * full but is was originally empty
+			 */
+			if (dst_full && htx_is_empty(dst))
+				src->flags |= HTX_FL_PARSING_ERROR;
+		}
+	}
+
+	if (!(flags & HTX_XFER_KEEP_SRC_BLKS)) {
+		/* True xfer performed, remove copied block from <src> */
+		struct htx_blk *blk2;
+
+		/* Remove all fully copied blocks */
+		if (!blk)
+			htx_drain(src, src->data);
+		else {
+			for (blk2 = htx_get_head_blk(src); blk2 && blk2 != blk; blk2 = htx_remove_blk(src, blk2));
+
+			/* If copy was stopped on a DATA block and the last destination
+			 * block is not NULL, it means a partial copy was performed. So
+			 * cut the source block accordingly
+			 */
+			if (last_dstblk && blk2 && htx_get_blk_type(blk2) == HTX_BLK_DATA) {
+				htx_cut_data_blk(src, blk2, last_dstblk_sz);
+			}
+		}
+	}
+
+	/* Everything was copied, transfert terminal HTX flags too */
+	if (!blk) {
+		dst->flags |= (src->flags & (HTX_FL_EOM|HTX_FL_PARSING_ERROR|HTX_FL_PROCESSING_ERROR));
+		src->flags = 0;
+	}
+
+	return ret;
+}
+
 /* Transfer HTX blocks from <src> to <dst>, stopping once the first block of the
  * type <mark> is transferred (typically EOH or EOT) or when <count> bytes were
  * moved (including payload and meta-data). It returns the number of bytes moved
  * and the last HTX block inserted in <dst>.
+ *
+ * DEPRECATED
  */
 struct htx_ret htx_xfer_blks(struct htx *dst, struct htx *src, uint32_t count,
 			     enum htx_blk_type mark)
@@ -1181,3 +1335,68 @@ int htx_append_msg(struct htx *dst, const struct htx *src)
 	htx_truncate(dst, offset);
 	return 0;
 }
+
+
+/* If possible, trasnfer HTX blocks from <src> to a small buffer. This function
+ * allocate the small buffer and makes <dst> point on it. If <dst> is not empty
+ * or if <src> contains to many data, NULL is returned. If the allocation
+ * failed, NULL is returned. Otherwise <dst> is returned.  <flags> instructs how
+ * the transfer must be performed.
+ */
+struct buffer *__htx_xfer_to_small_buffer(struct buffer *dst, struct buffer *src, unsigned int flags)
+{
+	struct htx *dst_htx;
+	struct htx *src_htx = htxbuf(src);
+	size_t sz = (sizeof(struct htx) + htx_used_space(src_htx));
+
+	if (dst->size || sz > global.tune.bufsize_small  || !b_alloc_small(dst))
+		return NULL;
+	dst_htx = htx_from_buf(dst);
+	htx_xfer(dst_htx, src_htx, src_htx->size, flags);
+	htx_to_buf(dst_htx, dst);
+	return dst;
+}
+
+/* If possible, trasnfer HTX blocks from <src> to a large buffer. This function
+ * allocate the small buffer and makes <dst> point on it. If <dst> is not empty
+ * or if <src> contains to many data, NULL is returned. If the allocation
+ * failed, NULL is returned. Otherwise <dst> is returned.  <flags> instructs how
+ * the transfer must be performed.
+ */
+struct buffer *__htx_xfer_to_large_buffer(struct buffer *dst, struct buffer *src, unsigned int flags)
+{
+	struct htx *dst_htx;
+	struct htx *src_htx = htxbuf(src);
+	size_t sz = (sizeof(struct htx) + htx_used_space(src_htx));
+
+	if (dst->size || sz > global.tune.bufsize_large || !b_alloc_large(dst))
+		return NULL;
+	dst_htx = htx_from_buf(dst);
+	htx_xfer(dst_htx, src_htx, src_htx->size, flags);
+	htx_to_buf(dst_htx, dst);
+	return dst;
+}
+
+/* Move HTX blocks from <src> to <dst>. Relies on __htx_xfer_to_small_buffer() */
+struct buffer *htx_move_to_small_buffer(struct buffer *dst, struct buffer *src)
+{
+	return __htx_xfer_to_small_buffer(dst, src, HTX_XFER_DEFAULT);
+}
+
+/* Move HTX blocks from <src> to <dst>. Relies on __htx_xfer_to_large_buffer() */
+struct buffer *htx_move_to_large_buffer(struct buffer *dst, struct buffer *src)
+{
+	return __htx_xfer_to_large_buffer(dst, src, HTX_XFER_DEFAULT);
+}
+
+/* Copy HTX blocks from <src> to <dst>. Relies on __htx_xfer_to_small_buffer() */
+struct buffer *htx_copy_to_small_buffer(struct buffer *dst, struct buffer *src)
+{
+	return __htx_xfer_to_small_buffer(dst, src, HTX_XFER_KEEP_SRC_BLKS);
+}
+
+/* Copy HTX blocks from <src> to <dst>. Relies on __htx_xfer_to_large_buffer() */
+struct buffer *htx_copy_to_large_buffer(struct buffer *dst, struct buffer *src)
+{
+	return __htx_xfer_to_large_buffer(dst, src, HTX_XFER_KEEP_SRC_BLKS);
+}

src/log.c

@@ -2913,6 +2913,7 @@ static inline void __send_log_set_metadata_sd(struct ist *metadata, char *sd, si
 struct process_send_log_ctx {
 	struct session *sess;
 	struct stream *stream;
+	struct log_profile *profile;
 	struct log_orig origin;
 };
 
@@ -2942,6 +2943,10 @@ static inline void _process_send_log_override(struct process_send_log_ctx *ctx,
 	enum log_orig_id orig = (ctx) ? ctx->origin.id : LOG_ORIG_UNSPEC;
 	uint16_t orig_fl = (ctx) ? ctx->origin.flags : LOG_ORIG_FL_NONE;
 
+	/* ctx->profile gets priority over logger profile */
+	if (ctx && ctx->profile)
+		prof = ctx->profile;
+
 	BUG_ON(!prof);
 
 	if (!b_is_null(&prof->log_tag))
@@ -3095,8 +3100,8 @@ static void process_send_log(struct process_send_log_ctx *ctx,
 
 			nblogger += 1;
 
-			/* logger may use a profile to override a few things */
+			/* caller or default logger may use a profile to override a few things */
-			if (unlikely(logger->prof))
+			if (unlikely(logger->prof || (ctx && ctx->profile)))
 				_process_send_log_override(ctx, logger, hdr, message, size, nblogger);
 			else
 				_process_send_log_final(logger, hdr, message, size, nblogger);
@@ -5200,17 +5205,11 @@ out:
 
 }
 
-/*
+static void do_log_ctx(struct process_send_log_ctx *ctx)
- * opportunistic log when at least the session is known to exist
- * <s> may be NULL
- *
- * Will not log if the frontend has no log defined. By default it will
- * try to emit the log as INFO, unless the stream already exists and
- * set-log-level was used.
- */
-void do_log(struct session *sess, struct stream *s, struct log_orig origin)
 {
-	struct process_send_log_ctx ctx;
+	struct stream *s = ctx->stream;
+	struct session *sess = ctx->sess;
+	struct log_orig origin = ctx->origin;
 	int size;
 	int sd_size = 0;
 	int level = -1;
@@ -5242,11 +5241,27 @@ void do_log(struct session *sess, struct stream *s, struct log_orig origin)
 
 	size = sess_build_logline_orig(sess, s, logline, global.max_syslog_len, &sess->fe->logformat, origin);
 
+	__send_log(ctx, &sess->fe->loggers, &sess->fe->log_tag, level,
+		   logline, size, logline_rfc5424, sd_size);
+}
+
+/*
+ * opportunistic log when at least the session is known to exist
+ * <s> may be NULL
+ *
+ * Will not log if the frontend has no log defined. By default it will
+ * try to emit the log as INFO, unless the stream already exists and
+ * set-log-level was used.
+ */
+void do_log(struct session *sess, struct stream *s, struct log_orig origin)
+{
+	struct process_send_log_ctx ctx;
+
 	ctx.origin = origin;
 	ctx.sess = sess;
 	ctx.stream = s;
-	__send_log(&ctx, &sess->fe->loggers, &sess->fe->log_tag, level,
+	ctx.profile = NULL;
-		   logline, size, logline_rfc5424, sd_size);
+	do_log_ctx(&ctx);
 }
 
 /*
@@ -5297,6 +5312,7 @@ void strm_log(struct stream *s, struct log_orig origin)
 	ctx.origin = origin;
 	ctx.sess = sess;
 	ctx.stream = s;
+	ctx.profile = NULL;
 	__send_log(&ctx, &sess->fe->loggers, &sess->fe->log_tag, level,
 		   logline, size, logline_rfc5424, sd_size);
 	s->logs.logwait = 0;
@@ -5364,6 +5380,7 @@ void _sess_log(struct session *sess, int embryonic)
 	ctx.origin = orig;
 	ctx.sess = sess;
 	ctx.stream = NULL;
+	ctx.profile = NULL;
 	__send_log(&ctx, &sess->fe->loggers,
 	           &sess->fe->log_tag, level,
 		   logline, size, logline_rfc5424, sd_size);
@@ -6910,24 +6927,87 @@ static int px_parse_log_steps(char **args, int section_type, struct proxy *curpx
 static enum act_return do_log_action(struct act_rule *rule, struct proxy *px,
                                      struct session *sess, struct stream *s, int flags)
 {
+	struct process_send_log_ctx ctx;
+
 	/* do_log() expects valid session pointer */
 	BUG_ON(sess == NULL);
 
-	do_log(sess, s, log_orig(rule->arg.expr_int.value, LOG_ORIG_FL_NONE));
+	ctx.origin = log_orig(rule->arg.do_log.orig, LOG_ORIG_FL_NONE);
+	ctx.sess = sess;
+	ctx.stream = s;
+	ctx.profile = rule->arg.do_log.profile;
+
+	do_log_ctx(&ctx);
 	return ACT_RET_CONT;
 }
 
-/* Parse a "do_log" action. It doesn't take any argument
+static int do_log_action_check(struct act_rule *rule, struct proxy *px, char **err)
+{
+	if (rule->arg.do_log.profile_name) {
+		struct log_profile *prof;
+
+		prof = log_profile_find_by_name(rule->arg.do_log.profile_name);
+		if (!prof) {
+			memprintf(err, "do-log action: profile '%s' is invalid", rule->arg.do_log.profile_name);
+			ha_free(&rule->arg.do_log.profile_name);
+			return 0;
+		}
+
+		ha_free(&rule->arg.do_log.profile_name);
+
+		if (!log_profile_postcheck(px, prof, err)) {
+			memprintf(err, "do-log action on %s %s uses incompatible log-profile '%s': %s",  proxy_type_str(px), px->id, prof->id, *err);
+			return 0;
+		}
+		rule->arg.do_log.profile = prof;
+	}
+	return 1; // success
+}
+
+static void do_log_action_release(struct act_rule *rule)
+{
+	ha_free(&rule->arg.do_log.profile_name);
+}
+
+
+/* Parse a "do_log" action. It takes optional "log-profile" argument to
+ * specifically use a given log-profile when generating the log message
+ *
  * May be used from places where per-context actions are usually registered
  */
 enum act_parse_ret do_log_parse_act(enum log_orig_id id,
                                     const char **args, int *orig_arg, struct proxy *px,
                                     struct act_rule *rule, char **err)
 {
+	int cur_arg = *orig_arg;
+
 	rule->action_ptr = do_log_action;
 	rule->action = ACT_CUSTOM;
-	rule->release_ptr = NULL;
+	rule->check_ptr = do_log_action_check;
-	rule->arg.expr_int.value = id;
+	rule->release_ptr = do_log_action_release;
+	rule->arg.do_log.orig = id;
+
+	while (*args[*orig_arg]) {
+		if (!strcmp(args[*orig_arg], "profile")) {
+			if (!*args[*orig_arg + 1]) {
+				memprintf(err,
+					  "action '%s': 'profile' expects argument.",
+					  args[cur_arg-1]);
+				return ACT_RET_PRS_ERR;
+			}
+			rule->arg.do_log.profile_name = strdup(args[*orig_arg + 1]);
+			if (!rule->arg.do_log.profile_name) {
+				memprintf(err,
+					  "action '%s': memory error when setting 'profile'",
+				           args[cur_arg-1]);
+				return ACT_RET_PRS_ERR;
+			}
+			*orig_arg += 2;
+		}
+		else
+			break;
+	}
+
 	return ACT_RET_PRS_OK;
 }
 

src/mux_h2.c

@@ -489,6 +489,9 @@ static int h2_be_glitches_threshold           =     0; /* backend's max glitches
 static int h2_fe_glitches_threshold           =     0; /* frontend's max glitches: unlimited */
 static uint h2_be_rxbuf                       =     0; /* backend's default total rxbuf (bytes) */
 static uint h2_fe_rxbuf                       =     0; /* frontend's default total rxbuf (bytes) */
+static unsigned int h2_be_max_frames_at_once  =     0; /* backend value: 0=no limit */
+static unsigned int h2_fe_max_frames_at_once  =     0; /* frontend value: 0=no limit */
+static unsigned int h2_fe_max_rst_at_once     =     0; /* frontend value: 0=no limit */
 static unsigned int h2_settings_max_concurrent_streams    = 100; /* default value */
 static unsigned int h2_be_settings_max_concurrent_streams =   0; /* backend value */
 static unsigned int h2_fe_settings_max_concurrent_streams =   0; /* frontend value */
@@ -4239,6 +4242,8 @@ static void h2_process_demux(struct h2c *h2c)
 	struct h2_fh hdr;
 	unsigned int padlen = 0;
 	int32_t old_iw = h2c->miw;
+	uint frames_budget = 0;
+	uint rst_budget = 0;
 
 	TRACE_ENTER(H2_EV_H2C_WAKE, h2c->conn);
 
@@ -4327,6 +4332,14 @@ static void h2_process_demux(struct h2c *h2c)
 		}
 	}
 
+	if (h2c->flags & H2_CF_IS_BACK) {
+		frames_budget = h2_be_max_frames_at_once;
+	}
+	else {
+		frames_budget = h2_fe_max_frames_at_once;
+		rst_budget    = h2_fe_max_rst_at_once;
+	}
+
 	/* process as many incoming frames as possible below */
 	while (1) {
 		int ret = 0;
@@ -4629,6 +4642,29 @@ static void h2_process_demux(struct h2c *h2c)
 				h2c->st0 = H2_CS_FRAME_H;
 			}
 		}
+
+		/* If more frames remain in the buffer, let's first check if we've
+		 * depleted the frames processing budget. Consuming the RST budget
+		 * makes the tasklet go to TL_BULK to make it less prioritary than
+		 * other processing since it's often used by attacks, while other
+		 * frame types just yield normally.
+		 */
+		if (b_data(&h2c->dbuf)) {
+			if (h2c->dft == H2_FT_RST_STREAM && (rst_budget && !--rst_budget)) {
+				/* we've consumed all RST frames permitted by
+				 * the budget, we have to yield now.
+				 */
+				tasklet_wakeup(h2c->wait_event.tasklet, 0);
+				break;
+			}
+			else if ((frames_budget && !--frames_budget)) {
+				/* we've consumed all frames permitted by the
+				 * budget, we have to yield now.
+				 */
+				tasklet_wakeup(h2c->wait_event.tasklet);
+				break;
+			}
+		}
 	}
 
 	if (h2c_update_strm_rx_win(h2c) &&
@@ -7830,7 +7866,6 @@ static size_t h2_rcv_buf(struct stconn *sc, struct buffer *buf, size_t count, in
 	struct htx *h2s_htx = NULL;
 	struct htx *buf_htx = NULL;
 	struct buffer *rxbuf = NULL;
-	struct htx_ret htxret;
 	size_t ret = 0;
 	uint prev_h2c_flags = h2c->flags;
 	unsigned long long prev_body_len = h2s->body_len;
@@ -7865,17 +7900,7 @@ static size_t h2_rcv_buf(struct stconn *sc, struct buffer *buf, size_t count, in
 		goto end;
 	}
 
-	htxret = htx_xfer_blks(buf_htx, h2s_htx, count, HTX_BLK_UNUSED);
+	count -= htx_xfer(buf_htx, h2s_htx, count, HTX_XFER_DEFAULT);
-	count -= htxret.ret;
-
-	if (h2s_htx->flags & HTX_FL_PARSING_ERROR) {
-		buf_htx->flags |= HTX_FL_PARSING_ERROR;
-		if (htx_is_empty(buf_htx))
-			se_fl_set(h2s->sd, SE_FL_EOI);
-	}
-	else if (htx_is_empty(h2s_htx)) {
-		buf_htx->flags |= (h2s_htx->flags & HTX_FL_EOM);
-	}
 
 	htx_to_buf(buf_htx, buf);
 	htx_to_buf(h2s_htx, rxbuf);
@@ -7904,13 +7929,7 @@ static size_t h2_rcv_buf(struct stconn *sc, struct buffer *buf, size_t count, in
 
 	/* tell the stream layer whether there are data left or not */
 	if (h2s_rxbuf_cnt(h2s)) {
-		/* Note that parsing errors can also arrive here, we may need
-		 * to propagate errors upstream otherwise no new activity will
-		 * unblock them.
-		 */
 		se_fl_set(h2s->sd, SE_FL_RCV_MORE | SE_FL_WANT_ROOM);
-		if (h2s_htx && h2s_htx->flags & HTX_FL_PARSING_ERROR)
-			h2s_propagate_term_flags(h2c, h2s);
 		BUG_ON_HOT(!buf->data);
 	}
 	else {
@@ -8800,6 +8819,30 @@ static int h2_parse_max_total_streams(char **args, int section_type, struct prox
 	return 0;
 }
 
+/* config parser for global "tune.h2.{be.,fe.,}max-{frames,rst}-at-once" */
+static int h2_parse_max_frames_at_once(char **args, int section_type, struct proxy *curpx,
+                                       const struct proxy *defpx, const char *file, int line,
+                                       char **err)
+{
+	uint *vptr;
+
+	/* backend/frontend/default */
+	if (strcmp(args[0], "tune.h2.be.max-frames-at-once") == 0)
+		vptr = &h2_be_max_frames_at_once;
+	else if (strcmp(args[0], "tune.h2.fe.max-frames-at-once") == 0)
+		vptr = &h2_fe_max_frames_at_once;
+	else if (strcmp(args[0], "tune.h2.fe.max-rst-at-once") == 0)
+		vptr = &h2_fe_max_rst_at_once;
+	else
+		BUG_ON(1, "unhandled keyword");
+
+	if (too_many_args(1, args, err, NULL))
+		return -1;
+
+	*vptr = atoi(args[1]);
+	return 0;
+}
+
 /* config parser for global "tune.h2.max-frame-size" */
 static int h2_parse_max_frame_size(char **args, int section_type, struct proxy *curpx,
                                    const struct proxy *defpx, const char *file, int line,
@@ -8898,10 +8941,13 @@ static struct cfg_kw_list cfg_kws = {ILH, {
 	{ CFG_GLOBAL, "tune.h2.be.glitches-threshold",  h2_parse_glitches_threshold     },
 	{ CFG_GLOBAL, "tune.h2.be.initial-window-size", h2_parse_initial_window_size    },
 	{ CFG_GLOBAL, "tune.h2.be.max-concurrent-streams", h2_parse_max_concurrent_streams },
+	{ CFG_GLOBAL, "tune.h2.be.max-frames-at-once",  h2_parse_max_frames_at_once     },
 	{ CFG_GLOBAL, "tune.h2.be.rxbuf",               h2_parse_rxbuf                  },
 	{ CFG_GLOBAL, "tune.h2.fe.glitches-threshold",  h2_parse_glitches_threshold     },
 	{ CFG_GLOBAL, "tune.h2.fe.initial-window-size", h2_parse_initial_window_size    },
 	{ CFG_GLOBAL, "tune.h2.fe.max-concurrent-streams", h2_parse_max_concurrent_streams },
+	{ CFG_GLOBAL, "tune.h2.fe.max-frames-at-once",  h2_parse_max_frames_at_once     },
+	{ CFG_GLOBAL, "tune.h2.fe.max-rst-at-once",     h2_parse_max_frames_at_once     },
 	{ CFG_GLOBAL, "tune.h2.fe.max-total-streams",   h2_parse_max_total_streams      },
 	{ CFG_GLOBAL, "tune.h2.fe.rxbuf",               h2_parse_rxbuf                  },
 	{ CFG_GLOBAL, "tune.h2.header-table-size",      h2_parse_header_table_size      },

src/proxy.c

@@ -1676,11 +1676,17 @@ int proxy_finalize(struct proxy *px, int *err_code)
 		}
 
 		if (bind_conf->mux_proto) {
+			int is_quic;
+
+			if ((bind_conf->options & (BC_O_USE_SOCK_DGRAM | BC_O_USE_XPRT_STREAM)) == (BC_O_USE_SOCK_DGRAM | BC_O_USE_XPRT_STREAM))
+				is_quic = 1;
+			else
+				is_quic = 0;
 			/* it is possible that an incorrect mux was referenced
 			 * due to the proxy's mode not being taken into account
 			 * on first pass. Let's adjust it now.
 			 */
-			mux_ent = conn_get_best_mux_entry(bind_conf->mux_proto->token, PROTO_SIDE_FE, mode);
+			mux_ent = conn_get_best_mux_entry(bind_conf->mux_proto->token, PROTO_SIDE_FE, is_quic, mode);
 
 			if (!mux_ent || !isteq(mux_ent->token, bind_conf->mux_proto->token)) {
 				ha_alert("%s '%s' : MUX protocol '%.*s' is not usable for 'bind %s' at [%s:%d].\n",
@@ -2672,6 +2678,8 @@ int proxy_finalize(struct proxy *px, int *err_code)
 		*err_code |= ERR_WARN;
 	}
 
+	*err_code |= proxy_check_http_errors(px);
+
 	if (px->mode != PR_MODE_HTTP && !(px->options & PR_O_HTTP_UPG)) {
 		int optnum;
 
@@ -2877,7 +2885,7 @@ int proxy_finalize(struct proxy *px, int *err_code)
 		 * due to the proxy's mode not being taken into account
 		 * on first pass. Let's adjust it now.
 		 */
-		mux_ent = conn_get_best_mux_entry(newsrv->mux_proto->token, PROTO_SIDE_BE, mode);
+		mux_ent = conn_get_best_mux_entry(newsrv->mux_proto->token, PROTO_SIDE_BE, srv_is_quic(newsrv), mode);
 
 		if (!mux_ent || !isteq(mux_ent->token, newsrv->mux_proto->token)) {
 			ha_alert("%s '%s' : MUX protocol '%.*s' is not usable for server '%s' at [%s:%d].\n",
@@ -4916,10 +4924,6 @@ static int cli_parse_add_backend(char **args, char *payload, struct appctx *appc
 		           def_name, proxy_mode_str(defpx->mode)));
 		return 1;
 	}
-	if (!LIST_ISEMPTY(&defpx->conf.errors)) {
-		cli_dynerr(appctx, memprintf(&msg, "Dynamic backends cannot inherit from default proxy '%s' because it references HTTP errors.\n", def_name));
-		return 1;
-	}
 
 	thread_isolate();
 

src/qmux_http.c

@@ -44,7 +44,7 @@ size_t qcs_http_rcv_buf(struct qcs *qcs, struct buffer *buf, size_t count,
 		goto end;
 	}
 
-	htx_xfer_blks(cs_htx, qcs_htx, count, HTX_BLK_UNUSED);
+	htx_xfer(cs_htx, qcs_htx, count, HTX_XFER_DEFAULT);
 	BUG_ON(qcs_htx->flags & HTX_FL_PARSING_ERROR);
 
 	/* Copy EOM from src to dst buffer if all data copied. */

src/qpack-dec.c

@@ -61,10 +61,10 @@
 static uint64_t qpack_get_varint(const unsigned char **buf, uint64_t *len_in, int b)
 {
 	uint64_t ret = 0;
-	int len = *len_in;
+	uint64_t len = *len_in;
 	const uint8_t *raw = *buf;
-	uint64_t v, max = ~0;
+	uint64_t v, limit = (1ULL << 62) - 1;
-	uint8_t shift = 0;
+	int shift = 0;
 
 	if (len == 0)
 		goto too_short;
@@ -77,24 +77,26 @@ static uint64_t qpack_get_varint(const unsigned char **buf, uint64_t *len_in, in
 	do {
 		if (!len)
 			goto too_short;
+
 		v = *raw++;
 		len--;
-		if (v & 127) { // make UBSan happy
+		/* This check is sufficient to prevent any overflow
-			if ((v & 127) > max)
+		 * and implicitly limits shift to 63.
-				goto too_large;
+		 */
-			ret += (v & 127) << shift;
+		if ((v & 127) > (limit - ret) >> shift)
-		}
+			goto too_large;
-		max >>= 7;
+
+		ret += (v & 127) << shift;
 		shift += 7;
 	} while (v & 128);
 
- end:
+end:
 	*buf = raw;
 	*len_in = len;
 	return ret;
 
- too_large:
+too_large:
- too_short:
+too_short:
 	*len_in = (uint64_t)-1;
 	return 0;
 }
@@ -402,7 +404,10 @@ int qpack_decode_fs(const unsigned char *raw, uint64_t len, struct buffer *tmp,
 			n = efl_type & 0x20;
 			static_tbl = efl_type & 0x10;
 			index = qpack_get_varint(&raw, &len, 4);
-			if (len == (uint64_t)-1) {
+			/* There must be at least one byte available for <h> value after this
+			 * decoding before the next call to qpack_get_varint().
+			 */
+			if ((int64_t)len <= 0) {
 				qpack_debug_printf(stderr, "##ERR@%d\n", __LINE__);
 				ret = -QPACK_RET_TRUNCATED;
 				goto out;
@@ -474,7 +479,10 @@ int qpack_decode_fs(const unsigned char *raw, uint64_t len, struct buffer *tmp,
 			n = *raw & 0x10;
 			hname = *raw & 0x08;
 			name_len = qpack_get_varint(&raw, &len, 3);
-			if (len == (uint64_t)-1 || len < name_len) {
+			/* There must be at least one byte available for <hvalue> after this
+			 * decoding before the next call to qpack_get_varint().
+			 */
+			if ((int64_t)len < (int64_t)name_len + 1) {
 				qpack_debug_printf(stderr, "##ERR@%d\n", __LINE__);
 				ret = -QPACK_RET_TRUNCATED;
 				goto out;

src/quic_stream.c

@@ -16,8 +16,6 @@ DECLARE_STATIC_TYPED_POOL(pool_head_quic_stream_desc, "qc_stream_desc", struct q
 DECLARE_STATIC_TYPED_POOL(pool_head_quic_stream_buf, "qc_stream_buf", struct qc_stream_buf);
 DECLARE_STATIC_TYPED_POOL(pool_head_quic_stream_ack, "qc_stream_ack", struct qc_stream_ack);
 
-static struct pool_head *pool_head_sbuf;
-
 static void qc_stream_buf_free(struct qc_stream_desc *stream,
                                struct qc_stream_buf **stream_buf)
 {
@@ -39,13 +37,10 @@ static void qc_stream_buf_free(struct qc_stream_desc *stream,
 		room = b_data(buf);
 	}
 
-	if ((*stream_buf)->sbuf) {
+	b_free(buf);
-		pool_free(pool_head_sbuf, buf->area);
+	if (!(*stream_buf)->sbuf) {
-	}
-	else {
 		bdata_ctr_del(&stream->data, b_data(buf));
 		bdata_ctr_bdec(&stream->data);
-		b_free(buf);
 		offer_buffers(NULL, 1);
 	}
 	pool_free(pool_head_quic_stream_buf, *stream_buf);
@@ -412,10 +407,7 @@ void qc_stream_desc_free(struct qc_stream_desc *stream, int closing)
 			pool_free(pool_head_quic_stream_ack, ack);
 		}
 
-		if (buf->sbuf)
+		b_free(&buf->buf);
-			pool_free(pool_head_sbuf, buf->buf.area);
-		else
-			b_free(&buf->buf);
 
 		eb64_delete(&buf->offset_node);
 		pool_free(pool_head_quic_stream_buf, buf);
@@ -461,7 +453,7 @@ struct buffer *qc_stream_buf_alloc(struct qc_stream_desc *stream,
 	stream->buf->buf = BUF_NULL;
 	stream->buf->offset_node.key = offset;
 
-	if (!small) {
+	if (!small || !global.tune.bufsize_small) {
 		stream->buf->sbuf = 0;
 		if (!b_alloc(&stream->buf->buf, DB_MUX_TX)) {
 			pool_free(pool_head_quic_stream_buf, stream->buf);
@@ -470,16 +462,12 @@ struct buffer *qc_stream_buf_alloc(struct qc_stream_desc *stream,
 		}
 	}
 	else {
-		char *area;
+		if (!b_alloc_small(&stream->buf->buf)) {
-
-		if (!(area = pool_alloc(pool_head_sbuf))) {
 			pool_free(pool_head_quic_stream_buf, stream->buf);
 			stream->buf = NULL;
 			return NULL;
 		}
-
 		stream->buf->sbuf = 1;
-		stream->buf->buf = b_make(area, global.tune.bufsize_small, 0, 0);
 	}
 
 	eb64_insert(&stream->buf_tree, &stream->buf->offset_node);
@@ -502,7 +490,7 @@ struct buffer *qc_stream_buf_realloc(struct qc_stream_desc *stream)
 	BUG_ON(b_data(&stream->buf->buf));
 
 	/* Release buffer */
-	pool_free(pool_head_sbuf, stream->buf->buf.area);
+	b_free(&stream->buf->buf);
 	stream->buf->buf = BUF_NULL;
 	stream->buf->sbuf = 0;
 
@@ -536,23 +524,3 @@ void qc_stream_buf_release(struct qc_stream_desc *stream)
 	if (stream->notify_room && room)
 		stream->notify_room(stream, room);
 }
-
-static int create_sbuf_pool(void)
-{
-	if (global.tune.bufsize_small > global.tune.bufsize) {
-		ha_warning("invalid small buffer size %d bytes which is greater to default bufsize %d bytes.\n",
-		           global.tune.bufsize_small, global.tune.bufsize);
-		return ERR_FATAL|ERR_ABORT;
-	}
-
-	pool_head_sbuf = create_pool("sbuf", global.tune.bufsize_small,
-	                             MEM_F_SHARED|MEM_F_EXACT);
-	if (!pool_head_sbuf) {
-		ha_warning("error on small buffer pool allocation.\n");
-		return ERR_FATAL|ERR_ABORT;
-	}
-
-	return ERR_NONE;
-}
-
-REGISTER_POST_CHECK(create_sbuf_pool);

src/server.c

@@ -37,6 +37,7 @@
 #include <haproxy/namespace.h>
 #include <haproxy/port_range.h>
 #include <haproxy/protocol.h>
+#include <haproxy/proto_tcp.h>
 #include <haproxy/proxy.h>
 #include <haproxy/queue.h>
 #include <haproxy/quic_tp.h>
@@ -2938,7 +2939,9 @@ void srv_settings_cpy(struct server *srv, const struct server *src, int srv_tmpl
 	}
 	srv->use_ssl                  = src->use_ssl;
 	srv->check.addr               = src->check.addr;
+	srv->check.proto              = src->check.proto;
 	srv->agent.addr               = src->agent.addr;
+	srv->agent.proto              = src->agent.proto;
 	srv->check.use_ssl            = src->check.use_ssl;
 	srv->check.port               = src->check.port;
 	if (src->check.sni != NULL)
@@ -4635,6 +4638,11 @@ out:
 			set_srv_agent_addr(s, &sk);
 		if (port)
 			set_srv_agent_port(s, new_port);
+		/* Agent currently only uses TCP */
+		if (sk.ss_family == AF_INET)
+			s->agent.proto = &proto_tcpv4;
+		else
+			s->agent.proto = &proto_tcpv6;
 	}
 	return NULL;
 }
@@ -4646,7 +4654,8 @@ out:
  */
 const char *srv_update_check_addr_port(struct server *s, const char *addr, const char *port)
 {
-	struct sockaddr_storage sk;
+	struct sockaddr_storage *sk = NULL;
+	struct protocol *proto = NULL;
 	struct buffer *msg;
 	int new_port;
 
@@ -4658,8 +4667,8 @@ const char *srv_update_check_addr_port(struct server *s, const char *addr, const
 		goto out;
 	}
 	if (addr) {
-		memset(&sk, 0, sizeof(struct sockaddr_storage));
+		sk = str2sa_range(addr, NULL, NULL, NULL, NULL, &proto, NULL, NULL, NULL, NULL, NULL, 0);
-		if (str2ip2(addr, &sk, 0) == NULL) {
+		if (sk == NULL) {
 			chunk_appendf(msg, "invalid addr '%s'", addr);
 			goto out;
 		}
@@ -4683,8 +4692,10 @@ out:
 	if (msg->data)
 		return msg->area;
 	else {
-		if (addr)
+		if (sk) {
-			s->check.addr = sk;
+			s->check.addr = *sk;
+			s->check.proto = proto;
+		}
 		if (port)
 			s->check.port = new_port;
 
@@ -6230,7 +6241,7 @@ static int cli_parse_add_server(char **args, char *payload, struct appctx *appct
 		int proto_mode = conn_pr_mode_to_proto_mode(be->mode);
 		const struct mux_proto_list *mux_ent;
 
-		mux_ent = conn_get_best_mux_entry(srv->mux_proto->token, PROTO_SIDE_BE, proto_mode);
+		mux_ent = conn_get_best_mux_entry(srv->mux_proto->token, PROTO_SIDE_BE, srv_is_quic(srv), proto_mode);
 
 		if (!mux_ent || !isteq(mux_ent->token, srv->mux_proto->token)) {
 			ha_alert("MUX protocol is not usable for server.\n");
@@ -7612,7 +7623,7 @@ static void srv_close_idle_conns(struct server *srv)
 
 REGISTER_SERVER_DEINIT(srv_close_idle_conns);
 
-/* config parser for global "tune.idle-pool.shared", accepts "on" or "off" */
+/* config parser for global "tune.idle-pool.shared", accepts "full", "on" or "off" */
 static int cfg_parse_idle_pool_shared(char **args, int section_type, struct proxy *curpx,
                                       const struct proxy *defpx, const char *file, int line,
                                       char **err)
@@ -7620,12 +7631,17 @@ static int cfg_parse_idle_pool_shared(char **args, int section_type, struct prox
 	if (too_many_args(1, args, err, NULL))
 		return -1;
 
-	if (strcmp(args[1], "on") == 0)
+	if (strcmp(args[1], "full") == 0) {
 		global.tune.options |= GTUNE_IDLE_POOL_SHARED;
-	else if (strcmp(args[1], "off") == 0)
+		global.tune.tg_takeover = FULL_THREADGROUP_TAKEOVER;
+	} else if (strcmp(args[1], "on") == 0) {
+		global.tune.options |= GTUNE_IDLE_POOL_SHARED;
+		global.tune.tg_takeover = RESTRICTED_THREADGROUP_TAKEOVER;
+	} else if (strcmp(args[1], "off") == 0) {
 		global.tune.options &= ~GTUNE_IDLE_POOL_SHARED;
-	else {
+		global.tune.tg_takeover = NO_THREADGROUP_TAKEOVER;
-		memprintf(err, "'%s' expects either 'on' or 'off' but got '%s'.", args[0], args[1]);
+	} else {
+		memprintf(err, "'%s' expects 'auto', 'on' or 'off' but got '%s'.", args[0], args[1]);
 		return -1;
 	}
 	return 0;

src/sock.c

@@ -162,8 +162,8 @@ struct connection *sock_accept_conn(struct listener *l, int *status)
 	case ENFILE:
 		if (p)
 			send_log(p, LOG_EMERG,
-			         "Proxy %s reached system FD limit (maxsock=%d). Please check system tunables.\n",
+			         "Proxy %s reached system FD limit (actconn=%d). Please check system tunables.\n",
-			         p->id, global.maxsock);
+			         p->id, actconn);
 		ret = CO_AC_PAUSE;
 		break;
 
@@ -179,8 +179,8 @@ struct connection *sock_accept_conn(struct listener *l, int *status)
 	case ENOMEM:
 		if (p)
 			send_log(p, LOG_EMERG,
-			         "Proxy %s reached system memory limit (maxsock=%d). Please check system tunables.\n",
+			         "Proxy %s reached system memory limit (actconn=%d). Please check system tunables.\n",
-			         p->id, global.maxsock);
+			         p->id, actconn);
 		ret = CO_AC_PAUSE;
 		break;
 

src/stconn.c

@@ -1190,7 +1190,7 @@ int sc_conn_recv(struct stconn *sc)
 		 * SE_FL_RCV_MORE on the SC if more space is needed.
 		 */
 		max = channel_recv_max(ic);
-		if ((ic->flags & CF_WROTE_DATA) && b_is_large(sc_ib(sc)))
+		if (b_is_small(sc_ib(sc)) || ((ic->flags & CF_WROTE_DATA) && b_is_large(sc_ib(sc))))
 			max = 0;
 		ret = CALL_MUX_WITH_RET(conn->mux, rcv_buf(sc, &ic->buf, max, cur_flags));
 
@@ -1496,16 +1496,21 @@ int sc_conn_send(struct stconn *sc)
 			if (s->txn->req.msg_state != HTTP_MSG_DONE || b_is_large(&oc->buf))
 				s->txn->flags &= ~TX_L7_RETRY;
 			else {
-				if (b_alloc(&s->txn->l7_buffer, DB_UNLIKELY) == NULL)
+				if (!(s->be->options2 & PR_O2_USE_SBUF_L7_RETRY) ||
-					s->txn->flags &= ~TX_L7_RETRY;
+				    !htx_copy_to_small_buffer(&s->txn->l7_buffer, &oc->buf)) {
-				else {
+					if (b_alloc(&s->txn->l7_buffer, DB_UNLIKELY) == NULL)
-					memcpy(b_orig(&s->txn->l7_buffer),
+						s->txn->flags &= ~TX_L7_RETRY;
-					       b_orig(&oc->buf),
+					else {
-					       b_size(&oc->buf));
+						memcpy(b_orig(&s->txn->l7_buffer),
-					s->txn->l7_buffer.head = co_data(oc);
+						       b_orig(&oc->buf),
-					b_add(&s->txn->l7_buffer, co_data(oc));
+						       b_size(&oc->buf));
+					}
 				}
 
+				if (s->txn->flags & TX_L7_RETRY) {
+					s->txn->l7_buffer.head = co_data(oc);
+					b_set_data(&s->txn->l7_buffer, co_data(oc));
+				}
 			}
 		}
 
@@ -1861,7 +1866,7 @@ int sc_applet_recv(struct stconn *sc)
 	 * SE_FL_RCV_MORE on the SC if more space is needed.
 	 */
 	max = channel_recv_max(ic);
-	if ((ic->flags & CF_WROTE_DATA) && b_is_large(sc_ib(sc)))
+	if (b_is_small(sc_ib(sc)) || ((ic->flags & CF_WROTE_DATA) && b_is_large(sc_ib(sc))))
 		max = 0;
 	ret = appctx_rcv_buf(sc, &ic->buf, max, flags);
 	if (sc_ep_test(sc, SE_FL_WANT_ROOM)) {

src/stream.c

@@ -2511,6 +2511,29 @@ struct task *process_stream(struct task *t, void *context, unsigned int state)
 			srv = objt_server(s->target);
 			if (scb->state == SC_ST_ASS && srv && srv->rdr_len && (s->flags & SF_REDIRECTABLE))
 				http_perform_server_redirect(s, scb);
+
+			if (unlikely((s->be->options2 & PR_O2_USE_SBUF_QUEUE) && scb->state == SC_ST_QUE)) {
+				struct buffer sbuf = BUF_NULL;
+
+				if (IS_HTX_STRM(s)) {
+					if (!htx_move_to_small_buffer(&sbuf, &req->buf))
+						break;
+				}
+				else {
+					if (b_size(&req->buf) == global.tune.bufsize_small ||
+					    b_data(&req->buf) > global.tune.bufsize_small)
+						break;
+					if (!b_alloc_small(&sbuf))
+						break;
+					b_xfer(&sbuf, &req->buf, b_data(&req->buf));
+				}
+
+				b_free(&req->buf);
+				offer_buffers(s, 1);
+				req->buf = sbuf;
+				DBG_TRACE_DEVEL("request moved to a small buffer", STRM_EV_STRM_PROC, s);
+			}
+
 		} while (scb->state == SC_ST_ASS);
 	}
 
@@ -3266,7 +3289,7 @@ static int check_tcp_switch_stream_mode(struct act_rule *rule, struct proxy *px,
 		px->options |= PR_O_HTTP_UPG;
 
 	if (mux_proto) {
-		mux_ent = conn_get_best_mux_entry(mux_proto->token, PROTO_SIDE_FE, mode);
+		mux_ent = conn_get_best_mux_entry(mux_proto->token, PROTO_SIDE_FE, 0, mode);
 		if (!mux_ent || !isteq(mux_ent->token, mux_proto->token)) {
 			memprintf(err, "MUX protocol '%.*s' is not compatible with the selected mode",
 				  (int)mux_proto->token.len, mux_proto->token.ptr);
@@ -3274,7 +3297,7 @@ static int check_tcp_switch_stream_mode(struct act_rule *rule, struct proxy *px,
 		}
 	}
 	else {
-		mux_ent = conn_get_best_mux_entry(IST_NULL, PROTO_SIDE_FE, mode);
+		mux_ent = conn_get_best_mux_entry(IST_NULL, PROTO_SIDE_FE, 0, mode);
 		if (!mux_ent) {
 			memprintf(err, "Unable to find compatible MUX protocol with the selected mode");
 			return 0;

src/task.c

@@ -147,8 +147,7 @@ void __tasklet_wakeup_on(struct tasklet *tl, int thr)
 			LIST_APPEND(&th_ctx->tasklets[TL_BULK], &tl->list);
 			th_ctx->tl_class_mask |= 1 << TL_BULK;
 		}
-		else if ((struct task *)tl == th_ctx->current) {
+		else if ((struct task *)tl == th_ctx->current && !(tl->state & TASK_WOKEN_ANY)) {
-			_HA_ATOMIC_OR(&tl->state, TASK_SELF_WAKING);
 			LIST_APPEND(&th_ctx->tasklets[TL_BULK], &tl->list);
 			th_ctx->tl_class_mask |= 1 << TL_BULK;
 		}
@@ -157,8 +156,8 @@ void __tasklet_wakeup_on(struct tasklet *tl, int thr)
 			th_ctx->tl_class_mask |= 1 << TL_URGENT;
 		}
 		else {
-			LIST_APPEND(&th_ctx->tasklets[th_ctx->current_queue], &tl->list);
+			LIST_APPEND(&th_ctx->tasklets[TL_NORMAL], &tl->list);
-			th_ctx->tl_class_mask |= 1 << th_ctx->current_queue;
+			th_ctx->tl_class_mask |= 1 << TL_NORMAL;
 		}
 		_HA_ATOMIC_INC(&th_ctx->rq_total);
 	} else {
@@ -186,8 +185,7 @@ struct list *__tasklet_wakeup_after(struct list *head, struct tasklet *tl)
 			LIST_INSERT(&th_ctx->tasklets[TL_BULK], &tl->list);
 			th_ctx->tl_class_mask |= 1 << TL_BULK;
 		}
-		else if ((struct task *)tl == th_ctx->current) {
+		else if ((struct task *)tl == th_ctx->current && !(tl->state & TASK_WOKEN_ANY)) {
-			_HA_ATOMIC_OR(&tl->state, TASK_SELF_WAKING);
 			LIST_INSERT(&th_ctx->tasklets[TL_BULK], &tl->list);
 			th_ctx->tl_class_mask |= 1 << TL_BULK;
 		}
@@ -196,8 +194,8 @@ struct list *__tasklet_wakeup_after(struct list *head, struct tasklet *tl)
 			th_ctx->tl_class_mask |= 1 << TL_URGENT;
 		}
 		else {
-			LIST_INSERT(&th_ctx->tasklets[th_ctx->current_queue], &tl->list);
+			LIST_INSERT(&th_ctx->tasklets[TL_NORMAL], &tl->list);
-			th_ctx->tl_class_mask |= 1 << th_ctx->current_queue;
+			th_ctx->tl_class_mask |= 1 << TL_NORMAL;
 		}
 	}
 	else {
@@ -563,14 +561,22 @@ unsigned int run_tasks_from_lists(unsigned int budgets[])
 			continue;
 		}
 
-		budgets[queue]--;
-		activity[tid].ctxsw++;
-
 		t = (struct task *)LIST_ELEM(tl_queues[queue].n, struct tasklet *, list);
+
+		/* check if this task has already run during this loop */
+		if ((uint16_t)t->last_run == (uint16_t)activity[tid].loops) {
+			budget_mask &= ~(1 << queue);
+			queue++;
+			continue;
+		}
+		t->last_run = activity[tid].loops;
 		ctx = t->context;
 		process = t->process;
 		t->calls++;
 
+		budgets[queue]--;
+		activity[tid].ctxsw++;
+
 		th_ctx->lock_wait_total = 0;
 		th_ctx->mem_wait_total = 0;
 		th_ctx->locked_total = 0;
@@ -723,8 +729,8 @@ void process_runnable_tasks()
 	struct task *t;
 	const unsigned int default_weights[TL_CLASSES] = {
 		[TL_URGENT] = 64, // ~50% of CPU bandwidth for I/O
-		[TL_NORMAL] = 48, // ~37% of CPU bandwidth for tasks
+		[TL_NORMAL] = 60, // ~47% of CPU bandwidth for tasks
-		[TL_BULK]   = 16, // ~13% of CPU bandwidth for self-wakers
+		[TL_BULK]   = 4,  // ~3% of CPU bandwidth for self-wakers
 		[TL_HEAVY]  = 1,  // never more than 1 heavy task at once
 	};
 	unsigned int max[TL_CLASSES]; // max to be run per class
@@ -734,7 +740,7 @@ void process_runnable_tasks()
 	int max_processed;
 	int lpicked, gpicked;
 	int heavy_queued = 0;
-	int budget;
+	int budget, done;
 
 	_HA_ATOMIC_AND(&th_ctx->flags, ~TH_FL_STUCK); // this thread is still running
 
@@ -904,10 +910,11 @@ void process_runnable_tasks()
 	}
 
 	/* execute tasklets in each queue */
-	max_processed -= run_tasks_from_lists(max);
+	done = run_tasks_from_lists(max);
+	max_processed -= done;
 
 	/* some tasks may have woken other ones up */
-	if (max_processed > 0 && thread_has_tasks())
+	if (done && max_processed > 0 && thread_has_tasks())
 		goto not_done_yet;
 
  leave:

src/tcp_rules.c

@@ -1257,7 +1257,8 @@ static int tcp_parse_tcp_rep(char **args, int section_type, struct proxy *curpx,
 		}
 
 		/* the following function directly emits the warning */
-		warnif_misplaced_tcp_res_cont(curpx, file, line, args[0], args[1]);
+		if (warnif_misplaced_tcp_res_cont(curpx, file, line, args[0], args[1]))
+			warn++;
 		LIST_APPEND(&curpx->tcp_rep.inspect_rules, &rule->list);
 	}
 	else {
@@ -1377,7 +1378,8 @@ static int tcp_parse_tcp_req(char **args, int section_type, struct proxy *curpx,
 		}
 
 		/* the following function directly emits the warning */
-		warnif_misplaced_tcp_req_cont(curpx, file, line, args[0], args[1]);
+		if (warnif_misplaced_tcp_req_cont(curpx, file, line, args[0], args[1]))
+			warn++;
 		LIST_APPEND(&curpx->tcp_req.inspect_rules, &rule->list);
 	}
 	else if (strcmp(args[1], "connection") == 0) {
@@ -1422,7 +1424,8 @@ static int tcp_parse_tcp_req(char **args, int section_type, struct proxy *curpx,
 		}
 
 		/* the following function directly emits the warning */
-		warnif_misplaced_tcp_req_conn(curpx, file, line, args[0], args[1]);
+		if (warnif_misplaced_tcp_req_conn(curpx, file, line, args[0], args[1]))
+			warn++;
 		LIST_APPEND(&curpx->tcp_req.l4_rules, &rule->list);
 	}
 	else if (strcmp(args[1], "session") == 0) {
@@ -1466,7 +1469,8 @@ static int tcp_parse_tcp_req(char **args, int section_type, struct proxy *curpx,
 		}
 
 		/* the following function directly emits the warning */
-		warnif_misplaced_tcp_req_sess(curpx, file, line, args[0], args[1]);
+		if (warnif_misplaced_tcp_req_sess(curpx, file, line, args[0], args[1]))
+			warn++;
 		LIST_APPEND(&curpx->tcp_req.l5_rules, &rule->list);
 	}
 	else {

src/tcpcheck.c

@@ -40,6 +40,7 @@
 #include <haproxy/check.h>
 #include <haproxy/chunk.h>
 #include <haproxy/connection.h>
+#include <haproxy/dynbuf.h>
 #include <haproxy/errors.h>
 #include <haproxy/global.h>
 #include <haproxy/h1.h>
@@ -1427,9 +1428,15 @@ enum tcpcheck_eval_ret tcpcheck_eval_connect(struct check *check, struct tcpchec
 			check->mux_proto = NULL;
 	}
 	else {
-		proto = s ?
+		if (check->proto)
-		  protocol_lookup(conn->dst->ss_family, s->addr_type.proto_type, s->alt_proto) :
+			proto = check->proto;
-		  protocol_lookup(conn->dst->ss_family, PROTO_TYPE_STREAM, 0);
+		else {
+			if (is_addr(&connect->addr))
+				proto = protocol_lookup(conn->dst->ss_family, PROTO_TYPE_STREAM, 0);
+			else
+				proto = protocol_lookup(conn->dst->ss_family, s->addr_type.proto_type, s->alt_proto);
+
+		}
 	}
 
 	port = 0;
@@ -1659,7 +1666,8 @@ enum tcpcheck_eval_ret tcpcheck_eval_send(struct check *check, struct tcpcheck_r
 		goto out;
 	}
 
-	if (!check_get_buf(check, &check->bo)) {
+  retry:
+	if (!check_get_buf(check, &check->bo, (check->state & CHK_ST_USE_SMALL_BUFF))) {
 		check->state |= CHK_ST_OUT_ALLOC;
 		ret = TCPCHK_EVAL_WAIT;
 		TRACE_STATE("waiting for output buffer allocation", CHK_EV_TCPCHK_SND|CHK_EV_TX_DATA|CHK_EV_TX_BLK, check);
@@ -1679,6 +1687,13 @@ enum tcpcheck_eval_ret tcpcheck_eval_send(struct check *check, struct tcpcheck_r
 	case TCPCHK_SEND_STRING:
 	case TCPCHK_SEND_BINARY:
 		if (istlen(send->data) >= b_size(&check->bo)) {
+			if (b_is_small(&check->bo)) {
+				check->state &= ~CHK_ST_USE_SMALL_BUFF;
+				check_release_buf(check, &check->bo);
+				TRACE_DEVEL("Send fail with small buffer retry with default one", CHK_EV_TCPCHK_SND|CHK_EV_TX_DATA, check);
+				goto retry;
+			}
+
 			chunk_printf(&trash, "tcp-check send : string too large (%u) for buffer size (%u) at step %d",
 				     (unsigned int)istlen(send->data), (unsigned int)b_size(&check->bo),
 				     tcpcheck_get_step_id(check, rule));
@@ -1689,6 +1704,7 @@ enum tcpcheck_eval_ret tcpcheck_eval_send(struct check *check, struct tcpcheck_r
 		b_putist(&check->bo, send->data);
 		break;
 	case TCPCHK_SEND_STRING_LF:
+		BUG_ON(check->state & CHK_ST_USE_SMALL_BUFF);
 		check->bo.data = sess_build_logline(check->sess, NULL, b_orig(&check->bo), b_size(&check->bo), &rule->send.fmt);
 		if (!b_data(&check->bo))
 			goto out;
@@ -1696,7 +1712,8 @@ enum tcpcheck_eval_ret tcpcheck_eval_send(struct check *check, struct tcpcheck_r
 	case TCPCHK_SEND_BINARY_LF: {
 		int len = b_size(&check->bo);
 
-		tmp = alloc_trash_chunk();
+		BUG_ON(check->state & CHK_ST_USE_SMALL_BUFF);
+		tmp = alloc_trash_chunk_sz(len);
 		if (!tmp)
 			goto error_lf;
 		tmp->data = sess_build_logline(check->sess, NULL, b_orig(tmp), b_size(tmp), &rule->send.fmt);
@@ -1713,7 +1730,7 @@ enum tcpcheck_eval_ret tcpcheck_eval_send(struct check *check, struct tcpcheck_r
 		struct ist meth, uri, vsn, clen, body;
 		unsigned int slflags = 0;
 
-		tmp = alloc_trash_chunk();
+		tmp = alloc_trash_chunk_sz(b_size(&check->bo));
 		if (!tmp)
 			goto error_htx;
 
@@ -1838,6 +1855,12 @@ enum tcpcheck_eval_ret tcpcheck_eval_send(struct check *check, struct tcpcheck_r
 		htx_reset(htx);
 		htx_to_buf(htx, &check->bo);
 	}
+	if (b_is_small(&check->bo)) {
+		check->state &= ~CHK_ST_USE_SMALL_BUFF;
+		check_release_buf(check, &check->bo);
+		TRACE_DEVEL("Send fail with small buffer retry with default one", CHK_EV_TCPCHK_SND|CHK_EV_TX_DATA, check);
+		goto retry;
+	}
 	chunk_printf(&trash, "tcp-check send : failed to build HTTP request at step %d",
 		     tcpcheck_get_step_id(check, rule));
 	TRACE_ERROR("failed to build HTTP request", CHK_EV_TCPCHK_SND|CHK_EV_TX_DATA|CHK_EV_TCPCHK_ERR, check);
@@ -1884,7 +1907,7 @@ enum tcpcheck_eval_ret tcpcheck_eval_recv(struct check *check, struct tcpcheck_r
 		goto wait_more_data;
 	}
 
-	if (!check_get_buf(check, &check->bi)) {
+	if (!check_get_buf(check, &check->bi, 0)) {
 		check->state |= CHK_ST_IN_ALLOC;
 		TRACE_STATE("waiting for input buffer allocation", CHK_EV_RX_DATA|CHK_EV_RX_BLK, check);
 		goto wait_more_data;
@@ -4067,6 +4090,8 @@ static int check_proxy_tcpcheck(struct proxy *px)
 		}
 	}
 
+	/* Allow small buffer use by default. All send rules must be compatible */
+	px->tcpcheck_rules.flags |= (global.tune.bufsize_small ? TCPCHK_RULES_MAY_USE_SBUF : 0);
 
 	/* Remove all comment rules. To do so, when a such rule is found, the
 	 * comment is assigned to the following rule(s).
@@ -4096,6 +4121,25 @@ static int check_proxy_tcpcheck(struct proxy *px)
 			ha_free(&comment);
 			break;
 		case TCPCHK_ACT_SEND:
+			/* Disable small buffer use for rules using LF stirngs or too large data */
+			switch (chk->send.type) {
+			case TCPCHK_SEND_STRING:
+			case TCPCHK_SEND_BINARY:
+				if (istlen(chk->send.data) >= global.tune.bufsize_small)
+					px->tcpcheck_rules.flags &= ~TCPCHK_RULES_MAY_USE_SBUF;
+				break;
+			case TCPCHK_SEND_STRING_LF:
+			case TCPCHK_SEND_BINARY_LF:
+				px->tcpcheck_rules.flags &= ~TCPCHK_RULES_MAY_USE_SBUF;
+				break;
+			case TCPCHK_SEND_HTTP:
+				if ((chk->send.http.flags & TCPCHK_SND_HTTP_FL_BODY_FMT) ||
+				    (istlen(chk->send.http.body) >= global.tune.bufsize_small))
+					px->tcpcheck_rules.flags &= ~TCPCHK_RULES_MAY_USE_SBUF;
+			default:
+				break;
+			}
+			__fallthrough;
 		case TCPCHK_ACT_EXPECT:
 			if (!chk->comment && comment)
 				chk->comment = strdup(comment);

src/tools.c

@@ -6057,6 +6057,9 @@ static int dl_collect_libs_cb(struct dl_phdr_info *info, size_t size, void *data
 		/* else it's a VDSO or similar and we're not interested */
 		goto leave;
 
+	if (!fname)
+		goto leave;
+
 	load_file_into_tar(&ctx->storage, &ctx->size, ctx->prefix, fname, NULL, "haproxy-libs-dump");
 
 	/* try to load equivalent debug symbols for absolute paths  */