Just like for the listener, the frontend is session-wide so let's move
it to the session. There are a lot of places which were changed but the
changes are minimal in fact.
There is now a pointer to the session in the stream, which is NULL
for now. The session pool is created as well. Some parts will move
from the stream to the session now.
With HTTP/2, we'll have to support multiplexed streams. A stream is in
fact the largest part of what we currently call a session, it has buffers,
logs, etc.
In order to catch any error, this commit removes any reference to the
struct session and tries to rename most "session" occurrences in function
names to "stream" and "sess" to "strm" when that's related to a session.
The files stream.{c,h} were added and session.{c,h} removed.
The session will be reintroduced later and a few parts of the stream
will progressively be moved overthere. It will more or less contain
only what we need in an embryonic session.
Sample fetch functions and converters will have to change a bit so
that they'll use an L5 (session) instead of what's currently called
"L4" which is in fact L6 for now.
Once all changes are completed, we should see approximately this :
L7 - http_txn
L6 - stream
L5 - session
L4 - connection | applet
There will be at most one http_txn per stream, and a same session will
possibly be referenced by multiple streams. A connection will point to
a session and to a stream. The session will hold all the information
we need to keep even when we don't yet have a stream.
Some more cleanup is needed because some code was already far from
being clean. The server queue management still refers to sessions at
many places while comments talk about connections. This will have to
be cleaned up once we have a server-side connection pool manager.
Stream flags "SN_*" still need to be renamed, it doesn't seem like
any of them will need to move to the session.
This library is designed to emit a zlib-compatible stream with no
memory usage and to favor resource savings over compression ratio.
While zlib requires 256 kB of RAM per compression context (and can only
support 4000 connections per GB of RAM), the stateless compression
offered by libslz does not need to retain buffers between subsequent
calls. In theory this slightly reduces the compression ratio but in
practice it does not have that much of an effect since the zlib
window is limited to 32kB.
Libslz is available at :
http://git.1wt.eu/web?p=libslz.git
It was designed for web compression and provides a lot of savings
over zlib in haproxy. Here are the preliminary results on a single
core of a core2-quad 3.0 GHz in 32-bit for only 300 concurrent
sessions visiting the home page of www.haproxy.org (76 kB) with
the default 16kB buffers :
BW In BW Out BW Saved Ratio memory VSZ/RSS
zlib 237 Mbps 92 Mbps 145 Mbps 2.58 84M / 69M
slz 733 Mbps 380 Mbps 353 Mbps 1.93 5.9M / 4.2M
So while the compression ratio is lower, the bandwidth savings are
much more important due to the significantly lower compression cost
which allows to consume even more data from the servers. In the
example above, zlib became the bottleneck at 24% of the output
bandwidth. Also the difference in memory usage is obvious.
More tests run on a single core of a core i5-3320M, with 500 concurrent
users and the default 16kB buffers :
At 100% CPU (no limit) :
BW In BW Out BW Saved Ratio memory VSZ/RSS hits/s
zlib 480 Mbps 188 Mbps 292 Mbps 2.55 130M / 101M 744
slz 1700 Mbps 810 Mbps 890 Mbps 2.10 23.7M / 9.7M 2382
At 85% CPU (limited) :
BW In BW Out BW Saved Ratio memory VSZ/RSS hits/s
zlib 1240 Mbps 976 Mbps 264 Mbps 1.27 130M / 100M 1738
slz 1600 Mbps 976 Mbps 624 Mbps 1.64 23.7M / 9.7M 2210
The most important benefit really happens when the CPU usage is
limited by "maxcompcpuusage" or the BW limited by "maxcomprate" :
in order to preserve resources, haproxy throttles the compression
ratio until usage is within limits. Since slz is much cheaper, the
average compression ratio is much higher and the input bandwidth
is quite higher for one Gbps output.
Other tests made with some reference files :
BW In BW Out BW Saved Ratio hits/s
daniels.html zlib 1320 Mbps 163 Mbps 1157 Mbps 8.10 1925
slz 3600 Mbps 580 Mbps 3020 Mbps 6.20 5300
tv.com/listing zlib 980 Mbps 124 Mbps 856 Mbps 7.90 310
slz 3300 Mbps 553 Mbps 2747 Mbps 5.97 1100
jquery.min.js zlib 430 Mbps 180 Mbps 250 Mbps 2.39 547
slz 1470 Mbps 764 Mbps 706 Mbps 1.92 1815
bootstrap.min.css zlib 790 Mbps 165 Mbps 625 Mbps 4.79 777
slz 2450 Mbps 650 Mbps 1800 Mbps 3.77 2400
So on top of saving a lot of memory, slz is constantly 2.5-3.5 times
faster than zlib and results in providing more savings for a fixed CPU
usage. For links smaller than 100 Mbps, zlib still provides a better
compression ratio, at the expense of a much higher CPU usage.
Larger input files provide slightly higher bandwidth for both libs, at
the expense of a bit more memory usage for zlib (it converges to 256kB
per connection).
This function used to take a zlib-specific flag as argument to indicate
whether a buffer flush or end of contents was met, let's split it in two
so that we don't depend on zlib anymore.
Thanks to MSIE/IIS, the "deflate" name is ambigous. According to the RFC
it's a zlib-wrapped deflate stream, but IIS used to send only a raw deflate
stream, which is the only format MSIE understands for "deflate". The other
widely used browsers do support both formats. For this reason some people
prefer to emit a raw deflate stream on "deflate" to serve more users even
it that means violating the standards. Haproxy only follows the standard,
so they cannot do this.
This patch makes it possible to have one algorithm name in the configuration
and another one in the protocol. This will make it possible to have a new
configuration token to add a different algorithm so that users can decide if
they want a raw deflate or the standard one.
There's no reason for exporting identity_* nor deflate_*, they're only
used in the same file. Mark them static, it will make it easier to add
other algorithms.
Till now we used to rely on a fixed maximum chunk size. Thanks to last
commit we're now free to adjust the chunk's length before sending the
data, so we don't have to use 6 digits all the time anymore, and if
one wants buffers larger than 16 MB it is now possible.
This class is accessible via the TXN object. It is created only if
the attached proxy have HTTP mode. It contain all the HTTP
manipulation functions:
- req_get_headers
- req_del_header
- req_rep_header
- req_rep_value
- req_add_header
- req_set_header
- req_set_method
- req_set_path
- req_set_query
- req_set_uri
- res_get_headers
- res_del_header
- res_rep_header
- res_rep_value
- res_add_header
- res_set_header
This function is a callback for HTTP actions. This function
creates the replacement string from a build_logline() format
and transform the header.
This patch split this function in two part. With this modification,
the header transformation and the replacement string are separed.
We can now transform the header with another replacement string
source than a build_logline() format.
The first part is the replacement engine. It take a replacement action
number and a replacement string and process the action.
The second part is the function which is called by the 'http-request
action' to replace a request line part. This function makes the
string used as replacement.
This split permits to use the replacement engine in other parts of the
code than the request action. The Lua use it for his own http action.
This will be useful later to state that some listeners have to use
certain decoders (typically an HTTP/2 decoder) regardless of the
regular processing applied to other listeners. For now it simply
defaults to the frontend's default target, and it is used by the
session.
Listerner->timeout is a vestigal thing going back to 2007 or so. It
used to only be used by stats and peers frontends to hold a pointer
to the proxy's client timeout. Now that we use regular frontends, we
don't use it anymore.
Some services such as peers and CLI pre-set the target applet immediately
during accept(), and for this reason they're forced to have a dedicated
accept() function which does not even properly follow everything the regular
one does (eg: sndbuf/rcvbuf/linger/nodelay are not set, etc).
Let's store the default target when known into the frontend's config so that
it's session_accept() which automatically sets it.
The default value is stored in a special struct that describe the
map. This default value is parsed with special parser. This is
useless because HAProxy provides a space to store the default
value (the args) and HAProxy provides also standard parser for
the input types (args again).
This patch remove this special storage and replace it by an argument.
In other way the args of maps are declared as the expected type in
place of strings.
It's now called conn_sock_drain() to make it clear that it only reads
at the sock layer and not at the data layer. The function was too big
to remain inlined and it's used at a few places where size counts.
Currently si_idle_conn_null_cb() has to perform some low-level checks
over the file descriptor and the connection configuration that should
only belong to conn_drain(). Let's move these controls there. The
function now automatically checks for errors and hangups on the file
descriptor for example, and disables recv polling if there's no drain
function at the control layer.
This function is an equivalent to send() which operates over a connection
instead of a file descriptor. It checks that the control layer is ready
and that it's allowed to send. If automatically enables polling if it
cannot send. It simplifies the return checks by returning zero in all
cases where it cannot send so that the caller only has to care about
negative values indicating errors.
This will save callers from having to care about conn->xprt and xprt->shutw.
Note that shutw() takes a second argument indicating whether it's a clean or
a hard shutw. This is used by SSL which tries to close cleanly in most cases.
Here we provide two versions, conn_data_shutw() which performs the clean
close, and conn_data_shutw_hard() which does the unclean one.
This function was not used yet and was only supposed to mark the connection
as shutdown for write. Unfortunately at other places in stream_interface.c,
we're seeing a bit of layering violations with attempts to perform the shutdown
on the fd directly. Let's make this function call shutdown() itself so that
the callers only have to care about the connection.
This struct used to carry only a sample fetch function. Thanks to
lua_pushuserdata(), we don't need to have the Lua engine allocate
that struct for us and we can simply push our pointer onto the stack.
This makes the code more readable by removing several occurrences of
"f->f->". Just like the previous patch, it comes with the nice effect
of saving about 1.3% of performance when fetching samples from Lua.
Since last cleanups, this one was only used to carry a struct channel.
Removing it makes the code a bit cleaner (no more chn->chn) and easier
to follow (no more abstraction for a common type). Interestingly it
happens to also make the Lua code slightly faster (about 1.5%) when
using channels, probably thanks to less pointer dereferences and maybe
the use of lua_pushlightuserdata().
Now that we can get the session from the channel, let's simplify the
prototype of session_alloc_recv_buffer() to only require the channel.
Both the caller and the function are now simplified.
The purpose of these two macros will be to pass via the session to
find the relevant stream interfaces so that we don't need to store
the ->cons nor ->prod pointers anymore. Currently they're only defined
so that all references could be removed.
Note that many places need a second pass of clean up so that we don't
have any chn_prod(&s->req) anymore and only &s->si[0] instead, and
conversely for the 3 other cases.
At a few places we need to find one stream interface from the other one.
Instead of passing via the channel, we simply use the session as an
intermediary, which simply results in applying an offset to the pointer.
We go back to the session to get the owner. Here again it's very easy
and is just a matter of relative offsets. Since the owner always exists
and always points to the session's task, we can remove some unneeded
tests.
In order to plan removal of si->ib / si->ob, we now check the side of the
stream interface and find the session, then the requested channel. In
practice it's just an offset applied to the pointer based on the flag.
This new flag "SI_FL_ISBACK" is set only on the back SI and is cleared
on the front SI. That way it's possible only by looking at the SI to
know what side it is.
We'll soon remove direct references to the channels from the stream
interface since everything belongs to the same session, so let's
first not dereference si->ib / si->ob anymore and use macros instead.
The channels were pointers to outside structs and this is not needed
anymore since the buffers have moved, but this complicates operations.
Move them back into the session so that both channels and stream interfaces
are always allocated for a session. Some places (some early sample fetch
functions) used to validate that a channel was NULL prior to dereferencing
it. Now instead we check if chn->buf is NULL and we force it to remain NULL
until the channel is initialized.
In some cases we don't want to known if a fetch or converter
fails. We just want a valid string. After this patch, we
have two sets of fetches and two sets of converters. There are:
txn.f, txn.sf, txn.c, txn.sc. The version prefixed by 's' always
returns strings for any type, and returns an empty string in the
error case or when the data are not available. This is particularly
useful when manipulating headers or cookies.
This patch implements a wrapper to give access to the converters
in the Lua code. The converters are used with the transaction.
The automatically created function are prefixed by "conv_".
HAProxy proposes many sample fetches. It is possible that the
automatic registration of the sample fetches causes a collision
with an existing Lua function. This patch sets a namespace for
the sample fetches.
The function buffer_contig_space() returns the contiguous space avalaible
to add data (at the end of the input side) while the function
hlua_channel_send_yield() needs to insert data starting at p. Here we
introduce a new function bi_space_for_replace() which returns the amount
of space that can be inserted at the head of the input side with one of
the buffer_replace* functions.
This patch proposes a function that returns the space avalaible after buf->p.
If we are writing in the request buffer, we are not waked up
when the data are forwarded because it is useles. The request
analyzers are waked up only when data is incoming. So, if the
request buffer is full, we set the WAKE_ON_WRITE flag.
Before this patch, each yield in a Lua action set a flags to be
waked up when some activity were detected on the response channel.
This behavior causes loop in the analyzer process.
This patch set the wake up on response buffer activity only if we
really want to be waked up on this activity.
This flag indicate that the current yield is returned by the Lua
execution task control. If this flag is set, the current task may
quit but will be set in the run queue to be re-executed immediatly.
This patch modify the "hlua_yieldk()" function, it adds an argument
that contain a field containing yield options.
This is used to ensure that the task doesn't become a zombie
when the Lua returns a yield. The yield wrapper ensure that an
timer used for waking up the task will be set.
The timer is reseted to TICK_ETERNITY if the Lua execution is
done.
This first patch permits to cofigure the Lua execution exipiration.
This expiration is configured but it is not yet avalaible, it will
be add in a future patch.
In the future, the lua execution must return scheduling informations.
We want more than one flag, so I convert an integer used with an
enum into an interer used as bitfield.
The channel class permits manipulation of channels. A channel is
an FIFO buffer between the client and the server. This class provides
function to read, write, forward, destroy and alter data between
the input and the ouput of the buffer.
This patch adds the TCP I/O functionnality. The class implemented
provides the same functions than the "lua socket" project. This
make network compatibility with another LUA project. The documentation
is located here:
http://w3.impa.br/~diego/software/luasocket/tcp.html
This version of sleep is based on a coroutine. A sleeping
task is started and a signal is registered. This sleep version
must disapear to be replaced by a version using the internal
timers.
This patch adds the browsing of all the HAProxy fetches and
create associated LUA functions. The HAProxy internal fetches
can be used in LUA trough the class "TXN".
Note that the symbols "-", "+" and "." in the name of current
sample fetch are rewrited as "_" in LUA because ".", "-" and "+"
are operators.
This class of functions permit to access to all the functions
associated with the transaction like http header, HAProxy internal
fetches, etc ...
This patch puts the skeleton of this class. The class will be
enhanced later.
This system permits to execute some lua function after than HAProxy
complete his initialisation. These functions are executed between
the end of the configuration parsing and check and the begin of the
scheduler.
This system permits to send signals between lua tasks. A main lua stack can
register the signal in a coprocess. When the coprocess finish his job, it
send a signal, and the associated task is wakes. If the main lua execution
stack stop (with or without errors), the list or pending signals is purged.
This is the first step of the lua integration. We add the useful
files in the HAProxy project. These files contains the main
includes, the Makefile options and empty initialisation function.
Is is the LUA skeleton.
We now have functions to retrieve one block and one line from
either the input or the output part of a buffer. They return
up to two (pointer,length) values in case the buffer wraps.
Later, the processing of some actions needs to be interrupted and resumed
later. This patch permit to resume the actions. The actions that needs
to run with the resume mode are not yet avalaible. It will be soon with
Lua patches. So the code added by this patch is untestable for the moment.
The list of "tcp_exec_req_rules" cannot resme because is called by the
unresumable function "accept_session".
This patch introduces an action keyword registration system for TCP
rulesets similar to what is available for HTTP rulesets. This sytem
will be useful with lua.
These modifications are done for resolving cross-dependent
includes in the upcoming LUA code.
<proto/channel.h> misses <types/channel.h>.
<types/acl.h> doesn't use <types/session.h> because the session
is already declared in the file as undefined pointer.
appsession.c misses <unistd.h> to use "write()".
Declare undefined pointer "struct session" for <types/proxy.h>
and <types/queue.h>. These includes dont need the detail of this
struct.
The functions "val_payload_lv" and "val_hdr" are useful with
lua. The lua automatic binding for sample fetchs needs to
compare check functions.
The "arg_type_names" permit to display error messages.
Some usages of the converters need to know the attached session. The Lua
needs the session for retrieving his running context. This patch adds
the "session" as an argument of the converters prototype.
This behavior is already existing for the "WAIT_HTTP" analyzer,
this patch just extends the system to any analyzer that would
be waked up on response activity.
When the destination IP is dynamically set, we can't use the "target"
to define the proto. This patch ensures that we always use the protocol
associated with the address family. The proto field was removed from
the server and check structs.
Actually, HAProxy uses the function "process_runnable_tasks" and
"wake_expired_tasks" to get the next task which can expires.
If a task is added with "task_schedule" or other method during
the execution of an other task, the expiration of this new task
is not taken into account, and the execution of this task can be
too late.
Actualy, HAProxy seems to be no sensitive to this bug.
This fix moves the call to process_runnable_tasks() before the timeout
calculation and ensures that all wakeups are processed together. Only
wake_expired_tasks() needs to return a timeout now.
Until now, the TLS ticket keys couldn't have been configured and
shared between multiple instances or multiple servers running HAproxy.
The result was that if a request got a TLS ticket from one instance/server
and it hits another one afterwards, it will have to go through the full
SSL handshake and negotation.
This patch enables adding a ticket file to the bind line, which will be
used for all SSL contexts created from that bind line. We can use the
same file on all instances or servers to mitigate this issue and have
consistent TLS tickets assigned. Clients will no longer have to negotiate
every time they change the handling process.
Signed-off-by: Nenad Merdanovic <nmerdan@anine.io>
As found by Thierry Fournier, if a task manages to kill another one and
if this other task is the next one in the run queue, we can do whatever
including crashing, because the scheduler restarts from the saved next
task. For now, there is no such concept of a task killing another one,
but with Lua it will come.
A solution consists in always performing the lookup of the first task in
the scheduler's loop, but it's expensive and costs around 2% of the
performance.
Another solution consists in keeping a global next run queue node and
ensuring that when this task gets removed, it updates this pointer to
the next one. This allows to simplify the code a bit and in the end to
slightly increase the performance (0.3-0.5%). The mechanism might still
be usable if we later migrate to a multi-threaded scheduler.
The patch "MEDIUM: args: increase arg type to 5 bits and limit arg count
to 5" (dbc79d0a) increased the number of types supported, but forgot to
remove the ARG6/ARG7 macros.
This option disables SSL session reuse when SSL is used to communicate with
the server. It will force the server to perform a full handshake for every
new connection. It's probably only useful for benchmarking, troubleshooting,
and for paranoid users.
This patch adds a new option which allows configuration of the maximum
log level of messages for which email alerts will be sent.
The default is alert which is more restrictive than
the current code which sends email alerts for all priorities.
That behaviour may be configured using the new configuration
option to set the maximum level to notice or greater.
email-alert level notice
Signed-off-by: Simon Horman <horms@verge.net.au>
On Linux since 2.6.37, it's possible to set the socket timeout for
pending outgoing data, with an accuracy of 1 millisecond. This is
pretty handy to deal with dead connections to clients and or servers.
For now we only implement it on the frontend side (bind line) so
that when a client disappears from the net, we're able to quickly
get rid of its connection and possibly release a server connection.
This can be useful with long-lived connections where an application
level timeout is not suited because long pauses are expected (remote
terminals, connection pools, etc).
Thanks to Thijs Houtenbos and John Eckersberg for the suggestion.
This currently does nothing beyond parsing the configuration
and storing in the proxy as there is no implementation of email alerts.
Signed-off-by: Simon Horman <horms@verge.net.au>
As mailer and mailers structures and allow parsing of
a mailers section into those structures.
These structures will subsequently be freed as it is
not yet possible to use reference them in the configuration.
Signed-off-by: Simon Horman <horms@verge.net.au>
The motivation for this is to make checks more independent of each
other to allow further reuse of their infrastructure.
For nowserver->check and server->agent still always use the same values
for the addr and proto fields so this patch should not introduce any
behavioural changes.
Signed-off-by: Simon Horman <horms@verge.net.au>
Refactor init_check so that an error string is returned
rather than alerts being printed by it. Also
init_check to checks.c and provide a prototype to allow
it to be used from multiple C files.
Signed-off-by: Simon Horman <horms@verge.net.au>
This commit implements the following new actions :
- "set-method" rewrites the request method with the result of the
evaluation of format string <fmt>. There should be very few valid reasons
for having to do so as this is more likely to break something than to fix
it.
- "set-path" rewrites the request path with the result of the evaluation of
format string <fmt>. The query string, if any, is left intact. If a
scheme and authority is found before the path, they are left intact as
well. If the request doesn't have a path ("*"), this one is replaced with
the format. This can be used to prepend a directory component in front of
a path for example. See also "set-query" and "set-uri".
Example :
# prepend the host name before the path
http-request set-path /%[hdr(host)]%[path]
- "set-query" rewrites the request's query string which appears after the
first question mark ("?") with the result of the evaluation of format
string <fmt>. The part prior to the question mark is left intact. If the
request doesn't contain a question mark and the new value is not empty,
then one is added at the end of the URI, followed by the new value. If
a question mark was present, it will never be removed even if the value
is empty. This can be used to add or remove parameters from the query
string. See also "set-query" and "set-uri".
Example :
# replace "%3D" with "=" in the query string
http-request set-query %[query,regsub(%3D,=,g)]
- "set-uri" rewrites the request URI with the result of the evaluation of
format string <fmt>. The scheme, authority, path and query string are all
replaced at once. This can be used to rewrite hosts in front of proxies,
or to perform complex modifications to the URI such as moving parts
between the path and the query string. See also "set-path" and
"set-query".
All of them are handled by the same parser and the same exec function,
which is why they're merged all together. For once, instead of adding
even more entries to the huge switch/case, we used the new facility to
register action keywords. A number of the existing ones should probably
move there as well.
This function (and its sister regex_exec_match2()) abstract the regex
execution but make it impossible to pass flags to the regex engine.
Currently we don't use them but we'll need to support REG_NOTBOL soon
(to indicate that we're not at the beginning of a line). So let's add
support for this flag and update the API accordingly.
This one will be used when a regex is expected. It is automatically
resolved after the parsing and compiled into a regex. Some optional
flags are supported in the type-specific flags that should be set by
the optional arg checker. One is used during the regex compilation :
ARGF_REG_ICASE to ignore case.
These flags are meant to be used by arg checkers to pass out-of-band
information related to some args. A typical use is to indicate how a
regex is expected to be compiled/matched based on other arguments.
These flags are initialized to zero by default and it is up to the args
checkers to set them if needed.
We'll soon need to add new argument types, and we don't use the current
limit of 7 arguments, so let's increase the arg type size to 5 bits and
reduce the arg count to 5 (3 max are used today).
This is in order to add new types. This patch does not change anything
else. Two remaining (harmless) occurrences of a count of 8 instead of 7
were fixed by this patch : empty_arg_list[] and the for() loop counting
args.
This function will be used to perform CRC32 computations. This one wa
loosely inspired from crc32b found here, and focuses on size and speed
at the same time :
http://www.hackersdelight.org/hdcodetxt/crc.c.txt
Much faster table-based versions exist but are pointless for our usage
here, this hash already sustains gigabit speed which is far faster than
what we'd ever need. Better preserve the CPU's cache instead.
If a memory size limit is enforced using "-n" on the command line and
one or both of maxconn / maxsslconn are not set, instead of using the
build-time values, haproxy now computes the number of sessions that can
be allocated depending on a number of parameters among which :
- global.maxconn (if set)
- global.maxsslconn (if set)
- maxzlibmem
- tune.ssl.cachesize
- presence of SSL in at least one frontend (bind lines)
- presence of SSL in at least one backend (server lines)
- tune.bufsize
- tune.cookie_len
The purpose is to ensure that not haproxy will not run out of memory
when maxing out all parameters. If neither maxconn nor maxsslconn are
used, it will consider that 100% of the sessions involve SSL on sides
where it's supported. That means that it will typically optimize maxconn
for SSL offloading or SSL bridging on all connections. This generally
means that the simple act of enabling SSL in a frontend or in a backend
will significantly reduce the global maxconn but in exchange of that, it
will guarantee that it will not fail.
All metrics may be enforced using #defines to accomodate variations in
SSL libraries or various allocation sizes.
An SSL connection takes some memory when it exists and during handshakes.
We measured up to 16kB for an established endpoint, and up to 76 extra kB
during a handshake. The SSL layer stores these values into the global
struct during initialization. If other SSL libs are used, it's easy to
change these values. Anyway they'll only be used as gross estimates in
order to guess the max number of SSL conns that can be established when
memory is constrained and the limit is not set.
We'll need to know the number of SSL connections, their use and their
cost soon. In order to avoid getting tons of ifdefs everywhere, always
export SSL information in the global section. We add two flags to know
whether or not SSL is used in a frontend and in a backend.
It applies to the channel and it doesn't erase outgoing data, only
pending unread data, which is strictly equivalent to what recv()
does with MSG_TRUNC, so that new name is more accurate and intuitive.
This name more accurately reminds that it applies to a channel and not
to a buffer, and that what is returned may be used as a max number of
bytes to pass to recv().
This applies to the channel, not the buffer, so let's fix this name.
Warning, the function's name happens to be the same as the old one
which was mistakenly used during 1.5.
This function's name was poorly chosen and is confusing to the point of
being suspiciously used at some places. The operations it does always
consider the ability to forward pending input data before receiving new
data. This is not obvious at all, especially at some places where it was
used when consuming outgoing data to know if the buffer has any chance
to ever get the missing data. The code needs to be re-audited with that
in mind. Care must be taken with existing code since the polarity of the
function was switched with the renaming.
channel_reserved is confusingly named. It is used to know whether or
not the rewrite area is left intact for situations where we want to
ensure we can use it before proceeding. Let's rename it to fix this
confusion.
Option http-send-name-header is still hurting. If a POST request has to be
redispatched when this option is used, and the next server's name is larger
than the initial one, and the POST body fills the buffer, it becomes
impossible to rewrite the server's name in the buffer when redispatching.
In 1.4, this is worse, the process may crash because of a negative size
computation for the memmove().
The only solution to fix this is to refrain from eating the reserve before
we're certain that we won't modify the buffer anymore. And the condition for
that is that the connection is established.
This patch introduces "channel_may_send()" which helps to detect whether it's
safe to eat the reserve or not. This condition is used by channel_in_transit()
introduced by recent patches.
This patch series must be backported into 1.5, and a simpler version must be
backported into 1.4 where fixing the bug is much easier since there were no
channels by then. Note that in 1.4 the severity is major.
This function returns the amount of bytes in transit in a channel's buffer,
which is the amount of outgoing data plus the amount of incoming data bound
to the forward limit.
We know that all incoming data are going to be purged if to_forward
is greater than them, not only if greater than the buffer size. This
buf has no direct impact on this version, but it participates to some
bugs affecting http-send-name-header since 1.4. This fix will have to
be backported down to 1.4 albeit in a different form.
The buffer_max_len() function is subject to an integer overflow in this
calculus :
int ret = global.tune.maxrewrite - chn->to_forward - chn->buf->o;
- chn->to_forward may be up to 2^31 - 1
- chn->buf->o may be up to chn->buf->size
- global.tune.maxrewrite is by definition smaller than chn->buf->size
Thus here we can subtract (2^31 + buf->o) (highly negative) from something
slightly positive, and result in ret being larger than expected.
Fortunately in 1.5 and 1.6, this is only used by bi_avail() which itself
is used by applets which do not set high values for to_forward so this
problem does not happen there. However in 1.4 the equivalent computation
was used to limit the size of a read and can result in a read overflow
when combined with the nasty http-send-name-header feature.
This fix must be backported to 1.5 and 1.4.
Some users reported that the default max hostname length of 32 is too
short in some environments. This patch does two things :
- it relies on the system's max hostname length as found in MAXHOSTNAMELEN
if it is set. This is the most logical thing to do as the system libs
generally present the appropriate value supported by the system. This
value is 64 on Linux and 256 on Solaris, to give a few examples.
- otherwise it defaults to 64
It is still possible to override this value by defining MAX_HOSTNAME_LEN at
build time. After some observation time, this patch may be backported to
1.5 if it does not cause any build issue, as it is harmless and may help
some users.
This is equivalent to what was done in commit 48936af ("[MINOR] log:
ability to override the syslog tag") but this time instead of doing
this globally, it does it per proxy. The purpose is to be able to use
a separate log tag for various proxies (eg: make it easier to route
log messages depending on the customer).
Since commit 3dd6a25 ("MINOR: stream-int: retrieve session pointer from
stream-int"), we can get the session from the task, so let's get rid of
this less obvious function.
commit 9ede66b0 introduced an environment variable (HAPROXY_SERVER_CURCONN) that
was supposed to be dynamically updated, but it was set only once, during its
initialization.
Most of the code provided in this previous patch has been rewritten in order to
easily update the environment variables without reallocating memory during each
check.
Now, HAPROXY_SERVER_CURCONN will contain the current number of connections on
the server at the time of the check.
bi_swpbuf() swaps the buffer passed in argument with the one attached to
the channel, but only if this last one is empty. The idea is to avoid a
copy when buffers can simply be swapped.
This setting is used to limit memory usage without causing the alloc
failures caused by "-m". Unexpectedly, tests have shown a performance
boost of up to about 18% on HTTP traffic when limiting the number of
buffers to about 10% of the amount of concurrent connections.
tune.buffers.limit <number>
Sets a hard limit on the number of buffers which may be allocated per process.
The default value is zero which means unlimited. The minimum non-zero value
will always be greater than "tune.buffers.reserve" and should ideally always
be about twice as large. Forcing this value can be particularly useful to
limit the amount of memory a process may take, while retaining a sane
behaviour. When this limit is reached, sessions which need a buffer wait for
another one to be released by another session. Since buffers are dynamically
allocated and released, the waiting time is very short and not perceptible
provided that limits remain reasonable. In fact sometimes reducing the limit
may even increase performance by increasing the CPU cache's efficiency. Tests
have shown good results on average HTTP traffic with a limit to 1/10 of the
expected global maxconn setting, which also significantly reduces memory
usage. The memory savings come from the fact that a number of connections
will not allocate 2*tune.bufsize. It is best not to touch this value unless
advised to do so by an haproxy core developer.
We've already experimented with three wake up algorithms when releasing
buffers : the first naive one used to wake up far too many sessions,
causing many of them not to get any buffer. The second approach which
was still in use prior to this patch consisted in waking up either 1
or 2 sessions depending on the number of FDs we had released. And this
was still inaccurate. The third one tried to cover the accuracy issues
of the second and took into consideration the number of FDs the sessions
would be willing to use, but most of the time we ended up waking up too
many of them for nothing, or deadlocking by lack of buffers.
This patch completely removes the need to allocate two buffers at once.
Instead it splits allocations into critical and non-critical ones and
implements a reserve in the pool for this. The deadlock situation happens
when all buffers are be allocated for requests pending in a maxconn-limited
server queue, because then there's no more way to allocate buffers for
responses, and these responses are critical to release the servers's
connection in order to release the pending requests. In fact maxconn on
a server creates a dependence between sessions and particularly between
oldest session's responses and latest session's requests. Thus, it is
mandatory to get a free buffer for a response in order to release a
server connection which will permit to release a request buffer.
Since we definitely have non-symmetrical buffers, we need to implement
this logic in the buffer allocation mechanism. What this commit does is
implement a reserve of buffers which can only be allocated for responses
and that will never be allocated for requests. This is made possible by
the requester indicating how much margin it wants to leave after the
allocation succeeds. Thus it is a cooperative allocation mechanism : the
requester (process_session() in general) prefers not to get a buffer in
order to respect other's need for response buffers. The session management
code always knows if a buffer will be used for requests or responses, so
that is not difficult :
- either there's an applet on the initiator side and we really need
the request buffer (since currently the applet is called in the
context of the session)
- or we have a connection and we really need the response buffer (in
order to support building and sending an error message back)
This reserve ensures that we don't take all allocatable buffers for
requests waiting in a queue. The downside is that all the extra buffers
are really allocated to ensure they can be allocated. But with small
values it is not an issue.
With this change, we don't observe any more deadlocks even when running
with maxconn 1 on a server under severely constrained memory conditions.
The code becomes a bit tricky, it relies on the scheduler's run queue to
estimate how many sessions are already expected to run so that it doesn't
wake up everyone with too few resources. A better solution would probably
consist in having two queues, one for urgent requests and one for normal
requests. A failed allocation for a session dealing with an error, a
connection event, or the need for a response (or request when there's an
applet on the left) would go to the urgent request queue, while other
requests would go to the other queue. Urgent requests would be served
from 1 entry in the pool, while the regular ones would be served only
according to the reserve. Despite not yet having this, it works
remarkably well.
This mechanism is quite efficient, we don't perform too many wake up calls
anymore. For 1 million sessions elapsed during massive memory contention,
we observe about 4.5M calls to process_session() compared to 4.0M without
memory constraints. Previously we used to observe up to 16M calls, which
rougly means 12M failures.
During a test run under high memory constraints (limit enforced to 27 MB
instead of the 58 MB normally needed), performance used to drop by 53% prior
to this patch. Now with this patch instead it *increases* by about 1.5%.
The best effect of this change is that by limiting the memory usage to about
2/3 to 3/4 of what is needed by default, it's possible to increase performance
by up to about 18% mainly due to the fact that pools are reused more often
and remain hot in the CPU cache (observed on regular HTTP traffic with 20k
objects, buffers.limit = maxconn/10, buffers.reserve = limit/2).
Below is an example of scenario which used to cause a deadlock previously :
- connection is received
- two buffers are allocated in process_session() then released
- one is allocated when receiving an HTTP request
- the second buffer is allocated then released in process_session()
for request parsing then connection establishment.
- poll() says we can send, so the request buffer is sent and released
- process session gets notified that the connection is now established
and allocates two buffers then releases them
- all other sessions do the same till one cannot get the request buffer
without hitting the margin
- and now the server responds. stream_interface allocates the response
buffer and manages to get it since it's higher priority being for a
response.
- but process_session() cannot allocate the request buffer anymore
=> We could end up with all buffers used by responses so that none may
be allocated for a request in process_session().
When the applet processing leaves the session context, the test will have
to be changed so that we always allocate a response buffer regardless of
the left side (eg: H2->H1 gateway). A final improvement would consists in
being able to only retry the failed I/O operation without waking up a
task, but to date all experiments to achieve this have proven not to be
reliable enough.
When a session_alloc_buffers() fails to allocate one or two buffers,
it subscribes the session to buffer_wq, and waits for another session
to release buffers. It's then removed from the queue and woken up with
TASK_WAKE_RES, and can attempt its allocation again.
We decide to try to wake as many waiters as we release buffers so
that if we release 2 and two waiters need only once, they both have
their chance. We must never come to the situation where we don't wake
enough tasks up.
It's common to release buffers after the completion of an I/O callback,
which can happen even if the I/O could not be performed due to half a
failure on memory allocation. In this situation, we don't want to move
out of the wait queue the session that was just added, otherwise it
will never get any buffer. Thus, we only force ourselves out of the
queue when freeing the session.
Note: at the moment, since session_alloc_buffers() is not used, no task
is subscribed to the wait queue.
This patch introduces session_alloc_recv_buffer(), session_alloc_buffers()
and session_release_buffers() whose purpose will be to allocate missing
buffers and release unneeded ones around the process_session() and during
I/O operations.
I/O callbacks only need a single buffer for recv operations and none
for send. However we still want to ensure that we don't pick the last
buffer. That's what session_alloc_recv_buffer() is for.
This allocator is atomic in that it always ensures we can get 2 buffers
or fails. Here, if any of the buffers is not ready and cannot be
allocated, the operation is cancelled. The purpose is to guarantee that
we don't enter into the deadlock where all buffers are allocated by the
same size of all sessions.
A queue will have to be implemented for failed allocations. For now
they're just reported as failures.
This function is used to allocate a buffer and ensure that we leave
some margin after it in the pool. The function is not obvious. While
we allocate only one buffer, we want to ensure that at least two remain
available after our allocation. The purpose is to ensure we'll never
enter a deadlock where all sessions allocate exactly one buffer, and
none of them will be able to allocate the second buffer needed to build
a response in order to release the first one.
We also take care of remaining fast in the the fast path by first
checking whether or not there is enough margin, in which case we only
rely on b_alloc_fast() which is guaranteed to succeed. Otherwise we
take the slow path using pool_refill_alloc().
This function allocates a buffer and replaces *buf with this buffer. If
no memory is available, &buf_wanted is used instead. No control is made
to check if *buf already pointed to another buffer. The allocated buffer
is returned, or NULL in case no memory is available. The difference with
b_alloc() is that this function only picks from the pool and never calls
malloc(), so it can fail even if some memory is available. It is the
caller's job to refill the buffer pool if needed.
Till now we'd consider a buffer full even if it had size==0 due to pointing
to buf.size. Now we change this : if buf_wanted is present, it means that we
have already tried to allocate a buffer but failed. Thus the buffer must be
considered full so that we stop trying to poll for reads on it. Otherwise if
it's empty, it's buf_empty and we report !full since we may allocate it on
the fly.
Doing so ensures that even when no memory is available, we leave the
channel in a sane condition. There's a special case in proto_http.c
regarding the compression, we simply pre-allocate the tmpbuf to point
to the dummy buffer. Not reusing &buf_empty for this allows the rest
of the code to differenciate an empty buffer that's not used from an
empty buffer that results from a failed allocation which has the same
semantics as a buffer full.
Channels are now created with a valid pointer to a buffer before the
buffer is allocated. This buffer is a global one called "buf_empty" and
of size zero. Thus it prevents any activity from being performed on
the buffer and still ensures that chn->buf may always be dereferenced.
b_free() also resets the buffer to &buf_empty, and was split into
b_drop() which does not reset the buffer.
We don't call pool_free2(pool2_buffers) anymore, we only call b_free()
to do the job. This ensures that we can start to centralize the releasing
of buffers.
b_alloc() now allocates a buffer and initializes it to the size specified
in the pool minus the size of the struct buffer itself. This ensures that
callers do not need to care about buffer details anymore. Also this never
applies memory poisonning, which is slow and useless on buffers.
We'll soon need to be able to switch buffers without touching the
channel, so let's move buffer initialization out of channel_init().
We had the same in compressoin.c.
Till now this function would only allocate one entry at a time. But with
dynamic buffers we'll like to allocate the number of missing entries to
properly refill the pool.
Let's modify it to take a minimum amount of available entries. This means
that when we know we need at least a number of available entries, we can
ask to allocate all of them at once. It also ensures that we don't move
the pointers back and forth between the caller and the pool, and that we
don't call pool_gc2() for each failed malloc. Instead, it's called only
once and the malloc is only allowed to fail once.
pool_alloc2() used to pick the entry from the pool, fall back to
pool_refill_alloc(), and to perform the poisonning itself, which
pool_refill_alloc() was also doing. While this led to optimal
code size, it imposes memory poisonning on the buffers as well,
which is extremely slow on large buffers.
This patch cuts the allocator in 3 layers :
- a layer to pick the first entry from the pool without falling back to
pool_refill_alloc() : pool_get_first()
- a layer to allocate a dirty area by falling back to pool_refill_alloc()
but never performing the poisonning : pool_alloc_dirty()
- pool_alloc2() which calls the latter and optionally poisons the area
No functional changes were made.
Remove the code dealing with the old dual-linked lists imported from
librt that has remained unused for the last 8 years. Now everything
uses the linux-like circular lists instead.
Since commit 656c5fa7e8 ("BUILD: ssl: disable OCSP when using
boringssl) the OCSP code is bypassed when OPENSSL_IS_BORINGSSL
is defined. The correct thing to do here is to use OPENSSL_NO_OCSP
instead, which is defined for this exact purpose in
openssl/opensslfeatures.h.
This makes haproxy forward compatible if boringssl ever introduces
full OCSP support with the additional benefit that it links fine
against a OCSP-disabled openssl.
Signed-off-by: Lukas Tribus <luky-37@hotmail.com>
Till now, when memory poisonning was enabled, it used to be done only
after a calloc(). But sometimes it's not enough to detect unexpected
sharing, so let's ensure that we now poison every allocation once it's
in place. Note that enabling poisonning significantly hurts performance
(it can typically half the overall performance).
A memory optimization can use the same pattern expression for many
equal pattern list (same parse method, index method and index_smp
method).
The pattern expression is returned by "pattern_new_expr", but this
function dont indicate if the returned pattern is already in use.
So, the caller function reload the list of patterns in addition with
the existing patterns. This behavior is not a problem with tree indexed
pattern, but it grows the lists indexed patterns.
This fix add a "reuse" flag in return of the function "pattern_new_expr".
If the flag is set, I suppose that the patterns are already loaded.
This fix must be backported into 1.5.
