erl_driver
An Erlang driver is a library containing a set of native driver callback functions that the Erlang VM calls when certain events occur. There may be multiple instances of a driver, each instance is associated with an Erlang port.
Warning!
Use this functionality with extreme care!
A driver callback is executed as a direct extension of the native code of the VM. Execution is not made in a safe environment. The VM can not provide the same services as provided when executing Erlang code, such as preemptive scheduling or memory protection. If the driver callback function doesn't behave well, the whole VM will misbehave.
A driver callback that crash will crash the whole VM.
An erroneously implemented driver callback might cause a VM internal state inconsistency which may cause a crash of the VM, or miscellaneous misbehaviors of the VM at any point after the call to the driver callback.
A driver callback that do lengthy work before returning will degrade responsiveness of the VM, and may cause miscellaneous strange behaviors. Such strange behaviors include, but are not limited to, extreme memory usage, and bad load balancing between schedulers. Strange behaviors that might occur due to lengthy work may also vary between OTP releases.
As of erts version 5.5.3 the driver interface has been extended (see extended marker). The extended interface introduce version management, the possibility to pass capability flags (see driver flags) to the runtime system at driver initialization, and some new driver API functions.
Note!
As of erts version 5.9 old drivers have to be recompiled and have to use the extended interface. They also have to be adjusted to the 64-bit capable driver interface.
The driver calls back to the emulator, using the API
functions declared in erl_driver.h
. They are used for
outputting data from the driver, using timers, etc.
Each driver instance is associated with a port. Every port
has a port owner process. Communication with the port is normally
done through the port owner process. Most of the functions take
the port
handle as an argument. This identifies the driver
instance. Note that this port handle must be stored by the driver,
it is not given when the driver is called from the emulator (see
driver_entry).
Some of the functions take a parameter of type
ErlDrvBinary
, a driver binary. It should be both
allocated and freed by the caller. Using a binary directly avoids
one extra copying of data.
Many of the output functions have a "header buffer", with
hbuf
and hlen
parameters. This buffer is sent as a
list before the binary (or list, depending on port mode) that is
sent. This is convenient when matching on messages received from
the port. (Although in the latest versions of Erlang, there is
the binary syntax, that enables you to match on the beginning of
a binary.)
In the runtime system with SMP support, drivers are locked either
on driver level or port level (driver instance level). By default
driver level locking will be used, i.e., only one emulator thread
will execute code in the driver at a time. If port level locking
is used, multiple emulator threads may execute code in the driver
at the same time. There will only be one thread at a time calling
driver call-backs corresponding to the same port, though. In order
to enable port level locking set the ERL_DRV_FLAG_USE_PORT_LOCKING
driver flag in
the driver_entry
used by the driver. When port level locking is used it is the
responsibility of the driver writer to synchronize all accesses
to data shared by the ports (driver instances).
Most drivers written before the runtime system with SMP support existed will be able to run in the runtime system with SMP support without being rewritten if driver level locking is used.
Note!
It is assumed that drivers do not access other drivers. If drivers should access each other they have to provide their own mechanism for thread safe synchronization. Such "inter driver communication" is strongly discouraged.
Previously, in the runtime system without SMP support, specific driver call-backs were always called from the same thread. This is not the case in the runtime system with SMP support. Regardless of locking scheme used, calls to driver call-backs may be made from different threads, e.g., two consecutive calls to exactly the same call-back for exactly the same port may be made from two different threads. This will for most drivers not be a problem, but it might. Drivers that depend on all call-backs being called in the same thread, have to be rewritten before being used in the runtime system with SMP support.
Note!
Regardless of locking scheme used, calls to driver call-backs may be made from different threads.
Most functions in this API are not thread-safe, i.e., they may not be called from an arbitrary thread. Functions that are not documented as thread-safe may only be called from driver call-backs or function calls descending from a driver call-back call. Note that driver call-backs may be called from different threads. This, however, is not a problem for any function in this API, since the emulator has control over these threads.
Warning!
Functions not explicitly documented as thread safe are not thread safe. Also note that some functions are only thread safe when used in a runtime system with SMP support.
A function not explicitly documented as thread safe may at some point in time have a thread safe implementation in the runtime system. Such an implementation may however change to a thread unsafe implementation at any time without any notice at all.
Only use functions explicitly documented as thread safe from arbitrary threads.
As mentioned in the warning text at
the beginning of this document it is of vital importance that a driver callback
does return relatively fast. It is hard to give an exact maximum amount
of time that a driver callback is allowed to work, but as a rule of thumb
a well behaving driver callback should return before a millisecond has
passed. This can be achieved using different approaches.
If you have full control over the code that are to execute in the driver
callback, the best approach is to divide the work into multiple chunks of
work and trigger multiple calls to the
timeout callback using
zero timeouts. The
erl_drv_consume_timeslice()
function can be useful in order to determine when to trigger such
timeout callback calls. It might, however, not always be possible to
implement it this way, e.g. when calling third party libraries. In this
case you typically want to dispatch the work to another thread.
Information about thread primitives can be found below.
FUNCTIONALITY
All functions that a driver needs to do with Erlang are performed through driver API functions. There are functions for the following functionality:
Every driver instance has an associated queue. This queue is a
SysIOVec
that works as a buffer. It's mostly used for
the driver to buffer data that should be written to a device,
it is a byte stream. If the port owner process closes the
driver, and the queue is not empty, the driver will not be
closed. This enables the driver to flush its buffers before
closing.
The queue can be manipulated from arbitrary threads if a port data lock is used. See documentation of the ErlDrvPDL type for more information.
open_port/2
. The vector function and the
function taking a driver binary are faster, because they avoid
copying the data buffer. There is also a fast way of sending
terms from the driver, without going through the binary term
format.A POSIX thread like API for multi-threading is provided. The Erlang driver thread API only provide a subset of the functionality provided by the POSIX thread API. The subset provided is more or less the basic functionality needed for multi-threaded programming:
The Erlang driver thread API can be used in conjunction with the POSIX thread API on UN-ices and with the Windows native thread API on Windows. The Erlang driver thread API has the advantage of being portable, but there might exist situations where you want to use functionality from the POSIX thread API or the Windows native thread API.
The Erlang driver thread API only returns error codes when it is reasonable to recover from an error condition. If it isn't reasonable to recover from an error condition, the whole runtime system is terminated. For example, if a create mutex operation fails, an error code is returned, but if a lock operation on a mutex fails, the whole runtime system is terminated.
Note that there exists no "condition variable wait with timeout" in
the Erlang driver thread API. This is due to issues with
pthread_cond_timedwait()
. When the system clock suddenly
is changed, it isn't always guaranteed that you will wake up from
the call as expected. An Erlang runtime system has to be able to
cope with sudden changes of the system clock. Therefore, we have
omitted it from the Erlang driver thread API. In the Erlang driver
case, timeouts can and should be handled with the timer functionality
of the Erlang driver API.
In order for the Erlang driver thread API to function, thread support has to be enabled in the runtime system. An Erlang driver can check if thread support is enabled by use of driver_system_info(). Note that some functions in the Erlang driver API are thread-safe only when the runtime system has SMP support, also this information can be retrieved via driver_system_info(). Also note that a lot of functions in the Erlang driver API are not thread-safe regardless of whether SMP support is enabled or not. If a function isn't documented as thread-safe it is not thread-safe.
NOTE: When executing in an emulator thread, it is very important that you unlock all locks you have locked before letting the thread out of your control; otherwise, you are very likely to deadlock the whole emulator. If you need to use thread specific data in an emulator thread, only have the thread specific data set while the thread is under your control, and clear the thread specific data before you let the thread out of your control.
In the future there will probably be debug functionality
integrated with the Erlang driver thread API. All functions
that create entities take a name
argument. Currently
the name
argument is unused, but it will be used when
the debug functionality has been implemented. If you name all
entities created well, the debug functionality will be able
to give you better error reports.
A driver can add and later remove drivers.
A driver can monitor a process that does not own a port.
Version management is enabled for drivers that have set the
extended_marker
field of their
driver_entry
to ERL_DRV_EXTENDED_MARKER
. erl_driver.h
defines
ERL_DRV_EXTENDED_MARKER
,
ERL_DRV_EXTENDED_MAJOR_VERSION
, and
ERL_DRV_EXTENDED_MINOR_VERSION
.
ERL_DRV_EXTENDED_MAJOR_VERSION
will be incremented when
driver incompatible changes are made to the Erlang runtime
system. Normally it will suffice to recompile drivers when the
ERL_DRV_EXTENDED_MAJOR_VERSION
has changed, but it
could, under rare circumstances, mean that drivers have to
be slightly modified. If so, this will of course be documented.
ERL_DRV_EXTENDED_MINOR_VERSION
will be incremented when
new features are added. The runtime system uses the minor version
of the driver to determine what features to use.
The runtime system will refuse to load a driver if the major
versions differ, or if the major versions are equal and the
minor version used by the driver is greater than the one used
by the runtime system.
The emulator will refuse to load a driver that does not use the extended driver interface, to allow for 64-bit capable drivers, since incompatible type changes for the callbacks output, control and call were introduced in release R15B. A driver written with the old types would compile with warnings and when called return garbage sizes to the emulator causing it to read random memory and create huge incorrect result blobs.
Therefore it is not enough to just recompile drivers written with version management for pre-R15B types; the types have to be changed in the driver suggesting other rewrites especially regarding size variables. Investigate all warnings when recompiling!
Also, the API driver functions driver_output*
,
driver_vec_to_buf
, driver_alloc/realloc*
and the driver_*
queue functions were changed to have
larger length arguments and return values. This is a
lesser problem since code that passes smaller types
will get them auto converted in the calls and as long as
the driver does not handle sizes that overflow an int
all will work as before.
REWRITES FOR 64-BIT DRIVER INTERFACE
For erts-5.9 two new integer types ErlDrvSizeT and ErlDrvSSizeT were introduced that can hold 64-bit sizes if necessary.
To not update a driver and just recompile it probably works when building for a 32-bit machine creating a false sense of security. Hopefully that will generate many important warnings. But when recompiling the same driver later on for a 64-bit machine there will be warnings and almost certainly crashes. So it is a BAD idea to postpone updating the driver and not fixing the warnings!
When recompiling with gcc
use the -Wstrict-prototypes
flag to get better warnings. Try to find a similar flag if you
are using some other compiler.
Here follows a checklist for rewriting a pre erts-5.9 driver, most important first.
Rewrite driver callback
control
to use return type ErlDrvSSizeT
instead of int
.
Rewrite driver callback
call
to use return type ErlDrvSSizeT
instead of int
.
Note!
These changes are essential to not crash the emulator or worse cause malfunction. Without them a driver may return garbage in the high 32 bits to the emulator causing it to build a huge result from random bytes either crashing on memory allocation or succeeding with a random result from the driver call.
Driver callback
output
now gets ErlDrvSizeT
as 3rd argument instead
of previously int
.
Driver callback
control
now gets ErlDrvSizeT
as 4th and 6th arguments instead
of previously int
.
Driver callback
call
now gets ErlDrvSizeT
as 4th and 6th arguments instead
of previously int
.
Sane compiler's calling conventions probably make these changes
necessary only for a driver to handle data chunks that require
64-bit size fields (mostly larger than 2 GB since that is what
an int
of 32 bits can hold). But it is possible to think
of non-sane calling conventions that would make the driver
callbacks mix up the arguments causing malfunction.
Note!
The argument type change is from signed to unsigned which may cause problems for e.g. loop termination conditions or error conditions if you just change the types all over the place.
size
field in ErlIOVec
The size
field in
ErlIOVec
has been changed to ErlDrvSizeT
from int
.
Check all code that use that field.
Automatic type casting probably makes these changes necessary only for a driver that encounters sizes larger than 32 bits.
Note!
The size
field changed from signed to unsigned which
may cause problems for e.g. loop termination conditions or
error conditions if you just change the types all over the place.
Many driver API functions have changed argument type
and/or return value to ErlDrvSizeT
from mostly int
.
Automatic type casting probably makes these changes necessary only
for a driver that encounters sizes larger than 32 bits.
Note!
This is a change from signed to unsigned which may cause problems for e.g. loop termination conditions and error conditions if you just change the types all over the place.
DATA TYPES
An unsigned integer type to be used as size_t
A signed integer type the size of ErlDrvSizeT
typedef struct ErlDrvSysInfo { int driver_major_version; int driver_minor_version; char *erts_version; char *otp_release; int thread_support; int smp_support; int async_threads; int scheduler_threads; int nif_major_version; int nif_minor_version; } ErlDrvSysInfo;
The ErlDrvSysInfo
structure is used for storage of
information about the Erlang runtime system.
driver_system_info()
will write the system information when passed a reference to
a ErlDrvSysInfo
structure. A description of the
fields in the structure follows:
driver_major_version
driver_minor_version
erts_version
otp_release
thread_support
!= 0
if the runtime system has thread support;
otherwise, 0
.
smp_support
!= 0
if the runtime system has SMP support;
otherwise, 0
.
async_threads
scheduler_threads
nif_major_version
ERL_NIF_MAJOR_VERSION
when the runtime system was compiled.
nif_minor_version
ERL_NIF_MINOR_VERSION
when the runtime system was compiled.
typedef struct ErlDrvBinary { ErlDrvSint orig_size; char orig_bytes[]; } ErlDrvBinary;
The ErlDrvBinary
structure is a binary, as sent
between the emulator and the driver. All binaries are
reference counted; when driver_binary_free
is called,
the reference count is decremented, when it reaches zero,
the binary is deallocated. The orig_size
is the size
of the binary, and orig_bytes
is the buffer. The
ErlDrvBinary
does not have a fixed size, its size is
orig_size + 2 * sizeof(int)
.
Note!
The refc
field has been removed. The reference count of
an ErlDrvBinary
is now stored elsewhere. The
reference count of an ErlDrvBinary
can be accessed via
driver_binary_get_refc(),
driver_binary_inc_refc(),
and
driver_binary_dec_refc().
Some driver calls, such as driver_enq_binary
,
increment the driver reference count, and others, such as
driver_deq
decrement it.
Using a driver binary instead of a normal buffer, is often faster, since the emulator doesn't need to copy the data, only the pointer is used.
A driver binary allocated in the driver, with
driver_alloc_binary
, should be freed in the driver (unless otherwise stated),
with driver_free_binary
. (Note that this doesn't
necessarily deallocate it, if the driver is still referred
in the emulator, the ref-count will not go to zero.)
Driver binaries are used in the driver_output2
and
driver_outputv
calls, and in the queue. Also the
driver call-back outputv uses driver
binaries.
If the driver for some reason or another, wants to keep a
driver binary around, in a static variable for instance, the
reference count should be incremented,
and the binary can later be freed in the stop call-back, with
driver_free_binary
.
Note that since a driver binary is shared by the driver and the emulator, a binary received from the emulator or sent to the emulator, must not be changed by the driver.
Since erts version 5.5 (OTP release R11B), orig_bytes is guaranteed to be properly aligned for storage of an array of doubles (usually 8-byte aligned).
The ErlDrvData
is a handle to driver-specific data,
passed to the driver call-backs. It is a pointer, and is
most often type cast to a specific pointer in the driver.
This is a system I/O vector, as used by writev
on
unix and WSASend
on Win32. It is used in
ErlIOVec
.
typedef struct ErlIOVec { int vsize; ErlDrvSizeT size; SysIOVec* iov; ErlDrvBinary** binv; } ErlIOVec;
The I/O vector used by the emulator and drivers, is a list
of binaries, with a SysIOVec
pointing to the buffers
of the binaries. It is used in driver_outputv
and the
outputv
driver call-back. Also, the driver queue is an
ErlIOVec
.
When a driver creates a monitor for a process, a
ErlDrvMonitor
is filled in. This is an opaque
data-type which can be assigned to but not compared without
using the supplied compare function (i.e. it behaves like a struct).
The driver writer should provide the memory for storing the
monitor when calling driver_monitor_process. The
address of the data is not stored outside of the driver, so
the ErlDrvMonitor
can be used as any other datum, it
can be copied, moved in memory, forgotten etc.
The ErlDrvNowData
structure holds a timestamp
consisting of three values measured from some arbitrary
point in the past. The three structure members are:
If certain port specific data have to be accessed from other threads than those calling the driver call-backs, a port data lock can be used in order to synchronize the operations on the data. Currently, the only port specific data that the emulator associates with the port data lock is the driver queue.
Normally a driver instance does not have a port data lock. If the driver instance wants to use a port data lock, it has to create the port data lock by calling driver_pdl_create(). NOTE: Once the port data lock has been created, every access to data associated with the port data lock has to be done while having the port data lock locked. The port data lock is locked, and unlocked, respectively, by use of driver_pdl_lock(), and driver_pdl_unlock().
A port data lock is reference counted, and when the reference count reaches zero, it will be destroyed. The emulator will at least increment the reference count once when the lock is created and decrement it once when the port associated with the lock terminates. The emulator will also increment the reference count when an async job is enqueued and decrement it after an async job has been invoked. Besides this, it is the responsibility of the driver to ensure that the reference count does not reach zero before the last use of the lock by the driver has been made. The reference count can be read, incremented, and decremented, respectively, by use of driver_pdl_get_refc(), driver_pdl_inc_refc(), and driver_pdl_dec_refc().
Thread identifier.
See also: erl_drv_thread_create(), erl_drv_thread_exit(), erl_drv_thread_join(), erl_drv_thread_self(), and erl_drv_equal_tids().
int suggested_stack_size;
Thread options structure passed to erl_drv_thread_create(). Currently the following fields exist:
See also: erl_drv_thread_opts_create(), erl_drv_thread_opts_destroy(), and erl_drv_thread_create().
Mutual exclusion lock. Used for synchronizing access to shared data. Only one thread at a time can lock a mutex.
See also: erl_drv_mutex_create(), erl_drv_mutex_destroy(), erl_drv_mutex_lock(), erl_drv_mutex_trylock(), and erl_drv_mutex_unlock().
Condition variable. Used when threads need to wait for a specific condition to appear before continuing execution. Condition variables need to be used with associated mutexes.
See also: erl_drv_cond_create(), erl_drv_cond_destroy(), erl_drv_cond_signal(), erl_drv_cond_broadcast(), and erl_drv_cond_wait().
Read/write lock. Used to allow multiple threads to read shared data while only allowing one thread to write the same data. Multiple threads can read lock an rwlock at the same time, while only one thread can read/write lock an rwlock at a time.
See also: erl_drv_rwlock_create(), erl_drv_rwlock_destroy(), erl_drv_rwlock_rlock(), erl_drv_rwlock_tryrlock(), erl_drv_rwlock_runlock(), erl_drv_rwlock_rwlock(), erl_drv_rwlock_tryrwlock(), and erl_drv_rwlock_rwunlock().
Key which thread specific data can be associated with.
See also: erl_drv_tsd_key_create(), erl_drv_tsd_key_destroy(), erl_drv_tsd_set(), and erl_drv_tsd_get().
Functions
void driver_system_info(ErlDrvSysInfo *sys_info_ptr, size_t size)
This function will write information about the Erlang runtime
system into the
ErlDrvSysInfo
structure referred to by the first argument. The second
argument should be the size of the
ErlDrvSysInfo
structure, i.e., sizeof(ErlDrvSysInfo)
.
See the documentation of the ErlDrvSysInfo structure for information about specific fields.
int driver_output(ErlDrvPort port, char *buf, ErlDrvSizeT len)
The driver_output
function is used to send data from
the driver up to the emulator. The data will be received as
terms or binary data, depending on how the driver port was
opened.
The data is queued in the port owner process' message queue. Note that this does not yield to the emulator. (Since the driver and the emulator run in the same thread.)
The parameter buf
points to the data to send, and
len
is the number of bytes.
The return value for all output functions is 0. (Unless the driver is used for distribution, in which case it can fail and return -1. For normal use, the output function always returns 0.)
int driver_output2(ErlDrvPort port, char *hbuf, ErlDrvSizeT hlen, char *buf, ErlDrvSizeT len)
The driver_output2
function first sends hbuf
(length in hlen
) data as a list, regardless of port
settings. Then buf
is sent as a binary or list.
E.g. if hlen
is 3 then the port owner process will
receive [H1, H2, H3 | T]
.
The point of sending data as a list header, is to facilitate matching on the data received.
The return value is 0 for normal use.
int driver_output_binary(ErlDrvPort port, char *hbuf, ErlDrvSizeT hlen, ErlDrvBinary* bin, ErlDrvSizeT offset, ErlDrvSizeT len)
This function sends data to port owner process from a
driver binary, it has a header buffer (hbuf
and hlen
) just like driver_output2
. The
hbuf
parameter can be NULL
.
The parameter offset
is an offset into the binary and
len
is the number of bytes to send.
Driver binaries are created with driver_alloc_binary
.
The data in the header is sent as a list and the binary as an Erlang binary in the tail of the list.
E.g. if hlen
is 2, then the port owner process will
receive [H1, H2 | <<T>>]
.
The return value is 0 for normal use.
Note that, using the binary syntax in Erlang, the driver application can match the header directly from the binary, so the header can be put in the binary, and hlen can be set to 0.
int driver_outputv(ErlDrvPort port, char* hbuf, ErlDrvSizeT hlen, ErlIOVec *ev, ErlDrvSizeT skip)
This function sends data from an IO vector, ev
, to
the port owner process. It has a header buffer (hbuf
and hlen
), just like driver_output2
.
The skip
parameter is a number of bytes to skip of
the ev
vector from the head.
You get vectors of ErlIOVec
type from the driver
queue (see below), and the outputv driver entry
function. You can also make them yourself, if you want to
send several ErlDrvBinary
buffers at once. Often
it is faster to use driver_output
or
driver_output_binary
.
E.g. if hlen
is 2 and ev
points to an array of
three binaries, the port owner process will receive [H1, H2, <<B1>>, <<B2>> | <<B3>>]
.
The return value is 0 for normal use.
The comment for driver_output_binary
applies for
driver_outputv
too.
ErlDrvSizeT driver_vec_to_buf(ErlIOVec *ev, char *buf, ErlDrvSizeT len)
This function collects several segments of data, referenced
by ev
, by copying them in order to the buffer
buf
, of the size len
.
If the data is to be sent from the driver to the port owner
process, it is faster to use driver_outputv
.
The return value is the space left in the buffer, i.e. if
the ev
contains less than len
bytes it's the
difference, and if ev
contains len
bytes or
more, it's 0. This is faster if there is more than one header byte,
since the binary syntax can construct integers directly from
the binary.
int driver_set_timer(ErlDrvPort port, unsigned long time)
This function sets a timer on the driver, which will count
down and call the driver when it is timed out. The
time
parameter is the time in milliseconds before the
timer expires.
When the timer reaches 0 and expires, the driver entry function timeout is called.
Note that there is only one timer on each driver instance; setting a new timer will replace an older one.
Return value is 0 (-1 only when the timeout
driver
function is NULL
).
int driver_cancel_timer(ErlDrvPort port)
This function cancels a timer set with
driver_set_timer
.
The return value is 0.
int driver_read_timer(ErlDrvPort port, unsigned long *time_left)
This function reads the current time of a timer, and places
the result in time_left
. This is the time in
milliseconds, before the timeout will occur.
The return value is 0.
int driver_get_now(ErlDrvNowData *now)
This function reads a timestamp into the memory pointed to by
the parameter now
. See the description of ErlDrvNowData for
specification of its fields.
The return value is 0 unless the now
pointer is not
valid, in which case it is < 0.
int driver_select(ErlDrvPort port, ErlDrvEvent event, int mode, int on)
This function is used by drivers to provide the emulator with events to check for. This enables the emulator to call the driver when something has happened asynchronously.
The event
argument identifies an OS-specific event object.
On Unix systems, the functions select
/poll
are used. The
event object must be a socket or pipe (or other object that
select
/poll
can use).
On windows, the Win32 API function WaitForMultipleObjects
is used. This places other restrictions on the event object.
Refer to the Win32 SDK documentation.
On Enea OSE, the receive function is used. See the for more details.
The on
parameter should be 1
for setting events
and 0
for clearing them.
The mode
argument is a bitwise-or combination of
ERL_DRV_READ
, ERL_DRV_WRITE
and ERL_DRV_USE
.
The first two specify whether to wait for read events and/or write
events. A fired read event will call
ready_input
while a fired write event will call
ready_output.
Note!
Some OS (Windows and Enea OSE) do not differentiate between read and write events.
The call-back for a fired event then only depends on the value of mode
.
ERL_DRV_USE
specifies if we are using the event object or if we want to close it.
On an emulator with SMP support, it is not safe to clear all events
and then close the event object after driver_select
has
returned. Another thread may still be using the event object
internally. To safely close an event object call
driver_select
with ERL_DRV_USE
and on==0
. That
will clear all events and then call
stop_select
when it is safe to close the event object.
ERL_DRV_USE
should be set together with the first event
for an event object. It is harmless to set ERL_DRV_USE
even though it already has been done. Clearing all events but keeping
ERL_DRV_USE
set will indicate that we are using the event
object and probably will set events for it again.
Note!
ERL_DRV_USE was added in OTP release R13. Old drivers will still work
as before. But it is recommended to update them to use ERL_DRV_USE
and
stop_select
to make sure that event objects are closed in a safe way.
The return value is 0 (failure, -1, only if the
ready_input
/ready_output
is
NULL
).
void * driver_alloc(ErlDrvSizeT size)
This function allocates a memory block of the size specified
in size
, and returns it. This only fails on out of
memory, in that case NULL
is returned. (This is most
often a wrapper for malloc
).
Memory allocated must be explicitly freed with a corresponding
call to driver_free
(unless otherwise stated).
This function is thread-safe.
void * driver_realloc(void *ptr, ErlDrvSizeT size)
This function resizes a memory block, either in place, or by
allocating a new block, copying the data and freeing the old
block. A pointer is returned to the reallocated memory. On
failure (out of memory), NULL
is returned. (This is
most often a wrapper for realloc
.)
This function is thread-safe.
void driver_free(void *ptr)
This function frees the memory pointed to by ptr
. The
memory should have been allocated with
driver_alloc
. All allocated memory should be
deallocated, just once. There is no garbage collection in
drivers.
This function is thread-safe.
ErlDrvBinary * driver_alloc_binary(ErlDrvSizeT size)
This function allocates a driver binary with a memory block
of at least size
bytes, and returns a pointer to it,
or NULL on failure (out of memory). When a driver binary has
been sent to the emulator, it must not be altered. Every
allocated binary should be freed by a corresponding call to
driver_free_binary
(unless otherwise stated).
Note that a driver binary has an internal reference counter,
this means that calling driver_free_binary
it may not
actually dispose of it. If it's sent to the emulator, it may
be referenced there.
The driver binary has a field, orig_bytes
, which
marks the start of the data in the binary.
This function is thread-safe.
ErlDrvBinary * driver_realloc_binary(ErlDrvBinary *bin, ErlDrvSizeT size)
This function resizes a driver binary, while keeping the
data. The resized driver binary is returned. On failure (out
of memory), NULL
is returned.
This function is only thread-safe when the emulator with SMP support is used.
void driver_free_binary(ErlDrvBinary *bin)
This function frees a driver binary bin
, allocated
previously with driver_alloc_binary
. Since binaries
in Erlang are reference counted, the binary may still be
around.
This function is only thread-safe when the emulator with SMP support is used.
long driver_binary_get_refc(ErlDrvBinary *bin)
Returns current reference count on bin
.
This function is only thread-safe when the emulator with SMP support is used.
long driver_binary_inc_refc(ErlDrvBinary *bin)
Increments the reference count on bin
and returns
the reference count reached after the increment.
This function is only thread-safe when the emulator with SMP support is used.
long driver_binary_dec_refc(ErlDrvBinary *bin)
Decrements the reference count on bin
and returns
the reference count reached after the decrement.
This function is only thread-safe when the emulator with SMP support is used.
Note!
You should normally decrement the reference count of a
driver binary by calling
driver_free_binary().
driver_binary_dec_refc()
does not free
the binary if the reference count reaches zero. Only
use driver_binary_dec_refc()
when you are sure
not to reach a reference count of zero.
int driver_enq(ErlDrvPort port, char* buf, ErlDrvSizeT len)
This function enqueues data in the driver queue. The data in
buf
is copied (len
bytes) and placed at the
end of the driver queue. The driver queue is normally used
in a FIFO way.
The driver queue is available to queue output from the emulator to the driver (data from the driver to the emulator is queued by the emulator in normal erlang message queues). This can be useful if the driver has to wait for slow devices etc, and wants to yield back to the emulator. The driver queue is implemented as an ErlIOVec.
When the queue contains data, the driver won't close, until the queue is empty.
The return value is 0.
This function can be called from an arbitrary thread if a
port data lock
associated with the port
is locked by the calling
thread during the call.
int driver_pushq(ErlDrvPort port, char* buf, ErlDrvSizeT len)
This function puts data at the head of the driver queue. The
data in buf
is copied (len
bytes) and placed
at the beginning of the queue.
The return value is 0.
This function can be called from an arbitrary thread if a
port data lock
associated with the port
is locked by the calling
thread during the call.
ErlDrvSizeT driver_deq(ErlDrvPort port, ErlDrvSizeT size)
This function dequeues data by moving the head pointer
forward in the driver queue by size
bytes. The data
in the queue will be deallocated.
The return value is the number of bytes remaining in the queue or -1 on failure.
This function can be called from an arbitrary thread if a
port data lock
associated with the port
is locked by the calling
thread during the call.
ErlDrvSizeT driver_sizeq(ErlDrvPort port)
This function returns the number of bytes currently in the driver queue.
This function can be called from an arbitrary thread if a
port data lock
associated with the port
is locked by the calling
thread during the call.
int driver_enq_bin(ErlDrvPort port, ErlDrvBinary *bin, ErlDrvSizeT offset, ErlDrvSizeT len)
This function enqueues a driver binary in the driver
queue. The data in bin
at offset
with length
len
is placed at the end of the queue. This function
is most often faster than driver_enq
, because the
data doesn't have to be copied.
This function can be called from an arbitrary thread if a
port data lock
associated with the port
is locked by the calling
thread during the call.
The return value is 0.
int driver_pushq_bin(ErlDrvPort port, ErlDrvBinary *bin, ErlDrvSizeT offset, ErlDrvSizeT len)
This function puts data in the binary bin
, at
offset
with length len
at the head of the
driver queue. It is most often faster than
driver_pushq
, because the data doesn't have to be
copied.
This function can be called from an arbitrary thread if a
port data lock
associated with the port
is locked by the calling
thread during the call.
The return value is 0.
ErlDrvSizeT driver_peekqv(ErlDrvPort port, ErlIOVec *ev)
This function retrieves the driver queue into a supplied
ErlIOVec
ev
. It also returns the queue size.
This is one of two ways to get data out of the queue.
If ev
is NULL
all ones i.e. -1
type cast to
ErlDrvSizeT
is returned.
Nothing is removed from the queue by this function, that must be done
with driver_deq
.
This function can be called from an arbitrary thread if a
port data lock
associated with the port
is locked by the calling
thread during the call.
SysIOVec * driver_peekq(ErlDrvPort port, int *vlen)
This function retrieves the driver queue as a pointer to an
array of SysIOVec
s. It also returns the number of
elements in vlen
. This is one of two ways to get data
out of the queue.
Nothing is removed from the queue by this function, that must be done
with driver_deq
.
The returned array is suitable to use with the Unix system
call writev
.
This function can be called from an arbitrary thread if a
port data lock
associated with the port
is locked by the calling
thread during the call.
int driver_enqv(ErlDrvPort port, ErlIOVec *ev, ErlDrvSizeT skip)
This function enqueues the data in ev
, skipping the
first skip
bytes of it, at the end of the driver
queue. It is faster than driver_enq
, because the data
doesn't have to be copied.
The return value is 0.
This function can be called from an arbitrary thread if a
port data lock
associated with the port
is locked by the calling
thread during the call.
int driver_pushqv(ErlDrvPort port, ErlIOVec *ev, ErlDrvSizeT skip)
This function puts the data in ev
, skipping the first
skip
bytes of it, at the head of the driver queue.
It is faster than driver_pushq
, because the data
doesn't have to be copied.
The return value is 0.
This function can be called from an arbitrary thread if a
port data lock
associated with the port
is locked by the calling
thread during the call.
ErlDrvPDL driver_pdl_create(ErlDrvPort port)
This function creates a port data lock associated with
the port
. NOTE: Once a port data lock has
been created, it has to be locked during all operations
on the driver queue of the port
.
On success a newly created port data lock is returned. On
failure NULL
is returned. driver_pdl_create()
will
fail if port
is invalid or if a port data lock already has
been associated with the port
.
void driver_pdl_lock(ErlDrvPDL pdl)
This function locks the port data lock passed as argument
(pdl
).
This function is thread-safe.
void driver_pdl_unlock(ErlDrvPDL pdl)
This function unlocks the port data lock passed as argument
(pdl
).
This function is thread-safe.
long driver_pdl_get_refc(ErlDrvPDL pdl)
This function returns the current reference count of
the port data lock passed as argument (pdl
).
This function is thread-safe.
long driver_pdl_inc_refc(ErlDrvPDL pdl)
This function increments the reference count of
the port data lock passed as argument (pdl
).
The current reference count after the increment has been performed is returned.
This function is thread-safe.
long driver_pdl_dec_refc(ErlDrvPDL pdl)
This function decrements the reference count of
the port data lock passed as argument (pdl
).
The current reference count after the decrement has been performed is returned.
This function is thread-safe.
int driver_monitor_process(ErlDrvPort port, ErlDrvTermData process, ErlDrvMonitor *monitor)
Start monitoring a process from a driver. When a process is
monitored, a process exit will result in a call to the
provided process_exit call-back
in the ErlDrvEntry
structure. The ErlDrvMonitor
structure is filled in, for later
removal or compare.
The process
parameter should be the return value of an
earlier call to driver_caller or driver_connected call.
The function returns 0 on success, < 0 if no call-back is provided and > 0 if the process is no longer alive.
int driver_demonitor_process(ErlDrvPort port, const ErlDrvMonitor *monitor)
This function cancels a monitor created earlier.
The function returns 0 if a monitor was removed and > 0 if the monitor did no longer exist.
ErlDrvTermData driver_get_monitored_process(ErlDrvPort port, const ErlDrvMonitor *monitor)
The function returns the process id associated with a living
monitor. It can be used in the process_exit
call-back to
get the process identification for the exiting process.
The function returns driver_term_nil
if the monitor
no longer exists.
int driver_compare_monitors(const ErlDrvMonitor *monitor1, const ErlDrvMonitor *monitor2)
This function is used to compare two ErlDrvMonitor
s. It
can also be used to imply some artificial order on monitors,
for whatever reason.
The function returns 0 if monitor1
and
monitor2
are equal, < 0 if monitor1
is less
than monitor2
and > 0 if monitor1
is greater
than monitor2
.
void add_driver_entry(ErlDrvEntry *de)
This function adds a driver entry to the list of drivers
known by Erlang. The init function of the de
parameter is called.
Note!
To use this function for adding drivers residing in
dynamically loaded code is dangerous. If the driver code
for the added driver resides in the same dynamically
loaded module (i.e. .so
file) as a normal
dynamically loaded driver (loaded with the erl_ddll
interface), the caller should call driver_lock_driver before
adding driver entries.
Use of this function is generally deprecated.
int remove_driver_entry(ErlDrvEntry *de)
This function removes a driver entry de
previously
added with add_driver_entry
.
Driver entries added by the erl_ddll
erlang interface can
not be removed by using this interface.
char * erl_errno_id(int error)
This function returns the atom name of the erlang error,
given the error number in error
. Error atoms are:
einval
, enoent
, etc. It can be used to make
error terms from the driver.
void erl_drv_busy_msgq_limits(ErlDrvPort port, ErlDrvSizeT *low, ErlDrvSizeT *high)
Sets and gets limits that will be used for controling the busy state of the port message queue.
The port message queue will be set into a busy
state when the amount of command data queued on the
message queue reaches the high
limit. The port
message queue will be set into a not busy state when the
amount of command data queued on the message queue falls
below the low
limit. Command data is in this
context data passed to the port using either
Port ! {Owner, {command, Data}}
, or
port_command/[2,3]
. Note that these limits
only concerns command data that have not yet reached the
port. The busy port
feature can be used for data that has reached the port.
Valid limits are values in the range
[ERL_DRV_BUSY_MSGQ_LIM_MIN, ERL_DRV_BUSY_MSGQ_LIM_MAX]
.
Limits will be automatically adjusted to be sane. That is,
the system will adjust values so that the low limit used is
lower than or equal to the high limit used. By default the high
limit will be 8 kB and the low limit will be 4 kB.
By passing a pointer to an integer variable containing
the value ERL_DRV_BUSY_MSGQ_READ_ONLY
, currently used
limit will be read and written back to the integer variable.
A new limit can be set by passing a pointer to an integer
variable containing a valid limit. The passed value will be
written to the internal limit. The internal limit will then
be adjusted. After this the adjusted limit will be written
back to the integer variable from which the new value was
read. Values are in bytes.
The busy message queue feature can be disabled either
by setting the ERL_DRV_FLAG_NO_BUSY_MSGQ
driver flag
in the driver_entry
used by the driver, or by calling this function with
ERL_DRV_BUSY_MSGQ_DISABLED
as a limit (either low or
high). When this feature has been disabled it cannot be
enabled again. When reading the limits both of them
will be ERL_DRV_BUSY_MSGQ_DISABLED
, if this
feature has been disabled.
Processes sending command data to the port will be suspended if either the port is busy or if the port message queue is busy. Suspended processes will be resumed when neither the port is busy, nor the port message queue is busy.
For information about busy port functionality see the documentation of the set_busy_port() function.
void set_busy_port(ErlDrvPort port, int on)
This function set and unset the busy state of the port. If
on
is non-zero, the port is set to busy, if it's zero the port
is set to not busy. You typically want to combine
this feature with the busy
port message queue functionality.
Processes sending command data to the port will be suspended
if either the port is busy or if the port message queue
is busy. Suspended processes will be resumed when neither the
port is busy, nor the port message queue is busy. Command data
is in this context data passed to the port using either
Port ! {Owner, {command, Data}}
, or
port_command/[2,3]
.
If the ERL_DRV_FLAG_SOFT_BUSY has been set in the driver_entry, data can be forced into the driver via port_command(Port, Data, [force]) even though the driver has signaled that it is busy.
For information about busy port message queue functionality see the documentation of the erl_drv_busy_msgq_limits() function.
void set_port_control_flags(ErlDrvPort port, int flags)
This function sets flags for how the control driver entry
function will return data to the port owner process. (The
control
function is called from port_control/3
in erlang.)
Currently there are only two meaningful values for
flags
: 0 means that data is returned in a list, and
PORT_CONTROL_FLAG_BINARY
means data is returned as
a binary from control
.
int driver_failure_eof(ErlDrvPort port)
This function signals to erlang that the driver has
encountered an EOF and should be closed, unless the port was
opened with the eof
option, in that case eof is sent
to the port. Otherwise, the port is closed and an
'EXIT'
message is sent to the port owner process.
The return value is 0.
int driver_failure_atom(ErlDrvPort port, char *string)
int driver_failure_posix(ErlDrvPort port, int error)
int driver_failure(ErlDrvPort port, int error)
These functions signal to Erlang that the driver has
encountered an error and should be closed. The port is
closed and the tuple {'EXIT', error, Err}
, is sent to
the port owner process, where error is an error atom
(driver_failure_atom
and
driver_failure_posix
), or an integer
(driver_failure
).
The driver should fail only when in severe error situations,
when the driver cannot possibly keep open, for instance
buffer allocation gets out of memory. For normal errors
it is more appropriate to send error codes with
driver_output
.
The return value is 0.
ErlDrvTermData driver_connected(ErlDrvPort port)
This function returns the port owner process.
Note that this function is not thread-safe, not even when the emulator with SMP support is used.
ErlDrvTermData driver_caller(ErlDrvPort port)
This function returns the process id of the process that
made the current call to the driver. The process id can be
used with driver_send_term
to send back data to the
caller. driver_caller()
only returns valid data
when currently executing in one of the following driver
callbacks:
open_port/2
.erlang:send/2
, and
erlang:port_command/2
erlang:send/2
, and
erlang:port_command/2
erlang:port_control/3
erlang:port_call/3
Note that this function is not thread-safe, not even when the emulator with SMP support is used.
int erl_drv_output_term(ErlDrvTermData port, ErlDrvTermData* term, int n)
This functions sends data in the special driver term format to the port owner process. This is a fast way to deliver term data from a driver. It also needs no binary conversion, so the port owner process receives data as normal Erlang terms. The erl_drv_send_term() functions can be used for sending to any arbitrary process on the local node.
Note!
Note that the port
parameter is not
an ordinary port handle, but a port handle converted using
driver_mk_port()
.
The term
parameter points to an array of
ErlDrvTermData
, with n
elements. This array
contains terms described in the driver term format. Every
term consists of one to four elements in the array. The
term first has a term type, and then arguments. The
port
parameter specifies the sending port.
Tuples, maps and lists (with the exception of strings, see below), are built in reverse polish notation, so that to build a tuple, the elements are given first, and then the tuple term, with a count. Likewise for lists and maps.
A tuple must be specified with the number of elements. (The
elements precede the ERL_DRV_TUPLE
term.)
A list must be specified with the number of elements,
including the tail, which is the last term preceding
ERL_DRV_LIST
.
A map must be specified with the number of key-value pairs N
.
The key-value pairs must precede the ERL_DRV_MAP
in this order:
key1,value1,key2,value2,...,keyN,valueN
.
Duplicate keys are not allowed.
The special term ERL_DRV_STRING_CONS
is used to
"splice" in a string in a list, a string given this way is
not a list per se, but the elements are elements of the
surrounding list.
Term type Argument(s) =========================================== ERL_DRV_NIL ERL_DRV_ATOM ErlDrvTermData atom (from driver_mk_atom(char *string)) ERL_DRV_INT ErlDrvSInt integer ERL_DRV_UINT ErlDrvUInt integer ERL_DRV_INT64 ErlDrvSInt64 *integer_ptr ERL_DRV_UINT64 ErlDrvUInt64 *integer_ptr ERL_DRV_PORT ErlDrvTermData port (from driver_mk_port(ErlDrvPort port)) ERL_DRV_BINARY ErlDrvBinary *bin, ErlDrvUInt len, ErlDrvUInt offset ERL_DRV_BUF2BINARY char *buf, ErlDrvUInt len ERL_DRV_STRING char *str, int len ERL_DRV_TUPLE int sz ERL_DRV_LIST int sz ERL_DRV_PID ErlDrvTermData pid (from driver_connected(ErlDrvPort port) or driver_caller(ErlDrvPort port)) ERL_DRV_STRING_CONS char *str, int len ERL_DRV_FLOAT double *dbl ERL_DRV_EXT2TERM char *buf, ErlDrvUInt len ERL_DRV_MAP int sz
The unsigned integer data type ErlDrvUInt
and the
signed integer data type ErlDrvSInt
are 64 bits wide
on a 64 bit runtime system and 32 bits wide on a 32 bit
runtime system. They were introduced in erts version 5.6,
and replaced some of the int
arguments in the list above.
The unsigned integer data type ErlDrvUInt64
and the
signed integer data type ErlDrvSInt64
are always 64 bits
wide. They were introduced in erts version 5.7.4.
To build the tuple {tcp, Port, [100 | Binary]}
, the
following call could be made.
ErlDrvBinary* bin = ... ErlDrvPort port = ... ErlDrvTermData spec[] = { ERL_DRV_ATOM, driver_mk_atom("tcp"), ERL_DRV_PORT, driver_mk_port(drvport), ERL_DRV_INT, 100, ERL_DRV_BINARY, bin, 50, 0, ERL_DRV_LIST, 2, ERL_DRV_TUPLE, 3, }; erl_drv_output_term(driver_mk_port(drvport), spec, sizeof(spec) / sizeof(spec[0]));
Where bin
is a driver binary of length at least 50
and drvport
is a port handle. Note that the ERL_DRV_LIST
comes after the elements of the list, likewise the
ERL_DRV_TUPLE
.
The term ERL_DRV_STRING_CONS
is a way to construct
strings. It works differently from how ERL_DRV_STRING
works. ERL_DRV_STRING_CONS
builds a string list in
reverse order, (as opposed to how ERL_DRV_LIST
works), concatenating the strings added to a list. The tail
must be given before ERL_DRV_STRING_CONS
.
The ERL_DRV_STRING
constructs a string, and ends
it. (So it's the same as ERL_DRV_NIL
followed by
ERL_DRV_STRING_CONS
.)
/* to send [x, "abc", y] to the port: */ ErlDrvTermData spec[] = { ERL_DRV_ATOM, driver_mk_atom("x"), ERL_DRV_STRING, (ErlDrvTermData)"abc", 3, ERL_DRV_ATOM, driver_mk_atom("y"), ERL_DRV_NIL, ERL_DRV_LIST, 4 }; erl_drv_output_term(driver_mk_port(drvport), spec, sizeof(spec) / sizeof(spec[0]));
/* to send "abc123" to the port: */ ErlDrvTermData spec[] = { ERL_DRV_NIL, /* with STRING_CONS, the tail comes first */ ERL_DRV_STRING_CONS, (ErlDrvTermData)"123", 3, ERL_DRV_STRING_CONS, (ErlDrvTermData)"abc", 3, }; erl_drv_output_term(driver_mk_port(drvport), spec, sizeof(spec) / sizeof(spec[0]));
The ERL_DRV_EXT2TERM
term type is used for passing a
term encoded with the
external format,
i.e., a term that has been encoded by
erlang:term_to_binary,
erl_interface, etc.
For example, if binp
is a pointer to an ErlDrvBinary
that contains the term {17, 4711}
encoded with the
external format
and you want to wrap it in a two tuple with the tag my_tag
,
i.e., {my_tag, {17, 4711}}
, you can do as follows:
ErlDrvTermData spec[] = { ERL_DRV_ATOM, driver_mk_atom("my_tag"), ERL_DRV_EXT2TERM, (ErlDrvTermData) binp->orig_bytes, binp->orig_size ERL_DRV_TUPLE, 2, }; erl_drv_output_term(driver_mk_port(drvport), spec, sizeof(spec) / sizeof(spec[0]));
To build the map #{key1 => 100, key2 => {200, 300}}
, the
following call could be made.
ErlDrvPort port = ... ErlDrvTermData spec[] = { ERL_DRV_ATOM, driver_mk_atom("key1"), ERL_DRV_INT, 100, ERL_DRV_ATOM, driver_mk_atom("key2"), ERL_DRV_INT, 200, ERL_DRV_INT, 300, ERL_DRV_TUPLE, 2, ERL_DRV_MAP, 2 }; erl_drv_output_term(driver_mk_port(drvport), spec, sizeof(spec) / sizeof(spec[0]));
If you want to pass a binary and don't already have the content
of the binary in an ErlDrvBinary
, you can benefit from using
ERL_DRV_BUF2BINARY
instead of creating an ErlDrvBinary
via driver_alloc_binary()
and then pass the binary via
ERL_DRV_BINARY
. The runtime system will often allocate
binaries smarter if ERL_DRV_BUF2BINARY
is used.
However, if the content of the binary to pass already resides in
an ErlDrvBinary
, it is normally better to pass the binary
using ERL_DRV_BINARY
and the ErlDrvBinary
in question.
The ERL_DRV_UINT
, ERL_DRV_BUF2BINARY
, and
ERL_DRV_EXT2TERM
term types were introduced in the 5.6
version of erts.
This function is only thread-safe when the emulator with SMP support is used.
int driver_output_term(ErlDrvPort port, ErlDrvTermData* term, int n)
Warning!
driver_output_term()
is deprecated and will
be removed in the OTP-R17 release. Use
erl_drv_output_term()
instead.
The parameters term
and n
do the same thing
as in erl_drv_output_term().
Note that this function is not thread-safe, not even when the emulator with SMP support is used.
ErlDrvTermData driver_mk_atom(char* string)
This function returns an atom given a name
string
. The atom is created and won't change, so the
return value may be saved and reused, which is faster than
looking up the atom several times.
Note that this function is not thread-safe, not even when the emulator with SMP support is used.
ErlDrvTermData driver_mk_port(ErlDrvPort port)
This function converts a port handle to the erlang term format, usable in the erl_drv_output_term(), and erl_drv_send_term() functions.
Note that this function is not thread-safe, not even when the emulator with SMP support is used.
int erl_drv_send_term(ErlDrvTermData port, ErlDrvTermData receiver, ErlDrvTermData* term, int n)
This function is the only way for a driver to send data to
other processes than the port owner process. The
receiver
parameter specifies the process to receive
the data.
Note!
Note that the port
parameter is not
an ordinary port handle, but a port handle converted using
driver_mk_port()
.
The parameters port
, term
and n
do the same thing
as in erl_drv_output_term().
This function is only thread-safe when the emulator with SMP support is used.
int driver_send_term(ErlDrvPort port, ErlDrvTermData receiver, ErlDrvTermData* term, int n)
Warning!
driver_send_term()
is deprecated and will
be removed in the OTP-R17 release. Use
erl_drv_send_term()
instead.
Also note that parameters of driver_send_term()
cannot be properly checked by the runtime system when
executed by arbitrary threads. This may cause the
driver_send_term()
function not to fail when
it should.
The parameters term
and n
do the same thing
as in erl_drv_output_term().
This function is only thread-safe when the emulator with SMP support is used.
long driver_async (ErlDrvPort port, unsigned int* key, void (*async_invoke)(void*), void* async_data, void (*async_free)(void*))
This function performs an asynchronous call. The function
async_invoke
is invoked in a thread separate from the
emulator thread. This enables the driver to perform
time-consuming, blocking operations without blocking the
emulator.
The async thread pool size can be set with the
+A
command line argument of erl(1).
If no async thread pool is available, the call is made
synchronously in the thread calling driver_async()
. The
current number of async threads in the async thread pool can be
retrieved via
driver_system_info().
If there is a thread pool available, a thread will be
used. If the key
argument is null, the threads from the
pool are used in a round-robin way, each call to
driver_async
uses the next thread in the pool. With the
key
argument set, this behaviour is changed. The two
same values of *key
always get the same thread.
To make sure that a driver instance always uses the same thread, the following call can be used:
unsigned int myKey = driver_async_port_key(myPort); r = driver_async(myPort, &myKey, myData, myFunc);
It is enough to initialize myKey
once for each
driver instance.
If a thread is already working, the calls will be queued up and executed in order. Using the same thread for each driver instance ensures that the calls will be made in sequence.
The async_data
is the argument to the functions
async_invoke
and async_free
. It's typically a
pointer to a structure that contains a pipe or event that
can be used to signal that the async operation completed.
The data should be freed in async_free
.
When the async operation is done, ready_async driver
entry function is called. If ready_async
is null in
the driver entry, the async_free
function is called
instead.
The return value is a handle to the asynchronous task.
Note!
As of erts version 5.5.4.3 the default stack size for threads in the async-thread pool is 16 kilowords, i.e., 64 kilobyte on 32-bit architectures. This small default size has been chosen since the amount of async-threads might be quite large. The default stack size is enough for drivers delivered with Erlang/OTP, but might not be sufficiently large for other dynamically linked in drivers that use the driver_async() functionality. A suggested stack size for threads in the async-thread pool can be configured via the +a command line argument of erl(1).
unsigned int driver_async_port_key (ErlDrvPort port)
This function calculates a key for later use in driver_async(). The keys are evenly distributed so that a fair mapping between port id's and async thread id's is achieved.
Note!
Before OTP-R16, the actual port id could be used as a key with proper casting, but after the rewrite of the port subsystem, this is no longer the case. With this function, you can achieve the same distribution based on port id's as before OTP-R16.
int driver_lock_driver(ErlDrvPort port)
This function locks the driver used by the port port
in memory for the rest of the emulator process'
lifetime. After this call, the driver behaves as one of Erlang's
statically linked in drivers.
ErlDrvPort driver_create_port(ErlDrvPort port, ErlDrvTermData owner_pid, char* name, ErlDrvData drv_data)
This function creates a new port executing the same driver code as the port creating the new port. A short description of the arguments:
port
owner_pid
driver_caller(port)
as owner_pid
.name
drv_data
The caller of driver_create_port()
is allowed to
manipulate the newly created port when driver_create_port()
has returned. When
port level locking
is used, the creating port is, however, only allowed to
manipulate the newly created port until the current driver
call-back that was called by the emulator returns.
Note!
When port level locking is used, the creating port is only allowed to manipulate the newly created port until the current driver call-back returns.
int erl_drv_thread_create(char *name, ErlDrvTid *tid, void * (*func)(void *), void *arg, ErlDrvThreadOpts *opts)
Arguments:
name
tid
func
arg
func
function.opts
NULL
.This function creates a new thread. On success 0
is returned;
otherwise, an errno
value is returned to indicate the error.
The newly created thread will begin executing in the function pointed
to by func
, and func
will be passed arg
as
argument. When erl_drv_thread_create()
returns the thread
identifier of the newly created thread will be available in
*tid
. opts
can be either a NULL
pointer, or a
pointer to an
ErlDrvThreadOpts
structure. If opts
is a NULL
pointer, default options
will be used; otherwise, the passed options will be used.
Warning!
You are not allowed to allocate the ErlDrvThreadOpts structure by yourself. It has to be allocated and initialized by erl_drv_thread_opts_create().
The created thread will terminate either when func
returns
or if
erl_drv_thread_exit()
is called by the thread. The exit value of the thread is either
returned from func
or passed as argument to
erl_drv_thread_exit().
The driver creating the thread has the responsibility of joining the
thread, via
erl_drv_thread_join(),
before the driver is unloaded. It is not possible to create
"detached" threads, i.e., threads that don't need to be joined.
Warning!
All created threads need to be joined by the driver before it is unloaded. If the driver fails to join all threads created before it is unloaded, the runtime system will most likely crash when the code of the driver is unloaded.
This function is thread-safe.
ErlDrvThreadOpts * erl_drv_thread_opts_create(char *name)
Arguments:
name
This function allocates and initialize a thread option
structure. On failure NULL
is returned. A thread option
structure is used for passing options to
erl_drv_thread_create().
If the structure isn't modified before it is passed to
erl_drv_thread_create(),
the default values will be used.
Warning!
You are not allowed to allocate the
ErlDrvThreadOpts
structure by yourself. It has to be allocated and
initialized by erl_drv_thread_opts_create()
.
This function is thread-safe.
void erl_drv_thread_opts_destroy(ErlDrvThreadOpts *opts)
Arguments:
opts
This function destroys thread options previously created by erl_drv_thread_opts_create().
This function is thread-safe.
void erl_drv_thread_exit(void *exit_value)
Arguments:
exit_value
NULL
.This function terminates the calling thread with the exit value passed as argument. You are only allowed to terminate threads created with erl_drv_thread_create(). The exit value can later be retrieved by another thread via erl_drv_thread_join().
This function is thread-safe.
int erl_drv_thread_join(ErlDrvTid tid, void **exit_value)
Arguments:
tid
exit_value
NULL
.This function joins the calling thread with another thread, i.e.,
the calling thread is blocked until the thread identified by
tid
has terminated. On success 0
is returned;
otherwise, an errno
value is returned to indicate the error.
A thread can only be joined once. The behavior of joining
more than once is undefined, an emulator crash is likely. If
exit_value == NULL
, the exit value of the terminated thread
will be ignored; otherwise, the exit value of the terminated thread
will be stored at *exit_value
.
This function is thread-safe.
ErlDrvTid erl_drv_thread_self(void)
This function returns the thread identifier of the calling thread.
This function is thread-safe.
int erl_drv_equal_tids(ErlDrvTid tid1, ErlDrvTid tid2)
Arguments:
tid1
tid2
This function compares two thread identifiers for equality,
and returns 0
it they aren't equal, and
a value not equal to 0
if they are equal.
Note!
A Thread identifier may be reused very quickly after
a thread has terminated. Therefore, if a thread
corresponding to one of the involved thread identifiers
has terminated since the thread identifier was saved,
the result of erl_drv_equal_tids()
might not give
the expected result.
This function is thread-safe.
ErlDrvMutex * erl_drv_mutex_create(char *name)
Arguments:
name
This function creates a mutex and returns a pointer to it. On
failure NULL
is returned. The driver creating the mutex
has the responsibility of destroying it before the driver is
unloaded.
This function is thread-safe.
void erl_drv_mutex_destroy(ErlDrvMutex *mtx)
Arguments:
mtx
This function destroys a mutex previously created by erl_drv_mutex_create(). The mutex has to be in an unlocked state before being destroyed.
This function is thread-safe.
void erl_drv_mutex_lock(ErlDrvMutex *mtx)
Arguments:
mtx
This function locks a mutex. The calling thread will be blocked until the mutex has been locked. A thread which currently has locked the mutex may not lock the same mutex again.
Warning!
If you leave a mutex locked in an emulator thread when you let the thread out of your control, you will very likely deadlock the whole emulator.
This function is thread-safe.
int erl_drv_mutex_trylock(ErlDrvMutex *mtx)
Arguments:
mtx
This function tries to lock a mutex. If successful 0
,
is returned; otherwise, EBUSY
is returned. A thread
which currently has locked the mutex may not try to
lock the same mutex again.
Warning!
If you leave a mutex locked in an emulator thread when you let the thread out of your control, you will very likely deadlock the whole emulator.
This function is thread-safe.
void erl_drv_mutex_unlock(ErlDrvMutex *mtx)
Arguments:
mtx
This function unlocks a mutex. The mutex currently has to be locked by the calling thread.
This function is thread-safe.
ErlDrvCond * erl_drv_cond_create(char *name)
Arguments:
name
This function creates a condition variable and returns a
pointer to it. On failure NULL
is returned. The driver
creating the condition variable has the responsibility of
destroying it before the driver is unloaded.
This function is thread-safe.
void erl_drv_cond_destroy(ErlDrvCond *cnd)
Arguments:
cnd
This function destroys a condition variable previously created by erl_drv_cond_create().
This function is thread-safe.
void erl_drv_cond_signal(ErlDrvCond *cnd)
Arguments:
cnd
This function signals on a condition variable. That is, if other threads are waiting on the condition variable being signaled, one of them will be woken.
This function is thread-safe.
void erl_drv_cond_broadcast(ErlDrvCond *cnd)
Arguments:
cnd
This function broadcasts on a condition variable. That is, if other threads are waiting on the condition variable being broadcast on, all of them will be woken.
This function is thread-safe.
void erl_drv_cond_wait(ErlDrvCond *cnd, ErlDrvMutex *mtx)
Arguments:
cnd
mtx
This function waits on a condition variable. The calling thread is blocked until another thread wakes it by signaling or broadcasting on the condition variable. Before the calling thread is blocked it unlocks the mutex passed as argument, and when the calling thread is woken it locks the same mutex before returning. That is, the mutex currently has to be locked by the calling thread when calling this function.
Note!
erl_drv_cond_wait()
might return even though
no-one has signaled or broadcast on the condition
variable. Code calling erl_drv_cond_wait()
should
always be prepared for erl_drv_cond_wait()
returning even though the condition that the thread was
waiting for hasn't occurred. That is, when returning from
erl_drv_cond_wait()
always check if the condition
has occurred, and if not call erl_drv_cond_wait()
again.
This function is thread-safe.
ErlDrvRWLock * erl_drv_rwlock_create(char *name)
Arguments:
name
This function creates an rwlock and returns a pointer to it. On
failure NULL
is returned. The driver creating the rwlock
has the responsibility of destroying it before the driver is
unloaded.
This function is thread-safe.
void erl_drv_rwlock_destroy(ErlDrvRWLock *rwlck)
Arguments:
rwlck
This function destroys an rwlock previously created by erl_drv_rwlock_create(). The rwlock has to be in an unlocked state before being destroyed.
This function is thread-safe.
void erl_drv_rwlock_rlock(ErlDrvRWLock *rwlck)
Arguments:
rwlck
This function read locks an rwlock. The calling thread will be blocked until the rwlock has been read locked. A thread which currently has read or read/write locked the rwlock may not lock the same rwlock again.
Warning!
If you leave an rwlock locked in an emulator thread when you let the thread out of your control, you will very likely deadlock the whole emulator.
This function is thread-safe.
int erl_drv_rwlock_tryrlock(ErlDrvRWLock *rwlck)
Arguments:
rwlck
This function tries to read lock an rwlock. If successful
0
, is returned; otherwise, EBUSY
is returned.
A thread which currently has read or read/write locked the
rwlock may not try to lock the same rwlock again.
Warning!
If you leave an rwlock locked in an emulator thread when you let the thread out of your control, you will very likely deadlock the whole emulator.
This function is thread-safe.
void erl_drv_rwlock_runlock(ErlDrvRWLock *rwlck)
Arguments:
rwlck
This function read unlocks an rwlock. The rwlock currently has to be read locked by the calling thread.
This function is thread-safe.
void erl_drv_rwlock_rwlock(ErlDrvRWLock *rwlck)
Arguments:
rwlck
This function read/write locks an rwlock. The calling thread will be blocked until the rwlock has been read/write locked. A thread which currently has read or read/write locked the rwlock may not lock the same rwlock again.
Warning!
If you leave an rwlock locked in an emulator thread when you let the thread out of your control, you will very likely deadlock the whole emulator.
This function is thread-safe.
int erl_drv_rwlock_tryrwlock(ErlDrvRWLock *rwlck)
Arguments:
rwlck
This function tries to read/write lock an rwlock. If successful
0
, is returned; otherwise, EBUSY
is returned.
A thread which currently has read or read/write locked the
rwlock may not try to lock the same rwlock again.
Warning!
If you leave an rwlock locked in an emulator thread when you let the thread out of your control, you will very likely deadlock the whole emulator.
This function is thread-safe.
void erl_drv_rwlock_rwunlock(ErlDrvRWLock *rwlck)
Arguments:
rwlck
This function read/write unlocks an rwlock. The rwlock currently has to be read/write locked by the calling thread.
This function is thread-safe.
int erl_drv_tsd_key_create(char *name, ErlDrvTSDKey *key)
Arguments:
name
key
This function creates a thread specific data key. On success
0
is returned; otherwise, an errno
value is returned
to indicate the error. The driver creating the key has the
responsibility of destroying it before the driver is unloaded.
This function is thread-safe.
void erl_drv_tsd_key_destroy(ErlDrvTSDKey key)
Arguments:
key
This function destroys a thread specific data key
previously created by
erl_drv_tsd_key_create().
All thread specific data using this key in all threads
have to be cleared (see
erl_drv_tsd_set())
prior to the call to erl_drv_tsd_key_destroy()
.
Warning!
A destroyed key is very likely to be reused soon. Therefore, if you fail to clear the thread specific data using this key in a thread prior to destroying the key, you will very likely get unexpected errors in other parts of the system.
This function is thread-safe.
void erl_drv_tsd_set(ErlDrvTSDKey key, void *data)
Arguments:
key
data
key
in calling thread.
This function sets thread specific data associated with
key
for the calling thread. You are only allowed to set
thread specific data for threads while they are fully under your
control. For example, if you set thread specific data in a thread
calling a driver call-back function, it has to be cleared, i.e.
set to NULL
, before returning from the driver call-back
function.
Warning!
If you fail to clear thread specific data in an emulator thread before letting it out of your control, you might not ever be able to clear this data with later unexpected errors in other parts of the system as a result.
This function is thread-safe.
void * erl_drv_tsd_get(ErlDrvTSDKey key)
Arguments:
key
This function returns the thread specific data
associated with key
for the calling thread.
If no data has been associated with key
for
the calling thread, NULL
is returned.
This function is thread-safe.
int erl_drv_putenv(char *key, char *value)
Arguments:
key
value
This function sets the value of an environment variable.
It returns 0
on success, and a value != 0
on
failure.
Note!
The result of passing the empty string ("") as a value is platform dependent. On some platforms the value of the variable is set to the empty string, on others, the environment variable is removed.
Warning!
Do not use libc's putenv
or similar
C library interfaces from a driver.
This function is thread-safe.
int erl_drv_getenv(char *key, char *value, size_t *value_size)
Arguments:
key
value
value_size
This function retrieves the value of an environment variable.
When called, *value_size
should contain the size of
the value
buffer. On success 0
is returned,
the value of the environment variable has been written to
the value
buffer, and *value_size
contains the
string length (excluding the terminating null character) of
the value written to the value
buffer. On failure,
i.e., no such environment variable was found, a value less than
0
is returned. When the size of the value
buffer is too small, a value greater than 0
is returned
and *value_size
has been set to the buffer size needed.
Warning!
Do not use libc's getenv
or similar
C library interfaces from a driver.
This function is thread-safe.
int erl_drv_consume_timeslice(ErlDrvPort port, int percent)
Arguments:
port
percent
Give the runtime system a hint about how much CPU time the
current driver callback call has consumed since last hint, or
since the start of the callback if no previous hint has been given.
The time is given as a fraction, in percent, of a full time-slice
that a port is allowed to execute before it should surrender the
CPU to other runnable ports or processes. Valid range is
[1, 100]
. The scheduling time-slice is not an exact entity,
but can usually be approximated to about 1 millisecond.
Note that it is up to the runtime system to determine if and
how to use this information. Implementations on some platforms
may use other means in order to determine the consumed fraction
of the time-slice. Lengthy driver callbacks should regardless of
this frequently call the erl_drv_consume_timeslice()
function in order to determine if it is allowed to continue
execution or not.
erl_drv_consume_timeslice()
returns a non-zero value
if the time-slice has been exhausted, and zero if the callback is
allowed to continue execution. If a non-zero value is
returned the driver callback should return as soon as possible in
order for the port to be able to yield.
This function is provided to better support co-operative scheduling, improve system responsiveness, and to make it easier to prevent misbehaviors of the VM due to a port monopolizing a scheduler thread. It can be used when dividing length work into a number of repeated driver callback calls without the need to use threads. Also see the important warning text at the beginning of this document.
char * erl_drv_cond_name(ErlDrvCond *cnd)
Arguments:
cnd
Returns a pointer to the name of the condition.
Note!
This function is intended for debugging purposes only.
char * erl_drv_mutex_name(ErlDrvMutex *mtx)
Arguments:
mtx
Returns a pointer to the name of the mutex.
Note!
This function is intended for debugging purposes only.
char * erl_drv_rwlock_name(ErlDrvRWLock *rwlck)
Arguments:
rwlck
Returns a pointer to the name of the r/w-lock.
Note!
This function is intended for debugging purposes only.
char * erl_drv_thread_name(ErlDrvTid tid)
Arguments:
tid
Returns a pointer to the name of the thread.
Note!
This function is intended for debugging purposes only.
SEE ALSO
driver_entry(3), erl_ddll(3), erlang(3)
An Alternative Distribution Driver (ERTS User's Guide Ch. 3)