This patch improves the handling of the corner case where a client
vanishes while a session request is in flight (CREATE_REQUESTED but
not yet AVAILABLE). This corner case could be sporadically observed with
the init_loop test on base-linux.
In the original version, the session would eventually be delivered but
never picked up by anyone. Such a stale session still uses resources that
should better be released. In the new version, the parent checks for the
liveliness of the client whenever a session is delivered. If there is no
client of the session, a close request is immediately issued to the
server. The session state must be preserved until the close requests has
been answered.
This patch moves the closing of a child's PD session from the 'Child'
destructor to the 'close_all_sessions' method. This way, the child's
PD quota is immediately returned as soon as init flags a child as
'abandoned', which removes jitter from init's RAM-state reports.
The patch is supposed to make the 'init_loop.run' test much happier.
This patch ensures that session-create requests are handled after close
requests, which solves the corner case where one single-session server
receives a close and create request at the same time. E.g., when
expanding a partition with Sculpt, the gpt-write tool is removed and
part_blk is started (to execute resize2fs on top) in one step. Both
interact with the USB-block driver, which is a single-session server.
Fixes#2877
Internally, Genode::Session_label has a typedef from String<capacity()> to
String. To have this typedef public is especially useful when reading a
label from an XML node. This can then be written as
! Session_label label = node.attribute_value("label", Session_label::String());
instead of
! Session_label label = node.attribute_value("label", String<160>());
which would be less generic (Session_label cannot be used directly as there
is no appropriate ascii_to implementation).
Issue #2840
Whenever an environment session was provided by an asynchronous service,
e.g., the depot_rom of the sculpt scenario, the session quota was not
transferred to the server at session-creation time. This resulted in a
slow depletion of the server's quota over time. This patch ensures that
the delivery of session quota is consistent with the information
reported to the server as session argument.
The 'Child::_revert_quota_and_destroy' assumes to be called from the
client's context, which is normally the case when destroying sessions.
However, if a client's session outlives the client (because the
asynchronous close request to the server is still pending), the session
cleanup is performed in the context of the server. Here, the
'session_response' implementation wrongly called
'_revert_quota_and_destroy' to the effect that the session quota was
withdrawn from the server (good) but subsequently transferred back to
the server (bad). The patch replaces the call of
'_revert_quota_and_destroy' with only the first - correct - part of the
transaction.
In case of an Alloc_return error, don't free the address at 0. If this
is a managed dataspace, there could be valid attachments which then gets
lost.
Make decreasing alignment search working. Continue with lower alignment
restrictions in case the optimal alignment wish can't be fulfilled.
When an environment session is provided by a async service such as a
sibling component, the session metadata must be preserved until end of
the lifetime of the session at the server has been acknowledged by the
server. Since the session meta data of env sessions are always part of
the 'Child' object, the destruction of this object must be deferred
until this point.
This patch makes the 'List_model' utility robust against duplicated
occurrences of node IDs in the supplied XML data. If two or more XML
nodes correspond to the same model element, the existing element is
updated with the information of the subsequent XML nodes.
Issue #2778
The "create_grub2" script functionality is moved to the g2fg contrib sources
and also the GRUB2 head image (tool/grub2-head.img). The head image is now
partitioned as GPT and bootable in BIOS legacy mode.
The sole existence of shared-object dependencies lead to fatal
static-constructor errors before. Now, ldso checks if the ctors section
of objects in the init list are non-empty before whining.
Issue #2759
When the former trace buffer implementation wrapped, the last entry
according to commit order couldn't be detected anymore. Now, the last
committed entry is always followed by an entry with length 0.
As a downside of this, there are now two meanings of "last" entry: It
means either that the entry marks the empty padding after the entry with
the highest memory address or that it actually marks the end of the
buffer according to commit order. This is an example state of the buffer
with the two types of "last" entry:
last last
+-------------+------------+---+---------+-------------+------------+---+-------+
| len3 data3 | len4 data4 | 0 | empty | len1 data1 | len2 data2 | 0 | empty |
+-------------+------------+---+---------+-------------+------------+---+-------+
If the entry with the highest memory address fits perfectly, the first
type of "last" entry is not needed:
last
+------------+--------------------+---+-------+-------------+-------------------+
| len3 data3 | len4 data4 | 0 | empty | len1 data1 | len2 data2 |
+------------+--------------------+---+-------+-------------+-------------------+
If the buffer didn't wrap so far, there is only one "last" entry that
has both meanings:
last
+--------------------------+------------+-------------+---+---------------------+
| len1 data1 | len2 data2 | len3 data3 | 0 | empty |
+--------------------------+------------+-------------+---+---------------------+
Issue #2735
Co-authored-by: Martin Stein <martin.stein@genode-labs.com>
The Ada runtime can now be set with the ADA_RTS variable. The ada
library builds a (currently) minimal runtime from the gcc sources that
come with Genode (to stay consistent with the used compiler) and sets
the runtime path accordingly. It is build as a shared library ada.lib.so
which needs to be added to the build files.
I split the existing Ada test into program and library, and moved it
to libports as it depends on the runtime library residing in this
repository too.
Fixes#2748
Integrate the code of the Alarm framework directly into the Timeout
framework. The former Alarm-framework methods are all private to the
corresponding classes of the Timeout framework and get prefixed with
'_alarm__'. The latter avoids name clashes and makes it easier to
simplify the code later.
Issue #2704
Switch port I/O based PCI config space access to memory-mapped IO. The
base address of the PCI configuration space is acquired by mapping the
ACPI ROM and reading the first <bdf> node. An exception is thrown if the
first <bdf> node is not for PCI domain zero or if multiple <bdf> nodes
exist. This is to reduce complexity and also because multiple PCI
domains are rare.
The PCI configuration space is accessed via I/O mem dataspace which is
created in the platform_drv root and then passed on to the PCI session,
device components and finally to the actual PCI config access instances.
The memory access code is implemented in a way to make it work with Muen
subject monitor (SM) device emulation and also general x86 targets. On
Muen, the simplified device emulation code (which works also for Linux)
always returns 0xffff in EAX to indicate a non-existing device.
Therefore, EAX is enforced in the assembly templates.
Fixes#2547
AVL trees can't be copied with the default copy constructor as the
parent pointer of the first item of both of the resulting trees would
point to the original tree. Copying an AVL node, however, generally
violates the integrity of the corresponding tree. The copy constructor
of Avl_tree is used in some places but in those places it can be
replaced easily. So, this commit deletes the copy constructor of
Avl_node_base which makes Avl_node and Avl_tree non-copyable.
Issue #2654
On platforms without nic or audio driver support, the corresponding
functions to return the correct names should return a meaningful
name instead of an empty list. Otherwise, in an erroneous run-script
that calls those functions to append names to the boot modules
the whole 'bin' directory is appended instead, which leads to big
image files or errors when linking the boot modules to core.
Previously, base/trace/buffer.h included base/thread.h which includes
base/trace/logger.h which includes base/trace/buffer.h.
Removed the base/thread.h include in base/trace/buffer.h as it is not
needed.
Issue #2654
With this patch, init responds to the exit of a child by closing all
sessions of the child. E.g., if a child is a GUI application, its
nitpicker session is closed at the time of exit, not at the time when
the start node disappears from init's configuration.
Since this change requires a modification of the 'Genode::Child' class,
it takes the chance to make the child-destruction less brutal. The
new version ensures that all threads of the destructed subsystem are
destructed before other sessions, in particular PD sessions. This
eliminates spurious page-fault warnings during the child destruction.
On Fiasco.OC, closing the CPU session of a thread while being called by
the thread causes a deadlock. Hence, we skip the eager destruction of
CPU sessions on this kernel.
Related to issue #2659
Previously, the dst_len value was not decreased after each character that was
written to the dst buffer. This way, if the content length was greater than
dst_len, decoded_content wrote to memory out of bounds.
Issue #2644
Do not leave space for a terminating '0' at the end of the dst buffer in
decoded_content as the method does not write this '0'. The caller of the
method shall take care of it instead.
Issue #2644
This patch enables the warnings -Wextra, -Weffc++, and -Werror for
compiling Genode components. It thereby helps us to detect bugs like
uninitialized member variables or missing virtual destructors at compile
time. The warning level is defined via the new 'CC_CXX_WARN_STRICT'
variable. For targets that compile 3rd-party code where this warning
level is not applicable, the variable may be explictly set to an empty
value in the corresponding build-description file.
Issue #465
The patch adjust the code of the base, base-<kernel>, and os repository.
To adapt existing components to fix violations of the best practices
suggested by "Effective C++" as reported by the -Weffc++ compiler
argument. The changes follow the patterns outlined below:
* A class with virtual functions can no longer publicly inherit base
classed without a vtable. The inherited object may either be moved
to a member variable, or inherited privately. The latter would be
used for classes that inherit 'List::Element' or 'Avl_node'. In order
to enable the 'List' and 'Avl_tree' to access the meta data, the
'List' must become a friend.
* Instead of adding a virtual destructor to abstract base classes,
we inherit the new 'Interface' class, which contains a virtual
destructor. This way, single-line abstract base classes can stay
as compact as they are now. The 'Interface' utility resides in
base/include/util/interface.h.
* With the new warnings enabled, all member variables must be explicitly
initialized. Basic types may be initialized with '='. All other types
are initialized with braces '{ ... }' or as class initializers. If
basic types and non-basic types appear in a row, it is nice to only
use the brace syntax (also for basic types) and align the braces.
* If a class contains pointers as members, it must now also provide a
copy constructor and assignment operator. In the most cases, one
would make them private, effectively disallowing the objects to be
copied. Unfortunately, this warning cannot be fixed be inheriting
our existing 'Noncopyable' class (the compiler fails to detect that
the inheriting class cannot be copied and still gives the error).
For now, we have to manually add declarations for both the copy
constructor and assignment operator as private class members. Those
declarations should be prepended with a comment like this:
/*
* Noncopyable
*/
Thread(Thread const &);
Thread &operator = (Thread const &);
In the future, we should revisit these places and try to replace
the pointers with references. In the presence of at least one
reference member, the compiler would no longer implicitly generate
a copy constructor. So we could remove the manual declaration.
Issue #465
The run tool now by default checks configurations with target-specific
XML schemata. Each component may define a config schema file in its
target.mk via the CONFIG_XSD variable. When the run tool has checked an
configuration of an init instance, it additionally goes through the
start nodes of the config. For each start node it checks whether there
is an XSD file that matches. If so, the run tool also checks the config
of the start node (if existant). This is done recursively. I.e., also
the child configs of a sub-init of a sub-init of the top-level init
receive a config check.
Issue #2600
In the past, we had the lock of the context we found for the rest of the scope
of pending_signal. Now we use for_each_locked to find the context and the
lock is released as soon as we return from for_each_locked. Thus, we need to
lock the context again before returning it to avoid that the copy constructor
during the return reads values that are currently changing.
Fixes#2532
As noted above the former enum for the local-attachment address we
discovered address clashes on current Linux installations, esp. 32-bit
runtime on 64-bit Linux. The local_attach_addr is now configurable in
the run script and the memory maps heuristics were removed.
* Instead of always re-load page-tables when a thread context is switched
only do this when another user PD's thread is the next target,
core-threads are always executed within the last PD's page-table set
* remove the concept of the mode transition
* instead map the exception vector once in bootstrap code into kernel's
memory segment
* when a new page directory is constructed for a user PD, copy over the
top-level kernel segment entries on RISCV and X86, on ARM we use a designated
page directory register for the kernel segment
* transfer the current CPU id from bootstrap to core/kernel in a register
to ease first stack address calculation
* align cpu context member of threads and vms, because of x86 constraints
regarding the stack-pointer loading
* introduce Align_at template for members with alignment constraints
* let the x86 hardware do part of the context saving in ISS, by passing
the thread context into the TSS before leaving to user-land
* use one exception vector for all ARM platforms including Arm_v6
Fix#2091
* introduce new syscall (core-only) to create privileged threads
* take the privilege level of the thread into account
when doing a context switch
* map kernel segment as accessable for privileged code only
Ref #2091
* introduces central memory map for core/kernel
* on 32-bit platforms the kernel/core starts at 0x80000000
* on 64-bit platforms the kernel/core starts at 0xffffffc000000000
* mark kernel/core mappings as global ones (tagged TLB)
* move the exception vector to begin of core's binary,
thereby bootstrap knows from where to map it appropriately
* do not map boot modules into core anymore
* constrain core's virtual heap memory area
* differentiate in between user's and core's main thread's UTCB,
which now resides inside the kernel segment
Ref #2091
In the past, a signal context, that was chosen for handling by
'Signal_receiver::pending_signal and always triggered again before
the next call of 'pending_signal', caused all other contexts behind
in the list to starve. This was the case because 'pending_signal'
always took the first pending context in its context list.
We avoid this problem now by handling pending signals in a round-robin
fashion instead.
Ref #2532
Ensure that the timer does not handle timeouts again within 1000
microseconds after the last handling of timeouts. This makes denial of
service attacks harder. This commit does not limit the rate of timeout
signals handled inside the timer but it causes the timer to do it less
often. If a client continuously installs a very small timeout at the
timer it still causes a signal to be submitted to the timer each time
and some extra CPU time to be spent in the internal handling method. But
only every 1000 microseconds this internal handling causes user timeouts
to trigger.
If we would want to limit also the call of the internal handling method
to ensure that CPU time is spent beside the RPCs only every 1000
microseconds, things would get more complex. For instance, on NOVA
Time_source::schedule_timeout(0) must be called each time a new timeout
gets installed and becomes head of the scheduling queue. We cannot
simply overwrite the already running timeout with the new one.
Ref #2490
This patch merges two similar rules, which create content at 'include'
into a single rule. This prevents a possible race condition when
creating archives in parallel.
We moved the stack-area segment 128 MiB behind text and data to comply
with assumptions in the kernel ELF loader.
This commit also reenables static binaries on linux and removes the
unused stack_area.stdlib.ld script.
Fixes#2521
In nested scenarios like driver_manager.run, the initial session quota
for IO_PORT, IO_PORT, and IRQ sessions is expectedly insufficient.
However, the condition is properly handled by re-attemping the request
with a slightly increased quota. Still, core prints a warning each time
the request is denied for quota reasons, which spams the log. This patch
removes the non-critical message.
This should actually never happen. However if it happens, be a bit robuster
and don't provide the memory for re-use (which causes tons of other trouble
afterwards).
Issue #2505
There are hardware timers whose frequency can't be expressed as
ticks-per-microsecond integer-value because only a ticks-per-millisecond
integer-value is precise enough. We don't want to use expensive
floating-point values here but nonetheless want to translate from ticks
to time with microseconds precision. Thus, we split the input in two and
translate both parts separately. This way, we can raise precision by
shifting the values to their optimal bit position. Afterwards, the results
are shifted back and merged together again.
As this algorithm is not so trivial anymore and used by at least three
timer drivers (base-hw/x86_64, base-hw/cortex_a9, timer/pit), move it to a
generic header to avoid redundancy.
Ref #2400
When building Genode on a Linux system running in a Xen Dom0, the 'xen'
run target can run a Genode scenario in a Xen DomU.
Usage: in build/x86_*/etc/build.conf, define:
RUN_OPT = --include boot_dir/$(KERNEL) --include image/iso --include power_on/xen --include log/xen --include power_off/xen
The Xen DomU runs in HVM mode and loads Genode from an ISO image. Serial
log output is printed to the console and graphical output is shown in an
SDL window.
The Xen DomU ist managed using the 'xl' command line tool and it is
possible to add configuration options in the 'xen_args' variable in a run
script. Common options are:
- disabling the graphical output:
append xen_args { sdl="0" }
- configuring a network device:
append xen_args { vif=\["model=e1000,mac=02:00:00:00:01:01,bridge=xenbr0"\] }
- configuring USB input devices:
append xen_args { usbdevice=\["mouse","keyboard"\] }
Note: the 'xl' tool requires super-user permissions and interactive
password input can be troublesome in combination with 'expect' and is not
practical for automatic tests. For this reason, the current implementation
assumes that no password input is needed when running 'sudo xl', which can
be achieved by creating a file '/etc/sudoers.d/xl' with the content
'user ALL=(root) NOPASSWD: /usr/sbin/xl'
(where 'user' is the Linux user name).
Fixes#2504
When running core as the kernel inside every component, a separate
stack area for core is needed that is different from the user-land
component's one.
Ref #2091
Acquire Signal_context objects locks via Object_pool::apply() in the
context of the entrpyoint thread, instead in the context of the calling
thread.
Fixes#2485
- Fix fatal exception handling so that stack traces are dumped
- Add 'include/nim' directories to Nim module search path
- Enable release optimizations for release builds
Fix#2493
This patch removes the assertion about the unexpected call of
'block_for_signal' within core. On Linux, this call is actually
expected because of the handling of SIGCHLD signals by core.
A boot module with size 0 previously made Core crash with a page fault in
Region_map_component::attach. This patch prevents the creation of ROM-FS
entries for such modules.
Ref #2490
For most base platforms (except linux and sel4), the initialization of
boot modules is the same. Thus, merge this default implementation in the
new unit base/src/core/platform_rom_modules.cc.
Ref #2490
In Region_map_component::attach, storing the metadata for a region may
throw an exception. Catch it and throw an Invalid_dataspace exception.
Ref #2490
This is helpful for disabling messages in etc/tools.conf by
setting it to e.g.
MSG_LINK = @true ""
This results in much shorter and less cluttered logs in automatic
builds.
- factor out Rm_client::pager lambda code into utility
Region_map_component::create_map_item
- use utility to find/lookup physical addresses to be mapped eagerly
Issue #2209
Platform_pd "_pd" uses a allocator for, which relies on the mapped RAM
dataspace within core. Unfortunately the RAM dataspaces are already freed up
during _ram_ds_factory destruction, which may lead to trouble if accessed
afterwards.
Issue #2451
This patch sets the -march complile flag in spec/arm_v7a.mk, which
enables us to build depot archives for the 'arm_v7a' architecture.
It also removes copy-pasted comments that offer no valuable insights but
contain grammar errors.
This patch decouples the error handling of the quota transfers
and the actual session creation. In the previous version, an error in
the 'initiate_request' phase would leave the local scope via an
exception without disarming the transfer guard objects. This way,
the guard destructors would attempt the returning of session quota in
addition to the explicit call of '_revert_quota_and_destroy' as done in
the error handling of the 'initiate_request' operation.
In the presence of a session-creation error in the 'initiate_request'
phase, session quota would eventually be returned twice. This patch
removes the intertwined error handling of both phases in a way that the
guards of the first phase (quota transfer) are no longer present in the
second phase (initiate_request).
This patch makes sure that the initial PD session limit (as defined by
the client-provided session quota) is preserved over the entire lifetime
of the PD session. That means, it cannot be transferred to other PD
sessions. Otherwise, it may be impossive to hand back all the static
session quota to the PD-session client at session-destruction time
because parts of the initial quota would no longer belong to the
session.
Note that the initial limit can still be used for allocations within the
PD session as those allocations are automatically reverted at
session-destruction time.
The implementations of the lock and C++ guards tests depend on
thread-execution priorities, which produces false negatives of the whole
thread test on platforms without priority support.
The recently implemented capability resource trading scheme unfortunately
broke the automated capability memory upgrade mechanism needed by base-hw
kernel/core. This commit splits the capability memory upgrade mechanism
from the PD session ram_quota upgrade, and moves that functionality
into a separate Pd_session::Native_pd interface.
Ref #2398
A dataspace capability request to a ROM service may invalidate any
previously issued dataspace. Therefor no requests should be made while a
session dataspace is mapped. Reducing calls to the session also improves
performance where servicing a ROM request has a significant cost.
Fix#2418
The 'Stack_area_ram_session' is now a 'Stack_area_ram_allocator', which
simplifies the code and remove a dependency from the 'Ram_session'
interface, which we want to remove after all.
Issue #2407
By supplying a statically allocated initial block to the slab allocator
for signal contexts, we become able to construct a 'Signal_broker' (the
back end for the PD's signalling API) without any dynamic memory
allocation. This is a precondition for using the PD as meta-data
allocator for its contained signal broker (meta data allocations must
not happen before the PD construction is complete).
Issue #2407
By separating the session-interface concerns from the mechanics of the
dataspace creation, the code becomes simpler to follow, and the RAM
session can be more easily merged with the PD session in a subsequent
step.
Issue #2407
This patch allows core's 'Signal_transmitter' implementation to sidestep
the 'Env::Pd' interface and thereby adhere to a stricter layering within
core. The 'Signal_transmitter' now uses - on kernels that depend on it -
a dedicated (and fairly freestanding) RPC proxy mechanism for signal
deliver, instead of channeling signals through the 'Pd_session::submit'
RPC function.
Previously, the Genode::Timer::curr_time always used the
Timer_session::elapsed_ms RPC as back end. Now, Genode::Timer reads
this remote time only in a periodic fashion independently from the calls
to Genode::Timer::curr_time. If now one calls Genode::Timer::curr_time,
the function takes the last read remote time value and adapts it using
the timestamp difference since the remote-time read. The conversion
factor from timestamps to time is estimated on every remote-time read
using the last read remote-time value and the timestamp difference since
the last remote time read.
This commit also re-works the timeout test. The test now has two stages.
In the first stage, it tests fast polling of the
Genode::Timer::curr_time. This stage checks the error between locally
interpolated and timer-driver time as well as wether the locally
interpolated time is monotone and sufficiently homogeneous. In the
second stage several periodic and one-shot timeouts are scheduled at
once. This stage checks if the timeouts trigger sufficiently precise.
This commit adds the new Kernel::time syscall to base-hw. The syscall is
solely used by the Genode::Timer on base-hw as substitute for the
timestamp. This is because on ARM, the timestamp function uses the ARM
performance counter that stops counting when the WFI (wait for
interrupt) instruction is active. This instruction, however is used by
the base-hw idle contexts that get active when no user thread needs to
be scheduled. Thus, the ARM performance counter is not a good choice for
time interpolation and we use the kernel internal time instead.
With this commit, the timeout library becomes a basic library. That means
that it is linked against the LDSO which then provides it to the program it
serves. Furthermore, you can't use the timeout library anymore without the
LDSO because through the kernel-dependent LDSO make-files we can achieve a
kernel-dependent timeout implementation.
This commit introduces a structured Duration type that shall successively
replace the use of Microseconds, Milliseconds, and integer types for duration
values.
Open issues:
* The timeout test fails on Raspberry PI because of precision errors in the
first stage. However, this does not render the framework unusable in general
on the RPI but merely is an issue when speaking of microseconds precision.
* If we run on ARM with another Kernel than HW the timestamp speed may
continuously vary from almost 0 up to CPU speed. The Timer, however,
only uses interpolation if the timestamp speed remained stable (12.5%
tolerance) for at least 3 observation periods. Currently, one period is
100ms, so its 300ms. As long as this is not the case,
Timer_session::elapsed_ms is called instead.
Anyway, it might happen that the CPU load was stable for some time so
interpolation becomes active and now the timestamp speed drops. In the
worst case, we would now have 100ms of slowed down time. The bad thing
about it would be, that this also affects the timeout of the period.
Thus, it might "freeze" the local time for more than 100ms.
On the other hand, if the timestamp speed suddenly raises after some
stable time, interpolated time can get too fast. This would shorten the
period but nonetheless may result in drifting away into the far future.
Now we would have the problem that we can't deliver the real time
anymore until it has caught up because the output of Timer::curr_time
shall be monotone. So, effectively local time might "freeze" again for
more than 100ms.
It would be a solution to not use the Trace::timestamp on ARM w/o HW but
a function whose return value causes the Timer to never use
interpolation because of its stability policy.
Fixes#2400
This patch make sure that a once managed parent RPC object will always be
dissolved if an exception during the remaining child construction
occurs. The original version would miss the dissolve call if one of the
subsequent members throws an exception at construction time.
This patch eases the debugging of situations where a session-object
constructor wrongly throws an exception type not specified in the
'Local_service::Factory' interface.
This patch reduces the number of exception types by facilitating
globally defined exceptions for common usage patterns shared by most
services. In particular, RPC functions that demand a session-resource
upgrade not longer reflect this condition via a session-specific
exception but via the 'Out_of_ram' or 'Out_of_caps' types.
Furthermore, the 'Parent::Service_denied', 'Parent::Unavailable',
'Root::Invalid_args', 'Root::Unavailable', 'Service::Invalid_args',
'Service::Unavailable', and 'Local_service::Factory::Denied' types have
been replaced by the single 'Service_denied' exception type defined in
'session/session.h'.
This consolidation eases the error handling (there are fewer exceptions
to handle), alleviates the need to convert exceptions along the
session-creation call chain, and avoids possible aliasing problems
(catching the wrong type with the same name but living in a different
scope).
This patch mirrors the accounting and trading scheme that Genode employs
for physical memory to the accounting of capability allocations.
Capability quotas must now be explicitly assigned to subsystems by
specifying a 'caps=<amount>' attribute to init's start nodes.
Analogously to RAM quotas, cap quotas can be traded between clients and
servers as part of the session protocol. The capability budget of each
component is maintained by the component's corresponding PD session at
core.
At the current stage, the accounting is applied to RPC capabilities,
signal-context capabilities, and dataspace capabilities. Capabilities
that are dynamically allocated via core's CPU and TRACE service are not
yet covered. Also, the capabilities allocated by resource multiplexers
outside of core (like nitpicker) must be accounted by the respective
servers, which is not covered yet.
If a component runs out of capabilities, core's PD service prints a
warning to the log. To observe the consumption of capabilities per
component in detail, the PD service is equipped with a diagnostic
mode, which can be enabled via the 'diag' attribute in the target
node of init's routing rules. E.g., the following route enables the
diagnostic mode for the PD session of the "timer" component:
<default-route>
<service name="PD" unscoped_label="timer">
<parent diag="yes"/>
</service>
...
</default-route>
For subsystems based on a sub-init instance, init can be configured
to report the capability-quota information of its subsystems by
adding the attribute 'child_caps="yes"' to init's '<report>'
config node. Init's own capability quota can be reported by adding
the attribute 'init_caps="yes"'.
Fixes#2398
This patch reworks the implementation of core's RAM service to make use
of the 'Session_object' and to remove the distinction between the
"metadata" quota and the managed RAM quota. With the new implementation,
the session implicitly allocates its metadata from its own account. So
there is not need to handle 'Out_of_metadata' and 'Quota_exceeded' via
different exceptions. Instead, the new version solely uses the
'Out_of_ram' exception.
Furthermore, the 'Allocator::Out_of_memory' exception has become an alias
for 'Out_of_ram', which simplifies the error handling.
Issue #2398
The 'Session_object' unifies several aspects of server-component
implementations:
* It keeps track of session quotas and is equipped with standardized
interfaces (Quota_guard) to upgrade (and in the future potentially
downgrade) session quotas in a uniform way.
* It follows the pattern of modern RPC objects / signal handlers that
manage/dissolve themselves at the entrypoint given as constructor
argument. Thereby, the relationship with its entrypoint is always
coupled with the lifetime of the session-component object.
* It stores the session label, which was previously done manually by
most but not all server-component implementations.
* It stores the session 'diag' flag.
* It is equipped with output methods 'diag', 'error', and 'warning'.
All messages printed from the context of a session component is
automatically prefixed with the session type and client label.
Messages passed via 'diag' are only printed if the 'diag' flag of
the session is set.
Issue #2398
The 'diag' flag can be defined by a target node of a route in init's
configuration. It is propagated as session argument to the server, which
may evaluate the flag to enable diagnostic output for the corresponding
session.
Issue #2398
This patch makes use of the new 'Quota_transfer::Account' by the service
types in base/service.h and uses 'Quota_transfer' objects in
base/child.cc and init/server.cc.
Furthermore, it decouples the notion of an 'Async_service' from
'Child_service'. Init's 'Routed_service' is no longer a 'Child_service'
but is based on the new 'Async_service' instead.
With this patch in place, quota transfers do no longer implicitly use
'Ram_session_client' objects. So transfers can in principle originate
from component-local 'Ram_session_component' objects, e.g., as used by
noux. Therefore, this patch removes a strumbling block for turning noux
into a single threaded component in the future.
Issue #2398
The 'Quota_transfer' helper facilitated the implementation of quota
transfers between components in a transactional manner. It is designated
for framework-internal use (replacing the 'Transfer' class in child.h).
However, since it is also useful for init, we make it publicly
available.
The 'Quota_transfer::Account' class serves as an interface representing
the donor or receiver of quotas (parent, service, client).
Issue #2398
This patch replaces the 'Parent::Quota_exceeded',
'Service::Quota_exceeded', and 'Root::Quota_exceeded' exceptions
by the single 'Insufficient_ram_quota' exception type.
Furthermore, the 'Parent' interface distinguished now between
'Out_of_ram' (the child's RAM is exhausted) from
'Insufficient_ram_quota' (the child's RAM donation does not suffice to
establish the session).
This eliminates ambiguities and removes the need to convert exception
types along the path of the session creation.
Issue #2398
This patch adds sanity checks to the RPC entrypoint that detect attempts
to manage or dissolve the same RPC object twice. This is not always a
bug. I.e., if RPC objects are implemented in the modern way where the
object manages/dissolves itself. As the generic framework code (in
particular root/component.h) cannot rely on this pattern, it has to
call manage/dissolve for session objects anyway. For modern session
objects, this double attempt would result in a serious error (double
insertion into the object pool's AVL tree).
Issue #2398
This patch replaces the former use of size_t with the use of the
'Ram_quota' type to improve type safety (in particular to avoid
accidentally mixing up RAM quotas with cap quotas).
Issue #2398
The 'Ram_allocator' interface contains the subset of the RAM session
interface that is needed to satisfy the needs of the 'Heap' and
'Sliced_heap'. Its small size makes it ideal for intercepting memory
allocations as done by the new 'Constrained_ram_allocator' wrapper
class, which is meant to replace the existing 'base/allocator_guard.h'
and 'os/ram_session_guard.h'.
Issue #2398
This patch augments the existing session/session.h with useful types for
the session creation:
* The new 'Insufficient_ram_quota' and 'Insufficient_cap_quota'
exceptions are meant to supersede the old 'Quota_exceeded' exception
of the 'Parent' and 'Root' interfaces.
* The 'Session::Resources' struct subsumes the information about the
session quota provided by the client.
* The boolean 'Session::Diag' type will allow sessions to operate in a
diagnostic mode.
* The existing 'Session_label' is not also available under the alias
'Session::Label'.
* A few helper functions ease the extraction of typed session arguments
from the session-argument string.
Issue #2398
This accessor is useful to eagerly expand the slab with new slab blocks,
side stepping the slab's built-in policy for the allocation of new slab
blocks.
This is particularly important when using the slab for allocating the
cap space meta-data for the base-hw kernel. To guarantee that the slab
gets never exhausted in the kernel, it is expanded before entering the
kernel.
With the introduction of the 'Out_of_caps' exception type, the slab
needs to consider exceptions during the call of '_new_slab_block' by
reverting the 'nested' state.
This commit moves the headers residing in `repos/base/include/spec/*/drivers`
to `repos/base/include/drivers/defs` or repos/base/include/drivers/uart`
respectively. The first one contains definitions about board-specific MMIO
iand RAM addresses, or IRQ lines. While the latter contains device driver
code for UART devices. Those definitions are used by driver implementations
in `repos/base-hw`, `repos/os`, and `repos/dde-linux`, which now need to
include them more explicitely.
This work is a step in the direction of reducing 'SPEC' identifiers overall.
Ref #2403
For asynchronously provided sessions, the parent has to maintain the
session state as long as the server hasn't explicitly responded to a
close request. For this reason, the lifetime of such session states is
bound to the server, not the client.
When the server responds to a close request, the session state gets
freed. The 'session_response' implementation does not immediately
destroy the session state but delegates the destruction to a client-side
callback, which thereby also notifies the client. However, the code did
not consider the case where the client has completely vanished at
session-response time. In this case, we need to drop the session state
immediately.
Fixes#2391
Originally, the spec files for less specific SPEC values were include
via the 'select_from_repositories' function. This implies that BASE_DIR
must always be present in the list of 'REPOSITORIES'. Otherwise the
spec files won't be found. By explicitly including sub specs from
'$(BASE_DIR)/mk', we lift this restriction.
The <build-dir>/bin/ directory used to contain symbolic links to the
unstripped build results. However, since the upcoming depot tool
extracts the content of binary archives from bin/, the resulting
archives would contain overly large unstripped binaries, which is
undesired. On the other hand, always stripping the build results is not
a good option either because we rely of symbol information during
debugging.
This patch changes the installation of build results such that a new
'debug/' directory is populated besides the existing 'bin/' directory.
The debug directory contains symbolic links to the unstripped build
results whereas the bin directory contains stripped binaries that are
palatable for packaging (depot tool) and for assembling boot images (run
tool).
By installing the core object to bin/, we follow the same convention as
for regular binaries. This, in turn, enables us to ship core in a
regular binary archive. The patch also adjusts the run tool to pick up
the core object from bin/ for the final linking stage.
The use of 'select_from_repositories' for locating the linker script for
dynamically-linked executables only works if 'BASE_DIR' appears in the
list of 'REPOSITORIES'. This is the case when using the build system in
the traditional way but it is not desired when building binary archives
of individual components.
The base class of Registered must provide a virtual destructor to enable
safe deletion with just a base class pointer. This requirement can be
lifted by using Registered_no_delete in places where the deletion
property is not needed.
Fixes#2331
Ldso now does not automatically execute static constructors of the
binary and shared libraries the binary depends on. If static
construction is required (e.g., if a shared library with constructor is
used or a compilation unit contains global statics) the component needs
to execute the constructors explicitly in Component::construct() via
Genode::Env::exec_static_constructors().
In the case of libc components this is done by the libc startup code
(i.e., the Component::construct() implementation in the libc).
The loading of shared objects at runtime is not affected by this change
and constructors of those objects are executed immediately.
Fixes#2332
This patch destructs the environment sessions for the binary and the
dynamic linker along with the other environment sessions to avoid a
warning about reverting quota that occurs when attempting to close
these sessions too late.
This patch addresses the corner cases where an environment session
could not be routed, i.e., if an environment LOG log session is
routed to a non-existing child.
This patch extends the constructor of 'Local_connection' with an
optional 'label' argument, which was previously passed implicitly as
part of the 'args' argument. Keeping the label separate from 'args'
enables us to distinguish the client-specified label from a label that
resulted from a server-side label as it is used when rewriting a label
of an environment session (i.e., the binary name) in init's routing
policy. In principle, this patch eliminates the need for init's
explicite handling of the binary name via the '<binary>' node, or
at least allows us to simplity the binary-node handling.
The race may happen when element objects get destructed by another thread then
the thread handling the for_each loop. In this case it may happen that the
object is already destructed (left the ~Element destructor) but the thread
handling the loop touches the invalid memory afterwards (the Element lock).
detected during issue #2299Fixes#2320
