Init's state reports are updated whenever an interesting part of init's
internal state changes (e.g., when sessions are established, or when
children are started/removed). However, until now, a change of a skipped
heartbeat counter was not taken as trigger for state-report updates.
In scenarios where no other intersting event happened, the last reported
state did no reflect the current heartbeat state. In particular, when
the last report was issued during the construction of a new child just
before the child became able to respond to heartbeat requests, the stale
report hinted at heartbeat problems that were just an initialization
artifact. This problem became visible on some Qemu platform where the
child startup takes a long time.
The patch tracks the observed skipped-heartbeat counter and triggers a
report whenever the counter value changes.
Issue #3079
By adding an additional synchronization point in the form of a matched
log message, this patch makes the timing behavior of the "test changing
provided services" step more deterministic. Without it, the scheduling
of OKL4 and base-hw resulted in a merge of two config updates into one.
Issue #3079
Packets whose data is stored within the Packet_descriptor itself
but not as payload, .e.g Usb::Packet_descriptor, are valid packets
after all. So loosen the packet valid check for zero-sized packets
is reasonable.
Fixes#3076.
Our overall nightly test time greatly decreases when the timeouts for
the single tests are not that over-pessimistic. Using the slowest
platforms as reference, this commit reduces the test timeouts.
Ref #3027
For reading the MAC address we try first to read it from the <policy> tag, and
when it is not defined in the <policy> tag, we allocate a MAC. But there was
no handling of the case that there is no appropriate <policy> tag. In this
case we want to create the session with an allocated MAC also.
This patch makes init robust against invalid routing rules that lack a
name attribute in the '<service>' node. This situation may occur when
interactively editing routes in Sculpt.
Some application code is dereferencing the pointer returned by
'packet_content' at packet streams without checking that it is valid.
Throw an exception rather than return a null pointer, except for
zero-length packets, which have somewhat implicit invalid content and
that we believe to be properly handled in all current cases.
The client-side of a packet stream cannot take corrective action if the
server-side is sending packets with invalid content, but the servers
that provide packet streams should catch this exception to detect
misbehaving clients.
Ref #3059
Change the root of a session request into an explicit path and apply the
label-to-path conversion using the "path_prefix" policy attribute. This
is in addition to only applying a root change with a "path" attribute.
Ref #3031Fix#3056
The bulk buffer is now 64Byte-aligned so that the allocated
packets get aligned likewise (assumed the packet allocator uses an
appropriately aligned block size). This ensures that each packet
starts at a new cache line on common platforms.
Issue #3053
This patch improves the detection of new appearing top-most windows.
Such a window should prompt the decorator to bring the corresponding
nitpicker view(s) to the front of the view stack. The original
implementation relied on hints provided by the layouter (the 'topped'
attribute). With the patch, the decorator tracks the top-most window by
itself, which improves the robustness.
As a second improvement, the patch defers the destruction of windows to
the point when all other window operations are completed. This hides
intermediate states when replacing one window by another in one step,
which is typical for console-like scenarios. Hence, this patch should
eliminate flickering artifacts when switching from one virtual console
to another.
Issue #3031
* Account all RAM/CAP quota of a session except quota for metadata used in
core. The latter is considered when asking if a session can afford to make
an operation but it does not get accounted to always be able to pay back all
quota when a session closes. The general accounting mechanism is moved from
atop of the allocators down to the level of RAM/RM session operations.
* report statistics about session objects and quota if <report stats="yes"
quota="yes"/> is configured. (default is yes if <report> is present)
Issue #2953
Like suggested by RFC 2663, reprogram the dissolve timeout of a TCP link
state to 2 times the maximum segment lifetime (by default 1 minute) when
receiving a matching packet with the FIN flag set, or with the ACK flag
set to acknowledge a FIN of the remote side.
Mark a link state as closed (no further reprogramming of the dissolve
timeout) and set the dissolve timeout to 2 times the maximum segment
lifetime when receiving a packet with the RESET flag set.
Issue #2953
If the NIC router has insufficient CAP or RAM quota for the creation of
a state object for an interface, it tries to destroy a certain amount of
existing state objects of this interface to free resources. Afterwards,
it retries handling the current packet once. If it does fail again, the
router drops the packet.
Issue #2953
Ensure that a '\0' always appears at the end of the ouput of the XML
generator. Previously, exceptions during the Node(...) constructor
might have prevented this. This commit also extends the xml_generator
test to drive a harder test on exceptions in the Xml_generator.
Issue #2953
When the functor provided to the Node constructor throws an exception,
do revert all changes in reverse order. Previously, the changes made
to the parent node were not considered by the exception handler which
caused unnecessary characters to remain in the out buffer for each
reverted node.
Issue #2953
The VFS server does not support file-system with one single-file plugin
providing the root. This is because the "leaf_path" is not universally
implemented to handle the path "/". This fix is simply to skip a
"leaf_path" check when opening the path "/".
Ref #2919
The PS/2 driver retries to get mouse-reset results for 700 ms, sleeping
after each attempt for 10 ms. So, the driver needs a Timer session now.
Fixes#2713
This patch addresses a situation where _process_packets was called as a
side effect of watch notification (that was processed during an unlink
RPC operation). This scenario (triggered by the fs_query test)
ultimately ended up in a deadlock. Io/watch reponse handlers should
never re-enter the application logic.
The new base/xsd/config.inc defines generic XSD types such as 'Boolean' or
'Session_label'. It can be included in config XSD files by using:
! <xs:include schemaLocation="file://${GENODE_CONFIG_INC}"/>
The string ${GENODE_CONFIG_INC} is replaced by the run tool with the
above mentionened file path.
Issue #2897
Seen on X250
Description from https://www.win.tue.nl/~aeb/linux/kbd/scancodes-1.html
The ten grey keys Insert, Home, PgUp, Delete, End, PgDn, Up, Left,
Down, Right are supposed to function regardless of the state of Shift
and NumLock keys. But for an old AT keyboard the keypad keys would
produce digits when Numlock was on or Shift was down. Therefore, in
order to fool old programs, fake scancodes are sent: when LShift is
down, and Insert is pressed, e0 aa e0 52 is sent; upon release of
Insert e0 d2 e0 2a is sent. In other words, a fake LShift-up and fake
LShift-down are inserted.
Fixes#2888
Add additional parsing modes to the sequence decoder to detect and
discard unhandled sequences for ECMA-48, DEC private, and Xterm.
Add new behavior for cursor movement, cursor hiding, character deletion,
and line-wrapping.
Fix#2923
When the cached_fs_rom saturates the packet stream of its File_system
session it will call the session request handler recursively as pending
transfers are completed. This is bad because the content of the XML node
currently being processed will change.
The session request handler can no longer be called directly, but the
"schedule" method will submit a signal to the request handler, and
requests will be processed after the current operation has completed.
Calling 'handle_io_response()' in a regular VFS function (in contrast to a
post-signal hook) can cause problems if the caller of the VFS function holds
a lock which prevents the io response handler from returning.
With this commit, the user of the VFS becomes responsible for unblocking
threads which might be blocking after a failed 'queue_read()', 'queue_sync()'
or 'write()' call.
Fixes#2896
This new vfs_import plugin allows a VFS instance to be populated during
construction using a sub-VFS configured in an '<import>' configuration
node. This allows the ram_fs File_system server to be replaced by the
VFS server by reimplementing the ram_fs 'content' feature. At the
moment the copying of symlinks is not enabled, and the resources
obtained by the import file-system may not be freed after the import is
finished.
Fix#2906
The ram plugin to the VFS must send notifications to watch handles on
files when they are removed. This brings the VFS server to parity with
the ram_fs server for the sake of notifications.
Move run/fs_report to gems and use the vfs server and the vfs init
plugin in the test.
Ref #2902
This patch removes the detection of statically linked executables from
the base framework. It thereby fixes the corner cases encountered with
Sculpt when obtaining the binaries of the runtime from the depot_rom
service that is hosted within the runtime.
Statically linked binaries and hybrid Linux/Genode (lx_hybrid) binaries
can still be started by relabeling the ROM-session route of "ld.lib.so"
to the binary name, pretending that the binary is the dynamic linker.
This can be achieved via init's label rewriting mechanism:
<route>
<service name="ROM" unscoped_label="ld.lib.so">
<parent label="test-platform"/> </service>
</route>
However, as this is quite cryptic and would need to be applied for all
lx_hybrid components, the patch adds a shortcut to init's configuration.
One can simply add the 'ld="no"' attribute to the <start> node of the
corresponding component:
<start name="test-platform" ld="no"/>
Fixes#2866
When mounting a ROM module with binary="no", the
vfs will detect the 0-termination to calculate the
file size instead of using the dataspace size.
Fixes#2903
Process I/O packets in batches. If a batch is processed and there are
still packets pending, send a signal locally to the packet handler and
return to the entrypoint signal dispatcher. This prevents clients from
starving each other, which happens when a client continuously submits
packets at a faster rate than the server can process.
Fix#2900
Test of File_system I/O scheduling. It appears that a client not
submitting I/O packets slower than the VFS server can process will
starve other clients.
Ref #2900
This component is contrasted with the fs_rom server that serves
independent dataspaces to each client. Using a cache was not possible
until the region map session supported the creation of read-only
attachments.
Test at run/read_only_rom.
Ref #1633Fix#2760
The verbosity mode of the NIC bridge can be toggled with the verbose attribute
(default value shown):
! <config verbose="no" />
If enabled, the NIC bridge logs sent and received packets as well as the
lifetime of interfaces connected to the bridge.
Issue #2899
The test triggers the following assertion in the kernel, which will
remain unfixed as the kernel is no longer developed:
Assertion queue_state.is_set(queue_state_t::send) failed in file
.../pistachio/kernel/src/api/v4/tcb.h, line 727
This assertion is presumably similar / related to issue #1495.
Quietly insert forward declaration of a Input::Binding class, and make
it a friend of Input::Event and Input::Session_client. This is to allow
non-C++ language bindings (Nim) to access private members by providing
their own implementation of the Binding class.
Fix#2889
The legacy PIT timer driver is too jittery for the character-repeat
test. This patch disables the test for platforms using this device as
user-level time source, which are at the moment OKL4, Pistachio, and
seL4.
Also remove 'requires_installation_of', while also checking sbin
directories in 'have_installed'. The run scripts have been adjusted
accordingly.
Fixes#2853
The 200-second timeout of the init test triggers on some test platform
(in particular on qemu) while it is overly pessimistic on others.
This patch aims at stabilizing the timing behavior of the test across
the tested kernels/machines.
This is a follow-up patch of "init: avoid too eager child restart". On
each config update of init, init re-applies child-specific configuration
changes. In the case of an already exited child, this re-evaluation
wrongly marked such a child as abandoned because the child's environment
sessions do no longer exist. Abandoning the child, in turn, triggers the
destruction and subseqent restart (because the <start> node of the
configuration still exists). The latter is bad for two reasons.
First, the exit state of the original instance becomes lost. Second, the
restart may have unexpected side effects due to sessions created by the
new instance. I.e., when resizing a partition in sculpt, init would
wrongly restart the gpt-write tool after the tool successfully exited.
This collides with a newly started instance of part_blk/resize2fs, which
now competes with the second gpt-write instance for the exclusive access
of the targeted block device.
The patch prevents init from re-applying configurations to exited
children. The accompanied test case covers the corner case.
Do not simply wait for the good ping test to finish, but for the other
flood tests to trigger the RAM exhaustion. This makes the test more
robust with slower platforms or a different timing.
Issue #2857
The log messages covered by verbose_packet_drop were previously
configured by the verbose attribute. This isn't the case anymore. Now,
you can configure them as follows:
! <config verbose_packet_drop="no" ... >
! <domain verbose_packet_drop="no" ... />
! <config/>
The new attribute determines whether to log each packet drop and the
rational behind it. The <config> value affects all domains without a
<domain> local value.
Issue #2857
The default value of each <domain> is the <config> value. However, if
a <domain> local value is set, the <config> value doesn't affect this
value at all.
Fixes#2874
The ICMP-Echo-server functionality of the router has the following
configuration attributes (default values shown):
! <config icmp_echo_server="yes">
! <domain icmp_echo_server="yes" ... />
! </config>
The icmp_echo_server attribute configures whether the router answers ICMP Echo
requests that address the router. The <config> value affects all domains
without a <domain> local value.
Issue #2874
When a domain is updated to a new component config, the two criteria for
keeping an ARP waiting state of a remote domain are whether the remote
domain still exists and whether its IP config is still the same. This
means that a domain must also dissolve all of its remote ARP waiting
states if its IP config changes (without an update of the component
config). This wasn't the case until now.
Issue #2840
The chroot server was conceived to automatically place File_system
sessions into segregated root directories by converting session labels
to paths. If multiple children needed to be grouped under the same path,
a 'merge' policy would truncate the session label before path conversion.
Now that init can rewrite session labels and thus reproduce truncation,
the chroot merge feature is redundant and can be removed.
Fix#2846
Removed the modified mark from handles that have been written to when
they are synced, otherwise a notification would be sent again when the
handle is closed.
Ref #2839
* Do not log events that are not critical (deadly) to the NIC router if not
configured to be verbose,
* Print almost all log lines with a prefix of the domain name they are
related to,
* And, do not use Genode::error and Genode::warning as they make it hard to
read the log with the domain name prefixes.
Fixes#2840
Introduce the uplink tag:
! <config>
! <uplink label="wifi" domain="uplink">
! <uplink label="wired" domain="wired_bridge">
! <uplink domain="wired_bridge">
! <config/>
For each uplink tag, the NIC router requests a NIC session with the
corresponding label or an empty label if there is no label attribute.
These NIC sessions get attached to the domain that is set in their
uplink tag as soon as the domain appears. This means their lifetime is
not bound to the domain. Uplink NIC sessions can be safely moved from
one domain to another without being closed by reconfiguring the
corresponding domain attribute.
Attention: This may render previously valid NIC router configurations
useless. A domain named "uplink" doesn't automatically request a NIC
session anymore. To fix these configurations, just add
! <uplink domain="uplink"/>
or
! <uplink label="[LABEL]" domain="uplink"/>
as direct subtag of the <config> tag.
Issue #2840
The term was used for the old configuration during the handling of a new
configuration but in other places it was already called old_config.
Issue #2840
Dissolve and destroy the invalid domain first before deinitializing all
domains for the next round. This way, the deinitialization is not done twice
for the invalid domain.
Issue #2840
Due to a bug in the deinitialization of the DHCP server of a domain (the
reference to the destroyed object was not cleared), the NIC router could
end up in a page fault caused by a double-free at the heap. This also fixes
the previously missing dissolving of the "DNS-server-from" relation to a
remote domain.
Issue #2840
Currently has three clients that continuously create new UDP/TCP/ICMP
connections through the NIC router with NAT to the outer world and they
get never closed. A fourth client does normal ping through the same
domain to the outer world that must succeed even after the RAM quota of
the other session at the router is exhausted. The test is restricted to
Qemu to not being at risk to flood real networks.
Issue #2857
This patch weakens the aggressive restart of a child with incomplete
environment sessions. The restart check is performed each time
the init configuration changes. In sculpt, this is not a rare special
case anymore but a frequent case when using the depot_rom as provider
for environment ROM sessions. In particular when starting a chain of
inter-depending children, the sculpt-manager quickly generates a
sequence of configurations with successively added start nodes.
The addition of the 'Cap_consumer' feature to app/dummy increased the
static RAM demand of the component to a point where 1 MiB no longer
suffices on sel4 (on this kernel, the base library uses a larger amount
of statically allocated data compared to the others). The tweaks enable
init.run and init_loop.run to work on sel4 on x86-64.
Selecting an alternate interface setting, even if it is the same as the
current one, apparently makes the INQUIRY command fail with USB devices
like 'SanDisk Ultra Fit' (0781:5583) and 'Corsair Flash Voyager'
(1b1c:1a03) when the USB block driver is restarted.
Fixes#2860
* Catch Quota_guard exceptions wherever we do 'new (_alloc)' in Interface
and in case drop the packet that caused the exception
* Modify Interface::handle_config_2 to not use the guarded allocator of the
NIC session quota as we cannot handle a RAM exhaustion well at this point
in time. Instead use a Constructible member in Interface so that a needed
RAM adds up to the calculation in the create_session implementation where
an exhaustion is handled via a service denial.
Issue #2857
The interface destructor called pure virtual methods at least when
cancelling ARP- waiting states. The implementations were made by the
deriving classes Session_component respectively Uplink. This led to an
abort of the NIC router as the destruction of the derived class was
already done. A similar problem already occured in the past during the
construction of Interface and was back then solved by introducing a
separate init() method. This commit, however, solved the problem by
making Interface a member of the other classes. Therefore, the init()
method could be removed again. Furthermore, the interface polica could be
moved from Session_component_base to Session_component. The commit also
had to generalize the way the link state of an interface is determined.
Fixes#2856
While a child is abandoned, we must limit the start of anothers with
the same name. Otherwise - of the child has startup problems - a number
of abandoned children with the same name may queue up. This becomes a
problem whenever the child destruction depends on an asynchronous
service that provides an env session for the children. If the service is
unable to keep up with the session requests (both create and close),
the queue of abandoned children becomes unbounded. Limiting the child
creation rate to one abandoned child per name mitigates this problem.
This patch reduces the latency of state reports when children are
removed or added, thereby, accellerating the feedback loop between a
management component and init during the staged startup or removal of
inter-dependent components.
This test monitors the RAM quota of a dynamic init and a server hosted
within the dynamic init in the presence of a repeatedly created and
destructed client.
Previously we were doing the initialization once over all domains,
remembered which of them became invalid and destroyed those afterwards.
This isn't sufficient. As soon as one domain becomes invalid we have to
dissolve/destroy this one, deinitialize all other domains again (as they
could contain references to the invalid domain) and retry to initialize
them from the beginning. We proceed with this until we have one run
without a domain becoming invalid. Then we can be sure that the last
initialization run did not create references to any invalid domain.
Issue #2840
The generic helper Avl_string_tree of the NIC router is currently only
used for finding domains via their names, but in the future it can be
used for finding uplinks by their labels also. Additionally, it enables
us to throw an exception when inserting two elements with the same
identifier.
Issue #2840
* Get rid of the base classes Rule and Leaf_rule,
* Make log output about initiated or invalid routing rules conform to the rest
of the router log, and
* Ensure that each type of routing rule when being invalid invalidates its
whole domain.
Issue #2840
On Linux, we have a tap device as NIC back end but there is no one to
ping to in the subnet of the tap device. On FOC, the tests seem to trigger
a bug in the destruction of components with parent.exit(X); .
Fixes#2848
The condition was too rigid. In the case where no motion occurred in
between the press and release events of the magic button, the delayed
press event would not be delivered. This - in turn - confused other
components (like nitpicker) down the input chain.
for such classes where it should be safe and where we have seen issues.
Disabling in general bus master DMA causes on some machines hard hangs, e.g.
because the USB handover protocol was violated.
Fixes#2835
In contrast to most information of init's state reports, which can be
monitored at a relatively low rate (like 2 seconds in Sculpt's runtime),
resource requests call for an immediate response by the consumer of the
report. Otherwise the requesting child stays unnecessarily blocked until
the next rate-limited state report is due. This patch adds a fast lane
for such low-latency state updates to init.
Do not send nor buffer packets at interfaces with link state "down". This
prevents that packets that were routed to one network (allowed to see them),
due to a sudden link down/up, are leaked to another network that is not
allowed to see them.
When updating an interface in the NIC router to a new configuration and
the domain name of the interface has not changed but the domain
disappeared, the NIC router did not detach from the old domain correctly
which led to broken remnants of interface state objects (like connection
states).
Adapt the nic_router_uplinks run script to work with the fix.
fs_report truncated the report file on construction of the report
session to mimic an "empty report". This prevented existing use cases
where the initial state was already written to the file system by other
means.
I doubt that use cases exist where the inital empty-report mimic is
needed and (if I'm wrong) there is still the race of the consumer
reading the file just before the report session is constructed.
Reporters that want to enforce an empty initial report should send it
explicitly.
Clients may wish to act on missing files. In any case the fs_rom
needs to reopen a watch handle when a file is deleted, and this
sort of change to the internal state machine is propelled by
client RPC requests.
Fix#2839
The new 'displays' report contains information about the used
framebuffer backend. It is issued when nitpicker has successfully passed
all initialization, in particular the connection to the framebuffer
driver. Hence, it can be taken as an indicator of whether the framebuffer
is available or not (e.g., caused by a faulty driver).
When re-configuring the NIC router, determine for each domain if at least one
interface stays with the domain. If a domain fullfills this and has a
dynamic IP config (received via a DHCP client), keep the IP config.
To achieve this, the following changes have been made to the existing NIC
router code:
* Split-up Interface::handle_config into three steps:
1) Determine for each interface if its domain can keep its IP config or
or if it has to mark it invalid. This must be done before (re-)attaching
any interface because during "attach" several decisions are made based on
the validity of the IP config of corresponding the domain.
(E.g. whether to participate in sending DHCP DISCOVERs {IP config
invalid} or whether to participate in sending pending ARP REQUESTs
{IP config valid} ).
2) Detach, attach, or re-attach each interface according to the
configuration. This must be done before re-considering the temporary
state objects of each interface because the latter might have effects
on the interfaces of remote domains which must then be in place already.
3) Re-consider temporary state objects of each interface. (E.g. transport
layer connection states)
* Re-work IP-config setter in a way that it works as follows:
1) If the old IP config is valid, let all local interfaces as well as remote
interfaces that depend on the IP config of the domain detach from the old
IP config.
2) Overwrite with new IP config
3) If the new IP config is valid, let all local interfaces as well as remote
interfaces that depend on the IP config of the domain attach to the new
IP config.
Issue #2815
The new attribute config.domain.label has effect only at the uplink
domain-tag. It determines which label the NIC router shall use when
requesting the NIC session for the uplink domain. If value of this
attribute changes at the uplink domain-tag, the NIC router closes and
re-requests the NIC session of the uplink with the new label.
Issue #2815
This tests the case that behind a NIC router domain there are not only
directly connected clients of the subnet but also Layer 2 indirections like
a switch or a bridge which might be interesting for example for testing the
NIC routers DHCP server implementation.
Fixes#2837
Set DHCP-DISCOVER retry timeout and DHCP-REQUEST retry timeout from 10
seconds to 2 seconds. This prevents problems with tests where the NIC
driver comes up slowly and the first packets of the DHCP client get
dropped.
Issue #2837
The Ethernet destination MAC address of a DHCP reply is not the same as the
DHCP client MAC address. The DHCP server of the NIC router did not take care
of this by now.
Issue #2837
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.
Merge the 'Packet_handle' into the 'Rom_root' and use an Id_space to
match File_system handles to session objects rather than a list. This
avoids the need to override 'Root_component::_destroy_session'.
Fix#2833
When exhausted of RAM quota while starting children, init used to throw
an uncaught 'Out_of_ram' exception as this condition was considered
fatal. However, this behavior is undesired when init is used in a highly
dynamic yet long-running fashion like sculpt's runtime subsystem. This
change keeps init running despite the error condition, giving the user
the chance to relieve the resource pressure.
The 'View_stack::draw_rec' method limited the redraw to parts of the
view that were explicitly marked as dirty. This does not produce the
desired result when stacking multiple transparent views. Here, the
background views must be drawn regardless of whether they are marked as
dirty or not.
The file may also be changed by other components, so a open-close cycle
for write() is more robust. For example, Vim removes the original file
and recreates with the new content.
The '_active_mode' must not be changed at any time except when the
client asks for the 'Framebuffer::mode'. Otherwise, the dimensions of
dataspace used by the client is not always consistent with the mode
information as gathered by the client.
On real hardware, the tests expect an IPv4 subnet such that UDP requests
to 10.0.0.2 port 12345 get answered with an ICMP destination port unreachable.
Issue #2775
support USB NIC
Issue #2788
Set DHCP discover timeout to 1 second because, for some reason, the first
DHCP discover attempt of the NIC router on the PandaBoard times out with the
nightly test infrastructure.
Issue #2788
Adaption to mac-address allocation changes
Normally, the NIC sessions are independent from the domain tags.
However, by now the uplink session, in contrast to the sessions of the
other domains, is still not a server but a client. This means that only
the NIC router itself can decide when to open and close uplink sessions
and how many. Thus, with this commit, we break with the pattern that
session lifetime is independent from domains by letting the NIC router
create the uplink session when the uplink domain appears and close the
session when the domain disappears.
Fixes#2795
Since the router MAC is allocated like the donwlink MACs it can't happen
anymore that these MACs clash, for instance due to nested routers. Thus,
the range of the MAC allocators of nested routers must not be exclusive
anymore which deprecates the 'mac_first' configuration attribute.
Issue #2795
Allocate a virtual MAC address at runtime that is used as router
Ethernet-identity for all downlink domains. This makes the downlink
domains independent from the uplink session.
Issue #2795
The old MAC allocator had several drawbacks:
* the address base was a public static that could and must have been written
directly from outside the class
* the in-use-flag array was based on unsigned values consuming 4 bytes each
for only one bit of information
* it was a public header that we actually don't want to expose to all
components but only to the few networking components
* it used the not-so-safe bit notation for integer members of GCC
The new version fixes all these drawbacks.
Issue #2795
Instead of handing over the maximum available size to the packet data
accessors, hand over a size guard that keeps track of the packets
boundaries.
This commit also moves the size-guard utilitiy header of Ping and NIC
Router to the include/net directory making it a part of the net library.
It applies the new approach to all net-lib users in the basic repositories.
Ping looses its configurability regarding the ICMP data size as this would
require an additional method in the size guard which would be used only by
Ping.
The size guard was also re-worked to fit the fact that a packet can
bring a tail as well as a header (Ethernet).
Issue #2788
For now it is enough to differentiate the most commonly used file
system on Genode, e.g. Ext2 for the Genode partition and FAT32 for
(U)EFI partitions.
Issue #2803.
The component will now always try to parse the MBR as well as the GPT
(in this order). It will bail out if both are considered valid, using
GPT/MBR hybrid tables is not supported.
Fixes#2803.
The Ethernet payload may be followed by padding of variable length and
the FCS (Frame Check Sequence). Thus, we should consider the value
"Ethernet-frame size minus Ethernet-header size" to be only the maximum
size of the encapsulated IP packet. But until now, we considered it to
be also the actual size of the encapsulated IP packet. This commit fixes
the problem for all affected components of the Genode base-repository.
Fixes#2775
This reduces the redundant implementations of checksum calculation to
one generic implementation, makes the checksum interface conform over
all protocols, and brings performance optimizations. For instance,
the checksum is now calculated directly in big endian which saves us
most of the previously done byte-re-ordering.
Issue #2775
Replace packet method 'T *data' by the new methods 'T &reinterpret_data'
for parsing or modifying existing sub-protocol packets and 'T
&construct_at_data' for composing a new sub-protocol packet. This has
the advantage that, when composing a new packet, the default constructor
that zero-fills the packet is always called first.
Fixes#2751
This commit changes the 'Input::Event' type to be more safe and to
deliver symbolic character information along with press events.
Issue #2761Fixes#2786
On the Raspberry PI, the 2 seconds of round time in the polling test
were not sufficient to reach the goal of at least 1000 successful polls.
Thus, the commit sets the round time to 2.5 seconds which doesn't hurt to
much but allows the RPI to just make it.
Fixes#2779
This patch largely reverts the feature of selecting parts of input nodes
from within the output node (as originally introduced by commit
7263cae5a18b4f1f2293d031f9bafcf05ba51146). The selection of content
should be consistently performed by input nodes instead. The principle
ability of copying input nodes verbatim into the output stays available.
Issue #2691
The new 'attribute' and 'value' attributes of input nodes
can be used to select input sub nodes that match the presence and value
of the specified attribute.
Issue #2691
This patch replaces the terminal's formerly built-in fonts with the new
VFS-based font handling.
To avoid the copying of the terminal's font configuration across run
scripts, this patch adds the new terminal/pkg runtime package, which
includes everything needed for instantiating a terminal: the actual
terminal component, the library dependencies (vfs_ttf, which in turn
depends on the libc), a font (bitstream-vera), and a reasonable default
configuration.
Fixes#2758
The 'stat' method is called for all paths, not just the specific file
system node of the ROM module. The ROM update is needed only in the
latter case.
Otherwise, when always updating the ROM on stat, stat calls on the VFS
become very expensive in the presence of a mounted ROM module if the ROM
is obtained from fs_rom (which re-watches the file and all its
individual path elements whenever the 'update' RPC function is called).
Reduce the size and forward compatibility of VFS file-system
constructors by passing an object holding accessors for 'Genode::Env',
'Genode::Allocator', response handlers, and the root file-system.
Fix#2742
When working with GPIO interrupts on i.MX6SX for Ethernet PHYs
it became obvious that the GPIO driver repeatedly receives interrupts
for the same event, because it acknowledges the interrupt before a
client has handled the event.
Ref #2750
This driver component provides support for using consumer NVMe storage
devices, i.e. it omits name space managment and will always use the
first name space, on Genode. For now it defaults to a reasonable low
configuration:
- 1 I/O queue (completion/submission tuple)
- 128 entries in the I/O queue
- 4096 as the only I/O transaction memory page size
Fixes#2747.
The 'default' attribute is useful to change the default value for those
protocol attributes that are not explicitely set in the configuration of
the component.
Issue #2738
Each supported protocol now has an attribute with the name of the protocol in
the config tag. Each of these attributes accepts one of four possible values:
* no - do not print out this protocol
* name - print only the protocol name
* default - print a short summary of the most important header values
* all - print all available header values
Example:
! <config eth="name"
! arp="all"
! ipv4="default"
! dhcp="no"
! icmp="all"
! udp="no"
! tcp="default"
! ... />
Corresponding output example:
! ETH IPV4 10.0.2.212 > 10.0.2.55 time 7158 ms (Δ 11 ms)
! ETH IPV4 10.0.2.55 > 10.0.2.201 TCP 80 > 49154 flags ' time 6976 ms (Δ 5 ms)
! ETH ARP hw 1 prot 2048 hwsz 6 protsz 4 op 1 srcmac 02:02:02:02:02:01 srcip 10.0.2.212 dstmac 00:00:00:00:00:00 dstip 10.0.2.55 time 7074 ms (Δ 98 ms)
Issue #2738
Nitpicker's 'Session:focus' call used to trigger a one-off focus change
at call time. This focus change did not pass the same code paths as a
focus change triggered by a "focus" ROM update, which led to
inconsistencies.
This patch changes the implementation of 'Session::focus' such that the
relationship of the caller and the focused session is preserved after
call time. Whenever the calling session is focused in the future, the
specified session will receive the focus instead. So 'Session::focus'
represents no longer a single operation but propagates the information
about the inter-session relationship. This information is taken into
account whenever the focus is evaluated regardless of how the change is
triggered.
This makes the focus handling in scenarios like the window manager more
robust.
Issue #2746
Relative motion events with a motion vector of (0,0) should not exists.
They cause jittery movements of nitpicker's pointer position. This
patch filters out such events.
This tests ping with simple IP forwarding, ping with NAPT as well as
forwarding of ICMP "Destination Unreachable" messages through the NIC
router.
Issue #2732
The 'ping' component continuously sends ICMP Echo requests to a given IP host
and waits for the corresponding ICMP Echo replies. For each successfull ICMP
Echo handshake it prints a short statistic. By now, it can be used only with a
static IP configuration. The size of the ICMP data field can be configured. It
gets filled with the letters of the alphabet ('a' to 'z') repeatedly.
Issue #2732
Originally, the timeout framework caused deadlocks when acquiring the same
lock from different timeout handlers. This use case is now tested in the
timeout test.
Fixes#2704
To handle all pending timeouts in the context of scheduling a timeout
was only necessary because the Timeout framework once made use of the
Alarm framework. The method Alarm_scheduler::schedule_absolute took an
absolute deadline as argument and we couldn't change this beause the
Alarm framework was also used without the Timeout framework. We had to
calculate this absolute deadline with the now time of the Timeout
framework but the Alarm framework has its own now time that is always a
bit behind the one of the Timeout framework. This lead to bad decisisons
when finding the right position for the new timeout. Now, we can call
schedule_absolute with a relative duration and thereby fix the problem.
When we schedule an absolute timeout without considering the small time
difference, the end-time for the timeout that is calculated using the
local time value is also smaller than the expected end-time. This can
also lead to directly triggering timeouts that should have triggered
with a certain delay.
As it is not trivial to update the local time value while scheduling a
timeout _without_ calling other timeout handlers, we simply raise the
duration of the new timeout by the age of the local time value.
Issue #2704
This fixes the problem that large timeouts, when rescheduled, are interpreted
to be from the last now_period instead of, what would be right, the next
now_period. This occured if there were multiple pending alarms at the head of
the queue and the reschedule of the first one was done with the other outdated
deadlines still in place.
Issue #2704
Instead of taking the absolute deadline of a timeout as argument from
outside (where it is calculated with a freshly requested now time), we
now take a relative duration as argument and calculate the deadline with
the scheduler-internal now time (which can be a little bit outdated).
This enables us to schedule timeouts without updating the internal now time
and thereby handle all pending timeouts.
Issue #2704
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
In the domain class there were several places where output was generated
not conforming to the typical output format of the router ("[domain]
event: parameters").
Issue #2670
When having an interface that yet is not attached to a domain, then a new
configuration comes in and the interface receives a domain name (via the
policy tag) but the corresponding domain doesn't exist, an exception
Domain_tree::No_match is thrown but was not caught and handled until now.
Issue #2670
This follows the guidelines in RFC 5508 to enable forwarding of ICMP
"Destination Unreachable" that correspond to an existing link state in
the NIC router. It also serves as blueprint for forwarding ICMP error
messages in general (They are merely not enabled because we don't test
them).
Issue #2732
By now, the 'verbose packets' output when sending packets was printed after
finish sending the packet. This makes following the packet flow harder if you
have multiple components that print such information.
Issue #2732
The mac_first attribute tells the MAC-address allocator of the router
from which MAC address to start allocating. This is useful, for
instance, if you have nested nic_routers. In this case, identical
MAC-allocator settings have led to name clashes in the past, so, you
want to be able to configure them differently.
Issue #2732
This follows the guidelines in RFC 5508 to enable ICMP echo through a NAPT
channel of the NIC router. It serves also as blueprint for ICMP queries in
general (they are merely not enabled because we don't test them by now).
Issue #2732
The 'verbose packets' output previously was not generated for Interfaces
without a domain. But this is desirable as the router nonetheless
receives packets at such interfaces. This is now fixed and such output
is simply prefixed with a "[?]" denoting that the interface has no
domain.
Issue #2732
We missed to zero-out the ECN field in IPv4 packets. We don't use the ECN
field but there might be old data left in the packet RAM allocated by the
NIC packet streams. If we don't zero-out ECN it might leak old data.
Issue #2732
Add a new 'Vfs_watch_handle' type to the VFS interface. This handle type
will pass a handle context up through the I/O handler to the application
when a notification event occurs.
Watch support implemented for RAM and File_system plugins, all other
file-systems return WATCH_ERR_STATIC by default.
Test at run/fs_rom_update_ram and run/fs_rom_update_fs.
Fix#1934
This patch improves the `Text_painter` utility that is commonly used by
native Genode components to render text:
- Support for subpixel positioning
- Generic interface for accessing font data
- Basic UTF-8 support
Since the change decouples the font format from the 'Text_painter' and
changes the API to use the sub-pixel accurate 'Text_painter::Position'
type, all users of the utility require an adaptation.
Fixes#2716
This file system is meant as a building block for pseudo file systems
that host a directory of several small files where each corresponds to
an attribute of the pseudo file system.
By letting the 'Dir_file_system' accept an arbitrary 'File_system'
as root directory, we can use the 'Dir_file_system' as a building
block for creating other file-system types.
This patch enables the use of the VFS from VFS plugins by passing a
reference of the root directory to the constructors of file-system
instances. Since it changes the signature of 'Vfs::Dir_file_system',
any code that uses the VFS directly requires an adaptation.
Fixes#2701
If the remote DNS server address value of a DHCP server changes, the affected
interfaces do a link down/up to inform all DHCP clients that they should
re-request their DHCP info.
Issue #2730
The dns_server_from attribute of the dhcp-server tag has effect only if
the dns_server attribute of the same tag is not set. If this is the
case, the dns_server_from attribute states the domain from whose IP
config to take the DNS server address. This is useful, for instance, if
the stated domain receives the address of a local DNS server via DHCP.
Whenever the IP config of the stated domain becomes invalid, the DHCP
server switches to a mode where it drops all requests unanswered until
the IP config becomes valid again.
Issue #2730
Until now, the DHCP server of a domain was re-constructed each time the
IP config changed. This is not necessary as a domain that acts as DHCP
server must have a static IP config as it would be senseless to act as
DHCP server and client at the same time. Now, a configured DHCP server
is constructed only when the Domain gets constructed and stays alive
until the domain gets destructed. Furthermore, we now throw Domain::Invalid
if there is no static IP config plus a DHCP server configured. However, by
now, this exception is not caught as it is not trivial to destruct the
domain at this point.
Issue #2730
The Interface constructor previously tried to attach to a domain. This
might include sending a DHCP request to get the domain a valid IP config.
But in order to achieve this, the constructor used a pure virtual method
of Interface which crashes due to the unfinished vtable. To fix this bug,
the attach attempt was moved to a new Interface::init method.
Issue #2730
Instead of Pointer<T>::set use assignment operator with implicit constructor
from T-reference. Instead of Pointer<T>::unset use assignment operator with
Pointer<T>(). Instead of Pointer<T>::deref provide () operator.
Issue #2730
The router reacts as follows to a configuration change:
1) Construct new internal configuration representation (the old one stays
in place to be able to do comparisons in the following steps)
2) Iterate through all user-dependent objects (interfaces, link states, ARP
information, DHCP information) and re-check which remain valid with the
new configuration and which must be dismissed.
3) Adapt the objects that remain valid to the new configuration (re-write
references) and remove or detach the dismissed objects.
4) Do a link state DOWN at each interface and a link state UP at each
interface that remains attached to a domain.
5) Replace the old internal configuration representation with the new one
This way, the router keeps as much user dependent states as possible
while going through a configuration change. Thus, overwriting the old
configuration with an exact copy of itself is (almost) transparent to
clients of the router. Almost, because there are things the router must
do on every configuration handling, like re-scheduling the expiration
timeouts of links.
Ref #2670
The for_each method of the List wrapper remembers the next list item
before calling the functor on the current one, so, the current one can
be destroyed during the functor.
Ref #2670
Clients can connect at any time to the NIC router. The interfaces (sessions)
get attached to the appropriate domain as soon as it appears. This implies
that interfaces can also be detached from a domain without beeing destructed
when the domain disappears. All user dependent states of an interface such as
the link states, DHCP allocations and ARP information get lost when the
interface gets detached.
Ref #2670
This separates the decision wether to log the received and sent packets
from the 'verbose' attribute. This information is now only logged if
'verbose_packets' is switched on. If 'verbose' is switched on, only
routing decisions and optional hints are printed.
Ref #2670
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
This patch adds a safety check to nit_fb to ensures that nit_fb never
runs out of RAM. Should the available RAM not suffice for resizing the
virtual framebuffer to a new mode, it keeps the current mode.
The Dir_file_system uses static cast to convert handles from the
application to a plugin local type. For this reason, only the local
handle type may be returned from 'opendir' or 'open'. This fixes the
unexpected behavior when opening directories as files.
Fix#2533
This patch removes the notion of partial writes from the file-system
servers. Since write operations are asynchronously submitted, they are
expected to succeed completely, except for I/O errors. I/O errors are
propagated with the write acknowledgement but those are usually handled
out of band at the client side. Partial writes must never occur because
they would go undetected by clients, which usually don't wait for the
completion of each single write operation.
Until now, most file-system servers returned the number of written bytes
in the acknowledgement packet. If a server managed to write a part of
the request only, it issued the acknowledgement immediately where it
should have cared about writing the remaining part first.
The patch detects such misbehaving server-side code. If partial writes
unexpectedly occur, it prints a message and leaves the corresponding
request unacknowdleged.
Issue #2672
File_system clients may now watch files and directories for changes by
opening a 'Watch_handle' rather than submitting a 'CONTENT_CHANGED'
packet to the server. When a change happens at a node with an open
Watch_handle a CONTENT_CHANGED packet will be sent from the server to
the client. This serializes registration with other handle operations
and separates I/O handle state from notification handle state.
Test at run/fs_rom_update.
Ref #1934
Catch out of RAM and capability exceptions and return error values.
Abort opening a composite directory at Dir_file_system where an
opendir call on any child file-system returns an OUT_OF_RAM or
OUT_OF_CAPS error.
Ref #2642
This change enables the use of negative values for the 'initial_width'
and 'initial_height' attributes to specify values that are relative to
the screen size. This is consistent with the meaning of the 'width' and
'height' attributes.
This patch introduces the subnodes <provides>, <requires>, and
<content> to the <runtime> node. All <rom> sessions that are
expected from the depot appear within the <content> node, which
sets them nicely apart from <rom> sessions that may be required
as runtime arguments.
Note that the <requires> and <provides> nodes do not appear in the
patch because the existing depot_deploy tool does not interpret this
information (the pkg/test-fs_report runtime does not provide any
service, and the timer session is provided as a common route).
The new 'Terminal_session::size_changed_sigh' RPC function registers a
signal handler that is triggered each time when the terminal size
changes. It enables the client to adjust itself to the new size by
subsequently calling the 'size' RPC function. Of all terminal servers,
only the graphical terminal triggers this signal.
- handle line wraps in 'ech()'
- take (1,1) origin into account in 'hpa()' and 'vpa()'
- unify handling of SGR escape sequences of different lengths
- accept the '[?2004h' and '[?2004l' escape sequences (used by midnight commander)
Fixes#2671
The 'Expanding_reporter' wrapper for the 'Reporter' eliminates
the burden of handling 'Xml_generator::Buffer_exceeded' exceptions
from components that generate reports.
Fixes#2655
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
The 'trace_logger' component can be used to easily gather, process and export
different types of tracing data. Which subjects to select is configurable via
session label policies and thread names. Which data to collect from the
selected subjects can be configured for each subject individually, for groups
of subjects, or for all subjects. The gathered data can be exported as log
output.
This is an example configuration of the 'trace_logger' component which shows
the default value for each attribute except the policy.thread and
policy.label:
! <config verbose="no"
! session_ram="10M"
! session_arg_buffer="4K"
! session_parent_levels="0"
! period_sec="5"
! activity="no"
! affinity="no"
! default_policy="null"
! default_buffer="4K">
!
! <policy label="init -> timer" />
! <policy label_suffix=" -> ram_fs" />
! <policy label_prefix="init -> encryption -> "
! thread="worker"
! buffer="4K"
! policy="null" />
! </config>
For more details see os/src/app/trace_logger/README.
Fixes#2654
Building this policy was disabled by a REQUIRES = riscv_toolchain_bugfix
to prevent compiler crashes on nightly test builds. Since the latest
RISCV toolchain update, test builds for RISCV crash at much more places
which makes this single work-around senseless.
Issue #2654
The 'Buffered_xml' utility is used by three components and a fourth is
on the way. To avoid another duplication of the code, this patch makes
it publicly available at 'os/buffered_xml.h'.
This patch replaces the 'Heap' by a 'Sliced_heap' to avoid the
allocation of a 16 KiB memory block when 'dummy' acts as a server. On
seL4, such an allocation would exceed the 1M quota as assigned by the
init.run script.
The 'initial_width' and 'initial_height' attributes were added to
accomodate the use-case to set the initial dimensions whenever 'nit_fb'
is used in a dynamic fashion, e.g, in combination with a window manager.
These attributes may not be mixed with the 'width' and 'height'
attributes, which are mostly used when a static size configuration is
desired.
When a NIC session is destructed at the router, we have to remove all ARP
cache entries that match the MAC address of that session. Otherwise the
outdated entries might be re-applied later, leading to wrong destination
MAC addresses in routed packets.
Fixes#2637
By specifying the 'config' of a '<runtime>' as an attribute, we can
distinguish the case where the config is obtained from a ROM session
from the case where the config is specified inline as a '<config>' node.
This fix handles the case where the focused domain loses its focus
because the currently focused client vanishes. In this case, the focus
will be undefined and the non-focused views of the domain become
tinted again. The refresh should take effect immediately as soon as the
client vanishes.
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
which executed in the current measurement period but are now dead.
Keep the Entry object up to next period, where the recent_execution_time will
become 0 since it will not execute definitely.
Issue #2638
Require x86_64 because memory/adress space limitations on x86_32
restrict the use-cases on such a platform anyway. Doing that,
we can also assume that memory adresses are always 64bit long and
do not have to handle 32bit adresses.
According to the creator of the net-stat lib, this lib was a mere debugging
tool that is not used anymore nor worth the work of updating the it to
modern Genode coding paradigms. Also, there exist no tests for the lib.
Instead of having a method validate_size in each packet class, check
sizes in the data accessor of the surrounding packet class. This packet
accessor is the one that casts the data pointer to the desired data type
so it is sensible that it also checks whether the desired type would
exceed the available RAM before doing the cast. This also fits nicely
the fact that for the top-level packet-class of a packet, the size must
not be checked (which was previously done).
Issue #465
This is a follow-up commit to "Increase default warning level", which
overrides Genode's new default warning level for targets contained in
higher-level repositories. By explicitly whitelisting all those targets,
we can selectively adjust them to the new strictness over time - by
looking out for 'CC_CXX_WARN_STRICT' in the target description files.
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 warning "no interface connected to domain" was introduced when only one NIC
session at a time could be connected to a domain. It should help to track
packet drops that were caused by startup timing issues between servers and
clients. However, a user should watch the "NIC sessions" value of a domain
(verbose_domain_state) instead when debugging packet loss. With support for
multiple sessions per domain, even a non-empty domain may still miss the
session that connects the desired server.
Fix#2629
Previously, all packets that the router wanted to sent were first prepared to
their final state and then copied at once into the packet stream RAM. This is
fine for packets that the router only passes through with modifying merely
a few values. But for packets that the router writes from scratch on its own,
it is better to compose the packet directly in the packet stream RAM.
Fix#2626
Normally, Interface::send always takes the base and size of the RAM region
where a packet was composed and copies this finished packet at once into the
packet stream RAM. But we want to be able to also compose packets directly in
the packet stream RAM, so that no memcpy is needed. Thus, Interface::send now
takes a functor that describes how to compose the packet, then allocates the
packet stream RAM and applies the functor to this RAM. there is also a version
of Interface::send that provides the old behavior but with the new back end.
This way, we stay backwards-compatible.
Issue #2626
When composing an ARP packet for sending, it's pointless to use the Arp_packet
constructor as the constructor only checks whether the packet is malformed.
Issue #2618
The NIC router can now be configured to periodically send reports.
Configuration example (shows default values):
<config>
<report interval_sec="5" bytes="yes" config="yes">
</config>
If the 'report' tag is not available, no reports are send.
The attributes of the 'report' tag:
'bytes' : Boolean : Whether to report sent bytes and received bytes per
domain
'config' : Boolean : Whether to report ipv4 interface and gateway per
domain
'interval_sec' : 1..3600 : Interval of sending reports in seconds
Issue #2614
In the context of link state objects we often used the term "close" were we
actually meant "dissolve". The term "close" originated from the TCP connection
state and is still used in TCP links in the correct manner.
Issue #2609
Act as hub for the interfaces at a domain. This also changes the roles of the
Domain and Interface classes. By now the Interface held the data structures for
the ARP cache, foreign ARP waiters, and the searchtrees for layer 3 links. All
these structures have moved to the Domain while the memory allocations and
lifetime management for the contents of these structures still come from from
the according Interface object. The mentioned data structures were also adapted
to fit the fact that they now may maintain objects of different interfaces.
Issue #2609
If an IPv4 packet targets an IP local to the domain it comes from and doesn't
target the routers IP of that domain, forward it to all other interfaces of
the domain without considering any other routing.
Issue #2609
Improve ARP handling code in general:
Make the several cases and their handling more clear by using a more
readable if/else statement structure. Drop gratuitous ARP requests.
Domain-local ARP:
Handle ARP packets that target local IPs other than the routers IP
(forward them to all other interfaces of the domain).
Issue #2609
IP allocations were renamed DHCP allocations without fixing the according
places in log messages and comments. This commit rectifies this omission.
Issue #2609
This patch makes service-announce messages depend on the configured
verbosity. It also omits "parent provides" title messages if no new
parent services are added during a config update.
Detect loops by walking hard-links at two different speeds and checking
for lapping. Tar link walking is no longer a recursive procedure.
Caught a loop created by GNU tar 1.29.
Fix#2611
If a client decides to spontaneously send a DHCP DISCOVER again, even though
he has received a still valid IP config from the router, we don't want to
discard the DISCOVER like it was done before but discard the IP config
assignment and offer a new one.
Issue #2534
When this flag is set in the config tag, the NIC router will print a
short information to the log for each general state change of a domain.
This includes currently the IP-configuration state and the number of
connected NIC sessions. This a useful addition as the normal verbose
flag's purpose is a very deep insight into almost every activity in the
router, which is cool for debugging sophisticated problems but normally
floods the log and therefore discards this option for, e.g., desktop
systems. In such systems, the new verbosity is pretty discreet but
already gives a good hint on why packets may get dropped by the router
although the routing rules are correct.
Issue #2534
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
Added a new constructor that takes the entrypoint as constructor
argument. The original constructor retrieves the entrypoint from the
Genode environment. This does not allow to use a different entrypoint.
Replace former rtt_sec attribute of the <config> tag by more specific
(and still optional) attributes for timeouts used in the NIC router
(these are also the default values):
<config dhcp_discover_timeout_sec="10"
dhcp_request_timeout_sec="10"
dhcp_offer_timeout_sec="10"
udp_idle_timeout_sec="30"
tcp_idle_timeout_sec="600"
tcp_max_segm_lifetime_sec="30">
Details about the new attributes can be found in the README of the router.
Issue #2590
On x86 64 bit with SeL4, the test needs around 80MB that must be
completely composed of 4KB-pages due to current limitations of the SeL4
port. Thus, Core must flush the page table caches pretty often during
the test which is an expensive high-prior operation and makes it
impossible to provide a highly precise time.
Multi-wraps
-----------
Previously, on every new timeout, we programmed registers LR=timeout and
CMP=0. The counter than counted from LR down to 0, triggered the IRQ,
jumped back to LR, and counted down again. If one installed small
timeouts (< 1000 us), it was likely that the counter wrapped multiple
times before we were able to read it out. Initially, this was not a big
issue as the additional wraps were simply ignored and the amount of time
lost through this was not big. But when we want to do correct rate
limitation, multiple wraps cause an overflow in the additional
calculations, and this has a big effect on the resulting time value.
Thus, we now program the counter to start from ~0 and count down to 0.
We set CMP=~0-timeout so that the timer still triggers the IRQ at the right
time. The counter continues counting down after the IRQ has triggered until
we install a new timeout. We do not consider anymore that the counter wraps.
The maximum timeout is set to half the maximum counter value, so, we should
be able to install a new timeout before the counter wraps.
Rate limit for time updates
---------------------------
In the time span between two interrupts we have to remember how many ticks
we have already added to the time value. This is because at each call of
curr_time we can only see how many ticks have passed since the last call of
schedule_timeout and not since the last call of curr_time. But we want to
limit the rate of time updates in curr_time. With the member for ticks that
were already added since the last call to schedule_timeout we can then
calculate how many are yet to be added.
This patch supplements the existing 'hover' report with the information
whether or not the user has recently moved the pointer. This works
analogously to how the 'focus' report features the information about
recent button/keyboard activity.
Together, the 'hover' and 'focus' reports may be combined to observe
prolonged user inactivity, e.g. to activate a lock screen.
This patch enables nitpicker to use an external focus policy instead of
the traditional builtin click-to-focus policy. The external focus policy
is obtained from a 'focus' ROM. The focus ROM is expected to have a
'label' attribute with the value set to the label of the to-be focused
client.
This patch revises the implementation of nitpicker in the following
respects:
- Split the implementation into smaller files,
- Consistently use the 'Nitpicker' namespace,
- Avoid the use of format strings,
- Retire old (and hackish) debug mode,
- Removal of unused timer connection,
- Merging 'Session' into 'Session_component',
- Merging 'Mode' into 'User_state',
- Adding the notions of 'View_owner' and 'Focus' as interfaces,
- Untangle 'User_state' and 'View_stack'
This patch adds a 'Color::print' method as counterpart to the 'ascii_to'
function. If the color is opaque (alpha is 255), its output has the form
"#rrggbb". If the color has a distinct alpha value, the output has the
form "#rrggbbaa". The new version of the 'ascii_to' overload for 'Color'
is able to deal with both forms.
This patch supplements init's service-forwarding mechanism to propagate
the insufficient RAM/cap quota conditions from the server to the client.
Without it, the client's session request stays pending infinitely.
This is a follow-up patch to "init: periodic state updates if sensible".
In situations where the report rate is deliberately limited via the
'delay_ms' attribute while also reporting child-resource stats, we don't
want generate reports at a fixed rate of one second. This patch limits
the rate according to the 'delay_ms' value.
The input filter used to temporarily close all input sessions upon its
reconfiguration. In most cases, the same set of sessions is
re-established immediately afterwards. However, at the server (driver)
side, the closing of the session implicitly disables the input-event
queue. Hence events generated by the hardware while the session is
closed are dropped. This becomes a noticeable problem when using the
recently added <rom> modifier feature for handling capslock. The change
of the ROM always triggers the re-configuration of the input filter.
When pressing capslock and other keys at a high rate, press/release
events may get lost.
This patch solves this problem by maintaining all input sessions that
are defined in both the old and new configuration. It thereby removes
the short duration where the input event queues are temporarily disabled
at the drivers.
This makes '/' and the actual root of VFS distinguishable. A VFS root
may contain one ore more '/' entries for each file system. 'opendir' for
the VFS root opens all file systems via 'open_composite_dir', while
'opendir' for '/' only returns a VFS handle.
Fixes#2569
This patch adds a sanity check to the Event::type accessor. If the key
code of a given PRESS or RELEASE event is out of the valid range, it
reports an INVALID event. This way, client side code does not need to
deal with such edge cases. E.g., on Lenovo notebooks, the ps2 driver
reports strange key events when pressing shift-pageup/pagedown,
violating the general assumption that there is a release event for each
press event. By flagging these events as INVALID, the client-side logic
stays intact.
Check at the VFS server that the capability cost of sessions do not
exceed the session quota donation. Raise the default initial capability
donation for File_system connections.
Fix#2553
Previously the destructors of classes derived from Source were not
called when destructing a Source because Source missed the declaration
of a virtual destructor. This caused at least problems when
reconstructing a Chargen_source with its Chargen_repeater that includes
a One_shot_timeout. In this case, the members of the Alarm behind the
timeout were not reset and the next time beeing used the Alarm appeared
to be still active which wasn't true.
Fixes#2570
A new configuration may implicitly change the domain color of the
currently focused session. We need to refresh the report to trigger an
update of the status bar.
This commit replaces the old xray_trigger component by a new component
called global_keys_handler. For details, please refer to the issue text
and the accompanied README file.
Fixes#2554
Whenever a childs is terminated the exit value is propagate through a
new state report. Thereby it becomes possibly for a managing component
to react upon the terminating condition of a child.
Issue #2558.
If the attribute 'interface' is not set in a 'domain' tag, the router tries to
dynamically receive and maintain an IP configuration for that domain by using
DHCP in the client role at all interfaces that connect to the domain. In the
DHCP discover phase, the router simply chooses the first DHCP offer that
arrives. So, no comparison of different DHCP offers is done. In the DHCP
request phase, the server is expected to provide an IP address, a gateway, a
subnet mask, and an IP lease time to the router. If anything substantial goes
wrong during a DHCP exchange, the router discards the outcome of the exchange
and goes back to the DHCP discover phase. At any time where there is no valid
IP configuration present at a domain, the domain does only act as DHCP client
and all other router functionality is disabled for the domain. A domain cannot
act as DHCP client and DHCP server at once. So, a 'domain' tag must either
have an 'interface' attribute or must not contain a 'dhcp-server' tag.
Ref #2534
An IPv4 config (for a domain/interface of the router) consists of
an IPv4 address, a subnet prefix specifier, an optional gateway
IPv4 address, and some flags that declare whether these fields and
the config as a whole are valid. To make the handling of those
tightly connected values easier and less error prone, we encapsulate
them in a new class.
Ref #2534
Under certain circumstances we don't want inits state report to become too
outdated even if there is no change to its config or the sessions of its
children. This is the case if init is requested to provide a capability or RAM
info of it's children via its state report. Now, init automatically updates
the state report with each 1000 ms if the attribute 'child_caps' or
'child_ram' is positively set in the 'report' tag.
Timing itself costs time. Thus, the stressfull timeout phase of the
test is not exactly as long as set but a little bit longer. This is why the
fast timeouts are able to trigger more often than they are expected to
(the timer has a static timeout-rate limit). Normally we consider this effect
through an error tolerance of 10%. But at least on foc x86_32 (PIT with very
low max timeout), timing is so expensive that 10% is not enough. We have to
raise it to 11%.
This patch propages the 'Service_denied' condition of forwarded sessions
to the parent. Without it, the invalid session request stays pending
infinitely, which leads to the problem described in issue #2542. It
turns out that suggested solution given in the issue text is actually
not needed when applying this fix.
Fixes#2542
The ROM filter did not handle the situation where the generated content
exceeds the size of the initially allocated dataspace for the target
buffer. This patch wraps the XML generation in a retry loop that
expands the buffer as needed.
This patch makes the specification of screen coordinates more flexible.
First, the 'origin' attribute allows one to refer to either of the four
screen corners without knowing the screen size. Second, the 'width'
and 'height' values now accept negative values, which are relative to
the screen size.
The 'File_system::Connection' already performs an on-demand session
upgrade should the server report an 'Out_of_caps' or 'Out_of_ram'
condition. So file-system clients are normally relieved from handling
those exceptions. However, the upgrade was limited to two attempts per
operation (which amounts to 16 KiB). When using the Rump VFS plugin in
the VFS server, this amount does not always suffice. So the exception is
reflected to the client. I observed this problem as a message "unhandled
error" printed by fs_rom. This patch removes the upgrade limit such that
a greedy file-system server becomes iteratively upgraded until it stops
arguing or the client's RAM is exhausted.
This was an error output-line for each affected packet previously but it
is pretty normal for the router to receive packets whose network layer
protocol it doesn't know . In the default case, these packets shall be
ignored silently.
Ref #2490
One can configure the NIC router to act as DHCP server at interfaces of a
domain by adding the <dhcp> tag to the configuration of the domain like
this:
<domain name="vbox" interface="10.0.1.1/24">
<dhcp-server ip_first="10.0.1.80"
ip_last="10.0.1.100"
ip_lease_time_sec="3600"
dns_server="10.0.0.2"/>
...
</domain>
The attributes ip_first and ip_last define the available IPv4 address
range while ip_lease_time_sec defines the lifetime of an IPv4 address
assignment in seconds. The IPv4 address range must be in the subnet
defined by the interface attribute of the domain tag and must not cover
the IPv4 address in this attribute. The dns_server attribute gives the
IPv4 address of the DNS server that might also be in another subnet.
The lifetime of an offered assignment is the configured round trip time of
the router while the ip_lease_time_sec is applied only if the offer is
requested by the client in time.
The ports/run/virtualbox_nic_router.run script is an example of how to
use the new DHCP server functionality.
Ref #2490
Previously, garbage collect was only done when an incoming packet passed the
Ethernet checks. Now it is really done first when receiving a packet at an
interface.
Ref #2490
If the router has no gateway attribute for a domain (means that the router
itself is the gateway), and it gets an ARP request for a foreign IP, it shall
answer with its own IP.
Ref #2490
Do not use two times the RTT for the lifetime of links but use it as
it is configured to simplify the usage of the router. Internally, use
Microseconds/Duration type instead of plain integers.
Ref #2490
The nic_dump uses a wrapper for all supported protocols that
takes a packet and a verbosity configuration. The wrapper object can
than be used as argument for a Genode log function and prints the
packet's contents according to the given configuration. The
configuration is a distinct class to enable the reuse of one instance
for different packets.
There are currently 4 possible configurations for each protocol:
* NONE (no output for this protocol)
* SHORT (only the protocol name)
* COMPACT (the most important information densely packed)
* COMPREHENSIVE (all header information of this protocol)
Ref #2490
Provide utilities for appending new options to an existing DHCP packet
and a utility for finding existing options that returns a typed option
object. Remove old version that return untyped options.
Ref #2490
Apply the style rule that an accessor is named similar to the the underlying
value. Provide read and write accessors for each mandatory header attribute.
Fix some incorrect structure in the headers like with the flags field
in Ipv4_packet.
Ref #2490
Encapsulate the enum into a struct so that it is named
Ethernet_frame::Type::Enum, give it the correct storage type
uint16_t, and remove those values that are (AFAIK) not used by
now (genode, world).
Ref #2490
Do not stop routing if the transport layer protocol is unknown but
continue with trying IP routing instead. The latter was already
done when no transport routing could be applied but for unknown transport
protocols we caught the exception at the wrong place.
Ref #2490
No starvation of timeout signals
--------------------------------
Add several timeouts < 1ms to the stress test and check that timeout
handling doesn't become significantly unfair (starvation) in this situation
where some timeouts trigger nmuch faster than they get handled.
Rate limiting for timeout handling in timer
-------------------------------------------
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
We did not set the correct now_period previously but it wasn't conspicuous
because the bug triggered not before a full period had passed which on most
platforms is a pretty long time.
Ref #2490
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 fixes an aliasing problem of the 'close' method signature
that prevented the Input::Root_component::close method to be called.
This way, the event-queue state was not reset at session-close time,
which prevented a subsequent session-creation request to succeed. With
the patch, input servers like ps2_drv, usb_drv that rely on the
Input::Root_component support the dynamic re-opening of sessions. This
happens in particular when using a dynamically configured input filter.
We update the alarm-scheduler time with results of
Timer::Connection::curr_time when we schedule new timeouts but when
handling the signal from the Timer server we updated the alarm-scheduler
time with the result of Timer::Connection::elapsed_us. Mixing times
like this could cause a non-monotone time value in the alarm scheduler.
The alarm scheduler then thought that the time value wrapped and
triggered all timeouts immediately. The problem was fixed by always
using Timer::Connection::curr_time as time source.
Ref #2490
This patch applies the handling of cursor keys, function keys, and page
up/down keys even if no keymap is defined. This is the case when using
the terminal with character events produced by the input filter.
Create periodic and one-shot timeouts with the maximum duration
to see if triggers any corner-case bugs. They must not trigger during
the test.
Ref #2490
If we add an absolute timeout to the back-end alarm-scheduler we must first
call 'handle' at the scheduler to update its internal time value.
Otherwise, it might happen that we add a timeout who's deadline is so big that
it normally belongs to the next time-counter period but the scheduler thinks
that it belongs to the current period as its time is older than the one used
to calculate the deadline.
Ref #2490
When we have two time values of an unsigned integer type and we create
the difference and want to know wether it is positive or negative within
the same value we loose at least one half of the value range for casting
to signed integers. This was the case in the alarm scheduler when
checking wether an alarm already triggered. Even worse, we casted from
'unsigned long' to 'signed int' which caused further loss on at least
x86_64. Thus, big timeouts like ~0UL falsely triggered directly.
Now, we use an extra boolean value to remember in which period of the
time counter we are and to which period of the time counter the deadline
of an alarm belongs. This boolean switches its value each time the time
counter wraps. This way, we can avoid any casting by checking wether the
current time is of the same period as the deadline of the alarm that we
inspect. If so, the alarm is pending if "current time >= alarm
deadline", otherwise it is pending if "current time < alarm deadline".
Ref #2490
If the PIT timer driver gets activated too slow (e.g. because of a bad priority
configuration), it might miss counter wraps and would than produce sudden time
jumps. The driver now detects this problem dynamically, warns about it and
adapts the affected values to avoid time jumps.
Ref #2400
The NIC router always reports the link state "Up" (true) because
the effective link state depends on the targeted remote interface
and thus on the individual routing for each packet. Consequently,
also the signal handler for state changes gets ignored.
Ref #2490
IP stacks may treat a network interface as "down" when it states a MAC
address with the I/G bit (bit 40) set to "Group" (value 0) instead of
"Individual" (value 1). This was observed with a TinyCore 8 inside a
Virtualbox VM. Thus, the previously choosen 03:03:03:03:03:00 as base
for the MAC address allocator is bad. Now we use the 02:02:02:02:02:00
instead. This also ensures that the MAC addresses are not marked as
"Universal" but as "Local" (bit 41, value 1) which is correct in general
as the router allocates MAC addresses only for virtual networks.
Ref #2490
The timer driver should always be of the highest priority to avoid
problem with timers that have low max-counter values like the PIT
with only 53 ms.
Ref #2400
The NIC dump component didn't support forwarding of link states and link-state
signals until now. Furthermore, it now prints MAC address and link state
on session creation and on every link state change.
Ref #2490
Previously, the uplink session was created on component startup while the
creation of the downlink session is timed by the client component. This
created a time span in which packets from the uplink were dropped at the
nic_dump. Now the uplink session-request is done by the session component
of the downlink.
Ref #2490
Add a "writeable" policy option to the ahci_drv and part_blk Block
servers and default from writeable to ready-only. Should a policy
permit write acesss the session request argument "writeable" may still
downgrade a session to ready-only.
Fix#2469
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
Due to the simplicity of the algorithm that translated from timer ticks
to time, we lost microseconds precision although the timer allows for it.
Ref #2400
When synchronizing with the remote time source, we have to take care that the
measured time difference cannot become null because its real value is smaller
than the measurement granularity. Since the granularity is one microsecond, we
simply go on polling timestamp and time until the microsecond has passed.
This busy waiting should be no problem for the system for two reasons. First,
it is limited to a relatively small amount of time and second, a busy lock
does not happen because the time source that is responsible for the limiting
factor is explicitely called on each poll.
Ref #2400
The VFS library can be used in single-threaded or multi-threaded
environments and depending on that, signals are handled by the same thread
which uses the VFS library or possibly by a different thread. If a VFS
plugin needs to block to wait for a signal, there is currently no way
which works reliably in both environments.
For this reason, this commit makes the interface of the VFS library
nonblocking, similar to the File_system session interface.
The most important changes are:
- Directories are created and opened with the 'opendir()' function and the
directory entries are read with the recently introduced 'queue_read()'
and 'complete_read()' functions.
- Symbolic links are created and opened with the 'openlink()' function and
the link target is read with the 'queue_read()' and 'complete_read()'
functions and written with the 'write()' function.
- The 'write()' function does not wait for signals anymore. This can have
the effect that data written by a VFS library user has not been
processed by a file system server yet when the library user asks for the
size of the file or closes it (both done with RPC functions at the file
system server). For this reason, a user of the VFS library should
request synchronization before calling 'stat()' or 'close()'. To make
sure that a file system server has processed all write request packets
which a client submitted before the synchronization request,
synchronization is now requested at the file system server with a
synchronization packet instead of an RPC function. Because of this
change, the synchronization interface of the VFS library is now split
into 'queue_sync()' and 'complete_sync()' functions.
Fixes#2399
This patch changes init's service forwarding such that pending requests
are kept unanswered as long as the requested service is not present
(yet). In dynamic-init scenarios, this is needed in situtions where the
dynamic init is known to eventually provide the service but the internal
subsystem is not ready yet. Previously, a client that attempted to
request a session in this early phase would get a 'Service_denied'
exception. By deferring the forwarding in this situation, the behaviour
becomes deterministic.
If a matching '<service>' exists but there is no matching policy sub
node, the request is answered with 'Service_denied' - as expected.
Currently, init does not test wether a service is abandoned on a new
configuration if the service was routed via an any-child route. Trigger
this behaviour in the init test.
Ref #2483
The calibration of the interpolation parameters was previously only done
periodically every 500 ms. Together with the fact that the parameters
had to be stable for at least 3 calibration steps to enable
interpolation, it took at least 1.5 seconds after establishing a
connection to get microseconds-precise time values.
This is a problem for some drivers that directly start to poll time.
Thus, the timer connection now does a calibration burst as soon as it
switches to the modern mode (the mode with microseconds precision).
During this phase it does several (currently 9) calibration steps
without a delay inbetween. It is assumed that this is fast enough to not
get interrupted by scheduling. Thus, despite being small, the measured
values should be very stable which is why the burst should in most cases
be sufficient to get the interpolation initialized.
Ref #2400
When in modern mode (with local time interpolation), the timer
connection used to maximize the left shifting of its
timestamp-to-microseconds factor. The higher the shift the more precise
is the translation from timestamps to microseconds. If the timestamp
values used for determining the best shift were small - i.e. the delay
between the calibration steps were small - we may got a pretty big
shift. If we then used the shift with bigger timestamp values - i.e.
called curr_time seldom or raised calibration delays - the big shift
value became a problem. The framework had to scale down all measured
timestamps and time values temporarily to stay operative until the next
calibration step.
Thus, we now raise the shift only that much that the resulting factor
fullfills a given minimum. This keeps it as low as possible according
to the precision requirement. Currently, this requirement is set to 8
meaning that the shifted factor shall be at least 2^8 = 256.
Ref #2400
As the timer session now provides a method 'elapsed_us', there is no more need
for doing any internal calculations with values of milliseconds.
Ref #2400
As timer sessions are not expected to be microseconds precise (because
of RPC latency and scheduling), the session interface provided only a
method 'elapsed_ms' although the back end of this method in the timer
driver works with microseconds.
However, in some cases it makes sense to have a method 'elapsed_us'. The
values it returns might be milliseconds away from the "real" time but it
allows you to work with delays smaller than a millisecond without
getting a zero delta value.
This commit is motivated by the need for fast bursts of calibration
steps for the time interpolation in the new timer connection.
Ref #2400
The run script did not consider the routing for the environment ROM
sessions for the test-iso component. It routed all ROM sessions -
including the ones for the executable and the dynamic linker - to
fs_rom. The patch also adds the cap quota definitions required since
version 17.05 and fixes a whitespace inconsistency between the test
program and the run script.
Thanks to Steven Harp for reporting!
This is expected by hardware terminals, ie., terminal programs connected
to null-modem serial connections. Otherwise, the next line starts at the
column right after the last line.
The new version of the test exercises the combination of fs_report with
ram_fs and fs_rom as a more flexible alternative to report_rom.
It covers two corner cases that remained unaddressed by fs_rom and
ram_fs so far: First, the late installation of a ROM-update signal
handler at fs_rom right before the content of the file is modified.
Second, the case where the requested file is not present on the file
system at the creation time of the ROM session. Here, the ram_fs missed
to inform listeners for the compound directory about the later created
file.
This patch ensures that fs_rom delivers a ROM-update notification in the
case where the underlying file was changed in-between requesting the
initial ROM content and registering the signal handler.
With the introduction of the CONTENT_CHANGED notifications delivered via
the packet stream, the assumption that no more than one READ packet is
in flight at all times does no longer hold. If the fs server responds
to a CONTENT_CHANGED packet while the fs_rom expects the completion of a
read request, the '_update_dataspace' method would prematurely return,
leaving the dataspace unpopulated. This patch solves the problem by
specifically waiting for the completion of the read request.
Session_requester inherits from Dynamic_rom_session::Content_producer
which specifies the Buffer_capacity_exceeded exception which is thrown
on insufficient buffer space.
On platforms that use the PIT timer driver, 'elapsed_ms' is pretty
inprecise/unsteady (up to 3 ms deviation) for a reason that is not
clearly determined yet. On Fiasco and Fiasco.OC, that use kernel timing,
it is the same. So, on these platforms, our locally interpolated time
seems to be fine but the reference time is bad. Until this is fixed, we
raise the error tolerance for these platforms in the run script.
Ref #2400
Appending a suffix to report filenames was behavior inherited from
fs_log, it prevents creating files where directories need to be created
later. But unlike logs, only a subset of the hierarchy will report and
those that do append a component-local label, so the risk of collision
is low.
By removing the suffix fs_rom can serve reports back as ROM just as
report_rom does.
Ref #2422
In the timeout framework, we maintain a translation factor value to
translate between time and timestamps. To raise precision we scale-up
the factor when we calculate it and scale-down the result of its
appliance later again. This up and down scaling is achieved through
left and right shifting. Until now, the shift width was statically
choosen. However, some platforms need a big shift width and others a
smaller one. The one static shift width couldn't cover all platforms
which caused overflows or precision problems.
Now, the shift width is choosen optimally for the actual translation
factor each time it gets re-calculated. This way, we can take care that
the shift always renders the best precision level without the risk for
overflows.
Ref #2400
The result-buffer related members of the fast polling test are
the same for each buffered result type. Thus, we can make the
code easier by providing them through a struct.
Ref #2400
This patch increases init's preserved RAM and capability quota to
account for a current limitation of init with respect to the creation of
sessions to parent services:
In contrast to regular routed services, sessions to parent services are
created via 'Env::session'. The implementation of 'Env::session'
automatically upgrades session quotas on demand, which is the desired
behavior for regular 'Connection' objects. However, for sessions
established on the behalf of init's children, we would need to reflect
the error condition to the child instead of resolving it locally within
init (by subsidizing the session with init's quota). This patch leaves
this issue unresolved but fixes the symptom for the bomb test. It is
meant as an interim solution until the handling of parent sessions is
revised.
On QEMU, NOVA uses the pretty unstable TSC emulation as primary time
source. Thus, timeouts do not trigger with the common precision (< 50
ms). Use an error tolerance of 200 ms for this platform constellation.
Ref #2400
Apparently this construct leads to a compiler errors like
error: second operand to the conditional operator is of type ‘void’, but
the third operand is neither a throw-expression nor of type ‘void’
The fast polling test uses one timer session for raw 'elapsed_ms' calls
and another one for potentially interpolated 'curr_time' calls. It then
compares the two results against each other. However, until now, the
test did not consider that the duration of the session construction may
create a remarkable shift between the local times of the two sessions.
This shift is now determined and compensated before doing any
comparison.
Ref #2400
The multiple-handlers test was checking if handlers at one signal were
activated in a fair manner. But on Qemu, the error tolerance of one was
too small in rare cases (2 of 100 runs). However, having multiple
handlers for the same signal context can be considered deprecated
anyway. With the recommended Signal_handler wrapper for signal sessions,
you can't use this feature. Thus, we removed the multiple-handlers test.
Fixes#2450
We incorrectly used 'unsigned long' (which is 32 or 64 bit depending on
the CPU architecture) for a timestamp (which is always 64 bit) in the
timer-connection implementation.
Ref #2435
On platforms were we do not have local time interpolation we can simply
skip the first test stage in the timeout test. This way, we can at least
test the rest.
Fixes#2435
On ARM, we do not have a component-local hardware time-source. The ARM
performance counter has no reliable frequency as the ARM idle command
halts the counter. Thus, we do not do local time interpolation on ARM.
Except we're on the HW kernel. In this case we can read out the kernel
time instead.
Ref #2435
The explicit relative location of the file instructed both target builds
to generate ../main.o which gloriously fails with parallel builds. The
produced range of error messages was astonishing ranging from "file
truncated" to "TLS reference in ../main.o mismatches non-TLS reference
in ../main.o".
If a child is allowed to constrain physical memory allocations but left
the 'phys_start' and 'phys_size' session arguments blank, init applies
builtin constraints for allocating DMA buffers.
The only component that makes use of the physical-memory constraint
feature is the platform driver. Since the built-in heuristics are
applied to the platform driver's environment RAM session, all
allocations performed by the platform driver satisfy the DMA
constraints.
To justify building-in these heuristics into init as opposed to
supplying the values as configuration arguments, the values differ
between 32 and 64 bit. The configuration approach would raise the need
to differentiate init configurations for both cases, which are
completely identical otherwise.
Issue #2407
This commit removes support for limitation of RAM allocations from the
platform_drv. A subsequent commit adds this feature to init.
Issue #2398
Issue #2407
With the capability-quota mechanism, the terminal-session won't always
be constructed completely on the first try (we may run out of caps in
the middle of the construction). Therefore, all members of the object
must be properly destructable. Furthermore, the patch replaces the
sliced heap by a heap to avoid allocating a new dataspace for each line
of the cell array.
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
