At least 64bit Seoul dies with Region_conflict reliable and reproducible.
When during startup of Seoul some Genode code (caused by executing some
constructors) try to attach a region, the region manager code in the rm_session
will try to place the attachment at the smallest large enough aligned free
virtual region.
For now, I observed one attachment causing trouble (but not knowing who causes
this - it does also not really matter). The questionable region is 0x4000 of
size for 32bit and 0x8000 of size for 64bit.
To steer the region manager a bit, we try now following trick:
With this commit the load address of the binary for 32 and 64 bit is moved
close to the end of the virtual address space, but leaving enough free virtual
space for the above observed attachment (and a bit more).
The region manager code now will try to fill up the virtual region behind
the binary up to the end of the virtual address space, effectively letting the
lower virtual region untouched - hopefully.
Works for now, but it will break again - for sure.
Fixes#519
If a script is executed which uses a interpreter that does not exist the
construction of the child fails and potentially leaks memory because the
wrong delete operator is called.
Therefore the binary dataspace of the script and the binary dataspace of
the interpreter are now checked before a new child will be created.
Fixes#812.
This patch extends the 'Parent::session()' and 'Root::session()'
functions with an additional 'affinity' parameter, which is inteded to
express the preferred affinity of the new session. For CPU sessions
provided by core, the values will be used to select the set of CPUs
assigned to the CPU session. For other services, the session affinity
information can be utilized to optimize the locality of the server
thread with the client. For example, to enable the IRQ session to route
an IRQ to the CPU core on which the corresponding device driver (the IRQ
client) is running.
This patch introduces new types for expressing CPU affinities. Instead
of dealing with physical CPU numbers, affinities are expressed as
rectangles in a grid of virtual CPU nodes. This clears the way to
conveniently assign sets of adjacent CPUs to subsystems, each of them
managing their respective viewport of the coordinate space.
By using 2D Cartesian coordinates, the locality of CPU nodes can be
modeled for different topologies such as SMP (simple Nx1 grid), grids of
NUMA nodes, or ring topologies.
This patch eliminates the "no attachment at..." warnings, which
were caused by a use-after-free problem of dataspaces. When a
dataspace was destroyed, the users of the dataspace were not
informed and therefore could not revert possible attachments to
RM sessions. The fix introduces a callback mechanism that allows
dataspace users (i.e., RM regions) to register for the event that
a dataspace vanishes.
The following types of dataspaces are handled:
* RAM dataspaces
* ROM dataspaces
* The process binary
* The binary of the dynamic linker
* Args dataspace
* Sysio dataspace
* Env dataspace
* managed RM dataspaces
The handling of ROM dataspaces is still not complete. When forking,
the ROM dataspace of the parent process gets just reused without
creating proper meta data ('Dataspace_info') for the forked process.
Similar issues might arise from other special dataspaces (e.g.,
args, env, sysio).
This patch removes all "no attachment at..." warnings except for
one (an attachment at 0).
Issue #485
The 'check_dev_tty()' function calls 'ttyname()', which calls the pthread
stub function 'pthread_main_np()', which prints a 'not implemented'
message. Calling 'check_dev_tty()' doesn't seem to be necessary, so this
patch removes the call.
Issue #815.
Previous commit denies the creation of regions larger then the dataspace.
Noux does it by setting the default size to the dataspace size without
subtracting the offset.
Fixes#591
Forgetting to restore the old utcb content results in hard to debug bugs.
Save only the amount of word items which are actually on the UTCB.
Issue #806
Avoids the message
cxx: operator delete (void *) called - not implemented. A working implementation is available in the 'stdcxx' library
during a " new ..." which causes exceptions. Happens for seoul in disk.cc
Issue #806
Split the netperf run script into 3 ones so that it can be used more easily
in an automated run.
netperf.run - use native nic driver (x86) or usb2.0 (arndale, panda)
netperf_usb30.run - use native nic driver (x86) or usb3.0
netperf_bridge.run - use native nic driver (x86) or usb3.0 (arndale) and bridge
Issue #794
Sometimes the ports are not freed up quick enough by the host system after the
first test finished. The port restriction is mainly required for qemu, so don't
use it for bare metal hardware tests.
lwip reports via getsockopt the size of the default size of the receive buffer
to the netperf server. lwip returns 2GB and netperf server uses this value to
allocate some buffers - which of course fails with out of memory.
Reduces the "default size" to some smaller value.
With the commit we are not forced anymore to (but still can) use specific
netperf client options regarding memory allocations of the receive buffer.
MAERTS is STREAM backwards and effectively lets the netserver sends the packets
to the netperf client. So, TCP_STREAM measure the receive performance of the
lwIP stack on Genode and TCP_MAERTS the send performance of the lwIP stack
on Genode.
The PWD variable contains the current working directory of the original
location where 'make -C' is executed, not the directory specified as
argument of '-C'. The tools referenced by ports/libports, however,
expect PWD to point to the root of the respective repository.
This patch splits the download of signatures from the download of the
archive to improve robustness. This way, signature files will be
downloaded even if the corresponding archive is already in place.
Issue #748
This patch makes the handling of failed integrity checks of 3rd-party
packages more robust. Previously, a once failed 'make prepare PKG=curl'
would not leave any trace of the verification state. Hence, a successive
attempt to perform the 'make prepare' step again succeeded even if the
signature check failed.
To solve this problem, the outcome of a successful signature check is
represented by a tag file called 'download/<archive-name>.verified'.
Because the rule for extracting the archive depends on the .verify tag
file, the extraction step is not performed until the signature check
succeeds.
Issue #748
With this patch, 'liquid_framebuffer' can be reconfigured at runtime.
The configuration arguments are now provided as XML attributes, matching
those for 'nit_fb'. Furthermore, two new configuration options are added:
<config ...
resize_handle="off" - show/hide a resize handle widget in the lower
right window corner
decoration="on" - show/hide window decoration
(title bar and borders)
/>
Fixes#740Fixes#14
The previous version of the PS/2 mouse backend manged mouse motion
events in a strange way, effectively throwing away most information
about the motion vector. Furthermore, the tracking of the mouse-button
states were missing. So drag-and-drop in a guest OS won't work. The new
version fixes those issues. For the transformation of input events to
PS/2 packets, a the Genode::Register facility is used. This greatly
simplifies the code.
This patch replaces the error-prone manual locking with the use of the
'Synced_interface' for the motherboard and the VCPU dispatcher. It also
removes all globally visible locks. Locks are now explicitly passed to
subsystems when needed.
In this version of the transition the Hip structure from Genode is reused,
@nfeskes seoul_libc_support is used for the string functions and the
nul/config.h is replaced by just using a constant value in the one place where
the file was needed.
Related to #666.
With the change of the Timer::Session interface, all scenarios that use
the timer use core's SIGNAL service. So we need to route sessions
accordingly.
In addition to the adaptation to the changed timer, this patch removes
some stale examples that predate the run tool and are no longer used.
This patch simplifies the way of how Genode's base libraries are
organized. Originally, the base API was implemented in the form of many
small libraries such as 'thread', 'env', 'server', etc. Most of them
used to consist of only a small number of files. Because those libraries
are incorporated in any build, the checking of their inter-dependencies
made the build process more verbose than desired. Also, the number of
libraries and their roles (core only, non-core only, shared by both core
and non-core) were not easy to capture.
Hereby, the base libraries have been reduced to the following few
libraries:
- startup.mk contains the startup code for normal Genode processes.
On some platform, core is able to use the library as well.
- base-common.mk contains the parts of the base library that are
identical by core and non-core processes.
- base.mk contains the complete base API implementation for non-core
processes
Consequently, the 'LIBS' declaration in 'target.mk' files becomes
simpler as well. In the most simple case, only the 'base' library must
be mentioned.
Fixes#18
Cap_sessions and portals created via the sessions are nowadays freed up during
c++ object destruction. Because of that the exception portals for a vCPU thread
get be revoked as soon as the cap_session object leaves its scope.
Keep one cap_session for the whole lifetime of the vmm to avoid disappearing
exception portals.
Related to #582.
The setup now uses nitpicker and nit_fb to display several instances of
vancouver. The guest OS binaries must be supplied in the
'<build-dir>/bin' directory manually.
Furthermore, the patch lets launchpad pass Block, Nic, and Rtc to the
parent.
Vancouver can now assign block devices to guests using the Block
interface. The machine has to be configured to use a specified drive,
which could be theoretically routed to different partitions or services
via policy definitions. Currently the USB driver only supports one
device. Genode's AHCI driver is untested.
If the session quota is too low, random pagefaults can occur on the
stack.
According to @Nils-TUD, it is necessary to protect the DiskCommit
messages with a lock against deadlocking with the timer. Observations
showed that this mitigates some problems with Gentoo on real hardware.
Vancouver is now able to use the Intel 82576 device model from NUL to
give VMs access to the network via the nic_bridge service. In order to
integrate the device model, it had to be renamed to i82576 due to XML
limitations. This is done by a patch applied via the 'make prepare'
mechanism.
Although current network card models in Vancouver panic if they can't
get a MAC address, the OP_GET_MAC hostop now fails gracefully in the
case where no nic_drv or nic_bridge is available.
The guest VM can now be provided with a framebuffer and keyboard input.
Mouse positioning of the guest is a problem. Because the PS2 model applies
some calculations to the movement values, it can happen that overflows mess
with the cursor. Therefore the handling was changed and only movements of 1
and -1 are sent. Since absolute positioning is not possible with PS2, we
have to live with this limitation until USB HID is implemented.
For the framebuffer size in Vancouver the configuration value in the machine
XML node is used. It is possible to map the corresponding memory area
directly to the guest, regardless if it is from nitpicker,
liquid_framebuffer or vesa_drv. The guest is provided with two modes (text
mode 3 and graphics mode 0x114 (0x314 in Linux).
Pressing LWIN+END while a VM has focus resets the virtual machine. Also,
RESET and DEBUG key presses will not be forwarded to the VM anymore.
It is possible to dump a VM's state by pressing LWIN+INS keys.
The text console is able to detect idle mode, unmaps the buffer from the
guest and stops interpreting. Upon the next pagefault in this area, it
resumes operation again. The code uses a simple checksum mechanism instead
of a large buffer and memcmp to detect an idle text console. False
positives don't matter very much.
When an EPT/NPT fault occurs during IDT vectoring, the original event must
be reinjected. Additionally we may have to inject an IRQ window if another
event is already pending.
With this patch, the 'Signal_receiver::dissolve()' function does not return
as long as the signal context to be dissolved is still referenced by one
or more 'Signal' objects. This is supposed to delay the destruction of the
signal context while it is still in use.
Fixes#594.
With this patch the destruction of Noux 'Child' objects gets delayed
further until the exit signal has been dispatched. This prevents the
self-destruction of the signal dispatcher, which is a member of the
'Child' object.
Fixes#603.
Several users of the signal API used custom convenience classes to
invoke signal-handling functions on the reception of incoming signals.
The 'Signal_dispatcher' pattern turned out to be particularly useful. To
avoid the duplication of this code across the code base, this patch
adds the interface to 'base/signal.h'.
Furthermore, the patch changes the 'Signal::num()' return type from int
to unsigned because negative numbers are meaningless here.
Fixes#511
When matching the 'label' session argument using '<if-args>' in a
routing table, we can omit the child name prefix because it is always
the same for all sessions originating from the child anyway. Therefore,
this patch adds a special case for matching session labels. It makes the
expression of label-specific routing more intuitive.
Add functionality to lookup an object and lock it. Additional the case is
handled that a object may be already in-destruction and the lookup will deny
returning the object.
The object_pool generalize the lookup and lock functionality of the rpc_server
and serve as base for following up patches to fix dangling pointer issues.
File_io_channel now includes ioctl() because pseudo devices which are
implemented as a file-system, e.g. /dev/tty, are controlled via ioctl()
requests. The method is exported to the Vfs through the Vfs_io_channel
class.
The random file-system provides an arc4 based urandom implementation
which is needed for OpenSSL.
NOTE: the Arc4random class currently _does not collect enough_ random
bytes!
We will reuse the terminal connection in the stdio filesystem
to implement /dev/tty. Therefor we need to access the terminal
from different locations which is simplified if it is provided
by a singleton.
With this patch the make install target for Noux applications gets defined
in the 'NOUX_INSTALL_TARGET' make variable with 'install-strip' as default.
Fixes#503.
Since the number of the fds in the select call is not necassarily equal
to the number of fds SYSCALL_SELECT returns we have to reset the number
of fds given as argument to select().
This keeps OpenSSH from reading from a non-ready socket.
Fixes#494.
Currently the lookup of a specific record in a tar archive may take a long
time, because the records get searched in sequence. With this patch a tree
structure representing the directory layout gets created when loading the
tar file for speeding up record lookups.
Fixes#491.
The previous implementation disregards the fact that we actually have
to use libc's plugin mechanism if we play with fds. So in the end the
libc did not know to which plugin the fd belonged.
Fixes#493.
Implies support for the ARMv6 architecture through 'base-hw'.
Get rid of 'base/include/drivers' expect of 'base/include/drivers/uart'.
Merge with the support for trustzone on VEA9X4 that came from
Stefan Kalkowski.
Leave board drivers in 'base/include/platform'.
Rework structure of the other drivers that were moved to
'base_hw/src/core' and those that came with the trustzone support.
Beautify further stuff in 'base_hw'.
Test 'nested_init' with 'hw_imx31' (hardware) and 'hw_panda_a2' (hardware),
'demo' and 'signal' with 'hw_pbxa9' (qemu) and 'hw_vea9x4'
(hardware, no trustzone), and 'vmm' with 'hw_vea9x4'
(hardware, with trustzone).
Since FD_ZERO() resets a whole fd_set (which is 128 bytes) using it to
reset dst_fds will override otherwise used memory if the memory was
allocated dynamically and is less than sizeof (fd_set). So instead of
using this macro we reset the fd_set manually.
The Io_receptor is now used to unblock certain I/O channels from lwip's
callback-function.
There was also a bug in which all ready-to-ready fds were overriden by
the ready-to-write ones.
The Socket_io_channel class now uses the Io_channel_backend to provide
the network related methods. In addition the Socket_io_channel_registry
was replaced with a simpler implementation which uses Io_receptors to
unblock I/O channels from the callback-function of lwip.
The eglgears application depends on 'sqrt'. With the old tool chain,
this symbol was resolved through the dependencies of the 'gallium'
library. This does not work anymore for the new tool chain.
A similar issue arised for avplay, where we need to explicitly
state the av components in the LIBS declaration of the target.
After this commit "make prepare" uses HTTP, HTTPS, or FTP where possible
fvor downloading third-party source codes. This prevents problems with
strict firewall rules where only selected ports are usable.
Unfortunately, git.l4android.org does not support Git via HTTP and,
therefore, the sources need a working Git port (9418).
Fixes#443.
There are certain programs that need gettimeofday(), e.g.
network-related tools like ping(1) etc. but also filesystem-related
programs like find(1) etc. and of course time-related programs like
date(1).
As there is currently no interface in Genode for actually using clock
devices like RTC on x86 (though there is a driver for it) we "abuse"
the timeout_scheduler thread to at least provide flow-of-time.
Noux: add clock_gettime() implementation
For now, only CLOCK_SECOND is supported.
Noux: add utimes() dummy
Fixes#401
This patch moves the thread iterator function declarations in GDB
monitor's 'Cpu_session_component' above the 'CPU session interface'
comment, because these functions are not part of the CPU session
interface anymore.
Fixes#10.
