This patch equips Sculpt with the ability to customize the system image
in very flexible ways.
All customizable aspects of the image have been relocated from the
former sculpt.run script and the accompanied gems/run/sculpt/ directory
to a new location - the sculpt/ directory - which can exist in any
repository. The directory at repos/gems/sculpt/ serves as reference.
The sculpt directory can host any number of <name>-<board>.sculpt files,
each containing a list of ingredients to be incorporated into the
Sculpt system image. The <name> can be specified to the sculpt.run
script. E.g., the following command refers to the 'default-pc.sculpt'
file:
make run/sculpt KERNEL=nova BOARD=pc SCULPT=default
If no 'SCULPT' argument is supplied, the value 'default' is used.
A .sculpt file refers to a selection of files found at various
subdirectries named after their respective purpose. In particular, There
exists a subdirectory for each file in Sculpt's config fs, like
nitpicker, drivers... The .sculpt file selects the alternative to use
by a simple tag-value notation.
drivers: pc
The supported tags are as follows.
*Optional* selection of /config files. If not specified, those files are
omitted, which prompts Sculpt to manage those configurations
automatically or via the Leitzentrale GUI:
fonts
nic_router
event_filter
wifi
runtime
gpu_drv
Selection of mandatory /config files. If not specified, the respective
'default' alternative will be used.
nitpicker
deploy
fb_drv
clipboard
drivers
numlock_remap
leitzentrale
usb
system
ram_fs
Furthermore, the .sculpt file supports the optional selection of
supplemental content such as a set of launchers.
launches: nano3d system_shell
Another type of content are the set of blessed pubkey/download files
used for installing and verifying software on target.
With the new version, it has become possible to supply a depot with the
the system image. The depot content is assembled according to the 'pkg'
attributes found in launcher files and the selected deploy config.
The resulting depot is incorporated into the system image as 'depot.tar'
archive. It can be supplied to the Sculpt system by mounting it into the
ram fs as done by the 'ram_fs/depot' configuration for the ram fs.
It is possible to add additional boot modules to the system image. There
are two options.
build: <list of targets>
This tag prompts the sculpt.run script to build the specified targets
directly using the Genode build system and add the created artifacts
into the system image as boot modules.
import: <list of depot src or pkg archives>
This tag instructs Sculpt to supply the specifid depot-archive content
as boot modules to the system image. This change eliminates the need for
board-specific pkg/sculpt-<board> archives. The board-specific
specializations can now be placed directly into the respective .sculpt
files by using 'import:'.
To make the use of Sculpt as testbed during development more convenient,
the log output of the drivers, leitzentrale, and runtime subsystems
can be redirected to core using the optional 'LOG=core' argument, e.g.,
make run/sculpt KERNEL=linux BOARD=linux LOG=core
The former pkg/sculpt-installation and pkg/sculpt-installation-pc
archives have been replaced by pkg/sculpt_distribution-pc, which
references the generic pkg/sculpt_distribution archive. Those pkgs are
solely used for publishing / distribution purposes.
Fixes#4369
Override 'try_alloc/free' because ethernet frame headers are 14 bytes
(src/dst mac (12) + ethertype (2)) causing the IP header to be 2 byte
aligned, leading to problems on platforms that require load/store
operations to be naturally aligned when reading, for example, 4 byte IP
addresses. Therefore, we align the allocation to 2 bytes, so the IP
header is aligned to 4.
issue #4312
After VirtIO::Queue refactoring buffers no longer share the same
dataspace as VirtIO rings. This makes optimal buffer calculations a lot
easier. In this case 64 buffers 2kB each will need precisely 128kB of
RAM. Previous value of 2016 will just waste 768b.
Fixed#4347
The key changes in this patch are:
* Buffer allocation is moved into a separate Buffer_pool helper. The
implementation of the buffer allocation strategy does not change.
The helper allocates a single RAM dataspace and splits it in multiple,
equally sized chunks.
* Management of main descriptor ring is enacapsulated in Descriptor_ring
helper class.
* Use separate RAM dataspaces for descriptor rings and buffers.
Previously both of them were packed into a single dataspace. This
might have been more RAM efficient, but IMO it made the code uglier and
harder to understand.
* All of the VirtIO::Queue members are now initialized on the class member
initializer list. This is possible due to previously listed changes.
* Since all VirtIO::Queue members are initalized on member initalizer
list, some additional ones can be marked as const, ex _avail, _used ring
pointers.
* Move descriptor writing code into a common method used by both
write_data and write_data_read_reply members. This avoids some code
duplication between those methods.
* Get rid of request_irq argument that most public VirtIO::Queue methods
accept. None of the existing drivers use it and I doubt this will
change any time soon.
* Use Genode namespace by default in Virtio.
This patch also fixes at least one bug that I discovered while working
on VirtIO block device driver. Namely, when chaining descriptors only the
first descriptor in the chain should be exposed in the available ring.
Issue #4347
Share datastructures for clock, power and reset related configurations
per device. In the generic platform driver component these structures
are kept empty. Driver derivates can fill the clocks settings, power and
reset switches with life. The former Driver::Env gets removed.
Fix#4338
This change of the inner working of the platform driver for ARM allows
clients to have permanent open sessions, as long as a policy node matches
the client. If devices disappear from the policy resp. from the set of
available devices (hotplug), the devices ROM of the session gets updated,
and a corresponding device session gets closed. If the device remains
untouched in the configuration but other devices appeared/disappeared, the
device session is not affected.
Ref #4330
* Track all caps and ram quotas of the sub-sessions properly
* Release DMA buffers, it is not done implicitely when destroying
the Constrained_ram_allocator
* Do not replenish quota before really releasing memory from
the allocator
Issue #4330
The Session_component must be destroyed before updating the device
model because the Session_component must also release all previously
acquired devices. If the device model is updated before, the devices
might have been removed.
Issue #4330
This implements the necessary bits to provide 2D framebuffer support on
top of VirtIO GPU device as implemented in Qemu. I don't know if any
other implementation of this specific device exists.
Compared to the ramfb driver which already exists in Genode Virtio FB driver
has one major benefit. It allows Qemu window to be dynamically resized at
runtime. The driver will treat this as resolution change and act accordingly.
Ramfb driver can currently only use the hardcoded 1024x768 screen size. Changing
screen resolution might not sound like a big deal, but it is rather useful to
run Genode on Qemu in full screen mode.
This patch changes the 'Allocator' interface to the use of 'Attempt'
return values instead of using exceptions for propagating errors.
To largely uphold compatibility with components using the original
exception-based interface - in particluar use cases where an 'Allocator'
is passed to the 'new' operator - the traditional 'alloc' is still
supported. But it existes merely as a wrapper around the new
'try_alloc'.
Issue #4324
Accidentally, the size of IOMEM dataspace got accounted within the
implementation of the platform driver for ARM. Instead we should
only account a bit for the metadata and paging of the I/O memory.
Fix#4307
The custom ack avail handler is required for zero-copy nic drivers (e.g.
the zynq nic driver), which must release the corresponding DMA buffers.
Fixesgenodelabs/genode#4277
This component can service Qemu VirtIO mouse, keyboard and tablet
devices. The implementation is based on VirtIO 1.1 device spec. Its
described in section 5.8 "Input Device".
Issue #4282
'Out_of_ram' was so far the only exception a client had to deal with
during buffer managment. Allocating memory, however, does not only
consume RAM quota but CAP quota as well.
This commit tries to mitigate that shortcoming by reflecting the
'Out_of_caps' state back to the client. Furthermore it allows for
resource accounting on certain client allocations, e.g. buffers.
Fixes#4284.
The platform driver uses a 'Constrained_ram_allocator' to allocate
meta-data on behave of a client. It uses the PD session as
'Ram_allocator' back end that in return is implemented via the
'Expanding_pd_session_client'.
Whenever the PD client itselft comes into resource shortage it will
ask its parent unconditionally. However, depending on the integration,
such a request may be left unanswered.
This commit introduces a check to prevent that situation from
occurring. In case the platform driver notices the resource shortage
it will reflect that back to the client.
Issue #4284.
The various mapping methods are modelled after the requirements of
the Intel GPUs or rather the Mesa driver back end.
With upcoming support for other driver back ends, we need to
sequeeze their requirements in as well. For now hijack 'map_buffer'
to provide for specifying the kind of attributes the client needs.
For now all buffers mapped in the GGTT for Intel GPUs are treated
as RW.
Issue #4265.
This call allows for checking if the given execution buffer has been
completed and complements the completion signal. Initially the GPU
multiplexer always sent such a signal when the currently scheduled
execution buffer has been completed. During enablement of the 'iris'
driver it became necessary to properly check of sequence number.
In case of the Intel GPU multiplexer the sequence numbers are
continous, which prompted the greater-than-or-equal check in the
DRM back end. By hidding this implementation detail behind the
interface, GPU drivers are free to deal with sequence numbers any
way they like and allows for polling in the client, where the
completion signal is now more of a progress signal.
Issue #4265.
The current info implementation (as RPC) is limited in a few ways:
* The amount of data that may be transferred is constrained by the
underlying base platform
* Most information never changes during run time but is copied
nonetheless
* The information differs depending on the used GPU device and
in its current implementation only contains Intel GPU specific
details
With this commit the 'info' RPC call is replaced with the
'info_dataspace' call that transfers the capability for the dataspace
containing the information only. This is complemented by a client
local 'attached_info' call that allows for getting typed access to
the information. The layout of the information is moved to its own
and GPU-specific header file, e.g., 'gpu/info_intel.h'
Issue #4265.
Rather than using the dataspace capability directly, let the client
choose its own local identifier that is linked to the underlying
capability.
Fixes#4265.
Right now the warning about failure to forward packet from driver to
uplink RX connection reads:
"exception while trying to forward packet from driverto Uplink
connection TX"
Add missing space between "driver" and "to".
Issue #4264
32KB is a rather small value. The driver can cope with it now, but
it does not perform as well as it should. This visible especially
in scenarions like nic_router_flood where we still often hit
synchronous wait path. Bump the size to 256kB.
Issue #4264
The problem can be seen when running nic_router_flood scenarion on arm
qemu_virt boards. With the amount of data this scenario tries to send
the driver quickly complains it has failed to push data into TX VirtIO
queue. After this warning message is printed nothing really happens and
after a while the test scenario fails.
The fact that we can't write all available data to the device is not
unexpected. VirtIO queue size is slected at initialization time and we
don't change it during driver lifetime. It can be tweaked via driver
config, but this does not change the fact that we'll always be able to
produce more data packets than we have free space in the VirtIO queue.
IMO the expected behavior of the driver in such case should be to:
1. Notify the device there is data to process.
2. Wait for the device to process at least part of it.
3. Retry sending queued packets.
One could expect returning Transmit_result::RETRY from _drv_transmit_pkt
would produce such result. Unfortunately it seems that Uplink_client_base
treats RETRY return value as indication of link being down. It'll retry
sending the packet only after the device notifies it the link is once
again up. This is the reason why nothing happens when running
nic_router_flood on top of virtio_nic driver. The link never goes down
in this case so once we fill the TX VirtIO queue and tell the base class
to retry the send, we'll be stuck waiting for link up change event
which will never arrive.
To fix this problem, when sending a packet to the device fails, do a
synchrnonus TX VirtIO queue flush (tell device there is data to process
and wait until its done with it).
With this fix in place nic_router_flood test scenario passes on both arm
qemu_virt boards.
Issue #4264
This commit contains features and buf fixes:
* Catch errors during resource allocation
* Because Mesa tries to allocate fence (hardware) registers for each
batch buffer execution, do not allocate new fences for buffer objects
that are already fenced
* Add support for global hardware status page. Each context additionally
has a per-process hardware status page, which we used to set the
global hardware status page during Vgpu switch. This was obviously
wrong. There is only one global hardware status page (set once during
initialization) and a distinct per-process page for contexts.
* Write the sequence number of the currently executing batch buffer to
dword 52 of the per-process hardware status page. We use the pipe line
command with QW_WRITE (quad word write), GLOBAL_GTT_IVB disabled
(address space is per-process address space), and STORE_DATA_INDEX
enabled (write goes to offset of hardware status page). This command
used to write to the scratch page. But Linux now uses the first
reserved word of the per-process hardware status page.
* Add Gen9+ WaEnableGapsTsvCreditFix workaround. This sets the "GAPS TSV
Credit fix Enable" bit of the Arbiter control register (GARBCNTLREG)
as described by the documentation this bit should be set by the BIOS
but is not on most Gen9/9.5 platforms. Not setting this bit leads to
random GPU hangs.
* Increase the context size from 20 to 22 pages for Gen9. On Gen8 the
hardware context is 20 pages (1 hardware status page + 19 ring context
register pages). On Gen9 the size of the ring context registers has
increased by two pages to 21 pages or 81.3125 KBytes as the IGD
documentation states.
* The logical ring size in the ring buffer control of the execlist
context has to be programmed with number of pages - 1. So 0 is 1 page.
We programmed the actual number of pages before, leading to ring
buffer execution of NOOPs if page behind our ring buffer was empty or
GPU hangs if there was data on the page.
issue #4260
According to spec the tail pointer points to the next qword instructions
which will be used by the software.
p 1354, Doc Ref # IHD-OS-BDW-Vol 2c-11.15
issue #4254
