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
Superpages (2M, 1G) are not supported by now, but partially copied over code
from base-hw was around. Remove unused register definitions and remove
non-working super page code do avoid confusion.
issue #4254
Size argument of ggtt free range check is ignored, which leads to
overlapping allocations inside, which leads to unavailable IO-MEM exceptions
thrown by core.
issue #4254
The platform services is intented to be used by dde_linux's intel_fb_drv
in order to initlialize displays.
* implement and announce platform session
* limit accessible GTT and aperture of client to 64 MB
* forward display engine IRQs to platform client
* move all PCI resources to 'Igd::Resources' class in order to make them
accessible by the platform service and the GPU driver
* fix fence register allocation for id zero (return true)
issue #4233
For mesa-21 the client takes care and manages
the virtual address space of the vGPU by itself and the intel/gpu driver
can't add silently a guard page anymore. Move the patch to the drm/ioctl
of the former mesa version.
Issue #4148#4233
_unmap_dataspace_ggtt requires the cap of Ggtt::Mapping (ring_map, ctx_map)
in order to find the right metadata and to free up the ggtt entries. Also the
pte range is removed already if the metadata was found.
Issue #4148#4233
A reset domain can consist of one or several reset-pins
denoted by name that are assigned to a device.
When the device gets acquired via the Platform RPC API,
the pins are de-asserted, and asserted again when the
device gets released.
A configuration looks like the following:
<device name="mipi_dsi>
<reset-domain name="mipi_dsi_pclk"/>
...
</device>
Fixes#4171
Introduces the notion of a transaction that consists of one or more
messages. Whereby a message has a read or write direction and consists
of one or more bytes.
Issue #4170Fixes#4169
Report via platform_info the capabilities of the kernel, e.g. ACPI and MSI.
With the commit the try-catch pattern on IRQ session creation by the platform
driver is avoided.
Issue #4016
* The device XML information dataspace is only provided,
when the client's policy states `info="yes"`
* The device XM information gets changed to include the
physical resource names (I/O memory and IRQ addresses)
instead of virtual ids and page offset
Fix#4077
This API rework eases the access to memory-mapped I/O registers and
interrupts when using the platform driver. It introduces the notions of
- Platform::Device - one device obtained from a platform session
- Platform::Device::Mmio - locally-mapped MMIO registers of a device
- Platform::Device::Irq - interface for receiving device interrupts
The patch touches several drivers. Some drivers would require a
significant structural change to adopt the new API (e.g., net/virtio,
dde_linux drivers, imx gpio). In these cases, the patch adds
compatibility shims meant to be temporary. In other cases (e.g., imx
i2c), the adaptation was simple enough to carry through.
Fixes#4075
This patch extends the 'Platform_session::alloc_dma_buffer' interface
with a 'Cache' argument that corresponds to the argument accepted by
'Ram_allocator::alloc', which is used by the platform driver under the
hood.
Since the x86 platform driver used to be hardwired to allocate DMA
buffers as UNCACHED, I adjusted all drivers by specifying the UNCACHED
argument. Right now, this is needed as a hint for core to steer the
allocation of I/O page tables. Once we eliminate the need for such hints
(by introducing an explicit 'Region_map::attach_dma' operation), we can
revisit the drivers individually because cached DMA buffers should
generally be fine on the x86 architecture.
Issue #2243
This change avoids many repetetive Genode:: prefixes, making the code
easier to read. The patch also includes a few consistency fixes
regarding include guards and file headers. It also renames
Platform_device::String to Platform_device::Device:name.
Issue #2243
