The USB S3 virtualization logic is complicated and heavily affected by system
timing. To make the debugging work easily, change related log level to facilitate
quick bug fixing.
Tracked-On: #1893
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
irqfd only supports msix devices. In the current code a mevent is
added to poll the callfd from userspace to support intx devices.
This patch removes the support for non-msix devices since they are
not used in the current device model.
Tracked-On: #1877
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
Teardown callback is provided when mevent_add is called and it is
used to free the virtio-net resources.
Tracked-On: #1877
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
Acked-by: Yin Fengwei <fengwei.yin@intel.com>
Teardown callback is provided when mevent_add is called. A ref_count
is added to virtio_console data structure. Teardown callback needs to
free the resources of the backend and when the ref_count of the
virtio_console is zero the whole virtio-console is freed.
Tracked-On: #1877
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
Acked-by: Yin Fengwei <fengwei.yin@intel.com>
Teardown callback is provided when mevent_add is called and it is
used to free the virtio-input resources.
Tracked-On: #1877
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
Acked-by: Yin Fengwei <fengwei.yin@intel.com>
On Intel Apllo Lake platform, the VBus will drop after the SOC suspended,
hence during the SOS resuming process, there will be a disconnecting event
and a connecting event sent to each native USB device in order.
This patch will use new strategy for S3 resuming emultion.
1. The DM set PORTSC register to 'no device attached' state when S3
suspending started,
2. SOS resuming starts and do nothing for device disconnecting event,
3. 'Cache' device connecting event and don't report it to UOS,
4. UOS believe no device attached due to PORTSC register state and
begin to clear the resource allocated for the device before S3
suspending,
5. DM receives the Disable Slot command from UOS and report the 'cached'
device connecting event to UOS, hence trigger the emulation behavior
for the device.
The purpose of this strategy is to let UOS resuming proceed as quickly as
possible, which means the UI will be turned on quickly to user.
Tracked-On: #1893
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
According to xHCI spec 5.4.1.1, this command hasn't ask to clear
PORTSC for R/S handling. So the PORTSC should always present the
physical usb bus status.
The old implementation changes the PORTSC according whether the
pci_xhci_dev_emu struct is allocated, it is not consistent with
spec and this patch is used to fix it.
Tracked-On: #1893
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
Comparing two USB devices' path is frequently used operation, abstract
it as an seperated function for convenience.
Tracked-On: #1893
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
This workaround will delay the UOS resuming for 5 seconds, this
behavior is not right and should be replaced by dynamic style.
Tracked-On: #1893
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
Always enforce data size when using BAR access functions.
Currently, the size serves as a hint to the BAR access functions and
these functions are expected to behave accordingly. Some of the access
functions, e.g. virtio ones, don't always truncate the data but expect
the caller to take care of the data size. This causes problems with
OVMF's virtio drivers during I/O instruction emulation because RAX can
contain junk bits that shouldn't be written to the device.
v1 -> v2:
- improve readability
Tracked-On: #1935
Signed-off-by: Peter Fang <peter.fang@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
Physical NIC mac address is used for generate UOS mac address.
This patch uses a new parameters to pass this information
instead of vm name.
Tracked-On: #1987
Signed-off-by: Jie Deng <jie.deng@intel.com>
Acked-by: Yan, Like <like.yan@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
EV_MSC is configured as INPUT_PASS_TO_ALL in input driver. There is
a loop in the use case of virtio-input. They should not be forwarded
back to BE since they have already been sent to native driver before
sending to FE.
Tracked-On: #2006
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
After mutiple usb devices send enable slot commnad and get slot id each
other, address device setup flow is not sorted by slot id. According to
current design, it will casue assert failure. This patch takes off this
restriction.
Tracked-On: #2017
Signed-off-by: Yang Liang <liang3.yang@intel.com>
Reviewed-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
Pass teardown callback when add mevent in mei mediator code.
Which could avoid run_callback calling after the related data
structure is freed.
Tracked-On: #1877
Signed-off-by: Yin Fengwei <fengwei.yin@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
There is one race issue between mevent callback (which is called
in mevent_dispatch thread) and mevent_delete (which could be called
in dev thread). And the callback is called after mevent_delete.
libevent have the exactly same issue. The issue is decripted here:
https://github.com/libevent/libevent/blob/master/whatsnew-2.1.txt
The fixing is:
We introduce a teardown callback to mevent and make sure there is
no race issue between callback and teardown call.
This patch updates the mevent API and the caller as well.
Tracked-On: #1877
Signed-off-by: Yin Fengwei <fengwei.yin@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
If the USB kernel drives and devices of SOS are not available and
functional ready after UOS is resumed from frozen state, the comming
USB command from UOS will cause a disaster because the DM has no
resourse to to emulation.
This patch add a dynamic time span (could change according to number
of native USB devices attached) to give USB drivers and devices of
SOS an opportunity to be ready.
Tracked-On: #1893
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
During S3 process, the dev in the function pci_xhci_device_doorbell
may be set to NULL when disconnection is detected by DM. This patch
is used to remove this risk.
Tracked-On: #1893
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
The 'slot_allocated[]' field of pci_xhci_vdev is used as flag of
slot allocation. In current design, this variable is not set to
false in disconnection callback 'pci_xhci_native_usb_dev_disconn_cb'.
This patch is used to fix it.
Tracked-On: #1894
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
In the old implementation, function vm_get_suspend_mode is used
as the interface to check whether the UOS is in the S3, That is
not saft way.
This patch use Port Link State (PLS) to substitute the old logic.
According to xHCI spec 5.4.8, the PLS should be U3 during the
UOS is in the S3 state.
Tracked-On: #1893
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
According to xHCI 5.4.8, the Port Link State Change (PLC) bit
should not be set to 1 during suspend process. This patch is used
to fix it.
Tracked-On: #1893
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
Under current implementation, when USB control transfer failed
with Endpoint Stall error, there is no response reporting to UOS.
This logic will result of timeout in UOS kernel and hence a longer
enumeration process.
This patch reports the Stall Error to UOS by xHCI Completion Event,
which will fix this issue.
Tracked-On: #1895
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
Device trap has great impact on latency of real time (RT) tasks.
This patch provide a virtio poll mode to avoid trap.
According to the virtio spec, backend devices can declare the
notification is not needed so that frontend will never trap.
This means the backends make commitment to the frontends they have a
poll mechanism which don’t need any frontends notification.
This patch uses a periodic timer to give backends pseudo notifications
so that drive them processing data in their virtqueues. People should
choose a appropriate notification peroid interval to use this poll
mode. Too big interval may cause virtqueue processing latency while
too small interval may cause high SOS CPU usage. The suggested interval
is between 100us to 1ms.
The poll mode is not enabled by default and traditional trap
notification mode will be used. To use poll mode for RT with interval
1ms. You can add following acrn-dm parameter.
--virtio_poll 1000000
Tracked-On: #1956
Signed-off-by: Jie Deng <jie.deng@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
It is preferred to state the absence of a return value explicitly in the
doxygen-stile comments. Currently there are different styles of doing this,
including:
@return None
@return NULL
@return void
@return N/A
This patch unifies the above with `@return None`.
Tracked-On: #1595
Signed-off-by: Junjie Mao <junjie.mao@intel.com>
v4 -> v5
Replace strcpy with strncpy
Refine resource free if error happens
Remove meaningless *in_progress* label for pthread wait/signal
Rename the coreu thread routine name (virtio_coreu_thread)
v3 -> v4
Move setsocketopt to connect_to_daemon
Cleanup coreu file descriptor after close
Reconnect CoreU daemon if the socket descriptor is invalid
v2 -> v3
Move the daemon connection to vdev_init
Diagram the CoreU virtualization architecture
Create a seperate thread for sending and receving the CoreU message
Change the socket name
v1 -> v2
No change
v1
Initial CoreU back-end driver
Signed-off-by: Yu Shiqiang <shiqiang.yu@intel.com>
Simple atomic add/dec do no guarantee reference count full
synchronization without a lock. Compare and swap operations
are required for correct implementation.
Tracked-On: #1875
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Acked-by: Yin Fengwei <fengwei.yin@intel.com>
Fix the prefix comparison, we need to compare on prefix length, not
on devpath len, otherwise we always fail.
Tracked-On: #1848
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Acked-by: Acked-by: Yu Wang <yu1.wang@intel.com>
Edge triggered mevent on sysfs file is triggered
immediately after mevent_add(), hence this has to be ignored.
The issue that is resolved that the 'first' flag was
global and wasn't reset on each mevent_add().
Tracked-On: #1846
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Acked-by: Acked-by: Yu Wang <yu1.wang@intel.com>
(EV_MSC, MSC_TIMESTAMP) is added to each frame just before the
SYN event since kernel 4.15. EV_MSC is configured as
INPUT_PASS_TO_ALL. In the use case of virtio-input, there is
a loop as follows:
- A mt frame with (EV_MSC, MSC_TIMESTAMP) is passed to FE.
- FE will call virtinput_status to pass (EV_MSC, MSC_TIMESTAMP)
back to BE.
- BE writes this event to evdev. Because (EV_MSC, MSC_TIMESTAMP)
is configured as INPUT_PASS_TO_ALL, it will be written into
the event buffer of evdev then be read out by BE without
SYN followed.
- Each mt frame will introduce one (EV_MSC, MSC_TIMESTAMP).
Later the frame becomes larger and larger...
This patch fixed above issue by ignoring MSC_TIMESTAMP from guest.
Besides that timestamp is added for every status event from guest
before writing to evdev.
Tracked-On: #1670
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
The FW reset is currently detected from two points upon
a read failure from native read and from the reset
handler.
The fix removes the detection from the mevent rx callback,
leaving a single detection point.
To prevent reset hiccup, hw_ready is not set if a full rescan is
performed, it will be set only when virtio FW will request the FW reset.
Tracked-On: #1632
Signed-off-by: Aviad Nissel <aviad.nissel@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Simplify the flow by adding mutex_type variable
and call pthread_mutexattr_destroy() on the error path.
Tracked-On: #1630
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
- remove ASSERT & DEASSET IRQ line IOCTLs
- remove PULSE IRQ line IOCTLs, use set/clear
IRQ line instead.
- Use IC_SET_IRQLINE to set or clear IRQ line
Tracked-On: #861
Signed-off-by: Yonghua Huang <yonghua.huang@intel.com>
Reviewed-by: Jason Chen CJ <jason.cj.chen@intel.com>
Reviewed-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
Virtio-console device specific configuration is readonly, callback
for the cfgwrite is not required. This patch removed the unused
virtio_console_cfgwrite.
Tracked-On: #1364
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
Remove the changing operation for 'xdev->native_ports[].state' in
pci_xhci_cmd_disable_slot. The operation should not be done in this
funciton.
Tracked-On: #1589
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
On certain SOC, during the system suspend and resume process, connecting
and disconnecting events may happen. In previous implementation, DM didn't
clear PED bit in the xHCI PORTSC register, this will induce many invalid
polling operations in UOS and fail to enumerate one or more USB devices.
Tracked-On: #1589
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
The function for port unassigning: pci_xhci_clr_native_port_assigned
should reset the 'info' member to all zero when it is called.
Tracked-On: #1434
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
This issue is result from reporting interrupt when the interrupt ability
of virtual xHCI is disabled.
As Event Ring State Machine showed in Figure 20 of xHCI spec, following
rules should be followed:
1 when RS bit of USBCMD register is zero, Event Ring should not be accessed;
2 when INTE bit of USBCMD register is zero, Interrupt should not be sent.
Tracked-On: #1566
Signed-off-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liang Yang <liang3.yang@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
When usb device numbers reach up to XHCI_MAX_SLOTS. The slot_allocated
array will get out of range. This patch is used to fix this issue.
Tracked-On: #1479
Signed-off-by: Liang Yang <liang3.yang@intel.com>
Reviewed-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
After excap pointer is assigned, it should be checked whether it's
possible to get assignment for NULL pointer or not. This patch
fixes this issue.
Tracked-On: #1479
Signed-off-by: Liang Yang <liang3.yang@intel.com>
Reviewed-by: Xiaoguang Wu <xiaoguang.wu@intel.com>
Reviewed-by: Liu Shuo <shuo.a.liu@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
sscanf and strcpy are banned according to the security requirements.
So replace them with their safe alternative.
Tracked-on: #1496
Signed-off-by: Jie Deng <jie.deng@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
