This patch add one new parameter --rtvm to indicate if the guest is a RTVM or not.
For RTVM, it may be not interference by SOS.
Tracked-On: #2865
Signed-off-by: Kaige Fu <kaige.fu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
This flag indicates that if the vm is RTVM.
And if a vm has GUEST_FLAG_LAPIC_PASSTHROUGH flag set,
we must set the GUEST_FLAG_RT at the same time.
Tracked-On: #2865
Signed-off-by: Kaige Fu <kaige.fu@intel.com>
Length of vmname is restricted to MAX_VM_OS_NAME_LEN. Keep it
coincide with VM name array size in vm_config.
Tracked-On: #2851
Signed-off-by: Tao Yuhong <yuhong.tao@intel.com>
Acked-by: Yin Fengwei <fengwei.yin@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
devicemodel/include/public/acrn_common.h should be identical with
hypervisor/include/public/acrn_common.h, so we can use a soft link
to hypervisor's acrn_common.h for devicemodel.
Tracked-On: #2851
Signed-off-by: Tao Yuhong <yuhong.tao@intel.com>
smbios table is not required so remove it.
Tracked-On: #2577
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
Guest OS (e.g. Linux) may rely on a gap in E820 map in the 32-bit memory
space to determine the MMIO space for its PCI devices. Leave this gap
when building E820 map to keep the guest's PCI subsystem working.
After commit 7752d5cfe3d11ca0bb9c673ec38bd78ba6578f8e, Linux kernel no
longer requires the MMCONFIG region to be reserved in the E820 map.
Nonetheless, keep it in the reserved region to be on the safe side.
Tracked-On: #2843
Signed-off-by: Peter Fang <peter.fang@intel.com>
Acked-by: Yin Fengwei <fengwei.yin@intel.com>
OVMF requires a more descriptive mechanism than RTC CMOS to retrieve
ACRN's memory layout, so we now pass the E820 map to it, starting at
0xEF000 (ROM area).
ACRN currently uses [4GB, 5GB) as its 64-bit PCI host aperture. This is
inconsistent with OVMF's assumption of its platform's memory layout,
because it derives the size of high memory from RTC CMOS, which is
incapable of describing the 64-bit PCI hole.
By default, OVMF uses RTC CMOS 0x5b/0x5c/0x5d to determine the size of
high memory. This value only tells OVMF how much memory is above 4GB,
but not the platform's memory layout above 4GB.
Using RTC CMOS works for QEMU, because QEMU places its 64-bit PCI host
aperture above its highmem. Therefore, OVMF can always assume highmem is
located at [4GB, 4GB + highmem), which is not where ACRN's highmem is
located. For example, if we have 1GB of usable memory above 4GB, ACRN
will place it at [5GB, 6GB).
This change allows OVMF to correctly identify the guest's memory layout.
It will consider any reserved region above 4GB as 64-bit PCI host
aperture.
MP table, SMBIOS and ACPI tables are all located above 0xF0000 so it is
guaranteed that there is no overlap. There can only be a maximum of 128
E820 entries.
v1 -> v2:
- provide more explanation to this commit
- add signature before E820 map for OVMF backward compatibility
Tracked-On: #2792
Signed-off-by: Peter Fang <peter.fang@intel.com>
Reviewed-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Yin Fengwei <fengwei.yin@intel.com>
While building the E820 map for UOS, [lowmem, lowmem_limit) and [5GB,
highmem) can be empty. Remove the empty entries if they appear.
Tracked-On: #2792
Signed-off-by: Peter Fang <peter.fang@intel.com>
Reviewed-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Yin Fengwei <fengwei.yin@intel.com>
- fix comments
- update the first RAM region to [0, 0xA0000) because [0xA0000, 1MB) is
designated as video memory and ROM area
- use 2GB as lowmem_limit
Tracked-On: #2792
Signed-off-by: Peter Fang <peter.fang@intel.com>
Reviewed-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Yin Fengwei <fengwei.yin@intel.com>
- use strncmp() instead of comparing string pointers to make no
assumptions about the toolchain's literal pool
- re-shuffle the functions so they're consistent with mem.h
- make non-public functions static
- increase code re-use
Tracked-On: #2792
Signed-off-by: Peter Fang <peter.fang@intel.com>
Reviewed-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Yin Fengwei <fengwei.yin@intel.com>
Physcial bdf is needed in hypervisor to find the right pci dev.
Tracked-On: #2788
Signed-off-by: Binbin Wu <binbin.wu@intel.com>
Reviewed-by: Jason Chen CJ <jason.cj.chen@intel.com>
There was a workaround in DM that allocates PCI 64bit mem bar in 32bit mem space
if the bar size is within 32MB.
After the workaround being removed, there is an issue to enter fastboot
mode for inappropriate handling of 64bit mem bar in guest driver.
The patch bring the workaround back, and skip the workaround when the guest
is booted by OVMF.
Revert the patch after the guest fixs the issue of handling 64bit mem bar in
fastboot mode.
Tracked-On: #2677
Signed-off-by: Binbin Wu <binbin.wu@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
1G PCI hole is added just after 4G address which is used as the
PCI high MMIO address space. Guest high memory is mapped from 5G
address for both EPT and device model user space address. Guest
e820 table and API vm_map_gpa are updated accordingly.
Tracked-On: #2577
Signed-off-by: Jian Jun Chen <jian.jun.chen@intel.com>
Reviewed-by: Jason Chen CJ <jason.cj.chen@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
New option "--lapic_pt" added to create VM with local apic passthrough, for
realtime scenarios.
When the option is set, a VM is created with LAPIC_PASSTHROUGH.
The option is not set by default.
Tracked-On: #2351
Signed-off-by: Yan, Like <like.yan@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
- don't generate io apic entry for vm with lapic pt.
Tracked-On: #2351
Signed-off-by: Yan, Like <like.yan@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
vHPET is used as a source of system timer by UEFI (e.g. OVMF).
This provides an alternative to using vPIT, which OVMF assumes is always
connected to vPIC.
This is ported from Bhyve, with a few changes:
- move to user space, using acrn_timer
- enable timers only when necessary
Origin: FreeBSD
License: BSD-3-Clause
URL: https://svnweb.freebsd.org/
commit: 326257
Purpose: Adding vHPET support.
Maintained-by: External
Tracked-On: #2319
Signed-off-by: Peter Fang <peter.fang@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
It is possible for multiple timeouts to occur in one mevent epoll
iteration. Providing the number of timer expirations to the timer
callback handlers can be useful. E.g., this could improve emulation of
timing-sensitive hardware components.
Tracked-On: #2319
Signed-off-by: Peter Fang <peter.fang@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
Allow specifying the initial expiration in absolute time.
Tracked-On: #2319
Signed-off-by: Peter Fang <peter.fang@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
It was designed to be used in non-blocking mode to prevent the mevent
thread from blocking itself indefinitely, but it was created in
blocking mode.
Tracked-On: #2319
Signed-off-by: Peter Fang <peter.fang@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
Currently VHM driver is opened by using the device file of /dev/acrn_vhm.
But for the upstream purpose it is renamed to /dev/acrn_hsm.
So we need to check that the device file is acrn_vhm or acrn_hsm and then
open the VHM driver.
Tracked-On: projectacrn/acrn-hypervisor#2356
Acked-by: Anthony Xu <anthony.xu@intel.com>
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
We will mark ioreq as COMPLETE in VHM via IC_NOTIFY_REQUEST_FINISH
ioctl. vm_notify_request_done can do it. So remove this one.
Tracked-On: #1821
Signed-off-by: Shuo Liu <shuo.a.liu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
VHM will provide a ioctl to clear all IO requests' status. This is
useful to handle ioreqs in VM normal reboot and emergency reboot.
Tracked-On: #1821
Signed-off-by: Shuo Liu <shuo.a.liu@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
Also remove the '\n' from the hugetlb's file name.
Tracked-On: #2133
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Reviewed-by: Yonghua Huang <yonghua.huang@intel.com>
The patch fix some string operations issues and also improve readability
of several snippet.
Tracked-On: #2133
Signed-off-by: Shuo A Liu <shuo.a.liu@intel.com>
Reviewed-by: Yonghua Huang <yonghua.huang@intel.com>
Now one macro is added to define the alignment requirement.
>#define __aligned(x) __attribute__((aligned(x)))
Some code uses the __aligned(x) to define the alignment while the other
code uses the original alignment definition.
So they are unified.
Tracked-On: projectacrn/acrn-hypervisor#2131
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
When issue happen, we could identify which thread is impacted.
This could help stability issue debugging.
Tracked-On: #2037
Signed-off-by: Yin Fengwei <fengwei.yin@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>
Use '--ovmf <OVMF image location>' when launching acrn-dm.
Tracked-On: #1832
Signed-off-by: Peter Fang <peter.fang@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
Generic infrastructure for loading BIOS/ROM and providing EPT pages at
High BIOS region.
The size of High BIOS is rounded up to a multiple of 2MB.
v2 -> v3:
- refine mmap_hugetlbfs* to reduce code replication
v1 -> v2:
- make this code generic instead of OVMF-specific
Tracked-On: #1832
Signed-off-by: Peter Fang <peter.fang@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
This is preparation for the upcoming High BIOS work.
Tracked-On: #1832
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>
To fix the issue that watchdog reset hang issue. At previous change,
we assume the reset comes from guest. But watchdog reset or dm killed
by signal is different. It's possible the vm_loop stick on ioreq
attaching.
The new fixing has two parts:
- fixing from kernel side to remove the race issue in ioreq attach,
req_complete_notify and client destory.
- Move the client destroy before waiting of vm_loop exit.
Tracked-On: #1986
Signed-off-by: Yin Fengwei <fengwei.yin@intel.com>
Reviewed-by: Liu Shuo <shuo.a.liu@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>
Peter, Thomas and Shuo raised one race issue in mevent_del. It
happens like following:
Thread mevent_dispatch Thread
mevent_delete_event
epoll_ctl_del
free(evp)
mevent_handle with freed evp
The fixing is adding sync between mevent_delete_event and
mevent_handle in mevent_dispatch.
Thread mevent_dispatch Thread
mevent_delete_event
add evp to del_list
notify mevent_dispatch
return
mevent_handle
Remove evp from del_list
Remove evp from epoll_fd
closefd()
free(evp)
Tracked-On: #1877
Signed-off-by: Yin Fengwei <fengwei.yin@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
The tpm patch delete the "break" for CMD_OPT_DEBUGEXIT branch.
Tracked-On: #1978
Signed-off-by: Yin Fengwei <fengwei.yin@intel.com>
Acked-by: Eddie Dong <eddie.dong@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>
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>
When thread_cancel() is called to exit pthread, we can't know
when the thread exit done unless pthread_join() is used to
wait for thread exit.
Tracked-On: #1868
Signed-off-by: Yin Fengwei <fengwei.yin@intel.com>
Acked-by: Anthony Xu <anthony.xu@intel.com>
vcpu thread need to exit when doing VM full reset and VM shutdown.
Change the code to align the logic. This also can fix race condition
between ioreq client destruction and accessing.
Tracked-On: #1857
Signed-off-by: Fengwei Yin <fengwei.yin@intel.com>
Signed-off-by: Shuo Liu <shuo.a.liu@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
this patch fixes such an information leaking issue:
in case that after a UOS is destroyed, its memroy
will be reclaimed and maybe re-allocated for a new UOS,
then the previous UOS sensitive data in memory may
be leaked to the new UOS.
Tracked-On: #1825
Signed-off-by: Yonghua Huang <yonghua.huang@intel.com>
Signed-off-by: Tianhua Sun <tianhuax.s.sun@intel.com>
Acked-by: Eddie Dong <eddie.dong@intel.com>
modify and add interface to set interrupt monitor params;
it can be set by acrn-dm input arguments like following:
--intr_monitor 10000,10,1,100
the 4 params order is:
interrupt storm theshold per second,
probe period to get interrupt data,
pass-through devices' interrupt injection delay time,
delay duration, after it, stop injection & restore to normal
Tracked-On: #1724
Signed-off-by: Minggui Cao <minggui.cao@intel.com>
Acked-by: Yin Fengwei <fengwei.yin@intel.com>
paddr_guest2host has same function with dm_gpa2hva. And There is no
usage of dm_gpa2hva. Remove it.
Tracked-On: #1595
Signed-off-by: Shuo Liu <shuo.a.liu@intel.com>
Acked-by: Fengwei Yin <fengwei.yin@intel.com>
Acked-by: Yu Wang <yu1.wang@intel.com>
For CS of UOS, we would like to pass all related info (cs attribute,
limit, base) from DM.
Tracked-On: #1231
Signed-off-by: Yin Fengwei <fengwei.yin@intel.com>
Acked-by: Eddie Dong <Eddie.dong@intel.com>
Kaige Fu