And start collecting qemu log earlier, so it isn't lost in case of a
timeout during domain startup.
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
It fails on larger initramfs (~250MB one) and sometimes even smaller
depending on memory size/memory map, let Linux do it.
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Bloat-o-meter reports:
add/remove: 0/0 grow/shrink: 0/2 up/down: 0/-29 (-29)
Function old new delta
_probe_mask_msr 99 94 -5
init_amd 2418 2394 -24
but this under-reports because .fixup doesn't contain sized/typed symbols.
This also drops two "mov -EFAULT, %reg; jmp ...;" sequences too, so the net
saving is -50.
wrmsr_amd_safe()'s return value is only checked against 0 (if at all), and
because of this, the compiler can now avoid manifesting the 0/-EFAULT
constants entirely, and the %[fault] label simply lands on the right basic
block.
Convert to Xen style while rewriting.
No functional change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
MISRA C:2012 Amendment 3 (which is under consideration, but not used by Xen
yet) is more particular about having the storage class specifier first.
No functional change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
The fixup section is only used by x86, and we're working to remove it there
too. Logic in the fixup section is unconnected to it's origin site, and
interferes with backtraces/etc.
Remove the section from the architectures which don't use it.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
Reviewed-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Acked-by: Shawn Anastasio <sanastasio@raptorengineering.com>
It makes no sense to request for preprocessor only output and also request
object file generation. Fix the _libxl.api-for-check target to only use
-E (preprocessor output).
Also Clang 20.0 reports an error if both options are used.
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com>
Fixes: 2862bf5b6c ('libxl: enforce prohibitions of internal callers')
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Anthony PERARD <anthony.perard@vates.tech>
The following sequence of events may lead a debug build of Xen to crash
when using the null scheduler:
1. domain creation (e.g. d1) failed due to bad configuration
2. complete_domain_destroy() was deferred
3. domain creation (e.g. d2) succeeds
At this point, d2 is running, while the zombie d1 is not fully cleaned
up:
(XEN) Online Cpus: 0-3
(XEN) Cpupool 0:
(XEN) Cpus: 0-3
(XEN) Scheduling granularity: cpu, 1 CPU per sched-resource
(XEN) Scheduler: null Scheduler (null)
(XEN) cpus_free = 3
(XEN) Domain info:
(XEN) Domain: 0
(XEN) 1: [0.0] pcpu=0
(XEN) 2: [0.1] pcpu=1
(XEN) Domain: 1
(XEN) 3: [1.0] pcpu=2
(XEN) Domain: 2
(XEN) 4: [2.0] pcpu=2
4. complete_domain_destroy() gets called for d1 and triggers the
following:
(XEN) Xen call trace:
(XEN) [<00000a0000322ed4>] null.c#unit_deassign+0x2d8/0xb70 (PC)
(XEN) [<00000a000032457c>] null.c#null_unit_remove+0x670/0xba8 (LR)
(XEN) [<00000a000032457c>] null.c#null_unit_remove+0x670/0xba8
(XEN) [<00000a0000336404>] sched_destroy_vcpu+0x354/0x8fc
(XEN) [<00000a0000227324>] domain.c#complete_domain_destroy+0x11c/0x49c
(XEN) [<00000a000029fbd0>] rcupdate.c#rcu_do_batch+0x94/0x3d0
(XEN) [<00000a00002a10c0>] rcupdate.c#__rcu_process_callbacks+0x160/0x5f4
(XEN) [<00000a00002a1e60>] rcupdate.c#rcu_process_callbacks+0xcc/0x1b0
(XEN) [<00000a00002a3460>] softirq.c#__do_softirq+0x1f4/0x3d8
(XEN) [<00000a00002a37c4>] do_softirq+0x14/0x1c
(XEN) [<00000a0000465260>] traps.c#check_for_pcpu_work+0x30/0xb8
(XEN) [<00000a000046bb08>] leave_hypervisor_to_guest+0x28/0x198
(XEN) [<00000a0000409c84>] entry.o#guest_sync_slowpath+0xac/0xd8
(XEN)
(XEN) ****************************************
(XEN) Panic on CPU 0:
(XEN) Assertion 'npc->unit == unit' failed at common/sched/null.c:383
(XEN) ****************************************
Fix by skipping unit_deassign() when the unit to be removed does not
match the pcpu's currently assigned unit.
Fixes: c2eae2614c ("sched/null: avoid crash after failed domU creation")
Signed-off-by: Stewart Hildebrand <stewart.hildebrand@amd.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
It gives the wrong impression of there being a type change, when
get_s_time() really returns s_time_t itself (kind of naturally given its
name).
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
We can now easily use __SIZEOF_LONG__ instead. For this to also work in
the test harness, move hvmloader's STR() to common-macros.h.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Anthony PERARD <anthony.perard@vates.tech>
The new toolchain baseline knows the VMX instructions; no need to carry the
workaround in the code.
Inline __vmxoff() into it's single caller.
Updated formatting in the wrappers to consistent.
No functional change.
Resolves: https://gitlab.com/xen-project/xen/-/work_items/202
Signed-off-by: Denis Mukhin <dmukhin@ford.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
There's no benefit in having process_shm_chosen() next to process_shm().
The former is just a helper to pass "/chosen" node to the latter for
hwdom case. Drop process_shm_chosen() and instead use process_shm()
passing NULL as node parameter, which will result in searching for and
using /chosen to find shm node (the DT full path search is done in
process_shm() to avoid expensive lookup if !CONFIG_STATIC_SHM). This
will simplify future handling of hw/control domain separation.
Signed-off-by: Michal Orzel <michal.orzel@amd.com>
Reviewed-by: Bertrand Marquis <bertrand.marquis@arm.com>
Static shared memory requires device-tree boot. At the moment, booting
with ACPI enabled and CONFIG_STATIC_SHM=y results in a data abort when
dereferencing node in process_shm() because dt_host is always NULL.
Fixes: 09c0a8976a ("xen/arm: enable statically shared memory on Dom0")
Signed-off-by: Michal Orzel <michal.orzel@amd.com>
Reviewed-by: Bertrand Marquis <bertrand.marquis@arm.com>
Putting a few bytes ahead of page tables isn't very efficient; there's
a gap almost worth a full page. To avoid re-ordering of items in the
source file, simply put the few small items in sub-section 1, for them
to end up after the page tables, followed (in the final binary) by non-
page-aligned items from other CUs.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Jason Andryuk <jason.andryuk@amd.com>
In particular if we're running virtualized, the underlying hypervisor
(which may be another Xen) may not surface MTRRs, and offer PAT only.
Fixes: 5a281883cd ("Hardcode many cpu features for x86/64 -- we know 64-bit")
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Roger Pau Monné <roger.pau@citrix.com>
Unlike stated in the offending commit's description,
load_system_tables() wasn't the only thing left to retain from the
earlier restore_rest_processor_state(). Note that MTRR state was still
reloaded via mtrr_aps_sync_end(), but that happens quite a bit later in
the resume process.
While there also do Misra-related tidying for the function itself: The
function being used from assembly only means it doesn't need to have a
declaration, but wants to be asmlinkage.
Fixes: 4304ff420e ("x86/S3: Drop {save,restore}_rest_processor_state() completely")
Reported-by: Roger Pau Monné <roger.pau@citrix.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Roger Pau Monné <roger.pau@citrix.com>
mtrr_set_all() has quite a bit of overhead, which is entirely useless
when set_mtrr_state() really does nothing. Furthermore, with
mtrr_state.def_type never initialized from hardware, post_set()'s
unconditional writing of the MSR means would leave us running in UC
mode after the sync.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Roger Pau Monné <roger.pau@citrix.com>
This is Intel i7-7567U in NUC 7i7BNH. This one is an older one, with no
firmware updates (last update from 2023) and no microcode udpates
either. While this firmware supports UEFI, network boot works only in
legacy mode - thus legacy is used here (via iPXE, instead of grub2.efi).
Testing legacy boot path may be a useful thing on its own.
Add the same set of tests as on ADL runner.
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Hardware runners that use legacy boot use iPXE instead of grub2. Create
boot.ipxe for those too - with exact same options.
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Evtchn fifos are not needed on smaller systems; the older interface is
lightweight and sufficient. Also, event_fifo causes runtime anonymous
memory allocations, which are undesirable. Additionally, it exposes an
extra interface to the guest, which is also undesirable unless
necessary.
Make it possible to disable evtchn fifo.
Signed-off-by: Stefano Stabellini <stefano.stabellini@amd.com>
Acked-by: Michal Orzel <michal.orzel@amd.com>
With the new toolchain baseline, we can make use of asm goto() in certain
places, and the VMXON invocation is one example.
This removes the logic to set up rc (including a fixup section where bactraces
have no connection to the invoking function), the logic to decode it,
including the default case which was dead but not visibly-so to the compiler.
No functional change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
This logic was inserted by commit 447f613c54 ("lzo: update LZO compression
to current upstream version") but was only relevant for the TMEM logic, so
should have been deleted in commit c492e19fdd ("xen: remove tmem from
hypervisor")
Fixes: c492e19fdd ("xen: remove tmem from hypervisor")
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
As reported by ECLAIR scan, MISRA requires declaration to be visible
(R8.4). This is not the case for ioremap().
Fixes: 2cd02c27d3 ("arm/mpu: Implement stubs for ioremap_attr on MPU")
Signed-off-by: Michal Orzel <michal.orzel@amd.com>
Reviewed-by: Bertrand Marquis <bertrand.marquis@arm.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
The little endian implementation of bitmap_to_xenctl_bitmap leads to
unnecessary xmallocs and xfrees. Given that Xen only supports little
endian architectures, it is worth optimizing.
This patch removes the need for the xmalloc on little endian
architectures.
Remove clamp_last_byte as it is only called once and only needs to
modify one byte. Inline it.
Signed-off-by: Stefano Stabellini <stefano.stabellini@amd.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
This commit introduces the skeleton for the MPU memory management
subsystem that allows the compilation on Arm64.
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
ARM MPU system doesn't need to use paging memory pool, as MPU memory
mapping table at most takes only one 4KB page, which is enough to
manage the maximum 255 MPU memory regions, for all EL2 stage 1
translation and EL1 stage 2 translation.
Introduce ARCH_PAGING_MEMPOOL Kconfig common symbol, selected for Arm
MMU systems and x86. Removed stubs from RISC-V now that the common code
provide them and the functions are not gonna be used.
Wrap the code inside 'construct_domU' that deal with p2m paging
allocation in a new function 'domain_p2m_set_allocation', protected
by ARCH_PAGING_MEMPOOL, this is done in this way to prevent polluting
the former function with #ifdefs and improve readability
Introduce arch_{get,set}_paging_mempool_size stubs for architecture
with !ARCH_PAGING_MEMPOOL.
Remove 'struct paging_domain' from Arm 'struct arch_domain' when the
field is not required.
Signed-off-by: Penny Zheng <penny.zheng@arm.com>
Signed-off-by: Wei Chen <wei.chen@arm.com>
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com> # arm
Reviewed-by: Oleksii Kurochko <oleksii.kurochko@gmail.com> # riscv
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Acked-by: Jan Beulich <jbeulich@suse.com>
Implement ioremap_attr() stub for MPU system; the
implementation of ioremap() is the same between MMU
and MPU system, and it relies on ioremap_attr(), so
move the definition from mmu/pt.c to arm/mm.c.
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
The MPU system requires static memory to work, select that
when building this memory management subsystem.
While there, provide a restriction for the ARM_EFI Kconfig
parameter to be built only when !MPU, the EFI stub is not
used as there are no implementation of UEFI services for
armv8-r.
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
Introduce frame_table in order to provide the implementation of
virt_to_page for MPU system, move the MMU variant in mmu/mm.h.
Introduce FRAMETABLE_NR that is required for 'pdx_group_valid' in
pdx.c, but leave the initialisation of the frame table to a later
stage.
Define FRAMETABLE_SIZE for MPU to support up to 1TB of ram at this
stage, as the only current implementation of armv8-r aarch64, which
is cortex R82, can support 1TB or 256TB (r82 TRM r3p1
ID_AA64MMFR0_EL1.PARange).
Take the occasion to sort alphabetically the headers following
the Xen code style and add the emacs footer in mpu/mm.c.
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
virt_to_maddr and maddr_to_virt are used widely in Xen code. So
even there is no VMSA in MPU system, we keep the interface in MPU to
to avoid changing the existing common code.
In order to do that, move the virt_to_maddr() and maddr_to_virt()
definitions to mmu/mm.h, move the include of memory management
subsystems (MMU/MPU) on a different place because the mentioned
helpers needs visibility of some macro in asm/mm.h.
Finally implement virt_to_maddr() and maddr_to_virt() for MPU systems
under mpu/mm.h, the MPU version of virt/maddr conversion is simple since
VA==PA.
Signed-off-by: Penny Zheng <penny.zheng@arm.com>
Signed-off-by: Wei Chen <wei.chen@arm.com>
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
The define HYPERVISOR_VIRT_START is required by the common code,
even if MPU system doesn't use virtual memory, define it in
mpu/layout.h in order to reuse existing code.
Disable a check in the linker script for arm for !MMU systems.
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
When purging old-style annotations, MPU code was left unmodified. Fix
it.
Fixes: 221c66f4f2 ("Arm: purge ENTRY(), ENDPROC(), and ALIGN")
Signed-off-by: Michal Orzel <michal.orzel@amd.com>
Acked-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Adding Q suffixes to FXSAVE/FXRSTOR did break the 32-bit build. Don't go
back though, as the hand-coded 0x48 there weren't quite right either for
the 32-bit case (they might well cause confusion when looking at the
disassembly). Instead arrange for the compiler to DCE respective asm()-s,
by short-circuiting REX_* to zero.
Fixes: 5a33ea2800 ("x86emul: drop open-coding of REX.W prefixes")
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Use infrastructure from xen/linkage.h instead of the custom legacy
macros that we're in the process of phasing out.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Currently, only the device tree method is available to locate and perform
pre-initialization steps for the interrupt controller (at the moment, only
one interrupt controller is going to be supported). When `acpi_disabled`
is true, the system will scan for a node with the "interrupt-controller"
property and then call `device_init()` to validate if it is an expected
interrupt controller and if yes then save this node for further usage.
If `acpi_disabled` is false, the system will panic, as ACPI support is not
yet implemented for RISC-V.
Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Introduce preinitialization stuff for the RISC-V Advanced Platform-Level
Interrupt Controller (APLIC) in Xen:
- Implementing the APLIC pre-initialization function (`aplic_preinit()`),
ensuring that only one APLIC instance is supported in S mode.
- Initialize APLIC's correspoinding DT node.
- Declaring the DT device match table for APLIC.
- Setting `aplic_info.hw_version` during its declaration.
- Declaring an APLIC device.
Since Microchip originally developed aplic.c [1], an internal discussion
with them led to the decision to use the MIT license instead of the default
GPL-2.0-only.
[1] https://gitlab.com/xen-project/people/olkur/xen/-/commit/7cfb4bd4748ca268142497ac5c327d2766fb342d
Co-developed-by: Romain Caritey <Romain.Caritey@microchip.com>
Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
By default, the `aia` option is set to "none" which selects the SiFive PLIC for
handling wired interrupts. However, since PLIC is now considered obsolete and
will not be supported by Xen now, APLIC and IMSIC are selected instead to manage
both wired interrupts and MSIs.
Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
preinit_xen_time() does two things:
1. Parse timebase-frequency properpy of /cpus node to initialize cpu_khz
variable.
2. Initialize boot_clock_cycles with the current time counter value to
have starting point for Xen.
timebase-frequency is read as a uint32_t because it is unlikely that the
timer will run at more than 4 GHz. If timebase-frequency exceeds 4 GHz,
a panic() is triggered, since dt_property_read_u32() will return 0 if
the size of the timebase-frequency property is greater than the size of
the output variable.
Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
They're no longer used. This also makes it unnecessary to #undef two of
them in the linker script.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Julien Grall <jgrall@amazon.com>
Tested-by: Luca Fancellu <luca.fancellu@arm.com> # arm
Marek Marczykowski-Górecki