For XPTI enabled VMs, restore_all_guest() copies most of the active L4
pagetable into the per-CPU root_pgt. The L4 is read via the
directmap.
Unfortunately, there was no accounting for the fact that PV guests
have short directmaps. For an L4 living beyond the 5T boundary (3.5T
for CONFIG_BIGMEM), Xen will copy arbitrary data from guest-controlled
virtual address space into the per-CPU root_pgt, instead of the
intended L4.
Xen has per-domain mappings for the LDT and GDT, with two pages
reserved for Xen's use in the GDT but only one in use. Reuse GDT page
15 to map the effective L4. The GDT limit keeps implicit accesses out
of this page.
We update this mapping in make_cr3(), the only place a non-empty value
of v->arch.cr3 is written into a guest's vcpu struct, for two reasons.
First is reference counting: this makes it easier to argue that the
reference for v->arch.cr3 itself covers the reference for the mapping.
Secondly, this guarantees that all updates, even updates made in
shadow mode, result in an update to the mapping. During live
migration, it's possible for guest_get_eff_kern_l1e() to result in a
modified value of v->arch.cr3 (sketch below); placing the mapping
update at make_cr3() catches this, while other candidate locations
like write_ptbase() and toggle_guest_mode() don't.
Note that we introduce the idea of an "effective L4" here, to mean
"guest L4 after any transformations": XPTI copies either the actual
guest L4 if not shadowed, or the shadow L4 (transformed) if shadowed.
Sketch: guest_get_eff_kern_l1e() has a pair of toggle_guest_pt() calls
which happen if the guest is in user mode. During live migration, a
log-dirty clean will do a domain_pause and then call
shadow_blow_tables(), unpinning the shadow of the guest user root
pagetable. After execution resumes, the first toggle_guest_pt() could
then cause that page to become unshadowed. The second
toggle_guest_pt() would then re-shadow it, possibly with a different
backing shadow page.
This is XSA-516 / CVE-2026-79609.
Fixes: 5784de3e20 ("x86: Meltdown band-aid against malicious 64-bit PV guests")
Assisted-by: Claude Code:claude-fable-5-1
Signed-off-by: George Dunlap <gwd@xenproject.org>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
The identified commit inverted the condition under which Xen chooses
to set the Zenbleed chickenbit.
Undo that, and leave a better comment explaining why.
Found by AI review of the referenced commit.
This is XSA-515 / CVE-2026-79608.
Fixes: d0c75dc4c0 ("x86/amd: Fix race editing DE_CFG")
Assisted-by: Sashiko + Opus 5.0
Reported-by: George Dunlap <dunlapg@umich.edu>
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Error path of failed sidtab_map() leaves global sidtab in shutdown state and
from that point system is unable to allocate new SIDs in sidtab_context_to_sid().
To restore sidtab state new function sidtab_cancel_shutdown() is introduced,
which only resets shutdown flag, rewinding effect of sidtab_shutdown().
Signed-off-by: Sergiy Kibrik <Sergiy_Kibrik@epam.com>
Acked-by: Daniel P. Smith <dpsmith@apertussoutions.com>
We already have two instances of this, and we're about to gain a third.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: George Dunlap <gwd@xenproject.org>
As of the referenced commit there's no reason to open-code the loop
anymore. With that replaced by a single call, there's then also no point
in using the local "mfn" in this code.
Amends: 5260e8fb93 ("xen: re-define assign_pages and introduce a new function assign_page")
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: George Dunlap <gwd@xenproject.org>
NextRIP is consumed on VMRUN as the return address pushed onto the
stack for injected soft exceptions/interrupts. nsvm_vmcb_prepare4vmrun()
builds n2vmcb from the virtual VMCB on every entry to the L2 guest, but
never copied this field, leaving n2vmcb->nextrip stale. Injecting an
event in this state can wedge the L2 vCPU, as the APM does not define
behaviour for a VMRUN with a garbage NextRIP.
Nested virt is only offered on NRIPS-capable hardware, and NRIPS is
required to be advertised to L1 (see start_nested_svm()), so the L1
hypervisor is required to keep NextRIP correct. Therefore no
conditional logic is needed here; an unconditional sync suffices.
Suggested-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Chunjie Zhu <chunjie.zhu@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Do not wait indefinitely for XHCI to configure console connection. If
the cable is not plugged in, timeout after about 1s and continue boot.
If the cable is plugged in later, the console will be activated then
(see final part of dbc_ensure_running()). When timeout occurs, print a
warning message - it will be saved in the console ring buffer, to be
later retrieved from dom0, or possibly later printed when the cable is
finally connected. But, do not print the warning when waiting after
controller reset, as that may have been called from from console
dbc_putc() and printing here would result in (possibly infinite)
recursion.
Try the timeout to be about 1s, which is way above time needed for
configuring already connected cable. But, since this part runs very
early in the Xen startup, time is not calibrated yet, and functions like
NOW() do not work yet. Use raw TSC and estimate needed time to be at
least 1s on a fast CPU (5GHz), and possibly longer on a slower one. This
is not very accurate method, but fortunately it doesn't need to be.
This also helps with similar situation on resuming from suspend - if
cable got unplugged in the meantime, system will not hang waiting for
the console anymore. And finally, on some systems, when dom0 is also
allowed to access the USB controller, it may put the device into low
power mode that in some older controllers (for example Kaby Lake)
disables also debug capability. In such a case, adding a timeout allows
the system to resume, then dom0 can wake up the controller and Xen will
resume XHCI console.
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Valgrind reports that 21 bytes are "still reachable" from the (XEN_LOG_DIR "/console") allocation:
HEAP SUMMARY:
in use at exit: 21 bytes in 1 blocks
total heap usage: 9 allocs, 8 frees, 4,799 bytes allocated
Since the dynamic memory allocation for the default log directory path happens before the process
forks into the background, the parent and intermediate processes exit during daemonize()
with the memory still reachable.
Move the strdup() allocation down below the daemonize() block. This ensures only the final
background daemon allocates the default path, matching the lifetime of the free(log_dir)
cleanup loop at the exit of main().
With this change, Valgrind reports a clean heap summary
HEAP SUMMARY:
in use at exit: 0 bytes in 0 blocks
total heap usage: 8 allocs, 8 frees, 4,778 bytes allocated
Signed-off-by: Andrew Mbugua <andrewprecious388@gmail.com>
Reviewed-by: Anthony PERARD <anthony.perard@vates.tech>
xentop parses -d/--delay and -i/--iterations with atoi(), so invalid
input is silently accepted with surprising results: a fractional delay
such as "-d 2.5" is truncated to 2; "-d abc" parses as 0, which in
batch mode turns the output loop into a busy loop; "-d -1" wraps to an
effective delay of about 136 years; and "-i 0" (or any unparsable
iterations count) decrements an unsigned counter from zero, running
for about 2^32 iterations. atoi() also has undefined behaviour on
out-of-range input.
Parse both options with strtoull() instead, and reject anything that
is not a plain decimal integer in range: a sign, a fractional part,
trailing junk or overflow now produce an error and exit rather than a
silently wrong value.
Compatibility considerations: "--delay 0" remains accepted, since
updating as fast as possible is a plausible deliberate choice and
works today; "--iterations 0" is rejected, since running the loop
2^32 times cannot be what the caller meant. The interactive 'D'
prompt already validates its input and is unchanged. The only
previously useful invocation this breaks is a fractional delay, which
now fails loudly instead of silently rounding down - which is the
point of the change.
A patch documenting the --delay truncation was posted in 2010 but
never applied:
Link: https://lore.kernel.org/xen-devel/01ea26d2420e3562eb30.1292604768@chilopoda.uk.xensource.com/
Signed-off-by: Matthias Goergens <matthias.goergens@gmail.com>
Reviewed-by: Anthony PERARD <anthony.perard@vates.tech>
xenstat_collect_networks() looks for a Linux bridge (a
/sys/class/net/*/bridge directory) so that, with bonding, dom0 can be
given the bridge's counters. When there is none, as on hosts using Open
vSwitch (the XCP-ng default), routed networking or only libvirt's
virbr0 (which is excluded on purpose), getBridge() leaves devBridge
empty. strstr(iface, "") then matches every interface, and devNoBridge
is just "p", so every VIF with a standard name (vifN.M, vifN.M-emu)
takes the bridge branch instead of being attached to its domain. Every
domain ends up with zero networks, and xentop shows NETS 0.
Only take the bridge branch when a bridge was actually found.
Fixes: 1fd93d4f8c ("xenstat: Linux dom0 statistics for case we use network bonding")
Assisted-by: Claude-Code:claude-opus-5-5
Signed-off-by: Olivier Lambert <olivier.lambert@vates.tech>
Tested-by: Jason Andryuk <jason.andryuk@amd.com>
Reviewed-by: Jason Andryuk <jason.andryuk@amd.com>
Acked-by: Anthony PERARD <anthony.perard@vates.tech>
Add a new directory linux-bridge-vlan with example files showing
how to configure systemd-networkd to support a bridge VLAN
configuration.
Signed-off-by: Leigh Brown <leigh@solinno.co.uk>
Acked-by: Anthony PERARD <anthony.perard@vates.tech>
Reviewed-by: Jason Andryuk <jason.andryuk@amd.com>
Update add_to_bridge shell function to read the vlan parameter from
xenstore and set the bridge VLAN configuration for the VID.
Add additional helper functions to parse the vlan specification,
which consists of one or more of the following:
a) single VLAN (e.g. 10).
b) contiguous range of VLANs (e.g. 10-15).
c) discontiguous range with base, increment and count
(e.g. 100+10x9 which gives VLAN IDs 100, 110, ... 190).
A single VLAN can be suffixed with "p" to indicate the PVID, or
"u" to indicate untagged. A range of VLANs can be suffixed with
"u" to indicate untagged. A complex example would be:
vlan=1p/10-15/20-25u
This capability requires the iproute2 bridge command to be
installed. An error will be generated if the vlan parameter is
set and the bridge command is not available.
Signed-off-by: Leigh Brown <leigh@solinno.co.uk>
Acked-by: Anthony PERARD <anthony.perard@vates.tech>
Reviewed-by: Jason Andryuk <jason.andryuk@amd.com>
RFC 791 sets the minimum acceptable mtu to 68:
Every internet module must be able to forward a datagram of 68
octets without further fragmentation. This is because an internet
header may be up to 60 octets, and the minimum fragment is 8 octets.
In the Linux kernel source:
include/net/ip.h:#define IPV4_MIN_MTU 68 /* RFC 791 */
While there is no reason to expect a value <68 to be read from xenstore or
from the bridge make the value test require a number >= this minimum.
Signed-off-by: James Dingwall <james@dingwall.me.uk>
Reviewed-by: Anthony PERARD <anthony.perard@vates.tech>
The set_mtu() function of xen-network-common.sh currently has this code:
if [ ${type_if} = vif ]
then
local dev_=${dev#vif}
local domid=${dev_%.*}
local devid=${dev_#*.}
local FRONTEND_PATH="/local/domain/$domid/device/vif/$devid"
xenstore_write "$FRONTEND_PATH/mtu" ${mtu}
fi
This works fine if the device has its default name but if the xen config
defines the vifname parameter the FRONTEND_PATH is incorrectly constructed.
Learn the frontend path by reading the appropriate value from the backend.
Also change use of `...` to $(...) for a consistent style in the script.
Signed-off-by: James Dingwall <james@dingwall.me.uk>
Reviewed-by: Anthony PERARD <anthony.perard@vates.tech>
Wire the per-cpupool admission-control switch through libxl
to xl sched-rtds, and document it.
Add -s/--schedparam to list or set pool-wide RTDS scheduler
parameters, and -a/--admission to enable or disable admission
control for a cpupool ("-c pool -s -a 0/1").
Signed-off-by: Furkan Caliskan <frn1furkan10@gmail.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Acked-by: Anthony PERARD <anthony.perard@vates.tech>
Everything discard_initial_modules() touches is already common code:
bootinfo.modules, using_static_heap and fw_unreserved_regions(). Move it
out of Arm into common/device-tree/bootinfo.c, next to the boot allocator
handling of the same memory, and declare it in xen/bootinfo.h, so that
other device-tree based architectures can reuse it instead of growing a
copy of their own.
The arch-specific remove_early_mappings() is not moved along with it:
how the early FDT mapping is torn down is up to the architecture. As the
DTB is one of the boot modules being released, the mapping has to go once
they are discarded, so call it from start_xen() right after
discard_initial_modules(), which keeps the common function free of any
arch dependency.
While moving it, use unsigned int for the loop counter and replace the
"goto out" by a plain return, as there is nothing left to do at the label.
No functional change.
Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
We require memory to allocate for a domain as RAM to be a multiple of a
page size. However, we neither document this nor sanity test. Specifying
memory size that does not conform to this requirement fails the domain
memory allocation in the non-obvious way that is difficult to parse for
the user (printing over-allocation messages followed by the panic that Xen
could not allocate the requested amount of memory).
While there, fix indentation from tabs to spaces for "memory" parameter in
booting.txt.
Signed-off-by: Michal Orzel <michal.orzel@amd.com>
Reviewed-by: Alejandro Vallejo <alejandro.garciavallejo@amd.com>
find_text_region() tests whether an arbitrary address lies inside
[text_start, text_end). The address and those bounds are not known to
point into one object, so relational operators on the pointers are
undefined behaviour and violate Misra C:2012 rule 18.3.
The bounds are cast to unsigned long because addr already has that
type. The type is wide enough to hold a pointer, so the comparison
result stays the same.
No functional changes.
Signed-off-by: Dmytro Prokopchuk <dmytro_prokopchuk1@epam.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
When sorting and merging overlapping ranges in
pfn_pdx_compression_setup(), the merged range is set to end where the
second range ends. If the second range is fully contained in the first,
this truncates the first range, and the tail of it is then neither
compressible nor translated correctly.
Keep the end of the merged range as the maximum of both ends.
On x86 the ranges come from the SRAT memory affinity entries, and
overlapping entries for the same node are tolerated with a warning by the
NUMA code. The caller's subsequent coverage check catches the truncated
range, so the effect is that PDX compression is disabled with a "RAM
region ... not covered" message rather than memory being mistranslated.
Add a test case that fails without this change.
Found with the ESBMC bounded model checker, and confirmed by running the
failing case natively against the unmodified code.
Fixes: c5c45bcbd6 ("pdx: introduce a new compression algorithm based on region offsets")
Assisted-by: Claude Code:claude-opus-5-5 # finding the issue with ESBMC, patch creation
Signed-off-by: Weiqi Wang <weiqi.wang-2@postgrad.manchester.ac.uk>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
vstimecmp is a 64-bit CSR independently of XLEN, which is why it is
written with csr_write64(). On RV32 that macro splits the value into
the vstimecmp/vstimecmph pair, so passing ULONG_MAX (0xffffffff there)
writes all ones to the low half and zero to the high half, leaving the
CSR at 0x00000000ffffffff rather than at its maximum. A VS-timer irq
would then become pending as soon as (time + htimedelta) reaches 2^32,
which is exactly what the code is trying to avoid.
Use ~0 instead: csr_write64() converts its argument to uint64_t, and the
conversion of the signed -1 sign-extends to all ones, so both halves get
0xffffffff. On RV64 the result is equal to ULONG_MAX, so no functional
change there.
Fixes: 25e0327306 ("xen/riscv: allow Xen to use SSTC while hiding it from guests")
Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Reviewed-by: Baptiste Le Duc <baptiste.le-duc@vates.tech>
Acked-by: Jan Beulich <jbeulich@suse.com>
vmid_handle_vmenter() reports that no flush is needed when VMIDs are
unavailable (e.g. vmid=off, or hardware with no more than one VMID bit).
Every domain then runs under VMID 0 with nothing flushed in between, so as
soon as a hart runs more than one domain, a domain entered there can use
the G-stage translations left behind by the domain which ran before it.
Report that a flush is needed on every entry instead.
Update the comment in p2m_handle_vmenter() accordingly: without VMIDs the
VS-stage TLB doesn't need a separate flush either, as the G-stage flush now
done on every entry already covers it.
While at it, spell the other early return as a bool literal.
Fixes: bff3b9ea46 ("xen/riscv: introduce VMID allocation and manegement")
Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Reviewed-by: Baptiste Le Duc <baptiste.le-duc@vates.tech>
Acked-by: Jan Beulich <jbeulich@suse.com>
kernel_image64_probe() never set info->type, so every domain kept the
DOMAIN_32BIT value a zeroed KERNEL_INFO_INIT leaves behind, even though
Xen on RISC-V is only built for RV64 and accepts nothing but a 64-bit
Image.
Nothing consumes d->type on RISC-V yet, so there is no functional change.
Fixes: f6a20eda92 ("xen/riscv: add Linux kernel loading support")
Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Reviewed-by: Baptiste Le Duc <baptiste.le-duc@vates.tech>
Acked-by: Jan Beulich <jbeulich@suse.com>
Both constants were inherited from OpenSBI and their names are misleading.
The length of an instruction is encoded in its low bits: bits[1:0] != 11
means 16 bits, otherwise bits[4:2] != 111 means 32 bits, and anything else
is 48 bits or wider. 0x3 and 0x1c are the patterns compared against in
these two tests, not masks selecting a 16-bit or a 32-bit instruction. As
each constant is needed in one place only, once INSN_IS_16BIT() and
INSN_IS_32BIT() encapsulate the tests, drop them and spell the values out
in those two macros instead.
While at it, make INSN_LEN() cover the case both macros already knew about:
an encoding of 48 bits or wider now yields 0 rather than being silently
taken for a 32-bit instruction. No ratified extension defines an instruction
of such a length, so rather than open-coding a decoder for encodings which
cannot legitimately occur, leave it to the caller to treat the 0 as an
illegal instruction.
No functional change: the only present caller applies INSN_LEN() to a bug
frame, i.e. to an instruction emitted by Xen itself, which is either 16 or
32 bits wide.
Suggested-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
Reviewed-by: Baptiste Le Duc <baptiste.le-duc@vates.tech>
Acked-by: Jan Beulich <jbeulich@suse.com>
While avoiding casts would be preferable here, the subtraction ends up
violating "Subtraction between pointers shall only be applied to pointers
that address elements of the same array". Use long arithmetic instead.
No functional change intended.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Dmytro Prokopchuk <dmytro_prokopchuk1@epam.com>
While casting to pointer types may be more natural there, the subtraction
then ends up violating "Subtraction between pointers shall only be applied
to pointers that address elements of the same array". Use long arithmetic
instead.
No functional change intended.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Dmytro Prokopchuk <dmytro_prokopchuk1@epam.com>
Add a Kconfig option that lets an integrator hard-code the number of
GICv3 List Registers Xen uses. The default (0) keeps reading the count
from ICH_VTR_EL2.ListRegs at boot. A non-zero value is validated
against the hardware count in gicv3_hyp_init() and replaces
gicv3_info.nr_lrs.
gicv3_hyp_init() now panics if CONFIG_GICV3_NR_LRS exceeds the
hardware count, and zeroes all hardware LRs (once per CPU) as defensive
hardening, so any interrupt left in an LR that Xen will not manage
cannot be picked up by the GIC.
gicv3_ich_read_lr()/gicv3_ich_write_lr() now reject out-of-range
indices with an error message, ASSERT_UNREACHABLE() and WARN() instead
of silently returning RAZ/WI; reaching this path indicates a bug.
Signed-off-by: Ayan Kumar Halder <ayan.kumar.halder@amd.com>
Signed-off-by: Michal Orzel <michal.orzel@amd.com>
Reviewed-by: Luca Fancellu <luca.fancellu@arm.com>
Reviewed-by: Bertrand Marquis <bertrand.marquis@arm.com>
The arch_init_idle_domain() initializes struct arch_csw idle_csw as:
static const struct arch_csw idle_csw = {
.from = paravirt_ctxt_switch_from,
.to = paravirt_ctxt_switch_to,
.tail = idle_loop,
};
but only place where d->arch.ctxt_switch->{to,from}() are used is
__context_switch() where they are both guarded by !is_idle_domain(pd)
which makes the callbacks in idle_csw unreachable.
Introduce a new poison value to be used in unreachable function hooks, and
set the from/to fields of idle_csw to it. Use a different value from the
currently defined ones for easy differentiation.
Signed-off-by: Grygorii Strashko <grygorii_strashko@epam.com>
Signed-off-by: Roger Pau Monné <roger@xenproject.org>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
XENFEAT_pae_pgdir_above_4gb is only used by 32bit PV guests, hence limit
its setting in XENVER_get_features to the is_pv_domain() guarded section,
with the rest of the PV specific features. Add an additional build time
guard to short-circuit the setting of XENFEAT_pae_pgdir_above_4gb if the
hypervisor has no PV 32bit support.
Signed-off-by: Roger Pau Monné <roger@xenproject.org>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Since the firmware may initialize eSPIs before Xen, and without
CONFIG_GICV3_ESPI enabled, Xen would not reinitialize them properly
during boot. In such cases, once the GIC is re-enabled in Xen,
interrupts may be received that cannot be handled.
To ensure proper operation on hardware with eSPI feature, even when the eSPI
config is disabled, gicv3_dist_espi_common_init() should be invoked
regardless of whether CONFIG_GICV3_ESPI is enabled or not. This will not
affect hardware without eSPI support, as the function checks if the
hardware supports eSPIs by reading the GICD_TYPER.ESPI field (using
GICD_TYPER_ESPIS_NUM macro), which indicates whether the extended SPI
range is supported. If the hardware does not support eSPI, the function
will not perform any actions.
There are no functional changes for setups where CONFIG_GICV3_ESPI=y.
Suggested-by: Julien Grall <jgrall@amazon.com>
Signed-off-by: Leonid Komarianskyi <leonid_komarianskyi@epam.com>
Acked-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Mykola Kvach <mykola_kvach@epam.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
Xen exposes ID_AA64PFR0_EL1.GIC and ID_PFR1.GIC from domain_cpuinfo,
which is initialized from the sanitized host CPU feature state. This
does not necessarily match the virtual interrupt controller configured
for a domain.
A vGICv2 domain can therefore observe a nonzero GIC field when the host
supports the GIC system register interface, even though Xen disables that
interface for the domain. On a GICv4.1-capable host, a vGICv3 domain can
observe encoding 0b0011, although Xen exposes only its vGICv3 model.
Derive the fields from the domain's vGIC version instead. Expose 0b0000
for vGICv2 and 0b0001 for vGICv3. This covers ID_AA64PFR0_EL1 and the
ID_PFR1_EL1 alias in AArch64 state, as well as ID_PFR1 accessed through
CP15 in AArch32 state. Leave the alias unchanged when AArch32 is
unavailable.
Fixes: 07b9acea11 ("xen/arm: Add handler for ID registers on arm64")
Fixes: 8f81064a07 ("xen/arm: Add handler for cp15 ID registers")
Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
The allocated_irqs bitmap in the existing vGIC implementation stores eSPI
allocation bits immediately after the regular vIRQ bits.
vgic_reserve_virq() converts an eSPI INTID to this compressed bitmap index,
but vgic_free_virq() used the raw INTID.
Freeing INTID 4096 therefore clears bit 4096 instead of the first eSPI bit.
This writes beyond allocated_irqs and leaves the intended eSPI bit set.
Valid eSPIs reach this path during DOMCTL bind failure cleanup and unbind,
and during vPL011 teardown.
Add virq_to_idx(), the inverse of idx_to_virq(), and use it when reserving
and freeing vIRQs. Use vgic_is_valid_line() for the validity checks when
reserving and freeing vIRQs in both vGIC implementations.
Fixes: bdde400c6e ("xen/arm: vgic: add resource management for extended SPIs")
Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
GICv3 eSPI support makes nr_irqs span the architectural INTID namespace
through ESPI_MAX_INTID, but descriptor storage is sparse. local_irq_desc[]
and irq_desc[] cover INTIDs below NR_IRQS, while espi_desc[] covers eSPIs.
INTIDs 1024 through 4095 have no backing descriptors.
Validation based only on nr_irqs accepts an INTID in this gap.
__irq_to_desc() then indexes beyond irq_desc[], and callers may lock or
update unrelated Xen memory.
Reject INTIDs that the GIC reports as unimplemented in setup_irq() before
looking up a descriptor. irq_set_spi_type() can run before the implemented
GIC line counts are available, so validate descriptor-backed ranges there
before looking up a descriptor.
Use the same descriptor range check in irq_set_spi_type() and the
assertion in __irq_to_desc() to keep them in sync. Log the IRQ number
when setup_irq() rejects an invalid line.
Fixes: 98f7060b9e ("xen/arm/irq: add handling for IRQs in the eSPI range")
Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
is_espi() currently changes its result according to CONFIG_GICV3_ESPI.
Without eSPI support, it returns false, and its assertion fails if
an eSPI INTID is passed. Callers therefore use it both to identify
eSPIs and to exclude eSPI handling when support is disabled.
Make is_espi() report only whether an INTID is in the architectural
eSPI range. Check for eSPI support at the call sites.
Use BUG_ON() if the GIC reports an eSPI without compiled-in support,
matching the handling of unsupported LPIs. Treat virtual eSPI lookup
without support as unreachable, with a NULL return from the
espi_to_pending() stub.
Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
Reviewed-by: Volodymyr Babchuk <volodymyr_babchuk@epam.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
Several Arm firmware initialization paths discard irq_set_type()'s return
value, violating MISRA C Rule 17.7. If trigger configuration fails,
initialization continues with an IRQ that was not configured as requested.
GTDT and MADT retain rejected timer and maintenance INTIDs.
check_timer_irq_cfg() and release_irq() later perform unconditional
descriptor lookups on those values. Xen has no backing descriptors for
INTIDs 1024 through 4095, so retaining one can cause an out-of-bounds
access.
Check the return value in the GTDT, MADT, SPCR, and FF-A paths. Store timer
INTIDs only after successful trigger configuration, make GTDT parsing
failure fatal, and stop notification setup when trigger configuration
fails. Keep the NS16550 and PL011 consoles available by falling back to
polling, with a warning, if their IRQ trigger configuration fails.
Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
Folding with an immediately successive region can't be done by merely
lowering the start of the region. The size of the region needs increasing
as well.
Fixes: 5427134eae ("x86: populate PVHv2 Dom0 physical memory map")
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Jason Andryuk <jason.andryuk@amd.com>
Reviewed-by: Roger Pau Monné <roger@xenproject.org>
Use of AllocateMaxAddress requires that the variable pointed to by the
last argument of ->AllocatePages() is initialized. For cfgfile buffers we
don't need AllocateMaxAddress though, at which point initialization of
"addr" also isn't necessary anymore.
Mirror the lack of address constraint also to the main / central buffer
allocation in read_file().
Fixes: df75f77092 ("EFI: avoid OOB config file reads")
Assisted-by: Sashiko + Opus 5.0
Reported-by: George Dunlap <dunlapg@umich.edu>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Daniel P. Smith <dpsmith@apertussolutions.com>
While _set_status_v2() indeed doesn't alter the grant_entry_header_t it
is handed a pointer to, _set_status_v1() does. Drop the const from the
parameter of the latter (and then necessarily also from _set_status()'s),
while adding const to the local variable of the former.
No functional change.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Roger Pau Monné <roger@xenproject.org>
Casting away const-ness (or volatile-ness) is never a good idea, and
there's no need to in cmpxchg16b().
No functional change intended.
Fixes: e0116acbbd ("x86: add cmpxchg16b support")
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Mark it as such, and then remove the code following at its sole call site,
for Eclair flagging that as a Misra rule 2.1 (unreachable code) violation.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Nicola Vetrini <nicola.vetrini@bugseng.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
The use of unreachable(), when unreachability is visible to Eclair (and
compilers), is deemed a violation. Drop the redundant statement.
No functional change.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
runq_tickle() picks an idle pCPU without taking CSCHED2_MIGRATE_RESIST
into account, but runq_candidate() on that pCPU still applies it. With
the unit's credit within CSCHED2_MIGRATE_RESIST of CSCHED2_CREDIT_MIN,
the tickled pCPU declines the unit because it is not its master, while
the master has already skipped it as deferred to the tickled pCPU. Both
go idle and the unit stays runnable on the runqueue with nobody to run
it.
Seen as a permanent PVH dom0 stall under nested virtualization, where
dom0 overruns its slice far enough to be clamped at CSCHED2_CREDIT_MIN.
Don't resist migration to the pCPU that was tickled for the unit.
Fixes: 069cf39fb1 ("xen: credit2: make tickling more deterministic")
Signed-off-by: Jiaqing Zhao <Zhao.Jiaqing@amd.com>
Reviewed-by: George Dunlap <gwd@xenproject.org>
Reviewed-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jason Andryuk <jason.andryuk@amd.com>
During core scheduler initialization, each registered scheduler
is sanity tested in two steps:
- it must provide required callbacks (e.g. init, do_schedule).
- if global_init callback is present, it must succeed.
If any of the steps fail, scheduler entry is cleared in global
'schedulers' array.
However, in the current implementation, if verification fails during
the first step, the scheduler entry is cleared but verification
sequence is not interrupted. This lead to NULL pointer dereference
when subsequent required callbacks verified, and possible during
the second step.
The patch fixes the crashes by skipping consequent checks if one of
the required callbacks are not present for the scheduler.
Fixes: ec25b3c026 ("xen/sched: only allow schedulers with all mandatory functions available")
Signed-off-by: Ruslan Ruslichenko <Ruslan_Ruslichenko@epam.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
All Zen or newer CPU which support C3 shares cache. Its not necessary to
flush the caches in software before entering C3. This will cause drop in
performance for the cores which share some caches. ARB_DIS is not used
with current AMD C state implementation. So set related flags correctly.
Signed-off-by: Deepak Sharma <deepak.sharma@amd.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Origin: https://git.kernel.org/torvalds/p/a8fb40966f19
[Port to Xen. Include Hygon.]
Assisted-by: Claude-Code:claude-opus-5-5
Signed-off-by: Olivier Lambert <olivier.lambert@vates.tech>
Reviewed-by: Jason Andryuk <jason.andryuk@amd.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Both functions already assume IRQs to be enabled upon entry, by e.g. their
acquiring of the domain's event channel lock. Hence like e.g.
hvm_migrate_pirq() (also called from here) does, saving/restoring of
EFLAGS.IF isn't necessary (because of the functions called, we can't
really avoid the saving there).
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Roger Pau Monné <roger@xenproject.org>
ioreq_server_init() takes a reference on the emulating domain before
allocating the server's range sets, but does not release it when that
allocation fails. The leaked reference prevents the emulating domain
from ever being fully destroyed.
Route this failure through the existing put_domain() cleanup.
Fixes: ccb42e0ca6 ("x86/hvm: take a reference on ioreq server emulating domain")
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@amd.com>
Reviewed-by: Mykola Kvach <mykola_kvach@epam.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
paddr_to_pte()'s "permissions" parameter, the matching local in
setup_initial_mapping() and p2m_set_permission() don't only deal with
permission bits: they also handle PTE_VALID, PTE_USER, PTE_ACCESSED and
PTE_DIRTY.
Rename them to "pte_flags" and p2m_set_pte_flags() respectively.
No functional change.
Requested-by: Jan Beulich <jbeulich@suse.com>
Assisted-by: Claude:claude-opus-5
Signed-off-by: Baptiste Le Duc <baptiste.le-duc@vates.tech>
Acked-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
When settings HVM_PARAM_IDENT_PT, skip domain pausing when :
- there is no vcpu
- unrestricted guest capability is used
Signed-off-by: Teddy Astie <teddy.astie@vates.tech>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
UPDATE_ENTRY/update_intpte always returns true since 1bc30c076a
"x86/mm: {paging, sh}_{cmpxchg, write}_guest_entry() cannot fault"
Signed-off-by: Kevin Lampis <kevin.lampis@citrix.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Zihintpause is in required_extensions[] so Xen panics when it is missing,
while it does not strictly require hardware support for it.
The PAUSE hint (Zihintpause extension) is encoded as FENCE W, 0
(0x0100000F). In accordance with the RISC-V Unprivileged ISA, HINTs are
encoded in the space of valid standard instructions. On platforms without
Zihintpause support, executing 0x0100000F is treated as a standard FENCE
with an empty successor set, which acts as a NOP and never generates an
illegal instruction trap.
Zihintpause is therefore purely an optimization hint. Drop it from
required_extensions[] so Xen can still boot on systems that don't advertise
it, and print a warning in that case.
Assisted-by: Claude:claude-opus-5
Signed-off-by: Baptiste Le Duc <baptiste.le-duc@vates.tech>
Reviewed-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
George Dunlap