Currently, only liveupdate request can be delayed. The request can only
be performed by a privileged connection (e.g. dom0). So it is fine to
have no limits.
In a follow-up patch we will want to delay request for unprivileged
connection as well. So it is best to apply a limit.
For now and for simplicity, only a single request can be delayed
for a given unprivileged connection.
Take the opportunity to tweak the prototype and provide a way to
bypass the quota check. This would be useful when the function
is called from the restore code.
Signed-off-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Luca Fancellu <luca.fancellu@arm.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
call_delayed() is currently assuming that conn->in is NULL when
handling delayed request. However, the connection is not paused.
Therefore new request can be processed and conn->in may be non-NULL
if we have only received a partial request.
Furthermore, as we overwrite conn->in, the current partial request
will not be transferred. This will result to corrupt the connection.
Rather than updating conn->in, stash the LU request in lu_status and
let each callback for delayed request to update conn->in when
necessary.
To keep a sane interface, the code to write the "OK" response the
LU request is moved in xenstored_core.c.
Fixes: c5ca1404b4 ("tools/xenstore: add support for delaying execution of a xenstore request")
Fixes: ed6eebf17d ("tools/xenstore: dump the xenstore state for live update")
Signed-off-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
At the moment, dump_state_buffered_data() is taking two connections
in parameters (one is the connection to dump, the other is the
connection used to request LU). The naming doesn't help to
distinguish (c vs conn) them and this already lead to several mistake
while modifying the function.
To remove the confusion, introduce an help lu_get_connection() that
will return the connection used to request LU and use it
in place of the existing parameter.
Signed-off-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Luca Fancellu <luca.fancellu@arm.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
I would like to help reviewing Xenstored patches. It is more convenient
to find them if I am CCed.
Signed-off-by: Julien Grall <julien@xen.org>
Acked-by: Juergen Gross <jgross@suse.com>
This reverts commit aad7b5c11d.
The change from OvmfX64 to OvmfXen causes a change in behaviour, whereby
OvmfXen maps its shared info page at the top of address space. When trying to
migrate such a domain, XENMEM_maximum_gpfn returns a very large value. This
has uncovered multiple issues:
1) The userspace hypercall wrappers truncate all return values to int on
Linux and Solaris, even on 64bit. This needs fixing in libxenctrl.
2) 32bit toolstacks can't migrate any domain with RAM above the 2^40 mark,
because of virtual address constraints. This needs fixing in OVMF.
Fixes for both of these aren't completely trivial. Revert the change to
unblock staging in the meantime.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Anthony PERARD <anthony.perard@citrix.com>
Commit 871e51d2d4 changed the sign on the xenlight error types (making
the values negative, same as the C-generated constants), but failed to
remove the code changing the sign before casting to Error(). This
results in error strings like "libxl error: <x>", rather than the
correct message. Fix all occurrances of this by running:
gofmt -w -r 'Error(-ret) -> Error(ret)' xenlight.go
from tools/golang/xenlight.
Signed-off-by: Nick Rosbrook <rosbrookn@ainfosec.com>
Acked-by: George Dunlap <george.dunlap@citrix.com>
As a matter of style, it is strange to see capitalized receiver names,
due to the significance of capitalized symbols in Go (although there is
in fact nothing special about a capitalized receiver name). Fix this in
xenlight.go by running:
gofmt -w -r 'Ctx -> ctx' xenlight.go
from tools/golang/xenlight. There is no functional change.
Signed-off-by: Nick Rosbrook <rosbrookn@ainfosec.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
Currently, when marshalig Go types with keyed union fields, we assign the
value of the struct (e.g. DomainBuildInfoTypeUnionHvm) which implements the
interface of the keyed union field (e.g. DomainBuildInfoTypeUnion).
As-is, this means that if a populated DomainBuildInfo is marshaled to
e.g. JSON, unmarshaling back to DomainBuildInfo will fail.
When the encoding/json is unmarshaling data into a Go type, and
encounters a JSON object, it basically can either marshal the data into
an empty interface, a map, or a struct. It cannot, however, marshal data
into an interface with at least one method defined on it (e.g.
DomainBuildInfoTypeUnion). Before this check is done, however, the
decoder will check if the Go type is a pointer, and dereference it if
so. It will then use the type of this value as the "target" type.
This means that if the TypeUnion field is populated with a
DomainBuildInfoTypeUnion, the decoder will see a non-empty interface and
fail. If the TypeUnion field is populated with a
*DomainBuildInfoTypeUnionHvm, it dereferences the pointer and sees a
struct instead, allowing decoding to continue normally.
Since there does not appear to be a strict need for NOT using pointers
in these fields, update code generation to set keyed union fields to
pointers of their implementing structs.
Signed-off-by: Nick Rosbrook <rosbrookn@ainfosec.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
For structs that have a keyed union, e.g. DomainBuildInfo, the TypeUnion
field must be exported so that package users can get/set the fields
within. This means that users are aware of the existence of the
interface type used in those fields (see [1]), so it is awkward that the
interface itself is not exported. However, the single method within the
interface must remain unexported so that users cannot mistakenly "implement"
those interfaces.
Since there seems to be no reason to do otherwise, export the keyed
union interface types.
[1] https://pkg.go.dev/xenbits.xenproject.org/git-http/xen.git/tools/golang/xenlight?tab=doc#DeviceUsbdev
Signed-off-by: Nick Rosbrook <rosbrookn@ainfosec.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
The current implementation of StringList.toC does not correctly account
for how libxl_string_list is expected to be laid out in C, which is clear
when one looks at libxl_string_list_length in libxl.c. In particular,
StringList.toC does not account for the extra memory that should be
allocated for the "sentinel" entry. And, when using the "slice trick" to
create a slice that can address C memory, the unsafe.Pointer conversion
should be on a C.libxl_string_list, not *C.libxl_string_list.
Fix these problems by (1) allocating an extra slot in the slice used to
address the C memory, and explicity set the last entry to nil so the C
memory will be zeroed out, and (2) dereferencing csl in the
unsafe.Pointer conversion.
Signed-off-by: Nick Rosbrook <rosbrookn@ainfosec.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
Re-generate code to reflect changes to libxl_types.idl from the
following commits:
0570d7f276 x86/msr: introduce an option for compatible MSR behavior selection
7e5cffcd1e viridian: allow vCPU hotplug for Windows VMs
9835246710 viridian: remove implicit limit of 64 VPs per partition
Signed-off-by: Nick Rosbrook <rosbrookn@ainfosec.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
Force WB type for grants and foreign pages. Those are usually mapped
over unpopulated physical ranges in the p2m, and those ranges would
usually be UC in the MTRR state, which is unlikely to be the correct
cache attribute. It's also cumbersome (or even impossible) for the
guest to be setting the MTRR type for all those mappings as WB, as
MTRR ranges are finite.
Note that this is not an issue on AMD because WB cache attribute is
already set on grants and foreign mappings in the p2m and MTRR types
are ignored. Also on AMD Xen cannot force a cache attribute because of
the lack of ignore PAT equivalent, so the behavior here slightly
diverges between AMD and Intel (or EPT vs NPT/shadow).
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
This is an EPT specific function, so it shouldn't live in the generic
mtrr file. Such movement is also needed for future work that will
require passing a p2m_type_t parameter to epte_get_entry_emt, and
making that type visible to the mtrr users is cumbersome and
unneeded.
Moving epte_get_entry_emt out of mtrr.c requires making the helper to
get the MTRR type of an address from the mtrr state public. While
there rename the function to start with the mtrr prefix, like other
mtrr related functions.
While there fix some of the types of the function parameters.
No functional change intended.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
... so tests can peek at the internals, without the structure being generally
available to users of the library.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
MSR_ARCH_CAPS is still not supported for guests yet (other than the hardware
domain), until the toolstack learns how to construct an MSR policy.
However, we want access to the host ARCH_CAPS_TSX_CTRL value in particular for
testing purposes.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
The logic to disallow writes to the TSC is out-of-place, and should be in
check_resource_access() rather than in resource_access().
Split the existing allow_access_msr() into two - msr_{read,write}_allowed() -
and move all permissions checks here.
Furthermore, guard access to MSR_IA32_CMT_{EVTSEL,CTR} to prohibit their use
on hardware which is lacking the QoS Monitoring feature. Introduce
cpu_has_pqe to help with the logic.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Commit cf8c4d3d13 made some preparation to have one day
variable-length-array argument, but didn't declare the array in the
function prototype the same way as in the function definition. And now
GCC 11 complains about it.
Fixes: cf8c4d3d13 ("tools/libs/foreignmemory: pull array length argument to map forward")
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
This section's contents do not represent part of actual hypervisor text,
so shouldn't be included in what is_kernel_inittext() or (while still
booting) is_active_kernel_text() report "true" for. Keep them in
.init.text though, as there's no real reason to have a separate section
for this in the final binary.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
With pt_vcpu_lock() no longer acquiring the pt_migrate lock, parties
iterating the list and acting on the timers of the list entries will no
longer be kept from entering their loops by create_periodic_time()'s
holding of that lock. Therefore at least init_timer() needs calling
ahead of list insertion, but keep this and set_timer() together.
Fixes: 8113b02f0b ("x86/vpt: do not take pt_migrate rwlock in some cases")
Reported-by: Igor Druzhinin <igor.druzhinin@citrix.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
"change" command as been removed in QEMU 6.0. We can use
"blockdev-change-medium" instead.
Using `id` with "blockdev-change-medium" requires a change to the QEMU
command line, introduced by:
"libxl: Use -device for cd-rom drives"
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Jason Andryuk <jandryuk@gmail.com>
`device` parameter is deprecated since QEMU 2.8.
This requires changes to the command line introduced by:
"libxl: Use -device for cd-rom drives"
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Jason Andryuk <jandryuk@gmail.com>
We are going to want to check QEMU's version in other places where we
can use libxl__ev_qmp_send.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Jason Andryuk <jandryuk@gmail.com>
We are supposed to read the version information only when qmp_ev is in
state "Connected" (that correspond to state==qmp_state_connected),
assert it so that the function isn't used too early.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Jason Andryuk <jandryuk@gmail.com>
This allows to set an `id` on the device instead of only the drive. We
are going to need the `id` with the "eject" and
"blockdev-change-media" QMP command as using `device` parameter on
those is deprecated. (`device` is the `id` of the `-drive` on the
command line).
We set the same `id` on both -device and -drive as QEMU doesn't
complain and we can then either do "eject id=$id" or "eject
device=$id".
Using "-drive + -device" instead of only "-drive" has been
available since at least QEMU 0.15, and seems to be the preferred way as it
separates the host part (-drive which describe the disk image location
and format) from the guest part (-device which describe the emulated
device). More information in qemu.git/docs/qdev-device-use.txt .
Changing the command line during migration for the cdrom seems fine.
Also the documentation about migration in QEMU explains that the device
state ID is "been formed from a bus name and device address", so
second IDE bus and first device address on bus is still thus and
doesn't matter if written "-drive if=ide,index=2" or "-drive
ide-cd,bus=ide.1,unit=0".
See qemu.git/docs/devel/migration.rst .
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Jason Andryuk <jandryuk@gmail.com>
The command "cpu-add" for CPU hotplug is deprecated and has been
removed from QEMU 6.0 (April 2021). We need to add cpus with the
command "device_add" now.
In order to find out which parameters to pass to "device_add" we first
make a call to "query-hotpluggable-cpus" which list the cpus drivers
and properties.
The algorithm to figure out which CPU to add, and by extension if any
CPU needs to be hotplugged, is in the function that adds the cpus.
Because of that, the command "query-hotpluggable-cpus" is always
called, even when not needed.
In case we are using a version of QEMU older than 2.7 (Sept 2016)
which don't have "query-hotpluggable-cpus", we fallback to using
"cpu-add".
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Jason Andryuk <jandryuk@gmail.com>
Short-form boolean options are deprecated in QEMU 6.0.
Upstream commit that deprecate those: ccd3b3b8112b ("qemu-option: warn
for short-form boolean options").
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Jason Andryuk <jandryuk@gmail.com>
We use the deprecated QMP command "query-cpus" which is removed in the
QEMU 6.0 release. There's a replacement which is "query-cpus-fast",
and have been available since QEMU 2.12 (April 2018).
This patch try the new command first and when the command isn't
available, it fall back to the deprecated one so libxl still works
with older QEMU versions.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Jason Andryuk <jandryuk@gmail.com>
This confuses disassemblers, at the very least. Move
.altinstr_replacement to .init.text. The previously redundant ALIGN()
now gets converted to page alignment, such that the hypervisor mapping
won't have this as executable (it'll instead get mapped r/w, which I'm
told is intended to be adjusted at some point).
Note that for the actual patching logic's purposes this part of
.init.text _has_ to live after _einittext (or before _sinittext), or
else branch_insn_requires_update() would produce wrong results.
Also, to have .altinstr_replacement have consistent attributes in the
object files, add "x" to the one instance where it was missing.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Julien Grall <jgrall@amazon.com>
The section .proc.info lives in .rodata as it doesn't contain any
executable code. However, the section is still marked as executable
as the consequence .rodata will also be marked executable.
Xen doesn't use the ELF permissions to decide the page-table mapping
permission. However, this will confuse disassemblers.
'#execinstr' is now removed on all the pushsection dealing with
.proc.info
Signed-off-by: Jan Beulich <jbeulich@suse.com>
[julieng: Rework the commit message]
Acked-by: Julien Grall <jgrall@amazon.com>
I have touched most of the Xen libraries in the past, and there is a
clear lack of reviewer band width in the tools area.
Add myself as a tools/libs reviewer for that reason.
Signed-off-by: Juergen Gross <jgross@suse.com>
Acked-by: Ian Jackson <iwj@xenproject.org>
The reasoning for commit 7bd8989ab7 ("tools/libs/guest: fix max_pfn
setting in map_p2m()") was wrong.
The max_pfn field in shared_info is misnamed, it has the semantics of
num_pfns, which is hidden at least partially in Linux, as the kernel is
(wrongly) treating it like the highest used pfn in some places.
So revert above commit.
Fixes: 7bd8989ab7 ("tools/libs/guest: fix max_pfn setting in map_p2m()")
Signed-off-by: Juergen Gross <jgross@suse.com>
Since XSA-228 (commit 02cbeeb620 "gnttab: split maptrack lock
to make it fulfill its purpose again"), v->maptrack_head,
v->maptrack_tail and the content of the freelist are accessed with
the lock v->maptrack_freelist_lock held.
Therefore it is not necessary to update the fields using cmpxchg()
and also read them atomically.
Note that there are two cases where v->maptrack_tail is accessed without
the lock. They both happen in get_maptrack_handle() when initializing
the free list of the current vCPU. Therefore there is no possible race.
The code is now reworked to remove any use of cmpxch() and read_atomic()
when accessing the fields v->maptrack_{head, tail} as wel as the
freelist.
Take the opportunity to add a comment on top of the lock definition
and explain what it protects.
Signed-off-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
The Arm ARM's description of MSR (ARM DDI 0406C.d section B9.3.12)
doesn't even allow for plain "SPSR" here, and while gas accepts this, it
takes it to mean SPSR_cf. Yet surely all of SPSR wants updating on this
path, not just the lowest and highest 8 bits.
Fixes: dfcffb128b ("xen/arm32: SPSR_hyp/SPSR")
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
There are some internals in the libxenstore interface which should be
removed.
Move those functions into xs_lib.c and the related definitions into
xs_lib.h. Remove the functions from the mapfile. Add xs_lib.o to
xenstore_client as some of the internal functions are needed there.
Bump the libxenstore version to 4.0 as the change is incompatible.
Note that the removed functions should not result in any problem as
they ought to be used by xenstored or xenstore_client only.
Avoid an enum as part of a structure as the size of an enum is
compiler implementation dependent.
Signed-off-by: Juergen Gross <jgross@suse.com>
Acked-by: Ian Jackson <iwj@xenproject.org>
Clang 10 reports
domain.c:1328:10: error: variable 'cr3_mfn' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]
if ( !compat )
^~~~~~~
domain.c:1334:34: note: uninitialized use occurs here
cr3_page = get_page_from_mfn(cr3_mfn, d);
^~~~~~~
domain.c:1328:5: note: remove the 'if' if its condition is always true
if ( !compat )
^~~~~~~~~~~~~~
domain.c:1042:18: note: initialize the variable 'cr3_mfn' to silence this warning
mfn_t cr3_mfn;
^
= 0
domain.c:1189:14: error: variable 'fail' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]
if ( !compat )
^~~~~~~
domain.c:1211:9: note: uninitialized use occurs here
fail |= v->arch.pv.gdt_ents != c(gdt_ents);
^~~~
domain.c:1189:9: note: remove the 'if' if its condition is always true
if ( !compat )
^~~~~~~~~~~~~~
domain.c:1187:18: note: initialize the variable 'fail' to silence this warning
bool fail;
^
= false
despite this being a build with -O2 in effect, and despite "compat"
being constant "false" when CONFIG_COMPAT (and hence CONFIG_PV32) is not
defined, as it gets set at the top of the function from the result of
is_pv_32bit_domain().
Re-arrange the two "offending" if()s such that when COMPAT=n the
respective variables will be seen as unconditionally initialized. The
original aim was to have the !compat cases first, though.
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
SPSR_hyp is not meant to be accessed from Hyp mode (EL2); accesses
trigger UNPREDICTABLE behaviour. Xen should read/write SPSR instead.
See: ARM DDI 0487D.b page G8-5993.
This fixes booting Xen/arm32 on QEMU.
Signed-off-by: Stefano Stabellini <stefano.stabellini@xilinx.com>
Reviewed-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Tested-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
The June 2021 microcode is formally de-featuring TSX on the older Skylake
client CPUs. The workaround from the March 2019 microcode is being dropped,
and replaced with additions to MSR_TSX_FORCE_ABORT to hide the HLE/RTM CPUID
bits.
With this microcode in place, TSX is disabled by default on these CPUs.
Backwards compatibility is provided in the same way as for TAA - RTM force
aborts, rather than suffering #UD, and the CPUID bits can be hidden to recover
performance.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Add arch-specific makefiles and configs needed to build for
riscv. Also add a minimal head.S that is a simple infinite loop.
head.o can be built with
$ make XEN_TARGET_ARCH=riscv64 SUBSYSTEMS=xen -C xen tiny64_defconfig
$ make XEN_TARGET_ARCH=riscv64 SUBSYSTEMS=xen -C xen TARGET=riscv64/head.o
No other TARGET is supported at the moment.
Signed-off-by: Connor Davis <connojdavis@gmail.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Acked-by: Bobby Eshleman <bobbyeshleman@gmail.com>
Defaulting to yes only for X86 and ARM reduces the requirements
for a minimal build when porting new architectures.
Signed-off-by: Connor Davis <connojdavis@gmail.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Better reflect reality: Andrew and Jan are active maintainers
and I review patches. Keep myself as a reviewer so I can help
with historical context &c.
Signed-off-by: Tim Deegan <tim@xen.org>
Acked-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
First and foremost - such timeouts were not signaled to callers, making
them believe they're fine to e.g. free previously unmapped pages.
Mirror VT-d's behavior: A fixed number of loop iterations is not a
suitable way to detect timeouts in an environment (CPU and bus speeds)
independent manner anyway. Furthermore, leaving an in-progress operation
pending when it appears to take too long is problematic: If a command
completed later, the signaling of its completion may instead be
understood to signal a subsequently started command's completion.
Log excessively long processing times (with a progressive threshold) to
have some indication of problems in this area. Allow callers to specify
a non-default timeout bias for this logging, using the same values as
VT-d does, which in particular means a (by default) much larger value
for device IO TLB invalidation.
This is part of XSA-373 / CVE-2021-28692.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul@xen.org>
No caller cared about send_iommu_command() indicating unavailability of
a slot. Hence if a sufficient number prior commands timed out, we did
blindly assume that the requested command was submitted to the IOMMU
when really it wasn't. This could mean both a hanging system (waiting
for a command to complete that was never seen by the IOMMU) or blindly
propagating success back to callers, making them believe they're fine
to e.g. free previously unmapped pages.
Fold the three involved functions into one, add spin waiting for an
available slot along the lines of VT-d's qinval_next_index(), and as a
consequence drop all error indicator return types/values.
This is part of XSA-373 / CVE-2021-28692.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul@xen.org>
The user chosen setting for MSR_TSX_CTRL needs restoring after S3.
All APs get the correct setting via start_secondary(), but the BSP was missed
out.
This is XSA-377 / CVE-2021-28690.
Fixes: 8c4330818f ("x86/spec-ctrl: Mitigate the TSX Asynchronous Abort sidechannel")
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Modern x86 processors have far-better-than-architecturally-guaranteed self
modifying code detection. Typically, when a write hits an instruction in
flight, a Machine Clear occurs to flush stale content in the frontend and
backend.
For self modifying code, before a write which hits an instruction in flight
retires, the frontend can speculatively decode and execute the old instruction
stream. Speculation of this form can suffer from type confusion in registers,
and potentially leak data.
Furthermore, updates are typically byte-wise, rather than atomic. Depending
on timing, speculation can race ahead multiple times between individual
writes, and execute the transiently-malformed instruction stream.
Xen has stubs which are used in certain cases for emulation purposes. Inhibit
speculation between updating the stub and executing it.
This is XSA-375 / CVE-2021-0089.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Leaving an in-progress operation pending when it appears to take too
long is problematic: If e.g. a QI command completed later, the write to
the "poll slot" may instead be understood to signal a subsequently
started command's completion. Also our accounting of the timeout period
was actually wrong: We included the time it took for the command to
actually make it to the front of the queue, which could be heavily
affected by guests other than the one for which the flush is being
performed.
Do away with all timeout detection on all flush related code paths.
Log excessively long processing times (with a progressive threshold) to
have some indication of problems in this area.
Additionally log (once) if qinval_next_index() didn't immediately find
an available slot. Together with the earlier change sizing the queue(s)
dynamically, we should now have a guarantee that with our fully
synchronous model any demand for slots can actually be satisfied.
This is part of XSA-373 / CVE-2021-28692.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul@xen.org>
With the present synchronous model, we need two slots for every
operation (the operation itself and a wait command). There can be one
such pair of commands pending per CPU. To ensure that under all normal
circumstances a slot is always available when one is requested, size the
command ring according to the number of present CPUs.
This is part of XSA-373 / CVE-2021-28692.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul@xen.org>
Julien Grall
Julien Grall
Andrew Cooper
Nick Rosbrook
Roger Pau Monné
Anthony PERARD
Jan Beulich
Juergen Gross
Stefano Stabellini
Connor Davis
Tim Deegan