The DT functions (dt_read_number(), device_tree_get_reg(), fdt_get_mem_rsv())
currently accept or return 64-bit values.
In future when we support 32-bit physical address, these DT functions are
expected to accept/return 32-bit or 64-bit values (depending on the width of
physical address). Also, we wish to detect if any truncation has occurred
(i.e. while parsing 32-bit physical addresses from 64-bit values read from DT).
device_tree_get_reg() should now be able to return paddr_t. This is invoked by
various callers to get DT address and size.
For fdt_get_mem_rsv(), we have introduced a wrapper named
fdt_get_mem_rsv_paddr() which will invoke fdt_get_mem_rsv() and translate
uint64_t to paddr_t. The reason being we cannot modify fdt_get_mem_rsv() as it
has been imported from external source.
For dt_read_number(), we have also introduced a wrapper named dt_read_paddr()
dt_read_paddr() to read physical addresses. We chose not to modify the original
function as it is used in places where it needs to specifically read 64-bit
values from dt (For e.g. dt_property_read_u64()).
Xen prints warning when it detects truncation in cases where it is not able to
return error.
Also, replaced u32/u64 with uint32_t/uint64_t in the functions touched
by the code changes.
Signed-off-by: Ayan Kumar Halder <ayan.kumar.halder@amd.com>
Reviewed-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
rangeset_{xxx}_range() functions are invoked with 'start' and 'size' as
arguments which are either 'uint64_t' or 'paddr_t'. However, the function
accepts 'unsigned long' for 'start' and 'size'. 'unsigned long' is 32 bits for
Arm32. Thus, there is an implicit downcasting from 'uint64_t'/'paddr_t' to
'unsigned long' when invoking rangeset_{xxx}_range().
So, it may seem there is a possibility of lose of data due to truncation.
In reality, 'start' and 'size' are always page aligned. And Arm32 currently
supports 40 bits as the width of physical address.
So if the addresses are page aligned, the last 12 bits contain zeroes.
Thus, we could instead pass page frame number which will contain 28 bits (40-12
on Arm32) and this can be represented using 'unsigned long'.
On Arm64, this change will not induce any adverse side effect as the max
supported width of physical address is 48 bits. Thus, the width of 'gfn'
(ie 48 - 12 = 36) can be represented using 'unsigned long' (which is 64 bits
wide).
Signed-off-by: Ayan Kumar Halder <ayan.kumar.halder@amd.com>
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
Acked-by: Julien Grall <jgrall@amazon.com>
Introduced in commit 6ddfaabcee
("x86/hvm/ioreq: simplify code and use consistent naming").
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Reviewed-by: Paul Durrant <paul@xen.org>
When traversing the list of pci devices assigned to a domain cope with
some of them not having the vpci struct allocated. It should be
possible for the hardware domain to have read-only devices assigned
that are not handled by vPCI, such support will be added by further
patches.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Making use of filechk means that we don't have to use
$(move-if-changed,). It also means that will have sometimes "UPD .." in
the build output when the target changed, rather than having "GEN ..."
all the time when "xlat.lst" happen to have a more recent modification
timestamp.
While there, replace `grep -v` by `sed '//d'` to avoid an extra
fork and pipe when building.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Luca Fancellu <luca.fancellu@arm.com>
Tested-by: Luca Fancellu <luca.fancellu@arm.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
PV domains won't use it, and even HVM ones won't when OOS is turned off
for them. There's therefore no point in putting extra pressure on the
(limited) pool of memory.
While there also zap the sh_type_to_size[] entry when OOS is disabled
altogether.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Roger Pau Monné <roger.pau@citrix.com>
"0 or 1 ... to indicate whether it is capable or incapable, respectively"
is luckily just swapped words. Making this shorter will
make the reading easier.
Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
Reviewed-by: Juergen Gross <jgross@suse.com>
Presumably by copy-and-paste we've accumulated a number of instances of
$(@D)/$(@F), which really is nothing else than $@. The split form only
needs using when we want to e.g. insert a leading . at the beginning of
the file name portion of the full name.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Bertrand Marquis <bertrand.marquis@arm.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
The 'i' iterator index stores a PDX, not a PFN, and hence the initial
assignation of start (which stores a PFN) needs a conversion from PFN
to PDX.
This is harmless currently, as the PDX compression skips the bottom
MAX_ORDER bits which cover the low 1MB, but still do the conversion
from PDX to PFN for type correctness.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Use tuple comparison to check the cppcheck version.
Take the occasion to harden the regex, escaping the dots so that we
check for them instead of generic characters.
Suggested-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
MSR_ARCH_CAPS data is now included in featureset information. Replace
opencoded checks with regular feature ones.
No functional change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
The current cpu_has_tsx_ctrl tristate is serving double pupose; to signal the
first pass through tsx_init(), and the availability of MSR_TSX_CTRL.
Drop the variable, replacing it with a once boolean, and altering
cpu_has_tsx_ctrl to come out of the feature information.
No functional change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
One area where Xen deviates from Linux is that test_bit() forces a volatile
read. This leads to poor code generation, because the optimiser cannot merge
bit operations on the same word.
Drop the use of test_bit(), and write the expressions in regular C. This
removes the include of bitops.h (which is a frequent source of header
tangles), and it offers the optimiser far more flexibility.
Bloat-o-meter reports a net change of:
add/remove: 0/0 grow/shrink: 21/87 up/down: 641/-2751 (-2110)
with half of that in x86_emulate() alone. vmx_ctxt_switch_to() seems to be
the fastpath with the greatest delta at -24, where the optimiser has
successfully removed the branch hidden in cpu_has_msr_tsc_aux.
No functional change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
We already have common and default feature adjustment helpers. Introduce one
for max featuresets too.
Offer MSR_ARCH_CAPS unconditionally in the max policy, and stop clobbering the
data inherited from the Host policy. This will be necessary to level a VM
safely for migration. Annotate the ARCH_CAPS CPUID bit as special. Note:
ARCH_CAPS is still max-only for now, so will not be inhereted by the default
policies.
With this done, the special case for dom0 can be shrunk to just resampling the
Host policy (as ARCH_CAPS isn't visible by default yet).
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Extend x86_cpu_policy_fill_native() with a read of ARCH_CAPS based on the
CPUID information just read, removing the specially handling in
calculate_raw_cpu_policy().
Right now, the only use of x86_cpu_policy_fill_native() outside of Xen is the
unit tests. Getting MSR data in this context is left to whomever first
encounters a genuine need to have it.
Extend generic_identify() to read ARCH_CAPS into x86_capability[], which is
fed into the Host Policy. This in turn means there's no need to special case
arch_caps in calculate_host_policy().
No practical change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Seed the default visibility from the dom0 special case, which for the most
part just exposes the *_NO bits. EIBRS is the one non-*_NO bit, which is
"just" a status bit to the guest indicating a change in implemention of IBRS
which is already fully supported.
Insert a block dependency from the ARCH_CAPS CPUID bit to the entire content
of the MSR. This is because MSRs have no structure information similar to
CPUID, and used by x86_cpu_policy_clear_out_of_range_leaves(), in order to
bulk-clear inaccessable words.
The overall CPUID bit is still max-only, so all of MSR_ARCH_CAPS is hidden in
the default policies.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Bits through 24 are already defined, meaning that we're not far off needing
the second word. Put both in right away.
As both halves are present now, the arch_caps field is full width. Adjust the
unit test, which notices.
The bool bitfield names in the arch_caps union are unused, and somewhat out of
date. They'll shortly be automatically generated.
Add CPUID and MSR prefixes to the ./xen-cpuid verbose output, now that there
are a mix of the two.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
We are about to move MSR_ARCH_CAPS into featureset, but the order of
operations (copy raw policy, then copy x86_capabilitiles[] in) will end up
clobbering the ARCH_CAPS value.
Some toolstacks use this information to handle TSX compatibility across the
CPUs and microcode versions where support was removed.
To avoid this transient breakage, read from raw_cpu_policy rather than
modifying it in place. This logic will be removed entirely in due course.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Right now, dom0's feature configuration is split between between the common
path and a dom0-specific one. This mostly is by accident, and causes some
very subtle bugs.
First, start by clearly defining init_dom0_cpuid_policy() to be the domain
that Xen builds automatically. The late hwdom case is still constructed in a
mostly normal way, with the control domain having full discretion over the CPU
policy.
Identifying this highlights a latent bug - the two halves of the MSR_ARCH_CAPS
bodge are asymmetric with respect to the hardware domain. This means that
shim, or a control-only dom0 sees the MSR_ARCH_CAPS CPUID bit but none of the
MSR content. This in turn declares the hardware to be retpoline-safe by
failing to advertise the {R,}RSBA bits appropriately. Restrict this logic to
the hardware domain, although the special case will cease to exist shortly.
For the CPUID Faulting adjustment, the comment in ctxt_switch_levelling()
isn't actually relevant. Provide a better explanation.
Move the recalculate_cpuid_policy() call outside of the dom0-cpuid= case.
This is no change for now, but will become necessary shortly.
Finally, place the second half of the MSR_ARCH_CAPS bodge after the
recalculate_cpuid_policy() call. This is necessary to avoid transiently
breaking the hardware domain's view while the handling is cleaned up. This
special case will cease to exist shortly.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Change of directory to xen/lib/x86 isn't needed to be shown. If
something gets updated, make will print the command line.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Luca Fancellu <luca.fancellu@arm.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
There's no real need for the associated probing - we can easily convert
to a uniform value without knowing the specific behavior (note also that
the respective comments weren't fully correct and have gone stale).
No difference in generated code.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
XSM is a subsystem where it is equally important of how and where its hooks are
called as is the implementation of the hooks. The people best suited for
evaluating the how and where are the XSM maintainers and reviewers. This
creates a challenge as the hooks are used throughout the hypervisor for which
the XSM maintainers and reviewers are not, and should not be, a reviewer for
each of these subsystems in the MAINTAINERS file. Though the MAINTAINERS file
does support the use of regex matches, 'K' identifier, that are applied to both
the commit message and the commit delta. Adding the 'K' identifier will declare
that any patch relating to XSM require the input from the XSM maintainers and
reviewers. For those that use the get_maintianers script, the 'K' identifier
will automatically add the XSM maintainers and reviewers. Any one not using
get_maintainers, it will be their responsibility to ensure that if their work
touches and XSM hook, to ensure the XSM maintainers and reviewers are copied.
This patch adds a pair of regex expressions to the XSM section. The first is
`xsm_.*` which seeks to match XSM hooks in the commit's delta. The second is
`\b(xsm|XSM)\b` which seeks to match strictly the words xsm or XSM and should
not capture words with a substring of "xsm".
Signed-off-by: Daniel P. Smith <dpsmith@apertussolutions.com>
Acked-by: Julien Grall <jgrall@amazon.com>
When a domain parameter is provided to pci_get_pdev() the search
function would match against the bdf, without taking the segment into
account.
Fix this and also account for the passed segment.
Fixes: 8cf6e07389 ('PCI: simplify (and thus correct) pci_get_pdev{,_by_domain}()')
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
When creating a domU, but the creation fails, there is a corner case that may
lead to a crash in the null scheduler when running a debug build of Xen.
(XEN) ****************************************
(XEN) Panic on CPU 0:
(XEN) Assertion 'npc->unit == unit' failed at common/sched/null.c:379
(XEN) ****************************************
The events leading to the crash are:
* null_unit_insert() was invoked with the unit offline. Since the unit was
offline, unit_assign() was not called, and null_unit_insert() returned.
* Later during domain creation, the unit was onlined
* Eventually, domain creation failed due to bad configuration
* null_unit_remove() was invoked with the unit still online. Since the unit was
online, it called unit_deassign() and triggered an ASSERT.
To fix this, only call unit_deassign() when npc->unit is non-NULL in
null_unit_remove.
Signed-off-by: Stewart Hildebrand <stewart.hildebrand@amd.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Acked-by: Dario Faggioli <dfaggioli@suse.com>
We currently have "Per Domain Paths" as an empty section, whereas it
looks like "General Paths" was not indended to include all the
following sections.
Signed-off-by: Yann Dirson <yann.dirson@vates.fr>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Calling build twice in the same environment will fail because the
directory 'binaries' was already created before. Use mkdir -p to ignore
an existing directory and move on to the actual build.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
The command used in the example is different from the command used in
the Gitlab CI pipelines. Adjust it to simulate what will be used by CI.
This is essentially the build script, which is invoked with a number of
expected environment variables such as CC, CXX and debug.
In addition the input should not be a tty, which disables colors from
meson and interactive questions from kconfig.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
repository in the reports
Currently the cppcheck report entries shows the relative file path
from the /xen folder of the repository instead of the base folder.
In order to ease the checks, for example, when looking a git diff
output and the report, use the repository folder as base.
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
Tested-by: Stefano Stabellini <sstabellini@kernel.org>
Allow the use of Cppcheck version above 2.7, exception for 2.8 which
is known and documented do be broken.
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
Tested-by: Stefano Stabellini <sstabellini@kernel.org>
Currently Cppcheck has a limitation that prevents to use make with
parallel build and have a parallel Cppcheck invocation on each
translation unit (the .c files), because of spurious internal errors.
The issue comes from the fact that when using the build directory,
Cppcheck saves temporary files as <filename>.c.<many-extensions>, but
this doesn't work well when files with the same name are being
analysed at the same time, leading to race conditions.
Fix the issue creating, under the build directory, the same directory
structure of the file being analysed to avoid any clash.
Fixes: 02b26c02c7 ("xen/scripts: add cppcheck tool to the xen-analysis.py script")
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
According to openpty(3) it is required to include <pty.h> to get the
prototypes for openpty() and login_tty(). But this is not what the
function AX_CHECK_PTYFUNCS actually does. It makes no attempt to include
the required header.
The two source files which call openpty() and login_tty() already contain
the conditionals to include the required header.
Remove the bogus m4 file to fix build with clang, which complains about
calls to undeclared functions.
Remove usage of INCLUDE_LIBUTIL_H in libxl_bootloader.c, it is already
covered by inclusion of libxl_osdep.h.
Remove usage of PTYFUNCS_LIBS in libxl/Makefile, it is already covered
by UTIL_LIBS from config/StdGNU.mk.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Reviewed-by: Anthony PERARD <anthony.perard@citrix.com>
On a fedora system, if you run `sudo sh install.sh` you break your
system. The installation clobbers /var/run, a symlink to /run.
A subsequent boot fails when /var/run and /run are different since
accesses through /var/run can't find items that now only exist in /run
and vice-versa.
Skip populating /var/run/xen during make install.
The directory is already created by some scripts. Adjust all remaining
scripts to create XEN_RUN_DIR at runtime.
Use the shell variable XEN_RUN_DIR instead of hardcoded paths.
XEN_RUN_STORED is covered by XEN_RUN_DIR because xenstored is usually
started afterwards.
Reported-by: Jason Andryuk <jandryuk@gmail.com>
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Tested-by: Jason Andryuk <jandryuk@gmail.com>
Reviewed-by: Anthony PERARD <anthony.perard@citrix.com>
clang complains about the signed type:
implicit truncation from 'int' to a one-bit wide bit-field changes value from 1 to -1 [-Wsingle-bit-bitfield-constant-conversion]
The potential ABI change in libxenvchan is covered by the Xen version based SONAME.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Reviewed-by: Juergen Gross <jgross@suse.com>
Acked-by: Anthony PERARD <anthony.perard@citrix.com>
Because CpuIdUserDis is reported in CPUID itself, the extended leaf
containing that bit must be retrieved before calling c_early_init()
Signed-off-by: Alejandro Vallejo <alejandro.vallejo@cloud.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Move vendor-specific checks to the vendor-specific callers. While at it
move the synth cap setters to the callers too, as it's needed for a later
patch and it's not a functional change either.
No functional change.
Signed-off-by: Alejandro Vallejo <alejandro.vallejo@cloud.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
The current logic to set SSBD on AMD Family 17h and Hygon Family 18h
processors requires that the setting of SSBD is coordinated at a core
level, as the setting is shared between threads. Logic was introduced
to keep track of how many threads require SSBD active in order to
coordinate it, such logic relies on using a per-core counter of
threads that have SSBD active.
Given the current logic, it's possible for a guest to under or
overflow the thread counter, because each write to VIRT_SPEC_CTRL.SSBD
by the guest gets propagated to the helper that does the per-core
active accounting. Overflowing the counter is not so much of an
issue, as this would just make SSBD sticky.
Underflowing however is more problematic: on non-debug Xen builds a
guest can perform empty writes to VIRT_SPEC_CTRL that would cause the
counter to underflow and thus the value gets saturated to the max
value of unsigned int. At which points attempts from any thread to
set VIRT_SPEC_CTRL.SSBD won't get propagated to the hardware anymore,
because the logic will see that the counter is greater than 1 and
assume that SSBD is already active, effectively loosing the setting
of SSBD and the protection it provides.
Fix this by introducing a per-CPU variable that keeps track of whether
the current thread has legacy SSBD active or not, and thus only
attempt to propagate the value to the hardware once the thread
selected value changes.
This is XSA-431 / CVE-2022-42336
Fixes: b2030e6730 ('amd/virt_ssbd: set SSBD at vCPU context switch')
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
The upcoming Leap 15.5 will come without a binary named 'python'.
Prepare the suse images for that change.
Starting with Xen 4.14 python3 can be used for build.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
The diffutils package is a hard requirement for building xen.
It was dropped from the Tumbleweed base image in the past 12 months.
Building with --enable-docs does now require the gs tool.
Add both packages to the suse dockerfiles.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
This adds another physical runner to Gitlab-CI, running similar set of
jobs that the Adler Lake one.
The machine specifically is
MinisForum UM773 Lite with AMD Ryzen 7 7735HS
The PV passthrough test is skipped as currently it fails on this system
with:
(d1) Can't find new memory area for initrd needed due to E820 map conflict
The S3 test is skipped as it currently fails - the system seems to
suspend properly (power LED blinks), but when woken up the power LED
gets back to solid on and the fan spins at top speed and but otherwise there is no
signs of if life from the system (no output on the console, HDMI or
anything else).
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
This makes the test script easier reusable for different runners, where
console may be connected differently. Include both console= option and
configuration for specific chosen console too (like com1= here) in the
'CONSOLE_OPTS' variable.
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
When adding new featureset words, it is convenient to split the work into
several patches. However, GCC 12 spotted that the way we prefer to split the
work results in a real (transient) breakage whereby the policy <-> featureset
helpers perform out-of-bounds accesses on the featureset array.
Fix this by having gen-cpuid.py calculate FEATURESET_NR_ENTRIES from the
comments describing the word blocks, rather than from the XEN_CPUFEATURE()
with the greatest value.
For simplicty, require that the word blocks appear in order. This can be
revisted if we find a good reason to have blocks out of order.
No functional change.
Reported-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
IRQ is of unsigned type so %u should be used. When printing domain id,
%pd should be the correct format to maintain the consistency.
Also, wherever possible, reduce the number of split lines for printk().
Signed-off-by: Michal Orzel <michal.orzel@amd.com>
Reviewed-by: Henry Wang <Henry.Wang@arm.com>
Reviewed-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Ayan Kumar Halder <ayan.kumar.halder@amd.com>
From map_dt_irq_to_domain() we are assigning a return code of
map_irq_to_domain() to a variable without checking it for an error.
Fix it by propagating the return code directly since this is the last
call.
Fixes: 467e5cbb2f ("xen: arm: consolidate mmio and irq mapping to dom0")
Signed-off-by: Michal Orzel <michal.orzel@amd.com>
Reviewed-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Henry Wang <Henry.Wang@arm.com>
Reviewed-by: Ayan Kumar Halder <ayan.kumar.halder@amd.com>
At the moment, even in case of a SMMU being I/O coherent, we clean the
updated PT as a result of not advertising the coherency feature. SMMUv3
coherency feature means that page table walks, accesses to memory
structures and queues are I/O coherent (refer ARM IHI 0070 E.A, 3.15).
Follow the same steps that were done for SMMU v1,v2 driver by the commit:
080dcb781e
The same restrictions apply, meaning that in order to advertise coherent
table walk platform feature, all the SMMU devices need to report coherency
feature. This is because the page tables (we are sharing them with CPU)
are populated before any device assignment and in case of a device being
behind non-coherent SMMU, we would have to scan the tables and clean
the cache.
It is to be noted that the SBSA/BSA (refer ARM DEN0094C 1.0C, section D)
requires that all SMMUv3 devices support I/O coherency.
Signed-off-by: Michal Orzel <michal.orzel@amd.com>
Reviewed-by: Ayan Kumar Halder <ayan.kumar.halder@amd.com>
Reviewed-by: Rahul Singh <rahul.singh@arm.com>
[julien: Add __ro_after_init]
Reviewed-by: Julien Grall <jgrall@amazon.com>
Ayan Kumar Halder