Current code that allocates memory and populates the p2m for PVH Dom0
doesn't take the address alignment into account, this can lead to high
order allocations that start on a non-aligned address to be broken
down into lower order entries on the p2m page tables.
Fix this by taking into account the p2m page sizes and alignment
requirements when allocating the memory and populating the p2m.
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Tested-by: Andrew Cooper <andrew.cooper3@citrix.com>
Switch parse_ept_param() to use the parse_boolean() infrastructure for more
consistency with related command line parameters. Rename opt_pml_enabled to
opt_ept_pml for consistency with opt_ept_ad, and switch it to being bool
Drop the leading comment for parse_ept_param(). It is stale, and just repeats
the command line documentation.
For the command line documentation, rewrite it largely from scratch, updating
to the latest metadata style. Document A/D first, including a note about
AVR41, and modify PML to note its dependency on A/D.
No practical changes to behaviour.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Kevin Tian <kevin.tian@intel.com>
change_type_range() invalidates gfn ranges to lazily change the type
of a range of gfns, and also modifies the logdirty rangesets of that
p2m. At the moment, it clips both down by the hostp2m.
While this will result in correct behavior, it's not entirely efficient,
since invalidated entries outside that range will, on fault, simply be
modified back to "empty" before faulting normally again.
Separate out the calculation of the two ranges. Keep using the
hostp2m's max_mapped_pfn to clip the logdirty ranges, but use the
current p2m's max_mapped_pfn to further clip the invalidation range
for alternate p2ms.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Signed-off-by: Razvan Cojocaru <rcojocaru@bitdefender.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Tested-by: Tamas K Lengyel <tamas@tklengyel.com>
Acked-by: George Dunlap <george.dunlap@citrix.com>
The logdirty rangesets of the altp2ms need to be kept in sync with the
hostp2m. This means when iterating through the altp2ms, we need to
use the host p2m to clip the rangeset, not the indiviual altp2m's
value.
This change also:
- Documents that the end is non-inclusive
- Calculates an "inclusive" value for the end once, rather than
open-coding the modification, and (worse) back-modifying updates so
that the calculation ends up correct
- Clarifies the logic deciding whether to call
change_entry_type_global() or change_entry_type_range()
- Handles the case where start >= hostp2m->max_mapped_pfn
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Signed-off-by: Razvan Cojocaru <rcojocaru@bitdefender.com>
Tested-by: Tamas K Lengyel <tamas@tklengyel.com>
When an new altp2m view is created very early on guest boot, the
display will freeze (although the guest will run normally). This
may also happen on resizing the display. The reason is the way
Xen currently (mis)handles logdirty VGA: it intentionally
misconfigures VGA pages so that they will fault.
The problem is that it only does this in the host p2m. Once we
switch to a new altp2m, the misconfigured entries will no longer
fault, so the display will not be updated.
This patch:
* updates ept_handle_misconfig() to use the active altp2m instead
of the hostp2m;
* modifies p2m_change_entry_type_global(),
p2m_memory_type_changed(), p2m_change_type_range() and
p2m_finish_type_change() to propagate their changes to all
valid altp2ms.
With the introduction of altp2m fields in p2m_memory_type_changed()
the whole function has been put under CONFIG_HVM.
Suggested-by: George Dunlap <george.dunlap@citrix.com>
Signed-off-by: Razvan Cojocaru <rcojocaru@bitdefender.com>
Tested-by: Tamas K Lengyel <tamas@tklengyel.com>
Acked-by: George Dunlap <george.dunlap@citrix.com>
Refactor p2m_reset_altp2m() so that it can be used to remove
redundant codepaths, fixing the locking while we're at it.
The previous code now replaced by p2m_reset_altp2m(d, i,
ALTP2M_DEACTIVATE) calls did not set p2m->min_remapped_gfn
and p2m->max_remapped_gfn because in those cases the altp2m
idx was disabled; so before getting used again,
p2m_init_altp2m_ept() would get called, which resets them.
Always setting them in p2m_reset_altp2m(), while redundant,
is preferable to an extra conditional.
Signed-off-by: Razvan Cojocaru <rcojocaru@bitdefender.com>
Tested-by: Tamas K Lengyel <tamas@tklengyel.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
For now, only do allocation/deallocation; keeping them in sync
will be done in subsequent patches.
Logdirty synchronization will only be done for active altp2ms;
so allocate logdirty rangesets (copying the host logdirty
rangeset) when an altp2m is activated, and free it when
deactivated.
Signed-off-by: Razvan Cojocaru <rcojocaru@bitdefender.com>
Tested-by: Tamas K Lengyel <tamas@tklengyel.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
Apparently older versions of gcc, when building with debug=n, the
uninitialized variable logic gets confused. Distros on which a
non-debug build will fail include:
- Centos 7
- Debian Jessie
- Ubuntu Trusty
It seems to be one particular path confusing the logic; so just set it
on that path to keep the compiler happy, while still catching other
potential paths where it might be unset.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Untrusted device models must be killed by uid rather than by pid for
safety. To do this reliably, we need another uid, not used for any
other purpose, from which to make the kill system call.
When using xen-qemuuser-range-base, we can repurpose
xen-qemuuser-range-base itself as a UID from which to kill other
devicemodel uids (since domain ID 0 should never have a device model
associated with it).
However, we'd like people to be able to use the device_model_user
feature without also defining xen-qemuuser-range-base (which requires
the ability to 'reserve' 32k+ user IDs).
To that end, introduce the xen-qemuuser-reaper id. When killing by
UID, first look for and use that ID if available; then fall back to
xen-qemuuser-range-base.
Document the new call in docs/features/qemu-deprivilege.pandoc.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Using kill(-1) to killing an untrusted dm process with the real uid
equal to the dm_uid isn't guaranteed to succeed: the process in
question may be able to kill the reaper process after the setresuid()
and before the kill().
Instead, set the real uid to the QEMU user for domain 0
(QEMU_USER_RANGE_BASE + 0). The reaper process will still be able to
kill the dm process, but not vice versa.
This, in turn, requires locking to make sure that only one reaper
process is using that uid at a time; otherwise one reaper process may
kill the other reaper process.
Create a lockfile in RUNDIR/dm-reaper-lock, and grab the lock before
executing kill.
In the event that we can't get the lock for some reason, go ahead with
the kill using dm_uid for both real and effective UIDs. This isn't
guaranteed to work, but it's no worse than not trying to kill the
process at all.
NB that this effectively requires admins using device_model_user to
also define xen_qemuuser_range_base; this will be addressed in
subsequent patches.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
The privcmd fd that a dm_restrict'ed QEMU has gives it permission to
one specific domain ID. This domain ID will probably eventually be
used again. It is therefore necessary to make absolutely sure that a
rogue QEMU process cannot hang around after its domain has exited.
Killing QEMU by pid is insufficient in this situation, because QEMU
may be able to fork() to escape killing. It is surprisingly tricky to
kill a process which can call fork() without races; the only reliable
way is to use kill(-1) to kill all processes with a given uid.
We can use this method only when we're sure that there's only one QEMU
instance per uid. Add a dm_uid into the domain_build_state struct,
and set it in libxl__domain_get_device_model_uid() when it's safe to
kill by UID. Store this in xenstore next to device-model-pid.
On domain destroy, check to see if device-model-uid is present in
xenstore. If so, fork off a reaper process, setuid to that uid, and
do kill(-9) to kill all uids of that type. Otherwise, carry on
destroying by pid.
While we're here, make libxl__destroy_device_model() consistently:
1. Return an error when anything fails
2. But continue to do as much clean-up as possible
NOTE that this is not yet completely safe: with ruid == dm_uid, the
device model may be able to kill(-9) the 'reaper' process before the
reaper process can kill it. Further patches will address this.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Or at least, give it an asynchronous interface so that we can make it
actually asynchronous in subsequent patches.
Create state structures and callback function signatures. Add the
state structure to libxl__destroy_domid_state. Break
libxl__destroy_domid down into two functions.
No functional change intended.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Removing the qmp connection is logically part of the device model
destruction; having the caller destroy it is a mild layering
violation.
Move libxl__qmp_cleanup() into libxl__destroy_device_model(). This
will make it easier when we make devicemodel destruction asynchronous.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
At the moment, we check for equivalence to literal "root" before
deciding whether to add the `runas` command-line option to QEMU. This
is unsatisfactory for several reasons.
First, just because the string doesn't match "root" doesn't mean the
final uid won't end up being zero; in particular, the range_base
calculations may end up producing "0:NNN", which would be root in any
case.
Secondly, it's almost certainly a configuration error if the resulting
uid ends up to be zero; rather than silently do what was specified but
probably not intended, throw an error.
To fix this, check for root once in
libxl__domain_get_device_model_uid. If the result is root, return an
error; if appropriate, set `runas`.
After that, assume that the presence of state->dm_runas implies that a
`runas` argument should be constructed.
One side effect of this is to check whether device_model_user exists
before passing it to qemu, resulting in better error reporting.
While we're here:
- Refactor the function to use the "goto out" idiom
- Use 'rc' rather than 'ret', in line with CODING_STYLE
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
A number of subsequent patches rely on as-yet undefined behavior for
what the `device_model_user` parameter does. Rather than implement it
incorrectly (or randomly), or remove the feature, describe an expected
usage for the feature. Further patches will make decisions based on
this expected usage.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Bring conventions more in line with libxl__xs_read_checked():
- If found, return 0 and set pointer to non-NULL
- If not found, return 0 and set pointer to NULL
- On error, return libxl-style error number.
Update documentation to match.
Use CODING_STYLE compliant `r` rather than `ret`.
On error, log the error code before returning instead of discarding
it.
Now that it only returns 0 or errno, update caller error checks to be
`if (ret)` rather than `if (ret < 0)`.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
QEMU_USER_BASE allows a user to specify the UID to use when running
the devicemodel for a specific domain number. Unfortunately, this is
not really practical: It requires nearly 32,000 entries in
/etc/passwd. QEMU_USER_RANGE_BASE is much more practical.
Remove support for QEMU_USER_BASE.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
To reliably kill an untrusted devicemodel, we need to know not only
its pid, but its uid. In preparation for this, move the userid
determination logic into a helper function.
Create a new field, `dm_runas`, in libxl__domain_build_state to store
the value during domain creation.
This change also removes unnecessary duplication of the argument
construction code.
While here, clean up some minor CODING_STYLE infractions (space
between * and variable name).
No functional change intended.
While here, delete some trailing whitespace.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Modified to provide building of the xenalyze binary for both ARM and
x86 platforms. The xenalyze binary is now built as part of the BIN
list for both platforms.
Signed-off-by: Benjamin Sanda <ben.sanda@dornerworks.com>
Signed-off-by: Julien Grall <julien.grall@arm.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
Now that we allow a privileged domain to map tracing buffer, initialize
them so a user can effectively trace Xen.
Signed-off-by: Benjamin Sanda <ben.sanda@dornerworks.com>
[julien: rework commit message]
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Andrii Anisov <andrii_anisov@epam.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
For auto-translated domain, the only way to map a page to itself is the
using the foreign map API. The current code does not allow mapping page from
special page (such as DOMID_XEN).
As xentrace buffers are shared using DOMID_XEN, it is not possible to use
tracing for Arm.
This could be solved by using the helper get_pg_owner(). This helper will
be able to get a reference on DOMID_XEN and therefore allow mapping for
privileged domain.
This patch replace the call to rcu_lock_domain_by_any_id() with
get_pg_owner(). For consistency, all the call to rcu_unlock_domain are
replaced by put_pg_owner().
Signed-off-by: Julien grall <julien.grall@arm.com>
Reviewed-by: Andrii Anisov <andrii_anisov@epam.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
DOMID_XEN is used to share pages beloging to the hypervisor
(e.g trace buffers). Unlike other domains, DOMID_XEN is a non-auto
translated domain and therefore does not have a P2M.
This patch adds a special case for DOMID_XEN in get_page_from_gfn. We
may want to provide "non-auto translated helpers" in the future if we
see more case.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Andrii Anisov <andrii_anisov@epam.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
Currently, foreign mappings can only be read-write. A follow-up patch will
extend foreign mapping for Xen backend memory (via XEN_DOMID), some of
that memory should only be read accessible for the mapping domain.
Introduce a new p2m_type to cater read-only foreign mappings. For now,
the decision between the two foreign mapping type is based on the type
of the guest page mapped.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Andrii Anisov <andrii_anisov@epam.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
A follow-up patch will introduce another type of foreign mapping. Rename
the type to make clear it is only used for read-write mapping.
No functional changes intended.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Andrii Anisov <andrii_anisov@epam.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
In a follow-up patch, we will need to handle get_page_from_gfn
differently for DOMID_XEN. To keep the code simple move the current
content in a new separate helper p2m_get_page_from_gfn.
Note the new helper is not anymore a static inline function as the helper
is quite complex.
Finally, take the opportunity to use typesafe gfn as the change is
minor.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Andrii Anisov <andrii_anisov@epam.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
get_pg_owner() and put_pg_owner() will be necessary in a follow-up
commit to support xentrace on Arm. So move the helper to common code.
Take the opportunity to clean-up a bit the code moved.
Signed-off-by: Benjamin Sanda <ben.sanda@dornerworks.com>
[julien: Rework commit title / turn put_pg_owner to a static inline]
Signed-off-by: Julien Grall <julien.grall@arm.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Note that simd_packed_fp for the opcode space 0f38 major opcodes 14 and
15 is not really correct, but sufficient for the purposes here. Further
adjustments may later be needed for the down conversion unsigned
saturating VPMOV* insns, first and foremost for the different Disp8
scaling those ones use.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
This is to better reflect that it's an abbreviation for "broadcast,
rounding, or SAE" rather than just "broadcast".
Take the opportunity and also add SDM naming comments to both union vex
and union evex.
Requested-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
And rename to have only one prefix underscore where applicable.
No functional change.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
Stop blacklisting ZynqMP's power management node. It is now possible
since we allow the hardware domain to issue HVC/SMC calls to firmware.
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Signed-off-by: Stefano Stabellini <stefanos@xilinx.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Acked-by: Julien Grall <julien.grall@arm.com>
zynqmp_eemi uses the defined functions and structs to decide whether to
make a call to the firmware, or to simply return a predefined value.
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Signed-off-by: Stefano Stabellini <stefanos@xilinx.com>
Introduce zynqmp_eemi: a function responsible for implementing access
controls over the firmware calls. Only calls that are allowed are
forwarded to the firmware.
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Signed-off-by: Stefano Stabellini <stefanos@xilinx.com>
Acked-by: Julien Grall <julien.grall@arm.com>
Introduce platform_smc as a way to handle firmware calls that Xen does
not know about in a platform specific way. This is particularly useful
for implementing the SiP (SoC implementation specific) service calls.
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Signed-off-by: Stefano Stabellini <stefanos@xilinx.com>
Acked-by: Julien Grall <julien.grall@arm.com>
At the moment, Xen is relocated towards the end of the memory. While
this has the advantage to free space in low memory, the code is not
compliant with the break-before-make because it requires to switch
between two sets of page-table. This is not entirely trivial to fix as
it would require us to go through an identity mapping and disabling MMU.
Furthermore, it looks like that some platform (such as the Hikey960)
may not be able to bring-up secondary CPUs if the entry is too high.
While Xen should be quite tiny (< 2MB), the current algorithm to
allocate Dom0 memory will allocate memory chunks of at least 128MB.
Those memory chunks will always be 128MB. This means that depending on
where the modules are loaded, an extra 128MB may disappear.
As there are up to 4 modules (initramfs, XSM, kernel, DTB) loaded in
low memory. The problem is not entirely new as you could already waste
512MB of low-memory. The right solution would be to fix the allocation
algorithm. But this is independent from this patch.
For user in control of the memory (such as in U-boot), all modules
should be loaded as much as possible together or outside low-memory (i.e
above 4GB). For other users (i.e Grub/UEFI), I believe the bootloader is
already keeping everything together.
Based on the above, it would be fine to stop relocating Xen. This has
the advantage to simplify the code and should speed-up the boot as
relocation is not necessary anymore.
Note that the break-before-make issue is not fixed by this patch.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reported-by: Matthew Daley <mattd@bugfuzz.com>
Tested-by: Matthew Daley <mattd@bugfuzz.com>
Tested-by: Andrii Anisov <andrii_anisov@epam.com>
Reviewed-by: Andrii Anisov <andrii_anisov@epam.com>
Signed-off-by: Stefano Stabellini <sstabellini@kernel.org>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
At the moment, the implementation of Set/Way operations will go through
all the entries of the guest P2M and flush them. However, this is very
expensive and may render unusable a guest OS using them.
For instance, Linux 32-bit will use Set/Way operations during secondary
CPU bring-up. As the implementation is really expensive, it may be possible
to hit the CPU bring-up timeout.
To limit the Set/Way impact, we track what pages has been of the guest
has been accessed between batch of Set/Way operations. This is done
using bit[0] (aka valid bit) of the P2M entry.
This patch adds a new per-arch helper is introduced to perform actions just
before the guest is first unpaused. This will be used to invalidate the
P2M to track access from the start of the guest.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Set/Way operations are used to perform maintenance on a given cache.
At the moment, Set/Way operations are not trapped and therefore a guest
OS will directly act on the local cache. However, a vCPU may migrate to
another pCPU in the middle of the processor. This will result to have
cache with stale data (Set/Way are not propagated) potentially causing
crash. This may be the cause of heisenbug noticed in Osstest [1].
Furthermore, Set/Way operations are not available on system cache. This
means that OS, such as Linux 32-bit, relying on those operations to
fully clean the cache before disabling MMU may break because data may
sits in system caches and not in RAM.
For more details about Set/Way, see the talk "The Art of Virtualizing
Cache Maintenance" given at Xen Summit 2018 [2].
In the context of Xen, we need to trap Set/Way operations and emulate
them. From the Arm Arm (B1.14.4 in DDI 046C.c), Set/Way operations are
difficult to virtualized. So we can assume that a guest OS using them will
suffer the consequence (i.e slowness) until developer removes all the usage
of Set/Way.
As the software is not allowed to infer the Set/Way to Physical Address
mapping, Xen will need to go through the guest P2M and clean &
invalidate all the entries mapped.
Because Set/Way happen in batch (a loop on all Set/Way of a cache), Xen
would need to go through the P2M for every instructions. This is quite
expensive and would severely impact the guest OS. The implementation is
re-using the KVM policy to limit the number of flush:
- If we trap a Set/Way operations, we enable VM trapping (i.e
HVC_EL2.TVM) to detect cache being turned on/off, and do a full
clean.
- We clean the caches when turning on and off
- Once the caches are enabled, we stop trapping VM instructions
[1] https://lists.xenproject.org/archives/html/xen-devel/2017-09/msg03191.html
[2] https://fr.slideshare.net/xen_com_mgr/virtualizing-cache
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
A follow-up patch will require to emulate some accesses to system
registers trapped by HCR_EL2.TVM. When set, all NS EL1 writes to the
virtual memory control registers will be trapped to the hypervisor.
This patch adds the infrastructure to passthrough the access to the host
registers.
Note that HCR_EL2.TVM will be set in a follow-up patch dynamically.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
A follow-up patch will require to emulate some accesses to some
co-processors registers trapped by HCR_EL2.TVM. When set, all NS EL1 writes
to the virtual memory control registers will be trapped to the hypervisor.
This patch adds the infrastructure to passthrough the access to host
registers. For convenience a bunch of helpers have been added to
generate the different helpers.
Note that HCR_EL2.TVM will be set in a follow-up patch dynamically.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
p2m_cache_flush_range does not yet support preemption, this may be an
issue as cleaning the cache can take a long time. While the current
caller (XEN_DOMCTL_cacheflush) does not stricly require preemption, this
will be necessary for new caller in a follow-up patch.
The preemption implemented is quite simple, a counter is incremented by:
- 1 on region skipped
- 10 for each page requiring a flush
When the counter reach 512 or above, we will check if preemption is
needed. If not, the counter will be reset to 0. If yes, the function
will stop, update start (to allow resuming later on) and return
-ERESTART. This allows the caller to decide how the preemption will be
done.
For now, XEN_DOMCTL_cacheflush will continue to ignore the preemption.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
The final remnanat of PVRDTSCP is that we would emulate RDTSCP even on
hardware which lacked the instruction. RDTSCP is available on almost all
64-bit x86 hardware.
Remove this emulation, drop the TSC_MODE_PVRDTSCP constant, and allow RDTSCP
in a PV guest's CPUID policy.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
With PVRDTSCP mode removed, handling of MSR_TSC_AUX can move into the common
code. Move its storage into struct vcpu_msrs (dropping the HVM-specific
msr_tsc_aux), and add an RDPID feature check as this bit also enumerates the
presence of the MSR.
Introduce cpu_has_rdpid along with the synthesized cpu_has_msr_tsc_aux to
correct the context switch paths, as MSR_TSC_AUX is enumerated by either
RDTSCP or RDPID.
Drop hvm_msr_tsc_aux() entirely, and use v->arch.msrs->tsc_aux directly.
Update hvm_load_cpu_ctxt() to check that the incoming ctxt.msr_tsc_aux isn't
out of range. In practice, no previous version of Xen ever wrote an
out-of-range value. Add MSR_TSC_AUX to the list of MSRs migrated for PV
guests, but leave the HVM path using the existing space in hvm_hw_cpu.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Brian Woods <brian.woods@amd.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Roger Pau Monne
Andrew Cooper
Razvan Cojocaru
George Dunlap
Benjamin Sanda
Julien Grall
Jan Beulich
Stefano Stabellini
Edgar E. Iglesias