Andrew Cooper [Mon, 26 Mar 2018 08:02:34 +0000 (09:02 +0100)]
x86/traps: Fix %dr6 handing in #DB handler
Most bits in %dr6 accumulate, rather than being set directly based on the
current source of #DB. Have the handler follow the manuals guidance, which
avoids leaking hypervisor debugging activities into guest context.
This is part of XSA-260 / CVE-2018-8897
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
Andrew Cooper [Tue, 8 May 2018 12:45:45 +0000 (13:45 +0100)]
x86/domain: Drop the only-written smap_check_policy infrastructure
c/s 4c5d78a10d "x86/pagewalk: Re-implement the pagetable walker" dropped the
consumer of smap_policy. Looking at c/s 31ae587e6f which introduced the
smap_check logic, it exists only to work around a bug in guest_walk_tables()
was resolved by the aformentioned commit.
Remove the unused variables and associated infrastructure.
Andrew Cooper [Tue, 20 Mar 2018 19:36:40 +0000 (19:36 +0000)]
x86/pv: Hide more EFER bits from PV guests
We don't advertise SVM in CPUID so a PV guest shouldn't be under the
impression that it can use SVM functionality, but despite this, it really
shouldn't see SVME set when reading EFER.
On Intel processors, 32bit PV guests don't see, and can't use SYSCALL.
Introduce EFER_KNOWN_MASK to whitelist the features Xen knows about, and use
this to clamp the guests view.
Take the opportunity to reuse the mask to simplify svm_vmcb_isvalid(), and
change "undefined" to "unknown" in the print message, as there is at least
EFER.TCE (Translation Cache Extension) defined but unknown to Xen.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Jan Beulich [Mon, 7 May 2018 07:12:16 +0000 (09:12 +0200)]
SVM: introduce a VM entry helper
Neither the register values copying nor the trace entry generation need
doing in assembly. The VMLOAD invocation can also be further deferred
(and centralized). Therefore replace the svm_asid_handle_vmrun()
invocation with one of the new helper.
Similarly move the VM exit side register value copying into
svm_vmexit_handler().
Now that we always make it out to guest context after VMLOAD,
svm_sync_vmcb() no longer overrides vmcb_needs_vmsave, making
svm_vmexit_handler() setting the field early unnecessary.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Jan Beulich [Mon, 7 May 2018 07:11:15 +0000 (09:11 +0200)]
SVM: re-work VMCB sync-ing
While the main problem to be addressed here is the issue of what so far
was named "vmcb_in_sync" starting out with the wrong value (should have
been true instead of false, to prevent performing a VMSAVE without ever
having VMLOADed the vCPU's state), go a step further and make the
sync-ed state a tristate: CPU and memory may be in sync or an update
may be required in either direction. Rename the field and introduce an
enum. Callers of svm_sync_vmcb() now indicate the intended new state
(with a slight "anomaly" when requesting VMLOAD: we could store
vmcb_needs_vmsave in those cases as the callers request, but the VMCB
really is in sync at that point, and hence there's no need to VMSAVE in
case we don't make it out to guest context), and all syncing goes
through that function.
With that, there's no need to VMLOAD the state perhaps multiple times;
all that's needed is loading it once before VM entry.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Avoid flushing the complete TLB when switching %cr3 for mitigation of
Meltdown by using the PCID feature if available.
We are using 4 PCID values for a 64 bit pv domain subject to XPTI and
2 values for the non-XPTI case:
- guest active and in kernel mode
- guest active and in user mode
- hypervisor active and guest in user mode (XPTI only)
- hypervisor active and guest in kernel mode (XPTI only)
We use PCID only if PCID _and_ INVPCID are supported. With PCID in use
we disable global pages in cr4. A command line parameter controls in
which cases PCID is being used.
As the non-XPTI case has shown not to perform better with PCID at least
on some machines the default is to use PCID only for domains subject to
XPTI.
With PCID enabled we always disable global pages. This avoids having to
either flush the complete TLB or do a cycle through all PCID values
when invalidating a single global page.
Signed-off-by: Juergen Gross <jgross@suse.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
xen/x86: use flag byte for decision whether xen_cr3 is valid
Today cpu_info->xen_cr3 is either 0 to indicate %cr3 doesn't need to
be switched on entry to Xen, or negative for keeping the value while
indicating not to restore %cr3, or positive in case %cr3 is to be
restored.
Switch to use a flag byte instead of a negative xen_cr3 value in order
to allow %cr3 values with the high bit set in case we want to keep TLB
entries when using the PCID feature.
This reduces the number of branches in interrupt handling and results
in better performance (e.g. parallel make of the Xen hypervisor on my
system was using about 3% less system time).
Signed-off-by: Juergen Gross <jgross@suse.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
xen/x86: disable global pages for domains with XPTI active
Instead of flushing the TLB from global pages when switching address
spaces with XPTI being active just disable global pages via %cr4
completely when a domain subject to XPTI is active. This avoids the
need for extra TLB flushes as loading %cr3 will remove all TLB
entries.
In order to avoid states with cr3/cr4 having inconsistent values
(e.g. global pages being activated while cr3 already specifies a XPTI
address space) move loading of the new cr4 value to write_ptbase()
(actually to switch_cr3_cr4() called by write_ptbase()).
This requires to use switch_cr3_cr4() instead of write_ptbase() when
building dom0 in order to avoid setting cr4 with cr4.smap set.
Signed-off-by: Juergen Gross <jgross@suse.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
Instead of switching XPTI globally on or off add a per-domain flag for
that purpose. This allows to modify the xpti boot parameter to support
running dom0 without Meltdown mitigations. Using "xpti=no-dom0" as boot
parameter will achieve that.
Move the xpti boot parameter handling to xen/arch/x86/pv/domain.c as
it is pv-domain specific.
Signed-off-by: Juergen Gross <jgross@suse.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
x86/xpti: avoid copying L4 page table contents when possible
For mitigation of Meltdown the current L4 page table is copied to the
cpu local root page table each time a 64 bit pv guest is entered.
Copying can be avoided in cases where the guest L4 page table hasn't
been modified while running the hypervisor, e.g. when handling
interrupts or any hypercall not modifying the L4 page table or %cr3.
So add a per-cpu flag indicating whether the copying should be
performed and set that flag only when loading a new %cr3 or modifying
the L4 page table. This includes synchronization of the cpu local
root page table with other cpus, so add a special synchronization flag
for that case.
A simple performance check (compiling the hypervisor via "make -j 4")
in dom0 with 4 vcpus shows a significant improvement:
- real time drops from 112 seconds to 103 seconds
- system time drops from 142 seconds to 131 seconds
Signed-off-by: Juergen Gross <jgross@suse.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
Jan Beulich [Thu, 3 May 2018 15:35:51 +0000 (17:35 +0200)]
x86: fix return value checks of set_guest_{machinecheck,nmi}_trapbounce
Commit 0142064421 ("x86/traps: move set_guest_{machine,nmi}_trapbounce")
converted the functions' return types from int to bool without also
correcting the checks in assembly code: The ABI does not guarantee sub-
32-bit return values to be promoted to 32 bits.
Take the liberty and also adjust the number of spaces used in the compat
code, such that both code sequences end up similar (they already are in
the non-compat case).
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
George Dunlap [Tue, 1 May 2018 17:13:27 +0000 (18:13 +0100)]
xen/schedule: Fix races in vcpu migration
The current sequence to initiate vcpu migration is inefficent and error-prone:
- The initiator sets VPF_migraging with the lock held, then drops the
lock and calls vcpu_sleep_nosync(), which immediately grabs the lock
again
- A number of places unnecessarily check for v->pause_flags in between
those two
- Every call to vcpu_migrate() must be prefaced with
vcpu_sleep_nosync() or introduce a race condition; this code
duplication is error-prone
- In the event that v->is_running is true at the beginning of
vcpu_migrate(), it's almost certain that vcpu_migrate() will end up
being called in context_switch() as well; we might as well simply
let it run there and save the duplicated effort (which will be
non-negligible).
The result is that Credit1 has several races which result in runqueue
<-> v->processor invariants being violated (triggering ASSERTs in
debug builds and strange bugs in production builds).
Instead, introduce vcpu_migrate_start() to initiate the process.
vcpu_migrate_start() is called with the scheduling lock held. It not
only sets VPF_migrating, but also calls vcpu_sleep_nosync_locked()
(which will automatically do nothing if there's nothing to do).
Rename vcpu_migrate() to vcpu_migrate_finish(). Check for v->is_running and
pause_flags & VPF_migrating at the top and return if appropriate.
Then the way to initiate migration is consistently:
George Dunlap [Wed, 2 May 2018 10:09:18 +0000 (11:09 +0100)]
xen: Introduce vcpu_sleep_nosync_locked()
There are a lot of places which release a lock before calling
vcpu_sleep_nosync(), which then just grabs the lock again. This is
not only a waste of time, but leads to more code duplication (since
you have to copy-and-paste recipes rather than calling a unified
function), which in turn leads to an increased chance of bugs.
Introduce vcpu_sleep_nosync_locked(), which can be called if you
already hold the schedule lock.
Lars Kurth [Tue, 1 May 2018 08:03:13 +0000 (09:03 +0100)]
MAINTAINERS: Add Brian Woods as Designated reviewer to AMD IOMMU and AMD SVM
This was discussed in an IRC discussion post the April x86 meeting.
On 27/4/18 Juergen gave a RAB via IRC
Cc: Lars Kurth <lars.kurth@citrix.com> Cc: Andrew Cooper <andrew.cooper3@citrix.com> Cc: George Dunlap <George.Dunlap@eu.citrix.com> Cc: Ian Jackson <ian.jackson@eu.citrix.com> Cc: Jan Beulich <jbeulich@suse.com> Cc: Julien Grall <julien.grall@arm.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Stefano Stabellini <sstabellini@kernel.org> Cc: Tim Deegan <tim@xen.org> Cc: Wei Liu <wei.liu2@citrix.com> Cc: Brian Woods <brian.woods@amd.com> Cc: Juergen Gross <jgross@suse.com> Signed-off-by: Lars Kurth <lars.kurth@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com> Acked-by: Brian Woods <brian.woods@amd.com> Acked-by: Wei Liu <wei.liu2@citrix.com> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Lars Kurth [Tue, 1 May 2018 08:03:12 +0000 (09:03 +0100)]
MAINTAINERS, get_maintainer.pl: Add Designated Reviewer (R:) role
The syntax has been copied from the Linux Maintainers file. I moved the following Linux
get_maintainer.pl patches to Xen, fixing up some merge issues (and a bug).
I also removed code related to
P: Person (obsolete)
which is in the Linux MAINTAINER's file, but not ours. I may not have
caught all instances though.
I have tested on a number of files using mock entries in MAINTAINERS
using ./scripts/get_maintainer.pl -f ...
I also tested --nor to disable the support and it worked as expected.
Cc: Lars Kurth <lars.kurth@citrix.com> Cc: Andrew Cooper <andrew.cooper3@citrix.com> Cc: George Dunlap <George.Dunlap@eu.citrix.com> Cc: Ian Jackson <ian.jackson@eu.citrix.com> Cc: Jan Beulich <jbeulich@suse.com> Cc: Julien Grall <julien.grall@arm.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Stefano Stabellini <sstabellini@kernel.org> Cc: Tim Deegan <tim@xen.org> Cc: Wei Liu <wei.liu2@citrix.com> Cc: Juergen Gross <jgross@suse.com> Signed-off-by: Lars Kurth <lars.kurth@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com> Acked-by: Wei Liu <wei.liu2@citrix.com> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
tools: prepend to PKG_CONFIG_PATH when configuring qemu
A user may choose to set his/her own PKG_CONFIG_PATH, which is useful in the
case of cross-compiling. We don't want to completely override the
PKG_CONFIG_PATH, just add to it.
Signed-off-by: Stewart Hildebrand <stewart.hildebrand@dornerworks.com> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Ian Jackson [Wed, 13 Dec 2017 12:14:28 +0000 (12:14 +0000)]
docs/process/release-checklist.txt: Say to push staging branch
Preparing a real release, not just an RC, involves making commits.
Typically, those will be on staging-$x. The tag will refer to them,
and the checklist already says to push them to xenbits.
But if the *branch* is not pushed, then people who just "git fetch"
won't get the tag because it refers to commits they don't have.
(Because of the strange rules git has about tag fetching.)
Worse, the same may be true of people who "git clone".
And anyway, those commits *should* be fed to staging-$x.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Acked-by: George Dunlap <george.dunlap@citrix.com>
Andrew Cooper [Thu, 22 Mar 2018 11:31:07 +0000 (11:31 +0000)]
x86/traps: Improve code generation for set_ist()
The IST field in an IDT entry is a 3 bit field, with 5 adjacent reserved bits
which we always write as zero. By expressing this as a byte field in a union,
we turn an invocation of enable_each_ist() from
While making changes here, tidy up the set_ist() declaration. Drop the
always_inline (I don't recall why I wrote it like that originally) and the ist
parameter need not be unsigned long (although it will be const-propagated in
practice).
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Juergen Gross <jgross@suse.com>
That is, Dom0's ACPI processor driver is in the process of uploading Px
and Cx data. Looking at the ticket lock state in CPU1's registers, it is
waiting for ticket 0x0000 to have its turn, while the supposed current
owner's ticket is 0x0001, which is an invalid state (and neither of the
other two CPUs holds the lock anyway). Hence I can only conclude that
cpuidle_init_cpu(1) ran on CPU 0 while some other CPU held the lock (the
unlock then put the lock in the state that CPU1 is observing).
Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Jan Beulich [Thu, 26 Apr 2018 16:02:37 +0000 (18:02 +0200)]
x86emul: adjust handling of AVX2 gathers
HVM's MMIO cache only has a capacity of three entries. Once running out
of entries, hvmemul_linear_mmio_access() will return
X86EMUL_UNHANDLEABLE. Since gathers are an iterative process anyway,
simply commit the portion of work done in this and hypothetical similar
cases, exiting back to guest context for the insn to be retried.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Ian Jackson [Thu, 26 Apr 2018 10:49:27 +0000 (11:49 +0100)]
docs/process/release-checklist.txt
Clarify what is expected of the release technician for SUPPORT.md: fix
the version number field. The support dates will be set by the
release manageer.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
This contains instructions and shell runes for the Release Technician,
who is the person doing the technical admin to construct the release,
fork branches, make tags and tarballs, etc.
Rename it to make this clearer.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Ian Jackson [Wed, 25 Apr 2018 12:48:27 +0000 (13:48 +0100)]
docs/parse-support-md: Do caveats properly (!)
Each document has its own objects in @insections. Only the first
RealSect encountered ends up in the main $toplevel_sectlist tree.
This means that trying to unify the Caveats information for all
version in the RealSect (the $insection) does not work. The caveats
for all versions that aren't the first one where this section was seen
end up in the @insections array during parsing of that version, but
not recorded in the main tree.
The result was that footnotes would only appear in the output for
versions which were the most recent version where that feature row or
category appeared. The other footnotes would be lost.
Instead, store HasCaveat in the sectnode. That means ri_Para needs to
find the sectnode.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Jan Beulich [Wed, 25 Apr 2018 12:39:41 +0000 (14:39 +0200)]
x86: fix slow int80 path after XPTI additions
For the int80 slow path to jump to handle_exception_saved, %r14 needs to
be set up suitably for XPTI purposes. This is because of the difference
in nature between the int80 path (which is synchronous WRT guest
actions) and the exception path which is potentially asynchronous.
This is XSA-259.
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Anthony PERARD [Thu, 8 Mar 2018 18:16:41 +0000 (18:16 +0000)]
libxl: Specify format of inserted cdrom
Without this extra parameter on the QMP command, QEMU will guess the
format of the new file.
This is XSA-258.
Reported-by: Anthony PERARD <anthony.perard@citrix.com> Signed-off-by: Anthony PERARD <anthony.perard@citrix.com> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Andrew Cooper [Tue, 24 Apr 2018 16:08:18 +0000 (17:08 +0100)]
x86/SVM: Fix intercepted {RD,WR}MSR for the SYS{CALL,ENTER} MSRs
By default, the SYSCALL MSRs are not intercepted, and accesses are completed
by hardware. The SYSENTER MSRs are intercepted for cross-vendor
purposes (albeit needlessly in the common case), and are fully emulated.
However, {RD,WR}MSR instructions which happen to be emulated (FEP,
introspection, or older versions of Xen which intercepted #UD), or when the
MSRs are explicitly intercepted (introspection), will be completed
incorrectly.
svm_msr_read_intercept() appears to return the correct values, but only
because of the default read-everything case (which is going to disappear), and
that in vcpu context, hardware should have the guest values in context.
Update the read path to explicitly sync the VMCB and complete the accesses,
rather than falling all the way through to the default case.
svm_msr_write_intercept() silently discard all updates. Synchronise the VMCB
for all applicable MSRs, and implement suitable checks. The actual behaviour
of AMD hardware is to truncate the SYSENTER and SFMASK MSRs at 32 bits, but
this isn't implemented yet to remain compatible with the cross-vendor case.
Drop one bit of trailing whitespace while modifing this area of the code.
Reported-by: Razvan Cojocaru <rcojocaru@bitdefender.com> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Andrew Cooper [Tue, 24 Apr 2018 09:44:02 +0000 (10:44 +0100)]
x86/spec_ctrl: Fix typo in ARCH_CAPS decode
Reported-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Acked-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Juergen Gross <jgross@suse.com>
When entering the hypervisor via the double fault handler resetting
xen_cr3 was missing. This led to switching to pv_cr3 when returning
from the next following exception, so repair this in order to allow
exception handling to work even after a double fault.
Signed-off-by: Juergen Gross <jgross@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com> Tested-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
This is wrong because ACPI regions overlap with RAM regions. The cause
of this issue is not setting a big enough MMIO hole and marking the
whole MMIO hole as reserved, when it actually contains several pieces:
- local APIC page.
- ACPI tables.
- HVM special pages.
Of those items only HVM special pages need to be marked as reserved in
order to advise the guest against using them for example for memory
hotplug.
After the fix the layout reported for the same guest is:
Ian Jackson [Tue, 17 Apr 2018 13:34:36 +0000 (14:34 +0100)]
docs/parse-support-md: Correctly handle footnotes for non-leaf sections
Non-leaf sections with footnotes must have a row of their own, for
just that section, because footnotes only appear if there is status
information.
In that case, the footnote applies to only the rows for that section
in the markdown document, ie that RealSect.
And of course for a leaf section that is true too.
So for footnoes we always want to use a rowspan of the number of
Status elements in the section. So (i) calculate this in
count_rows_sectlist and (ii) use it, instead of the total number of
rows including all the subsections', when writing out the footnote
ref.
This bug has been present in this script since the beginning.
Also, while we're here, suppress the rowspan if it would be 1.
Reported-by: Lars Kurth <lars.kurth@citrix.com> Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Ian Jackson [Thu, 12 Apr 2018 16:57:58 +0000 (17:57 +0100)]
SUPPORT.md, support matrix: Treat commentary before status as description
Running text in feature sections in the markdown document currently
might be (i) a caveat, qualifying or clarifying the support statement
(ii) a plain description of the feature.
Caveats can be version-specific and deserve the [*] annotation in the
relevant feature matrix cell. They must link to SUPPORT.html for the
specific version.
Descriptions are not version specific. In that case the [*]
annotation is visusal noise. Rather, it is better to make a hyperlink
out of the text which is being expanded on. The hyperlink can point
to any appropriate version.
There is a question about how to notate this distinction in
SUPPORT.md. After IRL discussion with George and Lars I propose that
we should put text which helps describe a feature (ie, which expands
on a section heading) after the heading but before the Status
indications; whereas, caveats and supplementary information about
the actual status, should follow the Status block.
This patch implements this distinction in the support matrix
generator. Only paragraphs containing _only_ italic content count as
descriptive; anything else is treated as a caveat.
In the code:
* Add a new entry to RealSect, HasDescription
* When parsing, track whether we are before or after the first Status
block in a new variable $has_feature.
* In ri_Para, set HasDescription set to the input document index
when we encounter text before the first feature.
* When writing a `heading' (ie, the table cell for a feature name)
look for HasDescription and make an appropriate hyperlink.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Jan Beulich [Mon, 23 Apr 2018 09:01:09 +0000 (11:01 +0200)]
x86/HVM: never retain emulated insn cache when exiting back to guest
Commit 5fcb26e69e ("x86/HVM: don't retain emulated insn cache when
exiting back to guest") didn't go quite far enough: The insn emulator
may itself decide to return X86EMUL_RETRY (currently for certain
CMPXCHG failures and AVX2 gather insns), in which case we'd also exit
back to guest context. Tie the caching to whether we have an I/O
completion pending, instead of x86_emulate()'s return value.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Paul Durrant <paul.durrant@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
CPUs may share an in-use channel. Hence clearing of a bit from the
cpumask (in hpet_broadcast_exit()) as well as setting one (in
hpet_broadcast_enter()) must not race evaluation of that same cpumask.
Therefore avoid evaluating the cpumask twice in hpet_detach_channel().
Otherwise cpumask_empty() may e.g.return false while the subsequent
cpumask_first() could return nr_cpu_ids, which then triggers the
assertion in cpumask_of() reached through set_channel_irq_affinity().
Signed-off-by: David Wang <davidwang@zhaoxin.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Juergen Gross <jgross@suse.com>
The 'RSB Alternative' bit in MSR_ARCH_CAPABILITIES may be set by a hypervisor
to indicate that the virtual machine may migrate to a processor which isn't
retpoline-safe. Introduce a shortened name (to reduce code volume), treat it
as authorative in retpoline_safe(), and print its value along with the other
ARCH_CAPS bits.
The exact processor models which do have RSB semantics which fall back to BTB
predictions are enumerated, and include Kabylake and Coffeelake. Leave a
printk() in the default case to help identify cases which aren't covered.
The exact microcode versions from Broadwell RSB-safety are taken from the
referenced microcode update file (adjusting for the known-bad microcode
versions). Despite the exact wording of the text, it is only Broadwell
processors which need a microcode check.
In practice, this means that all Broadwell hardware with up-to-date microcode
will use retpoline in preference to IBRS, which will be a performance
improvement for desktop and server systems which would previously always opt
for IBRS over retpoline.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Current interface to the gntdev in FreeBSD is wrong, and mostly worked
out of luck before the PTI FreeBSD fixes, when kernel and user-space
where sharing the same page tables.
On FreeBSD ioctls have the size of the passed struct encoded in the ioctl
number, because the generic ioctl handler in the OS takes care of
copying the data from user-space to kernel space, and then calls the
device specific ioctl handler. Thus using ioctl structs with variable
sizes is not possible.
The fix is to turn the array of structs at the end of
ioctl_gntdev_alloc_gref and ioctl_gntdev_map_grant_ref into pointers,
that can be properly accessed from the kernel gntdev driver using the
copyin/copyout functions. Note that this is exactly how it's done for
the privcmd driver.
Andrew Cooper [Tue, 17 Apr 2018 17:43:49 +0000 (18:43 +0100)]
x86: Use spec_ctrl_{enter,exit}_idle() in the S3/S5 path
The main purpose of this patch is to avoid opencoding the recovery logic at
the end, but also has the positive side effect of relaxing the SPEC_CTRL
mitigations when working to shut the final CPU down.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Jan Beulich [Wed, 18 Apr 2018 09:16:37 +0000 (11:16 +0200)]
x86/msr: further correct the emulation behaviour of MSR_PRED_CMD
Following commit a6aa678fa3 ("x86/msr: Correct the emulation behaviour
of MSR_PRED_CMD") we may end up writing the low bit with the wrong
value. While it's unlikely for a guest to want to write zero there, we
should still permit (this without incurring the overhead of an actual
barrier). Correcting this right away will also help whenever further
bits in the MSR might become defined.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Ian Jackson [Tue, 17 Apr 2018 16:53:01 +0000 (17:53 +0100)]
mktarball: For qemu upstream, use their scripts/archive-source.sh
qemu upstream uses git submodules. git archive does not work with git
submodules (and could not work properly with them, because this is one
of the many things it is inherently impossible to do correctly with
git submodules).
qemu upstream have worked around this by providing a rather scary
shell script which attempts to do roughly the right thing. It's close
enough that we can use it with only minor precautions.
Unfortunately this does mean that `mktarball' now executes the qemu
source code it was using, rather than merely shuffling it about, as it
did previously. I think this is a less bad ill than copying (and,
effectively, forking) the scary script.
CC: Wei Liu <wei.liu2@citrix.com> CC: George Dunlap <george.dunlap@eu.citrix.com> CC: Juergen Gross <jgross@suse.com> Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Acked-by: George Dunlap <george.dunlap@citrix.com>
Andrew Cooper [Fri, 23 Mar 2018 20:26:34 +0000 (20:26 +0000)]
x86/traps: Misc non-functional improvements to set_debugreg()
* Change 'int i' to being unsigned, and move it into its most narrow scope.
* Fold the access_ok() checks for %dr{0..3}. This halves the compiled size
of the function.
* Additional newlines in appropriate places.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Andrew Cooper [Fri, 23 Mar 2018 20:26:34 +0000 (20:26 +0000)]
x86/pv: Introduce and use x86emul_write_dr()
set_debugreg() has several bugs:
* %dr4/5 should function correctly as aliases of %dr6/7 when CR4.DE is clear.
* Attempting to set the upper 32 bits of %dr6/7 should fail with #GP[0]
rather than be silently corrected and complete.
* For emulation, the #UD and #GP[0] cases need properly distinguishing. Use
-ENODEV for #UD cases, leaving -EINVAL (bad bits) and -EPERM (not allowed to
use that valid bit) as before for hypercall callers.
* A write which clears %dr7.L/G leaves the IO shadow intact, meaning that
subsequent reads of %dr7 will see stale IO watchpoint configuration.
Implement x86emul_write_dr() as a thin wrapper around set_debugreg().
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Andrew Cooper [Fri, 23 Mar 2018 20:13:50 +0000 (20:13 +0000)]
x86/pv: Introduce and use x86emul_read_dr()
do_get_debugreg() has several bugs:
* The %cr4.de condition is inverted. %dr4/5 should be accessible only when
%cr4.de is disabled.
* When %cr4.de is disabled, emulation should yield #UD rather than complete
with zero.
* Using -EINVAL for errors is a broken ABI, as it overlaps with valid values
near the top of the address space.
Introduce a common x86emul_read_dr() handler (as we will eventually want to
add HVM support) which separates its success/failure indication from the data
value, and have do_get_debugreg() call into the handler.
The ABI of do_get_debugreg() remains broken, but switches from -EINVAL to
-ENODEV for compatibility with the changes in the following patch.
Take the opportunity to add a missing local variable block to x86_emulate.c
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Acked-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Andrew Cooper [Mon, 16 Apr 2018 10:56:00 +0000 (10:56 +0000)]
x86/msr: Correct the emulation behaviour of MSR_PRED_CMD
Experimentally, the behaviour of reserved bits in MSR_PRED_CMD changed between
beta and production microcode, and now raises a #GP fault for set reserved
bits. The AMD spec for future hardware also specifies this behaviour, and it
is the more sensible behaviour to implement.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Juergen Gross <jgross@suse.com>
The "end" address must be rounded down before shifting,
otherwise we will insert wrong page range to a heap if address isn't
page aligned.
It seems that a copy-paste mistake took place in the following commit: 0c12972e34b20a26f2b42044b98bf12db7ed62b6
xen/mm: Switch some of page_alloc.c to typesafe MFN
Jan Beulich [Mon, 16 Apr 2018 12:10:33 +0000 (14:10 +0200)]
x86: check feature flags after resume
Make sure no previously present features are missing after resume (and
the re-loading of microcode), to avoid later crashes or (likely silent)
hangs / live locks. This doesn't go beyond checking x86_capability[],
but this should be good enough for the immediate need of making sure
that the BIT mitigation MSRs are still available.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Jan Beulich [Mon, 16 Apr 2018 12:09:55 +0000 (14:09 +0200)]
x86: suppress BTI mitigations around S3 suspend/resume
NMI and #MC can occur at any time after S3 resume, yet the MSR_SPEC_CTRL
may become available only once we're reloaded microcode. Make
SPEC_CTRL_ENTRY_FROM_INTR_IST and DO_SPEC_CTRL_EXIT_TO_XEN no-ops for
the critical period of time.
Also set the MSR back to its intended value.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Jan Beulich [Mon, 16 Apr 2018 12:08:30 +0000 (14:08 +0200)]
x86: correct ordering of operations during S3 resume
Microcode loading needs to happen before re-enabling interrupts, in case
only updated microcode allows the use of e.g. the SPEC_{CTRL,CMD} MSRs.
Otoh it doesn't need to happen at all when we didn't suspend in the
first place. It needs to happen before spin_debug_enable() though, as it
acquires a lock and hence would otherwise make
common/spinlock.c:check_lock() unhappy. As micrcode loading can be
pretty verbose, also make sure it only runs after console_end_sync().
cpufreq_add_cpu() doesn't need calling on the only "goto enable_cpu"
path, which sits ahead of cpufreq_del_cpu().
Reported-by: Simon Gaiser <simon@invisiblethingslab.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Ian Jackson [Fri, 13 Apr 2018 13:55:27 +0000 (14:55 +0100)]
docs/gen-html-index: Make HTML::TreeBuilder::XPath optional again
7782db9260d4 "docs/gen-html-index: Extract titles from HTML documents"
requires HTML::TreeBuilder::XPath.
This is sadly not as widely available as I had hoped. Work around
this problem by making the use of this module optional: instead of
`use'ing at the toplevel, we `require' it in the eval. If it's not
present, then the title is simply not extracted and the filename is
used as before, which is tolerable.
Also add some debugging.
Reported-by: Doug Goldstein <cardoe@cardoe.com> Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Reviewed-by: Doug Goldstein <cardoe@cardoe.com> Tested-by: Doug Goldstein <cardoe@cardoe.com>
sndif: Add explicit back and front parameter negotiation
In order to provide explicit stream parameter negotiation between
backend and frontend the following changes are introduced in the protocol:
add XENSND_OP_HW_PARAM_QUERY request to read/update
configuration space for the parameter given: request passes
desired parameter interval (mask) and the response to this request
returns min/max interval (mask) for the parameter to be used.
Parameters supported by this request/response:
- format mask
- sample rate interval
- number of channels interval
- buffer size, interval, frames
- period size, interval, frames
sndif: Add explicit back and front synchronization
In order to provide explicit synchronization between backend and
frontend the following changes are introduced in the protocol:
- add new ring buffer for sending asynchronous events from
backend to frontend to report number of bytes played by the
frontend (XENSND_EVT_CUR_POS)
- introduce trigger events for playback control: start/stop/pause/resume
- add "req-" prefix to event-channel and ring-ref to unify naming
of the Xen event channels for requests and events
Ian Jackson [Thu, 12 Apr 2018 11:59:24 +0000 (12:59 +0100)]
docs/parse-support-md: Unify identical [*] in footnotes
A section in the SUPPORT.md may mention multiple
Status, something: Supported
and then have some text. The text is linked to from [*] footnotes
in the table. But, this means that each bit of text needs to
apply to multiple rows.
Before this commit this was a separate [*] after each applicable item.
But multiple apparently-different links to the same thing are annoying
for the reader.
So, in this commit we combine them. Formatting the result is not
entirely trivial.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
---
v3: New patch
v3.1: Drop `}' in multi-row [*] notes. I put this in to help work
around firefox bugs eg
https://bugzilla.mozilla.org/show_bug.cgi?id=244135
but I have been convinced it is not generally wanted.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Ian Jackson [Wed, 11 Apr 2018 10:42:27 +0000 (11:42 +0100)]
docs: Provide support-matrix-generate, to generate a support matrix in HTML
This archaeology script:
- figures out what the current and previous Xen versions were
- looks for appropriate git branches for them
- finds SUPPORT.md for each one
- feeds its findings to parse-support-md
We do not intend to integrate this into docs/Makefile, because it
relies on the git history. Instead, we will take the rune provided in
the head comment and paste a variant of it into an appropriate cronjob
on xenbits.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com> Acked-by: Lars Kurth <lars.kurth@citrix.com>
---
v3: Provide -D option.
Ian Jackson [Thu, 5 Apr 2018 17:12:21 +0000 (18:12 +0100)]
docs: Provide parse-support-md
This utility reads json format pandoc output, from parsing one or more
SUPPORT.md files, and generates an HTML table element containing the
principal version and feature information.
This is rather hairier than I anticipated when I started out; hence
the 400-odd-line Perl script.
Machinery to assemble the appropriate inputs for parse-support-md
will be in the next commit.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com> Acked-by: Lars Kurth <lars.kurth@citrix.com>
---
v2: New in this version of the series.
v3: Refactor to introduce RealSect
v3: Add [*] footnote to all applicable entries, not just the last
Ian Jackson [Fri, 6 Apr 2018 17:12:37 +0000 (18:12 +0100)]
docs/Makefile: Introduce GENERATE_PANDOC_RULE_RAW
We are going to want to format SUPPORT.md which does not match the
filename patterns in docs/. So provide a way to make an ad-hoc rule
using pandoc with the standard options.
No functional change in this patch.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com> Acked-by: Lars Kurth <lars.kurth@citrix.com>
Ian Jackson [Fri, 6 Apr 2018 17:16:35 +0000 (18:16 +0100)]
SUPPORT.md: Syntax: Provide a title rather than a spurious empty section
This commits (more or less) this file to be processed with pandoc,
rather than other markdown processors. There is, unfortunately, no
widely-accepted way to declare a title for the document.
I tested feeding the document to markdown(1) on Debian jessie and it
reproduced the % line as if it were simple text. I guess many other
markdown processors will do something similarly tolerable. My
internet searches did not discover a markdown processor that used
lines starting with % for something else.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com> Acked-by: George Dunlap <george.dunlap@citrix.com> Acked-by: Lars Kurth <lars.kurth@citrix.com>
Ian Jackson [Fri, 6 Apr 2018 14:20:22 +0000 (15:20 +0100)]
SUPPORT.md: Syntax: Fix a typo "States"
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com> Acked-by: George Dunlap <george.dunlap@citrix.com> Acked-by: Lars Kurth <lars.kurth@citrix.com>
Ian Jackson [Thu, 5 Apr 2018 16:19:31 +0000 (17:19 +0100)]
SUPPORT.md: Syntax: Fix some bullet lists
Continuations of bullet list items must be indented by exactly 4
spaces (according to pandoc_markdown(5) on Debian jessie).
This is most easily achieved by making the bullet list items have two
spaces before the `*'.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com> Acked-by: George Dunlap <george.dunlap@citrix.com> Acked-by: Lars Kurth <lars.kurth@citrix.com>
Jan Beulich [Wed, 11 Apr 2018 08:42:24 +0000 (10:42 +0200)]
x86/HVM: suppress I/O completion for port output
We don't break up port requests in case they cross emulation entity
boundaries, and a write to an I/O port is necessarily the last
operation of an instruction instance, so there's no need to re-invoke
the full emulation path upon receiving the result from an external
emulator.
In case we want to properly split port accesses in the future, this
change will need to be reverted, as it would prevent things working
correctly when e.g. the first part needs to go to an external emulator,
while the second part is to be handled internally.
While this addresses the reported problem of Windows paging out the
buffer underneath an in-process REP OUTS, it does not address the wider
problem of the re-issued insn (to the insn emulator) being prone to
raise an exception (#PF) during a replayed, previously successful memory
access (we only record prior MMIO accesses).
Leaving aside the problem tried to be worked around here, I think the
performance aspect alone is a good reason to change the behavior.
Also take the opportunity and change bool_t -> bool as
hvm_vcpu_io_need_completion()'s return type.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Paul Durrant <paul.durrant@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
c/s 74fd984ae "tools/libxl: Drop xc_domain_configuration_t from
libxl__domain_build_state" removed state->config completely but missed
some conversion libxl_arm.c.
Furthermore, not all the fields of xc_domain_configuration_t have a
corresponding field in libxl_domain_build_info. This is the case of
clock_frequency. As the field should not be exposed to the user, add a
corresponding field in libxl__domain_build_state. This require some
modification in the prototype of libxl__domain_make in order to have the
state.
For all the other fields, use the up-to-date version in
libxl_domain_build_info.
Boris Ostrovsky [Mon, 9 Apr 2018 14:24:59 +0000 (10:24 -0400)]
x86/PVH/libxl: Check whether Linux guest can handle RSDP at 4G boundary
Commit 4a5733771e6f ("libxl: put RSDP for PVH guest near 4GB") breaks
pre-4.17 Linux guests since they do not use start_info's rsdp_paddr
pointer and instread scan BIOS memory for RSDP signature.
Introduce XENFEAT_linux_rsdp_unrestricted feature flag that indicates
whether the guest can handle RSDP at locations pointed to by
rsdp_paddr.
Since only Linux PVH guests suffer from this problem (BSD has always
relied on rsdp_paddr) we check this flag just for those guests. If the
flag is not set we place RSDP in BIOS, as before.
Andrew Cooper [Mon, 9 Apr 2018 09:39:32 +0000 (09:39 +0000)]
x86/pv: Fix up erroneous segments for 32bit syscall entry
The existing FLAT_KERNEL_SS expands to the correct value, 0xe02b, but is the
wrong constant to use. Switch to FLAT_USER_SS32.
For compat domains however, the reported values are entirely bogus.
FLAT_USER_SS32 (value 0xe02b) is FLAT_RING3_CS in the 32bit ABI, while
FLAT_USER_CS32 (value 0xe023) is FLAT_RING1_DS with an RPL of 3.
The guests SYSCALL callback is invoked with a broken iret frame, and if left
unmodified by the guest, will fail on the way back out when Xen's iret tries
to load a code segment into %ss.
In practice, this is only a problem for 32bit PV guests on AMD hardware, as
Intel hardware doesn't permit the SYSCALL instruction outside of 64bit mode.
This appears to have been broken ever since 64bit support was added to Xen,
and has gone unnoticed because Linux doesn't use SYSCALL in 32bit builds.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Wei Liu <wei.liu2@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Drop the _mfn() wrappers now that page_to_mfn() returns the correct type.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Paul Durrant <paul.durrant@citrix.com> Release-acked-by: Juergen Gross <jgross@suse.com>
Julien Grall [Wed, 21 Feb 2018 13:46:27 +0000 (13:46 +0000)]
xen: Convert page_to_mfn and mfn_to_page to use typesafe MFN
Most of the users of page_to_mfn and mfn_to_page are either overriding
the macros to make them work with mfn_t or use mfn_x/_mfn because the
rest of the function use mfn_t.
So make page_to_mfn and mfn_to_page return mfn_t by default. The __*
version are now dropped as this patch will convert all the remaining
non-typesafe callers.
Only reasonable clean-ups are done in this patch. The rest will use
_mfn/mfn_x for the time being.
Lastly, domain_page_to_mfn is also converted to use mfn_t given that
most of the callers are now switched to _mfn(domain_page_to_mfn(...)).
Signed-off-by: Julien Grall <julien.grall@arm.com> Acked-by: Razvan Cojocaru <rcojocaru@bitdefender.com> Reviewed-by: Paul Durrant <paul.durrant@citrix.com> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Wei Liu <wei.liu2@citrix.com> Acked-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: George Dunlap <george.dunlap@citrix.com> Acked-by: Tim Deegan <tim@xen.org> Acked-by: Stefano Stabellini <sstabellini@kernel.org>
Julien Grall [Wed, 21 Feb 2018 13:46:26 +0000 (13:46 +0000)]
xen/grant: Switch common/grant_table.c to use typesafe MFN
At the same time replace MFN 0 by INVALID_MFN or drop the initializer
when it is not necessary. This will make clearer that the MFN
initialized is not valid.
Other than MFN 0 -> INVALID_MFN, no functional change intended.
Signed-off-by: Julien Grall <julien.grall@arm.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Wei Liu <wei.liu2@citrix.com> Acked-by: Stefano Stabellini <sstabellini@kernel.org>
projects / xen.git / log