During CPU down operation RCU callbacks are scheduled to finish
off some actions later as soon as CPU is fully dead (the same applies
to CPU up operation in case error path is taken). If in the same grace
period another CPU up operation is performed on the same CPU, RCU callback
will be called later on a CPU in a potentially wrong (already up again
instead of still being down) state leading to eventual state inconsistency
and/or crash.
In order to avoid it - flush RCU callbacks explicitly before starting the
next CPU up/down operation.
Signed-off-by: Igor Druzhinin <igor.druzhinin@citrix.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Xen's RCU implementation relies on no softirq handling taking place
while being in a RCU critical section. Add ASSERT()s in debug builds
in order to catch any violations.
For that purpose modify rcu_read_[un]lock() to use a dedicated percpu
counter additional to preempt_[en|dis]able() as this enables to test
that condition in __do_softirq() (ASSERT_NOT_IN_ATOMIC() is not
usable there due to __cpu_up() calling process_pending_softirqs()
while holding the cpu hotplug lock).
While at it switch the rcu_read_[un]lock() implementation to static
inline functions instead of macros.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Some keyhandlers are calling process_pending_softirqs() while holding
a rcu_read_lock(). This is wrong, as process_pending_softirqs() might
activate rcu calls which should not happen inside a rcu_read_lock().
For that purpose modify process_pending_softirqs() to not allow rcu
callback processing when a rcu_read_lock() is being held.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Today rcu_barrier() is calling stop_machine_run() to synchronize all
physical cpus in order to ensure all pending rcu calls have finished
when returning.
As stop_machine_run() is using tasklets this requires scheduling of
idle vcpus on all cpus imposing the need to call rcu_barrier() on idle
cpus only in case of core scheduling being active, as otherwise a
scheduling deadlock would occur.
There is no need at all to do the syncing of the cpus in tasklets, as
rcu activity is started in __do_softirq() called whenever softirq
activity is allowed. So rcu_barrier() can easily be modified to use
softirq for synchronization of the cpus no longer requiring any
scheduling activity.
As there already is a rcu softirq reuse that for the synchronization.
Remove the barrier element from struct rcu_data as it isn't used.
Finally switch rcu_barrier() to return void as it now can never fail.
Partially-based-on-patch-by: Igor Druzhinin <igor.druzhinin@citrix.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
When using atomic variables for synchronization barriers are needed
to ensure proper data serialization. Introduce smp_mb__before_atomic()
and smp_mb__after_atomic() as in the Linux kernel for that purpose.
Use the same definitions as in the Linux kernel.
Suggested-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Julien Grall <jgrall@amazon.com>
Intel CPUs permit both insns there while AMD ones don't.
While at it also
- drop the ring 0 check from SYSENTER handling - neither Intel's nor
AMD's insn pages have any indication of #GP(0) getting raised when
executed from ring 0, and trying it out in practice also confirms
the check shouldn't be there,
- move SYSENTER segment register writing until after the (in principle
able to fail) MSR reads.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Intel CPUs ignore operand size overrides here, while AMD ones don't.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Intel CPUs ignore operand size overrides here, while AMD ones don't.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Check whether there's a valid interrupt in VM_EXIT_INTR_INFO in order
to decide whether to update SVI in nvmx_update_apicv. If Ack on exit
is not being used VM_EXIT_INTR_INFO won't have a valid interrupt and
hence SVI shouldn't be updated to signal the interrupt is currently in
service because it won't be Acked.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
This reverts commit f96e1469ad.
The commit is wrong, as the whole point of nvmx_update_apicv is to
update the guest interrupt status field when the Ack on exit VMEXIT
control feature is enabled.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Offlining a cpu with core scheduling active can result in a hanging
system. Reason is the scheduling resource and unit of the to be removed
cpus needs to be split in order to remove the cpu from its cpupool and
move it to the idle scheduler. In case one of the involved cpus happens
to have received a sched slave event due to a vcpu former having been
running on that cpu being woken up again, it can happen that this cpu
will enter sched_wait_rendezvous_in() while its scheduling resource is
just about to be split. It might wait for ever for the other sibling
to join, which will never happen due to the resources already being
modified.
This can easily be avoided by:
- resetting the rendezvous counters of the idle unit which is kept
- checking for a new scheduling resource in sched_wait_rendezvous_in()
after reacquiring the scheduling lock and resetting the counters in
that case without scheduling another vcpu
- moving schedule resource modifications (in schedule_cpu_rm()) and
retrieving (schedule(), sched_slave() is fine already, others are not
critical) into locked regions
Reported-by: Igor Druzhinin <igor.druzhinin@citrix.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Dario Faggioli <dfaggioli@suse.com>
... to cover xenheap and PGC_extra pages.
PGC_extra pages are intended to hold data structures that are associated
with a domain and may be mapped by that domain. They should not be treated
as 'normal' guest pages (i.e. RAM or page tables). Hence, in many cases
where code currently tests is_xen_heap_page() it should also check for
the PGC_extra bit in 'count_info'.
This patch therefore defines is_special_page() to cover both cases and
converts tests of is_xen_heap_page() (or open coded tests of PGC_xen_heap)
to is_special_page() where the page is assigned to a domain.
Signed-off-by: Paul Durrant <paul@xen.org>
Acked-by: Tamas K Lengyel <tamas@tklengyel.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Julien Grall <julien@xen.org>
... now that it is safe to assign them.
This avoids relying on libxl (or whatever toolstack is in use) setting
max_pages up with sufficient 'slop' to allow all necessary ioreq server
pages to be allocated.
Signed-off-by: Paul Durrant <paul@xen.org>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
This patch adds a new page_list_head into struct domain to hold PGC_extra
pages. This avoids them getting confused with 'normal' domheap pages where
the domain's page_list is walked.
A new dump loop is also added to dump_pageframe_info() to unconditionally
dump the 'extra page list'.
Signed-off-by: Paul Durrant <paul@xen.org>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Julien Grall <julien@xen.org>
When onlining a cpu cpupool_cpu_add() checks whether all siblings of
the new cpu are free in order to decide whether to add it to cpupool0.
In case the added cpu is not the last sibling to be onlined this test
is wrong as it only checks for all online siblings to be free. The
test should include the check for the number of siblings having
reached the scheduling granularity of cpupool0, too.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Dario Faggioli <dfaggioli@suse.com>
Currently the xl dmesg output on Hygon platforms will be
"(XEN) CPU0: AMD Fam18h machine check reporting enabled",
which is misleading as AMD does not have family 18h (Hygon
negotiated with AMD to confirm that only Hygon has family 18h).
To correct this, add Hygon machine check type and vendor string.
Signed-off-by: Pu Wen <puwen@hygon.cn>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
struct evtchn_set_priority uses uint32_t type for event channel port.
Replace the type with evtchn_port_t. Such change is also done in Linux.
Signed-off-by: Yan Yankovskyi <yyankovskyi@gmail.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
The allocation can just return NULL. Return an error value early instead
of crashing later on.
Signed-off-by: Hongyan Xia <hongyxia@amazon.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
This causes gcc (yes, and clang) to emit phony targets for each dependency.
This means that when a header file is deleted, the C files which *used*
to include it will no longer stop building with bogus out-of-date
dependencies like this:
make[5]: *** No rule to make target
'/home/dwmw2/git/xen/xen/include/asm/hvm/svm/amd-iommu-proto.h',
needed by 'p2m.o'. Stop.
Based on -MP post-dating -MP by many years it is assumed that the
behavior of -MP isn't the defualt just out of extreme caution. We're
sufficiently convinced that there are no undue side effects of this.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
A return value of zero means no PV drivers. Restore a hunk which was removed.
Fixes commit b183e180bc
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Acked-by: Wei Liu <wl@xen.org>
Drop microcode_init_{intel,amd}(), export {intel,amd}_ucode_ops, and use a
switch statement in early_microcode_init() rather than probing each vendor in
turn. This allows the microcode_ops pointer to become local to core.c.
As there are no external users of microcode_ops, there is no need for
collect_cpu_info() to implement sanity checks. Move applicable checks to
early_microcode_init() so they are performed once, rather than repeatedly.
The Intel logic guarding the read of MSR_PLATFORM_ID is contrary to the SDM,
which states that the MSR has been architectural since the Pentium Pro
(06-01-xx), and lists no family/model restrictions in the pseudo-code for
microcode loading. Either way, Xen's 64bit-only nature already makes this
check redundant.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
This reduces the complexity of processor.h, particularly the need to include
public/xen.h. Substitute processor.h includes for microcode.h in some
sources, and add microcode.h includes in others.
Only 4 of the function declarations are actually called externally. Move the
vendor init declarations to private.h
No functional change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Split the existing asm/microcode.h in half, keeping the per-cpu cpu_sig
available to external users, and moving everything else into private.h
Take the opportunity to trim and clean up the include lists for all 3 source
files, all of which include rather more than necessary.
No functional change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Neither microcode_free_patch() nor early_microcode_update_cpu() have external
callers. Make them static.
early_microcode_update_cpu()'s sole caller is following a use of
microcode_ops, making the error path dead. Drop it as well.
No functional change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
... and, if it is not created, make the top level 'device' node in
xenstore writable by the guest instead.
The purpose and semantics of the suspend event channel node are explained
in xenstore-paths.pandoc [1]. It was originally introduced in xend by
commit 17636f47a4 "Teach xc_save to use event-channel-based domain
suspend if available.". Note that, because, the top-level frontend
'device' node was created writable by the guest in xend, there was no
need to explicitly create the 'suspend-event-channel' node as a writable
node.
However, libxl creates the 'device' node as read-only by the guest and so
explicit creation of the 'suspend-event-channel' node is necessary to make
it usable. This unfortunately has the side-effect of making some old
Windows PV drivers [2] cease to function. This is because they scan the top
level 'device' node, find the 'suspend' node and expect it to contain the
usual sub-nodes describing a PV frontend. When this is found not to be the
case, enumeration ceases and (because the 'suspend' node is observed before
the 'vbd' node) no system disk is enumerated. Windows will then crash with
bugcheck code 0x7B (missing system disk).
This patch adds a boolean 'xend_suspend_evtchn_compat' field into
libxl_create_info and a similarly named option in xl.cfg to set it.
If the value is true then the xenstore node is not created. Instead the
old xend behaviour of making top level device node writable by the guest is
re-instated. If the value is false (the default) then the current libxl
behaviour persists.
xenstore-paths.pandoc is also modified to say that the suspend event
channel node may not exist and, if it does not exist, then the guest may
create it. A note is also added concerning the writability of the top
level device node.
[1] https://xenbits.xen.org/gitweb/?p=xen.git;a=blob;f=docs/misc/xenstore-paths.pandoc;hb=HEAD#l177
[2] https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/5/html/para-virtualized_windows_drivers_guide/sect-para-virtualized_windows_drivers_guide-installing_and_configuring_the_para_virtualized_drivers-installing_the_para_virtualized_drivers
NOTE: While adding the new LIBXL_HAVE_CREATEINFO_... definition into
libxl.h, this patch corrects the previous stanza which erroneously
implies libxl_domain_create_info is a function.
Signed-off-by: Paul Durrant <paul@xen.org>
Reviewed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Due to recent reshuffling of header include paths mem_sharing no longer
compiles. Fix it by moving mem_sharing_domain declaration to location it
is used in.
Signed-off-by: Tamas K Lengyel <tamas@tklengyel.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
args.preempted is meaningless here as it doesn't signal whether the
hypercall was preempted before. Use start_extent instead which is
correct (as long as the hypercall was invoked in a "normal" way).
Signed-off-by: Igor Druzhinin <igor.druzhinin@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Roger Pau Monné <roger.pau@citrix.com>
Along the lines of commit d0b3ab0a0f46 ("libfdt: Fix undefined behaviour
in fdt_offset_ptr()"), _fdt_splice() similarly may not use pointer
arithmetic to do overflow checks.
[upstream commit 73d6e9ecb4179b510408bc526240f829262df361]
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Julien Grall <jgrall@amazon.com>
Using pointer arithmetic to generate a pointer outside a known object is,
technically, undefined behaviour in C. Unfortunately, we were using that
in fdt_offset_ptr() to detect overflows.
To fix this we need to do our bounds / overflow checking on the offsets
before constructing pointers from them.
Reported-by: David Binderman <dcb314@hotmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
[upstream commit d0b3ab0a0f46ac929b4713da46f7fdcd893dd3bd]
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Julien Grall <jgrall@amazon.com>
Drop the public header #include as not needed by the header itself. Add
one that was missing, and move all inside the inclusion guard.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Drop #include-s not needed by the header itself, and add smaller scope
ones instead. Put the ones needed into whichever other files actually
need them.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul@xen.org>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Drop #include-s not needed by the header itself as well as one include
of the header which isn't needed. Put the one needed into the file
actually requiring it.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Drop an #include not needed by the header itself. While verifying the
header (now) builds standalone, I noticed an omission in a public header
which gets taken care of here as well.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Drop #include-s not needed by the header itself. Put the ones needed
into whichever other files actually need them.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul@xen.org>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Drop #include-s not needed by the header itself. Put the ones needed
into whichever other files actually need them.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul@xen.org>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
c/s c47984aabe "nvmx: implement support for MSR bitmaps" introduced a use of
map_domain_page() which may get used in the middle of context switch.
This is not safe, and causes Xen to deadlock on the mapcache lock:
(XEN) Xen call trace:
(XEN) [<ffff82d08022d6ae>] R _spin_lock+0x34/0x5e
(XEN) [<ffff82d0803219d7>] F map_domain_page+0x250/0x527
(XEN) [<ffff82d080356332>] F do_page_fault+0x420/0x780
(XEN) [<ffff82d08038da3d>] F x86_64/entry.S#handle_exception_saved+0x68/0x94
(XEN) [<ffff82d08031729f>] F __find_next_zero_bit+0x28/0x69
(XEN) [<ffff82d080321a4d>] F map_domain_page+0x2c6/0x527
(XEN) [<ffff82d08029eeb2>] F nvmx_update_exec_control+0x1d7/0x323
(XEN) [<ffff82d080299f5a>] F vmx_update_cpu_exec_control+0x23/0x40
(XEN) [<ffff82d08029a3f7>] F arch/x86/hvm/vmx/vmx.c#vmx_ctxt_switch_from+0xb7/0x121
(XEN) [<ffff82d08031d796>] F arch/x86/domain.c#__context_switch+0x124/0x4a9
(XEN) [<ffff82d080320925>] F context_switch+0x154/0x62c
(XEN) [<ffff82d080252f3e>] F common/sched/core.c#sched_context_switch+0x16a/0x175
(XEN) [<ffff82d080253877>] F common/sched/core.c#schedule+0x2ad/0x2bc
(XEN) [<ffff82d08022cc97>] F common/softirq.c#__do_softirq+0xb7/0xc8
(XEN) [<ffff82d08022cd38>] F do_softirq+0x18/0x1a
(XEN) [<ffff82d0802a2fbb>] F vmx_asm_do_vmentry+0x2b/0x30
Convert the domheap page into being a xenheap page.
Fixes: c47984aabe - nvmx: implement support for MSR bitmaps
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Dividing by HZ/10 just to subsequently multiply by HZ again in all uses
of the respective variable is pretty pointlessly introducing rounding
(really: truncation) errors. While transforming the respective
expressions it became apparent that "result" would be left unused except
for its use as function return value. As the sole caller of the function
doesn't look at the returned value, simply convert the function to have
"void" return type.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Plain (unsigned) integer division simply truncates the results. The
overall errors are smaller though if we use proper rounding. (Extend
this to the purely cosmetic aspect of time.c's freq_string(), which
before this change I've frequently observed to report e.g. NN.999MHz
HPET clock speeds.)
While adding the rounding logic, also switch to using an unsigned
constant for the other, original half of bus_cycle's calculation.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
In case Xen ever gains preemption support the spinlock coding's
placement of preempt_disable() and preempt_enable() should be outside
of the locked section.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Similar to spinlocks preemption should be disabled while holding a
rwlock.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
First and foremost the comment talking about potential underflow being
taken care of by using signed long type variables was true only on
32-bit, which we've not been supporting for quite some time. Drop the
comment and change all involved types to unsigned. Take the opportunity
and also replace bus_cycle's fixed width type.
Additionally there's no point using an "arbitrary (but long enough)
timeout" here. Just use the maximum possible value; Linux does so too,
just as an additional data point.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Without having them at least at the xen/Makefile level they're (close
to?) inaccessible. As I'm uncertain about their utility at the top
level, I'm leaving it at that for now.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Wei Liu <wl@xen.org>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
amd_iommu_get_paging_mode() expects a count, not a "maximum possible"
value. Prior to b4f042236a dropping the reference, the use of our mis-
named "max_page" in amd_iommu_domain_init() may have lead to such a
misunderstanding. In an attempt to avoid such confusion in the future,
rename the function's parameter and - while at it - convert it to an
inline function.
Also replace a literal 4 by an expression tying it to a wider use
constant, just like amd_iommu_quarantine_init() does.
Fixes: ea38867831 ("x86 / iommu: set up a scratch page in the quarantine domain")
Fixes: b4f042236a ("AMD/IOMMU: Cease using a dynamic height for the IOMMU pagetables")
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Igor Druzhinin